diff --git "a/trainer_state.json" "b/trainer_state.json" new file mode 100644--- /dev/null +++ "b/trainer_state.json" @@ -0,0 +1,249522 @@ +{ + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 1.0, + "eval_steps": 500, + "global_step": 41582, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0, + "learning_rate": 1.6025641025641026e-08, + "loss": 1.9538, + "step": 1 + }, + { + "epoch": 0.0, + "learning_rate": 3.205128205128205e-08, + "loss": 1.9164, + "step": 2 + }, + { + "epoch": 0.0, + "learning_rate": 4.8076923076923085e-08, + "loss": 2.171, + "step": 3 + }, + { + "epoch": 0.0, + "learning_rate": 6.41025641025641e-08, + "loss": 2.1104, + "step": 4 + }, + { + "epoch": 0.0, + "learning_rate": 8.012820512820512e-08, + "loss": 2.1573, + "step": 5 + }, + { + "epoch": 0.0, + "learning_rate": 9.615384615384617e-08, + "loss": 2.0093, + "step": 6 + }, + { + "epoch": 0.0, + "learning_rate": 1.1217948717948719e-07, + "loss": 1.981, + "step": 7 + }, + { + "epoch": 0.0, + "learning_rate": 1.282051282051282e-07, + "loss": 2.1724, + "step": 8 + }, + { + "epoch": 0.0, + "learning_rate": 1.4423076923076925e-07, + "loss": 1.7192, + "step": 9 + }, + { + "epoch": 0.0, + "learning_rate": 1.6025641025641025e-07, + "loss": 1.7537, + "step": 10 + }, + { + "epoch": 0.0, + "learning_rate": 1.762820512820513e-07, + "loss": 2.26, + "step": 11 + }, + { + "epoch": 0.0, + "learning_rate": 1.9230769230769234e-07, + "loss": 1.6521, + "step": 12 + }, + { + "epoch": 0.0, + "learning_rate": 2.0833333333333333e-07, + "loss": 1.9375, + "step": 13 + }, + { + "epoch": 0.0, + "learning_rate": 2.2435897435897438e-07, + "loss": 1.8024, + "step": 14 + }, + { + "epoch": 0.0, + "learning_rate": 2.403846153846154e-07, + "loss": 2.0405, + "step": 15 + }, + { + "epoch": 0.0, + "learning_rate": 2.564102564102564e-07, + "loss": 1.7396, + "step": 16 + }, + { + "epoch": 0.0, + "learning_rate": 2.7243589743589746e-07, + "loss": 1.7965, + "step": 17 + }, + { + "epoch": 0.0, + "learning_rate": 2.884615384615385e-07, + "loss": 1.9289, + "step": 18 + }, + { + "epoch": 0.0, + "learning_rate": 3.044871794871795e-07, + "loss": 1.6023, + "step": 19 + }, + { + "epoch": 0.0, + "learning_rate": 3.205128205128205e-07, + "loss": 1.8467, + "step": 20 + }, + { + "epoch": 0.0, + "learning_rate": 3.3653846153846154e-07, + "loss": 2.1368, + "step": 21 + }, + { + "epoch": 0.0, + "learning_rate": 3.525641025641026e-07, + "loss": 1.8213, + "step": 22 + }, + { + "epoch": 0.0, + "learning_rate": 3.6858974358974363e-07, + "loss": 2.1428, + "step": 23 + }, + { + "epoch": 0.0, + "learning_rate": 3.846153846153847e-07, + "loss": 1.8322, + "step": 24 + }, + { + "epoch": 0.0, + "learning_rate": 4.006410256410257e-07, + "loss": 2.1171, + "step": 25 + }, + { + "epoch": 0.0, + "learning_rate": 4.1666666666666667e-07, + "loss": 2.0083, + "step": 26 + }, + { + "epoch": 0.0, + "learning_rate": 4.326923076923077e-07, + "loss": 2.1985, + "step": 27 + }, + { + "epoch": 0.0, + "learning_rate": 4.4871794871794876e-07, + "loss": 2.1939, + "step": 28 + }, + { + "epoch": 0.0, + "learning_rate": 4.647435897435898e-07, + "loss": 2.2956, + "step": 29 + }, + { + "epoch": 0.0, + "learning_rate": 4.807692307692308e-07, + "loss": 2.1718, + "step": 30 + }, + { + "epoch": 0.0, + "learning_rate": 4.967948717948718e-07, + "loss": 1.6943, + "step": 31 + }, + { + "epoch": 0.0, + "learning_rate": 5.128205128205128e-07, + "loss": 2.062, + "step": 32 + }, + { + "epoch": 0.0, + "learning_rate": 5.288461538461539e-07, + "loss": 1.8579, + "step": 33 + }, + { + "epoch": 0.0, + "learning_rate": 5.448717948717949e-07, + "loss": 1.9138, + "step": 34 + }, + { + "epoch": 0.0, + "learning_rate": 5.60897435897436e-07, + "loss": 1.9889, + "step": 35 + }, + { + "epoch": 0.0, + "learning_rate": 5.76923076923077e-07, + "loss": 1.9861, + "step": 36 + }, + { + "epoch": 0.0, + "learning_rate": 5.92948717948718e-07, + "loss": 2.0173, + "step": 37 + }, + { + "epoch": 0.0, + "learning_rate": 6.08974358974359e-07, + "loss": 2.0672, + "step": 38 + }, + { + "epoch": 0.0, + "learning_rate": 6.25e-07, + "loss": 1.9626, + "step": 39 + }, + { + "epoch": 0.0, + "learning_rate": 6.41025641025641e-07, + "loss": 1.8808, + "step": 40 + }, + { + "epoch": 0.0, + "learning_rate": 6.570512820512821e-07, + "loss": 1.5862, + "step": 41 + }, + { + "epoch": 0.0, + "learning_rate": 6.730769230769231e-07, + "loss": 1.8408, + "step": 42 + }, + { + "epoch": 0.0, + "learning_rate": 6.891025641025642e-07, + "loss": 1.9521, + "step": 43 + }, + { + "epoch": 0.0, + "learning_rate": 7.051282051282052e-07, + "loss": 1.7897, + "step": 44 + }, + { + "epoch": 0.0, + "learning_rate": 7.211538461538461e-07, + "loss": 1.7244, + "step": 45 + }, + { + "epoch": 0.0, + "learning_rate": 7.371794871794873e-07, + "loss": 1.9017, + "step": 46 + }, + { + "epoch": 0.0, + "learning_rate": 7.532051282051282e-07, + "loss": 2.0269, + "step": 47 + }, + { + "epoch": 0.0, + "learning_rate": 7.692307692307694e-07, + "loss": 2.0405, + "step": 48 + }, + { + "epoch": 0.0, + "learning_rate": 7.852564102564103e-07, + "loss": 2.0407, + "step": 49 + }, + { + "epoch": 0.0, + "learning_rate": 8.012820512820515e-07, + "loss": 1.8196, + "step": 50 + }, + { + "epoch": 0.0, + "learning_rate": 8.173076923076924e-07, + "loss": 1.9247, + "step": 51 + }, + { + "epoch": 0.0, + "learning_rate": 8.333333333333333e-07, + "loss": 1.8378, + "step": 52 + }, + { + "epoch": 0.0, + "learning_rate": 8.493589743589745e-07, + "loss": 1.7581, + "step": 53 + }, + { + "epoch": 0.0, + "learning_rate": 8.653846153846154e-07, + "loss": 1.8895, + "step": 54 + }, + { + "epoch": 0.0, + "learning_rate": 8.814102564102566e-07, + "loss": 1.8145, + "step": 55 + }, + { + "epoch": 0.0, + "learning_rate": 8.974358974358975e-07, + "loss": 1.7175, + "step": 56 + }, + { + "epoch": 0.0, + "learning_rate": 9.134615384615385e-07, + "loss": 1.9985, + "step": 57 + }, + { + "epoch": 0.0, + "learning_rate": 9.294871794871796e-07, + "loss": 1.5251, + "step": 58 + }, + { + "epoch": 0.0, + "learning_rate": 9.455128205128205e-07, + "loss": 1.6362, + "step": 59 + }, + { + "epoch": 0.0, + "learning_rate": 9.615384615384617e-07, + "loss": 1.9595, + "step": 60 + }, + { + "epoch": 0.0, + "learning_rate": 9.775641025641027e-07, + "loss": 1.7476, + "step": 61 + }, + { + "epoch": 0.0, + "learning_rate": 9.935897435897436e-07, + "loss": 1.9022, + "step": 62 + }, + { + "epoch": 0.0, + "learning_rate": 1.0096153846153846e-06, + "loss": 1.7147, + "step": 63 + }, + { + "epoch": 0.0, + "learning_rate": 1.0256410256410257e-06, + "loss": 1.776, + "step": 64 + }, + { + "epoch": 0.0, + "learning_rate": 1.0416666666666667e-06, + "loss": 1.4594, + "step": 65 + }, + { + "epoch": 0.0, + "learning_rate": 1.0576923076923078e-06, + "loss": 1.9547, + "step": 66 + }, + { + "epoch": 0.0, + "learning_rate": 1.0737179487179488e-06, + "loss": 1.8702, + "step": 67 + }, + { + "epoch": 0.0, + "learning_rate": 1.0897435897435899e-06, + "loss": 1.7985, + "step": 68 + }, + { + "epoch": 0.0, + "learning_rate": 1.105769230769231e-06, + "loss": 1.7049, + "step": 69 + }, + { + "epoch": 0.0, + "learning_rate": 1.121794871794872e-06, + "loss": 1.6199, + "step": 70 + }, + { + "epoch": 0.0, + "learning_rate": 1.137820512820513e-06, + "loss": 1.7192, + "step": 71 + }, + { + "epoch": 0.0, + "learning_rate": 1.153846153846154e-06, + "loss": 1.5246, + "step": 72 + }, + { + "epoch": 0.0, + "learning_rate": 1.1698717948717949e-06, + "loss": 2.0134, + "step": 73 + }, + { + "epoch": 0.0, + "learning_rate": 1.185897435897436e-06, + "loss": 1.8061, + "step": 74 + }, + { + "epoch": 0.0, + "learning_rate": 1.201923076923077e-06, + "loss": 1.9516, + "step": 75 + }, + { + "epoch": 0.0, + "learning_rate": 1.217948717948718e-06, + "loss": 1.9362, + "step": 76 + }, + { + "epoch": 0.0, + "learning_rate": 1.233974358974359e-06, + "loss": 1.5765, + "step": 77 + }, + { + "epoch": 0.0, + "learning_rate": 1.25e-06, + "loss": 1.8221, + "step": 78 + }, + { + "epoch": 0.0, + "learning_rate": 1.2660256410256411e-06, + "loss": 2.2832, + "step": 79 + }, + { + "epoch": 0.0, + "learning_rate": 1.282051282051282e-06, + "loss": 1.5049, + "step": 80 + }, + { + "epoch": 0.0, + "learning_rate": 1.2980769230769232e-06, + "loss": 1.6785, + "step": 81 + }, + { + "epoch": 0.0, + "learning_rate": 1.3141025641025643e-06, + "loss": 1.9647, + "step": 82 + }, + { + "epoch": 0.0, + "learning_rate": 1.3301282051282053e-06, + "loss": 1.7894, + "step": 83 + }, + { + "epoch": 0.0, + "learning_rate": 1.3461538461538462e-06, + "loss": 1.9051, + "step": 84 + }, + { + "epoch": 0.0, + "learning_rate": 1.3621794871794872e-06, + "loss": 1.5444, + "step": 85 + }, + { + "epoch": 0.0, + "learning_rate": 1.3782051282051285e-06, + "loss": 1.6199, + "step": 86 + }, + { + "epoch": 0.0, + "learning_rate": 1.3942307692307693e-06, + "loss": 1.5615, + "step": 87 + }, + { + "epoch": 0.0, + "learning_rate": 1.4102564102564104e-06, + "loss": 1.8761, + "step": 88 + }, + { + "epoch": 0.0, + "learning_rate": 1.4262820512820514e-06, + "loss": 1.6453, + "step": 89 + }, + { + "epoch": 0.0, + "learning_rate": 1.4423076923076922e-06, + "loss": 1.8099, + "step": 90 + }, + { + "epoch": 0.0, + "learning_rate": 1.4583333333333335e-06, + "loss": 1.5529, + "step": 91 + }, + { + "epoch": 0.0, + "learning_rate": 1.4743589743589745e-06, + "loss": 1.7383, + "step": 92 + }, + { + "epoch": 0.0, + "learning_rate": 1.4903846153846156e-06, + "loss": 1.6257, + "step": 93 + }, + { + "epoch": 0.0, + "learning_rate": 1.5064102564102564e-06, + "loss": 1.3988, + "step": 94 + }, + { + "epoch": 0.0, + "learning_rate": 1.5224358974358975e-06, + "loss": 1.2811, + "step": 95 + }, + { + "epoch": 0.0, + "learning_rate": 1.5384615384615387e-06, + "loss": 1.6967, + "step": 96 + }, + { + "epoch": 0.0, + "learning_rate": 1.5544871794871796e-06, + "loss": 2.0577, + "step": 97 + }, + { + "epoch": 0.0, + "learning_rate": 1.5705128205128206e-06, + "loss": 1.5674, + "step": 98 + }, + { + "epoch": 0.0, + "learning_rate": 1.5865384615384616e-06, + "loss": 1.6837, + "step": 99 + }, + { + "epoch": 0.0, + "learning_rate": 1.602564102564103e-06, + "loss": 1.4381, + "step": 100 + }, + { + "epoch": 0.0, + "learning_rate": 1.6185897435897437e-06, + "loss": 1.5557, + "step": 101 + }, + { + "epoch": 0.0, + "learning_rate": 1.6346153846153848e-06, + "loss": 1.4078, + "step": 102 + }, + { + "epoch": 0.0, + "learning_rate": 1.6506410256410258e-06, + "loss": 1.3764, + "step": 103 + }, + { + "epoch": 0.0, + "learning_rate": 1.6666666666666667e-06, + "loss": 1.5025, + "step": 104 + }, + { + "epoch": 0.0, + "learning_rate": 1.682692307692308e-06, + "loss": 1.6121, + "step": 105 + }, + { + "epoch": 0.0, + "learning_rate": 1.698717948717949e-06, + "loss": 1.4806, + "step": 106 + }, + { + "epoch": 0.0, + "learning_rate": 1.7147435897435898e-06, + "loss": 1.3984, + "step": 107 + }, + { + "epoch": 0.0, + "learning_rate": 1.7307692307692308e-06, + "loss": 1.1344, + "step": 108 + }, + { + "epoch": 0.0, + "learning_rate": 1.7467948717948719e-06, + "loss": 1.2661, + "step": 109 + }, + { + "epoch": 0.0, + "learning_rate": 1.7628205128205131e-06, + "loss": 1.2154, + "step": 110 + }, + { + "epoch": 0.0, + "learning_rate": 1.778846153846154e-06, + "loss": 1.1215, + "step": 111 + }, + { + "epoch": 0.0, + "learning_rate": 1.794871794871795e-06, + "loss": 1.4262, + "step": 112 + }, + { + "epoch": 0.0, + "learning_rate": 1.810897435897436e-06, + "loss": 1.2954, + "step": 113 + }, + { + "epoch": 0.0, + "learning_rate": 1.826923076923077e-06, + "loss": 1.2884, + "step": 114 + }, + { + "epoch": 0.0, + "learning_rate": 1.8429487179487182e-06, + "loss": 1.4665, + "step": 115 + }, + { + "epoch": 0.0, + "learning_rate": 1.8589743589743592e-06, + "loss": 1.4191, + "step": 116 + }, + { + "epoch": 0.0, + "learning_rate": 1.8750000000000003e-06, + "loss": 1.36, + "step": 117 + }, + { + "epoch": 0.0, + "learning_rate": 1.891025641025641e-06, + "loss": 1.3036, + "step": 118 + }, + { + "epoch": 0.0, + "learning_rate": 1.9070512820512821e-06, + "loss": 1.2402, + "step": 119 + }, + { + "epoch": 0.0, + "learning_rate": 1.9230769230769234e-06, + "loss": 1.4345, + "step": 120 + }, + { + "epoch": 0.0, + "learning_rate": 1.9391025641025644e-06, + "loss": 1.4774, + "step": 121 + }, + { + "epoch": 0.0, + "learning_rate": 1.9551282051282055e-06, + "loss": 1.289, + "step": 122 + }, + { + "epoch": 0.0, + "learning_rate": 1.971153846153846e-06, + "loss": 1.3362, + "step": 123 + }, + { + "epoch": 0.0, + "learning_rate": 1.987179487179487e-06, + "loss": 1.2849, + "step": 124 + }, + { + "epoch": 0.0, + "learning_rate": 2.0032051282051286e-06, + "loss": 1.6488, + "step": 125 + }, + { + "epoch": 0.0, + "learning_rate": 2.0192307692307692e-06, + "loss": 1.2244, + "step": 126 + }, + { + "epoch": 0.0, + "learning_rate": 2.0352564102564103e-06, + "loss": 1.6202, + "step": 127 + }, + { + "epoch": 0.0, + "learning_rate": 2.0512820512820513e-06, + "loss": 1.5624, + "step": 128 + }, + { + "epoch": 0.0, + "learning_rate": 2.0673076923076924e-06, + "loss": 1.3592, + "step": 129 + }, + { + "epoch": 0.0, + "learning_rate": 2.0833333333333334e-06, + "loss": 1.5368, + "step": 130 + }, + { + "epoch": 0.0, + "learning_rate": 2.0993589743589745e-06, + "loss": 1.5123, + "step": 131 + }, + { + "epoch": 0.0, + "learning_rate": 2.1153846153846155e-06, + "loss": 1.254, + "step": 132 + }, + { + "epoch": 0.0, + "learning_rate": 2.1314102564102566e-06, + "loss": 1.3201, + "step": 133 + }, + { + "epoch": 0.0, + "learning_rate": 2.1474358974358976e-06, + "loss": 1.4213, + "step": 134 + }, + { + "epoch": 0.0, + "learning_rate": 2.1634615384615387e-06, + "loss": 1.3141, + "step": 135 + }, + { + "epoch": 0.0, + "learning_rate": 2.1794871794871797e-06, + "loss": 1.5302, + "step": 136 + }, + { + "epoch": 0.0, + "learning_rate": 2.1955128205128208e-06, + "loss": 1.3298, + "step": 137 + }, + { + "epoch": 0.0, + "learning_rate": 2.211538461538462e-06, + "loss": 1.493, + "step": 138 + }, + { + "epoch": 0.0, + "learning_rate": 2.227564102564103e-06, + "loss": 1.3836, + "step": 139 + }, + { + "epoch": 0.0, + "learning_rate": 2.243589743589744e-06, + "loss": 1.5426, + "step": 140 + }, + { + "epoch": 0.0, + "learning_rate": 2.259615384615385e-06, + "loss": 1.3297, + "step": 141 + }, + { + "epoch": 0.0, + "learning_rate": 2.275641025641026e-06, + "loss": 1.4131, + "step": 142 + }, + { + "epoch": 0.0, + "learning_rate": 2.2916666666666666e-06, + "loss": 1.3647, + "step": 143 + }, + { + "epoch": 0.0, + "learning_rate": 2.307692307692308e-06, + "loss": 1.3118, + "step": 144 + }, + { + "epoch": 0.0, + "learning_rate": 2.323717948717949e-06, + "loss": 1.313, + "step": 145 + }, + { + "epoch": 0.0, + "learning_rate": 2.3397435897435897e-06, + "loss": 1.1575, + "step": 146 + }, + { + "epoch": 0.0, + "learning_rate": 2.355769230769231e-06, + "loss": 1.6193, + "step": 147 + }, + { + "epoch": 0.0, + "learning_rate": 2.371794871794872e-06, + "loss": 1.6139, + "step": 148 + }, + { + "epoch": 0.0, + "learning_rate": 2.3878205128205133e-06, + "loss": 1.5207, + "step": 149 + }, + { + "epoch": 0.0, + "learning_rate": 2.403846153846154e-06, + "loss": 1.2527, + "step": 150 + }, + { + "epoch": 0.0, + "learning_rate": 2.419871794871795e-06, + "loss": 1.3543, + "step": 151 + }, + { + "epoch": 0.0, + "learning_rate": 2.435897435897436e-06, + "loss": 1.1464, + "step": 152 + }, + { + "epoch": 0.0, + "learning_rate": 2.451923076923077e-06, + "loss": 1.3847, + "step": 153 + }, + { + "epoch": 0.0, + "learning_rate": 2.467948717948718e-06, + "loss": 1.4224, + "step": 154 + }, + { + "epoch": 0.0, + "learning_rate": 2.483974358974359e-06, + "loss": 1.4157, + "step": 155 + }, + { + "epoch": 0.0, + "learning_rate": 2.5e-06, + "loss": 1.3659, + "step": 156 + }, + { + "epoch": 0.0, + "learning_rate": 2.5160256410256413e-06, + "loss": 1.4306, + "step": 157 + }, + { + "epoch": 0.0, + "learning_rate": 2.5320512820512823e-06, + "loss": 1.2266, + "step": 158 + }, + { + "epoch": 0.0, + "learning_rate": 2.548076923076923e-06, + "loss": 1.2395, + "step": 159 + }, + { + "epoch": 0.0, + "learning_rate": 2.564102564102564e-06, + "loss": 1.185, + "step": 160 + }, + { + "epoch": 0.0, + "learning_rate": 2.5801282051282054e-06, + "loss": 1.2446, + "step": 161 + }, + { + "epoch": 0.0, + "learning_rate": 2.5961538461538465e-06, + "loss": 1.1803, + "step": 162 + }, + { + "epoch": 0.0, + "learning_rate": 2.6121794871794875e-06, + "loss": 1.3554, + "step": 163 + }, + { + "epoch": 0.0, + "learning_rate": 2.6282051282051286e-06, + "loss": 1.4014, + "step": 164 + }, + { + "epoch": 0.0, + "learning_rate": 2.6442307692307696e-06, + "loss": 1.5778, + "step": 165 + }, + { + "epoch": 0.0, + "learning_rate": 2.6602564102564107e-06, + "loss": 1.3804, + "step": 166 + }, + { + "epoch": 0.0, + "learning_rate": 2.6762820512820513e-06, + "loss": 1.1362, + "step": 167 + }, + { + "epoch": 0.0, + "learning_rate": 2.6923076923076923e-06, + "loss": 1.3089, + "step": 168 + }, + { + "epoch": 0.0, + "learning_rate": 2.7083333333333334e-06, + "loss": 1.5136, + "step": 169 + }, + { + "epoch": 0.0, + "learning_rate": 2.7243589743589744e-06, + "loss": 1.4478, + "step": 170 + }, + { + "epoch": 0.0, + "learning_rate": 2.740384615384616e-06, + "loss": 1.4632, + "step": 171 + }, + { + "epoch": 0.0, + "learning_rate": 2.756410256410257e-06, + "loss": 1.2353, + "step": 172 + }, + { + "epoch": 0.0, + "learning_rate": 2.772435897435898e-06, + "loss": 1.3857, + "step": 173 + }, + { + "epoch": 0.0, + "learning_rate": 2.7884615384615386e-06, + "loss": 1.5345, + "step": 174 + }, + { + "epoch": 0.0, + "learning_rate": 2.8044871794871797e-06, + "loss": 1.0988, + "step": 175 + }, + { + "epoch": 0.0, + "learning_rate": 2.8205128205128207e-06, + "loss": 1.2711, + "step": 176 + }, + { + "epoch": 0.0, + "learning_rate": 2.8365384615384617e-06, + "loss": 1.3319, + "step": 177 + }, + { + "epoch": 0.0, + "learning_rate": 2.852564102564103e-06, + "loss": 1.427, + "step": 178 + }, + { + "epoch": 0.0, + "learning_rate": 2.8685897435897434e-06, + "loss": 1.4368, + "step": 179 + }, + { + "epoch": 0.0, + "learning_rate": 2.8846153846153845e-06, + "loss": 1.1033, + "step": 180 + }, + { + "epoch": 0.0, + "learning_rate": 2.900641025641026e-06, + "loss": 1.1681, + "step": 181 + }, + { + "epoch": 0.0, + "learning_rate": 2.916666666666667e-06, + "loss": 1.4772, + "step": 182 + }, + { + "epoch": 0.0, + "learning_rate": 2.932692307692308e-06, + "loss": 1.3296, + "step": 183 + }, + { + "epoch": 0.0, + "learning_rate": 2.948717948717949e-06, + "loss": 1.3241, + "step": 184 + }, + { + "epoch": 0.0, + "learning_rate": 2.96474358974359e-06, + "loss": 1.3156, + "step": 185 + }, + { + "epoch": 0.0, + "learning_rate": 2.980769230769231e-06, + "loss": 1.4291, + "step": 186 + }, + { + "epoch": 0.0, + "learning_rate": 2.9967948717948718e-06, + "loss": 1.2777, + "step": 187 + }, + { + "epoch": 0.0, + "learning_rate": 3.012820512820513e-06, + "loss": 1.1893, + "step": 188 + }, + { + "epoch": 0.0, + "learning_rate": 3.028846153846154e-06, + "loss": 1.5126, + "step": 189 + }, + { + "epoch": 0.0, + "learning_rate": 3.044871794871795e-06, + "loss": 1.3923, + "step": 190 + }, + { + "epoch": 0.0, + "learning_rate": 3.0608974358974364e-06, + "loss": 1.2989, + "step": 191 + }, + { + "epoch": 0.0, + "learning_rate": 3.0769230769230774e-06, + "loss": 1.3042, + "step": 192 + }, + { + "epoch": 0.0, + "learning_rate": 3.0929487179487185e-06, + "loss": 1.2437, + "step": 193 + }, + { + "epoch": 0.0, + "learning_rate": 3.108974358974359e-06, + "loss": 1.5545, + "step": 194 + }, + { + "epoch": 0.0, + "learning_rate": 3.125e-06, + "loss": 1.5278, + "step": 195 + }, + { + "epoch": 0.0, + "learning_rate": 3.141025641025641e-06, + "loss": 1.3082, + "step": 196 + }, + { + "epoch": 0.0, + "learning_rate": 3.1570512820512822e-06, + "loss": 1.5546, + "step": 197 + }, + { + "epoch": 0.0, + "learning_rate": 3.1730769230769233e-06, + "loss": 1.2257, + "step": 198 + }, + { + "epoch": 0.0, + "learning_rate": 3.1891025641025643e-06, + "loss": 1.3822, + "step": 199 + }, + { + "epoch": 0.0, + "learning_rate": 3.205128205128206e-06, + "loss": 1.4296, + "step": 200 + }, + { + "epoch": 0.0, + "learning_rate": 3.2211538461538464e-06, + "loss": 1.5637, + "step": 201 + }, + { + "epoch": 0.0, + "learning_rate": 3.2371794871794875e-06, + "loss": 1.3505, + "step": 202 + }, + { + "epoch": 0.0, + "learning_rate": 3.2532051282051285e-06, + "loss": 1.4393, + "step": 203 + }, + { + "epoch": 0.0, + "learning_rate": 3.2692307692307696e-06, + "loss": 1.52, + "step": 204 + }, + { + "epoch": 0.0, + "learning_rate": 3.2852564102564106e-06, + "loss": 1.2799, + "step": 205 + }, + { + "epoch": 0.0, + "learning_rate": 3.3012820512820517e-06, + "loss": 1.4847, + "step": 206 + }, + { + "epoch": 0.0, + "learning_rate": 3.3173076923076923e-06, + "loss": 1.3072, + "step": 207 + }, + { + "epoch": 0.01, + "learning_rate": 3.3333333333333333e-06, + "loss": 1.1343, + "step": 208 + }, + { + "epoch": 0.01, + "learning_rate": 3.3493589743589744e-06, + "loss": 1.1716, + "step": 209 + }, + { + "epoch": 0.01, + "learning_rate": 3.365384615384616e-06, + "loss": 1.4641, + "step": 210 + }, + { + "epoch": 0.01, + "learning_rate": 3.381410256410257e-06, + "loss": 1.4082, + "step": 211 + }, + { + "epoch": 0.01, + "learning_rate": 3.397435897435898e-06, + "loss": 1.4428, + "step": 212 + }, + { + "epoch": 0.01, + "learning_rate": 3.413461538461539e-06, + "loss": 1.3146, + "step": 213 + }, + { + "epoch": 0.01, + "learning_rate": 3.4294871794871796e-06, + "loss": 1.2138, + "step": 214 + }, + { + "epoch": 0.01, + "learning_rate": 3.4455128205128206e-06, + "loss": 1.3853, + "step": 215 + }, + { + "epoch": 0.01, + "learning_rate": 3.4615384615384617e-06, + "loss": 1.3733, + "step": 216 + }, + { + "epoch": 0.01, + "learning_rate": 3.4775641025641027e-06, + "loss": 1.5065, + "step": 217 + }, + { + "epoch": 0.01, + "learning_rate": 3.4935897435897438e-06, + "loss": 1.583, + "step": 218 + }, + { + "epoch": 0.01, + "learning_rate": 3.509615384615385e-06, + "loss": 1.3676, + "step": 219 + }, + { + "epoch": 0.01, + "learning_rate": 3.5256410256410263e-06, + "loss": 1.3626, + "step": 220 + }, + { + "epoch": 0.01, + "learning_rate": 3.5416666666666673e-06, + "loss": 1.4496, + "step": 221 + }, + { + "epoch": 0.01, + "learning_rate": 3.557692307692308e-06, + "loss": 1.4055, + "step": 222 + }, + { + "epoch": 0.01, + "learning_rate": 3.573717948717949e-06, + "loss": 1.1764, + "step": 223 + }, + { + "epoch": 0.01, + "learning_rate": 3.58974358974359e-06, + "loss": 1.2525, + "step": 224 + }, + { + "epoch": 0.01, + "learning_rate": 3.605769230769231e-06, + "loss": 1.3222, + "step": 225 + }, + { + "epoch": 0.01, + "learning_rate": 3.621794871794872e-06, + "loss": 1.207, + "step": 226 + }, + { + "epoch": 0.01, + "learning_rate": 3.6378205128205128e-06, + "loss": 1.2839, + "step": 227 + }, + { + "epoch": 0.01, + "learning_rate": 3.653846153846154e-06, + "loss": 1.1866, + "step": 228 + }, + { + "epoch": 0.01, + "learning_rate": 3.669871794871795e-06, + "loss": 1.4367, + "step": 229 + }, + { + "epoch": 0.01, + "learning_rate": 3.6858974358974363e-06, + "loss": 1.366, + "step": 230 + }, + { + "epoch": 0.01, + "learning_rate": 3.7019230769230774e-06, + "loss": 1.5141, + "step": 231 + }, + { + "epoch": 0.01, + "learning_rate": 3.7179487179487184e-06, + "loss": 1.0181, + "step": 232 + }, + { + "epoch": 0.01, + "learning_rate": 3.7339743589743595e-06, + "loss": 1.2666, + "step": 233 + }, + { + "epoch": 0.01, + "learning_rate": 3.7500000000000005e-06, + "loss": 1.3622, + "step": 234 + }, + { + "epoch": 0.01, + "learning_rate": 3.766025641025641e-06, + "loss": 1.2849, + "step": 235 + }, + { + "epoch": 0.01, + "learning_rate": 3.782051282051282e-06, + "loss": 1.4603, + "step": 236 + }, + { + "epoch": 0.01, + "learning_rate": 3.7980769230769232e-06, + "loss": 1.4119, + "step": 237 + }, + { + "epoch": 0.01, + "learning_rate": 3.8141025641025643e-06, + "loss": 1.2073, + "step": 238 + }, + { + "epoch": 0.01, + "learning_rate": 3.830128205128206e-06, + "loss": 1.2492, + "step": 239 + }, + { + "epoch": 0.01, + "learning_rate": 3.846153846153847e-06, + "loss": 1.1187, + "step": 240 + }, + { + "epoch": 0.01, + "learning_rate": 3.862179487179488e-06, + "loss": 1.2268, + "step": 241 + }, + { + "epoch": 0.01, + "learning_rate": 3.878205128205129e-06, + "loss": 1.3603, + "step": 242 + }, + { + "epoch": 0.01, + "learning_rate": 3.89423076923077e-06, + "loss": 1.4297, + "step": 243 + }, + { + "epoch": 0.01, + "learning_rate": 3.910256410256411e-06, + "loss": 1.2315, + "step": 244 + }, + { + "epoch": 0.01, + "learning_rate": 3.926282051282051e-06, + "loss": 1.4049, + "step": 245 + }, + { + "epoch": 0.01, + "learning_rate": 3.942307692307692e-06, + "loss": 1.3065, + "step": 246 + }, + { + "epoch": 0.01, + "learning_rate": 3.958333333333333e-06, + "loss": 1.144, + "step": 247 + }, + { + "epoch": 0.01, + "learning_rate": 3.974358974358974e-06, + "loss": 1.2998, + "step": 248 + }, + { + "epoch": 0.01, + "learning_rate": 3.990384615384616e-06, + "loss": 1.338, + "step": 249 + }, + { + "epoch": 0.01, + "learning_rate": 4.006410256410257e-06, + "loss": 1.2309, + "step": 250 + }, + { + "epoch": 0.01, + "learning_rate": 4.022435897435898e-06, + "loss": 1.5311, + "step": 251 + }, + { + "epoch": 0.01, + "learning_rate": 4.0384615384615385e-06, + "loss": 1.405, + "step": 252 + }, + { + "epoch": 0.01, + "learning_rate": 4.0544871794871795e-06, + "loss": 1.0932, + "step": 253 + }, + { + "epoch": 0.01, + "learning_rate": 4.070512820512821e-06, + "loss": 1.2875, + "step": 254 + }, + { + "epoch": 0.01, + "learning_rate": 4.086538461538462e-06, + "loss": 1.1919, + "step": 255 + }, + { + "epoch": 0.01, + "learning_rate": 4.102564102564103e-06, + "loss": 1.1871, + "step": 256 + }, + { + "epoch": 0.01, + "learning_rate": 4.118589743589744e-06, + "loss": 1.1259, + "step": 257 + }, + { + "epoch": 0.01, + "learning_rate": 4.134615384615385e-06, + "loss": 1.2517, + "step": 258 + }, + { + "epoch": 0.01, + "learning_rate": 4.150641025641026e-06, + "loss": 1.2699, + "step": 259 + }, + { + "epoch": 0.01, + "learning_rate": 4.166666666666667e-06, + "loss": 1.3064, + "step": 260 + }, + { + "epoch": 0.01, + "learning_rate": 4.182692307692308e-06, + "loss": 1.1922, + "step": 261 + }, + { + "epoch": 0.01, + "learning_rate": 4.198717948717949e-06, + "loss": 1.2645, + "step": 262 + }, + { + "epoch": 0.01, + "learning_rate": 4.21474358974359e-06, + "loss": 1.2117, + "step": 263 + }, + { + "epoch": 0.01, + "learning_rate": 4.230769230769231e-06, + "loss": 1.1021, + "step": 264 + }, + { + "epoch": 0.01, + "learning_rate": 4.246794871794872e-06, + "loss": 1.2691, + "step": 265 + }, + { + "epoch": 0.01, + "learning_rate": 4.262820512820513e-06, + "loss": 1.184, + "step": 266 + }, + { + "epoch": 0.01, + "learning_rate": 4.278846153846154e-06, + "loss": 1.6006, + "step": 267 + }, + { + "epoch": 0.01, + "learning_rate": 4.294871794871795e-06, + "loss": 1.1442, + "step": 268 + }, + { + "epoch": 0.01, + "learning_rate": 4.310897435897436e-06, + "loss": 1.45, + "step": 269 + }, + { + "epoch": 0.01, + "learning_rate": 4.326923076923077e-06, + "loss": 1.0404, + "step": 270 + }, + { + "epoch": 0.01, + "learning_rate": 4.342948717948718e-06, + "loss": 1.2756, + "step": 271 + }, + { + "epoch": 0.01, + "learning_rate": 4.358974358974359e-06, + "loss": 1.3808, + "step": 272 + }, + { + "epoch": 0.01, + "learning_rate": 4.3750000000000005e-06, + "loss": 1.2727, + "step": 273 + }, + { + "epoch": 0.01, + "learning_rate": 4.3910256410256415e-06, + "loss": 1.4646, + "step": 274 + }, + { + "epoch": 0.01, + "learning_rate": 4.4070512820512826e-06, + "loss": 1.4188, + "step": 275 + }, + { + "epoch": 0.01, + "learning_rate": 4.423076923076924e-06, + "loss": 1.2114, + "step": 276 + }, + { + "epoch": 0.01, + "learning_rate": 4.439102564102565e-06, + "loss": 1.3336, + "step": 277 + }, + { + "epoch": 0.01, + "learning_rate": 4.455128205128206e-06, + "loss": 1.3193, + "step": 278 + }, + { + "epoch": 0.01, + "learning_rate": 4.471153846153847e-06, + "loss": 1.391, + "step": 279 + }, + { + "epoch": 0.01, + "learning_rate": 4.487179487179488e-06, + "loss": 1.3494, + "step": 280 + }, + { + "epoch": 0.01, + "learning_rate": 4.503205128205129e-06, + "loss": 1.1302, + "step": 281 + }, + { + "epoch": 0.01, + "learning_rate": 4.51923076923077e-06, + "loss": 1.1659, + "step": 282 + }, + { + "epoch": 0.01, + "learning_rate": 4.535256410256411e-06, + "loss": 1.254, + "step": 283 + }, + { + "epoch": 0.01, + "learning_rate": 4.551282051282052e-06, + "loss": 1.1251, + "step": 284 + }, + { + "epoch": 0.01, + "learning_rate": 4.567307692307692e-06, + "loss": 1.3989, + "step": 285 + }, + { + "epoch": 0.01, + "learning_rate": 4.583333333333333e-06, + "loss": 1.1763, + "step": 286 + }, + { + "epoch": 0.01, + "learning_rate": 4.599358974358974e-06, + "loss": 1.1381, + "step": 287 + }, + { + "epoch": 0.01, + "learning_rate": 4.615384615384616e-06, + "loss": 1.0707, + "step": 288 + }, + { + "epoch": 0.01, + "learning_rate": 4.631410256410257e-06, + "loss": 1.3445, + "step": 289 + }, + { + "epoch": 0.01, + "learning_rate": 4.647435897435898e-06, + "loss": 1.2192, + "step": 290 + }, + { + "epoch": 0.01, + "learning_rate": 4.663461538461539e-06, + "loss": 1.2312, + "step": 291 + }, + { + "epoch": 0.01, + "learning_rate": 4.6794871794871795e-06, + "loss": 1.0758, + "step": 292 + }, + { + "epoch": 0.01, + "learning_rate": 4.6955128205128205e-06, + "loss": 1.2916, + "step": 293 + }, + { + "epoch": 0.01, + "learning_rate": 4.711538461538462e-06, + "loss": 1.1983, + "step": 294 + }, + { + "epoch": 0.01, + "learning_rate": 4.727564102564103e-06, + "loss": 1.1122, + "step": 295 + }, + { + "epoch": 0.01, + "learning_rate": 4.743589743589744e-06, + "loss": 1.4513, + "step": 296 + }, + { + "epoch": 0.01, + "learning_rate": 4.759615384615385e-06, + "loss": 1.1572, + "step": 297 + }, + { + "epoch": 0.01, + "learning_rate": 4.775641025641027e-06, + "loss": 1.1511, + "step": 298 + }, + { + "epoch": 0.01, + "learning_rate": 4.791666666666668e-06, + "loss": 1.2424, + "step": 299 + }, + { + "epoch": 0.01, + "learning_rate": 4.807692307692308e-06, + "loss": 1.0537, + "step": 300 + }, + { + "epoch": 0.01, + "learning_rate": 4.823717948717949e-06, + "loss": 1.2127, + "step": 301 + }, + { + "epoch": 0.01, + "learning_rate": 4.83974358974359e-06, + "loss": 1.4414, + "step": 302 + }, + { + "epoch": 0.01, + "learning_rate": 4.855769230769231e-06, + "loss": 1.2005, + "step": 303 + }, + { + "epoch": 0.01, + "learning_rate": 4.871794871794872e-06, + "loss": 1.2645, + "step": 304 + }, + { + "epoch": 0.01, + "learning_rate": 4.887820512820513e-06, + "loss": 1.3248, + "step": 305 + }, + { + "epoch": 0.01, + "learning_rate": 4.903846153846154e-06, + "loss": 1.1797, + "step": 306 + }, + { + "epoch": 0.01, + "learning_rate": 4.919871794871795e-06, + "loss": 1.2666, + "step": 307 + }, + { + "epoch": 0.01, + "learning_rate": 4.935897435897436e-06, + "loss": 1.09, + "step": 308 + }, + { + "epoch": 0.01, + "learning_rate": 4.951923076923077e-06, + "loss": 1.2639, + "step": 309 + }, + { + "epoch": 0.01, + "learning_rate": 4.967948717948718e-06, + "loss": 1.253, + "step": 310 + }, + { + "epoch": 0.01, + "learning_rate": 4.983974358974359e-06, + "loss": 1.1904, + "step": 311 + }, + { + "epoch": 0.01, + "learning_rate": 5e-06, + "loss": 1.3708, + "step": 312 + }, + { + "epoch": 0.01, + "learning_rate": 5.0160256410256415e-06, + "loss": 1.308, + "step": 313 + }, + { + "epoch": 0.01, + "learning_rate": 5.0320512820512825e-06, + "loss": 0.9401, + "step": 314 + }, + { + "epoch": 0.01, + "learning_rate": 5.0480769230769235e-06, + "loss": 1.0321, + "step": 315 + }, + { + "epoch": 0.01, + "learning_rate": 5.064102564102565e-06, + "loss": 1.339, + "step": 316 + }, + { + "epoch": 0.01, + "learning_rate": 5.080128205128206e-06, + "loss": 1.3381, + "step": 317 + }, + { + "epoch": 0.01, + "learning_rate": 5.096153846153846e-06, + "loss": 1.3769, + "step": 318 + }, + { + "epoch": 0.01, + "learning_rate": 5.112179487179487e-06, + "loss": 1.3505, + "step": 319 + }, + { + "epoch": 0.01, + "learning_rate": 5.128205128205128e-06, + "loss": 1.2309, + "step": 320 + }, + { + "epoch": 0.01, + "learning_rate": 5.144230769230769e-06, + "loss": 1.1881, + "step": 321 + }, + { + "epoch": 0.01, + "learning_rate": 5.160256410256411e-06, + "loss": 1.4355, + "step": 322 + }, + { + "epoch": 0.01, + "learning_rate": 5.176282051282052e-06, + "loss": 1.2001, + "step": 323 + }, + { + "epoch": 0.01, + "learning_rate": 5.192307692307693e-06, + "loss": 1.3279, + "step": 324 + }, + { + "epoch": 0.01, + "learning_rate": 5.208333333333334e-06, + "loss": 1.2572, + "step": 325 + }, + { + "epoch": 0.01, + "learning_rate": 5.224358974358975e-06, + "loss": 1.5655, + "step": 326 + }, + { + "epoch": 0.01, + "learning_rate": 5.240384615384616e-06, + "loss": 1.1751, + "step": 327 + }, + { + "epoch": 0.01, + "learning_rate": 5.256410256410257e-06, + "loss": 1.4463, + "step": 328 + }, + { + "epoch": 0.01, + "learning_rate": 5.272435897435898e-06, + "loss": 1.3512, + "step": 329 + }, + { + "epoch": 0.01, + "learning_rate": 5.288461538461539e-06, + "loss": 1.2351, + "step": 330 + }, + { + "epoch": 0.01, + "learning_rate": 5.30448717948718e-06, + "loss": 1.2957, + "step": 331 + }, + { + "epoch": 0.01, + "learning_rate": 5.320512820512821e-06, + "loss": 1.2848, + "step": 332 + }, + { + "epoch": 0.01, + "learning_rate": 5.3365384615384615e-06, + "loss": 1.3063, + "step": 333 + }, + { + "epoch": 0.01, + "learning_rate": 5.3525641025641026e-06, + "loss": 1.3276, + "step": 334 + }, + { + "epoch": 0.01, + "learning_rate": 5.368589743589744e-06, + "loss": 1.1031, + "step": 335 + }, + { + "epoch": 0.01, + "learning_rate": 5.384615384615385e-06, + "loss": 1.2975, + "step": 336 + }, + { + "epoch": 0.01, + "learning_rate": 5.400641025641026e-06, + "loss": 1.2999, + "step": 337 + }, + { + "epoch": 0.01, + "learning_rate": 5.416666666666667e-06, + "loss": 1.3291, + "step": 338 + }, + { + "epoch": 0.01, + "learning_rate": 5.432692307692308e-06, + "loss": 1.2853, + "step": 339 + }, + { + "epoch": 0.01, + "learning_rate": 5.448717948717949e-06, + "loss": 1.3752, + "step": 340 + }, + { + "epoch": 0.01, + "learning_rate": 5.46474358974359e-06, + "loss": 1.1681, + "step": 341 + }, + { + "epoch": 0.01, + "learning_rate": 5.480769230769232e-06, + "loss": 1.5012, + "step": 342 + }, + { + "epoch": 0.01, + "learning_rate": 5.496794871794873e-06, + "loss": 1.3131, + "step": 343 + }, + { + "epoch": 0.01, + "learning_rate": 5.512820512820514e-06, + "loss": 1.1902, + "step": 344 + }, + { + "epoch": 0.01, + "learning_rate": 5.528846153846155e-06, + "loss": 1.1683, + "step": 345 + }, + { + "epoch": 0.01, + "learning_rate": 5.544871794871796e-06, + "loss": 1.5064, + "step": 346 + }, + { + "epoch": 0.01, + "learning_rate": 5.560897435897436e-06, + "loss": 1.5743, + "step": 347 + }, + { + "epoch": 0.01, + "learning_rate": 5.576923076923077e-06, + "loss": 1.5116, + "step": 348 + }, + { + "epoch": 0.01, + "learning_rate": 5.592948717948718e-06, + "loss": 1.0228, + "step": 349 + }, + { + "epoch": 0.01, + "learning_rate": 5.608974358974359e-06, + "loss": 1.4396, + "step": 350 + }, + { + "epoch": 0.01, + "learning_rate": 5.625e-06, + "loss": 1.2594, + "step": 351 + }, + { + "epoch": 0.01, + "learning_rate": 5.641025641025641e-06, + "loss": 1.0588, + "step": 352 + }, + { + "epoch": 0.01, + "learning_rate": 5.6570512820512824e-06, + "loss": 1.4938, + "step": 353 + }, + { + "epoch": 0.01, + "learning_rate": 5.6730769230769235e-06, + "loss": 0.9964, + "step": 354 + }, + { + "epoch": 0.01, + "learning_rate": 5.6891025641025645e-06, + "loss": 1.2989, + "step": 355 + }, + { + "epoch": 0.01, + "learning_rate": 5.705128205128206e-06, + "loss": 1.3335, + "step": 356 + }, + { + "epoch": 0.01, + "learning_rate": 5.721153846153847e-06, + "loss": 0.972, + "step": 357 + }, + { + "epoch": 0.01, + "learning_rate": 5.737179487179487e-06, + "loss": 1.0118, + "step": 358 + }, + { + "epoch": 0.01, + "learning_rate": 5.753205128205128e-06, + "loss": 1.0868, + "step": 359 + }, + { + "epoch": 0.01, + "learning_rate": 5.769230769230769e-06, + "loss": 1.1659, + "step": 360 + }, + { + "epoch": 0.01, + "learning_rate": 5.785256410256412e-06, + "loss": 1.138, + "step": 361 + }, + { + "epoch": 0.01, + "learning_rate": 5.801282051282052e-06, + "loss": 1.072, + "step": 362 + }, + { + "epoch": 0.01, + "learning_rate": 5.817307692307693e-06, + "loss": 1.2063, + "step": 363 + }, + { + "epoch": 0.01, + "learning_rate": 5.833333333333334e-06, + "loss": 1.3416, + "step": 364 + }, + { + "epoch": 0.01, + "learning_rate": 5.849358974358975e-06, + "loss": 1.1905, + "step": 365 + }, + { + "epoch": 0.01, + "learning_rate": 5.865384615384616e-06, + "loss": 1.0609, + "step": 366 + }, + { + "epoch": 0.01, + "learning_rate": 5.881410256410257e-06, + "loss": 1.2586, + "step": 367 + }, + { + "epoch": 0.01, + "learning_rate": 5.897435897435898e-06, + "loss": 1.2201, + "step": 368 + }, + { + "epoch": 0.01, + "learning_rate": 5.913461538461539e-06, + "loss": 1.3154, + "step": 369 + }, + { + "epoch": 0.01, + "learning_rate": 5.92948717948718e-06, + "loss": 1.218, + "step": 370 + }, + { + "epoch": 0.01, + "learning_rate": 5.945512820512821e-06, + "loss": 1.4568, + "step": 371 + }, + { + "epoch": 0.01, + "learning_rate": 5.961538461538462e-06, + "loss": 1.18, + "step": 372 + }, + { + "epoch": 0.01, + "learning_rate": 5.9775641025641025e-06, + "loss": 1.1976, + "step": 373 + }, + { + "epoch": 0.01, + "learning_rate": 5.9935897435897436e-06, + "loss": 1.1972, + "step": 374 + }, + { + "epoch": 0.01, + "learning_rate": 6.009615384615385e-06, + "loss": 1.1698, + "step": 375 + }, + { + "epoch": 0.01, + "learning_rate": 6.025641025641026e-06, + "loss": 1.4051, + "step": 376 + }, + { + "epoch": 0.01, + "learning_rate": 6.041666666666667e-06, + "loss": 1.2125, + "step": 377 + }, + { + "epoch": 0.01, + "learning_rate": 6.057692307692308e-06, + "loss": 1.0419, + "step": 378 + }, + { + "epoch": 0.01, + "learning_rate": 6.073717948717949e-06, + "loss": 1.0489, + "step": 379 + }, + { + "epoch": 0.01, + "learning_rate": 6.08974358974359e-06, + "loss": 1.2799, + "step": 380 + }, + { + "epoch": 0.01, + "learning_rate": 6.105769230769232e-06, + "loss": 1.4617, + "step": 381 + }, + { + "epoch": 0.01, + "learning_rate": 6.121794871794873e-06, + "loss": 1.2751, + "step": 382 + }, + { + "epoch": 0.01, + "learning_rate": 6.137820512820514e-06, + "loss": 1.2783, + "step": 383 + }, + { + "epoch": 0.01, + "learning_rate": 6.153846153846155e-06, + "loss": 1.0087, + "step": 384 + }, + { + "epoch": 0.01, + "learning_rate": 6.169871794871796e-06, + "loss": 1.3148, + "step": 385 + }, + { + "epoch": 0.01, + "learning_rate": 6.185897435897437e-06, + "loss": 1.3588, + "step": 386 + }, + { + "epoch": 0.01, + "learning_rate": 6.201923076923078e-06, + "loss": 1.33, + "step": 387 + }, + { + "epoch": 0.01, + "learning_rate": 6.217948717948718e-06, + "loss": 1.0655, + "step": 388 + }, + { + "epoch": 0.01, + "learning_rate": 6.233974358974359e-06, + "loss": 1.0166, + "step": 389 + }, + { + "epoch": 0.01, + "learning_rate": 6.25e-06, + "loss": 1.1426, + "step": 390 + }, + { + "epoch": 0.01, + "learning_rate": 6.266025641025641e-06, + "loss": 1.2688, + "step": 391 + }, + { + "epoch": 0.01, + "learning_rate": 6.282051282051282e-06, + "loss": 1.0576, + "step": 392 + }, + { + "epoch": 0.01, + "learning_rate": 6.2980769230769234e-06, + "loss": 1.0216, + "step": 393 + }, + { + "epoch": 0.01, + "learning_rate": 6.3141025641025645e-06, + "loss": 0.9943, + "step": 394 + }, + { + "epoch": 0.01, + "learning_rate": 6.3301282051282055e-06, + "loss": 1.0964, + "step": 395 + }, + { + "epoch": 0.01, + "learning_rate": 6.3461538461538466e-06, + "loss": 1.1716, + "step": 396 + }, + { + "epoch": 0.01, + "learning_rate": 6.362179487179488e-06, + "loss": 1.2416, + "step": 397 + }, + { + "epoch": 0.01, + "learning_rate": 6.378205128205129e-06, + "loss": 1.2919, + "step": 398 + }, + { + "epoch": 0.01, + "learning_rate": 6.394230769230769e-06, + "loss": 1.1502, + "step": 399 + }, + { + "epoch": 0.01, + "learning_rate": 6.410256410256412e-06, + "loss": 0.9034, + "step": 400 + }, + { + "epoch": 0.01, + "learning_rate": 6.426282051282053e-06, + "loss": 1.2893, + "step": 401 + }, + { + "epoch": 0.01, + "learning_rate": 6.442307692307693e-06, + "loss": 1.1716, + "step": 402 + }, + { + "epoch": 0.01, + "learning_rate": 6.458333333333334e-06, + "loss": 1.1029, + "step": 403 + }, + { + "epoch": 0.01, + "learning_rate": 6.474358974358975e-06, + "loss": 1.176, + "step": 404 + }, + { + "epoch": 0.01, + "learning_rate": 6.490384615384616e-06, + "loss": 1.0914, + "step": 405 + }, + { + "epoch": 0.01, + "learning_rate": 6.506410256410257e-06, + "loss": 1.3051, + "step": 406 + }, + { + "epoch": 0.01, + "learning_rate": 6.522435897435898e-06, + "loss": 1.262, + "step": 407 + }, + { + "epoch": 0.01, + "learning_rate": 6.538461538461539e-06, + "loss": 1.1453, + "step": 408 + }, + { + "epoch": 0.01, + "learning_rate": 6.55448717948718e-06, + "loss": 0.9694, + "step": 409 + }, + { + "epoch": 0.01, + "learning_rate": 6.570512820512821e-06, + "loss": 1.3627, + "step": 410 + }, + { + "epoch": 0.01, + "learning_rate": 6.586538461538462e-06, + "loss": 1.0662, + "step": 411 + }, + { + "epoch": 0.01, + "learning_rate": 6.602564102564103e-06, + "loss": 1.1889, + "step": 412 + }, + { + "epoch": 0.01, + "learning_rate": 6.6185897435897435e-06, + "loss": 1.3712, + "step": 413 + }, + { + "epoch": 0.01, + "learning_rate": 6.6346153846153846e-06, + "loss": 1.0901, + "step": 414 + }, + { + "epoch": 0.01, + "learning_rate": 6.650641025641026e-06, + "loss": 1.077, + "step": 415 + }, + { + "epoch": 0.01, + "learning_rate": 6.666666666666667e-06, + "loss": 1.2691, + "step": 416 + }, + { + "epoch": 0.01, + "learning_rate": 6.682692307692308e-06, + "loss": 1.1895, + "step": 417 + }, + { + "epoch": 0.01, + "learning_rate": 6.698717948717949e-06, + "loss": 1.0688, + "step": 418 + }, + { + "epoch": 0.01, + "learning_rate": 6.71474358974359e-06, + "loss": 1.0427, + "step": 419 + }, + { + "epoch": 0.01, + "learning_rate": 6.730769230769232e-06, + "loss": 1.0806, + "step": 420 + }, + { + "epoch": 0.01, + "learning_rate": 6.746794871794873e-06, + "loss": 1.3337, + "step": 421 + }, + { + "epoch": 0.01, + "learning_rate": 6.762820512820514e-06, + "loss": 1.1585, + "step": 422 + }, + { + "epoch": 0.01, + "learning_rate": 6.778846153846155e-06, + "loss": 1.2199, + "step": 423 + }, + { + "epoch": 0.01, + "learning_rate": 6.794871794871796e-06, + "loss": 1.1405, + "step": 424 + }, + { + "epoch": 0.01, + "learning_rate": 6.810897435897437e-06, + "loss": 1.1994, + "step": 425 + }, + { + "epoch": 0.01, + "learning_rate": 6.826923076923078e-06, + "loss": 1.1168, + "step": 426 + }, + { + "epoch": 0.01, + "learning_rate": 6.842948717948719e-06, + "loss": 1.0426, + "step": 427 + }, + { + "epoch": 0.01, + "learning_rate": 6.858974358974359e-06, + "loss": 1.1563, + "step": 428 + }, + { + "epoch": 0.01, + "learning_rate": 6.875e-06, + "loss": 1.206, + "step": 429 + }, + { + "epoch": 0.01, + "learning_rate": 6.891025641025641e-06, + "loss": 1.1217, + "step": 430 + }, + { + "epoch": 0.01, + "learning_rate": 6.907051282051282e-06, + "loss": 1.3817, + "step": 431 + }, + { + "epoch": 0.01, + "learning_rate": 6.923076923076923e-06, + "loss": 1.2151, + "step": 432 + }, + { + "epoch": 0.01, + "learning_rate": 6.9391025641025644e-06, + "loss": 1.0752, + "step": 433 + }, + { + "epoch": 0.01, + "learning_rate": 6.9551282051282055e-06, + "loss": 1.2786, + "step": 434 + }, + { + "epoch": 0.01, + "learning_rate": 6.9711538461538465e-06, + "loss": 1.2442, + "step": 435 + }, + { + "epoch": 0.01, + "learning_rate": 6.9871794871794876e-06, + "loss": 0.9404, + "step": 436 + }, + { + "epoch": 0.01, + "learning_rate": 7.003205128205129e-06, + "loss": 1.2063, + "step": 437 + }, + { + "epoch": 0.01, + "learning_rate": 7.01923076923077e-06, + "loss": 1.2681, + "step": 438 + }, + { + "epoch": 0.01, + "learning_rate": 7.0352564102564116e-06, + "loss": 1.215, + "step": 439 + }, + { + "epoch": 0.01, + "learning_rate": 7.051282051282053e-06, + "loss": 1.2347, + "step": 440 + }, + { + "epoch": 0.01, + "learning_rate": 7.067307692307694e-06, + "loss": 1.2478, + "step": 441 + }, + { + "epoch": 0.01, + "learning_rate": 7.083333333333335e-06, + "loss": 1.2075, + "step": 442 + }, + { + "epoch": 0.01, + "learning_rate": 7.099358974358975e-06, + "loss": 1.0692, + "step": 443 + }, + { + "epoch": 0.01, + "learning_rate": 7.115384615384616e-06, + "loss": 1.059, + "step": 444 + }, + { + "epoch": 0.01, + "learning_rate": 7.131410256410257e-06, + "loss": 1.1322, + "step": 445 + }, + { + "epoch": 0.01, + "learning_rate": 7.147435897435898e-06, + "loss": 1.3712, + "step": 446 + }, + { + "epoch": 0.01, + "learning_rate": 7.163461538461539e-06, + "loss": 1.2223, + "step": 447 + }, + { + "epoch": 0.01, + "learning_rate": 7.17948717948718e-06, + "loss": 1.1803, + "step": 448 + }, + { + "epoch": 0.01, + "learning_rate": 7.195512820512821e-06, + "loss": 1.1191, + "step": 449 + }, + { + "epoch": 0.01, + "learning_rate": 7.211538461538462e-06, + "loss": 1.1959, + "step": 450 + }, + { + "epoch": 0.01, + "learning_rate": 7.227564102564103e-06, + "loss": 1.0533, + "step": 451 + }, + { + "epoch": 0.01, + "learning_rate": 7.243589743589744e-06, + "loss": 1.0127, + "step": 452 + }, + { + "epoch": 0.01, + "learning_rate": 7.259615384615385e-06, + "loss": 1.2319, + "step": 453 + }, + { + "epoch": 0.01, + "learning_rate": 7.2756410256410255e-06, + "loss": 1.2411, + "step": 454 + }, + { + "epoch": 0.01, + "learning_rate": 7.291666666666667e-06, + "loss": 1.3419, + "step": 455 + }, + { + "epoch": 0.01, + "learning_rate": 7.307692307692308e-06, + "loss": 1.1042, + "step": 456 + }, + { + "epoch": 0.01, + "learning_rate": 7.323717948717949e-06, + "loss": 1.1138, + "step": 457 + }, + { + "epoch": 0.01, + "learning_rate": 7.33974358974359e-06, + "loss": 1.3629, + "step": 458 + }, + { + "epoch": 0.01, + "learning_rate": 7.355769230769232e-06, + "loss": 1.0237, + "step": 459 + }, + { + "epoch": 0.01, + "learning_rate": 7.371794871794873e-06, + "loss": 1.2258, + "step": 460 + }, + { + "epoch": 0.01, + "learning_rate": 7.387820512820514e-06, + "loss": 1.1482, + "step": 461 + }, + { + "epoch": 0.01, + "learning_rate": 7.403846153846155e-06, + "loss": 1.3115, + "step": 462 + }, + { + "epoch": 0.01, + "learning_rate": 7.419871794871796e-06, + "loss": 0.9838, + "step": 463 + }, + { + "epoch": 0.01, + "learning_rate": 7.435897435897437e-06, + "loss": 1.0598, + "step": 464 + }, + { + "epoch": 0.01, + "learning_rate": 7.451923076923078e-06, + "loss": 1.0941, + "step": 465 + }, + { + "epoch": 0.01, + "learning_rate": 7.467948717948719e-06, + "loss": 0.9008, + "step": 466 + }, + { + "epoch": 0.01, + "learning_rate": 7.48397435897436e-06, + "loss": 1.1612, + "step": 467 + }, + { + "epoch": 0.01, + "learning_rate": 7.500000000000001e-06, + "loss": 1.3597, + "step": 468 + }, + { + "epoch": 0.01, + "learning_rate": 7.516025641025641e-06, + "loss": 0.9616, + "step": 469 + }, + { + "epoch": 0.01, + "learning_rate": 7.532051282051282e-06, + "loss": 1.1248, + "step": 470 + }, + { + "epoch": 0.01, + "learning_rate": 7.548076923076923e-06, + "loss": 1.4105, + "step": 471 + }, + { + "epoch": 0.01, + "learning_rate": 7.564102564102564e-06, + "loss": 1.0354, + "step": 472 + }, + { + "epoch": 0.01, + "learning_rate": 7.580128205128205e-06, + "loss": 1.449, + "step": 473 + }, + { + "epoch": 0.01, + "learning_rate": 7.5961538461538465e-06, + "loss": 1.3207, + "step": 474 + }, + { + "epoch": 0.01, + "learning_rate": 7.6121794871794875e-06, + "loss": 1.4124, + "step": 475 + }, + { + "epoch": 0.01, + "learning_rate": 7.6282051282051286e-06, + "loss": 1.2421, + "step": 476 + }, + { + "epoch": 0.01, + "learning_rate": 7.64423076923077e-06, + "loss": 1.1161, + "step": 477 + }, + { + "epoch": 0.01, + "learning_rate": 7.660256410256411e-06, + "loss": 1.3364, + "step": 478 + }, + { + "epoch": 0.01, + "learning_rate": 7.676282051282052e-06, + "loss": 1.2533, + "step": 479 + }, + { + "epoch": 0.01, + "learning_rate": 7.692307692307694e-06, + "loss": 1.0699, + "step": 480 + }, + { + "epoch": 0.01, + "learning_rate": 7.708333333333334e-06, + "loss": 1.0422, + "step": 481 + }, + { + "epoch": 0.01, + "learning_rate": 7.724358974358976e-06, + "loss": 1.3994, + "step": 482 + }, + { + "epoch": 0.01, + "learning_rate": 7.740384615384616e-06, + "loss": 1.4755, + "step": 483 + }, + { + "epoch": 0.01, + "learning_rate": 7.756410256410258e-06, + "loss": 1.2507, + "step": 484 + }, + { + "epoch": 0.01, + "learning_rate": 7.772435897435898e-06, + "loss": 1.099, + "step": 485 + }, + { + "epoch": 0.01, + "learning_rate": 7.78846153846154e-06, + "loss": 1.4083, + "step": 486 + }, + { + "epoch": 0.01, + "learning_rate": 7.80448717948718e-06, + "loss": 1.2108, + "step": 487 + }, + { + "epoch": 0.01, + "learning_rate": 7.820512820512822e-06, + "loss": 1.0294, + "step": 488 + }, + { + "epoch": 0.01, + "learning_rate": 7.836538461538462e-06, + "loss": 1.14, + "step": 489 + }, + { + "epoch": 0.01, + "learning_rate": 7.852564102564102e-06, + "loss": 1.1175, + "step": 490 + }, + { + "epoch": 0.01, + "learning_rate": 7.868589743589744e-06, + "loss": 1.3487, + "step": 491 + }, + { + "epoch": 0.01, + "learning_rate": 7.884615384615384e-06, + "loss": 1.2527, + "step": 492 + }, + { + "epoch": 0.01, + "learning_rate": 7.900641025641026e-06, + "loss": 1.3033, + "step": 493 + }, + { + "epoch": 0.01, + "learning_rate": 7.916666666666667e-06, + "loss": 0.9275, + "step": 494 + }, + { + "epoch": 0.01, + "learning_rate": 7.932692307692308e-06, + "loss": 1.162, + "step": 495 + }, + { + "epoch": 0.01, + "learning_rate": 7.948717948717949e-06, + "loss": 1.2722, + "step": 496 + }, + { + "epoch": 0.01, + "learning_rate": 7.96474358974359e-06, + "loss": 0.9354, + "step": 497 + }, + { + "epoch": 0.01, + "learning_rate": 7.980769230769232e-06, + "loss": 1.0671, + "step": 498 + }, + { + "epoch": 0.01, + "learning_rate": 7.996794871794873e-06, + "loss": 1.1206, + "step": 499 + }, + { + "epoch": 0.01, + "learning_rate": 8.012820512820515e-06, + "loss": 1.1491, + "step": 500 + }, + { + "epoch": 0.01, + "learning_rate": 8.028846153846155e-06, + "loss": 1.0781, + "step": 501 + }, + { + "epoch": 0.01, + "learning_rate": 8.044871794871797e-06, + "loss": 1.1764, + "step": 502 + }, + { + "epoch": 0.01, + "learning_rate": 8.060897435897437e-06, + "loss": 1.4623, + "step": 503 + }, + { + "epoch": 0.01, + "learning_rate": 8.076923076923077e-06, + "loss": 0.982, + "step": 504 + }, + { + "epoch": 0.01, + "learning_rate": 8.092948717948719e-06, + "loss": 1.3779, + "step": 505 + }, + { + "epoch": 0.01, + "learning_rate": 8.108974358974359e-06, + "loss": 1.0912, + "step": 506 + }, + { + "epoch": 0.01, + "learning_rate": 8.125000000000001e-06, + "loss": 1.4112, + "step": 507 + }, + { + "epoch": 0.01, + "learning_rate": 8.141025641025641e-06, + "loss": 1.2558, + "step": 508 + }, + { + "epoch": 0.01, + "learning_rate": 8.157051282051283e-06, + "loss": 0.9946, + "step": 509 + }, + { + "epoch": 0.01, + "learning_rate": 8.173076923076923e-06, + "loss": 1.413, + "step": 510 + }, + { + "epoch": 0.01, + "learning_rate": 8.189102564102565e-06, + "loss": 1.2679, + "step": 511 + }, + { + "epoch": 0.01, + "learning_rate": 8.205128205128205e-06, + "loss": 1.2314, + "step": 512 + }, + { + "epoch": 0.01, + "learning_rate": 8.221153846153847e-06, + "loss": 1.1047, + "step": 513 + }, + { + "epoch": 0.01, + "learning_rate": 8.237179487179487e-06, + "loss": 1.0648, + "step": 514 + }, + { + "epoch": 0.01, + "learning_rate": 8.253205128205128e-06, + "loss": 1.2759, + "step": 515 + }, + { + "epoch": 0.01, + "learning_rate": 8.26923076923077e-06, + "loss": 1.4021, + "step": 516 + }, + { + "epoch": 0.01, + "learning_rate": 8.285256410256411e-06, + "loss": 1.0899, + "step": 517 + }, + { + "epoch": 0.01, + "learning_rate": 8.301282051282052e-06, + "loss": 1.0388, + "step": 518 + }, + { + "epoch": 0.01, + "learning_rate": 8.317307692307694e-06, + "loss": 1.1325, + "step": 519 + }, + { + "epoch": 0.01, + "learning_rate": 8.333333333333334e-06, + "loss": 1.1843, + "step": 520 + }, + { + "epoch": 0.01, + "learning_rate": 8.349358974358976e-06, + "loss": 1.3203, + "step": 521 + }, + { + "epoch": 0.01, + "learning_rate": 8.365384615384616e-06, + "loss": 0.9634, + "step": 522 + }, + { + "epoch": 0.01, + "learning_rate": 8.381410256410258e-06, + "loss": 1.1196, + "step": 523 + }, + { + "epoch": 0.01, + "learning_rate": 8.397435897435898e-06, + "loss": 1.1965, + "step": 524 + }, + { + "epoch": 0.01, + "learning_rate": 8.41346153846154e-06, + "loss": 1.149, + "step": 525 + }, + { + "epoch": 0.01, + "learning_rate": 8.42948717948718e-06, + "loss": 1.2187, + "step": 526 + }, + { + "epoch": 0.01, + "learning_rate": 8.445512820512822e-06, + "loss": 1.1607, + "step": 527 + }, + { + "epoch": 0.01, + "learning_rate": 8.461538461538462e-06, + "loss": 1.2259, + "step": 528 + }, + { + "epoch": 0.01, + "learning_rate": 8.477564102564102e-06, + "loss": 1.3048, + "step": 529 + }, + { + "epoch": 0.01, + "learning_rate": 8.493589743589744e-06, + "loss": 1.316, + "step": 530 + }, + { + "epoch": 0.01, + "learning_rate": 8.509615384615384e-06, + "loss": 1.1058, + "step": 531 + }, + { + "epoch": 0.01, + "learning_rate": 8.525641025641026e-06, + "loss": 1.397, + "step": 532 + }, + { + "epoch": 0.01, + "learning_rate": 8.541666666666666e-06, + "loss": 1.2585, + "step": 533 + }, + { + "epoch": 0.01, + "learning_rate": 8.557692307692308e-06, + "loss": 0.9887, + "step": 534 + }, + { + "epoch": 0.01, + "learning_rate": 8.573717948717949e-06, + "loss": 1.1306, + "step": 535 + }, + { + "epoch": 0.01, + "learning_rate": 8.58974358974359e-06, + "loss": 1.2361, + "step": 536 + }, + { + "epoch": 0.01, + "learning_rate": 8.605769230769232e-06, + "loss": 1.1534, + "step": 537 + }, + { + "epoch": 0.01, + "learning_rate": 8.621794871794873e-06, + "loss": 1.1039, + "step": 538 + }, + { + "epoch": 0.01, + "learning_rate": 8.637820512820514e-06, + "loss": 1.2351, + "step": 539 + }, + { + "epoch": 0.01, + "learning_rate": 8.653846153846155e-06, + "loss": 1.2716, + "step": 540 + }, + { + "epoch": 0.01, + "learning_rate": 8.669871794871797e-06, + "loss": 0.965, + "step": 541 + }, + { + "epoch": 0.01, + "learning_rate": 8.685897435897437e-06, + "loss": 1.3699, + "step": 542 + }, + { + "epoch": 0.01, + "learning_rate": 8.701923076923079e-06, + "loss": 1.1236, + "step": 543 + }, + { + "epoch": 0.01, + "learning_rate": 8.717948717948719e-06, + "loss": 1.2112, + "step": 544 + }, + { + "epoch": 0.01, + "learning_rate": 8.733974358974359e-06, + "loss": 1.0647, + "step": 545 + }, + { + "epoch": 0.01, + "learning_rate": 8.750000000000001e-06, + "loss": 1.3791, + "step": 546 + }, + { + "epoch": 0.01, + "learning_rate": 8.766025641025641e-06, + "loss": 1.292, + "step": 547 + }, + { + "epoch": 0.01, + "learning_rate": 8.782051282051283e-06, + "loss": 1.0585, + "step": 548 + }, + { + "epoch": 0.01, + "learning_rate": 8.798076923076923e-06, + "loss": 1.2454, + "step": 549 + }, + { + "epoch": 0.01, + "learning_rate": 8.814102564102565e-06, + "loss": 1.2955, + "step": 550 + }, + { + "epoch": 0.01, + "learning_rate": 8.830128205128205e-06, + "loss": 1.3526, + "step": 551 + }, + { + "epoch": 0.01, + "learning_rate": 8.846153846153847e-06, + "loss": 1.3597, + "step": 552 + }, + { + "epoch": 0.01, + "learning_rate": 8.862179487179487e-06, + "loss": 1.2796, + "step": 553 + }, + { + "epoch": 0.01, + "learning_rate": 8.87820512820513e-06, + "loss": 1.153, + "step": 554 + }, + { + "epoch": 0.01, + "learning_rate": 8.89423076923077e-06, + "loss": 1.1938, + "step": 555 + }, + { + "epoch": 0.01, + "learning_rate": 8.910256410256411e-06, + "loss": 1.11, + "step": 556 + }, + { + "epoch": 0.01, + "learning_rate": 8.926282051282053e-06, + "loss": 1.0653, + "step": 557 + }, + { + "epoch": 0.01, + "learning_rate": 8.942307692307693e-06, + "loss": 1.1155, + "step": 558 + }, + { + "epoch": 0.01, + "learning_rate": 8.958333333333334e-06, + "loss": 1.153, + "step": 559 + }, + { + "epoch": 0.01, + "learning_rate": 8.974358974358976e-06, + "loss": 1.2388, + "step": 560 + }, + { + "epoch": 0.01, + "learning_rate": 8.990384615384616e-06, + "loss": 1.2952, + "step": 561 + }, + { + "epoch": 0.01, + "learning_rate": 9.006410256410258e-06, + "loss": 0.9649, + "step": 562 + }, + { + "epoch": 0.01, + "learning_rate": 9.022435897435898e-06, + "loss": 1.287, + "step": 563 + }, + { + "epoch": 0.01, + "learning_rate": 9.03846153846154e-06, + "loss": 1.2681, + "step": 564 + }, + { + "epoch": 0.01, + "learning_rate": 9.05448717948718e-06, + "loss": 1.1516, + "step": 565 + }, + { + "epoch": 0.01, + "learning_rate": 9.070512820512822e-06, + "loss": 1.1678, + "step": 566 + }, + { + "epoch": 0.01, + "learning_rate": 9.086538461538462e-06, + "loss": 1.1312, + "step": 567 + }, + { + "epoch": 0.01, + "learning_rate": 9.102564102564104e-06, + "loss": 1.3251, + "step": 568 + }, + { + "epoch": 0.01, + "learning_rate": 9.118589743589744e-06, + "loss": 1.3024, + "step": 569 + }, + { + "epoch": 0.01, + "learning_rate": 9.134615384615384e-06, + "loss": 1.2942, + "step": 570 + }, + { + "epoch": 0.01, + "learning_rate": 9.150641025641026e-06, + "loss": 1.3577, + "step": 571 + }, + { + "epoch": 0.01, + "learning_rate": 9.166666666666666e-06, + "loss": 1.2206, + "step": 572 + }, + { + "epoch": 0.01, + "learning_rate": 9.182692307692308e-06, + "loss": 1.0085, + "step": 573 + }, + { + "epoch": 0.01, + "learning_rate": 9.198717948717949e-06, + "loss": 1.4219, + "step": 574 + }, + { + "epoch": 0.01, + "learning_rate": 9.21474358974359e-06, + "loss": 1.1577, + "step": 575 + }, + { + "epoch": 0.01, + "learning_rate": 9.230769230769232e-06, + "loss": 0.8437, + "step": 576 + }, + { + "epoch": 0.01, + "learning_rate": 9.246794871794873e-06, + "loss": 1.2896, + "step": 577 + }, + { + "epoch": 0.01, + "learning_rate": 9.262820512820514e-06, + "loss": 1.0145, + "step": 578 + }, + { + "epoch": 0.01, + "learning_rate": 9.278846153846155e-06, + "loss": 1.1443, + "step": 579 + }, + { + "epoch": 0.01, + "learning_rate": 9.294871794871796e-06, + "loss": 1.27, + "step": 580 + }, + { + "epoch": 0.01, + "learning_rate": 9.310897435897437e-06, + "loss": 1.2551, + "step": 581 + }, + { + "epoch": 0.01, + "learning_rate": 9.326923076923079e-06, + "loss": 1.2958, + "step": 582 + }, + { + "epoch": 0.01, + "learning_rate": 9.342948717948719e-06, + "loss": 1.2024, + "step": 583 + }, + { + "epoch": 0.01, + "learning_rate": 9.358974358974359e-06, + "loss": 1.197, + "step": 584 + }, + { + "epoch": 0.01, + "learning_rate": 9.375000000000001e-06, + "loss": 1.2403, + "step": 585 + }, + { + "epoch": 0.01, + "learning_rate": 9.391025641025641e-06, + "loss": 1.2835, + "step": 586 + }, + { + "epoch": 0.01, + "learning_rate": 9.407051282051283e-06, + "loss": 1.2071, + "step": 587 + }, + { + "epoch": 0.01, + "learning_rate": 9.423076923076923e-06, + "loss": 1.2126, + "step": 588 + }, + { + "epoch": 0.01, + "learning_rate": 9.439102564102565e-06, + "loss": 1.2585, + "step": 589 + }, + { + "epoch": 0.01, + "learning_rate": 9.455128205128205e-06, + "loss": 1.1184, + "step": 590 + }, + { + "epoch": 0.01, + "learning_rate": 9.471153846153847e-06, + "loss": 1.3123, + "step": 591 + }, + { + "epoch": 0.01, + "learning_rate": 9.487179487179487e-06, + "loss": 1.1937, + "step": 592 + }, + { + "epoch": 0.01, + "learning_rate": 9.50320512820513e-06, + "loss": 1.1913, + "step": 593 + }, + { + "epoch": 0.01, + "learning_rate": 9.51923076923077e-06, + "loss": 1.1435, + "step": 594 + }, + { + "epoch": 0.01, + "learning_rate": 9.535256410256411e-06, + "loss": 1.2038, + "step": 595 + }, + { + "epoch": 0.01, + "learning_rate": 9.551282051282053e-06, + "loss": 1.5369, + "step": 596 + }, + { + "epoch": 0.01, + "learning_rate": 9.567307692307693e-06, + "loss": 1.0098, + "step": 597 + }, + { + "epoch": 0.01, + "learning_rate": 9.583333333333335e-06, + "loss": 1.2558, + "step": 598 + }, + { + "epoch": 0.01, + "learning_rate": 9.599358974358976e-06, + "loss": 1.3229, + "step": 599 + }, + { + "epoch": 0.01, + "learning_rate": 9.615384615384616e-06, + "loss": 1.2014, + "step": 600 + }, + { + "epoch": 0.01, + "learning_rate": 9.631410256410258e-06, + "loss": 1.3267, + "step": 601 + }, + { + "epoch": 0.01, + "learning_rate": 9.647435897435898e-06, + "loss": 1.2633, + "step": 602 + }, + { + "epoch": 0.01, + "learning_rate": 9.66346153846154e-06, + "loss": 1.3408, + "step": 603 + }, + { + "epoch": 0.01, + "learning_rate": 9.67948717948718e-06, + "loss": 1.0763, + "step": 604 + }, + { + "epoch": 0.01, + "learning_rate": 9.695512820512822e-06, + "loss": 1.1176, + "step": 605 + }, + { + "epoch": 0.01, + "learning_rate": 9.711538461538462e-06, + "loss": 1.1935, + "step": 606 + }, + { + "epoch": 0.01, + "learning_rate": 9.727564102564104e-06, + "loss": 1.0024, + "step": 607 + }, + { + "epoch": 0.01, + "learning_rate": 9.743589743589744e-06, + "loss": 1.1876, + "step": 608 + }, + { + "epoch": 0.01, + "learning_rate": 9.759615384615386e-06, + "loss": 1.0333, + "step": 609 + }, + { + "epoch": 0.01, + "learning_rate": 9.775641025641026e-06, + "loss": 1.3527, + "step": 610 + }, + { + "epoch": 0.01, + "learning_rate": 9.791666666666666e-06, + "loss": 1.3001, + "step": 611 + }, + { + "epoch": 0.01, + "learning_rate": 9.807692307692308e-06, + "loss": 1.2325, + "step": 612 + }, + { + "epoch": 0.01, + "learning_rate": 9.823717948717948e-06, + "loss": 1.4669, + "step": 613 + }, + { + "epoch": 0.01, + "learning_rate": 9.83974358974359e-06, + "loss": 1.1053, + "step": 614 + }, + { + "epoch": 0.01, + "learning_rate": 9.855769230769232e-06, + "loss": 1.2874, + "step": 615 + }, + { + "epoch": 0.01, + "learning_rate": 9.871794871794872e-06, + "loss": 0.9135, + "step": 616 + }, + { + "epoch": 0.01, + "learning_rate": 9.887820512820514e-06, + "loss": 1.3558, + "step": 617 + }, + { + "epoch": 0.01, + "learning_rate": 9.903846153846155e-06, + "loss": 1.292, + "step": 618 + }, + { + "epoch": 0.01, + "learning_rate": 9.919871794871796e-06, + "loss": 1.2289, + "step": 619 + }, + { + "epoch": 0.01, + "learning_rate": 9.935897435897437e-06, + "loss": 1.0531, + "step": 620 + }, + { + "epoch": 0.01, + "learning_rate": 9.951923076923079e-06, + "loss": 1.0696, + "step": 621 + }, + { + "epoch": 0.01, + "learning_rate": 9.967948717948719e-06, + "loss": 0.8705, + "step": 622 + }, + { + "epoch": 0.01, + "learning_rate": 9.98397435897436e-06, + "loss": 1.0606, + "step": 623 + }, + { + "epoch": 0.02, + "learning_rate": 1e-05, + "loss": 1.2437, + "step": 624 + }, + { + "epoch": 0.02, + "learning_rate": 1.0016025641025641e-05, + "loss": 1.0472, + "step": 625 + }, + { + "epoch": 0.02, + "learning_rate": 1.0032051282051283e-05, + "loss": 1.1461, + "step": 626 + }, + { + "epoch": 0.02, + "learning_rate": 1.0048076923076923e-05, + "loss": 1.3593, + "step": 627 + }, + { + "epoch": 0.02, + "learning_rate": 1.0064102564102565e-05, + "loss": 1.0608, + "step": 628 + }, + { + "epoch": 0.02, + "learning_rate": 1.0080128205128205e-05, + "loss": 1.2378, + "step": 629 + }, + { + "epoch": 0.02, + "learning_rate": 1.0096153846153847e-05, + "loss": 1.3029, + "step": 630 + }, + { + "epoch": 0.02, + "learning_rate": 1.0112179487179487e-05, + "loss": 1.0011, + "step": 631 + }, + { + "epoch": 0.02, + "learning_rate": 1.012820512820513e-05, + "loss": 1.1237, + "step": 632 + }, + { + "epoch": 0.02, + "learning_rate": 1.014423076923077e-05, + "loss": 1.1889, + "step": 633 + }, + { + "epoch": 0.02, + "learning_rate": 1.0160256410256411e-05, + "loss": 1.0168, + "step": 634 + }, + { + "epoch": 0.02, + "learning_rate": 1.0176282051282051e-05, + "loss": 1.1731, + "step": 635 + }, + { + "epoch": 0.02, + "learning_rate": 1.0192307692307692e-05, + "loss": 1.2264, + "step": 636 + }, + { + "epoch": 0.02, + "learning_rate": 1.0208333333333334e-05, + "loss": 1.072, + "step": 637 + }, + { + "epoch": 0.02, + "learning_rate": 1.0224358974358974e-05, + "loss": 1.1968, + "step": 638 + }, + { + "epoch": 0.02, + "learning_rate": 1.0240384615384616e-05, + "loss": 1.2545, + "step": 639 + }, + { + "epoch": 0.02, + "learning_rate": 1.0256410256410256e-05, + "loss": 1.1809, + "step": 640 + }, + { + "epoch": 0.02, + "learning_rate": 1.0272435897435898e-05, + "loss": 1.2892, + "step": 641 + }, + { + "epoch": 0.02, + "learning_rate": 1.0288461538461538e-05, + "loss": 1.1453, + "step": 642 + }, + { + "epoch": 0.02, + "learning_rate": 1.030448717948718e-05, + "loss": 1.1241, + "step": 643 + }, + { + "epoch": 0.02, + "learning_rate": 1.0320512820512822e-05, + "loss": 1.1201, + "step": 644 + }, + { + "epoch": 0.02, + "learning_rate": 1.0336538461538464e-05, + "loss": 1.0595, + "step": 645 + }, + { + "epoch": 0.02, + "learning_rate": 1.0352564102564104e-05, + "loss": 1.2147, + "step": 646 + }, + { + "epoch": 0.02, + "learning_rate": 1.0368589743589746e-05, + "loss": 1.3046, + "step": 647 + }, + { + "epoch": 0.02, + "learning_rate": 1.0384615384615386e-05, + "loss": 1.2298, + "step": 648 + }, + { + "epoch": 0.02, + "learning_rate": 1.0400641025641028e-05, + "loss": 1.155, + "step": 649 + }, + { + "epoch": 0.02, + "learning_rate": 1.0416666666666668e-05, + "loss": 1.3139, + "step": 650 + }, + { + "epoch": 0.02, + "learning_rate": 1.043269230769231e-05, + "loss": 1.2136, + "step": 651 + }, + { + "epoch": 0.02, + "learning_rate": 1.044871794871795e-05, + "loss": 1.2458, + "step": 652 + }, + { + "epoch": 0.02, + "learning_rate": 1.0464743589743592e-05, + "loss": 1.2306, + "step": 653 + }, + { + "epoch": 0.02, + "learning_rate": 1.0480769230769232e-05, + "loss": 1.1397, + "step": 654 + }, + { + "epoch": 0.02, + "learning_rate": 1.0496794871794872e-05, + "loss": 0.8854, + "step": 655 + }, + { + "epoch": 0.02, + "learning_rate": 1.0512820512820514e-05, + "loss": 1.1736, + "step": 656 + }, + { + "epoch": 0.02, + "learning_rate": 1.0528846153846154e-05, + "loss": 1.322, + "step": 657 + }, + { + "epoch": 0.02, + "learning_rate": 1.0544871794871796e-05, + "loss": 1.3417, + "step": 658 + }, + { + "epoch": 0.02, + "learning_rate": 1.0560897435897437e-05, + "loss": 1.1573, + "step": 659 + }, + { + "epoch": 0.02, + "learning_rate": 1.0576923076923078e-05, + "loss": 1.3269, + "step": 660 + }, + { + "epoch": 0.02, + "learning_rate": 1.0592948717948719e-05, + "loss": 0.9123, + "step": 661 + }, + { + "epoch": 0.02, + "learning_rate": 1.060897435897436e-05, + "loss": 1.2631, + "step": 662 + }, + { + "epoch": 0.02, + "learning_rate": 1.0625e-05, + "loss": 1.1752, + "step": 663 + }, + { + "epoch": 0.02, + "learning_rate": 1.0641025641025643e-05, + "loss": 0.9409, + "step": 664 + }, + { + "epoch": 0.02, + "learning_rate": 1.0657051282051283e-05, + "loss": 1.3452, + "step": 665 + }, + { + "epoch": 0.02, + "learning_rate": 1.0673076923076923e-05, + "loss": 1.1283, + "step": 666 + }, + { + "epoch": 0.02, + "learning_rate": 1.0689102564102565e-05, + "loss": 1.0247, + "step": 667 + }, + { + "epoch": 0.02, + "learning_rate": 1.0705128205128205e-05, + "loss": 0.9357, + "step": 668 + }, + { + "epoch": 0.02, + "learning_rate": 1.0721153846153847e-05, + "loss": 0.9421, + "step": 669 + }, + { + "epoch": 0.02, + "learning_rate": 1.0737179487179487e-05, + "loss": 1.1709, + "step": 670 + }, + { + "epoch": 0.02, + "learning_rate": 1.0753205128205129e-05, + "loss": 1.1868, + "step": 671 + }, + { + "epoch": 0.02, + "learning_rate": 1.076923076923077e-05, + "loss": 1.2959, + "step": 672 + }, + { + "epoch": 0.02, + "learning_rate": 1.0785256410256411e-05, + "loss": 1.0635, + "step": 673 + }, + { + "epoch": 0.02, + "learning_rate": 1.0801282051282051e-05, + "loss": 1.2493, + "step": 674 + }, + { + "epoch": 0.02, + "learning_rate": 1.0817307692307693e-05, + "loss": 1.2343, + "step": 675 + }, + { + "epoch": 0.02, + "learning_rate": 1.0833333333333334e-05, + "loss": 1.0217, + "step": 676 + }, + { + "epoch": 0.02, + "learning_rate": 1.0849358974358974e-05, + "loss": 1.0055, + "step": 677 + }, + { + "epoch": 0.02, + "learning_rate": 1.0865384615384616e-05, + "loss": 0.915, + "step": 678 + }, + { + "epoch": 0.02, + "learning_rate": 1.0881410256410256e-05, + "loss": 1.0113, + "step": 679 + }, + { + "epoch": 0.02, + "learning_rate": 1.0897435897435898e-05, + "loss": 1.404, + "step": 680 + }, + { + "epoch": 0.02, + "learning_rate": 1.0913461538461538e-05, + "loss": 1.1131, + "step": 681 + }, + { + "epoch": 0.02, + "learning_rate": 1.092948717948718e-05, + "loss": 1.2703, + "step": 682 + }, + { + "epoch": 0.02, + "learning_rate": 1.0945512820512822e-05, + "loss": 1.0828, + "step": 683 + }, + { + "epoch": 0.02, + "learning_rate": 1.0961538461538464e-05, + "loss": 1.182, + "step": 684 + }, + { + "epoch": 0.02, + "learning_rate": 1.0977564102564104e-05, + "loss": 1.2516, + "step": 685 + }, + { + "epoch": 0.02, + "learning_rate": 1.0993589743589746e-05, + "loss": 1.0121, + "step": 686 + }, + { + "epoch": 0.02, + "learning_rate": 1.1009615384615386e-05, + "loss": 0.9948, + "step": 687 + }, + { + "epoch": 0.02, + "learning_rate": 1.1025641025641028e-05, + "loss": 1.0139, + "step": 688 + }, + { + "epoch": 0.02, + "learning_rate": 1.1041666666666668e-05, + "loss": 1.3951, + "step": 689 + }, + { + "epoch": 0.02, + "learning_rate": 1.105769230769231e-05, + "loss": 1.1741, + "step": 690 + }, + { + "epoch": 0.02, + "learning_rate": 1.107371794871795e-05, + "loss": 1.2621, + "step": 691 + }, + { + "epoch": 0.02, + "learning_rate": 1.1089743589743592e-05, + "loss": 1.1121, + "step": 692 + }, + { + "epoch": 0.02, + "learning_rate": 1.1105769230769232e-05, + "loss": 1.292, + "step": 693 + }, + { + "epoch": 0.02, + "learning_rate": 1.1121794871794872e-05, + "loss": 1.1853, + "step": 694 + }, + { + "epoch": 0.02, + "learning_rate": 1.1137820512820514e-05, + "loss": 1.209, + "step": 695 + }, + { + "epoch": 0.02, + "learning_rate": 1.1153846153846154e-05, + "loss": 1.1879, + "step": 696 + }, + { + "epoch": 0.02, + "learning_rate": 1.1169871794871796e-05, + "loss": 1.2187, + "step": 697 + }, + { + "epoch": 0.02, + "learning_rate": 1.1185897435897437e-05, + "loss": 1.0501, + "step": 698 + }, + { + "epoch": 0.02, + "learning_rate": 1.1201923076923078e-05, + "loss": 1.1549, + "step": 699 + }, + { + "epoch": 0.02, + "learning_rate": 1.1217948717948719e-05, + "loss": 1.2581, + "step": 700 + }, + { + "epoch": 0.02, + "learning_rate": 1.123397435897436e-05, + "loss": 1.2525, + "step": 701 + }, + { + "epoch": 0.02, + "learning_rate": 1.125e-05, + "loss": 1.2385, + "step": 702 + }, + { + "epoch": 0.02, + "learning_rate": 1.1266025641025643e-05, + "loss": 1.413, + "step": 703 + }, + { + "epoch": 0.02, + "learning_rate": 1.1282051282051283e-05, + "loss": 1.2558, + "step": 704 + }, + { + "epoch": 0.02, + "learning_rate": 1.1298076923076923e-05, + "loss": 1.3251, + "step": 705 + }, + { + "epoch": 0.02, + "learning_rate": 1.1314102564102565e-05, + "loss": 1.1343, + "step": 706 + }, + { + "epoch": 0.02, + "learning_rate": 1.1330128205128205e-05, + "loss": 1.1297, + "step": 707 + }, + { + "epoch": 0.02, + "learning_rate": 1.1346153846153847e-05, + "loss": 1.2558, + "step": 708 + }, + { + "epoch": 0.02, + "learning_rate": 1.1362179487179487e-05, + "loss": 1.2566, + "step": 709 + }, + { + "epoch": 0.02, + "learning_rate": 1.1378205128205129e-05, + "loss": 1.1762, + "step": 710 + }, + { + "epoch": 0.02, + "learning_rate": 1.139423076923077e-05, + "loss": 1.1678, + "step": 711 + }, + { + "epoch": 0.02, + "learning_rate": 1.1410256410256411e-05, + "loss": 1.1742, + "step": 712 + }, + { + "epoch": 0.02, + "learning_rate": 1.1426282051282051e-05, + "loss": 1.2385, + "step": 713 + }, + { + "epoch": 0.02, + "learning_rate": 1.1442307692307693e-05, + "loss": 1.2803, + "step": 714 + }, + { + "epoch": 0.02, + "learning_rate": 1.1458333333333333e-05, + "loss": 1.0962, + "step": 715 + }, + { + "epoch": 0.02, + "learning_rate": 1.1474358974358974e-05, + "loss": 1.1288, + "step": 716 + }, + { + "epoch": 0.02, + "learning_rate": 1.1490384615384616e-05, + "loss": 1.2248, + "step": 717 + }, + { + "epoch": 0.02, + "learning_rate": 1.1506410256410256e-05, + "loss": 1.1416, + "step": 718 + }, + { + "epoch": 0.02, + "learning_rate": 1.1522435897435898e-05, + "loss": 1.3362, + "step": 719 + }, + { + "epoch": 0.02, + "learning_rate": 1.1538461538461538e-05, + "loss": 1.2404, + "step": 720 + }, + { + "epoch": 0.02, + "learning_rate": 1.155448717948718e-05, + "loss": 1.2183, + "step": 721 + }, + { + "epoch": 0.02, + "learning_rate": 1.1570512820512823e-05, + "loss": 1.2678, + "step": 722 + }, + { + "epoch": 0.02, + "learning_rate": 1.1586538461538464e-05, + "loss": 1.2262, + "step": 723 + }, + { + "epoch": 0.02, + "learning_rate": 1.1602564102564104e-05, + "loss": 1.3106, + "step": 724 + }, + { + "epoch": 0.02, + "learning_rate": 1.1618589743589746e-05, + "loss": 1.0195, + "step": 725 + }, + { + "epoch": 0.02, + "learning_rate": 1.1634615384615386e-05, + "loss": 1.2501, + "step": 726 + }, + { + "epoch": 0.02, + "learning_rate": 1.1650641025641028e-05, + "loss": 1.267, + "step": 727 + }, + { + "epoch": 0.02, + "learning_rate": 1.1666666666666668e-05, + "loss": 1.2748, + "step": 728 + }, + { + "epoch": 0.02, + "learning_rate": 1.168269230769231e-05, + "loss": 1.1084, + "step": 729 + }, + { + "epoch": 0.02, + "learning_rate": 1.169871794871795e-05, + "loss": 1.0534, + "step": 730 + }, + { + "epoch": 0.02, + "learning_rate": 1.1714743589743592e-05, + "loss": 1.2566, + "step": 731 + }, + { + "epoch": 0.02, + "learning_rate": 1.1730769230769232e-05, + "loss": 1.4034, + "step": 732 + }, + { + "epoch": 0.02, + "learning_rate": 1.1746794871794874e-05, + "loss": 1.0362, + "step": 733 + }, + { + "epoch": 0.02, + "learning_rate": 1.1762820512820514e-05, + "loss": 1.2638, + "step": 734 + }, + { + "epoch": 0.02, + "learning_rate": 1.1778846153846154e-05, + "loss": 1.1432, + "step": 735 + }, + { + "epoch": 0.02, + "learning_rate": 1.1794871794871796e-05, + "loss": 1.1567, + "step": 736 + }, + { + "epoch": 0.02, + "learning_rate": 1.1810897435897436e-05, + "loss": 1.0847, + "step": 737 + }, + { + "epoch": 0.02, + "learning_rate": 1.1826923076923078e-05, + "loss": 1.119, + "step": 738 + }, + { + "epoch": 0.02, + "learning_rate": 1.1842948717948719e-05, + "loss": 1.118, + "step": 739 + }, + { + "epoch": 0.02, + "learning_rate": 1.185897435897436e-05, + "loss": 1.3421, + "step": 740 + }, + { + "epoch": 0.02, + "learning_rate": 1.1875e-05, + "loss": 1.3484, + "step": 741 + }, + { + "epoch": 0.02, + "learning_rate": 1.1891025641025643e-05, + "loss": 1.1505, + "step": 742 + }, + { + "epoch": 0.02, + "learning_rate": 1.1907051282051283e-05, + "loss": 1.1092, + "step": 743 + }, + { + "epoch": 0.02, + "learning_rate": 1.1923076923076925e-05, + "loss": 1.1108, + "step": 744 + }, + { + "epoch": 0.02, + "learning_rate": 1.1939102564102565e-05, + "loss": 1.2633, + "step": 745 + }, + { + "epoch": 0.02, + "learning_rate": 1.1955128205128205e-05, + "loss": 1.228, + "step": 746 + }, + { + "epoch": 0.02, + "learning_rate": 1.1971153846153847e-05, + "loss": 1.259, + "step": 747 + }, + { + "epoch": 0.02, + "learning_rate": 1.1987179487179487e-05, + "loss": 1.0895, + "step": 748 + }, + { + "epoch": 0.02, + "learning_rate": 1.2003205128205129e-05, + "loss": 1.2435, + "step": 749 + }, + { + "epoch": 0.02, + "learning_rate": 1.201923076923077e-05, + "loss": 1.1701, + "step": 750 + }, + { + "epoch": 0.02, + "learning_rate": 1.2035256410256411e-05, + "loss": 0.8938, + "step": 751 + }, + { + "epoch": 0.02, + "learning_rate": 1.2051282051282051e-05, + "loss": 1.0245, + "step": 752 + }, + { + "epoch": 0.02, + "learning_rate": 1.2067307692307693e-05, + "loss": 1.2733, + "step": 753 + }, + { + "epoch": 0.02, + "learning_rate": 1.2083333333333333e-05, + "loss": 1.1323, + "step": 754 + }, + { + "epoch": 0.02, + "learning_rate": 1.2099358974358975e-05, + "loss": 1.2435, + "step": 755 + }, + { + "epoch": 0.02, + "learning_rate": 1.2115384615384615e-05, + "loss": 1.2315, + "step": 756 + }, + { + "epoch": 0.02, + "learning_rate": 1.2131410256410256e-05, + "loss": 0.8903, + "step": 757 + }, + { + "epoch": 0.02, + "learning_rate": 1.2147435897435898e-05, + "loss": 1.0333, + "step": 758 + }, + { + "epoch": 0.02, + "learning_rate": 1.2163461538461538e-05, + "loss": 1.0496, + "step": 759 + }, + { + "epoch": 0.02, + "learning_rate": 1.217948717948718e-05, + "loss": 1.2698, + "step": 760 + }, + { + "epoch": 0.02, + "learning_rate": 1.2195512820512823e-05, + "loss": 1.139, + "step": 761 + }, + { + "epoch": 0.02, + "learning_rate": 1.2211538461538463e-05, + "loss": 1.4843, + "step": 762 + }, + { + "epoch": 0.02, + "learning_rate": 1.2227564102564105e-05, + "loss": 1.0862, + "step": 763 + }, + { + "epoch": 0.02, + "learning_rate": 1.2243589743589746e-05, + "loss": 1.1243, + "step": 764 + }, + { + "epoch": 0.02, + "learning_rate": 1.2259615384615386e-05, + "loss": 0.9964, + "step": 765 + }, + { + "epoch": 0.02, + "learning_rate": 1.2275641025641028e-05, + "loss": 1.1947, + "step": 766 + }, + { + "epoch": 0.02, + "learning_rate": 1.2291666666666668e-05, + "loss": 1.2302, + "step": 767 + }, + { + "epoch": 0.02, + "learning_rate": 1.230769230769231e-05, + "loss": 1.4749, + "step": 768 + }, + { + "epoch": 0.02, + "learning_rate": 1.232371794871795e-05, + "loss": 1.3738, + "step": 769 + }, + { + "epoch": 0.02, + "learning_rate": 1.2339743589743592e-05, + "loss": 1.1808, + "step": 770 + }, + { + "epoch": 0.02, + "learning_rate": 1.2355769230769232e-05, + "loss": 1.4324, + "step": 771 + }, + { + "epoch": 0.02, + "learning_rate": 1.2371794871794874e-05, + "loss": 1.2143, + "step": 772 + }, + { + "epoch": 0.02, + "learning_rate": 1.2387820512820514e-05, + "loss": 1.1823, + "step": 773 + }, + { + "epoch": 0.02, + "learning_rate": 1.2403846153846156e-05, + "loss": 1.247, + "step": 774 + }, + { + "epoch": 0.02, + "learning_rate": 1.2419871794871796e-05, + "loss": 1.2654, + "step": 775 + }, + { + "epoch": 0.02, + "learning_rate": 1.2435897435897436e-05, + "loss": 1.2242, + "step": 776 + }, + { + "epoch": 0.02, + "learning_rate": 1.2451923076923078e-05, + "loss": 1.04, + "step": 777 + }, + { + "epoch": 0.02, + "learning_rate": 1.2467948717948719e-05, + "loss": 1.1779, + "step": 778 + }, + { + "epoch": 0.02, + "learning_rate": 1.248397435897436e-05, + "loss": 0.9934, + "step": 779 + }, + { + "epoch": 0.02, + "learning_rate": 1.25e-05, + "loss": 1.2835, + "step": 780 + }, + { + "epoch": 0.02, + "learning_rate": 1.2516025641025642e-05, + "loss": 1.1535, + "step": 781 + }, + { + "epoch": 0.02, + "learning_rate": 1.2532051282051283e-05, + "loss": 1.4232, + "step": 782 + }, + { + "epoch": 0.02, + "learning_rate": 1.2548076923076925e-05, + "loss": 1.4085, + "step": 783 + }, + { + "epoch": 0.02, + "learning_rate": 1.2564102564102565e-05, + "loss": 1.3189, + "step": 784 + }, + { + "epoch": 0.02, + "learning_rate": 1.2580128205128207e-05, + "loss": 1.0653, + "step": 785 + }, + { + "epoch": 0.02, + "learning_rate": 1.2596153846153847e-05, + "loss": 1.1931, + "step": 786 + }, + { + "epoch": 0.02, + "learning_rate": 1.2612179487179487e-05, + "loss": 1.013, + "step": 787 + }, + { + "epoch": 0.02, + "learning_rate": 1.2628205128205129e-05, + "loss": 1.1359, + "step": 788 + }, + { + "epoch": 0.02, + "learning_rate": 1.264423076923077e-05, + "loss": 0.9323, + "step": 789 + }, + { + "epoch": 0.02, + "learning_rate": 1.2660256410256411e-05, + "loss": 1.3668, + "step": 790 + }, + { + "epoch": 0.02, + "learning_rate": 1.2676282051282051e-05, + "loss": 1.0643, + "step": 791 + }, + { + "epoch": 0.02, + "learning_rate": 1.2692307692307693e-05, + "loss": 1.3334, + "step": 792 + }, + { + "epoch": 0.02, + "learning_rate": 1.2708333333333333e-05, + "loss": 1.1463, + "step": 793 + }, + { + "epoch": 0.02, + "learning_rate": 1.2724358974358975e-05, + "loss": 1.2339, + "step": 794 + }, + { + "epoch": 0.02, + "learning_rate": 1.2740384615384615e-05, + "loss": 1.2516, + "step": 795 + }, + { + "epoch": 0.02, + "learning_rate": 1.2756410256410257e-05, + "loss": 1.0689, + "step": 796 + }, + { + "epoch": 0.02, + "learning_rate": 1.2772435897435898e-05, + "loss": 1.2584, + "step": 797 + }, + { + "epoch": 0.02, + "learning_rate": 1.2788461538461538e-05, + "loss": 1.0283, + "step": 798 + }, + { + "epoch": 0.02, + "learning_rate": 1.280448717948718e-05, + "loss": 1.2066, + "step": 799 + }, + { + "epoch": 0.02, + "learning_rate": 1.2820512820512823e-05, + "loss": 1.159, + "step": 800 + }, + { + "epoch": 0.02, + "learning_rate": 1.2836538461538463e-05, + "loss": 1.3461, + "step": 801 + }, + { + "epoch": 0.02, + "learning_rate": 1.2852564102564105e-05, + "loss": 1.088, + "step": 802 + }, + { + "epoch": 0.02, + "learning_rate": 1.2868589743589746e-05, + "loss": 1.2026, + "step": 803 + }, + { + "epoch": 0.02, + "learning_rate": 1.2884615384615386e-05, + "loss": 0.9784, + "step": 804 + }, + { + "epoch": 0.02, + "learning_rate": 1.2900641025641028e-05, + "loss": 1.3239, + "step": 805 + }, + { + "epoch": 0.02, + "learning_rate": 1.2916666666666668e-05, + "loss": 1.4252, + "step": 806 + }, + { + "epoch": 0.02, + "learning_rate": 1.293269230769231e-05, + "loss": 1.0566, + "step": 807 + }, + { + "epoch": 0.02, + "learning_rate": 1.294871794871795e-05, + "loss": 1.2712, + "step": 808 + }, + { + "epoch": 0.02, + "learning_rate": 1.2964743589743592e-05, + "loss": 1.2467, + "step": 809 + }, + { + "epoch": 0.02, + "learning_rate": 1.2980769230769232e-05, + "loss": 1.1858, + "step": 810 + }, + { + "epoch": 0.02, + "learning_rate": 1.2996794871794874e-05, + "loss": 1.1728, + "step": 811 + }, + { + "epoch": 0.02, + "learning_rate": 1.3012820512820514e-05, + "loss": 1.393, + "step": 812 + }, + { + "epoch": 0.02, + "learning_rate": 1.3028846153846156e-05, + "loss": 0.975, + "step": 813 + }, + { + "epoch": 0.02, + "learning_rate": 1.3044871794871796e-05, + "loss": 1.3003, + "step": 814 + }, + { + "epoch": 0.02, + "learning_rate": 1.3060897435897436e-05, + "loss": 1.3415, + "step": 815 + }, + { + "epoch": 0.02, + "learning_rate": 1.3076923076923078e-05, + "loss": 1.2685, + "step": 816 + }, + { + "epoch": 0.02, + "learning_rate": 1.3092948717948718e-05, + "loss": 1.2476, + "step": 817 + }, + { + "epoch": 0.02, + "learning_rate": 1.310897435897436e-05, + "loss": 1.2621, + "step": 818 + }, + { + "epoch": 0.02, + "learning_rate": 1.3125e-05, + "loss": 1.4171, + "step": 819 + }, + { + "epoch": 0.02, + "learning_rate": 1.3141025641025642e-05, + "loss": 1.1058, + "step": 820 + }, + { + "epoch": 0.02, + "learning_rate": 1.3157051282051283e-05, + "loss": 1.068, + "step": 821 + }, + { + "epoch": 0.02, + "learning_rate": 1.3173076923076925e-05, + "loss": 1.1448, + "step": 822 + }, + { + "epoch": 0.02, + "learning_rate": 1.3189102564102565e-05, + "loss": 0.9693, + "step": 823 + }, + { + "epoch": 0.02, + "learning_rate": 1.3205128205128207e-05, + "loss": 0.9219, + "step": 824 + }, + { + "epoch": 0.02, + "learning_rate": 1.3221153846153847e-05, + "loss": 1.1539, + "step": 825 + }, + { + "epoch": 0.02, + "learning_rate": 1.3237179487179487e-05, + "loss": 1.0712, + "step": 826 + }, + { + "epoch": 0.02, + "learning_rate": 1.3253205128205129e-05, + "loss": 1.3317, + "step": 827 + }, + { + "epoch": 0.02, + "learning_rate": 1.3269230769230769e-05, + "loss": 1.2582, + "step": 828 + }, + { + "epoch": 0.02, + "learning_rate": 1.3285256410256411e-05, + "loss": 1.0004, + "step": 829 + }, + { + "epoch": 0.02, + "learning_rate": 1.3301282051282051e-05, + "loss": 1.3036, + "step": 830 + }, + { + "epoch": 0.02, + "learning_rate": 1.3317307692307693e-05, + "loss": 1.1305, + "step": 831 + }, + { + "epoch": 0.02, + "learning_rate": 1.3333333333333333e-05, + "loss": 1.2392, + "step": 832 + }, + { + "epoch": 0.02, + "learning_rate": 1.3349358974358975e-05, + "loss": 1.2592, + "step": 833 + }, + { + "epoch": 0.02, + "learning_rate": 1.3365384615384615e-05, + "loss": 1.2527, + "step": 834 + }, + { + "epoch": 0.02, + "learning_rate": 1.3381410256410257e-05, + "loss": 1.2713, + "step": 835 + }, + { + "epoch": 0.02, + "learning_rate": 1.3397435897435897e-05, + "loss": 1.3415, + "step": 836 + }, + { + "epoch": 0.02, + "learning_rate": 1.341346153846154e-05, + "loss": 1.1736, + "step": 837 + }, + { + "epoch": 0.02, + "learning_rate": 1.342948717948718e-05, + "loss": 1.4003, + "step": 838 + }, + { + "epoch": 0.02, + "learning_rate": 1.3445512820512823e-05, + "loss": 1.1223, + "step": 839 + }, + { + "epoch": 0.02, + "learning_rate": 1.3461538461538463e-05, + "loss": 1.2768, + "step": 840 + }, + { + "epoch": 0.02, + "learning_rate": 1.3477564102564105e-05, + "loss": 1.1883, + "step": 841 + }, + { + "epoch": 0.02, + "learning_rate": 1.3493589743589745e-05, + "loss": 1.0252, + "step": 842 + }, + { + "epoch": 0.02, + "learning_rate": 1.3509615384615387e-05, + "loss": 1.3542, + "step": 843 + }, + { + "epoch": 0.02, + "learning_rate": 1.3525641025641028e-05, + "loss": 1.2649, + "step": 844 + }, + { + "epoch": 0.02, + "learning_rate": 1.3541666666666668e-05, + "loss": 1.1715, + "step": 845 + }, + { + "epoch": 0.02, + "learning_rate": 1.355769230769231e-05, + "loss": 1.0753, + "step": 846 + }, + { + "epoch": 0.02, + "learning_rate": 1.357371794871795e-05, + "loss": 1.3133, + "step": 847 + }, + { + "epoch": 0.02, + "learning_rate": 1.3589743589743592e-05, + "loss": 0.9736, + "step": 848 + }, + { + "epoch": 0.02, + "learning_rate": 1.3605769230769232e-05, + "loss": 0.9785, + "step": 849 + }, + { + "epoch": 0.02, + "learning_rate": 1.3621794871794874e-05, + "loss": 1.2361, + "step": 850 + }, + { + "epoch": 0.02, + "learning_rate": 1.3637820512820514e-05, + "loss": 1.401, + "step": 851 + }, + { + "epoch": 0.02, + "learning_rate": 1.3653846153846156e-05, + "loss": 1.1001, + "step": 852 + }, + { + "epoch": 0.02, + "learning_rate": 1.3669871794871796e-05, + "loss": 1.106, + "step": 853 + }, + { + "epoch": 0.02, + "learning_rate": 1.3685897435897438e-05, + "loss": 1.1717, + "step": 854 + }, + { + "epoch": 0.02, + "learning_rate": 1.3701923076923078e-05, + "loss": 1.1425, + "step": 855 + }, + { + "epoch": 0.02, + "learning_rate": 1.3717948717948718e-05, + "loss": 1.3325, + "step": 856 + }, + { + "epoch": 0.02, + "learning_rate": 1.373397435897436e-05, + "loss": 1.1074, + "step": 857 + }, + { + "epoch": 0.02, + "learning_rate": 1.375e-05, + "loss": 1.0714, + "step": 858 + }, + { + "epoch": 0.02, + "learning_rate": 1.3766025641025642e-05, + "loss": 0.9768, + "step": 859 + }, + { + "epoch": 0.02, + "learning_rate": 1.3782051282051283e-05, + "loss": 1.0747, + "step": 860 + }, + { + "epoch": 0.02, + "learning_rate": 1.3798076923076924e-05, + "loss": 1.4055, + "step": 861 + }, + { + "epoch": 0.02, + "learning_rate": 1.3814102564102565e-05, + "loss": 1.078, + "step": 862 + }, + { + "epoch": 0.02, + "learning_rate": 1.3830128205128207e-05, + "loss": 0.9701, + "step": 863 + }, + { + "epoch": 0.02, + "learning_rate": 1.3846153846153847e-05, + "loss": 1.216, + "step": 864 + }, + { + "epoch": 0.02, + "learning_rate": 1.3862179487179489e-05, + "loss": 1.4006, + "step": 865 + }, + { + "epoch": 0.02, + "learning_rate": 1.3878205128205129e-05, + "loss": 1.0334, + "step": 866 + }, + { + "epoch": 0.02, + "learning_rate": 1.3894230769230769e-05, + "loss": 1.0574, + "step": 867 + }, + { + "epoch": 0.02, + "learning_rate": 1.3910256410256411e-05, + "loss": 1.3486, + "step": 868 + }, + { + "epoch": 0.02, + "learning_rate": 1.3926282051282051e-05, + "loss": 1.1224, + "step": 869 + }, + { + "epoch": 0.02, + "learning_rate": 1.3942307692307693e-05, + "loss": 1.2438, + "step": 870 + }, + { + "epoch": 0.02, + "learning_rate": 1.3958333333333333e-05, + "loss": 1.1604, + "step": 871 + }, + { + "epoch": 0.02, + "learning_rate": 1.3974358974358975e-05, + "loss": 1.1749, + "step": 872 + }, + { + "epoch": 0.02, + "learning_rate": 1.3990384615384615e-05, + "loss": 1.4907, + "step": 873 + }, + { + "epoch": 0.02, + "learning_rate": 1.4006410256410257e-05, + "loss": 1.0963, + "step": 874 + }, + { + "epoch": 0.02, + "learning_rate": 1.4022435897435897e-05, + "loss": 1.2153, + "step": 875 + }, + { + "epoch": 0.02, + "learning_rate": 1.403846153846154e-05, + "loss": 1.1814, + "step": 876 + }, + { + "epoch": 0.02, + "learning_rate": 1.405448717948718e-05, + "loss": 1.2046, + "step": 877 + }, + { + "epoch": 0.02, + "learning_rate": 1.4070512820512823e-05, + "loss": 1.249, + "step": 878 + }, + { + "epoch": 0.02, + "learning_rate": 1.4086538461538463e-05, + "loss": 1.23, + "step": 879 + }, + { + "epoch": 0.02, + "learning_rate": 1.4102564102564105e-05, + "loss": 1.0741, + "step": 880 + }, + { + "epoch": 0.02, + "learning_rate": 1.4118589743589745e-05, + "loss": 1.2773, + "step": 881 + }, + { + "epoch": 0.02, + "learning_rate": 1.4134615384615387e-05, + "loss": 1.1469, + "step": 882 + }, + { + "epoch": 0.02, + "learning_rate": 1.4150641025641027e-05, + "loss": 1.068, + "step": 883 + }, + { + "epoch": 0.02, + "learning_rate": 1.416666666666667e-05, + "loss": 0.8944, + "step": 884 + }, + { + "epoch": 0.02, + "learning_rate": 1.418269230769231e-05, + "loss": 1.327, + "step": 885 + }, + { + "epoch": 0.02, + "learning_rate": 1.419871794871795e-05, + "loss": 1.2017, + "step": 886 + }, + { + "epoch": 0.02, + "learning_rate": 1.4214743589743592e-05, + "loss": 1.3529, + "step": 887 + }, + { + "epoch": 0.02, + "learning_rate": 1.4230769230769232e-05, + "loss": 1.2851, + "step": 888 + }, + { + "epoch": 0.02, + "learning_rate": 1.4246794871794874e-05, + "loss": 1.3631, + "step": 889 + }, + { + "epoch": 0.02, + "learning_rate": 1.4262820512820514e-05, + "loss": 1.2404, + "step": 890 + }, + { + "epoch": 0.02, + "learning_rate": 1.4278846153846156e-05, + "loss": 1.3024, + "step": 891 + }, + { + "epoch": 0.02, + "learning_rate": 1.4294871794871796e-05, + "loss": 1.148, + "step": 892 + }, + { + "epoch": 0.02, + "learning_rate": 1.4310897435897438e-05, + "loss": 1.1593, + "step": 893 + }, + { + "epoch": 0.02, + "learning_rate": 1.4326923076923078e-05, + "loss": 1.4134, + "step": 894 + }, + { + "epoch": 0.02, + "learning_rate": 1.434294871794872e-05, + "loss": 1.2109, + "step": 895 + }, + { + "epoch": 0.02, + "learning_rate": 1.435897435897436e-05, + "loss": 1.4204, + "step": 896 + }, + { + "epoch": 0.02, + "learning_rate": 1.4375e-05, + "loss": 1.1637, + "step": 897 + }, + { + "epoch": 0.02, + "learning_rate": 1.4391025641025642e-05, + "loss": 1.3659, + "step": 898 + }, + { + "epoch": 0.02, + "learning_rate": 1.4407051282051283e-05, + "loss": 1.3391, + "step": 899 + }, + { + "epoch": 0.02, + "learning_rate": 1.4423076923076924e-05, + "loss": 0.9049, + "step": 900 + }, + { + "epoch": 0.02, + "learning_rate": 1.4439102564102565e-05, + "loss": 1.0576, + "step": 901 + }, + { + "epoch": 0.02, + "learning_rate": 1.4455128205128207e-05, + "loss": 1.3248, + "step": 902 + }, + { + "epoch": 0.02, + "learning_rate": 1.4471153846153847e-05, + "loss": 1.0896, + "step": 903 + }, + { + "epoch": 0.02, + "learning_rate": 1.4487179487179489e-05, + "loss": 1.3628, + "step": 904 + }, + { + "epoch": 0.02, + "learning_rate": 1.4503205128205129e-05, + "loss": 1.2255, + "step": 905 + }, + { + "epoch": 0.02, + "learning_rate": 1.451923076923077e-05, + "loss": 1.1131, + "step": 906 + }, + { + "epoch": 0.02, + "learning_rate": 1.4535256410256411e-05, + "loss": 1.0406, + "step": 907 + }, + { + "epoch": 0.02, + "learning_rate": 1.4551282051282051e-05, + "loss": 1.2486, + "step": 908 + }, + { + "epoch": 0.02, + "learning_rate": 1.4567307692307693e-05, + "loss": 0.9595, + "step": 909 + }, + { + "epoch": 0.02, + "learning_rate": 1.4583333333333333e-05, + "loss": 1.0321, + "step": 910 + }, + { + "epoch": 0.02, + "learning_rate": 1.4599358974358975e-05, + "loss": 1.2561, + "step": 911 + }, + { + "epoch": 0.02, + "learning_rate": 1.4615384615384615e-05, + "loss": 1.1091, + "step": 912 + }, + { + "epoch": 0.02, + "learning_rate": 1.4631410256410257e-05, + "loss": 1.4594, + "step": 913 + }, + { + "epoch": 0.02, + "learning_rate": 1.4647435897435897e-05, + "loss": 0.8632, + "step": 914 + }, + { + "epoch": 0.02, + "learning_rate": 1.466346153846154e-05, + "loss": 1.1593, + "step": 915 + }, + { + "epoch": 0.02, + "learning_rate": 1.467948717948718e-05, + "loss": 1.3565, + "step": 916 + }, + { + "epoch": 0.02, + "learning_rate": 1.4695512820512823e-05, + "loss": 1.0567, + "step": 917 + }, + { + "epoch": 0.02, + "learning_rate": 1.4711538461538463e-05, + "loss": 1.278, + "step": 918 + }, + { + "epoch": 0.02, + "learning_rate": 1.4727564102564105e-05, + "loss": 1.1662, + "step": 919 + }, + { + "epoch": 0.02, + "learning_rate": 1.4743589743589745e-05, + "loss": 1.29, + "step": 920 + }, + { + "epoch": 0.02, + "learning_rate": 1.4759615384615387e-05, + "loss": 0.9986, + "step": 921 + }, + { + "epoch": 0.02, + "learning_rate": 1.4775641025641027e-05, + "loss": 1.0212, + "step": 922 + }, + { + "epoch": 0.02, + "learning_rate": 1.479166666666667e-05, + "loss": 1.3483, + "step": 923 + }, + { + "epoch": 0.02, + "learning_rate": 1.480769230769231e-05, + "loss": 0.9879, + "step": 924 + }, + { + "epoch": 0.02, + "learning_rate": 1.482371794871795e-05, + "loss": 1.2092, + "step": 925 + }, + { + "epoch": 0.02, + "learning_rate": 1.4839743589743592e-05, + "loss": 1.258, + "step": 926 + }, + { + "epoch": 0.02, + "learning_rate": 1.4855769230769232e-05, + "loss": 1.2322, + "step": 927 + }, + { + "epoch": 0.02, + "learning_rate": 1.4871794871794874e-05, + "loss": 1.2328, + "step": 928 + }, + { + "epoch": 0.02, + "learning_rate": 1.4887820512820514e-05, + "loss": 1.4027, + "step": 929 + }, + { + "epoch": 0.02, + "learning_rate": 1.4903846153846156e-05, + "loss": 1.2374, + "step": 930 + }, + { + "epoch": 0.02, + "learning_rate": 1.4919871794871796e-05, + "loss": 1.2301, + "step": 931 + }, + { + "epoch": 0.02, + "learning_rate": 1.4935897435897438e-05, + "loss": 1.335, + "step": 932 + }, + { + "epoch": 0.02, + "learning_rate": 1.4951923076923078e-05, + "loss": 1.245, + "step": 933 + }, + { + "epoch": 0.02, + "learning_rate": 1.496794871794872e-05, + "loss": 1.3863, + "step": 934 + }, + { + "epoch": 0.02, + "learning_rate": 1.498397435897436e-05, + "loss": 1.2159, + "step": 935 + }, + { + "epoch": 0.02, + "learning_rate": 1.5000000000000002e-05, + "loss": 1.1317, + "step": 936 + }, + { + "epoch": 0.02, + "learning_rate": 1.5016025641025642e-05, + "loss": 1.2765, + "step": 937 + }, + { + "epoch": 0.02, + "learning_rate": 1.5032051282051282e-05, + "loss": 1.0871, + "step": 938 + }, + { + "epoch": 0.02, + "learning_rate": 1.5048076923076924e-05, + "loss": 1.1013, + "step": 939 + }, + { + "epoch": 0.02, + "learning_rate": 1.5064102564102565e-05, + "loss": 0.9938, + "step": 940 + }, + { + "epoch": 0.02, + "learning_rate": 1.5080128205128206e-05, + "loss": 1.3721, + "step": 941 + }, + { + "epoch": 0.02, + "learning_rate": 1.5096153846153847e-05, + "loss": 0.8369, + "step": 942 + }, + { + "epoch": 0.02, + "learning_rate": 1.5112179487179489e-05, + "loss": 1.3132, + "step": 943 + }, + { + "epoch": 0.02, + "learning_rate": 1.5128205128205129e-05, + "loss": 1.1455, + "step": 944 + }, + { + "epoch": 0.02, + "learning_rate": 1.514423076923077e-05, + "loss": 1.2895, + "step": 945 + }, + { + "epoch": 0.02, + "learning_rate": 1.516025641025641e-05, + "loss": 1.164, + "step": 946 + }, + { + "epoch": 0.02, + "learning_rate": 1.5176282051282053e-05, + "loss": 1.2671, + "step": 947 + }, + { + "epoch": 0.02, + "learning_rate": 1.5192307692307693e-05, + "loss": 1.246, + "step": 948 + }, + { + "epoch": 0.02, + "learning_rate": 1.5208333333333333e-05, + "loss": 1.2016, + "step": 949 + }, + { + "epoch": 0.02, + "learning_rate": 1.5224358974358975e-05, + "loss": 1.2411, + "step": 950 + }, + { + "epoch": 0.02, + "learning_rate": 1.5240384615384615e-05, + "loss": 1.1066, + "step": 951 + }, + { + "epoch": 0.02, + "learning_rate": 1.5256410256410257e-05, + "loss": 1.1998, + "step": 952 + }, + { + "epoch": 0.02, + "learning_rate": 1.52724358974359e-05, + "loss": 1.118, + "step": 953 + }, + { + "epoch": 0.02, + "learning_rate": 1.528846153846154e-05, + "loss": 1.135, + "step": 954 + }, + { + "epoch": 0.02, + "learning_rate": 1.530448717948718e-05, + "loss": 1.3047, + "step": 955 + }, + { + "epoch": 0.02, + "learning_rate": 1.5320512820512823e-05, + "loss": 1.1776, + "step": 956 + }, + { + "epoch": 0.02, + "learning_rate": 1.5336538461538463e-05, + "loss": 1.0731, + "step": 957 + }, + { + "epoch": 0.02, + "learning_rate": 1.5352564102564103e-05, + "loss": 1.1849, + "step": 958 + }, + { + "epoch": 0.02, + "learning_rate": 1.5368589743589747e-05, + "loss": 1.4809, + "step": 959 + }, + { + "epoch": 0.02, + "learning_rate": 1.5384615384615387e-05, + "loss": 1.1407, + "step": 960 + }, + { + "epoch": 0.02, + "learning_rate": 1.5400641025641027e-05, + "loss": 1.2749, + "step": 961 + }, + { + "epoch": 0.02, + "learning_rate": 1.5416666666666668e-05, + "loss": 1.3091, + "step": 962 + }, + { + "epoch": 0.02, + "learning_rate": 1.543269230769231e-05, + "loss": 1.1063, + "step": 963 + }, + { + "epoch": 0.02, + "learning_rate": 1.544871794871795e-05, + "loss": 1.0388, + "step": 964 + }, + { + "epoch": 0.02, + "learning_rate": 1.546474358974359e-05, + "loss": 1.1179, + "step": 965 + }, + { + "epoch": 0.02, + "learning_rate": 1.5480769230769232e-05, + "loss": 1.1963, + "step": 966 + }, + { + "epoch": 0.02, + "learning_rate": 1.5496794871794872e-05, + "loss": 1.291, + "step": 967 + }, + { + "epoch": 0.02, + "learning_rate": 1.5512820512820516e-05, + "loss": 1.3837, + "step": 968 + }, + { + "epoch": 0.02, + "learning_rate": 1.5528846153846156e-05, + "loss": 1.1665, + "step": 969 + }, + { + "epoch": 0.02, + "learning_rate": 1.5544871794871796e-05, + "loss": 1.3376, + "step": 970 + }, + { + "epoch": 0.02, + "learning_rate": 1.5560897435897436e-05, + "loss": 1.2821, + "step": 971 + }, + { + "epoch": 0.02, + "learning_rate": 1.557692307692308e-05, + "loss": 1.3695, + "step": 972 + }, + { + "epoch": 0.02, + "learning_rate": 1.559294871794872e-05, + "loss": 1.2777, + "step": 973 + }, + { + "epoch": 0.02, + "learning_rate": 1.560897435897436e-05, + "loss": 1.3066, + "step": 974 + }, + { + "epoch": 0.02, + "learning_rate": 1.5625e-05, + "loss": 1.0925, + "step": 975 + }, + { + "epoch": 0.02, + "learning_rate": 1.5641025641025644e-05, + "loss": 1.1295, + "step": 976 + }, + { + "epoch": 0.02, + "learning_rate": 1.5657051282051284e-05, + "loss": 1.3146, + "step": 977 + }, + { + "epoch": 0.02, + "learning_rate": 1.5673076923076924e-05, + "loss": 1.1551, + "step": 978 + }, + { + "epoch": 0.02, + "learning_rate": 1.5689102564102565e-05, + "loss": 1.0994, + "step": 979 + }, + { + "epoch": 0.02, + "learning_rate": 1.5705128205128205e-05, + "loss": 1.3853, + "step": 980 + }, + { + "epoch": 0.02, + "learning_rate": 1.5721153846153848e-05, + "loss": 1.241, + "step": 981 + }, + { + "epoch": 0.02, + "learning_rate": 1.573717948717949e-05, + "loss": 1.4943, + "step": 982 + }, + { + "epoch": 0.02, + "learning_rate": 1.575320512820513e-05, + "loss": 1.2179, + "step": 983 + }, + { + "epoch": 0.02, + "learning_rate": 1.576923076923077e-05, + "loss": 1.0759, + "step": 984 + }, + { + "epoch": 0.02, + "learning_rate": 1.5785256410256412e-05, + "loss": 0.9679, + "step": 985 + }, + { + "epoch": 0.02, + "learning_rate": 1.5801282051282053e-05, + "loss": 1.109, + "step": 986 + }, + { + "epoch": 0.02, + "learning_rate": 1.5817307692307693e-05, + "loss": 1.1615, + "step": 987 + }, + { + "epoch": 0.02, + "learning_rate": 1.5833333333333333e-05, + "loss": 1.3012, + "step": 988 + }, + { + "epoch": 0.02, + "learning_rate": 1.5849358974358973e-05, + "loss": 1.0902, + "step": 989 + }, + { + "epoch": 0.02, + "learning_rate": 1.5865384615384617e-05, + "loss": 1.3745, + "step": 990 + }, + { + "epoch": 0.02, + "learning_rate": 1.5881410256410257e-05, + "loss": 1.0515, + "step": 991 + }, + { + "epoch": 0.02, + "learning_rate": 1.5897435897435897e-05, + "loss": 1.2372, + "step": 992 + }, + { + "epoch": 0.02, + "learning_rate": 1.5913461538461537e-05, + "loss": 1.267, + "step": 993 + }, + { + "epoch": 0.02, + "learning_rate": 1.592948717948718e-05, + "loss": 1.1558, + "step": 994 + }, + { + "epoch": 0.02, + "learning_rate": 1.594551282051282e-05, + "loss": 1.0116, + "step": 995 + }, + { + "epoch": 0.02, + "learning_rate": 1.5961538461538465e-05, + "loss": 1.1538, + "step": 996 + }, + { + "epoch": 0.02, + "learning_rate": 1.5977564102564105e-05, + "loss": 0.8837, + "step": 997 + }, + { + "epoch": 0.02, + "learning_rate": 1.5993589743589745e-05, + "loss": 1.1889, + "step": 998 + }, + { + "epoch": 0.02, + "learning_rate": 1.6009615384615385e-05, + "loss": 1.4821, + "step": 999 + }, + { + "epoch": 0.02, + "learning_rate": 1.602564102564103e-05, + "loss": 1.0711, + "step": 1000 + }, + { + "epoch": 0.02, + "learning_rate": 1.604166666666667e-05, + "loss": 1.2182, + "step": 1001 + }, + { + "epoch": 0.02, + "learning_rate": 1.605769230769231e-05, + "loss": 1.2039, + "step": 1002 + }, + { + "epoch": 0.02, + "learning_rate": 1.607371794871795e-05, + "loss": 1.2145, + "step": 1003 + }, + { + "epoch": 0.02, + "learning_rate": 1.6089743589743593e-05, + "loss": 1.2066, + "step": 1004 + }, + { + "epoch": 0.02, + "learning_rate": 1.6105769230769233e-05, + "loss": 1.1998, + "step": 1005 + }, + { + "epoch": 0.02, + "learning_rate": 1.6121794871794874e-05, + "loss": 1.1259, + "step": 1006 + }, + { + "epoch": 0.02, + "learning_rate": 1.6137820512820514e-05, + "loss": 1.0331, + "step": 1007 + }, + { + "epoch": 0.02, + "learning_rate": 1.6153846153846154e-05, + "loss": 1.1603, + "step": 1008 + }, + { + "epoch": 0.02, + "learning_rate": 1.6169871794871798e-05, + "loss": 1.2544, + "step": 1009 + }, + { + "epoch": 0.02, + "learning_rate": 1.6185897435897438e-05, + "loss": 1.2241, + "step": 1010 + }, + { + "epoch": 0.02, + "learning_rate": 1.6201923076923078e-05, + "loss": 1.1364, + "step": 1011 + }, + { + "epoch": 0.02, + "learning_rate": 1.6217948717948718e-05, + "loss": 1.4473, + "step": 1012 + }, + { + "epoch": 0.02, + "learning_rate": 1.6233974358974362e-05, + "loss": 1.0418, + "step": 1013 + }, + { + "epoch": 0.02, + "learning_rate": 1.6250000000000002e-05, + "loss": 1.2076, + "step": 1014 + }, + { + "epoch": 0.02, + "learning_rate": 1.6266025641025642e-05, + "loss": 1.2408, + "step": 1015 + }, + { + "epoch": 0.02, + "learning_rate": 1.6282051282051282e-05, + "loss": 1.2319, + "step": 1016 + }, + { + "epoch": 0.02, + "learning_rate": 1.6298076923076923e-05, + "loss": 1.2146, + "step": 1017 + }, + { + "epoch": 0.02, + "learning_rate": 1.6314102564102566e-05, + "loss": 1.0656, + "step": 1018 + }, + { + "epoch": 0.02, + "learning_rate": 1.6330128205128206e-05, + "loss": 1.2549, + "step": 1019 + }, + { + "epoch": 0.02, + "learning_rate": 1.6346153846153847e-05, + "loss": 1.034, + "step": 1020 + }, + { + "epoch": 0.02, + "learning_rate": 1.6362179487179487e-05, + "loss": 1.4244, + "step": 1021 + }, + { + "epoch": 0.02, + "learning_rate": 1.637820512820513e-05, + "loss": 1.4229, + "step": 1022 + }, + { + "epoch": 0.02, + "learning_rate": 1.639423076923077e-05, + "loss": 1.0142, + "step": 1023 + }, + { + "epoch": 0.02, + "learning_rate": 1.641025641025641e-05, + "loss": 1.0816, + "step": 1024 + }, + { + "epoch": 0.02, + "learning_rate": 1.642628205128205e-05, + "loss": 1.364, + "step": 1025 + }, + { + "epoch": 0.02, + "learning_rate": 1.6442307692307695e-05, + "loss": 1.2939, + "step": 1026 + }, + { + "epoch": 0.02, + "learning_rate": 1.6458333333333335e-05, + "loss": 1.3222, + "step": 1027 + }, + { + "epoch": 0.02, + "learning_rate": 1.6474358974358975e-05, + "loss": 1.2804, + "step": 1028 + }, + { + "epoch": 0.02, + "learning_rate": 1.6490384615384615e-05, + "loss": 1.206, + "step": 1029 + }, + { + "epoch": 0.02, + "learning_rate": 1.6506410256410255e-05, + "loss": 1.1043, + "step": 1030 + }, + { + "epoch": 0.02, + "learning_rate": 1.65224358974359e-05, + "loss": 0.9979, + "step": 1031 + }, + { + "epoch": 0.02, + "learning_rate": 1.653846153846154e-05, + "loss": 1.097, + "step": 1032 + }, + { + "epoch": 0.02, + "learning_rate": 1.655448717948718e-05, + "loss": 1.2871, + "step": 1033 + }, + { + "epoch": 0.02, + "learning_rate": 1.6570512820512823e-05, + "loss": 1.2588, + "step": 1034 + }, + { + "epoch": 0.02, + "learning_rate": 1.6586538461538463e-05, + "loss": 1.093, + "step": 1035 + }, + { + "epoch": 0.02, + "learning_rate": 1.6602564102564103e-05, + "loss": 1.4462, + "step": 1036 + }, + { + "epoch": 0.02, + "learning_rate": 1.6618589743589747e-05, + "loss": 1.039, + "step": 1037 + }, + { + "epoch": 0.02, + "learning_rate": 1.6634615384615387e-05, + "loss": 1.3481, + "step": 1038 + }, + { + "epoch": 0.02, + "learning_rate": 1.6650641025641027e-05, + "loss": 1.2281, + "step": 1039 + }, + { + "epoch": 0.03, + "learning_rate": 1.6666666666666667e-05, + "loss": 1.2737, + "step": 1040 + }, + { + "epoch": 0.03, + "learning_rate": 1.668269230769231e-05, + "loss": 1.13, + "step": 1041 + }, + { + "epoch": 0.03, + "learning_rate": 1.669871794871795e-05, + "loss": 1.3116, + "step": 1042 + }, + { + "epoch": 0.03, + "learning_rate": 1.671474358974359e-05, + "loss": 1.0846, + "step": 1043 + }, + { + "epoch": 0.03, + "learning_rate": 1.673076923076923e-05, + "loss": 1.118, + "step": 1044 + }, + { + "epoch": 0.03, + "learning_rate": 1.6746794871794875e-05, + "loss": 1.0307, + "step": 1045 + }, + { + "epoch": 0.03, + "learning_rate": 1.6762820512820515e-05, + "loss": 0.9823, + "step": 1046 + }, + { + "epoch": 0.03, + "learning_rate": 1.6778846153846156e-05, + "loss": 1.2747, + "step": 1047 + }, + { + "epoch": 0.03, + "learning_rate": 1.6794871794871796e-05, + "loss": 1.2088, + "step": 1048 + }, + { + "epoch": 0.03, + "learning_rate": 1.6810897435897436e-05, + "loss": 1.2097, + "step": 1049 + }, + { + "epoch": 0.03, + "learning_rate": 1.682692307692308e-05, + "loss": 0.9878, + "step": 1050 + }, + { + "epoch": 0.03, + "learning_rate": 1.684294871794872e-05, + "loss": 1.2073, + "step": 1051 + }, + { + "epoch": 0.03, + "learning_rate": 1.685897435897436e-05, + "loss": 1.05, + "step": 1052 + }, + { + "epoch": 0.03, + "learning_rate": 1.6875e-05, + "loss": 1.2462, + "step": 1053 + }, + { + "epoch": 0.03, + "learning_rate": 1.6891025641025644e-05, + "loss": 1.2335, + "step": 1054 + }, + { + "epoch": 0.03, + "learning_rate": 1.6907051282051284e-05, + "loss": 1.105, + "step": 1055 + }, + { + "epoch": 0.03, + "learning_rate": 1.6923076923076924e-05, + "loss": 1.2117, + "step": 1056 + }, + { + "epoch": 0.03, + "learning_rate": 1.6939102564102564e-05, + "loss": 1.2315, + "step": 1057 + }, + { + "epoch": 0.03, + "learning_rate": 1.6955128205128205e-05, + "loss": 1.0831, + "step": 1058 + }, + { + "epoch": 0.03, + "learning_rate": 1.6971153846153848e-05, + "loss": 1.2033, + "step": 1059 + }, + { + "epoch": 0.03, + "learning_rate": 1.698717948717949e-05, + "loss": 1.3953, + "step": 1060 + }, + { + "epoch": 0.03, + "learning_rate": 1.700320512820513e-05, + "loss": 1.4625, + "step": 1061 + }, + { + "epoch": 0.03, + "learning_rate": 1.701923076923077e-05, + "loss": 1.0707, + "step": 1062 + }, + { + "epoch": 0.03, + "learning_rate": 1.7035256410256412e-05, + "loss": 1.1274, + "step": 1063 + }, + { + "epoch": 0.03, + "learning_rate": 1.7051282051282053e-05, + "loss": 1.2945, + "step": 1064 + }, + { + "epoch": 0.03, + "learning_rate": 1.7067307692307693e-05, + "loss": 1.3943, + "step": 1065 + }, + { + "epoch": 0.03, + "learning_rate": 1.7083333333333333e-05, + "loss": 1.1173, + "step": 1066 + }, + { + "epoch": 0.03, + "learning_rate": 1.7099358974358977e-05, + "loss": 1.1752, + "step": 1067 + }, + { + "epoch": 0.03, + "learning_rate": 1.7115384615384617e-05, + "loss": 1.249, + "step": 1068 + }, + { + "epoch": 0.03, + "learning_rate": 1.7131410256410257e-05, + "loss": 1.1844, + "step": 1069 + }, + { + "epoch": 0.03, + "learning_rate": 1.7147435897435897e-05, + "loss": 1.2427, + "step": 1070 + }, + { + "epoch": 0.03, + "learning_rate": 1.7163461538461537e-05, + "loss": 1.0092, + "step": 1071 + }, + { + "epoch": 0.03, + "learning_rate": 1.717948717948718e-05, + "loss": 0.9577, + "step": 1072 + }, + { + "epoch": 0.03, + "learning_rate": 1.7195512820512825e-05, + "loss": 1.2859, + "step": 1073 + }, + { + "epoch": 0.03, + "learning_rate": 1.7211538461538465e-05, + "loss": 1.3183, + "step": 1074 + }, + { + "epoch": 0.03, + "learning_rate": 1.7227564102564105e-05, + "loss": 1.376, + "step": 1075 + }, + { + "epoch": 0.03, + "learning_rate": 1.7243589743589745e-05, + "loss": 1.1586, + "step": 1076 + }, + { + "epoch": 0.03, + "learning_rate": 1.7259615384615385e-05, + "loss": 1.0983, + "step": 1077 + }, + { + "epoch": 0.03, + "learning_rate": 1.727564102564103e-05, + "loss": 1.1441, + "step": 1078 + }, + { + "epoch": 0.03, + "learning_rate": 1.729166666666667e-05, + "loss": 1.1361, + "step": 1079 + }, + { + "epoch": 0.03, + "learning_rate": 1.730769230769231e-05, + "loss": 1.1763, + "step": 1080 + }, + { + "epoch": 0.03, + "learning_rate": 1.732371794871795e-05, + "loss": 1.2607, + "step": 1081 + }, + { + "epoch": 0.03, + "learning_rate": 1.7339743589743593e-05, + "loss": 0.8642, + "step": 1082 + }, + { + "epoch": 0.03, + "learning_rate": 1.7355769230769233e-05, + "loss": 0.972, + "step": 1083 + }, + { + "epoch": 0.03, + "learning_rate": 1.7371794871794873e-05, + "loss": 1.2604, + "step": 1084 + }, + { + "epoch": 0.03, + "learning_rate": 1.7387820512820514e-05, + "loss": 1.2129, + "step": 1085 + }, + { + "epoch": 0.03, + "learning_rate": 1.7403846153846157e-05, + "loss": 1.1742, + "step": 1086 + }, + { + "epoch": 0.03, + "learning_rate": 1.7419871794871797e-05, + "loss": 1.2444, + "step": 1087 + }, + { + "epoch": 0.03, + "learning_rate": 1.7435897435897438e-05, + "loss": 1.2756, + "step": 1088 + }, + { + "epoch": 0.03, + "learning_rate": 1.7451923076923078e-05, + "loss": 1.3372, + "step": 1089 + }, + { + "epoch": 0.03, + "learning_rate": 1.7467948717948718e-05, + "loss": 1.1342, + "step": 1090 + }, + { + "epoch": 0.03, + "learning_rate": 1.748397435897436e-05, + "loss": 1.1363, + "step": 1091 + }, + { + "epoch": 0.03, + "learning_rate": 1.7500000000000002e-05, + "loss": 1.3915, + "step": 1092 + }, + { + "epoch": 0.03, + "learning_rate": 1.7516025641025642e-05, + "loss": 1.3064, + "step": 1093 + }, + { + "epoch": 0.03, + "learning_rate": 1.7532051282051282e-05, + "loss": 1.1681, + "step": 1094 + }, + { + "epoch": 0.03, + "learning_rate": 1.7548076923076926e-05, + "loss": 1.4063, + "step": 1095 + }, + { + "epoch": 0.03, + "learning_rate": 1.7564102564102566e-05, + "loss": 0.9645, + "step": 1096 + }, + { + "epoch": 0.03, + "learning_rate": 1.7580128205128206e-05, + "loss": 1.2369, + "step": 1097 + }, + { + "epoch": 0.03, + "learning_rate": 1.7596153846153846e-05, + "loss": 1.3164, + "step": 1098 + }, + { + "epoch": 0.03, + "learning_rate": 1.7612179487179487e-05, + "loss": 1.3215, + "step": 1099 + }, + { + "epoch": 0.03, + "learning_rate": 1.762820512820513e-05, + "loss": 1.3073, + "step": 1100 + }, + { + "epoch": 0.03, + "learning_rate": 1.764423076923077e-05, + "loss": 1.1623, + "step": 1101 + }, + { + "epoch": 0.03, + "learning_rate": 1.766025641025641e-05, + "loss": 1.1935, + "step": 1102 + }, + { + "epoch": 0.03, + "learning_rate": 1.767628205128205e-05, + "loss": 1.1595, + "step": 1103 + }, + { + "epoch": 0.03, + "learning_rate": 1.7692307692307694e-05, + "loss": 1.136, + "step": 1104 + }, + { + "epoch": 0.03, + "learning_rate": 1.7708333333333335e-05, + "loss": 1.3046, + "step": 1105 + }, + { + "epoch": 0.03, + "learning_rate": 1.7724358974358975e-05, + "loss": 1.3894, + "step": 1106 + }, + { + "epoch": 0.03, + "learning_rate": 1.7740384615384615e-05, + "loss": 1.074, + "step": 1107 + }, + { + "epoch": 0.03, + "learning_rate": 1.775641025641026e-05, + "loss": 1.1588, + "step": 1108 + }, + { + "epoch": 0.03, + "learning_rate": 1.77724358974359e-05, + "loss": 1.284, + "step": 1109 + }, + { + "epoch": 0.03, + "learning_rate": 1.778846153846154e-05, + "loss": 1.3518, + "step": 1110 + }, + { + "epoch": 0.03, + "learning_rate": 1.780448717948718e-05, + "loss": 0.9886, + "step": 1111 + }, + { + "epoch": 0.03, + "learning_rate": 1.7820512820512823e-05, + "loss": 0.9491, + "step": 1112 + }, + { + "epoch": 0.03, + "learning_rate": 1.7836538461538463e-05, + "loss": 1.4863, + "step": 1113 + }, + { + "epoch": 0.03, + "learning_rate": 1.7852564102564107e-05, + "loss": 1.1829, + "step": 1114 + }, + { + "epoch": 0.03, + "learning_rate": 1.7868589743589747e-05, + "loss": 1.2949, + "step": 1115 + }, + { + "epoch": 0.03, + "learning_rate": 1.7884615384615387e-05, + "loss": 1.1914, + "step": 1116 + }, + { + "epoch": 0.03, + "learning_rate": 1.7900641025641027e-05, + "loss": 1.2202, + "step": 1117 + }, + { + "epoch": 0.03, + "learning_rate": 1.7916666666666667e-05, + "loss": 1.3728, + "step": 1118 + }, + { + "epoch": 0.03, + "learning_rate": 1.793269230769231e-05, + "loss": 1.0856, + "step": 1119 + }, + { + "epoch": 0.03, + "learning_rate": 1.794871794871795e-05, + "loss": 1.1882, + "step": 1120 + }, + { + "epoch": 0.03, + "learning_rate": 1.796474358974359e-05, + "loss": 1.2001, + "step": 1121 + }, + { + "epoch": 0.03, + "learning_rate": 1.798076923076923e-05, + "loss": 1.1222, + "step": 1122 + }, + { + "epoch": 0.03, + "learning_rate": 1.7996794871794875e-05, + "loss": 1.3825, + "step": 1123 + }, + { + "epoch": 0.03, + "learning_rate": 1.8012820512820515e-05, + "loss": 1.0243, + "step": 1124 + }, + { + "epoch": 0.03, + "learning_rate": 1.8028846153846156e-05, + "loss": 1.4666, + "step": 1125 + }, + { + "epoch": 0.03, + "learning_rate": 1.8044871794871796e-05, + "loss": 1.1345, + "step": 1126 + }, + { + "epoch": 0.03, + "learning_rate": 1.8060897435897436e-05, + "loss": 1.2714, + "step": 1127 + }, + { + "epoch": 0.03, + "learning_rate": 1.807692307692308e-05, + "loss": 1.2481, + "step": 1128 + }, + { + "epoch": 0.03, + "learning_rate": 1.809294871794872e-05, + "loss": 1.2921, + "step": 1129 + }, + { + "epoch": 0.03, + "learning_rate": 1.810897435897436e-05, + "loss": 1.2755, + "step": 1130 + }, + { + "epoch": 0.03, + "learning_rate": 1.8125e-05, + "loss": 1.2666, + "step": 1131 + }, + { + "epoch": 0.03, + "learning_rate": 1.8141025641025644e-05, + "loss": 1.1566, + "step": 1132 + }, + { + "epoch": 0.03, + "learning_rate": 1.8157051282051284e-05, + "loss": 1.3301, + "step": 1133 + }, + { + "epoch": 0.03, + "learning_rate": 1.8173076923076924e-05, + "loss": 1.2122, + "step": 1134 + }, + { + "epoch": 0.03, + "learning_rate": 1.8189102564102564e-05, + "loss": 1.3025, + "step": 1135 + }, + { + "epoch": 0.03, + "learning_rate": 1.8205128205128208e-05, + "loss": 1.1605, + "step": 1136 + }, + { + "epoch": 0.03, + "learning_rate": 1.8221153846153848e-05, + "loss": 1.3351, + "step": 1137 + }, + { + "epoch": 0.03, + "learning_rate": 1.8237179487179488e-05, + "loss": 0.9839, + "step": 1138 + }, + { + "epoch": 0.03, + "learning_rate": 1.825320512820513e-05, + "loss": 1.2086, + "step": 1139 + }, + { + "epoch": 0.03, + "learning_rate": 1.826923076923077e-05, + "loss": 1.1348, + "step": 1140 + }, + { + "epoch": 0.03, + "learning_rate": 1.8285256410256412e-05, + "loss": 1.184, + "step": 1141 + }, + { + "epoch": 0.03, + "learning_rate": 1.8301282051282052e-05, + "loss": 1.1468, + "step": 1142 + }, + { + "epoch": 0.03, + "learning_rate": 1.8317307692307693e-05, + "loss": 1.1834, + "step": 1143 + }, + { + "epoch": 0.03, + "learning_rate": 1.8333333333333333e-05, + "loss": 1.0399, + "step": 1144 + }, + { + "epoch": 0.03, + "learning_rate": 1.8349358974358976e-05, + "loss": 1.1462, + "step": 1145 + }, + { + "epoch": 0.03, + "learning_rate": 1.8365384615384617e-05, + "loss": 1.2575, + "step": 1146 + }, + { + "epoch": 0.03, + "learning_rate": 1.8381410256410257e-05, + "loss": 1.3008, + "step": 1147 + }, + { + "epoch": 0.03, + "learning_rate": 1.8397435897435897e-05, + "loss": 1.3056, + "step": 1148 + }, + { + "epoch": 0.03, + "learning_rate": 1.841346153846154e-05, + "loss": 1.1225, + "step": 1149 + }, + { + "epoch": 0.03, + "learning_rate": 1.842948717948718e-05, + "loss": 1.3262, + "step": 1150 + }, + { + "epoch": 0.03, + "learning_rate": 1.8445512820512824e-05, + "loss": 1.3117, + "step": 1151 + }, + { + "epoch": 0.03, + "learning_rate": 1.8461538461538465e-05, + "loss": 1.091, + "step": 1152 + }, + { + "epoch": 0.03, + "learning_rate": 1.8477564102564105e-05, + "loss": 0.9532, + "step": 1153 + }, + { + "epoch": 0.03, + "learning_rate": 1.8493589743589745e-05, + "loss": 1.1966, + "step": 1154 + }, + { + "epoch": 0.03, + "learning_rate": 1.850961538461539e-05, + "loss": 1.293, + "step": 1155 + }, + { + "epoch": 0.03, + "learning_rate": 1.852564102564103e-05, + "loss": 1.3953, + "step": 1156 + }, + { + "epoch": 0.03, + "learning_rate": 1.854166666666667e-05, + "loss": 1.106, + "step": 1157 + }, + { + "epoch": 0.03, + "learning_rate": 1.855769230769231e-05, + "loss": 1.2534, + "step": 1158 + }, + { + "epoch": 0.03, + "learning_rate": 1.857371794871795e-05, + "loss": 1.0298, + "step": 1159 + }, + { + "epoch": 0.03, + "learning_rate": 1.8589743589743593e-05, + "loss": 1.1465, + "step": 1160 + }, + { + "epoch": 0.03, + "learning_rate": 1.8605769230769233e-05, + "loss": 1.2744, + "step": 1161 + }, + { + "epoch": 0.03, + "learning_rate": 1.8621794871794873e-05, + "loss": 1.1241, + "step": 1162 + }, + { + "epoch": 0.03, + "learning_rate": 1.8637820512820514e-05, + "loss": 1.1692, + "step": 1163 + }, + { + "epoch": 0.03, + "learning_rate": 1.8653846153846157e-05, + "loss": 1.32, + "step": 1164 + }, + { + "epoch": 0.03, + "learning_rate": 1.8669871794871797e-05, + "loss": 1.0272, + "step": 1165 + }, + { + "epoch": 0.03, + "learning_rate": 1.8685897435897438e-05, + "loss": 1.204, + "step": 1166 + }, + { + "epoch": 0.03, + "learning_rate": 1.8701923076923078e-05, + "loss": 1.0095, + "step": 1167 + }, + { + "epoch": 0.03, + "learning_rate": 1.8717948717948718e-05, + "loss": 1.354, + "step": 1168 + }, + { + "epoch": 0.03, + "learning_rate": 1.873397435897436e-05, + "loss": 1.2202, + "step": 1169 + }, + { + "epoch": 0.03, + "learning_rate": 1.8750000000000002e-05, + "loss": 1.2493, + "step": 1170 + }, + { + "epoch": 0.03, + "learning_rate": 1.8766025641025642e-05, + "loss": 1.3089, + "step": 1171 + }, + { + "epoch": 0.03, + "learning_rate": 1.8782051282051282e-05, + "loss": 1.4443, + "step": 1172 + }, + { + "epoch": 0.03, + "learning_rate": 1.8798076923076926e-05, + "loss": 1.4023, + "step": 1173 + }, + { + "epoch": 0.03, + "learning_rate": 1.8814102564102566e-05, + "loss": 1.0788, + "step": 1174 + }, + { + "epoch": 0.03, + "learning_rate": 1.8830128205128206e-05, + "loss": 1.4082, + "step": 1175 + }, + { + "epoch": 0.03, + "learning_rate": 1.8846153846153846e-05, + "loss": 1.1345, + "step": 1176 + }, + { + "epoch": 0.03, + "learning_rate": 1.886217948717949e-05, + "loss": 1.2945, + "step": 1177 + }, + { + "epoch": 0.03, + "learning_rate": 1.887820512820513e-05, + "loss": 1.3056, + "step": 1178 + }, + { + "epoch": 0.03, + "learning_rate": 1.889423076923077e-05, + "loss": 1.0623, + "step": 1179 + }, + { + "epoch": 0.03, + "learning_rate": 1.891025641025641e-05, + "loss": 0.9963, + "step": 1180 + }, + { + "epoch": 0.03, + "learning_rate": 1.892628205128205e-05, + "loss": 1.0568, + "step": 1181 + }, + { + "epoch": 0.03, + "learning_rate": 1.8942307692307694e-05, + "loss": 1.1637, + "step": 1182 + }, + { + "epoch": 0.03, + "learning_rate": 1.8958333333333334e-05, + "loss": 1.2266, + "step": 1183 + }, + { + "epoch": 0.03, + "learning_rate": 1.8974358974358975e-05, + "loss": 1.3306, + "step": 1184 + }, + { + "epoch": 0.03, + "learning_rate": 1.8990384615384615e-05, + "loss": 1.2611, + "step": 1185 + }, + { + "epoch": 0.03, + "learning_rate": 1.900641025641026e-05, + "loss": 1.3534, + "step": 1186 + }, + { + "epoch": 0.03, + "learning_rate": 1.90224358974359e-05, + "loss": 1.0419, + "step": 1187 + }, + { + "epoch": 0.03, + "learning_rate": 1.903846153846154e-05, + "loss": 1.1463, + "step": 1188 + }, + { + "epoch": 0.03, + "learning_rate": 1.905448717948718e-05, + "loss": 0.8268, + "step": 1189 + }, + { + "epoch": 0.03, + "learning_rate": 1.9070512820512823e-05, + "loss": 1.0599, + "step": 1190 + }, + { + "epoch": 0.03, + "learning_rate": 1.9086538461538463e-05, + "loss": 1.3067, + "step": 1191 + }, + { + "epoch": 0.03, + "learning_rate": 1.9102564102564106e-05, + "loss": 0.9027, + "step": 1192 + }, + { + "epoch": 0.03, + "learning_rate": 1.9118589743589747e-05, + "loss": 1.0295, + "step": 1193 + }, + { + "epoch": 0.03, + "learning_rate": 1.9134615384615387e-05, + "loss": 1.2402, + "step": 1194 + }, + { + "epoch": 0.03, + "learning_rate": 1.9150641025641027e-05, + "loss": 1.3678, + "step": 1195 + }, + { + "epoch": 0.03, + "learning_rate": 1.916666666666667e-05, + "loss": 1.0075, + "step": 1196 + }, + { + "epoch": 0.03, + "learning_rate": 1.918269230769231e-05, + "loss": 1.0553, + "step": 1197 + }, + { + "epoch": 0.03, + "learning_rate": 1.919871794871795e-05, + "loss": 1.4548, + "step": 1198 + }, + { + "epoch": 0.03, + "learning_rate": 1.921474358974359e-05, + "loss": 1.1941, + "step": 1199 + }, + { + "epoch": 0.03, + "learning_rate": 1.923076923076923e-05, + "loss": 1.1166, + "step": 1200 + }, + { + "epoch": 0.03, + "learning_rate": 1.9246794871794875e-05, + "loss": 1.1436, + "step": 1201 + }, + { + "epoch": 0.03, + "learning_rate": 1.9262820512820515e-05, + "loss": 1.119, + "step": 1202 + }, + { + "epoch": 0.03, + "learning_rate": 1.9278846153846155e-05, + "loss": 1.3263, + "step": 1203 + }, + { + "epoch": 0.03, + "learning_rate": 1.9294871794871796e-05, + "loss": 1.5347, + "step": 1204 + }, + { + "epoch": 0.03, + "learning_rate": 1.931089743589744e-05, + "loss": 1.0763, + "step": 1205 + }, + { + "epoch": 0.03, + "learning_rate": 1.932692307692308e-05, + "loss": 1.2518, + "step": 1206 + }, + { + "epoch": 0.03, + "learning_rate": 1.934294871794872e-05, + "loss": 1.1188, + "step": 1207 + }, + { + "epoch": 0.03, + "learning_rate": 1.935897435897436e-05, + "loss": 1.2971, + "step": 1208 + }, + { + "epoch": 0.03, + "learning_rate": 1.9375e-05, + "loss": 1.122, + "step": 1209 + }, + { + "epoch": 0.03, + "learning_rate": 1.9391025641025644e-05, + "loss": 1.1307, + "step": 1210 + }, + { + "epoch": 0.03, + "learning_rate": 1.9407051282051284e-05, + "loss": 1.104, + "step": 1211 + }, + { + "epoch": 0.03, + "learning_rate": 1.9423076923076924e-05, + "loss": 1.1814, + "step": 1212 + }, + { + "epoch": 0.03, + "learning_rate": 1.9439102564102564e-05, + "loss": 1.1465, + "step": 1213 + }, + { + "epoch": 0.03, + "learning_rate": 1.9455128205128208e-05, + "loss": 1.1183, + "step": 1214 + }, + { + "epoch": 0.03, + "learning_rate": 1.9471153846153848e-05, + "loss": 1.0705, + "step": 1215 + }, + { + "epoch": 0.03, + "learning_rate": 1.9487179487179488e-05, + "loss": 1.2661, + "step": 1216 + }, + { + "epoch": 0.03, + "learning_rate": 1.950320512820513e-05, + "loss": 1.3433, + "step": 1217 + }, + { + "epoch": 0.03, + "learning_rate": 1.9519230769230772e-05, + "loss": 1.0297, + "step": 1218 + }, + { + "epoch": 0.03, + "learning_rate": 1.9535256410256412e-05, + "loss": 1.426, + "step": 1219 + }, + { + "epoch": 0.03, + "learning_rate": 1.9551282051282052e-05, + "loss": 0.9806, + "step": 1220 + }, + { + "epoch": 0.03, + "learning_rate": 1.9567307692307693e-05, + "loss": 1.1133, + "step": 1221 + }, + { + "epoch": 0.03, + "learning_rate": 1.9583333333333333e-05, + "loss": 1.2104, + "step": 1222 + }, + { + "epoch": 0.03, + "learning_rate": 1.9599358974358976e-05, + "loss": 1.1411, + "step": 1223 + }, + { + "epoch": 0.03, + "learning_rate": 1.9615384615384617e-05, + "loss": 1.3382, + "step": 1224 + }, + { + "epoch": 0.03, + "learning_rate": 1.9631410256410257e-05, + "loss": 1.2595, + "step": 1225 + }, + { + "epoch": 0.03, + "learning_rate": 1.9647435897435897e-05, + "loss": 1.1339, + "step": 1226 + }, + { + "epoch": 0.03, + "learning_rate": 1.966346153846154e-05, + "loss": 1.1595, + "step": 1227 + }, + { + "epoch": 0.03, + "learning_rate": 1.967948717948718e-05, + "loss": 1.0189, + "step": 1228 + }, + { + "epoch": 0.03, + "learning_rate": 1.9695512820512824e-05, + "loss": 1.2835, + "step": 1229 + }, + { + "epoch": 0.03, + "learning_rate": 1.9711538461538465e-05, + "loss": 1.322, + "step": 1230 + }, + { + "epoch": 0.03, + "learning_rate": 1.9727564102564105e-05, + "loss": 1.1665, + "step": 1231 + }, + { + "epoch": 0.03, + "learning_rate": 1.9743589743589745e-05, + "loss": 1.1738, + "step": 1232 + }, + { + "epoch": 0.03, + "learning_rate": 1.975961538461539e-05, + "loss": 1.1692, + "step": 1233 + }, + { + "epoch": 0.03, + "learning_rate": 1.977564102564103e-05, + "loss": 1.1725, + "step": 1234 + }, + { + "epoch": 0.03, + "learning_rate": 1.979166666666667e-05, + "loss": 1.3031, + "step": 1235 + }, + { + "epoch": 0.03, + "learning_rate": 1.980769230769231e-05, + "loss": 1.299, + "step": 1236 + }, + { + "epoch": 0.03, + "learning_rate": 1.982371794871795e-05, + "loss": 1.1124, + "step": 1237 + }, + { + "epoch": 0.03, + "learning_rate": 1.9839743589743593e-05, + "loss": 0.9778, + "step": 1238 + }, + { + "epoch": 0.03, + "learning_rate": 1.9855769230769233e-05, + "loss": 1.1593, + "step": 1239 + }, + { + "epoch": 0.03, + "learning_rate": 1.9871794871794873e-05, + "loss": 1.296, + "step": 1240 + }, + { + "epoch": 0.03, + "learning_rate": 1.9887820512820513e-05, + "loss": 1.2494, + "step": 1241 + }, + { + "epoch": 0.03, + "learning_rate": 1.9903846153846157e-05, + "loss": 1.176, + "step": 1242 + }, + { + "epoch": 0.03, + "learning_rate": 1.9919871794871797e-05, + "loss": 1.005, + "step": 1243 + }, + { + "epoch": 0.03, + "learning_rate": 1.9935897435897437e-05, + "loss": 1.2495, + "step": 1244 + }, + { + "epoch": 0.03, + "learning_rate": 1.9951923076923078e-05, + "loss": 1.359, + "step": 1245 + }, + { + "epoch": 0.03, + "learning_rate": 1.996794871794872e-05, + "loss": 1.1797, + "step": 1246 + }, + { + "epoch": 0.03, + "learning_rate": 1.998397435897436e-05, + "loss": 1.4007, + "step": 1247 + }, + { + "epoch": 0.03, + "learning_rate": 2e-05, + "loss": 1.4174, + "step": 1248 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999999969666176e-05, + "loss": 0.9298, + "step": 1249 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999999878664707e-05, + "loss": 0.9737, + "step": 1250 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999999726995587e-05, + "loss": 1.2171, + "step": 1251 + }, + { + "epoch": 0.03, + "learning_rate": 1.999999951465882e-05, + "loss": 1.4125, + "step": 1252 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999999241654413e-05, + "loss": 1.3462, + "step": 1253 + }, + { + "epoch": 0.03, + "learning_rate": 1.999999890798236e-05, + "loss": 1.0398, + "step": 1254 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999998513642666e-05, + "loss": 1.306, + "step": 1255 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999998058635333e-05, + "loss": 1.0299, + "step": 1256 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999997542960364e-05, + "loss": 1.0835, + "step": 1257 + }, + { + "epoch": 0.03, + "learning_rate": 1.999999696661776e-05, + "loss": 1.3478, + "step": 1258 + }, + { + "epoch": 0.03, + "learning_rate": 1.999999632960753e-05, + "loss": 1.4534, + "step": 1259 + }, + { + "epoch": 0.03, + "learning_rate": 1.999999563192967e-05, + "loss": 1.2274, + "step": 1260 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999994873584193e-05, + "loss": 1.2098, + "step": 1261 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999994054571096e-05, + "loss": 1.5152, + "step": 1262 + }, + { + "epoch": 0.03, + "learning_rate": 1.999999317489039e-05, + "loss": 1.1578, + "step": 1263 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999992234542078e-05, + "loss": 1.2717, + "step": 1264 + }, + { + "epoch": 0.03, + "learning_rate": 1.999999123352616e-05, + "loss": 1.3293, + "step": 1265 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999990171842654e-05, + "loss": 1.4355, + "step": 1266 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999989049491554e-05, + "loss": 1.2815, + "step": 1267 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999987866472878e-05, + "loss": 1.2314, + "step": 1268 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999986622786626e-05, + "loss": 1.0135, + "step": 1269 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999985318432804e-05, + "loss": 1.5132, + "step": 1270 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999983953411427e-05, + "loss": 0.9785, + "step": 1271 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999982527722498e-05, + "loss": 1.2947, + "step": 1272 + }, + { + "epoch": 0.03, + "learning_rate": 1.999998104136603e-05, + "loss": 1.4669, + "step": 1273 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999979494342022e-05, + "loss": 1.44, + "step": 1274 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999977886650496e-05, + "loss": 1.1507, + "step": 1275 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999976218291455e-05, + "loss": 1.2684, + "step": 1276 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999974489264913e-05, + "loss": 0.8852, + "step": 1277 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999972699570876e-05, + "loss": 1.4039, + "step": 1278 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999970849209354e-05, + "loss": 1.3602, + "step": 1279 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999968938180364e-05, + "loss": 1.3781, + "step": 1280 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999966966483914e-05, + "loss": 1.3144, + "step": 1281 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999964934120016e-05, + "loss": 1.2759, + "step": 1282 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999962841088684e-05, + "loss": 1.2749, + "step": 1283 + }, + { + "epoch": 0.03, + "learning_rate": 1.999996068738993e-05, + "loss": 1.2478, + "step": 1284 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999958473023767e-05, + "loss": 1.1803, + "step": 1285 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999956197990205e-05, + "loss": 1.1089, + "step": 1286 + }, + { + "epoch": 0.03, + "learning_rate": 1.999995386228926e-05, + "loss": 1.2705, + "step": 1287 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999951465920953e-05, + "loss": 1.2163, + "step": 1288 + }, + { + "epoch": 0.03, + "learning_rate": 1.999994900888529e-05, + "loss": 1.371, + "step": 1289 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999946491182284e-05, + "loss": 0.9926, + "step": 1290 + }, + { + "epoch": 0.03, + "learning_rate": 1.999994391281196e-05, + "loss": 1.0113, + "step": 1291 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999941273774327e-05, + "loss": 1.1501, + "step": 1292 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999938574069403e-05, + "loss": 1.2975, + "step": 1293 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999935813697204e-05, + "loss": 1.1903, + "step": 1294 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999932992657744e-05, + "loss": 1.3187, + "step": 1295 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999930110951043e-05, + "loss": 1.0846, + "step": 1296 + }, + { + "epoch": 0.03, + "learning_rate": 1.999992716857712e-05, + "loss": 1.2711, + "step": 1297 + }, + { + "epoch": 0.03, + "learning_rate": 1.999992416553599e-05, + "loss": 1.1284, + "step": 1298 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999921101827674e-05, + "loss": 1.2498, + "step": 1299 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999917977452187e-05, + "loss": 1.2353, + "step": 1300 + }, + { + "epoch": 0.03, + "learning_rate": 1.999991479240955e-05, + "loss": 1.2387, + "step": 1301 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999911546699785e-05, + "loss": 1.3747, + "step": 1302 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999908240322905e-05, + "loss": 0.986, + "step": 1303 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999904873278933e-05, + "loss": 1.435, + "step": 1304 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999901445567893e-05, + "loss": 1.2871, + "step": 1305 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999897957189802e-05, + "loss": 1.321, + "step": 1306 + }, + { + "epoch": 0.03, + "learning_rate": 1.999989440814468e-05, + "loss": 1.1524, + "step": 1307 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999890798432556e-05, + "loss": 1.0598, + "step": 1308 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999887128053444e-05, + "loss": 1.2085, + "step": 1309 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999883397007366e-05, + "loss": 1.3088, + "step": 1310 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999879605294353e-05, + "loss": 1.4097, + "step": 1311 + }, + { + "epoch": 0.03, + "learning_rate": 1.999987575291442e-05, + "loss": 1.3317, + "step": 1312 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999871839867593e-05, + "loss": 1.1172, + "step": 1313 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999867866153894e-05, + "loss": 1.2164, + "step": 1314 + }, + { + "epoch": 0.03, + "learning_rate": 1.999986383177335e-05, + "loss": 1.3177, + "step": 1315 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999859736725984e-05, + "loss": 1.2634, + "step": 1316 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999855581011823e-05, + "loss": 1.3597, + "step": 1317 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999851364630886e-05, + "loss": 1.2813, + "step": 1318 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999847087583206e-05, + "loss": 1.3083, + "step": 1319 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999842749868808e-05, + "loss": 1.4245, + "step": 1320 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999838351487714e-05, + "loss": 1.277, + "step": 1321 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999833892439952e-05, + "loss": 1.1994, + "step": 1322 + }, + { + "epoch": 0.03, + "learning_rate": 1.999982937272555e-05, + "loss": 1.2631, + "step": 1323 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999824792344536e-05, + "loss": 1.0925, + "step": 1324 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999820151296936e-05, + "loss": 1.2981, + "step": 1325 + }, + { + "epoch": 0.03, + "learning_rate": 1.999981544958278e-05, + "loss": 1.0645, + "step": 1326 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999810687202098e-05, + "loss": 1.3291, + "step": 1327 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999805864154913e-05, + "loss": 1.1396, + "step": 1328 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999800980441258e-05, + "loss": 1.1837, + "step": 1329 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999796036061164e-05, + "loss": 0.9636, + "step": 1330 + }, + { + "epoch": 0.03, + "learning_rate": 1.999979103101466e-05, + "loss": 1.1766, + "step": 1331 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999785965301776e-05, + "loss": 1.2311, + "step": 1332 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999780838922542e-05, + "loss": 1.2466, + "step": 1333 + }, + { + "epoch": 0.03, + "learning_rate": 1.999977565187699e-05, + "loss": 1.1779, + "step": 1334 + }, + { + "epoch": 0.03, + "learning_rate": 1.999977040416515e-05, + "loss": 1.2761, + "step": 1335 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999765095787055e-05, + "loss": 1.2989, + "step": 1336 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999759726742738e-05, + "loss": 1.2165, + "step": 1337 + }, + { + "epoch": 0.03, + "learning_rate": 1.999975429703223e-05, + "loss": 1.2526, + "step": 1338 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999748806655567e-05, + "loss": 0.9768, + "step": 1339 + }, + { + "epoch": 0.03, + "learning_rate": 1.999974325561278e-05, + "loss": 0.9567, + "step": 1340 + }, + { + "epoch": 0.03, + "learning_rate": 1.99997376439039e-05, + "loss": 1.3212, + "step": 1341 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999731971528965e-05, + "loss": 1.2293, + "step": 1342 + }, + { + "epoch": 0.03, + "learning_rate": 1.999972623848801e-05, + "loss": 1.1117, + "step": 1343 + }, + { + "epoch": 0.03, + "learning_rate": 1.999972044478107e-05, + "loss": 1.2992, + "step": 1344 + }, + { + "epoch": 0.03, + "learning_rate": 1.999971459040817e-05, + "loss": 1.596, + "step": 1345 + }, + { + "epoch": 0.03, + "learning_rate": 1.999970867536936e-05, + "loss": 1.2953, + "step": 1346 + }, + { + "epoch": 0.03, + "learning_rate": 1.999970269966467e-05, + "loss": 1.3181, + "step": 1347 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999696663294133e-05, + "loss": 1.1556, + "step": 1348 + }, + { + "epoch": 0.03, + "learning_rate": 1.999969056625779e-05, + "loss": 1.037, + "step": 1349 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999684408555673e-05, + "loss": 1.2297, + "step": 1350 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999678190187825e-05, + "loss": 1.0344, + "step": 1351 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999671911154285e-05, + "loss": 1.3408, + "step": 1352 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999665571455083e-05, + "loss": 1.2202, + "step": 1353 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999659171090263e-05, + "loss": 1.0181, + "step": 1354 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999652710059862e-05, + "loss": 1.3968, + "step": 1355 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999646188363925e-05, + "loss": 1.2858, + "step": 1356 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999639606002482e-05, + "loss": 1.2412, + "step": 1357 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999632962975578e-05, + "loss": 1.1295, + "step": 1358 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999626259283255e-05, + "loss": 1.1696, + "step": 1359 + }, + { + "epoch": 0.03, + "learning_rate": 1.999961949492555e-05, + "loss": 0.836, + "step": 1360 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999612669902505e-05, + "loss": 1.2428, + "step": 1361 + }, + { + "epoch": 0.03, + "learning_rate": 1.999960578421416e-05, + "loss": 1.1958, + "step": 1362 + }, + { + "epoch": 0.03, + "learning_rate": 1.999959883786056e-05, + "loss": 1.1991, + "step": 1363 + }, + { + "epoch": 0.03, + "learning_rate": 1.999959183084175e-05, + "loss": 1.1968, + "step": 1364 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999584763157763e-05, + "loss": 1.0284, + "step": 1365 + }, + { + "epoch": 0.03, + "learning_rate": 1.999957763480865e-05, + "loss": 1.1723, + "step": 1366 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999570445794455e-05, + "loss": 1.1535, + "step": 1367 + }, + { + "epoch": 0.03, + "learning_rate": 1.999956319611521e-05, + "loss": 1.2751, + "step": 1368 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999555885770973e-05, + "loss": 1.1175, + "step": 1369 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999548514761785e-05, + "loss": 1.2042, + "step": 1370 + }, + { + "epoch": 0.03, + "learning_rate": 1.999954108308768e-05, + "loss": 0.8889, + "step": 1371 + }, + { + "epoch": 0.03, + "learning_rate": 1.999953359074872e-05, + "loss": 1.0001, + "step": 1372 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999526037744935e-05, + "loss": 1.0883, + "step": 1373 + }, + { + "epoch": 0.03, + "learning_rate": 1.999951842407638e-05, + "loss": 1.1374, + "step": 1374 + }, + { + "epoch": 0.03, + "learning_rate": 1.99995107497431e-05, + "loss": 1.1038, + "step": 1375 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999503014745138e-05, + "loss": 1.2209, + "step": 1376 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999495219082544e-05, + "loss": 1.224, + "step": 1377 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999487362755366e-05, + "loss": 1.5973, + "step": 1378 + }, + { + "epoch": 0.03, + "learning_rate": 1.999947944576365e-05, + "loss": 1.3464, + "step": 1379 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999471468107444e-05, + "loss": 1.1595, + "step": 1380 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999463429786794e-05, + "loss": 1.2573, + "step": 1381 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999455330801752e-05, + "loss": 1.3009, + "step": 1382 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999447171152367e-05, + "loss": 1.0309, + "step": 1383 + }, + { + "epoch": 0.03, + "learning_rate": 1.999943895083869e-05, + "loss": 1.3895, + "step": 1384 + }, + { + "epoch": 0.03, + "learning_rate": 1.999943066986077e-05, + "loss": 1.2544, + "step": 1385 + }, + { + "epoch": 0.03, + "learning_rate": 1.999942232821865e-05, + "loss": 1.2299, + "step": 1386 + }, + { + "epoch": 0.03, + "learning_rate": 1.999941392591239e-05, + "loss": 1.2764, + "step": 1387 + }, + { + "epoch": 0.03, + "learning_rate": 1.999940546294204e-05, + "loss": 1.2536, + "step": 1388 + }, + { + "epoch": 0.03, + "learning_rate": 1.999939693930765e-05, + "loss": 1.3883, + "step": 1389 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999388355009266e-05, + "loss": 1.0295, + "step": 1390 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999379710046947e-05, + "loss": 1.1164, + "step": 1391 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999371004420744e-05, + "loss": 1.3313, + "step": 1392 + }, + { + "epoch": 0.03, + "learning_rate": 1.999936223813071e-05, + "loss": 1.043, + "step": 1393 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999353411176893e-05, + "loss": 1.001, + "step": 1394 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999344523559353e-05, + "loss": 1.1436, + "step": 1395 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999335575278143e-05, + "loss": 1.2262, + "step": 1396 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999326566333318e-05, + "loss": 1.2918, + "step": 1397 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999317496724925e-05, + "loss": 1.1104, + "step": 1398 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999308366453028e-05, + "loss": 1.3828, + "step": 1399 + }, + { + "epoch": 0.03, + "learning_rate": 1.999929917551768e-05, + "loss": 0.9822, + "step": 1400 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999289923918934e-05, + "loss": 1.2053, + "step": 1401 + }, + { + "epoch": 0.03, + "learning_rate": 1.999928061165685e-05, + "loss": 1.1398, + "step": 1402 + }, + { + "epoch": 0.03, + "learning_rate": 1.999927123873148e-05, + "loss": 1.1845, + "step": 1403 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999261805142885e-05, + "loss": 1.2833, + "step": 1404 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999252310891123e-05, + "loss": 1.5314, + "step": 1405 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999242755976246e-05, + "loss": 1.2922, + "step": 1406 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999233140398317e-05, + "loss": 1.264, + "step": 1407 + }, + { + "epoch": 0.03, + "learning_rate": 1.999922346415739e-05, + "loss": 1.2557, + "step": 1408 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999213727253526e-05, + "loss": 1.2299, + "step": 1409 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999203929686786e-05, + "loss": 1.1035, + "step": 1410 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999194071457228e-05, + "loss": 1.0328, + "step": 1411 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999184152564907e-05, + "loss": 1.3728, + "step": 1412 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999174173009893e-05, + "loss": 1.215, + "step": 1413 + }, + { + "epoch": 0.03, + "learning_rate": 1.999916413279224e-05, + "loss": 1.0902, + "step": 1414 + }, + { + "epoch": 0.03, + "learning_rate": 1.999915403191201e-05, + "loss": 0.9522, + "step": 1415 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999143870369265e-05, + "loss": 1.1598, + "step": 1416 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999133648164065e-05, + "loss": 1.2061, + "step": 1417 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999123365296473e-05, + "loss": 1.5061, + "step": 1418 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999113021766553e-05, + "loss": 1.1618, + "step": 1419 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999102617574366e-05, + "loss": 1.3579, + "step": 1420 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999092152719974e-05, + "loss": 1.1855, + "step": 1421 + }, + { + "epoch": 0.03, + "learning_rate": 1.999908162720344e-05, + "loss": 1.0284, + "step": 1422 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999071041024835e-05, + "loss": 0.9711, + "step": 1423 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999060394184214e-05, + "loss": 1.3943, + "step": 1424 + }, + { + "epoch": 0.03, + "learning_rate": 1.999904968668165e-05, + "loss": 1.2027, + "step": 1425 + }, + { + "epoch": 0.03, + "learning_rate": 1.99990389185172e-05, + "loss": 1.192, + "step": 1426 + }, + { + "epoch": 0.03, + "learning_rate": 1.999902808969093e-05, + "loss": 1.2666, + "step": 1427 + }, + { + "epoch": 0.03, + "learning_rate": 1.999901720020291e-05, + "loss": 1.0164, + "step": 1428 + }, + { + "epoch": 0.03, + "learning_rate": 1.999900625005321e-05, + "loss": 0.9871, + "step": 1429 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998995239241883e-05, + "loss": 1.035, + "step": 1430 + }, + { + "epoch": 0.03, + "learning_rate": 1.999898416776901e-05, + "loss": 1.0287, + "step": 1431 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998973035634648e-05, + "loss": 1.314, + "step": 1432 + }, + { + "epoch": 0.03, + "learning_rate": 1.999896184283887e-05, + "loss": 1.1673, + "step": 1433 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998950589381743e-05, + "loss": 1.1883, + "step": 1434 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998939275263335e-05, + "loss": 1.2013, + "step": 1435 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998927900483714e-05, + "loss": 1.1683, + "step": 1436 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998916465042948e-05, + "loss": 1.0864, + "step": 1437 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998904968941107e-05, + "loss": 0.8972, + "step": 1438 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998893412178263e-05, + "loss": 0.949, + "step": 1439 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998881794754484e-05, + "loss": 1.1225, + "step": 1440 + }, + { + "epoch": 0.03, + "learning_rate": 1.999887011666984e-05, + "loss": 1.0102, + "step": 1441 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998858377924408e-05, + "loss": 1.2887, + "step": 1442 + }, + { + "epoch": 0.03, + "learning_rate": 1.999884657851825e-05, + "loss": 1.1755, + "step": 1443 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998834718451444e-05, + "loss": 1.0595, + "step": 1444 + }, + { + "epoch": 0.03, + "learning_rate": 1.999882279772406e-05, + "loss": 1.2128, + "step": 1445 + }, + { + "epoch": 0.03, + "learning_rate": 1.999881081633616e-05, + "loss": 1.0735, + "step": 1446 + }, + { + "epoch": 0.03, + "learning_rate": 1.999879877428784e-05, + "loss": 0.8989, + "step": 1447 + }, + { + "epoch": 0.03, + "learning_rate": 1.999878667157915e-05, + "loss": 1.2106, + "step": 1448 + }, + { + "epoch": 0.03, + "learning_rate": 1.999877450821018e-05, + "loss": 1.2713, + "step": 1449 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998762284180996e-05, + "loss": 1.084, + "step": 1450 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998749999491667e-05, + "loss": 1.2648, + "step": 1451 + }, + { + "epoch": 0.03, + "learning_rate": 1.999873765414228e-05, + "loss": 1.3176, + "step": 1452 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998725248132902e-05, + "loss": 1.1616, + "step": 1453 + }, + { + "epoch": 0.03, + "learning_rate": 1.999871278146361e-05, + "loss": 0.9472, + "step": 1454 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998700254134477e-05, + "loss": 1.1634, + "step": 1455 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998687666145585e-05, + "loss": 1.0374, + "step": 1456 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998675017497e-05, + "loss": 1.2128, + "step": 1457 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998662308188813e-05, + "loss": 1.0847, + "step": 1458 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998649538221092e-05, + "loss": 1.3024, + "step": 1459 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998636707593913e-05, + "loss": 0.9743, + "step": 1460 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998623816307357e-05, + "loss": 1.372, + "step": 1461 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998610864361506e-05, + "loss": 0.8961, + "step": 1462 + }, + { + "epoch": 0.04, + "learning_rate": 1.999859785175643e-05, + "loss": 1.1006, + "step": 1463 + }, + { + "epoch": 0.04, + "learning_rate": 1.999858477849222e-05, + "loss": 1.1144, + "step": 1464 + }, + { + "epoch": 0.04, + "learning_rate": 1.999857164456894e-05, + "loss": 1.0424, + "step": 1465 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998558449986684e-05, + "loss": 1.2081, + "step": 1466 + }, + { + "epoch": 0.04, + "learning_rate": 1.999854519474552e-05, + "loss": 1.2114, + "step": 1467 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998531878845536e-05, + "loss": 0.8261, + "step": 1468 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998518502286815e-05, + "loss": 1.1029, + "step": 1469 + }, + { + "epoch": 0.04, + "learning_rate": 1.999850506506943e-05, + "loss": 1.1489, + "step": 1470 + }, + { + "epoch": 0.04, + "learning_rate": 1.999849156719347e-05, + "loss": 1.3614, + "step": 1471 + }, + { + "epoch": 0.04, + "learning_rate": 1.999847800865901e-05, + "loss": 1.0812, + "step": 1472 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998464389466132e-05, + "loss": 1.1438, + "step": 1473 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998450709614928e-05, + "loss": 1.2605, + "step": 1474 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998436969105475e-05, + "loss": 1.49, + "step": 1475 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998423167937852e-05, + "loss": 1.0843, + "step": 1476 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998409306112153e-05, + "loss": 1.264, + "step": 1477 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998395383628457e-05, + "loss": 1.1757, + "step": 1478 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998381400486843e-05, + "loss": 1.1594, + "step": 1479 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998367356687405e-05, + "loss": 1.3397, + "step": 1480 + }, + { + "epoch": 0.04, + "learning_rate": 1.999835325223022e-05, + "loss": 1.0883, + "step": 1481 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998339087115378e-05, + "loss": 1.3774, + "step": 1482 + }, + { + "epoch": 0.04, + "learning_rate": 1.999832486134297e-05, + "loss": 1.0554, + "step": 1483 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998310574913074e-05, + "loss": 1.3375, + "step": 1484 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998296227825778e-05, + "loss": 1.0207, + "step": 1485 + }, + { + "epoch": 0.04, + "learning_rate": 1.999828182008117e-05, + "loss": 1.3285, + "step": 1486 + }, + { + "epoch": 0.04, + "learning_rate": 1.999826735167934e-05, + "loss": 1.3071, + "step": 1487 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998252822620373e-05, + "loss": 1.2052, + "step": 1488 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998238232904355e-05, + "loss": 1.2561, + "step": 1489 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998223582531386e-05, + "loss": 1.3053, + "step": 1490 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998208871501537e-05, + "loss": 1.3821, + "step": 1491 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998194099814913e-05, + "loss": 1.1122, + "step": 1492 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998179267471592e-05, + "loss": 1.2885, + "step": 1493 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998164374471673e-05, + "loss": 1.2222, + "step": 1494 + }, + { + "epoch": 0.04, + "learning_rate": 1.999814942081524e-05, + "loss": 1.0195, + "step": 1495 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998134406502384e-05, + "loss": 1.2304, + "step": 1496 + }, + { + "epoch": 0.04, + "learning_rate": 1.99981193315332e-05, + "loss": 1.548, + "step": 1497 + }, + { + "epoch": 0.04, + "learning_rate": 1.999810419590778e-05, + "loss": 1.0833, + "step": 1498 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998088999626212e-05, + "loss": 1.1097, + "step": 1499 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998073742688592e-05, + "loss": 0.9328, + "step": 1500 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998058425095007e-05, + "loss": 1.3589, + "step": 1501 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998043046845555e-05, + "loss": 1.3803, + "step": 1502 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998027607940328e-05, + "loss": 1.301, + "step": 1503 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998012108379417e-05, + "loss": 1.2146, + "step": 1504 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997996548162918e-05, + "loss": 1.4443, + "step": 1505 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997980927290928e-05, + "loss": 0.8283, + "step": 1506 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997965245763538e-05, + "loss": 1.3056, + "step": 1507 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997949503580844e-05, + "loss": 1.3017, + "step": 1508 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997933700742945e-05, + "loss": 1.2811, + "step": 1509 + }, + { + "epoch": 0.04, + "learning_rate": 1.999791783724993e-05, + "loss": 1.2512, + "step": 1510 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997901913101903e-05, + "loss": 1.2545, + "step": 1511 + }, + { + "epoch": 0.04, + "learning_rate": 1.999788592829895e-05, + "loss": 1.2617, + "step": 1512 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997869882841183e-05, + "loss": 1.3856, + "step": 1513 + }, + { + "epoch": 0.04, + "learning_rate": 1.999785377672869e-05, + "loss": 1.191, + "step": 1514 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997837609961563e-05, + "loss": 0.9877, + "step": 1515 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997821382539914e-05, + "loss": 1.0913, + "step": 1516 + }, + { + "epoch": 0.04, + "learning_rate": 1.999780509446383e-05, + "loss": 1.1406, + "step": 1517 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997788745733415e-05, + "loss": 1.1424, + "step": 1518 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997772336348767e-05, + "loss": 1.2358, + "step": 1519 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997755866309988e-05, + "loss": 1.2086, + "step": 1520 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997739335617178e-05, + "loss": 1.2091, + "step": 1521 + }, + { + "epoch": 0.04, + "learning_rate": 1.999772274427043e-05, + "loss": 1.0431, + "step": 1522 + }, + { + "epoch": 0.04, + "learning_rate": 1.999770609226985e-05, + "loss": 0.9224, + "step": 1523 + }, + { + "epoch": 0.04, + "learning_rate": 1.999768937961554e-05, + "loss": 0.9866, + "step": 1524 + }, + { + "epoch": 0.04, + "learning_rate": 1.99976726063076e-05, + "loss": 1.6001, + "step": 1525 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997655772346132e-05, + "loss": 1.3625, + "step": 1526 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997638877731238e-05, + "loss": 1.3264, + "step": 1527 + }, + { + "epoch": 0.04, + "learning_rate": 1.999762192246302e-05, + "loss": 1.1827, + "step": 1528 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997604906541583e-05, + "loss": 1.1229, + "step": 1529 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997587829967027e-05, + "loss": 1.2224, + "step": 1530 + }, + { + "epoch": 0.04, + "learning_rate": 1.999757069273946e-05, + "loss": 1.1226, + "step": 1531 + }, + { + "epoch": 0.04, + "learning_rate": 1.999755349485898e-05, + "loss": 1.1768, + "step": 1532 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997536236325694e-05, + "loss": 1.2553, + "step": 1533 + }, + { + "epoch": 0.04, + "learning_rate": 1.999751891713971e-05, + "loss": 1.2499, + "step": 1534 + }, + { + "epoch": 0.04, + "learning_rate": 1.999750153730113e-05, + "loss": 0.9505, + "step": 1535 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997484096810054e-05, + "loss": 1.1665, + "step": 1536 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997466595666603e-05, + "loss": 1.0724, + "step": 1537 + }, + { + "epoch": 0.04, + "learning_rate": 1.999744903387087e-05, + "loss": 1.2432, + "step": 1538 + }, + { + "epoch": 0.04, + "learning_rate": 1.999743141142296e-05, + "loss": 1.362, + "step": 1539 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997413728322992e-05, + "loss": 0.9656, + "step": 1540 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997395984571063e-05, + "loss": 1.1188, + "step": 1541 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997378180167285e-05, + "loss": 1.1111, + "step": 1542 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997360315111764e-05, + "loss": 1.2277, + "step": 1543 + }, + { + "epoch": 0.04, + "learning_rate": 1.999734238940461e-05, + "loss": 1.4206, + "step": 1544 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997324403045932e-05, + "loss": 1.5407, + "step": 1545 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997306356035838e-05, + "loss": 1.2568, + "step": 1546 + }, + { + "epoch": 0.04, + "learning_rate": 1.999728824837444e-05, + "loss": 1.2583, + "step": 1547 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997270080061843e-05, + "loss": 1.4372, + "step": 1548 + }, + { + "epoch": 0.04, + "learning_rate": 1.999725185109816e-05, + "loss": 1.5356, + "step": 1549 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997233561483503e-05, + "loss": 0.9879, + "step": 1550 + }, + { + "epoch": 0.04, + "learning_rate": 1.999721521121798e-05, + "loss": 1.4874, + "step": 1551 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997196800301704e-05, + "loss": 1.1298, + "step": 1552 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997178328734787e-05, + "loss": 1.1505, + "step": 1553 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997159796517342e-05, + "loss": 1.0961, + "step": 1554 + }, + { + "epoch": 0.04, + "learning_rate": 1.999714120364948e-05, + "loss": 1.0551, + "step": 1555 + }, + { + "epoch": 0.04, + "learning_rate": 1.999712255013131e-05, + "loss": 1.385, + "step": 1556 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997103835962952e-05, + "loss": 1.0201, + "step": 1557 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997085061144514e-05, + "loss": 1.2267, + "step": 1558 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997066225676115e-05, + "loss": 1.4575, + "step": 1559 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997047329557867e-05, + "loss": 1.3146, + "step": 1560 + }, + { + "epoch": 0.04, + "learning_rate": 1.999702837278988e-05, + "loss": 1.1236, + "step": 1561 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997009355372276e-05, + "loss": 1.0099, + "step": 1562 + }, + { + "epoch": 0.04, + "learning_rate": 1.999699027730517e-05, + "loss": 1.1163, + "step": 1563 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996971138588674e-05, + "loss": 1.162, + "step": 1564 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996951939222903e-05, + "loss": 1.2046, + "step": 1565 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996932679207977e-05, + "loss": 1.3977, + "step": 1566 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996913358544016e-05, + "loss": 1.1414, + "step": 1567 + }, + { + "epoch": 0.04, + "learning_rate": 1.999689397723113e-05, + "loss": 1.3121, + "step": 1568 + }, + { + "epoch": 0.04, + "learning_rate": 1.999687453526944e-05, + "loss": 0.9957, + "step": 1569 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996855032659065e-05, + "loss": 1.2726, + "step": 1570 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996835469400117e-05, + "loss": 1.196, + "step": 1571 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996815845492722e-05, + "loss": 1.1427, + "step": 1572 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996796160937e-05, + "loss": 1.3286, + "step": 1573 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996776415733063e-05, + "loss": 1.0953, + "step": 1574 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996756609881036e-05, + "loss": 1.0898, + "step": 1575 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996736743381037e-05, + "loss": 1.4616, + "step": 1576 + }, + { + "epoch": 0.04, + "learning_rate": 1.999671681623319e-05, + "loss": 1.1986, + "step": 1577 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996696828437613e-05, + "loss": 1.3526, + "step": 1578 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996676779994425e-05, + "loss": 1.3286, + "step": 1579 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996656670903753e-05, + "loss": 1.1903, + "step": 1580 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996636501165718e-05, + "loss": 1.3202, + "step": 1581 + }, + { + "epoch": 0.04, + "learning_rate": 1.999661627078044e-05, + "loss": 1.3501, + "step": 1582 + }, + { + "epoch": 0.04, + "learning_rate": 1.999659597974804e-05, + "loss": 1.3214, + "step": 1583 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996575628068645e-05, + "loss": 1.2824, + "step": 1584 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996555215742376e-05, + "loss": 0.9093, + "step": 1585 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996534742769355e-05, + "loss": 1.4066, + "step": 1586 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996514209149713e-05, + "loss": 1.0518, + "step": 1587 + }, + { + "epoch": 0.04, + "learning_rate": 1.999649361488357e-05, + "loss": 1.2264, + "step": 1588 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996472959971053e-05, + "loss": 1.316, + "step": 1589 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996452244412285e-05, + "loss": 1.1441, + "step": 1590 + }, + { + "epoch": 0.04, + "learning_rate": 1.999643146820739e-05, + "loss": 1.3814, + "step": 1591 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996410631356496e-05, + "loss": 1.0183, + "step": 1592 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996389733859737e-05, + "loss": 1.2566, + "step": 1593 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996368775717228e-05, + "loss": 1.3472, + "step": 1594 + }, + { + "epoch": 0.04, + "learning_rate": 1.99963477569291e-05, + "loss": 1.1127, + "step": 1595 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996326677495482e-05, + "loss": 1.1608, + "step": 1596 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996305537416503e-05, + "loss": 0.9711, + "step": 1597 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996284336692286e-05, + "loss": 1.5858, + "step": 1598 + }, + { + "epoch": 0.04, + "learning_rate": 1.999626307532297e-05, + "loss": 1.2504, + "step": 1599 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996241753308673e-05, + "loss": 1.078, + "step": 1600 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996220370649527e-05, + "loss": 1.2236, + "step": 1601 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996198927345665e-05, + "loss": 1.2336, + "step": 1602 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996177423397215e-05, + "loss": 1.1514, + "step": 1603 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996155858804307e-05, + "loss": 1.1294, + "step": 1604 + }, + { + "epoch": 0.04, + "learning_rate": 1.999613423356707e-05, + "loss": 1.1286, + "step": 1605 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996112547685645e-05, + "loss": 1.1069, + "step": 1606 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996090801160153e-05, + "loss": 1.0826, + "step": 1607 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996068993990727e-05, + "loss": 1.17, + "step": 1608 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996047126177506e-05, + "loss": 1.3794, + "step": 1609 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996025197720615e-05, + "loss": 1.164, + "step": 1610 + }, + { + "epoch": 0.04, + "learning_rate": 1.999600320862019e-05, + "loss": 1.2854, + "step": 1611 + }, + { + "epoch": 0.04, + "learning_rate": 1.999598115887637e-05, + "loss": 1.2476, + "step": 1612 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995959048489276e-05, + "loss": 0.997, + "step": 1613 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995936877459053e-05, + "loss": 1.2626, + "step": 1614 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995914645785835e-05, + "loss": 1.2998, + "step": 1615 + }, + { + "epoch": 0.04, + "learning_rate": 1.999589235346975e-05, + "loss": 1.2178, + "step": 1616 + }, + { + "epoch": 0.04, + "learning_rate": 1.999587000051094e-05, + "loss": 1.4003, + "step": 1617 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995847586909537e-05, + "loss": 1.1863, + "step": 1618 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995825112665676e-05, + "loss": 0.9618, + "step": 1619 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995802577779498e-05, + "loss": 1.0537, + "step": 1620 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995779982251134e-05, + "loss": 1.1113, + "step": 1621 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995757326080727e-05, + "loss": 1.1626, + "step": 1622 + }, + { + "epoch": 0.04, + "learning_rate": 1.999573460926841e-05, + "loss": 1.286, + "step": 1623 + }, + { + "epoch": 0.04, + "learning_rate": 1.999571183181432e-05, + "loss": 1.1563, + "step": 1624 + }, + { + "epoch": 0.04, + "learning_rate": 1.99956889937186e-05, + "loss": 1.0709, + "step": 1625 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995666094981385e-05, + "loss": 1.2502, + "step": 1626 + }, + { + "epoch": 0.04, + "learning_rate": 1.999564313560282e-05, + "loss": 1.1349, + "step": 1627 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995620115583033e-05, + "loss": 1.1328, + "step": 1628 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995597034922173e-05, + "loss": 1.2225, + "step": 1629 + }, + { + "epoch": 0.04, + "learning_rate": 1.999557389362037e-05, + "loss": 1.2245, + "step": 1630 + }, + { + "epoch": 0.04, + "learning_rate": 1.999555069167778e-05, + "loss": 1.0787, + "step": 1631 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995527429094534e-05, + "loss": 1.0548, + "step": 1632 + }, + { + "epoch": 0.04, + "learning_rate": 1.999550410587077e-05, + "loss": 1.2431, + "step": 1633 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995480722006636e-05, + "loss": 1.1841, + "step": 1634 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995457277502272e-05, + "loss": 1.0643, + "step": 1635 + }, + { + "epoch": 0.04, + "learning_rate": 1.999543377235782e-05, + "loss": 1.2386, + "step": 1636 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995410206573422e-05, + "loss": 1.2395, + "step": 1637 + }, + { + "epoch": 0.04, + "learning_rate": 1.999538658014922e-05, + "loss": 1.0132, + "step": 1638 + }, + { + "epoch": 0.04, + "learning_rate": 1.999536289308536e-05, + "loss": 1.0032, + "step": 1639 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995339145381982e-05, + "loss": 1.2502, + "step": 1640 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995315337039237e-05, + "loss": 1.0798, + "step": 1641 + }, + { + "epoch": 0.04, + "learning_rate": 1.999529146805726e-05, + "loss": 1.1327, + "step": 1642 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995267538436206e-05, + "loss": 0.9814, + "step": 1643 + }, + { + "epoch": 0.04, + "learning_rate": 1.999524354817621e-05, + "loss": 1.1889, + "step": 1644 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995219497277424e-05, + "loss": 1.0412, + "step": 1645 + }, + { + "epoch": 0.04, + "learning_rate": 1.999519538573999e-05, + "loss": 0.9677, + "step": 1646 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995171213564062e-05, + "loss": 1.5616, + "step": 1647 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995146980749777e-05, + "loss": 1.3324, + "step": 1648 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995122687297287e-05, + "loss": 1.0287, + "step": 1649 + }, + { + "epoch": 0.04, + "learning_rate": 1.999509833320674e-05, + "loss": 1.2907, + "step": 1650 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995073918478283e-05, + "loss": 1.1806, + "step": 1651 + }, + { + "epoch": 0.04, + "learning_rate": 1.999504944311206e-05, + "loss": 1.1641, + "step": 1652 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995024907108227e-05, + "loss": 1.0836, + "step": 1653 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995000310466927e-05, + "loss": 1.2907, + "step": 1654 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994975653188315e-05, + "loss": 1.3462, + "step": 1655 + }, + { + "epoch": 0.04, + "learning_rate": 1.999495093527253e-05, + "loss": 1.1054, + "step": 1656 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994926156719732e-05, + "loss": 1.3337, + "step": 1657 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994901317530067e-05, + "loss": 0.9677, + "step": 1658 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994876417703688e-05, + "loss": 1.0734, + "step": 1659 + }, + { + "epoch": 0.04, + "learning_rate": 1.999485145724074e-05, + "loss": 1.4204, + "step": 1660 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994826436141386e-05, + "loss": 1.4053, + "step": 1661 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994801354405768e-05, + "loss": 1.3255, + "step": 1662 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994776212034042e-05, + "loss": 1.0396, + "step": 1663 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994751009026355e-05, + "loss": 1.3863, + "step": 1664 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994725745382873e-05, + "loss": 1.4062, + "step": 1665 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994700421103734e-05, + "loss": 1.2917, + "step": 1666 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994675036189098e-05, + "loss": 1.2273, + "step": 1667 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994649590639124e-05, + "loss": 1.4328, + "step": 1668 + }, + { + "epoch": 0.04, + "learning_rate": 1.999462408445396e-05, + "loss": 1.1903, + "step": 1669 + }, + { + "epoch": 0.04, + "learning_rate": 1.999459851763376e-05, + "loss": 1.4098, + "step": 1670 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994572890178685e-05, + "loss": 1.1899, + "step": 1671 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994547202088886e-05, + "loss": 1.201, + "step": 1672 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994521453364523e-05, + "loss": 1.256, + "step": 1673 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994495644005746e-05, + "loss": 1.0309, + "step": 1674 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994469774012715e-05, + "loss": 1.2348, + "step": 1675 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994443843385583e-05, + "loss": 1.2027, + "step": 1676 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994417852124516e-05, + "loss": 1.0225, + "step": 1677 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994391800229666e-05, + "loss": 1.1187, + "step": 1678 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994365687701195e-05, + "loss": 1.1893, + "step": 1679 + }, + { + "epoch": 0.04, + "learning_rate": 1.999433951453925e-05, + "loss": 1.2468, + "step": 1680 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994313280744003e-05, + "loss": 1.17, + "step": 1681 + }, + { + "epoch": 0.04, + "learning_rate": 1.999428698631561e-05, + "loss": 0.9506, + "step": 1682 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994260631254224e-05, + "loss": 1.3184, + "step": 1683 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994234215560014e-05, + "loss": 0.9853, + "step": 1684 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994207739233132e-05, + "loss": 1.2646, + "step": 1685 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994181202273745e-05, + "loss": 1.1842, + "step": 1686 + }, + { + "epoch": 0.04, + "learning_rate": 1.999415460468201e-05, + "loss": 1.2029, + "step": 1687 + }, + { + "epoch": 0.04, + "learning_rate": 1.999412794645809e-05, + "loss": 1.2292, + "step": 1688 + }, + { + "epoch": 0.04, + "learning_rate": 1.999410122760215e-05, + "loss": 1.1331, + "step": 1689 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994074448114342e-05, + "loss": 1.0557, + "step": 1690 + }, + { + "epoch": 0.04, + "learning_rate": 1.999404760799484e-05, + "loss": 1.2339, + "step": 1691 + }, + { + "epoch": 0.04, + "learning_rate": 1.99940207072438e-05, + "loss": 1.1052, + "step": 1692 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993993745861387e-05, + "loss": 1.1025, + "step": 1693 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993966723847766e-05, + "loss": 1.1689, + "step": 1694 + }, + { + "epoch": 0.04, + "learning_rate": 1.99939396412031e-05, + "loss": 0.992, + "step": 1695 + }, + { + "epoch": 0.04, + "learning_rate": 1.999391249792755e-05, + "loss": 1.0353, + "step": 1696 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993885294021285e-05, + "loss": 1.0605, + "step": 1697 + }, + { + "epoch": 0.04, + "learning_rate": 1.999385802948447e-05, + "loss": 1.2503, + "step": 1698 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993830704317268e-05, + "loss": 1.0279, + "step": 1699 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993803318519845e-05, + "loss": 1.4314, + "step": 1700 + }, + { + "epoch": 0.04, + "learning_rate": 1.999377587209237e-05, + "loss": 1.2053, + "step": 1701 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993748365035004e-05, + "loss": 1.0882, + "step": 1702 + }, + { + "epoch": 0.04, + "learning_rate": 1.999372079734792e-05, + "loss": 1.0142, + "step": 1703 + }, + { + "epoch": 0.04, + "learning_rate": 1.999369316903128e-05, + "loss": 1.2953, + "step": 1704 + }, + { + "epoch": 0.04, + "learning_rate": 1.999366548008526e-05, + "loss": 1.1222, + "step": 1705 + }, + { + "epoch": 0.04, + "learning_rate": 1.999363773051002e-05, + "loss": 1.2126, + "step": 1706 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993609920305725e-05, + "loss": 1.2799, + "step": 1707 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993582049472554e-05, + "loss": 1.148, + "step": 1708 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993554118010672e-05, + "loss": 1.0326, + "step": 1709 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993526125920245e-05, + "loss": 1.0591, + "step": 1710 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993498073201447e-05, + "loss": 1.1312, + "step": 1711 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993469959854445e-05, + "loss": 1.1502, + "step": 1712 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993441785879414e-05, + "loss": 1.2209, + "step": 1713 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993413551276523e-05, + "loss": 1.2113, + "step": 1714 + }, + { + "epoch": 0.04, + "learning_rate": 1.999338525604594e-05, + "loss": 1.2889, + "step": 1715 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993356900187843e-05, + "loss": 1.2754, + "step": 1716 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993328483702393e-05, + "loss": 1.039, + "step": 1717 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993300006589775e-05, + "loss": 1.3981, + "step": 1718 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993271468850156e-05, + "loss": 1.3661, + "step": 1719 + }, + { + "epoch": 0.04, + "learning_rate": 1.999324287048371e-05, + "loss": 1.1223, + "step": 1720 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993214211490607e-05, + "loss": 0.9549, + "step": 1721 + }, + { + "epoch": 0.04, + "learning_rate": 1.999318549187103e-05, + "loss": 1.324, + "step": 1722 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993156711625142e-05, + "loss": 1.2834, + "step": 1723 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993127870753123e-05, + "loss": 1.2842, + "step": 1724 + }, + { + "epoch": 0.04, + "learning_rate": 1.999309896925515e-05, + "loss": 1.1268, + "step": 1725 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993070007131395e-05, + "loss": 1.2287, + "step": 1726 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993040984382036e-05, + "loss": 1.0734, + "step": 1727 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993011901007245e-05, + "loss": 1.2117, + "step": 1728 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992982757007203e-05, + "loss": 1.1609, + "step": 1729 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992953552382085e-05, + "loss": 1.0775, + "step": 1730 + }, + { + "epoch": 0.04, + "learning_rate": 1.999292428713207e-05, + "loss": 1.177, + "step": 1731 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992894961257332e-05, + "loss": 1.1433, + "step": 1732 + }, + { + "epoch": 0.04, + "learning_rate": 1.999286557475805e-05, + "loss": 0.9191, + "step": 1733 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992836127634402e-05, + "loss": 1.1799, + "step": 1734 + }, + { + "epoch": 0.04, + "learning_rate": 1.999280661988657e-05, + "loss": 1.0223, + "step": 1735 + }, + { + "epoch": 0.04, + "learning_rate": 1.999277705151473e-05, + "loss": 1.0298, + "step": 1736 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992747422519063e-05, + "loss": 0.9261, + "step": 1737 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992717732899746e-05, + "loss": 1.3458, + "step": 1738 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992687982656963e-05, + "loss": 1.1749, + "step": 1739 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992658171790893e-05, + "loss": 1.029, + "step": 1740 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992628300301714e-05, + "loss": 1.3708, + "step": 1741 + }, + { + "epoch": 0.04, + "learning_rate": 1.999259836818961e-05, + "loss": 1.3416, + "step": 1742 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992568375454762e-05, + "loss": 1.423, + "step": 1743 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992538322097352e-05, + "loss": 1.344, + "step": 1744 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992508208117565e-05, + "loss": 1.3201, + "step": 1745 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992478033515578e-05, + "loss": 1.0255, + "step": 1746 + }, + { + "epoch": 0.04, + "learning_rate": 1.999244779829158e-05, + "loss": 1.1973, + "step": 1747 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992417502445746e-05, + "loss": 1.3356, + "step": 1748 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992387145978268e-05, + "loss": 1.2815, + "step": 1749 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992356728889332e-05, + "loss": 1.0958, + "step": 1750 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992326251179112e-05, + "loss": 1.0594, + "step": 1751 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992295712847802e-05, + "loss": 1.3108, + "step": 1752 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992265113895587e-05, + "loss": 1.1563, + "step": 1753 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992234454322643e-05, + "loss": 1.4469, + "step": 1754 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992203734129167e-05, + "loss": 1.398, + "step": 1755 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992172953315343e-05, + "loss": 1.0958, + "step": 1756 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992142111881353e-05, + "loss": 1.3333, + "step": 1757 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992111209827386e-05, + "loss": 1.3599, + "step": 1758 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992080247153633e-05, + "loss": 1.002, + "step": 1759 + }, + { + "epoch": 0.04, + "learning_rate": 1.999204922386028e-05, + "loss": 1.4716, + "step": 1760 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992018139947512e-05, + "loss": 1.1596, + "step": 1761 + }, + { + "epoch": 0.04, + "learning_rate": 1.999198699541552e-05, + "loss": 1.2858, + "step": 1762 + }, + { + "epoch": 0.04, + "learning_rate": 1.99919557902645e-05, + "loss": 1.1266, + "step": 1763 + }, + { + "epoch": 0.04, + "learning_rate": 1.999192452449463e-05, + "loss": 1.2816, + "step": 1764 + }, + { + "epoch": 0.04, + "learning_rate": 1.99918931981061e-05, + "loss": 1.4381, + "step": 1765 + }, + { + "epoch": 0.04, + "learning_rate": 1.999186181109911e-05, + "loss": 1.1332, + "step": 1766 + }, + { + "epoch": 0.04, + "learning_rate": 1.999183036347384e-05, + "loss": 1.0418, + "step": 1767 + }, + { + "epoch": 0.04, + "learning_rate": 1.999179885523049e-05, + "loss": 1.2007, + "step": 1768 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991767286369243e-05, + "loss": 1.129, + "step": 1769 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991735656890293e-05, + "loss": 1.3762, + "step": 1770 + }, + { + "epoch": 0.04, + "learning_rate": 1.999170396679384e-05, + "loss": 1.3065, + "step": 1771 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991672216080064e-05, + "loss": 1.3875, + "step": 1772 + }, + { + "epoch": 0.04, + "learning_rate": 1.999164040474916e-05, + "loss": 1.0404, + "step": 1773 + }, + { + "epoch": 0.04, + "learning_rate": 1.999160853280133e-05, + "loss": 1.2175, + "step": 1774 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991576600236757e-05, + "loss": 1.1042, + "step": 1775 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991544607055642e-05, + "loss": 1.1721, + "step": 1776 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991512553258173e-05, + "loss": 1.1531, + "step": 1777 + }, + { + "epoch": 0.04, + "learning_rate": 1.999148043884455e-05, + "loss": 1.2018, + "step": 1778 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991448263814965e-05, + "loss": 1.1922, + "step": 1779 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991416028169612e-05, + "loss": 1.2639, + "step": 1780 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991383731908688e-05, + "loss": 1.0487, + "step": 1781 + }, + { + "epoch": 0.04, + "learning_rate": 1.999135137503239e-05, + "loss": 1.0809, + "step": 1782 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991318957540913e-05, + "loss": 1.2193, + "step": 1783 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991286479434456e-05, + "loss": 1.2676, + "step": 1784 + }, + { + "epoch": 0.04, + "learning_rate": 1.999125394071321e-05, + "loss": 1.2196, + "step": 1785 + }, + { + "epoch": 0.04, + "learning_rate": 1.999122134137738e-05, + "loss": 1.1113, + "step": 1786 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991188681427157e-05, + "loss": 1.1588, + "step": 1787 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991155960862743e-05, + "loss": 1.0193, + "step": 1788 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991123179684344e-05, + "loss": 1.2576, + "step": 1789 + }, + { + "epoch": 0.04, + "learning_rate": 1.999109033789214e-05, + "loss": 1.1174, + "step": 1790 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991057435486345e-05, + "loss": 1.1801, + "step": 1791 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991024472467156e-05, + "loss": 1.3964, + "step": 1792 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990991448834767e-05, + "loss": 1.5171, + "step": 1793 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990958364589388e-05, + "loss": 1.037, + "step": 1794 + }, + { + "epoch": 0.04, + "learning_rate": 1.999092521973121e-05, + "loss": 1.2175, + "step": 1795 + }, + { + "epoch": 0.04, + "learning_rate": 1.999089201426044e-05, + "loss": 1.1994, + "step": 1796 + }, + { + "epoch": 0.04, + "learning_rate": 1.999085874817728e-05, + "loss": 1.0735, + "step": 1797 + }, + { + "epoch": 0.04, + "learning_rate": 1.999082542148193e-05, + "loss": 1.1067, + "step": 1798 + }, + { + "epoch": 0.04, + "learning_rate": 1.999079203417459e-05, + "loss": 1.2166, + "step": 1799 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990758586255467e-05, + "loss": 1.3089, + "step": 1800 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990725077724763e-05, + "loss": 1.203, + "step": 1801 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990691508582678e-05, + "loss": 1.1431, + "step": 1802 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990657878829416e-05, + "loss": 1.335, + "step": 1803 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990624188465183e-05, + "loss": 1.284, + "step": 1804 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990590437490183e-05, + "loss": 1.2447, + "step": 1805 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990556625904623e-05, + "loss": 1.0872, + "step": 1806 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990522753708705e-05, + "loss": 1.2842, + "step": 1807 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990488820902633e-05, + "loss": 1.1221, + "step": 1808 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990454827486617e-05, + "loss": 1.0584, + "step": 1809 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990420773460864e-05, + "loss": 1.0515, + "step": 1810 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990386658825573e-05, + "loss": 1.3803, + "step": 1811 + }, + { + "epoch": 0.04, + "learning_rate": 1.999035248358096e-05, + "loss": 1.1667, + "step": 1812 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990318247727227e-05, + "loss": 1.1363, + "step": 1813 + }, + { + "epoch": 0.04, + "learning_rate": 1.999028395126458e-05, + "loss": 1.2292, + "step": 1814 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990249594193238e-05, + "loss": 1.2429, + "step": 1815 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990215176513394e-05, + "loss": 1.4517, + "step": 1816 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990180698225266e-05, + "loss": 1.3409, + "step": 1817 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990146159329065e-05, + "loss": 1.1584, + "step": 1818 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990111559824994e-05, + "loss": 1.4174, + "step": 1819 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990076899713268e-05, + "loss": 1.253, + "step": 1820 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990042178994093e-05, + "loss": 1.4334, + "step": 1821 + }, + { + "epoch": 0.04, + "learning_rate": 1.999000739766768e-05, + "loss": 0.8999, + "step": 1822 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989972555734244e-05, + "loss": 0.7869, + "step": 1823 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989937653193995e-05, + "loss": 1.2366, + "step": 1824 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989902690047146e-05, + "loss": 1.2647, + "step": 1825 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989867666293904e-05, + "loss": 1.11, + "step": 1826 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989832581934488e-05, + "loss": 1.2204, + "step": 1827 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989797436969103e-05, + "loss": 1.269, + "step": 1828 + }, + { + "epoch": 0.04, + "learning_rate": 1.998976223139797e-05, + "loss": 1.2394, + "step": 1829 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989726965221298e-05, + "loss": 1.0669, + "step": 1830 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989691638439305e-05, + "loss": 1.4404, + "step": 1831 + }, + { + "epoch": 0.04, + "learning_rate": 1.99896562510522e-05, + "loss": 1.3043, + "step": 1832 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989620803060198e-05, + "loss": 1.358, + "step": 1833 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989585294463518e-05, + "loss": 1.1279, + "step": 1834 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989549725262372e-05, + "loss": 1.1523, + "step": 1835 + }, + { + "epoch": 0.04, + "learning_rate": 1.998951409545698e-05, + "loss": 1.2829, + "step": 1836 + }, + { + "epoch": 0.04, + "learning_rate": 1.998947840504756e-05, + "loss": 1.0786, + "step": 1837 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989442654034315e-05, + "loss": 1.2523, + "step": 1838 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989406842417477e-05, + "loss": 1.4258, + "step": 1839 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989370970197257e-05, + "loss": 1.2632, + "step": 1840 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989335037373873e-05, + "loss": 1.3481, + "step": 1841 + }, + { + "epoch": 0.04, + "learning_rate": 1.998929904394754e-05, + "loss": 1.1567, + "step": 1842 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989262989918482e-05, + "loss": 1.3413, + "step": 1843 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989226875286913e-05, + "loss": 1.2313, + "step": 1844 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989190700053057e-05, + "loss": 1.2806, + "step": 1845 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989154464217128e-05, + "loss": 1.2001, + "step": 1846 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989118167779352e-05, + "loss": 1.1093, + "step": 1847 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989081810739943e-05, + "loss": 1.1225, + "step": 1848 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989045393099125e-05, + "loss": 1.424, + "step": 1849 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989008914857115e-05, + "loss": 1.3216, + "step": 1850 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988972376014142e-05, + "loss": 1.3144, + "step": 1851 + }, + { + "epoch": 0.04, + "learning_rate": 1.998893577657042e-05, + "loss": 1.2803, + "step": 1852 + }, + { + "epoch": 0.04, + "learning_rate": 1.998889911652618e-05, + "loss": 1.0994, + "step": 1853 + }, + { + "epoch": 0.04, + "learning_rate": 1.998886239588163e-05, + "loss": 0.9272, + "step": 1854 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988825614637006e-05, + "loss": 1.2288, + "step": 1855 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988788772792523e-05, + "loss": 1.3362, + "step": 1856 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988751870348414e-05, + "loss": 1.467, + "step": 1857 + }, + { + "epoch": 0.04, + "learning_rate": 1.998871490730489e-05, + "loss": 1.3418, + "step": 1858 + }, + { + "epoch": 0.04, + "learning_rate": 1.998867788366219e-05, + "loss": 1.293, + "step": 1859 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988640799420524e-05, + "loss": 1.2418, + "step": 1860 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988603654580128e-05, + "loss": 0.8629, + "step": 1861 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988566449141223e-05, + "loss": 1.0132, + "step": 1862 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988529183104033e-05, + "loss": 1.3406, + "step": 1863 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988491856468785e-05, + "loss": 1.5238, + "step": 1864 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988454469235708e-05, + "loss": 1.1498, + "step": 1865 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988417021405027e-05, + "loss": 1.3507, + "step": 1866 + }, + { + "epoch": 0.04, + "learning_rate": 1.998837951297697e-05, + "loss": 1.2078, + "step": 1867 + }, + { + "epoch": 0.04, + "learning_rate": 1.998834194395176e-05, + "loss": 1.0002, + "step": 1868 + }, + { + "epoch": 0.04, + "learning_rate": 1.998830431432963e-05, + "loss": 1.0904, + "step": 1869 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988266624110813e-05, + "loss": 1.0453, + "step": 1870 + }, + { + "epoch": 0.04, + "learning_rate": 1.998822887329553e-05, + "loss": 1.0943, + "step": 1871 + }, + { + "epoch": 0.05, + "learning_rate": 1.9988191061884005e-05, + "loss": 1.2296, + "step": 1872 + }, + { + "epoch": 0.05, + "learning_rate": 1.998815318987648e-05, + "loss": 1.1918, + "step": 1873 + }, + { + "epoch": 0.05, + "learning_rate": 1.9988115257273176e-05, + "loss": 1.1789, + "step": 1874 + }, + { + "epoch": 0.05, + "learning_rate": 1.998807726407433e-05, + "loss": 1.6163, + "step": 1875 + }, + { + "epoch": 0.05, + "learning_rate": 1.9988039210280167e-05, + "loss": 1.2692, + "step": 1876 + }, + { + "epoch": 0.05, + "learning_rate": 1.9988001095890918e-05, + "loss": 0.9071, + "step": 1877 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987962920906816e-05, + "loss": 1.2885, + "step": 1878 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987924685328094e-05, + "loss": 1.2003, + "step": 1879 + }, + { + "epoch": 0.05, + "learning_rate": 1.998788638915498e-05, + "loss": 1.1292, + "step": 1880 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987848032387707e-05, + "loss": 1.3486, + "step": 1881 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987809615026513e-05, + "loss": 1.3131, + "step": 1882 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987771137071628e-05, + "loss": 1.2304, + "step": 1883 + }, + { + "epoch": 0.05, + "learning_rate": 1.998773259852328e-05, + "loss": 1.2802, + "step": 1884 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987693999381715e-05, + "loss": 1.1903, + "step": 1885 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987655339647153e-05, + "loss": 1.193, + "step": 1886 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987616619319837e-05, + "loss": 1.1443, + "step": 1887 + }, + { + "epoch": 0.05, + "learning_rate": 1.99875778384e-05, + "loss": 1.1259, + "step": 1888 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987538996887877e-05, + "loss": 1.4745, + "step": 1889 + }, + { + "epoch": 0.05, + "learning_rate": 1.998750009478371e-05, + "loss": 1.0466, + "step": 1890 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987461132087722e-05, + "loss": 1.2784, + "step": 1891 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987422108800157e-05, + "loss": 1.4956, + "step": 1892 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987383024921256e-05, + "loss": 1.2368, + "step": 1893 + }, + { + "epoch": 0.05, + "learning_rate": 1.998734388045125e-05, + "loss": 1.1705, + "step": 1894 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987304675390376e-05, + "loss": 1.3026, + "step": 1895 + }, + { + "epoch": 0.05, + "learning_rate": 1.998726540973887e-05, + "loss": 1.0115, + "step": 1896 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987226083496983e-05, + "loss": 1.1741, + "step": 1897 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987186696664937e-05, + "loss": 1.2162, + "step": 1898 + }, + { + "epoch": 0.05, + "learning_rate": 1.998714724924298e-05, + "loss": 0.8906, + "step": 1899 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987107741231352e-05, + "loss": 1.2689, + "step": 1900 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987068172630288e-05, + "loss": 1.2885, + "step": 1901 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987028543440035e-05, + "loss": 1.26, + "step": 1902 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986988853660825e-05, + "loss": 1.2949, + "step": 1903 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986949103292904e-05, + "loss": 1.1598, + "step": 1904 + }, + { + "epoch": 0.05, + "learning_rate": 1.998690929233651e-05, + "loss": 1.209, + "step": 1905 + }, + { + "epoch": 0.05, + "learning_rate": 1.998686942079189e-05, + "loss": 1.25, + "step": 1906 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986829488659276e-05, + "loss": 1.2712, + "step": 1907 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986789495938922e-05, + "loss": 1.0456, + "step": 1908 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986749442631063e-05, + "loss": 1.4031, + "step": 1909 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986709328735944e-05, + "loss": 1.1542, + "step": 1910 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986669154253812e-05, + "loss": 1.149, + "step": 1911 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986628919184904e-05, + "loss": 1.1265, + "step": 1912 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986588623529467e-05, + "loss": 1.2883, + "step": 1913 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986548267287745e-05, + "loss": 1.151, + "step": 1914 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986507850459987e-05, + "loss": 1.3017, + "step": 1915 + }, + { + "epoch": 0.05, + "learning_rate": 1.998646737304643e-05, + "loss": 1.0265, + "step": 1916 + }, + { + "epoch": 0.05, + "learning_rate": 1.998642683504733e-05, + "loss": 1.1583, + "step": 1917 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986386236462926e-05, + "loss": 1.208, + "step": 1918 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986345577293464e-05, + "loss": 1.0269, + "step": 1919 + }, + { + "epoch": 0.05, + "learning_rate": 1.998630485753919e-05, + "loss": 1.0315, + "step": 1920 + }, + { + "epoch": 0.05, + "learning_rate": 1.998626407720036e-05, + "loss": 1.0303, + "step": 1921 + }, + { + "epoch": 0.05, + "learning_rate": 1.998622323627721e-05, + "loss": 1.2048, + "step": 1922 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986182334769992e-05, + "loss": 1.29, + "step": 1923 + }, + { + "epoch": 0.05, + "learning_rate": 1.998614137267896e-05, + "loss": 1.0369, + "step": 1924 + }, + { + "epoch": 0.05, + "learning_rate": 1.998610035000435e-05, + "loss": 1.2481, + "step": 1925 + }, + { + "epoch": 0.05, + "learning_rate": 1.998605926674642e-05, + "loss": 0.9256, + "step": 1926 + }, + { + "epoch": 0.05, + "learning_rate": 1.998601812290542e-05, + "loss": 1.0636, + "step": 1927 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985976918481595e-05, + "loss": 1.3652, + "step": 1928 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985935653475196e-05, + "loss": 1.2209, + "step": 1929 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985894327886476e-05, + "loss": 1.0949, + "step": 1930 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985852941715686e-05, + "loss": 1.0826, + "step": 1931 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985811494963074e-05, + "loss": 1.3164, + "step": 1932 + }, + { + "epoch": 0.05, + "learning_rate": 1.998576998762889e-05, + "loss": 1.2723, + "step": 1933 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985728419713388e-05, + "loss": 1.2027, + "step": 1934 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985686791216823e-05, + "loss": 1.0907, + "step": 1935 + }, + { + "epoch": 0.05, + "learning_rate": 1.998564510213944e-05, + "loss": 1.3467, + "step": 1936 + }, + { + "epoch": 0.05, + "learning_rate": 1.99856033524815e-05, + "loss": 1.3226, + "step": 1937 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985561542243256e-05, + "loss": 1.3482, + "step": 1938 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985519671424957e-05, + "loss": 1.0189, + "step": 1939 + }, + { + "epoch": 0.05, + "learning_rate": 1.998547774002686e-05, + "loss": 1.2258, + "step": 1940 + }, + { + "epoch": 0.05, + "learning_rate": 1.998543574804922e-05, + "loss": 1.067, + "step": 1941 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985393695492283e-05, + "loss": 1.1626, + "step": 1942 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985351582356316e-05, + "loss": 1.0991, + "step": 1943 + }, + { + "epoch": 0.05, + "learning_rate": 1.998530940864157e-05, + "loss": 0.9106, + "step": 1944 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985267174348298e-05, + "loss": 1.1418, + "step": 1945 + }, + { + "epoch": 0.05, + "learning_rate": 1.998522487947676e-05, + "loss": 1.358, + "step": 1946 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985182524027213e-05, + "loss": 1.2785, + "step": 1947 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985140107999908e-05, + "loss": 1.3319, + "step": 1948 + }, + { + "epoch": 0.05, + "learning_rate": 1.998509763139511e-05, + "loss": 1.2757, + "step": 1949 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985055094213072e-05, + "loss": 1.0719, + "step": 1950 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985012496454056e-05, + "loss": 1.1937, + "step": 1951 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984969838118316e-05, + "loss": 1.2681, + "step": 1952 + }, + { + "epoch": 0.05, + "learning_rate": 1.998492711920611e-05, + "loss": 1.3676, + "step": 1953 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984884339717704e-05, + "loss": 1.4491, + "step": 1954 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984841499653352e-05, + "loss": 1.1542, + "step": 1955 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984798599013315e-05, + "loss": 1.546, + "step": 1956 + }, + { + "epoch": 0.05, + "learning_rate": 1.998475563779785e-05, + "loss": 1.1895, + "step": 1957 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984712616007226e-05, + "loss": 1.1195, + "step": 1958 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984669533641698e-05, + "loss": 1.2093, + "step": 1959 + }, + { + "epoch": 0.05, + "learning_rate": 1.998462639070153e-05, + "loss": 1.4658, + "step": 1960 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984583187186975e-05, + "loss": 1.3322, + "step": 1961 + }, + { + "epoch": 0.05, + "learning_rate": 1.998453992309831e-05, + "loss": 1.1884, + "step": 1962 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984496598435788e-05, + "loss": 0.9419, + "step": 1963 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984453213199673e-05, + "loss": 1.2273, + "step": 1964 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984409767390226e-05, + "loss": 1.3857, + "step": 1965 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984366261007716e-05, + "loss": 1.2709, + "step": 1966 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984322694052408e-05, + "loss": 1.3552, + "step": 1967 + }, + { + "epoch": 0.05, + "learning_rate": 1.998427906652456e-05, + "loss": 1.1746, + "step": 1968 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984235378424434e-05, + "loss": 1.2729, + "step": 1969 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984191629752306e-05, + "loss": 1.2402, + "step": 1970 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984147820508435e-05, + "loss": 1.3581, + "step": 1971 + }, + { + "epoch": 0.05, + "learning_rate": 1.998410395069309e-05, + "loss": 1.0773, + "step": 1972 + }, + { + "epoch": 0.05, + "learning_rate": 1.998406002030653e-05, + "loss": 1.1036, + "step": 1973 + }, + { + "epoch": 0.05, + "learning_rate": 1.998401602934903e-05, + "loss": 1.1874, + "step": 1974 + }, + { + "epoch": 0.05, + "learning_rate": 1.998397197782085e-05, + "loss": 1.0052, + "step": 1975 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983927865722262e-05, + "loss": 1.1764, + "step": 1976 + }, + { + "epoch": 0.05, + "learning_rate": 1.998388369305353e-05, + "loss": 0.964, + "step": 1977 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983839459814925e-05, + "loss": 1.4723, + "step": 1978 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983795166006717e-05, + "loss": 1.1791, + "step": 1979 + }, + { + "epoch": 0.05, + "learning_rate": 1.998375081162917e-05, + "loss": 0.9764, + "step": 1980 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983706396682554e-05, + "loss": 1.0632, + "step": 1981 + }, + { + "epoch": 0.05, + "learning_rate": 1.998366192116714e-05, + "loss": 1.1694, + "step": 1982 + }, + { + "epoch": 0.05, + "learning_rate": 1.99836173850832e-05, + "loss": 1.3619, + "step": 1983 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983572788431e-05, + "loss": 0.883, + "step": 1984 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983528131210814e-05, + "loss": 1.4033, + "step": 1985 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983483413422907e-05, + "loss": 1.2219, + "step": 1986 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983438635067556e-05, + "loss": 1.2189, + "step": 1987 + }, + { + "epoch": 0.05, + "learning_rate": 1.998339379614503e-05, + "loss": 1.1268, + "step": 1988 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983348896655607e-05, + "loss": 1.3858, + "step": 1989 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983303936599553e-05, + "loss": 1.1675, + "step": 1990 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983258915977137e-05, + "loss": 1.117, + "step": 1991 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983213834788643e-05, + "loss": 1.203, + "step": 1992 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983168693034332e-05, + "loss": 1.2841, + "step": 1993 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983123490714492e-05, + "loss": 1.2332, + "step": 1994 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983078227829388e-05, + "loss": 1.1384, + "step": 1995 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983032904379296e-05, + "loss": 1.3798, + "step": 1996 + }, + { + "epoch": 0.05, + "learning_rate": 1.998298752036449e-05, + "loss": 1.4061, + "step": 1997 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982942075785247e-05, + "loss": 1.2928, + "step": 1998 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982896570641848e-05, + "loss": 1.1158, + "step": 1999 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982851004934557e-05, + "loss": 1.0281, + "step": 2000 + }, + { + "epoch": 0.05, + "learning_rate": 1.998280537866366e-05, + "loss": 1.1238, + "step": 2001 + }, + { + "epoch": 0.05, + "learning_rate": 1.998275969182943e-05, + "loss": 1.3676, + "step": 2002 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982713944432148e-05, + "loss": 1.2408, + "step": 2003 + }, + { + "epoch": 0.05, + "learning_rate": 1.998266813647208e-05, + "loss": 1.0548, + "step": 2004 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982622267949523e-05, + "loss": 1.1383, + "step": 2005 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982576338864738e-05, + "loss": 1.1389, + "step": 2006 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982530349218012e-05, + "loss": 1.1583, + "step": 2007 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982484299009624e-05, + "loss": 1.1855, + "step": 2008 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982438188239853e-05, + "loss": 1.1272, + "step": 2009 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982392016908975e-05, + "loss": 1.1502, + "step": 2010 + }, + { + "epoch": 0.05, + "learning_rate": 1.998234578501727e-05, + "loss": 0.8877, + "step": 2011 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982299492565028e-05, + "loss": 1.1843, + "step": 2012 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982253139552515e-05, + "loss": 1.2533, + "step": 2013 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982206725980026e-05, + "loss": 1.2554, + "step": 2014 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982160251847836e-05, + "loss": 1.3922, + "step": 2015 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982113717156225e-05, + "loss": 1.1313, + "step": 2016 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982067121905483e-05, + "loss": 1.121, + "step": 2017 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982020466095882e-05, + "loss": 1.195, + "step": 2018 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981973749727715e-05, + "loss": 1.252, + "step": 2019 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981926972801257e-05, + "loss": 1.276, + "step": 2020 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981880135316798e-05, + "loss": 0.939, + "step": 2021 + }, + { + "epoch": 0.05, + "learning_rate": 1.998183323727462e-05, + "loss": 1.3847, + "step": 2022 + }, + { + "epoch": 0.05, + "learning_rate": 1.998178627867501e-05, + "loss": 1.0828, + "step": 2023 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981739259518246e-05, + "loss": 1.4657, + "step": 2024 + }, + { + "epoch": 0.05, + "learning_rate": 1.998169217980462e-05, + "loss": 1.1301, + "step": 2025 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981645039534417e-05, + "loss": 0.9945, + "step": 2026 + }, + { + "epoch": 0.05, + "learning_rate": 1.998159783870792e-05, + "loss": 1.0607, + "step": 2027 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981550577325417e-05, + "loss": 1.2393, + "step": 2028 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981503255387193e-05, + "loss": 0.9362, + "step": 2029 + }, + { + "epoch": 0.05, + "learning_rate": 1.998145587289354e-05, + "loss": 1.0995, + "step": 2030 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981408429844738e-05, + "loss": 1.2246, + "step": 2031 + }, + { + "epoch": 0.05, + "learning_rate": 1.998136092624108e-05, + "loss": 1.0589, + "step": 2032 + }, + { + "epoch": 0.05, + "learning_rate": 1.998131336208285e-05, + "loss": 1.2622, + "step": 2033 + }, + { + "epoch": 0.05, + "learning_rate": 1.998126573737035e-05, + "loss": 1.0948, + "step": 2034 + }, + { + "epoch": 0.05, + "learning_rate": 1.998121805210385e-05, + "loss": 1.1073, + "step": 2035 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981170306283647e-05, + "loss": 1.3071, + "step": 2036 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981122499910034e-05, + "loss": 1.2556, + "step": 2037 + }, + { + "epoch": 0.05, + "learning_rate": 1.99810746329833e-05, + "loss": 1.4596, + "step": 2038 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981026705503732e-05, + "loss": 1.1562, + "step": 2039 + }, + { + "epoch": 0.05, + "learning_rate": 1.998097871747162e-05, + "loss": 1.3712, + "step": 2040 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980930668887264e-05, + "loss": 1.4042, + "step": 2041 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980882559750947e-05, + "loss": 1.1072, + "step": 2042 + }, + { + "epoch": 0.05, + "learning_rate": 1.998083439006296e-05, + "loss": 1.3348, + "step": 2043 + }, + { + "epoch": 0.05, + "learning_rate": 1.99807861598236e-05, + "loss": 1.2546, + "step": 2044 + }, + { + "epoch": 0.05, + "learning_rate": 1.998073786903316e-05, + "loss": 1.1685, + "step": 2045 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980689517691928e-05, + "loss": 1.3159, + "step": 2046 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980641105800204e-05, + "loss": 1.3517, + "step": 2047 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980592633358276e-05, + "loss": 1.1331, + "step": 2048 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980544100366438e-05, + "loss": 1.2817, + "step": 2049 + }, + { + "epoch": 0.05, + "learning_rate": 1.998049550682499e-05, + "loss": 1.2806, + "step": 2050 + }, + { + "epoch": 0.05, + "learning_rate": 1.998044685273422e-05, + "loss": 1.3432, + "step": 2051 + }, + { + "epoch": 0.05, + "learning_rate": 1.998039813809443e-05, + "loss": 1.2329, + "step": 2052 + }, + { + "epoch": 0.05, + "learning_rate": 1.998034936290591e-05, + "loss": 1.1697, + "step": 2053 + }, + { + "epoch": 0.05, + "learning_rate": 1.998030052716896e-05, + "loss": 1.4094, + "step": 2054 + }, + { + "epoch": 0.05, + "learning_rate": 1.998025163088387e-05, + "loss": 1.3047, + "step": 2055 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980202674050945e-05, + "loss": 1.2599, + "step": 2056 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980153656670475e-05, + "loss": 1.2817, + "step": 2057 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980104578742762e-05, + "loss": 1.312, + "step": 2058 + }, + { + "epoch": 0.05, + "learning_rate": 1.99800554402681e-05, + "loss": 1.0478, + "step": 2059 + }, + { + "epoch": 0.05, + "learning_rate": 1.998000624124679e-05, + "loss": 1.3487, + "step": 2060 + }, + { + "epoch": 0.05, + "learning_rate": 1.997995698167913e-05, + "loss": 1.2258, + "step": 2061 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979907661565418e-05, + "loss": 0.7188, + "step": 2062 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979858280905954e-05, + "loss": 1.1736, + "step": 2063 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979808839701037e-05, + "loss": 0.9546, + "step": 2064 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979759337950966e-05, + "loss": 1.2582, + "step": 2065 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979709775656048e-05, + "loss": 1.2673, + "step": 2066 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979660152816573e-05, + "loss": 1.249, + "step": 2067 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979610469432847e-05, + "loss": 1.2092, + "step": 2068 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979560725505175e-05, + "loss": 1.2486, + "step": 2069 + }, + { + "epoch": 0.05, + "learning_rate": 1.997951092103385e-05, + "loss": 1.4155, + "step": 2070 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979461056019185e-05, + "loss": 1.0769, + "step": 2071 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979411130461475e-05, + "loss": 1.192, + "step": 2072 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979361144361025e-05, + "loss": 0.8836, + "step": 2073 + }, + { + "epoch": 0.05, + "learning_rate": 1.997931109771814e-05, + "loss": 1.3314, + "step": 2074 + }, + { + "epoch": 0.05, + "learning_rate": 1.997926099053312e-05, + "loss": 1.1944, + "step": 2075 + }, + { + "epoch": 0.05, + "learning_rate": 1.997921082280627e-05, + "loss": 0.9478, + "step": 2076 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979160594537895e-05, + "loss": 1.1284, + "step": 2077 + }, + { + "epoch": 0.05, + "learning_rate": 1.99791103057283e-05, + "loss": 1.1049, + "step": 2078 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979059956377793e-05, + "loss": 1.1424, + "step": 2079 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979009546486675e-05, + "loss": 1.2033, + "step": 2080 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978959076055252e-05, + "loss": 1.2075, + "step": 2081 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978908545083833e-05, + "loss": 1.2156, + "step": 2082 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978857953572723e-05, + "loss": 1.1048, + "step": 2083 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978807301522226e-05, + "loss": 1.1932, + "step": 2084 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978756588932656e-05, + "loss": 1.1651, + "step": 2085 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978705815804312e-05, + "loss": 1.1857, + "step": 2086 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978654982137514e-05, + "loss": 1.164, + "step": 2087 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978604087932557e-05, + "loss": 1.2299, + "step": 2088 + }, + { + "epoch": 0.05, + "learning_rate": 1.997855313318976e-05, + "loss": 1.3297, + "step": 2089 + }, + { + "epoch": 0.05, + "learning_rate": 1.997850211790943e-05, + "loss": 0.9757, + "step": 2090 + }, + { + "epoch": 0.05, + "learning_rate": 1.997845104209187e-05, + "loss": 1.3782, + "step": 2091 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978399905737395e-05, + "loss": 1.2242, + "step": 2092 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978348708846318e-05, + "loss": 1.1981, + "step": 2093 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978297451418945e-05, + "loss": 1.4702, + "step": 2094 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978246133455584e-05, + "loss": 1.3175, + "step": 2095 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978194754956558e-05, + "loss": 1.2379, + "step": 2096 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978143315922166e-05, + "loss": 1.187, + "step": 2097 + }, + { + "epoch": 0.05, + "learning_rate": 1.997809181635273e-05, + "loss": 1.105, + "step": 2098 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978040256248555e-05, + "loss": 1.0836, + "step": 2099 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977988635609957e-05, + "loss": 1.2372, + "step": 2100 + }, + { + "epoch": 0.05, + "learning_rate": 1.997793695443725e-05, + "loss": 1.2333, + "step": 2101 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977885212730745e-05, + "loss": 1.1961, + "step": 2102 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977833410490758e-05, + "loss": 1.2605, + "step": 2103 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977781547717604e-05, + "loss": 1.3532, + "step": 2104 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977729624411594e-05, + "loss": 1.2236, + "step": 2105 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977677640573048e-05, + "loss": 1.3264, + "step": 2106 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977625596202275e-05, + "loss": 0.9771, + "step": 2107 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977573491299597e-05, + "loss": 1.3766, + "step": 2108 + }, + { + "epoch": 0.05, + "learning_rate": 1.997752132586533e-05, + "loss": 1.1056, + "step": 2109 + }, + { + "epoch": 0.05, + "learning_rate": 1.997746909989978e-05, + "loss": 1.1433, + "step": 2110 + }, + { + "epoch": 0.05, + "learning_rate": 1.997741681340328e-05, + "loss": 1.1928, + "step": 2111 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977364466376135e-05, + "loss": 1.0886, + "step": 2112 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977312058818665e-05, + "loss": 1.2434, + "step": 2113 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977259590731193e-05, + "loss": 1.223, + "step": 2114 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977207062114032e-05, + "loss": 1.3113, + "step": 2115 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977154472967504e-05, + "loss": 1.1644, + "step": 2116 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977101823291923e-05, + "loss": 1.1538, + "step": 2117 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977049113087615e-05, + "loss": 1.141, + "step": 2118 + }, + { + "epoch": 0.05, + "learning_rate": 1.99769963423549e-05, + "loss": 1.3947, + "step": 2119 + }, + { + "epoch": 0.05, + "learning_rate": 1.997694351109409e-05, + "loss": 1.3487, + "step": 2120 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976890619305515e-05, + "loss": 1.3768, + "step": 2121 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976837666989484e-05, + "loss": 1.3734, + "step": 2122 + }, + { + "epoch": 0.05, + "learning_rate": 1.997678465414633e-05, + "loss": 1.1276, + "step": 2123 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976731580776373e-05, + "loss": 1.426, + "step": 2124 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976678446879927e-05, + "loss": 1.2702, + "step": 2125 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976625252457322e-05, + "loss": 1.0613, + "step": 2126 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976571997508877e-05, + "loss": 1.179, + "step": 2127 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976518682034917e-05, + "loss": 1.1118, + "step": 2128 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976465306035767e-05, + "loss": 1.2379, + "step": 2129 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976411869511746e-05, + "loss": 1.2667, + "step": 2130 + }, + { + "epoch": 0.05, + "learning_rate": 1.997635837246318e-05, + "loss": 1.4673, + "step": 2131 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976304814890396e-05, + "loss": 0.8387, + "step": 2132 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976251196793716e-05, + "loss": 1.2914, + "step": 2133 + }, + { + "epoch": 0.05, + "learning_rate": 1.997619751817347e-05, + "loss": 1.1166, + "step": 2134 + }, + { + "epoch": 0.05, + "learning_rate": 1.997614377902998e-05, + "loss": 1.2873, + "step": 2135 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976089979363566e-05, + "loss": 1.1786, + "step": 2136 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976036119174566e-05, + "loss": 1.0623, + "step": 2137 + }, + { + "epoch": 0.05, + "learning_rate": 1.99759821984633e-05, + "loss": 1.1284, + "step": 2138 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975928217230096e-05, + "loss": 0.9966, + "step": 2139 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975874175475284e-05, + "loss": 0.9412, + "step": 2140 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975820073199185e-05, + "loss": 1.0487, + "step": 2141 + }, + { + "epoch": 0.05, + "learning_rate": 1.997576591040214e-05, + "loss": 1.2317, + "step": 2142 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975711687084462e-05, + "loss": 1.4405, + "step": 2143 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975657403246492e-05, + "loss": 1.2672, + "step": 2144 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975603058888553e-05, + "loss": 1.4067, + "step": 2145 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975548654010978e-05, + "loss": 1.0517, + "step": 2146 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975494188614094e-05, + "loss": 1.0578, + "step": 2147 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975439662698234e-05, + "loss": 1.1081, + "step": 2148 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975385076263727e-05, + "loss": 1.0225, + "step": 2149 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975330429310908e-05, + "loss": 1.2373, + "step": 2150 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975275721840105e-05, + "loss": 0.9574, + "step": 2151 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975220953851648e-05, + "loss": 1.1754, + "step": 2152 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975166125345872e-05, + "loss": 1.222, + "step": 2153 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975111236323112e-05, + "loss": 1.2474, + "step": 2154 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975056286783693e-05, + "loss": 1.2824, + "step": 2155 + }, + { + "epoch": 0.05, + "learning_rate": 1.997500127672796e-05, + "loss": 1.0551, + "step": 2156 + }, + { + "epoch": 0.05, + "learning_rate": 1.997494620615624e-05, + "loss": 1.3251, + "step": 2157 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974891075068864e-05, + "loss": 1.5054, + "step": 2158 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974835883466172e-05, + "loss": 1.306, + "step": 2159 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974780631348495e-05, + "loss": 1.1477, + "step": 2160 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974725318716172e-05, + "loss": 1.1739, + "step": 2161 + }, + { + "epoch": 0.05, + "learning_rate": 1.997466994556954e-05, + "loss": 1.2158, + "step": 2162 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974614511908927e-05, + "loss": 1.0533, + "step": 2163 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974559017734676e-05, + "loss": 1.1655, + "step": 2164 + }, + { + "epoch": 0.05, + "learning_rate": 1.997450346304712e-05, + "loss": 1.2914, + "step": 2165 + }, + { + "epoch": 0.05, + "learning_rate": 1.99744478478466e-05, + "loss": 1.5139, + "step": 2166 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974392172133447e-05, + "loss": 1.3917, + "step": 2167 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974336435908005e-05, + "loss": 1.1671, + "step": 2168 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974280639170613e-05, + "loss": 1.3317, + "step": 2169 + }, + { + "epoch": 0.05, + "learning_rate": 1.99742247819216e-05, + "loss": 1.4613, + "step": 2170 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974168864161317e-05, + "loss": 1.1618, + "step": 2171 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974112885890094e-05, + "loss": 1.4131, + "step": 2172 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974056847108278e-05, + "loss": 1.0784, + "step": 2173 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974000747816203e-05, + "loss": 1.5054, + "step": 2174 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973944588014208e-05, + "loss": 1.141, + "step": 2175 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973888367702643e-05, + "loss": 1.1693, + "step": 2176 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973832086881842e-05, + "loss": 1.2474, + "step": 2177 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973775745552146e-05, + "loss": 1.0313, + "step": 2178 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973719343713898e-05, + "loss": 1.1821, + "step": 2179 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973662881367442e-05, + "loss": 1.2809, + "step": 2180 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973606358513117e-05, + "loss": 1.166, + "step": 2181 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973549775151273e-05, + "loss": 1.1291, + "step": 2182 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973493131282245e-05, + "loss": 0.9918, + "step": 2183 + }, + { + "epoch": 0.05, + "learning_rate": 1.997343642690638e-05, + "loss": 1.0235, + "step": 2184 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973379662024025e-05, + "loss": 1.2507, + "step": 2185 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973322836635517e-05, + "loss": 1.1971, + "step": 2186 + }, + { + "epoch": 0.05, + "learning_rate": 1.997326595074121e-05, + "loss": 1.0875, + "step": 2187 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973209004341442e-05, + "loss": 1.1501, + "step": 2188 + }, + { + "epoch": 0.05, + "learning_rate": 1.997315199743656e-05, + "loss": 1.2332, + "step": 2189 + }, + { + "epoch": 0.05, + "learning_rate": 1.997309493002691e-05, + "loss": 1.1303, + "step": 2190 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973037802112843e-05, + "loss": 1.2423, + "step": 2191 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972980613694698e-05, + "loss": 1.3381, + "step": 2192 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972923364772824e-05, + "loss": 1.169, + "step": 2193 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972866055347572e-05, + "loss": 1.4242, + "step": 2194 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972808685419286e-05, + "loss": 1.0858, + "step": 2195 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972751254988317e-05, + "loss": 1.1791, + "step": 2196 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972693764055015e-05, + "loss": 1.3374, + "step": 2197 + }, + { + "epoch": 0.05, + "learning_rate": 1.997263621261972e-05, + "loss": 1.1588, + "step": 2198 + }, + { + "epoch": 0.05, + "learning_rate": 1.997257860068279e-05, + "loss": 1.116, + "step": 2199 + }, + { + "epoch": 0.05, + "learning_rate": 1.997252092824457e-05, + "loss": 1.1255, + "step": 2200 + }, + { + "epoch": 0.05, + "learning_rate": 1.997246319530541e-05, + "loss": 1.0728, + "step": 2201 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972405401865663e-05, + "loss": 0.8281, + "step": 2202 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972347547925678e-05, + "loss": 1.3296, + "step": 2203 + }, + { + "epoch": 0.05, + "learning_rate": 1.997228963348581e-05, + "loss": 1.0172, + "step": 2204 + }, + { + "epoch": 0.05, + "learning_rate": 1.99722316585464e-05, + "loss": 1.2402, + "step": 2205 + }, + { + "epoch": 0.05, + "learning_rate": 1.997217362310781e-05, + "loss": 1.4138, + "step": 2206 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972115527170385e-05, + "loss": 1.202, + "step": 2207 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972057370734483e-05, + "loss": 1.0876, + "step": 2208 + }, + { + "epoch": 0.05, + "learning_rate": 1.9971999153800453e-05, + "loss": 1.2665, + "step": 2209 + }, + { + "epoch": 0.05, + "learning_rate": 1.9971940876368653e-05, + "loss": 0.8231, + "step": 2210 + }, + { + "epoch": 0.05, + "learning_rate": 1.997188253843943e-05, + "loss": 1.295, + "step": 2211 + }, + { + "epoch": 0.05, + "learning_rate": 1.9971824140013143e-05, + "loss": 1.0155, + "step": 2212 + }, + { + "epoch": 0.05, + "learning_rate": 1.997176568109014e-05, + "loss": 1.3611, + "step": 2213 + }, + { + "epoch": 0.05, + "learning_rate": 1.9971707161670786e-05, + "loss": 1.142, + "step": 2214 + }, + { + "epoch": 0.05, + "learning_rate": 1.997164858175543e-05, + "loss": 1.3249, + "step": 2215 + }, + { + "epoch": 0.05, + "learning_rate": 1.997158994134442e-05, + "loss": 0.9191, + "step": 2216 + }, + { + "epoch": 0.05, + "learning_rate": 1.997153124043813e-05, + "loss": 1.1797, + "step": 2217 + }, + { + "epoch": 0.05, + "learning_rate": 1.99714724790369e-05, + "loss": 1.3748, + "step": 2218 + }, + { + "epoch": 0.05, + "learning_rate": 1.997141365714109e-05, + "loss": 1.0947, + "step": 2219 + }, + { + "epoch": 0.05, + "learning_rate": 1.9971354774751063e-05, + "loss": 1.0465, + "step": 2220 + }, + { + "epoch": 0.05, + "learning_rate": 1.9971295831867167e-05, + "loss": 1.0826, + "step": 2221 + }, + { + "epoch": 0.05, + "learning_rate": 1.9971236828489768e-05, + "loss": 1.0769, + "step": 2222 + }, + { + "epoch": 0.05, + "learning_rate": 1.997117776461922e-05, + "loss": 1.0474, + "step": 2223 + }, + { + "epoch": 0.05, + "learning_rate": 1.9971118640255883e-05, + "loss": 1.163, + "step": 2224 + }, + { + "epoch": 0.05, + "learning_rate": 1.9971059455400112e-05, + "loss": 1.25, + "step": 2225 + }, + { + "epoch": 0.05, + "learning_rate": 1.997100021005227e-05, + "loss": 1.2518, + "step": 2226 + }, + { + "epoch": 0.05, + "learning_rate": 1.9970940904212718e-05, + "loss": 1.2802, + "step": 2227 + }, + { + "epoch": 0.05, + "learning_rate": 1.997088153788181e-05, + "loss": 1.3734, + "step": 2228 + }, + { + "epoch": 0.05, + "learning_rate": 1.9970822111059906e-05, + "loss": 1.2748, + "step": 2229 + }, + { + "epoch": 0.05, + "learning_rate": 1.9970762623747373e-05, + "loss": 1.4175, + "step": 2230 + }, + { + "epoch": 0.05, + "learning_rate": 1.9970703075944567e-05, + "loss": 1.1462, + "step": 2231 + }, + { + "epoch": 0.05, + "learning_rate": 1.9970643467651853e-05, + "loss": 1.3421, + "step": 2232 + }, + { + "epoch": 0.05, + "learning_rate": 1.997058379886959e-05, + "loss": 1.0733, + "step": 2233 + }, + { + "epoch": 0.05, + "learning_rate": 1.9970524069598136e-05, + "loss": 1.1711, + "step": 2234 + }, + { + "epoch": 0.05, + "learning_rate": 1.9970464279837862e-05, + "loss": 1.4776, + "step": 2235 + }, + { + "epoch": 0.05, + "learning_rate": 1.9970404429589126e-05, + "loss": 1.1793, + "step": 2236 + }, + { + "epoch": 0.05, + "learning_rate": 1.997034451885229e-05, + "loss": 1.3512, + "step": 2237 + }, + { + "epoch": 0.05, + "learning_rate": 1.997028454762772e-05, + "loss": 1.0582, + "step": 2238 + }, + { + "epoch": 0.05, + "learning_rate": 1.997022451591578e-05, + "loss": 1.1144, + "step": 2239 + }, + { + "epoch": 0.05, + "learning_rate": 1.997016442371683e-05, + "loss": 1.2672, + "step": 2240 + }, + { + "epoch": 0.05, + "learning_rate": 1.997010427103124e-05, + "loss": 0.891, + "step": 2241 + }, + { + "epoch": 0.05, + "learning_rate": 1.997004405785937e-05, + "loss": 1.15, + "step": 2242 + }, + { + "epoch": 0.05, + "learning_rate": 1.996998378420159e-05, + "loss": 1.1352, + "step": 2243 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969923450058264e-05, + "loss": 1.4527, + "step": 2244 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969863055429754e-05, + "loss": 1.082, + "step": 2245 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969802600316433e-05, + "loss": 1.2919, + "step": 2246 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969742084718664e-05, + "loss": 1.0343, + "step": 2247 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969681508636814e-05, + "loss": 1.1832, + "step": 2248 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969620872071252e-05, + "loss": 1.1201, + "step": 2249 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969560175022343e-05, + "loss": 1.2526, + "step": 2250 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969499417490464e-05, + "loss": 1.1164, + "step": 2251 + }, + { + "epoch": 0.05, + "learning_rate": 1.996943859947597e-05, + "loss": 1.1318, + "step": 2252 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969377720979238e-05, + "loss": 1.0801, + "step": 2253 + }, + { + "epoch": 0.05, + "learning_rate": 1.996931678200064e-05, + "loss": 1.1075, + "step": 2254 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969255782540542e-05, + "loss": 1.2153, + "step": 2255 + }, + { + "epoch": 0.05, + "learning_rate": 1.996919472259931e-05, + "loss": 1.3427, + "step": 2256 + }, + { + "epoch": 0.05, + "learning_rate": 1.996913360217732e-05, + "loss": 1.1799, + "step": 2257 + }, + { + "epoch": 0.05, + "learning_rate": 1.996907242127494e-05, + "loss": 1.2546, + "step": 2258 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969011179892547e-05, + "loss": 1.0719, + "step": 2259 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968949878030503e-05, + "loss": 1.1519, + "step": 2260 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968888515689188e-05, + "loss": 1.2066, + "step": 2261 + }, + { + "epoch": 0.05, + "learning_rate": 1.996882709286897e-05, + "loss": 1.3854, + "step": 2262 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968765609570227e-05, + "loss": 1.2845, + "step": 2263 + }, + { + "epoch": 0.05, + "learning_rate": 1.996870406579332e-05, + "loss": 1.0381, + "step": 2264 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968642461538636e-05, + "loss": 1.222, + "step": 2265 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968580796806542e-05, + "loss": 1.1129, + "step": 2266 + }, + { + "epoch": 0.05, + "learning_rate": 1.996851907159741e-05, + "loss": 1.1528, + "step": 2267 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968457285911624e-05, + "loss": 1.0386, + "step": 2268 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968395439749545e-05, + "loss": 1.2187, + "step": 2269 + }, + { + "epoch": 0.05, + "learning_rate": 1.996833353311156e-05, + "loss": 1.0549, + "step": 2270 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968271565998038e-05, + "loss": 1.1971, + "step": 2271 + }, + { + "epoch": 0.05, + "learning_rate": 1.996820953840936e-05, + "loss": 0.8922, + "step": 2272 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968147450345895e-05, + "loss": 1.3639, + "step": 2273 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968085301808026e-05, + "loss": 1.3, + "step": 2274 + }, + { + "epoch": 0.05, + "learning_rate": 1.996802309279613e-05, + "loss": 1.0378, + "step": 2275 + }, + { + "epoch": 0.05, + "learning_rate": 1.996796082331058e-05, + "loss": 1.398, + "step": 2276 + }, + { + "epoch": 0.05, + "learning_rate": 1.996789849335176e-05, + "loss": 1.3834, + "step": 2277 + }, + { + "epoch": 0.05, + "learning_rate": 1.9967836102920043e-05, + "loss": 1.1708, + "step": 2278 + }, + { + "epoch": 0.05, + "learning_rate": 1.9967773652015803e-05, + "loss": 1.1428, + "step": 2279 + }, + { + "epoch": 0.05, + "learning_rate": 1.996771114063943e-05, + "loss": 1.013, + "step": 2280 + }, + { + "epoch": 0.05, + "learning_rate": 1.99676485687913e-05, + "loss": 1.2546, + "step": 2281 + }, + { + "epoch": 0.05, + "learning_rate": 1.996758593647179e-05, + "loss": 1.1756, + "step": 2282 + }, + { + "epoch": 0.05, + "learning_rate": 1.996752324368128e-05, + "loss": 1.3908, + "step": 2283 + }, + { + "epoch": 0.05, + "learning_rate": 1.996746049042015e-05, + "loss": 1.2352, + "step": 2284 + }, + { + "epoch": 0.05, + "learning_rate": 1.996739767668878e-05, + "loss": 1.3366, + "step": 2285 + }, + { + "epoch": 0.05, + "learning_rate": 1.9967334802487553e-05, + "loss": 1.5243, + "step": 2286 + }, + { + "epoch": 0.05, + "learning_rate": 1.9967271867816854e-05, + "loss": 1.1507, + "step": 2287 + }, + { + "epoch": 0.06, + "learning_rate": 1.996720887267706e-05, + "loss": 1.2259, + "step": 2288 + }, + { + "epoch": 0.06, + "learning_rate": 1.996714581706855e-05, + "loss": 1.2273, + "step": 2289 + }, + { + "epoch": 0.06, + "learning_rate": 1.9967082700991712e-05, + "loss": 1.0561, + "step": 2290 + }, + { + "epoch": 0.06, + "learning_rate": 1.996701952444693e-05, + "loss": 1.1782, + "step": 2291 + }, + { + "epoch": 0.06, + "learning_rate": 1.9966956287434586e-05, + "loss": 1.3364, + "step": 2292 + }, + { + "epoch": 0.06, + "learning_rate": 1.996689298995506e-05, + "loss": 1.3043, + "step": 2293 + }, + { + "epoch": 0.06, + "learning_rate": 1.996682963200874e-05, + "loss": 1.0144, + "step": 2294 + }, + { + "epoch": 0.06, + "learning_rate": 1.9966766213596007e-05, + "loss": 1.029, + "step": 2295 + }, + { + "epoch": 0.06, + "learning_rate": 1.9966702734717248e-05, + "loss": 1.0966, + "step": 2296 + }, + { + "epoch": 0.06, + "learning_rate": 1.996663919537285e-05, + "loss": 1.1824, + "step": 2297 + }, + { + "epoch": 0.06, + "learning_rate": 1.9966575595563195e-05, + "loss": 0.9523, + "step": 2298 + }, + { + "epoch": 0.06, + "learning_rate": 1.9966511935288674e-05, + "loss": 1.085, + "step": 2299 + }, + { + "epoch": 0.06, + "learning_rate": 1.996644821454966e-05, + "loss": 1.0529, + "step": 2300 + }, + { + "epoch": 0.06, + "learning_rate": 1.9966384433346557e-05, + "loss": 1.0739, + "step": 2301 + }, + { + "epoch": 0.06, + "learning_rate": 1.996632059167974e-05, + "loss": 1.1085, + "step": 2302 + }, + { + "epoch": 0.06, + "learning_rate": 1.9966256689549602e-05, + "loss": 1.0295, + "step": 2303 + }, + { + "epoch": 0.06, + "learning_rate": 1.996619272695653e-05, + "loss": 1.1435, + "step": 2304 + }, + { + "epoch": 0.06, + "learning_rate": 1.996612870390091e-05, + "loss": 1.2261, + "step": 2305 + }, + { + "epoch": 0.06, + "learning_rate": 1.996606462038313e-05, + "loss": 1.1044, + "step": 2306 + }, + { + "epoch": 0.06, + "learning_rate": 1.9966000476403583e-05, + "loss": 1.2906, + "step": 2307 + }, + { + "epoch": 0.06, + "learning_rate": 1.9965936271962652e-05, + "loss": 1.2965, + "step": 2308 + }, + { + "epoch": 0.06, + "learning_rate": 1.9965872007060733e-05, + "loss": 1.2405, + "step": 2309 + }, + { + "epoch": 0.06, + "learning_rate": 1.9965807681698208e-05, + "loss": 1.2389, + "step": 2310 + }, + { + "epoch": 0.06, + "learning_rate": 1.9965743295875475e-05, + "loss": 1.2646, + "step": 2311 + }, + { + "epoch": 0.06, + "learning_rate": 1.996567884959292e-05, + "loss": 1.2989, + "step": 2312 + }, + { + "epoch": 0.06, + "learning_rate": 1.9965614342850937e-05, + "loss": 1.2503, + "step": 2313 + }, + { + "epoch": 0.06, + "learning_rate": 1.9965549775649914e-05, + "loss": 0.9214, + "step": 2314 + }, + { + "epoch": 0.06, + "learning_rate": 1.9965485147990245e-05, + "loss": 1.1913, + "step": 2315 + }, + { + "epoch": 0.06, + "learning_rate": 1.9965420459872325e-05, + "loss": 1.0922, + "step": 2316 + }, + { + "epoch": 0.06, + "learning_rate": 1.996535571129654e-05, + "loss": 1.0311, + "step": 2317 + }, + { + "epoch": 0.06, + "learning_rate": 1.9965290902263286e-05, + "loss": 1.0531, + "step": 2318 + }, + { + "epoch": 0.06, + "learning_rate": 1.9965226032772956e-05, + "loss": 1.0591, + "step": 2319 + }, + { + "epoch": 0.06, + "learning_rate": 1.9965161102825944e-05, + "loss": 1.2383, + "step": 2320 + }, + { + "epoch": 0.06, + "learning_rate": 1.9965096112422647e-05, + "loss": 1.1815, + "step": 2321 + }, + { + "epoch": 0.06, + "learning_rate": 1.996503106156345e-05, + "loss": 1.1535, + "step": 2322 + }, + { + "epoch": 0.06, + "learning_rate": 1.9964965950248757e-05, + "loss": 1.3617, + "step": 2323 + }, + { + "epoch": 0.06, + "learning_rate": 1.9964900778478958e-05, + "loss": 1.1788, + "step": 2324 + }, + { + "epoch": 0.06, + "learning_rate": 1.996483554625445e-05, + "loss": 1.1909, + "step": 2325 + }, + { + "epoch": 0.06, + "learning_rate": 1.996477025357563e-05, + "loss": 1.0622, + "step": 2326 + }, + { + "epoch": 0.06, + "learning_rate": 1.9964704900442894e-05, + "loss": 1.1225, + "step": 2327 + }, + { + "epoch": 0.06, + "learning_rate": 1.9964639486856634e-05, + "loss": 1.2335, + "step": 2328 + }, + { + "epoch": 0.06, + "learning_rate": 1.996457401281725e-05, + "loss": 1.2943, + "step": 2329 + }, + { + "epoch": 0.06, + "learning_rate": 1.996450847832514e-05, + "loss": 1.1578, + "step": 2330 + }, + { + "epoch": 0.06, + "learning_rate": 1.9964442883380703e-05, + "loss": 1.3645, + "step": 2331 + }, + { + "epoch": 0.06, + "learning_rate": 1.996437722798433e-05, + "loss": 0.9834, + "step": 2332 + }, + { + "epoch": 0.06, + "learning_rate": 1.996431151213643e-05, + "loss": 1.325, + "step": 2333 + }, + { + "epoch": 0.06, + "learning_rate": 1.996424573583739e-05, + "loss": 1.4323, + "step": 2334 + }, + { + "epoch": 0.06, + "learning_rate": 1.9964179899087618e-05, + "loss": 1.2651, + "step": 2335 + }, + { + "epoch": 0.06, + "learning_rate": 1.996411400188751e-05, + "loss": 1.1651, + "step": 2336 + }, + { + "epoch": 0.06, + "learning_rate": 1.9964048044237467e-05, + "loss": 1.2449, + "step": 2337 + }, + { + "epoch": 0.06, + "learning_rate": 1.9963982026137886e-05, + "loss": 1.0384, + "step": 2338 + }, + { + "epoch": 0.06, + "learning_rate": 1.9963915947589173e-05, + "loss": 1.1768, + "step": 2339 + }, + { + "epoch": 0.06, + "learning_rate": 1.9963849808591723e-05, + "loss": 1.2975, + "step": 2340 + }, + { + "epoch": 0.06, + "learning_rate": 1.996378360914594e-05, + "loss": 1.2143, + "step": 2341 + }, + { + "epoch": 0.06, + "learning_rate": 1.9963717349252226e-05, + "loss": 1.1078, + "step": 2342 + }, + { + "epoch": 0.06, + "learning_rate": 1.9963651028910982e-05, + "loss": 1.154, + "step": 2343 + }, + { + "epoch": 0.06, + "learning_rate": 1.996358464812261e-05, + "loss": 1.3729, + "step": 2344 + }, + { + "epoch": 0.06, + "learning_rate": 1.9963518206887515e-05, + "loss": 1.1735, + "step": 2345 + }, + { + "epoch": 0.06, + "learning_rate": 1.99634517052061e-05, + "loss": 1.2155, + "step": 2346 + }, + { + "epoch": 0.06, + "learning_rate": 1.9963385143078764e-05, + "loss": 1.1702, + "step": 2347 + }, + { + "epoch": 0.06, + "learning_rate": 1.9963318520505915e-05, + "loss": 1.1582, + "step": 2348 + }, + { + "epoch": 0.06, + "learning_rate": 1.9963251837487955e-05, + "loss": 1.0822, + "step": 2349 + }, + { + "epoch": 0.06, + "learning_rate": 1.9963185094025293e-05, + "loss": 1.3111, + "step": 2350 + }, + { + "epoch": 0.06, + "learning_rate": 1.9963118290118327e-05, + "loss": 0.9702, + "step": 2351 + }, + { + "epoch": 0.06, + "learning_rate": 1.996305142576747e-05, + "loss": 1.2724, + "step": 2352 + }, + { + "epoch": 0.06, + "learning_rate": 1.9962984500973118e-05, + "loss": 1.3445, + "step": 2353 + }, + { + "epoch": 0.06, + "learning_rate": 1.9962917515735686e-05, + "loss": 1.0271, + "step": 2354 + }, + { + "epoch": 0.06, + "learning_rate": 1.9962850470055576e-05, + "loss": 1.105, + "step": 2355 + }, + { + "epoch": 0.06, + "learning_rate": 1.9962783363933193e-05, + "loss": 1.1905, + "step": 2356 + }, + { + "epoch": 0.06, + "learning_rate": 1.996271619736895e-05, + "loss": 1.282, + "step": 2357 + }, + { + "epoch": 0.06, + "learning_rate": 1.996264897036325e-05, + "loss": 1.2341, + "step": 2358 + }, + { + "epoch": 0.06, + "learning_rate": 1.99625816829165e-05, + "loss": 1.0941, + "step": 2359 + }, + { + "epoch": 0.06, + "learning_rate": 1.9962514335029116e-05, + "loss": 1.1013, + "step": 2360 + }, + { + "epoch": 0.06, + "learning_rate": 1.9962446926701498e-05, + "loss": 1.4484, + "step": 2361 + }, + { + "epoch": 0.06, + "learning_rate": 1.9962379457934058e-05, + "loss": 1.3022, + "step": 2362 + }, + { + "epoch": 0.06, + "learning_rate": 1.9962311928727205e-05, + "loss": 1.1376, + "step": 2363 + }, + { + "epoch": 0.06, + "learning_rate": 1.9962244339081347e-05, + "loss": 1.1467, + "step": 2364 + }, + { + "epoch": 0.06, + "learning_rate": 1.99621766889969e-05, + "loss": 1.2456, + "step": 2365 + }, + { + "epoch": 0.06, + "learning_rate": 1.9962108978474265e-05, + "loss": 1.2708, + "step": 2366 + }, + { + "epoch": 0.06, + "learning_rate": 1.996204120751386e-05, + "loss": 1.0528, + "step": 2367 + }, + { + "epoch": 0.06, + "learning_rate": 1.9961973376116096e-05, + "loss": 1.0056, + "step": 2368 + }, + { + "epoch": 0.06, + "learning_rate": 1.996190548428138e-05, + "loss": 1.0126, + "step": 2369 + }, + { + "epoch": 0.06, + "learning_rate": 1.996183753201013e-05, + "loss": 1.1438, + "step": 2370 + }, + { + "epoch": 0.06, + "learning_rate": 1.9961769519302752e-05, + "loss": 1.1236, + "step": 2371 + }, + { + "epoch": 0.06, + "learning_rate": 1.996170144615966e-05, + "loss": 1.143, + "step": 2372 + }, + { + "epoch": 0.06, + "learning_rate": 1.9961633312581276e-05, + "loss": 1.2376, + "step": 2373 + }, + { + "epoch": 0.06, + "learning_rate": 1.9961565118567997e-05, + "loss": 1.3076, + "step": 2374 + }, + { + "epoch": 0.06, + "learning_rate": 1.9961496864120252e-05, + "loss": 1.063, + "step": 2375 + }, + { + "epoch": 0.06, + "learning_rate": 1.9961428549238445e-05, + "loss": 1.2842, + "step": 2376 + }, + { + "epoch": 0.06, + "learning_rate": 1.9961360173922992e-05, + "loss": 1.0496, + "step": 2377 + }, + { + "epoch": 0.06, + "learning_rate": 1.9961291738174316e-05, + "loss": 1.194, + "step": 2378 + }, + { + "epoch": 0.06, + "learning_rate": 1.996122324199282e-05, + "loss": 1.3939, + "step": 2379 + }, + { + "epoch": 0.06, + "learning_rate": 1.996115468537893e-05, + "loss": 1.3464, + "step": 2380 + }, + { + "epoch": 0.06, + "learning_rate": 1.9961086068333054e-05, + "loss": 1.1947, + "step": 2381 + }, + { + "epoch": 0.06, + "learning_rate": 1.9961017390855616e-05, + "loss": 1.2144, + "step": 2382 + }, + { + "epoch": 0.06, + "learning_rate": 1.9960948652947025e-05, + "loss": 1.2524, + "step": 2383 + }, + { + "epoch": 0.06, + "learning_rate": 1.9960879854607702e-05, + "loss": 1.0524, + "step": 2384 + }, + { + "epoch": 0.06, + "learning_rate": 1.9960810995838063e-05, + "loss": 0.9946, + "step": 2385 + }, + { + "epoch": 0.06, + "learning_rate": 1.996074207663853e-05, + "loss": 1.1811, + "step": 2386 + }, + { + "epoch": 0.06, + "learning_rate": 1.9960673097009518e-05, + "loss": 1.2625, + "step": 2387 + }, + { + "epoch": 0.06, + "learning_rate": 1.9960604056951444e-05, + "loss": 1.3356, + "step": 2388 + }, + { + "epoch": 0.06, + "learning_rate": 1.9960534956464725e-05, + "loss": 1.1714, + "step": 2389 + }, + { + "epoch": 0.06, + "learning_rate": 1.9960465795549787e-05, + "loss": 0.7815, + "step": 2390 + }, + { + "epoch": 0.06, + "learning_rate": 1.9960396574207045e-05, + "loss": 1.2398, + "step": 2391 + }, + { + "epoch": 0.06, + "learning_rate": 1.996032729243692e-05, + "loss": 1.0819, + "step": 2392 + }, + { + "epoch": 0.06, + "learning_rate": 1.996025795023983e-05, + "loss": 1.0544, + "step": 2393 + }, + { + "epoch": 0.06, + "learning_rate": 1.99601885476162e-05, + "loss": 1.2668, + "step": 2394 + }, + { + "epoch": 0.06, + "learning_rate": 1.9960119084566446e-05, + "loss": 1.0203, + "step": 2395 + }, + { + "epoch": 0.06, + "learning_rate": 1.9960049561090995e-05, + "loss": 1.1284, + "step": 2396 + }, + { + "epoch": 0.06, + "learning_rate": 1.9959979977190264e-05, + "loss": 1.2702, + "step": 2397 + }, + { + "epoch": 0.06, + "learning_rate": 1.995991033286468e-05, + "loss": 1.2138, + "step": 2398 + }, + { + "epoch": 0.06, + "learning_rate": 1.995984062811466e-05, + "loss": 1.3992, + "step": 2399 + }, + { + "epoch": 0.06, + "learning_rate": 1.9959770862940632e-05, + "loss": 1.2191, + "step": 2400 + }, + { + "epoch": 0.06, + "learning_rate": 1.9959701037343017e-05, + "loss": 1.2975, + "step": 2401 + }, + { + "epoch": 0.06, + "learning_rate": 1.9959631151322235e-05, + "loss": 1.224, + "step": 2402 + }, + { + "epoch": 0.06, + "learning_rate": 1.9959561204878714e-05, + "loss": 1.4224, + "step": 2403 + }, + { + "epoch": 0.06, + "learning_rate": 1.995949119801288e-05, + "loss": 1.1141, + "step": 2404 + }, + { + "epoch": 0.06, + "learning_rate": 1.9959421130725158e-05, + "loss": 1.3074, + "step": 2405 + }, + { + "epoch": 0.06, + "learning_rate": 1.995935100301597e-05, + "loss": 1.3227, + "step": 2406 + }, + { + "epoch": 0.06, + "learning_rate": 1.9959280814885735e-05, + "loss": 1.1129, + "step": 2407 + }, + { + "epoch": 0.06, + "learning_rate": 1.9959210566334893e-05, + "loss": 1.0197, + "step": 2408 + }, + { + "epoch": 0.06, + "learning_rate": 1.995914025736386e-05, + "loss": 1.0797, + "step": 2409 + }, + { + "epoch": 0.06, + "learning_rate": 1.9959069887973067e-05, + "loss": 1.2876, + "step": 2410 + }, + { + "epoch": 0.06, + "learning_rate": 1.9958999458162934e-05, + "loss": 1.0056, + "step": 2411 + }, + { + "epoch": 0.06, + "learning_rate": 1.9958928967933903e-05, + "loss": 1.2787, + "step": 2412 + }, + { + "epoch": 0.06, + "learning_rate": 1.9958858417286385e-05, + "loss": 1.2976, + "step": 2413 + }, + { + "epoch": 0.06, + "learning_rate": 1.995878780622082e-05, + "loss": 1.4151, + "step": 2414 + }, + { + "epoch": 0.06, + "learning_rate": 1.995871713473763e-05, + "loss": 1.1974, + "step": 2415 + }, + { + "epoch": 0.06, + "learning_rate": 1.9958646402837247e-05, + "loss": 1.0966, + "step": 2416 + }, + { + "epoch": 0.06, + "learning_rate": 1.9958575610520095e-05, + "loss": 1.1647, + "step": 2417 + }, + { + "epoch": 0.06, + "learning_rate": 1.995850475778661e-05, + "loss": 1.1094, + "step": 2418 + }, + { + "epoch": 0.06, + "learning_rate": 1.995843384463722e-05, + "loss": 1.27, + "step": 2419 + }, + { + "epoch": 0.06, + "learning_rate": 1.9958362871072353e-05, + "loss": 1.1015, + "step": 2420 + }, + { + "epoch": 0.06, + "learning_rate": 1.995829183709244e-05, + "loss": 1.2049, + "step": 2421 + }, + { + "epoch": 0.06, + "learning_rate": 1.9958220742697915e-05, + "loss": 1.3574, + "step": 2422 + }, + { + "epoch": 0.06, + "learning_rate": 1.9958149587889202e-05, + "loss": 1.0958, + "step": 2423 + }, + { + "epoch": 0.06, + "learning_rate": 1.995807837266674e-05, + "loss": 1.0352, + "step": 2424 + }, + { + "epoch": 0.06, + "learning_rate": 1.995800709703096e-05, + "loss": 1.0593, + "step": 2425 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957935760982297e-05, + "loss": 1.2279, + "step": 2426 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957864364521177e-05, + "loss": 1.2163, + "step": 2427 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957792907648033e-05, + "loss": 1.338, + "step": 2428 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957721390363304e-05, + "loss": 1.2704, + "step": 2429 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957649812667422e-05, + "loss": 0.9583, + "step": 2430 + }, + { + "epoch": 0.06, + "learning_rate": 1.995757817456082e-05, + "loss": 1.3589, + "step": 2431 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957506476043934e-05, + "loss": 1.393, + "step": 2432 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957434717117198e-05, + "loss": 1.0756, + "step": 2433 + }, + { + "epoch": 0.06, + "learning_rate": 1.995736289778105e-05, + "loss": 0.8608, + "step": 2434 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957291018035918e-05, + "loss": 1.3435, + "step": 2435 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957219077882245e-05, + "loss": 1.3216, + "step": 2436 + }, + { + "epoch": 0.06, + "learning_rate": 1.995714707732047e-05, + "loss": 1.18, + "step": 2437 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957075016351023e-05, + "loss": 1.2547, + "step": 2438 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957002894974346e-05, + "loss": 1.0676, + "step": 2439 + }, + { + "epoch": 0.06, + "learning_rate": 1.995693071319087e-05, + "loss": 1.3508, + "step": 2440 + }, + { + "epoch": 0.06, + "learning_rate": 1.995685847100104e-05, + "loss": 1.0061, + "step": 2441 + }, + { + "epoch": 0.06, + "learning_rate": 1.995678616840529e-05, + "loss": 1.2265, + "step": 2442 + }, + { + "epoch": 0.06, + "learning_rate": 1.9956713805404063e-05, + "loss": 1.0982, + "step": 2443 + }, + { + "epoch": 0.06, + "learning_rate": 1.9956641381997795e-05, + "loss": 0.9878, + "step": 2444 + }, + { + "epoch": 0.06, + "learning_rate": 1.9956568898186924e-05, + "loss": 1.3228, + "step": 2445 + }, + { + "epoch": 0.06, + "learning_rate": 1.9956496353971894e-05, + "loss": 1.2034, + "step": 2446 + }, + { + "epoch": 0.06, + "learning_rate": 1.9956423749353142e-05, + "loss": 1.2911, + "step": 2447 + }, + { + "epoch": 0.06, + "learning_rate": 1.995635108433111e-05, + "loss": 1.2926, + "step": 2448 + }, + { + "epoch": 0.06, + "learning_rate": 1.9956278358906234e-05, + "loss": 1.1598, + "step": 2449 + }, + { + "epoch": 0.06, + "learning_rate": 1.995620557307896e-05, + "loss": 1.386, + "step": 2450 + }, + { + "epoch": 0.06, + "learning_rate": 1.9956132726849727e-05, + "loss": 1.1554, + "step": 2451 + }, + { + "epoch": 0.06, + "learning_rate": 1.9956059820218982e-05, + "loss": 1.0116, + "step": 2452 + }, + { + "epoch": 0.06, + "learning_rate": 1.9955986853187165e-05, + "loss": 1.2765, + "step": 2453 + }, + { + "epoch": 0.06, + "learning_rate": 1.9955913825754713e-05, + "loss": 1.3907, + "step": 2454 + }, + { + "epoch": 0.06, + "learning_rate": 1.9955840737922075e-05, + "loss": 1.1925, + "step": 2455 + }, + { + "epoch": 0.06, + "learning_rate": 1.9955767589689697e-05, + "loss": 1.1335, + "step": 2456 + }, + { + "epoch": 0.06, + "learning_rate": 1.9955694381058015e-05, + "loss": 1.2262, + "step": 2457 + }, + { + "epoch": 0.06, + "learning_rate": 1.9955621112027476e-05, + "loss": 1.0522, + "step": 2458 + }, + { + "epoch": 0.06, + "learning_rate": 1.995554778259853e-05, + "loss": 1.4071, + "step": 2459 + }, + { + "epoch": 0.06, + "learning_rate": 1.995547439277161e-05, + "loss": 1.1088, + "step": 2460 + }, + { + "epoch": 0.06, + "learning_rate": 1.9955400942547176e-05, + "loss": 1.2393, + "step": 2461 + }, + { + "epoch": 0.06, + "learning_rate": 1.9955327431925663e-05, + "loss": 1.2354, + "step": 2462 + }, + { + "epoch": 0.06, + "learning_rate": 1.9955253860907523e-05, + "loss": 1.3258, + "step": 2463 + }, + { + "epoch": 0.06, + "learning_rate": 1.9955180229493193e-05, + "loss": 1.1386, + "step": 2464 + }, + { + "epoch": 0.06, + "learning_rate": 1.995510653768313e-05, + "loss": 1.1051, + "step": 2465 + }, + { + "epoch": 0.06, + "learning_rate": 1.9955032785477778e-05, + "loss": 1.3526, + "step": 2466 + }, + { + "epoch": 0.06, + "learning_rate": 1.9954958972877584e-05, + "loss": 1.1848, + "step": 2467 + }, + { + "epoch": 0.06, + "learning_rate": 1.9954885099882992e-05, + "loss": 1.2547, + "step": 2468 + }, + { + "epoch": 0.06, + "learning_rate": 1.9954811166494457e-05, + "loss": 1.1242, + "step": 2469 + }, + { + "epoch": 0.06, + "learning_rate": 1.9954737172712422e-05, + "loss": 0.9796, + "step": 2470 + }, + { + "epoch": 0.06, + "learning_rate": 1.995466311853734e-05, + "loss": 1.2591, + "step": 2471 + }, + { + "epoch": 0.06, + "learning_rate": 1.9954589003969657e-05, + "loss": 1.2455, + "step": 2472 + }, + { + "epoch": 0.06, + "learning_rate": 1.9954514829009826e-05, + "loss": 1.1436, + "step": 2473 + }, + { + "epoch": 0.06, + "learning_rate": 1.9954440593658294e-05, + "loss": 1.3222, + "step": 2474 + }, + { + "epoch": 0.06, + "learning_rate": 1.9954366297915508e-05, + "loss": 1.2238, + "step": 2475 + }, + { + "epoch": 0.06, + "learning_rate": 1.995429194178193e-05, + "loss": 1.0085, + "step": 2476 + }, + { + "epoch": 0.06, + "learning_rate": 1.9954217525257998e-05, + "loss": 0.9833, + "step": 2477 + }, + { + "epoch": 0.06, + "learning_rate": 1.9954143048344175e-05, + "loss": 1.0989, + "step": 2478 + }, + { + "epoch": 0.06, + "learning_rate": 1.9954068511040905e-05, + "loss": 1.2696, + "step": 2479 + }, + { + "epoch": 0.06, + "learning_rate": 1.995399391334864e-05, + "loss": 1.2386, + "step": 2480 + }, + { + "epoch": 0.06, + "learning_rate": 1.995391925526784e-05, + "loss": 1.2418, + "step": 2481 + }, + { + "epoch": 0.06, + "learning_rate": 1.995384453679895e-05, + "loss": 1.2221, + "step": 2482 + }, + { + "epoch": 0.06, + "learning_rate": 1.995376975794243e-05, + "loss": 1.2436, + "step": 2483 + }, + { + "epoch": 0.06, + "learning_rate": 1.9953694918698726e-05, + "loss": 1.4713, + "step": 2484 + }, + { + "epoch": 0.06, + "learning_rate": 1.9953620019068297e-05, + "loss": 0.9926, + "step": 2485 + }, + { + "epoch": 0.06, + "learning_rate": 1.99535450590516e-05, + "loss": 1.224, + "step": 2486 + }, + { + "epoch": 0.06, + "learning_rate": 1.9953470038649083e-05, + "loss": 0.9814, + "step": 2487 + }, + { + "epoch": 0.06, + "learning_rate": 1.9953394957861206e-05, + "loss": 1.1055, + "step": 2488 + }, + { + "epoch": 0.06, + "learning_rate": 1.9953319816688422e-05, + "loss": 1.1454, + "step": 2489 + }, + { + "epoch": 0.06, + "learning_rate": 1.9953244615131187e-05, + "loss": 1.0043, + "step": 2490 + }, + { + "epoch": 0.06, + "learning_rate": 1.995316935318996e-05, + "loss": 1.18, + "step": 2491 + }, + { + "epoch": 0.06, + "learning_rate": 1.9953094030865197e-05, + "loss": 1.1535, + "step": 2492 + }, + { + "epoch": 0.06, + "learning_rate": 1.995301864815735e-05, + "loss": 1.046, + "step": 2493 + }, + { + "epoch": 0.06, + "learning_rate": 1.995294320506688e-05, + "loss": 1.245, + "step": 2494 + }, + { + "epoch": 0.06, + "learning_rate": 1.9952867701594246e-05, + "loss": 1.1369, + "step": 2495 + }, + { + "epoch": 0.06, + "learning_rate": 1.9952792137739908e-05, + "loss": 1.2742, + "step": 2496 + }, + { + "epoch": 0.06, + "learning_rate": 1.995271651350432e-05, + "loss": 1.1859, + "step": 2497 + }, + { + "epoch": 0.06, + "learning_rate": 1.995264082888794e-05, + "loss": 1.3286, + "step": 2498 + }, + { + "epoch": 0.06, + "learning_rate": 1.9952565083891224e-05, + "loss": 1.2407, + "step": 2499 + }, + { + "epoch": 0.06, + "learning_rate": 1.9952489278514644e-05, + "loss": 1.1585, + "step": 2500 + }, + { + "epoch": 0.06, + "learning_rate": 1.995241341275865e-05, + "loss": 1.0678, + "step": 2501 + }, + { + "epoch": 0.06, + "learning_rate": 1.9952337486623704e-05, + "loss": 1.2, + "step": 2502 + }, + { + "epoch": 0.06, + "learning_rate": 1.995226150011027e-05, + "loss": 1.0489, + "step": 2503 + }, + { + "epoch": 0.06, + "learning_rate": 1.9952185453218803e-05, + "loss": 1.1275, + "step": 2504 + }, + { + "epoch": 0.06, + "learning_rate": 1.995210934594977e-05, + "loss": 1.3451, + "step": 2505 + }, + { + "epoch": 0.06, + "learning_rate": 1.9952033178303632e-05, + "loss": 1.1761, + "step": 2506 + }, + { + "epoch": 0.06, + "learning_rate": 1.9951956950280844e-05, + "loss": 1.0789, + "step": 2507 + }, + { + "epoch": 0.06, + "learning_rate": 1.995188066188188e-05, + "loss": 1.1139, + "step": 2508 + }, + { + "epoch": 0.06, + "learning_rate": 1.995180431310719e-05, + "loss": 1.3337, + "step": 2509 + }, + { + "epoch": 0.06, + "learning_rate": 1.9951727903957252e-05, + "loss": 1.0814, + "step": 2510 + }, + { + "epoch": 0.06, + "learning_rate": 1.9951651434432516e-05, + "loss": 1.231, + "step": 2511 + }, + { + "epoch": 0.06, + "learning_rate": 1.9951574904533456e-05, + "loss": 1.1127, + "step": 2512 + }, + { + "epoch": 0.06, + "learning_rate": 1.9951498314260528e-05, + "loss": 1.1503, + "step": 2513 + }, + { + "epoch": 0.06, + "learning_rate": 1.9951421663614204e-05, + "loss": 1.2628, + "step": 2514 + }, + { + "epoch": 0.06, + "learning_rate": 1.9951344952594944e-05, + "loss": 1.3088, + "step": 2515 + }, + { + "epoch": 0.06, + "learning_rate": 1.9951268181203213e-05, + "loss": 1.2819, + "step": 2516 + }, + { + "epoch": 0.06, + "learning_rate": 1.995119134943948e-05, + "loss": 1.1908, + "step": 2517 + }, + { + "epoch": 0.06, + "learning_rate": 1.995111445730421e-05, + "loss": 1.0751, + "step": 2518 + }, + { + "epoch": 0.06, + "learning_rate": 1.9951037504797868e-05, + "loss": 1.3106, + "step": 2519 + }, + { + "epoch": 0.06, + "learning_rate": 1.9950960491920923e-05, + "loss": 1.1508, + "step": 2520 + }, + { + "epoch": 0.06, + "learning_rate": 1.995088341867384e-05, + "loss": 1.0571, + "step": 2521 + }, + { + "epoch": 0.06, + "learning_rate": 1.9950806285057092e-05, + "loss": 1.0869, + "step": 2522 + }, + { + "epoch": 0.06, + "learning_rate": 1.9950729091071137e-05, + "loss": 1.0221, + "step": 2523 + }, + { + "epoch": 0.06, + "learning_rate": 1.9950651836716453e-05, + "loss": 1.2567, + "step": 2524 + }, + { + "epoch": 0.06, + "learning_rate": 1.9950574521993504e-05, + "loss": 1.3287, + "step": 2525 + }, + { + "epoch": 0.06, + "learning_rate": 1.9950497146902757e-05, + "loss": 0.8231, + "step": 2526 + }, + { + "epoch": 0.06, + "learning_rate": 1.995041971144469e-05, + "loss": 1.4228, + "step": 2527 + }, + { + "epoch": 0.06, + "learning_rate": 1.9950342215619764e-05, + "loss": 1.2217, + "step": 2528 + }, + { + "epoch": 0.06, + "learning_rate": 1.995026465942845e-05, + "loss": 1.2568, + "step": 2529 + }, + { + "epoch": 0.06, + "learning_rate": 1.9950187042871226e-05, + "loss": 1.1743, + "step": 2530 + }, + { + "epoch": 0.06, + "learning_rate": 1.9950109365948552e-05, + "loss": 1.2516, + "step": 2531 + }, + { + "epoch": 0.06, + "learning_rate": 1.995003162866091e-05, + "loss": 1.4113, + "step": 2532 + }, + { + "epoch": 0.06, + "learning_rate": 1.9949953831008763e-05, + "loss": 0.9847, + "step": 2533 + }, + { + "epoch": 0.06, + "learning_rate": 1.994987597299259e-05, + "loss": 1.184, + "step": 2534 + }, + { + "epoch": 0.06, + "learning_rate": 1.9949798054612857e-05, + "loss": 1.2603, + "step": 2535 + }, + { + "epoch": 0.06, + "learning_rate": 1.994972007587004e-05, + "loss": 1.1028, + "step": 2536 + }, + { + "epoch": 0.06, + "learning_rate": 1.9949642036764612e-05, + "loss": 1.2643, + "step": 2537 + }, + { + "epoch": 0.06, + "learning_rate": 1.9949563937297045e-05, + "loss": 1.0217, + "step": 2538 + }, + { + "epoch": 0.06, + "learning_rate": 1.9949485777467817e-05, + "loss": 1.0692, + "step": 2539 + }, + { + "epoch": 0.06, + "learning_rate": 1.9949407557277394e-05, + "loss": 1.4136, + "step": 2540 + }, + { + "epoch": 0.06, + "learning_rate": 1.994932927672626e-05, + "loss": 1.0858, + "step": 2541 + }, + { + "epoch": 0.06, + "learning_rate": 1.9949250935814884e-05, + "loss": 1.2628, + "step": 2542 + }, + { + "epoch": 0.06, + "learning_rate": 1.9949172534543742e-05, + "loss": 0.9348, + "step": 2543 + }, + { + "epoch": 0.06, + "learning_rate": 1.994909407291331e-05, + "loss": 1.1784, + "step": 2544 + }, + { + "epoch": 0.06, + "learning_rate": 1.9949015550924065e-05, + "loss": 1.3948, + "step": 2545 + }, + { + "epoch": 0.06, + "learning_rate": 1.9948936968576483e-05, + "loss": 1.2836, + "step": 2546 + }, + { + "epoch": 0.06, + "learning_rate": 1.9948858325871044e-05, + "loss": 1.2498, + "step": 2547 + }, + { + "epoch": 0.06, + "learning_rate": 1.9948779622808215e-05, + "loss": 1.1404, + "step": 2548 + }, + { + "epoch": 0.06, + "learning_rate": 1.9948700859388482e-05, + "loss": 1.4851, + "step": 2549 + }, + { + "epoch": 0.06, + "learning_rate": 1.9948622035612326e-05, + "loss": 1.1854, + "step": 2550 + }, + { + "epoch": 0.06, + "learning_rate": 1.9948543151480216e-05, + "loss": 1.454, + "step": 2551 + }, + { + "epoch": 0.06, + "learning_rate": 1.9948464206992635e-05, + "loss": 0.9132, + "step": 2552 + }, + { + "epoch": 0.06, + "learning_rate": 1.9948385202150064e-05, + "loss": 1.2545, + "step": 2553 + }, + { + "epoch": 0.06, + "learning_rate": 1.9948306136952976e-05, + "loss": 1.1831, + "step": 2554 + }, + { + "epoch": 0.06, + "learning_rate": 1.9948227011401857e-05, + "loss": 1.2837, + "step": 2555 + }, + { + "epoch": 0.06, + "learning_rate": 1.9948147825497184e-05, + "loss": 1.3439, + "step": 2556 + }, + { + "epoch": 0.06, + "learning_rate": 1.9948068579239438e-05, + "loss": 1.1377, + "step": 2557 + }, + { + "epoch": 0.06, + "learning_rate": 1.99479892726291e-05, + "loss": 1.2446, + "step": 2558 + }, + { + "epoch": 0.06, + "learning_rate": 1.9947909905666652e-05, + "loss": 1.3043, + "step": 2559 + }, + { + "epoch": 0.06, + "learning_rate": 1.9947830478352578e-05, + "loss": 1.1344, + "step": 2560 + }, + { + "epoch": 0.06, + "learning_rate": 1.994775099068735e-05, + "loss": 1.3138, + "step": 2561 + }, + { + "epoch": 0.06, + "learning_rate": 1.994767144267146e-05, + "loss": 1.2799, + "step": 2562 + }, + { + "epoch": 0.06, + "learning_rate": 1.9947591834305383e-05, + "loss": 1.1522, + "step": 2563 + }, + { + "epoch": 0.06, + "learning_rate": 1.994751216558961e-05, + "loss": 1.221, + "step": 2564 + }, + { + "epoch": 0.06, + "learning_rate": 1.9947432436524622e-05, + "loss": 1.228, + "step": 2565 + }, + { + "epoch": 0.06, + "learning_rate": 1.9947352647110895e-05, + "loss": 1.0746, + "step": 2566 + }, + { + "epoch": 0.06, + "learning_rate": 1.9947272797348923e-05, + "loss": 1.021, + "step": 2567 + }, + { + "epoch": 0.06, + "learning_rate": 1.9947192887239188e-05, + "loss": 1.1166, + "step": 2568 + }, + { + "epoch": 0.06, + "learning_rate": 1.9947112916782174e-05, + "loss": 1.3145, + "step": 2569 + }, + { + "epoch": 0.06, + "learning_rate": 1.9947032885978365e-05, + "loss": 1.005, + "step": 2570 + }, + { + "epoch": 0.06, + "learning_rate": 1.9946952794828243e-05, + "loss": 1.2501, + "step": 2571 + }, + { + "epoch": 0.06, + "learning_rate": 1.99468726433323e-05, + "loss": 1.2482, + "step": 2572 + }, + { + "epoch": 0.06, + "learning_rate": 1.9946792431491023e-05, + "loss": 1.0943, + "step": 2573 + }, + { + "epoch": 0.06, + "learning_rate": 1.9946712159304894e-05, + "loss": 0.9317, + "step": 2574 + }, + { + "epoch": 0.06, + "learning_rate": 1.9946631826774404e-05, + "loss": 1.3588, + "step": 2575 + }, + { + "epoch": 0.06, + "learning_rate": 1.9946551433900033e-05, + "loss": 1.1779, + "step": 2576 + }, + { + "epoch": 0.06, + "learning_rate": 1.994647098068228e-05, + "loss": 1.1325, + "step": 2577 + }, + { + "epoch": 0.06, + "learning_rate": 1.9946390467121627e-05, + "loss": 1.1905, + "step": 2578 + }, + { + "epoch": 0.06, + "learning_rate": 1.9946309893218558e-05, + "loss": 1.326, + "step": 2579 + }, + { + "epoch": 0.06, + "learning_rate": 1.994622925897357e-05, + "loss": 1.2675, + "step": 2580 + }, + { + "epoch": 0.06, + "learning_rate": 1.9946148564387146e-05, + "loss": 1.4062, + "step": 2581 + }, + { + "epoch": 0.06, + "learning_rate": 1.994606780945978e-05, + "loss": 1.18, + "step": 2582 + }, + { + "epoch": 0.06, + "learning_rate": 1.994598699419196e-05, + "loss": 1.0826, + "step": 2583 + }, + { + "epoch": 0.06, + "learning_rate": 1.9945906118584173e-05, + "loss": 1.2486, + "step": 2584 + }, + { + "epoch": 0.06, + "learning_rate": 1.994582518263692e-05, + "loss": 1.0786, + "step": 2585 + }, + { + "epoch": 0.06, + "learning_rate": 1.994574418635068e-05, + "loss": 1.2196, + "step": 2586 + }, + { + "epoch": 0.06, + "learning_rate": 1.994566312972595e-05, + "loss": 1.1714, + "step": 2587 + }, + { + "epoch": 0.06, + "learning_rate": 1.994558201276322e-05, + "loss": 1.0965, + "step": 2588 + }, + { + "epoch": 0.06, + "learning_rate": 1.9945500835462986e-05, + "loss": 1.2661, + "step": 2589 + }, + { + "epoch": 0.06, + "learning_rate": 1.9945419597825736e-05, + "loss": 1.3832, + "step": 2590 + }, + { + "epoch": 0.06, + "learning_rate": 1.9945338299851965e-05, + "loss": 0.9792, + "step": 2591 + }, + { + "epoch": 0.06, + "learning_rate": 1.9945256941542163e-05, + "loss": 1.1608, + "step": 2592 + }, + { + "epoch": 0.06, + "learning_rate": 1.9945175522896832e-05, + "loss": 1.2667, + "step": 2593 + }, + { + "epoch": 0.06, + "learning_rate": 1.9945094043916456e-05, + "loss": 1.2954, + "step": 2594 + }, + { + "epoch": 0.06, + "learning_rate": 1.9945012504601537e-05, + "loss": 1.2, + "step": 2595 + }, + { + "epoch": 0.06, + "learning_rate": 1.9944930904952564e-05, + "loss": 1.1022, + "step": 2596 + }, + { + "epoch": 0.06, + "learning_rate": 1.9944849244970036e-05, + "loss": 1.4059, + "step": 2597 + }, + { + "epoch": 0.06, + "learning_rate": 1.9944767524654446e-05, + "loss": 1.0648, + "step": 2598 + }, + { + "epoch": 0.06, + "learning_rate": 1.9944685744006288e-05, + "loss": 1.3449, + "step": 2599 + }, + { + "epoch": 0.06, + "learning_rate": 1.9944603903026064e-05, + "loss": 1.2183, + "step": 2600 + }, + { + "epoch": 0.06, + "learning_rate": 1.9944522001714265e-05, + "loss": 1.4192, + "step": 2601 + }, + { + "epoch": 0.06, + "learning_rate": 1.9944440040071392e-05, + "loss": 1.201, + "step": 2602 + }, + { + "epoch": 0.06, + "learning_rate": 1.994435801809794e-05, + "loss": 1.2018, + "step": 2603 + }, + { + "epoch": 0.06, + "learning_rate": 1.9944275935794407e-05, + "loss": 1.2108, + "step": 2604 + }, + { + "epoch": 0.06, + "learning_rate": 1.994419379316129e-05, + "loss": 1.0895, + "step": 2605 + }, + { + "epoch": 0.06, + "learning_rate": 1.9944111590199088e-05, + "loss": 1.3996, + "step": 2606 + }, + { + "epoch": 0.06, + "learning_rate": 1.9944029326908304e-05, + "loss": 1.1697, + "step": 2607 + }, + { + "epoch": 0.06, + "learning_rate": 1.994394700328943e-05, + "loss": 1.176, + "step": 2608 + }, + { + "epoch": 0.06, + "learning_rate": 1.9943864619342968e-05, + "loss": 1.0287, + "step": 2609 + }, + { + "epoch": 0.06, + "learning_rate": 1.994378217506942e-05, + "loss": 1.2959, + "step": 2610 + }, + { + "epoch": 0.06, + "learning_rate": 1.9943699670469283e-05, + "loss": 1.1067, + "step": 2611 + }, + { + "epoch": 0.06, + "learning_rate": 1.994361710554306e-05, + "loss": 1.1735, + "step": 2612 + }, + { + "epoch": 0.06, + "learning_rate": 1.994353448029125e-05, + "loss": 1.1475, + "step": 2613 + }, + { + "epoch": 0.06, + "learning_rate": 1.9943451794714354e-05, + "loss": 1.1067, + "step": 2614 + }, + { + "epoch": 0.06, + "learning_rate": 1.9943369048812874e-05, + "loss": 1.2768, + "step": 2615 + }, + { + "epoch": 0.06, + "learning_rate": 1.994328624258732e-05, + "loss": 1.1209, + "step": 2616 + }, + { + "epoch": 0.06, + "learning_rate": 1.994320337603818e-05, + "loss": 1.1433, + "step": 2617 + }, + { + "epoch": 0.06, + "learning_rate": 1.9943120449165963e-05, + "loss": 1.2357, + "step": 2618 + }, + { + "epoch": 0.06, + "learning_rate": 1.994303746197118e-05, + "loss": 1.3435, + "step": 2619 + }, + { + "epoch": 0.06, + "learning_rate": 1.9942954414454322e-05, + "loss": 1.1074, + "step": 2620 + }, + { + "epoch": 0.06, + "learning_rate": 1.99428713066159e-05, + "loss": 0.9955, + "step": 2621 + }, + { + "epoch": 0.06, + "learning_rate": 1.9942788138456418e-05, + "loss": 1.2167, + "step": 2622 + }, + { + "epoch": 0.06, + "learning_rate": 1.9942704909976374e-05, + "loss": 1.0915, + "step": 2623 + }, + { + "epoch": 0.06, + "learning_rate": 1.9942621621176283e-05, + "loss": 1.3376, + "step": 2624 + }, + { + "epoch": 0.06, + "learning_rate": 1.994253827205664e-05, + "loss": 1.2252, + "step": 2625 + }, + { + "epoch": 0.06, + "learning_rate": 1.9942454862617962e-05, + "loss": 1.2732, + "step": 2626 + }, + { + "epoch": 0.06, + "learning_rate": 1.9942371392860747e-05, + "loss": 1.1492, + "step": 2627 + }, + { + "epoch": 0.06, + "learning_rate": 1.9942287862785502e-05, + "loss": 1.2649, + "step": 2628 + }, + { + "epoch": 0.06, + "learning_rate": 1.994220427239274e-05, + "loss": 1.1985, + "step": 2629 + }, + { + "epoch": 0.06, + "learning_rate": 1.9942120621682957e-05, + "loss": 1.5518, + "step": 2630 + }, + { + "epoch": 0.06, + "learning_rate": 1.9942036910656667e-05, + "loss": 1.0832, + "step": 2631 + }, + { + "epoch": 0.06, + "learning_rate": 1.994195313931438e-05, + "loss": 1.1494, + "step": 2632 + }, + { + "epoch": 0.06, + "learning_rate": 1.99418693076566e-05, + "loss": 1.2608, + "step": 2633 + }, + { + "epoch": 0.06, + "learning_rate": 1.994178541568384e-05, + "loss": 1.2028, + "step": 2634 + }, + { + "epoch": 0.06, + "learning_rate": 1.9941701463396606e-05, + "loss": 1.3096, + "step": 2635 + }, + { + "epoch": 0.06, + "learning_rate": 1.9941617450795406e-05, + "loss": 1.345, + "step": 2636 + }, + { + "epoch": 0.06, + "learning_rate": 1.994153337788075e-05, + "loss": 0.9655, + "step": 2637 + }, + { + "epoch": 0.06, + "learning_rate": 1.9941449244653154e-05, + "loss": 1.0938, + "step": 2638 + }, + { + "epoch": 0.06, + "learning_rate": 1.994136505111312e-05, + "loss": 1.2277, + "step": 2639 + }, + { + "epoch": 0.06, + "learning_rate": 1.9941280797261163e-05, + "loss": 1.0491, + "step": 2640 + }, + { + "epoch": 0.06, + "learning_rate": 1.9941196483097795e-05, + "loss": 1.0425, + "step": 2641 + }, + { + "epoch": 0.06, + "learning_rate": 1.9941112108623523e-05, + "loss": 1.3115, + "step": 2642 + }, + { + "epoch": 0.06, + "learning_rate": 1.9941027673838865e-05, + "loss": 1.11, + "step": 2643 + }, + { + "epoch": 0.06, + "learning_rate": 1.9940943178744333e-05, + "loss": 1.2247, + "step": 2644 + }, + { + "epoch": 0.06, + "learning_rate": 1.9940858623340432e-05, + "loss": 1.1266, + "step": 2645 + }, + { + "epoch": 0.06, + "learning_rate": 1.9940774007627684e-05, + "loss": 1.1302, + "step": 2646 + }, + { + "epoch": 0.06, + "learning_rate": 1.9940689331606596e-05, + "loss": 1.0713, + "step": 2647 + }, + { + "epoch": 0.06, + "learning_rate": 1.9940604595277687e-05, + "loss": 0.9037, + "step": 2648 + }, + { + "epoch": 0.06, + "learning_rate": 1.9940519798641462e-05, + "loss": 1.2569, + "step": 2649 + }, + { + "epoch": 0.06, + "learning_rate": 1.9940434941698447e-05, + "loss": 0.9692, + "step": 2650 + }, + { + "epoch": 0.06, + "learning_rate": 1.9940350024449148e-05, + "loss": 1.1053, + "step": 2651 + }, + { + "epoch": 0.06, + "learning_rate": 1.9940265046894086e-05, + "loss": 0.9163, + "step": 2652 + }, + { + "epoch": 0.06, + "learning_rate": 1.994018000903377e-05, + "loss": 1.1435, + "step": 2653 + }, + { + "epoch": 0.06, + "learning_rate": 1.9940094910868727e-05, + "loss": 1.136, + "step": 2654 + }, + { + "epoch": 0.06, + "learning_rate": 1.9940009752399462e-05, + "loss": 1.1934, + "step": 2655 + }, + { + "epoch": 0.06, + "learning_rate": 1.9939924533626492e-05, + "loss": 1.349, + "step": 2656 + }, + { + "epoch": 0.06, + "learning_rate": 1.9939839254550346e-05, + "loss": 1.0291, + "step": 2657 + }, + { + "epoch": 0.06, + "learning_rate": 1.993975391517153e-05, + "loss": 1.0511, + "step": 2658 + }, + { + "epoch": 0.06, + "learning_rate": 1.993966851549056e-05, + "loss": 1.2315, + "step": 2659 + }, + { + "epoch": 0.06, + "learning_rate": 1.9939583055507964e-05, + "loss": 0.9342, + "step": 2660 + }, + { + "epoch": 0.06, + "learning_rate": 1.9939497535224256e-05, + "loss": 1.0301, + "step": 2661 + }, + { + "epoch": 0.06, + "learning_rate": 1.9939411954639953e-05, + "loss": 1.1404, + "step": 2662 + }, + { + "epoch": 0.06, + "learning_rate": 1.9939326313755574e-05, + "loss": 1.1915, + "step": 2663 + }, + { + "epoch": 0.06, + "learning_rate": 1.9939240612571642e-05, + "loss": 1.051, + "step": 2664 + }, + { + "epoch": 0.06, + "learning_rate": 1.9939154851088678e-05, + "loss": 1.1013, + "step": 2665 + }, + { + "epoch": 0.06, + "learning_rate": 1.9939069029307193e-05, + "loss": 1.2427, + "step": 2666 + }, + { + "epoch": 0.06, + "learning_rate": 1.9938983147227716e-05, + "loss": 1.3365, + "step": 2667 + }, + { + "epoch": 0.06, + "learning_rate": 1.993889720485077e-05, + "loss": 1.1363, + "step": 2668 + }, + { + "epoch": 0.06, + "learning_rate": 1.993881120217687e-05, + "loss": 1.0303, + "step": 2669 + }, + { + "epoch": 0.06, + "learning_rate": 1.993872513920654e-05, + "loss": 1.1143, + "step": 2670 + }, + { + "epoch": 0.06, + "learning_rate": 1.9938639015940302e-05, + "loss": 1.1165, + "step": 2671 + }, + { + "epoch": 0.06, + "learning_rate": 1.993855283237868e-05, + "loss": 1.0765, + "step": 2672 + }, + { + "epoch": 0.06, + "learning_rate": 1.9938466588522194e-05, + "loss": 1.1006, + "step": 2673 + }, + { + "epoch": 0.06, + "learning_rate": 1.993838028437137e-05, + "loss": 1.1828, + "step": 2674 + }, + { + "epoch": 0.06, + "learning_rate": 1.993829391992673e-05, + "loss": 1.1022, + "step": 2675 + }, + { + "epoch": 0.06, + "learning_rate": 1.9938207495188797e-05, + "loss": 1.147, + "step": 2676 + }, + { + "epoch": 0.06, + "learning_rate": 1.99381210101581e-05, + "loss": 1.3818, + "step": 2677 + }, + { + "epoch": 0.06, + "learning_rate": 1.9938034464835155e-05, + "loss": 1.3395, + "step": 2678 + }, + { + "epoch": 0.06, + "learning_rate": 1.9937947859220497e-05, + "loss": 1.0075, + "step": 2679 + }, + { + "epoch": 0.06, + "learning_rate": 1.9937861193314648e-05, + "loss": 1.188, + "step": 2680 + }, + { + "epoch": 0.06, + "learning_rate": 1.9937774467118128e-05, + "loss": 0.9787, + "step": 2681 + }, + { + "epoch": 0.06, + "learning_rate": 1.9937687680631473e-05, + "loss": 1.2021, + "step": 2682 + }, + { + "epoch": 0.06, + "learning_rate": 1.99376008338552e-05, + "loss": 1.1253, + "step": 2683 + }, + { + "epoch": 0.06, + "learning_rate": 1.9937513926789843e-05, + "loss": 1.208, + "step": 2684 + }, + { + "epoch": 0.06, + "learning_rate": 1.9937426959435925e-05, + "loss": 1.1523, + "step": 2685 + }, + { + "epoch": 0.06, + "learning_rate": 1.9937339931793975e-05, + "loss": 1.1838, + "step": 2686 + }, + { + "epoch": 0.06, + "learning_rate": 1.9937252843864524e-05, + "loss": 1.1157, + "step": 2687 + }, + { + "epoch": 0.06, + "learning_rate": 1.9937165695648092e-05, + "loss": 1.3685, + "step": 2688 + }, + { + "epoch": 0.06, + "learning_rate": 1.9937078487145215e-05, + "loss": 1.1802, + "step": 2689 + }, + { + "epoch": 0.06, + "learning_rate": 1.993699121835642e-05, + "loss": 1.4028, + "step": 2690 + }, + { + "epoch": 0.06, + "learning_rate": 1.993690388928224e-05, + "loss": 1.2211, + "step": 2691 + }, + { + "epoch": 0.06, + "learning_rate": 1.9936816499923198e-05, + "loss": 1.32, + "step": 2692 + }, + { + "epoch": 0.06, + "learning_rate": 1.993672905027983e-05, + "loss": 1.1772, + "step": 2693 + }, + { + "epoch": 0.06, + "learning_rate": 1.9936641540352665e-05, + "loss": 1.2369, + "step": 2694 + }, + { + "epoch": 0.06, + "learning_rate": 1.993655397014223e-05, + "loss": 1.2757, + "step": 2695 + }, + { + "epoch": 0.06, + "learning_rate": 1.993646633964906e-05, + "loss": 1.4657, + "step": 2696 + }, + { + "epoch": 0.06, + "learning_rate": 1.9936378648873686e-05, + "loss": 1.0823, + "step": 2697 + }, + { + "epoch": 0.06, + "learning_rate": 1.993629089781664e-05, + "loss": 1.0947, + "step": 2698 + }, + { + "epoch": 0.06, + "learning_rate": 1.993620308647845e-05, + "loss": 1.0708, + "step": 2699 + }, + { + "epoch": 0.06, + "learning_rate": 1.9936115214859663e-05, + "loss": 0.9989, + "step": 2700 + }, + { + "epoch": 0.06, + "learning_rate": 1.9936027282960793e-05, + "loss": 1.2161, + "step": 2701 + }, + { + "epoch": 0.06, + "learning_rate": 1.9935939290782386e-05, + "loss": 1.1129, + "step": 2702 + }, + { + "epoch": 0.07, + "learning_rate": 1.9935851238324974e-05, + "loss": 1.1017, + "step": 2703 + }, + { + "epoch": 0.07, + "learning_rate": 1.9935763125589086e-05, + "loss": 1.352, + "step": 2704 + }, + { + "epoch": 0.07, + "learning_rate": 1.9935674952575263e-05, + "loss": 1.0667, + "step": 2705 + }, + { + "epoch": 0.07, + "learning_rate": 1.9935586719284036e-05, + "loss": 1.3274, + "step": 2706 + }, + { + "epoch": 0.07, + "learning_rate": 1.993549842571594e-05, + "loss": 1.279, + "step": 2707 + }, + { + "epoch": 0.07, + "learning_rate": 1.9935410071871514e-05, + "loss": 1.1984, + "step": 2708 + }, + { + "epoch": 0.07, + "learning_rate": 1.9935321657751295e-05, + "loss": 1.1417, + "step": 2709 + }, + { + "epoch": 0.07, + "learning_rate": 1.993523318335581e-05, + "loss": 1.3635, + "step": 2710 + }, + { + "epoch": 0.07, + "learning_rate": 1.9935144648685606e-05, + "loss": 1.2751, + "step": 2711 + }, + { + "epoch": 0.07, + "learning_rate": 1.9935056053741216e-05, + "loss": 1.1072, + "step": 2712 + }, + { + "epoch": 0.07, + "learning_rate": 1.993496739852318e-05, + "loss": 1.0877, + "step": 2713 + }, + { + "epoch": 0.07, + "learning_rate": 1.9934878683032028e-05, + "loss": 1.2175, + "step": 2714 + }, + { + "epoch": 0.07, + "learning_rate": 1.9934789907268307e-05, + "loss": 0.9763, + "step": 2715 + }, + { + "epoch": 0.07, + "learning_rate": 1.993470107123255e-05, + "loss": 1.1034, + "step": 2716 + }, + { + "epoch": 0.07, + "learning_rate": 1.9934612174925304e-05, + "loss": 1.2351, + "step": 2717 + }, + { + "epoch": 0.07, + "learning_rate": 1.9934523218347096e-05, + "loss": 1.1317, + "step": 2718 + }, + { + "epoch": 0.07, + "learning_rate": 1.9934434201498475e-05, + "loss": 1.1165, + "step": 2719 + }, + { + "epoch": 0.07, + "learning_rate": 1.993434512437998e-05, + "loss": 1.425, + "step": 2720 + }, + { + "epoch": 0.07, + "learning_rate": 1.9934255986992152e-05, + "loss": 1.2215, + "step": 2721 + }, + { + "epoch": 0.07, + "learning_rate": 1.9934166789335526e-05, + "loss": 1.1981, + "step": 2722 + }, + { + "epoch": 0.07, + "learning_rate": 1.9934077531410646e-05, + "loss": 1.2389, + "step": 2723 + }, + { + "epoch": 0.07, + "learning_rate": 1.9933988213218054e-05, + "loss": 1.2777, + "step": 2724 + }, + { + "epoch": 0.07, + "learning_rate": 1.9933898834758295e-05, + "loss": 1.2788, + "step": 2725 + }, + { + "epoch": 0.07, + "learning_rate": 1.9933809396031908e-05, + "loss": 1.1784, + "step": 2726 + }, + { + "epoch": 0.07, + "learning_rate": 1.9933719897039435e-05, + "loss": 1.2478, + "step": 2727 + }, + { + "epoch": 0.07, + "learning_rate": 1.993363033778142e-05, + "loss": 1.252, + "step": 2728 + }, + { + "epoch": 0.07, + "learning_rate": 1.9933540718258407e-05, + "loss": 1.2211, + "step": 2729 + }, + { + "epoch": 0.07, + "learning_rate": 1.9933451038470936e-05, + "loss": 1.0375, + "step": 2730 + }, + { + "epoch": 0.07, + "learning_rate": 1.9933361298419558e-05, + "loss": 1.1502, + "step": 2731 + }, + { + "epoch": 0.07, + "learning_rate": 1.9933271498104808e-05, + "loss": 1.1171, + "step": 2732 + }, + { + "epoch": 0.07, + "learning_rate": 1.9933181637527244e-05, + "loss": 1.2517, + "step": 2733 + }, + { + "epoch": 0.07, + "learning_rate": 1.9933091716687397e-05, + "loss": 1.296, + "step": 2734 + }, + { + "epoch": 0.07, + "learning_rate": 1.993300173558582e-05, + "loss": 1.2501, + "step": 2735 + }, + { + "epoch": 0.07, + "learning_rate": 1.9932911694223064e-05, + "loss": 0.8899, + "step": 2736 + }, + { + "epoch": 0.07, + "learning_rate": 1.993282159259966e-05, + "loss": 1.1293, + "step": 2737 + }, + { + "epoch": 0.07, + "learning_rate": 1.993273143071617e-05, + "loss": 1.0442, + "step": 2738 + }, + { + "epoch": 0.07, + "learning_rate": 1.993264120857313e-05, + "loss": 1.2439, + "step": 2739 + }, + { + "epoch": 0.07, + "learning_rate": 1.9932550926171096e-05, + "loss": 1.3845, + "step": 2740 + }, + { + "epoch": 0.07, + "learning_rate": 1.9932460583510613e-05, + "loss": 1.3389, + "step": 2741 + }, + { + "epoch": 0.07, + "learning_rate": 1.9932370180592226e-05, + "loss": 0.8131, + "step": 2742 + }, + { + "epoch": 0.07, + "learning_rate": 1.9932279717416484e-05, + "loss": 1.1499, + "step": 2743 + }, + { + "epoch": 0.07, + "learning_rate": 1.9932189193983937e-05, + "loss": 1.0869, + "step": 2744 + }, + { + "epoch": 0.07, + "learning_rate": 1.9932098610295136e-05, + "loss": 1.2651, + "step": 2745 + }, + { + "epoch": 0.07, + "learning_rate": 1.9932007966350627e-05, + "loss": 1.3912, + "step": 2746 + }, + { + "epoch": 0.07, + "learning_rate": 1.9931917262150966e-05, + "loss": 1.2192, + "step": 2747 + }, + { + "epoch": 0.07, + "learning_rate": 1.9931826497696694e-05, + "loss": 1.3758, + "step": 2748 + }, + { + "epoch": 0.07, + "learning_rate": 1.993173567298837e-05, + "loss": 1.3028, + "step": 2749 + }, + { + "epoch": 0.07, + "learning_rate": 1.993164478802654e-05, + "loss": 1.231, + "step": 2750 + }, + { + "epoch": 0.07, + "learning_rate": 1.9931553842811757e-05, + "loss": 1.2106, + "step": 2751 + }, + { + "epoch": 0.07, + "learning_rate": 1.9931462837344578e-05, + "loss": 1.0493, + "step": 2752 + }, + { + "epoch": 0.07, + "learning_rate": 1.9931371771625545e-05, + "loss": 1.4176, + "step": 2753 + }, + { + "epoch": 0.07, + "learning_rate": 1.9931280645655216e-05, + "loss": 1.27, + "step": 2754 + }, + { + "epoch": 0.07, + "learning_rate": 1.9931189459434143e-05, + "loss": 1.2217, + "step": 2755 + }, + { + "epoch": 0.07, + "learning_rate": 1.993109821296288e-05, + "loss": 1.1704, + "step": 2756 + }, + { + "epoch": 0.07, + "learning_rate": 1.993100690624198e-05, + "loss": 1.3452, + "step": 2757 + }, + { + "epoch": 0.07, + "learning_rate": 1.9930915539271996e-05, + "loss": 1.1668, + "step": 2758 + }, + { + "epoch": 0.07, + "learning_rate": 1.9930824112053485e-05, + "loss": 1.1457, + "step": 2759 + }, + { + "epoch": 0.07, + "learning_rate": 1.9930732624587e-05, + "loss": 1.2033, + "step": 2760 + }, + { + "epoch": 0.07, + "learning_rate": 1.9930641076873092e-05, + "loss": 1.158, + "step": 2761 + }, + { + "epoch": 0.07, + "learning_rate": 1.9930549468912326e-05, + "loss": 1.1236, + "step": 2762 + }, + { + "epoch": 0.07, + "learning_rate": 1.9930457800705253e-05, + "loss": 1.3999, + "step": 2763 + }, + { + "epoch": 0.07, + "learning_rate": 1.9930366072252424e-05, + "loss": 1.0881, + "step": 2764 + }, + { + "epoch": 0.07, + "learning_rate": 1.99302742835544e-05, + "loss": 1.1979, + "step": 2765 + }, + { + "epoch": 0.07, + "learning_rate": 1.9930182434611736e-05, + "loss": 1.224, + "step": 2766 + }, + { + "epoch": 0.07, + "learning_rate": 1.9930090525424996e-05, + "loss": 1.4566, + "step": 2767 + }, + { + "epoch": 0.07, + "learning_rate": 1.9929998555994732e-05, + "loss": 1.1494, + "step": 2768 + }, + { + "epoch": 0.07, + "learning_rate": 1.99299065263215e-05, + "loss": 1.2678, + "step": 2769 + }, + { + "epoch": 0.07, + "learning_rate": 1.992981443640586e-05, + "loss": 1.3307, + "step": 2770 + }, + { + "epoch": 0.07, + "learning_rate": 1.9929722286248374e-05, + "loss": 0.9858, + "step": 2771 + }, + { + "epoch": 0.07, + "learning_rate": 1.9929630075849597e-05, + "loss": 1.1846, + "step": 2772 + }, + { + "epoch": 0.07, + "learning_rate": 1.992953780521009e-05, + "loss": 1.0779, + "step": 2773 + }, + { + "epoch": 0.07, + "learning_rate": 1.9929445474330413e-05, + "loss": 1.2167, + "step": 2774 + }, + { + "epoch": 0.07, + "learning_rate": 1.992935308321113e-05, + "loss": 1.2973, + "step": 2775 + }, + { + "epoch": 0.07, + "learning_rate": 1.9929260631852792e-05, + "loss": 1.3343, + "step": 2776 + }, + { + "epoch": 0.07, + "learning_rate": 1.9929168120255962e-05, + "loss": 1.3347, + "step": 2777 + }, + { + "epoch": 0.07, + "learning_rate": 1.992907554842121e-05, + "loss": 1.3379, + "step": 2778 + }, + { + "epoch": 0.07, + "learning_rate": 1.9928982916349095e-05, + "loss": 1.2965, + "step": 2779 + }, + { + "epoch": 0.07, + "learning_rate": 1.9928890224040168e-05, + "loss": 1.1456, + "step": 2780 + }, + { + "epoch": 0.07, + "learning_rate": 1.9928797471495005e-05, + "loss": 1.4429, + "step": 2781 + }, + { + "epoch": 0.07, + "learning_rate": 1.992870465871416e-05, + "loss": 1.0807, + "step": 2782 + }, + { + "epoch": 0.07, + "learning_rate": 1.9928611785698205e-05, + "loss": 1.1559, + "step": 2783 + }, + { + "epoch": 0.07, + "learning_rate": 1.992851885244769e-05, + "loss": 0.9448, + "step": 2784 + }, + { + "epoch": 0.07, + "learning_rate": 1.9928425858963192e-05, + "loss": 1.2251, + "step": 2785 + }, + { + "epoch": 0.07, + "learning_rate": 1.9928332805245266e-05, + "loss": 1.1268, + "step": 2786 + }, + { + "epoch": 0.07, + "learning_rate": 1.9928239691294485e-05, + "loss": 1.3163, + "step": 2787 + }, + { + "epoch": 0.07, + "learning_rate": 1.9928146517111404e-05, + "loss": 1.1586, + "step": 2788 + }, + { + "epoch": 0.07, + "learning_rate": 1.9928053282696597e-05, + "loss": 1.1398, + "step": 2789 + }, + { + "epoch": 0.07, + "learning_rate": 1.9927959988050622e-05, + "loss": 1.1232, + "step": 2790 + }, + { + "epoch": 0.07, + "learning_rate": 1.9927866633174055e-05, + "loss": 1.0866, + "step": 2791 + }, + { + "epoch": 0.07, + "learning_rate": 1.992777321806745e-05, + "loss": 1.4433, + "step": 2792 + }, + { + "epoch": 0.07, + "learning_rate": 1.9927679742731382e-05, + "loss": 1.3035, + "step": 2793 + }, + { + "epoch": 0.07, + "learning_rate": 1.9927586207166417e-05, + "loss": 1.2119, + "step": 2794 + }, + { + "epoch": 0.07, + "learning_rate": 1.992749261137312e-05, + "loss": 1.0146, + "step": 2795 + }, + { + "epoch": 0.07, + "learning_rate": 1.9927398955352062e-05, + "loss": 1.2819, + "step": 2796 + }, + { + "epoch": 0.07, + "learning_rate": 1.992730523910381e-05, + "loss": 1.1513, + "step": 2797 + }, + { + "epoch": 0.07, + "learning_rate": 1.992721146262893e-05, + "loss": 1.0229, + "step": 2798 + }, + { + "epoch": 0.07, + "learning_rate": 1.9927117625927997e-05, + "loss": 1.1994, + "step": 2799 + }, + { + "epoch": 0.07, + "learning_rate": 1.992702372900157e-05, + "loss": 1.0122, + "step": 2800 + }, + { + "epoch": 0.07, + "learning_rate": 1.992692977185023e-05, + "loss": 0.8904, + "step": 2801 + }, + { + "epoch": 0.07, + "learning_rate": 1.9926835754474543e-05, + "loss": 1.338, + "step": 2802 + }, + { + "epoch": 0.07, + "learning_rate": 1.9926741676875076e-05, + "loss": 0.9712, + "step": 2803 + }, + { + "epoch": 0.07, + "learning_rate": 1.9926647539052403e-05, + "loss": 1.2781, + "step": 2804 + }, + { + "epoch": 0.07, + "learning_rate": 1.992655334100709e-05, + "loss": 1.1675, + "step": 2805 + }, + { + "epoch": 0.07, + "learning_rate": 1.9926459082739717e-05, + "loss": 1.1931, + "step": 2806 + }, + { + "epoch": 0.07, + "learning_rate": 1.9926364764250853e-05, + "loss": 0.8778, + "step": 2807 + }, + { + "epoch": 0.07, + "learning_rate": 1.9926270385541067e-05, + "loss": 1.0094, + "step": 2808 + }, + { + "epoch": 0.07, + "learning_rate": 1.9926175946610934e-05, + "loss": 0.8789, + "step": 2809 + }, + { + "epoch": 0.07, + "learning_rate": 1.9926081447461025e-05, + "loss": 1.3873, + "step": 2810 + }, + { + "epoch": 0.07, + "learning_rate": 1.9925986888091914e-05, + "loss": 1.3395, + "step": 2811 + }, + { + "epoch": 0.07, + "learning_rate": 1.9925892268504176e-05, + "loss": 1.263, + "step": 2812 + }, + { + "epoch": 0.07, + "learning_rate": 1.9925797588698385e-05, + "loss": 1.273, + "step": 2813 + }, + { + "epoch": 0.07, + "learning_rate": 1.9925702848675117e-05, + "loss": 1.0968, + "step": 2814 + }, + { + "epoch": 0.07, + "learning_rate": 1.992560804843494e-05, + "loss": 1.2035, + "step": 2815 + }, + { + "epoch": 0.07, + "learning_rate": 1.9925513187978437e-05, + "loss": 1.3805, + "step": 2816 + }, + { + "epoch": 0.07, + "learning_rate": 1.992541826730618e-05, + "loss": 1.1792, + "step": 2817 + }, + { + "epoch": 0.07, + "learning_rate": 1.992532328641874e-05, + "loss": 0.9896, + "step": 2818 + }, + { + "epoch": 0.07, + "learning_rate": 1.9925228245316703e-05, + "loss": 1.2525, + "step": 2819 + }, + { + "epoch": 0.07, + "learning_rate": 1.9925133144000643e-05, + "loss": 1.2403, + "step": 2820 + }, + { + "epoch": 0.07, + "learning_rate": 1.9925037982471128e-05, + "loss": 1.241, + "step": 2821 + }, + { + "epoch": 0.07, + "learning_rate": 1.9924942760728748e-05, + "loss": 1.1408, + "step": 2822 + }, + { + "epoch": 0.07, + "learning_rate": 1.992484747877407e-05, + "loss": 1.1274, + "step": 2823 + }, + { + "epoch": 0.07, + "learning_rate": 1.992475213660768e-05, + "loss": 1.0878, + "step": 2824 + }, + { + "epoch": 0.07, + "learning_rate": 1.992465673423015e-05, + "loss": 1.0456, + "step": 2825 + }, + { + "epoch": 0.07, + "learning_rate": 1.9924561271642066e-05, + "loss": 1.3197, + "step": 2826 + }, + { + "epoch": 0.07, + "learning_rate": 1.9924465748844e-05, + "loss": 1.1129, + "step": 2827 + }, + { + "epoch": 0.07, + "learning_rate": 1.992437016583654e-05, + "loss": 1.2676, + "step": 2828 + }, + { + "epoch": 0.07, + "learning_rate": 1.9924274522620254e-05, + "loss": 1.075, + "step": 2829 + }, + { + "epoch": 0.07, + "learning_rate": 1.9924178819195732e-05, + "loss": 1.0625, + "step": 2830 + }, + { + "epoch": 0.07, + "learning_rate": 1.992408305556355e-05, + "loss": 1.2859, + "step": 2831 + }, + { + "epoch": 0.07, + "learning_rate": 1.992398723172429e-05, + "loss": 1.0702, + "step": 2832 + }, + { + "epoch": 0.07, + "learning_rate": 1.9923891347678537e-05, + "loss": 1.1251, + "step": 2833 + }, + { + "epoch": 0.07, + "learning_rate": 1.9923795403426865e-05, + "loss": 0.9468, + "step": 2834 + }, + { + "epoch": 0.07, + "learning_rate": 1.9923699398969863e-05, + "loss": 1.1031, + "step": 2835 + }, + { + "epoch": 0.07, + "learning_rate": 1.9923603334308114e-05, + "loss": 1.0452, + "step": 2836 + }, + { + "epoch": 0.07, + "learning_rate": 1.9923507209442194e-05, + "loss": 1.1049, + "step": 2837 + }, + { + "epoch": 0.07, + "learning_rate": 1.992341102437269e-05, + "loss": 1.176, + "step": 2838 + }, + { + "epoch": 0.07, + "learning_rate": 1.9923314779100184e-05, + "loss": 1.2998, + "step": 2839 + }, + { + "epoch": 0.07, + "learning_rate": 1.9923218473625264e-05, + "loss": 1.05, + "step": 2840 + }, + { + "epoch": 0.07, + "learning_rate": 1.992312210794851e-05, + "loss": 1.1599, + "step": 2841 + }, + { + "epoch": 0.07, + "learning_rate": 1.992302568207051e-05, + "loss": 1.3498, + "step": 2842 + }, + { + "epoch": 0.07, + "learning_rate": 1.9922929195991848e-05, + "loss": 1.0729, + "step": 2843 + }, + { + "epoch": 0.07, + "learning_rate": 1.9922832649713108e-05, + "loss": 1.0264, + "step": 2844 + }, + { + "epoch": 0.07, + "learning_rate": 1.9922736043234872e-05, + "loss": 1.1737, + "step": 2845 + }, + { + "epoch": 0.07, + "learning_rate": 1.9922639376557734e-05, + "loss": 1.2971, + "step": 2846 + }, + { + "epoch": 0.07, + "learning_rate": 1.992254264968228e-05, + "loss": 1.1453, + "step": 2847 + }, + { + "epoch": 0.07, + "learning_rate": 1.9922445862609088e-05, + "loss": 1.0941, + "step": 2848 + }, + { + "epoch": 0.07, + "learning_rate": 1.9922349015338755e-05, + "loss": 1.1745, + "step": 2849 + }, + { + "epoch": 0.07, + "learning_rate": 1.992225210787186e-05, + "loss": 1.004, + "step": 2850 + }, + { + "epoch": 0.07, + "learning_rate": 1.9922155140209e-05, + "loss": 0.9156, + "step": 2851 + }, + { + "epoch": 0.07, + "learning_rate": 1.9922058112350754e-05, + "loss": 1.2625, + "step": 2852 + }, + { + "epoch": 0.07, + "learning_rate": 1.992196102429772e-05, + "loss": 1.2753, + "step": 2853 + }, + { + "epoch": 0.07, + "learning_rate": 1.992186387605048e-05, + "loss": 1.4243, + "step": 2854 + }, + { + "epoch": 0.07, + "learning_rate": 1.9921766667609624e-05, + "loss": 1.228, + "step": 2855 + }, + { + "epoch": 0.07, + "learning_rate": 1.9921669398975745e-05, + "loss": 1.0446, + "step": 2856 + }, + { + "epoch": 0.07, + "learning_rate": 1.992157207014943e-05, + "loss": 1.313, + "step": 2857 + }, + { + "epoch": 0.07, + "learning_rate": 1.9921474681131273e-05, + "loss": 1.2912, + "step": 2858 + }, + { + "epoch": 0.07, + "learning_rate": 1.9921377231921862e-05, + "loss": 1.2086, + "step": 2859 + }, + { + "epoch": 0.07, + "learning_rate": 1.992127972252179e-05, + "loss": 1.175, + "step": 2860 + }, + { + "epoch": 0.07, + "learning_rate": 1.9921182152931645e-05, + "loss": 1.2653, + "step": 2861 + }, + { + "epoch": 0.07, + "learning_rate": 1.992108452315202e-05, + "loss": 1.264, + "step": 2862 + }, + { + "epoch": 0.07, + "learning_rate": 1.9920986833183518e-05, + "loss": 1.1672, + "step": 2863 + }, + { + "epoch": 0.07, + "learning_rate": 1.9920889083026716e-05, + "loss": 1.1808, + "step": 2864 + }, + { + "epoch": 0.07, + "learning_rate": 1.9920791272682213e-05, + "loss": 1.0739, + "step": 2865 + }, + { + "epoch": 0.07, + "learning_rate": 1.9920693402150604e-05, + "loss": 1.2319, + "step": 2866 + }, + { + "epoch": 0.07, + "learning_rate": 1.992059547143248e-05, + "loss": 1.0768, + "step": 2867 + }, + { + "epoch": 0.07, + "learning_rate": 1.992049748052844e-05, + "loss": 1.0984, + "step": 2868 + }, + { + "epoch": 0.07, + "learning_rate": 1.992039942943907e-05, + "loss": 1.1257, + "step": 2869 + }, + { + "epoch": 0.07, + "learning_rate": 1.9920301318164978e-05, + "loss": 1.2341, + "step": 2870 + }, + { + "epoch": 0.07, + "learning_rate": 1.992020314670674e-05, + "loss": 1.0267, + "step": 2871 + }, + { + "epoch": 0.07, + "learning_rate": 1.9920104915064974e-05, + "loss": 1.205, + "step": 2872 + }, + { + "epoch": 0.07, + "learning_rate": 1.992000662324026e-05, + "loss": 1.0569, + "step": 2873 + }, + { + "epoch": 0.07, + "learning_rate": 1.9919908271233198e-05, + "loss": 1.4105, + "step": 2874 + }, + { + "epoch": 0.07, + "learning_rate": 1.991980985904439e-05, + "loss": 1.2395, + "step": 2875 + }, + { + "epoch": 0.07, + "learning_rate": 1.9919711386674425e-05, + "loss": 0.988, + "step": 2876 + }, + { + "epoch": 0.07, + "learning_rate": 1.9919612854123905e-05, + "loss": 1.1471, + "step": 2877 + }, + { + "epoch": 0.07, + "learning_rate": 1.991951426139343e-05, + "loss": 0.9718, + "step": 2878 + }, + { + "epoch": 0.07, + "learning_rate": 1.9919415608483592e-05, + "loss": 1.1917, + "step": 2879 + }, + { + "epoch": 0.07, + "learning_rate": 1.9919316895394993e-05, + "loss": 1.1319, + "step": 2880 + }, + { + "epoch": 0.07, + "learning_rate": 1.991921812212823e-05, + "loss": 1.1186, + "step": 2881 + }, + { + "epoch": 0.07, + "learning_rate": 1.9919119288683908e-05, + "loss": 1.0034, + "step": 2882 + }, + { + "epoch": 0.07, + "learning_rate": 1.991902039506262e-05, + "loss": 1.1219, + "step": 2883 + }, + { + "epoch": 0.07, + "learning_rate": 1.9918921441264966e-05, + "loss": 1.1212, + "step": 2884 + }, + { + "epoch": 0.07, + "learning_rate": 1.9918822427291554e-05, + "loss": 1.0243, + "step": 2885 + }, + { + "epoch": 0.07, + "learning_rate": 1.9918723353142973e-05, + "loss": 1.1831, + "step": 2886 + }, + { + "epoch": 0.07, + "learning_rate": 1.9918624218819837e-05, + "loss": 1.1598, + "step": 2887 + }, + { + "epoch": 0.07, + "learning_rate": 1.9918525024322738e-05, + "loss": 1.0843, + "step": 2888 + }, + { + "epoch": 0.07, + "learning_rate": 1.991842576965228e-05, + "loss": 1.5508, + "step": 2889 + }, + { + "epoch": 0.07, + "learning_rate": 1.9918326454809066e-05, + "loss": 1.1663, + "step": 2890 + }, + { + "epoch": 0.07, + "learning_rate": 1.99182270797937e-05, + "loss": 1.274, + "step": 2891 + }, + { + "epoch": 0.07, + "learning_rate": 1.991812764460678e-05, + "loss": 1.2883, + "step": 2892 + }, + { + "epoch": 0.07, + "learning_rate": 1.9918028149248914e-05, + "loss": 1.0024, + "step": 2893 + }, + { + "epoch": 0.07, + "learning_rate": 1.9917928593720705e-05, + "loss": 1.0513, + "step": 2894 + }, + { + "epoch": 0.07, + "learning_rate": 1.9917828978022755e-05, + "loss": 1.242, + "step": 2895 + }, + { + "epoch": 0.07, + "learning_rate": 1.991772930215567e-05, + "loss": 1.0805, + "step": 2896 + }, + { + "epoch": 0.07, + "learning_rate": 1.9917629566120054e-05, + "loss": 1.3197, + "step": 2897 + }, + { + "epoch": 0.07, + "learning_rate": 1.9917529769916513e-05, + "loss": 0.9643, + "step": 2898 + }, + { + "epoch": 0.07, + "learning_rate": 1.9917429913545653e-05, + "loss": 1.0703, + "step": 2899 + }, + { + "epoch": 0.07, + "learning_rate": 1.9917329997008075e-05, + "loss": 1.4519, + "step": 2900 + }, + { + "epoch": 0.07, + "learning_rate": 1.991723002030439e-05, + "loss": 1.218, + "step": 2901 + }, + { + "epoch": 0.07, + "learning_rate": 1.99171299834352e-05, + "loss": 1.1606, + "step": 2902 + }, + { + "epoch": 0.07, + "learning_rate": 1.991702988640112e-05, + "loss": 1.0983, + "step": 2903 + }, + { + "epoch": 0.07, + "learning_rate": 1.991692972920275e-05, + "loss": 0.9978, + "step": 2904 + }, + { + "epoch": 0.07, + "learning_rate": 1.99168295118407e-05, + "loss": 1.1708, + "step": 2905 + }, + { + "epoch": 0.07, + "learning_rate": 1.991672923431558e-05, + "loss": 1.0232, + "step": 2906 + }, + { + "epoch": 0.07, + "learning_rate": 1.9916628896627995e-05, + "loss": 1.1363, + "step": 2907 + }, + { + "epoch": 0.07, + "learning_rate": 1.9916528498778554e-05, + "loss": 1.0472, + "step": 2908 + }, + { + "epoch": 0.07, + "learning_rate": 1.9916428040767866e-05, + "loss": 1.1171, + "step": 2909 + }, + { + "epoch": 0.07, + "learning_rate": 1.9916327522596545e-05, + "loss": 1.3163, + "step": 2910 + }, + { + "epoch": 0.07, + "learning_rate": 1.9916226944265196e-05, + "loss": 1.4434, + "step": 2911 + }, + { + "epoch": 0.07, + "learning_rate": 1.9916126305774427e-05, + "loss": 0.8668, + "step": 2912 + }, + { + "epoch": 0.07, + "learning_rate": 1.9916025607124857e-05, + "loss": 1.0576, + "step": 2913 + }, + { + "epoch": 0.07, + "learning_rate": 1.991592484831709e-05, + "loss": 1.3031, + "step": 2914 + }, + { + "epoch": 0.07, + "learning_rate": 1.991582402935174e-05, + "loss": 1.3394, + "step": 2915 + }, + { + "epoch": 0.07, + "learning_rate": 1.9915723150229417e-05, + "loss": 1.1285, + "step": 2916 + }, + { + "epoch": 0.07, + "learning_rate": 1.9915622210950736e-05, + "loss": 1.2045, + "step": 2917 + }, + { + "epoch": 0.07, + "learning_rate": 1.9915521211516307e-05, + "loss": 1.1228, + "step": 2918 + }, + { + "epoch": 0.07, + "learning_rate": 1.991542015192674e-05, + "loss": 0.9894, + "step": 2919 + }, + { + "epoch": 0.07, + "learning_rate": 1.9915319032182655e-05, + "loss": 1.3127, + "step": 2920 + }, + { + "epoch": 0.07, + "learning_rate": 1.991521785228466e-05, + "loss": 1.3185, + "step": 2921 + }, + { + "epoch": 0.07, + "learning_rate": 1.9915116612233367e-05, + "loss": 1.2446, + "step": 2922 + }, + { + "epoch": 0.07, + "learning_rate": 1.99150153120294e-05, + "loss": 1.2214, + "step": 2923 + }, + { + "epoch": 0.07, + "learning_rate": 1.991491395167336e-05, + "loss": 1.0688, + "step": 2924 + }, + { + "epoch": 0.07, + "learning_rate": 1.9914812531165878e-05, + "loss": 1.4221, + "step": 2925 + }, + { + "epoch": 0.07, + "learning_rate": 1.9914711050507556e-05, + "loss": 1.5334, + "step": 2926 + }, + { + "epoch": 0.07, + "learning_rate": 1.9914609509699012e-05, + "loss": 1.4252, + "step": 2927 + }, + { + "epoch": 0.07, + "learning_rate": 1.991450790874087e-05, + "loss": 1.1965, + "step": 2928 + }, + { + "epoch": 0.07, + "learning_rate": 1.9914406247633735e-05, + "loss": 1.4009, + "step": 2929 + }, + { + "epoch": 0.07, + "learning_rate": 1.991430452637823e-05, + "loss": 1.0661, + "step": 2930 + }, + { + "epoch": 0.07, + "learning_rate": 1.991420274497497e-05, + "loss": 1.2507, + "step": 2931 + }, + { + "epoch": 0.07, + "learning_rate": 1.991410090342458e-05, + "loss": 1.0383, + "step": 2932 + }, + { + "epoch": 0.07, + "learning_rate": 1.9913999001727666e-05, + "loss": 1.0514, + "step": 2933 + }, + { + "epoch": 0.07, + "learning_rate": 1.9913897039884855e-05, + "loss": 1.4557, + "step": 2934 + }, + { + "epoch": 0.07, + "learning_rate": 1.991379501789676e-05, + "loss": 1.1353, + "step": 2935 + }, + { + "epoch": 0.07, + "learning_rate": 1.9913692935764006e-05, + "loss": 1.4686, + "step": 2936 + }, + { + "epoch": 0.07, + "learning_rate": 1.991359079348721e-05, + "loss": 1.1791, + "step": 2937 + }, + { + "epoch": 0.07, + "learning_rate": 1.9913488591066986e-05, + "loss": 1.2161, + "step": 2938 + }, + { + "epoch": 0.07, + "learning_rate": 1.9913386328503958e-05, + "loss": 0.9531, + "step": 2939 + }, + { + "epoch": 0.07, + "learning_rate": 1.991328400579875e-05, + "loss": 0.9558, + "step": 2940 + }, + { + "epoch": 0.07, + "learning_rate": 1.991318162295198e-05, + "loss": 1.1747, + "step": 2941 + }, + { + "epoch": 0.07, + "learning_rate": 1.9913079179964266e-05, + "loss": 1.0004, + "step": 2942 + }, + { + "epoch": 0.07, + "learning_rate": 1.9912976676836234e-05, + "loss": 0.9845, + "step": 2943 + }, + { + "epoch": 0.07, + "learning_rate": 1.9912874113568503e-05, + "loss": 1.4146, + "step": 2944 + }, + { + "epoch": 0.07, + "learning_rate": 1.9912771490161698e-05, + "loss": 1.2182, + "step": 2945 + }, + { + "epoch": 0.07, + "learning_rate": 1.9912668806616437e-05, + "loss": 1.0272, + "step": 2946 + }, + { + "epoch": 0.07, + "learning_rate": 1.9912566062933352e-05, + "loss": 1.1524, + "step": 2947 + }, + { + "epoch": 0.07, + "learning_rate": 1.9912463259113055e-05, + "loss": 1.2476, + "step": 2948 + }, + { + "epoch": 0.07, + "learning_rate": 1.9912360395156177e-05, + "loss": 1.1895, + "step": 2949 + }, + { + "epoch": 0.07, + "learning_rate": 1.9912257471063338e-05, + "loss": 1.1808, + "step": 2950 + }, + { + "epoch": 0.07, + "learning_rate": 1.9912154486835168e-05, + "loss": 0.9723, + "step": 2951 + }, + { + "epoch": 0.07, + "learning_rate": 1.9912051442472283e-05, + "loss": 1.2424, + "step": 2952 + }, + { + "epoch": 0.07, + "learning_rate": 1.9911948337975317e-05, + "loss": 1.2593, + "step": 2953 + }, + { + "epoch": 0.07, + "learning_rate": 1.9911845173344894e-05, + "loss": 1.1317, + "step": 2954 + }, + { + "epoch": 0.07, + "learning_rate": 1.9911741948581637e-05, + "loss": 1.403, + "step": 2955 + }, + { + "epoch": 0.07, + "learning_rate": 1.991163866368617e-05, + "loss": 1.4592, + "step": 2956 + }, + { + "epoch": 0.07, + "learning_rate": 1.991153531865912e-05, + "loss": 1.3453, + "step": 2957 + }, + { + "epoch": 0.07, + "learning_rate": 1.991143191350112e-05, + "loss": 1.0102, + "step": 2958 + }, + { + "epoch": 0.07, + "learning_rate": 1.9911328448212797e-05, + "loss": 1.1116, + "step": 2959 + }, + { + "epoch": 0.07, + "learning_rate": 1.991122492279477e-05, + "loss": 1.529, + "step": 2960 + }, + { + "epoch": 0.07, + "learning_rate": 1.9911121337247676e-05, + "loss": 1.3002, + "step": 2961 + }, + { + "epoch": 0.07, + "learning_rate": 1.991101769157214e-05, + "loss": 1.1794, + "step": 2962 + }, + { + "epoch": 0.07, + "learning_rate": 1.991091398576879e-05, + "loss": 1.3057, + "step": 2963 + }, + { + "epoch": 0.07, + "learning_rate": 1.9910810219838257e-05, + "loss": 1.1193, + "step": 2964 + }, + { + "epoch": 0.07, + "learning_rate": 1.991070639378117e-05, + "loss": 1.3483, + "step": 2965 + }, + { + "epoch": 0.07, + "learning_rate": 1.991060250759816e-05, + "loss": 1.234, + "step": 2966 + }, + { + "epoch": 0.07, + "learning_rate": 1.991049856128985e-05, + "loss": 0.9501, + "step": 2967 + }, + { + "epoch": 0.07, + "learning_rate": 1.991039455485688e-05, + "loss": 1.4129, + "step": 2968 + }, + { + "epoch": 0.07, + "learning_rate": 1.9910290488299875e-05, + "loss": 1.3968, + "step": 2969 + }, + { + "epoch": 0.07, + "learning_rate": 1.9910186361619473e-05, + "loss": 1.3337, + "step": 2970 + }, + { + "epoch": 0.07, + "learning_rate": 1.9910082174816296e-05, + "loss": 1.1247, + "step": 2971 + }, + { + "epoch": 0.07, + "learning_rate": 1.9909977927890988e-05, + "loss": 1.1279, + "step": 2972 + }, + { + "epoch": 0.07, + "learning_rate": 1.990987362084417e-05, + "loss": 1.343, + "step": 2973 + }, + { + "epoch": 0.07, + "learning_rate": 1.990976925367648e-05, + "loss": 1.1571, + "step": 2974 + }, + { + "epoch": 0.07, + "learning_rate": 1.990966482638855e-05, + "loss": 1.1233, + "step": 2975 + }, + { + "epoch": 0.07, + "learning_rate": 1.9909560338981014e-05, + "loss": 0.9958, + "step": 2976 + }, + { + "epoch": 0.07, + "learning_rate": 1.990945579145451e-05, + "loss": 1.1288, + "step": 2977 + }, + { + "epoch": 0.07, + "learning_rate": 1.990935118380967e-05, + "loss": 1.3597, + "step": 2978 + }, + { + "epoch": 0.07, + "learning_rate": 1.9909246516047123e-05, + "loss": 1.0838, + "step": 2979 + }, + { + "epoch": 0.07, + "learning_rate": 1.9909141788167506e-05, + "loss": 1.1283, + "step": 2980 + }, + { + "epoch": 0.07, + "learning_rate": 1.990903700017146e-05, + "loss": 1.028, + "step": 2981 + }, + { + "epoch": 0.07, + "learning_rate": 1.9908932152059618e-05, + "loss": 0.9797, + "step": 2982 + }, + { + "epoch": 0.07, + "learning_rate": 1.9908827243832615e-05, + "loss": 1.2993, + "step": 2983 + }, + { + "epoch": 0.07, + "learning_rate": 1.9908722275491084e-05, + "loss": 1.3622, + "step": 2984 + }, + { + "epoch": 0.07, + "learning_rate": 1.9908617247035673e-05, + "loss": 1.2159, + "step": 2985 + }, + { + "epoch": 0.07, + "learning_rate": 1.9908512158467007e-05, + "loss": 1.0172, + "step": 2986 + }, + { + "epoch": 0.07, + "learning_rate": 1.9908407009785727e-05, + "loss": 1.0418, + "step": 2987 + }, + { + "epoch": 0.07, + "learning_rate": 1.9908301800992475e-05, + "loss": 1.0528, + "step": 2988 + }, + { + "epoch": 0.07, + "learning_rate": 1.9908196532087886e-05, + "loss": 1.01, + "step": 2989 + }, + { + "epoch": 0.07, + "learning_rate": 1.99080912030726e-05, + "loss": 1.2708, + "step": 2990 + }, + { + "epoch": 0.07, + "learning_rate": 1.9907985813947253e-05, + "loss": 0.9678, + "step": 2991 + }, + { + "epoch": 0.07, + "learning_rate": 1.990788036471249e-05, + "loss": 1.198, + "step": 2992 + }, + { + "epoch": 0.07, + "learning_rate": 1.9907774855368946e-05, + "loss": 1.4029, + "step": 2993 + }, + { + "epoch": 0.07, + "learning_rate": 1.990766928591726e-05, + "loss": 1.1421, + "step": 2994 + }, + { + "epoch": 0.07, + "learning_rate": 1.9907563656358076e-05, + "loss": 0.7327, + "step": 2995 + }, + { + "epoch": 0.07, + "learning_rate": 1.9907457966692036e-05, + "loss": 0.823, + "step": 2996 + }, + { + "epoch": 0.07, + "learning_rate": 1.9907352216919775e-05, + "loss": 1.2992, + "step": 2997 + }, + { + "epoch": 0.07, + "learning_rate": 1.990724640704194e-05, + "loss": 1.1872, + "step": 2998 + }, + { + "epoch": 0.07, + "learning_rate": 1.9907140537059174e-05, + "loss": 1.4023, + "step": 2999 + }, + { + "epoch": 0.07, + "learning_rate": 1.990703460697211e-05, + "loss": 1.2608, + "step": 3000 + }, + { + "epoch": 0.07, + "learning_rate": 1.9906928616781404e-05, + "loss": 1.1445, + "step": 3001 + }, + { + "epoch": 0.07, + "learning_rate": 1.990682256648769e-05, + "loss": 1.0647, + "step": 3002 + }, + { + "epoch": 0.07, + "learning_rate": 1.9906716456091613e-05, + "loss": 1.2481, + "step": 3003 + }, + { + "epoch": 0.07, + "learning_rate": 1.990661028559382e-05, + "loss": 0.9911, + "step": 3004 + }, + { + "epoch": 0.07, + "learning_rate": 1.990650405499495e-05, + "loss": 1.195, + "step": 3005 + }, + { + "epoch": 0.07, + "learning_rate": 1.990639776429565e-05, + "loss": 1.1122, + "step": 3006 + }, + { + "epoch": 0.07, + "learning_rate": 1.9906291413496567e-05, + "loss": 1.301, + "step": 3007 + }, + { + "epoch": 0.07, + "learning_rate": 1.9906185002598343e-05, + "loss": 1.3021, + "step": 3008 + }, + { + "epoch": 0.07, + "learning_rate": 1.9906078531601624e-05, + "loss": 1.1708, + "step": 3009 + }, + { + "epoch": 0.07, + "learning_rate": 1.9905972000507057e-05, + "loss": 1.1952, + "step": 3010 + }, + { + "epoch": 0.07, + "learning_rate": 1.9905865409315288e-05, + "loss": 1.3352, + "step": 3011 + }, + { + "epoch": 0.07, + "learning_rate": 1.9905758758026966e-05, + "loss": 1.1377, + "step": 3012 + }, + { + "epoch": 0.07, + "learning_rate": 1.990565204664273e-05, + "loss": 1.0921, + "step": 3013 + }, + { + "epoch": 0.07, + "learning_rate": 1.9905545275163235e-05, + "loss": 0.9481, + "step": 3014 + }, + { + "epoch": 0.07, + "learning_rate": 1.990543844358913e-05, + "loss": 0.9814, + "step": 3015 + }, + { + "epoch": 0.07, + "learning_rate": 1.9905331551921056e-05, + "loss": 1.1091, + "step": 3016 + }, + { + "epoch": 0.07, + "learning_rate": 1.9905224600159668e-05, + "loss": 1.3147, + "step": 3017 + }, + { + "epoch": 0.07, + "learning_rate": 1.9905117588305612e-05, + "loss": 1.3284, + "step": 3018 + }, + { + "epoch": 0.07, + "learning_rate": 1.990501051635954e-05, + "loss": 1.0772, + "step": 3019 + }, + { + "epoch": 0.07, + "learning_rate": 1.9904903384322095e-05, + "loss": 1.0441, + "step": 3020 + }, + { + "epoch": 0.07, + "learning_rate": 1.990479619219393e-05, + "loss": 1.1116, + "step": 3021 + }, + { + "epoch": 0.07, + "learning_rate": 1.9904688939975697e-05, + "loss": 1.0686, + "step": 3022 + }, + { + "epoch": 0.07, + "learning_rate": 1.9904581627668047e-05, + "loss": 1.007, + "step": 3023 + }, + { + "epoch": 0.07, + "learning_rate": 1.990447425527163e-05, + "loss": 1.0049, + "step": 3024 + }, + { + "epoch": 0.07, + "learning_rate": 1.99043668227871e-05, + "loss": 1.2932, + "step": 3025 + }, + { + "epoch": 0.07, + "learning_rate": 1.99042593302151e-05, + "loss": 1.0543, + "step": 3026 + }, + { + "epoch": 0.07, + "learning_rate": 1.990415177755629e-05, + "loss": 1.1737, + "step": 3027 + }, + { + "epoch": 0.07, + "learning_rate": 1.9904044164811325e-05, + "loss": 1.3844, + "step": 3028 + }, + { + "epoch": 0.07, + "learning_rate": 1.9903936491980852e-05, + "loss": 1.0751, + "step": 3029 + }, + { + "epoch": 0.07, + "learning_rate": 1.9903828759065524e-05, + "loss": 1.2107, + "step": 3030 + }, + { + "epoch": 0.07, + "learning_rate": 1.9903720966065995e-05, + "loss": 1.1571, + "step": 3031 + }, + { + "epoch": 0.07, + "learning_rate": 1.9903613112982925e-05, + "loss": 1.2761, + "step": 3032 + }, + { + "epoch": 0.07, + "learning_rate": 1.990350519981696e-05, + "loss": 1.1794, + "step": 3033 + }, + { + "epoch": 0.07, + "learning_rate": 1.990339722656876e-05, + "loss": 1.0743, + "step": 3034 + }, + { + "epoch": 0.07, + "learning_rate": 1.990328919323898e-05, + "loss": 1.1765, + "step": 3035 + }, + { + "epoch": 0.07, + "learning_rate": 1.990318109982827e-05, + "loss": 1.2175, + "step": 3036 + }, + { + "epoch": 0.07, + "learning_rate": 1.990307294633729e-05, + "loss": 1.4043, + "step": 3037 + }, + { + "epoch": 0.07, + "learning_rate": 1.9902964732766702e-05, + "loss": 1.3545, + "step": 3038 + }, + { + "epoch": 0.07, + "learning_rate": 1.990285645911715e-05, + "loss": 1.1274, + "step": 3039 + }, + { + "epoch": 0.07, + "learning_rate": 1.99027481253893e-05, + "loss": 1.0591, + "step": 3040 + }, + { + "epoch": 0.07, + "learning_rate": 1.9902639731583805e-05, + "loss": 1.3056, + "step": 3041 + }, + { + "epoch": 0.07, + "learning_rate": 1.9902531277701323e-05, + "loss": 1.2696, + "step": 3042 + }, + { + "epoch": 0.07, + "learning_rate": 1.9902422763742515e-05, + "loss": 1.0475, + "step": 3043 + }, + { + "epoch": 0.07, + "learning_rate": 1.9902314189708037e-05, + "loss": 1.185, + "step": 3044 + }, + { + "epoch": 0.07, + "learning_rate": 1.9902205555598547e-05, + "loss": 1.2213, + "step": 3045 + }, + { + "epoch": 0.07, + "learning_rate": 1.9902096861414706e-05, + "loss": 1.0489, + "step": 3046 + }, + { + "epoch": 0.07, + "learning_rate": 1.9901988107157173e-05, + "loss": 0.939, + "step": 3047 + }, + { + "epoch": 0.07, + "learning_rate": 1.9901879292826604e-05, + "loss": 1.0235, + "step": 3048 + }, + { + "epoch": 0.07, + "learning_rate": 1.990177041842367e-05, + "loss": 1.3625, + "step": 3049 + }, + { + "epoch": 0.07, + "learning_rate": 1.9901661483949015e-05, + "loss": 0.902, + "step": 3050 + }, + { + "epoch": 0.07, + "learning_rate": 1.9901552489403312e-05, + "loss": 1.5139, + "step": 3051 + }, + { + "epoch": 0.07, + "learning_rate": 1.990144343478722e-05, + "loss": 1.1941, + "step": 3052 + }, + { + "epoch": 0.07, + "learning_rate": 1.99013343201014e-05, + "loss": 1.3088, + "step": 3053 + }, + { + "epoch": 0.07, + "learning_rate": 1.990122514534651e-05, + "loss": 1.1085, + "step": 3054 + }, + { + "epoch": 0.07, + "learning_rate": 1.9901115910523223e-05, + "loss": 1.0693, + "step": 3055 + }, + { + "epoch": 0.07, + "learning_rate": 1.9901006615632187e-05, + "loss": 1.0476, + "step": 3056 + }, + { + "epoch": 0.07, + "learning_rate": 1.9900897260674074e-05, + "loss": 0.9293, + "step": 3057 + }, + { + "epoch": 0.07, + "learning_rate": 1.9900787845649548e-05, + "loss": 1.4127, + "step": 3058 + }, + { + "epoch": 0.07, + "learning_rate": 1.9900678370559273e-05, + "loss": 1.2548, + "step": 3059 + }, + { + "epoch": 0.07, + "learning_rate": 1.990056883540391e-05, + "loss": 1.1507, + "step": 3060 + }, + { + "epoch": 0.07, + "learning_rate": 1.9900459240184125e-05, + "loss": 1.4143, + "step": 3061 + }, + { + "epoch": 0.07, + "learning_rate": 1.990034958490058e-05, + "loss": 1.0095, + "step": 3062 + }, + { + "epoch": 0.07, + "learning_rate": 1.9900239869553944e-05, + "loss": 1.3924, + "step": 3063 + }, + { + "epoch": 0.07, + "learning_rate": 1.990013009414488e-05, + "loss": 1.0, + "step": 3064 + }, + { + "epoch": 0.07, + "learning_rate": 1.9900020258674057e-05, + "loss": 1.0689, + "step": 3065 + }, + { + "epoch": 0.07, + "learning_rate": 1.989991036314214e-05, + "loss": 1.1643, + "step": 3066 + }, + { + "epoch": 0.07, + "learning_rate": 1.9899800407549796e-05, + "loss": 1.3174, + "step": 3067 + }, + { + "epoch": 0.07, + "learning_rate": 1.9899690391897694e-05, + "loss": 1.3571, + "step": 3068 + }, + { + "epoch": 0.07, + "learning_rate": 1.9899580316186493e-05, + "loss": 1.3145, + "step": 3069 + }, + { + "epoch": 0.07, + "learning_rate": 1.989947018041687e-05, + "loss": 1.2274, + "step": 3070 + }, + { + "epoch": 0.07, + "learning_rate": 1.9899359984589488e-05, + "loss": 1.1622, + "step": 3071 + }, + { + "epoch": 0.07, + "learning_rate": 1.9899249728705018e-05, + "loss": 1.3304, + "step": 3072 + }, + { + "epoch": 0.07, + "learning_rate": 1.989913941276413e-05, + "loss": 1.2568, + "step": 3073 + }, + { + "epoch": 0.07, + "learning_rate": 1.989902903676749e-05, + "loss": 1.1625, + "step": 3074 + }, + { + "epoch": 0.07, + "learning_rate": 1.989891860071577e-05, + "loss": 1.2997, + "step": 3075 + }, + { + "epoch": 0.07, + "learning_rate": 1.9898808104609638e-05, + "loss": 1.2471, + "step": 3076 + }, + { + "epoch": 0.07, + "learning_rate": 1.9898697548449766e-05, + "loss": 1.5495, + "step": 3077 + }, + { + "epoch": 0.07, + "learning_rate": 1.9898586932236826e-05, + "loss": 0.9048, + "step": 3078 + }, + { + "epoch": 0.07, + "learning_rate": 1.9898476255971487e-05, + "loss": 1.1858, + "step": 3079 + }, + { + "epoch": 0.07, + "learning_rate": 1.989836551965442e-05, + "loss": 1.1626, + "step": 3080 + }, + { + "epoch": 0.07, + "learning_rate": 1.98982547232863e-05, + "loss": 0.9427, + "step": 3081 + }, + { + "epoch": 0.07, + "learning_rate": 1.9898143866867792e-05, + "loss": 1.0937, + "step": 3082 + }, + { + "epoch": 0.07, + "learning_rate": 1.9898032950399575e-05, + "loss": 0.9847, + "step": 3083 + }, + { + "epoch": 0.07, + "learning_rate": 1.989792197388232e-05, + "loss": 1.1999, + "step": 3084 + }, + { + "epoch": 0.07, + "learning_rate": 1.9897810937316704e-05, + "loss": 1.1976, + "step": 3085 + }, + { + "epoch": 0.07, + "learning_rate": 1.9897699840703393e-05, + "loss": 1.1955, + "step": 3086 + }, + { + "epoch": 0.07, + "learning_rate": 1.9897588684043067e-05, + "loss": 0.9502, + "step": 3087 + }, + { + "epoch": 0.07, + "learning_rate": 1.98974774673364e-05, + "loss": 1.2999, + "step": 3088 + }, + { + "epoch": 0.07, + "learning_rate": 1.989736619058406e-05, + "loss": 1.1409, + "step": 3089 + }, + { + "epoch": 0.07, + "learning_rate": 1.9897254853786735e-05, + "loss": 1.242, + "step": 3090 + }, + { + "epoch": 0.07, + "learning_rate": 1.9897143456945087e-05, + "loss": 1.2099, + "step": 3091 + }, + { + "epoch": 0.07, + "learning_rate": 1.98970320000598e-05, + "loss": 1.1521, + "step": 3092 + }, + { + "epoch": 0.07, + "learning_rate": 1.9896920483131548e-05, + "loss": 1.1172, + "step": 3093 + }, + { + "epoch": 0.07, + "learning_rate": 1.9896808906161005e-05, + "loss": 1.2841, + "step": 3094 + }, + { + "epoch": 0.07, + "learning_rate": 1.9896697269148853e-05, + "loss": 1.1291, + "step": 3095 + }, + { + "epoch": 0.07, + "learning_rate": 1.9896585572095764e-05, + "loss": 1.1688, + "step": 3096 + }, + { + "epoch": 0.07, + "learning_rate": 1.9896473815002418e-05, + "loss": 0.8613, + "step": 3097 + }, + { + "epoch": 0.07, + "learning_rate": 1.9896361997869496e-05, + "loss": 1.0755, + "step": 3098 + }, + { + "epoch": 0.07, + "learning_rate": 1.9896250120697673e-05, + "loss": 1.2069, + "step": 3099 + }, + { + "epoch": 0.07, + "learning_rate": 1.9896138183487626e-05, + "loss": 1.1777, + "step": 3100 + }, + { + "epoch": 0.07, + "learning_rate": 1.989602618624004e-05, + "loss": 1.3963, + "step": 3101 + }, + { + "epoch": 0.07, + "learning_rate": 1.9895914128955588e-05, + "loss": 1.413, + "step": 3102 + }, + { + "epoch": 0.07, + "learning_rate": 1.9895802011634952e-05, + "loss": 1.235, + "step": 3103 + }, + { + "epoch": 0.07, + "learning_rate": 1.9895689834278813e-05, + "loss": 1.0433, + "step": 3104 + }, + { + "epoch": 0.07, + "learning_rate": 1.9895577596887855e-05, + "loss": 1.0205, + "step": 3105 + }, + { + "epoch": 0.07, + "learning_rate": 1.989546529946275e-05, + "loss": 1.0817, + "step": 3106 + }, + { + "epoch": 0.07, + "learning_rate": 1.989535294200419e-05, + "loss": 1.1621, + "step": 3107 + }, + { + "epoch": 0.07, + "learning_rate": 1.9895240524512845e-05, + "loss": 1.1116, + "step": 3108 + }, + { + "epoch": 0.07, + "learning_rate": 1.989512804698941e-05, + "loss": 1.0505, + "step": 3109 + }, + { + "epoch": 0.07, + "learning_rate": 1.9895015509434555e-05, + "loss": 1.2046, + "step": 3110 + }, + { + "epoch": 0.07, + "learning_rate": 1.9894902911848973e-05, + "loss": 1.0524, + "step": 3111 + }, + { + "epoch": 0.07, + "learning_rate": 1.9894790254233338e-05, + "loss": 1.0458, + "step": 3112 + }, + { + "epoch": 0.07, + "learning_rate": 1.989467753658834e-05, + "loss": 1.2267, + "step": 3113 + }, + { + "epoch": 0.07, + "learning_rate": 1.9894564758914662e-05, + "loss": 0.9762, + "step": 3114 + }, + { + "epoch": 0.07, + "learning_rate": 1.9894451921212986e-05, + "loss": 1.0453, + "step": 3115 + }, + { + "epoch": 0.07, + "learning_rate": 1.9894339023484e-05, + "loss": 1.4661, + "step": 3116 + }, + { + "epoch": 0.07, + "learning_rate": 1.989422606572838e-05, + "loss": 1.1839, + "step": 3117 + }, + { + "epoch": 0.07, + "learning_rate": 1.989411304794682e-05, + "loss": 1.2925, + "step": 3118 + }, + { + "epoch": 0.08, + "learning_rate": 1.9893999970140005e-05, + "loss": 1.2263, + "step": 3119 + }, + { + "epoch": 0.08, + "learning_rate": 1.989388683230862e-05, + "loss": 1.0451, + "step": 3120 + }, + { + "epoch": 0.08, + "learning_rate": 1.989377363445335e-05, + "loss": 1.2324, + "step": 3121 + }, + { + "epoch": 0.08, + "learning_rate": 1.9893660376574883e-05, + "loss": 1.2006, + "step": 3122 + }, + { + "epoch": 0.08, + "learning_rate": 1.9893547058673905e-05, + "loss": 1.2412, + "step": 3123 + }, + { + "epoch": 0.08, + "learning_rate": 1.9893433680751105e-05, + "loss": 1.2164, + "step": 3124 + }, + { + "epoch": 0.08, + "learning_rate": 1.9893320242807167e-05, + "loss": 1.3662, + "step": 3125 + }, + { + "epoch": 0.08, + "learning_rate": 1.9893206744842787e-05, + "loss": 1.5143, + "step": 3126 + }, + { + "epoch": 0.08, + "learning_rate": 1.9893093186858645e-05, + "loss": 1.4221, + "step": 3127 + }, + { + "epoch": 0.08, + "learning_rate": 1.9892979568855435e-05, + "loss": 1.0048, + "step": 3128 + }, + { + "epoch": 0.08, + "learning_rate": 1.9892865890833842e-05, + "loss": 1.4308, + "step": 3129 + }, + { + "epoch": 0.08, + "learning_rate": 1.9892752152794565e-05, + "loss": 1.1784, + "step": 3130 + }, + { + "epoch": 0.08, + "learning_rate": 1.989263835473828e-05, + "loss": 1.224, + "step": 3131 + }, + { + "epoch": 0.08, + "learning_rate": 1.9892524496665692e-05, + "loss": 1.1204, + "step": 3132 + }, + { + "epoch": 0.08, + "learning_rate": 1.989241057857748e-05, + "loss": 1.1385, + "step": 3133 + }, + { + "epoch": 0.08, + "learning_rate": 1.989229660047434e-05, + "loss": 1.6018, + "step": 3134 + }, + { + "epoch": 0.08, + "learning_rate": 1.9892182562356964e-05, + "loss": 1.1927, + "step": 3135 + }, + { + "epoch": 0.08, + "learning_rate": 1.9892068464226044e-05, + "loss": 1.4161, + "step": 3136 + }, + { + "epoch": 0.08, + "learning_rate": 1.989195430608227e-05, + "loss": 1.2779, + "step": 3137 + }, + { + "epoch": 0.08, + "learning_rate": 1.989184008792634e-05, + "loss": 1.0287, + "step": 3138 + }, + { + "epoch": 0.08, + "learning_rate": 1.989172580975894e-05, + "loss": 1.2094, + "step": 3139 + }, + { + "epoch": 0.08, + "learning_rate": 1.9891611471580767e-05, + "loss": 1.1819, + "step": 3140 + }, + { + "epoch": 0.08, + "learning_rate": 1.989149707339251e-05, + "loss": 0.8319, + "step": 3141 + }, + { + "epoch": 0.08, + "learning_rate": 1.989138261519487e-05, + "loss": 1.2724, + "step": 3142 + }, + { + "epoch": 0.08, + "learning_rate": 1.989126809698854e-05, + "loss": 1.2869, + "step": 3143 + }, + { + "epoch": 0.08, + "learning_rate": 1.989115351877421e-05, + "loss": 0.9372, + "step": 3144 + }, + { + "epoch": 0.08, + "learning_rate": 1.989103888055258e-05, + "loss": 1.2519, + "step": 3145 + }, + { + "epoch": 0.08, + "learning_rate": 1.9890924182324345e-05, + "loss": 1.1716, + "step": 3146 + }, + { + "epoch": 0.08, + "learning_rate": 1.9890809424090197e-05, + "loss": 1.218, + "step": 3147 + }, + { + "epoch": 0.08, + "learning_rate": 1.989069460585083e-05, + "loss": 1.145, + "step": 3148 + }, + { + "epoch": 0.08, + "learning_rate": 1.9890579727606955e-05, + "loss": 0.9387, + "step": 3149 + }, + { + "epoch": 0.08, + "learning_rate": 1.9890464789359253e-05, + "loss": 1.212, + "step": 3150 + }, + { + "epoch": 0.08, + "learning_rate": 1.989034979110843e-05, + "loss": 1.1072, + "step": 3151 + }, + { + "epoch": 0.08, + "learning_rate": 1.989023473285518e-05, + "loss": 1.3224, + "step": 3152 + }, + { + "epoch": 0.08, + "learning_rate": 1.98901196146002e-05, + "loss": 1.093, + "step": 3153 + }, + { + "epoch": 0.08, + "learning_rate": 1.9890004436344197e-05, + "loss": 1.0866, + "step": 3154 + }, + { + "epoch": 0.08, + "learning_rate": 1.9889889198087858e-05, + "loss": 1.3047, + "step": 3155 + }, + { + "epoch": 0.08, + "learning_rate": 1.988977389983189e-05, + "loss": 1.1151, + "step": 3156 + }, + { + "epoch": 0.08, + "learning_rate": 1.988965854157699e-05, + "loss": 1.1215, + "step": 3157 + }, + { + "epoch": 0.08, + "learning_rate": 1.9889543123323854e-05, + "loss": 1.0786, + "step": 3158 + }, + { + "epoch": 0.08, + "learning_rate": 1.988942764507319e-05, + "loss": 1.1492, + "step": 3159 + }, + { + "epoch": 0.08, + "learning_rate": 1.9889312106825694e-05, + "loss": 0.9301, + "step": 3160 + }, + { + "epoch": 0.08, + "learning_rate": 1.9889196508582066e-05, + "loss": 1.148, + "step": 3161 + }, + { + "epoch": 0.08, + "learning_rate": 1.988908085034301e-05, + "loss": 1.1464, + "step": 3162 + }, + { + "epoch": 0.08, + "learning_rate": 1.9888965132109228e-05, + "loss": 0.824, + "step": 3163 + }, + { + "epoch": 0.08, + "learning_rate": 1.988884935388142e-05, + "loss": 1.3482, + "step": 3164 + }, + { + "epoch": 0.08, + "learning_rate": 1.9888733515660287e-05, + "loss": 1.0842, + "step": 3165 + }, + { + "epoch": 0.08, + "learning_rate": 1.988861761744653e-05, + "loss": 1.2779, + "step": 3166 + }, + { + "epoch": 0.08, + "learning_rate": 1.9888501659240864e-05, + "loss": 1.0025, + "step": 3167 + }, + { + "epoch": 0.08, + "learning_rate": 1.988838564104398e-05, + "loss": 1.1954, + "step": 3168 + }, + { + "epoch": 0.08, + "learning_rate": 1.9888269562856588e-05, + "loss": 1.0079, + "step": 3169 + }, + { + "epoch": 0.08, + "learning_rate": 1.9888153424679387e-05, + "loss": 1.1714, + "step": 3170 + }, + { + "epoch": 0.08, + "learning_rate": 1.988803722651309e-05, + "loss": 1.2235, + "step": 3171 + }, + { + "epoch": 0.08, + "learning_rate": 1.9887920968358394e-05, + "loss": 1.3727, + "step": 3172 + }, + { + "epoch": 0.08, + "learning_rate": 1.9887804650216006e-05, + "loss": 1.0542, + "step": 3173 + }, + { + "epoch": 0.08, + "learning_rate": 1.9887688272086637e-05, + "loss": 1.3229, + "step": 3174 + }, + { + "epoch": 0.08, + "learning_rate": 1.9887571833970988e-05, + "loss": 1.2159, + "step": 3175 + }, + { + "epoch": 0.08, + "learning_rate": 1.9887455335869762e-05, + "loss": 1.1229, + "step": 3176 + }, + { + "epoch": 0.08, + "learning_rate": 1.9887338777783673e-05, + "loss": 1.3377, + "step": 3177 + }, + { + "epoch": 0.08, + "learning_rate": 1.9887222159713427e-05, + "loss": 1.1271, + "step": 3178 + }, + { + "epoch": 0.08, + "learning_rate": 1.9887105481659727e-05, + "loss": 1.3531, + "step": 3179 + }, + { + "epoch": 0.08, + "learning_rate": 1.9886988743623284e-05, + "loss": 1.2278, + "step": 3180 + }, + { + "epoch": 0.08, + "learning_rate": 1.988687194560481e-05, + "loss": 1.0959, + "step": 3181 + }, + { + "epoch": 0.08, + "learning_rate": 1.9886755087605004e-05, + "loss": 1.0827, + "step": 3182 + }, + { + "epoch": 0.08, + "learning_rate": 1.9886638169624583e-05, + "loss": 1.068, + "step": 3183 + }, + { + "epoch": 0.08, + "learning_rate": 1.9886521191664255e-05, + "loss": 1.0142, + "step": 3184 + }, + { + "epoch": 0.08, + "learning_rate": 1.9886404153724724e-05, + "loss": 1.2728, + "step": 3185 + }, + { + "epoch": 0.08, + "learning_rate": 1.988628705580671e-05, + "loss": 0.9313, + "step": 3186 + }, + { + "epoch": 0.08, + "learning_rate": 1.988616989791091e-05, + "loss": 1.1341, + "step": 3187 + }, + { + "epoch": 0.08, + "learning_rate": 1.9886052680038048e-05, + "loss": 1.4172, + "step": 3188 + }, + { + "epoch": 0.08, + "learning_rate": 1.988593540218883e-05, + "loss": 1.1241, + "step": 3189 + }, + { + "epoch": 0.08, + "learning_rate": 1.9885818064363968e-05, + "loss": 1.2129, + "step": 3190 + }, + { + "epoch": 0.08, + "learning_rate": 1.988570066656417e-05, + "loss": 0.9856, + "step": 3191 + }, + { + "epoch": 0.08, + "learning_rate": 1.9885583208790154e-05, + "loss": 1.1562, + "step": 3192 + }, + { + "epoch": 0.08, + "learning_rate": 1.988546569104263e-05, + "loss": 1.2175, + "step": 3193 + }, + { + "epoch": 0.08, + "learning_rate": 1.988534811332231e-05, + "loss": 1.0892, + "step": 3194 + }, + { + "epoch": 0.08, + "learning_rate": 1.9885230475629904e-05, + "loss": 1.2735, + "step": 3195 + }, + { + "epoch": 0.08, + "learning_rate": 1.9885112777966135e-05, + "loss": 1.2398, + "step": 3196 + }, + { + "epoch": 0.08, + "learning_rate": 1.988499502033171e-05, + "loss": 1.2153, + "step": 3197 + }, + { + "epoch": 0.08, + "learning_rate": 1.9884877202727345e-05, + "loss": 1.1135, + "step": 3198 + }, + { + "epoch": 0.08, + "learning_rate": 1.9884759325153757e-05, + "loss": 1.11, + "step": 3199 + }, + { + "epoch": 0.08, + "learning_rate": 1.988464138761166e-05, + "loss": 1.3269, + "step": 3200 + }, + { + "epoch": 0.08, + "learning_rate": 1.9884523390101768e-05, + "loss": 1.3444, + "step": 3201 + }, + { + "epoch": 0.08, + "learning_rate": 1.9884405332624793e-05, + "loss": 1.0808, + "step": 3202 + }, + { + "epoch": 0.08, + "learning_rate": 1.988428721518146e-05, + "loss": 1.1982, + "step": 3203 + }, + { + "epoch": 0.08, + "learning_rate": 1.988416903777248e-05, + "loss": 1.0242, + "step": 3204 + }, + { + "epoch": 0.08, + "learning_rate": 1.9884050800398574e-05, + "loss": 0.9854, + "step": 3205 + }, + { + "epoch": 0.08, + "learning_rate": 1.9883932503060452e-05, + "loss": 1.085, + "step": 3206 + }, + { + "epoch": 0.08, + "learning_rate": 1.9883814145758838e-05, + "loss": 1.0078, + "step": 3207 + }, + { + "epoch": 0.08, + "learning_rate": 1.988369572849445e-05, + "loss": 1.2107, + "step": 3208 + }, + { + "epoch": 0.08, + "learning_rate": 1.9883577251268005e-05, + "loss": 1.2854, + "step": 3209 + }, + { + "epoch": 0.08, + "learning_rate": 1.9883458714080222e-05, + "loss": 1.0601, + "step": 3210 + }, + { + "epoch": 0.08, + "learning_rate": 1.9883340116931817e-05, + "loss": 1.0375, + "step": 3211 + }, + { + "epoch": 0.08, + "learning_rate": 1.9883221459823515e-05, + "loss": 1.2446, + "step": 3212 + }, + { + "epoch": 0.08, + "learning_rate": 1.988310274275603e-05, + "loss": 1.089, + "step": 3213 + }, + { + "epoch": 0.08, + "learning_rate": 1.9882983965730086e-05, + "loss": 1.2216, + "step": 3214 + }, + { + "epoch": 0.08, + "learning_rate": 1.9882865128746406e-05, + "loss": 1.1938, + "step": 3215 + }, + { + "epoch": 0.08, + "learning_rate": 1.9882746231805705e-05, + "loss": 1.433, + "step": 3216 + }, + { + "epoch": 0.08, + "learning_rate": 1.988262727490871e-05, + "loss": 0.9577, + "step": 3217 + }, + { + "epoch": 0.08, + "learning_rate": 1.9882508258056136e-05, + "loss": 1.3565, + "step": 3218 + }, + { + "epoch": 0.08, + "learning_rate": 1.9882389181248708e-05, + "loss": 1.2975, + "step": 3219 + }, + { + "epoch": 0.08, + "learning_rate": 1.9882270044487155e-05, + "loss": 1.1952, + "step": 3220 + }, + { + "epoch": 0.08, + "learning_rate": 1.988215084777219e-05, + "loss": 1.2495, + "step": 3221 + }, + { + "epoch": 0.08, + "learning_rate": 1.9882031591104543e-05, + "loss": 1.2027, + "step": 3222 + }, + { + "epoch": 0.08, + "learning_rate": 1.9881912274484933e-05, + "loss": 1.1926, + "step": 3223 + }, + { + "epoch": 0.08, + "learning_rate": 1.9881792897914086e-05, + "loss": 1.283, + "step": 3224 + }, + { + "epoch": 0.08, + "learning_rate": 1.9881673461392723e-05, + "loss": 1.3044, + "step": 3225 + }, + { + "epoch": 0.08, + "learning_rate": 1.9881553964921574e-05, + "loss": 1.2667, + "step": 3226 + }, + { + "epoch": 0.08, + "learning_rate": 1.988143440850136e-05, + "loss": 1.1065, + "step": 3227 + }, + { + "epoch": 0.08, + "learning_rate": 1.9881314792132812e-05, + "loss": 1.1136, + "step": 3228 + }, + { + "epoch": 0.08, + "learning_rate": 1.9881195115816646e-05, + "loss": 1.3079, + "step": 3229 + }, + { + "epoch": 0.08, + "learning_rate": 1.9881075379553597e-05, + "loss": 1.0125, + "step": 3230 + }, + { + "epoch": 0.08, + "learning_rate": 1.988095558334439e-05, + "loss": 0.8832, + "step": 3231 + }, + { + "epoch": 0.08, + "learning_rate": 1.9880835727189742e-05, + "loss": 1.1678, + "step": 3232 + }, + { + "epoch": 0.08, + "learning_rate": 1.9880715811090394e-05, + "loss": 0.8897, + "step": 3233 + }, + { + "epoch": 0.08, + "learning_rate": 1.9880595835047062e-05, + "loss": 1.1114, + "step": 3234 + }, + { + "epoch": 0.08, + "learning_rate": 1.9880475799060486e-05, + "loss": 0.8719, + "step": 3235 + }, + { + "epoch": 0.08, + "learning_rate": 1.988035570313138e-05, + "loss": 1.1943, + "step": 3236 + }, + { + "epoch": 0.08, + "learning_rate": 1.9880235547260488e-05, + "loss": 1.2603, + "step": 3237 + }, + { + "epoch": 0.08, + "learning_rate": 1.9880115331448526e-05, + "loss": 1.0409, + "step": 3238 + }, + { + "epoch": 0.08, + "learning_rate": 1.9879995055696227e-05, + "loss": 1.2205, + "step": 3239 + }, + { + "epoch": 0.08, + "learning_rate": 1.9879874720004326e-05, + "loss": 1.2698, + "step": 3240 + }, + { + "epoch": 0.08, + "learning_rate": 1.9879754324373547e-05, + "loss": 1.3249, + "step": 3241 + }, + { + "epoch": 0.08, + "learning_rate": 1.9879633868804624e-05, + "loss": 1.321, + "step": 3242 + }, + { + "epoch": 0.08, + "learning_rate": 1.987951335329828e-05, + "loss": 1.3202, + "step": 3243 + }, + { + "epoch": 0.08, + "learning_rate": 1.9879392777855258e-05, + "loss": 1.1153, + "step": 3244 + }, + { + "epoch": 0.08, + "learning_rate": 1.9879272142476283e-05, + "loss": 1.2365, + "step": 3245 + }, + { + "epoch": 0.08, + "learning_rate": 1.9879151447162086e-05, + "loss": 1.0028, + "step": 3246 + }, + { + "epoch": 0.08, + "learning_rate": 1.98790306919134e-05, + "loss": 1.1679, + "step": 3247 + }, + { + "epoch": 0.08, + "learning_rate": 1.987890987673096e-05, + "loss": 1.2592, + "step": 3248 + }, + { + "epoch": 0.08, + "learning_rate": 1.98787890016155e-05, + "loss": 0.9814, + "step": 3249 + }, + { + "epoch": 0.08, + "learning_rate": 1.987866806656775e-05, + "loss": 1.1682, + "step": 3250 + }, + { + "epoch": 0.08, + "learning_rate": 1.987854707158844e-05, + "loss": 1.1986, + "step": 3251 + }, + { + "epoch": 0.08, + "learning_rate": 1.9878426016678313e-05, + "loss": 0.8789, + "step": 3252 + }, + { + "epoch": 0.08, + "learning_rate": 1.9878304901838097e-05, + "loss": 1.3907, + "step": 3253 + }, + { + "epoch": 0.08, + "learning_rate": 1.987818372706853e-05, + "loss": 1.3054, + "step": 3254 + }, + { + "epoch": 0.08, + "learning_rate": 1.9878062492370343e-05, + "loss": 0.9645, + "step": 3255 + }, + { + "epoch": 0.08, + "learning_rate": 1.9877941197744277e-05, + "loss": 1.1877, + "step": 3256 + }, + { + "epoch": 0.08, + "learning_rate": 1.9877819843191063e-05, + "loss": 1.2963, + "step": 3257 + }, + { + "epoch": 0.08, + "learning_rate": 1.9877698428711444e-05, + "loss": 1.013, + "step": 3258 + }, + { + "epoch": 0.08, + "learning_rate": 1.987757695430615e-05, + "loss": 1.1984, + "step": 3259 + }, + { + "epoch": 0.08, + "learning_rate": 1.9877455419975917e-05, + "loss": 1.2038, + "step": 3260 + }, + { + "epoch": 0.08, + "learning_rate": 1.9877333825721487e-05, + "loss": 1.3986, + "step": 3261 + }, + { + "epoch": 0.08, + "learning_rate": 1.9877212171543595e-05, + "loss": 0.8799, + "step": 3262 + }, + { + "epoch": 0.08, + "learning_rate": 1.987709045744298e-05, + "loss": 1.2828, + "step": 3263 + }, + { + "epoch": 0.08, + "learning_rate": 1.9876968683420384e-05, + "loss": 1.0622, + "step": 3264 + }, + { + "epoch": 0.08, + "learning_rate": 1.9876846849476537e-05, + "loss": 1.2284, + "step": 3265 + }, + { + "epoch": 0.08, + "learning_rate": 1.9876724955612188e-05, + "loss": 1.3659, + "step": 3266 + }, + { + "epoch": 0.08, + "learning_rate": 1.987660300182807e-05, + "loss": 1.1963, + "step": 3267 + }, + { + "epoch": 0.08, + "learning_rate": 1.9876480988124923e-05, + "loss": 1.0661, + "step": 3268 + }, + { + "epoch": 0.08, + "learning_rate": 1.9876358914503492e-05, + "loss": 0.9839, + "step": 3269 + }, + { + "epoch": 0.08, + "learning_rate": 1.9876236780964513e-05, + "loss": 1.0327, + "step": 3270 + }, + { + "epoch": 0.08, + "learning_rate": 1.987611458750873e-05, + "loss": 1.1827, + "step": 3271 + }, + { + "epoch": 0.08, + "learning_rate": 1.9875992334136878e-05, + "loss": 1.0274, + "step": 3272 + }, + { + "epoch": 0.08, + "learning_rate": 1.987587002084971e-05, + "loss": 1.0616, + "step": 3273 + }, + { + "epoch": 0.08, + "learning_rate": 1.9875747647647956e-05, + "loss": 1.3907, + "step": 3274 + }, + { + "epoch": 0.08, + "learning_rate": 1.987562521453237e-05, + "loss": 1.1253, + "step": 3275 + }, + { + "epoch": 0.08, + "learning_rate": 1.9875502721503683e-05, + "loss": 1.0695, + "step": 3276 + }, + { + "epoch": 0.08, + "learning_rate": 1.9875380168562645e-05, + "loss": 1.1801, + "step": 3277 + }, + { + "epoch": 0.08, + "learning_rate": 1.987525755571e-05, + "loss": 1.059, + "step": 3278 + }, + { + "epoch": 0.08, + "learning_rate": 1.987513488294649e-05, + "loss": 1.1104, + "step": 3279 + }, + { + "epoch": 0.08, + "learning_rate": 1.987501215027286e-05, + "loss": 1.4503, + "step": 3280 + }, + { + "epoch": 0.08, + "learning_rate": 1.9874889357689852e-05, + "loss": 1.2838, + "step": 3281 + }, + { + "epoch": 0.08, + "learning_rate": 1.9874766505198214e-05, + "loss": 1.3008, + "step": 3282 + }, + { + "epoch": 0.08, + "learning_rate": 1.9874643592798687e-05, + "loss": 1.2201, + "step": 3283 + }, + { + "epoch": 0.08, + "learning_rate": 1.9874520620492026e-05, + "loss": 0.9697, + "step": 3284 + }, + { + "epoch": 0.08, + "learning_rate": 1.987439758827897e-05, + "loss": 1.3372, + "step": 3285 + }, + { + "epoch": 0.08, + "learning_rate": 1.987427449616026e-05, + "loss": 1.3062, + "step": 3286 + }, + { + "epoch": 0.08, + "learning_rate": 1.9874151344136655e-05, + "loss": 0.9201, + "step": 3287 + }, + { + "epoch": 0.08, + "learning_rate": 1.9874028132208897e-05, + "loss": 1.1901, + "step": 3288 + }, + { + "epoch": 0.08, + "learning_rate": 1.9873904860377728e-05, + "loss": 1.1343, + "step": 3289 + }, + { + "epoch": 0.08, + "learning_rate": 1.9873781528643905e-05, + "loss": 1.2926, + "step": 3290 + }, + { + "epoch": 0.08, + "learning_rate": 1.9873658137008173e-05, + "loss": 1.3906, + "step": 3291 + }, + { + "epoch": 0.08, + "learning_rate": 1.9873534685471277e-05, + "loss": 1.3955, + "step": 3292 + }, + { + "epoch": 0.08, + "learning_rate": 1.9873411174033966e-05, + "loss": 1.0885, + "step": 3293 + }, + { + "epoch": 0.08, + "learning_rate": 1.9873287602696996e-05, + "loss": 1.409, + "step": 3294 + }, + { + "epoch": 0.08, + "learning_rate": 1.9873163971461108e-05, + "loss": 1.1119, + "step": 3295 + }, + { + "epoch": 0.08, + "learning_rate": 1.9873040280327062e-05, + "loss": 1.185, + "step": 3296 + }, + { + "epoch": 0.08, + "learning_rate": 1.98729165292956e-05, + "loss": 1.1023, + "step": 3297 + }, + { + "epoch": 0.08, + "learning_rate": 1.9872792718367476e-05, + "loss": 1.4462, + "step": 3298 + }, + { + "epoch": 0.08, + "learning_rate": 1.987266884754344e-05, + "loss": 1.2229, + "step": 3299 + }, + { + "epoch": 0.08, + "learning_rate": 1.9872544916824244e-05, + "loss": 1.1859, + "step": 3300 + }, + { + "epoch": 0.08, + "learning_rate": 1.987242092621064e-05, + "loss": 1.085, + "step": 3301 + }, + { + "epoch": 0.08, + "learning_rate": 1.987229687570338e-05, + "loss": 1.0637, + "step": 3302 + }, + { + "epoch": 0.08, + "learning_rate": 1.9872172765303222e-05, + "loss": 1.0233, + "step": 3303 + }, + { + "epoch": 0.08, + "learning_rate": 1.987204859501091e-05, + "loss": 1.2448, + "step": 3304 + }, + { + "epoch": 0.08, + "learning_rate": 1.98719243648272e-05, + "loss": 1.3138, + "step": 3305 + }, + { + "epoch": 0.08, + "learning_rate": 1.9871800074752848e-05, + "loss": 1.2717, + "step": 3306 + }, + { + "epoch": 0.08, + "learning_rate": 1.987167572478861e-05, + "loss": 1.1972, + "step": 3307 + }, + { + "epoch": 0.08, + "learning_rate": 1.9871551314935233e-05, + "loss": 1.2151, + "step": 3308 + }, + { + "epoch": 0.08, + "learning_rate": 1.9871426845193478e-05, + "loss": 1.2777, + "step": 3309 + }, + { + "epoch": 0.08, + "learning_rate": 1.98713023155641e-05, + "loss": 1.0068, + "step": 3310 + }, + { + "epoch": 0.08, + "learning_rate": 1.987117772604785e-05, + "loss": 1.3641, + "step": 3311 + }, + { + "epoch": 0.08, + "learning_rate": 1.987105307664549e-05, + "loss": 1.0964, + "step": 3312 + }, + { + "epoch": 0.08, + "learning_rate": 1.987092836735777e-05, + "loss": 1.3739, + "step": 3313 + }, + { + "epoch": 0.08, + "learning_rate": 1.9870803598185446e-05, + "loss": 1.1421, + "step": 3314 + }, + { + "epoch": 0.08, + "learning_rate": 1.9870678769129283e-05, + "loss": 1.3599, + "step": 3315 + }, + { + "epoch": 0.08, + "learning_rate": 1.9870553880190032e-05, + "loss": 1.3521, + "step": 3316 + }, + { + "epoch": 0.08, + "learning_rate": 1.9870428931368452e-05, + "loss": 1.2717, + "step": 3317 + }, + { + "epoch": 0.08, + "learning_rate": 1.9870303922665305e-05, + "loss": 1.0355, + "step": 3318 + }, + { + "epoch": 0.08, + "learning_rate": 1.987017885408134e-05, + "loss": 1.1133, + "step": 3319 + }, + { + "epoch": 0.08, + "learning_rate": 1.9870053725617326e-05, + "loss": 1.0002, + "step": 3320 + }, + { + "epoch": 0.08, + "learning_rate": 1.9869928537274016e-05, + "loss": 1.1105, + "step": 3321 + }, + { + "epoch": 0.08, + "learning_rate": 1.986980328905217e-05, + "loss": 1.4315, + "step": 3322 + }, + { + "epoch": 0.08, + "learning_rate": 1.986967798095255e-05, + "loss": 1.2098, + "step": 3323 + }, + { + "epoch": 0.08, + "learning_rate": 1.9869552612975917e-05, + "loss": 1.003, + "step": 3324 + }, + { + "epoch": 0.08, + "learning_rate": 1.9869427185123025e-05, + "loss": 1.2692, + "step": 3325 + }, + { + "epoch": 0.08, + "learning_rate": 1.9869301697394646e-05, + "loss": 1.1625, + "step": 3326 + }, + { + "epoch": 0.08, + "learning_rate": 1.9869176149791533e-05, + "loss": 1.1236, + "step": 3327 + }, + { + "epoch": 0.08, + "learning_rate": 1.9869050542314446e-05, + "loss": 1.3357, + "step": 3328 + }, + { + "epoch": 0.08, + "learning_rate": 1.9868924874964155e-05, + "loss": 0.9929, + "step": 3329 + }, + { + "epoch": 0.08, + "learning_rate": 1.9868799147741417e-05, + "loss": 1.2913, + "step": 3330 + }, + { + "epoch": 0.08, + "learning_rate": 1.9868673360646993e-05, + "loss": 1.1364, + "step": 3331 + }, + { + "epoch": 0.08, + "learning_rate": 1.986854751368165e-05, + "loss": 1.271, + "step": 3332 + }, + { + "epoch": 0.08, + "learning_rate": 1.9868421606846152e-05, + "loss": 1.0067, + "step": 3333 + }, + { + "epoch": 0.08, + "learning_rate": 1.986829564014126e-05, + "loss": 1.2354, + "step": 3334 + }, + { + "epoch": 0.08, + "learning_rate": 1.9868169613567742e-05, + "loss": 1.1528, + "step": 3335 + }, + { + "epoch": 0.08, + "learning_rate": 1.9868043527126358e-05, + "loss": 1.2872, + "step": 3336 + }, + { + "epoch": 0.08, + "learning_rate": 1.9867917380817875e-05, + "loss": 1.3183, + "step": 3337 + }, + { + "epoch": 0.08, + "learning_rate": 1.9867791174643057e-05, + "loss": 1.3589, + "step": 3338 + }, + { + "epoch": 0.08, + "learning_rate": 1.9867664908602676e-05, + "loss": 1.2338, + "step": 3339 + }, + { + "epoch": 0.08, + "learning_rate": 1.986753858269749e-05, + "loss": 1.0776, + "step": 3340 + }, + { + "epoch": 0.08, + "learning_rate": 1.9867412196928267e-05, + "loss": 1.1487, + "step": 3341 + }, + { + "epoch": 0.08, + "learning_rate": 1.986728575129578e-05, + "loss": 1.0335, + "step": 3342 + }, + { + "epoch": 0.08, + "learning_rate": 1.9867159245800785e-05, + "loss": 1.0647, + "step": 3343 + }, + { + "epoch": 0.08, + "learning_rate": 1.986703268044406e-05, + "loss": 1.1605, + "step": 3344 + }, + { + "epoch": 0.08, + "learning_rate": 1.9866906055226366e-05, + "loss": 1.2697, + "step": 3345 + }, + { + "epoch": 0.08, + "learning_rate": 1.9866779370148475e-05, + "loss": 1.3178, + "step": 3346 + }, + { + "epoch": 0.08, + "learning_rate": 1.9866652625211152e-05, + "loss": 1.2985, + "step": 3347 + }, + { + "epoch": 0.08, + "learning_rate": 1.9866525820415173e-05, + "loss": 1.4681, + "step": 3348 + }, + { + "epoch": 0.08, + "learning_rate": 1.9866398955761296e-05, + "loss": 1.1214, + "step": 3349 + }, + { + "epoch": 0.08, + "learning_rate": 1.98662720312503e-05, + "loss": 1.3029, + "step": 3350 + }, + { + "epoch": 0.08, + "learning_rate": 1.9866145046882958e-05, + "loss": 1.1077, + "step": 3351 + }, + { + "epoch": 0.08, + "learning_rate": 1.9866018002660027e-05, + "loss": 1.1612, + "step": 3352 + }, + { + "epoch": 0.08, + "learning_rate": 1.986589089858229e-05, + "loss": 1.2481, + "step": 3353 + }, + { + "epoch": 0.08, + "learning_rate": 1.9865763734650514e-05, + "loss": 1.2118, + "step": 3354 + }, + { + "epoch": 0.08, + "learning_rate": 1.9865636510865466e-05, + "loss": 1.4908, + "step": 3355 + }, + { + "epoch": 0.08, + "learning_rate": 1.9865509227227924e-05, + "loss": 1.1226, + "step": 3356 + }, + { + "epoch": 0.08, + "learning_rate": 1.986538188373866e-05, + "loss": 1.382, + "step": 3357 + }, + { + "epoch": 0.08, + "learning_rate": 1.986525448039844e-05, + "loss": 1.0121, + "step": 3358 + }, + { + "epoch": 0.08, + "learning_rate": 1.9865127017208045e-05, + "loss": 1.0778, + "step": 3359 + }, + { + "epoch": 0.08, + "learning_rate": 1.9864999494168245e-05, + "loss": 0.8284, + "step": 3360 + }, + { + "epoch": 0.08, + "learning_rate": 1.986487191127981e-05, + "loss": 1.0018, + "step": 3361 + }, + { + "epoch": 0.08, + "learning_rate": 1.9864744268543522e-05, + "loss": 1.2122, + "step": 3362 + }, + { + "epoch": 0.08, + "learning_rate": 1.9864616565960148e-05, + "loss": 1.1531, + "step": 3363 + }, + { + "epoch": 0.08, + "learning_rate": 1.9864488803530467e-05, + "loss": 1.1064, + "step": 3364 + }, + { + "epoch": 0.08, + "learning_rate": 1.986436098125525e-05, + "loss": 0.9668, + "step": 3365 + }, + { + "epoch": 0.08, + "learning_rate": 1.9864233099135278e-05, + "loss": 1.0782, + "step": 3366 + }, + { + "epoch": 0.08, + "learning_rate": 1.9864105157171326e-05, + "loss": 1.1172, + "step": 3367 + }, + { + "epoch": 0.08, + "learning_rate": 1.9863977155364164e-05, + "loss": 1.1502, + "step": 3368 + }, + { + "epoch": 0.08, + "learning_rate": 1.9863849093714572e-05, + "loss": 1.3407, + "step": 3369 + }, + { + "epoch": 0.08, + "learning_rate": 1.986372097222333e-05, + "loss": 1.204, + "step": 3370 + }, + { + "epoch": 0.08, + "learning_rate": 1.9863592790891216e-05, + "loss": 1.2942, + "step": 3371 + }, + { + "epoch": 0.08, + "learning_rate": 1.9863464549719e-05, + "loss": 1.1286, + "step": 3372 + }, + { + "epoch": 0.08, + "learning_rate": 1.9863336248707466e-05, + "loss": 1.1404, + "step": 3373 + }, + { + "epoch": 0.08, + "learning_rate": 1.986320788785739e-05, + "loss": 1.0506, + "step": 3374 + }, + { + "epoch": 0.08, + "learning_rate": 1.9863079467169552e-05, + "loss": 1.4086, + "step": 3375 + }, + { + "epoch": 0.08, + "learning_rate": 1.9862950986644732e-05, + "loss": 1.118, + "step": 3376 + }, + { + "epoch": 0.08, + "learning_rate": 1.986282244628371e-05, + "loss": 1.4782, + "step": 3377 + }, + { + "epoch": 0.08, + "learning_rate": 1.9862693846087258e-05, + "loss": 1.3111, + "step": 3378 + }, + { + "epoch": 0.08, + "learning_rate": 1.986256518605617e-05, + "loss": 1.0878, + "step": 3379 + }, + { + "epoch": 0.08, + "learning_rate": 1.986243646619121e-05, + "loss": 1.1166, + "step": 3380 + }, + { + "epoch": 0.08, + "learning_rate": 1.9862307686493177e-05, + "loss": 1.1733, + "step": 3381 + }, + { + "epoch": 0.08, + "learning_rate": 1.9862178846962837e-05, + "loss": 1.1557, + "step": 3382 + }, + { + "epoch": 0.08, + "learning_rate": 1.9862049947600977e-05, + "loss": 1.0683, + "step": 3383 + }, + { + "epoch": 0.08, + "learning_rate": 1.9861920988408382e-05, + "loss": 1.315, + "step": 3384 + }, + { + "epoch": 0.08, + "learning_rate": 1.9861791969385832e-05, + "loss": 0.8991, + "step": 3385 + }, + { + "epoch": 0.08, + "learning_rate": 1.986166289053411e-05, + "loss": 1.2394, + "step": 3386 + }, + { + "epoch": 0.08, + "learning_rate": 1.9861533751854e-05, + "loss": 1.0499, + "step": 3387 + }, + { + "epoch": 0.08, + "learning_rate": 1.9861404553346282e-05, + "loss": 1.2113, + "step": 3388 + }, + { + "epoch": 0.08, + "learning_rate": 1.9861275295011744e-05, + "loss": 1.2319, + "step": 3389 + }, + { + "epoch": 0.08, + "learning_rate": 1.9861145976851167e-05, + "loss": 1.1193, + "step": 3390 + }, + { + "epoch": 0.08, + "learning_rate": 1.9861016598865337e-05, + "loss": 1.3588, + "step": 3391 + }, + { + "epoch": 0.08, + "learning_rate": 1.986088716105504e-05, + "loss": 1.3342, + "step": 3392 + }, + { + "epoch": 0.08, + "learning_rate": 1.986075766342106e-05, + "loss": 1.021, + "step": 3393 + }, + { + "epoch": 0.08, + "learning_rate": 1.986062810596418e-05, + "loss": 1.1498, + "step": 3394 + }, + { + "epoch": 0.08, + "learning_rate": 1.986049848868519e-05, + "loss": 1.1668, + "step": 3395 + }, + { + "epoch": 0.08, + "learning_rate": 1.9860368811584875e-05, + "loss": 1.1701, + "step": 3396 + }, + { + "epoch": 0.08, + "learning_rate": 1.9860239074664023e-05, + "loss": 0.888, + "step": 3397 + }, + { + "epoch": 0.08, + "learning_rate": 1.9860109277923417e-05, + "loss": 1.0595, + "step": 3398 + }, + { + "epoch": 0.08, + "learning_rate": 1.985997942136385e-05, + "loss": 1.2411, + "step": 3399 + }, + { + "epoch": 0.08, + "learning_rate": 1.9859849504986105e-05, + "loss": 1.1705, + "step": 3400 + }, + { + "epoch": 0.08, + "learning_rate": 1.9859719528790974e-05, + "loss": 1.1833, + "step": 3401 + }, + { + "epoch": 0.08, + "learning_rate": 1.9859589492779244e-05, + "loss": 1.235, + "step": 3402 + }, + { + "epoch": 0.08, + "learning_rate": 1.9859459396951704e-05, + "loss": 1.0072, + "step": 3403 + }, + { + "epoch": 0.08, + "learning_rate": 1.985932924130914e-05, + "loss": 0.8306, + "step": 3404 + }, + { + "epoch": 0.08, + "learning_rate": 1.985919902585235e-05, + "loss": 0.8915, + "step": 3405 + }, + { + "epoch": 0.08, + "learning_rate": 1.9859068750582112e-05, + "loss": 1.1383, + "step": 3406 + }, + { + "epoch": 0.08, + "learning_rate": 1.9858938415499223e-05, + "loss": 1.3928, + "step": 3407 + }, + { + "epoch": 0.08, + "learning_rate": 1.985880802060448e-05, + "loss": 1.329, + "step": 3408 + }, + { + "epoch": 0.08, + "learning_rate": 1.9858677565898664e-05, + "loss": 1.1778, + "step": 3409 + }, + { + "epoch": 0.08, + "learning_rate": 1.9858547051382565e-05, + "loss": 1.3325, + "step": 3410 + }, + { + "epoch": 0.08, + "learning_rate": 1.985841647705699e-05, + "loss": 1.0085, + "step": 3411 + }, + { + "epoch": 0.08, + "learning_rate": 1.9858285842922715e-05, + "loss": 1.0922, + "step": 3412 + }, + { + "epoch": 0.08, + "learning_rate": 1.9858155148980538e-05, + "loss": 1.1746, + "step": 3413 + }, + { + "epoch": 0.08, + "learning_rate": 1.9858024395231256e-05, + "loss": 1.0548, + "step": 3414 + }, + { + "epoch": 0.08, + "learning_rate": 1.9857893581675657e-05, + "loss": 0.8241, + "step": 3415 + }, + { + "epoch": 0.08, + "learning_rate": 1.9857762708314535e-05, + "loss": 1.4037, + "step": 3416 + }, + { + "epoch": 0.08, + "learning_rate": 1.9857631775148686e-05, + "loss": 0.9721, + "step": 3417 + }, + { + "epoch": 0.08, + "learning_rate": 1.9857500782178905e-05, + "loss": 0.9983, + "step": 3418 + }, + { + "epoch": 0.08, + "learning_rate": 1.9857369729405982e-05, + "loss": 1.1956, + "step": 3419 + }, + { + "epoch": 0.08, + "learning_rate": 1.9857238616830718e-05, + "loss": 1.3582, + "step": 3420 + }, + { + "epoch": 0.08, + "learning_rate": 1.9857107444453907e-05, + "loss": 1.2405, + "step": 3421 + }, + { + "epoch": 0.08, + "learning_rate": 1.9856976212276344e-05, + "loss": 1.2253, + "step": 3422 + }, + { + "epoch": 0.08, + "learning_rate": 1.9856844920298824e-05, + "loss": 0.9412, + "step": 3423 + }, + { + "epoch": 0.08, + "learning_rate": 1.9856713568522143e-05, + "loss": 1.0365, + "step": 3424 + }, + { + "epoch": 0.08, + "learning_rate": 1.98565821569471e-05, + "loss": 1.6231, + "step": 3425 + }, + { + "epoch": 0.08, + "learning_rate": 1.985645068557449e-05, + "loss": 1.2377, + "step": 3426 + }, + { + "epoch": 0.08, + "learning_rate": 1.985631915440511e-05, + "loss": 1.0014, + "step": 3427 + }, + { + "epoch": 0.08, + "learning_rate": 1.985618756343977e-05, + "loss": 1.3176, + "step": 3428 + }, + { + "epoch": 0.08, + "learning_rate": 1.9856055912679252e-05, + "loss": 1.0669, + "step": 3429 + }, + { + "epoch": 0.08, + "learning_rate": 1.9855924202124358e-05, + "loss": 0.8745, + "step": 3430 + }, + { + "epoch": 0.08, + "learning_rate": 1.9855792431775894e-05, + "loss": 1.1747, + "step": 3431 + }, + { + "epoch": 0.08, + "learning_rate": 1.9855660601634656e-05, + "loss": 1.1414, + "step": 3432 + }, + { + "epoch": 0.08, + "learning_rate": 1.985552871170144e-05, + "loss": 1.1827, + "step": 3433 + }, + { + "epoch": 0.08, + "learning_rate": 1.9855396761977052e-05, + "loss": 1.2793, + "step": 3434 + }, + { + "epoch": 0.08, + "learning_rate": 1.985526475246229e-05, + "loss": 0.9774, + "step": 3435 + }, + { + "epoch": 0.08, + "learning_rate": 1.985513268315795e-05, + "loss": 1.0844, + "step": 3436 + }, + { + "epoch": 0.08, + "learning_rate": 1.9855000554064846e-05, + "loss": 1.4392, + "step": 3437 + }, + { + "epoch": 0.08, + "learning_rate": 1.985486836518377e-05, + "loss": 1.2547, + "step": 3438 + }, + { + "epoch": 0.08, + "learning_rate": 1.985473611651552e-05, + "loss": 1.2698, + "step": 3439 + }, + { + "epoch": 0.08, + "learning_rate": 1.9854603808060907e-05, + "loss": 1.5611, + "step": 3440 + }, + { + "epoch": 0.08, + "learning_rate": 1.985447143982073e-05, + "loss": 0.9801, + "step": 3441 + }, + { + "epoch": 0.08, + "learning_rate": 1.9854339011795795e-05, + "loss": 1.0231, + "step": 3442 + }, + { + "epoch": 0.08, + "learning_rate": 1.9854206523986904e-05, + "loss": 1.118, + "step": 3443 + }, + { + "epoch": 0.08, + "learning_rate": 1.9854073976394858e-05, + "loss": 0.9769, + "step": 3444 + }, + { + "epoch": 0.08, + "learning_rate": 1.9853941369020463e-05, + "loss": 1.1099, + "step": 3445 + }, + { + "epoch": 0.08, + "learning_rate": 1.9853808701864522e-05, + "loss": 1.1236, + "step": 3446 + }, + { + "epoch": 0.08, + "learning_rate": 1.9853675974927845e-05, + "loss": 1.1522, + "step": 3447 + }, + { + "epoch": 0.08, + "learning_rate": 1.985354318821123e-05, + "loss": 1.3556, + "step": 3448 + }, + { + "epoch": 0.08, + "learning_rate": 1.985341034171549e-05, + "loss": 1.2136, + "step": 3449 + }, + { + "epoch": 0.08, + "learning_rate": 1.9853277435441422e-05, + "loss": 1.2881, + "step": 3450 + }, + { + "epoch": 0.08, + "learning_rate": 1.985314446938984e-05, + "loss": 1.3032, + "step": 3451 + }, + { + "epoch": 0.08, + "learning_rate": 1.985301144356155e-05, + "loss": 1.3039, + "step": 3452 + }, + { + "epoch": 0.08, + "learning_rate": 1.9852878357957354e-05, + "loss": 1.2492, + "step": 3453 + }, + { + "epoch": 0.08, + "learning_rate": 1.9852745212578063e-05, + "loss": 1.1387, + "step": 3454 + }, + { + "epoch": 0.08, + "learning_rate": 1.9852612007424484e-05, + "loss": 1.0987, + "step": 3455 + }, + { + "epoch": 0.08, + "learning_rate": 1.9852478742497426e-05, + "loss": 1.1465, + "step": 3456 + }, + { + "epoch": 0.08, + "learning_rate": 1.98523454177977e-05, + "loss": 1.385, + "step": 3457 + }, + { + "epoch": 0.08, + "learning_rate": 1.985221203332611e-05, + "loss": 1.1903, + "step": 3458 + }, + { + "epoch": 0.08, + "learning_rate": 1.9852078589083465e-05, + "loss": 1.2577, + "step": 3459 + }, + { + "epoch": 0.08, + "learning_rate": 1.985194508507058e-05, + "loss": 0.9819, + "step": 3460 + }, + { + "epoch": 0.08, + "learning_rate": 1.985181152128826e-05, + "loss": 1.1412, + "step": 3461 + }, + { + "epoch": 0.08, + "learning_rate": 1.9851677897737314e-05, + "loss": 1.2467, + "step": 3462 + }, + { + "epoch": 0.08, + "learning_rate": 1.9851544214418555e-05, + "loss": 1.2953, + "step": 3463 + }, + { + "epoch": 0.08, + "learning_rate": 1.98514104713328e-05, + "loss": 1.1168, + "step": 3464 + }, + { + "epoch": 0.08, + "learning_rate": 1.9851276668480853e-05, + "loss": 1.144, + "step": 3465 + }, + { + "epoch": 0.08, + "learning_rate": 1.9851142805863523e-05, + "loss": 1.1334, + "step": 3466 + }, + { + "epoch": 0.08, + "learning_rate": 1.9851008883481634e-05, + "loss": 1.089, + "step": 3467 + }, + { + "epoch": 0.08, + "learning_rate": 1.9850874901335984e-05, + "loss": 1.001, + "step": 3468 + }, + { + "epoch": 0.08, + "learning_rate": 1.9850740859427396e-05, + "loss": 1.2782, + "step": 3469 + }, + { + "epoch": 0.08, + "learning_rate": 1.9850606757756683e-05, + "loss": 1.0145, + "step": 3470 + }, + { + "epoch": 0.08, + "learning_rate": 1.985047259632465e-05, + "loss": 1.162, + "step": 3471 + }, + { + "epoch": 0.08, + "learning_rate": 1.9850338375132125e-05, + "loss": 1.2687, + "step": 3472 + }, + { + "epoch": 0.08, + "learning_rate": 1.985020409417991e-05, + "loss": 1.222, + "step": 3473 + }, + { + "epoch": 0.08, + "learning_rate": 1.985006975346882e-05, + "loss": 1.114, + "step": 3474 + }, + { + "epoch": 0.08, + "learning_rate": 1.9849935352999677e-05, + "loss": 1.2991, + "step": 3475 + }, + { + "epoch": 0.08, + "learning_rate": 1.9849800892773293e-05, + "loss": 1.1694, + "step": 3476 + }, + { + "epoch": 0.08, + "learning_rate": 1.9849666372790484e-05, + "loss": 1.3567, + "step": 3477 + }, + { + "epoch": 0.08, + "learning_rate": 1.9849531793052064e-05, + "loss": 1.2414, + "step": 3478 + }, + { + "epoch": 0.08, + "learning_rate": 1.9849397153558852e-05, + "loss": 0.9827, + "step": 3479 + }, + { + "epoch": 0.08, + "learning_rate": 1.9849262454311663e-05, + "loss": 1.2546, + "step": 3480 + }, + { + "epoch": 0.08, + "learning_rate": 1.9849127695311318e-05, + "loss": 1.0623, + "step": 3481 + }, + { + "epoch": 0.08, + "learning_rate": 1.9848992876558633e-05, + "loss": 1.1277, + "step": 3482 + }, + { + "epoch": 0.08, + "learning_rate": 1.9848857998054424e-05, + "loss": 1.2176, + "step": 3483 + }, + { + "epoch": 0.08, + "learning_rate": 1.9848723059799508e-05, + "loss": 1.1478, + "step": 3484 + }, + { + "epoch": 0.08, + "learning_rate": 1.9848588061794708e-05, + "loss": 1.2536, + "step": 3485 + }, + { + "epoch": 0.08, + "learning_rate": 1.9848453004040838e-05, + "loss": 1.3562, + "step": 3486 + }, + { + "epoch": 0.08, + "learning_rate": 1.9848317886538724e-05, + "loss": 1.2655, + "step": 3487 + }, + { + "epoch": 0.08, + "learning_rate": 1.984818270928918e-05, + "loss": 1.2927, + "step": 3488 + }, + { + "epoch": 0.08, + "learning_rate": 1.9848047472293027e-05, + "loss": 1.1299, + "step": 3489 + }, + { + "epoch": 0.08, + "learning_rate": 1.9847912175551085e-05, + "loss": 1.2336, + "step": 3490 + }, + { + "epoch": 0.08, + "learning_rate": 1.984777681906418e-05, + "loss": 1.0657, + "step": 3491 + }, + { + "epoch": 0.08, + "learning_rate": 1.9847641402833127e-05, + "loss": 0.8817, + "step": 3492 + }, + { + "epoch": 0.08, + "learning_rate": 1.984750592685875e-05, + "loss": 1.1482, + "step": 3493 + }, + { + "epoch": 0.08, + "learning_rate": 1.9847370391141872e-05, + "loss": 1.181, + "step": 3494 + }, + { + "epoch": 0.08, + "learning_rate": 1.984723479568331e-05, + "loss": 1.0854, + "step": 3495 + }, + { + "epoch": 0.08, + "learning_rate": 1.9847099140483896e-05, + "loss": 1.1566, + "step": 3496 + }, + { + "epoch": 0.08, + "learning_rate": 1.9846963425544447e-05, + "loss": 1.3264, + "step": 3497 + }, + { + "epoch": 0.08, + "learning_rate": 1.9846827650865787e-05, + "loss": 1.136, + "step": 3498 + }, + { + "epoch": 0.08, + "learning_rate": 1.984669181644874e-05, + "loss": 1.105, + "step": 3499 + }, + { + "epoch": 0.08, + "learning_rate": 1.9846555922294123e-05, + "loss": 1.2352, + "step": 3500 + }, + { + "epoch": 0.08, + "learning_rate": 1.9846419968402776e-05, + "loss": 1.1141, + "step": 3501 + }, + { + "epoch": 0.08, + "learning_rate": 1.984628395477551e-05, + "loss": 1.0727, + "step": 3502 + }, + { + "epoch": 0.08, + "learning_rate": 1.9846147881413156e-05, + "loss": 1.4706, + "step": 3503 + }, + { + "epoch": 0.08, + "learning_rate": 1.984601174831654e-05, + "loss": 1.1797, + "step": 3504 + }, + { + "epoch": 0.08, + "learning_rate": 1.9845875555486488e-05, + "loss": 1.2108, + "step": 3505 + }, + { + "epoch": 0.08, + "learning_rate": 1.9845739302923822e-05, + "loss": 1.161, + "step": 3506 + }, + { + "epoch": 0.08, + "learning_rate": 1.9845602990629373e-05, + "loss": 1.4042, + "step": 3507 + }, + { + "epoch": 0.08, + "learning_rate": 1.984546661860397e-05, + "loss": 1.0007, + "step": 3508 + }, + { + "epoch": 0.08, + "learning_rate": 1.984533018684843e-05, + "loss": 0.9695, + "step": 3509 + }, + { + "epoch": 0.08, + "learning_rate": 1.9845193695363592e-05, + "loss": 0.983, + "step": 3510 + }, + { + "epoch": 0.08, + "learning_rate": 1.9845057144150276e-05, + "loss": 1.1065, + "step": 3511 + }, + { + "epoch": 0.08, + "learning_rate": 1.984492053320932e-05, + "loss": 1.1378, + "step": 3512 + }, + { + "epoch": 0.08, + "learning_rate": 1.9844783862541538e-05, + "loss": 1.2607, + "step": 3513 + }, + { + "epoch": 0.08, + "learning_rate": 1.9844647132147773e-05, + "loss": 1.0554, + "step": 3514 + }, + { + "epoch": 0.08, + "learning_rate": 1.984451034202885e-05, + "loss": 1.3154, + "step": 3515 + }, + { + "epoch": 0.08, + "learning_rate": 1.98443734921856e-05, + "loss": 1.2193, + "step": 3516 + }, + { + "epoch": 0.08, + "learning_rate": 1.9844236582618847e-05, + "loss": 1.0727, + "step": 3517 + }, + { + "epoch": 0.08, + "learning_rate": 1.984409961332943e-05, + "loss": 1.1449, + "step": 3518 + }, + { + "epoch": 0.08, + "learning_rate": 1.9843962584318172e-05, + "loss": 1.0504, + "step": 3519 + }, + { + "epoch": 0.08, + "learning_rate": 1.9843825495585912e-05, + "loss": 1.1332, + "step": 3520 + }, + { + "epoch": 0.08, + "learning_rate": 1.984368834713348e-05, + "loss": 1.3178, + "step": 3521 + }, + { + "epoch": 0.08, + "learning_rate": 1.98435511389617e-05, + "loss": 0.8221, + "step": 3522 + }, + { + "epoch": 0.08, + "learning_rate": 1.9843413871071416e-05, + "loss": 1.0476, + "step": 3523 + }, + { + "epoch": 0.08, + "learning_rate": 1.9843276543463452e-05, + "loss": 1.2939, + "step": 3524 + }, + { + "epoch": 0.08, + "learning_rate": 1.9843139156138648e-05, + "loss": 1.12, + "step": 3525 + }, + { + "epoch": 0.08, + "learning_rate": 1.9843001709097832e-05, + "loss": 1.2747, + "step": 3526 + }, + { + "epoch": 0.08, + "learning_rate": 1.9842864202341837e-05, + "loss": 1.2734, + "step": 3527 + }, + { + "epoch": 0.08, + "learning_rate": 1.9842726635871503e-05, + "loss": 1.3682, + "step": 3528 + }, + { + "epoch": 0.08, + "learning_rate": 1.984258900968766e-05, + "loss": 1.0686, + "step": 3529 + }, + { + "epoch": 0.08, + "learning_rate": 1.9842451323791146e-05, + "loss": 1.161, + "step": 3530 + }, + { + "epoch": 0.08, + "learning_rate": 1.9842313578182796e-05, + "loss": 1.1944, + "step": 3531 + }, + { + "epoch": 0.08, + "learning_rate": 1.9842175772863442e-05, + "loss": 1.0764, + "step": 3532 + }, + { + "epoch": 0.08, + "learning_rate": 1.9842037907833926e-05, + "loss": 0.9753, + "step": 3533 + }, + { + "epoch": 0.08, + "learning_rate": 1.984189998309508e-05, + "loss": 1.0499, + "step": 3534 + }, + { + "epoch": 0.09, + "learning_rate": 1.984176199864774e-05, + "loss": 1.1777, + "step": 3535 + }, + { + "epoch": 0.09, + "learning_rate": 1.9841623954492744e-05, + "loss": 1.5069, + "step": 3536 + }, + { + "epoch": 0.09, + "learning_rate": 1.9841485850630932e-05, + "loss": 1.1162, + "step": 3537 + }, + { + "epoch": 0.09, + "learning_rate": 1.984134768706314e-05, + "loss": 1.2669, + "step": 3538 + }, + { + "epoch": 0.09, + "learning_rate": 1.9841209463790204e-05, + "loss": 1.0144, + "step": 3539 + }, + { + "epoch": 0.09, + "learning_rate": 1.9841071180812972e-05, + "loss": 1.1647, + "step": 3540 + }, + { + "epoch": 0.09, + "learning_rate": 1.984093283813227e-05, + "loss": 1.2187, + "step": 3541 + }, + { + "epoch": 0.09, + "learning_rate": 1.9840794435748946e-05, + "loss": 1.2059, + "step": 3542 + }, + { + "epoch": 0.09, + "learning_rate": 1.9840655973663832e-05, + "loss": 1.3008, + "step": 3543 + }, + { + "epoch": 0.09, + "learning_rate": 1.984051745187778e-05, + "loss": 1.2457, + "step": 3544 + }, + { + "epoch": 0.09, + "learning_rate": 1.9840378870391616e-05, + "loss": 1.0222, + "step": 3545 + }, + { + "epoch": 0.09, + "learning_rate": 1.9840240229206194e-05, + "loss": 1.4281, + "step": 3546 + }, + { + "epoch": 0.09, + "learning_rate": 1.9840101528322346e-05, + "loss": 1.2215, + "step": 3547 + }, + { + "epoch": 0.09, + "learning_rate": 1.9839962767740918e-05, + "loss": 1.2869, + "step": 3548 + }, + { + "epoch": 0.09, + "learning_rate": 1.983982394746275e-05, + "loss": 1.2889, + "step": 3549 + }, + { + "epoch": 0.09, + "learning_rate": 1.9839685067488683e-05, + "loss": 0.8533, + "step": 3550 + }, + { + "epoch": 0.09, + "learning_rate": 1.9839546127819563e-05, + "loss": 1.1654, + "step": 3551 + }, + { + "epoch": 0.09, + "learning_rate": 1.983940712845623e-05, + "loss": 1.2603, + "step": 3552 + }, + { + "epoch": 0.09, + "learning_rate": 1.9839268069399528e-05, + "loss": 1.2903, + "step": 3553 + }, + { + "epoch": 0.09, + "learning_rate": 1.9839128950650302e-05, + "loss": 1.1163, + "step": 3554 + }, + { + "epoch": 0.09, + "learning_rate": 1.9838989772209395e-05, + "loss": 1.3261, + "step": 3555 + }, + { + "epoch": 0.09, + "learning_rate": 1.9838850534077652e-05, + "loss": 1.3496, + "step": 3556 + }, + { + "epoch": 0.09, + "learning_rate": 1.9838711236255914e-05, + "loss": 0.9758, + "step": 3557 + }, + { + "epoch": 0.09, + "learning_rate": 1.983857187874503e-05, + "loss": 1.076, + "step": 3558 + }, + { + "epoch": 0.09, + "learning_rate": 1.9838432461545848e-05, + "loss": 1.0953, + "step": 3559 + }, + { + "epoch": 0.09, + "learning_rate": 1.9838292984659207e-05, + "loss": 1.357, + "step": 3560 + }, + { + "epoch": 0.09, + "learning_rate": 1.983815344808596e-05, + "loss": 1.2419, + "step": 3561 + }, + { + "epoch": 0.09, + "learning_rate": 1.9838013851826948e-05, + "loss": 1.1771, + "step": 3562 + }, + { + "epoch": 0.09, + "learning_rate": 1.983787419588302e-05, + "loss": 1.0569, + "step": 3563 + }, + { + "epoch": 0.09, + "learning_rate": 1.983773448025502e-05, + "loss": 1.398, + "step": 3564 + }, + { + "epoch": 0.09, + "learning_rate": 1.9837594704943802e-05, + "loss": 1.4012, + "step": 3565 + }, + { + "epoch": 0.09, + "learning_rate": 1.983745486995021e-05, + "loss": 1.315, + "step": 3566 + }, + { + "epoch": 0.09, + "learning_rate": 1.9837314975275096e-05, + "loss": 1.1628, + "step": 3567 + }, + { + "epoch": 0.09, + "learning_rate": 1.9837175020919304e-05, + "loss": 1.2702, + "step": 3568 + }, + { + "epoch": 0.09, + "learning_rate": 1.9837035006883683e-05, + "loss": 0.8399, + "step": 3569 + }, + { + "epoch": 0.09, + "learning_rate": 1.983689493316909e-05, + "loss": 1.3342, + "step": 3570 + }, + { + "epoch": 0.09, + "learning_rate": 1.9836754799776363e-05, + "loss": 1.2081, + "step": 3571 + }, + { + "epoch": 0.09, + "learning_rate": 1.983661460670636e-05, + "loss": 1.1229, + "step": 3572 + }, + { + "epoch": 0.09, + "learning_rate": 1.9836474353959935e-05, + "loss": 1.2749, + "step": 3573 + }, + { + "epoch": 0.09, + "learning_rate": 1.9836334041537928e-05, + "loss": 1.3417, + "step": 3574 + }, + { + "epoch": 0.09, + "learning_rate": 1.9836193669441196e-05, + "loss": 1.2025, + "step": 3575 + }, + { + "epoch": 0.09, + "learning_rate": 1.9836053237670594e-05, + "loss": 1.0629, + "step": 3576 + }, + { + "epoch": 0.09, + "learning_rate": 1.9835912746226967e-05, + "loss": 1.0381, + "step": 3577 + }, + { + "epoch": 0.09, + "learning_rate": 1.9835772195111174e-05, + "loss": 1.1218, + "step": 3578 + }, + { + "epoch": 0.09, + "learning_rate": 1.9835631584324064e-05, + "loss": 1.197, + "step": 3579 + }, + { + "epoch": 0.09, + "learning_rate": 1.9835490913866492e-05, + "loss": 1.2396, + "step": 3580 + }, + { + "epoch": 0.09, + "learning_rate": 1.9835350183739305e-05, + "loss": 1.1915, + "step": 3581 + }, + { + "epoch": 0.09, + "learning_rate": 1.9835209393943366e-05, + "loss": 1.1786, + "step": 3582 + }, + { + "epoch": 0.09, + "learning_rate": 1.9835068544479523e-05, + "loss": 1.3173, + "step": 3583 + }, + { + "epoch": 0.09, + "learning_rate": 1.9834927635348635e-05, + "loss": 1.0867, + "step": 3584 + }, + { + "epoch": 0.09, + "learning_rate": 1.9834786666551555e-05, + "loss": 1.1894, + "step": 3585 + }, + { + "epoch": 0.09, + "learning_rate": 1.983464563808913e-05, + "loss": 1.1999, + "step": 3586 + }, + { + "epoch": 0.09, + "learning_rate": 1.9834504549962232e-05, + "loss": 1.0747, + "step": 3587 + }, + { + "epoch": 0.09, + "learning_rate": 1.9834363402171704e-05, + "loss": 1.0218, + "step": 3588 + }, + { + "epoch": 0.09, + "learning_rate": 1.9834222194718408e-05, + "loss": 1.0807, + "step": 3589 + }, + { + "epoch": 0.09, + "learning_rate": 1.9834080927603198e-05, + "loss": 1.2238, + "step": 3590 + }, + { + "epoch": 0.09, + "learning_rate": 1.983393960082693e-05, + "loss": 1.2094, + "step": 3591 + }, + { + "epoch": 0.09, + "learning_rate": 1.9833798214390467e-05, + "loss": 1.4621, + "step": 3592 + }, + { + "epoch": 0.09, + "learning_rate": 1.983365676829466e-05, + "loss": 1.0217, + "step": 3593 + }, + { + "epoch": 0.09, + "learning_rate": 1.9833515262540376e-05, + "loss": 1.0439, + "step": 3594 + }, + { + "epoch": 0.09, + "learning_rate": 1.9833373697128463e-05, + "loss": 1.34, + "step": 3595 + }, + { + "epoch": 0.09, + "learning_rate": 1.9833232072059786e-05, + "loss": 1.1905, + "step": 3596 + }, + { + "epoch": 0.09, + "learning_rate": 1.9833090387335205e-05, + "loss": 1.1753, + "step": 3597 + }, + { + "epoch": 0.09, + "learning_rate": 1.9832948642955574e-05, + "loss": 1.2424, + "step": 3598 + }, + { + "epoch": 0.09, + "learning_rate": 1.9832806838921757e-05, + "loss": 1.3023, + "step": 3599 + }, + { + "epoch": 0.09, + "learning_rate": 1.983266497523462e-05, + "loss": 1.2164, + "step": 3600 + }, + { + "epoch": 0.09, + "learning_rate": 1.9832523051895008e-05, + "loss": 1.1744, + "step": 3601 + }, + { + "epoch": 0.09, + "learning_rate": 1.9832381068903797e-05, + "loss": 1.1895, + "step": 3602 + }, + { + "epoch": 0.09, + "learning_rate": 1.983223902626184e-05, + "loss": 1.3267, + "step": 3603 + }, + { + "epoch": 0.09, + "learning_rate": 1.9832096923970002e-05, + "loss": 0.9729, + "step": 3604 + }, + { + "epoch": 0.09, + "learning_rate": 1.9831954762029146e-05, + "loss": 1.2, + "step": 3605 + }, + { + "epoch": 0.09, + "learning_rate": 1.983181254044013e-05, + "loss": 1.2906, + "step": 3606 + }, + { + "epoch": 0.09, + "learning_rate": 1.9831670259203825e-05, + "loss": 1.0739, + "step": 3607 + }, + { + "epoch": 0.09, + "learning_rate": 1.983152791832109e-05, + "loss": 1.2807, + "step": 3608 + }, + { + "epoch": 0.09, + "learning_rate": 1.983138551779278e-05, + "loss": 1.2906, + "step": 3609 + }, + { + "epoch": 0.09, + "learning_rate": 1.9831243057619774e-05, + "loss": 1.2018, + "step": 3610 + }, + { + "epoch": 0.09, + "learning_rate": 1.9831100537802923e-05, + "loss": 1.2193, + "step": 3611 + }, + { + "epoch": 0.09, + "learning_rate": 1.98309579583431e-05, + "loss": 1.2005, + "step": 3612 + }, + { + "epoch": 0.09, + "learning_rate": 1.983081531924117e-05, + "loss": 1.1598, + "step": 3613 + }, + { + "epoch": 0.09, + "learning_rate": 1.983067262049799e-05, + "loss": 1.2393, + "step": 3614 + }, + { + "epoch": 0.09, + "learning_rate": 1.9830529862114435e-05, + "loss": 1.1517, + "step": 3615 + }, + { + "epoch": 0.09, + "learning_rate": 1.983038704409137e-05, + "loss": 1.3183, + "step": 3616 + }, + { + "epoch": 0.09, + "learning_rate": 1.9830244166429654e-05, + "loss": 1.2282, + "step": 3617 + }, + { + "epoch": 0.09, + "learning_rate": 1.9830101229130162e-05, + "loss": 1.1776, + "step": 3618 + }, + { + "epoch": 0.09, + "learning_rate": 1.9829958232193757e-05, + "loss": 1.2482, + "step": 3619 + }, + { + "epoch": 0.09, + "learning_rate": 1.9829815175621306e-05, + "loss": 1.0267, + "step": 3620 + }, + { + "epoch": 0.09, + "learning_rate": 1.982967205941368e-05, + "loss": 1.4199, + "step": 3621 + }, + { + "epoch": 0.09, + "learning_rate": 1.9829528883571747e-05, + "loss": 1.1668, + "step": 3622 + }, + { + "epoch": 0.09, + "learning_rate": 1.9829385648096373e-05, + "loss": 1.019, + "step": 3623 + }, + { + "epoch": 0.09, + "learning_rate": 1.982924235298843e-05, + "loss": 1.1508, + "step": 3624 + }, + { + "epoch": 0.09, + "learning_rate": 1.9829098998248783e-05, + "loss": 1.3279, + "step": 3625 + }, + { + "epoch": 0.09, + "learning_rate": 1.9828955583878306e-05, + "loss": 0.9331, + "step": 3626 + }, + { + "epoch": 0.09, + "learning_rate": 1.9828812109877868e-05, + "loss": 1.408, + "step": 3627 + }, + { + "epoch": 0.09, + "learning_rate": 1.9828668576248336e-05, + "loss": 1.0447, + "step": 3628 + }, + { + "epoch": 0.09, + "learning_rate": 1.9828524982990586e-05, + "loss": 1.4166, + "step": 3629 + }, + { + "epoch": 0.09, + "learning_rate": 1.9828381330105487e-05, + "loss": 1.2759, + "step": 3630 + }, + { + "epoch": 0.09, + "learning_rate": 1.982823761759391e-05, + "loss": 0.9381, + "step": 3631 + }, + { + "epoch": 0.09, + "learning_rate": 1.9828093845456726e-05, + "loss": 1.2505, + "step": 3632 + }, + { + "epoch": 0.09, + "learning_rate": 1.982795001369481e-05, + "loss": 1.1383, + "step": 3633 + }, + { + "epoch": 0.09, + "learning_rate": 1.982780612230903e-05, + "loss": 1.4354, + "step": 3634 + }, + { + "epoch": 0.09, + "learning_rate": 1.9827662171300263e-05, + "loss": 1.0405, + "step": 3635 + }, + { + "epoch": 0.09, + "learning_rate": 1.9827518160669382e-05, + "loss": 1.2596, + "step": 3636 + }, + { + "epoch": 0.09, + "learning_rate": 1.982737409041726e-05, + "loss": 1.3045, + "step": 3637 + }, + { + "epoch": 0.09, + "learning_rate": 1.9827229960544773e-05, + "loss": 1.2532, + "step": 3638 + }, + { + "epoch": 0.09, + "learning_rate": 1.982708577105279e-05, + "loss": 1.1017, + "step": 3639 + }, + { + "epoch": 0.09, + "learning_rate": 1.9826941521942187e-05, + "loss": 1.4189, + "step": 3640 + }, + { + "epoch": 0.09, + "learning_rate": 1.9826797213213846e-05, + "loss": 1.2992, + "step": 3641 + }, + { + "epoch": 0.09, + "learning_rate": 1.9826652844868636e-05, + "loss": 1.2783, + "step": 3642 + }, + { + "epoch": 0.09, + "learning_rate": 1.9826508416907434e-05, + "loss": 1.3038, + "step": 3643 + }, + { + "epoch": 0.09, + "learning_rate": 1.9826363929331118e-05, + "loss": 1.0153, + "step": 3644 + }, + { + "epoch": 0.09, + "learning_rate": 1.9826219382140563e-05, + "loss": 0.8038, + "step": 3645 + }, + { + "epoch": 0.09, + "learning_rate": 1.9826074775336645e-05, + "loss": 1.0064, + "step": 3646 + }, + { + "epoch": 0.09, + "learning_rate": 1.9825930108920243e-05, + "loss": 1.3284, + "step": 3647 + }, + { + "epoch": 0.09, + "learning_rate": 1.9825785382892236e-05, + "loss": 1.3618, + "step": 3648 + }, + { + "epoch": 0.09, + "learning_rate": 1.9825640597253497e-05, + "loss": 1.23, + "step": 3649 + }, + { + "epoch": 0.09, + "learning_rate": 1.9825495752004912e-05, + "loss": 1.2349, + "step": 3650 + }, + { + "epoch": 0.09, + "learning_rate": 1.9825350847147352e-05, + "loss": 1.1117, + "step": 3651 + }, + { + "epoch": 0.09, + "learning_rate": 1.98252058826817e-05, + "loss": 1.0586, + "step": 3652 + }, + { + "epoch": 0.09, + "learning_rate": 1.9825060858608837e-05, + "loss": 1.0875, + "step": 3653 + }, + { + "epoch": 0.09, + "learning_rate": 1.982491577492964e-05, + "loss": 1.1847, + "step": 3654 + }, + { + "epoch": 0.09, + "learning_rate": 1.9824770631644986e-05, + "loss": 1.0408, + "step": 3655 + }, + { + "epoch": 0.09, + "learning_rate": 1.982462542875576e-05, + "loss": 1.36, + "step": 3656 + }, + { + "epoch": 0.09, + "learning_rate": 1.982448016626285e-05, + "loss": 0.9289, + "step": 3657 + }, + { + "epoch": 0.09, + "learning_rate": 1.9824334844167122e-05, + "loss": 1.2093, + "step": 3658 + }, + { + "epoch": 0.09, + "learning_rate": 1.9824189462469468e-05, + "loss": 1.2965, + "step": 3659 + }, + { + "epoch": 0.09, + "learning_rate": 1.982404402117077e-05, + "loss": 1.1078, + "step": 3660 + }, + { + "epoch": 0.09, + "learning_rate": 1.98238985202719e-05, + "loss": 1.2491, + "step": 3661 + }, + { + "epoch": 0.09, + "learning_rate": 1.9823752959773758e-05, + "loss": 1.2955, + "step": 3662 + }, + { + "epoch": 0.09, + "learning_rate": 1.9823607339677214e-05, + "loss": 1.2272, + "step": 3663 + }, + { + "epoch": 0.09, + "learning_rate": 1.9823461659983154e-05, + "loss": 1.2935, + "step": 3664 + }, + { + "epoch": 0.09, + "learning_rate": 1.9823315920692462e-05, + "loss": 1.3571, + "step": 3665 + }, + { + "epoch": 0.09, + "learning_rate": 1.9823170121806023e-05, + "loss": 1.2099, + "step": 3666 + }, + { + "epoch": 0.09, + "learning_rate": 1.982302426332472e-05, + "loss": 1.217, + "step": 3667 + }, + { + "epoch": 0.09, + "learning_rate": 1.9822878345249443e-05, + "loss": 1.0837, + "step": 3668 + }, + { + "epoch": 0.09, + "learning_rate": 1.982273236758107e-05, + "loss": 1.16, + "step": 3669 + }, + { + "epoch": 0.09, + "learning_rate": 1.9822586330320492e-05, + "loss": 1.1386, + "step": 3670 + }, + { + "epoch": 0.09, + "learning_rate": 1.9822440233468594e-05, + "loss": 1.2053, + "step": 3671 + }, + { + "epoch": 0.09, + "learning_rate": 1.982229407702626e-05, + "loss": 1.0452, + "step": 3672 + }, + { + "epoch": 0.09, + "learning_rate": 1.9822147860994378e-05, + "loss": 1.1069, + "step": 3673 + }, + { + "epoch": 0.09, + "learning_rate": 1.9822001585373835e-05, + "loss": 1.2675, + "step": 3674 + }, + { + "epoch": 0.09, + "learning_rate": 1.982185525016552e-05, + "loss": 1.3378, + "step": 3675 + }, + { + "epoch": 0.09, + "learning_rate": 1.9821708855370322e-05, + "loss": 1.0323, + "step": 3676 + }, + { + "epoch": 0.09, + "learning_rate": 1.9821562400989124e-05, + "loss": 1.0716, + "step": 3677 + }, + { + "epoch": 0.09, + "learning_rate": 1.9821415887022813e-05, + "loss": 1.1367, + "step": 3678 + }, + { + "epoch": 0.09, + "learning_rate": 1.982126931347229e-05, + "loss": 1.1805, + "step": 3679 + }, + { + "epoch": 0.09, + "learning_rate": 1.9821122680338428e-05, + "loss": 1.4016, + "step": 3680 + }, + { + "epoch": 0.09, + "learning_rate": 1.982097598762213e-05, + "loss": 1.2141, + "step": 3681 + }, + { + "epoch": 0.09, + "learning_rate": 1.982082923532428e-05, + "loss": 1.2014, + "step": 3682 + }, + { + "epoch": 0.09, + "learning_rate": 1.982068242344577e-05, + "loss": 1.1746, + "step": 3683 + }, + { + "epoch": 0.09, + "learning_rate": 1.9820535551987486e-05, + "loss": 1.2846, + "step": 3684 + }, + { + "epoch": 0.09, + "learning_rate": 1.9820388620950328e-05, + "loss": 1.2232, + "step": 3685 + }, + { + "epoch": 0.09, + "learning_rate": 1.9820241630335176e-05, + "loss": 1.3991, + "step": 3686 + }, + { + "epoch": 0.09, + "learning_rate": 1.982009458014293e-05, + "loss": 1.1344, + "step": 3687 + }, + { + "epoch": 0.09, + "learning_rate": 1.9819947470374484e-05, + "loss": 1.354, + "step": 3688 + }, + { + "epoch": 0.09, + "learning_rate": 1.981980030103072e-05, + "loss": 1.2241, + "step": 3689 + }, + { + "epoch": 0.09, + "learning_rate": 1.981965307211254e-05, + "loss": 1.2451, + "step": 3690 + }, + { + "epoch": 0.09, + "learning_rate": 1.9819505783620835e-05, + "loss": 0.6869, + "step": 3691 + }, + { + "epoch": 0.09, + "learning_rate": 1.98193584355565e-05, + "loss": 1.1328, + "step": 3692 + }, + { + "epoch": 0.09, + "learning_rate": 1.9819211027920423e-05, + "loss": 1.1857, + "step": 3693 + }, + { + "epoch": 0.09, + "learning_rate": 1.9819063560713507e-05, + "loss": 0.7526, + "step": 3694 + }, + { + "epoch": 0.09, + "learning_rate": 1.981891603393664e-05, + "loss": 1.1279, + "step": 3695 + }, + { + "epoch": 0.09, + "learning_rate": 1.981876844759072e-05, + "loss": 1.2984, + "step": 3696 + }, + { + "epoch": 0.09, + "learning_rate": 1.981862080167664e-05, + "loss": 1.0924, + "step": 3697 + }, + { + "epoch": 0.09, + "learning_rate": 1.9818473096195295e-05, + "loss": 1.3092, + "step": 3698 + }, + { + "epoch": 0.09, + "learning_rate": 1.9818325331147586e-05, + "loss": 1.1493, + "step": 3699 + }, + { + "epoch": 0.09, + "learning_rate": 1.9818177506534406e-05, + "loss": 1.1859, + "step": 3700 + }, + { + "epoch": 0.09, + "learning_rate": 1.981802962235665e-05, + "loss": 1.1257, + "step": 3701 + }, + { + "epoch": 0.09, + "learning_rate": 1.981788167861522e-05, + "loss": 1.0382, + "step": 3702 + }, + { + "epoch": 0.09, + "learning_rate": 1.9817733675311012e-05, + "loss": 0.9905, + "step": 3703 + }, + { + "epoch": 0.09, + "learning_rate": 1.9817585612444923e-05, + "loss": 1.1527, + "step": 3704 + }, + { + "epoch": 0.09, + "learning_rate": 1.981743749001785e-05, + "loss": 1.2544, + "step": 3705 + }, + { + "epoch": 0.09, + "learning_rate": 1.9817289308030695e-05, + "loss": 1.1673, + "step": 3706 + }, + { + "epoch": 0.09, + "learning_rate": 1.9817141066484354e-05, + "loss": 1.2776, + "step": 3707 + }, + { + "epoch": 0.09, + "learning_rate": 1.9816992765379725e-05, + "loss": 1.3172, + "step": 3708 + }, + { + "epoch": 0.09, + "learning_rate": 1.9816844404717712e-05, + "loss": 1.2625, + "step": 3709 + }, + { + "epoch": 0.09, + "learning_rate": 1.981669598449921e-05, + "loss": 1.0436, + "step": 3710 + }, + { + "epoch": 0.09, + "learning_rate": 1.9816547504725128e-05, + "loss": 1.201, + "step": 3711 + }, + { + "epoch": 0.09, + "learning_rate": 1.981639896539636e-05, + "loss": 1.0668, + "step": 3712 + }, + { + "epoch": 0.09, + "learning_rate": 1.9816250366513804e-05, + "loss": 1.1867, + "step": 3713 + }, + { + "epoch": 0.09, + "learning_rate": 1.981610170807837e-05, + "loss": 1.1883, + "step": 3714 + }, + { + "epoch": 0.09, + "learning_rate": 1.9815952990090954e-05, + "loss": 1.3026, + "step": 3715 + }, + { + "epoch": 0.09, + "learning_rate": 1.981580421255246e-05, + "loss": 1.3232, + "step": 3716 + }, + { + "epoch": 0.09, + "learning_rate": 1.9815655375463794e-05, + "loss": 0.9036, + "step": 3717 + }, + { + "epoch": 0.09, + "learning_rate": 1.981550647882585e-05, + "loss": 1.3379, + "step": 3718 + }, + { + "epoch": 0.09, + "learning_rate": 1.981535752263954e-05, + "loss": 1.1917, + "step": 3719 + }, + { + "epoch": 0.09, + "learning_rate": 1.9815208506905764e-05, + "loss": 1.2216, + "step": 3720 + }, + { + "epoch": 0.09, + "learning_rate": 1.981505943162543e-05, + "loss": 1.0202, + "step": 3721 + }, + { + "epoch": 0.09, + "learning_rate": 1.9814910296799436e-05, + "loss": 1.0739, + "step": 3722 + }, + { + "epoch": 0.09, + "learning_rate": 1.981476110242869e-05, + "loss": 1.1719, + "step": 3723 + }, + { + "epoch": 0.09, + "learning_rate": 1.9814611848514095e-05, + "loss": 1.1239, + "step": 3724 + }, + { + "epoch": 0.09, + "learning_rate": 1.981446253505656e-05, + "loss": 1.4358, + "step": 3725 + }, + { + "epoch": 0.09, + "learning_rate": 1.9814313162056993e-05, + "loss": 1.0402, + "step": 3726 + }, + { + "epoch": 0.09, + "learning_rate": 1.981416372951629e-05, + "loss": 1.0509, + "step": 3727 + }, + { + "epoch": 0.09, + "learning_rate": 1.981401423743537e-05, + "loss": 1.1562, + "step": 3728 + }, + { + "epoch": 0.09, + "learning_rate": 1.981386468581513e-05, + "loss": 1.085, + "step": 3729 + }, + { + "epoch": 0.09, + "learning_rate": 1.9813715074656482e-05, + "loss": 1.2488, + "step": 3730 + }, + { + "epoch": 0.09, + "learning_rate": 1.9813565403960337e-05, + "loss": 1.0333, + "step": 3731 + }, + { + "epoch": 0.09, + "learning_rate": 1.9813415673727594e-05, + "loss": 1.0295, + "step": 3732 + }, + { + "epoch": 0.09, + "learning_rate": 1.981326588395917e-05, + "loss": 1.4162, + "step": 3733 + }, + { + "epoch": 0.09, + "learning_rate": 1.9813116034655966e-05, + "loss": 1.097, + "step": 3734 + }, + { + "epoch": 0.09, + "learning_rate": 1.98129661258189e-05, + "loss": 1.2101, + "step": 3735 + }, + { + "epoch": 0.09, + "learning_rate": 1.9812816157448874e-05, + "loss": 1.1897, + "step": 3736 + }, + { + "epoch": 0.09, + "learning_rate": 1.98126661295468e-05, + "loss": 1.3112, + "step": 3737 + }, + { + "epoch": 0.09, + "learning_rate": 1.9812516042113588e-05, + "loss": 1.135, + "step": 3738 + }, + { + "epoch": 0.09, + "learning_rate": 1.981236589515015e-05, + "loss": 1.2199, + "step": 3739 + }, + { + "epoch": 0.09, + "learning_rate": 1.9812215688657397e-05, + "loss": 1.0542, + "step": 3740 + }, + { + "epoch": 0.09, + "learning_rate": 1.9812065422636238e-05, + "loss": 1.1445, + "step": 3741 + }, + { + "epoch": 0.09, + "learning_rate": 1.9811915097087587e-05, + "loss": 1.1847, + "step": 3742 + }, + { + "epoch": 0.09, + "learning_rate": 1.9811764712012355e-05, + "loss": 1.1641, + "step": 3743 + }, + { + "epoch": 0.09, + "learning_rate": 1.9811614267411453e-05, + "loss": 1.189, + "step": 3744 + }, + { + "epoch": 0.09, + "learning_rate": 1.9811463763285797e-05, + "loss": 1.0216, + "step": 3745 + }, + { + "epoch": 0.09, + "learning_rate": 1.98113131996363e-05, + "loss": 1.1246, + "step": 3746 + }, + { + "epoch": 0.09, + "learning_rate": 1.981116257646387e-05, + "loss": 1.3407, + "step": 3747 + }, + { + "epoch": 0.09, + "learning_rate": 1.9811011893769424e-05, + "loss": 1.1966, + "step": 3748 + }, + { + "epoch": 0.09, + "learning_rate": 1.981086115155388e-05, + "loss": 1.2847, + "step": 3749 + }, + { + "epoch": 0.09, + "learning_rate": 1.9810710349818147e-05, + "loss": 0.842, + "step": 3750 + }, + { + "epoch": 0.09, + "learning_rate": 1.981055948856314e-05, + "loss": 1.2604, + "step": 3751 + }, + { + "epoch": 0.09, + "learning_rate": 1.9810408567789777e-05, + "loss": 1.0982, + "step": 3752 + }, + { + "epoch": 0.09, + "learning_rate": 1.9810257587498974e-05, + "loss": 1.1704, + "step": 3753 + }, + { + "epoch": 0.09, + "learning_rate": 1.9810106547691648e-05, + "loss": 1.3462, + "step": 3754 + }, + { + "epoch": 0.09, + "learning_rate": 1.980995544836871e-05, + "loss": 1.2639, + "step": 3755 + }, + { + "epoch": 0.09, + "learning_rate": 1.980980428953108e-05, + "loss": 1.0187, + "step": 3756 + }, + { + "epoch": 0.09, + "learning_rate": 1.9809653071179678e-05, + "loss": 1.2938, + "step": 3757 + }, + { + "epoch": 0.09, + "learning_rate": 1.9809501793315412e-05, + "loss": 1.2653, + "step": 3758 + }, + { + "epoch": 0.09, + "learning_rate": 1.9809350455939215e-05, + "loss": 1.2501, + "step": 3759 + }, + { + "epoch": 0.09, + "learning_rate": 1.9809199059051987e-05, + "loss": 1.1597, + "step": 3760 + }, + { + "epoch": 0.09, + "learning_rate": 1.980904760265466e-05, + "loss": 1.2128, + "step": 3761 + }, + { + "epoch": 0.09, + "learning_rate": 1.9808896086748146e-05, + "loss": 1.1709, + "step": 3762 + }, + { + "epoch": 0.09, + "learning_rate": 1.980874451133337e-05, + "loss": 1.3722, + "step": 3763 + }, + { + "epoch": 0.09, + "learning_rate": 1.9808592876411245e-05, + "loss": 0.9504, + "step": 3764 + }, + { + "epoch": 0.09, + "learning_rate": 1.9808441181982695e-05, + "loss": 1.2076, + "step": 3765 + }, + { + "epoch": 0.09, + "learning_rate": 1.980828942804864e-05, + "loss": 1.2721, + "step": 3766 + }, + { + "epoch": 0.09, + "learning_rate": 1.9808137614610002e-05, + "loss": 0.7583, + "step": 3767 + }, + { + "epoch": 0.09, + "learning_rate": 1.9807985741667696e-05, + "loss": 1.1885, + "step": 3768 + }, + { + "epoch": 0.09, + "learning_rate": 1.980783380922265e-05, + "loss": 1.1961, + "step": 3769 + }, + { + "epoch": 0.09, + "learning_rate": 1.9807681817275783e-05, + "loss": 1.4094, + "step": 3770 + }, + { + "epoch": 0.09, + "learning_rate": 1.9807529765828016e-05, + "loss": 1.2511, + "step": 3771 + }, + { + "epoch": 0.09, + "learning_rate": 1.9807377654880274e-05, + "loss": 1.2169, + "step": 3772 + }, + { + "epoch": 0.09, + "learning_rate": 1.9807225484433478e-05, + "loss": 1.2725, + "step": 3773 + }, + { + "epoch": 0.09, + "learning_rate": 1.9807073254488554e-05, + "loss": 1.0974, + "step": 3774 + }, + { + "epoch": 0.09, + "learning_rate": 1.9806920965046422e-05, + "loss": 1.0724, + "step": 3775 + }, + { + "epoch": 0.09, + "learning_rate": 1.9806768616108006e-05, + "loss": 0.9694, + "step": 3776 + }, + { + "epoch": 0.09, + "learning_rate": 1.9806616207674235e-05, + "loss": 1.1218, + "step": 3777 + }, + { + "epoch": 0.09, + "learning_rate": 1.9806463739746026e-05, + "loss": 1.2595, + "step": 3778 + }, + { + "epoch": 0.09, + "learning_rate": 1.980631121232431e-05, + "loss": 1.1496, + "step": 3779 + }, + { + "epoch": 0.09, + "learning_rate": 1.9806158625410008e-05, + "loss": 0.9553, + "step": 3780 + }, + { + "epoch": 0.09, + "learning_rate": 1.9806005979004052e-05, + "loss": 1.1052, + "step": 3781 + }, + { + "epoch": 0.09, + "learning_rate": 1.980585327310736e-05, + "loss": 1.3049, + "step": 3782 + }, + { + "epoch": 0.09, + "learning_rate": 1.980570050772087e-05, + "loss": 1.1117, + "step": 3783 + }, + { + "epoch": 0.09, + "learning_rate": 1.9805547682845494e-05, + "loss": 1.283, + "step": 3784 + }, + { + "epoch": 0.09, + "learning_rate": 1.980539479848217e-05, + "loss": 0.9399, + "step": 3785 + }, + { + "epoch": 0.09, + "learning_rate": 1.980524185463182e-05, + "loss": 1.1704, + "step": 3786 + }, + { + "epoch": 0.09, + "learning_rate": 1.9805088851295375e-05, + "loss": 1.1191, + "step": 3787 + }, + { + "epoch": 0.09, + "learning_rate": 1.9804935788473765e-05, + "loss": 1.0827, + "step": 3788 + }, + { + "epoch": 0.09, + "learning_rate": 1.980478266616791e-05, + "loss": 1.3928, + "step": 3789 + }, + { + "epoch": 0.09, + "learning_rate": 1.980462948437875e-05, + "loss": 1.367, + "step": 3790 + }, + { + "epoch": 0.09, + "learning_rate": 1.9804476243107208e-05, + "loss": 1.0554, + "step": 3791 + }, + { + "epoch": 0.09, + "learning_rate": 1.9804322942354216e-05, + "loss": 1.042, + "step": 3792 + }, + { + "epoch": 0.09, + "learning_rate": 1.9804169582120697e-05, + "loss": 1.2142, + "step": 3793 + }, + { + "epoch": 0.09, + "learning_rate": 1.9804016162407593e-05, + "loss": 1.1082, + "step": 3794 + }, + { + "epoch": 0.09, + "learning_rate": 1.9803862683215828e-05, + "loss": 1.1118, + "step": 3795 + }, + { + "epoch": 0.09, + "learning_rate": 1.980370914454633e-05, + "loss": 1.0937, + "step": 3796 + }, + { + "epoch": 0.09, + "learning_rate": 1.980355554640004e-05, + "loss": 1.5366, + "step": 3797 + }, + { + "epoch": 0.09, + "learning_rate": 1.9803401888777882e-05, + "loss": 1.1193, + "step": 3798 + }, + { + "epoch": 0.09, + "learning_rate": 1.980324817168079e-05, + "loss": 1.0605, + "step": 3799 + }, + { + "epoch": 0.09, + "learning_rate": 1.98030943951097e-05, + "loss": 1.3087, + "step": 3800 + }, + { + "epoch": 0.09, + "learning_rate": 1.980294055906554e-05, + "loss": 1.28, + "step": 3801 + }, + { + "epoch": 0.09, + "learning_rate": 1.9802786663549244e-05, + "loss": 1.14, + "step": 3802 + }, + { + "epoch": 0.09, + "learning_rate": 1.980263270856175e-05, + "loss": 1.2759, + "step": 3803 + }, + { + "epoch": 0.09, + "learning_rate": 1.9802478694103987e-05, + "loss": 1.2957, + "step": 3804 + }, + { + "epoch": 0.09, + "learning_rate": 1.980232462017689e-05, + "loss": 1.1313, + "step": 3805 + }, + { + "epoch": 0.09, + "learning_rate": 1.98021704867814e-05, + "loss": 1.1654, + "step": 3806 + }, + { + "epoch": 0.09, + "learning_rate": 1.9802016293918445e-05, + "loss": 1.3486, + "step": 3807 + }, + { + "epoch": 0.09, + "learning_rate": 1.9801862041588966e-05, + "loss": 1.3425, + "step": 3808 + }, + { + "epoch": 0.09, + "learning_rate": 1.980170772979389e-05, + "loss": 0.8978, + "step": 3809 + }, + { + "epoch": 0.09, + "learning_rate": 1.9801553358534157e-05, + "loss": 1.1838, + "step": 3810 + }, + { + "epoch": 0.09, + "learning_rate": 1.980139892781071e-05, + "loss": 1.1718, + "step": 3811 + }, + { + "epoch": 0.09, + "learning_rate": 1.980124443762448e-05, + "loss": 1.2365, + "step": 3812 + }, + { + "epoch": 0.09, + "learning_rate": 1.9801089887976408e-05, + "loss": 1.1995, + "step": 3813 + }, + { + "epoch": 0.09, + "learning_rate": 1.9800935278867425e-05, + "loss": 1.0761, + "step": 3814 + }, + { + "epoch": 0.09, + "learning_rate": 1.9800780610298474e-05, + "loss": 1.2027, + "step": 3815 + }, + { + "epoch": 0.09, + "learning_rate": 1.9800625882270493e-05, + "loss": 1.2356, + "step": 3816 + }, + { + "epoch": 0.09, + "learning_rate": 1.980047109478442e-05, + "loss": 1.1967, + "step": 3817 + }, + { + "epoch": 0.09, + "learning_rate": 1.9800316247841194e-05, + "loss": 1.0426, + "step": 3818 + }, + { + "epoch": 0.09, + "learning_rate": 1.9800161341441757e-05, + "loss": 1.3202, + "step": 3819 + }, + { + "epoch": 0.09, + "learning_rate": 1.9800006375587043e-05, + "loss": 1.297, + "step": 3820 + }, + { + "epoch": 0.09, + "learning_rate": 1.9799851350277993e-05, + "loss": 1.1181, + "step": 3821 + }, + { + "epoch": 0.09, + "learning_rate": 1.9799696265515554e-05, + "loss": 1.1222, + "step": 3822 + }, + { + "epoch": 0.09, + "learning_rate": 1.979954112130066e-05, + "loss": 1.2047, + "step": 3823 + }, + { + "epoch": 0.09, + "learning_rate": 1.9799385917634256e-05, + "loss": 1.2133, + "step": 3824 + }, + { + "epoch": 0.09, + "learning_rate": 1.9799230654517284e-05, + "loss": 1.1684, + "step": 3825 + }, + { + "epoch": 0.09, + "learning_rate": 1.9799075331950687e-05, + "loss": 1.4166, + "step": 3826 + }, + { + "epoch": 0.09, + "learning_rate": 1.97989199499354e-05, + "loss": 1.3304, + "step": 3827 + }, + { + "epoch": 0.09, + "learning_rate": 1.9798764508472373e-05, + "loss": 1.1588, + "step": 3828 + }, + { + "epoch": 0.09, + "learning_rate": 1.9798609007562546e-05, + "loss": 1.0587, + "step": 3829 + }, + { + "epoch": 0.09, + "learning_rate": 1.9798453447206862e-05, + "loss": 1.2371, + "step": 3830 + }, + { + "epoch": 0.09, + "learning_rate": 1.9798297827406267e-05, + "loss": 1.2273, + "step": 3831 + }, + { + "epoch": 0.09, + "learning_rate": 1.9798142148161703e-05, + "loss": 1.0517, + "step": 3832 + }, + { + "epoch": 0.09, + "learning_rate": 1.9797986409474115e-05, + "loss": 1.2337, + "step": 3833 + }, + { + "epoch": 0.09, + "learning_rate": 1.9797830611344448e-05, + "loss": 1.6159, + "step": 3834 + }, + { + "epoch": 0.09, + "learning_rate": 1.979767475377365e-05, + "loss": 1.1551, + "step": 3835 + }, + { + "epoch": 0.09, + "learning_rate": 1.979751883676266e-05, + "loss": 1.1244, + "step": 3836 + }, + { + "epoch": 0.09, + "learning_rate": 1.979736286031243e-05, + "loss": 1.0045, + "step": 3837 + }, + { + "epoch": 0.09, + "learning_rate": 1.9797206824423904e-05, + "loss": 1.3017, + "step": 3838 + }, + { + "epoch": 0.09, + "learning_rate": 1.979705072909803e-05, + "loss": 1.1421, + "step": 3839 + }, + { + "epoch": 0.09, + "learning_rate": 1.979689457433575e-05, + "loss": 1.1579, + "step": 3840 + }, + { + "epoch": 0.09, + "learning_rate": 1.9796738360138016e-05, + "loss": 0.998, + "step": 3841 + }, + { + "epoch": 0.09, + "learning_rate": 1.9796582086505774e-05, + "loss": 1.1833, + "step": 3842 + }, + { + "epoch": 0.09, + "learning_rate": 1.9796425753439975e-05, + "loss": 1.0951, + "step": 3843 + }, + { + "epoch": 0.09, + "learning_rate": 1.9796269360941563e-05, + "loss": 1.4658, + "step": 3844 + }, + { + "epoch": 0.09, + "learning_rate": 1.979611290901149e-05, + "loss": 1.2243, + "step": 3845 + }, + { + "epoch": 0.09, + "learning_rate": 1.9795956397650708e-05, + "loss": 1.1901, + "step": 3846 + }, + { + "epoch": 0.09, + "learning_rate": 1.9795799826860155e-05, + "loss": 1.2244, + "step": 3847 + }, + { + "epoch": 0.09, + "learning_rate": 1.9795643196640794e-05, + "loss": 1.0512, + "step": 3848 + }, + { + "epoch": 0.09, + "learning_rate": 1.9795486506993572e-05, + "loss": 1.2165, + "step": 3849 + }, + { + "epoch": 0.09, + "learning_rate": 1.9795329757919433e-05, + "loss": 1.2181, + "step": 3850 + }, + { + "epoch": 0.09, + "learning_rate": 1.9795172949419332e-05, + "loss": 1.0975, + "step": 3851 + }, + { + "epoch": 0.09, + "learning_rate": 1.9795016081494222e-05, + "loss": 1.0437, + "step": 3852 + }, + { + "epoch": 0.09, + "learning_rate": 1.979485915414505e-05, + "loss": 1.3006, + "step": 3853 + }, + { + "epoch": 0.09, + "learning_rate": 1.9794702167372776e-05, + "loss": 1.2004, + "step": 3854 + }, + { + "epoch": 0.09, + "learning_rate": 1.9794545121178347e-05, + "loss": 1.0403, + "step": 3855 + }, + { + "epoch": 0.09, + "learning_rate": 1.9794388015562718e-05, + "loss": 1.0316, + "step": 3856 + }, + { + "epoch": 0.09, + "learning_rate": 1.979423085052684e-05, + "loss": 1.3162, + "step": 3857 + }, + { + "epoch": 0.09, + "learning_rate": 1.9794073626071664e-05, + "loss": 1.0036, + "step": 3858 + }, + { + "epoch": 0.09, + "learning_rate": 1.979391634219815e-05, + "loss": 0.9962, + "step": 3859 + }, + { + "epoch": 0.09, + "learning_rate": 1.9793758998907248e-05, + "loss": 1.2321, + "step": 3860 + }, + { + "epoch": 0.09, + "learning_rate": 1.9793601596199914e-05, + "loss": 1.1281, + "step": 3861 + }, + { + "epoch": 0.09, + "learning_rate": 1.9793444134077102e-05, + "loss": 0.8681, + "step": 3862 + }, + { + "epoch": 0.09, + "learning_rate": 1.9793286612539768e-05, + "loss": 1.042, + "step": 3863 + }, + { + "epoch": 0.09, + "learning_rate": 1.979312903158887e-05, + "loss": 1.1559, + "step": 3864 + }, + { + "epoch": 0.09, + "learning_rate": 1.979297139122536e-05, + "loss": 1.3863, + "step": 3865 + }, + { + "epoch": 0.09, + "learning_rate": 1.9792813691450193e-05, + "loss": 1.0326, + "step": 3866 + }, + { + "epoch": 0.09, + "learning_rate": 1.9792655932264334e-05, + "loss": 1.3268, + "step": 3867 + }, + { + "epoch": 0.09, + "learning_rate": 1.9792498113668733e-05, + "loss": 1.2537, + "step": 3868 + }, + { + "epoch": 0.09, + "learning_rate": 1.979234023566435e-05, + "loss": 1.4693, + "step": 3869 + }, + { + "epoch": 0.09, + "learning_rate": 1.979218229825214e-05, + "loss": 1.1266, + "step": 3870 + }, + { + "epoch": 0.09, + "learning_rate": 1.9792024301433065e-05, + "loss": 1.3904, + "step": 3871 + }, + { + "epoch": 0.09, + "learning_rate": 1.979186624520808e-05, + "loss": 1.1334, + "step": 3872 + }, + { + "epoch": 0.09, + "learning_rate": 1.9791708129578146e-05, + "loss": 0.9262, + "step": 3873 + }, + { + "epoch": 0.09, + "learning_rate": 1.9791549954544226e-05, + "loss": 1.2408, + "step": 3874 + }, + { + "epoch": 0.09, + "learning_rate": 1.979139172010727e-05, + "loss": 1.1711, + "step": 3875 + }, + { + "epoch": 0.09, + "learning_rate": 1.9791233426268245e-05, + "loss": 1.1925, + "step": 3876 + }, + { + "epoch": 0.09, + "learning_rate": 1.979107507302811e-05, + "loss": 1.0765, + "step": 3877 + }, + { + "epoch": 0.09, + "learning_rate": 1.9790916660387825e-05, + "loss": 1.1737, + "step": 3878 + }, + { + "epoch": 0.09, + "learning_rate": 1.9790758188348356e-05, + "loss": 1.0348, + "step": 3879 + }, + { + "epoch": 0.09, + "learning_rate": 1.9790599656910653e-05, + "loss": 1.0955, + "step": 3880 + }, + { + "epoch": 0.09, + "learning_rate": 1.9790441066075693e-05, + "loss": 1.1994, + "step": 3881 + }, + { + "epoch": 0.09, + "learning_rate": 1.9790282415844425e-05, + "loss": 1.0438, + "step": 3882 + }, + { + "epoch": 0.09, + "learning_rate": 1.9790123706217814e-05, + "loss": 1.4787, + "step": 3883 + }, + { + "epoch": 0.09, + "learning_rate": 1.978996493719683e-05, + "loss": 1.2399, + "step": 3884 + }, + { + "epoch": 0.09, + "learning_rate": 1.9789806108782427e-05, + "loss": 1.098, + "step": 3885 + }, + { + "epoch": 0.09, + "learning_rate": 1.9789647220975578e-05, + "loss": 1.1567, + "step": 3886 + }, + { + "epoch": 0.09, + "learning_rate": 1.978948827377724e-05, + "loss": 1.338, + "step": 3887 + }, + { + "epoch": 0.09, + "learning_rate": 1.9789329267188378e-05, + "loss": 1.1434, + "step": 3888 + }, + { + "epoch": 0.09, + "learning_rate": 1.9789170201209957e-05, + "loss": 1.0197, + "step": 3889 + }, + { + "epoch": 0.09, + "learning_rate": 1.9789011075842947e-05, + "loss": 1.2382, + "step": 3890 + }, + { + "epoch": 0.09, + "learning_rate": 1.9788851891088305e-05, + "loss": 1.2201, + "step": 3891 + }, + { + "epoch": 0.09, + "learning_rate": 1.9788692646947e-05, + "loss": 1.0249, + "step": 3892 + }, + { + "epoch": 0.09, + "learning_rate": 1.9788533343420004e-05, + "loss": 0.93, + "step": 3893 + }, + { + "epoch": 0.09, + "learning_rate": 1.978837398050828e-05, + "loss": 1.1505, + "step": 3894 + }, + { + "epoch": 0.09, + "learning_rate": 1.9788214558212788e-05, + "loss": 0.8855, + "step": 3895 + }, + { + "epoch": 0.09, + "learning_rate": 1.9788055076534504e-05, + "loss": 1.3018, + "step": 3896 + }, + { + "epoch": 0.09, + "learning_rate": 1.978789553547439e-05, + "loss": 1.1477, + "step": 3897 + }, + { + "epoch": 0.09, + "learning_rate": 1.9787735935033415e-05, + "loss": 1.3216, + "step": 3898 + }, + { + "epoch": 0.09, + "learning_rate": 1.978757627521255e-05, + "loss": 1.2212, + "step": 3899 + }, + { + "epoch": 0.09, + "learning_rate": 1.9787416556012766e-05, + "loss": 1.3089, + "step": 3900 + }, + { + "epoch": 0.09, + "learning_rate": 1.978725677743502e-05, + "loss": 0.9821, + "step": 3901 + }, + { + "epoch": 0.09, + "learning_rate": 1.9787096939480296e-05, + "loss": 1.0367, + "step": 3902 + }, + { + "epoch": 0.09, + "learning_rate": 1.9786937042149552e-05, + "loss": 1.4102, + "step": 3903 + }, + { + "epoch": 0.09, + "learning_rate": 1.9786777085443766e-05, + "loss": 1.2756, + "step": 3904 + }, + { + "epoch": 0.09, + "learning_rate": 1.9786617069363905e-05, + "loss": 1.1611, + "step": 3905 + }, + { + "epoch": 0.09, + "learning_rate": 1.9786456993910938e-05, + "loss": 1.0985, + "step": 3906 + }, + { + "epoch": 0.09, + "learning_rate": 1.978629685908584e-05, + "loss": 0.9221, + "step": 3907 + }, + { + "epoch": 0.09, + "learning_rate": 1.9786136664889583e-05, + "loss": 1.1814, + "step": 3908 + }, + { + "epoch": 0.09, + "learning_rate": 1.9785976411323134e-05, + "loss": 1.242, + "step": 3909 + }, + { + "epoch": 0.09, + "learning_rate": 1.9785816098387468e-05, + "loss": 1.1125, + "step": 3910 + }, + { + "epoch": 0.09, + "learning_rate": 1.978565572608356e-05, + "loss": 1.2218, + "step": 3911 + }, + { + "epoch": 0.09, + "learning_rate": 1.9785495294412378e-05, + "loss": 1.2168, + "step": 3912 + }, + { + "epoch": 0.09, + "learning_rate": 1.97853348033749e-05, + "loss": 1.2907, + "step": 3913 + }, + { + "epoch": 0.09, + "learning_rate": 1.9785174252972092e-05, + "loss": 1.3277, + "step": 3914 + }, + { + "epoch": 0.09, + "learning_rate": 1.978501364320494e-05, + "loss": 1.1172, + "step": 3915 + }, + { + "epoch": 0.09, + "learning_rate": 1.9784852974074408e-05, + "loss": 1.2806, + "step": 3916 + }, + { + "epoch": 0.09, + "learning_rate": 1.9784692245581474e-05, + "loss": 1.0796, + "step": 3917 + }, + { + "epoch": 0.09, + "learning_rate": 1.9784531457727116e-05, + "loss": 1.2569, + "step": 3918 + }, + { + "epoch": 0.09, + "learning_rate": 1.9784370610512303e-05, + "loss": 1.1661, + "step": 3919 + }, + { + "epoch": 0.09, + "learning_rate": 1.978420970393802e-05, + "loss": 1.4259, + "step": 3920 + }, + { + "epoch": 0.09, + "learning_rate": 1.9784048738005238e-05, + "loss": 1.2682, + "step": 3921 + }, + { + "epoch": 0.09, + "learning_rate": 1.978388771271493e-05, + "loss": 1.3335, + "step": 3922 + }, + { + "epoch": 0.09, + "learning_rate": 1.978372662806808e-05, + "loss": 1.1475, + "step": 3923 + }, + { + "epoch": 0.09, + "learning_rate": 1.9783565484065657e-05, + "loss": 1.2843, + "step": 3924 + }, + { + "epoch": 0.09, + "learning_rate": 1.978340428070865e-05, + "loss": 1.367, + "step": 3925 + }, + { + "epoch": 0.09, + "learning_rate": 1.978324301799803e-05, + "loss": 0.9308, + "step": 3926 + }, + { + "epoch": 0.09, + "learning_rate": 1.978308169593477e-05, + "loss": 1.3539, + "step": 3927 + }, + { + "epoch": 0.09, + "learning_rate": 1.978292031451986e-05, + "loss": 1.1337, + "step": 3928 + }, + { + "epoch": 0.09, + "learning_rate": 1.9782758873754267e-05, + "loss": 1.2716, + "step": 3929 + }, + { + "epoch": 0.09, + "learning_rate": 1.978259737363898e-05, + "loss": 1.3227, + "step": 3930 + }, + { + "epoch": 0.09, + "learning_rate": 1.978243581417498e-05, + "loss": 1.1028, + "step": 3931 + }, + { + "epoch": 0.09, + "learning_rate": 1.9782274195363237e-05, + "loss": 1.2408, + "step": 3932 + }, + { + "epoch": 0.09, + "learning_rate": 1.9782112517204743e-05, + "loss": 1.1708, + "step": 3933 + }, + { + "epoch": 0.09, + "learning_rate": 1.978195077970047e-05, + "loss": 1.1042, + "step": 3934 + }, + { + "epoch": 0.09, + "learning_rate": 1.97817889828514e-05, + "loss": 1.3475, + "step": 3935 + }, + { + "epoch": 0.09, + "learning_rate": 1.978162712665852e-05, + "loss": 1.2278, + "step": 3936 + }, + { + "epoch": 0.09, + "learning_rate": 1.9781465211122807e-05, + "loss": 1.2416, + "step": 3937 + }, + { + "epoch": 0.09, + "learning_rate": 1.978130323624525e-05, + "loss": 1.1578, + "step": 3938 + }, + { + "epoch": 0.09, + "learning_rate": 1.9781141202026823e-05, + "loss": 1.1103, + "step": 3939 + }, + { + "epoch": 0.09, + "learning_rate": 1.9780979108468513e-05, + "loss": 1.3764, + "step": 3940 + }, + { + "epoch": 0.09, + "learning_rate": 1.9780816955571304e-05, + "loss": 1.302, + "step": 3941 + }, + { + "epoch": 0.09, + "learning_rate": 1.978065474333618e-05, + "loss": 1.1719, + "step": 3942 + }, + { + "epoch": 0.09, + "learning_rate": 1.9780492471764124e-05, + "loss": 1.3542, + "step": 3943 + }, + { + "epoch": 0.09, + "learning_rate": 1.978033014085612e-05, + "loss": 1.2339, + "step": 3944 + }, + { + "epoch": 0.09, + "learning_rate": 1.9780167750613155e-05, + "loss": 1.3063, + "step": 3945 + }, + { + "epoch": 0.09, + "learning_rate": 1.9780005301036208e-05, + "loss": 1.346, + "step": 3946 + }, + { + "epoch": 0.09, + "learning_rate": 1.9779842792126278e-05, + "loss": 1.1977, + "step": 3947 + }, + { + "epoch": 0.09, + "learning_rate": 1.9779680223884335e-05, + "loss": 1.219, + "step": 3948 + }, + { + "epoch": 0.09, + "learning_rate": 1.9779517596311374e-05, + "loss": 0.9191, + "step": 3949 + }, + { + "epoch": 0.09, + "learning_rate": 1.977935490940838e-05, + "loss": 1.0749, + "step": 3950 + }, + { + "epoch": 0.1, + "learning_rate": 1.977919216317634e-05, + "loss": 1.271, + "step": 3951 + }, + { + "epoch": 0.1, + "learning_rate": 1.977902935761624e-05, + "loss": 1.0308, + "step": 3952 + }, + { + "epoch": 0.1, + "learning_rate": 1.9778866492729076e-05, + "loss": 0.9246, + "step": 3953 + }, + { + "epoch": 0.1, + "learning_rate": 1.977870356851582e-05, + "loss": 1.0893, + "step": 3954 + }, + { + "epoch": 0.1, + "learning_rate": 1.9778540584977474e-05, + "loss": 1.2988, + "step": 3955 + }, + { + "epoch": 0.1, + "learning_rate": 1.9778377542115022e-05, + "loss": 1.2406, + "step": 3956 + }, + { + "epoch": 0.1, + "learning_rate": 1.9778214439929453e-05, + "loss": 1.2406, + "step": 3957 + }, + { + "epoch": 0.1, + "learning_rate": 1.9778051278421758e-05, + "loss": 1.214, + "step": 3958 + }, + { + "epoch": 0.1, + "learning_rate": 1.9777888057592924e-05, + "loss": 1.0691, + "step": 3959 + }, + { + "epoch": 0.1, + "learning_rate": 1.9777724777443943e-05, + "loss": 1.2578, + "step": 3960 + }, + { + "epoch": 0.1, + "learning_rate": 1.977756143797581e-05, + "loss": 1.0871, + "step": 3961 + }, + { + "epoch": 0.1, + "learning_rate": 1.9777398039189507e-05, + "loss": 0.8711, + "step": 3962 + }, + { + "epoch": 0.1, + "learning_rate": 1.9777234581086032e-05, + "loss": 0.8509, + "step": 3963 + }, + { + "epoch": 0.1, + "learning_rate": 1.9777071063666372e-05, + "loss": 1.1931, + "step": 3964 + }, + { + "epoch": 0.1, + "learning_rate": 1.9776907486931526e-05, + "loss": 1.3237, + "step": 3965 + }, + { + "epoch": 0.1, + "learning_rate": 1.9776743850882476e-05, + "loss": 1.0924, + "step": 3966 + }, + { + "epoch": 0.1, + "learning_rate": 1.9776580155520225e-05, + "loss": 1.1656, + "step": 3967 + }, + { + "epoch": 0.1, + "learning_rate": 1.977641640084576e-05, + "loss": 1.3846, + "step": 3968 + }, + { + "epoch": 0.1, + "learning_rate": 1.9776252586860077e-05, + "loss": 1.3351, + "step": 3969 + }, + { + "epoch": 0.1, + "learning_rate": 1.977608871356417e-05, + "loss": 1.1325, + "step": 3970 + }, + { + "epoch": 0.1, + "learning_rate": 1.977592478095903e-05, + "loss": 1.4411, + "step": 3971 + }, + { + "epoch": 0.1, + "learning_rate": 1.9775760789045656e-05, + "loss": 1.0918, + "step": 3972 + }, + { + "epoch": 0.1, + "learning_rate": 1.977559673782504e-05, + "loss": 1.3796, + "step": 3973 + }, + { + "epoch": 0.1, + "learning_rate": 1.9775432627298175e-05, + "loss": 1.2191, + "step": 3974 + }, + { + "epoch": 0.1, + "learning_rate": 1.9775268457466065e-05, + "loss": 1.2164, + "step": 3975 + }, + { + "epoch": 0.1, + "learning_rate": 1.97751042283297e-05, + "loss": 1.1619, + "step": 3976 + }, + { + "epoch": 0.1, + "learning_rate": 1.9774939939890072e-05, + "loss": 0.9095, + "step": 3977 + }, + { + "epoch": 0.1, + "learning_rate": 1.9774775592148185e-05, + "loss": 1.1155, + "step": 3978 + }, + { + "epoch": 0.1, + "learning_rate": 1.9774611185105033e-05, + "loss": 0.9433, + "step": 3979 + }, + { + "epoch": 0.1, + "learning_rate": 1.9774446718761616e-05, + "loss": 1.1865, + "step": 3980 + }, + { + "epoch": 0.1, + "learning_rate": 1.977428219311893e-05, + "loss": 1.0407, + "step": 3981 + }, + { + "epoch": 0.1, + "learning_rate": 1.9774117608177968e-05, + "loss": 1.2666, + "step": 3982 + }, + { + "epoch": 0.1, + "learning_rate": 1.977395296393974e-05, + "loss": 1.0961, + "step": 3983 + }, + { + "epoch": 0.1, + "learning_rate": 1.9773788260405237e-05, + "loss": 0.9372, + "step": 3984 + }, + { + "epoch": 0.1, + "learning_rate": 1.9773623497575457e-05, + "loss": 1.0628, + "step": 3985 + }, + { + "epoch": 0.1, + "learning_rate": 1.9773458675451406e-05, + "loss": 1.1175, + "step": 3986 + }, + { + "epoch": 0.1, + "learning_rate": 1.9773293794034075e-05, + "loss": 1.2222, + "step": 3987 + }, + { + "epoch": 0.1, + "learning_rate": 1.977312885332447e-05, + "loss": 1.2347, + "step": 3988 + }, + { + "epoch": 0.1, + "learning_rate": 1.9772963853323594e-05, + "loss": 1.0307, + "step": 3989 + }, + { + "epoch": 0.1, + "learning_rate": 1.9772798794032445e-05, + "loss": 1.2879, + "step": 3990 + }, + { + "epoch": 0.1, + "learning_rate": 1.9772633675452022e-05, + "loss": 1.1432, + "step": 3991 + }, + { + "epoch": 0.1, + "learning_rate": 1.977246849758333e-05, + "loss": 1.2153, + "step": 3992 + }, + { + "epoch": 0.1, + "learning_rate": 1.9772303260427373e-05, + "loss": 1.3043, + "step": 3993 + }, + { + "epoch": 0.1, + "learning_rate": 1.977213796398515e-05, + "loss": 1.346, + "step": 3994 + }, + { + "epoch": 0.1, + "learning_rate": 1.977197260825766e-05, + "loss": 1.1596, + "step": 3995 + }, + { + "epoch": 0.1, + "learning_rate": 1.9771807193245913e-05, + "loss": 1.1693, + "step": 3996 + }, + { + "epoch": 0.1, + "learning_rate": 1.9771641718950914e-05, + "loss": 1.15, + "step": 3997 + }, + { + "epoch": 0.1, + "learning_rate": 1.9771476185373658e-05, + "loss": 1.2591, + "step": 3998 + }, + { + "epoch": 0.1, + "learning_rate": 1.9771310592515155e-05, + "loss": 1.3877, + "step": 3999 + }, + { + "epoch": 0.1, + "learning_rate": 1.9771144940376413e-05, + "loss": 1.3465, + "step": 4000 + }, + { + "epoch": 0.1, + "learning_rate": 1.977097922895843e-05, + "loss": 1.1264, + "step": 4001 + }, + { + "epoch": 0.1, + "learning_rate": 1.9770813458262212e-05, + "loss": 1.0788, + "step": 4002 + }, + { + "epoch": 0.1, + "learning_rate": 1.977064762828877e-05, + "loss": 1.1372, + "step": 4003 + }, + { + "epoch": 0.1, + "learning_rate": 1.977048173903911e-05, + "loss": 0.9994, + "step": 4004 + }, + { + "epoch": 0.1, + "learning_rate": 1.977031579051423e-05, + "loss": 0.9487, + "step": 4005 + }, + { + "epoch": 0.1, + "learning_rate": 1.9770149782715143e-05, + "loss": 1.132, + "step": 4006 + }, + { + "epoch": 0.1, + "learning_rate": 1.976998371564286e-05, + "loss": 1.1226, + "step": 4007 + }, + { + "epoch": 0.1, + "learning_rate": 1.9769817589298377e-05, + "loss": 0.9953, + "step": 4008 + }, + { + "epoch": 0.1, + "learning_rate": 1.9769651403682713e-05, + "loss": 1.0062, + "step": 4009 + }, + { + "epoch": 0.1, + "learning_rate": 1.976948515879687e-05, + "loss": 1.0562, + "step": 4010 + }, + { + "epoch": 0.1, + "learning_rate": 1.976931885464186e-05, + "loss": 1.3343, + "step": 4011 + }, + { + "epoch": 0.1, + "learning_rate": 1.976915249121869e-05, + "loss": 1.1165, + "step": 4012 + }, + { + "epoch": 0.1, + "learning_rate": 1.9768986068528372e-05, + "loss": 1.2466, + "step": 4013 + }, + { + "epoch": 0.1, + "learning_rate": 1.9768819586571907e-05, + "loss": 1.2114, + "step": 4014 + }, + { + "epoch": 0.1, + "learning_rate": 1.9768653045350318e-05, + "loss": 1.1687, + "step": 4015 + }, + { + "epoch": 0.1, + "learning_rate": 1.9768486444864605e-05, + "loss": 1.1834, + "step": 4016 + }, + { + "epoch": 0.1, + "learning_rate": 1.9768319785115785e-05, + "loss": 1.2555, + "step": 4017 + }, + { + "epoch": 0.1, + "learning_rate": 1.9768153066104863e-05, + "loss": 1.2701, + "step": 4018 + }, + { + "epoch": 0.1, + "learning_rate": 1.9767986287832857e-05, + "loss": 1.1206, + "step": 4019 + }, + { + "epoch": 0.1, + "learning_rate": 1.9767819450300773e-05, + "loss": 1.103, + "step": 4020 + }, + { + "epoch": 0.1, + "learning_rate": 1.976765255350963e-05, + "loss": 1.1683, + "step": 4021 + }, + { + "epoch": 0.1, + "learning_rate": 1.976748559746043e-05, + "loss": 1.1677, + "step": 4022 + }, + { + "epoch": 0.1, + "learning_rate": 1.9767318582154197e-05, + "loss": 1.3937, + "step": 4023 + }, + { + "epoch": 0.1, + "learning_rate": 1.9767151507591943e-05, + "loss": 1.0474, + "step": 4024 + }, + { + "epoch": 0.1, + "learning_rate": 1.9766984373774672e-05, + "loss": 1.2867, + "step": 4025 + }, + { + "epoch": 0.1, + "learning_rate": 1.976681718070341e-05, + "loss": 1.2069, + "step": 4026 + }, + { + "epoch": 0.1, + "learning_rate": 1.9766649928379158e-05, + "loss": 1.0346, + "step": 4027 + }, + { + "epoch": 0.1, + "learning_rate": 1.9766482616802943e-05, + "loss": 1.229, + "step": 4028 + }, + { + "epoch": 0.1, + "learning_rate": 1.9766315245975774e-05, + "loss": 1.1366, + "step": 4029 + }, + { + "epoch": 0.1, + "learning_rate": 1.9766147815898668e-05, + "loss": 1.2803, + "step": 4030 + }, + { + "epoch": 0.1, + "learning_rate": 1.9765980326572637e-05, + "loss": 1.0652, + "step": 4031 + }, + { + "epoch": 0.1, + "learning_rate": 1.9765812777998704e-05, + "loss": 1.2625, + "step": 4032 + }, + { + "epoch": 0.1, + "learning_rate": 1.976564517017788e-05, + "loss": 0.9341, + "step": 4033 + }, + { + "epoch": 0.1, + "learning_rate": 1.9765477503111187e-05, + "loss": 0.9626, + "step": 4034 + }, + { + "epoch": 0.1, + "learning_rate": 1.9765309776799638e-05, + "loss": 1.0815, + "step": 4035 + }, + { + "epoch": 0.1, + "learning_rate": 1.976514199124425e-05, + "loss": 0.9146, + "step": 4036 + }, + { + "epoch": 0.1, + "learning_rate": 1.9764974146446044e-05, + "loss": 1.3232, + "step": 4037 + }, + { + "epoch": 0.1, + "learning_rate": 1.9764806242406034e-05, + "loss": 1.2539, + "step": 4038 + }, + { + "epoch": 0.1, + "learning_rate": 1.9764638279125245e-05, + "loss": 1.3148, + "step": 4039 + }, + { + "epoch": 0.1, + "learning_rate": 1.976447025660469e-05, + "loss": 1.2934, + "step": 4040 + }, + { + "epoch": 0.1, + "learning_rate": 1.976430217484539e-05, + "loss": 1.3373, + "step": 4041 + }, + { + "epoch": 0.1, + "learning_rate": 1.9764134033848367e-05, + "loss": 1.121, + "step": 4042 + }, + { + "epoch": 0.1, + "learning_rate": 1.976396583361464e-05, + "loss": 1.1297, + "step": 4043 + }, + { + "epoch": 0.1, + "learning_rate": 1.9763797574145227e-05, + "loss": 0.9502, + "step": 4044 + }, + { + "epoch": 0.1, + "learning_rate": 1.9763629255441153e-05, + "loss": 1.362, + "step": 4045 + }, + { + "epoch": 0.1, + "learning_rate": 1.9763460877503435e-05, + "loss": 1.1415, + "step": 4046 + }, + { + "epoch": 0.1, + "learning_rate": 1.9763292440333096e-05, + "loss": 1.0573, + "step": 4047 + }, + { + "epoch": 0.1, + "learning_rate": 1.976312394393116e-05, + "loss": 1.1298, + "step": 4048 + }, + { + "epoch": 0.1, + "learning_rate": 1.9762955388298647e-05, + "loss": 1.1059, + "step": 4049 + }, + { + "epoch": 0.1, + "learning_rate": 1.976278677343658e-05, + "loss": 1.2259, + "step": 4050 + }, + { + "epoch": 0.1, + "learning_rate": 1.976261809934598e-05, + "loss": 1.3154, + "step": 4051 + }, + { + "epoch": 0.1, + "learning_rate": 1.9762449366027873e-05, + "loss": 1.1417, + "step": 4052 + }, + { + "epoch": 0.1, + "learning_rate": 1.9762280573483285e-05, + "loss": 1.2121, + "step": 4053 + }, + { + "epoch": 0.1, + "learning_rate": 1.9762111721713232e-05, + "loss": 1.036, + "step": 4054 + }, + { + "epoch": 0.1, + "learning_rate": 1.9761942810718747e-05, + "loss": 1.5061, + "step": 4055 + }, + { + "epoch": 0.1, + "learning_rate": 1.976177384050085e-05, + "loss": 1.1562, + "step": 4056 + }, + { + "epoch": 0.1, + "learning_rate": 1.9761604811060568e-05, + "loss": 1.1822, + "step": 4057 + }, + { + "epoch": 0.1, + "learning_rate": 1.9761435722398926e-05, + "loss": 1.2586, + "step": 4058 + }, + { + "epoch": 0.1, + "learning_rate": 1.976126657451695e-05, + "loss": 1.2695, + "step": 4059 + }, + { + "epoch": 0.1, + "learning_rate": 1.9761097367415663e-05, + "loss": 1.2015, + "step": 4060 + }, + { + "epoch": 0.1, + "learning_rate": 1.9760928101096095e-05, + "loss": 1.2314, + "step": 4061 + }, + { + "epoch": 0.1, + "learning_rate": 1.9760758775559275e-05, + "loss": 1.2611, + "step": 4062 + }, + { + "epoch": 0.1, + "learning_rate": 1.9760589390806225e-05, + "loss": 1.3313, + "step": 4063 + }, + { + "epoch": 0.1, + "learning_rate": 1.9760419946837972e-05, + "loss": 1.1495, + "step": 4064 + }, + { + "epoch": 0.1, + "learning_rate": 1.9760250443655554e-05, + "loss": 0.9887, + "step": 4065 + }, + { + "epoch": 0.1, + "learning_rate": 1.9760080881259986e-05, + "loss": 1.253, + "step": 4066 + }, + { + "epoch": 0.1, + "learning_rate": 1.9759911259652306e-05, + "loss": 1.142, + "step": 4067 + }, + { + "epoch": 0.1, + "learning_rate": 1.975974157883354e-05, + "loss": 0.9986, + "step": 4068 + }, + { + "epoch": 0.1, + "learning_rate": 1.9759571838804717e-05, + "loss": 0.905, + "step": 4069 + }, + { + "epoch": 0.1, + "learning_rate": 1.9759402039566865e-05, + "loss": 1.0464, + "step": 4070 + }, + { + "epoch": 0.1, + "learning_rate": 1.975923218112102e-05, + "loss": 1.0845, + "step": 4071 + }, + { + "epoch": 0.1, + "learning_rate": 1.9759062263468207e-05, + "loss": 1.2307, + "step": 4072 + }, + { + "epoch": 0.1, + "learning_rate": 1.975889228660946e-05, + "loss": 1.1209, + "step": 4073 + }, + { + "epoch": 0.1, + "learning_rate": 1.9758722250545808e-05, + "loss": 1.0425, + "step": 4074 + }, + { + "epoch": 0.1, + "learning_rate": 1.9758552155278283e-05, + "loss": 1.0062, + "step": 4075 + }, + { + "epoch": 0.1, + "learning_rate": 1.9758382000807915e-05, + "loss": 1.3064, + "step": 4076 + }, + { + "epoch": 0.1, + "learning_rate": 1.9758211787135744e-05, + "loss": 1.0175, + "step": 4077 + }, + { + "epoch": 0.1, + "learning_rate": 1.9758041514262795e-05, + "loss": 1.1763, + "step": 4078 + }, + { + "epoch": 0.1, + "learning_rate": 1.9757871182190102e-05, + "loss": 1.259, + "step": 4079 + }, + { + "epoch": 0.1, + "learning_rate": 1.9757700790918704e-05, + "loss": 1.1916, + "step": 4080 + }, + { + "epoch": 0.1, + "learning_rate": 1.9757530340449627e-05, + "loss": 1.3089, + "step": 4081 + }, + { + "epoch": 0.1, + "learning_rate": 1.9757359830783907e-05, + "loss": 1.2265, + "step": 4082 + }, + { + "epoch": 0.1, + "learning_rate": 1.9757189261922585e-05, + "loss": 1.1126, + "step": 4083 + }, + { + "epoch": 0.1, + "learning_rate": 1.9757018633866688e-05, + "loss": 1.3276, + "step": 4084 + }, + { + "epoch": 0.1, + "learning_rate": 1.975684794661725e-05, + "loss": 1.1261, + "step": 4085 + }, + { + "epoch": 0.1, + "learning_rate": 1.9756677200175316e-05, + "loss": 1.3904, + "step": 4086 + }, + { + "epoch": 0.1, + "learning_rate": 1.9756506394541915e-05, + "loss": 1.0471, + "step": 4087 + }, + { + "epoch": 0.1, + "learning_rate": 1.9756335529718086e-05, + "loss": 1.4049, + "step": 4088 + }, + { + "epoch": 0.1, + "learning_rate": 1.9756164605704862e-05, + "loss": 1.1977, + "step": 4089 + }, + { + "epoch": 0.1, + "learning_rate": 1.9755993622503283e-05, + "loss": 1.2796, + "step": 4090 + }, + { + "epoch": 0.1, + "learning_rate": 1.975582258011438e-05, + "loss": 1.2573, + "step": 4091 + }, + { + "epoch": 0.1, + "learning_rate": 1.9755651478539203e-05, + "loss": 1.01, + "step": 4092 + }, + { + "epoch": 0.1, + "learning_rate": 1.9755480317778782e-05, + "loss": 0.9411, + "step": 4093 + }, + { + "epoch": 0.1, + "learning_rate": 1.975530909783416e-05, + "loss": 1.0417, + "step": 4094 + }, + { + "epoch": 0.1, + "learning_rate": 1.9755137818706368e-05, + "loss": 1.1472, + "step": 4095 + }, + { + "epoch": 0.1, + "learning_rate": 1.9754966480396447e-05, + "loss": 0.9342, + "step": 4096 + }, + { + "epoch": 0.1, + "learning_rate": 1.975479508290544e-05, + "loss": 1.2258, + "step": 4097 + }, + { + "epoch": 0.1, + "learning_rate": 1.9754623626234387e-05, + "loss": 1.2559, + "step": 4098 + }, + { + "epoch": 0.1, + "learning_rate": 1.9754452110384328e-05, + "loss": 1.1215, + "step": 4099 + }, + { + "epoch": 0.1, + "learning_rate": 1.9754280535356302e-05, + "loss": 1.0146, + "step": 4100 + }, + { + "epoch": 0.1, + "learning_rate": 1.975410890115135e-05, + "loss": 0.8902, + "step": 4101 + }, + { + "epoch": 0.1, + "learning_rate": 1.9753937207770513e-05, + "loss": 1.2275, + "step": 4102 + }, + { + "epoch": 0.1, + "learning_rate": 1.9753765455214833e-05, + "loss": 1.0979, + "step": 4103 + }, + { + "epoch": 0.1, + "learning_rate": 1.975359364348535e-05, + "loss": 1.2845, + "step": 4104 + }, + { + "epoch": 0.1, + "learning_rate": 1.9753421772583116e-05, + "loss": 1.2247, + "step": 4105 + }, + { + "epoch": 0.1, + "learning_rate": 1.975324984250916e-05, + "loss": 0.9652, + "step": 4106 + }, + { + "epoch": 0.1, + "learning_rate": 1.9753077853264534e-05, + "loss": 1.0721, + "step": 4107 + }, + { + "epoch": 0.1, + "learning_rate": 1.9752905804850275e-05, + "loss": 1.2831, + "step": 4108 + }, + { + "epoch": 0.1, + "learning_rate": 1.9752733697267434e-05, + "loss": 1.0904, + "step": 4109 + }, + { + "epoch": 0.1, + "learning_rate": 1.975256153051705e-05, + "loss": 1.0685, + "step": 4110 + }, + { + "epoch": 0.1, + "learning_rate": 1.9752389304600174e-05, + "loss": 1.1086, + "step": 4111 + }, + { + "epoch": 0.1, + "learning_rate": 1.9752217019517838e-05, + "loss": 1.1066, + "step": 4112 + }, + { + "epoch": 0.1, + "learning_rate": 1.9752044675271098e-05, + "loss": 1.166, + "step": 4113 + }, + { + "epoch": 0.1, + "learning_rate": 1.9751872271860998e-05, + "loss": 1.1708, + "step": 4114 + }, + { + "epoch": 0.1, + "learning_rate": 1.9751699809288586e-05, + "loss": 1.3147, + "step": 4115 + }, + { + "epoch": 0.1, + "learning_rate": 1.9751527287554898e-05, + "loss": 1.3885, + "step": 4116 + }, + { + "epoch": 0.1, + "learning_rate": 1.9751354706660993e-05, + "loss": 1.1462, + "step": 4117 + }, + { + "epoch": 0.1, + "learning_rate": 1.975118206660791e-05, + "loss": 1.3026, + "step": 4118 + }, + { + "epoch": 0.1, + "learning_rate": 1.97510093673967e-05, + "loss": 1.1665, + "step": 4119 + }, + { + "epoch": 0.1, + "learning_rate": 1.975083660902841e-05, + "loss": 1.2931, + "step": 4120 + }, + { + "epoch": 0.1, + "learning_rate": 1.975066379150409e-05, + "loss": 0.8529, + "step": 4121 + }, + { + "epoch": 0.1, + "learning_rate": 1.975049091482478e-05, + "loss": 1.4119, + "step": 4122 + }, + { + "epoch": 0.1, + "learning_rate": 1.975031797899154e-05, + "loss": 1.1308, + "step": 4123 + }, + { + "epoch": 0.1, + "learning_rate": 1.975014498400541e-05, + "loss": 0.8979, + "step": 4124 + }, + { + "epoch": 0.1, + "learning_rate": 1.974997192986745e-05, + "loss": 1.1651, + "step": 4125 + }, + { + "epoch": 0.1, + "learning_rate": 1.97497988165787e-05, + "loss": 1.1937, + "step": 4126 + }, + { + "epoch": 0.1, + "learning_rate": 1.9749625644140214e-05, + "loss": 0.9872, + "step": 4127 + }, + { + "epoch": 0.1, + "learning_rate": 1.974945241255304e-05, + "loss": 1.0335, + "step": 4128 + }, + { + "epoch": 0.1, + "learning_rate": 1.9749279121818235e-05, + "loss": 1.1847, + "step": 4129 + }, + { + "epoch": 0.1, + "learning_rate": 1.974910577193685e-05, + "loss": 1.0793, + "step": 4130 + }, + { + "epoch": 0.1, + "learning_rate": 1.974893236290993e-05, + "loss": 1.2616, + "step": 4131 + }, + { + "epoch": 0.1, + "learning_rate": 1.974875889473853e-05, + "loss": 1.1061, + "step": 4132 + }, + { + "epoch": 0.1, + "learning_rate": 1.97485853674237e-05, + "loss": 1.2919, + "step": 4133 + }, + { + "epoch": 0.1, + "learning_rate": 1.9748411780966497e-05, + "loss": 0.992, + "step": 4134 + }, + { + "epoch": 0.1, + "learning_rate": 1.9748238135367973e-05, + "loss": 1.1719, + "step": 4135 + }, + { + "epoch": 0.1, + "learning_rate": 1.9748064430629185e-05, + "loss": 1.1444, + "step": 4136 + }, + { + "epoch": 0.1, + "learning_rate": 1.974789066675118e-05, + "loss": 1.2375, + "step": 4137 + }, + { + "epoch": 0.1, + "learning_rate": 1.9747716843735015e-05, + "loss": 1.0607, + "step": 4138 + }, + { + "epoch": 0.1, + "learning_rate": 1.9747542961581744e-05, + "loss": 1.1994, + "step": 4139 + }, + { + "epoch": 0.1, + "learning_rate": 1.9747369020292424e-05, + "loss": 1.2556, + "step": 4140 + }, + { + "epoch": 0.1, + "learning_rate": 1.974719501986811e-05, + "loss": 1.2271, + "step": 4141 + }, + { + "epoch": 0.1, + "learning_rate": 1.9747020960309852e-05, + "loss": 1.0578, + "step": 4142 + }, + { + "epoch": 0.1, + "learning_rate": 1.9746846841618714e-05, + "loss": 0.8795, + "step": 4143 + }, + { + "epoch": 0.1, + "learning_rate": 1.9746672663795748e-05, + "loss": 1.1304, + "step": 4144 + }, + { + "epoch": 0.1, + "learning_rate": 1.9746498426842012e-05, + "loss": 0.983, + "step": 4145 + }, + { + "epoch": 0.1, + "learning_rate": 1.9746324130758563e-05, + "loss": 1.2418, + "step": 4146 + }, + { + "epoch": 0.1, + "learning_rate": 1.9746149775546458e-05, + "loss": 1.1116, + "step": 4147 + }, + { + "epoch": 0.1, + "learning_rate": 1.9745975361206753e-05, + "loss": 1.3168, + "step": 4148 + }, + { + "epoch": 0.1, + "learning_rate": 1.974580088774051e-05, + "loss": 0.8389, + "step": 4149 + }, + { + "epoch": 0.1, + "learning_rate": 1.9745626355148782e-05, + "loss": 0.9652, + "step": 4150 + }, + { + "epoch": 0.1, + "learning_rate": 1.9745451763432634e-05, + "loss": 1.2404, + "step": 4151 + }, + { + "epoch": 0.1, + "learning_rate": 1.974527711259312e-05, + "loss": 1.4466, + "step": 4152 + }, + { + "epoch": 0.1, + "learning_rate": 1.9745102402631308e-05, + "loss": 1.3679, + "step": 4153 + }, + { + "epoch": 0.1, + "learning_rate": 1.9744927633548246e-05, + "loss": 1.1566, + "step": 4154 + }, + { + "epoch": 0.1, + "learning_rate": 1.9744752805345e-05, + "loss": 1.1641, + "step": 4155 + }, + { + "epoch": 0.1, + "learning_rate": 1.9744577918022637e-05, + "loss": 1.3744, + "step": 4156 + }, + { + "epoch": 0.1, + "learning_rate": 1.9744402971582202e-05, + "loss": 1.1508, + "step": 4157 + }, + { + "epoch": 0.1, + "learning_rate": 1.9744227966024776e-05, + "loss": 1.1274, + "step": 4158 + }, + { + "epoch": 0.1, + "learning_rate": 1.9744052901351403e-05, + "loss": 1.2074, + "step": 4159 + }, + { + "epoch": 0.1, + "learning_rate": 1.974387777756316e-05, + "loss": 1.0919, + "step": 4160 + }, + { + "epoch": 0.1, + "learning_rate": 1.9743702594661098e-05, + "loss": 1.2273, + "step": 4161 + }, + { + "epoch": 0.1, + "learning_rate": 1.9743527352646286e-05, + "loss": 1.0971, + "step": 4162 + }, + { + "epoch": 0.1, + "learning_rate": 1.9743352051519783e-05, + "loss": 1.307, + "step": 4163 + }, + { + "epoch": 0.1, + "learning_rate": 1.974317669128266e-05, + "loss": 0.8757, + "step": 4164 + }, + { + "epoch": 0.1, + "learning_rate": 1.974300127193597e-05, + "loss": 1.2304, + "step": 4165 + }, + { + "epoch": 0.1, + "learning_rate": 1.9742825793480785e-05, + "loss": 1.4134, + "step": 4166 + }, + { + "epoch": 0.1, + "learning_rate": 1.9742650255918167e-05, + "loss": 1.2719, + "step": 4167 + }, + { + "epoch": 0.1, + "learning_rate": 1.9742474659249183e-05, + "loss": 0.9023, + "step": 4168 + }, + { + "epoch": 0.1, + "learning_rate": 1.9742299003474898e-05, + "loss": 1.1553, + "step": 4169 + }, + { + "epoch": 0.1, + "learning_rate": 1.9742123288596375e-05, + "loss": 1.0515, + "step": 4170 + }, + { + "epoch": 0.1, + "learning_rate": 1.974194751461468e-05, + "loss": 1.0256, + "step": 4171 + }, + { + "epoch": 0.1, + "learning_rate": 1.974177168153088e-05, + "loss": 1.3173, + "step": 4172 + }, + { + "epoch": 0.1, + "learning_rate": 1.9741595789346047e-05, + "loss": 1.1424, + "step": 4173 + }, + { + "epoch": 0.1, + "learning_rate": 1.974141983806124e-05, + "loss": 1.2499, + "step": 4174 + }, + { + "epoch": 0.1, + "learning_rate": 1.9741243827677534e-05, + "loss": 1.2859, + "step": 4175 + }, + { + "epoch": 0.1, + "learning_rate": 1.974106775819599e-05, + "loss": 1.2603, + "step": 4176 + }, + { + "epoch": 0.1, + "learning_rate": 1.974089162961768e-05, + "loss": 1.1319, + "step": 4177 + }, + { + "epoch": 0.1, + "learning_rate": 1.974071544194367e-05, + "loss": 1.22, + "step": 4178 + }, + { + "epoch": 0.1, + "learning_rate": 1.974053919517503e-05, + "loss": 1.3079, + "step": 4179 + }, + { + "epoch": 0.1, + "learning_rate": 1.9740362889312835e-05, + "loss": 1.1824, + "step": 4180 + }, + { + "epoch": 0.1, + "learning_rate": 1.9740186524358144e-05, + "loss": 1.3085, + "step": 4181 + }, + { + "epoch": 0.1, + "learning_rate": 1.9740010100312036e-05, + "loss": 1.1587, + "step": 4182 + }, + { + "epoch": 0.1, + "learning_rate": 1.9739833617175575e-05, + "loss": 1.3957, + "step": 4183 + }, + { + "epoch": 0.1, + "learning_rate": 1.9739657074949835e-05, + "loss": 1.307, + "step": 4184 + }, + { + "epoch": 0.1, + "learning_rate": 1.973948047363589e-05, + "loss": 1.2213, + "step": 4185 + }, + { + "epoch": 0.1, + "learning_rate": 1.9739303813234802e-05, + "loss": 1.3626, + "step": 4186 + }, + { + "epoch": 0.1, + "learning_rate": 1.973912709374765e-05, + "loss": 0.9698, + "step": 4187 + }, + { + "epoch": 0.1, + "learning_rate": 1.973895031517551e-05, + "loss": 1.0851, + "step": 4188 + }, + { + "epoch": 0.1, + "learning_rate": 1.9738773477519446e-05, + "loss": 1.1663, + "step": 4189 + }, + { + "epoch": 0.1, + "learning_rate": 1.973859658078053e-05, + "loss": 0.8235, + "step": 4190 + }, + { + "epoch": 0.1, + "learning_rate": 1.9738419624959847e-05, + "loss": 1.1193, + "step": 4191 + }, + { + "epoch": 0.1, + "learning_rate": 1.9738242610058457e-05, + "loss": 1.0521, + "step": 4192 + }, + { + "epoch": 0.1, + "learning_rate": 1.973806553607744e-05, + "loss": 1.2095, + "step": 4193 + }, + { + "epoch": 0.1, + "learning_rate": 1.973788840301787e-05, + "loss": 1.3102, + "step": 4194 + }, + { + "epoch": 0.1, + "learning_rate": 1.973771121088082e-05, + "loss": 1.354, + "step": 4195 + }, + { + "epoch": 0.1, + "learning_rate": 1.9737533959667373e-05, + "loss": 1.3056, + "step": 4196 + }, + { + "epoch": 0.1, + "learning_rate": 1.9737356649378593e-05, + "loss": 1.1583, + "step": 4197 + }, + { + "epoch": 0.1, + "learning_rate": 1.973717928001556e-05, + "loss": 0.9625, + "step": 4198 + }, + { + "epoch": 0.1, + "learning_rate": 1.9737001851579353e-05, + "loss": 1.2548, + "step": 4199 + }, + { + "epoch": 0.1, + "learning_rate": 1.9736824364071047e-05, + "loss": 0.8012, + "step": 4200 + }, + { + "epoch": 0.1, + "learning_rate": 1.9736646817491714e-05, + "loss": 1.1276, + "step": 4201 + }, + { + "epoch": 0.1, + "learning_rate": 1.9736469211842437e-05, + "loss": 1.2214, + "step": 4202 + }, + { + "epoch": 0.1, + "learning_rate": 1.973629154712429e-05, + "loss": 1.3026, + "step": 4203 + }, + { + "epoch": 0.1, + "learning_rate": 1.9736113823338357e-05, + "loss": 0.9213, + "step": 4204 + }, + { + "epoch": 0.1, + "learning_rate": 1.9735936040485705e-05, + "loss": 0.9391, + "step": 4205 + }, + { + "epoch": 0.1, + "learning_rate": 1.9735758198567427e-05, + "loss": 1.1893, + "step": 4206 + }, + { + "epoch": 0.1, + "learning_rate": 1.9735580297584585e-05, + "loss": 1.4161, + "step": 4207 + }, + { + "epoch": 0.1, + "learning_rate": 1.9735402337538275e-05, + "loss": 0.9252, + "step": 4208 + }, + { + "epoch": 0.1, + "learning_rate": 1.973522431842956e-05, + "loss": 1.1599, + "step": 4209 + }, + { + "epoch": 0.1, + "learning_rate": 1.973504624025954e-05, + "loss": 1.3324, + "step": 4210 + }, + { + "epoch": 0.1, + "learning_rate": 1.9734868103029275e-05, + "loss": 1.3566, + "step": 4211 + }, + { + "epoch": 0.1, + "learning_rate": 1.9734689906739862e-05, + "loss": 1.4136, + "step": 4212 + }, + { + "epoch": 0.1, + "learning_rate": 1.973451165139237e-05, + "loss": 1.2555, + "step": 4213 + }, + { + "epoch": 0.1, + "learning_rate": 1.9734333336987886e-05, + "loss": 1.1421, + "step": 4214 + }, + { + "epoch": 0.1, + "learning_rate": 1.9734154963527495e-05, + "loss": 1.3242, + "step": 4215 + }, + { + "epoch": 0.1, + "learning_rate": 1.9733976531012272e-05, + "loss": 1.2957, + "step": 4216 + }, + { + "epoch": 0.1, + "learning_rate": 1.9733798039443304e-05, + "loss": 1.2592, + "step": 4217 + }, + { + "epoch": 0.1, + "learning_rate": 1.9733619488821673e-05, + "loss": 1.278, + "step": 4218 + }, + { + "epoch": 0.1, + "learning_rate": 1.973344087914846e-05, + "loss": 1.3851, + "step": 4219 + }, + { + "epoch": 0.1, + "learning_rate": 1.9733262210424752e-05, + "loss": 0.9235, + "step": 4220 + }, + { + "epoch": 0.1, + "learning_rate": 1.9733083482651634e-05, + "loss": 1.1435, + "step": 4221 + }, + { + "epoch": 0.1, + "learning_rate": 1.9732904695830186e-05, + "loss": 1.2525, + "step": 4222 + }, + { + "epoch": 0.1, + "learning_rate": 1.9732725849961496e-05, + "loss": 0.9781, + "step": 4223 + }, + { + "epoch": 0.1, + "learning_rate": 1.9732546945046643e-05, + "loss": 0.978, + "step": 4224 + }, + { + "epoch": 0.1, + "learning_rate": 1.9732367981086723e-05, + "loss": 1.4214, + "step": 4225 + }, + { + "epoch": 0.1, + "learning_rate": 1.9732188958082813e-05, + "loss": 1.1622, + "step": 4226 + }, + { + "epoch": 0.1, + "learning_rate": 1.9732009876035997e-05, + "loss": 1.0346, + "step": 4227 + }, + { + "epoch": 0.1, + "learning_rate": 1.9731830734947373e-05, + "loss": 1.0794, + "step": 4228 + }, + { + "epoch": 0.1, + "learning_rate": 1.9731651534818018e-05, + "loss": 1.1566, + "step": 4229 + }, + { + "epoch": 0.1, + "learning_rate": 1.9731472275649023e-05, + "loss": 1.126, + "step": 4230 + }, + { + "epoch": 0.1, + "learning_rate": 1.9731292957441476e-05, + "loss": 1.1618, + "step": 4231 + }, + { + "epoch": 0.1, + "learning_rate": 1.9731113580196458e-05, + "loss": 1.1877, + "step": 4232 + }, + { + "epoch": 0.1, + "learning_rate": 1.973093414391507e-05, + "loss": 0.9822, + "step": 4233 + }, + { + "epoch": 0.1, + "learning_rate": 1.9730754648598388e-05, + "loss": 1.4443, + "step": 4234 + }, + { + "epoch": 0.1, + "learning_rate": 1.9730575094247508e-05, + "loss": 0.9699, + "step": 4235 + }, + { + "epoch": 0.1, + "learning_rate": 1.973039548086352e-05, + "loss": 1.374, + "step": 4236 + }, + { + "epoch": 0.1, + "learning_rate": 1.9730215808447507e-05, + "loss": 1.2341, + "step": 4237 + }, + { + "epoch": 0.1, + "learning_rate": 1.9730036077000568e-05, + "loss": 1.1143, + "step": 4238 + }, + { + "epoch": 0.1, + "learning_rate": 1.9729856286523787e-05, + "loss": 0.9846, + "step": 4239 + }, + { + "epoch": 0.1, + "learning_rate": 1.9729676437018256e-05, + "loss": 1.0962, + "step": 4240 + }, + { + "epoch": 0.1, + "learning_rate": 1.972949652848507e-05, + "loss": 1.1563, + "step": 4241 + }, + { + "epoch": 0.1, + "learning_rate": 1.972931656092531e-05, + "loss": 1.4695, + "step": 4242 + }, + { + "epoch": 0.1, + "learning_rate": 1.9729136534340078e-05, + "loss": 0.9055, + "step": 4243 + }, + { + "epoch": 0.1, + "learning_rate": 1.9728956448730466e-05, + "loss": 1.1433, + "step": 4244 + }, + { + "epoch": 0.1, + "learning_rate": 1.972877630409756e-05, + "loss": 1.2643, + "step": 4245 + }, + { + "epoch": 0.1, + "learning_rate": 1.972859610044246e-05, + "loss": 1.128, + "step": 4246 + }, + { + "epoch": 0.1, + "learning_rate": 1.972841583776625e-05, + "loss": 1.167, + "step": 4247 + }, + { + "epoch": 0.1, + "learning_rate": 1.9728235516070037e-05, + "loss": 1.143, + "step": 4248 + }, + { + "epoch": 0.1, + "learning_rate": 1.97280551353549e-05, + "loss": 1.0797, + "step": 4249 + }, + { + "epoch": 0.1, + "learning_rate": 1.9727874695621946e-05, + "loss": 1.4772, + "step": 4250 + }, + { + "epoch": 0.1, + "learning_rate": 1.972769419687226e-05, + "loss": 1.2126, + "step": 4251 + }, + { + "epoch": 0.1, + "learning_rate": 1.9727513639106948e-05, + "loss": 1.2583, + "step": 4252 + }, + { + "epoch": 0.1, + "learning_rate": 1.9727333022327094e-05, + "loss": 1.2972, + "step": 4253 + }, + { + "epoch": 0.1, + "learning_rate": 1.9727152346533798e-05, + "loss": 1.1427, + "step": 4254 + }, + { + "epoch": 0.1, + "learning_rate": 1.9726971611728157e-05, + "loss": 1.296, + "step": 4255 + }, + { + "epoch": 0.1, + "learning_rate": 1.972679081791127e-05, + "loss": 1.1868, + "step": 4256 + }, + { + "epoch": 0.1, + "learning_rate": 1.972660996508423e-05, + "loss": 1.0269, + "step": 4257 + }, + { + "epoch": 0.1, + "learning_rate": 1.972642905324813e-05, + "loss": 1.3015, + "step": 4258 + }, + { + "epoch": 0.1, + "learning_rate": 1.9726248082404077e-05, + "loss": 1.168, + "step": 4259 + }, + { + "epoch": 0.1, + "learning_rate": 1.9726067052553167e-05, + "loss": 1.0571, + "step": 4260 + }, + { + "epoch": 0.1, + "learning_rate": 1.9725885963696495e-05, + "loss": 1.3385, + "step": 4261 + }, + { + "epoch": 0.1, + "learning_rate": 1.9725704815835156e-05, + "loss": 1.1957, + "step": 4262 + }, + { + "epoch": 0.1, + "learning_rate": 1.9725523608970257e-05, + "loss": 1.0528, + "step": 4263 + }, + { + "epoch": 0.1, + "learning_rate": 1.9725342343102897e-05, + "loss": 1.2385, + "step": 4264 + }, + { + "epoch": 0.1, + "learning_rate": 1.972516101823417e-05, + "loss": 1.1424, + "step": 4265 + }, + { + "epoch": 0.1, + "learning_rate": 1.9724979634365174e-05, + "loss": 1.2572, + "step": 4266 + }, + { + "epoch": 0.1, + "learning_rate": 1.9724798191497018e-05, + "loss": 1.052, + "step": 4267 + }, + { + "epoch": 0.1, + "learning_rate": 1.9724616689630798e-05, + "loss": 1.2275, + "step": 4268 + }, + { + "epoch": 0.1, + "learning_rate": 1.9724435128767615e-05, + "loss": 1.1372, + "step": 4269 + }, + { + "epoch": 0.1, + "learning_rate": 1.9724253508908574e-05, + "loss": 1.4383, + "step": 4270 + }, + { + "epoch": 0.1, + "learning_rate": 1.9724071830054772e-05, + "loss": 1.095, + "step": 4271 + }, + { + "epoch": 0.1, + "learning_rate": 1.9723890092207317e-05, + "loss": 0.9714, + "step": 4272 + }, + { + "epoch": 0.1, + "learning_rate": 1.9723708295367305e-05, + "loss": 0.9103, + "step": 4273 + }, + { + "epoch": 0.1, + "learning_rate": 1.972352643953584e-05, + "loss": 1.2003, + "step": 4274 + }, + { + "epoch": 0.1, + "learning_rate": 1.972334452471403e-05, + "loss": 1.0267, + "step": 4275 + }, + { + "epoch": 0.1, + "learning_rate": 1.9723162550902975e-05, + "loss": 1.0383, + "step": 4276 + }, + { + "epoch": 0.1, + "learning_rate": 1.972298051810378e-05, + "loss": 1.3918, + "step": 4277 + }, + { + "epoch": 0.1, + "learning_rate": 1.9722798426317552e-05, + "loss": 1.0688, + "step": 4278 + }, + { + "epoch": 0.1, + "learning_rate": 1.9722616275545388e-05, + "loss": 1.3367, + "step": 4279 + }, + { + "epoch": 0.1, + "learning_rate": 1.9722434065788398e-05, + "loss": 1.0373, + "step": 4280 + }, + { + "epoch": 0.1, + "learning_rate": 1.9722251797047692e-05, + "loss": 1.0962, + "step": 4281 + }, + { + "epoch": 0.1, + "learning_rate": 1.972206946932437e-05, + "loss": 1.1708, + "step": 4282 + }, + { + "epoch": 0.1, + "learning_rate": 1.972188708261954e-05, + "loss": 1.2236, + "step": 4283 + }, + { + "epoch": 0.1, + "learning_rate": 1.97217046369343e-05, + "loss": 1.1133, + "step": 4284 + }, + { + "epoch": 0.1, + "learning_rate": 1.9721522132269773e-05, + "loss": 1.046, + "step": 4285 + }, + { + "epoch": 0.1, + "learning_rate": 1.9721339568627055e-05, + "loss": 1.1636, + "step": 4286 + }, + { + "epoch": 0.1, + "learning_rate": 1.9721156946007256e-05, + "loss": 1.0914, + "step": 4287 + }, + { + "epoch": 0.1, + "learning_rate": 1.9720974264411484e-05, + "loss": 1.0646, + "step": 4288 + }, + { + "epoch": 0.1, + "learning_rate": 1.972079152384085e-05, + "loss": 1.2257, + "step": 4289 + }, + { + "epoch": 0.1, + "learning_rate": 1.972060872429646e-05, + "loss": 1.1626, + "step": 4290 + }, + { + "epoch": 0.1, + "learning_rate": 1.9720425865779423e-05, + "loss": 1.1943, + "step": 4291 + }, + { + "epoch": 0.1, + "learning_rate": 1.9720242948290847e-05, + "loss": 1.3366, + "step": 4292 + }, + { + "epoch": 0.1, + "learning_rate": 1.9720059971831843e-05, + "loss": 1.2758, + "step": 4293 + }, + { + "epoch": 0.1, + "learning_rate": 1.9719876936403524e-05, + "loss": 1.0966, + "step": 4294 + }, + { + "epoch": 0.1, + "learning_rate": 1.9719693842006997e-05, + "loss": 1.2493, + "step": 4295 + }, + { + "epoch": 0.1, + "learning_rate": 1.9719510688643374e-05, + "loss": 1.2564, + "step": 4296 + }, + { + "epoch": 0.1, + "learning_rate": 1.9719327476313762e-05, + "loss": 1.1034, + "step": 4297 + }, + { + "epoch": 0.1, + "learning_rate": 1.971914420501928e-05, + "loss": 0.932, + "step": 4298 + }, + { + "epoch": 0.1, + "learning_rate": 1.971896087476104e-05, + "loss": 1.3194, + "step": 4299 + }, + { + "epoch": 0.1, + "learning_rate": 1.9718777485540145e-05, + "loss": 1.1115, + "step": 4300 + }, + { + "epoch": 0.1, + "learning_rate": 1.9718594037357715e-05, + "loss": 1.2929, + "step": 4301 + }, + { + "epoch": 0.1, + "learning_rate": 1.9718410530214857e-05, + "loss": 1.2441, + "step": 4302 + }, + { + "epoch": 0.1, + "learning_rate": 1.9718226964112692e-05, + "loss": 1.3832, + "step": 4303 + }, + { + "epoch": 0.1, + "learning_rate": 1.971804333905233e-05, + "loss": 1.0668, + "step": 4304 + }, + { + "epoch": 0.1, + "learning_rate": 1.9717859655034882e-05, + "loss": 1.0911, + "step": 4305 + }, + { + "epoch": 0.1, + "learning_rate": 1.971767591206147e-05, + "loss": 1.2057, + "step": 4306 + }, + { + "epoch": 0.1, + "learning_rate": 1.97174921101332e-05, + "loss": 1.2518, + "step": 4307 + }, + { + "epoch": 0.1, + "learning_rate": 1.971730824925119e-05, + "loss": 1.366, + "step": 4308 + }, + { + "epoch": 0.1, + "learning_rate": 1.971712432941656e-05, + "loss": 1.1429, + "step": 4309 + }, + { + "epoch": 0.1, + "learning_rate": 1.9716940350630424e-05, + "loss": 1.3722, + "step": 4310 + }, + { + "epoch": 0.1, + "learning_rate": 1.9716756312893893e-05, + "loss": 1.2214, + "step": 4311 + }, + { + "epoch": 0.1, + "learning_rate": 1.9716572216208084e-05, + "loss": 1.3978, + "step": 4312 + }, + { + "epoch": 0.1, + "learning_rate": 1.971638806057412e-05, + "loss": 1.394, + "step": 4313 + }, + { + "epoch": 0.1, + "learning_rate": 1.9716203845993117e-05, + "loss": 0.9331, + "step": 4314 + }, + { + "epoch": 0.1, + "learning_rate": 1.9716019572466192e-05, + "loss": 1.1683, + "step": 4315 + }, + { + "epoch": 0.1, + "learning_rate": 1.9715835239994457e-05, + "loss": 1.0508, + "step": 4316 + }, + { + "epoch": 0.1, + "learning_rate": 1.9715650848579038e-05, + "loss": 1.5015, + "step": 4317 + }, + { + "epoch": 0.1, + "learning_rate": 1.971546639822105e-05, + "loss": 1.0827, + "step": 4318 + }, + { + "epoch": 0.1, + "learning_rate": 1.971528188892161e-05, + "loss": 1.2797, + "step": 4319 + }, + { + "epoch": 0.1, + "learning_rate": 1.971509732068184e-05, + "loss": 1.2657, + "step": 4320 + }, + { + "epoch": 0.1, + "learning_rate": 1.9714912693502865e-05, + "loss": 0.9845, + "step": 4321 + }, + { + "epoch": 0.1, + "learning_rate": 1.9714728007385795e-05, + "loss": 1.1964, + "step": 4322 + }, + { + "epoch": 0.1, + "learning_rate": 1.9714543262331755e-05, + "loss": 1.2619, + "step": 4323 + }, + { + "epoch": 0.1, + "learning_rate": 1.9714358458341868e-05, + "loss": 1.2215, + "step": 4324 + }, + { + "epoch": 0.1, + "learning_rate": 1.971417359541725e-05, + "loss": 1.1717, + "step": 4325 + }, + { + "epoch": 0.1, + "learning_rate": 1.971398867355903e-05, + "loss": 1.2867, + "step": 4326 + }, + { + "epoch": 0.1, + "learning_rate": 1.9713803692768324e-05, + "loss": 1.3909, + "step": 4327 + }, + { + "epoch": 0.1, + "learning_rate": 1.9713618653046258e-05, + "loss": 1.1726, + "step": 4328 + }, + { + "epoch": 0.1, + "learning_rate": 1.971343355439395e-05, + "loss": 1.096, + "step": 4329 + }, + { + "epoch": 0.1, + "learning_rate": 1.9713248396812524e-05, + "loss": 0.9918, + "step": 4330 + }, + { + "epoch": 0.1, + "learning_rate": 1.9713063180303107e-05, + "loss": 1.1161, + "step": 4331 + }, + { + "epoch": 0.1, + "learning_rate": 1.9712877904866818e-05, + "loss": 0.7287, + "step": 4332 + }, + { + "epoch": 0.1, + "learning_rate": 1.9712692570504788e-05, + "loss": 1.0633, + "step": 4333 + }, + { + "epoch": 0.1, + "learning_rate": 1.971250717721813e-05, + "loss": 1.3469, + "step": 4334 + }, + { + "epoch": 0.1, + "learning_rate": 1.971232172500798e-05, + "loss": 1.0052, + "step": 4335 + }, + { + "epoch": 0.1, + "learning_rate": 1.971213621387546e-05, + "loss": 1.4162, + "step": 4336 + }, + { + "epoch": 0.1, + "learning_rate": 1.971195064382169e-05, + "loss": 1.1728, + "step": 4337 + }, + { + "epoch": 0.1, + "learning_rate": 1.9711765014847804e-05, + "loss": 1.2459, + "step": 4338 + }, + { + "epoch": 0.1, + "learning_rate": 1.971157932695492e-05, + "loss": 1.1118, + "step": 4339 + }, + { + "epoch": 0.1, + "learning_rate": 1.9711393580144168e-05, + "loss": 0.9809, + "step": 4340 + }, + { + "epoch": 0.1, + "learning_rate": 1.971120777441668e-05, + "loss": 1.2582, + "step": 4341 + }, + { + "epoch": 0.1, + "learning_rate": 1.9711021909773574e-05, + "loss": 1.2211, + "step": 4342 + }, + { + "epoch": 0.1, + "learning_rate": 1.9710835986215985e-05, + "loss": 1.0333, + "step": 4343 + }, + { + "epoch": 0.1, + "learning_rate": 1.971065000374504e-05, + "loss": 1.4077, + "step": 4344 + }, + { + "epoch": 0.1, + "learning_rate": 1.9710463962361863e-05, + "loss": 1.1102, + "step": 4345 + }, + { + "epoch": 0.1, + "learning_rate": 1.9710277862067583e-05, + "loss": 0.9719, + "step": 4346 + }, + { + "epoch": 0.1, + "learning_rate": 1.9710091702863333e-05, + "loss": 1.0053, + "step": 4347 + }, + { + "epoch": 0.1, + "learning_rate": 1.9709905484750243e-05, + "loss": 1.1231, + "step": 4348 + }, + { + "epoch": 0.1, + "learning_rate": 1.9709719207729437e-05, + "loss": 1.0218, + "step": 4349 + }, + { + "epoch": 0.1, + "learning_rate": 1.970953287180205e-05, + "loss": 0.9018, + "step": 4350 + }, + { + "epoch": 0.1, + "learning_rate": 1.9709346476969213e-05, + "loss": 1.4224, + "step": 4351 + }, + { + "epoch": 0.1, + "learning_rate": 1.9709160023232052e-05, + "loss": 1.1809, + "step": 4352 + }, + { + "epoch": 0.1, + "learning_rate": 1.97089735105917e-05, + "loss": 1.0199, + "step": 4353 + }, + { + "epoch": 0.1, + "learning_rate": 1.9708786939049292e-05, + "loss": 1.0991, + "step": 4354 + }, + { + "epoch": 0.1, + "learning_rate": 1.9708600308605958e-05, + "loss": 1.5482, + "step": 4355 + }, + { + "epoch": 0.1, + "learning_rate": 1.9708413619262825e-05, + "loss": 1.3102, + "step": 4356 + }, + { + "epoch": 0.1, + "learning_rate": 1.9708226871021032e-05, + "loss": 1.4941, + "step": 4357 + }, + { + "epoch": 0.1, + "learning_rate": 1.970804006388171e-05, + "loss": 1.177, + "step": 4358 + }, + { + "epoch": 0.1, + "learning_rate": 1.9707853197845995e-05, + "loss": 1.1155, + "step": 4359 + }, + { + "epoch": 0.1, + "learning_rate": 1.9707666272915016e-05, + "loss": 1.2122, + "step": 4360 + }, + { + "epoch": 0.1, + "learning_rate": 1.970747928908991e-05, + "loss": 1.1873, + "step": 4361 + }, + { + "epoch": 0.1, + "learning_rate": 1.970729224637181e-05, + "loss": 1.3849, + "step": 4362 + }, + { + "epoch": 0.1, + "learning_rate": 1.9707105144761847e-05, + "loss": 1.1954, + "step": 4363 + }, + { + "epoch": 0.1, + "learning_rate": 1.9706917984261168e-05, + "loss": 1.0633, + "step": 4364 + }, + { + "epoch": 0.1, + "learning_rate": 1.9706730764870895e-05, + "loss": 1.1479, + "step": 4365 + }, + { + "epoch": 0.1, + "learning_rate": 1.9706543486592174e-05, + "loss": 1.1117, + "step": 4366 + }, + { + "epoch": 0.11, + "learning_rate": 1.9706356149426134e-05, + "loss": 1.2393, + "step": 4367 + }, + { + "epoch": 0.11, + "learning_rate": 1.9706168753373912e-05, + "loss": 0.9787, + "step": 4368 + }, + { + "epoch": 0.11, + "learning_rate": 1.9705981298436652e-05, + "loss": 1.1354, + "step": 4369 + }, + { + "epoch": 0.11, + "learning_rate": 1.9705793784615487e-05, + "loss": 1.0079, + "step": 4370 + }, + { + "epoch": 0.11, + "learning_rate": 1.970560621191155e-05, + "loss": 1.3889, + "step": 4371 + }, + { + "epoch": 0.11, + "learning_rate": 1.9705418580325984e-05, + "loss": 1.0785, + "step": 4372 + }, + { + "epoch": 0.11, + "learning_rate": 1.9705230889859927e-05, + "loss": 0.9773, + "step": 4373 + }, + { + "epoch": 0.11, + "learning_rate": 1.970504314051452e-05, + "loss": 1.3892, + "step": 4374 + }, + { + "epoch": 0.11, + "learning_rate": 1.970485533229089e-05, + "loss": 1.342, + "step": 4375 + }, + { + "epoch": 0.11, + "learning_rate": 1.9704667465190194e-05, + "loss": 1.0324, + "step": 4376 + }, + { + "epoch": 0.11, + "learning_rate": 1.970447953921356e-05, + "loss": 1.2285, + "step": 4377 + }, + { + "epoch": 0.11, + "learning_rate": 1.970429155436213e-05, + "loss": 0.9799, + "step": 4378 + }, + { + "epoch": 0.11, + "learning_rate": 1.9704103510637053e-05, + "loss": 1.3126, + "step": 4379 + }, + { + "epoch": 0.11, + "learning_rate": 1.9703915408039454e-05, + "loss": 1.1309, + "step": 4380 + }, + { + "epoch": 0.11, + "learning_rate": 1.9703727246570484e-05, + "loss": 1.322, + "step": 4381 + }, + { + "epoch": 0.11, + "learning_rate": 1.9703539026231288e-05, + "loss": 1.2421, + "step": 4382 + }, + { + "epoch": 0.11, + "learning_rate": 1.9703350747023e-05, + "loss": 1.2435, + "step": 4383 + }, + { + "epoch": 0.11, + "learning_rate": 1.9703162408946766e-05, + "loss": 1.1985, + "step": 4384 + }, + { + "epoch": 0.11, + "learning_rate": 1.9702974012003727e-05, + "loss": 0.9954, + "step": 4385 + }, + { + "epoch": 0.11, + "learning_rate": 1.9702785556195026e-05, + "loss": 0.8598, + "step": 4386 + }, + { + "epoch": 0.11, + "learning_rate": 1.970259704152181e-05, + "loss": 1.1997, + "step": 4387 + }, + { + "epoch": 0.11, + "learning_rate": 1.970240846798522e-05, + "loss": 0.8946, + "step": 4388 + }, + { + "epoch": 0.11, + "learning_rate": 1.97022198355864e-05, + "loss": 1.2342, + "step": 4389 + }, + { + "epoch": 0.11, + "learning_rate": 1.970203114432649e-05, + "loss": 1.1897, + "step": 4390 + }, + { + "epoch": 0.11, + "learning_rate": 1.9701842394206643e-05, + "loss": 1.2736, + "step": 4391 + }, + { + "epoch": 0.11, + "learning_rate": 1.9701653585228e-05, + "loss": 0.9338, + "step": 4392 + }, + { + "epoch": 0.11, + "learning_rate": 1.9701464717391708e-05, + "loss": 1.1156, + "step": 4393 + }, + { + "epoch": 0.11, + "learning_rate": 1.9701275790698906e-05, + "loss": 1.0912, + "step": 4394 + }, + { + "epoch": 0.11, + "learning_rate": 1.970108680515075e-05, + "loss": 1.245, + "step": 4395 + }, + { + "epoch": 0.11, + "learning_rate": 1.9700897760748382e-05, + "loss": 1.1675, + "step": 4396 + }, + { + "epoch": 0.11, + "learning_rate": 1.9700708657492948e-05, + "loss": 1.1761, + "step": 4397 + }, + { + "epoch": 0.11, + "learning_rate": 1.9700519495385597e-05, + "loss": 1.2198, + "step": 4398 + }, + { + "epoch": 0.11, + "learning_rate": 1.9700330274427478e-05, + "loss": 0.9571, + "step": 4399 + }, + { + "epoch": 0.11, + "learning_rate": 1.970014099461973e-05, + "loss": 1.0055, + "step": 4400 + }, + { + "epoch": 0.11, + "learning_rate": 1.9699951655963514e-05, + "loss": 1.3965, + "step": 4401 + }, + { + "epoch": 0.11, + "learning_rate": 1.969976225845997e-05, + "loss": 1.0633, + "step": 4402 + }, + { + "epoch": 0.11, + "learning_rate": 1.9699572802110254e-05, + "loss": 1.1633, + "step": 4403 + }, + { + "epoch": 0.11, + "learning_rate": 1.969938328691551e-05, + "loss": 1.2562, + "step": 4404 + }, + { + "epoch": 0.11, + "learning_rate": 1.9699193712876888e-05, + "loss": 1.2241, + "step": 4405 + }, + { + "epoch": 0.11, + "learning_rate": 1.969900407999554e-05, + "loss": 0.9953, + "step": 4406 + }, + { + "epoch": 0.11, + "learning_rate": 1.9698814388272612e-05, + "loss": 1.3472, + "step": 4407 + }, + { + "epoch": 0.11, + "learning_rate": 1.9698624637709263e-05, + "loss": 1.323, + "step": 4408 + }, + { + "epoch": 0.11, + "learning_rate": 1.9698434828306635e-05, + "loss": 1.2235, + "step": 4409 + }, + { + "epoch": 0.11, + "learning_rate": 1.969824496006589e-05, + "loss": 1.3433, + "step": 4410 + }, + { + "epoch": 0.11, + "learning_rate": 1.969805503298817e-05, + "loss": 1.2053, + "step": 4411 + }, + { + "epoch": 0.11, + "learning_rate": 1.9697865047074633e-05, + "loss": 0.9669, + "step": 4412 + }, + { + "epoch": 0.11, + "learning_rate": 1.9697675002326428e-05, + "loss": 1.1918, + "step": 4413 + }, + { + "epoch": 0.11, + "learning_rate": 1.969748489874471e-05, + "loss": 1.4008, + "step": 4414 + }, + { + "epoch": 0.11, + "learning_rate": 1.9697294736330632e-05, + "loss": 1.0009, + "step": 4415 + }, + { + "epoch": 0.11, + "learning_rate": 1.969710451508535e-05, + "loss": 1.2248, + "step": 4416 + }, + { + "epoch": 0.11, + "learning_rate": 1.9696914235010012e-05, + "loss": 1.211, + "step": 4417 + }, + { + "epoch": 0.11, + "learning_rate": 1.9696723896105778e-05, + "loss": 1.2737, + "step": 4418 + }, + { + "epoch": 0.11, + "learning_rate": 1.9696533498373804e-05, + "loss": 1.052, + "step": 4419 + }, + { + "epoch": 0.11, + "learning_rate": 1.969634304181524e-05, + "loss": 1.1215, + "step": 4420 + }, + { + "epoch": 0.11, + "learning_rate": 1.969615252643124e-05, + "loss": 1.2517, + "step": 4421 + }, + { + "epoch": 0.11, + "learning_rate": 1.9695961952222966e-05, + "loss": 1.0625, + "step": 4422 + }, + { + "epoch": 0.11, + "learning_rate": 1.9695771319191572e-05, + "loss": 1.1154, + "step": 4423 + }, + { + "epoch": 0.11, + "learning_rate": 1.9695580627338212e-05, + "loss": 1.3563, + "step": 4424 + }, + { + "epoch": 0.11, + "learning_rate": 1.9695389876664048e-05, + "loss": 1.1972, + "step": 4425 + }, + { + "epoch": 0.11, + "learning_rate": 1.9695199067170236e-05, + "loss": 1.1366, + "step": 4426 + }, + { + "epoch": 0.11, + "learning_rate": 1.9695008198857927e-05, + "loss": 1.1356, + "step": 4427 + }, + { + "epoch": 0.11, + "learning_rate": 1.9694817271728284e-05, + "loss": 1.1634, + "step": 4428 + }, + { + "epoch": 0.11, + "learning_rate": 1.9694626285782466e-05, + "loss": 0.9991, + "step": 4429 + }, + { + "epoch": 0.11, + "learning_rate": 1.969443524102163e-05, + "loss": 1.0852, + "step": 4430 + }, + { + "epoch": 0.11, + "learning_rate": 1.9694244137446938e-05, + "loss": 1.149, + "step": 4431 + }, + { + "epoch": 0.11, + "learning_rate": 1.9694052975059545e-05, + "loss": 1.1659, + "step": 4432 + }, + { + "epoch": 0.11, + "learning_rate": 1.9693861753860612e-05, + "loss": 1.1288, + "step": 4433 + }, + { + "epoch": 0.11, + "learning_rate": 1.96936704738513e-05, + "loss": 1.2029, + "step": 4434 + }, + { + "epoch": 0.11, + "learning_rate": 1.9693479135032772e-05, + "loss": 1.3055, + "step": 4435 + }, + { + "epoch": 0.11, + "learning_rate": 1.9693287737406183e-05, + "loss": 1.1014, + "step": 4436 + }, + { + "epoch": 0.11, + "learning_rate": 1.96930962809727e-05, + "loss": 1.256, + "step": 4437 + }, + { + "epoch": 0.11, + "learning_rate": 1.969290476573348e-05, + "loss": 1.3166, + "step": 4438 + }, + { + "epoch": 0.11, + "learning_rate": 1.9692713191689686e-05, + "loss": 0.8973, + "step": 4439 + }, + { + "epoch": 0.11, + "learning_rate": 1.9692521558842485e-05, + "loss": 1.1969, + "step": 4440 + }, + { + "epoch": 0.11, + "learning_rate": 1.9692329867193033e-05, + "loss": 0.8174, + "step": 4441 + }, + { + "epoch": 0.11, + "learning_rate": 1.9692138116742494e-05, + "loss": 1.4025, + "step": 4442 + }, + { + "epoch": 0.11, + "learning_rate": 1.969194630749203e-05, + "loss": 1.2471, + "step": 4443 + }, + { + "epoch": 0.11, + "learning_rate": 1.9691754439442815e-05, + "loss": 1.285, + "step": 4444 + }, + { + "epoch": 0.11, + "learning_rate": 1.9691562512596e-05, + "loss": 1.1569, + "step": 4445 + }, + { + "epoch": 0.11, + "learning_rate": 1.9691370526952756e-05, + "loss": 1.1345, + "step": 4446 + }, + { + "epoch": 0.11, + "learning_rate": 1.9691178482514248e-05, + "loss": 1.4266, + "step": 4447 + }, + { + "epoch": 0.11, + "learning_rate": 1.969098637928164e-05, + "loss": 1.3314, + "step": 4448 + }, + { + "epoch": 0.11, + "learning_rate": 1.9690794217256095e-05, + "loss": 1.2501, + "step": 4449 + }, + { + "epoch": 0.11, + "learning_rate": 1.9690601996438782e-05, + "loss": 1.4465, + "step": 4450 + }, + { + "epoch": 0.11, + "learning_rate": 1.9690409716830866e-05, + "loss": 1.0168, + "step": 4451 + }, + { + "epoch": 0.11, + "learning_rate": 1.9690217378433514e-05, + "loss": 1.3621, + "step": 4452 + }, + { + "epoch": 0.11, + "learning_rate": 1.969002498124789e-05, + "loss": 1.1588, + "step": 4453 + }, + { + "epoch": 0.11, + "learning_rate": 1.9689832525275166e-05, + "loss": 1.2762, + "step": 4454 + }, + { + "epoch": 0.11, + "learning_rate": 1.968964001051651e-05, + "loss": 1.0669, + "step": 4455 + }, + { + "epoch": 0.11, + "learning_rate": 1.9689447436973083e-05, + "loss": 1.2908, + "step": 4456 + }, + { + "epoch": 0.11, + "learning_rate": 1.9689254804646055e-05, + "loss": 1.1125, + "step": 4457 + }, + { + "epoch": 0.11, + "learning_rate": 1.9689062113536605e-05, + "loss": 1.1454, + "step": 4458 + }, + { + "epoch": 0.11, + "learning_rate": 1.9688869363645887e-05, + "loss": 1.2083, + "step": 4459 + }, + { + "epoch": 0.11, + "learning_rate": 1.968867655497508e-05, + "loss": 1.2053, + "step": 4460 + }, + { + "epoch": 0.11, + "learning_rate": 1.9688483687525354e-05, + "loss": 1.2191, + "step": 4461 + }, + { + "epoch": 0.11, + "learning_rate": 1.9688290761297873e-05, + "loss": 1.4603, + "step": 4462 + }, + { + "epoch": 0.11, + "learning_rate": 1.9688097776293815e-05, + "loss": 1.2506, + "step": 4463 + }, + { + "epoch": 0.11, + "learning_rate": 1.968790473251434e-05, + "loss": 1.0511, + "step": 4464 + }, + { + "epoch": 0.11, + "learning_rate": 1.9687711629960627e-05, + "loss": 1.2024, + "step": 4465 + }, + { + "epoch": 0.11, + "learning_rate": 1.968751846863385e-05, + "loss": 0.9217, + "step": 4466 + }, + { + "epoch": 0.11, + "learning_rate": 1.9687325248535176e-05, + "loss": 1.1547, + "step": 4467 + }, + { + "epoch": 0.11, + "learning_rate": 1.968713196966578e-05, + "loss": 1.1018, + "step": 4468 + }, + { + "epoch": 0.11, + "learning_rate": 1.9686938632026827e-05, + "loss": 0.9525, + "step": 4469 + }, + { + "epoch": 0.11, + "learning_rate": 1.96867452356195e-05, + "loss": 1.2919, + "step": 4470 + }, + { + "epoch": 0.11, + "learning_rate": 1.9686551780444968e-05, + "loss": 0.6065, + "step": 4471 + }, + { + "epoch": 0.11, + "learning_rate": 1.9686358266504406e-05, + "loss": 1.1403, + "step": 4472 + }, + { + "epoch": 0.11, + "learning_rate": 1.9686164693798984e-05, + "loss": 1.2415, + "step": 4473 + }, + { + "epoch": 0.11, + "learning_rate": 1.968597106232988e-05, + "loss": 1.0488, + "step": 4474 + }, + { + "epoch": 0.11, + "learning_rate": 1.9685777372098268e-05, + "loss": 1.3228, + "step": 4475 + }, + { + "epoch": 0.11, + "learning_rate": 1.968558362310532e-05, + "loss": 1.2923, + "step": 4476 + }, + { + "epoch": 0.11, + "learning_rate": 1.968538981535222e-05, + "loss": 1.3928, + "step": 4477 + }, + { + "epoch": 0.11, + "learning_rate": 1.9685195948840137e-05, + "loss": 1.08, + "step": 4478 + }, + { + "epoch": 0.11, + "learning_rate": 1.9685002023570244e-05, + "loss": 1.0403, + "step": 4479 + }, + { + "epoch": 0.11, + "learning_rate": 1.9684808039543727e-05, + "loss": 0.9094, + "step": 4480 + }, + { + "epoch": 0.11, + "learning_rate": 1.9684613996761755e-05, + "loss": 1.2633, + "step": 4481 + }, + { + "epoch": 0.11, + "learning_rate": 1.9684419895225506e-05, + "loss": 1.2603, + "step": 4482 + }, + { + "epoch": 0.11, + "learning_rate": 1.9684225734936162e-05, + "loss": 0.9197, + "step": 4483 + }, + { + "epoch": 0.11, + "learning_rate": 1.9684031515894898e-05, + "loss": 1.2703, + "step": 4484 + }, + { + "epoch": 0.11, + "learning_rate": 1.9683837238102893e-05, + "loss": 0.9746, + "step": 4485 + }, + { + "epoch": 0.11, + "learning_rate": 1.9683642901561324e-05, + "loss": 1.2219, + "step": 4486 + }, + { + "epoch": 0.11, + "learning_rate": 1.9683448506271376e-05, + "loss": 1.2516, + "step": 4487 + }, + { + "epoch": 0.11, + "learning_rate": 1.9683254052234217e-05, + "loss": 1.216, + "step": 4488 + }, + { + "epoch": 0.11, + "learning_rate": 1.9683059539451034e-05, + "loss": 1.1147, + "step": 4489 + }, + { + "epoch": 0.11, + "learning_rate": 1.9682864967923006e-05, + "loss": 1.17, + "step": 4490 + }, + { + "epoch": 0.11, + "learning_rate": 1.9682670337651316e-05, + "loss": 1.3852, + "step": 4491 + }, + { + "epoch": 0.11, + "learning_rate": 1.9682475648637145e-05, + "loss": 1.3759, + "step": 4492 + }, + { + "epoch": 0.11, + "learning_rate": 1.968228090088167e-05, + "loss": 1.1013, + "step": 4493 + }, + { + "epoch": 0.11, + "learning_rate": 1.968208609438607e-05, + "loss": 1.2606, + "step": 4494 + }, + { + "epoch": 0.11, + "learning_rate": 1.9681891229151537e-05, + "loss": 0.7107, + "step": 4495 + }, + { + "epoch": 0.11, + "learning_rate": 1.9681696305179243e-05, + "loss": 1.0955, + "step": 4496 + }, + { + "epoch": 0.11, + "learning_rate": 1.9681501322470377e-05, + "loss": 1.3122, + "step": 4497 + }, + { + "epoch": 0.11, + "learning_rate": 1.9681306281026122e-05, + "loss": 1.218, + "step": 4498 + }, + { + "epoch": 0.11, + "learning_rate": 1.9681111180847654e-05, + "loss": 1.1335, + "step": 4499 + }, + { + "epoch": 0.11, + "learning_rate": 1.968091602193616e-05, + "loss": 1.3235, + "step": 4500 + }, + { + "epoch": 0.11, + "learning_rate": 1.968072080429283e-05, + "loss": 1.0795, + "step": 4501 + }, + { + "epoch": 0.11, + "learning_rate": 1.9680525527918845e-05, + "loss": 1.2385, + "step": 4502 + }, + { + "epoch": 0.11, + "learning_rate": 1.968033019281539e-05, + "loss": 1.0875, + "step": 4503 + }, + { + "epoch": 0.11, + "learning_rate": 1.9680134798983642e-05, + "loss": 1.385, + "step": 4504 + }, + { + "epoch": 0.11, + "learning_rate": 1.9679939346424798e-05, + "loss": 1.3685, + "step": 4505 + }, + { + "epoch": 0.11, + "learning_rate": 1.9679743835140037e-05, + "loss": 0.9537, + "step": 4506 + }, + { + "epoch": 0.11, + "learning_rate": 1.9679548265130547e-05, + "loss": 1.2763, + "step": 4507 + }, + { + "epoch": 0.11, + "learning_rate": 1.967935263639752e-05, + "loss": 1.0877, + "step": 4508 + }, + { + "epoch": 0.11, + "learning_rate": 1.967915694894213e-05, + "loss": 1.1832, + "step": 4509 + }, + { + "epoch": 0.11, + "learning_rate": 1.9678961202765572e-05, + "loss": 1.5043, + "step": 4510 + }, + { + "epoch": 0.11, + "learning_rate": 1.9678765397869034e-05, + "loss": 1.0891, + "step": 4511 + }, + { + "epoch": 0.11, + "learning_rate": 1.9678569534253706e-05, + "loss": 1.1645, + "step": 4512 + }, + { + "epoch": 0.11, + "learning_rate": 1.9678373611920766e-05, + "loss": 1.2987, + "step": 4513 + }, + { + "epoch": 0.11, + "learning_rate": 1.9678177630871418e-05, + "loss": 1.1045, + "step": 4514 + }, + { + "epoch": 0.11, + "learning_rate": 1.967798159110684e-05, + "loss": 1.3272, + "step": 4515 + }, + { + "epoch": 0.11, + "learning_rate": 1.967778549262822e-05, + "loss": 1.1741, + "step": 4516 + }, + { + "epoch": 0.11, + "learning_rate": 1.9677589335436756e-05, + "loss": 1.0491, + "step": 4517 + }, + { + "epoch": 0.11, + "learning_rate": 1.967739311953363e-05, + "loss": 0.9775, + "step": 4518 + }, + { + "epoch": 0.11, + "learning_rate": 1.9677196844920045e-05, + "loss": 1.0767, + "step": 4519 + }, + { + "epoch": 0.11, + "learning_rate": 1.9677000511597175e-05, + "loss": 1.0729, + "step": 4520 + }, + { + "epoch": 0.11, + "learning_rate": 1.967680411956622e-05, + "loss": 1.3233, + "step": 4521 + }, + { + "epoch": 0.11, + "learning_rate": 1.9676607668828372e-05, + "loss": 1.1768, + "step": 4522 + }, + { + "epoch": 0.11, + "learning_rate": 1.967641115938482e-05, + "loss": 1.1725, + "step": 4523 + }, + { + "epoch": 0.11, + "learning_rate": 1.967621459123676e-05, + "loss": 1.2348, + "step": 4524 + }, + { + "epoch": 0.11, + "learning_rate": 1.967601796438538e-05, + "loss": 1.1754, + "step": 4525 + }, + { + "epoch": 0.11, + "learning_rate": 1.9675821278831875e-05, + "loss": 1.1082, + "step": 4526 + }, + { + "epoch": 0.11, + "learning_rate": 1.967562453457744e-05, + "loss": 0.9509, + "step": 4527 + }, + { + "epoch": 0.11, + "learning_rate": 1.9675427731623267e-05, + "loss": 1.3119, + "step": 4528 + }, + { + "epoch": 0.11, + "learning_rate": 1.9675230869970545e-05, + "loss": 1.2952, + "step": 4529 + }, + { + "epoch": 0.11, + "learning_rate": 1.967503394962048e-05, + "loss": 1.2323, + "step": 4530 + }, + { + "epoch": 0.11, + "learning_rate": 1.9674836970574253e-05, + "loss": 1.0404, + "step": 4531 + }, + { + "epoch": 0.11, + "learning_rate": 1.9674639932833068e-05, + "loss": 1.3897, + "step": 4532 + }, + { + "epoch": 0.11, + "learning_rate": 1.967444283639812e-05, + "loss": 1.3218, + "step": 4533 + }, + { + "epoch": 0.11, + "learning_rate": 1.9674245681270604e-05, + "loss": 1.0907, + "step": 4534 + }, + { + "epoch": 0.11, + "learning_rate": 1.9674048467451713e-05, + "loss": 1.1827, + "step": 4535 + }, + { + "epoch": 0.11, + "learning_rate": 1.9673851194942645e-05, + "loss": 0.8911, + "step": 4536 + }, + { + "epoch": 0.11, + "learning_rate": 1.9673653863744598e-05, + "loss": 1.1367, + "step": 4537 + }, + { + "epoch": 0.11, + "learning_rate": 1.9673456473858766e-05, + "loss": 1.2065, + "step": 4538 + }, + { + "epoch": 0.11, + "learning_rate": 1.967325902528635e-05, + "loss": 1.0483, + "step": 4539 + }, + { + "epoch": 0.11, + "learning_rate": 1.967306151802855e-05, + "loss": 1.3061, + "step": 4540 + }, + { + "epoch": 0.11, + "learning_rate": 1.967286395208656e-05, + "loss": 1.2498, + "step": 4541 + }, + { + "epoch": 0.11, + "learning_rate": 1.9672666327461575e-05, + "loss": 1.4686, + "step": 4542 + }, + { + "epoch": 0.11, + "learning_rate": 1.96724686441548e-05, + "loss": 1.2097, + "step": 4543 + }, + { + "epoch": 0.11, + "learning_rate": 1.967227090216744e-05, + "loss": 1.174, + "step": 4544 + }, + { + "epoch": 0.11, + "learning_rate": 1.967207310150068e-05, + "loss": 0.94, + "step": 4545 + }, + { + "epoch": 0.11, + "learning_rate": 1.967187524215573e-05, + "loss": 1.0746, + "step": 4546 + }, + { + "epoch": 0.11, + "learning_rate": 1.9671677324133783e-05, + "loss": 1.0434, + "step": 4547 + }, + { + "epoch": 0.11, + "learning_rate": 1.967147934743605e-05, + "loss": 1.0013, + "step": 4548 + }, + { + "epoch": 0.11, + "learning_rate": 1.9671281312063727e-05, + "loss": 1.3003, + "step": 4549 + }, + { + "epoch": 0.11, + "learning_rate": 1.9671083218018016e-05, + "loss": 1.3966, + "step": 4550 + }, + { + "epoch": 0.11, + "learning_rate": 1.9670885065300115e-05, + "loss": 1.1453, + "step": 4551 + }, + { + "epoch": 0.11, + "learning_rate": 1.967068685391123e-05, + "loss": 1.3101, + "step": 4552 + }, + { + "epoch": 0.11, + "learning_rate": 1.967048858385256e-05, + "loss": 1.2779, + "step": 4553 + }, + { + "epoch": 0.11, + "learning_rate": 1.9670290255125314e-05, + "loss": 1.2516, + "step": 4554 + }, + { + "epoch": 0.11, + "learning_rate": 1.967009186773069e-05, + "loss": 1.0038, + "step": 4555 + }, + { + "epoch": 0.11, + "learning_rate": 1.9669893421669895e-05, + "loss": 0.9594, + "step": 4556 + }, + { + "epoch": 0.11, + "learning_rate": 1.9669694916944125e-05, + "loss": 1.1664, + "step": 4557 + }, + { + "epoch": 0.11, + "learning_rate": 1.96694963535546e-05, + "loss": 1.1026, + "step": 4558 + }, + { + "epoch": 0.11, + "learning_rate": 1.966929773150251e-05, + "loss": 1.2701, + "step": 4559 + }, + { + "epoch": 0.11, + "learning_rate": 1.9669099050789063e-05, + "loss": 1.5062, + "step": 4560 + }, + { + "epoch": 0.11, + "learning_rate": 1.966890031141547e-05, + "loss": 1.14, + "step": 4561 + }, + { + "epoch": 0.11, + "learning_rate": 1.966870151338293e-05, + "loss": 1.1705, + "step": 4562 + }, + { + "epoch": 0.11, + "learning_rate": 1.9668502656692655e-05, + "loss": 1.0894, + "step": 4563 + }, + { + "epoch": 0.11, + "learning_rate": 1.9668303741345845e-05, + "loss": 1.2588, + "step": 4564 + }, + { + "epoch": 0.11, + "learning_rate": 1.9668104767343715e-05, + "loss": 1.3785, + "step": 4565 + }, + { + "epoch": 0.11, + "learning_rate": 1.9667905734687463e-05, + "loss": 0.9962, + "step": 4566 + }, + { + "epoch": 0.11, + "learning_rate": 1.9667706643378303e-05, + "loss": 1.1135, + "step": 4567 + }, + { + "epoch": 0.11, + "learning_rate": 1.966750749341744e-05, + "loss": 1.1107, + "step": 4568 + }, + { + "epoch": 0.11, + "learning_rate": 1.9667308284806083e-05, + "loss": 1.074, + "step": 4569 + }, + { + "epoch": 0.11, + "learning_rate": 1.9667109017545442e-05, + "loss": 1.1572, + "step": 4570 + }, + { + "epoch": 0.11, + "learning_rate": 1.9666909691636724e-05, + "loss": 1.1641, + "step": 4571 + }, + { + "epoch": 0.11, + "learning_rate": 1.966671030708114e-05, + "loss": 1.1116, + "step": 4572 + }, + { + "epoch": 0.11, + "learning_rate": 1.9666510863879897e-05, + "loss": 1.3084, + "step": 4573 + }, + { + "epoch": 0.11, + "learning_rate": 1.9666311362034203e-05, + "loss": 1.2264, + "step": 4574 + }, + { + "epoch": 0.11, + "learning_rate": 1.9666111801545274e-05, + "loss": 0.9438, + "step": 4575 + }, + { + "epoch": 0.11, + "learning_rate": 1.966591218241432e-05, + "loss": 1.0964, + "step": 4576 + }, + { + "epoch": 0.11, + "learning_rate": 1.9665712504642548e-05, + "loss": 1.0827, + "step": 4577 + }, + { + "epoch": 0.11, + "learning_rate": 1.9665512768231175e-05, + "loss": 1.1245, + "step": 4578 + }, + { + "epoch": 0.11, + "learning_rate": 1.9665312973181407e-05, + "loss": 0.9637, + "step": 4579 + }, + { + "epoch": 0.11, + "learning_rate": 1.966511311949446e-05, + "loss": 1.1301, + "step": 4580 + }, + { + "epoch": 0.11, + "learning_rate": 1.9664913207171544e-05, + "loss": 1.322, + "step": 4581 + }, + { + "epoch": 0.11, + "learning_rate": 1.966471323621387e-05, + "loss": 1.1507, + "step": 4582 + }, + { + "epoch": 0.11, + "learning_rate": 1.966451320662266e-05, + "loss": 1.4108, + "step": 4583 + }, + { + "epoch": 0.11, + "learning_rate": 1.966431311839912e-05, + "loss": 1.2133, + "step": 4584 + }, + { + "epoch": 0.11, + "learning_rate": 1.966411297154446e-05, + "loss": 1.3335, + "step": 4585 + }, + { + "epoch": 0.11, + "learning_rate": 1.9663912766059904e-05, + "loss": 1.466, + "step": 4586 + }, + { + "epoch": 0.11, + "learning_rate": 1.9663712501946665e-05, + "loss": 1.2607, + "step": 4587 + }, + { + "epoch": 0.11, + "learning_rate": 1.966351217920595e-05, + "loss": 1.1006, + "step": 4588 + }, + { + "epoch": 0.11, + "learning_rate": 1.9663311797838983e-05, + "loss": 1.2213, + "step": 4589 + }, + { + "epoch": 0.11, + "learning_rate": 1.9663111357846976e-05, + "loss": 1.0894, + "step": 4590 + }, + { + "epoch": 0.11, + "learning_rate": 1.9662910859231145e-05, + "loss": 1.1394, + "step": 4591 + }, + { + "epoch": 0.11, + "learning_rate": 1.9662710301992705e-05, + "loss": 1.3312, + "step": 4592 + }, + { + "epoch": 0.11, + "learning_rate": 1.9662509686132874e-05, + "loss": 1.2261, + "step": 4593 + }, + { + "epoch": 0.11, + "learning_rate": 1.9662309011652874e-05, + "loss": 1.2201, + "step": 4594 + }, + { + "epoch": 0.11, + "learning_rate": 1.9662108278553915e-05, + "loss": 1.1208, + "step": 4595 + }, + { + "epoch": 0.11, + "learning_rate": 1.966190748683722e-05, + "loss": 1.216, + "step": 4596 + }, + { + "epoch": 0.11, + "learning_rate": 1.9661706636504e-05, + "loss": 1.0466, + "step": 4597 + }, + { + "epoch": 0.11, + "learning_rate": 1.9661505727555482e-05, + "loss": 1.2294, + "step": 4598 + }, + { + "epoch": 0.11, + "learning_rate": 1.966130475999288e-05, + "loss": 1.3954, + "step": 4599 + }, + { + "epoch": 0.11, + "learning_rate": 1.9661103733817418e-05, + "loss": 1.063, + "step": 4600 + }, + { + "epoch": 0.11, + "learning_rate": 1.9660902649030308e-05, + "loss": 1.2706, + "step": 4601 + }, + { + "epoch": 0.11, + "learning_rate": 1.9660701505632773e-05, + "loss": 0.9562, + "step": 4602 + }, + { + "epoch": 0.11, + "learning_rate": 1.9660500303626037e-05, + "loss": 1.2149, + "step": 4603 + }, + { + "epoch": 0.11, + "learning_rate": 1.9660299043011316e-05, + "loss": 1.1893, + "step": 4604 + }, + { + "epoch": 0.11, + "learning_rate": 1.9660097723789837e-05, + "loss": 0.8784, + "step": 4605 + }, + { + "epoch": 0.11, + "learning_rate": 1.9659896345962815e-05, + "loss": 1.2204, + "step": 4606 + }, + { + "epoch": 0.11, + "learning_rate": 1.9659694909531474e-05, + "loss": 1.2198, + "step": 4607 + }, + { + "epoch": 0.11, + "learning_rate": 1.9659493414497034e-05, + "loss": 1.2789, + "step": 4608 + }, + { + "epoch": 0.11, + "learning_rate": 1.965929186086072e-05, + "loss": 1.2736, + "step": 4609 + }, + { + "epoch": 0.11, + "learning_rate": 1.965909024862376e-05, + "loss": 1.5329, + "step": 4610 + }, + { + "epoch": 0.11, + "learning_rate": 1.9658888577787367e-05, + "loss": 1.2598, + "step": 4611 + }, + { + "epoch": 0.11, + "learning_rate": 1.9658686848352767e-05, + "loss": 1.2665, + "step": 4612 + }, + { + "epoch": 0.11, + "learning_rate": 1.9658485060321188e-05, + "loss": 1.0351, + "step": 4613 + }, + { + "epoch": 0.11, + "learning_rate": 1.965828321369385e-05, + "loss": 1.2606, + "step": 4614 + }, + { + "epoch": 0.11, + "learning_rate": 1.965808130847198e-05, + "loss": 0.9969, + "step": 4615 + }, + { + "epoch": 0.11, + "learning_rate": 1.9657879344656804e-05, + "loss": 1.3651, + "step": 4616 + }, + { + "epoch": 0.11, + "learning_rate": 1.9657677322249548e-05, + "loss": 1.4032, + "step": 4617 + }, + { + "epoch": 0.11, + "learning_rate": 1.965747524125143e-05, + "loss": 1.1591, + "step": 4618 + }, + { + "epoch": 0.11, + "learning_rate": 1.9657273101663685e-05, + "loss": 1.1634, + "step": 4619 + }, + { + "epoch": 0.11, + "learning_rate": 1.9657070903487534e-05, + "loss": 1.3692, + "step": 4620 + }, + { + "epoch": 0.11, + "learning_rate": 1.9656868646724204e-05, + "loss": 1.1454, + "step": 4621 + }, + { + "epoch": 0.11, + "learning_rate": 1.9656666331374927e-05, + "loss": 1.2608, + "step": 4622 + }, + { + "epoch": 0.11, + "learning_rate": 1.9656463957440923e-05, + "loss": 1.0487, + "step": 4623 + }, + { + "epoch": 0.11, + "learning_rate": 1.9656261524923428e-05, + "loss": 1.1629, + "step": 4624 + }, + { + "epoch": 0.11, + "learning_rate": 1.9656059033823663e-05, + "loss": 1.1626, + "step": 4625 + }, + { + "epoch": 0.11, + "learning_rate": 1.965585648414286e-05, + "loss": 1.3891, + "step": 4626 + }, + { + "epoch": 0.11, + "learning_rate": 1.9655653875882243e-05, + "loss": 1.0268, + "step": 4627 + }, + { + "epoch": 0.11, + "learning_rate": 1.965545120904305e-05, + "loss": 1.4232, + "step": 4628 + }, + { + "epoch": 0.11, + "learning_rate": 1.9655248483626505e-05, + "loss": 1.2646, + "step": 4629 + }, + { + "epoch": 0.11, + "learning_rate": 1.9655045699633836e-05, + "loss": 1.1313, + "step": 4630 + }, + { + "epoch": 0.11, + "learning_rate": 1.9654842857066275e-05, + "loss": 0.9431, + "step": 4631 + }, + { + "epoch": 0.11, + "learning_rate": 1.9654639955925057e-05, + "loss": 1.2578, + "step": 4632 + }, + { + "epoch": 0.11, + "learning_rate": 1.965443699621141e-05, + "loss": 1.086, + "step": 4633 + }, + { + "epoch": 0.11, + "learning_rate": 1.9654233977926557e-05, + "loss": 1.3181, + "step": 4634 + }, + { + "epoch": 0.11, + "learning_rate": 1.9654030901071742e-05, + "loss": 0.9138, + "step": 4635 + }, + { + "epoch": 0.11, + "learning_rate": 1.9653827765648194e-05, + "loss": 0.9526, + "step": 4636 + }, + { + "epoch": 0.11, + "learning_rate": 1.9653624571657142e-05, + "loss": 1.327, + "step": 4637 + }, + { + "epoch": 0.11, + "learning_rate": 1.965342131909982e-05, + "loss": 1.2823, + "step": 4638 + }, + { + "epoch": 0.11, + "learning_rate": 1.9653218007977464e-05, + "loss": 0.9441, + "step": 4639 + }, + { + "epoch": 0.11, + "learning_rate": 1.9653014638291304e-05, + "loss": 1.3103, + "step": 4640 + }, + { + "epoch": 0.11, + "learning_rate": 1.9652811210042574e-05, + "loss": 1.2312, + "step": 4641 + }, + { + "epoch": 0.11, + "learning_rate": 1.9652607723232507e-05, + "loss": 1.0635, + "step": 4642 + }, + { + "epoch": 0.11, + "learning_rate": 1.9652404177862343e-05, + "loss": 1.4106, + "step": 4643 + }, + { + "epoch": 0.11, + "learning_rate": 1.965220057393331e-05, + "loss": 1.206, + "step": 4644 + }, + { + "epoch": 0.11, + "learning_rate": 1.965199691144665e-05, + "loss": 1.1975, + "step": 4645 + }, + { + "epoch": 0.11, + "learning_rate": 1.9651793190403592e-05, + "loss": 1.2233, + "step": 4646 + }, + { + "epoch": 0.11, + "learning_rate": 1.9651589410805375e-05, + "loss": 1.3019, + "step": 4647 + }, + { + "epoch": 0.11, + "learning_rate": 1.9651385572653235e-05, + "loss": 1.1984, + "step": 4648 + }, + { + "epoch": 0.11, + "learning_rate": 1.9651181675948413e-05, + "loss": 1.1857, + "step": 4649 + }, + { + "epoch": 0.11, + "learning_rate": 1.9650977720692138e-05, + "loss": 1.1604, + "step": 4650 + }, + { + "epoch": 0.11, + "learning_rate": 1.9650773706885652e-05, + "loss": 0.8405, + "step": 4651 + }, + { + "epoch": 0.11, + "learning_rate": 1.965056963453019e-05, + "loss": 1.4129, + "step": 4652 + }, + { + "epoch": 0.11, + "learning_rate": 1.9650365503626995e-05, + "loss": 1.3314, + "step": 4653 + }, + { + "epoch": 0.11, + "learning_rate": 1.96501613141773e-05, + "loss": 0.9969, + "step": 4654 + }, + { + "epoch": 0.11, + "learning_rate": 1.964995706618235e-05, + "loss": 1.0376, + "step": 4655 + }, + { + "epoch": 0.11, + "learning_rate": 1.9649752759643377e-05, + "loss": 1.0659, + "step": 4656 + }, + { + "epoch": 0.11, + "learning_rate": 1.964954839456162e-05, + "loss": 1.0353, + "step": 4657 + }, + { + "epoch": 0.11, + "learning_rate": 1.964934397093833e-05, + "loss": 1.1267, + "step": 4658 + }, + { + "epoch": 0.11, + "learning_rate": 1.964913948877474e-05, + "loss": 1.2832, + "step": 4659 + }, + { + "epoch": 0.11, + "learning_rate": 1.9648934948072086e-05, + "loss": 1.0906, + "step": 4660 + }, + { + "epoch": 0.11, + "learning_rate": 1.9648730348831613e-05, + "loss": 1.1621, + "step": 4661 + }, + { + "epoch": 0.11, + "learning_rate": 1.9648525691054563e-05, + "loss": 0.8808, + "step": 4662 + }, + { + "epoch": 0.11, + "learning_rate": 1.964832097474218e-05, + "loss": 0.9044, + "step": 4663 + }, + { + "epoch": 0.11, + "learning_rate": 1.9648116199895703e-05, + "loss": 1.1283, + "step": 4664 + }, + { + "epoch": 0.11, + "learning_rate": 1.964791136651637e-05, + "loss": 1.2062, + "step": 4665 + }, + { + "epoch": 0.11, + "learning_rate": 1.964770647460543e-05, + "loss": 1.3482, + "step": 4666 + }, + { + "epoch": 0.11, + "learning_rate": 1.9647501524164124e-05, + "loss": 1.0819, + "step": 4667 + }, + { + "epoch": 0.11, + "learning_rate": 1.96472965151937e-05, + "loss": 1.2965, + "step": 4668 + }, + { + "epoch": 0.11, + "learning_rate": 1.964709144769539e-05, + "loss": 1.1098, + "step": 4669 + }, + { + "epoch": 0.11, + "learning_rate": 1.964688632167045e-05, + "loss": 1.3647, + "step": 4670 + }, + { + "epoch": 0.11, + "learning_rate": 1.9646681137120117e-05, + "loss": 1.4558, + "step": 4671 + }, + { + "epoch": 0.11, + "learning_rate": 1.964647589404564e-05, + "loss": 1.1437, + "step": 4672 + }, + { + "epoch": 0.11, + "learning_rate": 1.9646270592448264e-05, + "loss": 1.0472, + "step": 4673 + }, + { + "epoch": 0.11, + "learning_rate": 1.964606523232923e-05, + "loss": 1.2385, + "step": 4674 + }, + { + "epoch": 0.11, + "learning_rate": 1.9645859813689792e-05, + "loss": 0.965, + "step": 4675 + }, + { + "epoch": 0.11, + "learning_rate": 1.964565433653119e-05, + "loss": 0.9498, + "step": 4676 + }, + { + "epoch": 0.11, + "learning_rate": 1.964544880085467e-05, + "loss": 1.3493, + "step": 4677 + }, + { + "epoch": 0.11, + "learning_rate": 1.964524320666148e-05, + "loss": 0.9589, + "step": 4678 + }, + { + "epoch": 0.11, + "learning_rate": 1.964503755395287e-05, + "loss": 1.3186, + "step": 4679 + }, + { + "epoch": 0.11, + "learning_rate": 1.9644831842730084e-05, + "loss": 1.0259, + "step": 4680 + }, + { + "epoch": 0.11, + "learning_rate": 1.964462607299437e-05, + "loss": 1.1601, + "step": 4681 + }, + { + "epoch": 0.11, + "learning_rate": 1.9644420244746984e-05, + "loss": 1.4553, + "step": 4682 + }, + { + "epoch": 0.11, + "learning_rate": 1.9644214357989167e-05, + "loss": 1.3479, + "step": 4683 + }, + { + "epoch": 0.11, + "learning_rate": 1.964400841272217e-05, + "loss": 1.1644, + "step": 4684 + }, + { + "epoch": 0.11, + "learning_rate": 1.964380240894724e-05, + "loss": 1.1357, + "step": 4685 + }, + { + "epoch": 0.11, + "learning_rate": 1.9643596346665634e-05, + "loss": 1.0372, + "step": 4686 + }, + { + "epoch": 0.11, + "learning_rate": 1.9643390225878593e-05, + "loss": 0.6651, + "step": 4687 + }, + { + "epoch": 0.11, + "learning_rate": 1.9643184046587373e-05, + "loss": 1.1973, + "step": 4688 + }, + { + "epoch": 0.11, + "learning_rate": 1.9642977808793223e-05, + "loss": 1.2243, + "step": 4689 + }, + { + "epoch": 0.11, + "learning_rate": 1.9642771512497395e-05, + "loss": 1.1084, + "step": 4690 + }, + { + "epoch": 0.11, + "learning_rate": 1.9642565157701138e-05, + "loss": 0.9117, + "step": 4691 + }, + { + "epoch": 0.11, + "learning_rate": 1.964235874440571e-05, + "loss": 1.1983, + "step": 4692 + }, + { + "epoch": 0.11, + "learning_rate": 1.964215227261236e-05, + "loss": 1.2081, + "step": 4693 + }, + { + "epoch": 0.11, + "learning_rate": 1.9641945742322337e-05, + "loss": 1.1236, + "step": 4694 + }, + { + "epoch": 0.11, + "learning_rate": 1.9641739153536897e-05, + "loss": 1.3272, + "step": 4695 + }, + { + "epoch": 0.11, + "learning_rate": 1.9641532506257297e-05, + "loss": 1.3763, + "step": 4696 + }, + { + "epoch": 0.11, + "learning_rate": 1.9641325800484786e-05, + "loss": 1.2763, + "step": 4697 + }, + { + "epoch": 0.11, + "learning_rate": 1.9641119036220616e-05, + "loss": 1.3561, + "step": 4698 + }, + { + "epoch": 0.11, + "learning_rate": 1.9640912213466045e-05, + "loss": 1.0885, + "step": 4699 + }, + { + "epoch": 0.11, + "learning_rate": 1.964070533222233e-05, + "loss": 1.1061, + "step": 4700 + }, + { + "epoch": 0.11, + "learning_rate": 1.9640498392490725e-05, + "loss": 1.0307, + "step": 4701 + }, + { + "epoch": 0.11, + "learning_rate": 1.964029139427248e-05, + "loss": 1.1171, + "step": 4702 + }, + { + "epoch": 0.11, + "learning_rate": 1.9640084337568855e-05, + "loss": 1.2108, + "step": 4703 + }, + { + "epoch": 0.11, + "learning_rate": 1.9639877222381106e-05, + "loss": 1.2053, + "step": 4704 + }, + { + "epoch": 0.11, + "learning_rate": 1.9639670048710492e-05, + "loss": 1.3, + "step": 4705 + }, + { + "epoch": 0.11, + "learning_rate": 1.9639462816558264e-05, + "loss": 1.3425, + "step": 4706 + }, + { + "epoch": 0.11, + "learning_rate": 1.9639255525925685e-05, + "loss": 1.1661, + "step": 4707 + }, + { + "epoch": 0.11, + "learning_rate": 1.963904817681401e-05, + "loss": 1.1471, + "step": 4708 + }, + { + "epoch": 0.11, + "learning_rate": 1.9638840769224492e-05, + "loss": 1.0979, + "step": 4709 + }, + { + "epoch": 0.11, + "learning_rate": 1.96386333031584e-05, + "loss": 1.3125, + "step": 4710 + }, + { + "epoch": 0.11, + "learning_rate": 1.963842577861698e-05, + "loss": 1.1539, + "step": 4711 + }, + { + "epoch": 0.11, + "learning_rate": 1.9638218195601507e-05, + "loss": 1.1949, + "step": 4712 + }, + { + "epoch": 0.11, + "learning_rate": 1.9638010554113228e-05, + "loss": 1.1681, + "step": 4713 + }, + { + "epoch": 0.11, + "learning_rate": 1.9637802854153403e-05, + "loss": 1.2809, + "step": 4714 + }, + { + "epoch": 0.11, + "learning_rate": 1.9637595095723298e-05, + "loss": 1.1832, + "step": 4715 + }, + { + "epoch": 0.11, + "learning_rate": 1.9637387278824168e-05, + "loss": 1.2431, + "step": 4716 + }, + { + "epoch": 0.11, + "learning_rate": 1.9637179403457278e-05, + "loss": 1.1855, + "step": 4717 + }, + { + "epoch": 0.11, + "learning_rate": 1.9636971469623888e-05, + "loss": 1.2719, + "step": 4718 + }, + { + "epoch": 0.11, + "learning_rate": 1.9636763477325255e-05, + "loss": 0.8807, + "step": 4719 + }, + { + "epoch": 0.11, + "learning_rate": 1.9636555426562653e-05, + "loss": 1.0728, + "step": 4720 + }, + { + "epoch": 0.11, + "learning_rate": 1.9636347317337328e-05, + "loss": 1.2316, + "step": 4721 + }, + { + "epoch": 0.11, + "learning_rate": 1.963613914965055e-05, + "loss": 1.1179, + "step": 4722 + }, + { + "epoch": 0.11, + "learning_rate": 1.9635930923503585e-05, + "loss": 1.2267, + "step": 4723 + }, + { + "epoch": 0.11, + "learning_rate": 1.9635722638897697e-05, + "loss": 1.4557, + "step": 4724 + }, + { + "epoch": 0.11, + "learning_rate": 1.963551429583414e-05, + "loss": 1.204, + "step": 4725 + }, + { + "epoch": 0.11, + "learning_rate": 1.963530589431419e-05, + "loss": 1.4509, + "step": 4726 + }, + { + "epoch": 0.11, + "learning_rate": 1.96350974343391e-05, + "loss": 0.9818, + "step": 4727 + }, + { + "epoch": 0.11, + "learning_rate": 1.9634888915910144e-05, + "loss": 0.8814, + "step": 4728 + }, + { + "epoch": 0.11, + "learning_rate": 1.963468033902858e-05, + "loss": 1.3378, + "step": 4729 + }, + { + "epoch": 0.11, + "learning_rate": 1.963447170369568e-05, + "loss": 1.3197, + "step": 4730 + }, + { + "epoch": 0.11, + "learning_rate": 1.9634263009912706e-05, + "loss": 1.1806, + "step": 4731 + }, + { + "epoch": 0.11, + "learning_rate": 1.9634054257680924e-05, + "loss": 1.0464, + "step": 4732 + }, + { + "epoch": 0.11, + "learning_rate": 1.96338454470016e-05, + "loss": 0.9047, + "step": 4733 + }, + { + "epoch": 0.11, + "learning_rate": 1.9633636577876004e-05, + "loss": 1.0932, + "step": 4734 + }, + { + "epoch": 0.11, + "learning_rate": 1.9633427650305396e-05, + "loss": 1.0054, + "step": 4735 + }, + { + "epoch": 0.11, + "learning_rate": 1.963321866429105e-05, + "loss": 1.1644, + "step": 4736 + }, + { + "epoch": 0.11, + "learning_rate": 1.9633009619834236e-05, + "loss": 0.7978, + "step": 4737 + }, + { + "epoch": 0.11, + "learning_rate": 1.9632800516936215e-05, + "loss": 0.9626, + "step": 4738 + }, + { + "epoch": 0.11, + "learning_rate": 1.963259135559826e-05, + "loss": 0.9209, + "step": 4739 + }, + { + "epoch": 0.11, + "learning_rate": 1.9632382135821638e-05, + "loss": 0.9643, + "step": 4740 + }, + { + "epoch": 0.11, + "learning_rate": 1.963217285760762e-05, + "loss": 1.3093, + "step": 4741 + }, + { + "epoch": 0.11, + "learning_rate": 1.9631963520957477e-05, + "loss": 1.0844, + "step": 4742 + }, + { + "epoch": 0.11, + "learning_rate": 1.9631754125872472e-05, + "loss": 1.1826, + "step": 4743 + }, + { + "epoch": 0.11, + "learning_rate": 1.9631544672353886e-05, + "loss": 1.2149, + "step": 4744 + }, + { + "epoch": 0.11, + "learning_rate": 1.963133516040298e-05, + "loss": 1.0785, + "step": 4745 + }, + { + "epoch": 0.11, + "learning_rate": 1.963112559002103e-05, + "loss": 1.1975, + "step": 4746 + }, + { + "epoch": 0.11, + "learning_rate": 1.9630915961209305e-05, + "loss": 1.5266, + "step": 4747 + }, + { + "epoch": 0.11, + "learning_rate": 1.963070627396908e-05, + "loss": 1.0877, + "step": 4748 + }, + { + "epoch": 0.11, + "learning_rate": 1.963049652830163e-05, + "loss": 1.0051, + "step": 4749 + }, + { + "epoch": 0.11, + "learning_rate": 1.9630286724208216e-05, + "loss": 1.4133, + "step": 4750 + }, + { + "epoch": 0.11, + "learning_rate": 1.963007686169012e-05, + "loss": 1.027, + "step": 4751 + }, + { + "epoch": 0.11, + "learning_rate": 1.9629866940748612e-05, + "loss": 1.1609, + "step": 4752 + }, + { + "epoch": 0.11, + "learning_rate": 1.9629656961384966e-05, + "loss": 1.1282, + "step": 4753 + }, + { + "epoch": 0.11, + "learning_rate": 1.9629446923600458e-05, + "loss": 1.2758, + "step": 4754 + }, + { + "epoch": 0.11, + "learning_rate": 1.9629236827396355e-05, + "loss": 1.4009, + "step": 4755 + }, + { + "epoch": 0.11, + "learning_rate": 1.9629026672773946e-05, + "loss": 0.8329, + "step": 4756 + }, + { + "epoch": 0.11, + "learning_rate": 1.962881645973449e-05, + "loss": 1.3018, + "step": 4757 + }, + { + "epoch": 0.11, + "learning_rate": 1.9628606188279273e-05, + "loss": 1.2549, + "step": 4758 + }, + { + "epoch": 0.11, + "learning_rate": 1.9628395858409565e-05, + "loss": 1.2861, + "step": 4759 + }, + { + "epoch": 0.11, + "learning_rate": 1.9628185470126645e-05, + "loss": 1.0659, + "step": 4760 + }, + { + "epoch": 0.11, + "learning_rate": 1.9627975023431788e-05, + "loss": 1.3307, + "step": 4761 + }, + { + "epoch": 0.11, + "learning_rate": 1.9627764518326274e-05, + "loss": 1.3203, + "step": 4762 + }, + { + "epoch": 0.11, + "learning_rate": 1.9627553954811374e-05, + "loss": 1.3233, + "step": 4763 + }, + { + "epoch": 0.11, + "learning_rate": 1.962734333288837e-05, + "loss": 1.1733, + "step": 4764 + }, + { + "epoch": 0.11, + "learning_rate": 1.9627132652558538e-05, + "loss": 1.0741, + "step": 4765 + }, + { + "epoch": 0.11, + "learning_rate": 1.9626921913823156e-05, + "loss": 1.0086, + "step": 4766 + }, + { + "epoch": 0.11, + "learning_rate": 1.9626711116683504e-05, + "loss": 1.1723, + "step": 4767 + }, + { + "epoch": 0.11, + "learning_rate": 1.962650026114086e-05, + "loss": 1.0182, + "step": 4768 + }, + { + "epoch": 0.11, + "learning_rate": 1.9626289347196503e-05, + "loss": 1.2081, + "step": 4769 + }, + { + "epoch": 0.11, + "learning_rate": 1.9626078374851715e-05, + "loss": 0.9506, + "step": 4770 + }, + { + "epoch": 0.11, + "learning_rate": 1.962586734410777e-05, + "loss": 1.0984, + "step": 4771 + }, + { + "epoch": 0.11, + "learning_rate": 1.9625656254965954e-05, + "loss": 1.1581, + "step": 4772 + }, + { + "epoch": 0.11, + "learning_rate": 1.9625445107427542e-05, + "loss": 1.0535, + "step": 4773 + }, + { + "epoch": 0.11, + "learning_rate": 1.962523390149382e-05, + "loss": 1.2524, + "step": 4774 + }, + { + "epoch": 0.11, + "learning_rate": 1.9625022637166074e-05, + "loss": 1.2945, + "step": 4775 + }, + { + "epoch": 0.11, + "learning_rate": 1.9624811314445575e-05, + "loss": 1.3279, + "step": 4776 + }, + { + "epoch": 0.11, + "learning_rate": 1.962459993333361e-05, + "loss": 1.1023, + "step": 4777 + }, + { + "epoch": 0.11, + "learning_rate": 1.9624388493831462e-05, + "loss": 1.082, + "step": 4778 + }, + { + "epoch": 0.11, + "learning_rate": 1.962417699594041e-05, + "loss": 1.5204, + "step": 4779 + }, + { + "epoch": 0.11, + "learning_rate": 1.962396543966174e-05, + "loss": 1.2737, + "step": 4780 + }, + { + "epoch": 0.11, + "learning_rate": 1.962375382499674e-05, + "loss": 1.2124, + "step": 4781 + }, + { + "epoch": 0.12, + "learning_rate": 1.9623542151946683e-05, + "loss": 1.0515, + "step": 4782 + }, + { + "epoch": 0.12, + "learning_rate": 1.9623330420512865e-05, + "loss": 1.1519, + "step": 4783 + }, + { + "epoch": 0.12, + "learning_rate": 1.962311863069656e-05, + "loss": 1.358, + "step": 4784 + }, + { + "epoch": 0.12, + "learning_rate": 1.9622906782499063e-05, + "loss": 1.0475, + "step": 4785 + }, + { + "epoch": 0.12, + "learning_rate": 1.962269487592165e-05, + "loss": 1.096, + "step": 4786 + }, + { + "epoch": 0.12, + "learning_rate": 1.9622482910965613e-05, + "loss": 1.3789, + "step": 4787 + }, + { + "epoch": 0.12, + "learning_rate": 1.9622270887632234e-05, + "loss": 1.3161, + "step": 4788 + }, + { + "epoch": 0.12, + "learning_rate": 1.96220588059228e-05, + "loss": 1.3444, + "step": 4789 + }, + { + "epoch": 0.12, + "learning_rate": 1.9621846665838598e-05, + "loss": 1.027, + "step": 4790 + }, + { + "epoch": 0.12, + "learning_rate": 1.962163446738092e-05, + "loss": 1.1269, + "step": 4791 + }, + { + "epoch": 0.12, + "learning_rate": 1.9621422210551047e-05, + "loss": 1.0663, + "step": 4792 + }, + { + "epoch": 0.12, + "learning_rate": 1.9621209895350267e-05, + "loss": 1.1215, + "step": 4793 + }, + { + "epoch": 0.12, + "learning_rate": 1.962099752177987e-05, + "loss": 1.0642, + "step": 4794 + }, + { + "epoch": 0.12, + "learning_rate": 1.962078508984114e-05, + "loss": 1.15, + "step": 4795 + }, + { + "epoch": 0.12, + "learning_rate": 1.9620572599535378e-05, + "loss": 0.6905, + "step": 4796 + }, + { + "epoch": 0.12, + "learning_rate": 1.962036005086386e-05, + "loss": 1.0548, + "step": 4797 + }, + { + "epoch": 0.12, + "learning_rate": 1.962014744382788e-05, + "loss": 1.1944, + "step": 4798 + }, + { + "epoch": 0.12, + "learning_rate": 1.961993477842873e-05, + "loss": 1.1425, + "step": 4799 + }, + { + "epoch": 0.12, + "learning_rate": 1.9619722054667698e-05, + "loss": 0.946, + "step": 4800 + }, + { + "epoch": 0.12, + "learning_rate": 1.9619509272546073e-05, + "loss": 1.2261, + "step": 4801 + }, + { + "epoch": 0.12, + "learning_rate": 1.961929643206515e-05, + "loss": 1.052, + "step": 4802 + }, + { + "epoch": 0.12, + "learning_rate": 1.961908353322622e-05, + "loss": 1.3559, + "step": 4803 + }, + { + "epoch": 0.12, + "learning_rate": 1.961887057603057e-05, + "loss": 0.8719, + "step": 4804 + }, + { + "epoch": 0.12, + "learning_rate": 1.9618657560479497e-05, + "loss": 1.1661, + "step": 4805 + }, + { + "epoch": 0.12, + "learning_rate": 1.9618444486574287e-05, + "loss": 1.0755, + "step": 4806 + }, + { + "epoch": 0.12, + "learning_rate": 1.9618231354316243e-05, + "loss": 1.2375, + "step": 4807 + }, + { + "epoch": 0.12, + "learning_rate": 1.9618018163706644e-05, + "loss": 1.0344, + "step": 4808 + }, + { + "epoch": 0.12, + "learning_rate": 1.9617804914746795e-05, + "loss": 1.0888, + "step": 4809 + }, + { + "epoch": 0.12, + "learning_rate": 1.9617591607437988e-05, + "loss": 1.1976, + "step": 4810 + }, + { + "epoch": 0.12, + "learning_rate": 1.9617378241781512e-05, + "loss": 1.2272, + "step": 4811 + }, + { + "epoch": 0.12, + "learning_rate": 1.9617164817778666e-05, + "loss": 1.1926, + "step": 4812 + }, + { + "epoch": 0.12, + "learning_rate": 1.9616951335430744e-05, + "loss": 1.2239, + "step": 4813 + }, + { + "epoch": 0.12, + "learning_rate": 1.9616737794739036e-05, + "loss": 0.9656, + "step": 4814 + }, + { + "epoch": 0.12, + "learning_rate": 1.961652419570484e-05, + "loss": 1.0727, + "step": 4815 + }, + { + "epoch": 0.12, + "learning_rate": 1.9616310538329463e-05, + "loss": 1.1331, + "step": 4816 + }, + { + "epoch": 0.12, + "learning_rate": 1.9616096822614186e-05, + "loss": 1.1263, + "step": 4817 + }, + { + "epoch": 0.12, + "learning_rate": 1.961588304856031e-05, + "loss": 1.1229, + "step": 4818 + }, + { + "epoch": 0.12, + "learning_rate": 1.9615669216169138e-05, + "loss": 0.9918, + "step": 4819 + }, + { + "epoch": 0.12, + "learning_rate": 1.961545532544196e-05, + "loss": 1.1389, + "step": 4820 + }, + { + "epoch": 0.12, + "learning_rate": 1.9615241376380076e-05, + "loss": 1.228, + "step": 4821 + }, + { + "epoch": 0.12, + "learning_rate": 1.9615027368984785e-05, + "loss": 1.2474, + "step": 4822 + }, + { + "epoch": 0.12, + "learning_rate": 1.9614813303257385e-05, + "loss": 1.1745, + "step": 4823 + }, + { + "epoch": 0.12, + "learning_rate": 1.9614599179199172e-05, + "loss": 1.1749, + "step": 4824 + }, + { + "epoch": 0.12, + "learning_rate": 1.9614384996811448e-05, + "loss": 1.061, + "step": 4825 + }, + { + "epoch": 0.12, + "learning_rate": 1.961417075609551e-05, + "loss": 1.1913, + "step": 4826 + }, + { + "epoch": 0.12, + "learning_rate": 1.9613956457052665e-05, + "loss": 1.0646, + "step": 4827 + }, + { + "epoch": 0.12, + "learning_rate": 1.9613742099684204e-05, + "loss": 1.1292, + "step": 4828 + }, + { + "epoch": 0.12, + "learning_rate": 1.9613527683991433e-05, + "loss": 0.9165, + "step": 4829 + }, + { + "epoch": 0.12, + "learning_rate": 1.9613313209975645e-05, + "loss": 1.12, + "step": 4830 + }, + { + "epoch": 0.12, + "learning_rate": 1.9613098677638153e-05, + "loss": 1.2958, + "step": 4831 + }, + { + "epoch": 0.12, + "learning_rate": 1.961288408698025e-05, + "loss": 1.0622, + "step": 4832 + }, + { + "epoch": 0.12, + "learning_rate": 1.9612669438003242e-05, + "loss": 0.8668, + "step": 4833 + }, + { + "epoch": 0.12, + "learning_rate": 1.961245473070843e-05, + "loss": 1.1066, + "step": 4834 + }, + { + "epoch": 0.12, + "learning_rate": 1.9612239965097114e-05, + "loss": 1.0963, + "step": 4835 + }, + { + "epoch": 0.12, + "learning_rate": 1.96120251411706e-05, + "loss": 1.1542, + "step": 4836 + }, + { + "epoch": 0.12, + "learning_rate": 1.9611810258930188e-05, + "loss": 1.0768, + "step": 4837 + }, + { + "epoch": 0.12, + "learning_rate": 1.9611595318377184e-05, + "loss": 1.2344, + "step": 4838 + }, + { + "epoch": 0.12, + "learning_rate": 1.9611380319512895e-05, + "loss": 1.2215, + "step": 4839 + }, + { + "epoch": 0.12, + "learning_rate": 1.961116526233862e-05, + "loss": 1.0237, + "step": 4840 + }, + { + "epoch": 0.12, + "learning_rate": 1.961095014685567e-05, + "loss": 1.2385, + "step": 4841 + }, + { + "epoch": 0.12, + "learning_rate": 1.9610734973065342e-05, + "loss": 1.2895, + "step": 4842 + }, + { + "epoch": 0.12, + "learning_rate": 1.961051974096895e-05, + "loss": 0.7268, + "step": 4843 + }, + { + "epoch": 0.12, + "learning_rate": 1.961030445056779e-05, + "loss": 0.9772, + "step": 4844 + }, + { + "epoch": 0.12, + "learning_rate": 1.9610089101863173e-05, + "loss": 1.1954, + "step": 4845 + }, + { + "epoch": 0.12, + "learning_rate": 1.960987369485641e-05, + "loss": 0.9868, + "step": 4846 + }, + { + "epoch": 0.12, + "learning_rate": 1.9609658229548803e-05, + "loss": 1.2603, + "step": 4847 + }, + { + "epoch": 0.12, + "learning_rate": 1.960944270594166e-05, + "loss": 1.1276, + "step": 4848 + }, + { + "epoch": 0.12, + "learning_rate": 1.9609227124036285e-05, + "loss": 1.1406, + "step": 4849 + }, + { + "epoch": 0.12, + "learning_rate": 1.9609011483833993e-05, + "loss": 1.1539, + "step": 4850 + }, + { + "epoch": 0.12, + "learning_rate": 1.9608795785336085e-05, + "loss": 1.2234, + "step": 4851 + }, + { + "epoch": 0.12, + "learning_rate": 1.9608580028543875e-05, + "loss": 1.0133, + "step": 4852 + }, + { + "epoch": 0.12, + "learning_rate": 1.960836421345867e-05, + "loss": 0.9458, + "step": 4853 + }, + { + "epoch": 0.12, + "learning_rate": 1.960814834008178e-05, + "loss": 0.9204, + "step": 4854 + }, + { + "epoch": 0.12, + "learning_rate": 1.9607932408414513e-05, + "loss": 1.0282, + "step": 4855 + }, + { + "epoch": 0.12, + "learning_rate": 1.960771641845818e-05, + "loss": 1.2696, + "step": 4856 + }, + { + "epoch": 0.12, + "learning_rate": 1.960750037021409e-05, + "loss": 1.2335, + "step": 4857 + }, + { + "epoch": 0.12, + "learning_rate": 1.960728426368356e-05, + "loss": 1.207, + "step": 4858 + }, + { + "epoch": 0.12, + "learning_rate": 1.9607068098867894e-05, + "loss": 1.4598, + "step": 4859 + }, + { + "epoch": 0.12, + "learning_rate": 1.9606851875768404e-05, + "loss": 1.1719, + "step": 4860 + }, + { + "epoch": 0.12, + "learning_rate": 1.9606635594386403e-05, + "loss": 1.4129, + "step": 4861 + }, + { + "epoch": 0.12, + "learning_rate": 1.9606419254723205e-05, + "loss": 0.9829, + "step": 4862 + }, + { + "epoch": 0.12, + "learning_rate": 1.9606202856780122e-05, + "loss": 1.2571, + "step": 4863 + }, + { + "epoch": 0.12, + "learning_rate": 1.9605986400558462e-05, + "loss": 1.2788, + "step": 4864 + }, + { + "epoch": 0.12, + "learning_rate": 1.9605769886059546e-05, + "loss": 1.1576, + "step": 4865 + }, + { + "epoch": 0.12, + "learning_rate": 1.9605553313284682e-05, + "loss": 1.1699, + "step": 4866 + }, + { + "epoch": 0.12, + "learning_rate": 1.9605336682235185e-05, + "loss": 1.2138, + "step": 4867 + }, + { + "epoch": 0.12, + "learning_rate": 1.9605119992912368e-05, + "loss": 1.4715, + "step": 4868 + }, + { + "epoch": 0.12, + "learning_rate": 1.960490324531755e-05, + "loss": 1.2984, + "step": 4869 + }, + { + "epoch": 0.12, + "learning_rate": 1.960468643945204e-05, + "loss": 1.2713, + "step": 4870 + }, + { + "epoch": 0.12, + "learning_rate": 1.9604469575317158e-05, + "loss": 1.2002, + "step": 4871 + }, + { + "epoch": 0.12, + "learning_rate": 1.9604252652914222e-05, + "loss": 1.0376, + "step": 4872 + }, + { + "epoch": 0.12, + "learning_rate": 1.960403567224454e-05, + "loss": 1.2977, + "step": 4873 + }, + { + "epoch": 0.12, + "learning_rate": 1.9603818633309434e-05, + "loss": 1.0858, + "step": 4874 + }, + { + "epoch": 0.12, + "learning_rate": 1.9603601536110218e-05, + "loss": 1.1387, + "step": 4875 + }, + { + "epoch": 0.12, + "learning_rate": 1.9603384380648213e-05, + "loss": 1.1714, + "step": 4876 + }, + { + "epoch": 0.12, + "learning_rate": 1.9603167166924734e-05, + "loss": 0.9412, + "step": 4877 + }, + { + "epoch": 0.12, + "learning_rate": 1.9602949894941096e-05, + "loss": 1.1476, + "step": 4878 + }, + { + "epoch": 0.12, + "learning_rate": 1.960273256469862e-05, + "loss": 1.2296, + "step": 4879 + }, + { + "epoch": 0.12, + "learning_rate": 1.9602515176198623e-05, + "loss": 0.8598, + "step": 4880 + }, + { + "epoch": 0.12, + "learning_rate": 1.9602297729442428e-05, + "loss": 1.0648, + "step": 4881 + }, + { + "epoch": 0.12, + "learning_rate": 1.960208022443135e-05, + "loss": 1.0582, + "step": 4882 + }, + { + "epoch": 0.12, + "learning_rate": 1.960186266116671e-05, + "loss": 1.1108, + "step": 4883 + }, + { + "epoch": 0.12, + "learning_rate": 1.9601645039649828e-05, + "loss": 1.2818, + "step": 4884 + }, + { + "epoch": 0.12, + "learning_rate": 1.9601427359882025e-05, + "loss": 1.127, + "step": 4885 + }, + { + "epoch": 0.12, + "learning_rate": 1.9601209621864616e-05, + "loss": 1.0822, + "step": 4886 + }, + { + "epoch": 0.12, + "learning_rate": 1.960099182559893e-05, + "loss": 1.3453, + "step": 4887 + }, + { + "epoch": 0.12, + "learning_rate": 1.9600773971086286e-05, + "loss": 0.8728, + "step": 4888 + }, + { + "epoch": 0.12, + "learning_rate": 1.9600556058328003e-05, + "loss": 1.0077, + "step": 4889 + }, + { + "epoch": 0.12, + "learning_rate": 1.9600338087325407e-05, + "loss": 1.1574, + "step": 4890 + }, + { + "epoch": 0.12, + "learning_rate": 1.9600120058079816e-05, + "loss": 1.1397, + "step": 4891 + }, + { + "epoch": 0.12, + "learning_rate": 1.959990197059255e-05, + "loss": 1.4294, + "step": 4892 + }, + { + "epoch": 0.12, + "learning_rate": 1.9599683824864945e-05, + "loss": 1.3115, + "step": 4893 + }, + { + "epoch": 0.12, + "learning_rate": 1.9599465620898314e-05, + "loss": 1.116, + "step": 4894 + }, + { + "epoch": 0.12, + "learning_rate": 1.959924735869398e-05, + "loss": 1.1848, + "step": 4895 + }, + { + "epoch": 0.12, + "learning_rate": 1.959902903825327e-05, + "loss": 0.8818, + "step": 4896 + }, + { + "epoch": 0.12, + "learning_rate": 1.9598810659577512e-05, + "loss": 1.225, + "step": 4897 + }, + { + "epoch": 0.12, + "learning_rate": 1.959859222266803e-05, + "loss": 1.1026, + "step": 4898 + }, + { + "epoch": 0.12, + "learning_rate": 1.9598373727526137e-05, + "loss": 1.3019, + "step": 4899 + }, + { + "epoch": 0.12, + "learning_rate": 1.9598155174153174e-05, + "loss": 0.924, + "step": 4900 + }, + { + "epoch": 0.12, + "learning_rate": 1.9597936562550463e-05, + "loss": 1.307, + "step": 4901 + }, + { + "epoch": 0.12, + "learning_rate": 1.9597717892719326e-05, + "loss": 1.0936, + "step": 4902 + }, + { + "epoch": 0.12, + "learning_rate": 1.9597499164661095e-05, + "loss": 1.0152, + "step": 4903 + }, + { + "epoch": 0.12, + "learning_rate": 1.9597280378377087e-05, + "loss": 1.0341, + "step": 4904 + }, + { + "epoch": 0.12, + "learning_rate": 1.9597061533868643e-05, + "loss": 1.2687, + "step": 4905 + }, + { + "epoch": 0.12, + "learning_rate": 1.959684263113708e-05, + "loss": 1.2632, + "step": 4906 + }, + { + "epoch": 0.12, + "learning_rate": 1.9596623670183734e-05, + "loss": 1.2737, + "step": 4907 + }, + { + "epoch": 0.12, + "learning_rate": 1.9596404651009928e-05, + "loss": 1.2211, + "step": 4908 + }, + { + "epoch": 0.12, + "learning_rate": 1.959618557361699e-05, + "loss": 1.0173, + "step": 4909 + }, + { + "epoch": 0.12, + "learning_rate": 1.9595966438006253e-05, + "loss": 1.2564, + "step": 4910 + }, + { + "epoch": 0.12, + "learning_rate": 1.9595747244179048e-05, + "loss": 0.9357, + "step": 4911 + }, + { + "epoch": 0.12, + "learning_rate": 1.9595527992136697e-05, + "loss": 1.3041, + "step": 4912 + }, + { + "epoch": 0.12, + "learning_rate": 1.9595308681880537e-05, + "loss": 0.8582, + "step": 4913 + }, + { + "epoch": 0.12, + "learning_rate": 1.9595089313411892e-05, + "loss": 0.9005, + "step": 4914 + }, + { + "epoch": 0.12, + "learning_rate": 1.95948698867321e-05, + "loss": 1.333, + "step": 4915 + }, + { + "epoch": 0.12, + "learning_rate": 1.9594650401842493e-05, + "loss": 1.0478, + "step": 4916 + }, + { + "epoch": 0.12, + "learning_rate": 1.9594430858744393e-05, + "loss": 1.4672, + "step": 4917 + }, + { + "epoch": 0.12, + "learning_rate": 1.959421125743914e-05, + "loss": 1.1257, + "step": 4918 + }, + { + "epoch": 0.12, + "learning_rate": 1.9593991597928063e-05, + "loss": 1.1212, + "step": 4919 + }, + { + "epoch": 0.12, + "learning_rate": 1.9593771880212498e-05, + "loss": 0.9741, + "step": 4920 + }, + { + "epoch": 0.12, + "learning_rate": 1.9593552104293775e-05, + "loss": 1.1602, + "step": 4921 + }, + { + "epoch": 0.12, + "learning_rate": 1.9593332270173225e-05, + "loss": 0.9069, + "step": 4922 + }, + { + "epoch": 0.12, + "learning_rate": 1.959311237785219e-05, + "loss": 1.357, + "step": 4923 + }, + { + "epoch": 0.12, + "learning_rate": 1.9592892427331993e-05, + "loss": 1.2127, + "step": 4924 + }, + { + "epoch": 0.12, + "learning_rate": 1.9592672418613975e-05, + "loss": 1.1332, + "step": 4925 + }, + { + "epoch": 0.12, + "learning_rate": 1.9592452351699475e-05, + "loss": 1.192, + "step": 4926 + }, + { + "epoch": 0.12, + "learning_rate": 1.9592232226589817e-05, + "loss": 1.2969, + "step": 4927 + }, + { + "epoch": 0.12, + "learning_rate": 1.9592012043286342e-05, + "loss": 1.3605, + "step": 4928 + }, + { + "epoch": 0.12, + "learning_rate": 1.9591791801790393e-05, + "loss": 0.9299, + "step": 4929 + }, + { + "epoch": 0.12, + "learning_rate": 1.9591571502103294e-05, + "loss": 1.1807, + "step": 4930 + }, + { + "epoch": 0.12, + "learning_rate": 1.9591351144226386e-05, + "loss": 1.0548, + "step": 4931 + }, + { + "epoch": 0.12, + "learning_rate": 1.959113072816101e-05, + "loss": 1.391, + "step": 4932 + }, + { + "epoch": 0.12, + "learning_rate": 1.9590910253908493e-05, + "loss": 1.4093, + "step": 4933 + }, + { + "epoch": 0.12, + "learning_rate": 1.9590689721470188e-05, + "loss": 1.0743, + "step": 4934 + }, + { + "epoch": 0.12, + "learning_rate": 1.959046913084742e-05, + "loss": 1.2704, + "step": 4935 + }, + { + "epoch": 0.12, + "learning_rate": 1.9590248482041533e-05, + "loss": 1.109, + "step": 4936 + }, + { + "epoch": 0.12, + "learning_rate": 1.9590027775053864e-05, + "loss": 1.4226, + "step": 4937 + }, + { + "epoch": 0.12, + "learning_rate": 1.958980700988575e-05, + "loss": 1.3539, + "step": 4938 + }, + { + "epoch": 0.12, + "learning_rate": 1.9589586186538537e-05, + "loss": 1.1802, + "step": 4939 + }, + { + "epoch": 0.12, + "learning_rate": 1.9589365305013556e-05, + "loss": 1.4037, + "step": 4940 + }, + { + "epoch": 0.12, + "learning_rate": 1.9589144365312153e-05, + "loss": 1.0816, + "step": 4941 + }, + { + "epoch": 0.12, + "learning_rate": 1.9588923367435667e-05, + "loss": 1.3197, + "step": 4942 + }, + { + "epoch": 0.12, + "learning_rate": 1.9588702311385435e-05, + "loss": 1.3063, + "step": 4943 + }, + { + "epoch": 0.12, + "learning_rate": 1.9588481197162804e-05, + "loss": 1.0874, + "step": 4944 + }, + { + "epoch": 0.12, + "learning_rate": 1.9588260024769114e-05, + "loss": 0.9793, + "step": 4945 + }, + { + "epoch": 0.12, + "learning_rate": 1.9588038794205705e-05, + "loss": 0.9071, + "step": 4946 + }, + { + "epoch": 0.12, + "learning_rate": 1.958781750547392e-05, + "loss": 1.2353, + "step": 4947 + }, + { + "epoch": 0.12, + "learning_rate": 1.9587596158575102e-05, + "loss": 1.0659, + "step": 4948 + }, + { + "epoch": 0.12, + "learning_rate": 1.958737475351059e-05, + "loss": 1.3622, + "step": 4949 + }, + { + "epoch": 0.12, + "learning_rate": 1.9587153290281734e-05, + "loss": 1.3218, + "step": 4950 + }, + { + "epoch": 0.12, + "learning_rate": 1.9586931768889874e-05, + "loss": 1.2713, + "step": 4951 + }, + { + "epoch": 0.12, + "learning_rate": 1.958671018933635e-05, + "loss": 1.2848, + "step": 4952 + }, + { + "epoch": 0.12, + "learning_rate": 1.9586488551622516e-05, + "loss": 1.1904, + "step": 4953 + }, + { + "epoch": 0.12, + "learning_rate": 1.958626685574971e-05, + "loss": 1.1661, + "step": 4954 + }, + { + "epoch": 0.12, + "learning_rate": 1.9586045101719273e-05, + "loss": 1.1134, + "step": 4955 + }, + { + "epoch": 0.12, + "learning_rate": 1.9585823289532556e-05, + "loss": 1.0041, + "step": 4956 + }, + { + "epoch": 0.12, + "learning_rate": 1.9585601419190907e-05, + "loss": 1.1982, + "step": 4957 + }, + { + "epoch": 0.12, + "learning_rate": 1.958537949069567e-05, + "loss": 1.2999, + "step": 4958 + }, + { + "epoch": 0.12, + "learning_rate": 1.9585157504048184e-05, + "loss": 1.0874, + "step": 4959 + }, + { + "epoch": 0.12, + "learning_rate": 1.9584935459249807e-05, + "loss": 1.2002, + "step": 4960 + }, + { + "epoch": 0.12, + "learning_rate": 1.958471335630188e-05, + "loss": 1.245, + "step": 4961 + }, + { + "epoch": 0.12, + "learning_rate": 1.958449119520575e-05, + "loss": 1.1532, + "step": 4962 + }, + { + "epoch": 0.12, + "learning_rate": 1.9584268975962767e-05, + "loss": 1.415, + "step": 4963 + }, + { + "epoch": 0.12, + "learning_rate": 1.9584046698574282e-05, + "loss": 1.074, + "step": 4964 + }, + { + "epoch": 0.12, + "learning_rate": 1.9583824363041634e-05, + "loss": 1.2247, + "step": 4965 + }, + { + "epoch": 0.12, + "learning_rate": 1.958360196936618e-05, + "loss": 0.9751, + "step": 4966 + }, + { + "epoch": 0.12, + "learning_rate": 1.958337951754927e-05, + "loss": 1.36, + "step": 4967 + }, + { + "epoch": 0.12, + "learning_rate": 1.958315700759225e-05, + "loss": 1.3036, + "step": 4968 + }, + { + "epoch": 0.12, + "learning_rate": 1.9582934439496467e-05, + "loss": 1.0093, + "step": 4969 + }, + { + "epoch": 0.12, + "learning_rate": 1.9582711813263277e-05, + "loss": 1.1629, + "step": 4970 + }, + { + "epoch": 0.12, + "learning_rate": 1.958248912889403e-05, + "loss": 1.1343, + "step": 4971 + }, + { + "epoch": 0.12, + "learning_rate": 1.9582266386390075e-05, + "loss": 0.9814, + "step": 4972 + }, + { + "epoch": 0.12, + "learning_rate": 1.958204358575276e-05, + "loss": 1.0826, + "step": 4973 + }, + { + "epoch": 0.12, + "learning_rate": 1.9581820726983443e-05, + "loss": 1.1705, + "step": 4974 + }, + { + "epoch": 0.12, + "learning_rate": 1.9581597810083474e-05, + "loss": 0.9348, + "step": 4975 + }, + { + "epoch": 0.12, + "learning_rate": 1.9581374835054205e-05, + "loss": 0.8618, + "step": 4976 + }, + { + "epoch": 0.12, + "learning_rate": 1.9581151801896988e-05, + "loss": 1.1581, + "step": 4977 + }, + { + "epoch": 0.12, + "learning_rate": 1.9580928710613176e-05, + "loss": 0.9453, + "step": 4978 + }, + { + "epoch": 0.12, + "learning_rate": 1.9580705561204123e-05, + "loss": 1.0367, + "step": 4979 + }, + { + "epoch": 0.12, + "learning_rate": 1.9580482353671184e-05, + "loss": 1.1923, + "step": 4980 + }, + { + "epoch": 0.12, + "learning_rate": 1.9580259088015713e-05, + "loss": 0.8289, + "step": 4981 + }, + { + "epoch": 0.12, + "learning_rate": 1.958003576423906e-05, + "loss": 1.363, + "step": 4982 + }, + { + "epoch": 0.12, + "learning_rate": 1.9579812382342583e-05, + "loss": 1.3435, + "step": 4983 + }, + { + "epoch": 0.12, + "learning_rate": 1.9579588942327642e-05, + "loss": 1.1273, + "step": 4984 + }, + { + "epoch": 0.12, + "learning_rate": 1.9579365444195586e-05, + "loss": 1.04, + "step": 4985 + }, + { + "epoch": 0.12, + "learning_rate": 1.9579141887947772e-05, + "loss": 1.0231, + "step": 4986 + }, + { + "epoch": 0.12, + "learning_rate": 1.957891827358556e-05, + "loss": 1.035, + "step": 4987 + }, + { + "epoch": 0.12, + "learning_rate": 1.95786946011103e-05, + "loss": 1.0906, + "step": 4988 + }, + { + "epoch": 0.12, + "learning_rate": 1.957847087052336e-05, + "loss": 0.8557, + "step": 4989 + }, + { + "epoch": 0.12, + "learning_rate": 1.9578247081826083e-05, + "loss": 1.1466, + "step": 4990 + }, + { + "epoch": 0.12, + "learning_rate": 1.9578023235019836e-05, + "loss": 0.9967, + "step": 4991 + }, + { + "epoch": 0.12, + "learning_rate": 1.9577799330105973e-05, + "loss": 1.2334, + "step": 4992 + }, + { + "epoch": 0.12, + "learning_rate": 1.9577575367085857e-05, + "loss": 1.112, + "step": 4993 + }, + { + "epoch": 0.12, + "learning_rate": 1.9577351345960845e-05, + "loss": 1.0679, + "step": 4994 + }, + { + "epoch": 0.12, + "learning_rate": 1.957712726673229e-05, + "loss": 1.0044, + "step": 4995 + }, + { + "epoch": 0.12, + "learning_rate": 1.9576903129401563e-05, + "loss": 0.8891, + "step": 4996 + }, + { + "epoch": 0.12, + "learning_rate": 1.9576678933970015e-05, + "loss": 1.0278, + "step": 4997 + }, + { + "epoch": 0.12, + "learning_rate": 1.957645468043901e-05, + "loss": 1.1993, + "step": 4998 + }, + { + "epoch": 0.12, + "learning_rate": 1.95762303688099e-05, + "loss": 1.229, + "step": 4999 + }, + { + "epoch": 0.12, + "learning_rate": 1.957600599908406e-05, + "loss": 1.1005, + "step": 5000 + }, + { + "epoch": 0.12, + "learning_rate": 1.9575781571262843e-05, + "loss": 1.4382, + "step": 5001 + }, + { + "epoch": 0.12, + "learning_rate": 1.957555708534761e-05, + "loss": 1.1209, + "step": 5002 + }, + { + "epoch": 0.12, + "learning_rate": 1.9575332541339725e-05, + "loss": 1.2808, + "step": 5003 + }, + { + "epoch": 0.12, + "learning_rate": 1.9575107939240548e-05, + "loss": 1.2065, + "step": 5004 + }, + { + "epoch": 0.12, + "learning_rate": 1.9574883279051445e-05, + "loss": 0.8671, + "step": 5005 + }, + { + "epoch": 0.12, + "learning_rate": 1.957465856077378e-05, + "loss": 1.1488, + "step": 5006 + }, + { + "epoch": 0.12, + "learning_rate": 1.9574433784408912e-05, + "loss": 0.9967, + "step": 5007 + }, + { + "epoch": 0.12, + "learning_rate": 1.9574208949958203e-05, + "loss": 1.0101, + "step": 5008 + }, + { + "epoch": 0.12, + "learning_rate": 1.9573984057423022e-05, + "loss": 1.092, + "step": 5009 + }, + { + "epoch": 0.12, + "learning_rate": 1.9573759106804732e-05, + "loss": 1.2271, + "step": 5010 + }, + { + "epoch": 0.12, + "learning_rate": 1.95735340981047e-05, + "loss": 1.1964, + "step": 5011 + }, + { + "epoch": 0.12, + "learning_rate": 1.9573309031324284e-05, + "loss": 1.4116, + "step": 5012 + }, + { + "epoch": 0.12, + "learning_rate": 1.9573083906464856e-05, + "loss": 1.1229, + "step": 5013 + }, + { + "epoch": 0.12, + "learning_rate": 1.957285872352778e-05, + "loss": 1.1361, + "step": 5014 + }, + { + "epoch": 0.12, + "learning_rate": 1.957263348251442e-05, + "loss": 1.3642, + "step": 5015 + }, + { + "epoch": 0.12, + "learning_rate": 1.9572408183426145e-05, + "loss": 1.2537, + "step": 5016 + }, + { + "epoch": 0.12, + "learning_rate": 1.9572182826264322e-05, + "loss": 1.1105, + "step": 5017 + }, + { + "epoch": 0.12, + "learning_rate": 1.9571957411030318e-05, + "loss": 1.0431, + "step": 5018 + }, + { + "epoch": 0.12, + "learning_rate": 1.95717319377255e-05, + "loss": 1.4266, + "step": 5019 + }, + { + "epoch": 0.12, + "learning_rate": 1.9571506406351233e-05, + "loss": 1.1451, + "step": 5020 + }, + { + "epoch": 0.12, + "learning_rate": 1.957128081690889e-05, + "loss": 0.9904, + "step": 5021 + }, + { + "epoch": 0.12, + "learning_rate": 1.9571055169399837e-05, + "loss": 0.894, + "step": 5022 + }, + { + "epoch": 0.12, + "learning_rate": 1.9570829463825442e-05, + "loss": 1.3138, + "step": 5023 + }, + { + "epoch": 0.12, + "learning_rate": 1.957060370018708e-05, + "loss": 1.1278, + "step": 5024 + }, + { + "epoch": 0.12, + "learning_rate": 1.957037787848611e-05, + "loss": 1.099, + "step": 5025 + }, + { + "epoch": 0.12, + "learning_rate": 1.9570151998723918e-05, + "loss": 1.1605, + "step": 5026 + }, + { + "epoch": 0.12, + "learning_rate": 1.9569926060901857e-05, + "loss": 1.1013, + "step": 5027 + }, + { + "epoch": 0.12, + "learning_rate": 1.956970006502131e-05, + "loss": 1.0147, + "step": 5028 + }, + { + "epoch": 0.12, + "learning_rate": 1.956947401108364e-05, + "loss": 1.0492, + "step": 5029 + }, + { + "epoch": 0.12, + "learning_rate": 1.956924789909022e-05, + "loss": 1.1651, + "step": 5030 + }, + { + "epoch": 0.12, + "learning_rate": 1.956902172904243e-05, + "loss": 1.1775, + "step": 5031 + }, + { + "epoch": 0.12, + "learning_rate": 1.9568795500941635e-05, + "loss": 1.2214, + "step": 5032 + }, + { + "epoch": 0.12, + "learning_rate": 1.9568569214789205e-05, + "loss": 1.0905, + "step": 5033 + }, + { + "epoch": 0.12, + "learning_rate": 1.956834287058652e-05, + "loss": 1.3745, + "step": 5034 + }, + { + "epoch": 0.12, + "learning_rate": 1.9568116468334946e-05, + "loss": 1.2204, + "step": 5035 + }, + { + "epoch": 0.12, + "learning_rate": 1.9567890008035865e-05, + "loss": 1.1344, + "step": 5036 + }, + { + "epoch": 0.12, + "learning_rate": 1.956766348969064e-05, + "loss": 1.2708, + "step": 5037 + }, + { + "epoch": 0.12, + "learning_rate": 1.956743691330065e-05, + "loss": 1.1378, + "step": 5038 + }, + { + "epoch": 0.12, + "learning_rate": 1.9567210278867277e-05, + "loss": 1.1687, + "step": 5039 + }, + { + "epoch": 0.12, + "learning_rate": 1.9566983586391884e-05, + "loss": 1.1448, + "step": 5040 + }, + { + "epoch": 0.12, + "learning_rate": 1.9566756835875856e-05, + "loss": 1.2908, + "step": 5041 + }, + { + "epoch": 0.12, + "learning_rate": 1.956653002732056e-05, + "loss": 0.9729, + "step": 5042 + }, + { + "epoch": 0.12, + "learning_rate": 1.9566303160727386e-05, + "loss": 1.2196, + "step": 5043 + }, + { + "epoch": 0.12, + "learning_rate": 1.9566076236097695e-05, + "loss": 1.2093, + "step": 5044 + }, + { + "epoch": 0.12, + "learning_rate": 1.9565849253432863e-05, + "loss": 1.2236, + "step": 5045 + }, + { + "epoch": 0.12, + "learning_rate": 1.956562221273428e-05, + "loss": 1.0906, + "step": 5046 + }, + { + "epoch": 0.12, + "learning_rate": 1.9565395114003317e-05, + "loss": 1.18, + "step": 5047 + }, + { + "epoch": 0.12, + "learning_rate": 1.9565167957241353e-05, + "loss": 1.3513, + "step": 5048 + }, + { + "epoch": 0.12, + "learning_rate": 1.956494074244976e-05, + "loss": 1.0784, + "step": 5049 + }, + { + "epoch": 0.12, + "learning_rate": 1.9564713469629928e-05, + "loss": 1.2963, + "step": 5050 + }, + { + "epoch": 0.12, + "learning_rate": 1.9564486138783225e-05, + "loss": 1.2221, + "step": 5051 + }, + { + "epoch": 0.12, + "learning_rate": 1.9564258749911035e-05, + "loss": 1.1434, + "step": 5052 + }, + { + "epoch": 0.12, + "learning_rate": 1.9564031303014737e-05, + "loss": 1.46, + "step": 5053 + }, + { + "epoch": 0.12, + "learning_rate": 1.956380379809571e-05, + "loss": 1.1436, + "step": 5054 + }, + { + "epoch": 0.12, + "learning_rate": 1.9563576235155334e-05, + "loss": 0.9707, + "step": 5055 + }, + { + "epoch": 0.12, + "learning_rate": 1.9563348614194992e-05, + "loss": 1.2579, + "step": 5056 + }, + { + "epoch": 0.12, + "learning_rate": 1.9563120935216064e-05, + "loss": 1.1717, + "step": 5057 + }, + { + "epoch": 0.12, + "learning_rate": 1.956289319821993e-05, + "loss": 1.112, + "step": 5058 + }, + { + "epoch": 0.12, + "learning_rate": 1.9562665403207974e-05, + "loss": 1.3058, + "step": 5059 + }, + { + "epoch": 0.12, + "learning_rate": 1.9562437550181573e-05, + "loss": 1.3226, + "step": 5060 + }, + { + "epoch": 0.12, + "learning_rate": 1.9562209639142112e-05, + "loss": 1.1506, + "step": 5061 + }, + { + "epoch": 0.12, + "learning_rate": 1.9561981670090978e-05, + "loss": 0.8178, + "step": 5062 + }, + { + "epoch": 0.12, + "learning_rate": 1.956175364302955e-05, + "loss": 0.946, + "step": 5063 + }, + { + "epoch": 0.12, + "learning_rate": 1.9561525557959207e-05, + "loss": 1.1594, + "step": 5064 + }, + { + "epoch": 0.12, + "learning_rate": 1.9561297414881342e-05, + "loss": 1.1066, + "step": 5065 + }, + { + "epoch": 0.12, + "learning_rate": 1.9561069213797333e-05, + "loss": 1.1207, + "step": 5066 + }, + { + "epoch": 0.12, + "learning_rate": 1.9560840954708565e-05, + "loss": 1.0901, + "step": 5067 + }, + { + "epoch": 0.12, + "learning_rate": 1.9560612637616428e-05, + "loss": 1.0995, + "step": 5068 + }, + { + "epoch": 0.12, + "learning_rate": 1.95603842625223e-05, + "loss": 1.212, + "step": 5069 + }, + { + "epoch": 0.12, + "learning_rate": 1.9560155829427567e-05, + "loss": 1.1864, + "step": 5070 + }, + { + "epoch": 0.12, + "learning_rate": 1.955992733833362e-05, + "loss": 1.1923, + "step": 5071 + }, + { + "epoch": 0.12, + "learning_rate": 1.9559698789241844e-05, + "loss": 1.2725, + "step": 5072 + }, + { + "epoch": 0.12, + "learning_rate": 1.955947018215362e-05, + "loss": 0.9973, + "step": 5073 + }, + { + "epoch": 0.12, + "learning_rate": 1.955924151707034e-05, + "loss": 1.0769, + "step": 5074 + }, + { + "epoch": 0.12, + "learning_rate": 1.9559012793993394e-05, + "loss": 1.2189, + "step": 5075 + }, + { + "epoch": 0.12, + "learning_rate": 1.955878401292416e-05, + "loss": 1.2818, + "step": 5076 + }, + { + "epoch": 0.12, + "learning_rate": 1.955855517386404e-05, + "loss": 1.3795, + "step": 5077 + }, + { + "epoch": 0.12, + "learning_rate": 1.955832627681441e-05, + "loss": 1.0603, + "step": 5078 + }, + { + "epoch": 0.12, + "learning_rate": 1.9558097321776656e-05, + "loss": 1.4776, + "step": 5079 + }, + { + "epoch": 0.12, + "learning_rate": 1.955786830875218e-05, + "loss": 1.2626, + "step": 5080 + }, + { + "epoch": 0.12, + "learning_rate": 1.9557639237742366e-05, + "loss": 1.0615, + "step": 5081 + }, + { + "epoch": 0.12, + "learning_rate": 1.95574101087486e-05, + "loss": 1.341, + "step": 5082 + }, + { + "epoch": 0.12, + "learning_rate": 1.955718092177228e-05, + "loss": 1.2654, + "step": 5083 + }, + { + "epoch": 0.12, + "learning_rate": 1.9556951676814787e-05, + "loss": 0.8432, + "step": 5084 + }, + { + "epoch": 0.12, + "learning_rate": 1.9556722373877516e-05, + "loss": 1.2358, + "step": 5085 + }, + { + "epoch": 0.12, + "learning_rate": 1.9556493012961856e-05, + "loss": 1.093, + "step": 5086 + }, + { + "epoch": 0.12, + "learning_rate": 1.9556263594069204e-05, + "loss": 1.0469, + "step": 5087 + }, + { + "epoch": 0.12, + "learning_rate": 1.955603411720095e-05, + "loss": 1.2578, + "step": 5088 + }, + { + "epoch": 0.12, + "learning_rate": 1.9555804582358483e-05, + "loss": 1.0416, + "step": 5089 + }, + { + "epoch": 0.12, + "learning_rate": 1.9555574989543197e-05, + "loss": 1.1889, + "step": 5090 + }, + { + "epoch": 0.12, + "learning_rate": 1.9555345338756484e-05, + "loss": 1.4886, + "step": 5091 + }, + { + "epoch": 0.12, + "learning_rate": 1.9555115629999738e-05, + "loss": 1.2792, + "step": 5092 + }, + { + "epoch": 0.12, + "learning_rate": 1.9554885863274354e-05, + "loss": 1.2437, + "step": 5093 + }, + { + "epoch": 0.12, + "learning_rate": 1.9554656038581728e-05, + "loss": 1.1485, + "step": 5094 + }, + { + "epoch": 0.12, + "learning_rate": 1.955442615592325e-05, + "loss": 1.2628, + "step": 5095 + }, + { + "epoch": 0.12, + "learning_rate": 1.9554196215300314e-05, + "loss": 1.3926, + "step": 5096 + }, + { + "epoch": 0.12, + "learning_rate": 1.9553966216714318e-05, + "loss": 1.3007, + "step": 5097 + }, + { + "epoch": 0.12, + "learning_rate": 1.9553736160166657e-05, + "loss": 1.0409, + "step": 5098 + }, + { + "epoch": 0.12, + "learning_rate": 1.9553506045658723e-05, + "loss": 1.1828, + "step": 5099 + }, + { + "epoch": 0.12, + "learning_rate": 1.9553275873191916e-05, + "loss": 1.1873, + "step": 5100 + }, + { + "epoch": 0.12, + "learning_rate": 1.9553045642767632e-05, + "loss": 1.0485, + "step": 5101 + }, + { + "epoch": 0.12, + "learning_rate": 1.9552815354387267e-05, + "loss": 1.3673, + "step": 5102 + }, + { + "epoch": 0.12, + "learning_rate": 1.955258500805222e-05, + "loss": 1.3339, + "step": 5103 + }, + { + "epoch": 0.12, + "learning_rate": 1.9552354603763882e-05, + "loss": 1.1765, + "step": 5104 + }, + { + "epoch": 0.12, + "learning_rate": 1.955212414152366e-05, + "loss": 1.2319, + "step": 5105 + }, + { + "epoch": 0.12, + "learning_rate": 1.9551893621332944e-05, + "loss": 1.0598, + "step": 5106 + }, + { + "epoch": 0.12, + "learning_rate": 1.955166304319314e-05, + "loss": 1.3459, + "step": 5107 + }, + { + "epoch": 0.12, + "learning_rate": 1.9551432407105642e-05, + "loss": 0.9953, + "step": 5108 + }, + { + "epoch": 0.12, + "learning_rate": 1.9551201713071847e-05, + "loss": 1.0859, + "step": 5109 + }, + { + "epoch": 0.12, + "learning_rate": 1.955097096109316e-05, + "loss": 1.3046, + "step": 5110 + }, + { + "epoch": 0.12, + "learning_rate": 1.9550740151170978e-05, + "loss": 1.3002, + "step": 5111 + }, + { + "epoch": 0.12, + "learning_rate": 1.9550509283306703e-05, + "loss": 1.367, + "step": 5112 + }, + { + "epoch": 0.12, + "learning_rate": 1.9550278357501736e-05, + "loss": 1.3098, + "step": 5113 + }, + { + "epoch": 0.12, + "learning_rate": 1.9550047373757475e-05, + "loss": 1.1816, + "step": 5114 + }, + { + "epoch": 0.12, + "learning_rate": 1.954981633207532e-05, + "loss": 1.2876, + "step": 5115 + }, + { + "epoch": 0.12, + "learning_rate": 1.9549585232456682e-05, + "loss": 1.0906, + "step": 5116 + }, + { + "epoch": 0.12, + "learning_rate": 1.9549354074902955e-05, + "loss": 1.2316, + "step": 5117 + }, + { + "epoch": 0.12, + "learning_rate": 1.9549122859415538e-05, + "loss": 1.3151, + "step": 5118 + }, + { + "epoch": 0.12, + "learning_rate": 1.9548891585995842e-05, + "loss": 1.3295, + "step": 5119 + }, + { + "epoch": 0.12, + "learning_rate": 1.9548660254645265e-05, + "loss": 1.3252, + "step": 5120 + }, + { + "epoch": 0.12, + "learning_rate": 1.9548428865365214e-05, + "loss": 1.2435, + "step": 5121 + }, + { + "epoch": 0.12, + "learning_rate": 1.954819741815709e-05, + "loss": 1.3445, + "step": 5122 + }, + { + "epoch": 0.12, + "learning_rate": 1.95479659130223e-05, + "loss": 1.2886, + "step": 5123 + }, + { + "epoch": 0.12, + "learning_rate": 1.9547734349962246e-05, + "loss": 1.1877, + "step": 5124 + }, + { + "epoch": 0.12, + "learning_rate": 1.954750272897833e-05, + "loss": 1.3615, + "step": 5125 + }, + { + "epoch": 0.12, + "learning_rate": 1.9547271050071965e-05, + "loss": 1.1564, + "step": 5126 + }, + { + "epoch": 0.12, + "learning_rate": 1.9547039313244552e-05, + "loss": 1.3116, + "step": 5127 + }, + { + "epoch": 0.12, + "learning_rate": 1.9546807518497496e-05, + "loss": 1.0637, + "step": 5128 + }, + { + "epoch": 0.12, + "learning_rate": 1.9546575665832205e-05, + "loss": 1.2262, + "step": 5129 + }, + { + "epoch": 0.12, + "learning_rate": 1.954634375525008e-05, + "loss": 0.8833, + "step": 5130 + }, + { + "epoch": 0.12, + "learning_rate": 1.954611178675254e-05, + "loss": 1.3696, + "step": 5131 + }, + { + "epoch": 0.12, + "learning_rate": 1.9545879760340983e-05, + "loss": 1.2046, + "step": 5132 + }, + { + "epoch": 0.12, + "learning_rate": 1.9545647676016816e-05, + "loss": 1.4696, + "step": 5133 + }, + { + "epoch": 0.12, + "learning_rate": 1.9545415533781453e-05, + "loss": 0.9613, + "step": 5134 + }, + { + "epoch": 0.12, + "learning_rate": 1.9545183333636297e-05, + "loss": 1.1443, + "step": 5135 + }, + { + "epoch": 0.12, + "learning_rate": 1.954495107558276e-05, + "loss": 1.4208, + "step": 5136 + }, + { + "epoch": 0.12, + "learning_rate": 1.954471875962225e-05, + "loss": 0.9552, + "step": 5137 + }, + { + "epoch": 0.12, + "learning_rate": 1.9544486385756176e-05, + "loss": 1.2657, + "step": 5138 + }, + { + "epoch": 0.12, + "learning_rate": 1.9544253953985948e-05, + "loss": 1.0132, + "step": 5139 + }, + { + "epoch": 0.12, + "learning_rate": 1.9544021464312977e-05, + "loss": 1.2413, + "step": 5140 + }, + { + "epoch": 0.12, + "learning_rate": 1.954378891673867e-05, + "loss": 1.2184, + "step": 5141 + }, + { + "epoch": 0.12, + "learning_rate": 1.9543556311264445e-05, + "loss": 1.048, + "step": 5142 + }, + { + "epoch": 0.12, + "learning_rate": 1.9543323647891706e-05, + "loss": 1.497, + "step": 5143 + }, + { + "epoch": 0.12, + "learning_rate": 1.954309092662187e-05, + "loss": 1.1565, + "step": 5144 + }, + { + "epoch": 0.12, + "learning_rate": 1.9542858147456338e-05, + "loss": 1.149, + "step": 5145 + }, + { + "epoch": 0.12, + "learning_rate": 1.9542625310396538e-05, + "loss": 0.9787, + "step": 5146 + }, + { + "epoch": 0.12, + "learning_rate": 1.9542392415443873e-05, + "loss": 1.1758, + "step": 5147 + }, + { + "epoch": 0.12, + "learning_rate": 1.9542159462599755e-05, + "loss": 1.0609, + "step": 5148 + }, + { + "epoch": 0.12, + "learning_rate": 1.9541926451865604e-05, + "loss": 1.1953, + "step": 5149 + }, + { + "epoch": 0.12, + "learning_rate": 1.954169338324283e-05, + "loss": 1.1798, + "step": 5150 + }, + { + "epoch": 0.12, + "learning_rate": 1.9541460256732843e-05, + "loss": 0.9131, + "step": 5151 + }, + { + "epoch": 0.12, + "learning_rate": 1.954122707233706e-05, + "loss": 1.1766, + "step": 5152 + }, + { + "epoch": 0.12, + "learning_rate": 1.9540993830056898e-05, + "loss": 1.0842, + "step": 5153 + }, + { + "epoch": 0.12, + "learning_rate": 1.954076052989377e-05, + "loss": 1.0737, + "step": 5154 + }, + { + "epoch": 0.12, + "learning_rate": 1.9540527171849098e-05, + "loss": 1.0726, + "step": 5155 + }, + { + "epoch": 0.12, + "learning_rate": 1.9540293755924285e-05, + "loss": 1.4019, + "step": 5156 + }, + { + "epoch": 0.12, + "learning_rate": 1.9540060282120756e-05, + "loss": 1.2655, + "step": 5157 + }, + { + "epoch": 0.12, + "learning_rate": 1.9539826750439926e-05, + "loss": 1.2763, + "step": 5158 + }, + { + "epoch": 0.12, + "learning_rate": 1.9539593160883208e-05, + "loss": 0.9869, + "step": 5159 + }, + { + "epoch": 0.12, + "learning_rate": 1.9539359513452026e-05, + "loss": 1.1742, + "step": 5160 + }, + { + "epoch": 0.12, + "learning_rate": 1.953912580814779e-05, + "loss": 1.2716, + "step": 5161 + }, + { + "epoch": 0.12, + "learning_rate": 1.9538892044971925e-05, + "loss": 0.8932, + "step": 5162 + }, + { + "epoch": 0.12, + "learning_rate": 1.9538658223925843e-05, + "loss": 1.3649, + "step": 5163 + }, + { + "epoch": 0.12, + "learning_rate": 1.9538424345010968e-05, + "loss": 1.0918, + "step": 5164 + }, + { + "epoch": 0.12, + "learning_rate": 1.9538190408228713e-05, + "loss": 1.0209, + "step": 5165 + }, + { + "epoch": 0.12, + "learning_rate": 1.9537956413580504e-05, + "loss": 1.2687, + "step": 5166 + }, + { + "epoch": 0.12, + "learning_rate": 1.953772236106775e-05, + "loss": 1.3238, + "step": 5167 + }, + { + "epoch": 0.12, + "learning_rate": 1.9537488250691884e-05, + "loss": 1.2563, + "step": 5168 + }, + { + "epoch": 0.12, + "learning_rate": 1.953725408245432e-05, + "loss": 1.2016, + "step": 5169 + }, + { + "epoch": 0.12, + "learning_rate": 1.9537019856356478e-05, + "loss": 0.8997, + "step": 5170 + }, + { + "epoch": 0.12, + "learning_rate": 1.9536785572399778e-05, + "loss": 1.3549, + "step": 5171 + }, + { + "epoch": 0.12, + "learning_rate": 1.9536551230585643e-05, + "loss": 1.136, + "step": 5172 + }, + { + "epoch": 0.12, + "learning_rate": 1.9536316830915498e-05, + "loss": 1.1216, + "step": 5173 + }, + { + "epoch": 0.12, + "learning_rate": 1.953608237339076e-05, + "loss": 1.2777, + "step": 5174 + }, + { + "epoch": 0.12, + "learning_rate": 1.9535847858012855e-05, + "loss": 0.9929, + "step": 5175 + }, + { + "epoch": 0.12, + "learning_rate": 1.95356132847832e-05, + "loss": 0.8064, + "step": 5176 + }, + { + "epoch": 0.12, + "learning_rate": 1.9535378653703223e-05, + "loss": 1.1705, + "step": 5177 + }, + { + "epoch": 0.12, + "learning_rate": 1.953514396477435e-05, + "loss": 1.1453, + "step": 5178 + }, + { + "epoch": 0.12, + "learning_rate": 1.9534909217997996e-05, + "loss": 1.1999, + "step": 5179 + }, + { + "epoch": 0.12, + "learning_rate": 1.9534674413375595e-05, + "loss": 1.2435, + "step": 5180 + }, + { + "epoch": 0.12, + "learning_rate": 1.9534439550908566e-05, + "loss": 1.0329, + "step": 5181 + }, + { + "epoch": 0.12, + "learning_rate": 1.9534204630598334e-05, + "loss": 1.3059, + "step": 5182 + }, + { + "epoch": 0.12, + "learning_rate": 1.9533969652446326e-05, + "loss": 0.7381, + "step": 5183 + }, + { + "epoch": 0.12, + "learning_rate": 1.953373461645397e-05, + "loss": 1.026, + "step": 5184 + }, + { + "epoch": 0.12, + "learning_rate": 1.9533499522622682e-05, + "loss": 1.2321, + "step": 5185 + }, + { + "epoch": 0.12, + "learning_rate": 1.9533264370953898e-05, + "loss": 1.0691, + "step": 5186 + }, + { + "epoch": 0.12, + "learning_rate": 1.953302916144904e-05, + "loss": 0.9861, + "step": 5187 + }, + { + "epoch": 0.12, + "learning_rate": 1.953279389410954e-05, + "loss": 1.2741, + "step": 5188 + }, + { + "epoch": 0.12, + "learning_rate": 1.9532558568936816e-05, + "loss": 1.3643, + "step": 5189 + }, + { + "epoch": 0.12, + "learning_rate": 1.9532323185932306e-05, + "loss": 1.3149, + "step": 5190 + }, + { + "epoch": 0.12, + "learning_rate": 1.953208774509743e-05, + "loss": 1.0915, + "step": 5191 + }, + { + "epoch": 0.12, + "learning_rate": 1.953185224643362e-05, + "loss": 1.0592, + "step": 5192 + }, + { + "epoch": 0.12, + "learning_rate": 1.953161668994231e-05, + "loss": 1.1573, + "step": 5193 + }, + { + "epoch": 0.12, + "learning_rate": 1.953138107562492e-05, + "loss": 1.1446, + "step": 5194 + }, + { + "epoch": 0.12, + "learning_rate": 1.9531145403482884e-05, + "loss": 1.3184, + "step": 5195 + }, + { + "epoch": 0.12, + "learning_rate": 1.953090967351763e-05, + "loss": 1.4019, + "step": 5196 + }, + { + "epoch": 0.12, + "learning_rate": 1.9530673885730588e-05, + "loss": 1.1824, + "step": 5197 + }, + { + "epoch": 0.13, + "learning_rate": 1.9530438040123188e-05, + "loss": 1.1608, + "step": 5198 + }, + { + "epoch": 0.13, + "learning_rate": 1.953020213669686e-05, + "loss": 1.0063, + "step": 5199 + }, + { + "epoch": 0.13, + "learning_rate": 1.952996617545304e-05, + "loss": 1.2589, + "step": 5200 + }, + { + "epoch": 0.13, + "learning_rate": 1.952973015639316e-05, + "loss": 1.3227, + "step": 5201 + }, + { + "epoch": 0.13, + "learning_rate": 1.9529494079518647e-05, + "loss": 1.2036, + "step": 5202 + }, + { + "epoch": 0.13, + "learning_rate": 1.9529257944830933e-05, + "loss": 1.0009, + "step": 5203 + }, + { + "epoch": 0.13, + "learning_rate": 1.9529021752331455e-05, + "loss": 1.2649, + "step": 5204 + }, + { + "epoch": 0.13, + "learning_rate": 1.952878550202164e-05, + "loss": 1.2442, + "step": 5205 + }, + { + "epoch": 0.13, + "learning_rate": 1.9528549193902926e-05, + "loss": 0.7933, + "step": 5206 + }, + { + "epoch": 0.13, + "learning_rate": 1.952831282797675e-05, + "loss": 0.9525, + "step": 5207 + }, + { + "epoch": 0.13, + "learning_rate": 1.9528076404244537e-05, + "loss": 1.1349, + "step": 5208 + }, + { + "epoch": 0.13, + "learning_rate": 1.9527839922707724e-05, + "loss": 1.1373, + "step": 5209 + }, + { + "epoch": 0.13, + "learning_rate": 1.952760338336775e-05, + "loss": 1.1146, + "step": 5210 + }, + { + "epoch": 0.13, + "learning_rate": 1.9527366786226047e-05, + "loss": 0.9406, + "step": 5211 + }, + { + "epoch": 0.13, + "learning_rate": 1.952713013128405e-05, + "loss": 1.1636, + "step": 5212 + }, + { + "epoch": 0.13, + "learning_rate": 1.9526893418543193e-05, + "loss": 1.4278, + "step": 5213 + }, + { + "epoch": 0.13, + "learning_rate": 1.9526656648004918e-05, + "loss": 1.1489, + "step": 5214 + }, + { + "epoch": 0.13, + "learning_rate": 1.9526419819670658e-05, + "loss": 1.12, + "step": 5215 + }, + { + "epoch": 0.13, + "learning_rate": 1.952618293354185e-05, + "loss": 1.1494, + "step": 5216 + }, + { + "epoch": 0.13, + "learning_rate": 1.9525945989619924e-05, + "loss": 1.1103, + "step": 5217 + }, + { + "epoch": 0.13, + "learning_rate": 1.9525708987906334e-05, + "loss": 0.9832, + "step": 5218 + }, + { + "epoch": 0.13, + "learning_rate": 1.9525471928402502e-05, + "loss": 0.9478, + "step": 5219 + }, + { + "epoch": 0.13, + "learning_rate": 1.9525234811109874e-05, + "loss": 1.0487, + "step": 5220 + }, + { + "epoch": 0.13, + "learning_rate": 1.9524997636029886e-05, + "loss": 1.0919, + "step": 5221 + }, + { + "epoch": 0.13, + "learning_rate": 1.952476040316398e-05, + "loss": 1.1045, + "step": 5222 + }, + { + "epoch": 0.13, + "learning_rate": 1.952452311251359e-05, + "loss": 1.3034, + "step": 5223 + }, + { + "epoch": 0.13, + "learning_rate": 1.9524285764080166e-05, + "loss": 1.2951, + "step": 5224 + }, + { + "epoch": 0.13, + "learning_rate": 1.952404835786513e-05, + "loss": 1.2685, + "step": 5225 + }, + { + "epoch": 0.13, + "learning_rate": 1.9523810893869937e-05, + "loss": 1.0098, + "step": 5226 + }, + { + "epoch": 0.13, + "learning_rate": 1.9523573372096027e-05, + "loss": 1.0937, + "step": 5227 + }, + { + "epoch": 0.13, + "learning_rate": 1.9523335792544835e-05, + "loss": 1.2915, + "step": 5228 + }, + { + "epoch": 0.13, + "learning_rate": 1.9523098155217806e-05, + "loss": 1.3425, + "step": 5229 + }, + { + "epoch": 0.13, + "learning_rate": 1.9522860460116377e-05, + "loss": 1.1384, + "step": 5230 + }, + { + "epoch": 0.13, + "learning_rate": 1.9522622707241996e-05, + "loss": 1.1307, + "step": 5231 + }, + { + "epoch": 0.13, + "learning_rate": 1.9522384896596102e-05, + "loss": 1.1899, + "step": 5232 + }, + { + "epoch": 0.13, + "learning_rate": 1.9522147028180138e-05, + "loss": 0.8172, + "step": 5233 + }, + { + "epoch": 0.13, + "learning_rate": 1.952190910199555e-05, + "loss": 1.2428, + "step": 5234 + }, + { + "epoch": 0.13, + "learning_rate": 1.9521671118043776e-05, + "loss": 1.1946, + "step": 5235 + }, + { + "epoch": 0.13, + "learning_rate": 1.9521433076326267e-05, + "loss": 1.1995, + "step": 5236 + }, + { + "epoch": 0.13, + "learning_rate": 1.952119497684446e-05, + "loss": 1.2488, + "step": 5237 + }, + { + "epoch": 0.13, + "learning_rate": 1.9520956819599804e-05, + "loss": 1.4389, + "step": 5238 + }, + { + "epoch": 0.13, + "learning_rate": 1.9520718604593744e-05, + "loss": 1.1206, + "step": 5239 + }, + { + "epoch": 0.13, + "learning_rate": 1.9520480331827718e-05, + "loss": 1.0142, + "step": 5240 + }, + { + "epoch": 0.13, + "learning_rate": 1.952024200130318e-05, + "loss": 1.1852, + "step": 5241 + }, + { + "epoch": 0.13, + "learning_rate": 1.9520003613021577e-05, + "loss": 1.28, + "step": 5242 + }, + { + "epoch": 0.13, + "learning_rate": 1.9519765166984346e-05, + "loss": 1.1313, + "step": 5243 + }, + { + "epoch": 0.13, + "learning_rate": 1.951952666319294e-05, + "loss": 1.0728, + "step": 5244 + }, + { + "epoch": 0.13, + "learning_rate": 1.9519288101648807e-05, + "loss": 1.0391, + "step": 5245 + }, + { + "epoch": 0.13, + "learning_rate": 1.9519049482353393e-05, + "loss": 1.3595, + "step": 5246 + }, + { + "epoch": 0.13, + "learning_rate": 1.9518810805308142e-05, + "loss": 1.2821, + "step": 5247 + }, + { + "epoch": 0.13, + "learning_rate": 1.9518572070514507e-05, + "loss": 0.9234, + "step": 5248 + }, + { + "epoch": 0.13, + "learning_rate": 1.9518333277973932e-05, + "loss": 1.0532, + "step": 5249 + }, + { + "epoch": 0.13, + "learning_rate": 1.9518094427687866e-05, + "loss": 1.0198, + "step": 5250 + }, + { + "epoch": 0.13, + "learning_rate": 1.9517855519657764e-05, + "loss": 1.2829, + "step": 5251 + }, + { + "epoch": 0.13, + "learning_rate": 1.951761655388507e-05, + "loss": 1.069, + "step": 5252 + }, + { + "epoch": 0.13, + "learning_rate": 1.9517377530371236e-05, + "loss": 1.2512, + "step": 5253 + }, + { + "epoch": 0.13, + "learning_rate": 1.9517138449117706e-05, + "loss": 1.4147, + "step": 5254 + }, + { + "epoch": 0.13, + "learning_rate": 1.951689931012594e-05, + "loss": 1.2953, + "step": 5255 + }, + { + "epoch": 0.13, + "learning_rate": 1.9516660113397386e-05, + "loss": 0.8391, + "step": 5256 + }, + { + "epoch": 0.13, + "learning_rate": 1.9516420858933492e-05, + "loss": 1.0144, + "step": 5257 + }, + { + "epoch": 0.13, + "learning_rate": 1.951618154673571e-05, + "loss": 1.0745, + "step": 5258 + }, + { + "epoch": 0.13, + "learning_rate": 1.9515942176805492e-05, + "loss": 1.2335, + "step": 5259 + }, + { + "epoch": 0.13, + "learning_rate": 1.9515702749144293e-05, + "loss": 0.8136, + "step": 5260 + }, + { + "epoch": 0.13, + "learning_rate": 1.9515463263753566e-05, + "loss": 1.1869, + "step": 5261 + }, + { + "epoch": 0.13, + "learning_rate": 1.951522372063476e-05, + "loss": 1.1863, + "step": 5262 + }, + { + "epoch": 0.13, + "learning_rate": 1.951498411978933e-05, + "loss": 1.2062, + "step": 5263 + }, + { + "epoch": 0.13, + "learning_rate": 1.9514744461218725e-05, + "loss": 1.0624, + "step": 5264 + }, + { + "epoch": 0.13, + "learning_rate": 1.9514504744924405e-05, + "loss": 1.0659, + "step": 5265 + }, + { + "epoch": 0.13, + "learning_rate": 1.9514264970907825e-05, + "loss": 1.1449, + "step": 5266 + }, + { + "epoch": 0.13, + "learning_rate": 1.9514025139170435e-05, + "loss": 0.9231, + "step": 5267 + }, + { + "epoch": 0.13, + "learning_rate": 1.9513785249713697e-05, + "loss": 1.1388, + "step": 5268 + }, + { + "epoch": 0.13, + "learning_rate": 1.9513545302539056e-05, + "loss": 1.351, + "step": 5269 + }, + { + "epoch": 0.13, + "learning_rate": 1.9513305297647976e-05, + "loss": 1.0258, + "step": 5270 + }, + { + "epoch": 0.13, + "learning_rate": 1.951306523504191e-05, + "loss": 1.0735, + "step": 5271 + }, + { + "epoch": 0.13, + "learning_rate": 1.9512825114722314e-05, + "loss": 1.1507, + "step": 5272 + }, + { + "epoch": 0.13, + "learning_rate": 1.951258493669065e-05, + "loss": 1.1525, + "step": 5273 + }, + { + "epoch": 0.13, + "learning_rate": 1.9512344700948363e-05, + "loss": 1.2618, + "step": 5274 + }, + { + "epoch": 0.13, + "learning_rate": 1.9512104407496926e-05, + "loss": 1.1737, + "step": 5275 + }, + { + "epoch": 0.13, + "learning_rate": 1.9511864056337784e-05, + "loss": 1.2365, + "step": 5276 + }, + { + "epoch": 0.13, + "learning_rate": 1.9511623647472403e-05, + "loss": 1.3907, + "step": 5277 + }, + { + "epoch": 0.13, + "learning_rate": 1.9511383180902237e-05, + "loss": 1.4006, + "step": 5278 + }, + { + "epoch": 0.13, + "learning_rate": 1.9511142656628745e-05, + "loss": 0.9487, + "step": 5279 + }, + { + "epoch": 0.13, + "learning_rate": 1.951090207465339e-05, + "loss": 0.7964, + "step": 5280 + }, + { + "epoch": 0.13, + "learning_rate": 1.951066143497763e-05, + "loss": 1.0411, + "step": 5281 + }, + { + "epoch": 0.13, + "learning_rate": 1.951042073760292e-05, + "loss": 1.073, + "step": 5282 + }, + { + "epoch": 0.13, + "learning_rate": 1.951017998253073e-05, + "loss": 1.0217, + "step": 5283 + }, + { + "epoch": 0.13, + "learning_rate": 1.950993916976251e-05, + "loss": 1.2809, + "step": 5284 + }, + { + "epoch": 0.13, + "learning_rate": 1.9509698299299727e-05, + "loss": 1.0129, + "step": 5285 + }, + { + "epoch": 0.13, + "learning_rate": 1.9509457371143843e-05, + "loss": 1.1131, + "step": 5286 + }, + { + "epoch": 0.13, + "learning_rate": 1.9509216385296317e-05, + "loss": 1.2302, + "step": 5287 + }, + { + "epoch": 0.13, + "learning_rate": 1.950897534175861e-05, + "loss": 1.251, + "step": 5288 + }, + { + "epoch": 0.13, + "learning_rate": 1.950873424053219e-05, + "loss": 1.1626, + "step": 5289 + }, + { + "epoch": 0.13, + "learning_rate": 1.9508493081618515e-05, + "loss": 0.9884, + "step": 5290 + }, + { + "epoch": 0.13, + "learning_rate": 1.950825186501905e-05, + "loss": 1.2626, + "step": 5291 + }, + { + "epoch": 0.13, + "learning_rate": 1.9508010590735252e-05, + "loss": 1.1384, + "step": 5292 + }, + { + "epoch": 0.13, + "learning_rate": 1.9507769258768594e-05, + "loss": 1.1507, + "step": 5293 + }, + { + "epoch": 0.13, + "learning_rate": 1.9507527869120534e-05, + "loss": 1.2329, + "step": 5294 + }, + { + "epoch": 0.13, + "learning_rate": 1.9507286421792542e-05, + "loss": 1.3813, + "step": 5295 + }, + { + "epoch": 0.13, + "learning_rate": 1.950704491678608e-05, + "loss": 0.8342, + "step": 5296 + }, + { + "epoch": 0.13, + "learning_rate": 1.9506803354102607e-05, + "loss": 1.3332, + "step": 5297 + }, + { + "epoch": 0.13, + "learning_rate": 1.95065617337436e-05, + "loss": 1.124, + "step": 5298 + }, + { + "epoch": 0.13, + "learning_rate": 1.9506320055710517e-05, + "loss": 1.1837, + "step": 5299 + }, + { + "epoch": 0.13, + "learning_rate": 1.9506078320004825e-05, + "loss": 1.0801, + "step": 5300 + }, + { + "epoch": 0.13, + "learning_rate": 1.9505836526627994e-05, + "loss": 0.9769, + "step": 5301 + }, + { + "epoch": 0.13, + "learning_rate": 1.950559467558149e-05, + "loss": 0.8499, + "step": 5302 + }, + { + "epoch": 0.13, + "learning_rate": 1.9505352766866778e-05, + "loss": 1.3252, + "step": 5303 + }, + { + "epoch": 0.13, + "learning_rate": 1.9505110800485324e-05, + "loss": 1.0684, + "step": 5304 + }, + { + "epoch": 0.13, + "learning_rate": 1.95048687764386e-05, + "loss": 1.0586, + "step": 5305 + }, + { + "epoch": 0.13, + "learning_rate": 1.950462669472807e-05, + "loss": 1.1851, + "step": 5306 + }, + { + "epoch": 0.13, + "learning_rate": 1.950438455535521e-05, + "loss": 1.0729, + "step": 5307 + }, + { + "epoch": 0.13, + "learning_rate": 1.950414235832148e-05, + "loss": 1.0221, + "step": 5308 + }, + { + "epoch": 0.13, + "learning_rate": 1.9503900103628357e-05, + "loss": 0.9047, + "step": 5309 + }, + { + "epoch": 0.13, + "learning_rate": 1.950365779127731e-05, + "loss": 1.028, + "step": 5310 + }, + { + "epoch": 0.13, + "learning_rate": 1.95034154212698e-05, + "loss": 1.3031, + "step": 5311 + }, + { + "epoch": 0.13, + "learning_rate": 1.950317299360731e-05, + "loss": 1.1902, + "step": 5312 + }, + { + "epoch": 0.13, + "learning_rate": 1.9502930508291303e-05, + "loss": 1.1718, + "step": 5313 + }, + { + "epoch": 0.13, + "learning_rate": 1.950268796532325e-05, + "loss": 0.9719, + "step": 5314 + }, + { + "epoch": 0.13, + "learning_rate": 1.9502445364704626e-05, + "loss": 1.3627, + "step": 5315 + }, + { + "epoch": 0.13, + "learning_rate": 1.95022027064369e-05, + "loss": 1.2723, + "step": 5316 + }, + { + "epoch": 0.13, + "learning_rate": 1.9501959990521543e-05, + "loss": 1.2566, + "step": 5317 + }, + { + "epoch": 0.13, + "learning_rate": 1.9501717216960035e-05, + "loss": 1.1658, + "step": 5318 + }, + { + "epoch": 0.13, + "learning_rate": 1.950147438575384e-05, + "loss": 1.2736, + "step": 5319 + }, + { + "epoch": 0.13, + "learning_rate": 1.9501231496904435e-05, + "loss": 1.1144, + "step": 5320 + }, + { + "epoch": 0.13, + "learning_rate": 1.9500988550413294e-05, + "loss": 1.2655, + "step": 5321 + }, + { + "epoch": 0.13, + "learning_rate": 1.9500745546281893e-05, + "loss": 1.1455, + "step": 5322 + }, + { + "epoch": 0.13, + "learning_rate": 1.95005024845117e-05, + "loss": 1.0427, + "step": 5323 + }, + { + "epoch": 0.13, + "learning_rate": 1.9500259365104192e-05, + "loss": 1.1353, + "step": 5324 + }, + { + "epoch": 0.13, + "learning_rate": 1.9500016188060847e-05, + "loss": 1.445, + "step": 5325 + }, + { + "epoch": 0.13, + "learning_rate": 1.949977295338314e-05, + "loss": 1.3454, + "step": 5326 + }, + { + "epoch": 0.13, + "learning_rate": 1.949952966107254e-05, + "loss": 1.2961, + "step": 5327 + }, + { + "epoch": 0.13, + "learning_rate": 1.9499286311130533e-05, + "loss": 1.0307, + "step": 5328 + }, + { + "epoch": 0.13, + "learning_rate": 1.949904290355859e-05, + "loss": 1.4173, + "step": 5329 + }, + { + "epoch": 0.13, + "learning_rate": 1.9498799438358186e-05, + "loss": 1.2907, + "step": 5330 + }, + { + "epoch": 0.13, + "learning_rate": 1.9498555915530798e-05, + "loss": 1.1544, + "step": 5331 + }, + { + "epoch": 0.13, + "learning_rate": 1.949831233507791e-05, + "loss": 1.331, + "step": 5332 + }, + { + "epoch": 0.13, + "learning_rate": 1.9498068697000994e-05, + "loss": 1.0037, + "step": 5333 + }, + { + "epoch": 0.13, + "learning_rate": 1.949782500130153e-05, + "loss": 1.1953, + "step": 5334 + }, + { + "epoch": 0.13, + "learning_rate": 1.9497581247980993e-05, + "loss": 1.3638, + "step": 5335 + }, + { + "epoch": 0.13, + "learning_rate": 1.9497337437040867e-05, + "loss": 1.1707, + "step": 5336 + }, + { + "epoch": 0.13, + "learning_rate": 1.9497093568482626e-05, + "loss": 1.2879, + "step": 5337 + }, + { + "epoch": 0.13, + "learning_rate": 1.9496849642307754e-05, + "loss": 1.2071, + "step": 5338 + }, + { + "epoch": 0.13, + "learning_rate": 1.9496605658517728e-05, + "loss": 1.2692, + "step": 5339 + }, + { + "epoch": 0.13, + "learning_rate": 1.949636161711403e-05, + "loss": 1.2973, + "step": 5340 + }, + { + "epoch": 0.13, + "learning_rate": 1.949611751809814e-05, + "loss": 1.1925, + "step": 5341 + }, + { + "epoch": 0.13, + "learning_rate": 1.9495873361471538e-05, + "loss": 1.1368, + "step": 5342 + }, + { + "epoch": 0.13, + "learning_rate": 1.9495629147235705e-05, + "loss": 1.2706, + "step": 5343 + }, + { + "epoch": 0.13, + "learning_rate": 1.9495384875392125e-05, + "loss": 1.2448, + "step": 5344 + }, + { + "epoch": 0.13, + "learning_rate": 1.9495140545942277e-05, + "loss": 1.372, + "step": 5345 + }, + { + "epoch": 0.13, + "learning_rate": 1.9494896158887647e-05, + "loss": 1.1477, + "step": 5346 + }, + { + "epoch": 0.13, + "learning_rate": 1.9494651714229715e-05, + "loss": 1.3487, + "step": 5347 + }, + { + "epoch": 0.13, + "learning_rate": 1.949440721196996e-05, + "loss": 1.0379, + "step": 5348 + }, + { + "epoch": 0.13, + "learning_rate": 1.9494162652109876e-05, + "loss": 1.3368, + "step": 5349 + }, + { + "epoch": 0.13, + "learning_rate": 1.9493918034650934e-05, + "loss": 1.2603, + "step": 5350 + }, + { + "epoch": 0.13, + "learning_rate": 1.9493673359594633e-05, + "loss": 1.3247, + "step": 5351 + }, + { + "epoch": 0.13, + "learning_rate": 1.9493428626942443e-05, + "loss": 1.0998, + "step": 5352 + }, + { + "epoch": 0.13, + "learning_rate": 1.9493183836695854e-05, + "loss": 1.0881, + "step": 5353 + }, + { + "epoch": 0.13, + "learning_rate": 1.9492938988856354e-05, + "loss": 1.0728, + "step": 5354 + }, + { + "epoch": 0.13, + "learning_rate": 1.9492694083425424e-05, + "loss": 1.27, + "step": 5355 + }, + { + "epoch": 0.13, + "learning_rate": 1.949244912040455e-05, + "loss": 0.9878, + "step": 5356 + }, + { + "epoch": 0.13, + "learning_rate": 1.9492204099795223e-05, + "loss": 0.9812, + "step": 5357 + }, + { + "epoch": 0.13, + "learning_rate": 1.9491959021598927e-05, + "loss": 1.0534, + "step": 5358 + }, + { + "epoch": 0.13, + "learning_rate": 1.9491713885817147e-05, + "loss": 1.2055, + "step": 5359 + }, + { + "epoch": 0.13, + "learning_rate": 1.9491468692451373e-05, + "loss": 1.2852, + "step": 5360 + }, + { + "epoch": 0.13, + "learning_rate": 1.949122344150309e-05, + "loss": 1.4034, + "step": 5361 + }, + { + "epoch": 0.13, + "learning_rate": 1.9490978132973785e-05, + "loss": 1.1639, + "step": 5362 + }, + { + "epoch": 0.13, + "learning_rate": 1.9490732766864952e-05, + "loss": 1.2135, + "step": 5363 + }, + { + "epoch": 0.13, + "learning_rate": 1.9490487343178072e-05, + "loss": 1.1039, + "step": 5364 + }, + { + "epoch": 0.13, + "learning_rate": 1.949024186191464e-05, + "loss": 1.3047, + "step": 5365 + }, + { + "epoch": 0.13, + "learning_rate": 1.948999632307614e-05, + "loss": 0.9955, + "step": 5366 + }, + { + "epoch": 0.13, + "learning_rate": 1.9489750726664064e-05, + "loss": 1.2168, + "step": 5367 + }, + { + "epoch": 0.13, + "learning_rate": 1.9489505072679907e-05, + "loss": 0.9334, + "step": 5368 + }, + { + "epoch": 0.13, + "learning_rate": 1.9489259361125153e-05, + "loss": 1.245, + "step": 5369 + }, + { + "epoch": 0.13, + "learning_rate": 1.9489013592001293e-05, + "loss": 1.1139, + "step": 5370 + }, + { + "epoch": 0.13, + "learning_rate": 1.9488767765309818e-05, + "loss": 1.2335, + "step": 5371 + }, + { + "epoch": 0.13, + "learning_rate": 1.9488521881052225e-05, + "loss": 1.235, + "step": 5372 + }, + { + "epoch": 0.13, + "learning_rate": 1.948827593923e-05, + "loss": 1.2306, + "step": 5373 + }, + { + "epoch": 0.13, + "learning_rate": 1.9488029939844634e-05, + "loss": 1.1807, + "step": 5374 + }, + { + "epoch": 0.13, + "learning_rate": 1.9487783882897627e-05, + "loss": 0.9691, + "step": 5375 + }, + { + "epoch": 0.13, + "learning_rate": 1.9487537768390465e-05, + "loss": 1.173, + "step": 5376 + }, + { + "epoch": 0.13, + "learning_rate": 1.948729159632464e-05, + "loss": 1.1605, + "step": 5377 + }, + { + "epoch": 0.13, + "learning_rate": 1.9487045366701652e-05, + "loss": 1.3261, + "step": 5378 + }, + { + "epoch": 0.13, + "learning_rate": 1.9486799079522992e-05, + "loss": 1.1543, + "step": 5379 + }, + { + "epoch": 0.13, + "learning_rate": 1.948655273479015e-05, + "loss": 1.2654, + "step": 5380 + }, + { + "epoch": 0.13, + "learning_rate": 1.9486306332504626e-05, + "loss": 1.128, + "step": 5381 + }, + { + "epoch": 0.13, + "learning_rate": 1.948605987266791e-05, + "loss": 1.2729, + "step": 5382 + }, + { + "epoch": 0.13, + "learning_rate": 1.9485813355281503e-05, + "loss": 0.9937, + "step": 5383 + }, + { + "epoch": 0.13, + "learning_rate": 1.94855667803469e-05, + "loss": 1.2885, + "step": 5384 + }, + { + "epoch": 0.13, + "learning_rate": 1.948532014786559e-05, + "loss": 1.2599, + "step": 5385 + }, + { + "epoch": 0.13, + "learning_rate": 1.9485073457839072e-05, + "loss": 1.1631, + "step": 5386 + }, + { + "epoch": 0.13, + "learning_rate": 1.9484826710268847e-05, + "loss": 1.4379, + "step": 5387 + }, + { + "epoch": 0.13, + "learning_rate": 1.948457990515641e-05, + "loss": 1.3258, + "step": 5388 + }, + { + "epoch": 0.13, + "learning_rate": 1.948433304250326e-05, + "loss": 1.3006, + "step": 5389 + }, + { + "epoch": 0.13, + "learning_rate": 1.9484086122310887e-05, + "loss": 0.9539, + "step": 5390 + }, + { + "epoch": 0.13, + "learning_rate": 1.9483839144580797e-05, + "loss": 1.372, + "step": 5391 + }, + { + "epoch": 0.13, + "learning_rate": 1.9483592109314487e-05, + "loss": 1.1084, + "step": 5392 + }, + { + "epoch": 0.13, + "learning_rate": 1.948334501651345e-05, + "loss": 1.2256, + "step": 5393 + }, + { + "epoch": 0.13, + "learning_rate": 1.9483097866179194e-05, + "loss": 1.2649, + "step": 5394 + }, + { + "epoch": 0.13, + "learning_rate": 1.9482850658313213e-05, + "loss": 1.0582, + "step": 5395 + }, + { + "epoch": 0.13, + "learning_rate": 1.9482603392917006e-05, + "loss": 1.3217, + "step": 5396 + }, + { + "epoch": 0.13, + "learning_rate": 1.9482356069992075e-05, + "loss": 1.1399, + "step": 5397 + }, + { + "epoch": 0.13, + "learning_rate": 1.948210868953992e-05, + "loss": 1.2691, + "step": 5398 + }, + { + "epoch": 0.13, + "learning_rate": 1.9481861251562044e-05, + "loss": 1.1837, + "step": 5399 + }, + { + "epoch": 0.13, + "learning_rate": 1.9481613756059944e-05, + "loss": 1.0959, + "step": 5400 + }, + { + "epoch": 0.13, + "learning_rate": 1.9481366203035122e-05, + "loss": 1.1307, + "step": 5401 + }, + { + "epoch": 0.13, + "learning_rate": 1.9481118592489086e-05, + "loss": 0.8597, + "step": 5402 + }, + { + "epoch": 0.13, + "learning_rate": 1.948087092442333e-05, + "loss": 1.0169, + "step": 5403 + }, + { + "epoch": 0.13, + "learning_rate": 1.9480623198839362e-05, + "loss": 1.0203, + "step": 5404 + }, + { + "epoch": 0.13, + "learning_rate": 1.948037541573868e-05, + "loss": 0.8198, + "step": 5405 + }, + { + "epoch": 0.13, + "learning_rate": 1.9480127575122795e-05, + "loss": 1.1325, + "step": 5406 + }, + { + "epoch": 0.13, + "learning_rate": 1.94798796769932e-05, + "loss": 1.0105, + "step": 5407 + }, + { + "epoch": 0.13, + "learning_rate": 1.9479631721351412e-05, + "loss": 1.1731, + "step": 5408 + }, + { + "epoch": 0.13, + "learning_rate": 1.947938370819892e-05, + "loss": 0.7655, + "step": 5409 + }, + { + "epoch": 0.13, + "learning_rate": 1.947913563753724e-05, + "loss": 1.1297, + "step": 5410 + }, + { + "epoch": 0.13, + "learning_rate": 1.9478887509367876e-05, + "loss": 1.3901, + "step": 5411 + }, + { + "epoch": 0.13, + "learning_rate": 1.947863932369233e-05, + "loss": 0.9998, + "step": 5412 + }, + { + "epoch": 0.13, + "learning_rate": 1.947839108051211e-05, + "loss": 1.284, + "step": 5413 + }, + { + "epoch": 0.13, + "learning_rate": 1.9478142779828717e-05, + "loss": 1.1352, + "step": 5414 + }, + { + "epoch": 0.13, + "learning_rate": 1.9477894421643662e-05, + "loss": 1.1896, + "step": 5415 + }, + { + "epoch": 0.13, + "learning_rate": 1.9477646005958454e-05, + "loss": 1.0244, + "step": 5416 + }, + { + "epoch": 0.13, + "learning_rate": 1.9477397532774596e-05, + "loss": 1.2907, + "step": 5417 + }, + { + "epoch": 0.13, + "learning_rate": 1.9477149002093595e-05, + "loss": 1.4077, + "step": 5418 + }, + { + "epoch": 0.13, + "learning_rate": 1.947690041391696e-05, + "loss": 1.0301, + "step": 5419 + }, + { + "epoch": 0.13, + "learning_rate": 1.9476651768246203e-05, + "loss": 1.0392, + "step": 5420 + }, + { + "epoch": 0.13, + "learning_rate": 1.9476403065082824e-05, + "loss": 1.1911, + "step": 5421 + }, + { + "epoch": 0.13, + "learning_rate": 1.947615430442834e-05, + "loss": 1.2602, + "step": 5422 + }, + { + "epoch": 0.13, + "learning_rate": 1.9475905486284252e-05, + "loss": 1.3265, + "step": 5423 + }, + { + "epoch": 0.13, + "learning_rate": 1.947565661065208e-05, + "loss": 1.3326, + "step": 5424 + }, + { + "epoch": 0.13, + "learning_rate": 1.9475407677533322e-05, + "loss": 1.1922, + "step": 5425 + }, + { + "epoch": 0.13, + "learning_rate": 1.9475158686929498e-05, + "loss": 1.0962, + "step": 5426 + }, + { + "epoch": 0.13, + "learning_rate": 1.9474909638842115e-05, + "loss": 1.1271, + "step": 5427 + }, + { + "epoch": 0.13, + "learning_rate": 1.9474660533272684e-05, + "loss": 1.1226, + "step": 5428 + }, + { + "epoch": 0.13, + "learning_rate": 1.9474411370222717e-05, + "loss": 1.4189, + "step": 5429 + }, + { + "epoch": 0.13, + "learning_rate": 1.9474162149693724e-05, + "loss": 0.9961, + "step": 5430 + }, + { + "epoch": 0.13, + "learning_rate": 1.9473912871687216e-05, + "loss": 1.467, + "step": 5431 + }, + { + "epoch": 0.13, + "learning_rate": 1.947366353620471e-05, + "loss": 1.2849, + "step": 5432 + }, + { + "epoch": 0.13, + "learning_rate": 1.9473414143247712e-05, + "loss": 1.2288, + "step": 5433 + }, + { + "epoch": 0.13, + "learning_rate": 1.947316469281774e-05, + "loss": 1.274, + "step": 5434 + }, + { + "epoch": 0.13, + "learning_rate": 1.9472915184916307e-05, + "loss": 1.0841, + "step": 5435 + }, + { + "epoch": 0.13, + "learning_rate": 1.9472665619544927e-05, + "loss": 1.1438, + "step": 5436 + }, + { + "epoch": 0.13, + "learning_rate": 1.947241599670511e-05, + "loss": 1.3127, + "step": 5437 + }, + { + "epoch": 0.13, + "learning_rate": 1.9472166316398376e-05, + "loss": 1.4305, + "step": 5438 + }, + { + "epoch": 0.13, + "learning_rate": 1.9471916578626234e-05, + "loss": 1.5162, + "step": 5439 + }, + { + "epoch": 0.13, + "learning_rate": 1.9471666783390204e-05, + "loss": 1.373, + "step": 5440 + }, + { + "epoch": 0.13, + "learning_rate": 1.9471416930691798e-05, + "loss": 1.2396, + "step": 5441 + }, + { + "epoch": 0.13, + "learning_rate": 1.9471167020532533e-05, + "loss": 1.3076, + "step": 5442 + }, + { + "epoch": 0.13, + "learning_rate": 1.9470917052913928e-05, + "loss": 1.2362, + "step": 5443 + }, + { + "epoch": 0.13, + "learning_rate": 1.9470667027837497e-05, + "loss": 1.3085, + "step": 5444 + }, + { + "epoch": 0.13, + "learning_rate": 1.9470416945304754e-05, + "loss": 1.1534, + "step": 5445 + }, + { + "epoch": 0.13, + "learning_rate": 1.9470166805317217e-05, + "loss": 0.9303, + "step": 5446 + }, + { + "epoch": 0.13, + "learning_rate": 1.9469916607876412e-05, + "loss": 1.2621, + "step": 5447 + }, + { + "epoch": 0.13, + "learning_rate": 1.9469666352983845e-05, + "loss": 1.4804, + "step": 5448 + }, + { + "epoch": 0.13, + "learning_rate": 1.9469416040641045e-05, + "loss": 0.9582, + "step": 5449 + }, + { + "epoch": 0.13, + "learning_rate": 1.946916567084952e-05, + "loss": 0.7996, + "step": 5450 + }, + { + "epoch": 0.13, + "learning_rate": 1.9468915243610796e-05, + "loss": 1.2102, + "step": 5451 + }, + { + "epoch": 0.13, + "learning_rate": 1.9468664758926393e-05, + "loss": 1.4342, + "step": 5452 + }, + { + "epoch": 0.13, + "learning_rate": 1.9468414216797825e-05, + "loss": 1.3128, + "step": 5453 + }, + { + "epoch": 0.13, + "learning_rate": 1.9468163617226613e-05, + "loss": 1.1938, + "step": 5454 + }, + { + "epoch": 0.13, + "learning_rate": 1.9467912960214282e-05, + "loss": 1.3716, + "step": 5455 + }, + { + "epoch": 0.13, + "learning_rate": 1.9467662245762354e-05, + "loss": 1.1144, + "step": 5456 + }, + { + "epoch": 0.13, + "learning_rate": 1.946741147387234e-05, + "loss": 1.1794, + "step": 5457 + }, + { + "epoch": 0.13, + "learning_rate": 1.9467160644545767e-05, + "loss": 1.0301, + "step": 5458 + }, + { + "epoch": 0.13, + "learning_rate": 1.946690975778416e-05, + "loss": 1.1668, + "step": 5459 + }, + { + "epoch": 0.13, + "learning_rate": 1.946665881358904e-05, + "loss": 0.9894, + "step": 5460 + }, + { + "epoch": 0.13, + "learning_rate": 1.9466407811961927e-05, + "loss": 1.1963, + "step": 5461 + }, + { + "epoch": 0.13, + "learning_rate": 1.9466156752904344e-05, + "loss": 0.8595, + "step": 5462 + }, + { + "epoch": 0.13, + "learning_rate": 1.9465905636417815e-05, + "loss": 1.1554, + "step": 5463 + }, + { + "epoch": 0.13, + "learning_rate": 1.9465654462503862e-05, + "loss": 1.1545, + "step": 5464 + }, + { + "epoch": 0.13, + "learning_rate": 1.9465403231164005e-05, + "loss": 1.2134, + "step": 5465 + }, + { + "epoch": 0.13, + "learning_rate": 1.946515194239978e-05, + "loss": 1.2257, + "step": 5466 + }, + { + "epoch": 0.13, + "learning_rate": 1.9464900596212702e-05, + "loss": 1.2649, + "step": 5467 + }, + { + "epoch": 0.13, + "learning_rate": 1.94646491926043e-05, + "loss": 1.2918, + "step": 5468 + }, + { + "epoch": 0.13, + "learning_rate": 1.9464397731576093e-05, + "loss": 1.0443, + "step": 5469 + }, + { + "epoch": 0.13, + "learning_rate": 1.9464146213129615e-05, + "loss": 0.9819, + "step": 5470 + }, + { + "epoch": 0.13, + "learning_rate": 1.946389463726639e-05, + "loss": 1.0876, + "step": 5471 + }, + { + "epoch": 0.13, + "learning_rate": 1.9463643003987938e-05, + "loss": 1.1764, + "step": 5472 + }, + { + "epoch": 0.13, + "learning_rate": 1.9463391313295792e-05, + "loss": 1.1826, + "step": 5473 + }, + { + "epoch": 0.13, + "learning_rate": 1.9463139565191476e-05, + "loss": 1.3068, + "step": 5474 + }, + { + "epoch": 0.13, + "learning_rate": 1.946288775967652e-05, + "loss": 1.2165, + "step": 5475 + }, + { + "epoch": 0.13, + "learning_rate": 1.9462635896752448e-05, + "loss": 1.1932, + "step": 5476 + }, + { + "epoch": 0.13, + "learning_rate": 1.946238397642079e-05, + "loss": 1.1951, + "step": 5477 + }, + { + "epoch": 0.13, + "learning_rate": 1.9462131998683073e-05, + "loss": 1.1477, + "step": 5478 + }, + { + "epoch": 0.13, + "learning_rate": 1.9461879963540825e-05, + "loss": 1.2778, + "step": 5479 + }, + { + "epoch": 0.13, + "learning_rate": 1.9461627870995585e-05, + "loss": 0.9508, + "step": 5480 + }, + { + "epoch": 0.13, + "learning_rate": 1.9461375721048866e-05, + "loss": 0.9559, + "step": 5481 + }, + { + "epoch": 0.13, + "learning_rate": 1.9461123513702208e-05, + "loss": 1.1496, + "step": 5482 + }, + { + "epoch": 0.13, + "learning_rate": 1.946087124895714e-05, + "loss": 1.0986, + "step": 5483 + }, + { + "epoch": 0.13, + "learning_rate": 1.9460618926815195e-05, + "loss": 1.2949, + "step": 5484 + }, + { + "epoch": 0.13, + "learning_rate": 1.9460366547277892e-05, + "loss": 1.0339, + "step": 5485 + }, + { + "epoch": 0.13, + "learning_rate": 1.9460114110346775e-05, + "loss": 1.3389, + "step": 5486 + }, + { + "epoch": 0.13, + "learning_rate": 1.9459861616023373e-05, + "loss": 1.0571, + "step": 5487 + }, + { + "epoch": 0.13, + "learning_rate": 1.9459609064309212e-05, + "loss": 1.2071, + "step": 5488 + }, + { + "epoch": 0.13, + "learning_rate": 1.9459356455205828e-05, + "loss": 1.0362, + "step": 5489 + }, + { + "epoch": 0.13, + "learning_rate": 1.9459103788714756e-05, + "loss": 1.2518, + "step": 5490 + }, + { + "epoch": 0.13, + "learning_rate": 1.9458851064837526e-05, + "loss": 1.4334, + "step": 5491 + }, + { + "epoch": 0.13, + "learning_rate": 1.945859828357567e-05, + "loss": 0.877, + "step": 5492 + }, + { + "epoch": 0.13, + "learning_rate": 1.9458345444930723e-05, + "loss": 0.9463, + "step": 5493 + }, + { + "epoch": 0.13, + "learning_rate": 1.945809254890422e-05, + "loss": 1.1067, + "step": 5494 + }, + { + "epoch": 0.13, + "learning_rate": 1.9457839595497693e-05, + "loss": 1.3232, + "step": 5495 + }, + { + "epoch": 0.13, + "learning_rate": 1.9457586584712678e-05, + "loss": 1.0393, + "step": 5496 + }, + { + "epoch": 0.13, + "learning_rate": 1.945733351655071e-05, + "loss": 1.1556, + "step": 5497 + }, + { + "epoch": 0.13, + "learning_rate": 1.9457080391013325e-05, + "loss": 1.1884, + "step": 5498 + }, + { + "epoch": 0.13, + "learning_rate": 1.9456827208102057e-05, + "loss": 1.1962, + "step": 5499 + }, + { + "epoch": 0.13, + "learning_rate": 1.945657396781844e-05, + "loss": 0.9569, + "step": 5500 + }, + { + "epoch": 0.13, + "learning_rate": 1.9456320670164018e-05, + "loss": 1.2927, + "step": 5501 + }, + { + "epoch": 0.13, + "learning_rate": 1.945606731514032e-05, + "loss": 1.1008, + "step": 5502 + }, + { + "epoch": 0.13, + "learning_rate": 1.9455813902748887e-05, + "loss": 1.0175, + "step": 5503 + }, + { + "epoch": 0.13, + "learning_rate": 1.9455560432991253e-05, + "loss": 1.2872, + "step": 5504 + }, + { + "epoch": 0.13, + "learning_rate": 1.945530690586896e-05, + "loss": 0.9982, + "step": 5505 + }, + { + "epoch": 0.13, + "learning_rate": 1.9455053321383542e-05, + "loss": 1.0895, + "step": 5506 + }, + { + "epoch": 0.13, + "learning_rate": 1.945479967953654e-05, + "loss": 1.285, + "step": 5507 + }, + { + "epoch": 0.13, + "learning_rate": 1.9454545980329493e-05, + "loss": 1.28, + "step": 5508 + }, + { + "epoch": 0.13, + "learning_rate": 1.945429222376394e-05, + "loss": 1.1554, + "step": 5509 + }, + { + "epoch": 0.13, + "learning_rate": 1.945403840984142e-05, + "loss": 1.1815, + "step": 5510 + }, + { + "epoch": 0.13, + "learning_rate": 1.9453784538563473e-05, + "loss": 1.2211, + "step": 5511 + }, + { + "epoch": 0.13, + "learning_rate": 1.9453530609931635e-05, + "loss": 1.1542, + "step": 5512 + }, + { + "epoch": 0.13, + "learning_rate": 1.9453276623947454e-05, + "loss": 1.1372, + "step": 5513 + }, + { + "epoch": 0.13, + "learning_rate": 1.9453022580612468e-05, + "loss": 1.2229, + "step": 5514 + }, + { + "epoch": 0.13, + "learning_rate": 1.9452768479928214e-05, + "loss": 1.2891, + "step": 5515 + }, + { + "epoch": 0.13, + "learning_rate": 1.9452514321896242e-05, + "loss": 1.2004, + "step": 5516 + }, + { + "epoch": 0.13, + "learning_rate": 1.9452260106518083e-05, + "loss": 0.9485, + "step": 5517 + }, + { + "epoch": 0.13, + "learning_rate": 1.945200583379529e-05, + "loss": 1.2608, + "step": 5518 + }, + { + "epoch": 0.13, + "learning_rate": 1.9451751503729398e-05, + "loss": 1.0227, + "step": 5519 + }, + { + "epoch": 0.13, + "learning_rate": 1.9451497116321954e-05, + "loss": 1.1755, + "step": 5520 + }, + { + "epoch": 0.13, + "learning_rate": 1.94512426715745e-05, + "loss": 1.1446, + "step": 5521 + }, + { + "epoch": 0.13, + "learning_rate": 1.9450988169488577e-05, + "loss": 1.0093, + "step": 5522 + }, + { + "epoch": 0.13, + "learning_rate": 1.9450733610065736e-05, + "loss": 1.0826, + "step": 5523 + }, + { + "epoch": 0.13, + "learning_rate": 1.9450478993307517e-05, + "loss": 0.9083, + "step": 5524 + }, + { + "epoch": 0.13, + "learning_rate": 1.9450224319215464e-05, + "loss": 0.9265, + "step": 5525 + }, + { + "epoch": 0.13, + "learning_rate": 1.944996958779112e-05, + "loss": 1.3378, + "step": 5526 + }, + { + "epoch": 0.13, + "learning_rate": 1.9449714799036038e-05, + "loss": 1.1679, + "step": 5527 + }, + { + "epoch": 0.13, + "learning_rate": 1.9449459952951756e-05, + "loss": 1.0347, + "step": 5528 + }, + { + "epoch": 0.13, + "learning_rate": 1.9449205049539827e-05, + "loss": 1.0546, + "step": 5529 + }, + { + "epoch": 0.13, + "learning_rate": 1.944895008880179e-05, + "loss": 1.1382, + "step": 5530 + }, + { + "epoch": 0.13, + "learning_rate": 1.9448695070739193e-05, + "loss": 0.7577, + "step": 5531 + }, + { + "epoch": 0.13, + "learning_rate": 1.944843999535359e-05, + "loss": 0.8158, + "step": 5532 + }, + { + "epoch": 0.13, + "learning_rate": 1.944818486264652e-05, + "loss": 1.4673, + "step": 5533 + }, + { + "epoch": 0.13, + "learning_rate": 1.944792967261954e-05, + "loss": 0.9428, + "step": 5534 + }, + { + "epoch": 0.13, + "learning_rate": 1.9447674425274188e-05, + "loss": 1.1324, + "step": 5535 + }, + { + "epoch": 0.13, + "learning_rate": 1.9447419120612018e-05, + "loss": 0.9659, + "step": 5536 + }, + { + "epoch": 0.13, + "learning_rate": 1.944716375863458e-05, + "loss": 1.1171, + "step": 5537 + }, + { + "epoch": 0.13, + "learning_rate": 1.9446908339343422e-05, + "loss": 1.3929, + "step": 5538 + }, + { + "epoch": 0.13, + "learning_rate": 1.9446652862740092e-05, + "loss": 1.0894, + "step": 5539 + }, + { + "epoch": 0.13, + "learning_rate": 1.9446397328826145e-05, + "loss": 1.264, + "step": 5540 + }, + { + "epoch": 0.13, + "learning_rate": 1.944614173760312e-05, + "loss": 1.1806, + "step": 5541 + }, + { + "epoch": 0.13, + "learning_rate": 1.944588608907258e-05, + "loss": 1.0419, + "step": 5542 + }, + { + "epoch": 0.13, + "learning_rate": 1.944563038323607e-05, + "loss": 1.2597, + "step": 5543 + }, + { + "epoch": 0.13, + "learning_rate": 1.9445374620095142e-05, + "loss": 1.5319, + "step": 5544 + }, + { + "epoch": 0.13, + "learning_rate": 1.9445118799651348e-05, + "loss": 0.7728, + "step": 5545 + }, + { + "epoch": 0.13, + "learning_rate": 1.944486292190624e-05, + "loss": 1.2726, + "step": 5546 + }, + { + "epoch": 0.13, + "learning_rate": 1.944460698686137e-05, + "loss": 1.1796, + "step": 5547 + }, + { + "epoch": 0.13, + "learning_rate": 1.944435099451829e-05, + "loss": 1.1539, + "step": 5548 + }, + { + "epoch": 0.13, + "learning_rate": 1.9444094944878554e-05, + "loss": 1.003, + "step": 5549 + }, + { + "epoch": 0.13, + "learning_rate": 1.9443838837943717e-05, + "loss": 1.3619, + "step": 5550 + }, + { + "epoch": 0.13, + "learning_rate": 1.9443582673715328e-05, + "loss": 1.0009, + "step": 5551 + }, + { + "epoch": 0.13, + "learning_rate": 1.9443326452194948e-05, + "loss": 0.8326, + "step": 5552 + }, + { + "epoch": 0.13, + "learning_rate": 1.9443070173384126e-05, + "loss": 1.1072, + "step": 5553 + }, + { + "epoch": 0.13, + "learning_rate": 1.9442813837284416e-05, + "loss": 1.2416, + "step": 5554 + }, + { + "epoch": 0.13, + "learning_rate": 1.9442557443897382e-05, + "loss": 0.8832, + "step": 5555 + }, + { + "epoch": 0.13, + "learning_rate": 1.9442300993224568e-05, + "loss": 1.1304, + "step": 5556 + }, + { + "epoch": 0.13, + "learning_rate": 1.944204448526754e-05, + "loss": 1.2119, + "step": 5557 + }, + { + "epoch": 0.13, + "learning_rate": 1.9441787920027843e-05, + "loss": 1.1738, + "step": 5558 + }, + { + "epoch": 0.13, + "learning_rate": 1.944153129750704e-05, + "loss": 1.3102, + "step": 5559 + }, + { + "epoch": 0.13, + "learning_rate": 1.944127461770669e-05, + "loss": 1.3584, + "step": 5560 + }, + { + "epoch": 0.13, + "learning_rate": 1.9441017880628346e-05, + "loss": 1.3933, + "step": 5561 + }, + { + "epoch": 0.13, + "learning_rate": 1.9440761086273564e-05, + "loss": 1.2391, + "step": 5562 + }, + { + "epoch": 0.13, + "learning_rate": 1.944050423464391e-05, + "loss": 1.1717, + "step": 5563 + }, + { + "epoch": 0.13, + "learning_rate": 1.944024732574094e-05, + "loss": 1.1344, + "step": 5564 + }, + { + "epoch": 0.13, + "learning_rate": 1.9439990359566204e-05, + "loss": 1.088, + "step": 5565 + }, + { + "epoch": 0.13, + "learning_rate": 1.9439733336121267e-05, + "loss": 1.1801, + "step": 5566 + }, + { + "epoch": 0.13, + "learning_rate": 1.943947625540769e-05, + "loss": 1.0401, + "step": 5567 + }, + { + "epoch": 0.13, + "learning_rate": 1.9439219117427034e-05, + "loss": 1.2973, + "step": 5568 + }, + { + "epoch": 0.13, + "learning_rate": 1.943896192218085e-05, + "loss": 1.1071, + "step": 5569 + }, + { + "epoch": 0.13, + "learning_rate": 1.943870466967071e-05, + "loss": 1.2356, + "step": 5570 + }, + { + "epoch": 0.13, + "learning_rate": 1.9438447359898162e-05, + "loss": 1.0238, + "step": 5571 + }, + { + "epoch": 0.13, + "learning_rate": 1.943818999286478e-05, + "loss": 1.0032, + "step": 5572 + }, + { + "epoch": 0.13, + "learning_rate": 1.9437932568572116e-05, + "loss": 1.1281, + "step": 5573 + }, + { + "epoch": 0.13, + "learning_rate": 1.9437675087021736e-05, + "loss": 1.2847, + "step": 5574 + }, + { + "epoch": 0.13, + "learning_rate": 1.94374175482152e-05, + "loss": 1.0929, + "step": 5575 + }, + { + "epoch": 0.13, + "learning_rate": 1.9437159952154078e-05, + "loss": 1.1558, + "step": 5576 + }, + { + "epoch": 0.13, + "learning_rate": 1.9436902298839918e-05, + "loss": 1.3056, + "step": 5577 + }, + { + "epoch": 0.13, + "learning_rate": 1.9436644588274295e-05, + "loss": 1.0641, + "step": 5578 + }, + { + "epoch": 0.13, + "learning_rate": 1.9436386820458767e-05, + "loss": 1.1632, + "step": 5579 + }, + { + "epoch": 0.13, + "learning_rate": 1.9436128995394903e-05, + "loss": 1.1774, + "step": 5580 + }, + { + "epoch": 0.13, + "learning_rate": 1.943587111308426e-05, + "loss": 1.1265, + "step": 5581 + }, + { + "epoch": 0.13, + "learning_rate": 1.943561317352841e-05, + "loss": 0.9852, + "step": 5582 + }, + { + "epoch": 0.13, + "learning_rate": 1.9435355176728915e-05, + "loss": 1.1366, + "step": 5583 + }, + { + "epoch": 0.13, + "learning_rate": 1.9435097122687337e-05, + "loss": 1.1511, + "step": 5584 + }, + { + "epoch": 0.13, + "learning_rate": 1.9434839011405243e-05, + "loss": 1.2873, + "step": 5585 + }, + { + "epoch": 0.13, + "learning_rate": 1.94345808428842e-05, + "loss": 1.1611, + "step": 5586 + }, + { + "epoch": 0.13, + "learning_rate": 1.9434322617125777e-05, + "loss": 0.9807, + "step": 5587 + }, + { + "epoch": 0.13, + "learning_rate": 1.943406433413154e-05, + "loss": 1.1546, + "step": 5588 + }, + { + "epoch": 0.13, + "learning_rate": 1.9433805993903047e-05, + "loss": 1.4255, + "step": 5589 + }, + { + "epoch": 0.13, + "learning_rate": 1.9433547596441877e-05, + "loss": 1.0684, + "step": 5590 + }, + { + "epoch": 0.13, + "learning_rate": 1.9433289141749587e-05, + "loss": 1.0512, + "step": 5591 + }, + { + "epoch": 0.13, + "learning_rate": 1.9433030629827757e-05, + "loss": 1.1093, + "step": 5592 + }, + { + "epoch": 0.13, + "learning_rate": 1.9432772060677945e-05, + "loss": 1.1507, + "step": 5593 + }, + { + "epoch": 0.13, + "learning_rate": 1.9432513434301727e-05, + "loss": 1.0732, + "step": 5594 + }, + { + "epoch": 0.13, + "learning_rate": 1.9432254750700664e-05, + "loss": 1.2671, + "step": 5595 + }, + { + "epoch": 0.13, + "learning_rate": 1.943199600987633e-05, + "loss": 1.2503, + "step": 5596 + }, + { + "epoch": 0.13, + "learning_rate": 1.94317372118303e-05, + "loss": 1.3391, + "step": 5597 + }, + { + "epoch": 0.13, + "learning_rate": 1.943147835656414e-05, + "loss": 0.9568, + "step": 5598 + }, + { + "epoch": 0.13, + "learning_rate": 1.943121944407941e-05, + "loss": 1.1515, + "step": 5599 + }, + { + "epoch": 0.13, + "learning_rate": 1.9430960474377697e-05, + "loss": 1.1876, + "step": 5600 + }, + { + "epoch": 0.13, + "learning_rate": 1.9430701447460563e-05, + "loss": 1.1569, + "step": 5601 + }, + { + "epoch": 0.13, + "learning_rate": 1.9430442363329583e-05, + "loss": 1.3592, + "step": 5602 + }, + { + "epoch": 0.13, + "learning_rate": 1.9430183221986325e-05, + "loss": 1.0848, + "step": 5603 + }, + { + "epoch": 0.13, + "learning_rate": 1.9429924023432364e-05, + "loss": 1.0082, + "step": 5604 + }, + { + "epoch": 0.13, + "learning_rate": 1.9429664767669273e-05, + "loss": 0.95, + "step": 5605 + }, + { + "epoch": 0.13, + "learning_rate": 1.9429405454698624e-05, + "loss": 1.1464, + "step": 5606 + }, + { + "epoch": 0.13, + "learning_rate": 1.942914608452199e-05, + "loss": 1.1358, + "step": 5607 + }, + { + "epoch": 0.13, + "learning_rate": 1.9428886657140945e-05, + "loss": 0.9817, + "step": 5608 + }, + { + "epoch": 0.13, + "learning_rate": 1.942862717255706e-05, + "loss": 1.3817, + "step": 5609 + }, + { + "epoch": 0.13, + "learning_rate": 1.9428367630771915e-05, + "loss": 1.1604, + "step": 5610 + }, + { + "epoch": 0.13, + "learning_rate": 1.942810803178708e-05, + "loss": 1.019, + "step": 5611 + }, + { + "epoch": 0.13, + "learning_rate": 1.942784837560413e-05, + "loss": 1.1118, + "step": 5612 + }, + { + "epoch": 0.13, + "learning_rate": 1.942758866222464e-05, + "loss": 1.1354, + "step": 5613 + }, + { + "epoch": 0.14, + "learning_rate": 1.9427328891650194e-05, + "loss": 1.1548, + "step": 5614 + }, + { + "epoch": 0.14, + "learning_rate": 1.9427069063882355e-05, + "loss": 1.3057, + "step": 5615 + }, + { + "epoch": 0.14, + "learning_rate": 1.9426809178922706e-05, + "loss": 1.1377, + "step": 5616 + }, + { + "epoch": 0.14, + "learning_rate": 1.9426549236772826e-05, + "loss": 0.988, + "step": 5617 + }, + { + "epoch": 0.14, + "learning_rate": 1.9426289237434286e-05, + "loss": 1.2621, + "step": 5618 + }, + { + "epoch": 0.14, + "learning_rate": 1.9426029180908666e-05, + "loss": 1.3405, + "step": 5619 + }, + { + "epoch": 0.14, + "learning_rate": 1.9425769067197548e-05, + "loss": 1.0374, + "step": 5620 + }, + { + "epoch": 0.14, + "learning_rate": 1.9425508896302503e-05, + "loss": 0.8492, + "step": 5621 + }, + { + "epoch": 0.14, + "learning_rate": 1.942524866822511e-05, + "loss": 1.4294, + "step": 5622 + }, + { + "epoch": 0.14, + "learning_rate": 1.9424988382966955e-05, + "loss": 1.1936, + "step": 5623 + }, + { + "epoch": 0.14, + "learning_rate": 1.9424728040529612e-05, + "loss": 1.2615, + "step": 5624 + }, + { + "epoch": 0.14, + "learning_rate": 1.9424467640914655e-05, + "loss": 1.3634, + "step": 5625 + }, + { + "epoch": 0.14, + "learning_rate": 1.9424207184123677e-05, + "loss": 1.0091, + "step": 5626 + }, + { + "epoch": 0.14, + "learning_rate": 1.942394667015825e-05, + "loss": 1.3312, + "step": 5627 + }, + { + "epoch": 0.14, + "learning_rate": 1.942368609901995e-05, + "loss": 1.1473, + "step": 5628 + }, + { + "epoch": 0.14, + "learning_rate": 1.9423425470710367e-05, + "loss": 1.3645, + "step": 5629 + }, + { + "epoch": 0.14, + "learning_rate": 1.9423164785231078e-05, + "loss": 1.3409, + "step": 5630 + }, + { + "epoch": 0.14, + "learning_rate": 1.9422904042583664e-05, + "loss": 0.9827, + "step": 5631 + }, + { + "epoch": 0.14, + "learning_rate": 1.942264324276971e-05, + "loss": 1.0949, + "step": 5632 + }, + { + "epoch": 0.14, + "learning_rate": 1.942238238579079e-05, + "loss": 1.4197, + "step": 5633 + }, + { + "epoch": 0.14, + "learning_rate": 1.94221214716485e-05, + "loss": 1.0813, + "step": 5634 + }, + { + "epoch": 0.14, + "learning_rate": 1.9421860500344407e-05, + "loss": 1.1498, + "step": 5635 + }, + { + "epoch": 0.14, + "learning_rate": 1.9421599471880108e-05, + "loss": 1.4192, + "step": 5636 + }, + { + "epoch": 0.14, + "learning_rate": 1.942133838625718e-05, + "loss": 1.3142, + "step": 5637 + }, + { + "epoch": 0.14, + "learning_rate": 1.9421077243477208e-05, + "loss": 1.2685, + "step": 5638 + }, + { + "epoch": 0.14, + "learning_rate": 1.9420816043541775e-05, + "loss": 1.0727, + "step": 5639 + }, + { + "epoch": 0.14, + "learning_rate": 1.942055478645247e-05, + "loss": 1.1972, + "step": 5640 + }, + { + "epoch": 0.14, + "learning_rate": 1.9420293472210868e-05, + "loss": 1.1417, + "step": 5641 + }, + { + "epoch": 0.14, + "learning_rate": 1.942003210081857e-05, + "loss": 1.1735, + "step": 5642 + }, + { + "epoch": 0.14, + "learning_rate": 1.9419770672277146e-05, + "loss": 1.1139, + "step": 5643 + }, + { + "epoch": 0.14, + "learning_rate": 1.9419509186588196e-05, + "loss": 1.3354, + "step": 5644 + }, + { + "epoch": 0.14, + "learning_rate": 1.9419247643753294e-05, + "loss": 1.2294, + "step": 5645 + }, + { + "epoch": 0.14, + "learning_rate": 1.9418986043774036e-05, + "loss": 0.8933, + "step": 5646 + }, + { + "epoch": 0.14, + "learning_rate": 1.9418724386652e-05, + "loss": 1.2269, + "step": 5647 + }, + { + "epoch": 0.14, + "learning_rate": 1.9418462672388784e-05, + "loss": 1.1243, + "step": 5648 + }, + { + "epoch": 0.14, + "learning_rate": 1.9418200900985968e-05, + "loss": 1.1089, + "step": 5649 + }, + { + "epoch": 0.14, + "learning_rate": 1.941793907244514e-05, + "loss": 1.3455, + "step": 5650 + }, + { + "epoch": 0.14, + "learning_rate": 1.9417677186767894e-05, + "loss": 1.185, + "step": 5651 + }, + { + "epoch": 0.14, + "learning_rate": 1.941741524395581e-05, + "loss": 1.0942, + "step": 5652 + }, + { + "epoch": 0.14, + "learning_rate": 1.9417153244010493e-05, + "loss": 1.0127, + "step": 5653 + }, + { + "epoch": 0.14, + "learning_rate": 1.9416891186933516e-05, + "loss": 1.2604, + "step": 5654 + }, + { + "epoch": 0.14, + "learning_rate": 1.9416629072726475e-05, + "loss": 1.1568, + "step": 5655 + }, + { + "epoch": 0.14, + "learning_rate": 1.941636690139096e-05, + "loss": 1.3687, + "step": 5656 + }, + { + "epoch": 0.14, + "learning_rate": 1.9416104672928564e-05, + "loss": 1.3385, + "step": 5657 + }, + { + "epoch": 0.14, + "learning_rate": 1.9415842387340875e-05, + "loss": 1.2436, + "step": 5658 + }, + { + "epoch": 0.14, + "learning_rate": 1.9415580044629485e-05, + "loss": 1.0034, + "step": 5659 + }, + { + "epoch": 0.14, + "learning_rate": 1.9415317644795984e-05, + "loss": 1.2572, + "step": 5660 + }, + { + "epoch": 0.14, + "learning_rate": 1.941505518784197e-05, + "loss": 1.009, + "step": 5661 + }, + { + "epoch": 0.14, + "learning_rate": 1.9414792673769028e-05, + "loss": 1.0186, + "step": 5662 + }, + { + "epoch": 0.14, + "learning_rate": 1.9414530102578753e-05, + "loss": 1.1103, + "step": 5663 + }, + { + "epoch": 0.14, + "learning_rate": 1.9414267474272735e-05, + "loss": 1.1575, + "step": 5664 + }, + { + "epoch": 0.14, + "learning_rate": 1.9414004788852577e-05, + "loss": 1.1695, + "step": 5665 + }, + { + "epoch": 0.14, + "learning_rate": 1.9413742046319863e-05, + "loss": 1.2179, + "step": 5666 + }, + { + "epoch": 0.14, + "learning_rate": 1.941347924667619e-05, + "loss": 1.2394, + "step": 5667 + }, + { + "epoch": 0.14, + "learning_rate": 1.941321638992315e-05, + "loss": 1.3788, + "step": 5668 + }, + { + "epoch": 0.14, + "learning_rate": 1.9412953476062346e-05, + "loss": 1.0219, + "step": 5669 + }, + { + "epoch": 0.14, + "learning_rate": 1.9412690505095363e-05, + "loss": 1.0913, + "step": 5670 + }, + { + "epoch": 0.14, + "learning_rate": 1.9412427477023804e-05, + "loss": 1.4096, + "step": 5671 + }, + { + "epoch": 0.14, + "learning_rate": 1.941216439184926e-05, + "loss": 0.984, + "step": 5672 + }, + { + "epoch": 0.14, + "learning_rate": 1.9411901249573327e-05, + "loss": 1.3071, + "step": 5673 + }, + { + "epoch": 0.14, + "learning_rate": 1.9411638050197605e-05, + "loss": 1.3708, + "step": 5674 + }, + { + "epoch": 0.14, + "learning_rate": 1.9411374793723686e-05, + "loss": 1.0818, + "step": 5675 + }, + { + "epoch": 0.14, + "learning_rate": 1.9411111480153174e-05, + "loss": 1.3779, + "step": 5676 + }, + { + "epoch": 0.14, + "learning_rate": 1.941084810948766e-05, + "loss": 1.24, + "step": 5677 + }, + { + "epoch": 0.14, + "learning_rate": 1.9410584681728745e-05, + "loss": 1.2655, + "step": 5678 + }, + { + "epoch": 0.14, + "learning_rate": 1.9410321196878023e-05, + "loss": 1.3137, + "step": 5679 + }, + { + "epoch": 0.14, + "learning_rate": 1.94100576549371e-05, + "loss": 1.4518, + "step": 5680 + }, + { + "epoch": 0.14, + "learning_rate": 1.9409794055907568e-05, + "loss": 1.01, + "step": 5681 + }, + { + "epoch": 0.14, + "learning_rate": 1.9409530399791026e-05, + "loss": 1.2742, + "step": 5682 + }, + { + "epoch": 0.14, + "learning_rate": 1.9409266686589082e-05, + "loss": 1.1909, + "step": 5683 + }, + { + "epoch": 0.14, + "learning_rate": 1.940900291630333e-05, + "loss": 1.3117, + "step": 5684 + }, + { + "epoch": 0.14, + "learning_rate": 1.940873908893537e-05, + "loss": 1.2803, + "step": 5685 + }, + { + "epoch": 0.14, + "learning_rate": 1.94084752044868e-05, + "loss": 1.0642, + "step": 5686 + }, + { + "epoch": 0.14, + "learning_rate": 1.9408211262959225e-05, + "loss": 1.0633, + "step": 5687 + }, + { + "epoch": 0.14, + "learning_rate": 1.9407947264354242e-05, + "loss": 1.0678, + "step": 5688 + }, + { + "epoch": 0.14, + "learning_rate": 1.9407683208673462e-05, + "loss": 1.3679, + "step": 5689 + }, + { + "epoch": 0.14, + "learning_rate": 1.9407419095918477e-05, + "loss": 0.8689, + "step": 5690 + }, + { + "epoch": 0.14, + "learning_rate": 1.9407154926090896e-05, + "loss": 1.2676, + "step": 5691 + }, + { + "epoch": 0.14, + "learning_rate": 1.9406890699192316e-05, + "loss": 1.2171, + "step": 5692 + }, + { + "epoch": 0.14, + "learning_rate": 1.9406626415224346e-05, + "loss": 1.4104, + "step": 5693 + }, + { + "epoch": 0.14, + "learning_rate": 1.9406362074188584e-05, + "loss": 1.3234, + "step": 5694 + }, + { + "epoch": 0.14, + "learning_rate": 1.9406097676086634e-05, + "loss": 1.2102, + "step": 5695 + }, + { + "epoch": 0.14, + "learning_rate": 1.9405833220920104e-05, + "loss": 1.2316, + "step": 5696 + }, + { + "epoch": 0.14, + "learning_rate": 1.9405568708690595e-05, + "loss": 1.1726, + "step": 5697 + }, + { + "epoch": 0.14, + "learning_rate": 1.9405304139399715e-05, + "loss": 1.0073, + "step": 5698 + }, + { + "epoch": 0.14, + "learning_rate": 1.9405039513049067e-05, + "loss": 1.0108, + "step": 5699 + }, + { + "epoch": 0.14, + "learning_rate": 1.9404774829640254e-05, + "loss": 0.982, + "step": 5700 + }, + { + "epoch": 0.14, + "learning_rate": 1.9404510089174888e-05, + "loss": 1.1925, + "step": 5701 + }, + { + "epoch": 0.14, + "learning_rate": 1.9404245291654568e-05, + "loss": 1.2659, + "step": 5702 + }, + { + "epoch": 0.14, + "learning_rate": 1.94039804370809e-05, + "loss": 1.1612, + "step": 5703 + }, + { + "epoch": 0.14, + "learning_rate": 1.9403715525455503e-05, + "loss": 1.2525, + "step": 5704 + }, + { + "epoch": 0.14, + "learning_rate": 1.9403450556779972e-05, + "loss": 1.0984, + "step": 5705 + }, + { + "epoch": 0.14, + "learning_rate": 1.9403185531055915e-05, + "loss": 1.3052, + "step": 5706 + }, + { + "epoch": 0.14, + "learning_rate": 1.940292044828495e-05, + "loss": 1.4478, + "step": 5707 + }, + { + "epoch": 0.14, + "learning_rate": 1.9402655308468678e-05, + "loss": 0.9061, + "step": 5708 + }, + { + "epoch": 0.14, + "learning_rate": 1.9402390111608702e-05, + "loss": 1.3391, + "step": 5709 + }, + { + "epoch": 0.14, + "learning_rate": 1.940212485770664e-05, + "loss": 0.9128, + "step": 5710 + }, + { + "epoch": 0.14, + "learning_rate": 1.9401859546764097e-05, + "loss": 1.0675, + "step": 5711 + }, + { + "epoch": 0.14, + "learning_rate": 1.9401594178782686e-05, + "loss": 1.2353, + "step": 5712 + }, + { + "epoch": 0.14, + "learning_rate": 1.940132875376401e-05, + "loss": 1.0575, + "step": 5713 + }, + { + "epoch": 0.14, + "learning_rate": 1.940106327170969e-05, + "loss": 1.2067, + "step": 5714 + }, + { + "epoch": 0.14, + "learning_rate": 1.9400797732621325e-05, + "loss": 1.0534, + "step": 5715 + }, + { + "epoch": 0.14, + "learning_rate": 1.940053213650053e-05, + "loss": 1.183, + "step": 5716 + }, + { + "epoch": 0.14, + "learning_rate": 1.9400266483348924e-05, + "loss": 1.0075, + "step": 5717 + }, + { + "epoch": 0.14, + "learning_rate": 1.9400000773168107e-05, + "loss": 0.8111, + "step": 5718 + }, + { + "epoch": 0.14, + "learning_rate": 1.9399735005959702e-05, + "loss": 1.3712, + "step": 5719 + }, + { + "epoch": 0.14, + "learning_rate": 1.939946918172531e-05, + "loss": 1.2239, + "step": 5720 + }, + { + "epoch": 0.14, + "learning_rate": 1.9399203300466554e-05, + "loss": 1.1433, + "step": 5721 + }, + { + "epoch": 0.14, + "learning_rate": 1.9398937362185037e-05, + "loss": 1.2214, + "step": 5722 + }, + { + "epoch": 0.14, + "learning_rate": 1.9398671366882382e-05, + "loss": 0.9237, + "step": 5723 + }, + { + "epoch": 0.14, + "learning_rate": 1.9398405314560197e-05, + "loss": 1.2332, + "step": 5724 + }, + { + "epoch": 0.14, + "learning_rate": 1.9398139205220098e-05, + "loss": 1.0817, + "step": 5725 + }, + { + "epoch": 0.14, + "learning_rate": 1.93978730388637e-05, + "loss": 1.0901, + "step": 5726 + }, + { + "epoch": 0.14, + "learning_rate": 1.9397606815492617e-05, + "loss": 1.4387, + "step": 5727 + }, + { + "epoch": 0.14, + "learning_rate": 1.939734053510846e-05, + "loss": 1.2328, + "step": 5728 + }, + { + "epoch": 0.14, + "learning_rate": 1.939707419771285e-05, + "loss": 1.2794, + "step": 5729 + }, + { + "epoch": 0.14, + "learning_rate": 1.9396807803307405e-05, + "loss": 1.1713, + "step": 5730 + }, + { + "epoch": 0.14, + "learning_rate": 1.9396541351893734e-05, + "loss": 1.1861, + "step": 5731 + }, + { + "epoch": 0.14, + "learning_rate": 1.9396274843473455e-05, + "loss": 1.1007, + "step": 5732 + }, + { + "epoch": 0.14, + "learning_rate": 1.939600827804819e-05, + "loss": 1.243, + "step": 5733 + }, + { + "epoch": 0.14, + "learning_rate": 1.9395741655619554e-05, + "loss": 1.241, + "step": 5734 + }, + { + "epoch": 0.14, + "learning_rate": 1.939547497618916e-05, + "loss": 1.2553, + "step": 5735 + }, + { + "epoch": 0.14, + "learning_rate": 1.939520823975863e-05, + "loss": 0.9555, + "step": 5736 + }, + { + "epoch": 0.14, + "learning_rate": 1.9394941446329583e-05, + "loss": 1.0379, + "step": 5737 + }, + { + "epoch": 0.14, + "learning_rate": 1.9394674595903635e-05, + "loss": 1.1767, + "step": 5738 + }, + { + "epoch": 0.14, + "learning_rate": 1.9394407688482407e-05, + "loss": 1.1698, + "step": 5739 + }, + { + "epoch": 0.14, + "learning_rate": 1.9394140724067515e-05, + "loss": 1.4374, + "step": 5740 + }, + { + "epoch": 0.14, + "learning_rate": 1.9393873702660583e-05, + "loss": 1.2317, + "step": 5741 + }, + { + "epoch": 0.14, + "learning_rate": 1.9393606624263228e-05, + "loss": 0.947, + "step": 5742 + }, + { + "epoch": 0.14, + "learning_rate": 1.9393339488877073e-05, + "loss": 1.1237, + "step": 5743 + }, + { + "epoch": 0.14, + "learning_rate": 1.9393072296503733e-05, + "loss": 1.4612, + "step": 5744 + }, + { + "epoch": 0.14, + "learning_rate": 1.9392805047144833e-05, + "loss": 1.0513, + "step": 5745 + }, + { + "epoch": 0.14, + "learning_rate": 1.9392537740801997e-05, + "loss": 0.9053, + "step": 5746 + }, + { + "epoch": 0.14, + "learning_rate": 1.9392270377476843e-05, + "loss": 1.0323, + "step": 5747 + }, + { + "epoch": 0.14, + "learning_rate": 1.9392002957170994e-05, + "loss": 1.1564, + "step": 5748 + }, + { + "epoch": 0.14, + "learning_rate": 1.939173547988607e-05, + "loss": 1.1583, + "step": 5749 + }, + { + "epoch": 0.14, + "learning_rate": 1.9391467945623698e-05, + "loss": 1.4773, + "step": 5750 + }, + { + "epoch": 0.14, + "learning_rate": 1.93912003543855e-05, + "loss": 1.1811, + "step": 5751 + }, + { + "epoch": 0.14, + "learning_rate": 1.9390932706173095e-05, + "loss": 1.2284, + "step": 5752 + }, + { + "epoch": 0.14, + "learning_rate": 1.9390665000988113e-05, + "loss": 0.988, + "step": 5753 + }, + { + "epoch": 0.14, + "learning_rate": 1.9390397238832173e-05, + "loss": 1.3102, + "step": 5754 + }, + { + "epoch": 0.14, + "learning_rate": 1.9390129419706904e-05, + "loss": 1.1584, + "step": 5755 + }, + { + "epoch": 0.14, + "learning_rate": 1.9389861543613927e-05, + "loss": 0.9277, + "step": 5756 + }, + { + "epoch": 0.14, + "learning_rate": 1.938959361055487e-05, + "loss": 1.3555, + "step": 5757 + }, + { + "epoch": 0.14, + "learning_rate": 1.9389325620531356e-05, + "loss": 0.9279, + "step": 5758 + }, + { + "epoch": 0.14, + "learning_rate": 1.9389057573545012e-05, + "loss": 1.2585, + "step": 5759 + }, + { + "epoch": 0.14, + "learning_rate": 1.9388789469597464e-05, + "loss": 0.9508, + "step": 5760 + }, + { + "epoch": 0.14, + "learning_rate": 1.938852130869034e-05, + "loss": 1.3831, + "step": 5761 + }, + { + "epoch": 0.14, + "learning_rate": 1.9388253090825263e-05, + "loss": 1.2381, + "step": 5762 + }, + { + "epoch": 0.14, + "learning_rate": 1.9387984816003868e-05, + "loss": 1.258, + "step": 5763 + }, + { + "epoch": 0.14, + "learning_rate": 1.938771648422777e-05, + "loss": 1.4026, + "step": 5764 + }, + { + "epoch": 0.14, + "learning_rate": 1.938744809549861e-05, + "loss": 1.0846, + "step": 5765 + }, + { + "epoch": 0.14, + "learning_rate": 1.9387179649818007e-05, + "loss": 1.0542, + "step": 5766 + }, + { + "epoch": 0.14, + "learning_rate": 1.9386911147187596e-05, + "loss": 0.8847, + "step": 5767 + }, + { + "epoch": 0.14, + "learning_rate": 1.9386642587609002e-05, + "loss": 1.3717, + "step": 5768 + }, + { + "epoch": 0.14, + "learning_rate": 1.9386373971083854e-05, + "loss": 1.2658, + "step": 5769 + }, + { + "epoch": 0.14, + "learning_rate": 1.9386105297613782e-05, + "loss": 1.0157, + "step": 5770 + }, + { + "epoch": 0.14, + "learning_rate": 1.938583656720042e-05, + "loss": 1.1367, + "step": 5771 + }, + { + "epoch": 0.14, + "learning_rate": 1.9385567779845392e-05, + "loss": 0.9597, + "step": 5772 + }, + { + "epoch": 0.14, + "learning_rate": 1.9385298935550333e-05, + "loss": 1.5079, + "step": 5773 + }, + { + "epoch": 0.14, + "learning_rate": 1.9385030034316873e-05, + "loss": 0.9811, + "step": 5774 + }, + { + "epoch": 0.14, + "learning_rate": 1.9384761076146647e-05, + "loss": 0.9931, + "step": 5775 + }, + { + "epoch": 0.14, + "learning_rate": 1.938449206104128e-05, + "loss": 1.1504, + "step": 5776 + }, + { + "epoch": 0.14, + "learning_rate": 1.9384222989002405e-05, + "loss": 1.1326, + "step": 5777 + }, + { + "epoch": 0.14, + "learning_rate": 1.9383953860031658e-05, + "loss": 1.1987, + "step": 5778 + }, + { + "epoch": 0.14, + "learning_rate": 1.938368467413067e-05, + "loss": 1.1858, + "step": 5779 + }, + { + "epoch": 0.14, + "learning_rate": 1.9383415431301075e-05, + "loss": 1.1069, + "step": 5780 + }, + { + "epoch": 0.14, + "learning_rate": 1.9383146131544505e-05, + "loss": 1.2445, + "step": 5781 + }, + { + "epoch": 0.14, + "learning_rate": 1.9382876774862594e-05, + "loss": 0.9736, + "step": 5782 + }, + { + "epoch": 0.14, + "learning_rate": 1.938260736125698e-05, + "loss": 1.0604, + "step": 5783 + }, + { + "epoch": 0.14, + "learning_rate": 1.938233789072929e-05, + "loss": 1.1068, + "step": 5784 + }, + { + "epoch": 0.14, + "learning_rate": 1.9382068363281167e-05, + "loss": 1.2364, + "step": 5785 + }, + { + "epoch": 0.14, + "learning_rate": 1.938179877891424e-05, + "loss": 1.0661, + "step": 5786 + }, + { + "epoch": 0.14, + "learning_rate": 1.9381529137630145e-05, + "loss": 0.9964, + "step": 5787 + }, + { + "epoch": 0.14, + "learning_rate": 1.9381259439430517e-05, + "loss": 1.0719, + "step": 5788 + }, + { + "epoch": 0.14, + "learning_rate": 1.9380989684317e-05, + "loss": 1.2271, + "step": 5789 + }, + { + "epoch": 0.14, + "learning_rate": 1.9380719872291223e-05, + "loss": 1.019, + "step": 5790 + }, + { + "epoch": 0.14, + "learning_rate": 1.9380450003354825e-05, + "loss": 1.022, + "step": 5791 + }, + { + "epoch": 0.14, + "learning_rate": 1.9380180077509444e-05, + "loss": 0.9003, + "step": 5792 + }, + { + "epoch": 0.14, + "learning_rate": 1.9379910094756717e-05, + "loss": 1.1824, + "step": 5793 + }, + { + "epoch": 0.14, + "learning_rate": 1.937964005509828e-05, + "loss": 0.7782, + "step": 5794 + }, + { + "epoch": 0.14, + "learning_rate": 1.9379369958535773e-05, + "loss": 1.0045, + "step": 5795 + }, + { + "epoch": 0.14, + "learning_rate": 1.9379099805070836e-05, + "loss": 1.0595, + "step": 5796 + }, + { + "epoch": 0.14, + "learning_rate": 1.9378829594705106e-05, + "loss": 1.3355, + "step": 5797 + }, + { + "epoch": 0.14, + "learning_rate": 1.9378559327440222e-05, + "loss": 1.258, + "step": 5798 + }, + { + "epoch": 0.14, + "learning_rate": 1.9378289003277828e-05, + "loss": 1.1598, + "step": 5799 + }, + { + "epoch": 0.14, + "learning_rate": 1.9378018622219557e-05, + "loss": 1.387, + "step": 5800 + }, + { + "epoch": 0.14, + "learning_rate": 1.9377748184267055e-05, + "loss": 1.241, + "step": 5801 + }, + { + "epoch": 0.14, + "learning_rate": 1.9377477689421958e-05, + "loss": 1.3884, + "step": 5802 + }, + { + "epoch": 0.14, + "learning_rate": 1.937720713768591e-05, + "loss": 1.0918, + "step": 5803 + }, + { + "epoch": 0.14, + "learning_rate": 1.9376936529060556e-05, + "loss": 1.0069, + "step": 5804 + }, + { + "epoch": 0.14, + "learning_rate": 1.937666586354753e-05, + "loss": 0.8857, + "step": 5805 + }, + { + "epoch": 0.14, + "learning_rate": 1.9376395141148475e-05, + "loss": 0.8812, + "step": 5806 + }, + { + "epoch": 0.14, + "learning_rate": 1.937612436186504e-05, + "loss": 1.2061, + "step": 5807 + }, + { + "epoch": 0.14, + "learning_rate": 1.9375853525698866e-05, + "loss": 1.1024, + "step": 5808 + }, + { + "epoch": 0.14, + "learning_rate": 1.937558263265159e-05, + "loss": 1.1868, + "step": 5809 + }, + { + "epoch": 0.14, + "learning_rate": 1.9375311682724863e-05, + "loss": 1.1566, + "step": 5810 + }, + { + "epoch": 0.14, + "learning_rate": 1.937504067592032e-05, + "loss": 1.0299, + "step": 5811 + }, + { + "epoch": 0.14, + "learning_rate": 1.9374769612239617e-05, + "loss": 1.0025, + "step": 5812 + }, + { + "epoch": 0.14, + "learning_rate": 1.9374498491684388e-05, + "loss": 1.2365, + "step": 5813 + }, + { + "epoch": 0.14, + "learning_rate": 1.937422731425628e-05, + "loss": 1.0775, + "step": 5814 + }, + { + "epoch": 0.14, + "learning_rate": 1.9373956079956943e-05, + "loss": 1.157, + "step": 5815 + }, + { + "epoch": 0.14, + "learning_rate": 1.9373684788788018e-05, + "loss": 1.3265, + "step": 5816 + }, + { + "epoch": 0.14, + "learning_rate": 1.9373413440751153e-05, + "loss": 1.1755, + "step": 5817 + }, + { + "epoch": 0.14, + "learning_rate": 1.937314203584799e-05, + "loss": 1.1404, + "step": 5818 + }, + { + "epoch": 0.14, + "learning_rate": 1.9372870574080185e-05, + "loss": 1.3101, + "step": 5819 + }, + { + "epoch": 0.14, + "learning_rate": 1.9372599055449374e-05, + "loss": 0.9056, + "step": 5820 + }, + { + "epoch": 0.14, + "learning_rate": 1.9372327479957208e-05, + "loss": 1.2436, + "step": 5821 + }, + { + "epoch": 0.14, + "learning_rate": 1.937205584760534e-05, + "loss": 1.3438, + "step": 5822 + }, + { + "epoch": 0.14, + "learning_rate": 1.9371784158395415e-05, + "loss": 1.2212, + "step": 5823 + }, + { + "epoch": 0.14, + "learning_rate": 1.9371512412329078e-05, + "loss": 1.143, + "step": 5824 + }, + { + "epoch": 0.14, + "learning_rate": 1.9371240609407975e-05, + "loss": 1.1252, + "step": 5825 + }, + { + "epoch": 0.14, + "learning_rate": 1.9370968749633764e-05, + "loss": 1.244, + "step": 5826 + }, + { + "epoch": 0.14, + "learning_rate": 1.9370696833008087e-05, + "loss": 1.206, + "step": 5827 + }, + { + "epoch": 0.14, + "learning_rate": 1.9370424859532598e-05, + "loss": 1.2543, + "step": 5828 + }, + { + "epoch": 0.14, + "learning_rate": 1.9370152829208946e-05, + "loss": 1.1965, + "step": 5829 + }, + { + "epoch": 0.14, + "learning_rate": 1.9369880742038783e-05, + "loss": 1.1301, + "step": 5830 + }, + { + "epoch": 0.14, + "learning_rate": 1.9369608598023753e-05, + "loss": 1.0515, + "step": 5831 + }, + { + "epoch": 0.14, + "learning_rate": 1.9369336397165512e-05, + "loss": 1.1111, + "step": 5832 + }, + { + "epoch": 0.14, + "learning_rate": 1.9369064139465715e-05, + "loss": 1.1133, + "step": 5833 + }, + { + "epoch": 0.14, + "learning_rate": 1.9368791824926006e-05, + "loss": 0.9897, + "step": 5834 + }, + { + "epoch": 0.14, + "learning_rate": 1.936851945354804e-05, + "loss": 0.8985, + "step": 5835 + }, + { + "epoch": 0.14, + "learning_rate": 1.9368247025333472e-05, + "loss": 0.9349, + "step": 5836 + }, + { + "epoch": 0.14, + "learning_rate": 1.9367974540283954e-05, + "loss": 1.0602, + "step": 5837 + }, + { + "epoch": 0.14, + "learning_rate": 1.9367701998401137e-05, + "loss": 1.1977, + "step": 5838 + }, + { + "epoch": 0.14, + "learning_rate": 1.9367429399686672e-05, + "loss": 1.0115, + "step": 5839 + }, + { + "epoch": 0.14, + "learning_rate": 1.9367156744142218e-05, + "loss": 1.0623, + "step": 5840 + }, + { + "epoch": 0.14, + "learning_rate": 1.936688403176943e-05, + "loss": 0.9516, + "step": 5841 + }, + { + "epoch": 0.14, + "learning_rate": 1.936661126256996e-05, + "loss": 1.1613, + "step": 5842 + }, + { + "epoch": 0.14, + "learning_rate": 1.936633843654546e-05, + "loss": 1.1271, + "step": 5843 + }, + { + "epoch": 0.14, + "learning_rate": 1.9366065553697586e-05, + "loss": 1.148, + "step": 5844 + }, + { + "epoch": 0.14, + "learning_rate": 1.9365792614028002e-05, + "loss": 1.1733, + "step": 5845 + }, + { + "epoch": 0.14, + "learning_rate": 1.936551961753835e-05, + "loss": 1.2644, + "step": 5846 + }, + { + "epoch": 0.14, + "learning_rate": 1.9365246564230298e-05, + "loss": 0.9288, + "step": 5847 + }, + { + "epoch": 0.14, + "learning_rate": 1.9364973454105494e-05, + "loss": 1.2534, + "step": 5848 + }, + { + "epoch": 0.14, + "learning_rate": 1.9364700287165604e-05, + "loss": 1.1318, + "step": 5849 + }, + { + "epoch": 0.14, + "learning_rate": 1.9364427063412276e-05, + "loss": 1.0313, + "step": 5850 + }, + { + "epoch": 0.14, + "learning_rate": 1.9364153782847174e-05, + "loss": 0.9864, + "step": 5851 + }, + { + "epoch": 0.14, + "learning_rate": 1.936388044547195e-05, + "loss": 1.2152, + "step": 5852 + }, + { + "epoch": 0.14, + "learning_rate": 1.936360705128827e-05, + "loss": 0.945, + "step": 5853 + }, + { + "epoch": 0.14, + "learning_rate": 1.9363333600297784e-05, + "loss": 1.1649, + "step": 5854 + }, + { + "epoch": 0.14, + "learning_rate": 1.936306009250216e-05, + "loss": 1.127, + "step": 5855 + }, + { + "epoch": 0.14, + "learning_rate": 1.936278652790305e-05, + "loss": 1.0694, + "step": 5856 + }, + { + "epoch": 0.14, + "learning_rate": 1.9362512906502117e-05, + "loss": 1.2716, + "step": 5857 + }, + { + "epoch": 0.14, + "learning_rate": 1.9362239228301023e-05, + "loss": 1.1133, + "step": 5858 + }, + { + "epoch": 0.14, + "learning_rate": 1.9361965493301424e-05, + "loss": 1.2368, + "step": 5859 + }, + { + "epoch": 0.14, + "learning_rate": 1.936169170150498e-05, + "loss": 0.9978, + "step": 5860 + }, + { + "epoch": 0.14, + "learning_rate": 1.936141785291336e-05, + "loss": 0.9538, + "step": 5861 + }, + { + "epoch": 0.14, + "learning_rate": 1.936114394752821e-05, + "loss": 1.1437, + "step": 5862 + }, + { + "epoch": 0.14, + "learning_rate": 1.936086998535121e-05, + "loss": 1.0623, + "step": 5863 + }, + { + "epoch": 0.14, + "learning_rate": 1.936059596638401e-05, + "loss": 0.8912, + "step": 5864 + }, + { + "epoch": 0.14, + "learning_rate": 1.936032189062828e-05, + "loss": 1.3665, + "step": 5865 + }, + { + "epoch": 0.14, + "learning_rate": 1.9360047758085675e-05, + "loss": 1.2124, + "step": 5866 + }, + { + "epoch": 0.14, + "learning_rate": 1.9359773568757864e-05, + "loss": 1.0842, + "step": 5867 + }, + { + "epoch": 0.14, + "learning_rate": 1.9359499322646505e-05, + "loss": 1.1771, + "step": 5868 + }, + { + "epoch": 0.14, + "learning_rate": 1.935922501975327e-05, + "loss": 1.3186, + "step": 5869 + }, + { + "epoch": 0.14, + "learning_rate": 1.9358950660079815e-05, + "loss": 1.4651, + "step": 5870 + }, + { + "epoch": 0.14, + "learning_rate": 1.9358676243627807e-05, + "loss": 1.0491, + "step": 5871 + }, + { + "epoch": 0.14, + "learning_rate": 1.9358401770398912e-05, + "loss": 1.0979, + "step": 5872 + }, + { + "epoch": 0.14, + "learning_rate": 1.93581272403948e-05, + "loss": 1.3361, + "step": 5873 + }, + { + "epoch": 0.14, + "learning_rate": 1.9357852653617123e-05, + "loss": 1.0113, + "step": 5874 + }, + { + "epoch": 0.14, + "learning_rate": 1.9357578010067558e-05, + "loss": 1.1287, + "step": 5875 + }, + { + "epoch": 0.14, + "learning_rate": 1.9357303309747772e-05, + "loss": 1.1522, + "step": 5876 + }, + { + "epoch": 0.14, + "learning_rate": 1.9357028552659424e-05, + "loss": 1.2349, + "step": 5877 + }, + { + "epoch": 0.14, + "learning_rate": 1.9356753738804185e-05, + "loss": 1.2834, + "step": 5878 + }, + { + "epoch": 0.14, + "learning_rate": 1.9356478868183722e-05, + "loss": 1.2669, + "step": 5879 + }, + { + "epoch": 0.14, + "learning_rate": 1.9356203940799702e-05, + "loss": 1.2289, + "step": 5880 + }, + { + "epoch": 0.14, + "learning_rate": 1.9355928956653795e-05, + "loss": 1.1426, + "step": 5881 + }, + { + "epoch": 0.14, + "learning_rate": 1.9355653915747668e-05, + "loss": 1.2977, + "step": 5882 + }, + { + "epoch": 0.14, + "learning_rate": 1.9355378818082988e-05, + "loss": 1.2607, + "step": 5883 + }, + { + "epoch": 0.14, + "learning_rate": 1.9355103663661426e-05, + "loss": 1.026, + "step": 5884 + }, + { + "epoch": 0.14, + "learning_rate": 1.935482845248465e-05, + "loss": 1.2647, + "step": 5885 + }, + { + "epoch": 0.14, + "learning_rate": 1.935455318455433e-05, + "loss": 0.9971, + "step": 5886 + }, + { + "epoch": 0.14, + "learning_rate": 1.9354277859872136e-05, + "loss": 1.0043, + "step": 5887 + }, + { + "epoch": 0.14, + "learning_rate": 1.935400247843974e-05, + "loss": 1.2153, + "step": 5888 + }, + { + "epoch": 0.14, + "learning_rate": 1.935372704025881e-05, + "loss": 1.1488, + "step": 5889 + }, + { + "epoch": 0.14, + "learning_rate": 1.935345154533102e-05, + "loss": 1.2286, + "step": 5890 + }, + { + "epoch": 0.14, + "learning_rate": 1.9353175993658035e-05, + "loss": 1.2208, + "step": 5891 + }, + { + "epoch": 0.14, + "learning_rate": 1.9352900385241534e-05, + "loss": 1.04, + "step": 5892 + }, + { + "epoch": 0.14, + "learning_rate": 1.9352624720083187e-05, + "loss": 1.14, + "step": 5893 + }, + { + "epoch": 0.14, + "learning_rate": 1.9352348998184664e-05, + "loss": 1.206, + "step": 5894 + }, + { + "epoch": 0.14, + "learning_rate": 1.935207321954764e-05, + "loss": 1.0427, + "step": 5895 + }, + { + "epoch": 0.14, + "learning_rate": 1.9351797384173787e-05, + "loss": 1.1559, + "step": 5896 + }, + { + "epoch": 0.14, + "learning_rate": 1.9351521492064783e-05, + "loss": 1.1933, + "step": 5897 + }, + { + "epoch": 0.14, + "learning_rate": 1.9351245543222292e-05, + "loss": 1.1078, + "step": 5898 + }, + { + "epoch": 0.14, + "learning_rate": 1.9350969537648e-05, + "loss": 1.0463, + "step": 5899 + }, + { + "epoch": 0.14, + "learning_rate": 1.935069347534357e-05, + "loss": 1.4833, + "step": 5900 + }, + { + "epoch": 0.14, + "learning_rate": 1.935041735631068e-05, + "loss": 1.2907, + "step": 5901 + }, + { + "epoch": 0.14, + "learning_rate": 1.9350141180551014e-05, + "loss": 1.1895, + "step": 5902 + }, + { + "epoch": 0.14, + "learning_rate": 1.934986494806624e-05, + "loss": 1.2937, + "step": 5903 + }, + { + "epoch": 0.14, + "learning_rate": 1.934958865885803e-05, + "loss": 1.0293, + "step": 5904 + }, + { + "epoch": 0.14, + "learning_rate": 1.9349312312928066e-05, + "loss": 0.889, + "step": 5905 + }, + { + "epoch": 0.14, + "learning_rate": 1.9349035910278027e-05, + "loss": 1.0581, + "step": 5906 + }, + { + "epoch": 0.14, + "learning_rate": 1.934875945090958e-05, + "loss": 1.2008, + "step": 5907 + }, + { + "epoch": 0.14, + "learning_rate": 1.9348482934824413e-05, + "loss": 1.0311, + "step": 5908 + }, + { + "epoch": 0.14, + "learning_rate": 1.93482063620242e-05, + "loss": 1.1203, + "step": 5909 + }, + { + "epoch": 0.14, + "learning_rate": 1.9347929732510614e-05, + "loss": 1.1691, + "step": 5910 + }, + { + "epoch": 0.14, + "learning_rate": 1.934765304628534e-05, + "loss": 1.2022, + "step": 5911 + }, + { + "epoch": 0.14, + "learning_rate": 1.9347376303350056e-05, + "loss": 1.0078, + "step": 5912 + }, + { + "epoch": 0.14, + "learning_rate": 1.934709950370643e-05, + "loss": 0.7422, + "step": 5913 + }, + { + "epoch": 0.14, + "learning_rate": 1.9346822647356158e-05, + "loss": 1.1725, + "step": 5914 + }, + { + "epoch": 0.14, + "learning_rate": 1.9346545734300907e-05, + "loss": 0.8905, + "step": 5915 + }, + { + "epoch": 0.14, + "learning_rate": 1.934626876454236e-05, + "loss": 1.2284, + "step": 5916 + }, + { + "epoch": 0.14, + "learning_rate": 1.9345991738082202e-05, + "loss": 0.9654, + "step": 5917 + }, + { + "epoch": 0.14, + "learning_rate": 1.9345714654922112e-05, + "loss": 1.1309, + "step": 5918 + }, + { + "epoch": 0.14, + "learning_rate": 1.9345437515063768e-05, + "loss": 1.4053, + "step": 5919 + }, + { + "epoch": 0.14, + "learning_rate": 1.9345160318508853e-05, + "loss": 1.1531, + "step": 5920 + }, + { + "epoch": 0.14, + "learning_rate": 1.934488306525905e-05, + "loss": 1.4125, + "step": 5921 + }, + { + "epoch": 0.14, + "learning_rate": 1.9344605755316035e-05, + "loss": 1.1652, + "step": 5922 + }, + { + "epoch": 0.14, + "learning_rate": 1.93443283886815e-05, + "loss": 1.1578, + "step": 5923 + }, + { + "epoch": 0.14, + "learning_rate": 1.9344050965357117e-05, + "loss": 1.1542, + "step": 5924 + }, + { + "epoch": 0.14, + "learning_rate": 1.934377348534458e-05, + "loss": 0.9481, + "step": 5925 + }, + { + "epoch": 0.14, + "learning_rate": 1.9343495948645562e-05, + "loss": 1.1323, + "step": 5926 + }, + { + "epoch": 0.14, + "learning_rate": 1.9343218355261756e-05, + "loss": 1.2717, + "step": 5927 + }, + { + "epoch": 0.14, + "learning_rate": 1.9342940705194838e-05, + "loss": 1.1351, + "step": 5928 + }, + { + "epoch": 0.14, + "learning_rate": 1.9342662998446497e-05, + "loss": 0.9918, + "step": 5929 + }, + { + "epoch": 0.14, + "learning_rate": 1.934238523501842e-05, + "loss": 1.3163, + "step": 5930 + }, + { + "epoch": 0.14, + "learning_rate": 1.9342107414912287e-05, + "loss": 1.0142, + "step": 5931 + }, + { + "epoch": 0.14, + "learning_rate": 1.9341829538129787e-05, + "loss": 1.2873, + "step": 5932 + }, + { + "epoch": 0.14, + "learning_rate": 1.9341551604672607e-05, + "loss": 1.0066, + "step": 5933 + }, + { + "epoch": 0.14, + "learning_rate": 1.9341273614542427e-05, + "loss": 1.2753, + "step": 5934 + }, + { + "epoch": 0.14, + "learning_rate": 1.934099556774094e-05, + "loss": 0.9941, + "step": 5935 + }, + { + "epoch": 0.14, + "learning_rate": 1.9340717464269823e-05, + "loss": 1.4147, + "step": 5936 + }, + { + "epoch": 0.14, + "learning_rate": 1.9340439304130777e-05, + "loss": 1.2596, + "step": 5937 + }, + { + "epoch": 0.14, + "learning_rate": 1.9340161087325483e-05, + "loss": 1.2206, + "step": 5938 + }, + { + "epoch": 0.14, + "learning_rate": 1.9339882813855624e-05, + "loss": 0.9186, + "step": 5939 + }, + { + "epoch": 0.14, + "learning_rate": 1.9339604483722896e-05, + "loss": 1.1637, + "step": 5940 + }, + { + "epoch": 0.14, + "learning_rate": 1.9339326096928986e-05, + "loss": 1.1867, + "step": 5941 + }, + { + "epoch": 0.14, + "learning_rate": 1.933904765347558e-05, + "loss": 1.3661, + "step": 5942 + }, + { + "epoch": 0.14, + "learning_rate": 1.9338769153364365e-05, + "loss": 0.9654, + "step": 5943 + }, + { + "epoch": 0.14, + "learning_rate": 1.933849059659704e-05, + "loss": 0.9261, + "step": 5944 + }, + { + "epoch": 0.14, + "learning_rate": 1.9338211983175283e-05, + "loss": 1.1893, + "step": 5945 + }, + { + "epoch": 0.14, + "learning_rate": 1.9337933313100793e-05, + "loss": 1.3543, + "step": 5946 + }, + { + "epoch": 0.14, + "learning_rate": 1.933765458637526e-05, + "loss": 1.199, + "step": 5947 + }, + { + "epoch": 0.14, + "learning_rate": 1.933737580300037e-05, + "loss": 1.1631, + "step": 5948 + }, + { + "epoch": 0.14, + "learning_rate": 1.933709696297782e-05, + "loss": 1.1617, + "step": 5949 + }, + { + "epoch": 0.14, + "learning_rate": 1.9336818066309297e-05, + "loss": 1.0525, + "step": 5950 + }, + { + "epoch": 0.14, + "learning_rate": 1.9336539112996496e-05, + "loss": 1.1842, + "step": 5951 + }, + { + "epoch": 0.14, + "learning_rate": 1.9336260103041108e-05, + "loss": 1.136, + "step": 5952 + }, + { + "epoch": 0.14, + "learning_rate": 1.9335981036444823e-05, + "loss": 1.1576, + "step": 5953 + }, + { + "epoch": 0.14, + "learning_rate": 1.9335701913209342e-05, + "loss": 1.176, + "step": 5954 + }, + { + "epoch": 0.14, + "learning_rate": 1.933542273333635e-05, + "loss": 1.1654, + "step": 5955 + }, + { + "epoch": 0.14, + "learning_rate": 1.9335143496827546e-05, + "loss": 1.3269, + "step": 5956 + }, + { + "epoch": 0.14, + "learning_rate": 1.9334864203684622e-05, + "loss": 1.0233, + "step": 5957 + }, + { + "epoch": 0.14, + "learning_rate": 1.933458485390927e-05, + "loss": 1.0023, + "step": 5958 + }, + { + "epoch": 0.14, + "learning_rate": 1.9334305447503192e-05, + "loss": 1.218, + "step": 5959 + }, + { + "epoch": 0.14, + "learning_rate": 1.9334025984468075e-05, + "loss": 0.9704, + "step": 5960 + }, + { + "epoch": 0.14, + "learning_rate": 1.933374646480562e-05, + "loss": 1.3256, + "step": 5961 + }, + { + "epoch": 0.14, + "learning_rate": 1.9333466888517518e-05, + "loss": 1.2886, + "step": 5962 + }, + { + "epoch": 0.14, + "learning_rate": 1.9333187255605467e-05, + "loss": 1.2812, + "step": 5963 + }, + { + "epoch": 0.14, + "learning_rate": 1.9332907566071168e-05, + "loss": 1.2343, + "step": 5964 + }, + { + "epoch": 0.14, + "learning_rate": 1.933262781991631e-05, + "loss": 1.0633, + "step": 5965 + }, + { + "epoch": 0.14, + "learning_rate": 1.9332348017142598e-05, + "loss": 1.4706, + "step": 5966 + }, + { + "epoch": 0.14, + "learning_rate": 1.9332068157751724e-05, + "loss": 1.091, + "step": 5967 + }, + { + "epoch": 0.14, + "learning_rate": 1.933178824174539e-05, + "loss": 1.2252, + "step": 5968 + }, + { + "epoch": 0.14, + "learning_rate": 1.9331508269125286e-05, + "loss": 1.0633, + "step": 5969 + }, + { + "epoch": 0.14, + "learning_rate": 1.933122823989312e-05, + "loss": 1.1289, + "step": 5970 + }, + { + "epoch": 0.14, + "learning_rate": 1.933094815405059e-05, + "loss": 1.0714, + "step": 5971 + }, + { + "epoch": 0.14, + "learning_rate": 1.9330668011599388e-05, + "loss": 1.1788, + "step": 5972 + }, + { + "epoch": 0.14, + "learning_rate": 1.933038781254122e-05, + "loss": 1.0057, + "step": 5973 + }, + { + "epoch": 0.14, + "learning_rate": 1.933010755687778e-05, + "loss": 1.2269, + "step": 5974 + }, + { + "epoch": 0.14, + "learning_rate": 1.9329827244610775e-05, + "loss": 0.9989, + "step": 5975 + }, + { + "epoch": 0.14, + "learning_rate": 1.9329546875741904e-05, + "loss": 1.3602, + "step": 5976 + }, + { + "epoch": 0.14, + "learning_rate": 1.9329266450272864e-05, + "loss": 1.0848, + "step": 5977 + }, + { + "epoch": 0.14, + "learning_rate": 1.932898596820536e-05, + "loss": 1.1038, + "step": 5978 + }, + { + "epoch": 0.14, + "learning_rate": 1.9328705429541094e-05, + "loss": 1.0452, + "step": 5979 + }, + { + "epoch": 0.14, + "learning_rate": 1.9328424834281763e-05, + "loss": 1.0521, + "step": 5980 + }, + { + "epoch": 0.14, + "learning_rate": 1.9328144182429073e-05, + "loss": 1.2049, + "step": 5981 + }, + { + "epoch": 0.14, + "learning_rate": 1.932786347398473e-05, + "loss": 1.2522, + "step": 5982 + }, + { + "epoch": 0.14, + "learning_rate": 1.9327582708950427e-05, + "loss": 1.2587, + "step": 5983 + }, + { + "epoch": 0.14, + "learning_rate": 1.932730188732788e-05, + "loss": 0.9753, + "step": 5984 + }, + { + "epoch": 0.14, + "learning_rate": 1.9327021009118785e-05, + "loss": 1.4577, + "step": 5985 + }, + { + "epoch": 0.14, + "learning_rate": 1.9326740074324846e-05, + "loss": 0.9929, + "step": 5986 + }, + { + "epoch": 0.14, + "learning_rate": 1.9326459082947768e-05, + "loss": 1.6311, + "step": 5987 + }, + { + "epoch": 0.14, + "learning_rate": 1.932617803498926e-05, + "loss": 1.3353, + "step": 5988 + }, + { + "epoch": 0.14, + "learning_rate": 1.9325896930451022e-05, + "loss": 1.1575, + "step": 5989 + }, + { + "epoch": 0.14, + "learning_rate": 1.9325615769334755e-05, + "loss": 1.1762, + "step": 5990 + }, + { + "epoch": 0.14, + "learning_rate": 1.932533455164218e-05, + "loss": 0.9735, + "step": 5991 + }, + { + "epoch": 0.14, + "learning_rate": 1.9325053277374986e-05, + "loss": 1.115, + "step": 5992 + }, + { + "epoch": 0.14, + "learning_rate": 1.932477194653489e-05, + "loss": 0.9703, + "step": 5993 + }, + { + "epoch": 0.14, + "learning_rate": 1.93244905591236e-05, + "loss": 1.2312, + "step": 5994 + }, + { + "epoch": 0.14, + "learning_rate": 1.9324209115142815e-05, + "loss": 1.2174, + "step": 5995 + }, + { + "epoch": 0.14, + "learning_rate": 1.9323927614594245e-05, + "loss": 1.1816, + "step": 5996 + }, + { + "epoch": 0.14, + "learning_rate": 1.93236460574796e-05, + "loss": 1.1191, + "step": 5997 + }, + { + "epoch": 0.14, + "learning_rate": 1.932336444380059e-05, + "loss": 1.1025, + "step": 5998 + }, + { + "epoch": 0.14, + "learning_rate": 1.932308277355892e-05, + "loss": 1.3129, + "step": 5999 + }, + { + "epoch": 0.14, + "learning_rate": 1.93228010467563e-05, + "loss": 1.1838, + "step": 6000 + }, + { + "epoch": 0.14, + "learning_rate": 1.9322519263394438e-05, + "loss": 1.4698, + "step": 6001 + }, + { + "epoch": 0.14, + "learning_rate": 1.9322237423475044e-05, + "loss": 0.8619, + "step": 6002 + }, + { + "epoch": 0.14, + "learning_rate": 1.932195552699983e-05, + "loss": 1.2349, + "step": 6003 + }, + { + "epoch": 0.14, + "learning_rate": 1.9321673573970502e-05, + "loss": 1.1472, + "step": 6004 + }, + { + "epoch": 0.14, + "learning_rate": 1.9321391564388773e-05, + "loss": 1.2111, + "step": 6005 + }, + { + "epoch": 0.14, + "learning_rate": 1.932110949825636e-05, + "loss": 1.361, + "step": 6006 + }, + { + "epoch": 0.14, + "learning_rate": 1.932082737557496e-05, + "loss": 1.1571, + "step": 6007 + }, + { + "epoch": 0.14, + "learning_rate": 1.9320545196346295e-05, + "loss": 1.1447, + "step": 6008 + }, + { + "epoch": 0.14, + "learning_rate": 1.9320262960572077e-05, + "loss": 1.4029, + "step": 6009 + }, + { + "epoch": 0.14, + "learning_rate": 1.9319980668254016e-05, + "loss": 1.1499, + "step": 6010 + }, + { + "epoch": 0.14, + "learning_rate": 1.9319698319393822e-05, + "loss": 1.1832, + "step": 6011 + }, + { + "epoch": 0.14, + "learning_rate": 1.9319415913993214e-05, + "loss": 1.1835, + "step": 6012 + }, + { + "epoch": 0.14, + "learning_rate": 1.9319133452053898e-05, + "loss": 1.0577, + "step": 6013 + }, + { + "epoch": 0.14, + "learning_rate": 1.9318850933577592e-05, + "loss": 1.1808, + "step": 6014 + }, + { + "epoch": 0.14, + "learning_rate": 1.931856835856601e-05, + "loss": 1.2067, + "step": 6015 + }, + { + "epoch": 0.14, + "learning_rate": 1.9318285727020867e-05, + "loss": 1.138, + "step": 6016 + }, + { + "epoch": 0.14, + "learning_rate": 1.9318003038943873e-05, + "loss": 1.2247, + "step": 6017 + }, + { + "epoch": 0.14, + "learning_rate": 1.9317720294336747e-05, + "loss": 1.1345, + "step": 6018 + }, + { + "epoch": 0.14, + "learning_rate": 1.9317437493201207e-05, + "loss": 1.2274, + "step": 6019 + }, + { + "epoch": 0.14, + "learning_rate": 1.9317154635538964e-05, + "loss": 1.3889, + "step": 6020 + }, + { + "epoch": 0.14, + "learning_rate": 1.9316871721351732e-05, + "loss": 1.0427, + "step": 6021 + }, + { + "epoch": 0.14, + "learning_rate": 1.9316588750641233e-05, + "loss": 0.9812, + "step": 6022 + }, + { + "epoch": 0.14, + "learning_rate": 1.9316305723409184e-05, + "loss": 1.1154, + "step": 6023 + }, + { + "epoch": 0.14, + "learning_rate": 1.9316022639657296e-05, + "loss": 1.0002, + "step": 6024 + }, + { + "epoch": 0.14, + "learning_rate": 1.931573949938729e-05, + "loss": 1.3014, + "step": 6025 + }, + { + "epoch": 0.14, + "learning_rate": 1.9315456302600885e-05, + "loss": 1.1062, + "step": 6026 + }, + { + "epoch": 0.14, + "learning_rate": 1.93151730492998e-05, + "loss": 1.2708, + "step": 6027 + }, + { + "epoch": 0.14, + "learning_rate": 1.9314889739485747e-05, + "loss": 1.0026, + "step": 6028 + }, + { + "epoch": 0.14, + "learning_rate": 1.931460637316045e-05, + "loss": 1.1669, + "step": 6029 + }, + { + "epoch": 0.15, + "learning_rate": 1.931432295032563e-05, + "loss": 1.1157, + "step": 6030 + }, + { + "epoch": 0.15, + "learning_rate": 1.9314039470983e-05, + "loss": 1.3661, + "step": 6031 + }, + { + "epoch": 0.15, + "learning_rate": 1.9313755935134286e-05, + "loss": 1.1705, + "step": 6032 + }, + { + "epoch": 0.15, + "learning_rate": 1.9313472342781204e-05, + "loss": 1.2551, + "step": 6033 + }, + { + "epoch": 0.15, + "learning_rate": 1.9313188693925475e-05, + "loss": 1.0924, + "step": 6034 + }, + { + "epoch": 0.15, + "learning_rate": 1.931290498856882e-05, + "loss": 1.1629, + "step": 6035 + }, + { + "epoch": 0.15, + "learning_rate": 1.9312621226712962e-05, + "loss": 1.1059, + "step": 6036 + }, + { + "epoch": 0.15, + "learning_rate": 1.931233740835962e-05, + "loss": 1.4778, + "step": 6037 + }, + { + "epoch": 0.15, + "learning_rate": 1.9312053533510525e-05, + "loss": 1.2104, + "step": 6038 + }, + { + "epoch": 0.15, + "learning_rate": 1.9311769602167386e-05, + "loss": 1.1497, + "step": 6039 + }, + { + "epoch": 0.15, + "learning_rate": 1.9311485614331928e-05, + "loss": 1.1193, + "step": 6040 + }, + { + "epoch": 0.15, + "learning_rate": 1.931120157000588e-05, + "loss": 1.1146, + "step": 6041 + }, + { + "epoch": 0.15, + "learning_rate": 1.9310917469190965e-05, + "loss": 0.9642, + "step": 6042 + }, + { + "epoch": 0.15, + "learning_rate": 1.93106333118889e-05, + "loss": 1.1251, + "step": 6043 + }, + { + "epoch": 0.15, + "learning_rate": 1.9310349098101412e-05, + "loss": 1.0236, + "step": 6044 + }, + { + "epoch": 0.15, + "learning_rate": 1.931006482783023e-05, + "loss": 1.1905, + "step": 6045 + }, + { + "epoch": 0.15, + "learning_rate": 1.930978050107707e-05, + "loss": 1.167, + "step": 6046 + }, + { + "epoch": 0.15, + "learning_rate": 1.9309496117843663e-05, + "loss": 1.0581, + "step": 6047 + }, + { + "epoch": 0.15, + "learning_rate": 1.9309211678131733e-05, + "loss": 1.1289, + "step": 6048 + }, + { + "epoch": 0.15, + "learning_rate": 1.930892718194301e-05, + "loss": 1.0868, + "step": 6049 + }, + { + "epoch": 0.15, + "learning_rate": 1.930864262927921e-05, + "loss": 1.1717, + "step": 6050 + }, + { + "epoch": 0.15, + "learning_rate": 1.9308358020142065e-05, + "loss": 1.247, + "step": 6051 + }, + { + "epoch": 0.15, + "learning_rate": 1.9308073354533302e-05, + "loss": 1.2195, + "step": 6052 + }, + { + "epoch": 0.15, + "learning_rate": 1.930778863245465e-05, + "loss": 1.3185, + "step": 6053 + }, + { + "epoch": 0.15, + "learning_rate": 1.9307503853907832e-05, + "loss": 1.2362, + "step": 6054 + }, + { + "epoch": 0.15, + "learning_rate": 1.9307219018894577e-05, + "loss": 1.2632, + "step": 6055 + }, + { + "epoch": 0.15, + "learning_rate": 1.930693412741661e-05, + "loss": 1.0, + "step": 6056 + }, + { + "epoch": 0.15, + "learning_rate": 1.9306649179475666e-05, + "loss": 1.3021, + "step": 6057 + }, + { + "epoch": 0.15, + "learning_rate": 1.9306364175073474e-05, + "loss": 0.8104, + "step": 6058 + }, + { + "epoch": 0.15, + "learning_rate": 1.9306079114211755e-05, + "loss": 1.1199, + "step": 6059 + }, + { + "epoch": 0.15, + "learning_rate": 1.9305793996892244e-05, + "loss": 1.2126, + "step": 6060 + }, + { + "epoch": 0.15, + "learning_rate": 1.930550882311667e-05, + "loss": 1.2725, + "step": 6061 + }, + { + "epoch": 0.15, + "learning_rate": 1.930522359288676e-05, + "loss": 1.0506, + "step": 6062 + }, + { + "epoch": 0.15, + "learning_rate": 1.930493830620425e-05, + "loss": 1.3628, + "step": 6063 + }, + { + "epoch": 0.15, + "learning_rate": 1.9304652963070868e-05, + "loss": 1.0119, + "step": 6064 + }, + { + "epoch": 0.15, + "learning_rate": 1.9304367563488344e-05, + "loss": 1.1215, + "step": 6065 + }, + { + "epoch": 0.15, + "learning_rate": 1.9304082107458412e-05, + "loss": 1.0776, + "step": 6066 + }, + { + "epoch": 0.15, + "learning_rate": 1.93037965949828e-05, + "loss": 1.0716, + "step": 6067 + }, + { + "epoch": 0.15, + "learning_rate": 1.9303511026063244e-05, + "loss": 0.9437, + "step": 6068 + }, + { + "epoch": 0.15, + "learning_rate": 1.9303225400701475e-05, + "loss": 0.9962, + "step": 6069 + }, + { + "epoch": 0.15, + "learning_rate": 1.930293971889923e-05, + "loss": 1.4805, + "step": 6070 + }, + { + "epoch": 0.15, + "learning_rate": 1.930265398065823e-05, + "loss": 1.3928, + "step": 6071 + }, + { + "epoch": 0.15, + "learning_rate": 1.9302368185980218e-05, + "loss": 1.1597, + "step": 6072 + }, + { + "epoch": 0.15, + "learning_rate": 1.930208233486693e-05, + "loss": 1.0164, + "step": 6073 + }, + { + "epoch": 0.15, + "learning_rate": 1.9301796427320093e-05, + "loss": 1.0275, + "step": 6074 + }, + { + "epoch": 0.15, + "learning_rate": 1.9301510463341448e-05, + "loss": 1.1955, + "step": 6075 + }, + { + "epoch": 0.15, + "learning_rate": 1.9301224442932725e-05, + "loss": 1.0087, + "step": 6076 + }, + { + "epoch": 0.15, + "learning_rate": 1.930093836609566e-05, + "loss": 1.0081, + "step": 6077 + }, + { + "epoch": 0.15, + "learning_rate": 1.930065223283199e-05, + "loss": 1.1911, + "step": 6078 + }, + { + "epoch": 0.15, + "learning_rate": 1.9300366043143453e-05, + "loss": 0.7709, + "step": 6079 + }, + { + "epoch": 0.15, + "learning_rate": 1.930007979703178e-05, + "loss": 1.3073, + "step": 6080 + }, + { + "epoch": 0.15, + "learning_rate": 1.9299793494498714e-05, + "loss": 1.4881, + "step": 6081 + }, + { + "epoch": 0.15, + "learning_rate": 1.9299507135545986e-05, + "loss": 0.9238, + "step": 6082 + }, + { + "epoch": 0.15, + "learning_rate": 1.9299220720175338e-05, + "loss": 1.1881, + "step": 6083 + }, + { + "epoch": 0.15, + "learning_rate": 1.92989342483885e-05, + "loss": 1.1028, + "step": 6084 + }, + { + "epoch": 0.15, + "learning_rate": 1.929864772018722e-05, + "loss": 1.3524, + "step": 6085 + }, + { + "epoch": 0.15, + "learning_rate": 1.929836113557323e-05, + "loss": 1.3384, + "step": 6086 + }, + { + "epoch": 0.15, + "learning_rate": 1.929807449454827e-05, + "loss": 1.0625, + "step": 6087 + }, + { + "epoch": 0.15, + "learning_rate": 1.9297787797114078e-05, + "loss": 1.0892, + "step": 6088 + }, + { + "epoch": 0.15, + "learning_rate": 1.9297501043272395e-05, + "loss": 1.339, + "step": 6089 + }, + { + "epoch": 0.15, + "learning_rate": 1.929721423302496e-05, + "loss": 1.0017, + "step": 6090 + }, + { + "epoch": 0.15, + "learning_rate": 1.9296927366373515e-05, + "loss": 1.3002, + "step": 6091 + }, + { + "epoch": 0.15, + "learning_rate": 1.9296640443319793e-05, + "loss": 1.0258, + "step": 6092 + }, + { + "epoch": 0.15, + "learning_rate": 1.9296353463865544e-05, + "loss": 1.2834, + "step": 6093 + }, + { + "epoch": 0.15, + "learning_rate": 1.9296066428012508e-05, + "loss": 1.5381, + "step": 6094 + }, + { + "epoch": 0.15, + "learning_rate": 1.929577933576242e-05, + "loss": 0.9588, + "step": 6095 + }, + { + "epoch": 0.15, + "learning_rate": 1.9295492187117025e-05, + "loss": 1.1825, + "step": 6096 + }, + { + "epoch": 0.15, + "learning_rate": 1.929520498207807e-05, + "loss": 0.8539, + "step": 6097 + }, + { + "epoch": 0.15, + "learning_rate": 1.9294917720647287e-05, + "loss": 1.2107, + "step": 6098 + }, + { + "epoch": 0.15, + "learning_rate": 1.9294630402826425e-05, + "loss": 1.0377, + "step": 6099 + }, + { + "epoch": 0.15, + "learning_rate": 1.929434302861723e-05, + "loss": 1.2562, + "step": 6100 + }, + { + "epoch": 0.15, + "learning_rate": 1.9294055598021438e-05, + "loss": 1.1539, + "step": 6101 + }, + { + "epoch": 0.15, + "learning_rate": 1.92937681110408e-05, + "loss": 1.3309, + "step": 6102 + }, + { + "epoch": 0.15, + "learning_rate": 1.9293480567677056e-05, + "loss": 0.9353, + "step": 6103 + }, + { + "epoch": 0.15, + "learning_rate": 1.929319296793195e-05, + "loss": 1.075, + "step": 6104 + }, + { + "epoch": 0.15, + "learning_rate": 1.929290531180723e-05, + "loss": 1.3656, + "step": 6105 + }, + { + "epoch": 0.15, + "learning_rate": 1.929261759930464e-05, + "loss": 1.1984, + "step": 6106 + }, + { + "epoch": 0.15, + "learning_rate": 1.9292329830425922e-05, + "loss": 1.2508, + "step": 6107 + }, + { + "epoch": 0.15, + "learning_rate": 1.9292042005172823e-05, + "loss": 1.1552, + "step": 6108 + }, + { + "epoch": 0.15, + "learning_rate": 1.9291754123547095e-05, + "loss": 1.4227, + "step": 6109 + }, + { + "epoch": 0.15, + "learning_rate": 1.9291466185550482e-05, + "loss": 1.1341, + "step": 6110 + }, + { + "epoch": 0.15, + "learning_rate": 1.9291178191184724e-05, + "loss": 1.2621, + "step": 6111 + }, + { + "epoch": 0.15, + "learning_rate": 1.9290890140451576e-05, + "loss": 1.0116, + "step": 6112 + }, + { + "epoch": 0.15, + "learning_rate": 1.9290602033352783e-05, + "loss": 1.0277, + "step": 6113 + }, + { + "epoch": 0.15, + "learning_rate": 1.929031386989009e-05, + "loss": 1.0978, + "step": 6114 + }, + { + "epoch": 0.15, + "learning_rate": 1.929002565006525e-05, + "loss": 1.3203, + "step": 6115 + }, + { + "epoch": 0.15, + "learning_rate": 1.928973737388001e-05, + "loss": 0.9301, + "step": 6116 + }, + { + "epoch": 0.15, + "learning_rate": 1.928944904133612e-05, + "loss": 1.0974, + "step": 6117 + }, + { + "epoch": 0.15, + "learning_rate": 1.9289160652435326e-05, + "loss": 1.2512, + "step": 6118 + }, + { + "epoch": 0.15, + "learning_rate": 1.9288872207179378e-05, + "loss": 0.9942, + "step": 6119 + }, + { + "epoch": 0.15, + "learning_rate": 1.9288583705570026e-05, + "loss": 1.1737, + "step": 6120 + }, + { + "epoch": 0.15, + "learning_rate": 1.9288295147609023e-05, + "loss": 1.1957, + "step": 6121 + }, + { + "epoch": 0.15, + "learning_rate": 1.928800653329812e-05, + "loss": 1.4628, + "step": 6122 + }, + { + "epoch": 0.15, + "learning_rate": 1.9287717862639067e-05, + "loss": 1.1344, + "step": 6123 + }, + { + "epoch": 0.15, + "learning_rate": 1.928742913563361e-05, + "loss": 1.3994, + "step": 6124 + }, + { + "epoch": 0.15, + "learning_rate": 1.928714035228351e-05, + "loss": 1.216, + "step": 6125 + }, + { + "epoch": 0.15, + "learning_rate": 1.928685151259051e-05, + "loss": 1.1146, + "step": 6126 + }, + { + "epoch": 0.15, + "learning_rate": 1.9286562616556365e-05, + "loss": 0.9984, + "step": 6127 + }, + { + "epoch": 0.15, + "learning_rate": 1.9286273664182832e-05, + "loss": 1.1939, + "step": 6128 + }, + { + "epoch": 0.15, + "learning_rate": 1.928598465547166e-05, + "loss": 1.2524, + "step": 6129 + }, + { + "epoch": 0.15, + "learning_rate": 1.9285695590424604e-05, + "loss": 1.3543, + "step": 6130 + }, + { + "epoch": 0.15, + "learning_rate": 1.9285406469043417e-05, + "loss": 1.1197, + "step": 6131 + }, + { + "epoch": 0.15, + "learning_rate": 1.9285117291329853e-05, + "loss": 1.036, + "step": 6132 + }, + { + "epoch": 0.15, + "learning_rate": 1.9284828057285664e-05, + "loss": 1.2219, + "step": 6133 + }, + { + "epoch": 0.15, + "learning_rate": 1.9284538766912608e-05, + "loss": 1.5199, + "step": 6134 + }, + { + "epoch": 0.15, + "learning_rate": 1.928424942021244e-05, + "loss": 0.9023, + "step": 6135 + }, + { + "epoch": 0.15, + "learning_rate": 1.9283960017186916e-05, + "loss": 1.0442, + "step": 6136 + }, + { + "epoch": 0.15, + "learning_rate": 1.9283670557837787e-05, + "loss": 1.0651, + "step": 6137 + }, + { + "epoch": 0.15, + "learning_rate": 1.928338104216682e-05, + "loss": 1.3656, + "step": 6138 + }, + { + "epoch": 0.15, + "learning_rate": 1.9283091470175755e-05, + "loss": 1.1669, + "step": 6139 + }, + { + "epoch": 0.15, + "learning_rate": 1.928280184186636e-05, + "loss": 1.2, + "step": 6140 + }, + { + "epoch": 0.15, + "learning_rate": 1.928251215724039e-05, + "loss": 1.3671, + "step": 6141 + }, + { + "epoch": 0.15, + "learning_rate": 1.9282222416299604e-05, + "loss": 1.0008, + "step": 6142 + }, + { + "epoch": 0.15, + "learning_rate": 1.9281932619045756e-05, + "loss": 1.0444, + "step": 6143 + }, + { + "epoch": 0.15, + "learning_rate": 1.9281642765480605e-05, + "loss": 1.2192, + "step": 6144 + }, + { + "epoch": 0.15, + "learning_rate": 1.9281352855605914e-05, + "loss": 1.4565, + "step": 6145 + }, + { + "epoch": 0.15, + "learning_rate": 1.9281062889423436e-05, + "loss": 1.1753, + "step": 6146 + }, + { + "epoch": 0.15, + "learning_rate": 1.9280772866934932e-05, + "loss": 1.3095, + "step": 6147 + }, + { + "epoch": 0.15, + "learning_rate": 1.9280482788142162e-05, + "loss": 1.29, + "step": 6148 + }, + { + "epoch": 0.15, + "learning_rate": 1.928019265304689e-05, + "loss": 1.2316, + "step": 6149 + }, + { + "epoch": 0.15, + "learning_rate": 1.9279902461650866e-05, + "loss": 1.354, + "step": 6150 + }, + { + "epoch": 0.15, + "learning_rate": 1.9279612213955857e-05, + "loss": 0.9624, + "step": 6151 + }, + { + "epoch": 0.15, + "learning_rate": 1.9279321909963627e-05, + "loss": 1.4685, + "step": 6152 + }, + { + "epoch": 0.15, + "learning_rate": 1.9279031549675934e-05, + "loss": 1.2734, + "step": 6153 + }, + { + "epoch": 0.15, + "learning_rate": 1.9278741133094535e-05, + "loss": 1.359, + "step": 6154 + }, + { + "epoch": 0.15, + "learning_rate": 1.92784506602212e-05, + "loss": 1.1268, + "step": 6155 + }, + { + "epoch": 0.15, + "learning_rate": 1.9278160131057686e-05, + "loss": 1.1259, + "step": 6156 + }, + { + "epoch": 0.15, + "learning_rate": 1.9277869545605755e-05, + "loss": 1.1977, + "step": 6157 + }, + { + "epoch": 0.15, + "learning_rate": 1.9277578903867174e-05, + "loss": 1.2138, + "step": 6158 + }, + { + "epoch": 0.15, + "learning_rate": 1.9277288205843704e-05, + "loss": 1.2617, + "step": 6159 + }, + { + "epoch": 0.15, + "learning_rate": 1.9276997451537107e-05, + "loss": 1.026, + "step": 6160 + }, + { + "epoch": 0.15, + "learning_rate": 1.9276706640949148e-05, + "loss": 1.3976, + "step": 6161 + }, + { + "epoch": 0.15, + "learning_rate": 1.9276415774081593e-05, + "loss": 1.1565, + "step": 6162 + }, + { + "epoch": 0.15, + "learning_rate": 1.9276124850936206e-05, + "loss": 1.1088, + "step": 6163 + }, + { + "epoch": 0.15, + "learning_rate": 1.927583387151475e-05, + "loss": 1.0853, + "step": 6164 + }, + { + "epoch": 0.15, + "learning_rate": 1.9275542835818994e-05, + "loss": 1.0261, + "step": 6165 + }, + { + "epoch": 0.15, + "learning_rate": 1.92752517438507e-05, + "loss": 1.431, + "step": 6166 + }, + { + "epoch": 0.15, + "learning_rate": 1.9274960595611635e-05, + "loss": 1.2393, + "step": 6167 + }, + { + "epoch": 0.15, + "learning_rate": 1.9274669391103567e-05, + "loss": 1.1327, + "step": 6168 + }, + { + "epoch": 0.15, + "learning_rate": 1.927437813032826e-05, + "loss": 1.2166, + "step": 6169 + }, + { + "epoch": 0.15, + "learning_rate": 1.9274086813287484e-05, + "loss": 1.1558, + "step": 6170 + }, + { + "epoch": 0.15, + "learning_rate": 1.9273795439983002e-05, + "loss": 1.306, + "step": 6171 + }, + { + "epoch": 0.15, + "learning_rate": 1.9273504010416586e-05, + "loss": 1.242, + "step": 6172 + }, + { + "epoch": 0.15, + "learning_rate": 1.927321252459e-05, + "loss": 1.4928, + "step": 6173 + }, + { + "epoch": 0.15, + "learning_rate": 1.927292098250502e-05, + "loss": 0.943, + "step": 6174 + }, + { + "epoch": 0.15, + "learning_rate": 1.9272629384163406e-05, + "loss": 1.0241, + "step": 6175 + }, + { + "epoch": 0.15, + "learning_rate": 1.9272337729566933e-05, + "loss": 1.3757, + "step": 6176 + }, + { + "epoch": 0.15, + "learning_rate": 1.927204601871736e-05, + "loss": 0.8735, + "step": 6177 + }, + { + "epoch": 0.15, + "learning_rate": 1.9271754251616475e-05, + "loss": 1.027, + "step": 6178 + }, + { + "epoch": 0.15, + "learning_rate": 1.9271462428266034e-05, + "loss": 1.1473, + "step": 6179 + }, + { + "epoch": 0.15, + "learning_rate": 1.927117054866781e-05, + "loss": 1.2898, + "step": 6180 + }, + { + "epoch": 0.15, + "learning_rate": 1.927087861282358e-05, + "loss": 1.3543, + "step": 6181 + }, + { + "epoch": 0.15, + "learning_rate": 1.9270586620735102e-05, + "loss": 1.1565, + "step": 6182 + }, + { + "epoch": 0.15, + "learning_rate": 1.927029457240416e-05, + "loss": 1.23, + "step": 6183 + }, + { + "epoch": 0.15, + "learning_rate": 1.927000246783252e-05, + "loss": 1.3041, + "step": 6184 + }, + { + "epoch": 0.15, + "learning_rate": 1.926971030702196e-05, + "loss": 1.3578, + "step": 6185 + }, + { + "epoch": 0.15, + "learning_rate": 1.926941808997424e-05, + "loss": 1.1589, + "step": 6186 + }, + { + "epoch": 0.15, + "learning_rate": 1.9269125816691146e-05, + "loss": 1.1347, + "step": 6187 + }, + { + "epoch": 0.15, + "learning_rate": 1.9268833487174447e-05, + "loss": 1.1379, + "step": 6188 + }, + { + "epoch": 0.15, + "learning_rate": 1.9268541101425913e-05, + "loss": 1.08, + "step": 6189 + }, + { + "epoch": 0.15, + "learning_rate": 1.926824865944732e-05, + "loss": 1.1902, + "step": 6190 + }, + { + "epoch": 0.15, + "learning_rate": 1.9267956161240445e-05, + "loss": 1.1787, + "step": 6191 + }, + { + "epoch": 0.15, + "learning_rate": 1.9267663606807055e-05, + "loss": 1.1119, + "step": 6192 + }, + { + "epoch": 0.15, + "learning_rate": 1.9267370996148936e-05, + "loss": 1.03, + "step": 6193 + }, + { + "epoch": 0.15, + "learning_rate": 1.9267078329267853e-05, + "loss": 0.9803, + "step": 6194 + }, + { + "epoch": 0.15, + "learning_rate": 1.9266785606165585e-05, + "loss": 1.1274, + "step": 6195 + }, + { + "epoch": 0.15, + "learning_rate": 1.926649282684391e-05, + "loss": 1.2431, + "step": 6196 + }, + { + "epoch": 0.15, + "learning_rate": 1.9266199991304602e-05, + "loss": 1.177, + "step": 6197 + }, + { + "epoch": 0.15, + "learning_rate": 1.9265907099549438e-05, + "loss": 1.244, + "step": 6198 + }, + { + "epoch": 0.15, + "learning_rate": 1.9265614151580197e-05, + "loss": 0.773, + "step": 6199 + }, + { + "epoch": 0.15, + "learning_rate": 1.926532114739865e-05, + "loss": 1.2294, + "step": 6200 + }, + { + "epoch": 0.15, + "learning_rate": 1.9265028087006583e-05, + "loss": 1.017, + "step": 6201 + }, + { + "epoch": 0.15, + "learning_rate": 1.926473497040577e-05, + "loss": 1.0833, + "step": 6202 + }, + { + "epoch": 0.15, + "learning_rate": 1.9264441797597987e-05, + "loss": 1.2318, + "step": 6203 + }, + { + "epoch": 0.15, + "learning_rate": 1.9264148568585013e-05, + "loss": 0.9749, + "step": 6204 + }, + { + "epoch": 0.15, + "learning_rate": 1.9263855283368636e-05, + "loss": 1.0093, + "step": 6205 + }, + { + "epoch": 0.15, + "learning_rate": 1.9263561941950622e-05, + "loss": 1.2, + "step": 6206 + }, + { + "epoch": 0.15, + "learning_rate": 1.926326854433276e-05, + "loss": 1.3409, + "step": 6207 + }, + { + "epoch": 0.15, + "learning_rate": 1.926297509051682e-05, + "loss": 1.3049, + "step": 6208 + }, + { + "epoch": 0.15, + "learning_rate": 1.9262681580504597e-05, + "loss": 1.3014, + "step": 6209 + }, + { + "epoch": 0.15, + "learning_rate": 1.926238801429786e-05, + "loss": 1.107, + "step": 6210 + }, + { + "epoch": 0.15, + "learning_rate": 1.9262094391898394e-05, + "loss": 1.275, + "step": 6211 + }, + { + "epoch": 0.15, + "learning_rate": 1.926180071330798e-05, + "loss": 1.1377, + "step": 6212 + }, + { + "epoch": 0.15, + "learning_rate": 1.9261506978528397e-05, + "loss": 0.9766, + "step": 6213 + }, + { + "epoch": 0.15, + "learning_rate": 1.9261213187561433e-05, + "loss": 1.08, + "step": 6214 + }, + { + "epoch": 0.15, + "learning_rate": 1.9260919340408866e-05, + "loss": 1.0501, + "step": 6215 + }, + { + "epoch": 0.15, + "learning_rate": 1.926062543707248e-05, + "loss": 1.1233, + "step": 6216 + }, + { + "epoch": 0.15, + "learning_rate": 1.9260331477554053e-05, + "loss": 1.0987, + "step": 6217 + }, + { + "epoch": 0.15, + "learning_rate": 1.926003746185538e-05, + "loss": 1.242, + "step": 6218 + }, + { + "epoch": 0.15, + "learning_rate": 1.9259743389978233e-05, + "loss": 1.1149, + "step": 6219 + }, + { + "epoch": 0.15, + "learning_rate": 1.9259449261924405e-05, + "loss": 1.3766, + "step": 6220 + }, + { + "epoch": 0.15, + "learning_rate": 1.9259155077695673e-05, + "loss": 1.2989, + "step": 6221 + }, + { + "epoch": 0.15, + "learning_rate": 1.9258860837293824e-05, + "loss": 1.2362, + "step": 6222 + }, + { + "epoch": 0.15, + "learning_rate": 1.9258566540720645e-05, + "loss": 1.3075, + "step": 6223 + }, + { + "epoch": 0.15, + "learning_rate": 1.9258272187977924e-05, + "loss": 1.3009, + "step": 6224 + }, + { + "epoch": 0.15, + "learning_rate": 1.925797777906744e-05, + "loss": 1.2705, + "step": 6225 + }, + { + "epoch": 0.15, + "learning_rate": 1.9257683313990984e-05, + "loss": 1.1394, + "step": 6226 + }, + { + "epoch": 0.15, + "learning_rate": 1.9257388792750343e-05, + "loss": 1.203, + "step": 6227 + }, + { + "epoch": 0.15, + "learning_rate": 1.9257094215347298e-05, + "loss": 1.3583, + "step": 6228 + }, + { + "epoch": 0.15, + "learning_rate": 1.925679958178364e-05, + "loss": 1.1857, + "step": 6229 + }, + { + "epoch": 0.15, + "learning_rate": 1.9256504892061156e-05, + "loss": 0.9445, + "step": 6230 + }, + { + "epoch": 0.15, + "learning_rate": 1.9256210146181635e-05, + "loss": 1.1888, + "step": 6231 + }, + { + "epoch": 0.15, + "learning_rate": 1.9255915344146865e-05, + "loss": 1.0641, + "step": 6232 + }, + { + "epoch": 0.15, + "learning_rate": 1.925562048595863e-05, + "loss": 0.9493, + "step": 6233 + }, + { + "epoch": 0.15, + "learning_rate": 1.9255325571618728e-05, + "loss": 1.3133, + "step": 6234 + }, + { + "epoch": 0.15, + "learning_rate": 1.9255030601128942e-05, + "loss": 1.1684, + "step": 6235 + }, + { + "epoch": 0.15, + "learning_rate": 1.925473557449106e-05, + "loss": 1.0469, + "step": 6236 + }, + { + "epoch": 0.15, + "learning_rate": 1.9254440491706874e-05, + "loss": 1.1476, + "step": 6237 + }, + { + "epoch": 0.15, + "learning_rate": 1.9254145352778176e-05, + "loss": 1.2819, + "step": 6238 + }, + { + "epoch": 0.15, + "learning_rate": 1.9253850157706756e-05, + "loss": 1.2294, + "step": 6239 + }, + { + "epoch": 0.15, + "learning_rate": 1.92535549064944e-05, + "loss": 1.1421, + "step": 6240 + }, + { + "epoch": 0.15, + "learning_rate": 1.9253259599142903e-05, + "loss": 1.2511, + "step": 6241 + }, + { + "epoch": 0.15, + "learning_rate": 1.9252964235654058e-05, + "loss": 1.5443, + "step": 6242 + }, + { + "epoch": 0.15, + "learning_rate": 1.9252668816029656e-05, + "loss": 1.2313, + "step": 6243 + }, + { + "epoch": 0.15, + "learning_rate": 1.925237334027149e-05, + "loss": 1.2637, + "step": 6244 + }, + { + "epoch": 0.15, + "learning_rate": 1.925207780838135e-05, + "loss": 1.4158, + "step": 6245 + }, + { + "epoch": 0.15, + "learning_rate": 1.9251782220361027e-05, + "loss": 0.768, + "step": 6246 + }, + { + "epoch": 0.15, + "learning_rate": 1.925148657621232e-05, + "loss": 1.1972, + "step": 6247 + }, + { + "epoch": 0.15, + "learning_rate": 1.9251190875937024e-05, + "loss": 1.3503, + "step": 6248 + }, + { + "epoch": 0.15, + "learning_rate": 1.9250895119536924e-05, + "loss": 1.0771, + "step": 6249 + }, + { + "epoch": 0.15, + "learning_rate": 1.925059930701382e-05, + "loss": 1.1091, + "step": 6250 + }, + { + "epoch": 0.15, + "learning_rate": 1.9250303438369507e-05, + "loss": 1.1974, + "step": 6251 + }, + { + "epoch": 0.15, + "learning_rate": 1.9250007513605776e-05, + "loss": 1.3465, + "step": 6252 + }, + { + "epoch": 0.15, + "learning_rate": 1.9249711532724428e-05, + "loss": 1.0733, + "step": 6253 + }, + { + "epoch": 0.15, + "learning_rate": 1.9249415495727252e-05, + "loss": 1.1498, + "step": 6254 + }, + { + "epoch": 0.15, + "learning_rate": 1.9249119402616055e-05, + "loss": 1.1095, + "step": 6255 + }, + { + "epoch": 0.15, + "learning_rate": 1.924882325339262e-05, + "loss": 1.1492, + "step": 6256 + }, + { + "epoch": 0.15, + "learning_rate": 1.9248527048058753e-05, + "loss": 1.0495, + "step": 6257 + }, + { + "epoch": 0.15, + "learning_rate": 1.9248230786616244e-05, + "loss": 1.3174, + "step": 6258 + }, + { + "epoch": 0.15, + "learning_rate": 1.92479344690669e-05, + "loss": 1.2208, + "step": 6259 + }, + { + "epoch": 0.15, + "learning_rate": 1.9247638095412508e-05, + "loss": 1.0506, + "step": 6260 + }, + { + "epoch": 0.15, + "learning_rate": 1.9247341665654872e-05, + "loss": 0.938, + "step": 6261 + }, + { + "epoch": 0.15, + "learning_rate": 1.9247045179795788e-05, + "loss": 1.1823, + "step": 6262 + }, + { + "epoch": 0.15, + "learning_rate": 1.9246748637837057e-05, + "loss": 1.2412, + "step": 6263 + }, + { + "epoch": 0.15, + "learning_rate": 1.924645203978048e-05, + "loss": 1.0378, + "step": 6264 + }, + { + "epoch": 0.15, + "learning_rate": 1.924615538562785e-05, + "loss": 1.3362, + "step": 6265 + }, + { + "epoch": 0.15, + "learning_rate": 1.924585867538097e-05, + "loss": 1.234, + "step": 6266 + }, + { + "epoch": 0.15, + "learning_rate": 1.924556190904164e-05, + "loss": 1.0851, + "step": 6267 + }, + { + "epoch": 0.15, + "learning_rate": 1.924526508661166e-05, + "loss": 0.8029, + "step": 6268 + }, + { + "epoch": 0.15, + "learning_rate": 1.9244968208092833e-05, + "loss": 1.1255, + "step": 6269 + }, + { + "epoch": 0.15, + "learning_rate": 1.9244671273486962e-05, + "loss": 0.9916, + "step": 6270 + }, + { + "epoch": 0.15, + "learning_rate": 1.924437428279584e-05, + "loss": 1.1804, + "step": 6271 + }, + { + "epoch": 0.15, + "learning_rate": 1.9244077236021273e-05, + "loss": 1.2796, + "step": 6272 + }, + { + "epoch": 0.15, + "learning_rate": 1.9243780133165067e-05, + "loss": 1.2834, + "step": 6273 + }, + { + "epoch": 0.15, + "learning_rate": 1.924348297422902e-05, + "loss": 1.2521, + "step": 6274 + }, + { + "epoch": 0.15, + "learning_rate": 1.9243185759214938e-05, + "loss": 1.0562, + "step": 6275 + }, + { + "epoch": 0.15, + "learning_rate": 1.924288848812462e-05, + "loss": 1.2594, + "step": 6276 + }, + { + "epoch": 0.15, + "learning_rate": 1.9242591160959868e-05, + "loss": 1.3037, + "step": 6277 + }, + { + "epoch": 0.15, + "learning_rate": 1.9242293777722496e-05, + "loss": 0.9487, + "step": 6278 + }, + { + "epoch": 0.15, + "learning_rate": 1.9241996338414296e-05, + "loss": 0.9335, + "step": 6279 + }, + { + "epoch": 0.15, + "learning_rate": 1.9241698843037083e-05, + "loss": 1.1248, + "step": 6280 + }, + { + "epoch": 0.15, + "learning_rate": 1.9241401291592652e-05, + "loss": 1.1468, + "step": 6281 + }, + { + "epoch": 0.15, + "learning_rate": 1.9241103684082815e-05, + "loss": 1.0924, + "step": 6282 + }, + { + "epoch": 0.15, + "learning_rate": 1.924080602050938e-05, + "loss": 1.1411, + "step": 6283 + }, + { + "epoch": 0.15, + "learning_rate": 1.9240508300874145e-05, + "loss": 1.0309, + "step": 6284 + }, + { + "epoch": 0.15, + "learning_rate": 1.9240210525178916e-05, + "loss": 1.1652, + "step": 6285 + }, + { + "epoch": 0.15, + "learning_rate": 1.9239912693425506e-05, + "loss": 1.0422, + "step": 6286 + }, + { + "epoch": 0.15, + "learning_rate": 1.923961480561572e-05, + "loss": 1.2625, + "step": 6287 + }, + { + "epoch": 0.15, + "learning_rate": 1.9239316861751365e-05, + "loss": 1.3966, + "step": 6288 + }, + { + "epoch": 0.15, + "learning_rate": 1.9239018861834244e-05, + "loss": 0.9857, + "step": 6289 + }, + { + "epoch": 0.15, + "learning_rate": 1.9238720805866174e-05, + "loss": 1.1554, + "step": 6290 + }, + { + "epoch": 0.15, + "learning_rate": 1.9238422693848955e-05, + "loss": 1.097, + "step": 6291 + }, + { + "epoch": 0.15, + "learning_rate": 1.92381245257844e-05, + "loss": 1.3734, + "step": 6292 + }, + { + "epoch": 0.15, + "learning_rate": 1.9237826301674313e-05, + "loss": 1.1369, + "step": 6293 + }, + { + "epoch": 0.15, + "learning_rate": 1.9237528021520512e-05, + "loss": 1.1058, + "step": 6294 + }, + { + "epoch": 0.15, + "learning_rate": 1.9237229685324795e-05, + "loss": 1.3583, + "step": 6295 + }, + { + "epoch": 0.15, + "learning_rate": 1.9236931293088982e-05, + "loss": 1.2495, + "step": 6296 + }, + { + "epoch": 0.15, + "learning_rate": 1.923663284481488e-05, + "loss": 1.1067, + "step": 6297 + }, + { + "epoch": 0.15, + "learning_rate": 1.9236334340504298e-05, + "loss": 1.1255, + "step": 6298 + }, + { + "epoch": 0.15, + "learning_rate": 1.9236035780159045e-05, + "loss": 1.1364, + "step": 6299 + }, + { + "epoch": 0.15, + "learning_rate": 1.923573716378094e-05, + "loss": 1.3099, + "step": 6300 + }, + { + "epoch": 0.15, + "learning_rate": 1.923543849137179e-05, + "loss": 0.9607, + "step": 6301 + }, + { + "epoch": 0.15, + "learning_rate": 1.9235139762933402e-05, + "loss": 1.2269, + "step": 6302 + }, + { + "epoch": 0.15, + "learning_rate": 1.92348409784676e-05, + "loss": 1.1784, + "step": 6303 + }, + { + "epoch": 0.15, + "learning_rate": 1.9234542137976184e-05, + "loss": 1.1264, + "step": 6304 + }, + { + "epoch": 0.15, + "learning_rate": 1.9234243241460976e-05, + "loss": 0.9627, + "step": 6305 + }, + { + "epoch": 0.15, + "learning_rate": 1.9233944288923788e-05, + "loss": 1.231, + "step": 6306 + }, + { + "epoch": 0.15, + "learning_rate": 1.923364528036643e-05, + "loss": 0.933, + "step": 6307 + }, + { + "epoch": 0.15, + "learning_rate": 1.9233346215790717e-05, + "loss": 1.1681, + "step": 6308 + }, + { + "epoch": 0.15, + "learning_rate": 1.9233047095198466e-05, + "loss": 1.1139, + "step": 6309 + }, + { + "epoch": 0.15, + "learning_rate": 1.9232747918591488e-05, + "loss": 0.9688, + "step": 6310 + }, + { + "epoch": 0.15, + "learning_rate": 1.92324486859716e-05, + "loss": 1.2468, + "step": 6311 + }, + { + "epoch": 0.15, + "learning_rate": 1.923214939734062e-05, + "loss": 1.2157, + "step": 6312 + }, + { + "epoch": 0.15, + "learning_rate": 1.9231850052700358e-05, + "loss": 1.1574, + "step": 6313 + }, + { + "epoch": 0.15, + "learning_rate": 1.9231550652052635e-05, + "loss": 1.1226, + "step": 6314 + }, + { + "epoch": 0.15, + "learning_rate": 1.923125119539927e-05, + "loss": 1.1273, + "step": 6315 + }, + { + "epoch": 0.15, + "learning_rate": 1.9230951682742066e-05, + "loss": 1.1797, + "step": 6316 + }, + { + "epoch": 0.15, + "learning_rate": 1.9230652114082853e-05, + "loss": 1.3521, + "step": 6317 + }, + { + "epoch": 0.15, + "learning_rate": 1.923035248942345e-05, + "loss": 1.169, + "step": 6318 + }, + { + "epoch": 0.15, + "learning_rate": 1.9230052808765665e-05, + "loss": 1.2167, + "step": 6319 + }, + { + "epoch": 0.15, + "learning_rate": 1.9229753072111325e-05, + "loss": 1.1083, + "step": 6320 + }, + { + "epoch": 0.15, + "learning_rate": 1.9229453279462235e-05, + "loss": 1.3547, + "step": 6321 + }, + { + "epoch": 0.15, + "learning_rate": 1.9229153430820232e-05, + "loss": 1.1075, + "step": 6322 + }, + { + "epoch": 0.15, + "learning_rate": 1.9228853526187122e-05, + "loss": 0.9528, + "step": 6323 + }, + { + "epoch": 0.15, + "learning_rate": 1.9228553565564728e-05, + "loss": 1.254, + "step": 6324 + }, + { + "epoch": 0.15, + "learning_rate": 1.922825354895487e-05, + "loss": 1.2323, + "step": 6325 + }, + { + "epoch": 0.15, + "learning_rate": 1.922795347635937e-05, + "loss": 1.1203, + "step": 6326 + }, + { + "epoch": 0.15, + "learning_rate": 1.9227653347780046e-05, + "loss": 0.9265, + "step": 6327 + }, + { + "epoch": 0.15, + "learning_rate": 1.922735316321872e-05, + "loss": 1.2522, + "step": 6328 + }, + { + "epoch": 0.15, + "learning_rate": 1.922705292267721e-05, + "loss": 1.1742, + "step": 6329 + }, + { + "epoch": 0.15, + "learning_rate": 1.9226752626157345e-05, + "loss": 1.3789, + "step": 6330 + }, + { + "epoch": 0.15, + "learning_rate": 1.9226452273660938e-05, + "loss": 1.0268, + "step": 6331 + }, + { + "epoch": 0.15, + "learning_rate": 1.922615186518982e-05, + "loss": 0.9648, + "step": 6332 + }, + { + "epoch": 0.15, + "learning_rate": 1.9225851400745803e-05, + "loss": 1.133, + "step": 6333 + }, + { + "epoch": 0.15, + "learning_rate": 1.9225550880330718e-05, + "loss": 1.2609, + "step": 6334 + }, + { + "epoch": 0.15, + "learning_rate": 1.922525030394639e-05, + "loss": 1.1434, + "step": 6335 + }, + { + "epoch": 0.15, + "learning_rate": 1.9224949671594633e-05, + "loss": 1.2126, + "step": 6336 + }, + { + "epoch": 0.15, + "learning_rate": 1.9224648983277275e-05, + "loss": 1.0583, + "step": 6337 + }, + { + "epoch": 0.15, + "learning_rate": 1.9224348238996146e-05, + "loss": 0.9182, + "step": 6338 + }, + { + "epoch": 0.15, + "learning_rate": 1.922404743875306e-05, + "loss": 1.0518, + "step": 6339 + }, + { + "epoch": 0.15, + "learning_rate": 1.9223746582549853e-05, + "loss": 1.0561, + "step": 6340 + }, + { + "epoch": 0.15, + "learning_rate": 1.9223445670388345e-05, + "loss": 0.9805, + "step": 6341 + }, + { + "epoch": 0.15, + "learning_rate": 1.922314470227036e-05, + "loss": 1.2756, + "step": 6342 + }, + { + "epoch": 0.15, + "learning_rate": 1.9222843678197728e-05, + "loss": 1.1124, + "step": 6343 + }, + { + "epoch": 0.15, + "learning_rate": 1.9222542598172268e-05, + "loss": 1.0904, + "step": 6344 + }, + { + "epoch": 0.15, + "learning_rate": 1.9222241462195816e-05, + "loss": 1.0212, + "step": 6345 + }, + { + "epoch": 0.15, + "learning_rate": 1.922194027027019e-05, + "loss": 1.2565, + "step": 6346 + }, + { + "epoch": 0.15, + "learning_rate": 1.9221639022397224e-05, + "loss": 0.9767, + "step": 6347 + }, + { + "epoch": 0.15, + "learning_rate": 1.9221337718578744e-05, + "loss": 0.9295, + "step": 6348 + }, + { + "epoch": 0.15, + "learning_rate": 1.9221036358816576e-05, + "loss": 1.0101, + "step": 6349 + }, + { + "epoch": 0.15, + "learning_rate": 1.922073494311255e-05, + "loss": 1.0684, + "step": 6350 + }, + { + "epoch": 0.15, + "learning_rate": 1.9220433471468493e-05, + "loss": 1.0839, + "step": 6351 + }, + { + "epoch": 0.15, + "learning_rate": 1.9220131943886232e-05, + "loss": 1.1895, + "step": 6352 + }, + { + "epoch": 0.15, + "learning_rate": 1.9219830360367603e-05, + "loss": 0.9258, + "step": 6353 + }, + { + "epoch": 0.15, + "learning_rate": 1.9219528720914432e-05, + "loss": 1.0239, + "step": 6354 + }, + { + "epoch": 0.15, + "learning_rate": 1.9219227025528548e-05, + "loss": 1.3385, + "step": 6355 + }, + { + "epoch": 0.15, + "learning_rate": 1.921892527421178e-05, + "loss": 1.0677, + "step": 6356 + }, + { + "epoch": 0.15, + "learning_rate": 1.9218623466965966e-05, + "loss": 1.2605, + "step": 6357 + }, + { + "epoch": 0.15, + "learning_rate": 1.9218321603792928e-05, + "loss": 1.1533, + "step": 6358 + }, + { + "epoch": 0.15, + "learning_rate": 1.9218019684694502e-05, + "loss": 1.0601, + "step": 6359 + }, + { + "epoch": 0.15, + "learning_rate": 1.921771770967252e-05, + "loss": 1.1694, + "step": 6360 + }, + { + "epoch": 0.15, + "learning_rate": 1.921741567872881e-05, + "loss": 1.1232, + "step": 6361 + }, + { + "epoch": 0.15, + "learning_rate": 1.921711359186521e-05, + "loss": 1.038, + "step": 6362 + }, + { + "epoch": 0.15, + "learning_rate": 1.921681144908355e-05, + "loss": 1.2811, + "step": 6363 + }, + { + "epoch": 0.15, + "learning_rate": 1.921650925038566e-05, + "loss": 0.981, + "step": 6364 + }, + { + "epoch": 0.15, + "learning_rate": 1.9216206995773373e-05, + "loss": 1.2402, + "step": 6365 + }, + { + "epoch": 0.15, + "learning_rate": 1.9215904685248534e-05, + "loss": 1.0089, + "step": 6366 + }, + { + "epoch": 0.15, + "learning_rate": 1.9215602318812964e-05, + "loss": 0.9949, + "step": 6367 + }, + { + "epoch": 0.15, + "learning_rate": 1.9215299896468503e-05, + "loss": 1.1127, + "step": 6368 + }, + { + "epoch": 0.15, + "learning_rate": 1.9214997418216988e-05, + "loss": 1.0835, + "step": 6369 + }, + { + "epoch": 0.15, + "learning_rate": 1.9214694884060248e-05, + "loss": 1.2194, + "step": 6370 + }, + { + "epoch": 0.15, + "learning_rate": 1.9214392294000123e-05, + "loss": 1.2061, + "step": 6371 + }, + { + "epoch": 0.15, + "learning_rate": 1.9214089648038446e-05, + "loss": 1.207, + "step": 6372 + }, + { + "epoch": 0.15, + "learning_rate": 1.9213786946177057e-05, + "loss": 1.2777, + "step": 6373 + }, + { + "epoch": 0.15, + "learning_rate": 1.9213484188417788e-05, + "loss": 1.2359, + "step": 6374 + }, + { + "epoch": 0.15, + "learning_rate": 1.9213181374762476e-05, + "loss": 1.2416, + "step": 6375 + }, + { + "epoch": 0.15, + "learning_rate": 1.921287850521296e-05, + "loss": 1.0452, + "step": 6376 + }, + { + "epoch": 0.15, + "learning_rate": 1.921257557977108e-05, + "loss": 1.151, + "step": 6377 + }, + { + "epoch": 0.15, + "learning_rate": 1.921227259843867e-05, + "loss": 1.3945, + "step": 6378 + }, + { + "epoch": 0.15, + "learning_rate": 1.9211969561217566e-05, + "loss": 1.1218, + "step": 6379 + }, + { + "epoch": 0.15, + "learning_rate": 1.9211666468109612e-05, + "loss": 1.3716, + "step": 6380 + }, + { + "epoch": 0.15, + "learning_rate": 1.9211363319116647e-05, + "loss": 1.152, + "step": 6381 + }, + { + "epoch": 0.15, + "learning_rate": 1.9211060114240503e-05, + "loss": 1.107, + "step": 6382 + }, + { + "epoch": 0.15, + "learning_rate": 1.9210756853483027e-05, + "loss": 1.1319, + "step": 6383 + }, + { + "epoch": 0.15, + "learning_rate": 1.9210453536846053e-05, + "loss": 1.0164, + "step": 6384 + }, + { + "epoch": 0.15, + "learning_rate": 1.9210150164331423e-05, + "loss": 1.1207, + "step": 6385 + }, + { + "epoch": 0.15, + "learning_rate": 1.920984673594098e-05, + "loss": 1.0746, + "step": 6386 + }, + { + "epoch": 0.15, + "learning_rate": 1.9209543251676566e-05, + "loss": 1.0134, + "step": 6387 + }, + { + "epoch": 0.15, + "learning_rate": 1.9209239711540014e-05, + "loss": 1.0234, + "step": 6388 + }, + { + "epoch": 0.15, + "learning_rate": 1.9208936115533173e-05, + "loss": 1.1255, + "step": 6389 + }, + { + "epoch": 0.15, + "learning_rate": 1.9208632463657885e-05, + "loss": 1.0041, + "step": 6390 + }, + { + "epoch": 0.15, + "learning_rate": 1.9208328755915987e-05, + "loss": 1.188, + "step": 6391 + }, + { + "epoch": 0.15, + "learning_rate": 1.9208024992309325e-05, + "loss": 1.085, + "step": 6392 + }, + { + "epoch": 0.15, + "learning_rate": 1.9207721172839743e-05, + "loss": 1.0036, + "step": 6393 + }, + { + "epoch": 0.15, + "learning_rate": 1.920741729750908e-05, + "loss": 0.9756, + "step": 6394 + }, + { + "epoch": 0.15, + "learning_rate": 1.9207113366319185e-05, + "loss": 1.1709, + "step": 6395 + }, + { + "epoch": 0.15, + "learning_rate": 1.9206809379271892e-05, + "loss": 1.3076, + "step": 6396 + }, + { + "epoch": 0.15, + "learning_rate": 1.920650533636906e-05, + "loss": 1.199, + "step": 6397 + }, + { + "epoch": 0.15, + "learning_rate": 1.920620123761252e-05, + "loss": 1.3691, + "step": 6398 + }, + { + "epoch": 0.15, + "learning_rate": 1.9205897083004127e-05, + "loss": 1.3195, + "step": 6399 + }, + { + "epoch": 0.15, + "learning_rate": 1.920559287254572e-05, + "loss": 1.0901, + "step": 6400 + }, + { + "epoch": 0.15, + "learning_rate": 1.9205288606239144e-05, + "loss": 1.2808, + "step": 6401 + }, + { + "epoch": 0.15, + "learning_rate": 1.920498428408625e-05, + "loss": 1.2857, + "step": 6402 + }, + { + "epoch": 0.15, + "learning_rate": 1.9204679906088882e-05, + "loss": 1.1337, + "step": 6403 + }, + { + "epoch": 0.15, + "learning_rate": 1.9204375472248885e-05, + "loss": 1.2484, + "step": 6404 + }, + { + "epoch": 0.15, + "learning_rate": 1.9204070982568105e-05, + "loss": 1.1935, + "step": 6405 + }, + { + "epoch": 0.15, + "learning_rate": 1.9203766437048395e-05, + "loss": 1.1096, + "step": 6406 + }, + { + "epoch": 0.15, + "learning_rate": 1.9203461835691596e-05, + "loss": 1.082, + "step": 6407 + }, + { + "epoch": 0.15, + "learning_rate": 1.920315717849956e-05, + "loss": 1.1654, + "step": 6408 + }, + { + "epoch": 0.15, + "learning_rate": 1.9202852465474133e-05, + "loss": 1.1876, + "step": 6409 + }, + { + "epoch": 0.15, + "learning_rate": 1.9202547696617165e-05, + "loss": 1.2938, + "step": 6410 + }, + { + "epoch": 0.15, + "learning_rate": 1.9202242871930504e-05, + "loss": 1.0789, + "step": 6411 + }, + { + "epoch": 0.15, + "learning_rate": 1.9201937991416003e-05, + "loss": 1.2857, + "step": 6412 + }, + { + "epoch": 0.15, + "learning_rate": 1.9201633055075508e-05, + "loss": 1.1864, + "step": 6413 + }, + { + "epoch": 0.15, + "learning_rate": 1.920132806291087e-05, + "loss": 1.441, + "step": 6414 + }, + { + "epoch": 0.15, + "learning_rate": 1.9201023014923935e-05, + "loss": 1.117, + "step": 6415 + }, + { + "epoch": 0.15, + "learning_rate": 1.9200717911116564e-05, + "loss": 1.0859, + "step": 6416 + }, + { + "epoch": 0.15, + "learning_rate": 1.9200412751490595e-05, + "loss": 1.2382, + "step": 6417 + }, + { + "epoch": 0.15, + "learning_rate": 1.920010753604789e-05, + "loss": 1.0849, + "step": 6418 + }, + { + "epoch": 0.15, + "learning_rate": 1.9199802264790293e-05, + "loss": 1.0837, + "step": 6419 + }, + { + "epoch": 0.15, + "learning_rate": 1.9199496937719663e-05, + "loss": 1.0916, + "step": 6420 + }, + { + "epoch": 0.15, + "learning_rate": 1.919919155483785e-05, + "loss": 1.1459, + "step": 6421 + }, + { + "epoch": 0.15, + "learning_rate": 1.91988861161467e-05, + "loss": 0.9929, + "step": 6422 + }, + { + "epoch": 0.15, + "learning_rate": 1.9198580621648076e-05, + "loss": 1.1989, + "step": 6423 + }, + { + "epoch": 0.15, + "learning_rate": 1.9198275071343827e-05, + "loss": 1.1108, + "step": 6424 + }, + { + "epoch": 0.15, + "learning_rate": 1.9197969465235804e-05, + "loss": 0.8475, + "step": 6425 + }, + { + "epoch": 0.15, + "learning_rate": 1.9197663803325867e-05, + "loss": 1.147, + "step": 6426 + }, + { + "epoch": 0.15, + "learning_rate": 1.9197358085615866e-05, + "loss": 1.0467, + "step": 6427 + }, + { + "epoch": 0.15, + "learning_rate": 1.9197052312107655e-05, + "loss": 1.1788, + "step": 6428 + }, + { + "epoch": 0.15, + "learning_rate": 1.9196746482803092e-05, + "loss": 1.1762, + "step": 6429 + }, + { + "epoch": 0.15, + "learning_rate": 1.9196440597704033e-05, + "loss": 1.2397, + "step": 6430 + }, + { + "epoch": 0.15, + "learning_rate": 1.919613465681233e-05, + "loss": 1.0958, + "step": 6431 + }, + { + "epoch": 0.15, + "learning_rate": 1.9195828660129842e-05, + "loss": 1.0238, + "step": 6432 + }, + { + "epoch": 0.15, + "learning_rate": 1.9195522607658424e-05, + "loss": 1.1349, + "step": 6433 + }, + { + "epoch": 0.15, + "learning_rate": 1.9195216499399932e-05, + "loss": 1.1284, + "step": 6434 + }, + { + "epoch": 0.15, + "learning_rate": 1.9194910335356226e-05, + "loss": 1.1425, + "step": 6435 + }, + { + "epoch": 0.15, + "learning_rate": 1.9194604115529163e-05, + "loss": 1.1302, + "step": 6436 + }, + { + "epoch": 0.15, + "learning_rate": 1.91942978399206e-05, + "loss": 1.45, + "step": 6437 + }, + { + "epoch": 0.15, + "learning_rate": 1.919399150853239e-05, + "loss": 1.0979, + "step": 6438 + }, + { + "epoch": 0.15, + "learning_rate": 1.91936851213664e-05, + "loss": 1.0468, + "step": 6439 + }, + { + "epoch": 0.15, + "learning_rate": 1.9193378678424484e-05, + "loss": 1.1169, + "step": 6440 + }, + { + "epoch": 0.15, + "learning_rate": 1.91930721797085e-05, + "loss": 1.0813, + "step": 6441 + }, + { + "epoch": 0.15, + "learning_rate": 1.9192765625220312e-05, + "loss": 1.0153, + "step": 6442 + }, + { + "epoch": 0.15, + "learning_rate": 1.9192459014961774e-05, + "loss": 1.1423, + "step": 6443 + }, + { + "epoch": 0.15, + "learning_rate": 1.9192152348934753e-05, + "loss": 1.229, + "step": 6444 + }, + { + "epoch": 0.15, + "learning_rate": 1.9191845627141105e-05, + "loss": 1.0862, + "step": 6445 + }, + { + "epoch": 0.16, + "learning_rate": 1.919153884958269e-05, + "loss": 1.0901, + "step": 6446 + }, + { + "epoch": 0.16, + "learning_rate": 1.9191232016261373e-05, + "loss": 1.117, + "step": 6447 + }, + { + "epoch": 0.16, + "learning_rate": 1.9190925127179013e-05, + "loss": 1.0964, + "step": 6448 + }, + { + "epoch": 0.16, + "learning_rate": 1.9190618182337474e-05, + "loss": 1.2372, + "step": 6449 + }, + { + "epoch": 0.16, + "learning_rate": 1.919031118173861e-05, + "loss": 0.9812, + "step": 6450 + }, + { + "epoch": 0.16, + "learning_rate": 1.9190004125384293e-05, + "loss": 1.0542, + "step": 6451 + }, + { + "epoch": 0.16, + "learning_rate": 1.9189697013276386e-05, + "loss": 1.0379, + "step": 6452 + }, + { + "epoch": 0.16, + "learning_rate": 1.9189389845416744e-05, + "loss": 1.3485, + "step": 6453 + }, + { + "epoch": 0.16, + "learning_rate": 1.9189082621807235e-05, + "loss": 0.9629, + "step": 6454 + }, + { + "epoch": 0.16, + "learning_rate": 1.9188775342449725e-05, + "loss": 0.9939, + "step": 6455 + }, + { + "epoch": 0.16, + "learning_rate": 1.918846800734608e-05, + "loss": 1.3178, + "step": 6456 + }, + { + "epoch": 0.16, + "learning_rate": 1.9188160616498156e-05, + "loss": 1.4683, + "step": 6457 + }, + { + "epoch": 0.16, + "learning_rate": 1.9187853169907824e-05, + "loss": 1.0734, + "step": 6458 + }, + { + "epoch": 0.16, + "learning_rate": 1.9187545667576947e-05, + "loss": 1.2994, + "step": 6459 + }, + { + "epoch": 0.16, + "learning_rate": 1.9187238109507393e-05, + "loss": 1.1602, + "step": 6460 + }, + { + "epoch": 0.16, + "learning_rate": 1.9186930495701028e-05, + "loss": 1.1942, + "step": 6461 + }, + { + "epoch": 0.16, + "learning_rate": 1.918662282615971e-05, + "loss": 0.9932, + "step": 6462 + }, + { + "epoch": 0.16, + "learning_rate": 1.918631510088532e-05, + "loss": 0.9685, + "step": 6463 + }, + { + "epoch": 0.16, + "learning_rate": 1.9186007319879714e-05, + "loss": 1.2083, + "step": 6464 + }, + { + "epoch": 0.16, + "learning_rate": 1.918569948314476e-05, + "loss": 1.1845, + "step": 6465 + }, + { + "epoch": 0.16, + "learning_rate": 1.918539159068233e-05, + "loss": 1.0858, + "step": 6466 + }, + { + "epoch": 0.16, + "learning_rate": 1.9185083642494288e-05, + "loss": 0.8968, + "step": 6467 + }, + { + "epoch": 0.16, + "learning_rate": 1.9184775638582508e-05, + "loss": 1.1855, + "step": 6468 + }, + { + "epoch": 0.16, + "learning_rate": 1.918446757894885e-05, + "loss": 1.2978, + "step": 6469 + }, + { + "epoch": 0.16, + "learning_rate": 1.918415946359519e-05, + "loss": 1.2857, + "step": 6470 + }, + { + "epoch": 0.16, + "learning_rate": 1.9183851292523395e-05, + "loss": 1.202, + "step": 6471 + }, + { + "epoch": 0.16, + "learning_rate": 1.918354306573533e-05, + "loss": 1.0171, + "step": 6472 + }, + { + "epoch": 0.16, + "learning_rate": 1.9183234783232877e-05, + "loss": 1.0593, + "step": 6473 + }, + { + "epoch": 0.16, + "learning_rate": 1.9182926445017893e-05, + "loss": 1.1235, + "step": 6474 + }, + { + "epoch": 0.16, + "learning_rate": 1.9182618051092256e-05, + "loss": 1.0394, + "step": 6475 + }, + { + "epoch": 0.16, + "learning_rate": 1.9182309601457837e-05, + "loss": 1.1502, + "step": 6476 + }, + { + "epoch": 0.16, + "learning_rate": 1.91820010961165e-05, + "loss": 0.9986, + "step": 6477 + }, + { + "epoch": 0.16, + "learning_rate": 1.9181692535070128e-05, + "loss": 1.1762, + "step": 6478 + }, + { + "epoch": 0.16, + "learning_rate": 1.9181383918320584e-05, + "loss": 1.0762, + "step": 6479 + }, + { + "epoch": 0.16, + "learning_rate": 1.9181075245869744e-05, + "loss": 0.918, + "step": 6480 + }, + { + "epoch": 0.16, + "learning_rate": 1.918076651771948e-05, + "loss": 1.2218, + "step": 6481 + }, + { + "epoch": 0.16, + "learning_rate": 1.9180457733871666e-05, + "loss": 1.3095, + "step": 6482 + }, + { + "epoch": 0.16, + "learning_rate": 1.918014889432817e-05, + "loss": 1.2535, + "step": 6483 + }, + { + "epoch": 0.16, + "learning_rate": 1.9179839999090874e-05, + "loss": 0.9782, + "step": 6484 + }, + { + "epoch": 0.16, + "learning_rate": 1.917953104816165e-05, + "loss": 1.2448, + "step": 6485 + }, + { + "epoch": 0.16, + "learning_rate": 1.9179222041542366e-05, + "loss": 1.0558, + "step": 6486 + }, + { + "epoch": 0.16, + "learning_rate": 1.9178912979234902e-05, + "loss": 1.0782, + "step": 6487 + }, + { + "epoch": 0.16, + "learning_rate": 1.9178603861241133e-05, + "loss": 1.3055, + "step": 6488 + }, + { + "epoch": 0.16, + "learning_rate": 1.9178294687562934e-05, + "loss": 1.0469, + "step": 6489 + }, + { + "epoch": 0.16, + "learning_rate": 1.917798545820218e-05, + "loss": 1.2238, + "step": 6490 + }, + { + "epoch": 0.16, + "learning_rate": 1.9177676173160743e-05, + "loss": 1.2204, + "step": 6491 + }, + { + "epoch": 0.16, + "learning_rate": 1.9177366832440505e-05, + "loss": 1.0174, + "step": 6492 + }, + { + "epoch": 0.16, + "learning_rate": 1.917705743604334e-05, + "loss": 1.2184, + "step": 6493 + }, + { + "epoch": 0.16, + "learning_rate": 1.917674798397113e-05, + "loss": 1.0903, + "step": 6494 + }, + { + "epoch": 0.16, + "learning_rate": 1.9176438476225747e-05, + "loss": 1.0064, + "step": 6495 + }, + { + "epoch": 0.16, + "learning_rate": 1.917612891280907e-05, + "loss": 1.4643, + "step": 6496 + }, + { + "epoch": 0.16, + "learning_rate": 1.9175819293722978e-05, + "loss": 1.2443, + "step": 6497 + }, + { + "epoch": 0.16, + "learning_rate": 1.9175509618969347e-05, + "loss": 1.2514, + "step": 6498 + }, + { + "epoch": 0.16, + "learning_rate": 1.917519988855006e-05, + "loss": 1.3707, + "step": 6499 + }, + { + "epoch": 0.16, + "learning_rate": 1.917489010246699e-05, + "loss": 0.9303, + "step": 6500 + }, + { + "epoch": 0.16, + "learning_rate": 1.9174580260722023e-05, + "loss": 1.1765, + "step": 6501 + }, + { + "epoch": 0.16, + "learning_rate": 1.917427036331703e-05, + "loss": 1.1131, + "step": 6502 + }, + { + "epoch": 0.16, + "learning_rate": 1.9173960410253903e-05, + "loss": 1.1514, + "step": 6503 + }, + { + "epoch": 0.16, + "learning_rate": 1.9173650401534514e-05, + "loss": 1.1771, + "step": 6504 + }, + { + "epoch": 0.16, + "learning_rate": 1.9173340337160744e-05, + "loss": 0.6784, + "step": 6505 + }, + { + "epoch": 0.16, + "learning_rate": 1.917303021713448e-05, + "loss": 1.167, + "step": 6506 + }, + { + "epoch": 0.16, + "learning_rate": 1.9172720041457596e-05, + "loss": 1.2523, + "step": 6507 + }, + { + "epoch": 0.16, + "learning_rate": 1.9172409810131975e-05, + "loss": 1.0295, + "step": 6508 + }, + { + "epoch": 0.16, + "learning_rate": 1.9172099523159508e-05, + "loss": 1.3117, + "step": 6509 + }, + { + "epoch": 0.16, + "learning_rate": 1.9171789180542066e-05, + "loss": 1.2331, + "step": 6510 + }, + { + "epoch": 0.16, + "learning_rate": 1.9171478782281535e-05, + "loss": 1.1878, + "step": 6511 + }, + { + "epoch": 0.16, + "learning_rate": 1.9171168328379803e-05, + "loss": 1.2064, + "step": 6512 + }, + { + "epoch": 0.16, + "learning_rate": 1.9170857818838746e-05, + "loss": 1.1367, + "step": 6513 + }, + { + "epoch": 0.16, + "learning_rate": 1.9170547253660253e-05, + "loss": 1.0369, + "step": 6514 + }, + { + "epoch": 0.16, + "learning_rate": 1.917023663284621e-05, + "loss": 0.8623, + "step": 6515 + }, + { + "epoch": 0.16, + "learning_rate": 1.9169925956398497e-05, + "loss": 1.1062, + "step": 6516 + }, + { + "epoch": 0.16, + "learning_rate": 1.9169615224318998e-05, + "loss": 1.0378, + "step": 6517 + }, + { + "epoch": 0.16, + "learning_rate": 1.91693044366096e-05, + "loss": 1.3457, + "step": 6518 + }, + { + "epoch": 0.16, + "learning_rate": 1.9168993593272193e-05, + "loss": 1.2817, + "step": 6519 + }, + { + "epoch": 0.16, + "learning_rate": 1.9168682694308654e-05, + "loss": 0.9923, + "step": 6520 + }, + { + "epoch": 0.16, + "learning_rate": 1.9168371739720878e-05, + "loss": 1.2169, + "step": 6521 + }, + { + "epoch": 0.16, + "learning_rate": 1.9168060729510742e-05, + "loss": 1.0576, + "step": 6522 + }, + { + "epoch": 0.16, + "learning_rate": 1.916774966368014e-05, + "loss": 1.1331, + "step": 6523 + }, + { + "epoch": 0.16, + "learning_rate": 1.916743854223096e-05, + "loss": 1.0702, + "step": 6524 + }, + { + "epoch": 0.16, + "learning_rate": 1.9167127365165084e-05, + "loss": 1.1441, + "step": 6525 + }, + { + "epoch": 0.16, + "learning_rate": 1.9166816132484404e-05, + "loss": 1.1919, + "step": 6526 + }, + { + "epoch": 0.16, + "learning_rate": 1.9166504844190804e-05, + "loss": 1.1638, + "step": 6527 + }, + { + "epoch": 0.16, + "learning_rate": 1.9166193500286177e-05, + "loss": 1.2821, + "step": 6528 + }, + { + "epoch": 0.16, + "learning_rate": 1.916588210077241e-05, + "loss": 1.1917, + "step": 6529 + }, + { + "epoch": 0.16, + "learning_rate": 1.9165570645651392e-05, + "loss": 1.2864, + "step": 6530 + }, + { + "epoch": 0.16, + "learning_rate": 1.9165259134925015e-05, + "loss": 0.9739, + "step": 6531 + }, + { + "epoch": 0.16, + "learning_rate": 1.9164947568595164e-05, + "loss": 1.1647, + "step": 6532 + }, + { + "epoch": 0.16, + "learning_rate": 1.9164635946663732e-05, + "loss": 1.2826, + "step": 6533 + }, + { + "epoch": 0.16, + "learning_rate": 1.9164324269132608e-05, + "loss": 0.8987, + "step": 6534 + }, + { + "epoch": 0.16, + "learning_rate": 1.916401253600369e-05, + "loss": 1.0381, + "step": 6535 + }, + { + "epoch": 0.16, + "learning_rate": 1.9163700747278857e-05, + "loss": 1.0941, + "step": 6536 + }, + { + "epoch": 0.16, + "learning_rate": 1.916338890296001e-05, + "loss": 0.8977, + "step": 6537 + }, + { + "epoch": 0.16, + "learning_rate": 1.9163077003049037e-05, + "loss": 1.343, + "step": 6538 + }, + { + "epoch": 0.16, + "learning_rate": 1.9162765047547832e-05, + "loss": 0.873, + "step": 6539 + }, + { + "epoch": 0.16, + "learning_rate": 1.9162453036458287e-05, + "loss": 1.0781, + "step": 6540 + }, + { + "epoch": 0.16, + "learning_rate": 1.9162140969782292e-05, + "loss": 1.0747, + "step": 6541 + }, + { + "epoch": 0.16, + "learning_rate": 1.9161828847521743e-05, + "loss": 1.0507, + "step": 6542 + }, + { + "epoch": 0.16, + "learning_rate": 1.9161516669678536e-05, + "loss": 1.13, + "step": 6543 + }, + { + "epoch": 0.16, + "learning_rate": 1.9161204436254565e-05, + "loss": 1.3603, + "step": 6544 + }, + { + "epoch": 0.16, + "learning_rate": 1.9160892147251714e-05, + "loss": 1.1316, + "step": 6545 + }, + { + "epoch": 0.16, + "learning_rate": 1.916057980267189e-05, + "loss": 1.1401, + "step": 6546 + }, + { + "epoch": 0.16, + "learning_rate": 1.9160267402516982e-05, + "loss": 1.079, + "step": 6547 + }, + { + "epoch": 0.16, + "learning_rate": 1.915995494678889e-05, + "loss": 1.2372, + "step": 6548 + }, + { + "epoch": 0.16, + "learning_rate": 1.9159642435489498e-05, + "loss": 1.266, + "step": 6549 + }, + { + "epoch": 0.16, + "learning_rate": 1.9159329868620714e-05, + "loss": 1.2695, + "step": 6550 + }, + { + "epoch": 0.16, + "learning_rate": 1.915901724618443e-05, + "loss": 0.8897, + "step": 6551 + }, + { + "epoch": 0.16, + "learning_rate": 1.9158704568182543e-05, + "loss": 0.8113, + "step": 6552 + }, + { + "epoch": 0.16, + "learning_rate": 1.915839183461695e-05, + "loss": 1.1653, + "step": 6553 + }, + { + "epoch": 0.16, + "learning_rate": 1.9158079045489547e-05, + "loss": 1.1463, + "step": 6554 + }, + { + "epoch": 0.16, + "learning_rate": 1.9157766200802233e-05, + "loss": 1.1352, + "step": 6555 + }, + { + "epoch": 0.16, + "learning_rate": 1.9157453300556904e-05, + "loss": 1.3088, + "step": 6556 + }, + { + "epoch": 0.16, + "learning_rate": 1.9157140344755463e-05, + "loss": 0.9501, + "step": 6557 + }, + { + "epoch": 0.16, + "learning_rate": 1.9156827333399805e-05, + "loss": 1.0286, + "step": 6558 + }, + { + "epoch": 0.16, + "learning_rate": 1.9156514266491826e-05, + "loss": 0.9698, + "step": 6559 + }, + { + "epoch": 0.16, + "learning_rate": 1.915620114403343e-05, + "loss": 0.9666, + "step": 6560 + }, + { + "epoch": 0.16, + "learning_rate": 1.9155887966026516e-05, + "loss": 1.1789, + "step": 6561 + }, + { + "epoch": 0.16, + "learning_rate": 1.9155574732472983e-05, + "loss": 1.2624, + "step": 6562 + }, + { + "epoch": 0.16, + "learning_rate": 1.915526144337473e-05, + "loss": 1.0507, + "step": 6563 + }, + { + "epoch": 0.16, + "learning_rate": 1.9154948098733663e-05, + "loss": 1.2196, + "step": 6564 + }, + { + "epoch": 0.16, + "learning_rate": 1.9154634698551677e-05, + "loss": 1.0951, + "step": 6565 + }, + { + "epoch": 0.16, + "learning_rate": 1.9154321242830676e-05, + "loss": 1.0114, + "step": 6566 + }, + { + "epoch": 0.16, + "learning_rate": 1.915400773157256e-05, + "loss": 1.2919, + "step": 6567 + }, + { + "epoch": 0.16, + "learning_rate": 1.9153694164779234e-05, + "loss": 0.9366, + "step": 6568 + }, + { + "epoch": 0.16, + "learning_rate": 1.91533805424526e-05, + "loss": 1.054, + "step": 6569 + }, + { + "epoch": 0.16, + "learning_rate": 1.9153066864594558e-05, + "loss": 1.028, + "step": 6570 + }, + { + "epoch": 0.16, + "learning_rate": 1.9152753131207008e-05, + "loss": 1.065, + "step": 6571 + }, + { + "epoch": 0.16, + "learning_rate": 1.9152439342291865e-05, + "loss": 1.1801, + "step": 6572 + }, + { + "epoch": 0.16, + "learning_rate": 1.915212549785102e-05, + "loss": 1.0097, + "step": 6573 + }, + { + "epoch": 0.16, + "learning_rate": 1.9151811597886383e-05, + "loss": 1.2162, + "step": 6574 + }, + { + "epoch": 0.16, + "learning_rate": 1.915149764239986e-05, + "loss": 1.1649, + "step": 6575 + }, + { + "epoch": 0.16, + "learning_rate": 1.9151183631393352e-05, + "loss": 1.1565, + "step": 6576 + }, + { + "epoch": 0.16, + "learning_rate": 1.9150869564868764e-05, + "loss": 1.2173, + "step": 6577 + }, + { + "epoch": 0.16, + "learning_rate": 1.9150555442828004e-05, + "loss": 1.2714, + "step": 6578 + }, + { + "epoch": 0.16, + "learning_rate": 1.9150241265272976e-05, + "loss": 1.3674, + "step": 6579 + }, + { + "epoch": 0.16, + "learning_rate": 1.914992703220559e-05, + "loss": 1.1296, + "step": 6580 + }, + { + "epoch": 0.16, + "learning_rate": 1.9149612743627744e-05, + "loss": 1.2441, + "step": 6581 + }, + { + "epoch": 0.16, + "learning_rate": 1.9149298399541353e-05, + "loss": 1.15, + "step": 6582 + }, + { + "epoch": 0.16, + "learning_rate": 1.914898399994832e-05, + "loss": 1.1279, + "step": 6583 + }, + { + "epoch": 0.16, + "learning_rate": 1.9148669544850552e-05, + "loss": 1.1854, + "step": 6584 + }, + { + "epoch": 0.16, + "learning_rate": 1.9148355034249958e-05, + "loss": 1.2377, + "step": 6585 + }, + { + "epoch": 0.16, + "learning_rate": 1.9148040468148442e-05, + "loss": 1.4427, + "step": 6586 + }, + { + "epoch": 0.16, + "learning_rate": 1.914772584654792e-05, + "loss": 1.299, + "step": 6587 + }, + { + "epoch": 0.16, + "learning_rate": 1.9147411169450302e-05, + "loss": 1.1083, + "step": 6588 + }, + { + "epoch": 0.16, + "learning_rate": 1.9147096436857484e-05, + "loss": 0.8501, + "step": 6589 + }, + { + "epoch": 0.16, + "learning_rate": 1.9146781648771387e-05, + "loss": 0.9595, + "step": 6590 + }, + { + "epoch": 0.16, + "learning_rate": 1.9146466805193915e-05, + "loss": 1.2175, + "step": 6591 + }, + { + "epoch": 0.16, + "learning_rate": 1.9146151906126983e-05, + "loss": 1.3278, + "step": 6592 + }, + { + "epoch": 0.16, + "learning_rate": 1.9145836951572497e-05, + "loss": 1.1061, + "step": 6593 + }, + { + "epoch": 0.16, + "learning_rate": 1.9145521941532374e-05, + "loss": 0.8815, + "step": 6594 + }, + { + "epoch": 0.16, + "learning_rate": 1.9145206876008515e-05, + "loss": 1.3626, + "step": 6595 + }, + { + "epoch": 0.16, + "learning_rate": 1.9144891755002837e-05, + "loss": 0.9735, + "step": 6596 + }, + { + "epoch": 0.16, + "learning_rate": 1.9144576578517253e-05, + "loss": 1.2348, + "step": 6597 + }, + { + "epoch": 0.16, + "learning_rate": 1.9144261346553677e-05, + "loss": 1.3902, + "step": 6598 + }, + { + "epoch": 0.16, + "learning_rate": 1.9143946059114013e-05, + "loss": 0.98, + "step": 6599 + }, + { + "epoch": 0.16, + "learning_rate": 1.9143630716200184e-05, + "loss": 1.3568, + "step": 6600 + }, + { + "epoch": 0.16, + "learning_rate": 1.914331531781409e-05, + "loss": 1.261, + "step": 6601 + }, + { + "epoch": 0.16, + "learning_rate": 1.9142999863957662e-05, + "loss": 1.1439, + "step": 6602 + }, + { + "epoch": 0.16, + "learning_rate": 1.9142684354632802e-05, + "loss": 1.224, + "step": 6603 + }, + { + "epoch": 0.16, + "learning_rate": 1.9142368789841422e-05, + "loss": 1.2416, + "step": 6604 + }, + { + "epoch": 0.16, + "learning_rate": 1.9142053169585445e-05, + "loss": 1.0782, + "step": 6605 + }, + { + "epoch": 0.16, + "learning_rate": 1.914173749386678e-05, + "loss": 1.179, + "step": 6606 + }, + { + "epoch": 0.16, + "learning_rate": 1.9141421762687345e-05, + "loss": 1.1953, + "step": 6607 + }, + { + "epoch": 0.16, + "learning_rate": 1.9141105976049054e-05, + "loss": 1.0932, + "step": 6608 + }, + { + "epoch": 0.16, + "learning_rate": 1.9140790133953823e-05, + "loss": 1.059, + "step": 6609 + }, + { + "epoch": 0.16, + "learning_rate": 1.914047423640357e-05, + "loss": 0.9982, + "step": 6610 + }, + { + "epoch": 0.16, + "learning_rate": 1.9140158283400208e-05, + "loss": 1.2029, + "step": 6611 + }, + { + "epoch": 0.16, + "learning_rate": 1.9139842274945655e-05, + "loss": 1.1538, + "step": 6612 + }, + { + "epoch": 0.16, + "learning_rate": 1.913952621104183e-05, + "loss": 1.3101, + "step": 6613 + }, + { + "epoch": 0.16, + "learning_rate": 1.913921009169065e-05, + "loss": 1.0411, + "step": 6614 + }, + { + "epoch": 0.16, + "learning_rate": 1.913889391689403e-05, + "loss": 1.0723, + "step": 6615 + }, + { + "epoch": 0.16, + "learning_rate": 1.9138577686653893e-05, + "loss": 1.2598, + "step": 6616 + }, + { + "epoch": 0.16, + "learning_rate": 1.9138261400972152e-05, + "loss": 0.7621, + "step": 6617 + }, + { + "epoch": 0.16, + "learning_rate": 1.913794505985073e-05, + "loss": 1.0263, + "step": 6618 + }, + { + "epoch": 0.16, + "learning_rate": 1.9137628663291543e-05, + "loss": 1.0449, + "step": 6619 + }, + { + "epoch": 0.16, + "learning_rate": 1.9137312211296516e-05, + "loss": 1.0501, + "step": 6620 + }, + { + "epoch": 0.16, + "learning_rate": 1.913699570386756e-05, + "loss": 0.8667, + "step": 6621 + }, + { + "epoch": 0.16, + "learning_rate": 1.9136679141006603e-05, + "loss": 1.3949, + "step": 6622 + }, + { + "epoch": 0.16, + "learning_rate": 1.9136362522715563e-05, + "loss": 0.9646, + "step": 6623 + }, + { + "epoch": 0.16, + "learning_rate": 1.9136045848996357e-05, + "loss": 1.0832, + "step": 6624 + }, + { + "epoch": 0.16, + "learning_rate": 1.9135729119850913e-05, + "loss": 1.3173, + "step": 6625 + }, + { + "epoch": 0.16, + "learning_rate": 1.9135412335281152e-05, + "loss": 0.9367, + "step": 6626 + }, + { + "epoch": 0.16, + "learning_rate": 1.913509549528899e-05, + "loss": 0.9644, + "step": 6627 + }, + { + "epoch": 0.16, + "learning_rate": 1.913477859987635e-05, + "loss": 1.3505, + "step": 6628 + }, + { + "epoch": 0.16, + "learning_rate": 1.913446164904516e-05, + "loss": 1.3799, + "step": 6629 + }, + { + "epoch": 0.16, + "learning_rate": 1.913414464279734e-05, + "loss": 1.253, + "step": 6630 + }, + { + "epoch": 0.16, + "learning_rate": 1.913382758113481e-05, + "loss": 1.2037, + "step": 6631 + }, + { + "epoch": 0.16, + "learning_rate": 1.91335104640595e-05, + "loss": 1.0877, + "step": 6632 + }, + { + "epoch": 0.16, + "learning_rate": 1.9133193291573327e-05, + "loss": 0.874, + "step": 6633 + }, + { + "epoch": 0.16, + "learning_rate": 1.913287606367822e-05, + "loss": 0.8261, + "step": 6634 + }, + { + "epoch": 0.16, + "learning_rate": 1.91325587803761e-05, + "loss": 1.1354, + "step": 6635 + }, + { + "epoch": 0.16, + "learning_rate": 1.9132241441668903e-05, + "loss": 1.2277, + "step": 6636 + }, + { + "epoch": 0.16, + "learning_rate": 1.9131924047558536e-05, + "loss": 1.2152, + "step": 6637 + }, + { + "epoch": 0.16, + "learning_rate": 1.9131606598046936e-05, + "loss": 1.1957, + "step": 6638 + }, + { + "epoch": 0.16, + "learning_rate": 1.9131289093136027e-05, + "loss": 1.1771, + "step": 6639 + }, + { + "epoch": 0.16, + "learning_rate": 1.9130971532827737e-05, + "loss": 1.1487, + "step": 6640 + }, + { + "epoch": 0.16, + "learning_rate": 1.9130653917123986e-05, + "loss": 1.252, + "step": 6641 + }, + { + "epoch": 0.16, + "learning_rate": 1.9130336246026707e-05, + "loss": 1.0981, + "step": 6642 + }, + { + "epoch": 0.16, + "learning_rate": 1.9130018519537827e-05, + "loss": 1.2182, + "step": 6643 + }, + { + "epoch": 0.16, + "learning_rate": 1.9129700737659273e-05, + "loss": 1.0307, + "step": 6644 + }, + { + "epoch": 0.16, + "learning_rate": 1.912938290039297e-05, + "loss": 0.9065, + "step": 6645 + }, + { + "epoch": 0.16, + "learning_rate": 1.9129065007740848e-05, + "loss": 1.0708, + "step": 6646 + }, + { + "epoch": 0.16, + "learning_rate": 1.912874705970484e-05, + "loss": 1.1221, + "step": 6647 + }, + { + "epoch": 0.16, + "learning_rate": 1.9128429056286865e-05, + "loss": 1.0377, + "step": 6648 + }, + { + "epoch": 0.16, + "learning_rate": 1.912811099748886e-05, + "loss": 1.2275, + "step": 6649 + }, + { + "epoch": 0.16, + "learning_rate": 1.9127792883312756e-05, + "loss": 1.2064, + "step": 6650 + }, + { + "epoch": 0.16, + "learning_rate": 1.9127474713760473e-05, + "loss": 1.2987, + "step": 6651 + }, + { + "epoch": 0.16, + "learning_rate": 1.912715648883395e-05, + "loss": 1.3688, + "step": 6652 + }, + { + "epoch": 0.16, + "learning_rate": 1.912683820853512e-05, + "loss": 1.1044, + "step": 6653 + }, + { + "epoch": 0.16, + "learning_rate": 1.9126519872865906e-05, + "loss": 1.2339, + "step": 6654 + }, + { + "epoch": 0.16, + "learning_rate": 1.9126201481828244e-05, + "loss": 1.2427, + "step": 6655 + }, + { + "epoch": 0.16, + "learning_rate": 1.9125883035424062e-05, + "loss": 1.1993, + "step": 6656 + }, + { + "epoch": 0.16, + "learning_rate": 1.9125564533655297e-05, + "loss": 1.2252, + "step": 6657 + }, + { + "epoch": 0.16, + "learning_rate": 1.9125245976523876e-05, + "loss": 1.358, + "step": 6658 + }, + { + "epoch": 0.16, + "learning_rate": 1.9124927364031736e-05, + "loss": 1.2627, + "step": 6659 + }, + { + "epoch": 0.16, + "learning_rate": 1.9124608696180806e-05, + "loss": 1.3671, + "step": 6660 + }, + { + "epoch": 0.16, + "learning_rate": 1.912428997297302e-05, + "loss": 1.1922, + "step": 6661 + }, + { + "epoch": 0.16, + "learning_rate": 1.9123971194410317e-05, + "loss": 1.0812, + "step": 6662 + }, + { + "epoch": 0.16, + "learning_rate": 1.9123652360494624e-05, + "loss": 1.1929, + "step": 6663 + }, + { + "epoch": 0.16, + "learning_rate": 1.912333347122788e-05, + "loss": 1.2204, + "step": 6664 + }, + { + "epoch": 0.16, + "learning_rate": 1.9123014526612016e-05, + "loss": 1.1887, + "step": 6665 + }, + { + "epoch": 0.16, + "learning_rate": 1.9122695526648968e-05, + "loss": 0.8818, + "step": 6666 + }, + { + "epoch": 0.16, + "learning_rate": 1.9122376471340675e-05, + "loss": 1.2142, + "step": 6667 + }, + { + "epoch": 0.16, + "learning_rate": 1.912205736068907e-05, + "loss": 1.274, + "step": 6668 + }, + { + "epoch": 0.16, + "learning_rate": 1.9121738194696084e-05, + "loss": 0.8012, + "step": 6669 + }, + { + "epoch": 0.16, + "learning_rate": 1.912141897336366e-05, + "loss": 1.3834, + "step": 6670 + }, + { + "epoch": 0.16, + "learning_rate": 1.9121099696693733e-05, + "loss": 0.941, + "step": 6671 + }, + { + "epoch": 0.16, + "learning_rate": 1.9120780364688243e-05, + "loss": 1.2544, + "step": 6672 + }, + { + "epoch": 0.16, + "learning_rate": 1.912046097734912e-05, + "loss": 1.3484, + "step": 6673 + }, + { + "epoch": 0.16, + "learning_rate": 1.912014153467831e-05, + "loss": 1.2908, + "step": 6674 + }, + { + "epoch": 0.16, + "learning_rate": 1.911982203667774e-05, + "loss": 1.068, + "step": 6675 + }, + { + "epoch": 0.16, + "learning_rate": 1.9119502483349357e-05, + "loss": 1.2645, + "step": 6676 + }, + { + "epoch": 0.16, + "learning_rate": 1.9119182874695097e-05, + "loss": 1.3102, + "step": 6677 + }, + { + "epoch": 0.16, + "learning_rate": 1.91188632107169e-05, + "loss": 0.8104, + "step": 6678 + }, + { + "epoch": 0.16, + "learning_rate": 1.911854349141671e-05, + "loss": 1.14, + "step": 6679 + }, + { + "epoch": 0.16, + "learning_rate": 1.9118223716796453e-05, + "loss": 0.8094, + "step": 6680 + }, + { + "epoch": 0.16, + "learning_rate": 1.911790388685808e-05, + "loss": 1.3634, + "step": 6681 + }, + { + "epoch": 0.16, + "learning_rate": 1.9117584001603533e-05, + "loss": 1.4068, + "step": 6682 + }, + { + "epoch": 0.16, + "learning_rate": 1.9117264061034743e-05, + "loss": 1.1875, + "step": 6683 + }, + { + "epoch": 0.16, + "learning_rate": 1.911694406515366e-05, + "loss": 1.2675, + "step": 6684 + }, + { + "epoch": 0.16, + "learning_rate": 1.9116624013962218e-05, + "loss": 1.2642, + "step": 6685 + }, + { + "epoch": 0.16, + "learning_rate": 1.9116303907462365e-05, + "loss": 1.3054, + "step": 6686 + }, + { + "epoch": 0.16, + "learning_rate": 1.911598374565604e-05, + "loss": 1.0819, + "step": 6687 + }, + { + "epoch": 0.16, + "learning_rate": 1.911566352854519e-05, + "loss": 1.1154, + "step": 6688 + }, + { + "epoch": 0.16, + "learning_rate": 1.911534325613175e-05, + "loss": 1.1951, + "step": 6689 + }, + { + "epoch": 0.16, + "learning_rate": 1.9115022928417664e-05, + "loss": 1.1149, + "step": 6690 + }, + { + "epoch": 0.16, + "learning_rate": 1.911470254540488e-05, + "loss": 1.0767, + "step": 6691 + }, + { + "epoch": 0.16, + "learning_rate": 1.911438210709534e-05, + "loss": 1.2358, + "step": 6692 + }, + { + "epoch": 0.16, + "learning_rate": 1.9114061613490992e-05, + "loss": 1.107, + "step": 6693 + }, + { + "epoch": 0.16, + "learning_rate": 1.911374106459377e-05, + "loss": 0.9792, + "step": 6694 + }, + { + "epoch": 0.16, + "learning_rate": 1.9113420460405626e-05, + "loss": 1.352, + "step": 6695 + }, + { + "epoch": 0.16, + "learning_rate": 1.9113099800928502e-05, + "loss": 1.1096, + "step": 6696 + }, + { + "epoch": 0.16, + "learning_rate": 1.911277908616435e-05, + "loss": 0.8162, + "step": 6697 + }, + { + "epoch": 0.16, + "learning_rate": 1.9112458316115107e-05, + "loss": 1.1863, + "step": 6698 + }, + { + "epoch": 0.16, + "learning_rate": 1.9112137490782722e-05, + "loss": 1.3748, + "step": 6699 + }, + { + "epoch": 0.16, + "learning_rate": 1.911181661016914e-05, + "loss": 1.1531, + "step": 6700 + }, + { + "epoch": 0.16, + "learning_rate": 1.9111495674276317e-05, + "loss": 1.1162, + "step": 6701 + }, + { + "epoch": 0.16, + "learning_rate": 1.911117468310619e-05, + "loss": 0.9479, + "step": 6702 + }, + { + "epoch": 0.16, + "learning_rate": 1.9110853636660705e-05, + "loss": 0.9971, + "step": 6703 + }, + { + "epoch": 0.16, + "learning_rate": 1.9110532534941822e-05, + "loss": 1.0563, + "step": 6704 + }, + { + "epoch": 0.16, + "learning_rate": 1.9110211377951477e-05, + "loss": 1.2333, + "step": 6705 + }, + { + "epoch": 0.16, + "learning_rate": 1.910989016569162e-05, + "loss": 1.0857, + "step": 6706 + }, + { + "epoch": 0.16, + "learning_rate": 1.9109568898164203e-05, + "loss": 1.0204, + "step": 6707 + }, + { + "epoch": 0.16, + "learning_rate": 1.9109247575371177e-05, + "loss": 1.1636, + "step": 6708 + }, + { + "epoch": 0.16, + "learning_rate": 1.9108926197314486e-05, + "loss": 1.1111, + "step": 6709 + }, + { + "epoch": 0.16, + "learning_rate": 1.9108604763996084e-05, + "loss": 1.4598, + "step": 6710 + }, + { + "epoch": 0.16, + "learning_rate": 1.9108283275417918e-05, + "loss": 1.489, + "step": 6711 + }, + { + "epoch": 0.16, + "learning_rate": 1.910796173158194e-05, + "loss": 1.3755, + "step": 6712 + }, + { + "epoch": 0.16, + "learning_rate": 1.91076401324901e-05, + "loss": 1.328, + "step": 6713 + }, + { + "epoch": 0.16, + "learning_rate": 1.910731847814435e-05, + "loss": 1.1372, + "step": 6714 + }, + { + "epoch": 0.16, + "learning_rate": 1.910699676854664e-05, + "loss": 1.1087, + "step": 6715 + }, + { + "epoch": 0.16, + "learning_rate": 1.9106675003698928e-05, + "loss": 1.3109, + "step": 6716 + }, + { + "epoch": 0.16, + "learning_rate": 1.910635318360316e-05, + "loss": 1.2451, + "step": 6717 + }, + { + "epoch": 0.16, + "learning_rate": 1.9106031308261284e-05, + "loss": 1.1095, + "step": 6718 + }, + { + "epoch": 0.16, + "learning_rate": 1.9105709377675258e-05, + "loss": 1.0767, + "step": 6719 + }, + { + "epoch": 0.16, + "learning_rate": 1.9105387391847036e-05, + "loss": 1.035, + "step": 6720 + }, + { + "epoch": 0.16, + "learning_rate": 1.910506535077857e-05, + "loss": 1.0699, + "step": 6721 + }, + { + "epoch": 0.16, + "learning_rate": 1.910474325447182e-05, + "loss": 1.1247, + "step": 6722 + }, + { + "epoch": 0.16, + "learning_rate": 1.9104421102928727e-05, + "loss": 1.2497, + "step": 6723 + }, + { + "epoch": 0.16, + "learning_rate": 1.9104098896151256e-05, + "loss": 1.293, + "step": 6724 + }, + { + "epoch": 0.16, + "learning_rate": 1.910377663414136e-05, + "loss": 0.9764, + "step": 6725 + }, + { + "epoch": 0.16, + "learning_rate": 1.9103454316900987e-05, + "loss": 1.2886, + "step": 6726 + }, + { + "epoch": 0.16, + "learning_rate": 1.9103131944432102e-05, + "loss": 0.9039, + "step": 6727 + }, + { + "epoch": 0.16, + "learning_rate": 1.9102809516736655e-05, + "loss": 1.2521, + "step": 6728 + }, + { + "epoch": 0.16, + "learning_rate": 1.9102487033816604e-05, + "loss": 1.0952, + "step": 6729 + }, + { + "epoch": 0.16, + "learning_rate": 1.9102164495673906e-05, + "loss": 1.1271, + "step": 6730 + }, + { + "epoch": 0.16, + "learning_rate": 1.910184190231052e-05, + "loss": 1.1748, + "step": 6731 + }, + { + "epoch": 0.16, + "learning_rate": 1.9101519253728396e-05, + "loss": 1.0913, + "step": 6732 + }, + { + "epoch": 0.16, + "learning_rate": 1.91011965499295e-05, + "loss": 1.0106, + "step": 6733 + }, + { + "epoch": 0.16, + "learning_rate": 1.9100873790915776e-05, + "loss": 1.1884, + "step": 6734 + }, + { + "epoch": 0.16, + "learning_rate": 1.91005509766892e-05, + "loss": 0.945, + "step": 6735 + }, + { + "epoch": 0.16, + "learning_rate": 1.9100228107251715e-05, + "loss": 1.0657, + "step": 6736 + }, + { + "epoch": 0.16, + "learning_rate": 1.9099905182605294e-05, + "loss": 1.0902, + "step": 6737 + }, + { + "epoch": 0.16, + "learning_rate": 1.9099582202751885e-05, + "loss": 1.1624, + "step": 6738 + }, + { + "epoch": 0.16, + "learning_rate": 1.909925916769345e-05, + "loss": 0.8556, + "step": 6739 + }, + { + "epoch": 0.16, + "learning_rate": 1.9098936077431953e-05, + "loss": 1.1784, + "step": 6740 + }, + { + "epoch": 0.16, + "learning_rate": 1.9098612931969345e-05, + "loss": 1.2203, + "step": 6741 + }, + { + "epoch": 0.16, + "learning_rate": 1.90982897313076e-05, + "loss": 0.9454, + "step": 6742 + }, + { + "epoch": 0.16, + "learning_rate": 1.909796647544867e-05, + "loss": 1.0947, + "step": 6743 + }, + { + "epoch": 0.16, + "learning_rate": 1.9097643164394512e-05, + "loss": 1.1204, + "step": 6744 + }, + { + "epoch": 0.16, + "learning_rate": 1.9097319798147098e-05, + "loss": 0.9417, + "step": 6745 + }, + { + "epoch": 0.16, + "learning_rate": 1.9096996376708382e-05, + "loss": 1.0203, + "step": 6746 + }, + { + "epoch": 0.16, + "learning_rate": 1.909667290008033e-05, + "loss": 1.2212, + "step": 6747 + }, + { + "epoch": 0.16, + "learning_rate": 1.9096349368264904e-05, + "loss": 1.0424, + "step": 6748 + }, + { + "epoch": 0.16, + "learning_rate": 1.9096025781264064e-05, + "loss": 0.8541, + "step": 6749 + }, + { + "epoch": 0.16, + "learning_rate": 1.9095702139079774e-05, + "loss": 1.0966, + "step": 6750 + }, + { + "epoch": 0.16, + "learning_rate": 1.9095378441714e-05, + "loss": 0.9071, + "step": 6751 + }, + { + "epoch": 0.16, + "learning_rate": 1.9095054689168707e-05, + "loss": 1.2277, + "step": 6752 + }, + { + "epoch": 0.16, + "learning_rate": 1.9094730881445854e-05, + "loss": 1.0929, + "step": 6753 + }, + { + "epoch": 0.16, + "learning_rate": 1.9094407018547406e-05, + "loss": 1.0036, + "step": 6754 + }, + { + "epoch": 0.16, + "learning_rate": 1.909408310047533e-05, + "loss": 1.2614, + "step": 6755 + }, + { + "epoch": 0.16, + "learning_rate": 1.9093759127231594e-05, + "loss": 1.0275, + "step": 6756 + }, + { + "epoch": 0.16, + "learning_rate": 1.9093435098818162e-05, + "loss": 1.0375, + "step": 6757 + }, + { + "epoch": 0.16, + "learning_rate": 1.9093111015236993e-05, + "loss": 1.0742, + "step": 6758 + }, + { + "epoch": 0.16, + "learning_rate": 1.909278687649006e-05, + "loss": 1.0681, + "step": 6759 + }, + { + "epoch": 0.16, + "learning_rate": 1.909246268257933e-05, + "loss": 1.2701, + "step": 6760 + }, + { + "epoch": 0.16, + "learning_rate": 1.9092138433506763e-05, + "loss": 0.9476, + "step": 6761 + }, + { + "epoch": 0.16, + "learning_rate": 1.9091814129274335e-05, + "loss": 1.0874, + "step": 6762 + }, + { + "epoch": 0.16, + "learning_rate": 1.909148976988401e-05, + "loss": 1.1684, + "step": 6763 + }, + { + "epoch": 0.16, + "learning_rate": 1.9091165355337754e-05, + "loss": 0.9441, + "step": 6764 + }, + { + "epoch": 0.16, + "learning_rate": 1.9090840885637534e-05, + "loss": 1.1712, + "step": 6765 + }, + { + "epoch": 0.16, + "learning_rate": 1.9090516360785322e-05, + "loss": 1.3836, + "step": 6766 + }, + { + "epoch": 0.16, + "learning_rate": 1.9090191780783083e-05, + "loss": 1.1465, + "step": 6767 + }, + { + "epoch": 0.16, + "learning_rate": 1.908986714563279e-05, + "loss": 1.1716, + "step": 6768 + }, + { + "epoch": 0.16, + "learning_rate": 1.908954245533641e-05, + "loss": 1.0416, + "step": 6769 + }, + { + "epoch": 0.16, + "learning_rate": 1.9089217709895918e-05, + "loss": 1.0586, + "step": 6770 + }, + { + "epoch": 0.16, + "learning_rate": 1.908889290931328e-05, + "loss": 1.2281, + "step": 6771 + }, + { + "epoch": 0.16, + "learning_rate": 1.908856805359046e-05, + "loss": 1.021, + "step": 6772 + }, + { + "epoch": 0.16, + "learning_rate": 1.908824314272944e-05, + "loss": 0.9994, + "step": 6773 + }, + { + "epoch": 0.16, + "learning_rate": 1.9087918176732188e-05, + "loss": 1.1088, + "step": 6774 + }, + { + "epoch": 0.16, + "learning_rate": 1.908759315560067e-05, + "loss": 1.1349, + "step": 6775 + }, + { + "epoch": 0.16, + "learning_rate": 1.9087268079336865e-05, + "loss": 0.9504, + "step": 6776 + }, + { + "epoch": 0.16, + "learning_rate": 1.908694294794274e-05, + "loss": 1.2334, + "step": 6777 + }, + { + "epoch": 0.16, + "learning_rate": 1.9086617761420272e-05, + "loss": 1.211, + "step": 6778 + }, + { + "epoch": 0.16, + "learning_rate": 1.9086292519771433e-05, + "loss": 1.1645, + "step": 6779 + }, + { + "epoch": 0.16, + "learning_rate": 1.908596722299819e-05, + "loss": 1.4515, + "step": 6780 + }, + { + "epoch": 0.16, + "learning_rate": 1.908564187110252e-05, + "loss": 1.1857, + "step": 6781 + }, + { + "epoch": 0.16, + "learning_rate": 1.9085316464086403e-05, + "loss": 0.9452, + "step": 6782 + }, + { + "epoch": 0.16, + "learning_rate": 1.9084991001951804e-05, + "loss": 1.2917, + "step": 6783 + }, + { + "epoch": 0.16, + "learning_rate": 1.9084665484700704e-05, + "loss": 0.9672, + "step": 6784 + }, + { + "epoch": 0.16, + "learning_rate": 1.9084339912335074e-05, + "loss": 1.2566, + "step": 6785 + }, + { + "epoch": 0.16, + "learning_rate": 1.908401428485689e-05, + "loss": 1.1025, + "step": 6786 + }, + { + "epoch": 0.16, + "learning_rate": 1.908368860226813e-05, + "loss": 1.2441, + "step": 6787 + }, + { + "epoch": 0.16, + "learning_rate": 1.9083362864570768e-05, + "loss": 1.0743, + "step": 6788 + }, + { + "epoch": 0.16, + "learning_rate": 1.9083037071766775e-05, + "loss": 1.396, + "step": 6789 + }, + { + "epoch": 0.16, + "learning_rate": 1.9082711223858136e-05, + "loss": 1.2888, + "step": 6790 + }, + { + "epoch": 0.16, + "learning_rate": 1.9082385320846825e-05, + "loss": 1.2442, + "step": 6791 + }, + { + "epoch": 0.16, + "learning_rate": 1.9082059362734822e-05, + "loss": 1.0578, + "step": 6792 + }, + { + "epoch": 0.16, + "learning_rate": 1.9081733349524098e-05, + "loss": 1.108, + "step": 6793 + }, + { + "epoch": 0.16, + "learning_rate": 1.908140728121663e-05, + "loss": 1.1568, + "step": 6794 + }, + { + "epoch": 0.16, + "learning_rate": 1.9081081157814406e-05, + "loss": 0.8811, + "step": 6795 + }, + { + "epoch": 0.16, + "learning_rate": 1.9080754979319395e-05, + "loss": 1.3006, + "step": 6796 + }, + { + "epoch": 0.16, + "learning_rate": 1.9080428745733578e-05, + "loss": 1.1765, + "step": 6797 + }, + { + "epoch": 0.16, + "learning_rate": 1.908010245705894e-05, + "loss": 1.1826, + "step": 6798 + }, + { + "epoch": 0.16, + "learning_rate": 1.9079776113297454e-05, + "loss": 1.3808, + "step": 6799 + }, + { + "epoch": 0.16, + "learning_rate": 1.90794497144511e-05, + "loss": 1.1131, + "step": 6800 + }, + { + "epoch": 0.16, + "learning_rate": 1.9079123260521864e-05, + "loss": 1.2545, + "step": 6801 + }, + { + "epoch": 0.16, + "learning_rate": 1.907879675151172e-05, + "loss": 0.9239, + "step": 6802 + }, + { + "epoch": 0.16, + "learning_rate": 1.9078470187422657e-05, + "loss": 1.0056, + "step": 6803 + }, + { + "epoch": 0.16, + "learning_rate": 1.9078143568256644e-05, + "loss": 1.1326, + "step": 6804 + }, + { + "epoch": 0.16, + "learning_rate": 1.9077816894015672e-05, + "loss": 1.1665, + "step": 6805 + }, + { + "epoch": 0.16, + "learning_rate": 1.907749016470172e-05, + "loss": 1.4038, + "step": 6806 + }, + { + "epoch": 0.16, + "learning_rate": 1.9077163380316766e-05, + "loss": 1.4028, + "step": 6807 + }, + { + "epoch": 0.16, + "learning_rate": 1.9076836540862804e-05, + "loss": 1.1778, + "step": 6808 + }, + { + "epoch": 0.16, + "learning_rate": 1.9076509646341806e-05, + "loss": 1.1456, + "step": 6809 + }, + { + "epoch": 0.16, + "learning_rate": 1.907618269675576e-05, + "loss": 0.9157, + "step": 6810 + }, + { + "epoch": 0.16, + "learning_rate": 1.907585569210665e-05, + "loss": 1.0256, + "step": 6811 + }, + { + "epoch": 0.16, + "learning_rate": 1.9075528632396456e-05, + "loss": 1.027, + "step": 6812 + }, + { + "epoch": 0.16, + "learning_rate": 1.9075201517627163e-05, + "loss": 1.2442, + "step": 6813 + }, + { + "epoch": 0.16, + "learning_rate": 1.907487434780076e-05, + "loss": 1.1877, + "step": 6814 + }, + { + "epoch": 0.16, + "learning_rate": 1.9074547122919226e-05, + "loss": 0.8014, + "step": 6815 + }, + { + "epoch": 0.16, + "learning_rate": 1.9074219842984553e-05, + "loss": 1.2361, + "step": 6816 + }, + { + "epoch": 0.16, + "learning_rate": 1.907389250799872e-05, + "loss": 1.1236, + "step": 6817 + }, + { + "epoch": 0.16, + "learning_rate": 1.9073565117963714e-05, + "loss": 1.3866, + "step": 6818 + }, + { + "epoch": 0.16, + "learning_rate": 1.9073237672881527e-05, + "loss": 1.1225, + "step": 6819 + }, + { + "epoch": 0.16, + "learning_rate": 1.9072910172754137e-05, + "loss": 1.3251, + "step": 6820 + }, + { + "epoch": 0.16, + "learning_rate": 1.9072582617583538e-05, + "loss": 1.2012, + "step": 6821 + }, + { + "epoch": 0.16, + "learning_rate": 1.9072255007371715e-05, + "loss": 1.124, + "step": 6822 + }, + { + "epoch": 0.16, + "learning_rate": 1.907192734212065e-05, + "loss": 0.9797, + "step": 6823 + }, + { + "epoch": 0.16, + "learning_rate": 1.9071599621832338e-05, + "loss": 1.2392, + "step": 6824 + }, + { + "epoch": 0.16, + "learning_rate": 1.9071271846508765e-05, + "loss": 1.0622, + "step": 6825 + }, + { + "epoch": 0.16, + "learning_rate": 1.9070944016151923e-05, + "loss": 1.1795, + "step": 6826 + }, + { + "epoch": 0.16, + "learning_rate": 1.907061613076379e-05, + "loss": 1.0379, + "step": 6827 + }, + { + "epoch": 0.16, + "learning_rate": 1.907028819034637e-05, + "loss": 1.3303, + "step": 6828 + }, + { + "epoch": 0.16, + "learning_rate": 1.906996019490164e-05, + "loss": 1.0095, + "step": 6829 + }, + { + "epoch": 0.16, + "learning_rate": 1.9069632144431595e-05, + "loss": 1.1486, + "step": 6830 + }, + { + "epoch": 0.16, + "learning_rate": 1.9069304038938228e-05, + "loss": 1.1942, + "step": 6831 + }, + { + "epoch": 0.16, + "learning_rate": 1.9068975878423526e-05, + "loss": 1.3356, + "step": 6832 + }, + { + "epoch": 0.16, + "learning_rate": 1.906864766288948e-05, + "loss": 1.1615, + "step": 6833 + }, + { + "epoch": 0.16, + "learning_rate": 1.9068319392338082e-05, + "loss": 0.9063, + "step": 6834 + }, + { + "epoch": 0.16, + "learning_rate": 1.9067991066771322e-05, + "loss": 1.155, + "step": 6835 + }, + { + "epoch": 0.16, + "learning_rate": 1.9067662686191192e-05, + "loss": 1.1627, + "step": 6836 + }, + { + "epoch": 0.16, + "learning_rate": 1.9067334250599687e-05, + "loss": 1.136, + "step": 6837 + }, + { + "epoch": 0.16, + "learning_rate": 1.9067005759998797e-05, + "loss": 1.0684, + "step": 6838 + }, + { + "epoch": 0.16, + "learning_rate": 1.906667721439052e-05, + "loss": 1.4025, + "step": 6839 + }, + { + "epoch": 0.16, + "learning_rate": 1.906634861377684e-05, + "loss": 1.0889, + "step": 6840 + }, + { + "epoch": 0.16, + "learning_rate": 1.9066019958159755e-05, + "loss": 0.9783, + "step": 6841 + }, + { + "epoch": 0.16, + "learning_rate": 1.9065691247541264e-05, + "loss": 1.3393, + "step": 6842 + }, + { + "epoch": 0.16, + "learning_rate": 1.9065362481923354e-05, + "loss": 0.9854, + "step": 6843 + }, + { + "epoch": 0.16, + "learning_rate": 1.9065033661308022e-05, + "loss": 1.2401, + "step": 6844 + }, + { + "epoch": 0.16, + "learning_rate": 1.9064704785697264e-05, + "loss": 0.88, + "step": 6845 + }, + { + "epoch": 0.16, + "learning_rate": 1.906437585509307e-05, + "loss": 0.9362, + "step": 6846 + }, + { + "epoch": 0.16, + "learning_rate": 1.9064046869497445e-05, + "loss": 1.2729, + "step": 6847 + }, + { + "epoch": 0.16, + "learning_rate": 1.9063717828912376e-05, + "loss": 1.1059, + "step": 6848 + }, + { + "epoch": 0.16, + "learning_rate": 1.9063388733339863e-05, + "loss": 1.0445, + "step": 6849 + }, + { + "epoch": 0.16, + "learning_rate": 1.9063059582781905e-05, + "loss": 1.1629, + "step": 6850 + }, + { + "epoch": 0.16, + "learning_rate": 1.9062730377240497e-05, + "loss": 1.4265, + "step": 6851 + }, + { + "epoch": 0.16, + "learning_rate": 1.9062401116717635e-05, + "loss": 1.2033, + "step": 6852 + }, + { + "epoch": 0.16, + "learning_rate": 1.9062071801215315e-05, + "loss": 1.1102, + "step": 6853 + }, + { + "epoch": 0.16, + "learning_rate": 1.9061742430735538e-05, + "loss": 1.2722, + "step": 6854 + }, + { + "epoch": 0.16, + "learning_rate": 1.90614130052803e-05, + "loss": 1.1274, + "step": 6855 + }, + { + "epoch": 0.16, + "learning_rate": 1.90610835248516e-05, + "loss": 1.0298, + "step": 6856 + }, + { + "epoch": 0.16, + "learning_rate": 1.906075398945144e-05, + "loss": 0.9632, + "step": 6857 + }, + { + "epoch": 0.16, + "learning_rate": 1.9060424399081816e-05, + "loss": 1.0892, + "step": 6858 + }, + { + "epoch": 0.16, + "learning_rate": 1.906009475374473e-05, + "loss": 1.1282, + "step": 6859 + }, + { + "epoch": 0.16, + "learning_rate": 1.9059765053442176e-05, + "loss": 1.1807, + "step": 6860 + }, + { + "epoch": 0.16, + "learning_rate": 1.905943529817616e-05, + "loss": 1.2989, + "step": 6861 + }, + { + "epoch": 0.17, + "learning_rate": 1.9059105487948683e-05, + "loss": 1.2118, + "step": 6862 + }, + { + "epoch": 0.17, + "learning_rate": 1.9058775622761743e-05, + "loss": 1.2983, + "step": 6863 + }, + { + "epoch": 0.17, + "learning_rate": 1.905844570261734e-05, + "loss": 0.9729, + "step": 6864 + }, + { + "epoch": 0.17, + "learning_rate": 1.905811572751748e-05, + "loss": 1.2087, + "step": 6865 + }, + { + "epoch": 0.17, + "learning_rate": 1.9057785697464162e-05, + "loss": 1.0843, + "step": 6866 + }, + { + "epoch": 0.17, + "learning_rate": 1.905745561245939e-05, + "loss": 1.1048, + "step": 6867 + }, + { + "epoch": 0.17, + "learning_rate": 1.9057125472505162e-05, + "loss": 1.1836, + "step": 6868 + }, + { + "epoch": 0.17, + "learning_rate": 1.9056795277603485e-05, + "loss": 1.1266, + "step": 6869 + }, + { + "epoch": 0.17, + "learning_rate": 1.9056465027756362e-05, + "loss": 1.0955, + "step": 6870 + }, + { + "epoch": 0.17, + "learning_rate": 1.9056134722965797e-05, + "loss": 1.3271, + "step": 6871 + }, + { + "epoch": 0.17, + "learning_rate": 1.905580436323379e-05, + "loss": 1.4187, + "step": 6872 + }, + { + "epoch": 0.17, + "learning_rate": 1.9055473948562352e-05, + "loss": 0.7107, + "step": 6873 + }, + { + "epoch": 0.17, + "learning_rate": 1.905514347895348e-05, + "loss": 1.3674, + "step": 6874 + }, + { + "epoch": 0.17, + "learning_rate": 1.9054812954409185e-05, + "loss": 1.3635, + "step": 6875 + }, + { + "epoch": 0.17, + "learning_rate": 1.905448237493147e-05, + "loss": 1.1836, + "step": 6876 + }, + { + "epoch": 0.17, + "learning_rate": 1.9054151740522336e-05, + "loss": 1.128, + "step": 6877 + }, + { + "epoch": 0.17, + "learning_rate": 1.9053821051183797e-05, + "loss": 1.4442, + "step": 6878 + }, + { + "epoch": 0.17, + "learning_rate": 1.905349030691785e-05, + "loss": 1.1139, + "step": 6879 + }, + { + "epoch": 0.17, + "learning_rate": 1.9053159507726514e-05, + "loss": 1.214, + "step": 6880 + }, + { + "epoch": 0.17, + "learning_rate": 1.9052828653611786e-05, + "loss": 1.2239, + "step": 6881 + }, + { + "epoch": 0.17, + "learning_rate": 1.9052497744575675e-05, + "loss": 0.9861, + "step": 6882 + }, + { + "epoch": 0.17, + "learning_rate": 1.9052166780620188e-05, + "loss": 1.0726, + "step": 6883 + }, + { + "epoch": 0.17, + "learning_rate": 1.9051835761747336e-05, + "loss": 1.0053, + "step": 6884 + }, + { + "epoch": 0.17, + "learning_rate": 1.9051504687959126e-05, + "loss": 0.7545, + "step": 6885 + }, + { + "epoch": 0.17, + "learning_rate": 1.9051173559257567e-05, + "loss": 0.8167, + "step": 6886 + }, + { + "epoch": 0.17, + "learning_rate": 1.9050842375644664e-05, + "loss": 1.0554, + "step": 6887 + }, + { + "epoch": 0.17, + "learning_rate": 1.905051113712243e-05, + "loss": 1.2853, + "step": 6888 + }, + { + "epoch": 0.17, + "learning_rate": 1.9050179843692875e-05, + "loss": 1.1392, + "step": 6889 + }, + { + "epoch": 0.17, + "learning_rate": 1.9049848495358006e-05, + "loss": 1.1594, + "step": 6890 + }, + { + "epoch": 0.17, + "learning_rate": 1.9049517092119838e-05, + "loss": 0.8912, + "step": 6891 + }, + { + "epoch": 0.17, + "learning_rate": 1.9049185633980376e-05, + "loss": 1.3128, + "step": 6892 + }, + { + "epoch": 0.17, + "learning_rate": 1.9048854120941635e-05, + "loss": 1.2388, + "step": 6893 + }, + { + "epoch": 0.17, + "learning_rate": 1.904852255300562e-05, + "loss": 1.4992, + "step": 6894 + }, + { + "epoch": 0.17, + "learning_rate": 1.904819093017435e-05, + "loss": 1.1729, + "step": 6895 + }, + { + "epoch": 0.17, + "learning_rate": 1.9047859252449837e-05, + "loss": 1.17, + "step": 6896 + }, + { + "epoch": 0.17, + "learning_rate": 1.9047527519834085e-05, + "loss": 1.0303, + "step": 6897 + }, + { + "epoch": 0.17, + "learning_rate": 1.9047195732329117e-05, + "loss": 1.2643, + "step": 6898 + }, + { + "epoch": 0.17, + "learning_rate": 1.9046863889936936e-05, + "loss": 1.3232, + "step": 6899 + }, + { + "epoch": 0.17, + "learning_rate": 1.904653199265956e-05, + "loss": 0.9851, + "step": 6900 + }, + { + "epoch": 0.17, + "learning_rate": 1.9046200040499004e-05, + "loss": 1.3088, + "step": 6901 + }, + { + "epoch": 0.17, + "learning_rate": 1.904586803345728e-05, + "loss": 1.101, + "step": 6902 + }, + { + "epoch": 0.17, + "learning_rate": 1.90455359715364e-05, + "loss": 1.1081, + "step": 6903 + }, + { + "epoch": 0.17, + "learning_rate": 1.9045203854738386e-05, + "loss": 1.1171, + "step": 6904 + }, + { + "epoch": 0.17, + "learning_rate": 1.9044871683065243e-05, + "loss": 1.2028, + "step": 6905 + }, + { + "epoch": 0.17, + "learning_rate": 1.9044539456518992e-05, + "loss": 1.1752, + "step": 6906 + }, + { + "epoch": 0.17, + "learning_rate": 1.9044207175101648e-05, + "loss": 1.4409, + "step": 6907 + }, + { + "epoch": 0.17, + "learning_rate": 1.9043874838815225e-05, + "loss": 0.9374, + "step": 6908 + }, + { + "epoch": 0.17, + "learning_rate": 1.9043542447661744e-05, + "loss": 1.2209, + "step": 6909 + }, + { + "epoch": 0.17, + "learning_rate": 1.9043210001643215e-05, + "loss": 0.9682, + "step": 6910 + }, + { + "epoch": 0.17, + "learning_rate": 1.9042877500761657e-05, + "loss": 1.1087, + "step": 6911 + }, + { + "epoch": 0.17, + "learning_rate": 1.9042544945019094e-05, + "loss": 1.2792, + "step": 6912 + }, + { + "epoch": 0.17, + "learning_rate": 1.9042212334417533e-05, + "loss": 1.2523, + "step": 6913 + }, + { + "epoch": 0.17, + "learning_rate": 1.9041879668959e-05, + "loss": 1.3202, + "step": 6914 + }, + { + "epoch": 0.17, + "learning_rate": 1.9041546948645506e-05, + "loss": 1.2993, + "step": 6915 + }, + { + "epoch": 0.17, + "learning_rate": 1.9041214173479075e-05, + "loss": 1.1517, + "step": 6916 + }, + { + "epoch": 0.17, + "learning_rate": 1.9040881343461723e-05, + "loss": 1.2809, + "step": 6917 + }, + { + "epoch": 0.17, + "learning_rate": 1.9040548458595472e-05, + "loss": 0.9933, + "step": 6918 + }, + { + "epoch": 0.17, + "learning_rate": 1.904021551888234e-05, + "loss": 1.0053, + "step": 6919 + }, + { + "epoch": 0.17, + "learning_rate": 1.9039882524324346e-05, + "loss": 0.7768, + "step": 6920 + }, + { + "epoch": 0.17, + "learning_rate": 1.903954947492351e-05, + "loss": 1.2125, + "step": 6921 + }, + { + "epoch": 0.17, + "learning_rate": 1.9039216370681858e-05, + "loss": 1.2027, + "step": 6922 + }, + { + "epoch": 0.17, + "learning_rate": 1.9038883211601402e-05, + "loss": 1.0372, + "step": 6923 + }, + { + "epoch": 0.17, + "learning_rate": 1.903854999768417e-05, + "loss": 1.245, + "step": 6924 + }, + { + "epoch": 0.17, + "learning_rate": 1.903821672893218e-05, + "loss": 1.0535, + "step": 6925 + }, + { + "epoch": 0.17, + "learning_rate": 1.9037883405347454e-05, + "loss": 0.9704, + "step": 6926 + }, + { + "epoch": 0.17, + "learning_rate": 1.9037550026932018e-05, + "loss": 1.1757, + "step": 6927 + }, + { + "epoch": 0.17, + "learning_rate": 1.903721659368789e-05, + "loss": 1.1674, + "step": 6928 + }, + { + "epoch": 0.17, + "learning_rate": 1.9036883105617095e-05, + "loss": 1.3053, + "step": 6929 + }, + { + "epoch": 0.17, + "learning_rate": 1.9036549562721657e-05, + "loss": 0.9839, + "step": 6930 + }, + { + "epoch": 0.17, + "learning_rate": 1.9036215965003593e-05, + "loss": 1.2663, + "step": 6931 + }, + { + "epoch": 0.17, + "learning_rate": 1.9035882312464938e-05, + "loss": 1.1839, + "step": 6932 + }, + { + "epoch": 0.17, + "learning_rate": 1.903554860510771e-05, + "loss": 1.1087, + "step": 6933 + }, + { + "epoch": 0.17, + "learning_rate": 1.903521484293393e-05, + "loss": 1.0186, + "step": 6934 + }, + { + "epoch": 0.17, + "learning_rate": 1.903488102594563e-05, + "loss": 1.0997, + "step": 6935 + }, + { + "epoch": 0.17, + "learning_rate": 1.9034547154144832e-05, + "loss": 1.2503, + "step": 6936 + }, + { + "epoch": 0.17, + "learning_rate": 1.903421322753356e-05, + "loss": 1.244, + "step": 6937 + }, + { + "epoch": 0.17, + "learning_rate": 1.9033879246113842e-05, + "loss": 1.0569, + "step": 6938 + }, + { + "epoch": 0.17, + "learning_rate": 1.9033545209887703e-05, + "loss": 1.028, + "step": 6939 + }, + { + "epoch": 0.17, + "learning_rate": 1.903321111885717e-05, + "loss": 1.1366, + "step": 6940 + }, + { + "epoch": 0.17, + "learning_rate": 1.9032876973024273e-05, + "loss": 1.119, + "step": 6941 + }, + { + "epoch": 0.17, + "learning_rate": 1.903254277239103e-05, + "loss": 1.4487, + "step": 6942 + }, + { + "epoch": 0.17, + "learning_rate": 1.903220851695948e-05, + "loss": 1.2723, + "step": 6943 + }, + { + "epoch": 0.17, + "learning_rate": 1.9031874206731644e-05, + "loss": 1.4767, + "step": 6944 + }, + { + "epoch": 0.17, + "learning_rate": 1.903153984170955e-05, + "loss": 1.2025, + "step": 6945 + }, + { + "epoch": 0.17, + "learning_rate": 1.903120542189523e-05, + "loss": 1.2845, + "step": 6946 + }, + { + "epoch": 0.17, + "learning_rate": 1.903087094729071e-05, + "loss": 1.2253, + "step": 6947 + }, + { + "epoch": 0.17, + "learning_rate": 1.903053641789802e-05, + "loss": 1.0975, + "step": 6948 + }, + { + "epoch": 0.17, + "learning_rate": 1.903020183371919e-05, + "loss": 1.0206, + "step": 6949 + }, + { + "epoch": 0.17, + "learning_rate": 1.9029867194756248e-05, + "loss": 1.2983, + "step": 6950 + }, + { + "epoch": 0.17, + "learning_rate": 1.902953250101123e-05, + "loss": 1.2523, + "step": 6951 + }, + { + "epoch": 0.17, + "learning_rate": 1.902919775248616e-05, + "loss": 1.2929, + "step": 6952 + }, + { + "epoch": 0.17, + "learning_rate": 1.902886294918307e-05, + "loss": 1.0485, + "step": 6953 + }, + { + "epoch": 0.17, + "learning_rate": 1.902852809110399e-05, + "loss": 1.0637, + "step": 6954 + }, + { + "epoch": 0.17, + "learning_rate": 1.9028193178250958e-05, + "loss": 1.1916, + "step": 6955 + }, + { + "epoch": 0.17, + "learning_rate": 1.9027858210626e-05, + "loss": 1.0413, + "step": 6956 + }, + { + "epoch": 0.17, + "learning_rate": 1.9027523188231148e-05, + "loss": 1.1467, + "step": 6957 + }, + { + "epoch": 0.17, + "learning_rate": 1.902718811106844e-05, + "loss": 0.8498, + "step": 6958 + }, + { + "epoch": 0.17, + "learning_rate": 1.9026852979139903e-05, + "loss": 1.2003, + "step": 6959 + }, + { + "epoch": 0.17, + "learning_rate": 1.902651779244757e-05, + "loss": 0.7407, + "step": 6960 + }, + { + "epoch": 0.17, + "learning_rate": 1.902618255099348e-05, + "loss": 1.1009, + "step": 6961 + }, + { + "epoch": 0.17, + "learning_rate": 1.9025847254779662e-05, + "loss": 1.3357, + "step": 6962 + }, + { + "epoch": 0.17, + "learning_rate": 1.9025511903808152e-05, + "loss": 1.2272, + "step": 6963 + }, + { + "epoch": 0.17, + "learning_rate": 1.9025176498080987e-05, + "loss": 1.0908, + "step": 6964 + }, + { + "epoch": 0.17, + "learning_rate": 1.9024841037600196e-05, + "loss": 1.154, + "step": 6965 + }, + { + "epoch": 0.17, + "learning_rate": 1.9024505522367818e-05, + "loss": 1.2141, + "step": 6966 + }, + { + "epoch": 0.17, + "learning_rate": 1.9024169952385887e-05, + "loss": 1.2834, + "step": 6967 + }, + { + "epoch": 0.17, + "learning_rate": 1.902383432765644e-05, + "loss": 1.3468, + "step": 6968 + }, + { + "epoch": 0.17, + "learning_rate": 1.902349864818151e-05, + "loss": 1.0292, + "step": 6969 + }, + { + "epoch": 0.17, + "learning_rate": 1.9023162913963136e-05, + "loss": 1.1074, + "step": 6970 + }, + { + "epoch": 0.17, + "learning_rate": 1.902282712500336e-05, + "loss": 1.2417, + "step": 6971 + }, + { + "epoch": 0.17, + "learning_rate": 1.9022491281304214e-05, + "loss": 1.0833, + "step": 6972 + }, + { + "epoch": 0.17, + "learning_rate": 1.902215538286773e-05, + "loss": 1.2481, + "step": 6973 + }, + { + "epoch": 0.17, + "learning_rate": 1.9021819429695955e-05, + "loss": 1.2301, + "step": 6974 + }, + { + "epoch": 0.17, + "learning_rate": 1.9021483421790926e-05, + "loss": 1.0116, + "step": 6975 + }, + { + "epoch": 0.17, + "learning_rate": 1.9021147359154674e-05, + "loss": 1.1402, + "step": 6976 + }, + { + "epoch": 0.17, + "learning_rate": 1.9020811241789244e-05, + "loss": 1.1297, + "step": 6977 + }, + { + "epoch": 0.17, + "learning_rate": 1.9020475069696676e-05, + "loss": 0.9733, + "step": 6978 + }, + { + "epoch": 0.17, + "learning_rate": 1.9020138842879008e-05, + "loss": 1.1672, + "step": 6979 + }, + { + "epoch": 0.17, + "learning_rate": 1.9019802561338278e-05, + "loss": 0.9735, + "step": 6980 + }, + { + "epoch": 0.17, + "learning_rate": 1.901946622507653e-05, + "loss": 1.2457, + "step": 6981 + }, + { + "epoch": 0.17, + "learning_rate": 1.9019129834095798e-05, + "loss": 1.2589, + "step": 6982 + }, + { + "epoch": 0.17, + "learning_rate": 1.9018793388398128e-05, + "loss": 1.1115, + "step": 6983 + }, + { + "epoch": 0.17, + "learning_rate": 1.9018456887985558e-05, + "loss": 1.2237, + "step": 6984 + }, + { + "epoch": 0.17, + "learning_rate": 1.9018120332860135e-05, + "loss": 1.0925, + "step": 6985 + }, + { + "epoch": 0.17, + "learning_rate": 1.9017783723023895e-05, + "loss": 0.9928, + "step": 6986 + }, + { + "epoch": 0.17, + "learning_rate": 1.901744705847888e-05, + "loss": 1.2868, + "step": 6987 + }, + { + "epoch": 0.17, + "learning_rate": 1.901711033922714e-05, + "loss": 0.9624, + "step": 6988 + }, + { + "epoch": 0.17, + "learning_rate": 1.901677356527071e-05, + "loss": 1.2326, + "step": 6989 + }, + { + "epoch": 0.17, + "learning_rate": 1.9016436736611637e-05, + "loss": 1.4186, + "step": 6990 + }, + { + "epoch": 0.17, + "learning_rate": 1.901609985325196e-05, + "loss": 0.9648, + "step": 6991 + }, + { + "epoch": 0.17, + "learning_rate": 1.9015762915193727e-05, + "loss": 1.2041, + "step": 6992 + }, + { + "epoch": 0.17, + "learning_rate": 1.9015425922438984e-05, + "loss": 0.9495, + "step": 6993 + }, + { + "epoch": 0.17, + "learning_rate": 1.901508887498977e-05, + "loss": 1.2439, + "step": 6994 + }, + { + "epoch": 0.17, + "learning_rate": 1.9014751772848134e-05, + "loss": 1.3042, + "step": 6995 + }, + { + "epoch": 0.17, + "learning_rate": 1.9014414616016116e-05, + "loss": 1.2219, + "step": 6996 + }, + { + "epoch": 0.17, + "learning_rate": 1.9014077404495768e-05, + "loss": 1.3781, + "step": 6997 + }, + { + "epoch": 0.17, + "learning_rate": 1.901374013828913e-05, + "loss": 1.2651, + "step": 6998 + }, + { + "epoch": 0.17, + "learning_rate": 1.901340281739825e-05, + "loss": 1.2571, + "step": 6999 + }, + { + "epoch": 0.17, + "learning_rate": 1.901306544182518e-05, + "loss": 1.1168, + "step": 7000 + }, + { + "epoch": 0.17, + "learning_rate": 1.901272801157196e-05, + "loss": 1.3746, + "step": 7001 + }, + { + "epoch": 0.17, + "learning_rate": 1.9012390526640638e-05, + "loss": 1.1663, + "step": 7002 + }, + { + "epoch": 0.17, + "learning_rate": 1.9012052987033262e-05, + "loss": 1.0382, + "step": 7003 + }, + { + "epoch": 0.17, + "learning_rate": 1.9011715392751882e-05, + "loss": 1.4152, + "step": 7004 + }, + { + "epoch": 0.17, + "learning_rate": 1.9011377743798543e-05, + "loss": 1.1996, + "step": 7005 + }, + { + "epoch": 0.17, + "learning_rate": 1.9011040040175295e-05, + "loss": 1.3168, + "step": 7006 + }, + { + "epoch": 0.17, + "learning_rate": 1.901070228188419e-05, + "loss": 0.9661, + "step": 7007 + }, + { + "epoch": 0.17, + "learning_rate": 1.9010364468927267e-05, + "loss": 1.1812, + "step": 7008 + }, + { + "epoch": 0.17, + "learning_rate": 1.9010026601306586e-05, + "loss": 1.2173, + "step": 7009 + }, + { + "epoch": 0.17, + "learning_rate": 1.900968867902419e-05, + "loss": 1.1506, + "step": 7010 + }, + { + "epoch": 0.17, + "learning_rate": 1.9009350702082138e-05, + "loss": 1.1566, + "step": 7011 + }, + { + "epoch": 0.17, + "learning_rate": 1.900901267048247e-05, + "loss": 1.0736, + "step": 7012 + }, + { + "epoch": 0.17, + "learning_rate": 1.9008674584227242e-05, + "loss": 0.9469, + "step": 7013 + }, + { + "epoch": 0.17, + "learning_rate": 1.9008336443318505e-05, + "loss": 1.0078, + "step": 7014 + }, + { + "epoch": 0.17, + "learning_rate": 1.900799824775831e-05, + "loss": 1.2386, + "step": 7015 + }, + { + "epoch": 0.17, + "learning_rate": 1.900765999754871e-05, + "loss": 1.1979, + "step": 7016 + }, + { + "epoch": 0.17, + "learning_rate": 1.9007321692691748e-05, + "loss": 1.2573, + "step": 7017 + }, + { + "epoch": 0.17, + "learning_rate": 1.9006983333189493e-05, + "loss": 1.436, + "step": 7018 + }, + { + "epoch": 0.17, + "learning_rate": 1.9006644919043984e-05, + "loss": 1.1195, + "step": 7019 + }, + { + "epoch": 0.17, + "learning_rate": 1.9006306450257278e-05, + "loss": 0.8686, + "step": 7020 + }, + { + "epoch": 0.17, + "learning_rate": 1.9005967926831434e-05, + "loss": 1.0995, + "step": 7021 + }, + { + "epoch": 0.17, + "learning_rate": 1.90056293487685e-05, + "loss": 1.2078, + "step": 7022 + }, + { + "epoch": 0.17, + "learning_rate": 1.9005290716070525e-05, + "loss": 1.1243, + "step": 7023 + }, + { + "epoch": 0.17, + "learning_rate": 1.900495202873957e-05, + "loss": 1.1084, + "step": 7024 + }, + { + "epoch": 0.17, + "learning_rate": 1.9004613286777695e-05, + "loss": 1.0224, + "step": 7025 + }, + { + "epoch": 0.17, + "learning_rate": 1.9004274490186944e-05, + "loss": 1.0167, + "step": 7026 + }, + { + "epoch": 0.17, + "learning_rate": 1.9003935638969382e-05, + "loss": 1.1834, + "step": 7027 + }, + { + "epoch": 0.17, + "learning_rate": 1.9003596733127058e-05, + "loss": 1.0949, + "step": 7028 + }, + { + "epoch": 0.17, + "learning_rate": 1.9003257772662032e-05, + "loss": 1.1608, + "step": 7029 + }, + { + "epoch": 0.17, + "learning_rate": 1.9002918757576358e-05, + "loss": 1.288, + "step": 7030 + }, + { + "epoch": 0.17, + "learning_rate": 1.900257968787209e-05, + "loss": 1.2677, + "step": 7031 + }, + { + "epoch": 0.17, + "learning_rate": 1.9002240563551293e-05, + "loss": 1.1503, + "step": 7032 + }, + { + "epoch": 0.17, + "learning_rate": 1.900190138461602e-05, + "loss": 1.1327, + "step": 7033 + }, + { + "epoch": 0.17, + "learning_rate": 1.9001562151068327e-05, + "loss": 1.0334, + "step": 7034 + }, + { + "epoch": 0.17, + "learning_rate": 1.9001222862910272e-05, + "loss": 1.3574, + "step": 7035 + }, + { + "epoch": 0.17, + "learning_rate": 1.900088352014392e-05, + "loss": 1.1482, + "step": 7036 + }, + { + "epoch": 0.17, + "learning_rate": 1.900054412277132e-05, + "loss": 1.1486, + "step": 7037 + }, + { + "epoch": 0.17, + "learning_rate": 1.900020467079454e-05, + "loss": 0.9756, + "step": 7038 + }, + { + "epoch": 0.17, + "learning_rate": 1.899986516421563e-05, + "loss": 1.0841, + "step": 7039 + }, + { + "epoch": 0.17, + "learning_rate": 1.899952560303666e-05, + "loss": 1.2927, + "step": 7040 + }, + { + "epoch": 0.17, + "learning_rate": 1.899918598725968e-05, + "loss": 1.0684, + "step": 7041 + }, + { + "epoch": 0.17, + "learning_rate": 1.899884631688676e-05, + "loss": 1.4727, + "step": 7042 + }, + { + "epoch": 0.17, + "learning_rate": 1.8998506591919957e-05, + "loss": 1.1916, + "step": 7043 + }, + { + "epoch": 0.17, + "learning_rate": 1.8998166812361325e-05, + "loss": 1.0909, + "step": 7044 + }, + { + "epoch": 0.17, + "learning_rate": 1.8997826978212937e-05, + "loss": 1.08, + "step": 7045 + }, + { + "epoch": 0.17, + "learning_rate": 1.8997487089476844e-05, + "loss": 0.9879, + "step": 7046 + }, + { + "epoch": 0.17, + "learning_rate": 1.899714714615512e-05, + "loss": 1.2612, + "step": 7047 + }, + { + "epoch": 0.17, + "learning_rate": 1.8996807148249817e-05, + "loss": 1.34, + "step": 7048 + }, + { + "epoch": 0.17, + "learning_rate": 1.8996467095762997e-05, + "loss": 1.0789, + "step": 7049 + }, + { + "epoch": 0.17, + "learning_rate": 1.899612698869673e-05, + "loss": 1.3123, + "step": 7050 + }, + { + "epoch": 0.17, + "learning_rate": 1.899578682705308e-05, + "loss": 0.9472, + "step": 7051 + }, + { + "epoch": 0.17, + "learning_rate": 1.8995446610834104e-05, + "loss": 1.0531, + "step": 7052 + }, + { + "epoch": 0.17, + "learning_rate": 1.8995106340041875e-05, + "loss": 0.9276, + "step": 7053 + }, + { + "epoch": 0.17, + "learning_rate": 1.8994766014678444e-05, + "loss": 1.2955, + "step": 7054 + }, + { + "epoch": 0.17, + "learning_rate": 1.8994425634745885e-05, + "loss": 1.2935, + "step": 7055 + }, + { + "epoch": 0.17, + "learning_rate": 1.8994085200246263e-05, + "loss": 1.2142, + "step": 7056 + }, + { + "epoch": 0.17, + "learning_rate": 1.899374471118164e-05, + "loss": 1.1755, + "step": 7057 + }, + { + "epoch": 0.17, + "learning_rate": 1.8993404167554085e-05, + "loss": 1.1641, + "step": 7058 + }, + { + "epoch": 0.17, + "learning_rate": 1.8993063569365666e-05, + "loss": 1.0753, + "step": 7059 + }, + { + "epoch": 0.17, + "learning_rate": 1.899272291661844e-05, + "loss": 1.0425, + "step": 7060 + }, + { + "epoch": 0.17, + "learning_rate": 1.899238220931448e-05, + "loss": 1.1905, + "step": 7061 + }, + { + "epoch": 0.17, + "learning_rate": 1.899204144745585e-05, + "loss": 0.8165, + "step": 7062 + }, + { + "epoch": 0.17, + "learning_rate": 1.8991700631044624e-05, + "loss": 1.2816, + "step": 7063 + }, + { + "epoch": 0.17, + "learning_rate": 1.8991359760082864e-05, + "loss": 1.3222, + "step": 7064 + }, + { + "epoch": 0.17, + "learning_rate": 1.899101883457264e-05, + "loss": 1.2936, + "step": 7065 + }, + { + "epoch": 0.17, + "learning_rate": 1.8990677854516017e-05, + "loss": 1.2971, + "step": 7066 + }, + { + "epoch": 0.17, + "learning_rate": 1.8990336819915067e-05, + "loss": 1.311, + "step": 7067 + }, + { + "epoch": 0.17, + "learning_rate": 1.898999573077186e-05, + "loss": 1.161, + "step": 7068 + }, + { + "epoch": 0.17, + "learning_rate": 1.8989654587088463e-05, + "loss": 0.9758, + "step": 7069 + }, + { + "epoch": 0.17, + "learning_rate": 1.8989313388866944e-05, + "loss": 1.4087, + "step": 7070 + }, + { + "epoch": 0.17, + "learning_rate": 1.898897213610938e-05, + "loss": 1.1401, + "step": 7071 + }, + { + "epoch": 0.17, + "learning_rate": 1.8988630828817827e-05, + "loss": 1.1183, + "step": 7072 + }, + { + "epoch": 0.17, + "learning_rate": 1.8988289466994376e-05, + "loss": 1.2046, + "step": 7073 + }, + { + "epoch": 0.17, + "learning_rate": 1.898794805064108e-05, + "loss": 1.2855, + "step": 7074 + }, + { + "epoch": 0.17, + "learning_rate": 1.898760657976002e-05, + "loss": 1.1074, + "step": 7075 + }, + { + "epoch": 0.17, + "learning_rate": 1.8987265054353262e-05, + "loss": 1.008, + "step": 7076 + }, + { + "epoch": 0.17, + "learning_rate": 1.8986923474422885e-05, + "loss": 0.9645, + "step": 7077 + }, + { + "epoch": 0.17, + "learning_rate": 1.8986581839970956e-05, + "loss": 0.92, + "step": 7078 + }, + { + "epoch": 0.17, + "learning_rate": 1.8986240150999548e-05, + "loss": 1.3183, + "step": 7079 + }, + { + "epoch": 0.17, + "learning_rate": 1.898589840751073e-05, + "loss": 1.1931, + "step": 7080 + }, + { + "epoch": 0.17, + "learning_rate": 1.898555660950659e-05, + "loss": 1.0089, + "step": 7081 + }, + { + "epoch": 0.17, + "learning_rate": 1.898521475698918e-05, + "loss": 0.9544, + "step": 7082 + }, + { + "epoch": 0.17, + "learning_rate": 1.8984872849960597e-05, + "loss": 1.0444, + "step": 7083 + }, + { + "epoch": 0.17, + "learning_rate": 1.8984530888422897e-05, + "loss": 1.0253, + "step": 7084 + }, + { + "epoch": 0.17, + "learning_rate": 1.8984188872378163e-05, + "loss": 1.1385, + "step": 7085 + }, + { + "epoch": 0.17, + "learning_rate": 1.8983846801828468e-05, + "loss": 1.1549, + "step": 7086 + }, + { + "epoch": 0.17, + "learning_rate": 1.8983504676775886e-05, + "loss": 1.3016, + "step": 7087 + }, + { + "epoch": 0.17, + "learning_rate": 1.8983162497222497e-05, + "loss": 1.2396, + "step": 7088 + }, + { + "epoch": 0.17, + "learning_rate": 1.898282026317037e-05, + "loss": 1.2631, + "step": 7089 + }, + { + "epoch": 0.17, + "learning_rate": 1.898247797462159e-05, + "loss": 1.135, + "step": 7090 + }, + { + "epoch": 0.17, + "learning_rate": 1.898213563157823e-05, + "loss": 1.0542, + "step": 7091 + }, + { + "epoch": 0.17, + "learning_rate": 1.8981793234042362e-05, + "loss": 1.0502, + "step": 7092 + }, + { + "epoch": 0.17, + "learning_rate": 1.8981450782016068e-05, + "loss": 1.2073, + "step": 7093 + }, + { + "epoch": 0.17, + "learning_rate": 1.8981108275501422e-05, + "loss": 1.222, + "step": 7094 + }, + { + "epoch": 0.17, + "learning_rate": 1.898076571450051e-05, + "loss": 1.3214, + "step": 7095 + }, + { + "epoch": 0.17, + "learning_rate": 1.8980423099015402e-05, + "loss": 1.034, + "step": 7096 + }, + { + "epoch": 0.17, + "learning_rate": 1.898008042904818e-05, + "loss": 1.1968, + "step": 7097 + }, + { + "epoch": 0.17, + "learning_rate": 1.8979737704600923e-05, + "loss": 0.8613, + "step": 7098 + }, + { + "epoch": 0.17, + "learning_rate": 1.8979394925675708e-05, + "loss": 1.3163, + "step": 7099 + }, + { + "epoch": 0.17, + "learning_rate": 1.8979052092274615e-05, + "loss": 1.1816, + "step": 7100 + }, + { + "epoch": 0.17, + "learning_rate": 1.8978709204399726e-05, + "loss": 1.1456, + "step": 7101 + }, + { + "epoch": 0.17, + "learning_rate": 1.897836626205312e-05, + "loss": 1.1604, + "step": 7102 + }, + { + "epoch": 0.17, + "learning_rate": 1.897802326523688e-05, + "loss": 0.9915, + "step": 7103 + }, + { + "epoch": 0.17, + "learning_rate": 1.8977680213953084e-05, + "loss": 1.317, + "step": 7104 + }, + { + "epoch": 0.17, + "learning_rate": 1.897733710820381e-05, + "loss": 1.0852, + "step": 7105 + }, + { + "epoch": 0.17, + "learning_rate": 1.8976993947991145e-05, + "loss": 1.0808, + "step": 7106 + }, + { + "epoch": 0.17, + "learning_rate": 1.8976650733317172e-05, + "loss": 1.1751, + "step": 7107 + }, + { + "epoch": 0.17, + "learning_rate": 1.897630746418397e-05, + "loss": 1.2722, + "step": 7108 + }, + { + "epoch": 0.17, + "learning_rate": 1.8975964140593623e-05, + "loss": 1.2293, + "step": 7109 + }, + { + "epoch": 0.17, + "learning_rate": 1.8975620762548207e-05, + "loss": 1.5201, + "step": 7110 + }, + { + "epoch": 0.17, + "learning_rate": 1.8975277330049815e-05, + "loss": 1.3145, + "step": 7111 + }, + { + "epoch": 0.17, + "learning_rate": 1.8974933843100526e-05, + "loss": 1.253, + "step": 7112 + }, + { + "epoch": 0.17, + "learning_rate": 1.8974590301702425e-05, + "loss": 1.0487, + "step": 7113 + }, + { + "epoch": 0.17, + "learning_rate": 1.8974246705857594e-05, + "loss": 1.2141, + "step": 7114 + }, + { + "epoch": 0.17, + "learning_rate": 1.8973903055568122e-05, + "loss": 1.1366, + "step": 7115 + }, + { + "epoch": 0.17, + "learning_rate": 1.8973559350836092e-05, + "loss": 0.9535, + "step": 7116 + }, + { + "epoch": 0.17, + "learning_rate": 1.8973215591663584e-05, + "loss": 1.2593, + "step": 7117 + }, + { + "epoch": 0.17, + "learning_rate": 1.8972871778052688e-05, + "loss": 1.2778, + "step": 7118 + }, + { + "epoch": 0.17, + "learning_rate": 1.8972527910005492e-05, + "loss": 1.333, + "step": 7119 + }, + { + "epoch": 0.17, + "learning_rate": 1.8972183987524076e-05, + "loss": 1.1739, + "step": 7120 + }, + { + "epoch": 0.17, + "learning_rate": 1.8971840010610534e-05, + "loss": 0.9959, + "step": 7121 + }, + { + "epoch": 0.17, + "learning_rate": 1.8971495979266946e-05, + "loss": 1.2754, + "step": 7122 + }, + { + "epoch": 0.17, + "learning_rate": 1.8971151893495403e-05, + "loss": 1.0996, + "step": 7123 + }, + { + "epoch": 0.17, + "learning_rate": 1.897080775329799e-05, + "loss": 1.407, + "step": 7124 + }, + { + "epoch": 0.17, + "learning_rate": 1.89704635586768e-05, + "loss": 1.283, + "step": 7125 + }, + { + "epoch": 0.17, + "learning_rate": 1.8970119309633916e-05, + "loss": 1.1037, + "step": 7126 + }, + { + "epoch": 0.17, + "learning_rate": 1.896977500617143e-05, + "loss": 1.2679, + "step": 7127 + }, + { + "epoch": 0.17, + "learning_rate": 1.8969430648291425e-05, + "loss": 1.166, + "step": 7128 + }, + { + "epoch": 0.17, + "learning_rate": 1.8969086235995996e-05, + "loss": 1.1691, + "step": 7129 + }, + { + "epoch": 0.17, + "learning_rate": 1.896874176928723e-05, + "loss": 1.2094, + "step": 7130 + }, + { + "epoch": 0.17, + "learning_rate": 1.8968397248167218e-05, + "loss": 1.1487, + "step": 7131 + }, + { + "epoch": 0.17, + "learning_rate": 1.896805267263805e-05, + "loss": 1.031, + "step": 7132 + }, + { + "epoch": 0.17, + "learning_rate": 1.8967708042701814e-05, + "loss": 1.0002, + "step": 7133 + }, + { + "epoch": 0.17, + "learning_rate": 1.89673633583606e-05, + "loss": 1.2457, + "step": 7134 + }, + { + "epoch": 0.17, + "learning_rate": 1.896701861961651e-05, + "loss": 1.1201, + "step": 7135 + }, + { + "epoch": 0.17, + "learning_rate": 1.896667382647162e-05, + "loss": 1.2677, + "step": 7136 + }, + { + "epoch": 0.17, + "learning_rate": 1.8966328978928034e-05, + "loss": 0.9373, + "step": 7137 + }, + { + "epoch": 0.17, + "learning_rate": 1.8965984076987835e-05, + "loss": 1.2372, + "step": 7138 + }, + { + "epoch": 0.17, + "learning_rate": 1.896563912065312e-05, + "loss": 1.3106, + "step": 7139 + }, + { + "epoch": 0.17, + "learning_rate": 1.8965294109925984e-05, + "loss": 1.1289, + "step": 7140 + }, + { + "epoch": 0.17, + "learning_rate": 1.8964949044808513e-05, + "loss": 0.9933, + "step": 7141 + }, + { + "epoch": 0.17, + "learning_rate": 1.8964603925302813e-05, + "loss": 1.4096, + "step": 7142 + }, + { + "epoch": 0.17, + "learning_rate": 1.896425875141096e-05, + "loss": 1.2974, + "step": 7143 + }, + { + "epoch": 0.17, + "learning_rate": 1.8963913523135062e-05, + "loss": 0.7944, + "step": 7144 + }, + { + "epoch": 0.17, + "learning_rate": 1.896356824047721e-05, + "loss": 1.2744, + "step": 7145 + }, + { + "epoch": 0.17, + "learning_rate": 1.8963222903439495e-05, + "loss": 1.0622, + "step": 7146 + }, + { + "epoch": 0.17, + "learning_rate": 1.896287751202402e-05, + "loss": 1.0643, + "step": 7147 + }, + { + "epoch": 0.17, + "learning_rate": 1.896253206623287e-05, + "loss": 1.1892, + "step": 7148 + }, + { + "epoch": 0.17, + "learning_rate": 1.896218656606815e-05, + "loss": 0.8422, + "step": 7149 + }, + { + "epoch": 0.17, + "learning_rate": 1.8961841011531948e-05, + "loss": 1.1794, + "step": 7150 + }, + { + "epoch": 0.17, + "learning_rate": 1.896149540262637e-05, + "loss": 1.1391, + "step": 7151 + }, + { + "epoch": 0.17, + "learning_rate": 1.8961149739353502e-05, + "loss": 1.2437, + "step": 7152 + }, + { + "epoch": 0.17, + "learning_rate": 1.8960804021715445e-05, + "loss": 1.1167, + "step": 7153 + }, + { + "epoch": 0.17, + "learning_rate": 1.8960458249714303e-05, + "loss": 1.1706, + "step": 7154 + }, + { + "epoch": 0.17, + "learning_rate": 1.8960112423352167e-05, + "loss": 1.218, + "step": 7155 + }, + { + "epoch": 0.17, + "learning_rate": 1.8959766542631137e-05, + "loss": 1.0386, + "step": 7156 + }, + { + "epoch": 0.17, + "learning_rate": 1.895942060755331e-05, + "loss": 1.1849, + "step": 7157 + }, + { + "epoch": 0.17, + "learning_rate": 1.8959074618120784e-05, + "loss": 1.326, + "step": 7158 + }, + { + "epoch": 0.17, + "learning_rate": 1.8958728574335662e-05, + "loss": 1.1479, + "step": 7159 + }, + { + "epoch": 0.17, + "learning_rate": 1.8958382476200038e-05, + "loss": 1.2064, + "step": 7160 + }, + { + "epoch": 0.17, + "learning_rate": 1.8958036323716018e-05, + "loss": 1.1144, + "step": 7161 + }, + { + "epoch": 0.17, + "learning_rate": 1.8957690116885697e-05, + "loss": 1.2869, + "step": 7162 + }, + { + "epoch": 0.17, + "learning_rate": 1.895734385571118e-05, + "loss": 1.1935, + "step": 7163 + }, + { + "epoch": 0.17, + "learning_rate": 1.895699754019456e-05, + "loss": 1.3616, + "step": 7164 + }, + { + "epoch": 0.17, + "learning_rate": 1.8956651170337946e-05, + "loss": 1.0813, + "step": 7165 + }, + { + "epoch": 0.17, + "learning_rate": 1.8956304746143433e-05, + "loss": 1.0832, + "step": 7166 + }, + { + "epoch": 0.17, + "learning_rate": 1.895595826761313e-05, + "loss": 1.3723, + "step": 7167 + }, + { + "epoch": 0.17, + "learning_rate": 1.895561173474913e-05, + "loss": 1.0897, + "step": 7168 + }, + { + "epoch": 0.17, + "learning_rate": 1.8955265147553545e-05, + "loss": 1.191, + "step": 7169 + }, + { + "epoch": 0.17, + "learning_rate": 1.8954918506028467e-05, + "loss": 1.0585, + "step": 7170 + }, + { + "epoch": 0.17, + "learning_rate": 1.895457181017601e-05, + "loss": 1.0659, + "step": 7171 + }, + { + "epoch": 0.17, + "learning_rate": 1.895422505999827e-05, + "loss": 1.0614, + "step": 7172 + }, + { + "epoch": 0.17, + "learning_rate": 1.8953878255497353e-05, + "loss": 1.0684, + "step": 7173 + }, + { + "epoch": 0.17, + "learning_rate": 1.895353139667536e-05, + "loss": 1.1382, + "step": 7174 + }, + { + "epoch": 0.17, + "learning_rate": 1.8953184483534402e-05, + "loss": 0.9179, + "step": 7175 + }, + { + "epoch": 0.17, + "learning_rate": 1.895283751607658e-05, + "loss": 1.0878, + "step": 7176 + }, + { + "epoch": 0.17, + "learning_rate": 1.8952490494303997e-05, + "loss": 1.0662, + "step": 7177 + }, + { + "epoch": 0.17, + "learning_rate": 1.895214341821876e-05, + "loss": 1.2298, + "step": 7178 + }, + { + "epoch": 0.17, + "learning_rate": 1.895179628782297e-05, + "loss": 1.2275, + "step": 7179 + }, + { + "epoch": 0.17, + "learning_rate": 1.8951449103118743e-05, + "loss": 1.1665, + "step": 7180 + }, + { + "epoch": 0.17, + "learning_rate": 1.8951101864108178e-05, + "loss": 1.2272, + "step": 7181 + }, + { + "epoch": 0.17, + "learning_rate": 1.8950754570793384e-05, + "loss": 1.2068, + "step": 7182 + }, + { + "epoch": 0.17, + "learning_rate": 1.895040722317647e-05, + "loss": 1.3324, + "step": 7183 + }, + { + "epoch": 0.17, + "learning_rate": 1.8950059821259535e-05, + "loss": 1.3586, + "step": 7184 + }, + { + "epoch": 0.17, + "learning_rate": 1.8949712365044697e-05, + "loss": 0.9599, + "step": 7185 + }, + { + "epoch": 0.17, + "learning_rate": 1.8949364854534054e-05, + "loss": 1.2705, + "step": 7186 + }, + { + "epoch": 0.17, + "learning_rate": 1.8949017289729728e-05, + "loss": 1.2742, + "step": 7187 + }, + { + "epoch": 0.17, + "learning_rate": 1.8948669670633812e-05, + "loss": 0.7574, + "step": 7188 + }, + { + "epoch": 0.17, + "learning_rate": 1.8948321997248423e-05, + "loss": 1.1966, + "step": 7189 + }, + { + "epoch": 0.17, + "learning_rate": 1.894797426957567e-05, + "loss": 1.3261, + "step": 7190 + }, + { + "epoch": 0.17, + "learning_rate": 1.8947626487617663e-05, + "loss": 1.0293, + "step": 7191 + }, + { + "epoch": 0.17, + "learning_rate": 1.894727865137651e-05, + "loss": 1.2906, + "step": 7192 + }, + { + "epoch": 0.17, + "learning_rate": 1.894693076085432e-05, + "loss": 1.2143, + "step": 7193 + }, + { + "epoch": 0.17, + "learning_rate": 1.894658281605321e-05, + "loss": 1.0696, + "step": 7194 + }, + { + "epoch": 0.17, + "learning_rate": 1.8946234816975283e-05, + "loss": 1.0934, + "step": 7195 + }, + { + "epoch": 0.17, + "learning_rate": 1.894588676362265e-05, + "loss": 1.2717, + "step": 7196 + }, + { + "epoch": 0.17, + "learning_rate": 1.8945538655997433e-05, + "loss": 1.1491, + "step": 7197 + }, + { + "epoch": 0.17, + "learning_rate": 1.8945190494101734e-05, + "loss": 1.2289, + "step": 7198 + }, + { + "epoch": 0.17, + "learning_rate": 1.894484227793767e-05, + "loss": 1.0968, + "step": 7199 + }, + { + "epoch": 0.17, + "learning_rate": 1.894449400750735e-05, + "loss": 0.778, + "step": 7200 + }, + { + "epoch": 0.17, + "learning_rate": 1.894414568281289e-05, + "loss": 1.3327, + "step": 7201 + }, + { + "epoch": 0.17, + "learning_rate": 1.89437973038564e-05, + "loss": 1.5287, + "step": 7202 + }, + { + "epoch": 0.17, + "learning_rate": 1.8943448870639996e-05, + "loss": 1.2067, + "step": 7203 + }, + { + "epoch": 0.17, + "learning_rate": 1.8943100383165794e-05, + "loss": 1.1198, + "step": 7204 + }, + { + "epoch": 0.17, + "learning_rate": 1.8942751841435903e-05, + "loss": 1.0134, + "step": 7205 + }, + { + "epoch": 0.17, + "learning_rate": 1.894240324545244e-05, + "loss": 1.2027, + "step": 7206 + }, + { + "epoch": 0.17, + "learning_rate": 1.894205459521752e-05, + "loss": 1.1393, + "step": 7207 + }, + { + "epoch": 0.17, + "learning_rate": 1.894170589073326e-05, + "loss": 1.2032, + "step": 7208 + }, + { + "epoch": 0.17, + "learning_rate": 1.894135713200177e-05, + "loss": 1.0188, + "step": 7209 + }, + { + "epoch": 0.17, + "learning_rate": 1.8941008319025174e-05, + "loss": 1.0866, + "step": 7210 + }, + { + "epoch": 0.17, + "learning_rate": 1.894065945180558e-05, + "loss": 1.4068, + "step": 7211 + }, + { + "epoch": 0.17, + "learning_rate": 1.894031053034511e-05, + "loss": 1.1463, + "step": 7212 + }, + { + "epoch": 0.17, + "learning_rate": 1.893996155464588e-05, + "loss": 1.0026, + "step": 7213 + }, + { + "epoch": 0.17, + "learning_rate": 1.8939612524710003e-05, + "loss": 1.1965, + "step": 7214 + }, + { + "epoch": 0.17, + "learning_rate": 1.8939263440539602e-05, + "loss": 1.2573, + "step": 7215 + }, + { + "epoch": 0.17, + "learning_rate": 1.8938914302136792e-05, + "loss": 1.0433, + "step": 7216 + }, + { + "epoch": 0.17, + "learning_rate": 1.8938565109503687e-05, + "loss": 0.9979, + "step": 7217 + }, + { + "epoch": 0.17, + "learning_rate": 1.893821586264241e-05, + "loss": 1.0273, + "step": 7218 + }, + { + "epoch": 0.17, + "learning_rate": 1.8937866561555086e-05, + "loss": 0.916, + "step": 7219 + }, + { + "epoch": 0.17, + "learning_rate": 1.8937517206243828e-05, + "loss": 1.1024, + "step": 7220 + }, + { + "epoch": 0.17, + "learning_rate": 1.8937167796710752e-05, + "loss": 0.9741, + "step": 7221 + }, + { + "epoch": 0.17, + "learning_rate": 1.8936818332957983e-05, + "loss": 1.2474, + "step": 7222 + }, + { + "epoch": 0.17, + "learning_rate": 1.8936468814987637e-05, + "loss": 1.2192, + "step": 7223 + }, + { + "epoch": 0.17, + "learning_rate": 1.8936119242801837e-05, + "loss": 0.9855, + "step": 7224 + }, + { + "epoch": 0.17, + "learning_rate": 1.8935769616402702e-05, + "loss": 1.0841, + "step": 7225 + }, + { + "epoch": 0.17, + "learning_rate": 1.8935419935792358e-05, + "loss": 0.9832, + "step": 7226 + }, + { + "epoch": 0.17, + "learning_rate": 1.893507020097292e-05, + "loss": 1.4464, + "step": 7227 + }, + { + "epoch": 0.17, + "learning_rate": 1.8934720411946513e-05, + "loss": 1.369, + "step": 7228 + }, + { + "epoch": 0.17, + "learning_rate": 1.8934370568715262e-05, + "loss": 1.2178, + "step": 7229 + }, + { + "epoch": 0.17, + "learning_rate": 1.8934020671281285e-05, + "loss": 1.0486, + "step": 7230 + }, + { + "epoch": 0.17, + "learning_rate": 1.8933670719646706e-05, + "loss": 1.2198, + "step": 7231 + }, + { + "epoch": 0.17, + "learning_rate": 1.8933320713813647e-05, + "loss": 0.9901, + "step": 7232 + }, + { + "epoch": 0.17, + "learning_rate": 1.893297065378423e-05, + "loss": 1.0526, + "step": 7233 + }, + { + "epoch": 0.17, + "learning_rate": 1.8932620539560587e-05, + "loss": 1.2305, + "step": 7234 + }, + { + "epoch": 0.17, + "learning_rate": 1.8932270371144833e-05, + "loss": 1.2228, + "step": 7235 + }, + { + "epoch": 0.17, + "learning_rate": 1.8931920148539098e-05, + "loss": 1.2693, + "step": 7236 + }, + { + "epoch": 0.17, + "learning_rate": 1.89315698717455e-05, + "loss": 1.2537, + "step": 7237 + }, + { + "epoch": 0.17, + "learning_rate": 1.8931219540766174e-05, + "loss": 1.3028, + "step": 7238 + }, + { + "epoch": 0.17, + "learning_rate": 1.8930869155603236e-05, + "loss": 1.2291, + "step": 7239 + }, + { + "epoch": 0.17, + "learning_rate": 1.8930518716258816e-05, + "loss": 1.3456, + "step": 7240 + }, + { + "epoch": 0.17, + "learning_rate": 1.8930168222735038e-05, + "loss": 0.872, + "step": 7241 + }, + { + "epoch": 0.17, + "learning_rate": 1.8929817675034033e-05, + "loss": 1.3726, + "step": 7242 + }, + { + "epoch": 0.17, + "learning_rate": 1.8929467073157923e-05, + "loss": 1.4206, + "step": 7243 + }, + { + "epoch": 0.17, + "learning_rate": 1.8929116417108837e-05, + "loss": 1.2084, + "step": 7244 + }, + { + "epoch": 0.17, + "learning_rate": 1.89287657068889e-05, + "loss": 0.8182, + "step": 7245 + }, + { + "epoch": 0.17, + "learning_rate": 1.8928414942500243e-05, + "loss": 1.2323, + "step": 7246 + }, + { + "epoch": 0.17, + "learning_rate": 1.8928064123944994e-05, + "loss": 1.1696, + "step": 7247 + }, + { + "epoch": 0.17, + "learning_rate": 1.892771325122528e-05, + "loss": 1.2001, + "step": 7248 + }, + { + "epoch": 0.17, + "learning_rate": 1.8927362324343227e-05, + "loss": 1.2499, + "step": 7249 + }, + { + "epoch": 0.17, + "learning_rate": 1.892701134330097e-05, + "loss": 0.9826, + "step": 7250 + }, + { + "epoch": 0.17, + "learning_rate": 1.8926660308100632e-05, + "loss": 1.1292, + "step": 7251 + }, + { + "epoch": 0.17, + "learning_rate": 1.8926309218744348e-05, + "loss": 0.9542, + "step": 7252 + }, + { + "epoch": 0.17, + "learning_rate": 1.8925958075234242e-05, + "loss": 0.7851, + "step": 7253 + }, + { + "epoch": 0.17, + "learning_rate": 1.892560687757245e-05, + "loss": 1.2207, + "step": 7254 + }, + { + "epoch": 0.17, + "learning_rate": 1.8925255625761104e-05, + "loss": 1.2449, + "step": 7255 + }, + { + "epoch": 0.17, + "learning_rate": 1.8924904319802327e-05, + "loss": 1.1866, + "step": 7256 + }, + { + "epoch": 0.17, + "learning_rate": 1.8924552959698257e-05, + "loss": 0.9782, + "step": 7257 + }, + { + "epoch": 0.17, + "learning_rate": 1.892420154545102e-05, + "loss": 0.9386, + "step": 7258 + }, + { + "epoch": 0.17, + "learning_rate": 1.8923850077062756e-05, + "loss": 1.2166, + "step": 7259 + }, + { + "epoch": 0.17, + "learning_rate": 1.892349855453559e-05, + "loss": 0.9743, + "step": 7260 + }, + { + "epoch": 0.17, + "learning_rate": 1.8923146977871658e-05, + "loss": 1.0651, + "step": 7261 + }, + { + "epoch": 0.17, + "learning_rate": 1.892279534707309e-05, + "loss": 1.2979, + "step": 7262 + }, + { + "epoch": 0.17, + "learning_rate": 1.8922443662142024e-05, + "loss": 1.2144, + "step": 7263 + }, + { + "epoch": 0.17, + "learning_rate": 1.892209192308059e-05, + "loss": 1.1642, + "step": 7264 + }, + { + "epoch": 0.17, + "learning_rate": 1.8921740129890925e-05, + "loss": 1.1612, + "step": 7265 + }, + { + "epoch": 0.17, + "learning_rate": 1.8921388282575157e-05, + "loss": 1.2431, + "step": 7266 + }, + { + "epoch": 0.17, + "learning_rate": 1.8921036381135427e-05, + "loss": 1.4224, + "step": 7267 + }, + { + "epoch": 0.17, + "learning_rate": 1.8920684425573865e-05, + "loss": 1.3566, + "step": 7268 + }, + { + "epoch": 0.17, + "learning_rate": 1.892033241589261e-05, + "loss": 1.1344, + "step": 7269 + }, + { + "epoch": 0.17, + "learning_rate": 1.8919980352093802e-05, + "loss": 1.2079, + "step": 7270 + }, + { + "epoch": 0.17, + "learning_rate": 1.8919628234179566e-05, + "loss": 1.3249, + "step": 7271 + }, + { + "epoch": 0.17, + "learning_rate": 1.8919276062152043e-05, + "loss": 1.1848, + "step": 7272 + }, + { + "epoch": 0.17, + "learning_rate": 1.8918923836013373e-05, + "loss": 1.0517, + "step": 7273 + }, + { + "epoch": 0.17, + "learning_rate": 1.891857155576569e-05, + "loss": 1.1866, + "step": 7274 + }, + { + "epoch": 0.17, + "learning_rate": 1.8918219221411126e-05, + "loss": 1.1743, + "step": 7275 + }, + { + "epoch": 0.17, + "learning_rate": 1.891786683295183e-05, + "loss": 1.3616, + "step": 7276 + }, + { + "epoch": 0.18, + "learning_rate": 1.891751439038993e-05, + "loss": 1.2926, + "step": 7277 + }, + { + "epoch": 0.18, + "learning_rate": 1.891716189372757e-05, + "loss": 0.9133, + "step": 7278 + }, + { + "epoch": 0.18, + "learning_rate": 1.8916809342966886e-05, + "loss": 1.0471, + "step": 7279 + }, + { + "epoch": 0.18, + "learning_rate": 1.8916456738110013e-05, + "loss": 1.0668, + "step": 7280 + }, + { + "epoch": 0.18, + "learning_rate": 1.8916104079159096e-05, + "loss": 1.3084, + "step": 7281 + }, + { + "epoch": 0.18, + "learning_rate": 1.8915751366116275e-05, + "loss": 1.2921, + "step": 7282 + }, + { + "epoch": 0.18, + "learning_rate": 1.891539859898369e-05, + "loss": 0.9622, + "step": 7283 + }, + { + "epoch": 0.18, + "learning_rate": 1.891504577776347e-05, + "loss": 1.0229, + "step": 7284 + }, + { + "epoch": 0.18, + "learning_rate": 1.891469290245777e-05, + "loss": 0.9074, + "step": 7285 + }, + { + "epoch": 0.18, + "learning_rate": 1.8914339973068725e-05, + "loss": 1.0213, + "step": 7286 + }, + { + "epoch": 0.18, + "learning_rate": 1.8913986989598476e-05, + "loss": 1.0381, + "step": 7287 + }, + { + "epoch": 0.18, + "learning_rate": 1.8913633952049168e-05, + "loss": 1.3408, + "step": 7288 + }, + { + "epoch": 0.18, + "learning_rate": 1.8913280860422935e-05, + "loss": 1.0939, + "step": 7289 + }, + { + "epoch": 0.18, + "learning_rate": 1.8912927714721922e-05, + "loss": 1.1372, + "step": 7290 + }, + { + "epoch": 0.18, + "learning_rate": 1.8912574514948274e-05, + "loss": 1.1702, + "step": 7291 + }, + { + "epoch": 0.18, + "learning_rate": 1.8912221261104136e-05, + "loss": 1.1661, + "step": 7292 + }, + { + "epoch": 0.18, + "learning_rate": 1.8911867953191648e-05, + "loss": 1.4077, + "step": 7293 + }, + { + "epoch": 0.18, + "learning_rate": 1.891151459121295e-05, + "loss": 1.2576, + "step": 7294 + }, + { + "epoch": 0.18, + "learning_rate": 1.891116117517019e-05, + "loss": 1.1379, + "step": 7295 + }, + { + "epoch": 0.18, + "learning_rate": 1.8910807705065514e-05, + "loss": 1.3185, + "step": 7296 + }, + { + "epoch": 0.18, + "learning_rate": 1.891045418090106e-05, + "loss": 1.156, + "step": 7297 + }, + { + "epoch": 0.18, + "learning_rate": 1.8910100602678976e-05, + "loss": 1.1253, + "step": 7298 + }, + { + "epoch": 0.18, + "learning_rate": 1.8909746970401408e-05, + "loss": 1.0869, + "step": 7299 + }, + { + "epoch": 0.18, + "learning_rate": 1.89093932840705e-05, + "loss": 0.9303, + "step": 7300 + }, + { + "epoch": 0.18, + "learning_rate": 1.8909039543688404e-05, + "loss": 1.0798, + "step": 7301 + }, + { + "epoch": 0.18, + "learning_rate": 1.8908685749257257e-05, + "loss": 0.8744, + "step": 7302 + }, + { + "epoch": 0.18, + "learning_rate": 1.890833190077921e-05, + "loss": 1.1266, + "step": 7303 + }, + { + "epoch": 0.18, + "learning_rate": 1.8907977998256405e-05, + "loss": 1.2468, + "step": 7304 + }, + { + "epoch": 0.18, + "learning_rate": 1.8907624041690996e-05, + "loss": 1.2534, + "step": 7305 + }, + { + "epoch": 0.18, + "learning_rate": 1.890727003108513e-05, + "loss": 0.9448, + "step": 7306 + }, + { + "epoch": 0.18, + "learning_rate": 1.8906915966440945e-05, + "loss": 1.0659, + "step": 7307 + }, + { + "epoch": 0.18, + "learning_rate": 1.89065618477606e-05, + "loss": 1.1757, + "step": 7308 + }, + { + "epoch": 0.18, + "learning_rate": 1.890620767504624e-05, + "loss": 1.235, + "step": 7309 + }, + { + "epoch": 0.18, + "learning_rate": 1.890585344830001e-05, + "loss": 1.0755, + "step": 7310 + }, + { + "epoch": 0.18, + "learning_rate": 1.8905499167524063e-05, + "loss": 1.1687, + "step": 7311 + }, + { + "epoch": 0.18, + "learning_rate": 1.8905144832720547e-05, + "loss": 1.2037, + "step": 7312 + }, + { + "epoch": 0.18, + "learning_rate": 1.8904790443891612e-05, + "loss": 1.2126, + "step": 7313 + }, + { + "epoch": 0.18, + "learning_rate": 1.8904436001039408e-05, + "loss": 1.2581, + "step": 7314 + }, + { + "epoch": 0.18, + "learning_rate": 1.8904081504166084e-05, + "loss": 1.2054, + "step": 7315 + }, + { + "epoch": 0.18, + "learning_rate": 1.8903726953273794e-05, + "loss": 1.1658, + "step": 7316 + }, + { + "epoch": 0.18, + "learning_rate": 1.8903372348364688e-05, + "loss": 1.1961, + "step": 7317 + }, + { + "epoch": 0.18, + "learning_rate": 1.8903017689440914e-05, + "loss": 1.2479, + "step": 7318 + }, + { + "epoch": 0.18, + "learning_rate": 1.8902662976504627e-05, + "loss": 0.8631, + "step": 7319 + }, + { + "epoch": 0.18, + "learning_rate": 1.8902308209557976e-05, + "loss": 1.0881, + "step": 7320 + }, + { + "epoch": 0.18, + "learning_rate": 1.890195338860312e-05, + "loss": 1.1075, + "step": 7321 + }, + { + "epoch": 0.18, + "learning_rate": 1.89015985136422e-05, + "loss": 1.3082, + "step": 7322 + }, + { + "epoch": 0.18, + "learning_rate": 1.890124358467738e-05, + "loss": 1.1975, + "step": 7323 + }, + { + "epoch": 0.18, + "learning_rate": 1.8900888601710804e-05, + "loss": 0.9187, + "step": 7324 + }, + { + "epoch": 0.18, + "learning_rate": 1.8900533564744635e-05, + "loss": 1.2616, + "step": 7325 + }, + { + "epoch": 0.18, + "learning_rate": 1.890017847378102e-05, + "loss": 1.2533, + "step": 7326 + }, + { + "epoch": 0.18, + "learning_rate": 1.8899823328822118e-05, + "loss": 1.0556, + "step": 7327 + }, + { + "epoch": 0.18, + "learning_rate": 1.889946812987008e-05, + "loss": 1.2141, + "step": 7328 + }, + { + "epoch": 0.18, + "learning_rate": 1.8899112876927064e-05, + "loss": 1.0798, + "step": 7329 + }, + { + "epoch": 0.18, + "learning_rate": 1.8898757569995218e-05, + "loss": 1.2469, + "step": 7330 + }, + { + "epoch": 0.18, + "learning_rate": 1.889840220907671e-05, + "loss": 1.0669, + "step": 7331 + }, + { + "epoch": 0.18, + "learning_rate": 1.8898046794173684e-05, + "loss": 0.8215, + "step": 7332 + }, + { + "epoch": 0.18, + "learning_rate": 1.88976913252883e-05, + "loss": 1.0722, + "step": 7333 + }, + { + "epoch": 0.18, + "learning_rate": 1.889733580242272e-05, + "loss": 1.0519, + "step": 7334 + }, + { + "epoch": 0.18, + "learning_rate": 1.8896980225579095e-05, + "loss": 1.0727, + "step": 7335 + }, + { + "epoch": 0.18, + "learning_rate": 1.8896624594759583e-05, + "loss": 1.1309, + "step": 7336 + }, + { + "epoch": 0.18, + "learning_rate": 1.8896268909966343e-05, + "loss": 1.0739, + "step": 7337 + }, + { + "epoch": 0.18, + "learning_rate": 1.889591317120153e-05, + "loss": 1.1256, + "step": 7338 + }, + { + "epoch": 0.18, + "learning_rate": 1.8895557378467306e-05, + "loss": 1.1783, + "step": 7339 + }, + { + "epoch": 0.18, + "learning_rate": 1.889520153176583e-05, + "loss": 1.2679, + "step": 7340 + }, + { + "epoch": 0.18, + "learning_rate": 1.8894845631099255e-05, + "loss": 1.2919, + "step": 7341 + }, + { + "epoch": 0.18, + "learning_rate": 1.889448967646974e-05, + "loss": 1.0936, + "step": 7342 + }, + { + "epoch": 0.18, + "learning_rate": 1.8894133667879458e-05, + "loss": 1.1425, + "step": 7343 + }, + { + "epoch": 0.18, + "learning_rate": 1.8893777605330553e-05, + "loss": 1.1084, + "step": 7344 + }, + { + "epoch": 0.18, + "learning_rate": 1.889342148882519e-05, + "loss": 0.9154, + "step": 7345 + }, + { + "epoch": 0.18, + "learning_rate": 1.8893065318365534e-05, + "loss": 1.3649, + "step": 7346 + }, + { + "epoch": 0.18, + "learning_rate": 1.889270909395374e-05, + "loss": 1.0865, + "step": 7347 + }, + { + "epoch": 0.18, + "learning_rate": 1.8892352815591973e-05, + "loss": 0.8815, + "step": 7348 + }, + { + "epoch": 0.18, + "learning_rate": 1.8891996483282394e-05, + "loss": 1.0314, + "step": 7349 + }, + { + "epoch": 0.18, + "learning_rate": 1.8891640097027163e-05, + "loss": 1.1298, + "step": 7350 + }, + { + "epoch": 0.18, + "learning_rate": 1.8891283656828443e-05, + "loss": 1.3167, + "step": 7351 + }, + { + "epoch": 0.18, + "learning_rate": 1.8890927162688398e-05, + "loss": 0.9409, + "step": 7352 + }, + { + "epoch": 0.18, + "learning_rate": 1.8890570614609186e-05, + "loss": 1.0696, + "step": 7353 + }, + { + "epoch": 0.18, + "learning_rate": 1.8890214012592977e-05, + "loss": 0.9309, + "step": 7354 + }, + { + "epoch": 0.18, + "learning_rate": 1.8889857356641924e-05, + "loss": 1.1973, + "step": 7355 + }, + { + "epoch": 0.18, + "learning_rate": 1.8889500646758203e-05, + "loss": 1.3703, + "step": 7356 + }, + { + "epoch": 0.18, + "learning_rate": 1.8889143882943972e-05, + "loss": 1.242, + "step": 7357 + }, + { + "epoch": 0.18, + "learning_rate": 1.8888787065201393e-05, + "loss": 1.2375, + "step": 7358 + }, + { + "epoch": 0.18, + "learning_rate": 1.8888430193532638e-05, + "loss": 1.3278, + "step": 7359 + }, + { + "epoch": 0.18, + "learning_rate": 1.8888073267939865e-05, + "loss": 0.9872, + "step": 7360 + }, + { + "epoch": 0.18, + "learning_rate": 1.888771628842524e-05, + "loss": 1.184, + "step": 7361 + }, + { + "epoch": 0.18, + "learning_rate": 1.8887359254990937e-05, + "loss": 1.044, + "step": 7362 + }, + { + "epoch": 0.18, + "learning_rate": 1.8887002167639112e-05, + "loss": 1.2426, + "step": 7363 + }, + { + "epoch": 0.18, + "learning_rate": 1.8886645026371937e-05, + "loss": 1.1519, + "step": 7364 + }, + { + "epoch": 0.18, + "learning_rate": 1.8886287831191575e-05, + "loss": 1.0736, + "step": 7365 + }, + { + "epoch": 0.18, + "learning_rate": 1.8885930582100195e-05, + "loss": 1.1989, + "step": 7366 + }, + { + "epoch": 0.18, + "learning_rate": 1.8885573279099965e-05, + "loss": 1.0943, + "step": 7367 + }, + { + "epoch": 0.18, + "learning_rate": 1.888521592219305e-05, + "loss": 1.1614, + "step": 7368 + }, + { + "epoch": 0.18, + "learning_rate": 1.888485851138162e-05, + "loss": 1.3778, + "step": 7369 + }, + { + "epoch": 0.18, + "learning_rate": 1.8884501046667847e-05, + "loss": 1.009, + "step": 7370 + }, + { + "epoch": 0.18, + "learning_rate": 1.8884143528053894e-05, + "loss": 1.0933, + "step": 7371 + }, + { + "epoch": 0.18, + "learning_rate": 1.888378595554193e-05, + "loss": 1.052, + "step": 7372 + }, + { + "epoch": 0.18, + "learning_rate": 1.888342832913413e-05, + "loss": 1.1076, + "step": 7373 + }, + { + "epoch": 0.18, + "learning_rate": 1.888307064883266e-05, + "loss": 1.206, + "step": 7374 + }, + { + "epoch": 0.18, + "learning_rate": 1.8882712914639682e-05, + "loss": 1.2686, + "step": 7375 + }, + { + "epoch": 0.18, + "learning_rate": 1.8882355126557382e-05, + "loss": 1.3325, + "step": 7376 + }, + { + "epoch": 0.18, + "learning_rate": 1.8881997284587917e-05, + "loss": 1.3259, + "step": 7377 + }, + { + "epoch": 0.18, + "learning_rate": 1.8881639388733468e-05, + "loss": 1.2287, + "step": 7378 + }, + { + "epoch": 0.18, + "learning_rate": 1.8881281438996204e-05, + "loss": 1.2216, + "step": 7379 + }, + { + "epoch": 0.18, + "learning_rate": 1.888092343537829e-05, + "loss": 1.0623, + "step": 7380 + }, + { + "epoch": 0.18, + "learning_rate": 1.8880565377881904e-05, + "loss": 1.1648, + "step": 7381 + }, + { + "epoch": 0.18, + "learning_rate": 1.8880207266509215e-05, + "loss": 1.1503, + "step": 7382 + }, + { + "epoch": 0.18, + "learning_rate": 1.8879849101262398e-05, + "loss": 1.1832, + "step": 7383 + }, + { + "epoch": 0.18, + "learning_rate": 1.887949088214363e-05, + "loss": 1.2092, + "step": 7384 + }, + { + "epoch": 0.18, + "learning_rate": 1.887913260915507e-05, + "loss": 1.2529, + "step": 7385 + }, + { + "epoch": 0.18, + "learning_rate": 1.887877428229891e-05, + "loss": 1.0471, + "step": 7386 + }, + { + "epoch": 0.18, + "learning_rate": 1.887841590157731e-05, + "loss": 1.0236, + "step": 7387 + }, + { + "epoch": 0.18, + "learning_rate": 1.887805746699245e-05, + "loss": 1.2931, + "step": 7388 + }, + { + "epoch": 0.18, + "learning_rate": 1.8877698978546503e-05, + "loss": 1.2315, + "step": 7389 + }, + { + "epoch": 0.18, + "learning_rate": 1.8877340436241645e-05, + "loss": 1.3548, + "step": 7390 + }, + { + "epoch": 0.18, + "learning_rate": 1.8876981840080054e-05, + "loss": 0.9938, + "step": 7391 + }, + { + "epoch": 0.18, + "learning_rate": 1.8876623190063898e-05, + "loss": 1.1767, + "step": 7392 + }, + { + "epoch": 0.18, + "learning_rate": 1.8876264486195356e-05, + "loss": 1.0413, + "step": 7393 + }, + { + "epoch": 0.18, + "learning_rate": 1.887590572847661e-05, + "loss": 1.1472, + "step": 7394 + }, + { + "epoch": 0.18, + "learning_rate": 1.8875546916909826e-05, + "loss": 1.2544, + "step": 7395 + }, + { + "epoch": 0.18, + "learning_rate": 1.887518805149719e-05, + "loss": 1.1536, + "step": 7396 + }, + { + "epoch": 0.18, + "learning_rate": 1.8874829132240874e-05, + "loss": 1.0114, + "step": 7397 + }, + { + "epoch": 0.18, + "learning_rate": 1.887447015914306e-05, + "loss": 0.8818, + "step": 7398 + }, + { + "epoch": 0.18, + "learning_rate": 1.887411113220592e-05, + "loss": 1.0707, + "step": 7399 + }, + { + "epoch": 0.18, + "learning_rate": 1.8873752051431635e-05, + "loss": 1.058, + "step": 7400 + }, + { + "epoch": 0.18, + "learning_rate": 1.8873392916822387e-05, + "loss": 1.0658, + "step": 7401 + }, + { + "epoch": 0.18, + "learning_rate": 1.887303372838035e-05, + "loss": 1.3013, + "step": 7402 + }, + { + "epoch": 0.18, + "learning_rate": 1.88726744861077e-05, + "loss": 1.0764, + "step": 7403 + }, + { + "epoch": 0.18, + "learning_rate": 1.8872315190006624e-05, + "loss": 0.9945, + "step": 7404 + }, + { + "epoch": 0.18, + "learning_rate": 1.88719558400793e-05, + "loss": 1.4043, + "step": 7405 + }, + { + "epoch": 0.18, + "learning_rate": 1.8871596436327908e-05, + "loss": 1.211, + "step": 7406 + }, + { + "epoch": 0.18, + "learning_rate": 1.8871236978754622e-05, + "loss": 1.2828, + "step": 7407 + }, + { + "epoch": 0.18, + "learning_rate": 1.8870877467361633e-05, + "loss": 0.8602, + "step": 7408 + }, + { + "epoch": 0.18, + "learning_rate": 1.8870517902151114e-05, + "loss": 1.1846, + "step": 7409 + }, + { + "epoch": 0.18, + "learning_rate": 1.887015828312525e-05, + "loss": 0.9991, + "step": 7410 + }, + { + "epoch": 0.18, + "learning_rate": 1.8869798610286223e-05, + "loss": 1.3315, + "step": 7411 + }, + { + "epoch": 0.18, + "learning_rate": 1.8869438883636212e-05, + "loss": 1.0799, + "step": 7412 + }, + { + "epoch": 0.18, + "learning_rate": 1.8869079103177403e-05, + "loss": 1.1432, + "step": 7413 + }, + { + "epoch": 0.18, + "learning_rate": 1.8868719268911978e-05, + "loss": 1.1155, + "step": 7414 + }, + { + "epoch": 0.18, + "learning_rate": 1.886835938084212e-05, + "loss": 1.1298, + "step": 7415 + }, + { + "epoch": 0.18, + "learning_rate": 1.8867999438970007e-05, + "loss": 1.1204, + "step": 7416 + }, + { + "epoch": 0.18, + "learning_rate": 1.8867639443297832e-05, + "loss": 1.2727, + "step": 7417 + }, + { + "epoch": 0.18, + "learning_rate": 1.886727939382777e-05, + "loss": 1.2528, + "step": 7418 + }, + { + "epoch": 0.18, + "learning_rate": 1.8866919290562014e-05, + "loss": 1.1388, + "step": 7419 + }, + { + "epoch": 0.18, + "learning_rate": 1.886655913350274e-05, + "loss": 1.1834, + "step": 7420 + }, + { + "epoch": 0.18, + "learning_rate": 1.8866198922652143e-05, + "loss": 1.2228, + "step": 7421 + }, + { + "epoch": 0.18, + "learning_rate": 1.8865838658012397e-05, + "loss": 0.8533, + "step": 7422 + }, + { + "epoch": 0.18, + "learning_rate": 1.8865478339585697e-05, + "loss": 1.1909, + "step": 7423 + }, + { + "epoch": 0.18, + "learning_rate": 1.8865117967374225e-05, + "loss": 1.1486, + "step": 7424 + }, + { + "epoch": 0.18, + "learning_rate": 1.8864757541380165e-05, + "loss": 0.7253, + "step": 7425 + }, + { + "epoch": 0.18, + "learning_rate": 1.886439706160571e-05, + "loss": 1.4463, + "step": 7426 + }, + { + "epoch": 0.18, + "learning_rate": 1.8864036528053037e-05, + "loss": 1.207, + "step": 7427 + }, + { + "epoch": 0.18, + "learning_rate": 1.8863675940724344e-05, + "loss": 1.2057, + "step": 7428 + }, + { + "epoch": 0.18, + "learning_rate": 1.8863315299621812e-05, + "loss": 1.1764, + "step": 7429 + }, + { + "epoch": 0.18, + "learning_rate": 1.886295460474763e-05, + "loss": 0.9461, + "step": 7430 + }, + { + "epoch": 0.18, + "learning_rate": 1.886259385610399e-05, + "loss": 1.0402, + "step": 7431 + }, + { + "epoch": 0.18, + "learning_rate": 1.8862233053693074e-05, + "loss": 1.3162, + "step": 7432 + }, + { + "epoch": 0.18, + "learning_rate": 1.886187219751708e-05, + "loss": 1.3487, + "step": 7433 + }, + { + "epoch": 0.18, + "learning_rate": 1.8861511287578188e-05, + "loss": 1.2013, + "step": 7434 + }, + { + "epoch": 0.18, + "learning_rate": 1.8861150323878594e-05, + "loss": 1.0508, + "step": 7435 + }, + { + "epoch": 0.18, + "learning_rate": 1.8860789306420478e-05, + "loss": 1.1615, + "step": 7436 + }, + { + "epoch": 0.18, + "learning_rate": 1.8860428235206047e-05, + "loss": 1.0301, + "step": 7437 + }, + { + "epoch": 0.18, + "learning_rate": 1.8860067110237478e-05, + "loss": 1.0379, + "step": 7438 + }, + { + "epoch": 0.18, + "learning_rate": 1.8859705931516964e-05, + "loss": 0.9297, + "step": 7439 + }, + { + "epoch": 0.18, + "learning_rate": 1.88593446990467e-05, + "loss": 1.1671, + "step": 7440 + }, + { + "epoch": 0.18, + "learning_rate": 1.8858983412828877e-05, + "loss": 1.254, + "step": 7441 + }, + { + "epoch": 0.18, + "learning_rate": 1.885862207286568e-05, + "loss": 0.8738, + "step": 7442 + }, + { + "epoch": 0.18, + "learning_rate": 1.885826067915931e-05, + "loss": 0.8058, + "step": 7443 + }, + { + "epoch": 0.18, + "learning_rate": 1.8857899231711956e-05, + "loss": 1.1897, + "step": 7444 + }, + { + "epoch": 0.18, + "learning_rate": 1.8857537730525813e-05, + "loss": 1.1058, + "step": 7445 + }, + { + "epoch": 0.18, + "learning_rate": 1.885717617560307e-05, + "loss": 1.2619, + "step": 7446 + }, + { + "epoch": 0.18, + "learning_rate": 1.885681456694592e-05, + "loss": 1.1012, + "step": 7447 + }, + { + "epoch": 0.18, + "learning_rate": 1.8856452904556564e-05, + "loss": 1.3386, + "step": 7448 + }, + { + "epoch": 0.18, + "learning_rate": 1.885609118843719e-05, + "loss": 1.16, + "step": 7449 + }, + { + "epoch": 0.18, + "learning_rate": 1.8855729418589994e-05, + "loss": 1.0599, + "step": 7450 + }, + { + "epoch": 0.18, + "learning_rate": 1.8855367595017168e-05, + "loss": 1.236, + "step": 7451 + }, + { + "epoch": 0.18, + "learning_rate": 1.885500571772091e-05, + "loss": 1.3587, + "step": 7452 + }, + { + "epoch": 0.18, + "learning_rate": 1.8854643786703417e-05, + "loss": 1.1272, + "step": 7453 + }, + { + "epoch": 0.18, + "learning_rate": 1.8854281801966886e-05, + "loss": 1.138, + "step": 7454 + }, + { + "epoch": 0.18, + "learning_rate": 1.8853919763513504e-05, + "loss": 1.3992, + "step": 7455 + }, + { + "epoch": 0.18, + "learning_rate": 1.8853557671345477e-05, + "loss": 1.3474, + "step": 7456 + }, + { + "epoch": 0.18, + "learning_rate": 1.8853195525464997e-05, + "loss": 0.7988, + "step": 7457 + }, + { + "epoch": 0.18, + "learning_rate": 1.885283332587426e-05, + "loss": 1.4814, + "step": 7458 + }, + { + "epoch": 0.18, + "learning_rate": 1.885247107257547e-05, + "loss": 1.0337, + "step": 7459 + }, + { + "epoch": 0.18, + "learning_rate": 1.8852108765570814e-05, + "loss": 1.3432, + "step": 7460 + }, + { + "epoch": 0.18, + "learning_rate": 1.88517464048625e-05, + "loss": 1.2408, + "step": 7461 + }, + { + "epoch": 0.18, + "learning_rate": 1.8851383990452726e-05, + "loss": 1.1318, + "step": 7462 + }, + { + "epoch": 0.18, + "learning_rate": 1.885102152234368e-05, + "loss": 1.1507, + "step": 7463 + }, + { + "epoch": 0.18, + "learning_rate": 1.8850659000537575e-05, + "loss": 1.0351, + "step": 7464 + }, + { + "epoch": 0.18, + "learning_rate": 1.88502964250366e-05, + "loss": 1.2121, + "step": 7465 + }, + { + "epoch": 0.18, + "learning_rate": 1.884993379584296e-05, + "loss": 1.2029, + "step": 7466 + }, + { + "epoch": 0.18, + "learning_rate": 1.8849571112958854e-05, + "loss": 1.1858, + "step": 7467 + }, + { + "epoch": 0.18, + "learning_rate": 1.8849208376386477e-05, + "loss": 1.4289, + "step": 7468 + }, + { + "epoch": 0.18, + "learning_rate": 1.884884558612804e-05, + "loss": 1.2113, + "step": 7469 + }, + { + "epoch": 0.18, + "learning_rate": 1.8848482742185737e-05, + "loss": 1.1893, + "step": 7470 + }, + { + "epoch": 0.18, + "learning_rate": 1.8848119844561768e-05, + "loss": 1.1215, + "step": 7471 + }, + { + "epoch": 0.18, + "learning_rate": 1.8847756893258336e-05, + "loss": 1.2757, + "step": 7472 + }, + { + "epoch": 0.18, + "learning_rate": 1.8847393888277647e-05, + "loss": 1.156, + "step": 7473 + }, + { + "epoch": 0.18, + "learning_rate": 1.88470308296219e-05, + "loss": 1.1901, + "step": 7474 + }, + { + "epoch": 0.18, + "learning_rate": 1.8846667717293297e-05, + "loss": 1.0087, + "step": 7475 + }, + { + "epoch": 0.18, + "learning_rate": 1.884630455129404e-05, + "loss": 1.1818, + "step": 7476 + }, + { + "epoch": 0.18, + "learning_rate": 1.884594133162634e-05, + "loss": 1.0294, + "step": 7477 + }, + { + "epoch": 0.18, + "learning_rate": 1.8845578058292392e-05, + "loss": 0.9606, + "step": 7478 + }, + { + "epoch": 0.18, + "learning_rate": 1.8845214731294398e-05, + "loss": 1.079, + "step": 7479 + }, + { + "epoch": 0.18, + "learning_rate": 1.8844851350634573e-05, + "loss": 0.9072, + "step": 7480 + }, + { + "epoch": 0.18, + "learning_rate": 1.884448791631511e-05, + "loss": 0.7947, + "step": 7481 + }, + { + "epoch": 0.18, + "learning_rate": 1.8844124428338223e-05, + "loss": 1.3236, + "step": 7482 + }, + { + "epoch": 0.18, + "learning_rate": 1.8843760886706112e-05, + "loss": 1.1822, + "step": 7483 + }, + { + "epoch": 0.18, + "learning_rate": 1.8843397291420982e-05, + "loss": 0.862, + "step": 7484 + }, + { + "epoch": 0.18, + "learning_rate": 1.8843033642485044e-05, + "loss": 1.1017, + "step": 7485 + }, + { + "epoch": 0.18, + "learning_rate": 1.88426699399005e-05, + "loss": 1.2202, + "step": 7486 + }, + { + "epoch": 0.18, + "learning_rate": 1.8842306183669558e-05, + "loss": 1.1548, + "step": 7487 + }, + { + "epoch": 0.18, + "learning_rate": 1.884194237379442e-05, + "loss": 1.0343, + "step": 7488 + }, + { + "epoch": 0.18, + "learning_rate": 1.8841578510277302e-05, + "loss": 1.4019, + "step": 7489 + }, + { + "epoch": 0.18, + "learning_rate": 1.8841214593120405e-05, + "loss": 1.1858, + "step": 7490 + }, + { + "epoch": 0.18, + "learning_rate": 1.884085062232594e-05, + "loss": 1.2473, + "step": 7491 + }, + { + "epoch": 0.18, + "learning_rate": 1.8840486597896114e-05, + "loss": 0.8506, + "step": 7492 + }, + { + "epoch": 0.18, + "learning_rate": 1.8840122519833132e-05, + "loss": 1.2681, + "step": 7493 + }, + { + "epoch": 0.18, + "learning_rate": 1.8839758388139208e-05, + "loss": 1.1938, + "step": 7494 + }, + { + "epoch": 0.18, + "learning_rate": 1.883939420281655e-05, + "loss": 1.1003, + "step": 7495 + }, + { + "epoch": 0.18, + "learning_rate": 1.8839029963867362e-05, + "loss": 1.2684, + "step": 7496 + }, + { + "epoch": 0.18, + "learning_rate": 1.8838665671293866e-05, + "loss": 1.1123, + "step": 7497 + }, + { + "epoch": 0.18, + "learning_rate": 1.883830132509826e-05, + "loss": 1.0353, + "step": 7498 + }, + { + "epoch": 0.18, + "learning_rate": 1.883793692528276e-05, + "loss": 1.0368, + "step": 7499 + }, + { + "epoch": 0.18, + "learning_rate": 1.8837572471849574e-05, + "loss": 1.0526, + "step": 7500 + }, + { + "epoch": 0.18, + "learning_rate": 1.8837207964800914e-05, + "loss": 1.0863, + "step": 7501 + }, + { + "epoch": 0.18, + "learning_rate": 1.8836843404138992e-05, + "loss": 0.8394, + "step": 7502 + }, + { + "epoch": 0.18, + "learning_rate": 1.883647878986602e-05, + "loss": 1.013, + "step": 7503 + }, + { + "epoch": 0.18, + "learning_rate": 1.883611412198421e-05, + "loss": 0.9689, + "step": 7504 + }, + { + "epoch": 0.18, + "learning_rate": 1.8835749400495776e-05, + "loss": 0.9884, + "step": 7505 + }, + { + "epoch": 0.18, + "learning_rate": 1.8835384625402927e-05, + "loss": 1.2268, + "step": 7506 + }, + { + "epoch": 0.18, + "learning_rate": 1.8835019796707878e-05, + "loss": 1.0776, + "step": 7507 + }, + { + "epoch": 0.18, + "learning_rate": 1.883465491441284e-05, + "loss": 1.089, + "step": 7508 + }, + { + "epoch": 0.18, + "learning_rate": 1.8834289978520033e-05, + "loss": 1.1201, + "step": 7509 + }, + { + "epoch": 0.18, + "learning_rate": 1.8833924989031663e-05, + "loss": 1.0692, + "step": 7510 + }, + { + "epoch": 0.18, + "learning_rate": 1.883355994594995e-05, + "loss": 1.2871, + "step": 7511 + }, + { + "epoch": 0.18, + "learning_rate": 1.8833194849277105e-05, + "loss": 1.326, + "step": 7512 + }, + { + "epoch": 0.18, + "learning_rate": 1.8832829699015342e-05, + "loss": 1.2474, + "step": 7513 + }, + { + "epoch": 0.18, + "learning_rate": 1.8832464495166883e-05, + "loss": 1.1257, + "step": 7514 + }, + { + "epoch": 0.18, + "learning_rate": 1.8832099237733938e-05, + "loss": 1.2464, + "step": 7515 + }, + { + "epoch": 0.18, + "learning_rate": 1.8831733926718725e-05, + "loss": 1.1252, + "step": 7516 + }, + { + "epoch": 0.18, + "learning_rate": 1.883136856212346e-05, + "loss": 0.9669, + "step": 7517 + }, + { + "epoch": 0.18, + "learning_rate": 1.8831003143950357e-05, + "loss": 0.9136, + "step": 7518 + }, + { + "epoch": 0.18, + "learning_rate": 1.8830637672201637e-05, + "loss": 1.214, + "step": 7519 + }, + { + "epoch": 0.18, + "learning_rate": 1.8830272146879513e-05, + "loss": 1.1558, + "step": 7520 + }, + { + "epoch": 0.18, + "learning_rate": 1.882990656798621e-05, + "loss": 1.1664, + "step": 7521 + }, + { + "epoch": 0.18, + "learning_rate": 1.882954093552394e-05, + "loss": 1.1669, + "step": 7522 + }, + { + "epoch": 0.18, + "learning_rate": 1.8829175249494914e-05, + "loss": 1.2218, + "step": 7523 + }, + { + "epoch": 0.18, + "learning_rate": 1.8828809509901366e-05, + "loss": 1.0618, + "step": 7524 + }, + { + "epoch": 0.18, + "learning_rate": 1.8828443716745506e-05, + "loss": 0.9452, + "step": 7525 + }, + { + "epoch": 0.18, + "learning_rate": 1.8828077870029554e-05, + "loss": 1.0891, + "step": 7526 + }, + { + "epoch": 0.18, + "learning_rate": 1.882771196975573e-05, + "loss": 0.9591, + "step": 7527 + }, + { + "epoch": 0.18, + "learning_rate": 1.8827346015926253e-05, + "loss": 1.1774, + "step": 7528 + }, + { + "epoch": 0.18, + "learning_rate": 1.8826980008543343e-05, + "loss": 1.0724, + "step": 7529 + }, + { + "epoch": 0.18, + "learning_rate": 1.8826613947609225e-05, + "loss": 1.1875, + "step": 7530 + }, + { + "epoch": 0.18, + "learning_rate": 1.8826247833126118e-05, + "loss": 1.3468, + "step": 7531 + }, + { + "epoch": 0.18, + "learning_rate": 1.8825881665096237e-05, + "loss": 1.294, + "step": 7532 + }, + { + "epoch": 0.18, + "learning_rate": 1.8825515443521805e-05, + "loss": 1.3856, + "step": 7533 + }, + { + "epoch": 0.18, + "learning_rate": 1.8825149168405055e-05, + "loss": 1.1007, + "step": 7534 + }, + { + "epoch": 0.18, + "learning_rate": 1.8824782839748197e-05, + "loss": 1.0775, + "step": 7535 + }, + { + "epoch": 0.18, + "learning_rate": 1.8824416457553455e-05, + "loss": 1.1894, + "step": 7536 + }, + { + "epoch": 0.18, + "learning_rate": 1.8824050021823053e-05, + "loss": 1.2745, + "step": 7537 + }, + { + "epoch": 0.18, + "learning_rate": 1.882368353255922e-05, + "loss": 1.4699, + "step": 7538 + }, + { + "epoch": 0.18, + "learning_rate": 1.8823316989764172e-05, + "loss": 1.1876, + "step": 7539 + }, + { + "epoch": 0.18, + "learning_rate": 1.8822950393440135e-05, + "loss": 1.1864, + "step": 7540 + }, + { + "epoch": 0.18, + "learning_rate": 1.8822583743589335e-05, + "loss": 1.2361, + "step": 7541 + }, + { + "epoch": 0.18, + "learning_rate": 1.8822217040213994e-05, + "loss": 0.9592, + "step": 7542 + }, + { + "epoch": 0.18, + "learning_rate": 1.8821850283316336e-05, + "loss": 1.0158, + "step": 7543 + }, + { + "epoch": 0.18, + "learning_rate": 1.882148347289859e-05, + "loss": 1.3487, + "step": 7544 + }, + { + "epoch": 0.18, + "learning_rate": 1.8821116608962976e-05, + "loss": 1.2357, + "step": 7545 + }, + { + "epoch": 0.18, + "learning_rate": 1.8820749691511723e-05, + "loss": 0.9214, + "step": 7546 + }, + { + "epoch": 0.18, + "learning_rate": 1.8820382720547056e-05, + "loss": 1.4647, + "step": 7547 + }, + { + "epoch": 0.18, + "learning_rate": 1.8820015696071202e-05, + "loss": 1.2781, + "step": 7548 + }, + { + "epoch": 0.18, + "learning_rate": 1.881964861808639e-05, + "loss": 1.0865, + "step": 7549 + }, + { + "epoch": 0.18, + "learning_rate": 1.881928148659484e-05, + "loss": 0.9765, + "step": 7550 + }, + { + "epoch": 0.18, + "learning_rate": 1.8818914301598785e-05, + "loss": 1.3985, + "step": 7551 + }, + { + "epoch": 0.18, + "learning_rate": 1.8818547063100453e-05, + "loss": 1.2418, + "step": 7552 + }, + { + "epoch": 0.18, + "learning_rate": 1.881817977110207e-05, + "loss": 1.2221, + "step": 7553 + }, + { + "epoch": 0.18, + "learning_rate": 1.8817812425605864e-05, + "loss": 1.0047, + "step": 7554 + }, + { + "epoch": 0.18, + "learning_rate": 1.8817445026614063e-05, + "loss": 1.2176, + "step": 7555 + }, + { + "epoch": 0.18, + "learning_rate": 1.88170775741289e-05, + "loss": 1.1823, + "step": 7556 + }, + { + "epoch": 0.18, + "learning_rate": 1.8816710068152596e-05, + "loss": 1.19, + "step": 7557 + }, + { + "epoch": 0.18, + "learning_rate": 1.8816342508687388e-05, + "loss": 1.0226, + "step": 7558 + }, + { + "epoch": 0.18, + "learning_rate": 1.8815974895735505e-05, + "loss": 1.0148, + "step": 7559 + }, + { + "epoch": 0.18, + "learning_rate": 1.8815607229299177e-05, + "loss": 1.2011, + "step": 7560 + }, + { + "epoch": 0.18, + "learning_rate": 1.8815239509380633e-05, + "loss": 0.8898, + "step": 7561 + }, + { + "epoch": 0.18, + "learning_rate": 1.8814871735982102e-05, + "loss": 0.907, + "step": 7562 + }, + { + "epoch": 0.18, + "learning_rate": 1.8814503909105818e-05, + "loss": 1.0889, + "step": 7563 + }, + { + "epoch": 0.18, + "learning_rate": 1.8814136028754016e-05, + "loss": 0.788, + "step": 7564 + }, + { + "epoch": 0.18, + "learning_rate": 1.881376809492892e-05, + "loss": 1.277, + "step": 7565 + }, + { + "epoch": 0.18, + "learning_rate": 1.8813400107632763e-05, + "loss": 1.0645, + "step": 7566 + }, + { + "epoch": 0.18, + "learning_rate": 1.8813032066867788e-05, + "loss": 1.059, + "step": 7567 + }, + { + "epoch": 0.18, + "learning_rate": 1.881266397263622e-05, + "loss": 1.2238, + "step": 7568 + }, + { + "epoch": 0.18, + "learning_rate": 1.8812295824940284e-05, + "loss": 0.9914, + "step": 7569 + }, + { + "epoch": 0.18, + "learning_rate": 1.8811927623782227e-05, + "loss": 0.8806, + "step": 7570 + }, + { + "epoch": 0.18, + "learning_rate": 1.881155936916428e-05, + "loss": 1.2453, + "step": 7571 + }, + { + "epoch": 0.18, + "learning_rate": 1.8811191061088673e-05, + "loss": 1.3001, + "step": 7572 + }, + { + "epoch": 0.18, + "learning_rate": 1.8810822699557643e-05, + "loss": 1.1412, + "step": 7573 + }, + { + "epoch": 0.18, + "learning_rate": 1.8810454284573424e-05, + "loss": 1.374, + "step": 7574 + }, + { + "epoch": 0.18, + "learning_rate": 1.8810085816138254e-05, + "loss": 1.1798, + "step": 7575 + }, + { + "epoch": 0.18, + "learning_rate": 1.880971729425436e-05, + "loss": 1.0447, + "step": 7576 + }, + { + "epoch": 0.18, + "learning_rate": 1.880934871892399e-05, + "loss": 1.1301, + "step": 7577 + }, + { + "epoch": 0.18, + "learning_rate": 1.880898009014937e-05, + "loss": 1.2402, + "step": 7578 + }, + { + "epoch": 0.18, + "learning_rate": 1.8808611407932736e-05, + "loss": 1.0135, + "step": 7579 + }, + { + "epoch": 0.18, + "learning_rate": 1.880824267227633e-05, + "loss": 1.1433, + "step": 7580 + }, + { + "epoch": 0.18, + "learning_rate": 1.880787388318239e-05, + "loss": 1.2295, + "step": 7581 + }, + { + "epoch": 0.18, + "learning_rate": 1.880750504065315e-05, + "loss": 1.0031, + "step": 7582 + }, + { + "epoch": 0.18, + "learning_rate": 1.8807136144690846e-05, + "loss": 0.9584, + "step": 7583 + }, + { + "epoch": 0.18, + "learning_rate": 1.8806767195297724e-05, + "loss": 1.1021, + "step": 7584 + }, + { + "epoch": 0.18, + "learning_rate": 1.8806398192476014e-05, + "loss": 1.2153, + "step": 7585 + }, + { + "epoch": 0.18, + "learning_rate": 1.880602913622796e-05, + "loss": 1.1792, + "step": 7586 + }, + { + "epoch": 0.18, + "learning_rate": 1.8805660026555793e-05, + "loss": 1.1428, + "step": 7587 + }, + { + "epoch": 0.18, + "learning_rate": 1.880529086346176e-05, + "loss": 0.9431, + "step": 7588 + }, + { + "epoch": 0.18, + "learning_rate": 1.8804921646948103e-05, + "loss": 1.0666, + "step": 7589 + }, + { + "epoch": 0.18, + "learning_rate": 1.880455237701705e-05, + "loss": 1.109, + "step": 7590 + }, + { + "epoch": 0.18, + "learning_rate": 1.8804183053670856e-05, + "loss": 1.0801, + "step": 7591 + }, + { + "epoch": 0.18, + "learning_rate": 1.880381367691175e-05, + "loss": 1.2093, + "step": 7592 + }, + { + "epoch": 0.18, + "learning_rate": 1.880344424674198e-05, + "loss": 1.5647, + "step": 7593 + }, + { + "epoch": 0.18, + "learning_rate": 1.880307476316378e-05, + "loss": 1.0809, + "step": 7594 + }, + { + "epoch": 0.18, + "learning_rate": 1.88027052261794e-05, + "loss": 1.0756, + "step": 7595 + }, + { + "epoch": 0.18, + "learning_rate": 1.8802335635791076e-05, + "loss": 1.0846, + "step": 7596 + }, + { + "epoch": 0.18, + "learning_rate": 1.8801965992001052e-05, + "loss": 1.0007, + "step": 7597 + }, + { + "epoch": 0.18, + "learning_rate": 1.880159629481157e-05, + "loss": 1.2983, + "step": 7598 + }, + { + "epoch": 0.18, + "learning_rate": 1.8801226544224876e-05, + "loss": 1.2507, + "step": 7599 + }, + { + "epoch": 0.18, + "learning_rate": 1.880085674024321e-05, + "loss": 1.1085, + "step": 7600 + }, + { + "epoch": 0.18, + "learning_rate": 1.8800486882868816e-05, + "loss": 1.2504, + "step": 7601 + }, + { + "epoch": 0.18, + "learning_rate": 1.8800116972103935e-05, + "loss": 1.2544, + "step": 7602 + }, + { + "epoch": 0.18, + "learning_rate": 1.8799747007950818e-05, + "loss": 0.9192, + "step": 7603 + }, + { + "epoch": 0.18, + "learning_rate": 1.8799376990411708e-05, + "loss": 1.1629, + "step": 7604 + }, + { + "epoch": 0.18, + "learning_rate": 1.8799006919488844e-05, + "loss": 1.1264, + "step": 7605 + }, + { + "epoch": 0.18, + "learning_rate": 1.8798636795184473e-05, + "loss": 1.2362, + "step": 7606 + }, + { + "epoch": 0.18, + "learning_rate": 1.8798266617500845e-05, + "loss": 1.0777, + "step": 7607 + }, + { + "epoch": 0.18, + "learning_rate": 1.8797896386440202e-05, + "loss": 1.2769, + "step": 7608 + }, + { + "epoch": 0.18, + "learning_rate": 1.8797526102004792e-05, + "loss": 1.0421, + "step": 7609 + }, + { + "epoch": 0.18, + "learning_rate": 1.879715576419686e-05, + "loss": 1.2743, + "step": 7610 + }, + { + "epoch": 0.18, + "learning_rate": 1.879678537301865e-05, + "loss": 1.088, + "step": 7611 + }, + { + "epoch": 0.18, + "learning_rate": 1.8796414928472417e-05, + "loss": 1.2436, + "step": 7612 + }, + { + "epoch": 0.18, + "learning_rate": 1.87960444305604e-05, + "loss": 1.2301, + "step": 7613 + }, + { + "epoch": 0.18, + "learning_rate": 1.879567387928485e-05, + "loss": 1.1392, + "step": 7614 + }, + { + "epoch": 0.18, + "learning_rate": 1.8795303274648022e-05, + "loss": 1.1439, + "step": 7615 + }, + { + "epoch": 0.18, + "learning_rate": 1.8794932616652152e-05, + "loss": 1.1985, + "step": 7616 + }, + { + "epoch": 0.18, + "learning_rate": 1.8794561905299495e-05, + "loss": 1.0839, + "step": 7617 + }, + { + "epoch": 0.18, + "learning_rate": 1.8794191140592303e-05, + "loss": 1.1371, + "step": 7618 + }, + { + "epoch": 0.18, + "learning_rate": 1.8793820322532817e-05, + "loss": 1.1172, + "step": 7619 + }, + { + "epoch": 0.18, + "learning_rate": 1.8793449451123296e-05, + "loss": 1.1973, + "step": 7620 + }, + { + "epoch": 0.18, + "learning_rate": 1.8793078526365983e-05, + "loss": 1.3138, + "step": 7621 + }, + { + "epoch": 0.18, + "learning_rate": 1.879270754826313e-05, + "loss": 1.2559, + "step": 7622 + }, + { + "epoch": 0.18, + "learning_rate": 1.8792336516816988e-05, + "loss": 1.1111, + "step": 7623 + }, + { + "epoch": 0.18, + "learning_rate": 1.8791965432029812e-05, + "loss": 0.954, + "step": 7624 + }, + { + "epoch": 0.18, + "learning_rate": 1.8791594293903846e-05, + "loss": 1.1006, + "step": 7625 + }, + { + "epoch": 0.18, + "learning_rate": 1.8791223102441347e-05, + "loss": 1.224, + "step": 7626 + }, + { + "epoch": 0.18, + "learning_rate": 1.879085185764457e-05, + "loss": 1.1411, + "step": 7627 + }, + { + "epoch": 0.18, + "learning_rate": 1.8790480559515756e-05, + "loss": 1.1539, + "step": 7628 + }, + { + "epoch": 0.18, + "learning_rate": 1.8790109208057163e-05, + "loss": 1.1798, + "step": 7629 + }, + { + "epoch": 0.18, + "learning_rate": 1.878973780327105e-05, + "loss": 1.2631, + "step": 7630 + }, + { + "epoch": 0.18, + "learning_rate": 1.8789366345159662e-05, + "loss": 1.0031, + "step": 7631 + }, + { + "epoch": 0.18, + "learning_rate": 1.878899483372526e-05, + "loss": 1.1992, + "step": 7632 + }, + { + "epoch": 0.18, + "learning_rate": 1.878862326897009e-05, + "loss": 1.2983, + "step": 7633 + }, + { + "epoch": 0.18, + "learning_rate": 1.8788251650896407e-05, + "loss": 1.2242, + "step": 7634 + }, + { + "epoch": 0.18, + "learning_rate": 1.8787879979506474e-05, + "loss": 0.9957, + "step": 7635 + }, + { + "epoch": 0.18, + "learning_rate": 1.8787508254802538e-05, + "loss": 1.1202, + "step": 7636 + }, + { + "epoch": 0.18, + "learning_rate": 1.8787136476786857e-05, + "loss": 1.1056, + "step": 7637 + }, + { + "epoch": 0.18, + "learning_rate": 1.8786764645461684e-05, + "loss": 1.0379, + "step": 7638 + }, + { + "epoch": 0.18, + "learning_rate": 1.878639276082928e-05, + "loss": 1.204, + "step": 7639 + }, + { + "epoch": 0.18, + "learning_rate": 1.8786020822891892e-05, + "loss": 1.4236, + "step": 7640 + }, + { + "epoch": 0.18, + "learning_rate": 1.8785648831651788e-05, + "loss": 0.9448, + "step": 7641 + }, + { + "epoch": 0.18, + "learning_rate": 1.8785276787111216e-05, + "loss": 1.2151, + "step": 7642 + }, + { + "epoch": 0.18, + "learning_rate": 1.878490468927244e-05, + "loss": 1.043, + "step": 7643 + }, + { + "epoch": 0.18, + "learning_rate": 1.878453253813771e-05, + "loss": 1.0972, + "step": 7644 + }, + { + "epoch": 0.18, + "learning_rate": 1.8784160333709286e-05, + "loss": 1.0348, + "step": 7645 + }, + { + "epoch": 0.18, + "learning_rate": 1.878378807598943e-05, + "loss": 1.1865, + "step": 7646 + }, + { + "epoch": 0.18, + "learning_rate": 1.87834157649804e-05, + "loss": 1.2581, + "step": 7647 + }, + { + "epoch": 0.18, + "learning_rate": 1.8783043400684447e-05, + "loss": 1.0908, + "step": 7648 + }, + { + "epoch": 0.18, + "learning_rate": 1.878267098310384e-05, + "loss": 1.1793, + "step": 7649 + }, + { + "epoch": 0.18, + "learning_rate": 1.878229851224083e-05, + "loss": 1.4003, + "step": 7650 + }, + { + "epoch": 0.18, + "learning_rate": 1.8781925988097684e-05, + "loss": 1.1946, + "step": 7651 + }, + { + "epoch": 0.18, + "learning_rate": 1.8781553410676658e-05, + "loss": 1.2858, + "step": 7652 + }, + { + "epoch": 0.18, + "learning_rate": 1.8781180779980014e-05, + "loss": 1.0849, + "step": 7653 + }, + { + "epoch": 0.18, + "learning_rate": 1.878080809601001e-05, + "loss": 1.2307, + "step": 7654 + }, + { + "epoch": 0.18, + "learning_rate": 1.878043535876891e-05, + "loss": 1.2055, + "step": 7655 + }, + { + "epoch": 0.18, + "learning_rate": 1.878006256825897e-05, + "loss": 1.1329, + "step": 7656 + }, + { + "epoch": 0.18, + "learning_rate": 1.877968972448246e-05, + "loss": 1.3825, + "step": 7657 + }, + { + "epoch": 0.18, + "learning_rate": 1.8779316827441636e-05, + "loss": 1.1029, + "step": 7658 + }, + { + "epoch": 0.18, + "learning_rate": 1.8778943877138763e-05, + "loss": 1.1258, + "step": 7659 + }, + { + "epoch": 0.18, + "learning_rate": 1.87785708735761e-05, + "loss": 0.9922, + "step": 7660 + }, + { + "epoch": 0.18, + "learning_rate": 1.8778197816755915e-05, + "loss": 1.0327, + "step": 7661 + }, + { + "epoch": 0.18, + "learning_rate": 1.8777824706680466e-05, + "loss": 0.9973, + "step": 7662 + }, + { + "epoch": 0.18, + "learning_rate": 1.8777451543352022e-05, + "loss": 0.6975, + "step": 7663 + }, + { + "epoch": 0.18, + "learning_rate": 1.8777078326772843e-05, + "loss": 1.3561, + "step": 7664 + }, + { + "epoch": 0.18, + "learning_rate": 1.8776705056945195e-05, + "loss": 1.1141, + "step": 7665 + }, + { + "epoch": 0.18, + "learning_rate": 1.877633173387134e-05, + "loss": 0.8305, + "step": 7666 + }, + { + "epoch": 0.18, + "learning_rate": 1.8775958357553542e-05, + "loss": 1.3385, + "step": 7667 + }, + { + "epoch": 0.18, + "learning_rate": 1.8775584927994074e-05, + "loss": 1.3892, + "step": 7668 + }, + { + "epoch": 0.18, + "learning_rate": 1.8775211445195194e-05, + "loss": 1.1509, + "step": 7669 + }, + { + "epoch": 0.18, + "learning_rate": 1.877483790915917e-05, + "loss": 1.2136, + "step": 7670 + }, + { + "epoch": 0.18, + "learning_rate": 1.877446431988827e-05, + "loss": 1.1649, + "step": 7671 + }, + { + "epoch": 0.18, + "learning_rate": 1.877409067738476e-05, + "loss": 1.1757, + "step": 7672 + }, + { + "epoch": 0.18, + "learning_rate": 1.8773716981650904e-05, + "loss": 1.1609, + "step": 7673 + }, + { + "epoch": 0.18, + "learning_rate": 1.877334323268897e-05, + "loss": 1.1321, + "step": 7674 + }, + { + "epoch": 0.18, + "learning_rate": 1.8772969430501226e-05, + "loss": 0.799, + "step": 7675 + }, + { + "epoch": 0.18, + "learning_rate": 1.8772595575089943e-05, + "loss": 1.2161, + "step": 7676 + }, + { + "epoch": 0.18, + "learning_rate": 1.8772221666457385e-05, + "loss": 0.9383, + "step": 7677 + }, + { + "epoch": 0.18, + "learning_rate": 1.8771847704605818e-05, + "loss": 1.2051, + "step": 7678 + }, + { + "epoch": 0.18, + "learning_rate": 1.8771473689537516e-05, + "loss": 1.2462, + "step": 7679 + }, + { + "epoch": 0.18, + "learning_rate": 1.8771099621254748e-05, + "loss": 1.103, + "step": 7680 + }, + { + "epoch": 0.18, + "learning_rate": 1.8770725499759778e-05, + "loss": 0.8991, + "step": 7681 + }, + { + "epoch": 0.18, + "learning_rate": 1.8770351325054882e-05, + "loss": 1.2849, + "step": 7682 + }, + { + "epoch": 0.18, + "learning_rate": 1.876997709714233e-05, + "loss": 1.0836, + "step": 7683 + }, + { + "epoch": 0.18, + "learning_rate": 1.876960281602439e-05, + "loss": 1.0744, + "step": 7684 + }, + { + "epoch": 0.18, + "learning_rate": 1.8769228481703327e-05, + "loss": 1.0336, + "step": 7685 + }, + { + "epoch": 0.18, + "learning_rate": 1.876885409418142e-05, + "loss": 1.2075, + "step": 7686 + }, + { + "epoch": 0.18, + "learning_rate": 1.8768479653460936e-05, + "loss": 1.1874, + "step": 7687 + }, + { + "epoch": 0.18, + "learning_rate": 1.8768105159544152e-05, + "loss": 1.2694, + "step": 7688 + }, + { + "epoch": 0.18, + "learning_rate": 1.8767730612433334e-05, + "loss": 1.1466, + "step": 7689 + }, + { + "epoch": 0.18, + "learning_rate": 1.8767356012130758e-05, + "loss": 0.9434, + "step": 7690 + }, + { + "epoch": 0.18, + "learning_rate": 1.8766981358638693e-05, + "loss": 1.1205, + "step": 7691 + }, + { + "epoch": 0.18, + "learning_rate": 1.8766606651959417e-05, + "loss": 1.1102, + "step": 7692 + }, + { + "epoch": 0.19, + "learning_rate": 1.87662318920952e-05, + "loss": 1.0003, + "step": 7693 + }, + { + "epoch": 0.19, + "learning_rate": 1.876585707904831e-05, + "loss": 1.1582, + "step": 7694 + }, + { + "epoch": 0.19, + "learning_rate": 1.8765482212821035e-05, + "loss": 0.8883, + "step": 7695 + }, + { + "epoch": 0.19, + "learning_rate": 1.876510729341564e-05, + "loss": 1.1733, + "step": 7696 + }, + { + "epoch": 0.19, + "learning_rate": 1.87647323208344e-05, + "loss": 1.2802, + "step": 7697 + }, + { + "epoch": 0.19, + "learning_rate": 1.876435729507959e-05, + "loss": 1.0715, + "step": 7698 + }, + { + "epoch": 0.19, + "learning_rate": 1.8763982216153483e-05, + "loss": 1.3475, + "step": 7699 + }, + { + "epoch": 0.19, + "learning_rate": 1.876360708405836e-05, + "loss": 0.987, + "step": 7700 + }, + { + "epoch": 0.19, + "learning_rate": 1.8763231898796498e-05, + "loss": 1.0204, + "step": 7701 + }, + { + "epoch": 0.19, + "learning_rate": 1.8762856660370165e-05, + "loss": 1.4541, + "step": 7702 + }, + { + "epoch": 0.19, + "learning_rate": 1.8762481368781638e-05, + "loss": 1.2274, + "step": 7703 + }, + { + "epoch": 0.19, + "learning_rate": 1.8762106024033206e-05, + "loss": 1.1622, + "step": 7704 + }, + { + "epoch": 0.19, + "learning_rate": 1.8761730626127133e-05, + "loss": 1.3131, + "step": 7705 + }, + { + "epoch": 0.19, + "learning_rate": 1.87613551750657e-05, + "loss": 0.9956, + "step": 7706 + }, + { + "epoch": 0.19, + "learning_rate": 1.876097967085119e-05, + "loss": 1.291, + "step": 7707 + }, + { + "epoch": 0.19, + "learning_rate": 1.8760604113485873e-05, + "loss": 1.2376, + "step": 7708 + }, + { + "epoch": 0.19, + "learning_rate": 1.8760228502972034e-05, + "loss": 1.0084, + "step": 7709 + }, + { + "epoch": 0.19, + "learning_rate": 1.8759852839311946e-05, + "loss": 1.1191, + "step": 7710 + }, + { + "epoch": 0.19, + "learning_rate": 1.8759477122507896e-05, + "loss": 1.1243, + "step": 7711 + }, + { + "epoch": 0.19, + "learning_rate": 1.8759101352562154e-05, + "loss": 1.1503, + "step": 7712 + }, + { + "epoch": 0.19, + "learning_rate": 1.8758725529477008e-05, + "loss": 1.2238, + "step": 7713 + }, + { + "epoch": 0.19, + "learning_rate": 1.8758349653254733e-05, + "loss": 0.9606, + "step": 7714 + }, + { + "epoch": 0.19, + "learning_rate": 1.875797372389761e-05, + "loss": 1.0369, + "step": 7715 + }, + { + "epoch": 0.19, + "learning_rate": 1.8757597741407922e-05, + "loss": 1.1549, + "step": 7716 + }, + { + "epoch": 0.19, + "learning_rate": 1.8757221705787948e-05, + "loss": 0.8746, + "step": 7717 + }, + { + "epoch": 0.19, + "learning_rate": 1.875684561703997e-05, + "loss": 1.144, + "step": 7718 + }, + { + "epoch": 0.19, + "learning_rate": 1.8756469475166267e-05, + "loss": 1.2859, + "step": 7719 + }, + { + "epoch": 0.19, + "learning_rate": 1.8756093280169126e-05, + "loss": 0.8208, + "step": 7720 + }, + { + "epoch": 0.19, + "learning_rate": 1.8755717032050826e-05, + "loss": 1.102, + "step": 7721 + }, + { + "epoch": 0.19, + "learning_rate": 1.875534073081365e-05, + "loss": 1.1812, + "step": 7722 + }, + { + "epoch": 0.19, + "learning_rate": 1.875496437645988e-05, + "loss": 1.0212, + "step": 7723 + }, + { + "epoch": 0.19, + "learning_rate": 1.8754587968991803e-05, + "loss": 1.2372, + "step": 7724 + }, + { + "epoch": 0.19, + "learning_rate": 1.8754211508411697e-05, + "loss": 1.1924, + "step": 7725 + }, + { + "epoch": 0.19, + "learning_rate": 1.8753834994721852e-05, + "loss": 1.1008, + "step": 7726 + }, + { + "epoch": 0.19, + "learning_rate": 1.8753458427924546e-05, + "loss": 1.1768, + "step": 7727 + }, + { + "epoch": 0.19, + "learning_rate": 1.8753081808022065e-05, + "loss": 0.9606, + "step": 7728 + }, + { + "epoch": 0.19, + "learning_rate": 1.87527051350167e-05, + "loss": 0.9283, + "step": 7729 + }, + { + "epoch": 0.19, + "learning_rate": 1.8752328408910732e-05, + "loss": 0.9394, + "step": 7730 + }, + { + "epoch": 0.19, + "learning_rate": 1.8751951629706443e-05, + "loss": 1.0508, + "step": 7731 + }, + { + "epoch": 0.19, + "learning_rate": 1.8751574797406124e-05, + "loss": 0.9761, + "step": 7732 + }, + { + "epoch": 0.19, + "learning_rate": 1.875119791201206e-05, + "loss": 0.8316, + "step": 7733 + }, + { + "epoch": 0.19, + "learning_rate": 1.8750820973526535e-05, + "loss": 1.0895, + "step": 7734 + }, + { + "epoch": 0.19, + "learning_rate": 1.875044398195184e-05, + "loss": 1.148, + "step": 7735 + }, + { + "epoch": 0.19, + "learning_rate": 1.8750066937290256e-05, + "loss": 1.2793, + "step": 7736 + }, + { + "epoch": 0.19, + "learning_rate": 1.8749689839544076e-05, + "loss": 1.2559, + "step": 7737 + }, + { + "epoch": 0.19, + "learning_rate": 1.8749312688715587e-05, + "loss": 1.0834, + "step": 7738 + }, + { + "epoch": 0.19, + "learning_rate": 1.8748935484807075e-05, + "loss": 0.9964, + "step": 7739 + }, + { + "epoch": 0.19, + "learning_rate": 1.8748558227820828e-05, + "loss": 1.2116, + "step": 7740 + }, + { + "epoch": 0.19, + "learning_rate": 1.8748180917759134e-05, + "loss": 0.9329, + "step": 7741 + }, + { + "epoch": 0.19, + "learning_rate": 1.8747803554624287e-05, + "loss": 1.0985, + "step": 7742 + }, + { + "epoch": 0.19, + "learning_rate": 1.8747426138418574e-05, + "loss": 1.286, + "step": 7743 + }, + { + "epoch": 0.19, + "learning_rate": 1.874704866914428e-05, + "loss": 1.2622, + "step": 7744 + }, + { + "epoch": 0.19, + "learning_rate": 1.8746671146803707e-05, + "loss": 0.9177, + "step": 7745 + }, + { + "epoch": 0.19, + "learning_rate": 1.874629357139913e-05, + "loss": 1.2876, + "step": 7746 + }, + { + "epoch": 0.19, + "learning_rate": 1.874591594293285e-05, + "loss": 1.0801, + "step": 7747 + }, + { + "epoch": 0.19, + "learning_rate": 1.8745538261407157e-05, + "loss": 1.0139, + "step": 7748 + }, + { + "epoch": 0.19, + "learning_rate": 1.874516052682434e-05, + "loss": 1.1183, + "step": 7749 + }, + { + "epoch": 0.19, + "learning_rate": 1.8744782739186688e-05, + "loss": 1.1824, + "step": 7750 + }, + { + "epoch": 0.19, + "learning_rate": 1.8744404898496498e-05, + "loss": 1.0897, + "step": 7751 + }, + { + "epoch": 0.19, + "learning_rate": 1.874402700475606e-05, + "loss": 1.2907, + "step": 7752 + }, + { + "epoch": 0.19, + "learning_rate": 1.8743649057967667e-05, + "loss": 1.2078, + "step": 7753 + }, + { + "epoch": 0.19, + "learning_rate": 1.874327105813361e-05, + "loss": 1.2293, + "step": 7754 + }, + { + "epoch": 0.19, + "learning_rate": 1.874289300525619e-05, + "loss": 0.9378, + "step": 7755 + }, + { + "epoch": 0.19, + "learning_rate": 1.874251489933769e-05, + "loss": 1.2662, + "step": 7756 + }, + { + "epoch": 0.19, + "learning_rate": 1.8742136740380406e-05, + "loss": 1.0082, + "step": 7757 + }, + { + "epoch": 0.19, + "learning_rate": 1.874175852838664e-05, + "loss": 1.2988, + "step": 7758 + }, + { + "epoch": 0.19, + "learning_rate": 1.8741380263358676e-05, + "loss": 1.19, + "step": 7759 + }, + { + "epoch": 0.19, + "learning_rate": 1.8741001945298817e-05, + "loss": 1.0966, + "step": 7760 + }, + { + "epoch": 0.19, + "learning_rate": 1.8740623574209355e-05, + "loss": 1.1375, + "step": 7761 + }, + { + "epoch": 0.19, + "learning_rate": 1.8740245150092585e-05, + "loss": 1.1491, + "step": 7762 + }, + { + "epoch": 0.19, + "learning_rate": 1.8739866672950807e-05, + "loss": 1.1113, + "step": 7763 + }, + { + "epoch": 0.19, + "learning_rate": 1.873948814278631e-05, + "loss": 1.0469, + "step": 7764 + }, + { + "epoch": 0.19, + "learning_rate": 1.8739109559601393e-05, + "loss": 1.0858, + "step": 7765 + }, + { + "epoch": 0.19, + "learning_rate": 1.8738730923398357e-05, + "loss": 1.1724, + "step": 7766 + }, + { + "epoch": 0.19, + "learning_rate": 1.8738352234179492e-05, + "loss": 1.47, + "step": 7767 + }, + { + "epoch": 0.19, + "learning_rate": 1.8737973491947102e-05, + "loss": 1.1349, + "step": 7768 + }, + { + "epoch": 0.19, + "learning_rate": 1.8737594696703486e-05, + "loss": 1.0167, + "step": 7769 + }, + { + "epoch": 0.19, + "learning_rate": 1.8737215848450933e-05, + "loss": 1.099, + "step": 7770 + }, + { + "epoch": 0.19, + "learning_rate": 1.873683694719175e-05, + "loss": 1.2254, + "step": 7771 + }, + { + "epoch": 0.19, + "learning_rate": 1.8736457992928228e-05, + "loss": 1.3504, + "step": 7772 + }, + { + "epoch": 0.19, + "learning_rate": 1.873607898566267e-05, + "loss": 1.3424, + "step": 7773 + }, + { + "epoch": 0.19, + "learning_rate": 1.873569992539738e-05, + "loss": 1.1387, + "step": 7774 + }, + { + "epoch": 0.19, + "learning_rate": 1.8735320812134653e-05, + "loss": 0.9588, + "step": 7775 + }, + { + "epoch": 0.19, + "learning_rate": 1.8734941645876786e-05, + "loss": 1.2042, + "step": 7776 + }, + { + "epoch": 0.19, + "learning_rate": 1.8734562426626085e-05, + "loss": 0.9993, + "step": 7777 + }, + { + "epoch": 0.19, + "learning_rate": 1.8734183154384848e-05, + "loss": 1.1483, + "step": 7778 + }, + { + "epoch": 0.19, + "learning_rate": 1.8733803829155377e-05, + "loss": 1.0473, + "step": 7779 + }, + { + "epoch": 0.19, + "learning_rate": 1.873342445093997e-05, + "loss": 0.9925, + "step": 7780 + }, + { + "epoch": 0.19, + "learning_rate": 1.873304501974093e-05, + "loss": 1.1006, + "step": 7781 + }, + { + "epoch": 0.19, + "learning_rate": 1.8732665535560564e-05, + "loss": 1.0236, + "step": 7782 + }, + { + "epoch": 0.19, + "learning_rate": 1.873228599840117e-05, + "loss": 0.8212, + "step": 7783 + }, + { + "epoch": 0.19, + "learning_rate": 1.873190640826505e-05, + "loss": 1.1574, + "step": 7784 + }, + { + "epoch": 0.19, + "learning_rate": 1.8731526765154507e-05, + "loss": 1.1086, + "step": 7785 + }, + { + "epoch": 0.19, + "learning_rate": 1.8731147069071843e-05, + "loss": 0.9661, + "step": 7786 + }, + { + "epoch": 0.19, + "learning_rate": 1.8730767320019368e-05, + "loss": 1.2767, + "step": 7787 + }, + { + "epoch": 0.19, + "learning_rate": 1.8730387517999378e-05, + "loss": 0.9348, + "step": 7788 + }, + { + "epoch": 0.19, + "learning_rate": 1.873000766301418e-05, + "loss": 1.1975, + "step": 7789 + }, + { + "epoch": 0.19, + "learning_rate": 1.8729627755066082e-05, + "loss": 1.3048, + "step": 7790 + }, + { + "epoch": 0.19, + "learning_rate": 1.8729247794157382e-05, + "loss": 1.2479, + "step": 7791 + }, + { + "epoch": 0.19, + "learning_rate": 1.8728867780290393e-05, + "loss": 0.9198, + "step": 7792 + }, + { + "epoch": 0.19, + "learning_rate": 1.872848771346741e-05, + "loss": 0.9215, + "step": 7793 + }, + { + "epoch": 0.19, + "learning_rate": 1.8728107593690753e-05, + "loss": 1.1475, + "step": 7794 + }, + { + "epoch": 0.19, + "learning_rate": 1.8727727420962714e-05, + "loss": 1.1226, + "step": 7795 + }, + { + "epoch": 0.19, + "learning_rate": 1.872734719528561e-05, + "loss": 1.2062, + "step": 7796 + }, + { + "epoch": 0.19, + "learning_rate": 1.872696691666174e-05, + "loss": 1.2652, + "step": 7797 + }, + { + "epoch": 0.19, + "learning_rate": 1.872658658509342e-05, + "loss": 1.3338, + "step": 7798 + }, + { + "epoch": 0.19, + "learning_rate": 1.8726206200582945e-05, + "loss": 1.0807, + "step": 7799 + }, + { + "epoch": 0.19, + "learning_rate": 1.872582576313263e-05, + "loss": 1.1402, + "step": 7800 + }, + { + "epoch": 0.19, + "learning_rate": 1.8725445272744784e-05, + "loss": 1.1823, + "step": 7801 + }, + { + "epoch": 0.19, + "learning_rate": 1.8725064729421717e-05, + "loss": 1.1103, + "step": 7802 + }, + { + "epoch": 0.19, + "learning_rate": 1.8724684133165733e-05, + "loss": 1.2444, + "step": 7803 + }, + { + "epoch": 0.19, + "learning_rate": 1.872430348397914e-05, + "loss": 1.2422, + "step": 7804 + }, + { + "epoch": 0.19, + "learning_rate": 1.872392278186425e-05, + "loss": 1.224, + "step": 7805 + }, + { + "epoch": 0.19, + "learning_rate": 1.8723542026823375e-05, + "loss": 0.9818, + "step": 7806 + }, + { + "epoch": 0.19, + "learning_rate": 1.872316121885882e-05, + "loss": 1.3438, + "step": 7807 + }, + { + "epoch": 0.19, + "learning_rate": 1.8722780357972903e-05, + "loss": 1.3911, + "step": 7808 + }, + { + "epoch": 0.19, + "learning_rate": 1.8722399444167926e-05, + "loss": 1.0332, + "step": 7809 + }, + { + "epoch": 0.19, + "learning_rate": 1.87220184774462e-05, + "loss": 1.0338, + "step": 7810 + }, + { + "epoch": 0.19, + "learning_rate": 1.872163745781004e-05, + "loss": 1.2565, + "step": 7811 + }, + { + "epoch": 0.19, + "learning_rate": 1.872125638526176e-05, + "loss": 1.1329, + "step": 7812 + }, + { + "epoch": 0.19, + "learning_rate": 1.8720875259803667e-05, + "loss": 1.4226, + "step": 7813 + }, + { + "epoch": 0.19, + "learning_rate": 1.872049408143808e-05, + "loss": 1.1004, + "step": 7814 + }, + { + "epoch": 0.19, + "learning_rate": 1.87201128501673e-05, + "loss": 1.2028, + "step": 7815 + }, + { + "epoch": 0.19, + "learning_rate": 1.8719731565993647e-05, + "loss": 1.0919, + "step": 7816 + }, + { + "epoch": 0.19, + "learning_rate": 1.871935022891944e-05, + "loss": 1.2581, + "step": 7817 + }, + { + "epoch": 0.19, + "learning_rate": 1.871896883894698e-05, + "loss": 0.9007, + "step": 7818 + }, + { + "epoch": 0.19, + "learning_rate": 1.8718587396078588e-05, + "loss": 1.1893, + "step": 7819 + }, + { + "epoch": 0.19, + "learning_rate": 1.8718205900316578e-05, + "loss": 1.0782, + "step": 7820 + }, + { + "epoch": 0.19, + "learning_rate": 1.8717824351663263e-05, + "loss": 1.3409, + "step": 7821 + }, + { + "epoch": 0.19, + "learning_rate": 1.8717442750120956e-05, + "loss": 1.0876, + "step": 7822 + }, + { + "epoch": 0.19, + "learning_rate": 1.871706109569198e-05, + "loss": 1.2273, + "step": 7823 + }, + { + "epoch": 0.19, + "learning_rate": 1.871667938837864e-05, + "loss": 1.3759, + "step": 7824 + }, + { + "epoch": 0.19, + "learning_rate": 1.8716297628183257e-05, + "loss": 1.3857, + "step": 7825 + }, + { + "epoch": 0.19, + "learning_rate": 1.871591581510815e-05, + "loss": 1.1538, + "step": 7826 + }, + { + "epoch": 0.19, + "learning_rate": 1.8715533949155628e-05, + "loss": 1.2326, + "step": 7827 + }, + { + "epoch": 0.19, + "learning_rate": 1.871515203032801e-05, + "loss": 1.0743, + "step": 7828 + }, + { + "epoch": 0.19, + "learning_rate": 1.871477005862762e-05, + "loss": 1.2211, + "step": 7829 + }, + { + "epoch": 0.19, + "learning_rate": 1.8714388034056764e-05, + "loss": 1.1326, + "step": 7830 + }, + { + "epoch": 0.19, + "learning_rate": 1.8714005956617768e-05, + "loss": 1.0037, + "step": 7831 + }, + { + "epoch": 0.19, + "learning_rate": 1.871362382631295e-05, + "loss": 1.2251, + "step": 7832 + }, + { + "epoch": 0.19, + "learning_rate": 1.871324164314462e-05, + "loss": 1.0867, + "step": 7833 + }, + { + "epoch": 0.19, + "learning_rate": 1.871285940711511e-05, + "loss": 1.0381, + "step": 7834 + }, + { + "epoch": 0.19, + "learning_rate": 1.871247711822673e-05, + "loss": 1.0858, + "step": 7835 + }, + { + "epoch": 0.19, + "learning_rate": 1.8712094776481798e-05, + "loss": 1.5634, + "step": 7836 + }, + { + "epoch": 0.19, + "learning_rate": 1.8711712381882634e-05, + "loss": 1.2811, + "step": 7837 + }, + { + "epoch": 0.19, + "learning_rate": 1.8711329934431563e-05, + "loss": 1.3005, + "step": 7838 + }, + { + "epoch": 0.19, + "learning_rate": 1.87109474341309e-05, + "loss": 1.1314, + "step": 7839 + }, + { + "epoch": 0.19, + "learning_rate": 1.8710564880982975e-05, + "loss": 1.0266, + "step": 7840 + }, + { + "epoch": 0.19, + "learning_rate": 1.8710182274990093e-05, + "loss": 1.13, + "step": 7841 + }, + { + "epoch": 0.19, + "learning_rate": 1.8709799616154587e-05, + "loss": 1.1116, + "step": 7842 + }, + { + "epoch": 0.19, + "learning_rate": 1.8709416904478777e-05, + "loss": 1.4582, + "step": 7843 + }, + { + "epoch": 0.19, + "learning_rate": 1.870903413996498e-05, + "loss": 1.0811, + "step": 7844 + }, + { + "epoch": 0.19, + "learning_rate": 1.8708651322615522e-05, + "loss": 1.1129, + "step": 7845 + }, + { + "epoch": 0.19, + "learning_rate": 1.870826845243273e-05, + "loss": 1.0708, + "step": 7846 + }, + { + "epoch": 0.19, + "learning_rate": 1.8707885529418916e-05, + "loss": 1.1932, + "step": 7847 + }, + { + "epoch": 0.19, + "learning_rate": 1.870750255357641e-05, + "loss": 1.2514, + "step": 7848 + }, + { + "epoch": 0.19, + "learning_rate": 1.8707119524907535e-05, + "loss": 1.172, + "step": 7849 + }, + { + "epoch": 0.19, + "learning_rate": 1.8706736443414616e-05, + "loss": 1.4724, + "step": 7850 + }, + { + "epoch": 0.19, + "learning_rate": 1.870635330909997e-05, + "loss": 1.2157, + "step": 7851 + }, + { + "epoch": 0.19, + "learning_rate": 1.870597012196593e-05, + "loss": 1.2042, + "step": 7852 + }, + { + "epoch": 0.19, + "learning_rate": 1.8705586882014816e-05, + "loss": 1.1511, + "step": 7853 + }, + { + "epoch": 0.19, + "learning_rate": 1.8705203589248953e-05, + "loss": 1.3835, + "step": 7854 + }, + { + "epoch": 0.19, + "learning_rate": 1.8704820243670666e-05, + "loss": 0.9769, + "step": 7855 + }, + { + "epoch": 0.19, + "learning_rate": 1.8704436845282288e-05, + "loss": 1.1309, + "step": 7856 + }, + { + "epoch": 0.19, + "learning_rate": 1.8704053394086135e-05, + "loss": 1.4412, + "step": 7857 + }, + { + "epoch": 0.19, + "learning_rate": 1.8703669890084536e-05, + "loss": 0.9865, + "step": 7858 + }, + { + "epoch": 0.19, + "learning_rate": 1.8703286333279824e-05, + "loss": 1.0813, + "step": 7859 + }, + { + "epoch": 0.19, + "learning_rate": 1.8702902723674317e-05, + "loss": 1.3807, + "step": 7860 + }, + { + "epoch": 0.19, + "learning_rate": 1.8702519061270346e-05, + "loss": 1.0304, + "step": 7861 + }, + { + "epoch": 0.19, + "learning_rate": 1.870213534607024e-05, + "loss": 1.3365, + "step": 7862 + }, + { + "epoch": 0.19, + "learning_rate": 1.8701751578076328e-05, + "loss": 1.3433, + "step": 7863 + }, + { + "epoch": 0.19, + "learning_rate": 1.870136775729093e-05, + "loss": 0.8457, + "step": 7864 + }, + { + "epoch": 0.19, + "learning_rate": 1.8700983883716385e-05, + "loss": 1.2595, + "step": 7865 + }, + { + "epoch": 0.19, + "learning_rate": 1.8700599957355017e-05, + "loss": 1.3096, + "step": 7866 + }, + { + "epoch": 0.19, + "learning_rate": 1.8700215978209154e-05, + "loss": 0.8573, + "step": 7867 + }, + { + "epoch": 0.19, + "learning_rate": 1.8699831946281127e-05, + "loss": 0.9615, + "step": 7868 + }, + { + "epoch": 0.19, + "learning_rate": 1.869944786157327e-05, + "loss": 1.2221, + "step": 7869 + }, + { + "epoch": 0.19, + "learning_rate": 1.8699063724087905e-05, + "loss": 1.0951, + "step": 7870 + }, + { + "epoch": 0.19, + "learning_rate": 1.8698679533827367e-05, + "loss": 1.2189, + "step": 7871 + }, + { + "epoch": 0.19, + "learning_rate": 1.869829529079399e-05, + "loss": 1.4963, + "step": 7872 + }, + { + "epoch": 0.19, + "learning_rate": 1.8697910994990098e-05, + "loss": 1.2119, + "step": 7873 + }, + { + "epoch": 0.19, + "learning_rate": 1.869752664641802e-05, + "loss": 1.0896, + "step": 7874 + }, + { + "epoch": 0.19, + "learning_rate": 1.8697142245080102e-05, + "loss": 1.3613, + "step": 7875 + }, + { + "epoch": 0.19, + "learning_rate": 1.8696757790978668e-05, + "loss": 1.3259, + "step": 7876 + }, + { + "epoch": 0.19, + "learning_rate": 1.869637328411605e-05, + "loss": 1.0784, + "step": 7877 + }, + { + "epoch": 0.19, + "learning_rate": 1.8695988724494577e-05, + "loss": 1.3333, + "step": 7878 + }, + { + "epoch": 0.19, + "learning_rate": 1.8695604112116586e-05, + "loss": 1.3005, + "step": 7879 + }, + { + "epoch": 0.19, + "learning_rate": 1.869521944698441e-05, + "loss": 1.1542, + "step": 7880 + }, + { + "epoch": 0.19, + "learning_rate": 1.8694834729100386e-05, + "loss": 1.1629, + "step": 7881 + }, + { + "epoch": 0.19, + "learning_rate": 1.869444995846684e-05, + "loss": 1.0331, + "step": 7882 + }, + { + "epoch": 0.19, + "learning_rate": 1.8694065135086114e-05, + "loss": 1.0578, + "step": 7883 + }, + { + "epoch": 0.19, + "learning_rate": 1.8693680258960543e-05, + "loss": 1.2124, + "step": 7884 + }, + { + "epoch": 0.19, + "learning_rate": 1.8693295330092452e-05, + "loss": 1.1196, + "step": 7885 + }, + { + "epoch": 0.19, + "learning_rate": 1.869291034848419e-05, + "loss": 1.0044, + "step": 7886 + }, + { + "epoch": 0.19, + "learning_rate": 1.8692525314138082e-05, + "loss": 1.0858, + "step": 7887 + }, + { + "epoch": 0.19, + "learning_rate": 1.8692140227056468e-05, + "loss": 1.2623, + "step": 7888 + }, + { + "epoch": 0.19, + "learning_rate": 1.8691755087241686e-05, + "loss": 1.1641, + "step": 7889 + }, + { + "epoch": 0.19, + "learning_rate": 1.8691369894696064e-05, + "loss": 1.0789, + "step": 7890 + }, + { + "epoch": 0.19, + "learning_rate": 1.8690984649421954e-05, + "loss": 1.1961, + "step": 7891 + }, + { + "epoch": 0.19, + "learning_rate": 1.8690599351421675e-05, + "loss": 1.4141, + "step": 7892 + }, + { + "epoch": 0.19, + "learning_rate": 1.869021400069758e-05, + "loss": 1.2109, + "step": 7893 + }, + { + "epoch": 0.19, + "learning_rate": 1.8689828597252e-05, + "loss": 1.1577, + "step": 7894 + }, + { + "epoch": 0.19, + "learning_rate": 1.868944314108727e-05, + "loss": 1.2786, + "step": 7895 + }, + { + "epoch": 0.19, + "learning_rate": 1.8689057632205737e-05, + "loss": 1.1466, + "step": 7896 + }, + { + "epoch": 0.19, + "learning_rate": 1.8688672070609733e-05, + "loss": 1.3239, + "step": 7897 + }, + { + "epoch": 0.19, + "learning_rate": 1.8688286456301602e-05, + "loss": 1.2362, + "step": 7898 + }, + { + "epoch": 0.19, + "learning_rate": 1.868790078928368e-05, + "loss": 0.7968, + "step": 7899 + }, + { + "epoch": 0.19, + "learning_rate": 1.8687515069558303e-05, + "loss": 1.0285, + "step": 7900 + }, + { + "epoch": 0.19, + "learning_rate": 1.868712929712782e-05, + "loss": 1.1483, + "step": 7901 + }, + { + "epoch": 0.19, + "learning_rate": 1.8686743471994564e-05, + "loss": 1.218, + "step": 7902 + }, + { + "epoch": 0.19, + "learning_rate": 1.868635759416088e-05, + "loss": 1.2408, + "step": 7903 + }, + { + "epoch": 0.19, + "learning_rate": 1.868597166362911e-05, + "loss": 0.803, + "step": 7904 + }, + { + "epoch": 0.19, + "learning_rate": 1.868558568040159e-05, + "loss": 1.0948, + "step": 7905 + }, + { + "epoch": 0.19, + "learning_rate": 1.8685199644480663e-05, + "loss": 1.2306, + "step": 7906 + }, + { + "epoch": 0.19, + "learning_rate": 1.8684813555868675e-05, + "loss": 0.9911, + "step": 7907 + }, + { + "epoch": 0.19, + "learning_rate": 1.8684427414567964e-05, + "loss": 1.1407, + "step": 7908 + }, + { + "epoch": 0.19, + "learning_rate": 1.8684041220580876e-05, + "loss": 1.0221, + "step": 7909 + }, + { + "epoch": 0.19, + "learning_rate": 1.8683654973909754e-05, + "loss": 1.068, + "step": 7910 + }, + { + "epoch": 0.19, + "learning_rate": 1.868326867455694e-05, + "loss": 1.2965, + "step": 7911 + }, + { + "epoch": 0.19, + "learning_rate": 1.8682882322524777e-05, + "loss": 1.2615, + "step": 7912 + }, + { + "epoch": 0.19, + "learning_rate": 1.8682495917815607e-05, + "loss": 1.0951, + "step": 7913 + }, + { + "epoch": 0.19, + "learning_rate": 1.8682109460431775e-05, + "loss": 1.2653, + "step": 7914 + }, + { + "epoch": 0.19, + "learning_rate": 1.8681722950375634e-05, + "loss": 1.3543, + "step": 7915 + }, + { + "epoch": 0.19, + "learning_rate": 1.8681336387649516e-05, + "loss": 1.1755, + "step": 7916 + }, + { + "epoch": 0.19, + "learning_rate": 1.8680949772255774e-05, + "loss": 1.2501, + "step": 7917 + }, + { + "epoch": 0.19, + "learning_rate": 1.8680563104196753e-05, + "loss": 0.7897, + "step": 7918 + }, + { + "epoch": 0.19, + "learning_rate": 1.8680176383474796e-05, + "loss": 0.8447, + "step": 7919 + }, + { + "epoch": 0.19, + "learning_rate": 1.867978961009225e-05, + "loss": 1.181, + "step": 7920 + }, + { + "epoch": 0.19, + "learning_rate": 1.867940278405146e-05, + "loss": 1.1369, + "step": 7921 + }, + { + "epoch": 0.19, + "learning_rate": 1.8679015905354777e-05, + "loss": 1.1268, + "step": 7922 + }, + { + "epoch": 0.19, + "learning_rate": 1.867862897400455e-05, + "loss": 1.1277, + "step": 7923 + }, + { + "epoch": 0.19, + "learning_rate": 1.8678241990003116e-05, + "loss": 0.9596, + "step": 7924 + }, + { + "epoch": 0.19, + "learning_rate": 1.8677854953352835e-05, + "loss": 0.9112, + "step": 7925 + }, + { + "epoch": 0.19, + "learning_rate": 1.8677467864056045e-05, + "loss": 0.9035, + "step": 7926 + }, + { + "epoch": 0.19, + "learning_rate": 1.86770807221151e-05, + "loss": 1.021, + "step": 7927 + }, + { + "epoch": 0.19, + "learning_rate": 1.8676693527532344e-05, + "loss": 1.1633, + "step": 7928 + }, + { + "epoch": 0.19, + "learning_rate": 1.8676306280310134e-05, + "loss": 0.9803, + "step": 7929 + }, + { + "epoch": 0.19, + "learning_rate": 1.8675918980450812e-05, + "loss": 1.402, + "step": 7930 + }, + { + "epoch": 0.19, + "learning_rate": 1.867553162795673e-05, + "loss": 1.1856, + "step": 7931 + }, + { + "epoch": 0.19, + "learning_rate": 1.8675144222830242e-05, + "loss": 1.2756, + "step": 7932 + }, + { + "epoch": 0.19, + "learning_rate": 1.867475676507369e-05, + "loss": 1.0772, + "step": 7933 + }, + { + "epoch": 0.19, + "learning_rate": 1.867436925468943e-05, + "loss": 1.113, + "step": 7934 + }, + { + "epoch": 0.19, + "learning_rate": 1.8673981691679813e-05, + "loss": 1.1245, + "step": 7935 + }, + { + "epoch": 0.19, + "learning_rate": 1.867359407604719e-05, + "loss": 1.2827, + "step": 7936 + }, + { + "epoch": 0.19, + "learning_rate": 1.8673206407793908e-05, + "loss": 1.0085, + "step": 7937 + }, + { + "epoch": 0.19, + "learning_rate": 1.867281868692233e-05, + "loss": 1.213, + "step": 7938 + }, + { + "epoch": 0.19, + "learning_rate": 1.8672430913434794e-05, + "loss": 1.2722, + "step": 7939 + }, + { + "epoch": 0.19, + "learning_rate": 1.8672043087333662e-05, + "loss": 1.3088, + "step": 7940 + }, + { + "epoch": 0.19, + "learning_rate": 1.8671655208621287e-05, + "loss": 1.073, + "step": 7941 + }, + { + "epoch": 0.19, + "learning_rate": 1.8671267277300015e-05, + "loss": 1.2092, + "step": 7942 + }, + { + "epoch": 0.19, + "learning_rate": 1.867087929337221e-05, + "loss": 1.1242, + "step": 7943 + }, + { + "epoch": 0.19, + "learning_rate": 1.8670491256840212e-05, + "loss": 0.9775, + "step": 7944 + }, + { + "epoch": 0.19, + "learning_rate": 1.867010316770639e-05, + "loss": 1.1675, + "step": 7945 + }, + { + "epoch": 0.19, + "learning_rate": 1.866971502597309e-05, + "loss": 0.8912, + "step": 7946 + }, + { + "epoch": 0.19, + "learning_rate": 1.8669326831642666e-05, + "loss": 1.1646, + "step": 7947 + }, + { + "epoch": 0.19, + "learning_rate": 1.8668938584717473e-05, + "loss": 1.1514, + "step": 7948 + }, + { + "epoch": 0.19, + "learning_rate": 1.8668550285199872e-05, + "loss": 1.2358, + "step": 7949 + }, + { + "epoch": 0.19, + "learning_rate": 1.8668161933092218e-05, + "loss": 1.1242, + "step": 7950 + }, + { + "epoch": 0.19, + "learning_rate": 1.866777352839686e-05, + "loss": 0.9885, + "step": 7951 + }, + { + "epoch": 0.19, + "learning_rate": 1.8667385071116157e-05, + "loss": 0.9524, + "step": 7952 + }, + { + "epoch": 0.19, + "learning_rate": 1.8666996561252472e-05, + "loss": 0.9274, + "step": 7953 + }, + { + "epoch": 0.19, + "learning_rate": 1.8666607998808157e-05, + "loss": 1.0374, + "step": 7954 + }, + { + "epoch": 0.19, + "learning_rate": 1.866621938378557e-05, + "loss": 1.1034, + "step": 7955 + }, + { + "epoch": 0.19, + "learning_rate": 1.8665830716187064e-05, + "loss": 1.0136, + "step": 7956 + }, + { + "epoch": 0.19, + "learning_rate": 1.8665441996015006e-05, + "loss": 1.3235, + "step": 7957 + }, + { + "epoch": 0.19, + "learning_rate": 1.866505322327175e-05, + "loss": 0.9666, + "step": 7958 + }, + { + "epoch": 0.19, + "learning_rate": 1.8664664397959652e-05, + "loss": 0.8245, + "step": 7959 + }, + { + "epoch": 0.19, + "learning_rate": 1.866427552008107e-05, + "loss": 0.9167, + "step": 7960 + }, + { + "epoch": 0.19, + "learning_rate": 1.8663886589638372e-05, + "loss": 1.1034, + "step": 7961 + }, + { + "epoch": 0.19, + "learning_rate": 1.866349760663391e-05, + "loss": 1.1025, + "step": 7962 + }, + { + "epoch": 0.19, + "learning_rate": 1.8663108571070046e-05, + "loss": 1.2894, + "step": 7963 + }, + { + "epoch": 0.19, + "learning_rate": 1.8662719482949142e-05, + "loss": 1.105, + "step": 7964 + }, + { + "epoch": 0.19, + "learning_rate": 1.8662330342273554e-05, + "loss": 1.256, + "step": 7965 + }, + { + "epoch": 0.19, + "learning_rate": 1.8661941149045646e-05, + "loss": 1.1823, + "step": 7966 + }, + { + "epoch": 0.19, + "learning_rate": 1.8661551903267777e-05, + "loss": 1.0836, + "step": 7967 + }, + { + "epoch": 0.19, + "learning_rate": 1.866116260494231e-05, + "loss": 1.3001, + "step": 7968 + }, + { + "epoch": 0.19, + "learning_rate": 1.866077325407161e-05, + "loss": 1.2621, + "step": 7969 + }, + { + "epoch": 0.19, + "learning_rate": 1.8660383850658033e-05, + "loss": 1.0375, + "step": 7970 + }, + { + "epoch": 0.19, + "learning_rate": 1.8659994394703944e-05, + "loss": 1.3496, + "step": 7971 + }, + { + "epoch": 0.19, + "learning_rate": 1.8659604886211705e-05, + "loss": 1.2462, + "step": 7972 + }, + { + "epoch": 0.19, + "learning_rate": 1.8659215325183684e-05, + "loss": 1.2073, + "step": 7973 + }, + { + "epoch": 0.19, + "learning_rate": 1.865882571162224e-05, + "loss": 1.0639, + "step": 7974 + }, + { + "epoch": 0.19, + "learning_rate": 1.8658436045529735e-05, + "loss": 1.0867, + "step": 7975 + }, + { + "epoch": 0.19, + "learning_rate": 1.8658046326908533e-05, + "loss": 1.1222, + "step": 7976 + }, + { + "epoch": 0.19, + "learning_rate": 1.8657656555761003e-05, + "loss": 1.0409, + "step": 7977 + }, + { + "epoch": 0.19, + "learning_rate": 1.8657266732089508e-05, + "loss": 1.1782, + "step": 7978 + }, + { + "epoch": 0.19, + "learning_rate": 1.865687685589641e-05, + "loss": 0.9324, + "step": 7979 + }, + { + "epoch": 0.19, + "learning_rate": 1.865648692718408e-05, + "loss": 1.2098, + "step": 7980 + }, + { + "epoch": 0.19, + "learning_rate": 1.8656096945954877e-05, + "loss": 1.1745, + "step": 7981 + }, + { + "epoch": 0.19, + "learning_rate": 1.865570691221117e-05, + "loss": 1.2718, + "step": 7982 + }, + { + "epoch": 0.19, + "learning_rate": 1.8655316825955323e-05, + "loss": 1.0604, + "step": 7983 + }, + { + "epoch": 0.19, + "learning_rate": 1.8654926687189706e-05, + "loss": 1.0555, + "step": 7984 + }, + { + "epoch": 0.19, + "learning_rate": 1.8654536495916687e-05, + "loss": 1.1193, + "step": 7985 + }, + { + "epoch": 0.19, + "learning_rate": 1.865414625213863e-05, + "loss": 1.1086, + "step": 7986 + }, + { + "epoch": 0.19, + "learning_rate": 1.8653755955857898e-05, + "loss": 1.307, + "step": 7987 + }, + { + "epoch": 0.19, + "learning_rate": 1.8653365607076873e-05, + "loss": 1.3786, + "step": 7988 + }, + { + "epoch": 0.19, + "learning_rate": 1.8652975205797907e-05, + "loss": 1.198, + "step": 7989 + }, + { + "epoch": 0.19, + "learning_rate": 1.865258475202338e-05, + "loss": 1.1624, + "step": 7990 + }, + { + "epoch": 0.19, + "learning_rate": 1.8652194245755653e-05, + "loss": 1.1101, + "step": 7991 + }, + { + "epoch": 0.19, + "learning_rate": 1.86518036869971e-05, + "loss": 1.1827, + "step": 7992 + }, + { + "epoch": 0.19, + "learning_rate": 1.865141307575009e-05, + "loss": 1.0315, + "step": 7993 + }, + { + "epoch": 0.19, + "learning_rate": 1.8651022412016993e-05, + "loss": 1.2183, + "step": 7994 + }, + { + "epoch": 0.19, + "learning_rate": 1.8650631695800177e-05, + "loss": 1.2369, + "step": 7995 + }, + { + "epoch": 0.19, + "learning_rate": 1.8650240927102012e-05, + "loss": 1.277, + "step": 7996 + }, + { + "epoch": 0.19, + "learning_rate": 1.864985010592487e-05, + "loss": 1.1957, + "step": 7997 + }, + { + "epoch": 0.19, + "learning_rate": 1.8649459232271124e-05, + "loss": 1.3551, + "step": 7998 + }, + { + "epoch": 0.19, + "learning_rate": 1.864906830614314e-05, + "loss": 1.054, + "step": 7999 + }, + { + "epoch": 0.19, + "learning_rate": 1.8648677327543297e-05, + "loss": 1.0398, + "step": 8000 + }, + { + "epoch": 0.19, + "learning_rate": 1.864828629647396e-05, + "loss": 1.2725, + "step": 8001 + }, + { + "epoch": 0.19, + "learning_rate": 1.8647895212937504e-05, + "loss": 0.9358, + "step": 8002 + }, + { + "epoch": 0.19, + "learning_rate": 1.8647504076936306e-05, + "loss": 1.1778, + "step": 8003 + }, + { + "epoch": 0.19, + "learning_rate": 1.8647112888472732e-05, + "loss": 0.9425, + "step": 8004 + }, + { + "epoch": 0.19, + "learning_rate": 1.8646721647549156e-05, + "loss": 0.8007, + "step": 8005 + }, + { + "epoch": 0.19, + "learning_rate": 1.8646330354167955e-05, + "loss": 1.2192, + "step": 8006 + }, + { + "epoch": 0.19, + "learning_rate": 1.86459390083315e-05, + "loss": 1.0746, + "step": 8007 + }, + { + "epoch": 0.19, + "learning_rate": 1.8645547610042173e-05, + "loss": 0.9264, + "step": 8008 + }, + { + "epoch": 0.19, + "learning_rate": 1.8645156159302335e-05, + "loss": 1.2113, + "step": 8009 + }, + { + "epoch": 0.19, + "learning_rate": 1.864476465611437e-05, + "loss": 0.9415, + "step": 8010 + }, + { + "epoch": 0.19, + "learning_rate": 1.8644373100480652e-05, + "loss": 1.3746, + "step": 8011 + }, + { + "epoch": 0.19, + "learning_rate": 1.8643981492403557e-05, + "loss": 1.1971, + "step": 8012 + }, + { + "epoch": 0.19, + "learning_rate": 1.8643589831885456e-05, + "loss": 0.9863, + "step": 8013 + }, + { + "epoch": 0.19, + "learning_rate": 1.864319811892873e-05, + "loss": 1.338, + "step": 8014 + }, + { + "epoch": 0.19, + "learning_rate": 1.8642806353535757e-05, + "loss": 1.0905, + "step": 8015 + }, + { + "epoch": 0.19, + "learning_rate": 1.864241453570891e-05, + "loss": 1.1998, + "step": 8016 + }, + { + "epoch": 0.19, + "learning_rate": 1.8642022665450563e-05, + "loss": 1.0985, + "step": 8017 + }, + { + "epoch": 0.19, + "learning_rate": 1.86416307427631e-05, + "loss": 1.1255, + "step": 8018 + }, + { + "epoch": 0.19, + "learning_rate": 1.8641238767648895e-05, + "loss": 1.1737, + "step": 8019 + }, + { + "epoch": 0.19, + "learning_rate": 1.8640846740110327e-05, + "loss": 1.1461, + "step": 8020 + }, + { + "epoch": 0.19, + "learning_rate": 1.8640454660149774e-05, + "loss": 0.9427, + "step": 8021 + }, + { + "epoch": 0.19, + "learning_rate": 1.8640062527769615e-05, + "loss": 1.176, + "step": 8022 + }, + { + "epoch": 0.19, + "learning_rate": 1.863967034297223e-05, + "loss": 1.0431, + "step": 8023 + }, + { + "epoch": 0.19, + "learning_rate": 1.8639278105759998e-05, + "loss": 1.4376, + "step": 8024 + }, + { + "epoch": 0.19, + "learning_rate": 1.8638885816135297e-05, + "loss": 1.2639, + "step": 8025 + }, + { + "epoch": 0.19, + "learning_rate": 1.8638493474100507e-05, + "loss": 1.0151, + "step": 8026 + }, + { + "epoch": 0.19, + "learning_rate": 1.8638101079658007e-05, + "loss": 1.1338, + "step": 8027 + }, + { + "epoch": 0.19, + "learning_rate": 1.8637708632810185e-05, + "loss": 1.4276, + "step": 8028 + }, + { + "epoch": 0.19, + "learning_rate": 1.8637316133559412e-05, + "loss": 0.961, + "step": 8029 + }, + { + "epoch": 0.19, + "learning_rate": 1.8636923581908074e-05, + "loss": 1.2753, + "step": 8030 + }, + { + "epoch": 0.19, + "learning_rate": 1.8636530977858554e-05, + "loss": 1.1146, + "step": 8031 + }, + { + "epoch": 0.19, + "learning_rate": 1.863613832141323e-05, + "loss": 0.9295, + "step": 8032 + }, + { + "epoch": 0.19, + "learning_rate": 1.863574561257449e-05, + "loss": 1.1935, + "step": 8033 + }, + { + "epoch": 0.19, + "learning_rate": 1.8635352851344707e-05, + "loss": 1.0208, + "step": 8034 + }, + { + "epoch": 0.19, + "learning_rate": 1.8634960037726272e-05, + "loss": 1.0703, + "step": 8035 + }, + { + "epoch": 0.19, + "learning_rate": 1.8634567171721567e-05, + "loss": 1.0458, + "step": 8036 + }, + { + "epoch": 0.19, + "learning_rate": 1.8634174253332973e-05, + "loss": 0.9149, + "step": 8037 + }, + { + "epoch": 0.19, + "learning_rate": 1.8633781282562875e-05, + "loss": 1.0827, + "step": 8038 + }, + { + "epoch": 0.19, + "learning_rate": 1.8633388259413653e-05, + "loss": 1.1231, + "step": 8039 + }, + { + "epoch": 0.19, + "learning_rate": 1.8632995183887697e-05, + "loss": 1.4558, + "step": 8040 + }, + { + "epoch": 0.19, + "learning_rate": 1.8632602055987394e-05, + "loss": 1.2872, + "step": 8041 + }, + { + "epoch": 0.19, + "learning_rate": 1.8632208875715122e-05, + "loss": 1.023, + "step": 8042 + }, + { + "epoch": 0.19, + "learning_rate": 1.863181564307327e-05, + "loss": 1.0445, + "step": 8043 + }, + { + "epoch": 0.19, + "learning_rate": 1.8631422358064218e-05, + "loss": 1.1631, + "step": 8044 + }, + { + "epoch": 0.19, + "learning_rate": 1.8631029020690366e-05, + "loss": 1.0442, + "step": 8045 + }, + { + "epoch": 0.19, + "learning_rate": 1.8630635630954083e-05, + "loss": 1.0481, + "step": 8046 + }, + { + "epoch": 0.19, + "learning_rate": 1.8630242188857764e-05, + "loss": 1.3137, + "step": 8047 + }, + { + "epoch": 0.19, + "learning_rate": 1.86298486944038e-05, + "loss": 1.2463, + "step": 8048 + }, + { + "epoch": 0.19, + "learning_rate": 1.862945514759457e-05, + "loss": 1.5706, + "step": 8049 + }, + { + "epoch": 0.19, + "learning_rate": 1.862906154843247e-05, + "loss": 1.4473, + "step": 8050 + }, + { + "epoch": 0.19, + "learning_rate": 1.862866789691988e-05, + "loss": 1.0463, + "step": 8051 + }, + { + "epoch": 0.19, + "learning_rate": 1.8628274193059193e-05, + "loss": 1.3284, + "step": 8052 + }, + { + "epoch": 0.19, + "learning_rate": 1.8627880436852794e-05, + "loss": 1.1634, + "step": 8053 + }, + { + "epoch": 0.19, + "learning_rate": 1.8627486628303076e-05, + "loss": 1.3135, + "step": 8054 + }, + { + "epoch": 0.19, + "learning_rate": 1.8627092767412424e-05, + "loss": 1.1822, + "step": 8055 + }, + { + "epoch": 0.19, + "learning_rate": 1.862669885418323e-05, + "loss": 1.1594, + "step": 8056 + }, + { + "epoch": 0.19, + "learning_rate": 1.8626304888617883e-05, + "loss": 1.3331, + "step": 8057 + }, + { + "epoch": 0.19, + "learning_rate": 1.8625910870718775e-05, + "loss": 1.0156, + "step": 8058 + }, + { + "epoch": 0.19, + "learning_rate": 1.8625516800488294e-05, + "loss": 1.2379, + "step": 8059 + }, + { + "epoch": 0.19, + "learning_rate": 1.862512267792883e-05, + "loss": 1.1758, + "step": 8060 + }, + { + "epoch": 0.19, + "learning_rate": 1.8624728503042772e-05, + "loss": 1.2367, + "step": 8061 + }, + { + "epoch": 0.19, + "learning_rate": 1.8624334275832522e-05, + "loss": 1.1585, + "step": 8062 + }, + { + "epoch": 0.19, + "learning_rate": 1.862393999630046e-05, + "loss": 1.2077, + "step": 8063 + }, + { + "epoch": 0.19, + "learning_rate": 1.8623545664448987e-05, + "loss": 1.0952, + "step": 8064 + }, + { + "epoch": 0.19, + "learning_rate": 1.8623151280280486e-05, + "loss": 1.1435, + "step": 8065 + }, + { + "epoch": 0.19, + "learning_rate": 1.8622756843797356e-05, + "loss": 1.1412, + "step": 8066 + }, + { + "epoch": 0.19, + "learning_rate": 1.8622362355001987e-05, + "loss": 1.074, + "step": 8067 + }, + { + "epoch": 0.19, + "learning_rate": 1.8621967813896776e-05, + "loss": 1.1119, + "step": 8068 + }, + { + "epoch": 0.19, + "learning_rate": 1.8621573220484114e-05, + "loss": 1.0605, + "step": 8069 + }, + { + "epoch": 0.19, + "learning_rate": 1.8621178574766397e-05, + "loss": 0.8914, + "step": 8070 + }, + { + "epoch": 0.19, + "learning_rate": 1.8620783876746012e-05, + "loss": 0.945, + "step": 8071 + }, + { + "epoch": 0.19, + "learning_rate": 1.8620389126425365e-05, + "loss": 1.1406, + "step": 8072 + }, + { + "epoch": 0.19, + "learning_rate": 1.861999432380684e-05, + "loss": 1.3401, + "step": 8073 + }, + { + "epoch": 0.19, + "learning_rate": 1.861959946889284e-05, + "loss": 0.9595, + "step": 8074 + }, + { + "epoch": 0.19, + "learning_rate": 1.8619204561685756e-05, + "loss": 1.2465, + "step": 8075 + }, + { + "epoch": 0.19, + "learning_rate": 1.8618809602187987e-05, + "loss": 1.022, + "step": 8076 + }, + { + "epoch": 0.19, + "learning_rate": 1.8618414590401927e-05, + "loss": 1.313, + "step": 8077 + }, + { + "epoch": 0.19, + "learning_rate": 1.861801952632997e-05, + "loss": 1.1601, + "step": 8078 + }, + { + "epoch": 0.19, + "learning_rate": 1.8617624409974516e-05, + "loss": 0.8944, + "step": 8079 + }, + { + "epoch": 0.19, + "learning_rate": 1.8617229241337967e-05, + "loss": 1.1234, + "step": 8080 + }, + { + "epoch": 0.19, + "learning_rate": 1.861683402042271e-05, + "loss": 1.0918, + "step": 8081 + }, + { + "epoch": 0.19, + "learning_rate": 1.861643874723115e-05, + "loss": 1.3337, + "step": 8082 + }, + { + "epoch": 0.19, + "learning_rate": 1.861604342176568e-05, + "loss": 1.2949, + "step": 8083 + }, + { + "epoch": 0.19, + "learning_rate": 1.86156480440287e-05, + "loss": 0.9295, + "step": 8084 + }, + { + "epoch": 0.19, + "learning_rate": 1.8615252614022616e-05, + "loss": 1.1123, + "step": 8085 + }, + { + "epoch": 0.19, + "learning_rate": 1.8614857131749818e-05, + "loss": 1.117, + "step": 8086 + }, + { + "epoch": 0.19, + "learning_rate": 1.861446159721271e-05, + "loss": 1.1364, + "step": 8087 + }, + { + "epoch": 0.19, + "learning_rate": 1.8614066010413686e-05, + "loss": 1.322, + "step": 8088 + }, + { + "epoch": 0.19, + "learning_rate": 1.8613670371355154e-05, + "loss": 1.284, + "step": 8089 + }, + { + "epoch": 0.19, + "learning_rate": 1.8613274680039506e-05, + "loss": 1.2347, + "step": 8090 + }, + { + "epoch": 0.19, + "learning_rate": 1.8612878936469146e-05, + "loss": 1.1485, + "step": 8091 + }, + { + "epoch": 0.19, + "learning_rate": 1.861248314064648e-05, + "loss": 1.1534, + "step": 8092 + }, + { + "epoch": 0.19, + "learning_rate": 1.8612087292573903e-05, + "loss": 1.1115, + "step": 8093 + }, + { + "epoch": 0.19, + "learning_rate": 1.8611691392253814e-05, + "loss": 1.2828, + "step": 8094 + }, + { + "epoch": 0.19, + "learning_rate": 1.8611295439688622e-05, + "loss": 1.315, + "step": 8095 + }, + { + "epoch": 0.19, + "learning_rate": 1.8610899434880724e-05, + "loss": 1.1472, + "step": 8096 + }, + { + "epoch": 0.19, + "learning_rate": 1.8610503377832526e-05, + "loss": 1.1525, + "step": 8097 + }, + { + "epoch": 0.19, + "learning_rate": 1.861010726854643e-05, + "loss": 0.9189, + "step": 8098 + }, + { + "epoch": 0.19, + "learning_rate": 1.8609711107024838e-05, + "loss": 1.069, + "step": 8099 + }, + { + "epoch": 0.19, + "learning_rate": 1.8609314893270155e-05, + "loss": 0.9631, + "step": 8100 + }, + { + "epoch": 0.19, + "learning_rate": 1.860891862728478e-05, + "loss": 1.3242, + "step": 8101 + }, + { + "epoch": 0.19, + "learning_rate": 1.8608522309071128e-05, + "loss": 1.2563, + "step": 8102 + }, + { + "epoch": 0.19, + "learning_rate": 1.860812593863159e-05, + "loss": 1.0658, + "step": 8103 + }, + { + "epoch": 0.19, + "learning_rate": 1.860772951596858e-05, + "loss": 1.3909, + "step": 8104 + }, + { + "epoch": 0.19, + "learning_rate": 1.8607333041084494e-05, + "loss": 1.3258, + "step": 8105 + }, + { + "epoch": 0.19, + "learning_rate": 1.8606936513981745e-05, + "loss": 1.0201, + "step": 8106 + }, + { + "epoch": 0.19, + "learning_rate": 1.860653993466274e-05, + "loss": 1.0712, + "step": 8107 + }, + { + "epoch": 0.19, + "learning_rate": 1.860614330312988e-05, + "loss": 0.9474, + "step": 8108 + }, + { + "epoch": 0.2, + "learning_rate": 1.860574661938557e-05, + "loss": 1.0922, + "step": 8109 + }, + { + "epoch": 0.2, + "learning_rate": 1.8605349883432223e-05, + "loss": 0.9718, + "step": 8110 + }, + { + "epoch": 0.2, + "learning_rate": 1.8604953095272244e-05, + "loss": 1.1625, + "step": 8111 + }, + { + "epoch": 0.2, + "learning_rate": 1.8604556254908034e-05, + "loss": 0.8478, + "step": 8112 + }, + { + "epoch": 0.2, + "learning_rate": 1.8604159362342008e-05, + "loss": 1.1162, + "step": 8113 + }, + { + "epoch": 0.2, + "learning_rate": 1.860376241757657e-05, + "loss": 1.0782, + "step": 8114 + }, + { + "epoch": 0.2, + "learning_rate": 1.860336542061413e-05, + "loss": 1.2906, + "step": 8115 + }, + { + "epoch": 0.2, + "learning_rate": 1.8602968371457094e-05, + "loss": 1.2265, + "step": 8116 + }, + { + "epoch": 0.2, + "learning_rate": 1.8602571270107874e-05, + "loss": 1.5319, + "step": 8117 + }, + { + "epoch": 0.2, + "learning_rate": 1.8602174116568876e-05, + "loss": 1.3119, + "step": 8118 + }, + { + "epoch": 0.2, + "learning_rate": 1.8601776910842514e-05, + "loss": 1.0202, + "step": 8119 + }, + { + "epoch": 0.2, + "learning_rate": 1.860137965293119e-05, + "loss": 1.1047, + "step": 8120 + }, + { + "epoch": 0.2, + "learning_rate": 1.8600982342837325e-05, + "loss": 1.0546, + "step": 8121 + }, + { + "epoch": 0.2, + "learning_rate": 1.8600584980563316e-05, + "loss": 1.2364, + "step": 8122 + }, + { + "epoch": 0.2, + "learning_rate": 1.8600187566111583e-05, + "loss": 1.2058, + "step": 8123 + }, + { + "epoch": 0.2, + "learning_rate": 1.859979009948454e-05, + "loss": 1.1634, + "step": 8124 + }, + { + "epoch": 0.2, + "learning_rate": 1.859939258068459e-05, + "loss": 1.0519, + "step": 8125 + }, + { + "epoch": 0.2, + "learning_rate": 1.8598995009714145e-05, + "loss": 1.2186, + "step": 8126 + }, + { + "epoch": 0.2, + "learning_rate": 1.8598597386575622e-05, + "loss": 0.9456, + "step": 8127 + }, + { + "epoch": 0.2, + "learning_rate": 1.8598199711271433e-05, + "loss": 1.2474, + "step": 8128 + }, + { + "epoch": 0.2, + "learning_rate": 1.8597801983803987e-05, + "loss": 1.1509, + "step": 8129 + }, + { + "epoch": 0.2, + "learning_rate": 1.85974042041757e-05, + "loss": 1.2537, + "step": 8130 + }, + { + "epoch": 0.2, + "learning_rate": 1.8597006372388984e-05, + "loss": 0.9615, + "step": 8131 + }, + { + "epoch": 0.2, + "learning_rate": 1.859660848844625e-05, + "loss": 1.2219, + "step": 8132 + }, + { + "epoch": 0.2, + "learning_rate": 1.8596210552349914e-05, + "loss": 1.2688, + "step": 8133 + }, + { + "epoch": 0.2, + "learning_rate": 1.8595812564102393e-05, + "loss": 1.1773, + "step": 8134 + }, + { + "epoch": 0.2, + "learning_rate": 1.8595414523706102e-05, + "loss": 1.1503, + "step": 8135 + }, + { + "epoch": 0.2, + "learning_rate": 1.8595016431163448e-05, + "loss": 1.1181, + "step": 8136 + }, + { + "epoch": 0.2, + "learning_rate": 1.8594618286476852e-05, + "loss": 1.183, + "step": 8137 + }, + { + "epoch": 0.2, + "learning_rate": 1.8594220089648727e-05, + "loss": 1.2448, + "step": 8138 + }, + { + "epoch": 0.2, + "learning_rate": 1.8593821840681493e-05, + "loss": 1.051, + "step": 8139 + }, + { + "epoch": 0.2, + "learning_rate": 1.8593423539577565e-05, + "loss": 0.8314, + "step": 8140 + }, + { + "epoch": 0.2, + "learning_rate": 1.8593025186339352e-05, + "loss": 1.168, + "step": 8141 + }, + { + "epoch": 0.2, + "learning_rate": 1.859262678096928e-05, + "loss": 1.245, + "step": 8142 + }, + { + "epoch": 0.2, + "learning_rate": 1.8592228323469764e-05, + "loss": 1.1092, + "step": 8143 + }, + { + "epoch": 0.2, + "learning_rate": 1.8591829813843215e-05, + "loss": 1.1336, + "step": 8144 + }, + { + "epoch": 0.2, + "learning_rate": 1.859143125209206e-05, + "loss": 1.1551, + "step": 8145 + }, + { + "epoch": 0.2, + "learning_rate": 1.859103263821871e-05, + "loss": 0.9076, + "step": 8146 + }, + { + "epoch": 0.2, + "learning_rate": 1.8590633972225584e-05, + "loss": 0.913, + "step": 8147 + }, + { + "epoch": 0.2, + "learning_rate": 1.8590235254115105e-05, + "loss": 1.1825, + "step": 8148 + }, + { + "epoch": 0.2, + "learning_rate": 1.8589836483889687e-05, + "loss": 1.2182, + "step": 8149 + }, + { + "epoch": 0.2, + "learning_rate": 1.8589437661551756e-05, + "loss": 1.0662, + "step": 8150 + }, + { + "epoch": 0.2, + "learning_rate": 1.8589038787103725e-05, + "loss": 1.2932, + "step": 8151 + }, + { + "epoch": 0.2, + "learning_rate": 1.8588639860548012e-05, + "loss": 1.3683, + "step": 8152 + }, + { + "epoch": 0.2, + "learning_rate": 1.8588240881887044e-05, + "loss": 0.9929, + "step": 8153 + }, + { + "epoch": 0.2, + "learning_rate": 1.858784185112324e-05, + "loss": 1.2604, + "step": 8154 + }, + { + "epoch": 0.2, + "learning_rate": 1.8587442768259018e-05, + "loss": 1.188, + "step": 8155 + }, + { + "epoch": 0.2, + "learning_rate": 1.85870436332968e-05, + "loss": 0.9624, + "step": 8156 + }, + { + "epoch": 0.2, + "learning_rate": 1.858664444623901e-05, + "loss": 1.1321, + "step": 8157 + }, + { + "epoch": 0.2, + "learning_rate": 1.8586245207088068e-05, + "loss": 1.1351, + "step": 8158 + }, + { + "epoch": 0.2, + "learning_rate": 1.8585845915846396e-05, + "loss": 1.2636, + "step": 8159 + }, + { + "epoch": 0.2, + "learning_rate": 1.8585446572516416e-05, + "loss": 1.0903, + "step": 8160 + }, + { + "epoch": 0.2, + "learning_rate": 1.8585047177100546e-05, + "loss": 0.988, + "step": 8161 + }, + { + "epoch": 0.2, + "learning_rate": 1.858464772960122e-05, + "loss": 1.2293, + "step": 8162 + }, + { + "epoch": 0.2, + "learning_rate": 1.8584248230020855e-05, + "loss": 1.128, + "step": 8163 + }, + { + "epoch": 0.2, + "learning_rate": 1.858384867836187e-05, + "loss": 1.211, + "step": 8164 + }, + { + "epoch": 0.2, + "learning_rate": 1.8583449074626702e-05, + "loss": 0.9412, + "step": 8165 + }, + { + "epoch": 0.2, + "learning_rate": 1.8583049418817764e-05, + "loss": 1.2573, + "step": 8166 + }, + { + "epoch": 0.2, + "learning_rate": 1.8582649710937485e-05, + "loss": 1.0302, + "step": 8167 + }, + { + "epoch": 0.2, + "learning_rate": 1.8582249950988285e-05, + "loss": 1.0965, + "step": 8168 + }, + { + "epoch": 0.2, + "learning_rate": 1.8581850138972596e-05, + "loss": 1.0817, + "step": 8169 + }, + { + "epoch": 0.2, + "learning_rate": 1.8581450274892842e-05, + "loss": 1.0803, + "step": 8170 + }, + { + "epoch": 0.2, + "learning_rate": 1.8581050358751444e-05, + "loss": 1.1677, + "step": 8171 + }, + { + "epoch": 0.2, + "learning_rate": 1.8580650390550835e-05, + "loss": 1.4572, + "step": 8172 + }, + { + "epoch": 0.2, + "learning_rate": 1.858025037029344e-05, + "loss": 1.4151, + "step": 8173 + }, + { + "epoch": 0.2, + "learning_rate": 1.857985029798168e-05, + "loss": 0.8228, + "step": 8174 + }, + { + "epoch": 0.2, + "learning_rate": 1.8579450173617987e-05, + "loss": 1.1502, + "step": 8175 + }, + { + "epoch": 0.2, + "learning_rate": 1.857904999720479e-05, + "loss": 0.7184, + "step": 8176 + }, + { + "epoch": 0.2, + "learning_rate": 1.857864976874451e-05, + "loss": 1.3641, + "step": 8177 + }, + { + "epoch": 0.2, + "learning_rate": 1.8578249488239584e-05, + "loss": 1.1547, + "step": 8178 + }, + { + "epoch": 0.2, + "learning_rate": 1.8577849155692436e-05, + "loss": 1.0024, + "step": 8179 + }, + { + "epoch": 0.2, + "learning_rate": 1.8577448771105494e-05, + "loss": 0.9517, + "step": 8180 + }, + { + "epoch": 0.2, + "learning_rate": 1.8577048334481188e-05, + "loss": 1.1596, + "step": 8181 + }, + { + "epoch": 0.2, + "learning_rate": 1.8576647845821947e-05, + "loss": 1.3643, + "step": 8182 + }, + { + "epoch": 0.2, + "learning_rate": 1.85762473051302e-05, + "loss": 1.2284, + "step": 8183 + }, + { + "epoch": 0.2, + "learning_rate": 1.857584671240838e-05, + "loss": 1.3144, + "step": 8184 + }, + { + "epoch": 0.2, + "learning_rate": 1.8575446067658912e-05, + "loss": 1.0397, + "step": 8185 + }, + { + "epoch": 0.2, + "learning_rate": 1.8575045370884232e-05, + "loss": 1.1767, + "step": 8186 + }, + { + "epoch": 0.2, + "learning_rate": 1.8574644622086767e-05, + "loss": 1.1589, + "step": 8187 + }, + { + "epoch": 0.2, + "learning_rate": 1.8574243821268953e-05, + "loss": 1.0204, + "step": 8188 + }, + { + "epoch": 0.2, + "learning_rate": 1.8573842968433216e-05, + "loss": 1.3099, + "step": 8189 + }, + { + "epoch": 0.2, + "learning_rate": 1.857344206358199e-05, + "loss": 1.1989, + "step": 8190 + }, + { + "epoch": 0.2, + "learning_rate": 1.857304110671771e-05, + "loss": 1.2714, + "step": 8191 + }, + { + "epoch": 0.2, + "learning_rate": 1.8572640097842804e-05, + "loss": 1.1144, + "step": 8192 + }, + { + "epoch": 0.2, + "learning_rate": 1.8572239036959708e-05, + "loss": 1.151, + "step": 8193 + }, + { + "epoch": 0.2, + "learning_rate": 1.8571837924070853e-05, + "loss": 0.9139, + "step": 8194 + }, + { + "epoch": 0.2, + "learning_rate": 1.8571436759178675e-05, + "loss": 1.4493, + "step": 8195 + }, + { + "epoch": 0.2, + "learning_rate": 1.857103554228561e-05, + "loss": 0.9535, + "step": 8196 + }, + { + "epoch": 0.2, + "learning_rate": 1.857063427339408e-05, + "loss": 1.2795, + "step": 8197 + }, + { + "epoch": 0.2, + "learning_rate": 1.857023295250653e-05, + "loss": 1.154, + "step": 8198 + }, + { + "epoch": 0.2, + "learning_rate": 1.8569831579625396e-05, + "loss": 1.0899, + "step": 8199 + }, + { + "epoch": 0.2, + "learning_rate": 1.856943015475311e-05, + "loss": 1.1646, + "step": 8200 + }, + { + "epoch": 0.2, + "learning_rate": 1.8569028677892104e-05, + "loss": 1.0893, + "step": 8201 + }, + { + "epoch": 0.2, + "learning_rate": 1.856862714904482e-05, + "loss": 1.2237, + "step": 8202 + }, + { + "epoch": 0.2, + "learning_rate": 1.8568225568213683e-05, + "loss": 1.1282, + "step": 8203 + }, + { + "epoch": 0.2, + "learning_rate": 1.8567823935401145e-05, + "loss": 0.7746, + "step": 8204 + }, + { + "epoch": 0.2, + "learning_rate": 1.856742225060963e-05, + "loss": 1.3668, + "step": 8205 + }, + { + "epoch": 0.2, + "learning_rate": 1.8567020513841582e-05, + "loss": 1.263, + "step": 8206 + }, + { + "epoch": 0.2, + "learning_rate": 1.8566618725099436e-05, + "loss": 1.191, + "step": 8207 + }, + { + "epoch": 0.2, + "learning_rate": 1.8566216884385625e-05, + "loss": 1.3319, + "step": 8208 + }, + { + "epoch": 0.2, + "learning_rate": 1.8565814991702596e-05, + "loss": 1.0886, + "step": 8209 + }, + { + "epoch": 0.2, + "learning_rate": 1.8565413047052778e-05, + "loss": 1.0803, + "step": 8210 + }, + { + "epoch": 0.2, + "learning_rate": 1.8565011050438616e-05, + "loss": 1.2441, + "step": 8211 + }, + { + "epoch": 0.2, + "learning_rate": 1.8564609001862547e-05, + "loss": 1.3577, + "step": 8212 + }, + { + "epoch": 0.2, + "learning_rate": 1.8564206901327008e-05, + "loss": 1.3508, + "step": 8213 + }, + { + "epoch": 0.2, + "learning_rate": 1.856380474883444e-05, + "loss": 1.0925, + "step": 8214 + }, + { + "epoch": 0.2, + "learning_rate": 1.8563402544387287e-05, + "loss": 1.3546, + "step": 8215 + }, + { + "epoch": 0.2, + "learning_rate": 1.856300028798798e-05, + "loss": 1.0594, + "step": 8216 + }, + { + "epoch": 0.2, + "learning_rate": 1.8562597979638966e-05, + "loss": 1.3118, + "step": 8217 + }, + { + "epoch": 0.2, + "learning_rate": 1.8562195619342684e-05, + "loss": 1.1063, + "step": 8218 + }, + { + "epoch": 0.2, + "learning_rate": 1.8561793207101578e-05, + "loss": 1.0505, + "step": 8219 + }, + { + "epoch": 0.2, + "learning_rate": 1.856139074291808e-05, + "loss": 1.1487, + "step": 8220 + }, + { + "epoch": 0.2, + "learning_rate": 1.8560988226794645e-05, + "loss": 1.3607, + "step": 8221 + }, + { + "epoch": 0.2, + "learning_rate": 1.8560585658733707e-05, + "loss": 1.1712, + "step": 8222 + }, + { + "epoch": 0.2, + "learning_rate": 1.8560183038737706e-05, + "loss": 1.3286, + "step": 8223 + }, + { + "epoch": 0.2, + "learning_rate": 1.855978036680909e-05, + "loss": 1.148, + "step": 8224 + }, + { + "epoch": 0.2, + "learning_rate": 1.85593776429503e-05, + "loss": 1.1163, + "step": 8225 + }, + { + "epoch": 0.2, + "learning_rate": 1.8558974867163778e-05, + "loss": 0.9921, + "step": 8226 + }, + { + "epoch": 0.2, + "learning_rate": 1.855857203945197e-05, + "loss": 1.3269, + "step": 8227 + }, + { + "epoch": 0.2, + "learning_rate": 1.8558169159817316e-05, + "loss": 1.0105, + "step": 8228 + }, + { + "epoch": 0.2, + "learning_rate": 1.8557766228262265e-05, + "loss": 0.8189, + "step": 8229 + }, + { + "epoch": 0.2, + "learning_rate": 1.855736324478926e-05, + "loss": 0.9424, + "step": 8230 + }, + { + "epoch": 0.2, + "learning_rate": 1.8556960209400743e-05, + "loss": 1.0622, + "step": 8231 + }, + { + "epoch": 0.2, + "learning_rate": 1.8556557122099163e-05, + "loss": 1.1315, + "step": 8232 + }, + { + "epoch": 0.2, + "learning_rate": 1.855615398288696e-05, + "loss": 1.3393, + "step": 8233 + }, + { + "epoch": 0.2, + "learning_rate": 1.8555750791766588e-05, + "loss": 1.0584, + "step": 8234 + }, + { + "epoch": 0.2, + "learning_rate": 1.8555347548740487e-05, + "loss": 1.0881, + "step": 8235 + }, + { + "epoch": 0.2, + "learning_rate": 1.8554944253811103e-05, + "loss": 1.3192, + "step": 8236 + }, + { + "epoch": 0.2, + "learning_rate": 1.8554540906980884e-05, + "loss": 0.9091, + "step": 8237 + }, + { + "epoch": 0.2, + "learning_rate": 1.8554137508252278e-05, + "loss": 1.294, + "step": 8238 + }, + { + "epoch": 0.2, + "learning_rate": 1.8553734057627735e-05, + "loss": 1.3005, + "step": 8239 + }, + { + "epoch": 0.2, + "learning_rate": 1.8553330555109696e-05, + "loss": 1.2445, + "step": 8240 + }, + { + "epoch": 0.2, + "learning_rate": 1.8552927000700615e-05, + "loss": 1.1403, + "step": 8241 + }, + { + "epoch": 0.2, + "learning_rate": 1.8552523394402933e-05, + "loss": 1.0805, + "step": 8242 + }, + { + "epoch": 0.2, + "learning_rate": 1.855211973621911e-05, + "loss": 1.1011, + "step": 8243 + }, + { + "epoch": 0.2, + "learning_rate": 1.8551716026151584e-05, + "loss": 1.3042, + "step": 8244 + }, + { + "epoch": 0.2, + "learning_rate": 1.855131226420281e-05, + "loss": 1.1765, + "step": 8245 + }, + { + "epoch": 0.2, + "learning_rate": 1.8550908450375232e-05, + "loss": 1.3028, + "step": 8246 + }, + { + "epoch": 0.2, + "learning_rate": 1.8550504584671308e-05, + "loss": 1.3601, + "step": 8247 + }, + { + "epoch": 0.2, + "learning_rate": 1.855010066709348e-05, + "loss": 0.8476, + "step": 8248 + }, + { + "epoch": 0.2, + "learning_rate": 1.8549696697644205e-05, + "loss": 1.1443, + "step": 8249 + }, + { + "epoch": 0.2, + "learning_rate": 1.8549292676325935e-05, + "loss": 1.0806, + "step": 8250 + }, + { + "epoch": 0.2, + "learning_rate": 1.854888860314111e-05, + "loss": 1.0855, + "step": 8251 + }, + { + "epoch": 0.2, + "learning_rate": 1.8548484478092192e-05, + "loss": 1.348, + "step": 8252 + }, + { + "epoch": 0.2, + "learning_rate": 1.854808030118163e-05, + "loss": 1.2695, + "step": 8253 + }, + { + "epoch": 0.2, + "learning_rate": 1.8547676072411874e-05, + "loss": 0.8687, + "step": 8254 + }, + { + "epoch": 0.2, + "learning_rate": 1.854727179178538e-05, + "loss": 1.1956, + "step": 8255 + }, + { + "epoch": 0.2, + "learning_rate": 1.8546867459304595e-05, + "loss": 1.0228, + "step": 8256 + }, + { + "epoch": 0.2, + "learning_rate": 1.854646307497198e-05, + "loss": 1.1208, + "step": 8257 + }, + { + "epoch": 0.2, + "learning_rate": 1.8546058638789982e-05, + "loss": 0.7364, + "step": 8258 + }, + { + "epoch": 0.2, + "learning_rate": 1.854565415076105e-05, + "loss": 1.3751, + "step": 8259 + }, + { + "epoch": 0.2, + "learning_rate": 1.8545249610887653e-05, + "loss": 0.8483, + "step": 8260 + }, + { + "epoch": 0.2, + "learning_rate": 1.8544845019172234e-05, + "loss": 1.086, + "step": 8261 + }, + { + "epoch": 0.2, + "learning_rate": 1.854444037561725e-05, + "loss": 1.1679, + "step": 8262 + }, + { + "epoch": 0.2, + "learning_rate": 1.8544035680225153e-05, + "loss": 1.3676, + "step": 8263 + }, + { + "epoch": 0.2, + "learning_rate": 1.8543630932998408e-05, + "loss": 1.0052, + "step": 8264 + }, + { + "epoch": 0.2, + "learning_rate": 1.8543226133939457e-05, + "loss": 1.051, + "step": 8265 + }, + { + "epoch": 0.2, + "learning_rate": 1.8542821283050766e-05, + "loss": 1.135, + "step": 8266 + }, + { + "epoch": 0.2, + "learning_rate": 1.8542416380334786e-05, + "loss": 1.0432, + "step": 8267 + }, + { + "epoch": 0.2, + "learning_rate": 1.8542011425793976e-05, + "loss": 1.52, + "step": 8268 + }, + { + "epoch": 0.2, + "learning_rate": 1.8541606419430792e-05, + "loss": 0.9384, + "step": 8269 + }, + { + "epoch": 0.2, + "learning_rate": 1.854120136124769e-05, + "loss": 1.3107, + "step": 8270 + }, + { + "epoch": 0.2, + "learning_rate": 1.854079625124713e-05, + "loss": 1.3141, + "step": 8271 + }, + { + "epoch": 0.2, + "learning_rate": 1.8540391089431566e-05, + "loss": 1.1446, + "step": 8272 + }, + { + "epoch": 0.2, + "learning_rate": 1.8539985875803462e-05, + "loss": 1.3044, + "step": 8273 + }, + { + "epoch": 0.2, + "learning_rate": 1.8539580610365267e-05, + "loss": 1.2749, + "step": 8274 + }, + { + "epoch": 0.2, + "learning_rate": 1.8539175293119446e-05, + "loss": 1.0608, + "step": 8275 + }, + { + "epoch": 0.2, + "learning_rate": 1.853876992406846e-05, + "loss": 0.8558, + "step": 8276 + }, + { + "epoch": 0.2, + "learning_rate": 1.8538364503214766e-05, + "loss": 1.0684, + "step": 8277 + }, + { + "epoch": 0.2, + "learning_rate": 1.853795903056082e-05, + "loss": 1.3746, + "step": 8278 + }, + { + "epoch": 0.2, + "learning_rate": 1.8537553506109086e-05, + "loss": 1.0022, + "step": 8279 + }, + { + "epoch": 0.2, + "learning_rate": 1.8537147929862023e-05, + "loss": 1.1242, + "step": 8280 + }, + { + "epoch": 0.2, + "learning_rate": 1.853674230182209e-05, + "loss": 1.0716, + "step": 8281 + }, + { + "epoch": 0.2, + "learning_rate": 1.853633662199175e-05, + "loss": 1.2987, + "step": 8282 + }, + { + "epoch": 0.2, + "learning_rate": 1.8535930890373467e-05, + "loss": 1.3155, + "step": 8283 + }, + { + "epoch": 0.2, + "learning_rate": 1.8535525106969694e-05, + "loss": 0.7417, + "step": 8284 + }, + { + "epoch": 0.2, + "learning_rate": 1.85351192717829e-05, + "loss": 1.056, + "step": 8285 + }, + { + "epoch": 0.2, + "learning_rate": 1.8534713384815548e-05, + "loss": 1.1666, + "step": 8286 + }, + { + "epoch": 0.2, + "learning_rate": 1.8534307446070094e-05, + "loss": 1.1595, + "step": 8287 + }, + { + "epoch": 0.2, + "learning_rate": 1.8533901455549005e-05, + "loss": 1.2561, + "step": 8288 + }, + { + "epoch": 0.2, + "learning_rate": 1.853349541325474e-05, + "loss": 1.1954, + "step": 8289 + }, + { + "epoch": 0.2, + "learning_rate": 1.853308931918977e-05, + "loss": 1.1354, + "step": 8290 + }, + { + "epoch": 0.2, + "learning_rate": 1.8532683173356555e-05, + "loss": 1.0052, + "step": 8291 + }, + { + "epoch": 0.2, + "learning_rate": 1.8532276975757554e-05, + "loss": 1.1474, + "step": 8292 + }, + { + "epoch": 0.2, + "learning_rate": 1.8531870726395238e-05, + "loss": 1.2242, + "step": 8293 + }, + { + "epoch": 0.2, + "learning_rate": 1.8531464425272067e-05, + "loss": 1.0193, + "step": 8294 + }, + { + "epoch": 0.2, + "learning_rate": 1.853105807239051e-05, + "loss": 0.9425, + "step": 8295 + }, + { + "epoch": 0.2, + "learning_rate": 1.853065166775303e-05, + "loss": 1.1546, + "step": 8296 + }, + { + "epoch": 0.2, + "learning_rate": 1.8530245211362094e-05, + "loss": 1.1797, + "step": 8297 + }, + { + "epoch": 0.2, + "learning_rate": 1.8529838703220164e-05, + "loss": 1.1387, + "step": 8298 + }, + { + "epoch": 0.2, + "learning_rate": 1.852943214332971e-05, + "loss": 1.1331, + "step": 8299 + }, + { + "epoch": 0.2, + "learning_rate": 1.85290255316932e-05, + "loss": 1.3726, + "step": 8300 + }, + { + "epoch": 0.2, + "learning_rate": 1.8528618868313095e-05, + "loss": 1.129, + "step": 8301 + }, + { + "epoch": 0.2, + "learning_rate": 1.8528212153191868e-05, + "loss": 1.0583, + "step": 8302 + }, + { + "epoch": 0.2, + "learning_rate": 1.852780538633198e-05, + "loss": 1.0912, + "step": 8303 + }, + { + "epoch": 0.2, + "learning_rate": 1.8527398567735904e-05, + "loss": 1.3017, + "step": 8304 + }, + { + "epoch": 0.2, + "learning_rate": 1.8526991697406108e-05, + "loss": 1.318, + "step": 8305 + }, + { + "epoch": 0.2, + "learning_rate": 1.852658477534506e-05, + "loss": 0.9504, + "step": 8306 + }, + { + "epoch": 0.2, + "learning_rate": 1.8526177801555226e-05, + "loss": 1.1537, + "step": 8307 + }, + { + "epoch": 0.2, + "learning_rate": 1.8525770776039077e-05, + "loss": 1.3499, + "step": 8308 + }, + { + "epoch": 0.2, + "learning_rate": 1.8525363698799084e-05, + "loss": 0.7566, + "step": 8309 + }, + { + "epoch": 0.2, + "learning_rate": 1.852495656983771e-05, + "loss": 1.0429, + "step": 8310 + }, + { + "epoch": 0.2, + "learning_rate": 1.8524549389157432e-05, + "loss": 1.0989, + "step": 8311 + }, + { + "epoch": 0.2, + "learning_rate": 1.852414215676072e-05, + "loss": 1.2215, + "step": 8312 + }, + { + "epoch": 0.2, + "learning_rate": 1.8523734872650042e-05, + "loss": 1.0322, + "step": 8313 + }, + { + "epoch": 0.2, + "learning_rate": 1.8523327536827866e-05, + "loss": 1.0626, + "step": 8314 + }, + { + "epoch": 0.2, + "learning_rate": 1.8522920149296672e-05, + "loss": 1.4848, + "step": 8315 + }, + { + "epoch": 0.2, + "learning_rate": 1.852251271005892e-05, + "loss": 1.2855, + "step": 8316 + }, + { + "epoch": 0.2, + "learning_rate": 1.8522105219117093e-05, + "loss": 1.1199, + "step": 8317 + }, + { + "epoch": 0.2, + "learning_rate": 1.852169767647366e-05, + "loss": 1.0824, + "step": 8318 + }, + { + "epoch": 0.2, + "learning_rate": 1.8521290082131084e-05, + "loss": 1.1586, + "step": 8319 + }, + { + "epoch": 0.2, + "learning_rate": 1.852088243609185e-05, + "loss": 1.4044, + "step": 8320 + }, + { + "epoch": 0.2, + "learning_rate": 1.8520474738358426e-05, + "loss": 1.1794, + "step": 8321 + }, + { + "epoch": 0.2, + "learning_rate": 1.8520066988933286e-05, + "loss": 1.2893, + "step": 8322 + }, + { + "epoch": 0.2, + "learning_rate": 1.85196591878189e-05, + "loss": 0.981, + "step": 8323 + }, + { + "epoch": 0.2, + "learning_rate": 1.8519251335017753e-05, + "loss": 1.0644, + "step": 8324 + }, + { + "epoch": 0.2, + "learning_rate": 1.8518843430532306e-05, + "loss": 1.1786, + "step": 8325 + }, + { + "epoch": 0.2, + "learning_rate": 1.8518435474365045e-05, + "loss": 0.9506, + "step": 8326 + }, + { + "epoch": 0.2, + "learning_rate": 1.8518027466518434e-05, + "loss": 1.1911, + "step": 8327 + }, + { + "epoch": 0.2, + "learning_rate": 1.8517619406994956e-05, + "loss": 1.0676, + "step": 8328 + }, + { + "epoch": 0.2, + "learning_rate": 1.8517211295797085e-05, + "loss": 1.2909, + "step": 8329 + }, + { + "epoch": 0.2, + "learning_rate": 1.8516803132927296e-05, + "loss": 0.9372, + "step": 8330 + }, + { + "epoch": 0.2, + "learning_rate": 1.8516394918388068e-05, + "loss": 1.1021, + "step": 8331 + }, + { + "epoch": 0.2, + "learning_rate": 1.8515986652181873e-05, + "loss": 1.1401, + "step": 8332 + }, + { + "epoch": 0.2, + "learning_rate": 1.851557833431119e-05, + "loss": 1.2219, + "step": 8333 + }, + { + "epoch": 0.2, + "learning_rate": 1.8515169964778496e-05, + "loss": 1.263, + "step": 8334 + }, + { + "epoch": 0.2, + "learning_rate": 1.8514761543586268e-05, + "loss": 1.0141, + "step": 8335 + }, + { + "epoch": 0.2, + "learning_rate": 1.8514353070736987e-05, + "loss": 1.4663, + "step": 8336 + }, + { + "epoch": 0.2, + "learning_rate": 1.8513944546233126e-05, + "loss": 1.4865, + "step": 8337 + }, + { + "epoch": 0.2, + "learning_rate": 1.851353597007717e-05, + "loss": 1.3606, + "step": 8338 + }, + { + "epoch": 0.2, + "learning_rate": 1.851312734227159e-05, + "loss": 1.2025, + "step": 8339 + }, + { + "epoch": 0.2, + "learning_rate": 1.851271866281887e-05, + "loss": 1.0654, + "step": 8340 + }, + { + "epoch": 0.2, + "learning_rate": 1.8512309931721487e-05, + "loss": 1.0538, + "step": 8341 + }, + { + "epoch": 0.2, + "learning_rate": 1.8511901148981922e-05, + "loss": 1.3097, + "step": 8342 + }, + { + "epoch": 0.2, + "learning_rate": 1.851149231460266e-05, + "loss": 1.2439, + "step": 8343 + }, + { + "epoch": 0.2, + "learning_rate": 1.8511083428586172e-05, + "loss": 0.982, + "step": 8344 + }, + { + "epoch": 0.2, + "learning_rate": 1.851067449093494e-05, + "loss": 1.3051, + "step": 8345 + }, + { + "epoch": 0.2, + "learning_rate": 1.8510265501651454e-05, + "loss": 1.2874, + "step": 8346 + }, + { + "epoch": 0.2, + "learning_rate": 1.850985646073818e-05, + "loss": 1.2918, + "step": 8347 + }, + { + "epoch": 0.2, + "learning_rate": 1.8509447368197617e-05, + "loss": 1.1206, + "step": 8348 + }, + { + "epoch": 0.2, + "learning_rate": 1.8509038224032238e-05, + "loss": 0.9734, + "step": 8349 + }, + { + "epoch": 0.2, + "learning_rate": 1.850862902824452e-05, + "loss": 0.8313, + "step": 8350 + }, + { + "epoch": 0.2, + "learning_rate": 1.8508219780836952e-05, + "loss": 1.3226, + "step": 8351 + }, + { + "epoch": 0.2, + "learning_rate": 1.850781048181202e-05, + "loss": 1.2363, + "step": 8352 + }, + { + "epoch": 0.2, + "learning_rate": 1.8507401131172198e-05, + "loss": 1.3117, + "step": 8353 + }, + { + "epoch": 0.2, + "learning_rate": 1.850699172891998e-05, + "loss": 1.1956, + "step": 8354 + }, + { + "epoch": 0.2, + "learning_rate": 1.850658227505784e-05, + "loss": 1.2891, + "step": 8355 + }, + { + "epoch": 0.2, + "learning_rate": 1.850617276958827e-05, + "loss": 0.9628, + "step": 8356 + }, + { + "epoch": 0.2, + "learning_rate": 1.8505763212513747e-05, + "loss": 1.2339, + "step": 8357 + }, + { + "epoch": 0.2, + "learning_rate": 1.8505353603836756e-05, + "loss": 1.0428, + "step": 8358 + }, + { + "epoch": 0.2, + "learning_rate": 1.8504943943559793e-05, + "loss": 0.8993, + "step": 8359 + }, + { + "epoch": 0.2, + "learning_rate": 1.8504534231685332e-05, + "loss": 1.2602, + "step": 8360 + }, + { + "epoch": 0.2, + "learning_rate": 1.850412446821586e-05, + "loss": 1.3232, + "step": 8361 + }, + { + "epoch": 0.2, + "learning_rate": 1.8503714653153867e-05, + "loss": 1.2397, + "step": 8362 + }, + { + "epoch": 0.2, + "learning_rate": 1.8503304786501842e-05, + "loss": 1.1427, + "step": 8363 + }, + { + "epoch": 0.2, + "learning_rate": 1.850289486826226e-05, + "loss": 1.0434, + "step": 8364 + }, + { + "epoch": 0.2, + "learning_rate": 1.850248489843762e-05, + "loss": 1.2279, + "step": 8365 + }, + { + "epoch": 0.2, + "learning_rate": 1.85020748770304e-05, + "loss": 1.0918, + "step": 8366 + }, + { + "epoch": 0.2, + "learning_rate": 1.8501664804043092e-05, + "loss": 1.0062, + "step": 8367 + }, + { + "epoch": 0.2, + "learning_rate": 1.8501254679478188e-05, + "loss": 1.5109, + "step": 8368 + }, + { + "epoch": 0.2, + "learning_rate": 1.8500844503338166e-05, + "loss": 0.9626, + "step": 8369 + }, + { + "epoch": 0.2, + "learning_rate": 1.850043427562552e-05, + "loss": 0.8064, + "step": 8370 + }, + { + "epoch": 0.2, + "learning_rate": 1.8500023996342744e-05, + "loss": 0.9771, + "step": 8371 + }, + { + "epoch": 0.2, + "learning_rate": 1.8499613665492317e-05, + "loss": 1.0865, + "step": 8372 + }, + { + "epoch": 0.2, + "learning_rate": 1.8499203283076733e-05, + "loss": 1.3694, + "step": 8373 + }, + { + "epoch": 0.2, + "learning_rate": 1.8498792849098482e-05, + "loss": 1.2001, + "step": 8374 + }, + { + "epoch": 0.2, + "learning_rate": 1.8498382363560056e-05, + "loss": 0.801, + "step": 8375 + }, + { + "epoch": 0.2, + "learning_rate": 1.8497971826463943e-05, + "loss": 1.1535, + "step": 8376 + }, + { + "epoch": 0.2, + "learning_rate": 1.8497561237812632e-05, + "loss": 1.24, + "step": 8377 + }, + { + "epoch": 0.2, + "learning_rate": 1.849715059760862e-05, + "loss": 1.1556, + "step": 8378 + }, + { + "epoch": 0.2, + "learning_rate": 1.849673990585439e-05, + "loss": 1.3042, + "step": 8379 + }, + { + "epoch": 0.2, + "learning_rate": 1.8496329162552437e-05, + "loss": 1.2044, + "step": 8380 + }, + { + "epoch": 0.2, + "learning_rate": 1.8495918367705255e-05, + "loss": 1.204, + "step": 8381 + }, + { + "epoch": 0.2, + "learning_rate": 1.8495507521315333e-05, + "loss": 1.4922, + "step": 8382 + }, + { + "epoch": 0.2, + "learning_rate": 1.849509662338517e-05, + "loss": 1.0845, + "step": 8383 + }, + { + "epoch": 0.2, + "learning_rate": 1.8494685673917248e-05, + "loss": 0.9606, + "step": 8384 + }, + { + "epoch": 0.2, + "learning_rate": 1.849427467291407e-05, + "loss": 1.1278, + "step": 8385 + }, + { + "epoch": 0.2, + "learning_rate": 1.8493863620378123e-05, + "loss": 1.3199, + "step": 8386 + }, + { + "epoch": 0.2, + "learning_rate": 1.8493452516311905e-05, + "loss": 1.2522, + "step": 8387 + }, + { + "epoch": 0.2, + "learning_rate": 1.8493041360717908e-05, + "loss": 0.9882, + "step": 8388 + }, + { + "epoch": 0.2, + "learning_rate": 1.8492630153598625e-05, + "loss": 0.9697, + "step": 8389 + }, + { + "epoch": 0.2, + "learning_rate": 1.8492218894956555e-05, + "loss": 1.2091, + "step": 8390 + }, + { + "epoch": 0.2, + "learning_rate": 1.849180758479419e-05, + "loss": 1.325, + "step": 8391 + }, + { + "epoch": 0.2, + "learning_rate": 1.8491396223114024e-05, + "loss": 1.1625, + "step": 8392 + }, + { + "epoch": 0.2, + "learning_rate": 1.8490984809918555e-05, + "loss": 1.1247, + "step": 8393 + }, + { + "epoch": 0.2, + "learning_rate": 1.849057334521028e-05, + "loss": 1.1634, + "step": 8394 + }, + { + "epoch": 0.2, + "learning_rate": 1.8490161828991692e-05, + "loss": 1.2357, + "step": 8395 + }, + { + "epoch": 0.2, + "learning_rate": 1.848975026126529e-05, + "loss": 1.1839, + "step": 8396 + }, + { + "epoch": 0.2, + "learning_rate": 1.848933864203357e-05, + "loss": 1.232, + "step": 8397 + }, + { + "epoch": 0.2, + "learning_rate": 1.848892697129903e-05, + "loss": 1.1483, + "step": 8398 + }, + { + "epoch": 0.2, + "learning_rate": 1.8488515249064165e-05, + "loss": 1.1695, + "step": 8399 + }, + { + "epoch": 0.2, + "learning_rate": 1.8488103475331476e-05, + "loss": 1.1422, + "step": 8400 + }, + { + "epoch": 0.2, + "learning_rate": 1.848769165010346e-05, + "loss": 1.4388, + "step": 8401 + }, + { + "epoch": 0.2, + "learning_rate": 1.8487279773382613e-05, + "loss": 1.133, + "step": 8402 + }, + { + "epoch": 0.2, + "learning_rate": 1.8486867845171437e-05, + "loss": 1.1177, + "step": 8403 + }, + { + "epoch": 0.2, + "learning_rate": 1.848645586547243e-05, + "loss": 1.3261, + "step": 8404 + }, + { + "epoch": 0.2, + "learning_rate": 1.8486043834288093e-05, + "loss": 1.0307, + "step": 8405 + }, + { + "epoch": 0.2, + "learning_rate": 1.848563175162092e-05, + "loss": 1.1201, + "step": 8406 + }, + { + "epoch": 0.2, + "learning_rate": 1.848521961747342e-05, + "loss": 1.22, + "step": 8407 + }, + { + "epoch": 0.2, + "learning_rate": 1.8484807431848085e-05, + "loss": 1.3545, + "step": 8408 + }, + { + "epoch": 0.2, + "learning_rate": 1.848439519474742e-05, + "loss": 1.1627, + "step": 8409 + }, + { + "epoch": 0.2, + "learning_rate": 1.8483982906173928e-05, + "loss": 1.08, + "step": 8410 + }, + { + "epoch": 0.2, + "learning_rate": 1.8483570566130105e-05, + "loss": 1.178, + "step": 8411 + }, + { + "epoch": 0.2, + "learning_rate": 1.8483158174618456e-05, + "loss": 1.0941, + "step": 8412 + }, + { + "epoch": 0.2, + "learning_rate": 1.8482745731641478e-05, + "loss": 1.3238, + "step": 8413 + }, + { + "epoch": 0.2, + "learning_rate": 1.8482333237201678e-05, + "loss": 0.9516, + "step": 8414 + }, + { + "epoch": 0.2, + "learning_rate": 1.848192069130156e-05, + "loss": 1.1796, + "step": 8415 + }, + { + "epoch": 0.2, + "learning_rate": 1.8481508093943622e-05, + "loss": 1.3272, + "step": 8416 + }, + { + "epoch": 0.2, + "learning_rate": 1.848109544513037e-05, + "loss": 0.8931, + "step": 8417 + }, + { + "epoch": 0.2, + "learning_rate": 1.8480682744864306e-05, + "loss": 1.2551, + "step": 8418 + }, + { + "epoch": 0.2, + "learning_rate": 1.848026999314794e-05, + "loss": 1.2976, + "step": 8419 + }, + { + "epoch": 0.2, + "learning_rate": 1.8479857189983762e-05, + "loss": 1.1771, + "step": 8420 + }, + { + "epoch": 0.2, + "learning_rate": 1.847944433537429e-05, + "loss": 1.2269, + "step": 8421 + }, + { + "epoch": 0.2, + "learning_rate": 1.8479031429322022e-05, + "loss": 1.1349, + "step": 8422 + }, + { + "epoch": 0.2, + "learning_rate": 1.8478618471829467e-05, + "loss": 0.805, + "step": 8423 + }, + { + "epoch": 0.2, + "learning_rate": 1.8478205462899127e-05, + "loss": 1.0979, + "step": 8424 + }, + { + "epoch": 0.2, + "learning_rate": 1.847779240253351e-05, + "loss": 1.302, + "step": 8425 + }, + { + "epoch": 0.2, + "learning_rate": 1.8477379290735117e-05, + "loss": 1.4143, + "step": 8426 + }, + { + "epoch": 0.2, + "learning_rate": 1.8476966127506458e-05, + "loss": 0.9057, + "step": 8427 + }, + { + "epoch": 0.2, + "learning_rate": 1.8476552912850044e-05, + "loss": 1.1095, + "step": 8428 + }, + { + "epoch": 0.2, + "learning_rate": 1.8476139646768372e-05, + "loss": 1.207, + "step": 8429 + }, + { + "epoch": 0.2, + "learning_rate": 1.8475726329263958e-05, + "loss": 1.219, + "step": 8430 + }, + { + "epoch": 0.2, + "learning_rate": 1.8475312960339303e-05, + "loss": 1.1208, + "step": 8431 + }, + { + "epoch": 0.2, + "learning_rate": 1.8474899539996923e-05, + "loss": 1.0786, + "step": 8432 + }, + { + "epoch": 0.2, + "learning_rate": 1.8474486068239318e-05, + "loss": 1.0838, + "step": 8433 + }, + { + "epoch": 0.2, + "learning_rate": 1.8474072545068998e-05, + "loss": 1.2421, + "step": 8434 + }, + { + "epoch": 0.2, + "learning_rate": 1.8473658970488474e-05, + "loss": 1.2335, + "step": 8435 + }, + { + "epoch": 0.2, + "learning_rate": 1.8473245344500254e-05, + "loss": 1.2494, + "step": 8436 + }, + { + "epoch": 0.2, + "learning_rate": 1.8472831667106844e-05, + "loss": 1.2873, + "step": 8437 + }, + { + "epoch": 0.2, + "learning_rate": 1.8472417938310765e-05, + "loss": 1.2322, + "step": 8438 + }, + { + "epoch": 0.2, + "learning_rate": 1.8472004158114515e-05, + "loss": 0.9607, + "step": 8439 + }, + { + "epoch": 0.2, + "learning_rate": 1.8471590326520607e-05, + "loss": 1.304, + "step": 8440 + }, + { + "epoch": 0.2, + "learning_rate": 1.8471176443531555e-05, + "loss": 1.0671, + "step": 8441 + }, + { + "epoch": 0.2, + "learning_rate": 1.8470762509149867e-05, + "loss": 1.3289, + "step": 8442 + }, + { + "epoch": 0.2, + "learning_rate": 1.8470348523378056e-05, + "loss": 1.2382, + "step": 8443 + }, + { + "epoch": 0.2, + "learning_rate": 1.8469934486218634e-05, + "loss": 1.2574, + "step": 8444 + }, + { + "epoch": 0.2, + "learning_rate": 1.8469520397674113e-05, + "loss": 1.125, + "step": 8445 + }, + { + "epoch": 0.2, + "learning_rate": 1.8469106257747002e-05, + "loss": 1.0794, + "step": 8446 + }, + { + "epoch": 0.2, + "learning_rate": 1.846869206643981e-05, + "loss": 1.325, + "step": 8447 + }, + { + "epoch": 0.2, + "learning_rate": 1.846827782375506e-05, + "loss": 1.1787, + "step": 8448 + }, + { + "epoch": 0.2, + "learning_rate": 1.8467863529695264e-05, + "loss": 1.1191, + "step": 8449 + }, + { + "epoch": 0.2, + "learning_rate": 1.8467449184262927e-05, + "loss": 1.0924, + "step": 8450 + }, + { + "epoch": 0.2, + "learning_rate": 1.846703478746057e-05, + "loss": 1.0251, + "step": 8451 + }, + { + "epoch": 0.2, + "learning_rate": 1.84666203392907e-05, + "loss": 1.3647, + "step": 8452 + }, + { + "epoch": 0.2, + "learning_rate": 1.8466205839755843e-05, + "loss": 0.971, + "step": 8453 + }, + { + "epoch": 0.2, + "learning_rate": 1.84657912888585e-05, + "loss": 0.8847, + "step": 8454 + }, + { + "epoch": 0.2, + "learning_rate": 1.8465376686601196e-05, + "loss": 1.183, + "step": 8455 + }, + { + "epoch": 0.2, + "learning_rate": 1.846496203298644e-05, + "loss": 1.2275, + "step": 8456 + }, + { + "epoch": 0.2, + "learning_rate": 1.8464547328016756e-05, + "loss": 1.1581, + "step": 8457 + }, + { + "epoch": 0.2, + "learning_rate": 1.8464132571694648e-05, + "loss": 1.1302, + "step": 8458 + }, + { + "epoch": 0.2, + "learning_rate": 1.846371776402264e-05, + "loss": 1.2311, + "step": 8459 + }, + { + "epoch": 0.2, + "learning_rate": 1.8463302905003247e-05, + "loss": 1.2998, + "step": 8460 + }, + { + "epoch": 0.2, + "learning_rate": 1.8462887994638992e-05, + "loss": 1.2506, + "step": 8461 + }, + { + "epoch": 0.2, + "learning_rate": 1.846247303293238e-05, + "loss": 0.9863, + "step": 8462 + }, + { + "epoch": 0.2, + "learning_rate": 1.8462058019885937e-05, + "loss": 1.1795, + "step": 8463 + }, + { + "epoch": 0.2, + "learning_rate": 1.846164295550218e-05, + "loss": 1.069, + "step": 8464 + }, + { + "epoch": 0.2, + "learning_rate": 1.846122783978362e-05, + "loss": 1.1167, + "step": 8465 + }, + { + "epoch": 0.2, + "learning_rate": 1.846081267273278e-05, + "loss": 1.0614, + "step": 8466 + }, + { + "epoch": 0.2, + "learning_rate": 1.846039745435219e-05, + "loss": 1.1179, + "step": 8467 + }, + { + "epoch": 0.2, + "learning_rate": 1.845998218464435e-05, + "loss": 1.3042, + "step": 8468 + }, + { + "epoch": 0.2, + "learning_rate": 1.845956686361179e-05, + "loss": 1.252, + "step": 8469 + }, + { + "epoch": 0.2, + "learning_rate": 1.845915149125703e-05, + "loss": 1.1907, + "step": 8470 + }, + { + "epoch": 0.2, + "learning_rate": 1.8458736067582582e-05, + "loss": 1.1017, + "step": 8471 + }, + { + "epoch": 0.2, + "learning_rate": 1.8458320592590976e-05, + "loss": 1.2694, + "step": 8472 + }, + { + "epoch": 0.2, + "learning_rate": 1.8457905066284728e-05, + "loss": 1.2013, + "step": 8473 + }, + { + "epoch": 0.2, + "learning_rate": 1.845748948866636e-05, + "loss": 0.9785, + "step": 8474 + }, + { + "epoch": 0.2, + "learning_rate": 1.845707385973839e-05, + "loss": 1.0435, + "step": 8475 + }, + { + "epoch": 0.2, + "learning_rate": 1.8456658179503345e-05, + "loss": 1.0594, + "step": 8476 + }, + { + "epoch": 0.2, + "learning_rate": 1.845624244796374e-05, + "loss": 0.8331, + "step": 8477 + }, + { + "epoch": 0.2, + "learning_rate": 1.8455826665122107e-05, + "loss": 1.0277, + "step": 8478 + }, + { + "epoch": 0.2, + "learning_rate": 1.8455410830980958e-05, + "loss": 1.1341, + "step": 8479 + }, + { + "epoch": 0.2, + "learning_rate": 1.845499494554282e-05, + "loss": 1.3769, + "step": 8480 + }, + { + "epoch": 0.2, + "learning_rate": 1.845457900881022e-05, + "loss": 1.365, + "step": 8481 + }, + { + "epoch": 0.2, + "learning_rate": 1.8454163020785676e-05, + "loss": 1.0962, + "step": 8482 + }, + { + "epoch": 0.2, + "learning_rate": 1.8453746981471714e-05, + "loss": 1.2083, + "step": 8483 + }, + { + "epoch": 0.2, + "learning_rate": 1.8453330890870857e-05, + "loss": 1.0037, + "step": 8484 + }, + { + "epoch": 0.2, + "learning_rate": 1.845291474898563e-05, + "loss": 0.9851, + "step": 8485 + }, + { + "epoch": 0.2, + "learning_rate": 1.8452498555818556e-05, + "loss": 1.0258, + "step": 8486 + }, + { + "epoch": 0.2, + "learning_rate": 1.845208231137216e-05, + "loss": 1.207, + "step": 8487 + }, + { + "epoch": 0.2, + "learning_rate": 1.8451666015648976e-05, + "loss": 1.2012, + "step": 8488 + }, + { + "epoch": 0.2, + "learning_rate": 1.8451249668651513e-05, + "loss": 1.216, + "step": 8489 + }, + { + "epoch": 0.2, + "learning_rate": 1.8450833270382312e-05, + "loss": 1.0958, + "step": 8490 + }, + { + "epoch": 0.2, + "learning_rate": 1.8450416820843894e-05, + "loss": 1.1928, + "step": 8491 + }, + { + "epoch": 0.2, + "learning_rate": 1.845000032003878e-05, + "loss": 1.04, + "step": 8492 + }, + { + "epoch": 0.2, + "learning_rate": 1.8449583767969507e-05, + "loss": 0.9048, + "step": 8493 + }, + { + "epoch": 0.2, + "learning_rate": 1.8449167164638596e-05, + "loss": 1.2754, + "step": 8494 + }, + { + "epoch": 0.2, + "learning_rate": 1.844875051004857e-05, + "loss": 1.1877, + "step": 8495 + }, + { + "epoch": 0.2, + "learning_rate": 1.844833380420197e-05, + "loss": 1.0087, + "step": 8496 + }, + { + "epoch": 0.2, + "learning_rate": 1.8447917047101307e-05, + "loss": 1.117, + "step": 8497 + }, + { + "epoch": 0.2, + "learning_rate": 1.8447500238749124e-05, + "loss": 1.3166, + "step": 8498 + }, + { + "epoch": 0.2, + "learning_rate": 1.8447083379147944e-05, + "loss": 1.3303, + "step": 8499 + }, + { + "epoch": 0.2, + "learning_rate": 1.8446666468300292e-05, + "loss": 1.064, + "step": 8500 + }, + { + "epoch": 0.2, + "learning_rate": 1.8446249506208706e-05, + "loss": 0.9657, + "step": 8501 + }, + { + "epoch": 0.2, + "learning_rate": 1.844583249287571e-05, + "loss": 1.3327, + "step": 8502 + }, + { + "epoch": 0.2, + "learning_rate": 1.844541542830383e-05, + "loss": 0.995, + "step": 8503 + }, + { + "epoch": 0.2, + "learning_rate": 1.844499831249561e-05, + "loss": 0.9657, + "step": 8504 + }, + { + "epoch": 0.2, + "learning_rate": 1.8444581145453563e-05, + "loss": 1.3023, + "step": 8505 + }, + { + "epoch": 0.2, + "learning_rate": 1.8444163927180232e-05, + "loss": 1.094, + "step": 8506 + }, + { + "epoch": 0.2, + "learning_rate": 1.8443746657678146e-05, + "loss": 0.8603, + "step": 8507 + }, + { + "epoch": 0.2, + "learning_rate": 1.8443329336949835e-05, + "loss": 1.125, + "step": 8508 + }, + { + "epoch": 0.2, + "learning_rate": 1.8442911964997826e-05, + "loss": 1.3188, + "step": 8509 + }, + { + "epoch": 0.2, + "learning_rate": 1.844249454182466e-05, + "loss": 1.0757, + "step": 8510 + }, + { + "epoch": 0.2, + "learning_rate": 1.8442077067432867e-05, + "loss": 0.8957, + "step": 8511 + }, + { + "epoch": 0.2, + "learning_rate": 1.8441659541824975e-05, + "loss": 1.1931, + "step": 8512 + }, + { + "epoch": 0.2, + "learning_rate": 1.844124196500352e-05, + "loss": 1.4426, + "step": 8513 + }, + { + "epoch": 0.2, + "learning_rate": 1.844082433697104e-05, + "loss": 0.9953, + "step": 8514 + }, + { + "epoch": 0.2, + "learning_rate": 1.844040665773006e-05, + "loss": 1.3691, + "step": 8515 + }, + { + "epoch": 0.2, + "learning_rate": 1.843998892728312e-05, + "loss": 1.1639, + "step": 8516 + }, + { + "epoch": 0.2, + "learning_rate": 1.843957114563275e-05, + "loss": 1.0262, + "step": 8517 + }, + { + "epoch": 0.2, + "learning_rate": 1.8439153312781487e-05, + "loss": 1.1805, + "step": 8518 + }, + { + "epoch": 0.2, + "learning_rate": 1.8438735428731868e-05, + "loss": 1.3173, + "step": 8519 + }, + { + "epoch": 0.2, + "learning_rate": 1.8438317493486426e-05, + "loss": 1.0585, + "step": 8520 + }, + { + "epoch": 0.2, + "learning_rate": 1.8437899507047694e-05, + "loss": 1.1723, + "step": 8521 + }, + { + "epoch": 0.2, + "learning_rate": 1.843748146941821e-05, + "loss": 1.165, + "step": 8522 + }, + { + "epoch": 0.2, + "learning_rate": 1.8437063380600515e-05, + "loss": 1.284, + "step": 8523 + }, + { + "epoch": 0.2, + "learning_rate": 1.843664524059714e-05, + "loss": 1.1005, + "step": 8524 + }, + { + "epoch": 0.21, + "learning_rate": 1.843622704941062e-05, + "loss": 1.3263, + "step": 8525 + }, + { + "epoch": 0.21, + "learning_rate": 1.8435808807043498e-05, + "loss": 1.315, + "step": 8526 + }, + { + "epoch": 0.21, + "learning_rate": 1.8435390513498307e-05, + "loss": 0.956, + "step": 8527 + }, + { + "epoch": 0.21, + "learning_rate": 1.8434972168777585e-05, + "loss": 0.9659, + "step": 8528 + }, + { + "epoch": 0.21, + "learning_rate": 1.843455377288387e-05, + "loss": 1.0466, + "step": 8529 + }, + { + "epoch": 0.21, + "learning_rate": 1.8434135325819703e-05, + "loss": 1.0586, + "step": 8530 + }, + { + "epoch": 0.21, + "learning_rate": 1.843371682758762e-05, + "loss": 1.0492, + "step": 8531 + }, + { + "epoch": 0.21, + "learning_rate": 1.8433298278190163e-05, + "loss": 0.9091, + "step": 8532 + }, + { + "epoch": 0.21, + "learning_rate": 1.8432879677629865e-05, + "loss": 1.2449, + "step": 8533 + }, + { + "epoch": 0.21, + "learning_rate": 1.843246102590927e-05, + "loss": 1.1353, + "step": 8534 + }, + { + "epoch": 0.21, + "learning_rate": 1.843204232303092e-05, + "loss": 1.346, + "step": 8535 + }, + { + "epoch": 0.21, + "learning_rate": 1.843162356899735e-05, + "loss": 1.3087, + "step": 8536 + }, + { + "epoch": 0.21, + "learning_rate": 1.8431204763811108e-05, + "loss": 1.2247, + "step": 8537 + }, + { + "epoch": 0.21, + "learning_rate": 1.8430785907474727e-05, + "loss": 1.2451, + "step": 8538 + }, + { + "epoch": 0.21, + "learning_rate": 1.843036699999075e-05, + "loss": 1.341, + "step": 8539 + }, + { + "epoch": 0.21, + "learning_rate": 1.842994804136172e-05, + "loss": 1.3985, + "step": 8540 + }, + { + "epoch": 0.21, + "learning_rate": 1.8429529031590178e-05, + "loss": 1.021, + "step": 8541 + }, + { + "epoch": 0.21, + "learning_rate": 1.8429109970678664e-05, + "loss": 1.2475, + "step": 8542 + }, + { + "epoch": 0.21, + "learning_rate": 1.8428690858629724e-05, + "loss": 1.075, + "step": 8543 + }, + { + "epoch": 0.21, + "learning_rate": 1.8428271695445903e-05, + "loss": 1.2771, + "step": 8544 + }, + { + "epoch": 0.21, + "learning_rate": 1.8427852481129736e-05, + "loss": 0.9912, + "step": 8545 + }, + { + "epoch": 0.21, + "learning_rate": 1.842743321568377e-05, + "loss": 1.127, + "step": 8546 + }, + { + "epoch": 0.21, + "learning_rate": 1.842701389911055e-05, + "loss": 1.4185, + "step": 8547 + }, + { + "epoch": 0.21, + "learning_rate": 1.842659453141262e-05, + "loss": 0.9982, + "step": 8548 + }, + { + "epoch": 0.21, + "learning_rate": 1.842617511259252e-05, + "loss": 0.8869, + "step": 8549 + }, + { + "epoch": 0.21, + "learning_rate": 1.8425755642652797e-05, + "loss": 1.3254, + "step": 8550 + }, + { + "epoch": 0.21, + "learning_rate": 1.8425336121596e-05, + "loss": 1.2382, + "step": 8551 + }, + { + "epoch": 0.21, + "learning_rate": 1.842491654942467e-05, + "loss": 0.9223, + "step": 8552 + }, + { + "epoch": 0.21, + "learning_rate": 1.8424496926141354e-05, + "loss": 1.439, + "step": 8553 + }, + { + "epoch": 0.21, + "learning_rate": 1.8424077251748593e-05, + "loss": 1.1423, + "step": 8554 + }, + { + "epoch": 0.21, + "learning_rate": 1.8423657526248936e-05, + "loss": 1.2423, + "step": 8555 + }, + { + "epoch": 0.21, + "learning_rate": 1.8423237749644936e-05, + "loss": 1.0429, + "step": 8556 + }, + { + "epoch": 0.21, + "learning_rate": 1.8422817921939133e-05, + "loss": 1.0594, + "step": 8557 + }, + { + "epoch": 0.21, + "learning_rate": 1.8422398043134068e-05, + "loss": 1.2381, + "step": 8558 + }, + { + "epoch": 0.21, + "learning_rate": 1.84219781132323e-05, + "loss": 1.4345, + "step": 8559 + }, + { + "epoch": 0.21, + "learning_rate": 1.842155813223637e-05, + "loss": 0.9842, + "step": 8560 + }, + { + "epoch": 0.21, + "learning_rate": 1.842113810014883e-05, + "loss": 1.0804, + "step": 8561 + }, + { + "epoch": 0.21, + "learning_rate": 1.8420718016972227e-05, + "loss": 1.2216, + "step": 8562 + }, + { + "epoch": 0.21, + "learning_rate": 1.8420297882709107e-05, + "loss": 1.0747, + "step": 8563 + }, + { + "epoch": 0.21, + "learning_rate": 1.8419877697362015e-05, + "loss": 1.1712, + "step": 8564 + }, + { + "epoch": 0.21, + "learning_rate": 1.841945746093351e-05, + "loss": 1.1163, + "step": 8565 + }, + { + "epoch": 0.21, + "learning_rate": 1.841903717342614e-05, + "loss": 0.9202, + "step": 8566 + }, + { + "epoch": 0.21, + "learning_rate": 1.841861683484245e-05, + "loss": 1.1111, + "step": 8567 + }, + { + "epoch": 0.21, + "learning_rate": 1.8418196445184988e-05, + "loss": 1.2228, + "step": 8568 + }, + { + "epoch": 0.21, + "learning_rate": 1.841777600445631e-05, + "loss": 1.3838, + "step": 8569 + }, + { + "epoch": 0.21, + "learning_rate": 1.841735551265897e-05, + "loss": 1.1568, + "step": 8570 + }, + { + "epoch": 0.21, + "learning_rate": 1.841693496979551e-05, + "loss": 1.2119, + "step": 8571 + }, + { + "epoch": 0.21, + "learning_rate": 1.8416514375868482e-05, + "loss": 0.7434, + "step": 8572 + }, + { + "epoch": 0.21, + "learning_rate": 1.8416093730880446e-05, + "loss": 1.0575, + "step": 8573 + }, + { + "epoch": 0.21, + "learning_rate": 1.8415673034833945e-05, + "loss": 1.2272, + "step": 8574 + }, + { + "epoch": 0.21, + "learning_rate": 1.841525228773154e-05, + "loss": 1.1998, + "step": 8575 + }, + { + "epoch": 0.21, + "learning_rate": 1.8414831489575774e-05, + "loss": 0.9628, + "step": 8576 + }, + { + "epoch": 0.21, + "learning_rate": 1.841441064036921e-05, + "loss": 1.1989, + "step": 8577 + }, + { + "epoch": 0.21, + "learning_rate": 1.841398974011439e-05, + "loss": 0.913, + "step": 8578 + }, + { + "epoch": 0.21, + "learning_rate": 1.8413568788813874e-05, + "loss": 1.3126, + "step": 8579 + }, + { + "epoch": 0.21, + "learning_rate": 1.8413147786470217e-05, + "loss": 1.0829, + "step": 8580 + }, + { + "epoch": 0.21, + "learning_rate": 1.8412726733085974e-05, + "loss": 1.2116, + "step": 8581 + }, + { + "epoch": 0.21, + "learning_rate": 1.8412305628663693e-05, + "loss": 1.2277, + "step": 8582 + }, + { + "epoch": 0.21, + "learning_rate": 1.8411884473205935e-05, + "loss": 1.2304, + "step": 8583 + }, + { + "epoch": 0.21, + "learning_rate": 1.841146326671525e-05, + "loss": 0.9573, + "step": 8584 + }, + { + "epoch": 0.21, + "learning_rate": 1.84110420091942e-05, + "loss": 1.1791, + "step": 8585 + }, + { + "epoch": 0.21, + "learning_rate": 1.841062070064533e-05, + "loss": 1.1713, + "step": 8586 + }, + { + "epoch": 0.21, + "learning_rate": 1.8410199341071205e-05, + "loss": 0.8691, + "step": 8587 + }, + { + "epoch": 0.21, + "learning_rate": 1.840977793047438e-05, + "loss": 1.012, + "step": 8588 + }, + { + "epoch": 0.21, + "learning_rate": 1.8409356468857412e-05, + "loss": 0.8769, + "step": 8589 + }, + { + "epoch": 0.21, + "learning_rate": 1.8408934956222855e-05, + "loss": 1.3399, + "step": 8590 + }, + { + "epoch": 0.21, + "learning_rate": 1.8408513392573266e-05, + "loss": 1.1528, + "step": 8591 + }, + { + "epoch": 0.21, + "learning_rate": 1.8408091777911203e-05, + "loss": 1.3379, + "step": 8592 + }, + { + "epoch": 0.21, + "learning_rate": 1.840767011223923e-05, + "loss": 0.891, + "step": 8593 + }, + { + "epoch": 0.21, + "learning_rate": 1.8407248395559897e-05, + "loss": 1.1323, + "step": 8594 + }, + { + "epoch": 0.21, + "learning_rate": 1.840682662787576e-05, + "loss": 0.8704, + "step": 8595 + }, + { + "epoch": 0.21, + "learning_rate": 1.840640480918939e-05, + "loss": 1.2449, + "step": 8596 + }, + { + "epoch": 0.21, + "learning_rate": 1.8405982939503338e-05, + "loss": 1.2698, + "step": 8597 + }, + { + "epoch": 0.21, + "learning_rate": 1.8405561018820166e-05, + "loss": 0.9848, + "step": 8598 + }, + { + "epoch": 0.21, + "learning_rate": 1.8405139047142433e-05, + "loss": 1.2417, + "step": 8599 + }, + { + "epoch": 0.21, + "learning_rate": 1.8404717024472696e-05, + "loss": 1.25, + "step": 8600 + }, + { + "epoch": 0.21, + "learning_rate": 1.840429495081352e-05, + "loss": 1.1467, + "step": 8601 + }, + { + "epoch": 0.21, + "learning_rate": 1.8403872826167458e-05, + "loss": 0.9454, + "step": 8602 + }, + { + "epoch": 0.21, + "learning_rate": 1.840345065053708e-05, + "loss": 1.0963, + "step": 8603 + }, + { + "epoch": 0.21, + "learning_rate": 1.8403028423924943e-05, + "loss": 1.2354, + "step": 8604 + }, + { + "epoch": 0.21, + "learning_rate": 1.840260614633361e-05, + "loss": 1.0877, + "step": 8605 + }, + { + "epoch": 0.21, + "learning_rate": 1.8402183817765643e-05, + "loss": 1.3823, + "step": 8606 + }, + { + "epoch": 0.21, + "learning_rate": 1.84017614382236e-05, + "loss": 1.116, + "step": 8607 + }, + { + "epoch": 0.21, + "learning_rate": 1.840133900771005e-05, + "loss": 0.9364, + "step": 8608 + }, + { + "epoch": 0.21, + "learning_rate": 1.840091652622755e-05, + "loss": 0.8904, + "step": 8609 + }, + { + "epoch": 0.21, + "learning_rate": 1.8400493993778666e-05, + "loss": 1.0052, + "step": 8610 + }, + { + "epoch": 0.21, + "learning_rate": 1.840007141036596e-05, + "loss": 1.1104, + "step": 8611 + }, + { + "epoch": 0.21, + "learning_rate": 1.8399648775991996e-05, + "loss": 1.116, + "step": 8612 + }, + { + "epoch": 0.21, + "learning_rate": 1.839922609065934e-05, + "loss": 1.0493, + "step": 8613 + }, + { + "epoch": 0.21, + "learning_rate": 1.8398803354370554e-05, + "loss": 1.0497, + "step": 8614 + }, + { + "epoch": 0.21, + "learning_rate": 1.8398380567128208e-05, + "loss": 1.1735, + "step": 8615 + }, + { + "epoch": 0.21, + "learning_rate": 1.8397957728934855e-05, + "loss": 1.3646, + "step": 8616 + }, + { + "epoch": 0.21, + "learning_rate": 1.8397534839793073e-05, + "loss": 1.2393, + "step": 8617 + }, + { + "epoch": 0.21, + "learning_rate": 1.839711189970542e-05, + "loss": 1.4669, + "step": 8618 + }, + { + "epoch": 0.21, + "learning_rate": 1.8396688908674466e-05, + "loss": 1.2797, + "step": 8619 + }, + { + "epoch": 0.21, + "learning_rate": 1.8396265866702773e-05, + "loss": 1.004, + "step": 8620 + }, + { + "epoch": 0.21, + "learning_rate": 1.8395842773792914e-05, + "loss": 1.2599, + "step": 8621 + }, + { + "epoch": 0.21, + "learning_rate": 1.8395419629947448e-05, + "loss": 1.2605, + "step": 8622 + }, + { + "epoch": 0.21, + "learning_rate": 1.8394996435168945e-05, + "loss": 1.1848, + "step": 8623 + }, + { + "epoch": 0.21, + "learning_rate": 1.8394573189459977e-05, + "loss": 1.0967, + "step": 8624 + }, + { + "epoch": 0.21, + "learning_rate": 1.8394149892823106e-05, + "loss": 1.0339, + "step": 8625 + }, + { + "epoch": 0.21, + "learning_rate": 1.8393726545260903e-05, + "loss": 1.3336, + "step": 8626 + }, + { + "epoch": 0.21, + "learning_rate": 1.8393303146775933e-05, + "loss": 0.9006, + "step": 8627 + }, + { + "epoch": 0.21, + "learning_rate": 1.839287969737077e-05, + "loss": 1.0404, + "step": 8628 + }, + { + "epoch": 0.21, + "learning_rate": 1.839245619704798e-05, + "loss": 1.0316, + "step": 8629 + }, + { + "epoch": 0.21, + "learning_rate": 1.839203264581013e-05, + "loss": 1.191, + "step": 8630 + }, + { + "epoch": 0.21, + "learning_rate": 1.839160904365979e-05, + "loss": 0.8656, + "step": 8631 + }, + { + "epoch": 0.21, + "learning_rate": 1.8391185390599537e-05, + "loss": 1.1868, + "step": 8632 + }, + { + "epoch": 0.21, + "learning_rate": 1.8390761686631934e-05, + "loss": 1.3109, + "step": 8633 + }, + { + "epoch": 0.21, + "learning_rate": 1.8390337931759553e-05, + "loss": 1.0303, + "step": 8634 + }, + { + "epoch": 0.21, + "learning_rate": 1.8389914125984963e-05, + "loss": 1.1788, + "step": 8635 + }, + { + "epoch": 0.21, + "learning_rate": 1.8389490269310744e-05, + "loss": 1.0772, + "step": 8636 + }, + { + "epoch": 0.21, + "learning_rate": 1.8389066361739454e-05, + "loss": 0.9398, + "step": 8637 + }, + { + "epoch": 0.21, + "learning_rate": 1.8388642403273674e-05, + "loss": 1.2404, + "step": 8638 + }, + { + "epoch": 0.21, + "learning_rate": 1.8388218393915976e-05, + "loss": 1.0749, + "step": 8639 + }, + { + "epoch": 0.21, + "learning_rate": 1.8387794333668928e-05, + "loss": 1.105, + "step": 8640 + }, + { + "epoch": 0.21, + "learning_rate": 1.8387370222535102e-05, + "loss": 1.0465, + "step": 8641 + }, + { + "epoch": 0.21, + "learning_rate": 1.838694606051708e-05, + "loss": 1.2458, + "step": 8642 + }, + { + "epoch": 0.21, + "learning_rate": 1.8386521847617422e-05, + "loss": 1.0427, + "step": 8643 + }, + { + "epoch": 0.21, + "learning_rate": 1.8386097583838714e-05, + "loss": 1.0202, + "step": 8644 + }, + { + "epoch": 0.21, + "learning_rate": 1.8385673269183522e-05, + "loss": 1.0389, + "step": 8645 + }, + { + "epoch": 0.21, + "learning_rate": 1.8385248903654423e-05, + "loss": 1.0737, + "step": 8646 + }, + { + "epoch": 0.21, + "learning_rate": 1.838482448725399e-05, + "loss": 1.054, + "step": 8647 + }, + { + "epoch": 0.21, + "learning_rate": 1.83844000199848e-05, + "loss": 1.1298, + "step": 8648 + }, + { + "epoch": 0.21, + "learning_rate": 1.838397550184943e-05, + "loss": 1.2248, + "step": 8649 + }, + { + "epoch": 0.21, + "learning_rate": 1.838355093285045e-05, + "loss": 0.8066, + "step": 8650 + }, + { + "epoch": 0.21, + "learning_rate": 1.838312631299044e-05, + "loss": 1.0287, + "step": 8651 + }, + { + "epoch": 0.21, + "learning_rate": 1.838270164227197e-05, + "loss": 0.9838, + "step": 8652 + }, + { + "epoch": 0.21, + "learning_rate": 1.8382276920697623e-05, + "loss": 1.159, + "step": 8653 + }, + { + "epoch": 0.21, + "learning_rate": 1.8381852148269976e-05, + "loss": 1.2791, + "step": 8654 + }, + { + "epoch": 0.21, + "learning_rate": 1.8381427324991604e-05, + "loss": 0.8934, + "step": 8655 + }, + { + "epoch": 0.21, + "learning_rate": 1.8381002450865078e-05, + "loss": 1.2493, + "step": 8656 + }, + { + "epoch": 0.21, + "learning_rate": 1.8380577525892987e-05, + "loss": 1.2299, + "step": 8657 + }, + { + "epoch": 0.21, + "learning_rate": 1.8380152550077903e-05, + "loss": 1.191, + "step": 8658 + }, + { + "epoch": 0.21, + "learning_rate": 1.8379727523422404e-05, + "loss": 1.1405, + "step": 8659 + }, + { + "epoch": 0.21, + "learning_rate": 1.8379302445929068e-05, + "loss": 1.0741, + "step": 8660 + }, + { + "epoch": 0.21, + "learning_rate": 1.8378877317600477e-05, + "loss": 1.236, + "step": 8661 + }, + { + "epoch": 0.21, + "learning_rate": 1.8378452138439206e-05, + "loss": 1.0516, + "step": 8662 + }, + { + "epoch": 0.21, + "learning_rate": 1.8378026908447837e-05, + "loss": 1.3612, + "step": 8663 + }, + { + "epoch": 0.21, + "learning_rate": 1.8377601627628952e-05, + "loss": 1.0189, + "step": 8664 + }, + { + "epoch": 0.21, + "learning_rate": 1.8377176295985125e-05, + "loss": 1.0502, + "step": 8665 + }, + { + "epoch": 0.21, + "learning_rate": 1.837675091351894e-05, + "loss": 1.1491, + "step": 8666 + }, + { + "epoch": 0.21, + "learning_rate": 1.8376325480232977e-05, + "loss": 0.9922, + "step": 8667 + }, + { + "epoch": 0.21, + "learning_rate": 1.8375899996129823e-05, + "loss": 1.342, + "step": 8668 + }, + { + "epoch": 0.21, + "learning_rate": 1.8375474461212048e-05, + "loss": 1.0339, + "step": 8669 + }, + { + "epoch": 0.21, + "learning_rate": 1.837504887548224e-05, + "loss": 1.0144, + "step": 8670 + }, + { + "epoch": 0.21, + "learning_rate": 1.8374623238942985e-05, + "loss": 1.1474, + "step": 8671 + }, + { + "epoch": 0.21, + "learning_rate": 1.8374197551596857e-05, + "loss": 1.0251, + "step": 8672 + }, + { + "epoch": 0.21, + "learning_rate": 1.8373771813446442e-05, + "loss": 0.886, + "step": 8673 + }, + { + "epoch": 0.21, + "learning_rate": 1.8373346024494324e-05, + "loss": 1.0556, + "step": 8674 + }, + { + "epoch": 0.21, + "learning_rate": 1.8372920184743083e-05, + "loss": 1.1519, + "step": 8675 + }, + { + "epoch": 0.21, + "learning_rate": 1.8372494294195306e-05, + "loss": 1.0856, + "step": 8676 + }, + { + "epoch": 0.21, + "learning_rate": 1.8372068352853576e-05, + "loss": 1.2708, + "step": 8677 + }, + { + "epoch": 0.21, + "learning_rate": 1.837164236072048e-05, + "loss": 0.7649, + "step": 8678 + }, + { + "epoch": 0.21, + "learning_rate": 1.8371216317798595e-05, + "loss": 0.9482, + "step": 8679 + }, + { + "epoch": 0.21, + "learning_rate": 1.8370790224090508e-05, + "loss": 1.3113, + "step": 8680 + }, + { + "epoch": 0.21, + "learning_rate": 1.837036407959881e-05, + "loss": 1.0863, + "step": 8681 + }, + { + "epoch": 0.21, + "learning_rate": 1.836993788432608e-05, + "loss": 1.2249, + "step": 8682 + }, + { + "epoch": 0.21, + "learning_rate": 1.8369511638274906e-05, + "loss": 1.1773, + "step": 8683 + }, + { + "epoch": 0.21, + "learning_rate": 1.836908534144787e-05, + "loss": 1.0544, + "step": 8684 + }, + { + "epoch": 0.21, + "learning_rate": 1.8368658993847566e-05, + "loss": 1.1778, + "step": 8685 + }, + { + "epoch": 0.21, + "learning_rate": 1.8368232595476575e-05, + "loss": 1.1514, + "step": 8686 + }, + { + "epoch": 0.21, + "learning_rate": 1.8367806146337485e-05, + "loss": 1.4712, + "step": 8687 + }, + { + "epoch": 0.21, + "learning_rate": 1.8367379646432884e-05, + "loss": 1.129, + "step": 8688 + }, + { + "epoch": 0.21, + "learning_rate": 1.8366953095765356e-05, + "loss": 1.1425, + "step": 8689 + }, + { + "epoch": 0.21, + "learning_rate": 1.8366526494337497e-05, + "loss": 1.1578, + "step": 8690 + }, + { + "epoch": 0.21, + "learning_rate": 1.8366099842151883e-05, + "loss": 0.985, + "step": 8691 + }, + { + "epoch": 0.21, + "learning_rate": 1.8365673139211114e-05, + "loss": 1.0593, + "step": 8692 + }, + { + "epoch": 0.21, + "learning_rate": 1.8365246385517773e-05, + "loss": 1.2573, + "step": 8693 + }, + { + "epoch": 0.21, + "learning_rate": 1.8364819581074447e-05, + "loss": 1.1651, + "step": 8694 + }, + { + "epoch": 0.21, + "learning_rate": 1.8364392725883734e-05, + "loss": 1.6057, + "step": 8695 + }, + { + "epoch": 0.21, + "learning_rate": 1.836396581994821e-05, + "loss": 0.9873, + "step": 8696 + }, + { + "epoch": 0.21, + "learning_rate": 1.836353886327048e-05, + "loss": 1.2274, + "step": 8697 + }, + { + "epoch": 0.21, + "learning_rate": 1.8363111855853122e-05, + "loss": 1.061, + "step": 8698 + }, + { + "epoch": 0.21, + "learning_rate": 1.8362684797698732e-05, + "loss": 1.1183, + "step": 8699 + }, + { + "epoch": 0.21, + "learning_rate": 1.8362257688809904e-05, + "loss": 0.8621, + "step": 8700 + }, + { + "epoch": 0.21, + "learning_rate": 1.836183052918922e-05, + "loss": 1.102, + "step": 8701 + }, + { + "epoch": 0.21, + "learning_rate": 1.8361403318839283e-05, + "loss": 1.1228, + "step": 8702 + }, + { + "epoch": 0.21, + "learning_rate": 1.8360976057762677e-05, + "loss": 1.1635, + "step": 8703 + }, + { + "epoch": 0.21, + "learning_rate": 1.8360548745961994e-05, + "loss": 1.2666, + "step": 8704 + }, + { + "epoch": 0.21, + "learning_rate": 1.836012138343983e-05, + "loss": 0.9791, + "step": 8705 + }, + { + "epoch": 0.21, + "learning_rate": 1.8359693970198772e-05, + "loss": 1.0028, + "step": 8706 + }, + { + "epoch": 0.21, + "learning_rate": 1.835926650624142e-05, + "loss": 1.3706, + "step": 8707 + }, + { + "epoch": 0.21, + "learning_rate": 1.8358838991570363e-05, + "loss": 0.9912, + "step": 8708 + }, + { + "epoch": 0.21, + "learning_rate": 1.83584114261882e-05, + "loss": 1.136, + "step": 8709 + }, + { + "epoch": 0.21, + "learning_rate": 1.835798381009752e-05, + "loss": 0.9221, + "step": 8710 + }, + { + "epoch": 0.21, + "learning_rate": 1.8357556143300914e-05, + "loss": 0.9826, + "step": 8711 + }, + { + "epoch": 0.21, + "learning_rate": 1.8357128425800983e-05, + "loss": 1.0221, + "step": 8712 + }, + { + "epoch": 0.21, + "learning_rate": 1.835670065760032e-05, + "loss": 1.0136, + "step": 8713 + }, + { + "epoch": 0.21, + "learning_rate": 1.835627283870152e-05, + "loss": 1.1917, + "step": 8714 + }, + { + "epoch": 0.21, + "learning_rate": 1.8355844969107178e-05, + "loss": 0.8575, + "step": 8715 + }, + { + "epoch": 0.21, + "learning_rate": 1.835541704881989e-05, + "loss": 1.0263, + "step": 8716 + }, + { + "epoch": 0.21, + "learning_rate": 1.8354989077842253e-05, + "loss": 1.2871, + "step": 8717 + }, + { + "epoch": 0.21, + "learning_rate": 1.8354561056176857e-05, + "loss": 1.254, + "step": 8718 + }, + { + "epoch": 0.21, + "learning_rate": 1.835413298382631e-05, + "loss": 1.1738, + "step": 8719 + }, + { + "epoch": 0.21, + "learning_rate": 1.8353704860793202e-05, + "loss": 1.1478, + "step": 8720 + }, + { + "epoch": 0.21, + "learning_rate": 1.835327668708013e-05, + "loss": 1.1774, + "step": 8721 + }, + { + "epoch": 0.21, + "learning_rate": 1.8352848462689694e-05, + "loss": 1.1305, + "step": 8722 + }, + { + "epoch": 0.21, + "learning_rate": 1.835242018762449e-05, + "loss": 1.12, + "step": 8723 + }, + { + "epoch": 0.21, + "learning_rate": 1.835199186188712e-05, + "loss": 1.0322, + "step": 8724 + }, + { + "epoch": 0.21, + "learning_rate": 1.8351563485480177e-05, + "loss": 1.1999, + "step": 8725 + }, + { + "epoch": 0.21, + "learning_rate": 1.8351135058406266e-05, + "loss": 1.0281, + "step": 8726 + }, + { + "epoch": 0.21, + "learning_rate": 1.835070658066798e-05, + "loss": 1.1406, + "step": 8727 + }, + { + "epoch": 0.21, + "learning_rate": 1.8350278052267922e-05, + "loss": 1.2112, + "step": 8728 + }, + { + "epoch": 0.21, + "learning_rate": 1.834984947320869e-05, + "loss": 1.1267, + "step": 8729 + }, + { + "epoch": 0.21, + "learning_rate": 1.834942084349289e-05, + "loss": 1.1737, + "step": 8730 + }, + { + "epoch": 0.21, + "learning_rate": 1.8348992163123114e-05, + "loss": 1.2722, + "step": 8731 + }, + { + "epoch": 0.21, + "learning_rate": 1.8348563432101967e-05, + "loss": 1.0007, + "step": 8732 + }, + { + "epoch": 0.21, + "learning_rate": 1.834813465043205e-05, + "loss": 0.9783, + "step": 8733 + }, + { + "epoch": 0.21, + "learning_rate": 1.834770581811596e-05, + "loss": 1.1065, + "step": 8734 + }, + { + "epoch": 0.21, + "learning_rate": 1.8347276935156308e-05, + "loss": 0.7675, + "step": 8735 + }, + { + "epoch": 0.21, + "learning_rate": 1.8346848001555688e-05, + "loss": 1.3472, + "step": 8736 + }, + { + "epoch": 0.21, + "learning_rate": 1.8346419017316704e-05, + "loss": 0.9841, + "step": 8737 + }, + { + "epoch": 0.21, + "learning_rate": 1.8345989982441955e-05, + "loss": 1.0649, + "step": 8738 + }, + { + "epoch": 0.21, + "learning_rate": 1.8345560896934053e-05, + "loss": 1.1445, + "step": 8739 + }, + { + "epoch": 0.21, + "learning_rate": 1.8345131760795598e-05, + "loss": 1.1356, + "step": 8740 + }, + { + "epoch": 0.21, + "learning_rate": 1.8344702574029186e-05, + "loss": 1.2251, + "step": 8741 + }, + { + "epoch": 0.21, + "learning_rate": 1.834427333663743e-05, + "loss": 1.305, + "step": 8742 + }, + { + "epoch": 0.21, + "learning_rate": 1.8343844048622926e-05, + "loss": 0.9083, + "step": 8743 + }, + { + "epoch": 0.21, + "learning_rate": 1.8343414709988288e-05, + "loss": 1.1667, + "step": 8744 + }, + { + "epoch": 0.21, + "learning_rate": 1.8342985320736113e-05, + "loss": 1.347, + "step": 8745 + }, + { + "epoch": 0.21, + "learning_rate": 1.834255588086901e-05, + "loss": 1.2901, + "step": 8746 + }, + { + "epoch": 0.21, + "learning_rate": 1.834212639038958e-05, + "loss": 1.296, + "step": 8747 + }, + { + "epoch": 0.21, + "learning_rate": 1.834169684930043e-05, + "loss": 0.8466, + "step": 8748 + }, + { + "epoch": 0.21, + "learning_rate": 1.8341267257604173e-05, + "loss": 0.9782, + "step": 8749 + }, + { + "epoch": 0.21, + "learning_rate": 1.8340837615303405e-05, + "loss": 1.174, + "step": 8750 + }, + { + "epoch": 0.21, + "learning_rate": 1.834040792240074e-05, + "loss": 1.1015, + "step": 8751 + }, + { + "epoch": 0.21, + "learning_rate": 1.833997817889878e-05, + "loss": 1.3173, + "step": 8752 + }, + { + "epoch": 0.21, + "learning_rate": 1.8339548384800133e-05, + "loss": 1.0831, + "step": 8753 + }, + { + "epoch": 0.21, + "learning_rate": 1.833911854010741e-05, + "loss": 1.3814, + "step": 8754 + }, + { + "epoch": 0.21, + "learning_rate": 1.8338688644823216e-05, + "loss": 1.2365, + "step": 8755 + }, + { + "epoch": 0.21, + "learning_rate": 1.833825869895016e-05, + "loss": 1.2643, + "step": 8756 + }, + { + "epoch": 0.21, + "learning_rate": 1.8337828702490846e-05, + "loss": 1.0847, + "step": 8757 + }, + { + "epoch": 0.21, + "learning_rate": 1.8337398655447894e-05, + "loss": 0.9579, + "step": 8758 + }, + { + "epoch": 0.21, + "learning_rate": 1.8336968557823898e-05, + "loss": 1.2498, + "step": 8759 + }, + { + "epoch": 0.21, + "learning_rate": 1.8336538409621474e-05, + "loss": 1.0441, + "step": 8760 + }, + { + "epoch": 0.21, + "learning_rate": 1.833610821084324e-05, + "loss": 1.1396, + "step": 8761 + }, + { + "epoch": 0.21, + "learning_rate": 1.8335677961491793e-05, + "loss": 1.0945, + "step": 8762 + }, + { + "epoch": 0.21, + "learning_rate": 1.833524766156975e-05, + "loss": 1.1765, + "step": 8763 + }, + { + "epoch": 0.21, + "learning_rate": 1.833481731107972e-05, + "loss": 0.9363, + "step": 8764 + }, + { + "epoch": 0.21, + "learning_rate": 1.833438691002431e-05, + "loss": 1.4027, + "step": 8765 + }, + { + "epoch": 0.21, + "learning_rate": 1.833395645840614e-05, + "loss": 0.9006, + "step": 8766 + }, + { + "epoch": 0.21, + "learning_rate": 1.8333525956227815e-05, + "loss": 1.0883, + "step": 8767 + }, + { + "epoch": 0.21, + "learning_rate": 1.8333095403491946e-05, + "loss": 1.0632, + "step": 8768 + }, + { + "epoch": 0.21, + "learning_rate": 1.833266480020115e-05, + "loss": 0.9756, + "step": 8769 + }, + { + "epoch": 0.21, + "learning_rate": 1.8332234146358034e-05, + "loss": 1.172, + "step": 8770 + }, + { + "epoch": 0.21, + "learning_rate": 1.8331803441965218e-05, + "loss": 1.2195, + "step": 8771 + }, + { + "epoch": 0.21, + "learning_rate": 1.8331372687025305e-05, + "loss": 1.0392, + "step": 8772 + }, + { + "epoch": 0.21, + "learning_rate": 1.8330941881540917e-05, + "loss": 1.0793, + "step": 8773 + }, + { + "epoch": 0.21, + "learning_rate": 1.8330511025514662e-05, + "loss": 1.1498, + "step": 8774 + }, + { + "epoch": 0.21, + "learning_rate": 1.833008011894916e-05, + "loss": 1.2774, + "step": 8775 + }, + { + "epoch": 0.21, + "learning_rate": 1.8329649161847016e-05, + "loss": 1.1542, + "step": 8776 + }, + { + "epoch": 0.21, + "learning_rate": 1.8329218154210852e-05, + "loss": 1.1039, + "step": 8777 + }, + { + "epoch": 0.21, + "learning_rate": 1.8328787096043278e-05, + "loss": 1.286, + "step": 8778 + }, + { + "epoch": 0.21, + "learning_rate": 1.832835598734692e-05, + "loss": 0.9618, + "step": 8779 + }, + { + "epoch": 0.21, + "learning_rate": 1.8327924828124377e-05, + "loss": 1.0655, + "step": 8780 + }, + { + "epoch": 0.21, + "learning_rate": 1.8327493618378275e-05, + "loss": 1.112, + "step": 8781 + }, + { + "epoch": 0.21, + "learning_rate": 1.8327062358111228e-05, + "loss": 1.0432, + "step": 8782 + }, + { + "epoch": 0.21, + "learning_rate": 1.8326631047325855e-05, + "loss": 1.177, + "step": 8783 + }, + { + "epoch": 0.21, + "learning_rate": 1.8326199686024765e-05, + "loss": 1.1137, + "step": 8784 + }, + { + "epoch": 0.21, + "learning_rate": 1.8325768274210585e-05, + "loss": 1.1982, + "step": 8785 + }, + { + "epoch": 0.21, + "learning_rate": 1.8325336811885926e-05, + "loss": 1.1765, + "step": 8786 + }, + { + "epoch": 0.21, + "learning_rate": 1.8324905299053403e-05, + "loss": 1.066, + "step": 8787 + }, + { + "epoch": 0.21, + "learning_rate": 1.8324473735715643e-05, + "loss": 0.9608, + "step": 8788 + }, + { + "epoch": 0.21, + "learning_rate": 1.8324042121875255e-05, + "loss": 1.0466, + "step": 8789 + }, + { + "epoch": 0.21, + "learning_rate": 1.832361045753486e-05, + "loss": 1.1124, + "step": 8790 + }, + { + "epoch": 0.21, + "learning_rate": 1.8323178742697083e-05, + "loss": 1.0724, + "step": 8791 + }, + { + "epoch": 0.21, + "learning_rate": 1.8322746977364537e-05, + "loss": 0.9515, + "step": 8792 + }, + { + "epoch": 0.21, + "learning_rate": 1.8322315161539842e-05, + "loss": 1.0841, + "step": 8793 + }, + { + "epoch": 0.21, + "learning_rate": 1.8321883295225617e-05, + "loss": 0.9265, + "step": 8794 + }, + { + "epoch": 0.21, + "learning_rate": 1.8321451378424485e-05, + "loss": 0.8792, + "step": 8795 + }, + { + "epoch": 0.21, + "learning_rate": 1.8321019411139064e-05, + "loss": 1.4631, + "step": 8796 + }, + { + "epoch": 0.21, + "learning_rate": 1.8320587393371977e-05, + "loss": 1.1515, + "step": 8797 + }, + { + "epoch": 0.21, + "learning_rate": 1.8320155325125843e-05, + "loss": 1.507, + "step": 8798 + }, + { + "epoch": 0.21, + "learning_rate": 1.831972320640328e-05, + "loss": 1.2908, + "step": 8799 + }, + { + "epoch": 0.21, + "learning_rate": 1.831929103720692e-05, + "loss": 1.1878, + "step": 8800 + }, + { + "epoch": 0.21, + "learning_rate": 1.8318858817539372e-05, + "loss": 1.0231, + "step": 8801 + }, + { + "epoch": 0.21, + "learning_rate": 1.831842654740327e-05, + "loss": 1.182, + "step": 8802 + }, + { + "epoch": 0.21, + "learning_rate": 1.8317994226801227e-05, + "loss": 1.1799, + "step": 8803 + }, + { + "epoch": 0.21, + "learning_rate": 1.8317561855735867e-05, + "loss": 1.1517, + "step": 8804 + }, + { + "epoch": 0.21, + "learning_rate": 1.8317129434209818e-05, + "loss": 1.1619, + "step": 8805 + }, + { + "epoch": 0.21, + "learning_rate": 1.8316696962225704e-05, + "loss": 1.029, + "step": 8806 + }, + { + "epoch": 0.21, + "learning_rate": 1.8316264439786145e-05, + "loss": 1.2479, + "step": 8807 + }, + { + "epoch": 0.21, + "learning_rate": 1.8315831866893762e-05, + "loss": 1.0503, + "step": 8808 + }, + { + "epoch": 0.21, + "learning_rate": 1.8315399243551186e-05, + "loss": 1.2267, + "step": 8809 + }, + { + "epoch": 0.21, + "learning_rate": 1.831496656976104e-05, + "loss": 1.1589, + "step": 8810 + }, + { + "epoch": 0.21, + "learning_rate": 1.8314533845525943e-05, + "loss": 1.1367, + "step": 8811 + }, + { + "epoch": 0.21, + "learning_rate": 1.8314101070848527e-05, + "loss": 1.1195, + "step": 8812 + }, + { + "epoch": 0.21, + "learning_rate": 1.8313668245731417e-05, + "loss": 0.769, + "step": 8813 + }, + { + "epoch": 0.21, + "learning_rate": 1.8313235370177235e-05, + "loss": 1.135, + "step": 8814 + }, + { + "epoch": 0.21, + "learning_rate": 1.831280244418861e-05, + "loss": 1.3628, + "step": 8815 + }, + { + "epoch": 0.21, + "learning_rate": 1.8312369467768168e-05, + "loss": 0.9592, + "step": 8816 + }, + { + "epoch": 0.21, + "learning_rate": 1.8311936440918537e-05, + "loss": 1.3479, + "step": 8817 + }, + { + "epoch": 0.21, + "learning_rate": 1.831150336364234e-05, + "loss": 1.1617, + "step": 8818 + }, + { + "epoch": 0.21, + "learning_rate": 1.831107023594221e-05, + "loss": 1.1484, + "step": 8819 + }, + { + "epoch": 0.21, + "learning_rate": 1.831063705782077e-05, + "loss": 1.5007, + "step": 8820 + }, + { + "epoch": 0.21, + "learning_rate": 1.831020382928065e-05, + "loss": 1.2259, + "step": 8821 + }, + { + "epoch": 0.21, + "learning_rate": 1.830977055032448e-05, + "loss": 1.4316, + "step": 8822 + }, + { + "epoch": 0.21, + "learning_rate": 1.8309337220954882e-05, + "loss": 1.0029, + "step": 8823 + }, + { + "epoch": 0.21, + "learning_rate": 1.8308903841174493e-05, + "loss": 0.9783, + "step": 8824 + }, + { + "epoch": 0.21, + "learning_rate": 1.830847041098594e-05, + "loss": 1.1627, + "step": 8825 + }, + { + "epoch": 0.21, + "learning_rate": 1.8308036930391848e-05, + "loss": 1.1938, + "step": 8826 + }, + { + "epoch": 0.21, + "learning_rate": 1.8307603399394855e-05, + "loss": 1.1075, + "step": 8827 + }, + { + "epoch": 0.21, + "learning_rate": 1.830716981799758e-05, + "loss": 1.0793, + "step": 8828 + }, + { + "epoch": 0.21, + "learning_rate": 1.8306736186202664e-05, + "loss": 0.9168, + "step": 8829 + }, + { + "epoch": 0.21, + "learning_rate": 1.8306302504012732e-05, + "loss": 0.7712, + "step": 8830 + }, + { + "epoch": 0.21, + "learning_rate": 1.8305868771430418e-05, + "loss": 0.9478, + "step": 8831 + }, + { + "epoch": 0.21, + "learning_rate": 1.8305434988458348e-05, + "loss": 1.1615, + "step": 8832 + }, + { + "epoch": 0.21, + "learning_rate": 1.830500115509916e-05, + "loss": 1.1982, + "step": 8833 + }, + { + "epoch": 0.21, + "learning_rate": 1.8304567271355482e-05, + "loss": 1.1034, + "step": 8834 + }, + { + "epoch": 0.21, + "learning_rate": 1.830413333722995e-05, + "loss": 0.736, + "step": 8835 + }, + { + "epoch": 0.21, + "learning_rate": 1.8303699352725193e-05, + "loss": 1.3635, + "step": 8836 + }, + { + "epoch": 0.21, + "learning_rate": 1.8303265317843847e-05, + "loss": 1.045, + "step": 8837 + }, + { + "epoch": 0.21, + "learning_rate": 1.830283123258854e-05, + "loss": 1.492, + "step": 8838 + }, + { + "epoch": 0.21, + "learning_rate": 1.830239709696191e-05, + "loss": 1.399, + "step": 8839 + }, + { + "epoch": 0.21, + "learning_rate": 1.8301962910966592e-05, + "loss": 1.3118, + "step": 8840 + }, + { + "epoch": 0.21, + "learning_rate": 1.8301528674605216e-05, + "loss": 1.1799, + "step": 8841 + }, + { + "epoch": 0.21, + "learning_rate": 1.830109438788042e-05, + "loss": 1.2591, + "step": 8842 + }, + { + "epoch": 0.21, + "learning_rate": 1.830066005079483e-05, + "loss": 1.1137, + "step": 8843 + }, + { + "epoch": 0.21, + "learning_rate": 1.8300225663351098e-05, + "loss": 1.1479, + "step": 8844 + }, + { + "epoch": 0.21, + "learning_rate": 1.8299791225551842e-05, + "loss": 0.8736, + "step": 8845 + }, + { + "epoch": 0.21, + "learning_rate": 1.8299356737399707e-05, + "loss": 1.3649, + "step": 8846 + }, + { + "epoch": 0.21, + "learning_rate": 1.8298922198897328e-05, + "loss": 0.8491, + "step": 8847 + }, + { + "epoch": 0.21, + "learning_rate": 1.8298487610047337e-05, + "loss": 1.2188, + "step": 8848 + }, + { + "epoch": 0.21, + "learning_rate": 1.8298052970852378e-05, + "loss": 0.9963, + "step": 8849 + }, + { + "epoch": 0.21, + "learning_rate": 1.829761828131508e-05, + "loss": 1.1129, + "step": 8850 + }, + { + "epoch": 0.21, + "learning_rate": 1.8297183541438085e-05, + "loss": 0.984, + "step": 8851 + }, + { + "epoch": 0.21, + "learning_rate": 1.829674875122403e-05, + "loss": 0.6932, + "step": 8852 + }, + { + "epoch": 0.21, + "learning_rate": 1.829631391067555e-05, + "loss": 1.0989, + "step": 8853 + }, + { + "epoch": 0.21, + "learning_rate": 1.8295879019795283e-05, + "loss": 1.2617, + "step": 8854 + }, + { + "epoch": 0.21, + "learning_rate": 1.8295444078585872e-05, + "loss": 1.1386, + "step": 8855 + }, + { + "epoch": 0.21, + "learning_rate": 1.8295009087049954e-05, + "loss": 1.2227, + "step": 8856 + }, + { + "epoch": 0.21, + "learning_rate": 1.8294574045190165e-05, + "loss": 1.1689, + "step": 8857 + }, + { + "epoch": 0.21, + "learning_rate": 1.8294138953009145e-05, + "loss": 0.9661, + "step": 8858 + }, + { + "epoch": 0.21, + "learning_rate": 1.829370381050954e-05, + "loss": 1.2232, + "step": 8859 + }, + { + "epoch": 0.21, + "learning_rate": 1.829326861769398e-05, + "loss": 0.8805, + "step": 8860 + }, + { + "epoch": 0.21, + "learning_rate": 1.829283337456511e-05, + "loss": 1.2008, + "step": 8861 + }, + { + "epoch": 0.21, + "learning_rate": 1.8292398081125572e-05, + "loss": 1.1396, + "step": 8862 + }, + { + "epoch": 0.21, + "learning_rate": 1.8291962737378006e-05, + "loss": 1.265, + "step": 8863 + }, + { + "epoch": 0.21, + "learning_rate": 1.8291527343325052e-05, + "loss": 1.2367, + "step": 8864 + }, + { + "epoch": 0.21, + "learning_rate": 1.8291091898969352e-05, + "loss": 1.122, + "step": 8865 + }, + { + "epoch": 0.21, + "learning_rate": 1.8290656404313546e-05, + "loss": 1.0608, + "step": 8866 + }, + { + "epoch": 0.21, + "learning_rate": 1.8290220859360283e-05, + "loss": 1.2864, + "step": 8867 + }, + { + "epoch": 0.21, + "learning_rate": 1.8289785264112193e-05, + "loss": 1.0496, + "step": 8868 + }, + { + "epoch": 0.21, + "learning_rate": 1.828934961857193e-05, + "loss": 1.1087, + "step": 8869 + }, + { + "epoch": 0.21, + "learning_rate": 1.8288913922742134e-05, + "loss": 1.1062, + "step": 8870 + }, + { + "epoch": 0.21, + "learning_rate": 1.828847817662544e-05, + "loss": 1.0472, + "step": 8871 + }, + { + "epoch": 0.21, + "learning_rate": 1.8288042380224508e-05, + "loss": 1.0702, + "step": 8872 + }, + { + "epoch": 0.21, + "learning_rate": 1.8287606533541965e-05, + "loss": 0.8672, + "step": 8873 + }, + { + "epoch": 0.21, + "learning_rate": 1.8287170636580464e-05, + "loss": 1.2399, + "step": 8874 + }, + { + "epoch": 0.21, + "learning_rate": 1.8286734689342648e-05, + "loss": 1.084, + "step": 8875 + }, + { + "epoch": 0.21, + "learning_rate": 1.8286298691831164e-05, + "loss": 1.0131, + "step": 8876 + }, + { + "epoch": 0.21, + "learning_rate": 1.8285862644048654e-05, + "loss": 1.3528, + "step": 8877 + }, + { + "epoch": 0.21, + "learning_rate": 1.8285426545997764e-05, + "loss": 1.1265, + "step": 8878 + }, + { + "epoch": 0.21, + "learning_rate": 1.8284990397681137e-05, + "loss": 1.1502, + "step": 8879 + }, + { + "epoch": 0.21, + "learning_rate": 1.828455419910143e-05, + "loss": 0.9849, + "step": 8880 + }, + { + "epoch": 0.21, + "learning_rate": 1.8284117950261272e-05, + "loss": 1.1712, + "step": 8881 + }, + { + "epoch": 0.21, + "learning_rate": 1.8283681651163324e-05, + "loss": 0.9277, + "step": 8882 + }, + { + "epoch": 0.21, + "learning_rate": 1.8283245301810227e-05, + "loss": 1.0819, + "step": 8883 + }, + { + "epoch": 0.21, + "learning_rate": 1.8282808902204627e-05, + "loss": 1.245, + "step": 8884 + }, + { + "epoch": 0.21, + "learning_rate": 1.8282372452349177e-05, + "loss": 1.3393, + "step": 8885 + }, + { + "epoch": 0.21, + "learning_rate": 1.828193595224652e-05, + "loss": 0.9679, + "step": 8886 + }, + { + "epoch": 0.21, + "learning_rate": 1.8281499401899303e-05, + "loss": 1.0861, + "step": 8887 + }, + { + "epoch": 0.21, + "learning_rate": 1.828106280131018e-05, + "loss": 1.4061, + "step": 8888 + }, + { + "epoch": 0.21, + "learning_rate": 1.8280626150481797e-05, + "loss": 1.0973, + "step": 8889 + }, + { + "epoch": 0.21, + "learning_rate": 1.8280189449416805e-05, + "loss": 1.0228, + "step": 8890 + }, + { + "epoch": 0.21, + "learning_rate": 1.8279752698117846e-05, + "loss": 1.0196, + "step": 8891 + }, + { + "epoch": 0.21, + "learning_rate": 1.827931589658758e-05, + "loss": 1.0505, + "step": 8892 + }, + { + "epoch": 0.21, + "learning_rate": 1.8278879044828647e-05, + "loss": 1.0051, + "step": 8893 + }, + { + "epoch": 0.21, + "learning_rate": 1.8278442142843703e-05, + "loss": 1.0489, + "step": 8894 + }, + { + "epoch": 0.21, + "learning_rate": 1.82780051906354e-05, + "loss": 0.9179, + "step": 8895 + }, + { + "epoch": 0.21, + "learning_rate": 1.8277568188206386e-05, + "loss": 1.0048, + "step": 8896 + }, + { + "epoch": 0.21, + "learning_rate": 1.8277131135559312e-05, + "loss": 1.2603, + "step": 8897 + }, + { + "epoch": 0.21, + "learning_rate": 1.8276694032696835e-05, + "loss": 1.0877, + "step": 8898 + }, + { + "epoch": 0.21, + "learning_rate": 1.8276256879621594e-05, + "loss": 1.2554, + "step": 8899 + }, + { + "epoch": 0.21, + "learning_rate": 1.8275819676336256e-05, + "loss": 1.2485, + "step": 8900 + }, + { + "epoch": 0.21, + "learning_rate": 1.827538242284346e-05, + "loss": 1.073, + "step": 8901 + }, + { + "epoch": 0.21, + "learning_rate": 1.827494511914587e-05, + "loss": 1.2295, + "step": 8902 + }, + { + "epoch": 0.21, + "learning_rate": 1.8274507765246134e-05, + "loss": 1.036, + "step": 8903 + }, + { + "epoch": 0.21, + "learning_rate": 1.8274070361146906e-05, + "loss": 1.2987, + "step": 8904 + }, + { + "epoch": 0.21, + "learning_rate": 1.827363290685084e-05, + "loss": 1.2945, + "step": 8905 + }, + { + "epoch": 0.21, + "learning_rate": 1.8273195402360585e-05, + "loss": 1.034, + "step": 8906 + }, + { + "epoch": 0.21, + "learning_rate": 1.8272757847678804e-05, + "loss": 1.3148, + "step": 8907 + }, + { + "epoch": 0.21, + "learning_rate": 1.8272320242808143e-05, + "loss": 0.8879, + "step": 8908 + }, + { + "epoch": 0.21, + "learning_rate": 1.8271882587751264e-05, + "loss": 0.9999, + "step": 8909 + }, + { + "epoch": 0.21, + "learning_rate": 1.827144488251082e-05, + "loss": 1.2534, + "step": 8910 + }, + { + "epoch": 0.21, + "learning_rate": 1.827100712708946e-05, + "loss": 1.1666, + "step": 8911 + }, + { + "epoch": 0.21, + "learning_rate": 1.827056932148985e-05, + "loss": 1.266, + "step": 8912 + }, + { + "epoch": 0.21, + "learning_rate": 1.827013146571464e-05, + "loss": 1.2906, + "step": 8913 + }, + { + "epoch": 0.21, + "learning_rate": 1.8269693559766487e-05, + "loss": 1.2319, + "step": 8914 + }, + { + "epoch": 0.21, + "learning_rate": 1.826925560364805e-05, + "loss": 1.0436, + "step": 8915 + }, + { + "epoch": 0.21, + "learning_rate": 1.8268817597361983e-05, + "loss": 1.0892, + "step": 8916 + }, + { + "epoch": 0.21, + "learning_rate": 1.8268379540910946e-05, + "loss": 1.1818, + "step": 8917 + }, + { + "epoch": 0.21, + "learning_rate": 1.8267941434297594e-05, + "loss": 1.1621, + "step": 8918 + }, + { + "epoch": 0.21, + "learning_rate": 1.8267503277524588e-05, + "loss": 1.0911, + "step": 8919 + }, + { + "epoch": 0.21, + "learning_rate": 1.826706507059458e-05, + "loss": 1.1797, + "step": 8920 + }, + { + "epoch": 0.21, + "learning_rate": 1.8266626813510236e-05, + "loss": 1.0939, + "step": 8921 + }, + { + "epoch": 0.21, + "learning_rate": 1.826618850627421e-05, + "loss": 1.1124, + "step": 8922 + }, + { + "epoch": 0.21, + "learning_rate": 1.8265750148889165e-05, + "loss": 0.9934, + "step": 8923 + }, + { + "epoch": 0.21, + "learning_rate": 1.8265311741357753e-05, + "loss": 1.2366, + "step": 8924 + }, + { + "epoch": 0.21, + "learning_rate": 1.8264873283682645e-05, + "loss": 1.0063, + "step": 8925 + }, + { + "epoch": 0.21, + "learning_rate": 1.826443477586649e-05, + "loss": 1.1595, + "step": 8926 + }, + { + "epoch": 0.21, + "learning_rate": 1.8263996217911957e-05, + "loss": 0.9422, + "step": 8927 + }, + { + "epoch": 0.21, + "learning_rate": 1.82635576098217e-05, + "loss": 0.8693, + "step": 8928 + }, + { + "epoch": 0.21, + "learning_rate": 1.8263118951598383e-05, + "loss": 1.1578, + "step": 8929 + }, + { + "epoch": 0.21, + "learning_rate": 1.826268024324467e-05, + "loss": 1.0485, + "step": 8930 + }, + { + "epoch": 0.21, + "learning_rate": 1.8262241484763215e-05, + "loss": 1.3461, + "step": 8931 + }, + { + "epoch": 0.21, + "learning_rate": 1.8261802676156685e-05, + "loss": 1.3196, + "step": 8932 + }, + { + "epoch": 0.21, + "learning_rate": 1.8261363817427743e-05, + "loss": 1.2308, + "step": 8933 + }, + { + "epoch": 0.21, + "learning_rate": 1.826092490857905e-05, + "loss": 1.2416, + "step": 8934 + }, + { + "epoch": 0.21, + "learning_rate": 1.826048594961327e-05, + "loss": 1.2738, + "step": 8935 + }, + { + "epoch": 0.21, + "learning_rate": 1.8260046940533063e-05, + "loss": 0.9487, + "step": 8936 + }, + { + "epoch": 0.21, + "learning_rate": 1.825960788134109e-05, + "loss": 1.0568, + "step": 8937 + }, + { + "epoch": 0.21, + "learning_rate": 1.8259168772040027e-05, + "loss": 1.3244, + "step": 8938 + }, + { + "epoch": 0.21, + "learning_rate": 1.8258729612632525e-05, + "loss": 1.2881, + "step": 8939 + }, + { + "epoch": 0.21, + "learning_rate": 1.8258290403121252e-05, + "loss": 1.2656, + "step": 8940 + }, + { + "epoch": 0.22, + "learning_rate": 1.8257851143508875e-05, + "loss": 1.0682, + "step": 8941 + }, + { + "epoch": 0.22, + "learning_rate": 1.825741183379806e-05, + "loss": 1.1087, + "step": 8942 + }, + { + "epoch": 0.22, + "learning_rate": 1.8256972473991467e-05, + "loss": 1.2558, + "step": 8943 + }, + { + "epoch": 0.22, + "learning_rate": 1.8256533064091765e-05, + "loss": 1.0703, + "step": 8944 + }, + { + "epoch": 0.22, + "learning_rate": 1.8256093604101618e-05, + "loss": 1.2112, + "step": 8945 + }, + { + "epoch": 0.22, + "learning_rate": 1.8255654094023692e-05, + "loss": 1.3773, + "step": 8946 + }, + { + "epoch": 0.22, + "learning_rate": 1.8255214533860658e-05, + "loss": 1.2678, + "step": 8947 + }, + { + "epoch": 0.22, + "learning_rate": 1.8254774923615177e-05, + "loss": 1.021, + "step": 8948 + }, + { + "epoch": 0.22, + "learning_rate": 1.8254335263289917e-05, + "loss": 1.0108, + "step": 8949 + }, + { + "epoch": 0.22, + "learning_rate": 1.8253895552887547e-05, + "loss": 1.129, + "step": 8950 + }, + { + "epoch": 0.22, + "learning_rate": 1.8253455792410734e-05, + "loss": 0.9631, + "step": 8951 + }, + { + "epoch": 0.22, + "learning_rate": 1.825301598186215e-05, + "loss": 0.948, + "step": 8952 + }, + { + "epoch": 0.22, + "learning_rate": 1.8252576121244455e-05, + "loss": 1.2447, + "step": 8953 + }, + { + "epoch": 0.22, + "learning_rate": 1.825213621056032e-05, + "loss": 1.0605, + "step": 8954 + }, + { + "epoch": 0.22, + "learning_rate": 1.8251696249812424e-05, + "loss": 0.9609, + "step": 8955 + }, + { + "epoch": 0.22, + "learning_rate": 1.825125623900342e-05, + "loss": 0.9635, + "step": 8956 + }, + { + "epoch": 0.22, + "learning_rate": 1.8250816178135988e-05, + "loss": 1.421, + "step": 8957 + }, + { + "epoch": 0.22, + "learning_rate": 1.8250376067212794e-05, + "loss": 1.1944, + "step": 8958 + }, + { + "epoch": 0.22, + "learning_rate": 1.824993590623651e-05, + "loss": 1.1183, + "step": 8959 + }, + { + "epoch": 0.22, + "learning_rate": 1.8249495695209805e-05, + "loss": 1.3193, + "step": 8960 + }, + { + "epoch": 0.22, + "learning_rate": 1.824905543413535e-05, + "loss": 1.3053, + "step": 8961 + }, + { + "epoch": 0.22, + "learning_rate": 1.8248615123015816e-05, + "loss": 1.1635, + "step": 8962 + }, + { + "epoch": 0.22, + "learning_rate": 1.8248174761853876e-05, + "loss": 1.1463, + "step": 8963 + }, + { + "epoch": 0.22, + "learning_rate": 1.8247734350652197e-05, + "loss": 0.8702, + "step": 8964 + }, + { + "epoch": 0.22, + "learning_rate": 1.8247293889413456e-05, + "loss": 0.9878, + "step": 8965 + }, + { + "epoch": 0.22, + "learning_rate": 1.824685337814032e-05, + "loss": 1.1451, + "step": 8966 + }, + { + "epoch": 0.22, + "learning_rate": 1.8246412816835465e-05, + "loss": 1.082, + "step": 8967 + }, + { + "epoch": 0.22, + "learning_rate": 1.8245972205501565e-05, + "loss": 1.2178, + "step": 8968 + }, + { + "epoch": 0.22, + "learning_rate": 1.8245531544141292e-05, + "loss": 0.933, + "step": 8969 + }, + { + "epoch": 0.22, + "learning_rate": 1.8245090832757317e-05, + "loss": 1.1933, + "step": 8970 + }, + { + "epoch": 0.22, + "learning_rate": 1.8244650071352312e-05, + "loss": 1.1917, + "step": 8971 + }, + { + "epoch": 0.22, + "learning_rate": 1.824420925992896e-05, + "loss": 1.1712, + "step": 8972 + }, + { + "epoch": 0.22, + "learning_rate": 1.824376839848993e-05, + "loss": 1.1209, + "step": 8973 + }, + { + "epoch": 0.22, + "learning_rate": 1.824332748703789e-05, + "loss": 0.8528, + "step": 8974 + }, + { + "epoch": 0.22, + "learning_rate": 1.8242886525575524e-05, + "loss": 0.9981, + "step": 8975 + }, + { + "epoch": 0.22, + "learning_rate": 1.8242445514105505e-05, + "loss": 1.4416, + "step": 8976 + }, + { + "epoch": 0.22, + "learning_rate": 1.8242004452630507e-05, + "loss": 1.0266, + "step": 8977 + }, + { + "epoch": 0.22, + "learning_rate": 1.8241563341153203e-05, + "loss": 1.1984, + "step": 8978 + }, + { + "epoch": 0.22, + "learning_rate": 1.8241122179676282e-05, + "loss": 1.3389, + "step": 8979 + }, + { + "epoch": 0.22, + "learning_rate": 1.82406809682024e-05, + "loss": 1.0969, + "step": 8980 + }, + { + "epoch": 0.22, + "learning_rate": 1.8240239706734254e-05, + "loss": 1.0556, + "step": 8981 + }, + { + "epoch": 0.22, + "learning_rate": 1.8239798395274507e-05, + "loss": 1.0234, + "step": 8982 + }, + { + "epoch": 0.22, + "learning_rate": 1.8239357033825843e-05, + "loss": 0.919, + "step": 8983 + }, + { + "epoch": 0.22, + "learning_rate": 1.823891562239094e-05, + "loss": 1.1072, + "step": 8984 + }, + { + "epoch": 0.22, + "learning_rate": 1.823847416097247e-05, + "loss": 0.9662, + "step": 8985 + }, + { + "epoch": 0.22, + "learning_rate": 1.8238032649573116e-05, + "loss": 1.1358, + "step": 8986 + }, + { + "epoch": 0.22, + "learning_rate": 1.8237591088195556e-05, + "loss": 1.203, + "step": 8987 + }, + { + "epoch": 0.22, + "learning_rate": 1.8237149476842472e-05, + "loss": 1.0857, + "step": 8988 + }, + { + "epoch": 0.22, + "learning_rate": 1.8236707815516535e-05, + "loss": 0.9581, + "step": 8989 + }, + { + "epoch": 0.22, + "learning_rate": 1.8236266104220432e-05, + "loss": 1.0791, + "step": 8990 + }, + { + "epoch": 0.22, + "learning_rate": 1.8235824342956835e-05, + "loss": 1.3119, + "step": 8991 + }, + { + "epoch": 0.22, + "learning_rate": 1.8235382531728435e-05, + "loss": 1.1596, + "step": 8992 + }, + { + "epoch": 0.22, + "learning_rate": 1.8234940670537905e-05, + "loss": 1.4114, + "step": 8993 + }, + { + "epoch": 0.22, + "learning_rate": 1.8234498759387925e-05, + "loss": 1.1152, + "step": 8994 + }, + { + "epoch": 0.22, + "learning_rate": 1.823405679828118e-05, + "loss": 1.0278, + "step": 8995 + }, + { + "epoch": 0.22, + "learning_rate": 1.8233614787220345e-05, + "loss": 1.2729, + "step": 8996 + }, + { + "epoch": 0.22, + "learning_rate": 1.823317272620811e-05, + "loss": 1.0107, + "step": 8997 + }, + { + "epoch": 0.22, + "learning_rate": 1.8232730615247146e-05, + "loss": 1.0041, + "step": 8998 + }, + { + "epoch": 0.22, + "learning_rate": 1.8232288454340147e-05, + "loss": 1.1417, + "step": 8999 + }, + { + "epoch": 0.22, + "learning_rate": 1.823184624348979e-05, + "loss": 1.228, + "step": 9000 + }, + { + "epoch": 0.22, + "learning_rate": 1.823140398269876e-05, + "loss": 1.2288, + "step": 9001 + }, + { + "epoch": 0.22, + "learning_rate": 1.8230961671969735e-05, + "loss": 1.3379, + "step": 9002 + }, + { + "epoch": 0.22, + "learning_rate": 1.8230519311305404e-05, + "loss": 1.2474, + "step": 9003 + }, + { + "epoch": 0.22, + "learning_rate": 1.8230076900708447e-05, + "loss": 1.3221, + "step": 9004 + }, + { + "epoch": 0.22, + "learning_rate": 1.822963444018155e-05, + "loss": 1.3585, + "step": 9005 + }, + { + "epoch": 0.22, + "learning_rate": 1.8229191929727395e-05, + "loss": 1.1616, + "step": 9006 + }, + { + "epoch": 0.22, + "learning_rate": 1.8228749369348667e-05, + "loss": 1.1088, + "step": 9007 + }, + { + "epoch": 0.22, + "learning_rate": 1.8228306759048057e-05, + "loss": 1.2393, + "step": 9008 + }, + { + "epoch": 0.22, + "learning_rate": 1.822786409882824e-05, + "loss": 1.1715, + "step": 9009 + }, + { + "epoch": 0.22, + "learning_rate": 1.8227421388691912e-05, + "loss": 1.1637, + "step": 9010 + }, + { + "epoch": 0.22, + "learning_rate": 1.8226978628641748e-05, + "loss": 1.2441, + "step": 9011 + }, + { + "epoch": 0.22, + "learning_rate": 1.8226535818680445e-05, + "loss": 1.1305, + "step": 9012 + }, + { + "epoch": 0.22, + "learning_rate": 1.822609295881068e-05, + "loss": 1.0548, + "step": 9013 + }, + { + "epoch": 0.22, + "learning_rate": 1.8225650049035148e-05, + "loss": 1.2198, + "step": 9014 + }, + { + "epoch": 0.22, + "learning_rate": 1.822520708935653e-05, + "loss": 1.1808, + "step": 9015 + }, + { + "epoch": 0.22, + "learning_rate": 1.8224764079777514e-05, + "loss": 1.136, + "step": 9016 + }, + { + "epoch": 0.22, + "learning_rate": 1.822432102030079e-05, + "loss": 0.9578, + "step": 9017 + }, + { + "epoch": 0.22, + "learning_rate": 1.8223877910929048e-05, + "loss": 1.2275, + "step": 9018 + }, + { + "epoch": 0.22, + "learning_rate": 1.8223434751664964e-05, + "loss": 1.2581, + "step": 9019 + }, + { + "epoch": 0.22, + "learning_rate": 1.8222991542511247e-05, + "loss": 1.1553, + "step": 9020 + }, + { + "epoch": 0.22, + "learning_rate": 1.8222548283470565e-05, + "loss": 1.0707, + "step": 9021 + }, + { + "epoch": 0.22, + "learning_rate": 1.8222104974545623e-05, + "loss": 1.0882, + "step": 9022 + }, + { + "epoch": 0.22, + "learning_rate": 1.82216616157391e-05, + "loss": 1.0902, + "step": 9023 + }, + { + "epoch": 0.22, + "learning_rate": 1.8221218207053694e-05, + "loss": 1.3078, + "step": 9024 + }, + { + "epoch": 0.22, + "learning_rate": 1.8220774748492088e-05, + "loss": 1.1767, + "step": 9025 + }, + { + "epoch": 0.22, + "learning_rate": 1.8220331240056974e-05, + "loss": 1.0542, + "step": 9026 + }, + { + "epoch": 0.22, + "learning_rate": 1.821988768175105e-05, + "loss": 1.1366, + "step": 9027 + }, + { + "epoch": 0.22, + "learning_rate": 1.8219444073576993e-05, + "loss": 1.1642, + "step": 9028 + }, + { + "epoch": 0.22, + "learning_rate": 1.8219000415537508e-05, + "loss": 1.0177, + "step": 9029 + }, + { + "epoch": 0.22, + "learning_rate": 1.8218556707635277e-05, + "loss": 1.1037, + "step": 9030 + }, + { + "epoch": 0.22, + "learning_rate": 1.8218112949872993e-05, + "loss": 0.9562, + "step": 9031 + }, + { + "epoch": 0.22, + "learning_rate": 1.8217669142253358e-05, + "loss": 1.1509, + "step": 9032 + }, + { + "epoch": 0.22, + "learning_rate": 1.8217225284779054e-05, + "loss": 1.194, + "step": 9033 + }, + { + "epoch": 0.22, + "learning_rate": 1.8216781377452775e-05, + "loss": 1.3037, + "step": 9034 + }, + { + "epoch": 0.22, + "learning_rate": 1.8216337420277218e-05, + "loss": 0.8401, + "step": 9035 + }, + { + "epoch": 0.22, + "learning_rate": 1.8215893413255073e-05, + "loss": 1.168, + "step": 9036 + }, + { + "epoch": 0.22, + "learning_rate": 1.821544935638904e-05, + "loss": 0.9318, + "step": 9037 + }, + { + "epoch": 0.22, + "learning_rate": 1.8215005249681804e-05, + "loss": 1.1708, + "step": 9038 + }, + { + "epoch": 0.22, + "learning_rate": 1.8214561093136063e-05, + "loss": 1.1068, + "step": 9039 + }, + { + "epoch": 0.22, + "learning_rate": 1.8214116886754513e-05, + "loss": 1.0987, + "step": 9040 + }, + { + "epoch": 0.22, + "learning_rate": 1.821367263053985e-05, + "loss": 1.0687, + "step": 9041 + }, + { + "epoch": 0.22, + "learning_rate": 1.8213228324494765e-05, + "loss": 1.0309, + "step": 9042 + }, + { + "epoch": 0.22, + "learning_rate": 1.8212783968621955e-05, + "loss": 1.3836, + "step": 9043 + }, + { + "epoch": 0.22, + "learning_rate": 1.821233956292412e-05, + "loss": 1.1549, + "step": 9044 + }, + { + "epoch": 0.22, + "learning_rate": 1.8211895107403948e-05, + "loss": 1.1395, + "step": 9045 + }, + { + "epoch": 0.22, + "learning_rate": 1.821145060206414e-05, + "loss": 1.1779, + "step": 9046 + }, + { + "epoch": 0.22, + "learning_rate": 1.8211006046907396e-05, + "loss": 1.2816, + "step": 9047 + }, + { + "epoch": 0.22, + "learning_rate": 1.8210561441936406e-05, + "loss": 1.0769, + "step": 9048 + }, + { + "epoch": 0.22, + "learning_rate": 1.8210116787153875e-05, + "loss": 1.1657, + "step": 9049 + }, + { + "epoch": 0.22, + "learning_rate": 1.8209672082562496e-05, + "loss": 1.054, + "step": 9050 + }, + { + "epoch": 0.22, + "learning_rate": 1.8209227328164964e-05, + "loss": 1.3611, + "step": 9051 + }, + { + "epoch": 0.22, + "learning_rate": 1.8208782523963985e-05, + "loss": 1.1824, + "step": 9052 + }, + { + "epoch": 0.22, + "learning_rate": 1.820833766996225e-05, + "loss": 1.1742, + "step": 9053 + }, + { + "epoch": 0.22, + "learning_rate": 1.8207892766162463e-05, + "loss": 1.3076, + "step": 9054 + }, + { + "epoch": 0.22, + "learning_rate": 1.8207447812567322e-05, + "loss": 1.3178, + "step": 9055 + }, + { + "epoch": 0.22, + "learning_rate": 1.820700280917952e-05, + "loss": 0.9708, + "step": 9056 + }, + { + "epoch": 0.22, + "learning_rate": 1.820655775600177e-05, + "loss": 1.3103, + "step": 9057 + }, + { + "epoch": 0.22, + "learning_rate": 1.820611265303676e-05, + "loss": 1.2849, + "step": 9058 + }, + { + "epoch": 0.22, + "learning_rate": 1.8205667500287197e-05, + "loss": 1.1383, + "step": 9059 + }, + { + "epoch": 0.22, + "learning_rate": 1.8205222297755774e-05, + "loss": 1.0279, + "step": 9060 + }, + { + "epoch": 0.22, + "learning_rate": 1.82047770454452e-05, + "loss": 1.2342, + "step": 9061 + }, + { + "epoch": 0.22, + "learning_rate": 1.820433174335818e-05, + "loss": 0.9723, + "step": 9062 + }, + { + "epoch": 0.22, + "learning_rate": 1.8203886391497404e-05, + "loss": 1.1148, + "step": 9063 + }, + { + "epoch": 0.22, + "learning_rate": 1.8203440989865577e-05, + "loss": 1.1549, + "step": 9064 + }, + { + "epoch": 0.22, + "learning_rate": 1.8202995538465405e-05, + "loss": 0.8815, + "step": 9065 + }, + { + "epoch": 0.22, + "learning_rate": 1.8202550037299588e-05, + "loss": 1.0557, + "step": 9066 + }, + { + "epoch": 0.22, + "learning_rate": 1.8202104486370832e-05, + "loss": 1.2908, + "step": 9067 + }, + { + "epoch": 0.22, + "learning_rate": 1.820165888568183e-05, + "loss": 1.3454, + "step": 9068 + }, + { + "epoch": 0.22, + "learning_rate": 1.82012132352353e-05, + "loss": 1.2739, + "step": 9069 + }, + { + "epoch": 0.22, + "learning_rate": 1.8200767535033938e-05, + "loss": 0.9024, + "step": 9070 + }, + { + "epoch": 0.22, + "learning_rate": 1.8200321785080442e-05, + "loss": 0.9133, + "step": 9071 + }, + { + "epoch": 0.22, + "learning_rate": 1.819987598537753e-05, + "loss": 1.1537, + "step": 9072 + }, + { + "epoch": 0.22, + "learning_rate": 1.8199430135927896e-05, + "loss": 1.3408, + "step": 9073 + }, + { + "epoch": 0.22, + "learning_rate": 1.8198984236734246e-05, + "loss": 1.0212, + "step": 9074 + }, + { + "epoch": 0.22, + "learning_rate": 1.819853828779929e-05, + "loss": 0.9623, + "step": 9075 + }, + { + "epoch": 0.22, + "learning_rate": 1.819809228912573e-05, + "loss": 1.0476, + "step": 9076 + }, + { + "epoch": 0.22, + "learning_rate": 1.819764624071627e-05, + "loss": 1.2146, + "step": 9077 + }, + { + "epoch": 0.22, + "learning_rate": 1.8197200142573625e-05, + "loss": 1.2932, + "step": 9078 + }, + { + "epoch": 0.22, + "learning_rate": 1.819675399470049e-05, + "loss": 0.7674, + "step": 9079 + }, + { + "epoch": 0.22, + "learning_rate": 1.819630779709958e-05, + "loss": 0.8277, + "step": 9080 + }, + { + "epoch": 0.22, + "learning_rate": 1.819586154977359e-05, + "loss": 1.1243, + "step": 9081 + }, + { + "epoch": 0.22, + "learning_rate": 1.8195415252725242e-05, + "loss": 1.1294, + "step": 9082 + }, + { + "epoch": 0.22, + "learning_rate": 1.819496890595724e-05, + "loss": 1.2941, + "step": 9083 + }, + { + "epoch": 0.22, + "learning_rate": 1.8194522509472283e-05, + "loss": 1.1453, + "step": 9084 + }, + { + "epoch": 0.22, + "learning_rate": 1.819407606327309e-05, + "loss": 1.1261, + "step": 9085 + }, + { + "epoch": 0.22, + "learning_rate": 1.819362956736236e-05, + "loss": 1.1916, + "step": 9086 + }, + { + "epoch": 0.22, + "learning_rate": 1.819318302174281e-05, + "loss": 1.4185, + "step": 9087 + }, + { + "epoch": 0.22, + "learning_rate": 1.8192736426417146e-05, + "loss": 1.1288, + "step": 9088 + }, + { + "epoch": 0.22, + "learning_rate": 1.819228978138808e-05, + "loss": 1.0307, + "step": 9089 + }, + { + "epoch": 0.22, + "learning_rate": 1.8191843086658313e-05, + "loss": 1.2957, + "step": 9090 + }, + { + "epoch": 0.22, + "learning_rate": 1.8191396342230563e-05, + "loss": 1.2261, + "step": 9091 + }, + { + "epoch": 0.22, + "learning_rate": 1.819094954810754e-05, + "loss": 1.1643, + "step": 9092 + }, + { + "epoch": 0.22, + "learning_rate": 1.819050270429195e-05, + "loss": 1.0716, + "step": 9093 + }, + { + "epoch": 0.22, + "learning_rate": 1.8190055810786507e-05, + "loss": 1.3903, + "step": 9094 + }, + { + "epoch": 0.22, + "learning_rate": 1.8189608867593925e-05, + "loss": 0.8921, + "step": 9095 + }, + { + "epoch": 0.22, + "learning_rate": 1.818916187471691e-05, + "loss": 1.3528, + "step": 9096 + }, + { + "epoch": 0.22, + "learning_rate": 1.8188714832158174e-05, + "loss": 0.9959, + "step": 9097 + }, + { + "epoch": 0.22, + "learning_rate": 1.8188267739920433e-05, + "loss": 1.0841, + "step": 9098 + }, + { + "epoch": 0.22, + "learning_rate": 1.81878205980064e-05, + "loss": 0.9765, + "step": 9099 + }, + { + "epoch": 0.22, + "learning_rate": 1.8187373406418782e-05, + "loss": 1.0024, + "step": 9100 + }, + { + "epoch": 0.22, + "learning_rate": 1.8186926165160296e-05, + "loss": 1.0528, + "step": 9101 + }, + { + "epoch": 0.22, + "learning_rate": 1.8186478874233655e-05, + "loss": 1.1516, + "step": 9102 + }, + { + "epoch": 0.22, + "learning_rate": 1.818603153364157e-05, + "loss": 0.8837, + "step": 9103 + }, + { + "epoch": 0.22, + "learning_rate": 1.8185584143386764e-05, + "loss": 1.0377, + "step": 9104 + }, + { + "epoch": 0.22, + "learning_rate": 1.8185136703471937e-05, + "loss": 1.0297, + "step": 9105 + }, + { + "epoch": 0.22, + "learning_rate": 1.8184689213899816e-05, + "loss": 1.0848, + "step": 9106 + }, + { + "epoch": 0.22, + "learning_rate": 1.8184241674673106e-05, + "loss": 1.1206, + "step": 9107 + }, + { + "epoch": 0.22, + "learning_rate": 1.818379408579453e-05, + "loss": 1.2522, + "step": 9108 + }, + { + "epoch": 0.22, + "learning_rate": 1.81833464472668e-05, + "loss": 0.9437, + "step": 9109 + }, + { + "epoch": 0.22, + "learning_rate": 1.818289875909263e-05, + "loss": 1.2009, + "step": 9110 + }, + { + "epoch": 0.22, + "learning_rate": 1.8182451021274744e-05, + "loss": 0.9052, + "step": 9111 + }, + { + "epoch": 0.22, + "learning_rate": 1.8182003233815847e-05, + "loss": 1.16, + "step": 9112 + }, + { + "epoch": 0.22, + "learning_rate": 1.818155539671866e-05, + "loss": 1.0687, + "step": 9113 + }, + { + "epoch": 0.22, + "learning_rate": 1.8181107509985903e-05, + "loss": 1.0183, + "step": 9114 + }, + { + "epoch": 0.22, + "learning_rate": 1.818065957362029e-05, + "loss": 1.2458, + "step": 9115 + }, + { + "epoch": 0.22, + "learning_rate": 1.8180211587624543e-05, + "loss": 1.1329, + "step": 9116 + }, + { + "epoch": 0.22, + "learning_rate": 1.8179763552001374e-05, + "loss": 1.2134, + "step": 9117 + }, + { + "epoch": 0.22, + "learning_rate": 1.81793154667535e-05, + "loss": 1.1177, + "step": 9118 + }, + { + "epoch": 0.22, + "learning_rate": 1.8178867331883646e-05, + "loss": 1.1139, + "step": 9119 + }, + { + "epoch": 0.22, + "learning_rate": 1.8178419147394528e-05, + "loss": 0.8016, + "step": 9120 + }, + { + "epoch": 0.22, + "learning_rate": 1.8177970913288866e-05, + "loss": 1.1814, + "step": 9121 + }, + { + "epoch": 0.22, + "learning_rate": 1.8177522629569375e-05, + "loss": 1.1718, + "step": 9122 + }, + { + "epoch": 0.22, + "learning_rate": 1.817707429623878e-05, + "loss": 1.0477, + "step": 9123 + }, + { + "epoch": 0.22, + "learning_rate": 1.8176625913299797e-05, + "loss": 1.0257, + "step": 9124 + }, + { + "epoch": 0.22, + "learning_rate": 1.817617748075515e-05, + "loss": 1.0389, + "step": 9125 + }, + { + "epoch": 0.22, + "learning_rate": 1.8175728998607555e-05, + "loss": 1.2376, + "step": 9126 + }, + { + "epoch": 0.22, + "learning_rate": 1.8175280466859736e-05, + "loss": 1.065, + "step": 9127 + }, + { + "epoch": 0.22, + "learning_rate": 1.8174831885514415e-05, + "loss": 1.2808, + "step": 9128 + }, + { + "epoch": 0.22, + "learning_rate": 1.817438325457431e-05, + "loss": 0.8631, + "step": 9129 + }, + { + "epoch": 0.22, + "learning_rate": 1.817393457404214e-05, + "loss": 0.8597, + "step": 9130 + }, + { + "epoch": 0.22, + "learning_rate": 1.817348584392064e-05, + "loss": 1.0276, + "step": 9131 + }, + { + "epoch": 0.22, + "learning_rate": 1.8173037064212522e-05, + "loss": 1.4277, + "step": 9132 + }, + { + "epoch": 0.22, + "learning_rate": 1.817258823492051e-05, + "loss": 0.753, + "step": 9133 + }, + { + "epoch": 0.22, + "learning_rate": 1.8172139356047323e-05, + "loss": 1.2795, + "step": 9134 + }, + { + "epoch": 0.22, + "learning_rate": 1.8171690427595693e-05, + "loss": 0.9988, + "step": 9135 + }, + { + "epoch": 0.22, + "learning_rate": 1.817124144956834e-05, + "loss": 1.0467, + "step": 9136 + }, + { + "epoch": 0.22, + "learning_rate": 1.8170792421967986e-05, + "loss": 1.0032, + "step": 9137 + }, + { + "epoch": 0.22, + "learning_rate": 1.8170343344797354e-05, + "loss": 1.129, + "step": 9138 + }, + { + "epoch": 0.22, + "learning_rate": 1.8169894218059173e-05, + "loss": 1.2145, + "step": 9139 + }, + { + "epoch": 0.22, + "learning_rate": 1.8169445041756165e-05, + "loss": 1.3111, + "step": 9140 + }, + { + "epoch": 0.22, + "learning_rate": 1.8168995815891054e-05, + "loss": 0.964, + "step": 9141 + }, + { + "epoch": 0.22, + "learning_rate": 1.8168546540466567e-05, + "loss": 0.9981, + "step": 9142 + }, + { + "epoch": 0.22, + "learning_rate": 1.8168097215485434e-05, + "loss": 1.223, + "step": 9143 + }, + { + "epoch": 0.22, + "learning_rate": 1.816764784095037e-05, + "loss": 1.2319, + "step": 9144 + }, + { + "epoch": 0.22, + "learning_rate": 1.8167198416864113e-05, + "loss": 1.1186, + "step": 9145 + }, + { + "epoch": 0.22, + "learning_rate": 1.816674894322938e-05, + "loss": 1.2479, + "step": 9146 + }, + { + "epoch": 0.22, + "learning_rate": 1.8166299420048903e-05, + "loss": 1.3155, + "step": 9147 + }, + { + "epoch": 0.22, + "learning_rate": 1.8165849847325413e-05, + "loss": 1.0398, + "step": 9148 + }, + { + "epoch": 0.22, + "learning_rate": 1.816540022506163e-05, + "loss": 1.0264, + "step": 9149 + }, + { + "epoch": 0.22, + "learning_rate": 1.816495055326028e-05, + "loss": 0.971, + "step": 9150 + }, + { + "epoch": 0.22, + "learning_rate": 1.81645008319241e-05, + "loss": 1.1104, + "step": 9151 + }, + { + "epoch": 0.22, + "learning_rate": 1.8164051061055812e-05, + "loss": 1.249, + "step": 9152 + }, + { + "epoch": 0.22, + "learning_rate": 1.816360124065815e-05, + "loss": 0.9181, + "step": 9153 + }, + { + "epoch": 0.22, + "learning_rate": 1.8163151370733838e-05, + "loss": 1.4525, + "step": 9154 + }, + { + "epoch": 0.22, + "learning_rate": 1.8162701451285606e-05, + "loss": 1.0922, + "step": 9155 + }, + { + "epoch": 0.22, + "learning_rate": 1.8162251482316186e-05, + "loss": 1.1634, + "step": 9156 + }, + { + "epoch": 0.22, + "learning_rate": 1.8161801463828302e-05, + "loss": 1.3217, + "step": 9157 + }, + { + "epoch": 0.22, + "learning_rate": 1.8161351395824688e-05, + "loss": 1.2168, + "step": 9158 + }, + { + "epoch": 0.22, + "learning_rate": 1.8160901278308078e-05, + "loss": 1.215, + "step": 9159 + }, + { + "epoch": 0.22, + "learning_rate": 1.8160451111281202e-05, + "loss": 1.2334, + "step": 9160 + }, + { + "epoch": 0.22, + "learning_rate": 1.8160000894746783e-05, + "loss": 0.9267, + "step": 9161 + }, + { + "epoch": 0.22, + "learning_rate": 1.815955062870756e-05, + "loss": 1.2998, + "step": 9162 + }, + { + "epoch": 0.22, + "learning_rate": 1.8159100313166265e-05, + "loss": 1.5735, + "step": 9163 + }, + { + "epoch": 0.22, + "learning_rate": 1.815864994812562e-05, + "loss": 0.8993, + "step": 9164 + }, + { + "epoch": 0.22, + "learning_rate": 1.8158199533588374e-05, + "loss": 1.1398, + "step": 9165 + }, + { + "epoch": 0.22, + "learning_rate": 1.8157749069557246e-05, + "loss": 1.3307, + "step": 9166 + }, + { + "epoch": 0.22, + "learning_rate": 1.8157298556034972e-05, + "loss": 1.3066, + "step": 9167 + }, + { + "epoch": 0.22, + "learning_rate": 1.815684799302429e-05, + "loss": 1.0358, + "step": 9168 + }, + { + "epoch": 0.22, + "learning_rate": 1.8156397380527928e-05, + "loss": 1.1582, + "step": 9169 + }, + { + "epoch": 0.22, + "learning_rate": 1.815594671854862e-05, + "loss": 1.172, + "step": 9170 + }, + { + "epoch": 0.22, + "learning_rate": 1.8155496007089105e-05, + "loss": 0.9438, + "step": 9171 + }, + { + "epoch": 0.22, + "learning_rate": 1.8155045246152113e-05, + "loss": 1.286, + "step": 9172 + }, + { + "epoch": 0.22, + "learning_rate": 1.815459443574038e-05, + "loss": 1.1465, + "step": 9173 + }, + { + "epoch": 0.22, + "learning_rate": 1.8154143575856634e-05, + "loss": 0.9759, + "step": 9174 + }, + { + "epoch": 0.22, + "learning_rate": 1.8153692666503623e-05, + "loss": 1.2369, + "step": 9175 + }, + { + "epoch": 0.22, + "learning_rate": 1.8153241707684077e-05, + "loss": 1.2104, + "step": 9176 + }, + { + "epoch": 0.22, + "learning_rate": 1.815279069940073e-05, + "loss": 1.0527, + "step": 9177 + }, + { + "epoch": 0.22, + "learning_rate": 1.815233964165632e-05, + "loss": 1.1916, + "step": 9178 + }, + { + "epoch": 0.22, + "learning_rate": 1.8151888534453584e-05, + "loss": 1.1525, + "step": 9179 + }, + { + "epoch": 0.22, + "learning_rate": 1.8151437377795256e-05, + "loss": 1.3952, + "step": 9180 + }, + { + "epoch": 0.22, + "learning_rate": 1.8150986171684076e-05, + "loss": 0.9566, + "step": 9181 + }, + { + "epoch": 0.22, + "learning_rate": 1.8150534916122777e-05, + "loss": 1.0968, + "step": 9182 + }, + { + "epoch": 0.22, + "learning_rate": 1.8150083611114103e-05, + "loss": 1.2101, + "step": 9183 + }, + { + "epoch": 0.22, + "learning_rate": 1.814963225666079e-05, + "loss": 1.1749, + "step": 9184 + }, + { + "epoch": 0.22, + "learning_rate": 1.8149180852765572e-05, + "loss": 1.5987, + "step": 9185 + }, + { + "epoch": 0.22, + "learning_rate": 1.8148729399431187e-05, + "loss": 1.0535, + "step": 9186 + }, + { + "epoch": 0.22, + "learning_rate": 1.814827789666038e-05, + "loss": 1.2069, + "step": 9187 + }, + { + "epoch": 0.22, + "learning_rate": 1.8147826344455893e-05, + "loss": 1.308, + "step": 9188 + }, + { + "epoch": 0.22, + "learning_rate": 1.8147374742820455e-05, + "loss": 1.1302, + "step": 9189 + }, + { + "epoch": 0.22, + "learning_rate": 1.8146923091756813e-05, + "loss": 1.0636, + "step": 9190 + }, + { + "epoch": 0.22, + "learning_rate": 1.8146471391267704e-05, + "loss": 0.9192, + "step": 9191 + }, + { + "epoch": 0.22, + "learning_rate": 1.814601964135587e-05, + "loss": 1.276, + "step": 9192 + }, + { + "epoch": 0.22, + "learning_rate": 1.814556784202405e-05, + "loss": 1.1789, + "step": 9193 + }, + { + "epoch": 0.22, + "learning_rate": 1.8145115993274986e-05, + "loss": 1.2105, + "step": 9194 + }, + { + "epoch": 0.22, + "learning_rate": 1.814466409511142e-05, + "loss": 1.3096, + "step": 9195 + }, + { + "epoch": 0.22, + "learning_rate": 1.814421214753609e-05, + "loss": 0.9446, + "step": 9196 + }, + { + "epoch": 0.22, + "learning_rate": 1.8143760150551742e-05, + "loss": 0.9016, + "step": 9197 + }, + { + "epoch": 0.22, + "learning_rate": 1.8143308104161117e-05, + "loss": 1.1975, + "step": 9198 + }, + { + "epoch": 0.22, + "learning_rate": 1.814285600836696e-05, + "loss": 1.2822, + "step": 9199 + }, + { + "epoch": 0.22, + "learning_rate": 1.8142403863172007e-05, + "loss": 1.0363, + "step": 9200 + }, + { + "epoch": 0.22, + "learning_rate": 1.8141951668579005e-05, + "loss": 1.1702, + "step": 9201 + }, + { + "epoch": 0.22, + "learning_rate": 1.81414994245907e-05, + "loss": 1.1957, + "step": 9202 + }, + { + "epoch": 0.22, + "learning_rate": 1.814104713120983e-05, + "loss": 0.9861, + "step": 9203 + }, + { + "epoch": 0.22, + "learning_rate": 1.8140594788439142e-05, + "loss": 0.9559, + "step": 9204 + }, + { + "epoch": 0.22, + "learning_rate": 1.814014239628138e-05, + "loss": 1.2864, + "step": 9205 + }, + { + "epoch": 0.22, + "learning_rate": 1.8139689954739294e-05, + "loss": 1.1737, + "step": 9206 + }, + { + "epoch": 0.22, + "learning_rate": 1.813923746381562e-05, + "loss": 1.0586, + "step": 9207 + }, + { + "epoch": 0.22, + "learning_rate": 1.8138784923513107e-05, + "loss": 1.1691, + "step": 9208 + }, + { + "epoch": 0.22, + "learning_rate": 1.81383323338345e-05, + "loss": 1.0424, + "step": 9209 + }, + { + "epoch": 0.22, + "learning_rate": 1.8137879694782543e-05, + "loss": 1.2444, + "step": 9210 + }, + { + "epoch": 0.22, + "learning_rate": 1.8137427006359988e-05, + "loss": 1.1519, + "step": 9211 + }, + { + "epoch": 0.22, + "learning_rate": 1.8136974268569575e-05, + "loss": 1.1833, + "step": 9212 + }, + { + "epoch": 0.22, + "learning_rate": 1.8136521481414054e-05, + "loss": 1.0363, + "step": 9213 + }, + { + "epoch": 0.22, + "learning_rate": 1.813606864489617e-05, + "loss": 1.17, + "step": 9214 + }, + { + "epoch": 0.22, + "learning_rate": 1.8135615759018673e-05, + "loss": 1.13, + "step": 9215 + }, + { + "epoch": 0.22, + "learning_rate": 1.8135162823784303e-05, + "loss": 1.4041, + "step": 9216 + }, + { + "epoch": 0.22, + "learning_rate": 1.813470983919582e-05, + "loss": 0.9142, + "step": 9217 + }, + { + "epoch": 0.22, + "learning_rate": 1.8134256805255964e-05, + "loss": 1.1374, + "step": 9218 + }, + { + "epoch": 0.22, + "learning_rate": 1.8133803721967486e-05, + "loss": 1.1798, + "step": 9219 + }, + { + "epoch": 0.22, + "learning_rate": 1.8133350589333136e-05, + "loss": 1.081, + "step": 9220 + }, + { + "epoch": 0.22, + "learning_rate": 1.8132897407355657e-05, + "loss": 1.0611, + "step": 9221 + }, + { + "epoch": 0.22, + "learning_rate": 1.8132444176037803e-05, + "loss": 1.1724, + "step": 9222 + }, + { + "epoch": 0.22, + "learning_rate": 1.813199089538232e-05, + "loss": 1.088, + "step": 9223 + }, + { + "epoch": 0.22, + "learning_rate": 1.8131537565391967e-05, + "loss": 1.3296, + "step": 9224 + }, + { + "epoch": 0.22, + "learning_rate": 1.8131084186069485e-05, + "loss": 0.8865, + "step": 9225 + }, + { + "epoch": 0.22, + "learning_rate": 1.8130630757417628e-05, + "loss": 1.1734, + "step": 9226 + }, + { + "epoch": 0.22, + "learning_rate": 1.813017727943915e-05, + "loss": 1.0648, + "step": 9227 + }, + { + "epoch": 0.22, + "learning_rate": 1.8129723752136797e-05, + "loss": 1.1972, + "step": 9228 + }, + { + "epoch": 0.22, + "learning_rate": 1.8129270175513318e-05, + "loss": 1.1193, + "step": 9229 + }, + { + "epoch": 0.22, + "learning_rate": 1.8128816549571472e-05, + "loss": 1.0121, + "step": 9230 + }, + { + "epoch": 0.22, + "learning_rate": 1.812836287431401e-05, + "loss": 1.1486, + "step": 9231 + }, + { + "epoch": 0.22, + "learning_rate": 1.812790914974368e-05, + "loss": 1.1016, + "step": 9232 + }, + { + "epoch": 0.22, + "learning_rate": 1.812745537586324e-05, + "loss": 1.3455, + "step": 9233 + }, + { + "epoch": 0.22, + "learning_rate": 1.8127001552675436e-05, + "loss": 1.0835, + "step": 9234 + }, + { + "epoch": 0.22, + "learning_rate": 1.8126547680183027e-05, + "loss": 1.2245, + "step": 9235 + }, + { + "epoch": 0.22, + "learning_rate": 1.8126093758388764e-05, + "loss": 1.1651, + "step": 9236 + }, + { + "epoch": 0.22, + "learning_rate": 1.81256397872954e-05, + "loss": 1.1673, + "step": 9237 + }, + { + "epoch": 0.22, + "learning_rate": 1.8125185766905697e-05, + "loss": 1.2318, + "step": 9238 + }, + { + "epoch": 0.22, + "learning_rate": 1.8124731697222398e-05, + "loss": 1.1085, + "step": 9239 + }, + { + "epoch": 0.22, + "learning_rate": 1.812427757824826e-05, + "loss": 0.9675, + "step": 9240 + }, + { + "epoch": 0.22, + "learning_rate": 1.8123823409986044e-05, + "loss": 1.1761, + "step": 9241 + }, + { + "epoch": 0.22, + "learning_rate": 1.8123369192438508e-05, + "loss": 1.288, + "step": 9242 + }, + { + "epoch": 0.22, + "learning_rate": 1.8122914925608396e-05, + "loss": 1.0384, + "step": 9243 + }, + { + "epoch": 0.22, + "learning_rate": 1.812246060949847e-05, + "loss": 0.8279, + "step": 9244 + }, + { + "epoch": 0.22, + "learning_rate": 1.8122006244111484e-05, + "loss": 1.4444, + "step": 9245 + }, + { + "epoch": 0.22, + "learning_rate": 1.81215518294502e-05, + "loss": 0.9913, + "step": 9246 + }, + { + "epoch": 0.22, + "learning_rate": 1.812109736551737e-05, + "loss": 1.258, + "step": 9247 + }, + { + "epoch": 0.22, + "learning_rate": 1.812064285231575e-05, + "loss": 0.947, + "step": 9248 + }, + { + "epoch": 0.22, + "learning_rate": 1.8120188289848105e-05, + "loss": 1.2872, + "step": 9249 + }, + { + "epoch": 0.22, + "learning_rate": 1.8119733678117185e-05, + "loss": 1.171, + "step": 9250 + }, + { + "epoch": 0.22, + "learning_rate": 1.8119279017125753e-05, + "loss": 0.7936, + "step": 9251 + }, + { + "epoch": 0.22, + "learning_rate": 1.811882430687656e-05, + "loss": 1.3328, + "step": 9252 + }, + { + "epoch": 0.22, + "learning_rate": 1.8118369547372373e-05, + "loss": 1.0122, + "step": 9253 + }, + { + "epoch": 0.22, + "learning_rate": 1.811791473861595e-05, + "loss": 1.1009, + "step": 9254 + }, + { + "epoch": 0.22, + "learning_rate": 1.8117459880610043e-05, + "loss": 1.1666, + "step": 9255 + }, + { + "epoch": 0.22, + "learning_rate": 1.811700497335742e-05, + "loss": 1.4124, + "step": 9256 + }, + { + "epoch": 0.22, + "learning_rate": 1.8116550016860833e-05, + "loss": 1.0566, + "step": 9257 + }, + { + "epoch": 0.22, + "learning_rate": 1.811609501112305e-05, + "loss": 1.1665, + "step": 9258 + }, + { + "epoch": 0.22, + "learning_rate": 1.8115639956146823e-05, + "loss": 0.9575, + "step": 9259 + }, + { + "epoch": 0.22, + "learning_rate": 1.8115184851934922e-05, + "loss": 1.0787, + "step": 9260 + }, + { + "epoch": 0.22, + "learning_rate": 1.81147296984901e-05, + "loss": 1.0946, + "step": 9261 + }, + { + "epoch": 0.22, + "learning_rate": 1.8114274495815123e-05, + "loss": 1.1393, + "step": 9262 + }, + { + "epoch": 0.22, + "learning_rate": 1.811381924391275e-05, + "loss": 1.0914, + "step": 9263 + }, + { + "epoch": 0.22, + "learning_rate": 1.8113363942785747e-05, + "loss": 1.2894, + "step": 9264 + }, + { + "epoch": 0.22, + "learning_rate": 1.8112908592436872e-05, + "loss": 1.2409, + "step": 9265 + }, + { + "epoch": 0.22, + "learning_rate": 1.811245319286889e-05, + "loss": 1.2876, + "step": 9266 + }, + { + "epoch": 0.22, + "learning_rate": 1.811199774408456e-05, + "loss": 1.1161, + "step": 9267 + }, + { + "epoch": 0.22, + "learning_rate": 1.811154224608665e-05, + "loss": 1.2446, + "step": 9268 + }, + { + "epoch": 0.22, + "learning_rate": 1.8111086698877918e-05, + "loss": 1.1402, + "step": 9269 + }, + { + "epoch": 0.22, + "learning_rate": 1.8110631102461134e-05, + "loss": 0.9955, + "step": 9270 + }, + { + "epoch": 0.22, + "learning_rate": 1.8110175456839058e-05, + "loss": 1.2449, + "step": 9271 + }, + { + "epoch": 0.22, + "learning_rate": 1.8109719762014454e-05, + "loss": 1.2676, + "step": 9272 + }, + { + "epoch": 0.22, + "learning_rate": 1.810926401799009e-05, + "loss": 1.049, + "step": 9273 + }, + { + "epoch": 0.22, + "learning_rate": 1.8108808224768724e-05, + "loss": 1.1466, + "step": 9274 + }, + { + "epoch": 0.22, + "learning_rate": 1.8108352382353132e-05, + "loss": 1.1066, + "step": 9275 + }, + { + "epoch": 0.22, + "learning_rate": 1.8107896490746067e-05, + "loss": 1.2426, + "step": 9276 + }, + { + "epoch": 0.22, + "learning_rate": 1.810744054995031e-05, + "loss": 1.2843, + "step": 9277 + }, + { + "epoch": 0.22, + "learning_rate": 1.810698455996861e-05, + "loss": 1.1183, + "step": 9278 + }, + { + "epoch": 0.22, + "learning_rate": 1.8106528520803744e-05, + "loss": 1.1851, + "step": 9279 + }, + { + "epoch": 0.22, + "learning_rate": 1.8106072432458478e-05, + "loss": 1.0267, + "step": 9280 + }, + { + "epoch": 0.22, + "learning_rate": 1.8105616294935577e-05, + "loss": 1.369, + "step": 9281 + }, + { + "epoch": 0.22, + "learning_rate": 1.8105160108237805e-05, + "loss": 0.9934, + "step": 9282 + }, + { + "epoch": 0.22, + "learning_rate": 1.8104703872367934e-05, + "loss": 0.9505, + "step": 9283 + }, + { + "epoch": 0.22, + "learning_rate": 1.8104247587328733e-05, + "loss": 1.1788, + "step": 9284 + }, + { + "epoch": 0.22, + "learning_rate": 1.810379125312297e-05, + "loss": 0.9459, + "step": 9285 + }, + { + "epoch": 0.22, + "learning_rate": 1.8103334869753406e-05, + "loss": 1.2068, + "step": 9286 + }, + { + "epoch": 0.22, + "learning_rate": 1.8102878437222817e-05, + "loss": 1.1101, + "step": 9287 + }, + { + "epoch": 0.22, + "learning_rate": 1.8102421955533974e-05, + "loss": 1.1601, + "step": 9288 + }, + { + "epoch": 0.22, + "learning_rate": 1.810196542468964e-05, + "loss": 1.2409, + "step": 9289 + }, + { + "epoch": 0.22, + "learning_rate": 1.8101508844692586e-05, + "loss": 1.2335, + "step": 9290 + }, + { + "epoch": 0.22, + "learning_rate": 1.8101052215545586e-05, + "loss": 0.9633, + "step": 9291 + }, + { + "epoch": 0.22, + "learning_rate": 1.810059553725141e-05, + "loss": 1.2706, + "step": 9292 + }, + { + "epoch": 0.22, + "learning_rate": 1.8100138809812822e-05, + "loss": 1.2878, + "step": 9293 + }, + { + "epoch": 0.22, + "learning_rate": 1.80996820332326e-05, + "loss": 1.1421, + "step": 9294 + }, + { + "epoch": 0.22, + "learning_rate": 1.809922520751351e-05, + "loss": 1.2675, + "step": 9295 + }, + { + "epoch": 0.22, + "learning_rate": 1.8098768332658325e-05, + "loss": 1.1071, + "step": 9296 + }, + { + "epoch": 0.22, + "learning_rate": 1.8098311408669818e-05, + "loss": 1.1738, + "step": 9297 + }, + { + "epoch": 0.22, + "learning_rate": 1.809785443555076e-05, + "loss": 1.2264, + "step": 9298 + }, + { + "epoch": 0.22, + "learning_rate": 1.809739741330393e-05, + "loss": 1.1281, + "step": 9299 + }, + { + "epoch": 0.22, + "learning_rate": 1.809694034193209e-05, + "loss": 1.1799, + "step": 9300 + }, + { + "epoch": 0.22, + "learning_rate": 1.8096483221438016e-05, + "loss": 1.1401, + "step": 9301 + }, + { + "epoch": 0.22, + "learning_rate": 1.8096026051824483e-05, + "loss": 1.0267, + "step": 9302 + }, + { + "epoch": 0.22, + "learning_rate": 1.8095568833094267e-05, + "loss": 1.093, + "step": 9303 + }, + { + "epoch": 0.22, + "learning_rate": 1.8095111565250137e-05, + "loss": 0.9347, + "step": 9304 + }, + { + "epoch": 0.22, + "learning_rate": 1.8094654248294866e-05, + "loss": 1.2481, + "step": 9305 + }, + { + "epoch": 0.22, + "learning_rate": 1.8094196882231235e-05, + "loss": 0.9162, + "step": 9306 + }, + { + "epoch": 0.22, + "learning_rate": 1.809373946706202e-05, + "loss": 1.4426, + "step": 9307 + }, + { + "epoch": 0.22, + "learning_rate": 1.8093282002789983e-05, + "loss": 1.2643, + "step": 9308 + }, + { + "epoch": 0.22, + "learning_rate": 1.8092824489417913e-05, + "loss": 0.8394, + "step": 9309 + }, + { + "epoch": 0.22, + "learning_rate": 1.8092366926948578e-05, + "loss": 1.0754, + "step": 9310 + }, + { + "epoch": 0.22, + "learning_rate": 1.809190931538476e-05, + "loss": 1.1556, + "step": 9311 + }, + { + "epoch": 0.22, + "learning_rate": 1.8091451654729225e-05, + "loss": 1.1115, + "step": 9312 + }, + { + "epoch": 0.22, + "learning_rate": 1.8090993944984763e-05, + "loss": 1.2247, + "step": 9313 + }, + { + "epoch": 0.22, + "learning_rate": 1.8090536186154143e-05, + "loss": 1.1657, + "step": 9314 + }, + { + "epoch": 0.22, + "learning_rate": 1.8090078378240138e-05, + "loss": 1.197, + "step": 9315 + }, + { + "epoch": 0.22, + "learning_rate": 1.8089620521245534e-05, + "loss": 1.1769, + "step": 9316 + }, + { + "epoch": 0.22, + "learning_rate": 1.8089162615173103e-05, + "loss": 1.1737, + "step": 9317 + }, + { + "epoch": 0.22, + "learning_rate": 1.8088704660025626e-05, + "loss": 1.4701, + "step": 9318 + }, + { + "epoch": 0.22, + "learning_rate": 1.808824665580588e-05, + "loss": 1.0839, + "step": 9319 + }, + { + "epoch": 0.22, + "learning_rate": 1.8087788602516643e-05, + "loss": 1.042, + "step": 9320 + }, + { + "epoch": 0.22, + "learning_rate": 1.80873305001607e-05, + "loss": 1.147, + "step": 9321 + }, + { + "epoch": 0.22, + "learning_rate": 1.808687234874082e-05, + "loss": 0.9557, + "step": 9322 + }, + { + "epoch": 0.22, + "learning_rate": 1.8086414148259794e-05, + "loss": 1.0929, + "step": 9323 + }, + { + "epoch": 0.22, + "learning_rate": 1.8085955898720388e-05, + "loss": 1.2276, + "step": 9324 + }, + { + "epoch": 0.22, + "learning_rate": 1.808549760012539e-05, + "loss": 1.2291, + "step": 9325 + }, + { + "epoch": 0.22, + "learning_rate": 1.8085039252477584e-05, + "loss": 1.0909, + "step": 9326 + }, + { + "epoch": 0.22, + "learning_rate": 1.8084580855779746e-05, + "loss": 1.2606, + "step": 9327 + }, + { + "epoch": 0.22, + "learning_rate": 1.8084122410034655e-05, + "loss": 1.2609, + "step": 9328 + }, + { + "epoch": 0.22, + "learning_rate": 1.8083663915245096e-05, + "loss": 1.2167, + "step": 9329 + }, + { + "epoch": 0.22, + "learning_rate": 1.808320537141385e-05, + "loss": 1.1975, + "step": 9330 + }, + { + "epoch": 0.22, + "learning_rate": 1.8082746778543697e-05, + "loss": 1.2506, + "step": 9331 + }, + { + "epoch": 0.22, + "learning_rate": 1.808228813663742e-05, + "loss": 1.3042, + "step": 9332 + }, + { + "epoch": 0.22, + "learning_rate": 1.8081829445697806e-05, + "loss": 0.7941, + "step": 9333 + }, + { + "epoch": 0.22, + "learning_rate": 1.8081370705727632e-05, + "loss": 1.3186, + "step": 9334 + }, + { + "epoch": 0.22, + "learning_rate": 1.808091191672968e-05, + "loss": 1.1819, + "step": 9335 + }, + { + "epoch": 0.22, + "learning_rate": 1.8080453078706737e-05, + "loss": 1.2393, + "step": 9336 + }, + { + "epoch": 0.22, + "learning_rate": 1.807999419166159e-05, + "loss": 1.0499, + "step": 9337 + }, + { + "epoch": 0.22, + "learning_rate": 1.8079535255597014e-05, + "loss": 1.1333, + "step": 9338 + }, + { + "epoch": 0.22, + "learning_rate": 1.80790762705158e-05, + "loss": 1.3294, + "step": 9339 + }, + { + "epoch": 0.22, + "learning_rate": 1.807861723642073e-05, + "loss": 1.1193, + "step": 9340 + }, + { + "epoch": 0.22, + "learning_rate": 1.807815815331459e-05, + "loss": 1.2003, + "step": 9341 + }, + { + "epoch": 0.22, + "learning_rate": 1.8077699021200163e-05, + "loss": 1.2858, + "step": 9342 + }, + { + "epoch": 0.22, + "learning_rate": 1.8077239840080237e-05, + "loss": 1.1583, + "step": 9343 + }, + { + "epoch": 0.22, + "learning_rate": 1.80767806099576e-05, + "loss": 1.3369, + "step": 9344 + }, + { + "epoch": 0.22, + "learning_rate": 1.807632133083503e-05, + "loss": 1.1643, + "step": 9345 + }, + { + "epoch": 0.22, + "learning_rate": 1.807586200271532e-05, + "loss": 0.9354, + "step": 9346 + }, + { + "epoch": 0.22, + "learning_rate": 1.8075402625601254e-05, + "loss": 1.2351, + "step": 9347 + }, + { + "epoch": 0.22, + "learning_rate": 1.8074943199495622e-05, + "loss": 1.2462, + "step": 9348 + }, + { + "epoch": 0.22, + "learning_rate": 1.8074483724401206e-05, + "loss": 1.4263, + "step": 9349 + }, + { + "epoch": 0.22, + "learning_rate": 1.8074024200320797e-05, + "loss": 1.1729, + "step": 9350 + }, + { + "epoch": 0.22, + "learning_rate": 1.8073564627257182e-05, + "loss": 1.1864, + "step": 9351 + }, + { + "epoch": 0.22, + "learning_rate": 1.8073105005213154e-05, + "loss": 1.4111, + "step": 9352 + }, + { + "epoch": 0.22, + "learning_rate": 1.8072645334191493e-05, + "loss": 0.7571, + "step": 9353 + }, + { + "epoch": 0.22, + "learning_rate": 1.807218561419499e-05, + "loss": 1.1827, + "step": 9354 + }, + { + "epoch": 0.22, + "learning_rate": 1.8071725845226436e-05, + "loss": 1.1329, + "step": 9355 + }, + { + "epoch": 0.23, + "learning_rate": 1.8071266027288625e-05, + "loss": 1.2714, + "step": 9356 + }, + { + "epoch": 0.23, + "learning_rate": 1.8070806160384336e-05, + "loss": 1.0786, + "step": 9357 + }, + { + "epoch": 0.23, + "learning_rate": 1.8070346244516367e-05, + "loss": 1.0524, + "step": 9358 + }, + { + "epoch": 0.23, + "learning_rate": 1.8069886279687504e-05, + "loss": 1.0384, + "step": 9359 + }, + { + "epoch": 0.23, + "learning_rate": 1.806942626590054e-05, + "loss": 1.087, + "step": 9360 + }, + { + "epoch": 0.23, + "learning_rate": 1.8068966203158263e-05, + "loss": 1.1341, + "step": 9361 + }, + { + "epoch": 0.23, + "learning_rate": 1.8068506091463473e-05, + "loss": 0.9858, + "step": 9362 + }, + { + "epoch": 0.23, + "learning_rate": 1.8068045930818946e-05, + "loss": 1.0735, + "step": 9363 + }, + { + "epoch": 0.23, + "learning_rate": 1.806758572122749e-05, + "loss": 1.0752, + "step": 9364 + }, + { + "epoch": 0.23, + "learning_rate": 1.806712546269188e-05, + "loss": 1.1742, + "step": 9365 + }, + { + "epoch": 0.23, + "learning_rate": 1.806666515521492e-05, + "loss": 1.1113, + "step": 9366 + }, + { + "epoch": 0.23, + "learning_rate": 1.8066204798799406e-05, + "loss": 1.0441, + "step": 9367 + }, + { + "epoch": 0.23, + "learning_rate": 1.8065744393448118e-05, + "loss": 1.0846, + "step": 9368 + }, + { + "epoch": 0.23, + "learning_rate": 1.806528393916386e-05, + "loss": 1.2848, + "step": 9369 + }, + { + "epoch": 0.23, + "learning_rate": 1.806482343594942e-05, + "loss": 1.1487, + "step": 9370 + }, + { + "epoch": 0.23, + "learning_rate": 1.8064362883807593e-05, + "loss": 1.1756, + "step": 9371 + }, + { + "epoch": 0.23, + "learning_rate": 1.806390228274117e-05, + "loss": 1.3963, + "step": 9372 + }, + { + "epoch": 0.23, + "learning_rate": 1.8063441632752954e-05, + "loss": 1.0274, + "step": 9373 + }, + { + "epoch": 0.23, + "learning_rate": 1.8062980933845732e-05, + "loss": 1.0909, + "step": 9374 + }, + { + "epoch": 0.23, + "learning_rate": 1.80625201860223e-05, + "loss": 1.5676, + "step": 9375 + }, + { + "epoch": 0.23, + "learning_rate": 1.8062059389285455e-05, + "loss": 1.4391, + "step": 9376 + }, + { + "epoch": 0.23, + "learning_rate": 1.806159854363799e-05, + "loss": 1.2269, + "step": 9377 + }, + { + "epoch": 0.23, + "learning_rate": 1.806113764908271e-05, + "loss": 1.3826, + "step": 9378 + }, + { + "epoch": 0.23, + "learning_rate": 1.80606767056224e-05, + "loss": 0.8628, + "step": 9379 + }, + { + "epoch": 0.23, + "learning_rate": 1.8060215713259856e-05, + "loss": 1.1067, + "step": 9380 + }, + { + "epoch": 0.23, + "learning_rate": 1.8059754671997885e-05, + "loss": 1.0575, + "step": 9381 + }, + { + "epoch": 0.23, + "learning_rate": 1.8059293581839277e-05, + "loss": 1.0708, + "step": 9382 + }, + { + "epoch": 0.23, + "learning_rate": 1.805883244278683e-05, + "loss": 1.0415, + "step": 9383 + }, + { + "epoch": 0.23, + "learning_rate": 1.805837125484334e-05, + "loss": 1.3415, + "step": 9384 + }, + { + "epoch": 0.23, + "learning_rate": 1.805791001801161e-05, + "loss": 1.2222, + "step": 9385 + }, + { + "epoch": 0.23, + "learning_rate": 1.8057448732294432e-05, + "loss": 1.1946, + "step": 9386 + }, + { + "epoch": 0.23, + "learning_rate": 1.805698739769461e-05, + "loss": 1.2051, + "step": 9387 + }, + { + "epoch": 0.23, + "learning_rate": 1.8056526014214944e-05, + "loss": 1.246, + "step": 9388 + }, + { + "epoch": 0.23, + "learning_rate": 1.8056064581858224e-05, + "loss": 1.0842, + "step": 9389 + }, + { + "epoch": 0.23, + "learning_rate": 1.805560310062726e-05, + "loss": 0.7921, + "step": 9390 + }, + { + "epoch": 0.23, + "learning_rate": 1.805514157052484e-05, + "loss": 1.1799, + "step": 9391 + }, + { + "epoch": 0.23, + "learning_rate": 1.8054679991553777e-05, + "loss": 1.1203, + "step": 9392 + }, + { + "epoch": 0.23, + "learning_rate": 1.8054218363716862e-05, + "loss": 1.335, + "step": 9393 + }, + { + "epoch": 0.23, + "learning_rate": 1.80537566870169e-05, + "loss": 1.0622, + "step": 9394 + }, + { + "epoch": 0.23, + "learning_rate": 1.805329496145669e-05, + "loss": 1.2427, + "step": 9395 + }, + { + "epoch": 0.23, + "learning_rate": 1.805283318703903e-05, + "loss": 1.2837, + "step": 9396 + }, + { + "epoch": 0.23, + "learning_rate": 1.8052371363766733e-05, + "loss": 1.0389, + "step": 9397 + }, + { + "epoch": 0.23, + "learning_rate": 1.805190949164259e-05, + "loss": 1.1104, + "step": 9398 + }, + { + "epoch": 0.23, + "learning_rate": 1.8051447570669403e-05, + "loss": 1.0786, + "step": 9399 + }, + { + "epoch": 0.23, + "learning_rate": 1.805098560084998e-05, + "loss": 0.8644, + "step": 9400 + }, + { + "epoch": 0.23, + "learning_rate": 1.805052358218712e-05, + "loss": 1.0771, + "step": 9401 + }, + { + "epoch": 0.23, + "learning_rate": 1.8050061514683624e-05, + "loss": 1.1156, + "step": 9402 + }, + { + "epoch": 0.23, + "learning_rate": 1.8049599398342303e-05, + "loss": 1.0581, + "step": 9403 + }, + { + "epoch": 0.23, + "learning_rate": 1.8049137233165955e-05, + "loss": 1.1193, + "step": 9404 + }, + { + "epoch": 0.23, + "learning_rate": 1.8048675019157385e-05, + "loss": 1.0512, + "step": 9405 + }, + { + "epoch": 0.23, + "learning_rate": 1.8048212756319395e-05, + "loss": 1.1778, + "step": 9406 + }, + { + "epoch": 0.23, + "learning_rate": 1.804775044465479e-05, + "loss": 1.0359, + "step": 9407 + }, + { + "epoch": 0.23, + "learning_rate": 1.804728808416638e-05, + "loss": 1.2803, + "step": 9408 + }, + { + "epoch": 0.23, + "learning_rate": 1.8046825674856966e-05, + "loss": 1.1933, + "step": 9409 + }, + { + "epoch": 0.23, + "learning_rate": 1.8046363216729354e-05, + "loss": 0.8392, + "step": 9410 + }, + { + "epoch": 0.23, + "learning_rate": 1.8045900709786345e-05, + "loss": 1.214, + "step": 9411 + }, + { + "epoch": 0.23, + "learning_rate": 1.8045438154030752e-05, + "loss": 1.0719, + "step": 9412 + }, + { + "epoch": 0.23, + "learning_rate": 1.8044975549465376e-05, + "loss": 1.22, + "step": 9413 + }, + { + "epoch": 0.23, + "learning_rate": 1.8044512896093027e-05, + "loss": 0.9207, + "step": 9414 + }, + { + "epoch": 0.23, + "learning_rate": 1.8044050193916506e-05, + "loss": 1.1379, + "step": 9415 + }, + { + "epoch": 0.23, + "learning_rate": 1.8043587442938633e-05, + "loss": 1.2037, + "step": 9416 + }, + { + "epoch": 0.23, + "learning_rate": 1.80431246431622e-05, + "loss": 1.2229, + "step": 9417 + }, + { + "epoch": 0.23, + "learning_rate": 1.8042661794590023e-05, + "loss": 1.3398, + "step": 9418 + }, + { + "epoch": 0.23, + "learning_rate": 1.804219889722491e-05, + "loss": 1.1749, + "step": 9419 + }, + { + "epoch": 0.23, + "learning_rate": 1.804173595106967e-05, + "loss": 1.1743, + "step": 9420 + }, + { + "epoch": 0.23, + "learning_rate": 1.8041272956127104e-05, + "loss": 1.0121, + "step": 9421 + }, + { + "epoch": 0.23, + "learning_rate": 1.804080991240003e-05, + "loss": 1.0215, + "step": 9422 + }, + { + "epoch": 0.23, + "learning_rate": 1.8040346819891254e-05, + "loss": 1.1263, + "step": 9423 + }, + { + "epoch": 0.23, + "learning_rate": 1.8039883678603583e-05, + "loss": 1.1604, + "step": 9424 + }, + { + "epoch": 0.23, + "learning_rate": 1.803942048853983e-05, + "loss": 0.9181, + "step": 9425 + }, + { + "epoch": 0.23, + "learning_rate": 1.8038957249702806e-05, + "loss": 1.3098, + "step": 9426 + }, + { + "epoch": 0.23, + "learning_rate": 1.8038493962095317e-05, + "loss": 1.0833, + "step": 9427 + }, + { + "epoch": 0.23, + "learning_rate": 1.8038030625720173e-05, + "loss": 1.2701, + "step": 9428 + }, + { + "epoch": 0.23, + "learning_rate": 1.8037567240580194e-05, + "loss": 1.2797, + "step": 9429 + }, + { + "epoch": 0.23, + "learning_rate": 1.803710380667818e-05, + "loss": 1.0794, + "step": 9430 + }, + { + "epoch": 0.23, + "learning_rate": 1.803664032401695e-05, + "loss": 1.3706, + "step": 9431 + }, + { + "epoch": 0.23, + "learning_rate": 1.8036176792599313e-05, + "loss": 0.8986, + "step": 9432 + }, + { + "epoch": 0.23, + "learning_rate": 1.803571321242808e-05, + "loss": 1.1059, + "step": 9433 + }, + { + "epoch": 0.23, + "learning_rate": 1.803524958350607e-05, + "loss": 0.916, + "step": 9434 + }, + { + "epoch": 0.23, + "learning_rate": 1.8034785905836086e-05, + "loss": 1.0472, + "step": 9435 + }, + { + "epoch": 0.23, + "learning_rate": 1.8034322179420946e-05, + "loss": 0.9055, + "step": 9436 + }, + { + "epoch": 0.23, + "learning_rate": 1.8033858404263464e-05, + "loss": 0.9674, + "step": 9437 + }, + { + "epoch": 0.23, + "learning_rate": 1.8033394580366453e-05, + "loss": 1.4139, + "step": 9438 + }, + { + "epoch": 0.23, + "learning_rate": 1.8032930707732728e-05, + "loss": 1.2157, + "step": 9439 + }, + { + "epoch": 0.23, + "learning_rate": 1.8032466786365098e-05, + "loss": 0.8474, + "step": 9440 + }, + { + "epoch": 0.23, + "learning_rate": 1.8032002816266385e-05, + "loss": 1.1095, + "step": 9441 + }, + { + "epoch": 0.23, + "learning_rate": 1.80315387974394e-05, + "loss": 1.2534, + "step": 9442 + }, + { + "epoch": 0.23, + "learning_rate": 1.8031074729886955e-05, + "loss": 1.2111, + "step": 9443 + }, + { + "epoch": 0.23, + "learning_rate": 1.8030610613611874e-05, + "loss": 1.2265, + "step": 9444 + }, + { + "epoch": 0.23, + "learning_rate": 1.8030146448616965e-05, + "loss": 0.9855, + "step": 9445 + }, + { + "epoch": 0.23, + "learning_rate": 1.8029682234905044e-05, + "loss": 0.9909, + "step": 9446 + }, + { + "epoch": 0.23, + "learning_rate": 1.8029217972478933e-05, + "loss": 0.948, + "step": 9447 + }, + { + "epoch": 0.23, + "learning_rate": 1.8028753661341442e-05, + "loss": 0.8649, + "step": 9448 + }, + { + "epoch": 0.23, + "learning_rate": 1.8028289301495392e-05, + "loss": 1.3512, + "step": 9449 + }, + { + "epoch": 0.23, + "learning_rate": 1.80278248929436e-05, + "loss": 1.0493, + "step": 9450 + }, + { + "epoch": 0.23, + "learning_rate": 1.8027360435688883e-05, + "loss": 1.2964, + "step": 9451 + }, + { + "epoch": 0.23, + "learning_rate": 1.8026895929734057e-05, + "loss": 0.9716, + "step": 9452 + }, + { + "epoch": 0.23, + "learning_rate": 1.8026431375081943e-05, + "loss": 1.0027, + "step": 9453 + }, + { + "epoch": 0.23, + "learning_rate": 1.8025966771735354e-05, + "loss": 1.1965, + "step": 9454 + }, + { + "epoch": 0.23, + "learning_rate": 1.8025502119697117e-05, + "loss": 1.1142, + "step": 9455 + }, + { + "epoch": 0.23, + "learning_rate": 1.8025037418970047e-05, + "loss": 1.1129, + "step": 9456 + }, + { + "epoch": 0.23, + "learning_rate": 1.802457266955696e-05, + "loss": 1.1482, + "step": 9457 + }, + { + "epoch": 0.23, + "learning_rate": 1.8024107871460678e-05, + "loss": 1.0911, + "step": 9458 + }, + { + "epoch": 0.23, + "learning_rate": 1.802364302468402e-05, + "loss": 1.1517, + "step": 9459 + }, + { + "epoch": 0.23, + "learning_rate": 1.8023178129229808e-05, + "loss": 1.4224, + "step": 9460 + }, + { + "epoch": 0.23, + "learning_rate": 1.802271318510086e-05, + "loss": 1.2852, + "step": 9461 + }, + { + "epoch": 0.23, + "learning_rate": 1.8022248192300002e-05, + "loss": 0.9296, + "step": 9462 + }, + { + "epoch": 0.23, + "learning_rate": 1.8021783150830047e-05, + "loss": 0.9861, + "step": 9463 + }, + { + "epoch": 0.23, + "learning_rate": 1.8021318060693823e-05, + "loss": 1.0764, + "step": 9464 + }, + { + "epoch": 0.23, + "learning_rate": 1.802085292189415e-05, + "loss": 1.1877, + "step": 9465 + }, + { + "epoch": 0.23, + "learning_rate": 1.8020387734433848e-05, + "loss": 1.2315, + "step": 9466 + }, + { + "epoch": 0.23, + "learning_rate": 1.801992249831574e-05, + "loss": 1.0984, + "step": 9467 + }, + { + "epoch": 0.23, + "learning_rate": 1.801945721354265e-05, + "loss": 1.0026, + "step": 9468 + }, + { + "epoch": 0.23, + "learning_rate": 1.8018991880117397e-05, + "loss": 1.0491, + "step": 9469 + }, + { + "epoch": 0.23, + "learning_rate": 1.8018526498042805e-05, + "loss": 1.096, + "step": 9470 + }, + { + "epoch": 0.23, + "learning_rate": 1.80180610673217e-05, + "loss": 0.9105, + "step": 9471 + }, + { + "epoch": 0.23, + "learning_rate": 1.8017595587956907e-05, + "loss": 0.887, + "step": 9472 + }, + { + "epoch": 0.23, + "learning_rate": 1.8017130059951243e-05, + "loss": 1.1285, + "step": 9473 + }, + { + "epoch": 0.23, + "learning_rate": 1.801666448330754e-05, + "loss": 0.9099, + "step": 9474 + }, + { + "epoch": 0.23, + "learning_rate": 1.801619885802862e-05, + "loss": 1.1187, + "step": 9475 + }, + { + "epoch": 0.23, + "learning_rate": 1.8015733184117307e-05, + "loss": 1.452, + "step": 9476 + }, + { + "epoch": 0.23, + "learning_rate": 1.8015267461576425e-05, + "loss": 1.0791, + "step": 9477 + }, + { + "epoch": 0.23, + "learning_rate": 1.8014801690408798e-05, + "loss": 1.1477, + "step": 9478 + }, + { + "epoch": 0.23, + "learning_rate": 1.801433587061726e-05, + "loss": 1.0172, + "step": 9479 + }, + { + "epoch": 0.23, + "learning_rate": 1.8013870002204625e-05, + "loss": 1.0654, + "step": 9480 + }, + { + "epoch": 0.23, + "learning_rate": 1.801340408517373e-05, + "loss": 1.0846, + "step": 9481 + }, + { + "epoch": 0.23, + "learning_rate": 1.8012938119527396e-05, + "loss": 1.1082, + "step": 9482 + }, + { + "epoch": 0.23, + "learning_rate": 1.801247210526845e-05, + "loss": 1.1633, + "step": 9483 + }, + { + "epoch": 0.23, + "learning_rate": 1.8012006042399724e-05, + "loss": 1.0327, + "step": 9484 + }, + { + "epoch": 0.23, + "learning_rate": 1.8011539930924038e-05, + "loss": 1.2052, + "step": 9485 + }, + { + "epoch": 0.23, + "learning_rate": 1.8011073770844225e-05, + "loss": 1.2709, + "step": 9486 + }, + { + "epoch": 0.23, + "learning_rate": 1.8010607562163115e-05, + "loss": 1.0747, + "step": 9487 + }, + { + "epoch": 0.23, + "learning_rate": 1.8010141304883527e-05, + "loss": 1.1721, + "step": 9488 + }, + { + "epoch": 0.23, + "learning_rate": 1.8009674999008304e-05, + "loss": 1.1072, + "step": 9489 + }, + { + "epoch": 0.23, + "learning_rate": 1.800920864454026e-05, + "loss": 1.3344, + "step": 9490 + }, + { + "epoch": 0.23, + "learning_rate": 1.800874224148223e-05, + "loss": 1.216, + "step": 9491 + }, + { + "epoch": 0.23, + "learning_rate": 1.8008275789837047e-05, + "loss": 1.3672, + "step": 9492 + }, + { + "epoch": 0.23, + "learning_rate": 1.800780928960754e-05, + "loss": 1.3243, + "step": 9493 + }, + { + "epoch": 0.23, + "learning_rate": 1.8007342740796538e-05, + "loss": 1.4792, + "step": 9494 + }, + { + "epoch": 0.23, + "learning_rate": 1.800687614340687e-05, + "loss": 1.0275, + "step": 9495 + }, + { + "epoch": 0.23, + "learning_rate": 1.8006409497441364e-05, + "loss": 1.0469, + "step": 9496 + }, + { + "epoch": 0.23, + "learning_rate": 1.800594280290286e-05, + "loss": 0.8156, + "step": 9497 + }, + { + "epoch": 0.23, + "learning_rate": 1.8005476059794183e-05, + "loss": 1.0585, + "step": 9498 + }, + { + "epoch": 0.23, + "learning_rate": 1.800500926811816e-05, + "loss": 1.0903, + "step": 9499 + }, + { + "epoch": 0.23, + "learning_rate": 1.8004542427877635e-05, + "loss": 1.2071, + "step": 9500 + }, + { + "epoch": 0.23, + "learning_rate": 1.8004075539075435e-05, + "loss": 1.3456, + "step": 9501 + }, + { + "epoch": 0.23, + "learning_rate": 1.8003608601714388e-05, + "loss": 1.1415, + "step": 9502 + }, + { + "epoch": 0.23, + "learning_rate": 1.800314161579733e-05, + "loss": 1.0568, + "step": 9503 + }, + { + "epoch": 0.23, + "learning_rate": 1.8002674581327096e-05, + "loss": 1.2882, + "step": 9504 + }, + { + "epoch": 0.23, + "learning_rate": 1.800220749830652e-05, + "loss": 1.1875, + "step": 9505 + }, + { + "epoch": 0.23, + "learning_rate": 1.8001740366738426e-05, + "loss": 1.1128, + "step": 9506 + }, + { + "epoch": 0.23, + "learning_rate": 1.800127318662566e-05, + "loss": 1.0368, + "step": 9507 + }, + { + "epoch": 0.23, + "learning_rate": 1.8000805957971054e-05, + "loss": 1.143, + "step": 9508 + }, + { + "epoch": 0.23, + "learning_rate": 1.8000338680777437e-05, + "loss": 1.3417, + "step": 9509 + }, + { + "epoch": 0.23, + "learning_rate": 1.7999871355047647e-05, + "loss": 1.0094, + "step": 9510 + }, + { + "epoch": 0.23, + "learning_rate": 1.799940398078452e-05, + "loss": 1.2713, + "step": 9511 + }, + { + "epoch": 0.23, + "learning_rate": 1.799893655799089e-05, + "loss": 1.1229, + "step": 9512 + }, + { + "epoch": 0.23, + "learning_rate": 1.7998469086669593e-05, + "loss": 1.2741, + "step": 9513 + }, + { + "epoch": 0.23, + "learning_rate": 1.7998001566823466e-05, + "loss": 1.1529, + "step": 9514 + }, + { + "epoch": 0.23, + "learning_rate": 1.7997533998455342e-05, + "loss": 1.2429, + "step": 9515 + }, + { + "epoch": 0.23, + "learning_rate": 1.7997066381568066e-05, + "loss": 0.8868, + "step": 9516 + }, + { + "epoch": 0.23, + "learning_rate": 1.7996598716164464e-05, + "loss": 1.2475, + "step": 9517 + }, + { + "epoch": 0.23, + "learning_rate": 1.7996131002247382e-05, + "loss": 1.3541, + "step": 9518 + }, + { + "epoch": 0.23, + "learning_rate": 1.7995663239819652e-05, + "loss": 1.2548, + "step": 9519 + }, + { + "epoch": 0.23, + "learning_rate": 1.7995195428884114e-05, + "loss": 1.3485, + "step": 9520 + }, + { + "epoch": 0.23, + "learning_rate": 1.7994727569443607e-05, + "loss": 1.3221, + "step": 9521 + }, + { + "epoch": 0.23, + "learning_rate": 1.7994259661500967e-05, + "loss": 1.0581, + "step": 9522 + }, + { + "epoch": 0.23, + "learning_rate": 1.7993791705059037e-05, + "loss": 0.9394, + "step": 9523 + }, + { + "epoch": 0.23, + "learning_rate": 1.7993323700120648e-05, + "loss": 1.3287, + "step": 9524 + }, + { + "epoch": 0.23, + "learning_rate": 1.799285564668865e-05, + "loss": 1.0755, + "step": 9525 + }, + { + "epoch": 0.23, + "learning_rate": 1.7992387544765874e-05, + "loss": 1.1529, + "step": 9526 + }, + { + "epoch": 0.23, + "learning_rate": 1.799191939435516e-05, + "loss": 1.2192, + "step": 9527 + }, + { + "epoch": 0.23, + "learning_rate": 1.7991451195459356e-05, + "loss": 1.3262, + "step": 9528 + }, + { + "epoch": 0.23, + "learning_rate": 1.7990982948081296e-05, + "loss": 1.2581, + "step": 9529 + }, + { + "epoch": 0.23, + "learning_rate": 1.7990514652223818e-05, + "loss": 1.3175, + "step": 9530 + }, + { + "epoch": 0.23, + "learning_rate": 1.799004630788977e-05, + "loss": 0.9221, + "step": 9531 + }, + { + "epoch": 0.23, + "learning_rate": 1.798957791508199e-05, + "loss": 1.2691, + "step": 9532 + }, + { + "epoch": 0.23, + "learning_rate": 1.798910947380332e-05, + "loss": 1.1283, + "step": 9533 + }, + { + "epoch": 0.23, + "learning_rate": 1.7988640984056602e-05, + "loss": 1.0974, + "step": 9534 + }, + { + "epoch": 0.23, + "learning_rate": 1.798817244584468e-05, + "loss": 1.1536, + "step": 9535 + }, + { + "epoch": 0.23, + "learning_rate": 1.7987703859170393e-05, + "loss": 1.2127, + "step": 9536 + }, + { + "epoch": 0.23, + "learning_rate": 1.7987235224036584e-05, + "loss": 0.8983, + "step": 9537 + }, + { + "epoch": 0.23, + "learning_rate": 1.79867665404461e-05, + "loss": 1.0763, + "step": 9538 + }, + { + "epoch": 0.23, + "learning_rate": 1.7986297808401783e-05, + "loss": 1.1376, + "step": 9539 + }, + { + "epoch": 0.23, + "learning_rate": 1.7985829027906475e-05, + "loss": 1.0721, + "step": 9540 + }, + { + "epoch": 0.23, + "learning_rate": 1.798536019896302e-05, + "loss": 1.1752, + "step": 9541 + }, + { + "epoch": 0.23, + "learning_rate": 1.798489132157426e-05, + "loss": 1.1366, + "step": 9542 + }, + { + "epoch": 0.23, + "learning_rate": 1.7984422395743046e-05, + "loss": 0.999, + "step": 9543 + }, + { + "epoch": 0.23, + "learning_rate": 1.798395342147222e-05, + "loss": 1.0519, + "step": 9544 + }, + { + "epoch": 0.23, + "learning_rate": 1.7983484398764624e-05, + "loss": 1.1821, + "step": 9545 + }, + { + "epoch": 0.23, + "learning_rate": 1.798301532762311e-05, + "loss": 1.0424, + "step": 9546 + }, + { + "epoch": 0.23, + "learning_rate": 1.7982546208050515e-05, + "loss": 1.0095, + "step": 9547 + }, + { + "epoch": 0.23, + "learning_rate": 1.7982077040049692e-05, + "loss": 1.0635, + "step": 9548 + }, + { + "epoch": 0.23, + "learning_rate": 1.798160782362348e-05, + "loss": 1.067, + "step": 9549 + }, + { + "epoch": 0.23, + "learning_rate": 1.7981138558774737e-05, + "loss": 1.3233, + "step": 9550 + }, + { + "epoch": 0.23, + "learning_rate": 1.7980669245506304e-05, + "loss": 1.0837, + "step": 9551 + }, + { + "epoch": 0.23, + "learning_rate": 1.7980199883821026e-05, + "loss": 1.1364, + "step": 9552 + }, + { + "epoch": 0.23, + "learning_rate": 1.797973047372175e-05, + "loss": 1.1584, + "step": 9553 + }, + { + "epoch": 0.23, + "learning_rate": 1.797926101521133e-05, + "loss": 1.19, + "step": 9554 + }, + { + "epoch": 0.23, + "learning_rate": 1.7978791508292607e-05, + "loss": 1.1974, + "step": 9555 + }, + { + "epoch": 0.23, + "learning_rate": 1.7978321952968435e-05, + "loss": 0.9864, + "step": 9556 + }, + { + "epoch": 0.23, + "learning_rate": 1.7977852349241664e-05, + "loss": 1.1841, + "step": 9557 + }, + { + "epoch": 0.23, + "learning_rate": 1.797738269711513e-05, + "loss": 1.1613, + "step": 9558 + }, + { + "epoch": 0.23, + "learning_rate": 1.7976912996591698e-05, + "loss": 1.101, + "step": 9559 + }, + { + "epoch": 0.23, + "learning_rate": 1.797644324767421e-05, + "loss": 0.8185, + "step": 9560 + }, + { + "epoch": 0.23, + "learning_rate": 1.7975973450365515e-05, + "loss": 1.1199, + "step": 9561 + }, + { + "epoch": 0.23, + "learning_rate": 1.7975503604668468e-05, + "loss": 1.1907, + "step": 9562 + }, + { + "epoch": 0.23, + "learning_rate": 1.7975033710585914e-05, + "loss": 1.0056, + "step": 9563 + }, + { + "epoch": 0.23, + "learning_rate": 1.797456376812071e-05, + "loss": 1.1959, + "step": 9564 + }, + { + "epoch": 0.23, + "learning_rate": 1.79740937772757e-05, + "loss": 1.0677, + "step": 9565 + }, + { + "epoch": 0.23, + "learning_rate": 1.797362373805374e-05, + "loss": 1.1981, + "step": 9566 + }, + { + "epoch": 0.23, + "learning_rate": 1.797315365045768e-05, + "loss": 1.1969, + "step": 9567 + }, + { + "epoch": 0.23, + "learning_rate": 1.7972683514490372e-05, + "loss": 0.8398, + "step": 9568 + }, + { + "epoch": 0.23, + "learning_rate": 1.797221333015467e-05, + "loss": 1.2052, + "step": 9569 + }, + { + "epoch": 0.23, + "learning_rate": 1.797174309745342e-05, + "loss": 1.1798, + "step": 9570 + }, + { + "epoch": 0.23, + "learning_rate": 1.7971272816389486e-05, + "loss": 1.2396, + "step": 9571 + }, + { + "epoch": 0.23, + "learning_rate": 1.7970802486965712e-05, + "loss": 0.7173, + "step": 9572 + }, + { + "epoch": 0.23, + "learning_rate": 1.7970332109184953e-05, + "loss": 1.1449, + "step": 9573 + }, + { + "epoch": 0.23, + "learning_rate": 1.7969861683050064e-05, + "loss": 1.4264, + "step": 9574 + }, + { + "epoch": 0.23, + "learning_rate": 1.79693912085639e-05, + "loss": 1.0661, + "step": 9575 + }, + { + "epoch": 0.23, + "learning_rate": 1.7968920685729317e-05, + "loss": 1.0457, + "step": 9576 + }, + { + "epoch": 0.23, + "learning_rate": 1.796845011454916e-05, + "loss": 0.8882, + "step": 9577 + }, + { + "epoch": 0.23, + "learning_rate": 1.7967979495026296e-05, + "loss": 1.2502, + "step": 9578 + }, + { + "epoch": 0.23, + "learning_rate": 1.7967508827163573e-05, + "loss": 1.158, + "step": 9579 + }, + { + "epoch": 0.23, + "learning_rate": 1.7967038110963847e-05, + "loss": 1.3594, + "step": 9580 + }, + { + "epoch": 0.23, + "learning_rate": 1.7966567346429975e-05, + "loss": 1.1713, + "step": 9581 + }, + { + "epoch": 0.23, + "learning_rate": 1.7966096533564813e-05, + "loss": 1.3267, + "step": 9582 + }, + { + "epoch": 0.23, + "learning_rate": 1.7965625672371218e-05, + "loss": 0.9803, + "step": 9583 + }, + { + "epoch": 0.23, + "learning_rate": 1.7965154762852044e-05, + "loss": 1.1674, + "step": 9584 + }, + { + "epoch": 0.23, + "learning_rate": 1.796468380501015e-05, + "loss": 0.9467, + "step": 9585 + }, + { + "epoch": 0.23, + "learning_rate": 1.7964212798848393e-05, + "loss": 1.126, + "step": 9586 + }, + { + "epoch": 0.23, + "learning_rate": 1.7963741744369628e-05, + "loss": 1.2261, + "step": 9587 + }, + { + "epoch": 0.23, + "learning_rate": 1.796327064157672e-05, + "loss": 0.9586, + "step": 9588 + }, + { + "epoch": 0.23, + "learning_rate": 1.7962799490472518e-05, + "loss": 1.0571, + "step": 9589 + }, + { + "epoch": 0.23, + "learning_rate": 1.7962328291059886e-05, + "loss": 1.2281, + "step": 9590 + }, + { + "epoch": 0.23, + "learning_rate": 1.796185704334168e-05, + "loss": 0.9227, + "step": 9591 + }, + { + "epoch": 0.23, + "learning_rate": 1.7961385747320763e-05, + "loss": 0.9596, + "step": 9592 + }, + { + "epoch": 0.23, + "learning_rate": 1.7960914402999988e-05, + "loss": 1.1488, + "step": 9593 + }, + { + "epoch": 0.23, + "learning_rate": 1.796044301038222e-05, + "loss": 1.1811, + "step": 9594 + }, + { + "epoch": 0.23, + "learning_rate": 1.7959971569470318e-05, + "loss": 1.1185, + "step": 9595 + }, + { + "epoch": 0.23, + "learning_rate": 1.795950008026714e-05, + "loss": 0.8604, + "step": 9596 + }, + { + "epoch": 0.23, + "learning_rate": 1.7959028542775543e-05, + "loss": 1.3264, + "step": 9597 + }, + { + "epoch": 0.23, + "learning_rate": 1.7958556956998397e-05, + "loss": 1.0394, + "step": 9598 + }, + { + "epoch": 0.23, + "learning_rate": 1.7958085322938554e-05, + "loss": 1.0885, + "step": 9599 + }, + { + "epoch": 0.23, + "learning_rate": 1.795761364059888e-05, + "loss": 0.7371, + "step": 9600 + }, + { + "epoch": 0.23, + "learning_rate": 1.7957141909982237e-05, + "loss": 1.0283, + "step": 9601 + }, + { + "epoch": 0.23, + "learning_rate": 1.7956670131091486e-05, + "loss": 0.9005, + "step": 9602 + }, + { + "epoch": 0.23, + "learning_rate": 1.7956198303929486e-05, + "loss": 1.1923, + "step": 9603 + }, + { + "epoch": 0.23, + "learning_rate": 1.7955726428499107e-05, + "loss": 1.0525, + "step": 9604 + }, + { + "epoch": 0.23, + "learning_rate": 1.7955254504803205e-05, + "loss": 1.2892, + "step": 9605 + }, + { + "epoch": 0.23, + "learning_rate": 1.7954782532844643e-05, + "loss": 1.2068, + "step": 9606 + }, + { + "epoch": 0.23, + "learning_rate": 1.7954310512626287e-05, + "loss": 1.1887, + "step": 9607 + }, + { + "epoch": 0.23, + "learning_rate": 1.7953838444151004e-05, + "loss": 1.4203, + "step": 9608 + }, + { + "epoch": 0.23, + "learning_rate": 1.795336632742165e-05, + "loss": 1.0353, + "step": 9609 + }, + { + "epoch": 0.23, + "learning_rate": 1.7952894162441094e-05, + "loss": 1.3073, + "step": 9610 + }, + { + "epoch": 0.23, + "learning_rate": 1.79524219492122e-05, + "loss": 1.074, + "step": 9611 + }, + { + "epoch": 0.23, + "learning_rate": 1.795194968773783e-05, + "loss": 1.0461, + "step": 9612 + }, + { + "epoch": 0.23, + "learning_rate": 1.795147737802086e-05, + "loss": 1.2114, + "step": 9613 + }, + { + "epoch": 0.23, + "learning_rate": 1.7951005020064142e-05, + "loss": 1.2565, + "step": 9614 + }, + { + "epoch": 0.23, + "learning_rate": 1.795053261387054e-05, + "loss": 0.945, + "step": 9615 + }, + { + "epoch": 0.23, + "learning_rate": 1.7950060159442934e-05, + "loss": 1.4167, + "step": 9616 + }, + { + "epoch": 0.23, + "learning_rate": 1.7949587656784183e-05, + "loss": 1.1143, + "step": 9617 + }, + { + "epoch": 0.23, + "learning_rate": 1.7949115105897155e-05, + "loss": 1.283, + "step": 9618 + }, + { + "epoch": 0.23, + "learning_rate": 1.794864250678471e-05, + "loss": 1.4981, + "step": 9619 + }, + { + "epoch": 0.23, + "learning_rate": 1.7948169859449726e-05, + "loss": 1.0793, + "step": 9620 + }, + { + "epoch": 0.23, + "learning_rate": 1.794769716389506e-05, + "loss": 1.1949, + "step": 9621 + }, + { + "epoch": 0.23, + "learning_rate": 1.7947224420123587e-05, + "loss": 1.2645, + "step": 9622 + }, + { + "epoch": 0.23, + "learning_rate": 1.7946751628138175e-05, + "loss": 1.434, + "step": 9623 + }, + { + "epoch": 0.23, + "learning_rate": 1.7946278787941687e-05, + "loss": 1.0046, + "step": 9624 + }, + { + "epoch": 0.23, + "learning_rate": 1.7945805899536997e-05, + "loss": 1.0567, + "step": 9625 + }, + { + "epoch": 0.23, + "learning_rate": 1.7945332962926966e-05, + "loss": 0.9904, + "step": 9626 + }, + { + "epoch": 0.23, + "learning_rate": 1.7944859978114473e-05, + "loss": 1.0753, + "step": 9627 + }, + { + "epoch": 0.23, + "learning_rate": 1.7944386945102387e-05, + "loss": 1.0028, + "step": 9628 + }, + { + "epoch": 0.23, + "learning_rate": 1.7943913863893567e-05, + "loss": 1.2206, + "step": 9629 + }, + { + "epoch": 0.23, + "learning_rate": 1.7943440734490893e-05, + "loss": 1.5803, + "step": 9630 + }, + { + "epoch": 0.23, + "learning_rate": 1.7942967556897235e-05, + "loss": 1.0209, + "step": 9631 + }, + { + "epoch": 0.23, + "learning_rate": 1.794249433111546e-05, + "loss": 1.4433, + "step": 9632 + }, + { + "epoch": 0.23, + "learning_rate": 1.7942021057148438e-05, + "loss": 1.1838, + "step": 9633 + }, + { + "epoch": 0.23, + "learning_rate": 1.7941547734999043e-05, + "loss": 1.3533, + "step": 9634 + }, + { + "epoch": 0.23, + "learning_rate": 1.7941074364670146e-05, + "loss": 1.0245, + "step": 9635 + }, + { + "epoch": 0.23, + "learning_rate": 1.794060094616462e-05, + "loss": 0.9937, + "step": 9636 + }, + { + "epoch": 0.23, + "learning_rate": 1.7940127479485333e-05, + "loss": 1.1608, + "step": 9637 + }, + { + "epoch": 0.23, + "learning_rate": 1.7939653964635163e-05, + "loss": 0.6019, + "step": 9638 + }, + { + "epoch": 0.23, + "learning_rate": 1.7939180401616977e-05, + "loss": 1.2685, + "step": 9639 + }, + { + "epoch": 0.23, + "learning_rate": 1.7938706790433655e-05, + "loss": 1.1949, + "step": 9640 + }, + { + "epoch": 0.23, + "learning_rate": 1.7938233131088063e-05, + "loss": 1.2659, + "step": 9641 + }, + { + "epoch": 0.23, + "learning_rate": 1.793775942358308e-05, + "loss": 1.2578, + "step": 9642 + }, + { + "epoch": 0.23, + "learning_rate": 1.7937285667921577e-05, + "loss": 1.0785, + "step": 9643 + }, + { + "epoch": 0.23, + "learning_rate": 1.7936811864106425e-05, + "loss": 1.3398, + "step": 9644 + }, + { + "epoch": 0.23, + "learning_rate": 1.7936338012140507e-05, + "loss": 1.1041, + "step": 9645 + }, + { + "epoch": 0.23, + "learning_rate": 1.793586411202669e-05, + "loss": 0.8807, + "step": 9646 + }, + { + "epoch": 0.23, + "learning_rate": 1.7935390163767854e-05, + "loss": 1.1755, + "step": 9647 + }, + { + "epoch": 0.23, + "learning_rate": 1.793491616736687e-05, + "loss": 1.1256, + "step": 9648 + }, + { + "epoch": 0.23, + "learning_rate": 1.793444212282662e-05, + "loss": 1.0076, + "step": 9649 + }, + { + "epoch": 0.23, + "learning_rate": 1.7933968030149972e-05, + "loss": 0.965, + "step": 9650 + }, + { + "epoch": 0.23, + "learning_rate": 1.793349388933981e-05, + "loss": 1.0368, + "step": 9651 + }, + { + "epoch": 0.23, + "learning_rate": 1.7933019700399006e-05, + "loss": 1.0236, + "step": 9652 + }, + { + "epoch": 0.23, + "learning_rate": 1.7932545463330435e-05, + "loss": 1.42, + "step": 9653 + }, + { + "epoch": 0.23, + "learning_rate": 1.793207117813698e-05, + "loss": 1.0481, + "step": 9654 + }, + { + "epoch": 0.23, + "learning_rate": 1.793159684482151e-05, + "loss": 1.0818, + "step": 9655 + }, + { + "epoch": 0.23, + "learning_rate": 1.793112246338691e-05, + "loss": 1.1146, + "step": 9656 + }, + { + "epoch": 0.23, + "learning_rate": 1.7930648033836055e-05, + "loss": 1.1198, + "step": 9657 + }, + { + "epoch": 0.23, + "learning_rate": 1.7930173556171824e-05, + "loss": 1.3419, + "step": 9658 + }, + { + "epoch": 0.23, + "learning_rate": 1.7929699030397095e-05, + "loss": 1.3019, + "step": 9659 + }, + { + "epoch": 0.23, + "learning_rate": 1.792922445651475e-05, + "loss": 0.9782, + "step": 9660 + }, + { + "epoch": 0.23, + "learning_rate": 1.7928749834527662e-05, + "loss": 1.1203, + "step": 9661 + }, + { + "epoch": 0.23, + "learning_rate": 1.7928275164438715e-05, + "loss": 1.0605, + "step": 9662 + }, + { + "epoch": 0.23, + "learning_rate": 1.792780044625079e-05, + "loss": 0.9188, + "step": 9663 + }, + { + "epoch": 0.23, + "learning_rate": 1.792732567996676e-05, + "loss": 1.3265, + "step": 9664 + }, + { + "epoch": 0.23, + "learning_rate": 1.792685086558951e-05, + "loss": 0.9662, + "step": 9665 + }, + { + "epoch": 0.23, + "learning_rate": 1.7926376003121922e-05, + "loss": 1.137, + "step": 9666 + }, + { + "epoch": 0.23, + "learning_rate": 1.7925901092566874e-05, + "loss": 1.3383, + "step": 9667 + }, + { + "epoch": 0.23, + "learning_rate": 1.792542613392725e-05, + "loss": 0.9683, + "step": 9668 + }, + { + "epoch": 0.23, + "learning_rate": 1.7924951127205932e-05, + "loss": 1.0326, + "step": 9669 + }, + { + "epoch": 0.23, + "learning_rate": 1.7924476072405795e-05, + "loss": 1.335, + "step": 9670 + }, + { + "epoch": 0.23, + "learning_rate": 1.7924000969529727e-05, + "loss": 1.343, + "step": 9671 + }, + { + "epoch": 0.23, + "learning_rate": 1.792352581858061e-05, + "loss": 1.2886, + "step": 9672 + }, + { + "epoch": 0.23, + "learning_rate": 1.7923050619561325e-05, + "loss": 1.1714, + "step": 9673 + }, + { + "epoch": 0.23, + "learning_rate": 1.7922575372474755e-05, + "loss": 1.337, + "step": 9674 + }, + { + "epoch": 0.23, + "learning_rate": 1.7922100077323783e-05, + "loss": 1.1423, + "step": 9675 + }, + { + "epoch": 0.23, + "learning_rate": 1.7921624734111292e-05, + "loss": 1.0463, + "step": 9676 + }, + { + "epoch": 0.23, + "learning_rate": 1.792114934284017e-05, + "loss": 1.0561, + "step": 9677 + }, + { + "epoch": 0.23, + "learning_rate": 1.79206739035133e-05, + "loss": 1.2074, + "step": 9678 + }, + { + "epoch": 0.23, + "learning_rate": 1.7920198416133565e-05, + "loss": 1.1533, + "step": 9679 + }, + { + "epoch": 0.23, + "learning_rate": 1.7919722880703843e-05, + "loss": 1.404, + "step": 9680 + }, + { + "epoch": 0.23, + "learning_rate": 1.7919247297227027e-05, + "loss": 1.0674, + "step": 9681 + }, + { + "epoch": 0.23, + "learning_rate": 1.7918771665706e-05, + "loss": 0.8141, + "step": 9682 + }, + { + "epoch": 0.23, + "learning_rate": 1.7918295986143652e-05, + "loss": 1.2163, + "step": 9683 + }, + { + "epoch": 0.23, + "learning_rate": 1.7917820258542863e-05, + "loss": 1.1019, + "step": 9684 + }, + { + "epoch": 0.23, + "learning_rate": 1.7917344482906516e-05, + "loss": 1.098, + "step": 9685 + }, + { + "epoch": 0.23, + "learning_rate": 1.791686865923751e-05, + "loss": 1.1905, + "step": 9686 + }, + { + "epoch": 0.23, + "learning_rate": 1.791639278753872e-05, + "loss": 0.8812, + "step": 9687 + }, + { + "epoch": 0.23, + "learning_rate": 1.7915916867813037e-05, + "loss": 1.2535, + "step": 9688 + }, + { + "epoch": 0.23, + "learning_rate": 1.791544090006335e-05, + "loss": 1.1122, + "step": 9689 + }, + { + "epoch": 0.23, + "learning_rate": 1.7914964884292543e-05, + "loss": 1.2864, + "step": 9690 + }, + { + "epoch": 0.23, + "learning_rate": 1.7914488820503506e-05, + "loss": 1.3958, + "step": 9691 + }, + { + "epoch": 0.23, + "learning_rate": 1.7914012708699126e-05, + "loss": 1.0204, + "step": 9692 + }, + { + "epoch": 0.23, + "learning_rate": 1.7913536548882295e-05, + "loss": 1.1147, + "step": 9693 + }, + { + "epoch": 0.23, + "learning_rate": 1.7913060341055895e-05, + "loss": 1.3758, + "step": 9694 + }, + { + "epoch": 0.23, + "learning_rate": 1.7912584085222826e-05, + "loss": 1.2043, + "step": 9695 + }, + { + "epoch": 0.23, + "learning_rate": 1.7912107781385963e-05, + "loss": 1.1524, + "step": 9696 + }, + { + "epoch": 0.23, + "learning_rate": 1.791163142954821e-05, + "loss": 0.9341, + "step": 9697 + }, + { + "epoch": 0.23, + "learning_rate": 1.7911155029712444e-05, + "loss": 1.0356, + "step": 9698 + }, + { + "epoch": 0.23, + "learning_rate": 1.7910678581881566e-05, + "loss": 1.2729, + "step": 9699 + }, + { + "epoch": 0.23, + "learning_rate": 1.7910202086058458e-05, + "loss": 1.1715, + "step": 9700 + }, + { + "epoch": 0.23, + "learning_rate": 1.7909725542246018e-05, + "loss": 1.1119, + "step": 9701 + }, + { + "epoch": 0.23, + "learning_rate": 1.790924895044713e-05, + "loss": 0.9837, + "step": 9702 + }, + { + "epoch": 0.23, + "learning_rate": 1.7908772310664692e-05, + "loss": 1.0232, + "step": 9703 + }, + { + "epoch": 0.23, + "learning_rate": 1.790829562290159e-05, + "loss": 1.1114, + "step": 9704 + }, + { + "epoch": 0.23, + "learning_rate": 1.790781888716072e-05, + "loss": 1.1605, + "step": 9705 + }, + { + "epoch": 0.23, + "learning_rate": 1.7907342103444975e-05, + "loss": 1.2916, + "step": 9706 + }, + { + "epoch": 0.23, + "learning_rate": 1.790686527175724e-05, + "loss": 1.3559, + "step": 9707 + }, + { + "epoch": 0.23, + "learning_rate": 1.790638839210042e-05, + "loss": 1.0393, + "step": 9708 + }, + { + "epoch": 0.23, + "learning_rate": 1.7905911464477395e-05, + "loss": 1.2797, + "step": 9709 + }, + { + "epoch": 0.23, + "learning_rate": 1.790543448889107e-05, + "loss": 1.1321, + "step": 9710 + }, + { + "epoch": 0.23, + "learning_rate": 1.790495746534433e-05, + "loss": 1.2346, + "step": 9711 + }, + { + "epoch": 0.23, + "learning_rate": 1.7904480393840074e-05, + "loss": 1.1263, + "step": 9712 + }, + { + "epoch": 0.23, + "learning_rate": 1.7904003274381193e-05, + "loss": 1.1575, + "step": 9713 + }, + { + "epoch": 0.23, + "learning_rate": 1.7903526106970585e-05, + "loss": 1.126, + "step": 9714 + }, + { + "epoch": 0.23, + "learning_rate": 1.7903048891611142e-05, + "loss": 1.1717, + "step": 9715 + }, + { + "epoch": 0.23, + "learning_rate": 1.790257162830576e-05, + "loss": 0.9258, + "step": 9716 + }, + { + "epoch": 0.23, + "learning_rate": 1.7902094317057334e-05, + "loss": 1.24, + "step": 9717 + }, + { + "epoch": 0.23, + "learning_rate": 1.7901616957868766e-05, + "loss": 1.0855, + "step": 9718 + }, + { + "epoch": 0.23, + "learning_rate": 1.790113955074294e-05, + "loss": 1.3424, + "step": 9719 + }, + { + "epoch": 0.23, + "learning_rate": 1.7900662095682762e-05, + "loss": 0.9496, + "step": 9720 + }, + { + "epoch": 0.23, + "learning_rate": 1.7900184592691127e-05, + "loss": 1.0028, + "step": 9721 + }, + { + "epoch": 0.23, + "learning_rate": 1.7899707041770925e-05, + "loss": 1.1645, + "step": 9722 + }, + { + "epoch": 0.23, + "learning_rate": 1.7899229442925064e-05, + "loss": 1.0425, + "step": 9723 + }, + { + "epoch": 0.23, + "learning_rate": 1.7898751796156433e-05, + "loss": 1.0798, + "step": 9724 + }, + { + "epoch": 0.23, + "learning_rate": 1.7898274101467933e-05, + "loss": 1.1788, + "step": 9725 + }, + { + "epoch": 0.23, + "learning_rate": 1.789779635886246e-05, + "loss": 1.3186, + "step": 9726 + }, + { + "epoch": 0.23, + "learning_rate": 1.7897318568342917e-05, + "loss": 0.9813, + "step": 9727 + }, + { + "epoch": 0.23, + "learning_rate": 1.7896840729912198e-05, + "loss": 0.9718, + "step": 9728 + }, + { + "epoch": 0.23, + "learning_rate": 1.7896362843573204e-05, + "loss": 0.848, + "step": 9729 + }, + { + "epoch": 0.23, + "learning_rate": 1.7895884909328835e-05, + "loss": 0.9783, + "step": 9730 + }, + { + "epoch": 0.23, + "learning_rate": 1.789540692718199e-05, + "loss": 1.1555, + "step": 9731 + }, + { + "epoch": 0.23, + "learning_rate": 1.789492889713557e-05, + "loss": 1.0111, + "step": 9732 + }, + { + "epoch": 0.23, + "learning_rate": 1.7894450819192468e-05, + "loss": 1.1257, + "step": 9733 + }, + { + "epoch": 0.23, + "learning_rate": 1.7893972693355595e-05, + "loss": 0.9823, + "step": 9734 + }, + { + "epoch": 0.23, + "learning_rate": 1.7893494519627843e-05, + "loss": 1.2982, + "step": 9735 + }, + { + "epoch": 0.23, + "learning_rate": 1.7893016298012117e-05, + "loss": 1.1176, + "step": 9736 + }, + { + "epoch": 0.23, + "learning_rate": 1.7892538028511318e-05, + "loss": 1.1878, + "step": 9737 + }, + { + "epoch": 0.23, + "learning_rate": 1.7892059711128346e-05, + "loss": 1.1254, + "step": 9738 + }, + { + "epoch": 0.23, + "learning_rate": 1.789158134586611e-05, + "loss": 1.1152, + "step": 9739 + }, + { + "epoch": 0.23, + "learning_rate": 1.78911029327275e-05, + "loss": 1.2402, + "step": 9740 + }, + { + "epoch": 0.23, + "learning_rate": 1.7890624471715427e-05, + "loss": 1.4151, + "step": 9741 + }, + { + "epoch": 0.23, + "learning_rate": 1.7890145962832786e-05, + "loss": 1.1983, + "step": 9742 + }, + { + "epoch": 0.23, + "learning_rate": 1.788966740608249e-05, + "loss": 1.0872, + "step": 9743 + }, + { + "epoch": 0.23, + "learning_rate": 1.788918880146744e-05, + "loss": 0.9131, + "step": 9744 + }, + { + "epoch": 0.23, + "learning_rate": 1.7888710148990535e-05, + "loss": 1.0833, + "step": 9745 + }, + { + "epoch": 0.23, + "learning_rate": 1.788823144865468e-05, + "loss": 1.187, + "step": 9746 + }, + { + "epoch": 0.23, + "learning_rate": 1.7887752700462783e-05, + "loss": 1.0291, + "step": 9747 + }, + { + "epoch": 0.23, + "learning_rate": 1.7887273904417742e-05, + "loss": 1.0447, + "step": 9748 + }, + { + "epoch": 0.23, + "learning_rate": 1.7886795060522465e-05, + "loss": 1.215, + "step": 9749 + }, + { + "epoch": 0.23, + "learning_rate": 1.7886316168779862e-05, + "loss": 1.1787, + "step": 9750 + }, + { + "epoch": 0.23, + "learning_rate": 1.788583722919283e-05, + "loss": 1.0614, + "step": 9751 + }, + { + "epoch": 0.23, + "learning_rate": 1.7885358241764282e-05, + "loss": 1.2223, + "step": 9752 + }, + { + "epoch": 0.23, + "learning_rate": 1.7884879206497117e-05, + "loss": 1.0652, + "step": 9753 + }, + { + "epoch": 0.23, + "learning_rate": 1.7884400123394243e-05, + "loss": 1.0326, + "step": 9754 + }, + { + "epoch": 0.23, + "learning_rate": 1.788392099245857e-05, + "loss": 1.1715, + "step": 9755 + }, + { + "epoch": 0.23, + "learning_rate": 1.7883441813693006e-05, + "loss": 1.0548, + "step": 9756 + }, + { + "epoch": 0.23, + "learning_rate": 1.788296258710045e-05, + "loss": 1.0188, + "step": 9757 + }, + { + "epoch": 0.23, + "learning_rate": 1.7882483312683816e-05, + "loss": 1.1668, + "step": 9758 + }, + { + "epoch": 0.23, + "learning_rate": 1.788200399044601e-05, + "loss": 1.0866, + "step": 9759 + }, + { + "epoch": 0.23, + "learning_rate": 1.788152462038994e-05, + "loss": 1.1111, + "step": 9760 + }, + { + "epoch": 0.23, + "learning_rate": 1.7881045202518514e-05, + "loss": 1.0748, + "step": 9761 + }, + { + "epoch": 0.23, + "learning_rate": 1.7880565736834642e-05, + "loss": 1.4641, + "step": 9762 + }, + { + "epoch": 0.23, + "learning_rate": 1.788008622334123e-05, + "loss": 1.1938, + "step": 9763 + }, + { + "epoch": 0.23, + "learning_rate": 1.7879606662041186e-05, + "loss": 1.0061, + "step": 9764 + }, + { + "epoch": 0.23, + "learning_rate": 1.7879127052937425e-05, + "loss": 1.042, + "step": 9765 + }, + { + "epoch": 0.23, + "learning_rate": 1.787864739603285e-05, + "loss": 1.0384, + "step": 9766 + }, + { + "epoch": 0.23, + "learning_rate": 1.7878167691330376e-05, + "loss": 1.2603, + "step": 9767 + }, + { + "epoch": 0.23, + "learning_rate": 1.7877687938832915e-05, + "loss": 1.0258, + "step": 9768 + }, + { + "epoch": 0.23, + "learning_rate": 1.787720813854337e-05, + "loss": 1.0592, + "step": 9769 + }, + { + "epoch": 0.23, + "learning_rate": 1.7876728290464658e-05, + "loss": 1.1053, + "step": 9770 + }, + { + "epoch": 0.23, + "learning_rate": 1.787624839459969e-05, + "loss": 1.1005, + "step": 9771 + }, + { + "epoch": 0.24, + "learning_rate": 1.787576845095137e-05, + "loss": 1.1808, + "step": 9772 + }, + { + "epoch": 0.24, + "learning_rate": 1.7875288459522616e-05, + "loss": 1.2682, + "step": 9773 + }, + { + "epoch": 0.24, + "learning_rate": 1.7874808420316345e-05, + "loss": 1.3998, + "step": 9774 + }, + { + "epoch": 0.24, + "learning_rate": 1.787432833333546e-05, + "loss": 1.1639, + "step": 9775 + }, + { + "epoch": 0.24, + "learning_rate": 1.787384819858288e-05, + "loss": 1.1397, + "step": 9776 + }, + { + "epoch": 0.24, + "learning_rate": 1.7873368016061513e-05, + "loss": 1.3849, + "step": 9777 + }, + { + "epoch": 0.24, + "learning_rate": 1.787288778577427e-05, + "loss": 1.0127, + "step": 9778 + }, + { + "epoch": 0.24, + "learning_rate": 1.7872407507724072e-05, + "loss": 1.0201, + "step": 9779 + }, + { + "epoch": 0.24, + "learning_rate": 1.7871927181913832e-05, + "loss": 1.1408, + "step": 9780 + }, + { + "epoch": 0.24, + "learning_rate": 1.7871446808346458e-05, + "loss": 0.7853, + "step": 9781 + }, + { + "epoch": 0.24, + "learning_rate": 1.787096638702487e-05, + "loss": 1.1436, + "step": 9782 + }, + { + "epoch": 0.24, + "learning_rate": 1.7870485917951977e-05, + "loss": 1.2178, + "step": 9783 + }, + { + "epoch": 0.24, + "learning_rate": 1.78700054011307e-05, + "loss": 0.9038, + "step": 9784 + }, + { + "epoch": 0.24, + "learning_rate": 1.7869524836563953e-05, + "loss": 1.1343, + "step": 9785 + }, + { + "epoch": 0.24, + "learning_rate": 1.7869044224254648e-05, + "loss": 1.0861, + "step": 9786 + }, + { + "epoch": 0.24, + "learning_rate": 1.78685635642057e-05, + "loss": 0.9844, + "step": 9787 + }, + { + "epoch": 0.24, + "learning_rate": 1.786808285642003e-05, + "loss": 1.1042, + "step": 9788 + }, + { + "epoch": 0.24, + "learning_rate": 1.786760210090055e-05, + "loss": 1.1415, + "step": 9789 + }, + { + "epoch": 0.24, + "learning_rate": 1.7867121297650184e-05, + "loss": 1.1235, + "step": 9790 + }, + { + "epoch": 0.24, + "learning_rate": 1.786664044667184e-05, + "loss": 1.1364, + "step": 9791 + }, + { + "epoch": 0.24, + "learning_rate": 1.786615954796844e-05, + "loss": 1.1494, + "step": 9792 + }, + { + "epoch": 0.24, + "learning_rate": 1.7865678601542894e-05, + "loss": 1.0564, + "step": 9793 + }, + { + "epoch": 0.24, + "learning_rate": 1.7865197607398133e-05, + "loss": 0.9802, + "step": 9794 + }, + { + "epoch": 0.24, + "learning_rate": 1.7864716565537067e-05, + "loss": 1.0721, + "step": 9795 + }, + { + "epoch": 0.24, + "learning_rate": 1.7864235475962616e-05, + "loss": 1.0999, + "step": 9796 + }, + { + "epoch": 0.24, + "learning_rate": 1.7863754338677695e-05, + "loss": 0.9392, + "step": 9797 + }, + { + "epoch": 0.24, + "learning_rate": 1.786327315368523e-05, + "loss": 0.9086, + "step": 9798 + }, + { + "epoch": 0.24, + "learning_rate": 1.7862791920988132e-05, + "loss": 1.1778, + "step": 9799 + }, + { + "epoch": 0.24, + "learning_rate": 1.7862310640589328e-05, + "loss": 0.9707, + "step": 9800 + }, + { + "epoch": 0.24, + "learning_rate": 1.7861829312491733e-05, + "loss": 1.3141, + "step": 9801 + }, + { + "epoch": 0.24, + "learning_rate": 1.786134793669827e-05, + "loss": 1.2648, + "step": 9802 + }, + { + "epoch": 0.24, + "learning_rate": 1.7860866513211856e-05, + "loss": 1.2332, + "step": 9803 + }, + { + "epoch": 0.24, + "learning_rate": 1.7860385042035418e-05, + "loss": 0.9294, + "step": 9804 + }, + { + "epoch": 0.24, + "learning_rate": 1.785990352317187e-05, + "loss": 0.9626, + "step": 9805 + }, + { + "epoch": 0.24, + "learning_rate": 1.7859421956624135e-05, + "loss": 1.1131, + "step": 9806 + }, + { + "epoch": 0.24, + "learning_rate": 1.7858940342395138e-05, + "loss": 1.466, + "step": 9807 + }, + { + "epoch": 0.24, + "learning_rate": 1.7858458680487798e-05, + "loss": 1.3308, + "step": 9808 + }, + { + "epoch": 0.24, + "learning_rate": 1.7857976970905036e-05, + "loss": 1.1243, + "step": 9809 + }, + { + "epoch": 0.24, + "learning_rate": 1.785749521364978e-05, + "loss": 1.2685, + "step": 9810 + }, + { + "epoch": 0.24, + "learning_rate": 1.7857013408724945e-05, + "loss": 1.1024, + "step": 9811 + }, + { + "epoch": 0.24, + "learning_rate": 1.7856531556133457e-05, + "loss": 1.2064, + "step": 9812 + }, + { + "epoch": 0.24, + "learning_rate": 1.7856049655878242e-05, + "loss": 1.2715, + "step": 9813 + }, + { + "epoch": 0.24, + "learning_rate": 1.785556770796222e-05, + "loss": 1.0812, + "step": 9814 + }, + { + "epoch": 0.24, + "learning_rate": 1.785508571238832e-05, + "loss": 1.2257, + "step": 9815 + }, + { + "epoch": 0.24, + "learning_rate": 1.785460366915946e-05, + "loss": 0.8155, + "step": 9816 + }, + { + "epoch": 0.24, + "learning_rate": 1.7854121578278568e-05, + "loss": 1.3237, + "step": 9817 + }, + { + "epoch": 0.24, + "learning_rate": 1.7853639439748564e-05, + "loss": 1.3191, + "step": 9818 + }, + { + "epoch": 0.24, + "learning_rate": 1.785315725357238e-05, + "loss": 1.1158, + "step": 9819 + }, + { + "epoch": 0.24, + "learning_rate": 1.785267501975294e-05, + "loss": 1.0594, + "step": 9820 + }, + { + "epoch": 0.24, + "learning_rate": 1.7852192738293165e-05, + "loss": 1.1078, + "step": 9821 + }, + { + "epoch": 0.24, + "learning_rate": 1.7851710409195987e-05, + "loss": 1.2818, + "step": 9822 + }, + { + "epoch": 0.24, + "learning_rate": 1.7851228032464326e-05, + "loss": 1.3649, + "step": 9823 + }, + { + "epoch": 0.24, + "learning_rate": 1.785074560810111e-05, + "loss": 1.13, + "step": 9824 + }, + { + "epoch": 0.24, + "learning_rate": 1.785026313610927e-05, + "loss": 1.1285, + "step": 9825 + }, + { + "epoch": 0.24, + "learning_rate": 1.7849780616491726e-05, + "loss": 1.1848, + "step": 9826 + }, + { + "epoch": 0.24, + "learning_rate": 1.7849298049251415e-05, + "loss": 1.1523, + "step": 9827 + }, + { + "epoch": 0.24, + "learning_rate": 1.7848815434391254e-05, + "loss": 1.1613, + "step": 9828 + }, + { + "epoch": 0.24, + "learning_rate": 1.784833277191418e-05, + "loss": 1.1687, + "step": 9829 + }, + { + "epoch": 0.24, + "learning_rate": 1.784785006182311e-05, + "loss": 1.0217, + "step": 9830 + }, + { + "epoch": 0.24, + "learning_rate": 1.7847367304120987e-05, + "loss": 1.2261, + "step": 9831 + }, + { + "epoch": 0.24, + "learning_rate": 1.784688449881073e-05, + "loss": 1.3191, + "step": 9832 + }, + { + "epoch": 0.24, + "learning_rate": 1.784640164589527e-05, + "loss": 1.2993, + "step": 9833 + }, + { + "epoch": 0.24, + "learning_rate": 1.784591874537754e-05, + "loss": 0.982, + "step": 9834 + }, + { + "epoch": 0.24, + "learning_rate": 1.784543579726046e-05, + "loss": 1.4367, + "step": 9835 + }, + { + "epoch": 0.24, + "learning_rate": 1.784495280154697e-05, + "loss": 1.1342, + "step": 9836 + }, + { + "epoch": 0.24, + "learning_rate": 1.7844469758239998e-05, + "loss": 1.0694, + "step": 9837 + }, + { + "epoch": 0.24, + "learning_rate": 1.784398666734247e-05, + "loss": 1.1694, + "step": 9838 + }, + { + "epoch": 0.24, + "learning_rate": 1.784350352885732e-05, + "loss": 1.1918, + "step": 9839 + }, + { + "epoch": 0.24, + "learning_rate": 1.784302034278748e-05, + "loss": 1.2979, + "step": 9840 + }, + { + "epoch": 0.24, + "learning_rate": 1.7842537109135887e-05, + "loss": 1.1623, + "step": 9841 + }, + { + "epoch": 0.24, + "learning_rate": 1.7842053827905457e-05, + "loss": 1.1555, + "step": 9842 + }, + { + "epoch": 0.24, + "learning_rate": 1.7841570499099138e-05, + "loss": 0.8228, + "step": 9843 + }, + { + "epoch": 0.24, + "learning_rate": 1.784108712271985e-05, + "loss": 0.8083, + "step": 9844 + }, + { + "epoch": 0.24, + "learning_rate": 1.7840603698770528e-05, + "loss": 1.0629, + "step": 9845 + }, + { + "epoch": 0.24, + "learning_rate": 1.7840120227254115e-05, + "loss": 1.2239, + "step": 9846 + }, + { + "epoch": 0.24, + "learning_rate": 1.7839636708173533e-05, + "loss": 1.0201, + "step": 9847 + }, + { + "epoch": 0.24, + "learning_rate": 1.783915314153172e-05, + "loss": 1.0004, + "step": 9848 + }, + { + "epoch": 0.24, + "learning_rate": 1.7838669527331608e-05, + "loss": 0.8881, + "step": 9849 + }, + { + "epoch": 0.24, + "learning_rate": 1.783818586557613e-05, + "loss": 1.1855, + "step": 9850 + }, + { + "epoch": 0.24, + "learning_rate": 1.7837702156268223e-05, + "loss": 0.8826, + "step": 9851 + }, + { + "epoch": 0.24, + "learning_rate": 1.7837218399410822e-05, + "loss": 0.8524, + "step": 9852 + }, + { + "epoch": 0.24, + "learning_rate": 1.783673459500686e-05, + "loss": 1.2076, + "step": 9853 + }, + { + "epoch": 0.24, + "learning_rate": 1.783625074305927e-05, + "loss": 1.1251, + "step": 9854 + }, + { + "epoch": 0.24, + "learning_rate": 1.7835766843570995e-05, + "loss": 1.3075, + "step": 9855 + }, + { + "epoch": 0.24, + "learning_rate": 1.7835282896544963e-05, + "loss": 1.1051, + "step": 9856 + }, + { + "epoch": 0.24, + "learning_rate": 1.7834798901984113e-05, + "loss": 1.1486, + "step": 9857 + }, + { + "epoch": 0.24, + "learning_rate": 1.783431485989138e-05, + "loss": 1.0626, + "step": 9858 + }, + { + "epoch": 0.24, + "learning_rate": 1.7833830770269703e-05, + "loss": 1.2613, + "step": 9859 + }, + { + "epoch": 0.24, + "learning_rate": 1.7833346633122013e-05, + "loss": 1.224, + "step": 9860 + }, + { + "epoch": 0.24, + "learning_rate": 1.7832862448451256e-05, + "loss": 1.4211, + "step": 9861 + }, + { + "epoch": 0.24, + "learning_rate": 1.7832378216260365e-05, + "loss": 1.2055, + "step": 9862 + }, + { + "epoch": 0.24, + "learning_rate": 1.7831893936552277e-05, + "loss": 1.2401, + "step": 9863 + }, + { + "epoch": 0.24, + "learning_rate": 1.7831409609329927e-05, + "loss": 1.3013, + "step": 9864 + }, + { + "epoch": 0.24, + "learning_rate": 1.7830925234596265e-05, + "loss": 0.944, + "step": 9865 + }, + { + "epoch": 0.24, + "learning_rate": 1.7830440812354216e-05, + "loss": 1.2706, + "step": 9866 + }, + { + "epoch": 0.24, + "learning_rate": 1.782995634260672e-05, + "loss": 1.2025, + "step": 9867 + }, + { + "epoch": 0.24, + "learning_rate": 1.782947182535673e-05, + "loss": 1.316, + "step": 9868 + }, + { + "epoch": 0.24, + "learning_rate": 1.7828987260607168e-05, + "loss": 1.2011, + "step": 9869 + }, + { + "epoch": 0.24, + "learning_rate": 1.782850264836099e-05, + "loss": 0.9572, + "step": 9870 + }, + { + "epoch": 0.24, + "learning_rate": 1.7828017988621124e-05, + "loss": 1.2425, + "step": 9871 + }, + { + "epoch": 0.24, + "learning_rate": 1.782753328139051e-05, + "loss": 1.0895, + "step": 9872 + }, + { + "epoch": 0.24, + "learning_rate": 1.78270485266721e-05, + "loss": 1.3811, + "step": 9873 + }, + { + "epoch": 0.24, + "learning_rate": 1.782656372446882e-05, + "loss": 1.2156, + "step": 9874 + }, + { + "epoch": 0.24, + "learning_rate": 1.7826078874783622e-05, + "loss": 1.2629, + "step": 9875 + }, + { + "epoch": 0.24, + "learning_rate": 1.7825593977619443e-05, + "loss": 0.681, + "step": 9876 + }, + { + "epoch": 0.24, + "learning_rate": 1.7825109032979232e-05, + "loss": 1.2301, + "step": 9877 + }, + { + "epoch": 0.24, + "learning_rate": 1.782462404086592e-05, + "loss": 0.7773, + "step": 9878 + }, + { + "epoch": 0.24, + "learning_rate": 1.7824139001282453e-05, + "loss": 0.8432, + "step": 9879 + }, + { + "epoch": 0.24, + "learning_rate": 1.782365391423178e-05, + "loss": 1.2436, + "step": 9880 + }, + { + "epoch": 0.24, + "learning_rate": 1.782316877971683e-05, + "loss": 1.0761, + "step": 9881 + }, + { + "epoch": 0.24, + "learning_rate": 1.7822683597740568e-05, + "loss": 1.3887, + "step": 9882 + }, + { + "epoch": 0.24, + "learning_rate": 1.7822198368305916e-05, + "loss": 1.2877, + "step": 9883 + }, + { + "epoch": 0.24, + "learning_rate": 1.782171309141583e-05, + "loss": 1.166, + "step": 9884 + }, + { + "epoch": 0.24, + "learning_rate": 1.7821227767073246e-05, + "loss": 1.0026, + "step": 9885 + }, + { + "epoch": 0.24, + "learning_rate": 1.782074239528112e-05, + "loss": 1.2378, + "step": 9886 + }, + { + "epoch": 0.24, + "learning_rate": 1.782025697604238e-05, + "loss": 1.1127, + "step": 9887 + }, + { + "epoch": 0.24, + "learning_rate": 1.781977150935999e-05, + "loss": 1.2024, + "step": 9888 + }, + { + "epoch": 0.24, + "learning_rate": 1.781928599523688e-05, + "loss": 1.2424, + "step": 9889 + }, + { + "epoch": 0.24, + "learning_rate": 1.7818800433676e-05, + "loss": 1.29, + "step": 9890 + }, + { + "epoch": 0.24, + "learning_rate": 1.78183148246803e-05, + "loss": 1.2862, + "step": 9891 + }, + { + "epoch": 0.24, + "learning_rate": 1.781782916825272e-05, + "loss": 1.0056, + "step": 9892 + }, + { + "epoch": 0.24, + "learning_rate": 1.7817343464396212e-05, + "loss": 1.2412, + "step": 9893 + }, + { + "epoch": 0.24, + "learning_rate": 1.781685771311372e-05, + "loss": 1.3243, + "step": 9894 + }, + { + "epoch": 0.24, + "learning_rate": 1.781637191440819e-05, + "loss": 1.1467, + "step": 9895 + }, + { + "epoch": 0.24, + "learning_rate": 1.781588606828257e-05, + "loss": 1.0341, + "step": 9896 + }, + { + "epoch": 0.24, + "learning_rate": 1.7815400174739806e-05, + "loss": 1.1468, + "step": 9897 + }, + { + "epoch": 0.24, + "learning_rate": 1.7814914233782848e-05, + "loss": 1.1264, + "step": 9898 + }, + { + "epoch": 0.24, + "learning_rate": 1.7814428245414647e-05, + "loss": 1.2232, + "step": 9899 + }, + { + "epoch": 0.24, + "learning_rate": 1.7813942209638148e-05, + "loss": 1.2578, + "step": 9900 + }, + { + "epoch": 0.24, + "learning_rate": 1.7813456126456294e-05, + "loss": 1.1895, + "step": 9901 + }, + { + "epoch": 0.24, + "learning_rate": 1.7812969995872044e-05, + "loss": 1.1186, + "step": 9902 + }, + { + "epoch": 0.24, + "learning_rate": 1.7812483817888342e-05, + "loss": 1.3463, + "step": 9903 + }, + { + "epoch": 0.24, + "learning_rate": 1.781199759250814e-05, + "loss": 0.878, + "step": 9904 + }, + { + "epoch": 0.24, + "learning_rate": 1.781151131973438e-05, + "loss": 0.8188, + "step": 9905 + }, + { + "epoch": 0.24, + "learning_rate": 1.7811024999570023e-05, + "loss": 1.1325, + "step": 9906 + }, + { + "epoch": 0.24, + "learning_rate": 1.7810538632018013e-05, + "loss": 1.2035, + "step": 9907 + }, + { + "epoch": 0.24, + "learning_rate": 1.7810052217081306e-05, + "loss": 1.1876, + "step": 9908 + }, + { + "epoch": 0.24, + "learning_rate": 1.7809565754762844e-05, + "loss": 1.223, + "step": 9909 + }, + { + "epoch": 0.24, + "learning_rate": 1.7809079245065586e-05, + "loss": 1.2248, + "step": 9910 + }, + { + "epoch": 0.24, + "learning_rate": 1.780859268799248e-05, + "loss": 1.1418, + "step": 9911 + }, + { + "epoch": 0.24, + "learning_rate": 1.7808106083546478e-05, + "loss": 1.2041, + "step": 9912 + }, + { + "epoch": 0.24, + "learning_rate": 1.7807619431730536e-05, + "loss": 1.2125, + "step": 9913 + }, + { + "epoch": 0.24, + "learning_rate": 1.7807132732547603e-05, + "loss": 1.3944, + "step": 9914 + }, + { + "epoch": 0.24, + "learning_rate": 1.780664598600063e-05, + "loss": 1.072, + "step": 9915 + }, + { + "epoch": 0.24, + "learning_rate": 1.7806159192092575e-05, + "loss": 1.0038, + "step": 9916 + }, + { + "epoch": 0.24, + "learning_rate": 1.7805672350826384e-05, + "loss": 1.0925, + "step": 9917 + }, + { + "epoch": 0.24, + "learning_rate": 1.7805185462205018e-05, + "loss": 1.0905, + "step": 9918 + }, + { + "epoch": 0.24, + "learning_rate": 1.7804698526231426e-05, + "loss": 1.183, + "step": 9919 + }, + { + "epoch": 0.24, + "learning_rate": 1.7804211542908568e-05, + "loss": 0.9327, + "step": 9920 + }, + { + "epoch": 0.24, + "learning_rate": 1.780372451223939e-05, + "loss": 1.3233, + "step": 9921 + }, + { + "epoch": 0.24, + "learning_rate": 1.780323743422685e-05, + "loss": 1.038, + "step": 9922 + }, + { + "epoch": 0.24, + "learning_rate": 1.7802750308873907e-05, + "loss": 1.3324, + "step": 9923 + }, + { + "epoch": 0.24, + "learning_rate": 1.7802263136183514e-05, + "loss": 1.2305, + "step": 9924 + }, + { + "epoch": 0.24, + "learning_rate": 1.7801775916158623e-05, + "loss": 1.0058, + "step": 9925 + }, + { + "epoch": 0.24, + "learning_rate": 1.780128864880219e-05, + "loss": 1.2443, + "step": 9926 + }, + { + "epoch": 0.24, + "learning_rate": 1.780080133411718e-05, + "loss": 1.1999, + "step": 9927 + }, + { + "epoch": 0.24, + "learning_rate": 1.780031397210654e-05, + "loss": 1.189, + "step": 9928 + }, + { + "epoch": 0.24, + "learning_rate": 1.7799826562773227e-05, + "loss": 1.0139, + "step": 9929 + }, + { + "epoch": 0.24, + "learning_rate": 1.7799339106120205e-05, + "loss": 0.9581, + "step": 9930 + }, + { + "epoch": 0.24, + "learning_rate": 1.779885160215043e-05, + "loss": 1.0791, + "step": 9931 + }, + { + "epoch": 0.24, + "learning_rate": 1.7798364050866853e-05, + "loss": 0.9899, + "step": 9932 + }, + { + "epoch": 0.24, + "learning_rate": 1.779787645227243e-05, + "loss": 1.2439, + "step": 9933 + }, + { + "epoch": 0.24, + "learning_rate": 1.7797388806370132e-05, + "loss": 1.0344, + "step": 9934 + }, + { + "epoch": 0.24, + "learning_rate": 1.779690111316291e-05, + "loss": 1.0837, + "step": 9935 + }, + { + "epoch": 0.24, + "learning_rate": 1.779641337265372e-05, + "loss": 1.3707, + "step": 9936 + }, + { + "epoch": 0.24, + "learning_rate": 1.7795925584845525e-05, + "loss": 1.3436, + "step": 9937 + }, + { + "epoch": 0.24, + "learning_rate": 1.7795437749741283e-05, + "loss": 1.158, + "step": 9938 + }, + { + "epoch": 0.24, + "learning_rate": 1.7794949867343952e-05, + "loss": 1.0993, + "step": 9939 + }, + { + "epoch": 0.24, + "learning_rate": 1.77944619376565e-05, + "loss": 1.1618, + "step": 9940 + }, + { + "epoch": 0.24, + "learning_rate": 1.7793973960681877e-05, + "loss": 1.1634, + "step": 9941 + }, + { + "epoch": 0.24, + "learning_rate": 1.7793485936423045e-05, + "loss": 1.1931, + "step": 9942 + }, + { + "epoch": 0.24, + "learning_rate": 1.779299786488297e-05, + "loss": 1.1566, + "step": 9943 + }, + { + "epoch": 0.24, + "learning_rate": 1.7792509746064608e-05, + "loss": 1.3347, + "step": 9944 + }, + { + "epoch": 0.24, + "learning_rate": 1.7792021579970922e-05, + "loss": 1.0607, + "step": 9945 + }, + { + "epoch": 0.24, + "learning_rate": 1.7791533366604876e-05, + "loss": 1.1651, + "step": 9946 + }, + { + "epoch": 0.24, + "learning_rate": 1.779104510596943e-05, + "loss": 1.1226, + "step": 9947 + }, + { + "epoch": 0.24, + "learning_rate": 1.7790556798067543e-05, + "loss": 0.9888, + "step": 9948 + }, + { + "epoch": 0.24, + "learning_rate": 1.779006844290218e-05, + "loss": 1.2219, + "step": 9949 + }, + { + "epoch": 0.24, + "learning_rate": 1.7789580040476305e-05, + "loss": 0.9777, + "step": 9950 + }, + { + "epoch": 0.24, + "learning_rate": 1.778909159079288e-05, + "loss": 1.145, + "step": 9951 + }, + { + "epoch": 0.24, + "learning_rate": 1.7788603093854872e-05, + "loss": 1.3097, + "step": 9952 + }, + { + "epoch": 0.24, + "learning_rate": 1.7788114549665238e-05, + "loss": 1.1941, + "step": 9953 + }, + { + "epoch": 0.24, + "learning_rate": 1.7787625958226947e-05, + "loss": 1.3975, + "step": 9954 + }, + { + "epoch": 0.24, + "learning_rate": 1.7787137319542958e-05, + "loss": 1.4252, + "step": 9955 + }, + { + "epoch": 0.24, + "learning_rate": 1.778664863361624e-05, + "loss": 1.2569, + "step": 9956 + }, + { + "epoch": 0.24, + "learning_rate": 1.778615990044976e-05, + "loss": 1.2074, + "step": 9957 + }, + { + "epoch": 0.24, + "learning_rate": 1.7785671120046472e-05, + "loss": 0.8695, + "step": 9958 + }, + { + "epoch": 0.24, + "learning_rate": 1.7785182292409354e-05, + "loss": 1.18, + "step": 9959 + }, + { + "epoch": 0.24, + "learning_rate": 1.7784693417541364e-05, + "loss": 1.1554, + "step": 9960 + }, + { + "epoch": 0.24, + "learning_rate": 1.778420449544547e-05, + "loss": 1.1582, + "step": 9961 + }, + { + "epoch": 0.24, + "learning_rate": 1.7783715526124637e-05, + "loss": 1.0058, + "step": 9962 + }, + { + "epoch": 0.24, + "learning_rate": 1.7783226509581835e-05, + "loss": 1.1735, + "step": 9963 + }, + { + "epoch": 0.24, + "learning_rate": 1.778273744582003e-05, + "loss": 1.2103, + "step": 9964 + }, + { + "epoch": 0.24, + "learning_rate": 1.7782248334842185e-05, + "loss": 1.192, + "step": 9965 + }, + { + "epoch": 0.24, + "learning_rate": 1.7781759176651273e-05, + "loss": 1.1255, + "step": 9966 + }, + { + "epoch": 0.24, + "learning_rate": 1.7781269971250255e-05, + "loss": 1.1262, + "step": 9967 + }, + { + "epoch": 0.24, + "learning_rate": 1.77807807186421e-05, + "loss": 1.0299, + "step": 9968 + }, + { + "epoch": 0.24, + "learning_rate": 1.7780291418829785e-05, + "loss": 1.1675, + "step": 9969 + }, + { + "epoch": 0.24, + "learning_rate": 1.777980207181627e-05, + "loss": 1.2767, + "step": 9970 + }, + { + "epoch": 0.24, + "learning_rate": 1.7779312677604522e-05, + "loss": 1.0448, + "step": 9971 + }, + { + "epoch": 0.24, + "learning_rate": 1.7778823236197515e-05, + "loss": 1.2654, + "step": 9972 + }, + { + "epoch": 0.24, + "learning_rate": 1.777833374759822e-05, + "loss": 0.9369, + "step": 9973 + }, + { + "epoch": 0.24, + "learning_rate": 1.7777844211809602e-05, + "loss": 1.0414, + "step": 9974 + }, + { + "epoch": 0.24, + "learning_rate": 1.7777354628834632e-05, + "loss": 1.2452, + "step": 9975 + }, + { + "epoch": 0.24, + "learning_rate": 1.7776864998676284e-05, + "loss": 1.0891, + "step": 9976 + }, + { + "epoch": 0.24, + "learning_rate": 1.7776375321337523e-05, + "loss": 1.3259, + "step": 9977 + }, + { + "epoch": 0.24, + "learning_rate": 1.777588559682132e-05, + "loss": 1.1742, + "step": 9978 + }, + { + "epoch": 0.24, + "learning_rate": 1.777539582513065e-05, + "loss": 1.0532, + "step": 9979 + }, + { + "epoch": 0.24, + "learning_rate": 1.7774906006268482e-05, + "loss": 1.1881, + "step": 9980 + }, + { + "epoch": 0.24, + "learning_rate": 1.7774416140237788e-05, + "loss": 1.0277, + "step": 9981 + }, + { + "epoch": 0.24, + "learning_rate": 1.777392622704154e-05, + "loss": 1.0513, + "step": 9982 + }, + { + "epoch": 0.24, + "learning_rate": 1.7773436266682713e-05, + "loss": 0.8718, + "step": 9983 + }, + { + "epoch": 0.24, + "learning_rate": 1.7772946259164272e-05, + "loss": 1.2157, + "step": 9984 + }, + { + "epoch": 0.24, + "learning_rate": 1.7772456204489196e-05, + "loss": 0.8278, + "step": 9985 + }, + { + "epoch": 0.24, + "learning_rate": 1.777196610266046e-05, + "loss": 1.055, + "step": 9986 + }, + { + "epoch": 0.24, + "learning_rate": 1.777147595368103e-05, + "loss": 1.2056, + "step": 9987 + }, + { + "epoch": 0.24, + "learning_rate": 1.7770985757553883e-05, + "loss": 0.9681, + "step": 9988 + }, + { + "epoch": 0.24, + "learning_rate": 1.7770495514281995e-05, + "loss": 1.209, + "step": 9989 + }, + { + "epoch": 0.24, + "learning_rate": 1.7770005223868336e-05, + "loss": 1.438, + "step": 9990 + }, + { + "epoch": 0.24, + "learning_rate": 1.7769514886315887e-05, + "loss": 1.1944, + "step": 9991 + }, + { + "epoch": 0.24, + "learning_rate": 1.7769024501627617e-05, + "loss": 1.2231, + "step": 9992 + }, + { + "epoch": 0.24, + "learning_rate": 1.77685340698065e-05, + "loss": 1.3447, + "step": 9993 + }, + { + "epoch": 0.24, + "learning_rate": 1.7768043590855514e-05, + "loss": 1.0828, + "step": 9994 + }, + { + "epoch": 0.24, + "learning_rate": 1.7767553064777638e-05, + "loss": 1.1431, + "step": 9995 + }, + { + "epoch": 0.24, + "learning_rate": 1.7767062491575843e-05, + "loss": 1.0077, + "step": 9996 + }, + { + "epoch": 0.24, + "learning_rate": 1.776657187125311e-05, + "loss": 0.9601, + "step": 9997 + }, + { + "epoch": 0.24, + "learning_rate": 1.7766081203812408e-05, + "loss": 1.2845, + "step": 9998 + }, + { + "epoch": 0.24, + "learning_rate": 1.7765590489256717e-05, + "loss": 1.131, + "step": 9999 + }, + { + "epoch": 0.24, + "learning_rate": 1.776509972758902e-05, + "loss": 1.0621, + "step": 10000 + }, + { + "epoch": 0.24, + "learning_rate": 1.7764608918812286e-05, + "loss": 0.8107, + "step": 10001 + }, + { + "epoch": 0.24, + "learning_rate": 1.7764118062929497e-05, + "loss": 1.196, + "step": 10002 + }, + { + "epoch": 0.24, + "learning_rate": 1.7763627159943633e-05, + "loss": 1.0771, + "step": 10003 + }, + { + "epoch": 0.24, + "learning_rate": 1.7763136209857666e-05, + "loss": 1.2439, + "step": 10004 + }, + { + "epoch": 0.24, + "learning_rate": 1.7762645212674575e-05, + "loss": 1.296, + "step": 10005 + }, + { + "epoch": 0.24, + "learning_rate": 1.7762154168397345e-05, + "loss": 0.9552, + "step": 10006 + }, + { + "epoch": 0.24, + "learning_rate": 1.776166307702895e-05, + "loss": 0.9846, + "step": 10007 + }, + { + "epoch": 0.24, + "learning_rate": 1.7761171938572373e-05, + "loss": 0.9542, + "step": 10008 + }, + { + "epoch": 0.24, + "learning_rate": 1.776068075303059e-05, + "loss": 1.0828, + "step": 10009 + }, + { + "epoch": 0.24, + "learning_rate": 1.776018952040658e-05, + "loss": 1.329, + "step": 10010 + }, + { + "epoch": 0.24, + "learning_rate": 1.7759698240703328e-05, + "loss": 1.3364, + "step": 10011 + }, + { + "epoch": 0.24, + "learning_rate": 1.775920691392381e-05, + "loss": 1.1317, + "step": 10012 + }, + { + "epoch": 0.24, + "learning_rate": 1.7758715540071013e-05, + "loss": 1.1657, + "step": 10013 + }, + { + "epoch": 0.24, + "learning_rate": 1.7758224119147908e-05, + "loss": 1.0302, + "step": 10014 + }, + { + "epoch": 0.24, + "learning_rate": 1.7757732651157488e-05, + "loss": 1.0031, + "step": 10015 + }, + { + "epoch": 0.24, + "learning_rate": 1.7757241136102723e-05, + "loss": 1.1543, + "step": 10016 + }, + { + "epoch": 0.24, + "learning_rate": 1.7756749573986604e-05, + "loss": 1.3571, + "step": 10017 + }, + { + "epoch": 0.24, + "learning_rate": 1.775625796481211e-05, + "loss": 1.1406, + "step": 10018 + }, + { + "epoch": 0.24, + "learning_rate": 1.775576630858222e-05, + "loss": 1.1338, + "step": 10019 + }, + { + "epoch": 0.24, + "learning_rate": 1.7755274605299924e-05, + "loss": 1.2824, + "step": 10020 + }, + { + "epoch": 0.24, + "learning_rate": 1.77547828549682e-05, + "loss": 0.8469, + "step": 10021 + }, + { + "epoch": 0.24, + "learning_rate": 1.7754291057590033e-05, + "loss": 1.0654, + "step": 10022 + }, + { + "epoch": 0.24, + "learning_rate": 1.7753799213168404e-05, + "loss": 1.1939, + "step": 10023 + }, + { + "epoch": 0.24, + "learning_rate": 1.77533073217063e-05, + "loss": 1.3658, + "step": 10024 + }, + { + "epoch": 0.24, + "learning_rate": 1.7752815383206706e-05, + "loss": 0.8875, + "step": 10025 + }, + { + "epoch": 0.24, + "learning_rate": 1.7752323397672603e-05, + "loss": 1.1453, + "step": 10026 + }, + { + "epoch": 0.24, + "learning_rate": 1.775183136510698e-05, + "loss": 0.9049, + "step": 10027 + }, + { + "epoch": 0.24, + "learning_rate": 1.775133928551281e-05, + "loss": 1.0008, + "step": 10028 + }, + { + "epoch": 0.24, + "learning_rate": 1.77508471588931e-05, + "loss": 1.3553, + "step": 10029 + }, + { + "epoch": 0.24, + "learning_rate": 1.7750354985250817e-05, + "loss": 1.1313, + "step": 10030 + }, + { + "epoch": 0.24, + "learning_rate": 1.7749862764588954e-05, + "loss": 1.2299, + "step": 10031 + }, + { + "epoch": 0.24, + "learning_rate": 1.7749370496910497e-05, + "loss": 1.4342, + "step": 10032 + }, + { + "epoch": 0.24, + "learning_rate": 1.7748878182218432e-05, + "loss": 1.019, + "step": 10033 + }, + { + "epoch": 0.24, + "learning_rate": 1.7748385820515744e-05, + "loss": 1.0555, + "step": 10034 + }, + { + "epoch": 0.24, + "learning_rate": 1.7747893411805424e-05, + "loss": 1.2047, + "step": 10035 + }, + { + "epoch": 0.24, + "learning_rate": 1.7747400956090457e-05, + "loss": 1.203, + "step": 10036 + }, + { + "epoch": 0.24, + "learning_rate": 1.774690845337383e-05, + "loss": 0.9521, + "step": 10037 + }, + { + "epoch": 0.24, + "learning_rate": 1.774641590365853e-05, + "loss": 1.0993, + "step": 10038 + }, + { + "epoch": 0.24, + "learning_rate": 1.774592330694755e-05, + "loss": 1.0515, + "step": 10039 + }, + { + "epoch": 0.24, + "learning_rate": 1.7745430663243874e-05, + "loss": 1.2207, + "step": 10040 + }, + { + "epoch": 0.24, + "learning_rate": 1.7744937972550492e-05, + "loss": 1.0893, + "step": 10041 + }, + { + "epoch": 0.24, + "learning_rate": 1.774444523487039e-05, + "loss": 0.9978, + "step": 10042 + }, + { + "epoch": 0.24, + "learning_rate": 1.7743952450206567e-05, + "loss": 1.1412, + "step": 10043 + }, + { + "epoch": 0.24, + "learning_rate": 1.7743459618562e-05, + "loss": 1.2288, + "step": 10044 + }, + { + "epoch": 0.24, + "learning_rate": 1.7742966739939688e-05, + "loss": 1.2443, + "step": 10045 + }, + { + "epoch": 0.24, + "learning_rate": 1.7742473814342616e-05, + "loss": 0.8641, + "step": 10046 + }, + { + "epoch": 0.24, + "learning_rate": 1.774198084177378e-05, + "loss": 1.0467, + "step": 10047 + }, + { + "epoch": 0.24, + "learning_rate": 1.7741487822236163e-05, + "loss": 1.3507, + "step": 10048 + }, + { + "epoch": 0.24, + "learning_rate": 1.7740994755732762e-05, + "loss": 1.1627, + "step": 10049 + }, + { + "epoch": 0.24, + "learning_rate": 1.7740501642266568e-05, + "loss": 1.1456, + "step": 10050 + }, + { + "epoch": 0.24, + "learning_rate": 1.774000848184057e-05, + "loss": 0.9392, + "step": 10051 + }, + { + "epoch": 0.24, + "learning_rate": 1.773951527445776e-05, + "loss": 1.2286, + "step": 10052 + }, + { + "epoch": 0.24, + "learning_rate": 1.773902202012113e-05, + "loss": 1.1727, + "step": 10053 + }, + { + "epoch": 0.24, + "learning_rate": 1.7738528718833677e-05, + "loss": 1.1922, + "step": 10054 + }, + { + "epoch": 0.24, + "learning_rate": 1.773803537059839e-05, + "loss": 1.0975, + "step": 10055 + }, + { + "epoch": 0.24, + "learning_rate": 1.7737541975418263e-05, + "loss": 1.2005, + "step": 10056 + }, + { + "epoch": 0.24, + "learning_rate": 1.7737048533296288e-05, + "loss": 1.1146, + "step": 10057 + }, + { + "epoch": 0.24, + "learning_rate": 1.7736555044235458e-05, + "loss": 1.4248, + "step": 10058 + }, + { + "epoch": 0.24, + "learning_rate": 1.773606150823877e-05, + "loss": 0.9464, + "step": 10059 + }, + { + "epoch": 0.24, + "learning_rate": 1.7735567925309217e-05, + "loss": 1.2995, + "step": 10060 + }, + { + "epoch": 0.24, + "learning_rate": 1.773507429544979e-05, + "loss": 1.1223, + "step": 10061 + }, + { + "epoch": 0.24, + "learning_rate": 1.773458061866349e-05, + "loss": 1.086, + "step": 10062 + }, + { + "epoch": 0.24, + "learning_rate": 1.7734086894953306e-05, + "loss": 1.1831, + "step": 10063 + }, + { + "epoch": 0.24, + "learning_rate": 1.7733593124322238e-05, + "loss": 0.8553, + "step": 10064 + }, + { + "epoch": 0.24, + "learning_rate": 1.773309930677328e-05, + "loss": 1.14, + "step": 10065 + }, + { + "epoch": 0.24, + "learning_rate": 1.773260544230942e-05, + "loss": 0.9473, + "step": 10066 + }, + { + "epoch": 0.24, + "learning_rate": 1.7732111530933674e-05, + "loss": 1.2167, + "step": 10067 + }, + { + "epoch": 0.24, + "learning_rate": 1.773161757264902e-05, + "loss": 0.9765, + "step": 10068 + }, + { + "epoch": 0.24, + "learning_rate": 1.7731123567458457e-05, + "loss": 1.2943, + "step": 10069 + }, + { + "epoch": 0.24, + "learning_rate": 1.773062951536499e-05, + "loss": 1.2013, + "step": 10070 + }, + { + "epoch": 0.24, + "learning_rate": 1.7730135416371612e-05, + "loss": 1.0794, + "step": 10071 + }, + { + "epoch": 0.24, + "learning_rate": 1.7729641270481323e-05, + "loss": 1.2494, + "step": 10072 + }, + { + "epoch": 0.24, + "learning_rate": 1.7729147077697114e-05, + "loss": 1.1679, + "step": 10073 + }, + { + "epoch": 0.24, + "learning_rate": 1.7728652838021988e-05, + "loss": 1.0243, + "step": 10074 + }, + { + "epoch": 0.24, + "learning_rate": 1.7728158551458946e-05, + "loss": 0.9072, + "step": 10075 + }, + { + "epoch": 0.24, + "learning_rate": 1.772766421801098e-05, + "loss": 0.9995, + "step": 10076 + }, + { + "epoch": 0.24, + "learning_rate": 1.7727169837681098e-05, + "loss": 0.9781, + "step": 10077 + }, + { + "epoch": 0.24, + "learning_rate": 1.772667541047229e-05, + "loss": 1.3679, + "step": 10078 + }, + { + "epoch": 0.24, + "learning_rate": 1.772618093638756e-05, + "loss": 1.0771, + "step": 10079 + }, + { + "epoch": 0.24, + "learning_rate": 1.772568641542991e-05, + "loss": 1.0774, + "step": 10080 + }, + { + "epoch": 0.24, + "learning_rate": 1.7725191847602336e-05, + "loss": 1.336, + "step": 10081 + }, + { + "epoch": 0.24, + "learning_rate": 1.7724697232907838e-05, + "loss": 1.0819, + "step": 10082 + }, + { + "epoch": 0.24, + "learning_rate": 1.7724202571349423e-05, + "loss": 0.9306, + "step": 10083 + }, + { + "epoch": 0.24, + "learning_rate": 1.7723707862930086e-05, + "loss": 1.1595, + "step": 10084 + }, + { + "epoch": 0.24, + "learning_rate": 1.772321310765283e-05, + "loss": 1.3297, + "step": 10085 + }, + { + "epoch": 0.24, + "learning_rate": 1.772271830552066e-05, + "loss": 1.0553, + "step": 10086 + }, + { + "epoch": 0.24, + "learning_rate": 1.772222345653657e-05, + "loss": 0.9565, + "step": 10087 + }, + { + "epoch": 0.24, + "learning_rate": 1.772172856070357e-05, + "loss": 1.3471, + "step": 10088 + }, + { + "epoch": 0.24, + "learning_rate": 1.7721233618024657e-05, + "loss": 1.0855, + "step": 10089 + }, + { + "epoch": 0.24, + "learning_rate": 1.7720738628502836e-05, + "loss": 1.1829, + "step": 10090 + }, + { + "epoch": 0.24, + "learning_rate": 1.7720243592141107e-05, + "loss": 1.1093, + "step": 10091 + }, + { + "epoch": 0.24, + "learning_rate": 1.771974850894248e-05, + "loss": 1.2818, + "step": 10092 + }, + { + "epoch": 0.24, + "learning_rate": 1.7719253378909953e-05, + "loss": 1.2137, + "step": 10093 + }, + { + "epoch": 0.24, + "learning_rate": 1.7718758202046532e-05, + "loss": 1.2734, + "step": 10094 + }, + { + "epoch": 0.24, + "learning_rate": 1.771826297835522e-05, + "loss": 1.0546, + "step": 10095 + }, + { + "epoch": 0.24, + "learning_rate": 1.771776770783902e-05, + "loss": 0.8993, + "step": 10096 + }, + { + "epoch": 0.24, + "learning_rate": 1.7717272390500943e-05, + "loss": 0.9591, + "step": 10097 + }, + { + "epoch": 0.24, + "learning_rate": 1.771677702634399e-05, + "loss": 1.1, + "step": 10098 + }, + { + "epoch": 0.24, + "learning_rate": 1.771628161537116e-05, + "loss": 0.8969, + "step": 10099 + }, + { + "epoch": 0.24, + "learning_rate": 1.771578615758547e-05, + "loss": 1.2173, + "step": 10100 + }, + { + "epoch": 0.24, + "learning_rate": 1.771529065298992e-05, + "loss": 1.1782, + "step": 10101 + }, + { + "epoch": 0.24, + "learning_rate": 1.7714795101587512e-05, + "loss": 1.2319, + "step": 10102 + }, + { + "epoch": 0.24, + "learning_rate": 1.771429950338126e-05, + "loss": 1.1876, + "step": 10103 + }, + { + "epoch": 0.24, + "learning_rate": 1.7713803858374164e-05, + "loss": 1.1747, + "step": 10104 + }, + { + "epoch": 0.24, + "learning_rate": 1.771330816656924e-05, + "loss": 1.275, + "step": 10105 + }, + { + "epoch": 0.24, + "learning_rate": 1.7712812427969485e-05, + "loss": 0.9007, + "step": 10106 + }, + { + "epoch": 0.24, + "learning_rate": 1.7712316642577913e-05, + "loss": 0.9168, + "step": 10107 + }, + { + "epoch": 0.24, + "learning_rate": 1.7711820810397532e-05, + "loss": 1.2216, + "step": 10108 + }, + { + "epoch": 0.24, + "learning_rate": 1.7711324931431344e-05, + "loss": 1.0176, + "step": 10109 + }, + { + "epoch": 0.24, + "learning_rate": 1.7710829005682364e-05, + "loss": 1.1052, + "step": 10110 + }, + { + "epoch": 0.24, + "learning_rate": 1.7710333033153597e-05, + "loss": 1.0849, + "step": 10111 + }, + { + "epoch": 0.24, + "learning_rate": 1.7709837013848056e-05, + "loss": 1.2947, + "step": 10112 + }, + { + "epoch": 0.24, + "learning_rate": 1.7709340947768742e-05, + "loss": 1.218, + "step": 10113 + }, + { + "epoch": 0.24, + "learning_rate": 1.770884483491867e-05, + "loss": 1.1833, + "step": 10114 + }, + { + "epoch": 0.24, + "learning_rate": 1.7708348675300855e-05, + "loss": 1.342, + "step": 10115 + }, + { + "epoch": 0.24, + "learning_rate": 1.77078524689183e-05, + "loss": 0.9077, + "step": 10116 + }, + { + "epoch": 0.24, + "learning_rate": 1.7707356215774015e-05, + "loss": 1.1924, + "step": 10117 + }, + { + "epoch": 0.24, + "learning_rate": 1.7706859915871016e-05, + "loss": 1.4655, + "step": 10118 + }, + { + "epoch": 0.24, + "learning_rate": 1.7706363569212306e-05, + "loss": 1.1333, + "step": 10119 + }, + { + "epoch": 0.24, + "learning_rate": 1.7705867175800905e-05, + "loss": 1.1437, + "step": 10120 + }, + { + "epoch": 0.24, + "learning_rate": 1.7705370735639817e-05, + "loss": 1.1022, + "step": 10121 + }, + { + "epoch": 0.24, + "learning_rate": 1.770487424873206e-05, + "loss": 1.2378, + "step": 10122 + }, + { + "epoch": 0.24, + "learning_rate": 1.7704377715080643e-05, + "loss": 1.1556, + "step": 10123 + }, + { + "epoch": 0.24, + "learning_rate": 1.7703881134688576e-05, + "loss": 1.3496, + "step": 10124 + }, + { + "epoch": 0.24, + "learning_rate": 1.7703384507558876e-05, + "loss": 1.1708, + "step": 10125 + }, + { + "epoch": 0.24, + "learning_rate": 1.7702887833694554e-05, + "loss": 0.9726, + "step": 10126 + }, + { + "epoch": 0.24, + "learning_rate": 1.7702391113098625e-05, + "loss": 1.0495, + "step": 10127 + }, + { + "epoch": 0.24, + "learning_rate": 1.77018943457741e-05, + "loss": 1.1251, + "step": 10128 + }, + { + "epoch": 0.24, + "learning_rate": 1.7701397531723992e-05, + "loss": 0.9472, + "step": 10129 + }, + { + "epoch": 0.24, + "learning_rate": 1.7700900670951317e-05, + "loss": 1.1078, + "step": 10130 + }, + { + "epoch": 0.24, + "learning_rate": 1.7700403763459092e-05, + "loss": 1.295, + "step": 10131 + }, + { + "epoch": 0.24, + "learning_rate": 1.7699906809250324e-05, + "loss": 0.8686, + "step": 10132 + }, + { + "epoch": 0.24, + "learning_rate": 1.7699409808328034e-05, + "loss": 0.8965, + "step": 10133 + }, + { + "epoch": 0.24, + "learning_rate": 1.769891276069524e-05, + "loss": 0.9652, + "step": 10134 + }, + { + "epoch": 0.24, + "learning_rate": 1.7698415666354945e-05, + "loss": 1.1696, + "step": 10135 + }, + { + "epoch": 0.24, + "learning_rate": 1.7697918525310177e-05, + "loss": 1.0212, + "step": 10136 + }, + { + "epoch": 0.24, + "learning_rate": 1.769742133756395e-05, + "loss": 0.8103, + "step": 10137 + }, + { + "epoch": 0.24, + "learning_rate": 1.7696924103119277e-05, + "loss": 1.1527, + "step": 10138 + }, + { + "epoch": 0.24, + "learning_rate": 1.7696426821979176e-05, + "loss": 1.2498, + "step": 10139 + }, + { + "epoch": 0.24, + "learning_rate": 1.7695929494146662e-05, + "loss": 1.2305, + "step": 10140 + }, + { + "epoch": 0.24, + "learning_rate": 1.7695432119624756e-05, + "loss": 1.0248, + "step": 10141 + }, + { + "epoch": 0.24, + "learning_rate": 1.7694934698416472e-05, + "loss": 1.1818, + "step": 10142 + }, + { + "epoch": 0.24, + "learning_rate": 1.769443723052483e-05, + "loss": 1.1709, + "step": 10143 + }, + { + "epoch": 0.24, + "learning_rate": 1.7693939715952847e-05, + "loss": 0.9598, + "step": 10144 + }, + { + "epoch": 0.24, + "learning_rate": 1.769344215470354e-05, + "loss": 0.9322, + "step": 10145 + }, + { + "epoch": 0.24, + "learning_rate": 1.7692944546779932e-05, + "loss": 1.1831, + "step": 10146 + }, + { + "epoch": 0.24, + "learning_rate": 1.7692446892185037e-05, + "loss": 0.9762, + "step": 10147 + }, + { + "epoch": 0.24, + "learning_rate": 1.7691949190921877e-05, + "loss": 1.1391, + "step": 10148 + }, + { + "epoch": 0.24, + "learning_rate": 1.769145144299347e-05, + "loss": 1.4207, + "step": 10149 + }, + { + "epoch": 0.24, + "learning_rate": 1.7690953648402835e-05, + "loss": 1.1271, + "step": 10150 + }, + { + "epoch": 0.24, + "learning_rate": 1.7690455807152996e-05, + "loss": 1.3443, + "step": 10151 + }, + { + "epoch": 0.24, + "learning_rate": 1.768995791924697e-05, + "loss": 1.1227, + "step": 10152 + }, + { + "epoch": 0.24, + "learning_rate": 1.7689459984687777e-05, + "loss": 1.0112, + "step": 10153 + }, + { + "epoch": 0.24, + "learning_rate": 1.7688962003478438e-05, + "loss": 1.1021, + "step": 10154 + }, + { + "epoch": 0.24, + "learning_rate": 1.7688463975621975e-05, + "loss": 1.1796, + "step": 10155 + }, + { + "epoch": 0.24, + "learning_rate": 1.768796590112141e-05, + "loss": 0.9792, + "step": 10156 + }, + { + "epoch": 0.24, + "learning_rate": 1.7687467779979765e-05, + "loss": 1.1062, + "step": 10157 + }, + { + "epoch": 0.24, + "learning_rate": 1.7686969612200063e-05, + "loss": 1.2704, + "step": 10158 + }, + { + "epoch": 0.24, + "learning_rate": 1.7686471397785322e-05, + "loss": 1.2617, + "step": 10159 + }, + { + "epoch": 0.24, + "learning_rate": 1.7685973136738568e-05, + "loss": 0.9048, + "step": 10160 + }, + { + "epoch": 0.24, + "learning_rate": 1.7685474829062822e-05, + "loss": 0.9873, + "step": 10161 + }, + { + "epoch": 0.24, + "learning_rate": 1.768497647476111e-05, + "loss": 0.8171, + "step": 10162 + }, + { + "epoch": 0.24, + "learning_rate": 1.768447807383645e-05, + "loss": 1.2394, + "step": 10163 + }, + { + "epoch": 0.24, + "learning_rate": 1.768397962629187e-05, + "loss": 1.0983, + "step": 10164 + }, + { + "epoch": 0.24, + "learning_rate": 1.7683481132130396e-05, + "loss": 1.4342, + "step": 10165 + }, + { + "epoch": 0.24, + "learning_rate": 1.7682982591355043e-05, + "loss": 1.1573, + "step": 10166 + }, + { + "epoch": 0.24, + "learning_rate": 1.768248400396885e-05, + "loss": 1.0501, + "step": 10167 + }, + { + "epoch": 0.24, + "learning_rate": 1.7681985369974827e-05, + "loss": 1.0995, + "step": 10168 + }, + { + "epoch": 0.24, + "learning_rate": 1.7681486689376008e-05, + "loss": 1.1668, + "step": 10169 + }, + { + "epoch": 0.24, + "learning_rate": 1.7680987962175415e-05, + "loss": 1.1083, + "step": 10170 + }, + { + "epoch": 0.24, + "learning_rate": 1.7680489188376076e-05, + "loss": 1.3155, + "step": 10171 + }, + { + "epoch": 0.24, + "learning_rate": 1.7679990367981015e-05, + "loss": 1.0978, + "step": 10172 + }, + { + "epoch": 0.24, + "learning_rate": 1.767949150099326e-05, + "loss": 1.1721, + "step": 10173 + }, + { + "epoch": 0.24, + "learning_rate": 1.7678992587415834e-05, + "loss": 1.2089, + "step": 10174 + }, + { + "epoch": 0.24, + "learning_rate": 1.7678493627251768e-05, + "loss": 1.1503, + "step": 10175 + }, + { + "epoch": 0.24, + "learning_rate": 1.767799462050409e-05, + "loss": 1.1992, + "step": 10176 + }, + { + "epoch": 0.24, + "learning_rate": 1.767749556717582e-05, + "loss": 1.1614, + "step": 10177 + }, + { + "epoch": 0.24, + "learning_rate": 1.767699646726999e-05, + "loss": 1.0966, + "step": 10178 + }, + { + "epoch": 0.24, + "learning_rate": 1.7676497320789633e-05, + "loss": 1.3092, + "step": 10179 + }, + { + "epoch": 0.24, + "learning_rate": 1.767599812773777e-05, + "loss": 1.1955, + "step": 10180 + }, + { + "epoch": 0.24, + "learning_rate": 1.7675498888117428e-05, + "loss": 1.2709, + "step": 10181 + }, + { + "epoch": 0.24, + "learning_rate": 1.7674999601931645e-05, + "loss": 1.1775, + "step": 10182 + }, + { + "epoch": 0.24, + "learning_rate": 1.7674500269183444e-05, + "loss": 1.1622, + "step": 10183 + }, + { + "epoch": 0.24, + "learning_rate": 1.7674000889875854e-05, + "loss": 1.3477, + "step": 10184 + }, + { + "epoch": 0.24, + "learning_rate": 1.7673501464011905e-05, + "loss": 1.0955, + "step": 10185 + }, + { + "epoch": 0.24, + "learning_rate": 1.7673001991594625e-05, + "loss": 1.1423, + "step": 10186 + }, + { + "epoch": 0.24, + "learning_rate": 1.767250247262705e-05, + "loss": 1.1652, + "step": 10187 + }, + { + "epoch": 0.25, + "learning_rate": 1.7672002907112203e-05, + "loss": 1.3669, + "step": 10188 + }, + { + "epoch": 0.25, + "learning_rate": 1.7671503295053126e-05, + "loss": 0.9038, + "step": 10189 + }, + { + "epoch": 0.25, + "learning_rate": 1.767100363645284e-05, + "loss": 1.0626, + "step": 10190 + }, + { + "epoch": 0.25, + "learning_rate": 1.767050393131438e-05, + "loss": 1.3467, + "step": 10191 + }, + { + "epoch": 0.25, + "learning_rate": 1.7670004179640773e-05, + "loss": 1.1092, + "step": 10192 + }, + { + "epoch": 0.25, + "learning_rate": 1.766950438143506e-05, + "loss": 1.0655, + "step": 10193 + }, + { + "epoch": 0.25, + "learning_rate": 1.7669004536700263e-05, + "loss": 0.9524, + "step": 10194 + }, + { + "epoch": 0.25, + "learning_rate": 1.7668504645439424e-05, + "loss": 1.0992, + "step": 10195 + }, + { + "epoch": 0.25, + "learning_rate": 1.7668004707655568e-05, + "loss": 0.9169, + "step": 10196 + }, + { + "epoch": 0.25, + "learning_rate": 1.766750472335173e-05, + "loss": 1.1204, + "step": 10197 + }, + { + "epoch": 0.25, + "learning_rate": 1.7667004692530946e-05, + "loss": 1.0919, + "step": 10198 + }, + { + "epoch": 0.25, + "learning_rate": 1.766650461519625e-05, + "loss": 1.2708, + "step": 10199 + }, + { + "epoch": 0.25, + "learning_rate": 1.766600449135067e-05, + "loss": 1.1356, + "step": 10200 + }, + { + "epoch": 0.25, + "learning_rate": 1.7665504320997247e-05, + "loss": 1.3315, + "step": 10201 + }, + { + "epoch": 0.25, + "learning_rate": 1.766500410413901e-05, + "loss": 0.9987, + "step": 10202 + }, + { + "epoch": 0.25, + "learning_rate": 1.7664503840778996e-05, + "loss": 1.0445, + "step": 10203 + }, + { + "epoch": 0.25, + "learning_rate": 1.766400353092024e-05, + "loss": 0.9683, + "step": 10204 + }, + { + "epoch": 0.25, + "learning_rate": 1.7663503174565782e-05, + "loss": 1.0656, + "step": 10205 + }, + { + "epoch": 0.25, + "learning_rate": 1.766300277171865e-05, + "loss": 1.1055, + "step": 10206 + }, + { + "epoch": 0.25, + "learning_rate": 1.766250232238188e-05, + "loss": 1.2158, + "step": 10207 + }, + { + "epoch": 0.25, + "learning_rate": 1.766200182655851e-05, + "loss": 1.1793, + "step": 10208 + }, + { + "epoch": 0.25, + "learning_rate": 1.7661501284251582e-05, + "loss": 1.2844, + "step": 10209 + }, + { + "epoch": 0.25, + "learning_rate": 1.7661000695464125e-05, + "loss": 1.1918, + "step": 10210 + }, + { + "epoch": 0.25, + "learning_rate": 1.766050006019918e-05, + "loss": 1.1307, + "step": 10211 + }, + { + "epoch": 0.25, + "learning_rate": 1.7659999378459783e-05, + "loss": 1.038, + "step": 10212 + }, + { + "epoch": 0.25, + "learning_rate": 1.7659498650248973e-05, + "loss": 1.2957, + "step": 10213 + }, + { + "epoch": 0.25, + "learning_rate": 1.7658997875569782e-05, + "loss": 1.1327, + "step": 10214 + }, + { + "epoch": 0.25, + "learning_rate": 1.7658497054425254e-05, + "loss": 0.8573, + "step": 10215 + }, + { + "epoch": 0.25, + "learning_rate": 1.7657996186818427e-05, + "loss": 1.0209, + "step": 10216 + }, + { + "epoch": 0.25, + "learning_rate": 1.765749527275234e-05, + "loss": 1.1649, + "step": 10217 + }, + { + "epoch": 0.25, + "learning_rate": 1.7656994312230023e-05, + "loss": 0.9231, + "step": 10218 + }, + { + "epoch": 0.25, + "learning_rate": 1.765649330525453e-05, + "loss": 1.1506, + "step": 10219 + }, + { + "epoch": 0.25, + "learning_rate": 1.765599225182889e-05, + "loss": 1.0806, + "step": 10220 + }, + { + "epoch": 0.25, + "learning_rate": 1.7655491151956145e-05, + "loss": 1.0681, + "step": 10221 + }, + { + "epoch": 0.25, + "learning_rate": 1.7654990005639338e-05, + "loss": 0.8918, + "step": 10222 + }, + { + "epoch": 0.25, + "learning_rate": 1.7654488812881506e-05, + "loss": 1.0472, + "step": 10223 + }, + { + "epoch": 0.25, + "learning_rate": 1.765398757368569e-05, + "loss": 1.0847, + "step": 10224 + }, + { + "epoch": 0.25, + "learning_rate": 1.7653486288054934e-05, + "loss": 1.2942, + "step": 10225 + }, + { + "epoch": 0.25, + "learning_rate": 1.7652984955992275e-05, + "loss": 0.9138, + "step": 10226 + }, + { + "epoch": 0.25, + "learning_rate": 1.765248357750076e-05, + "loss": 1.3508, + "step": 10227 + }, + { + "epoch": 0.25, + "learning_rate": 1.765198215258342e-05, + "loss": 1.3215, + "step": 10228 + }, + { + "epoch": 0.25, + "learning_rate": 1.765148068124331e-05, + "loss": 1.2454, + "step": 10229 + }, + { + "epoch": 0.25, + "learning_rate": 1.7650979163483464e-05, + "loss": 1.2159, + "step": 10230 + }, + { + "epoch": 0.25, + "learning_rate": 1.765047759930693e-05, + "loss": 1.1771, + "step": 10231 + }, + { + "epoch": 0.25, + "learning_rate": 1.7649975988716745e-05, + "loss": 1.1583, + "step": 10232 + }, + { + "epoch": 0.25, + "learning_rate": 1.7649474331715957e-05, + "loss": 1.1418, + "step": 10233 + }, + { + "epoch": 0.25, + "learning_rate": 1.7648972628307607e-05, + "loss": 1.1929, + "step": 10234 + }, + { + "epoch": 0.25, + "learning_rate": 1.7648470878494736e-05, + "loss": 0.9831, + "step": 10235 + }, + { + "epoch": 0.25, + "learning_rate": 1.7647969082280397e-05, + "loss": 1.0126, + "step": 10236 + }, + { + "epoch": 0.25, + "learning_rate": 1.7647467239667623e-05, + "loss": 0.9627, + "step": 10237 + }, + { + "epoch": 0.25, + "learning_rate": 1.7646965350659466e-05, + "loss": 1.1391, + "step": 10238 + }, + { + "epoch": 0.25, + "learning_rate": 1.764646341525897e-05, + "loss": 1.1629, + "step": 10239 + }, + { + "epoch": 0.25, + "learning_rate": 1.764596143346918e-05, + "loss": 1.1908, + "step": 10240 + }, + { + "epoch": 0.25, + "learning_rate": 1.7645459405293136e-05, + "loss": 1.2353, + "step": 10241 + }, + { + "epoch": 0.25, + "learning_rate": 1.764495733073389e-05, + "loss": 1.3575, + "step": 10242 + }, + { + "epoch": 0.25, + "learning_rate": 1.764445520979449e-05, + "loss": 1.0808, + "step": 10243 + }, + { + "epoch": 0.25, + "learning_rate": 1.764395304247797e-05, + "loss": 1.3268, + "step": 10244 + }, + { + "epoch": 0.25, + "learning_rate": 1.764345082878739e-05, + "loss": 1.1546, + "step": 10245 + }, + { + "epoch": 0.25, + "learning_rate": 1.7642948568725796e-05, + "loss": 0.899, + "step": 10246 + }, + { + "epoch": 0.25, + "learning_rate": 1.7642446262296225e-05, + "loss": 0.9706, + "step": 10247 + }, + { + "epoch": 0.25, + "learning_rate": 1.7641943909501734e-05, + "loss": 0.9723, + "step": 10248 + }, + { + "epoch": 0.25, + "learning_rate": 1.7641441510345363e-05, + "loss": 0.891, + "step": 10249 + }, + { + "epoch": 0.25, + "learning_rate": 1.7640939064830166e-05, + "loss": 0.9772, + "step": 10250 + }, + { + "epoch": 0.25, + "learning_rate": 1.7640436572959188e-05, + "loss": 1.1641, + "step": 10251 + }, + { + "epoch": 0.25, + "learning_rate": 1.7639934034735476e-05, + "loss": 0.9643, + "step": 10252 + }, + { + "epoch": 0.25, + "learning_rate": 1.7639431450162084e-05, + "loss": 1.178, + "step": 10253 + }, + { + "epoch": 0.25, + "learning_rate": 1.7638928819242062e-05, + "loss": 1.4465, + "step": 10254 + }, + { + "epoch": 0.25, + "learning_rate": 1.7638426141978448e-05, + "loss": 1.1046, + "step": 10255 + }, + { + "epoch": 0.25, + "learning_rate": 1.7637923418374305e-05, + "loss": 1.0404, + "step": 10256 + }, + { + "epoch": 0.25, + "learning_rate": 1.7637420648432676e-05, + "loss": 0.9431, + "step": 10257 + }, + { + "epoch": 0.25, + "learning_rate": 1.763691783215661e-05, + "loss": 1.0662, + "step": 10258 + }, + { + "epoch": 0.25, + "learning_rate": 1.7636414969549163e-05, + "loss": 1.4452, + "step": 10259 + }, + { + "epoch": 0.25, + "learning_rate": 1.7635912060613377e-05, + "loss": 1.1125, + "step": 10260 + }, + { + "epoch": 0.25, + "learning_rate": 1.7635409105352313e-05, + "loss": 1.3728, + "step": 10261 + }, + { + "epoch": 0.25, + "learning_rate": 1.763490610376902e-05, + "loss": 1.0292, + "step": 10262 + }, + { + "epoch": 0.25, + "learning_rate": 1.7634403055866546e-05, + "loss": 1.308, + "step": 10263 + }, + { + "epoch": 0.25, + "learning_rate": 1.763389996164794e-05, + "loss": 1.0801, + "step": 10264 + }, + { + "epoch": 0.25, + "learning_rate": 1.7633396821116262e-05, + "loss": 1.1432, + "step": 10265 + }, + { + "epoch": 0.25, + "learning_rate": 1.763289363427456e-05, + "loss": 1.2462, + "step": 10266 + }, + { + "epoch": 0.25, + "learning_rate": 1.763239040112589e-05, + "loss": 1.0451, + "step": 10267 + }, + { + "epoch": 0.25, + "learning_rate": 1.7631887121673302e-05, + "loss": 1.1902, + "step": 10268 + }, + { + "epoch": 0.25, + "learning_rate": 1.7631383795919847e-05, + "loss": 1.0623, + "step": 10269 + }, + { + "epoch": 0.25, + "learning_rate": 1.7630880423868585e-05, + "loss": 1.3084, + "step": 10270 + }, + { + "epoch": 0.25, + "learning_rate": 1.7630377005522567e-05, + "loss": 1.2047, + "step": 10271 + }, + { + "epoch": 0.25, + "learning_rate": 1.7629873540884843e-05, + "loss": 1.3852, + "step": 10272 + }, + { + "epoch": 0.25, + "learning_rate": 1.7629370029958473e-05, + "loss": 1.4009, + "step": 10273 + }, + { + "epoch": 0.25, + "learning_rate": 1.762886647274651e-05, + "loss": 1.0911, + "step": 10274 + }, + { + "epoch": 0.25, + "learning_rate": 1.7628362869252005e-05, + "loss": 1.1731, + "step": 10275 + }, + { + "epoch": 0.25, + "learning_rate": 1.762785921947802e-05, + "loss": 1.2427, + "step": 10276 + }, + { + "epoch": 0.25, + "learning_rate": 1.7627355523427608e-05, + "loss": 0.9102, + "step": 10277 + }, + { + "epoch": 0.25, + "learning_rate": 1.762685178110382e-05, + "loss": 1.0508, + "step": 10278 + }, + { + "epoch": 0.25, + "learning_rate": 1.762634799250972e-05, + "loss": 1.0403, + "step": 10279 + }, + { + "epoch": 0.25, + "learning_rate": 1.762584415764836e-05, + "loss": 1.0839, + "step": 10280 + }, + { + "epoch": 0.25, + "learning_rate": 1.7625340276522798e-05, + "loss": 1.3532, + "step": 10281 + }, + { + "epoch": 0.25, + "learning_rate": 1.7624836349136088e-05, + "loss": 1.1165, + "step": 10282 + }, + { + "epoch": 0.25, + "learning_rate": 1.762433237549129e-05, + "loss": 1.2199, + "step": 10283 + }, + { + "epoch": 0.25, + "learning_rate": 1.762382835559146e-05, + "loss": 1.0711, + "step": 10284 + }, + { + "epoch": 0.25, + "learning_rate": 1.7623324289439656e-05, + "loss": 1.1328, + "step": 10285 + }, + { + "epoch": 0.25, + "learning_rate": 1.762282017703894e-05, + "loss": 1.2102, + "step": 10286 + }, + { + "epoch": 0.25, + "learning_rate": 1.7622316018392364e-05, + "loss": 0.9219, + "step": 10287 + }, + { + "epoch": 0.25, + "learning_rate": 1.762181181350299e-05, + "loss": 1.1334, + "step": 10288 + }, + { + "epoch": 0.25, + "learning_rate": 1.762130756237388e-05, + "loss": 1.3246, + "step": 10289 + }, + { + "epoch": 0.25, + "learning_rate": 1.7620803265008085e-05, + "loss": 1.1271, + "step": 10290 + }, + { + "epoch": 0.25, + "learning_rate": 1.7620298921408673e-05, + "loss": 1.1146, + "step": 10291 + }, + { + "epoch": 0.25, + "learning_rate": 1.7619794531578698e-05, + "loss": 1.2256, + "step": 10292 + }, + { + "epoch": 0.25, + "learning_rate": 1.761929009552122e-05, + "loss": 1.2603, + "step": 10293 + }, + { + "epoch": 0.25, + "learning_rate": 1.7618785613239305e-05, + "loss": 1.1981, + "step": 10294 + }, + { + "epoch": 0.25, + "learning_rate": 1.7618281084736005e-05, + "loss": 1.0437, + "step": 10295 + }, + { + "epoch": 0.25, + "learning_rate": 1.761777651001439e-05, + "loss": 1.0269, + "step": 10296 + }, + { + "epoch": 0.25, + "learning_rate": 1.7617271889077514e-05, + "loss": 1.0799, + "step": 10297 + }, + { + "epoch": 0.25, + "learning_rate": 1.7616767221928443e-05, + "loss": 1.159, + "step": 10298 + }, + { + "epoch": 0.25, + "learning_rate": 1.7616262508570235e-05, + "loss": 1.0424, + "step": 10299 + }, + { + "epoch": 0.25, + "learning_rate": 1.7615757749005954e-05, + "loss": 1.012, + "step": 10300 + }, + { + "epoch": 0.25, + "learning_rate": 1.7615252943238664e-05, + "loss": 1.3379, + "step": 10301 + }, + { + "epoch": 0.25, + "learning_rate": 1.7614748091271426e-05, + "loss": 1.2043, + "step": 10302 + }, + { + "epoch": 0.25, + "learning_rate": 1.76142431931073e-05, + "loss": 1.1336, + "step": 10303 + }, + { + "epoch": 0.25, + "learning_rate": 1.761373824874935e-05, + "loss": 1.0325, + "step": 10304 + }, + { + "epoch": 0.25, + "learning_rate": 1.7613233258200643e-05, + "loss": 1.1483, + "step": 10305 + }, + { + "epoch": 0.25, + "learning_rate": 1.7612728221464243e-05, + "loss": 1.0253, + "step": 10306 + }, + { + "epoch": 0.25, + "learning_rate": 1.761222313854321e-05, + "loss": 1.0978, + "step": 10307 + }, + { + "epoch": 0.25, + "learning_rate": 1.7611718009440604e-05, + "loss": 1.0866, + "step": 10308 + }, + { + "epoch": 0.25, + "learning_rate": 1.7611212834159505e-05, + "loss": 1.2367, + "step": 10309 + }, + { + "epoch": 0.25, + "learning_rate": 1.7610707612702963e-05, + "loss": 1.3066, + "step": 10310 + }, + { + "epoch": 0.25, + "learning_rate": 1.7610202345074045e-05, + "loss": 1.1524, + "step": 10311 + }, + { + "epoch": 0.25, + "learning_rate": 1.7609697031275824e-05, + "loss": 1.234, + "step": 10312 + }, + { + "epoch": 0.25, + "learning_rate": 1.760919167131136e-05, + "loss": 0.9781, + "step": 10313 + }, + { + "epoch": 0.25, + "learning_rate": 1.760868626518372e-05, + "loss": 1.2192, + "step": 10314 + }, + { + "epoch": 0.25, + "learning_rate": 1.7608180812895968e-05, + "loss": 1.1122, + "step": 10315 + }, + { + "epoch": 0.25, + "learning_rate": 1.760767531445118e-05, + "loss": 1.0282, + "step": 10316 + }, + { + "epoch": 0.25, + "learning_rate": 1.7607169769852404e-05, + "loss": 1.3971, + "step": 10317 + }, + { + "epoch": 0.25, + "learning_rate": 1.7606664179102725e-05, + "loss": 1.269, + "step": 10318 + }, + { + "epoch": 0.25, + "learning_rate": 1.7606158542205203e-05, + "loss": 1.1135, + "step": 10319 + }, + { + "epoch": 0.25, + "learning_rate": 1.7605652859162908e-05, + "loss": 1.2766, + "step": 10320 + }, + { + "epoch": 0.25, + "learning_rate": 1.7605147129978904e-05, + "loss": 1.0405, + "step": 10321 + }, + { + "epoch": 0.25, + "learning_rate": 1.760464135465626e-05, + "loss": 1.2183, + "step": 10322 + }, + { + "epoch": 0.25, + "learning_rate": 1.7604135533198047e-05, + "loss": 0.9944, + "step": 10323 + }, + { + "epoch": 0.25, + "learning_rate": 1.760362966560733e-05, + "loss": 1.0555, + "step": 10324 + }, + { + "epoch": 0.25, + "learning_rate": 1.7603123751887184e-05, + "loss": 1.0136, + "step": 10325 + }, + { + "epoch": 0.25, + "learning_rate": 1.760261779204067e-05, + "loss": 1.1687, + "step": 10326 + }, + { + "epoch": 0.25, + "learning_rate": 1.7602111786070867e-05, + "loss": 1.2625, + "step": 10327 + }, + { + "epoch": 0.25, + "learning_rate": 1.7601605733980837e-05, + "loss": 1.0073, + "step": 10328 + }, + { + "epoch": 0.25, + "learning_rate": 1.7601099635773654e-05, + "loss": 1.1487, + "step": 10329 + }, + { + "epoch": 0.25, + "learning_rate": 1.760059349145239e-05, + "loss": 1.3894, + "step": 10330 + }, + { + "epoch": 0.25, + "learning_rate": 1.760008730102011e-05, + "loss": 1.3104, + "step": 10331 + }, + { + "epoch": 0.25, + "learning_rate": 1.7599581064479886e-05, + "loss": 1.2925, + "step": 10332 + }, + { + "epoch": 0.25, + "learning_rate": 1.7599074781834796e-05, + "loss": 1.0034, + "step": 10333 + }, + { + "epoch": 0.25, + "learning_rate": 1.7598568453087902e-05, + "loss": 1.1985, + "step": 10334 + }, + { + "epoch": 0.25, + "learning_rate": 1.7598062078242283e-05, + "loss": 1.3996, + "step": 10335 + }, + { + "epoch": 0.25, + "learning_rate": 1.759755565730101e-05, + "loss": 1.0697, + "step": 10336 + }, + { + "epoch": 0.25, + "learning_rate": 1.759704919026715e-05, + "loss": 1.2138, + "step": 10337 + }, + { + "epoch": 0.25, + "learning_rate": 1.759654267714378e-05, + "loss": 1.1425, + "step": 10338 + }, + { + "epoch": 0.25, + "learning_rate": 1.7596036117933977e-05, + "loss": 1.1188, + "step": 10339 + }, + { + "epoch": 0.25, + "learning_rate": 1.7595529512640807e-05, + "loss": 1.1812, + "step": 10340 + }, + { + "epoch": 0.25, + "learning_rate": 1.7595022861267346e-05, + "loss": 0.9708, + "step": 10341 + }, + { + "epoch": 0.25, + "learning_rate": 1.7594516163816666e-05, + "loss": 1.1715, + "step": 10342 + }, + { + "epoch": 0.25, + "learning_rate": 1.7594009420291846e-05, + "loss": 1.4051, + "step": 10343 + }, + { + "epoch": 0.25, + "learning_rate": 1.7593502630695954e-05, + "loss": 1.1629, + "step": 10344 + }, + { + "epoch": 0.25, + "learning_rate": 1.759299579503207e-05, + "loss": 1.1, + "step": 10345 + }, + { + "epoch": 0.25, + "learning_rate": 1.7592488913303265e-05, + "loss": 1.1864, + "step": 10346 + }, + { + "epoch": 0.25, + "learning_rate": 1.7591981985512617e-05, + "loss": 1.1241, + "step": 10347 + }, + { + "epoch": 0.25, + "learning_rate": 1.7591475011663197e-05, + "loss": 1.4306, + "step": 10348 + }, + { + "epoch": 0.25, + "learning_rate": 1.759096799175809e-05, + "loss": 1.2431, + "step": 10349 + }, + { + "epoch": 0.25, + "learning_rate": 1.7590460925800358e-05, + "loss": 1.1896, + "step": 10350 + }, + { + "epoch": 0.25, + "learning_rate": 1.7589953813793088e-05, + "loss": 0.9144, + "step": 10351 + }, + { + "epoch": 0.25, + "learning_rate": 1.7589446655739352e-05, + "loss": 0.8418, + "step": 10352 + }, + { + "epoch": 0.25, + "learning_rate": 1.758893945164223e-05, + "loss": 1.4507, + "step": 10353 + }, + { + "epoch": 0.25, + "learning_rate": 1.7588432201504798e-05, + "loss": 1.1814, + "step": 10354 + }, + { + "epoch": 0.25, + "learning_rate": 1.7587924905330133e-05, + "loss": 1.3006, + "step": 10355 + }, + { + "epoch": 0.25, + "learning_rate": 1.7587417563121314e-05, + "loss": 1.1533, + "step": 10356 + }, + { + "epoch": 0.25, + "learning_rate": 1.7586910174881414e-05, + "loss": 1.168, + "step": 10357 + }, + { + "epoch": 0.25, + "learning_rate": 1.7586402740613514e-05, + "loss": 1.2554, + "step": 10358 + }, + { + "epoch": 0.25, + "learning_rate": 1.7585895260320696e-05, + "loss": 1.1846, + "step": 10359 + }, + { + "epoch": 0.25, + "learning_rate": 1.7585387734006033e-05, + "loss": 0.8164, + "step": 10360 + }, + { + "epoch": 0.25, + "learning_rate": 1.758488016167261e-05, + "loss": 1.0242, + "step": 10361 + }, + { + "epoch": 0.25, + "learning_rate": 1.75843725433235e-05, + "loss": 1.2568, + "step": 10362 + }, + { + "epoch": 0.25, + "learning_rate": 1.758386487896179e-05, + "loss": 1.1421, + "step": 10363 + }, + { + "epoch": 0.25, + "learning_rate": 1.7583357168590552e-05, + "loss": 1.2593, + "step": 10364 + }, + { + "epoch": 0.25, + "learning_rate": 1.758284941221287e-05, + "loss": 0.9065, + "step": 10365 + }, + { + "epoch": 0.25, + "learning_rate": 1.7582341609831824e-05, + "loss": 1.2738, + "step": 10366 + }, + { + "epoch": 0.25, + "learning_rate": 1.7581833761450495e-05, + "loss": 0.8894, + "step": 10367 + }, + { + "epoch": 0.25, + "learning_rate": 1.7581325867071967e-05, + "loss": 0.9853, + "step": 10368 + }, + { + "epoch": 0.25, + "learning_rate": 1.7580817926699312e-05, + "loss": 1.37, + "step": 10369 + }, + { + "epoch": 0.25, + "learning_rate": 1.758030994033562e-05, + "loss": 1.3255, + "step": 10370 + }, + { + "epoch": 0.25, + "learning_rate": 1.7579801907983975e-05, + "loss": 1.0792, + "step": 10371 + }, + { + "epoch": 0.25, + "learning_rate": 1.7579293829647454e-05, + "loss": 1.0618, + "step": 10372 + }, + { + "epoch": 0.25, + "learning_rate": 1.7578785705329137e-05, + "loss": 1.0847, + "step": 10373 + }, + { + "epoch": 0.25, + "learning_rate": 1.757827753503211e-05, + "loss": 1.1185, + "step": 10374 + }, + { + "epoch": 0.25, + "learning_rate": 1.7577769318759455e-05, + "loss": 1.1098, + "step": 10375 + }, + { + "epoch": 0.25, + "learning_rate": 1.7577261056514257e-05, + "loss": 1.2498, + "step": 10376 + }, + { + "epoch": 0.25, + "learning_rate": 1.7576752748299598e-05, + "loss": 1.1391, + "step": 10377 + }, + { + "epoch": 0.25, + "learning_rate": 1.7576244394118562e-05, + "loss": 1.307, + "step": 10378 + }, + { + "epoch": 0.25, + "learning_rate": 1.7575735993974235e-05, + "loss": 1.1708, + "step": 10379 + }, + { + "epoch": 0.25, + "learning_rate": 1.75752275478697e-05, + "loss": 1.2377, + "step": 10380 + }, + { + "epoch": 0.25, + "learning_rate": 1.757471905580804e-05, + "loss": 1.1695, + "step": 10381 + }, + { + "epoch": 0.25, + "learning_rate": 1.757421051779234e-05, + "loss": 1.0532, + "step": 10382 + }, + { + "epoch": 0.25, + "learning_rate": 1.757370193382569e-05, + "loss": 1.3055, + "step": 10383 + }, + { + "epoch": 0.25, + "learning_rate": 1.7573193303911167e-05, + "loss": 0.9577, + "step": 10384 + }, + { + "epoch": 0.25, + "learning_rate": 1.7572684628051866e-05, + "loss": 1.1802, + "step": 10385 + }, + { + "epoch": 0.25, + "learning_rate": 1.7572175906250866e-05, + "loss": 0.9482, + "step": 10386 + }, + { + "epoch": 0.25, + "learning_rate": 1.7571667138511255e-05, + "loss": 1.0595, + "step": 10387 + }, + { + "epoch": 0.25, + "learning_rate": 1.7571158324836123e-05, + "loss": 1.1338, + "step": 10388 + }, + { + "epoch": 0.25, + "learning_rate": 1.757064946522855e-05, + "loss": 1.0837, + "step": 10389 + }, + { + "epoch": 0.25, + "learning_rate": 1.7570140559691635e-05, + "loss": 1.119, + "step": 10390 + }, + { + "epoch": 0.25, + "learning_rate": 1.7569631608228454e-05, + "loss": 1.2277, + "step": 10391 + }, + { + "epoch": 0.25, + "learning_rate": 1.7569122610842097e-05, + "loss": 0.7813, + "step": 10392 + }, + { + "epoch": 0.25, + "learning_rate": 1.756861356753565e-05, + "loss": 0.8796, + "step": 10393 + }, + { + "epoch": 0.25, + "learning_rate": 1.7568104478312214e-05, + "loss": 1.0162, + "step": 10394 + }, + { + "epoch": 0.25, + "learning_rate": 1.756759534317486e-05, + "loss": 1.0685, + "step": 10395 + }, + { + "epoch": 0.25, + "learning_rate": 1.756708616212669e-05, + "loss": 1.1211, + "step": 10396 + }, + { + "epoch": 0.25, + "learning_rate": 1.7566576935170788e-05, + "loss": 1.1406, + "step": 10397 + }, + { + "epoch": 0.25, + "learning_rate": 1.756606766231024e-05, + "loss": 1.2986, + "step": 10398 + }, + { + "epoch": 0.25, + "learning_rate": 1.7565558343548142e-05, + "loss": 0.734, + "step": 10399 + }, + { + "epoch": 0.25, + "learning_rate": 1.756504897888758e-05, + "loss": 0.9499, + "step": 10400 + }, + { + "epoch": 0.25, + "learning_rate": 1.7564539568331648e-05, + "loss": 0.9888, + "step": 10401 + }, + { + "epoch": 0.25, + "learning_rate": 1.756403011188343e-05, + "loss": 1.0939, + "step": 10402 + }, + { + "epoch": 0.25, + "learning_rate": 1.756352060954602e-05, + "loss": 1.4031, + "step": 10403 + }, + { + "epoch": 0.25, + "learning_rate": 1.7563011061322513e-05, + "loss": 1.1291, + "step": 10404 + }, + { + "epoch": 0.25, + "learning_rate": 1.7562501467215997e-05, + "loss": 1.1724, + "step": 10405 + }, + { + "epoch": 0.25, + "learning_rate": 1.756199182722956e-05, + "loss": 1.3726, + "step": 10406 + }, + { + "epoch": 0.25, + "learning_rate": 1.75614821413663e-05, + "loss": 1.1733, + "step": 10407 + }, + { + "epoch": 0.25, + "learning_rate": 1.75609724096293e-05, + "loss": 1.0294, + "step": 10408 + }, + { + "epoch": 0.25, + "learning_rate": 1.756046263202167e-05, + "loss": 1.2418, + "step": 10409 + }, + { + "epoch": 0.25, + "learning_rate": 1.755995280854648e-05, + "loss": 0.7735, + "step": 10410 + }, + { + "epoch": 0.25, + "learning_rate": 1.7559442939206844e-05, + "loss": 1.3563, + "step": 10411 + }, + { + "epoch": 0.25, + "learning_rate": 1.755893302400584e-05, + "loss": 1.4366, + "step": 10412 + }, + { + "epoch": 0.25, + "learning_rate": 1.7558423062946566e-05, + "loss": 0.8523, + "step": 10413 + }, + { + "epoch": 0.25, + "learning_rate": 1.7557913056032126e-05, + "loss": 1.3649, + "step": 10414 + }, + { + "epoch": 0.25, + "learning_rate": 1.7557403003265597e-05, + "loss": 1.246, + "step": 10415 + }, + { + "epoch": 0.25, + "learning_rate": 1.7556892904650086e-05, + "loss": 1.0135, + "step": 10416 + }, + { + "epoch": 0.25, + "learning_rate": 1.755638276018868e-05, + "loss": 1.1487, + "step": 10417 + }, + { + "epoch": 0.25, + "learning_rate": 1.7555872569884478e-05, + "loss": 0.8895, + "step": 10418 + }, + { + "epoch": 0.25, + "learning_rate": 1.7555362333740574e-05, + "loss": 1.0618, + "step": 10419 + }, + { + "epoch": 0.25, + "learning_rate": 1.7554852051760068e-05, + "loss": 0.6501, + "step": 10420 + }, + { + "epoch": 0.25, + "learning_rate": 1.7554341723946047e-05, + "loss": 1.3147, + "step": 10421 + }, + { + "epoch": 0.25, + "learning_rate": 1.755383135030161e-05, + "loss": 1.1061, + "step": 10422 + }, + { + "epoch": 0.25, + "learning_rate": 1.755332093082986e-05, + "loss": 1.2252, + "step": 10423 + }, + { + "epoch": 0.25, + "learning_rate": 1.755281046553389e-05, + "loss": 1.2063, + "step": 10424 + }, + { + "epoch": 0.25, + "learning_rate": 1.7552299954416788e-05, + "loss": 0.8622, + "step": 10425 + }, + { + "epoch": 0.25, + "learning_rate": 1.7551789397481666e-05, + "loss": 1.2641, + "step": 10426 + }, + { + "epoch": 0.25, + "learning_rate": 1.7551278794731606e-05, + "loss": 1.3497, + "step": 10427 + }, + { + "epoch": 0.25, + "learning_rate": 1.7550768146169722e-05, + "loss": 1.5047, + "step": 10428 + }, + { + "epoch": 0.25, + "learning_rate": 1.7550257451799096e-05, + "loss": 1.1597, + "step": 10429 + }, + { + "epoch": 0.25, + "learning_rate": 1.754974671162284e-05, + "loss": 1.3467, + "step": 10430 + }, + { + "epoch": 0.25, + "learning_rate": 1.7549235925644046e-05, + "loss": 1.3466, + "step": 10431 + }, + { + "epoch": 0.25, + "learning_rate": 1.754872509386581e-05, + "loss": 0.6874, + "step": 10432 + }, + { + "epoch": 0.25, + "learning_rate": 1.7548214216291236e-05, + "loss": 0.9949, + "step": 10433 + }, + { + "epoch": 0.25, + "learning_rate": 1.754770329292342e-05, + "loss": 0.9717, + "step": 10434 + }, + { + "epoch": 0.25, + "learning_rate": 1.7547192323765468e-05, + "loss": 0.9962, + "step": 10435 + }, + { + "epoch": 0.25, + "learning_rate": 1.754668130882047e-05, + "loss": 1.1, + "step": 10436 + }, + { + "epoch": 0.25, + "learning_rate": 1.7546170248091535e-05, + "loss": 1.1396, + "step": 10437 + }, + { + "epoch": 0.25, + "learning_rate": 1.7545659141581762e-05, + "loss": 1.1123, + "step": 10438 + }, + { + "epoch": 0.25, + "learning_rate": 1.7545147989294246e-05, + "loss": 1.4092, + "step": 10439 + }, + { + "epoch": 0.25, + "learning_rate": 1.7544636791232094e-05, + "loss": 1.156, + "step": 10440 + }, + { + "epoch": 0.25, + "learning_rate": 1.7544125547398402e-05, + "loss": 0.7029, + "step": 10441 + }, + { + "epoch": 0.25, + "learning_rate": 1.754361425779628e-05, + "loss": 1.0191, + "step": 10442 + }, + { + "epoch": 0.25, + "learning_rate": 1.7543102922428824e-05, + "loss": 1.2574, + "step": 10443 + }, + { + "epoch": 0.25, + "learning_rate": 1.754259154129913e-05, + "loss": 1.1675, + "step": 10444 + }, + { + "epoch": 0.25, + "learning_rate": 1.7542080114410314e-05, + "loss": 1.3222, + "step": 10445 + }, + { + "epoch": 0.25, + "learning_rate": 1.754156864176547e-05, + "loss": 1.0974, + "step": 10446 + }, + { + "epoch": 0.25, + "learning_rate": 1.7541057123367702e-05, + "loss": 1.4883, + "step": 10447 + }, + { + "epoch": 0.25, + "learning_rate": 1.7540545559220116e-05, + "loss": 1.2224, + "step": 10448 + }, + { + "epoch": 0.25, + "learning_rate": 1.754003394932581e-05, + "loss": 0.9263, + "step": 10449 + }, + { + "epoch": 0.25, + "learning_rate": 1.75395222936879e-05, + "loss": 1.1089, + "step": 10450 + }, + { + "epoch": 0.25, + "learning_rate": 1.7539010592309472e-05, + "loss": 1.0056, + "step": 10451 + }, + { + "epoch": 0.25, + "learning_rate": 1.7538498845193645e-05, + "loss": 1.2081, + "step": 10452 + }, + { + "epoch": 0.25, + "learning_rate": 1.7537987052343517e-05, + "loss": 1.6204, + "step": 10453 + }, + { + "epoch": 0.25, + "learning_rate": 1.7537475213762196e-05, + "loss": 0.9764, + "step": 10454 + }, + { + "epoch": 0.25, + "learning_rate": 1.7536963329452784e-05, + "loss": 1.1863, + "step": 10455 + }, + { + "epoch": 0.25, + "learning_rate": 1.753645139941839e-05, + "loss": 1.2175, + "step": 10456 + }, + { + "epoch": 0.25, + "learning_rate": 1.753593942366212e-05, + "loss": 0.9719, + "step": 10457 + }, + { + "epoch": 0.25, + "learning_rate": 1.753542740218707e-05, + "loss": 1.0918, + "step": 10458 + }, + { + "epoch": 0.25, + "learning_rate": 1.7534915334996362e-05, + "loss": 1.0401, + "step": 10459 + }, + { + "epoch": 0.25, + "learning_rate": 1.7534403222093093e-05, + "loss": 0.9073, + "step": 10460 + }, + { + "epoch": 0.25, + "learning_rate": 1.7533891063480373e-05, + "loss": 0.8702, + "step": 10461 + }, + { + "epoch": 0.25, + "learning_rate": 1.7533378859161304e-05, + "loss": 1.1668, + "step": 10462 + }, + { + "epoch": 0.25, + "learning_rate": 1.7532866609139004e-05, + "loss": 1.0892, + "step": 10463 + }, + { + "epoch": 0.25, + "learning_rate": 1.753235431341657e-05, + "loss": 1.2317, + "step": 10464 + }, + { + "epoch": 0.25, + "learning_rate": 1.7531841971997113e-05, + "loss": 0.9978, + "step": 10465 + }, + { + "epoch": 0.25, + "learning_rate": 1.7531329584883747e-05, + "loss": 1.1001, + "step": 10466 + }, + { + "epoch": 0.25, + "learning_rate": 1.753081715207957e-05, + "loss": 1.2184, + "step": 10467 + }, + { + "epoch": 0.25, + "learning_rate": 1.75303046735877e-05, + "loss": 0.9107, + "step": 10468 + }, + { + "epoch": 0.25, + "learning_rate": 1.752979214941124e-05, + "loss": 0.8622, + "step": 10469 + }, + { + "epoch": 0.25, + "learning_rate": 1.7529279579553303e-05, + "loss": 1.1793, + "step": 10470 + }, + { + "epoch": 0.25, + "learning_rate": 1.7528766964017e-05, + "loss": 0.8966, + "step": 10471 + }, + { + "epoch": 0.25, + "learning_rate": 1.752825430280544e-05, + "loss": 1.1356, + "step": 10472 + }, + { + "epoch": 0.25, + "learning_rate": 1.752774159592173e-05, + "loss": 1.0104, + "step": 10473 + }, + { + "epoch": 0.25, + "learning_rate": 1.752722884336898e-05, + "loss": 1.1291, + "step": 10474 + }, + { + "epoch": 0.25, + "learning_rate": 1.7526716045150305e-05, + "loss": 1.2544, + "step": 10475 + }, + { + "epoch": 0.25, + "learning_rate": 1.7526203201268814e-05, + "loss": 1.1942, + "step": 10476 + }, + { + "epoch": 0.25, + "learning_rate": 1.752569031172762e-05, + "loss": 1.1869, + "step": 10477 + }, + { + "epoch": 0.25, + "learning_rate": 1.7525177376529828e-05, + "loss": 1.1056, + "step": 10478 + }, + { + "epoch": 0.25, + "learning_rate": 1.752466439567856e-05, + "loss": 1.1879, + "step": 10479 + }, + { + "epoch": 0.25, + "learning_rate": 1.752415136917692e-05, + "loss": 1.2369, + "step": 10480 + }, + { + "epoch": 0.25, + "learning_rate": 1.7523638297028024e-05, + "loss": 1.2163, + "step": 10481 + }, + { + "epoch": 0.25, + "learning_rate": 1.7523125179234987e-05, + "loss": 1.088, + "step": 10482 + }, + { + "epoch": 0.25, + "learning_rate": 1.752261201580092e-05, + "loss": 1.0726, + "step": 10483 + }, + { + "epoch": 0.25, + "learning_rate": 1.752209880672893e-05, + "loss": 0.8858, + "step": 10484 + }, + { + "epoch": 0.25, + "learning_rate": 1.7521585552022137e-05, + "loss": 1.0134, + "step": 10485 + }, + { + "epoch": 0.25, + "learning_rate": 1.7521072251683657e-05, + "loss": 1.2779, + "step": 10486 + }, + { + "epoch": 0.25, + "learning_rate": 1.75205589057166e-05, + "loss": 0.9504, + "step": 10487 + }, + { + "epoch": 0.25, + "learning_rate": 1.7520045514124077e-05, + "loss": 1.0387, + "step": 10488 + }, + { + "epoch": 0.25, + "learning_rate": 1.751953207690921e-05, + "loss": 0.8602, + "step": 10489 + }, + { + "epoch": 0.25, + "learning_rate": 1.751901859407511e-05, + "loss": 1.0846, + "step": 10490 + }, + { + "epoch": 0.25, + "learning_rate": 1.7518505065624896e-05, + "loss": 0.9061, + "step": 10491 + }, + { + "epoch": 0.25, + "learning_rate": 1.7517991491561676e-05, + "loss": 1.062, + "step": 10492 + }, + { + "epoch": 0.25, + "learning_rate": 1.751747787188857e-05, + "loss": 1.1512, + "step": 10493 + }, + { + "epoch": 0.25, + "learning_rate": 1.7516964206608697e-05, + "loss": 1.0755, + "step": 10494 + }, + { + "epoch": 0.25, + "learning_rate": 1.751645049572517e-05, + "loss": 1.1862, + "step": 10495 + }, + { + "epoch": 0.25, + "learning_rate": 1.75159367392411e-05, + "loss": 1.1086, + "step": 10496 + }, + { + "epoch": 0.25, + "learning_rate": 1.7515422937159613e-05, + "loss": 1.2179, + "step": 10497 + }, + { + "epoch": 0.25, + "learning_rate": 1.7514909089483825e-05, + "loss": 1.0512, + "step": 10498 + }, + { + "epoch": 0.25, + "learning_rate": 1.751439519621685e-05, + "loss": 1.191, + "step": 10499 + }, + { + "epoch": 0.25, + "learning_rate": 1.7513881257361806e-05, + "loss": 1.342, + "step": 10500 + }, + { + "epoch": 0.25, + "learning_rate": 1.7513367272921812e-05, + "loss": 1.1649, + "step": 10501 + }, + { + "epoch": 0.25, + "learning_rate": 1.7512853242899987e-05, + "loss": 0.9853, + "step": 10502 + }, + { + "epoch": 0.25, + "learning_rate": 1.7512339167299448e-05, + "loss": 0.9845, + "step": 10503 + }, + { + "epoch": 0.25, + "learning_rate": 1.751182504612331e-05, + "loss": 1.341, + "step": 10504 + }, + { + "epoch": 0.25, + "learning_rate": 1.75113108793747e-05, + "loss": 1.265, + "step": 10505 + }, + { + "epoch": 0.25, + "learning_rate": 1.7510796667056736e-05, + "loss": 0.9381, + "step": 10506 + }, + { + "epoch": 0.25, + "learning_rate": 1.7510282409172538e-05, + "loss": 0.9805, + "step": 10507 + }, + { + "epoch": 0.25, + "learning_rate": 1.7509768105725215e-05, + "loss": 1.3459, + "step": 10508 + }, + { + "epoch": 0.25, + "learning_rate": 1.75092537567179e-05, + "loss": 1.0069, + "step": 10509 + }, + { + "epoch": 0.25, + "learning_rate": 1.7508739362153708e-05, + "loss": 1.1119, + "step": 10510 + }, + { + "epoch": 0.25, + "learning_rate": 1.7508224922035758e-05, + "loss": 1.125, + "step": 10511 + }, + { + "epoch": 0.25, + "learning_rate": 1.750771043636718e-05, + "loss": 1.0903, + "step": 10512 + }, + { + "epoch": 0.25, + "learning_rate": 1.7507195905151083e-05, + "loss": 1.1588, + "step": 10513 + }, + { + "epoch": 0.25, + "learning_rate": 1.7506681328390592e-05, + "loss": 1.0035, + "step": 10514 + }, + { + "epoch": 0.25, + "learning_rate": 1.7506166706088835e-05, + "loss": 1.1678, + "step": 10515 + }, + { + "epoch": 0.25, + "learning_rate": 1.750565203824893e-05, + "loss": 0.9011, + "step": 10516 + }, + { + "epoch": 0.25, + "learning_rate": 1.7505137324874e-05, + "loss": 1.1379, + "step": 10517 + }, + { + "epoch": 0.25, + "learning_rate": 1.7504622565967166e-05, + "loss": 1.1641, + "step": 10518 + }, + { + "epoch": 0.25, + "learning_rate": 1.750410776153155e-05, + "loss": 1.2779, + "step": 10519 + }, + { + "epoch": 0.25, + "learning_rate": 1.7503592911570283e-05, + "loss": 1.1007, + "step": 10520 + }, + { + "epoch": 0.25, + "learning_rate": 1.750307801608648e-05, + "loss": 1.0578, + "step": 10521 + }, + { + "epoch": 0.25, + "learning_rate": 1.7502563075083267e-05, + "loss": 1.2183, + "step": 10522 + }, + { + "epoch": 0.25, + "learning_rate": 1.7502048088563767e-05, + "loss": 1.0759, + "step": 10523 + }, + { + "epoch": 0.25, + "learning_rate": 1.7501533056531106e-05, + "loss": 1.0547, + "step": 10524 + }, + { + "epoch": 0.25, + "learning_rate": 1.750101797898841e-05, + "loss": 1.2154, + "step": 10525 + }, + { + "epoch": 0.25, + "learning_rate": 1.7500502855938802e-05, + "loss": 1.203, + "step": 10526 + }, + { + "epoch": 0.25, + "learning_rate": 1.7499987687385407e-05, + "loss": 1.0873, + "step": 10527 + }, + { + "epoch": 0.25, + "learning_rate": 1.7499472473331353e-05, + "loss": 1.1315, + "step": 10528 + }, + { + "epoch": 0.25, + "learning_rate": 1.749895721377976e-05, + "loss": 0.9916, + "step": 10529 + }, + { + "epoch": 0.25, + "learning_rate": 1.749844190873376e-05, + "loss": 1.1765, + "step": 10530 + }, + { + "epoch": 0.25, + "learning_rate": 1.7497926558196476e-05, + "loss": 1.1176, + "step": 10531 + }, + { + "epoch": 0.25, + "learning_rate": 1.7497411162171034e-05, + "loss": 1.2856, + "step": 10532 + }, + { + "epoch": 0.25, + "learning_rate": 1.7496895720660566e-05, + "loss": 1.2279, + "step": 10533 + }, + { + "epoch": 0.25, + "learning_rate": 1.749638023366819e-05, + "loss": 1.1342, + "step": 10534 + }, + { + "epoch": 0.25, + "learning_rate": 1.749586470119704e-05, + "loss": 1.2052, + "step": 10535 + }, + { + "epoch": 0.25, + "learning_rate": 1.7495349123250242e-05, + "loss": 1.2417, + "step": 10536 + }, + { + "epoch": 0.25, + "learning_rate": 1.7494833499830924e-05, + "loss": 1.2677, + "step": 10537 + }, + { + "epoch": 0.25, + "learning_rate": 1.7494317830942214e-05, + "loss": 1.2789, + "step": 10538 + }, + { + "epoch": 0.25, + "learning_rate": 1.7493802116587245e-05, + "loss": 1.1987, + "step": 10539 + }, + { + "epoch": 0.25, + "learning_rate": 1.7493286356769135e-05, + "loss": 1.0704, + "step": 10540 + }, + { + "epoch": 0.25, + "learning_rate": 1.749277055149102e-05, + "loss": 1.1154, + "step": 10541 + }, + { + "epoch": 0.25, + "learning_rate": 1.7492254700756032e-05, + "loss": 1.0606, + "step": 10542 + }, + { + "epoch": 0.25, + "learning_rate": 1.7491738804567294e-05, + "loss": 1.044, + "step": 10543 + }, + { + "epoch": 0.25, + "learning_rate": 1.749122286292794e-05, + "loss": 1.1628, + "step": 10544 + }, + { + "epoch": 0.25, + "learning_rate": 1.74907068758411e-05, + "loss": 0.9755, + "step": 10545 + }, + { + "epoch": 0.25, + "learning_rate": 1.74901908433099e-05, + "loss": 1.2287, + "step": 10546 + }, + { + "epoch": 0.25, + "learning_rate": 1.7489674765337475e-05, + "loss": 1.0177, + "step": 10547 + }, + { + "epoch": 0.25, + "learning_rate": 1.7489158641926955e-05, + "loss": 1.1499, + "step": 10548 + }, + { + "epoch": 0.25, + "learning_rate": 1.7488642473081473e-05, + "loss": 1.1431, + "step": 10549 + }, + { + "epoch": 0.25, + "learning_rate": 1.748812625880416e-05, + "loss": 1.337, + "step": 10550 + }, + { + "epoch": 0.25, + "learning_rate": 1.748760999909814e-05, + "loss": 1.4824, + "step": 10551 + }, + { + "epoch": 0.25, + "learning_rate": 1.7487093693966553e-05, + "loss": 0.8611, + "step": 10552 + }, + { + "epoch": 0.25, + "learning_rate": 1.7486577343412534e-05, + "loss": 1.0575, + "step": 10553 + }, + { + "epoch": 0.25, + "learning_rate": 1.748606094743921e-05, + "loss": 1.1743, + "step": 10554 + }, + { + "epoch": 0.25, + "learning_rate": 1.748554450604971e-05, + "loss": 0.8339, + "step": 10555 + }, + { + "epoch": 0.25, + "learning_rate": 1.7485028019247176e-05, + "loss": 0.9617, + "step": 10556 + }, + { + "epoch": 0.25, + "learning_rate": 1.7484511487034738e-05, + "loss": 1.2516, + "step": 10557 + }, + { + "epoch": 0.25, + "learning_rate": 1.748399490941553e-05, + "loss": 1.1261, + "step": 10558 + }, + { + "epoch": 0.25, + "learning_rate": 1.7483478286392684e-05, + "loss": 1.0523, + "step": 10559 + }, + { + "epoch": 0.25, + "learning_rate": 1.7482961617969338e-05, + "loss": 1.0835, + "step": 10560 + }, + { + "epoch": 0.25, + "learning_rate": 1.748244490414862e-05, + "loss": 1.1271, + "step": 10561 + }, + { + "epoch": 0.25, + "learning_rate": 1.748192814493367e-05, + "loss": 1.2099, + "step": 10562 + }, + { + "epoch": 0.25, + "learning_rate": 1.748141134032762e-05, + "loss": 1.2031, + "step": 10563 + }, + { + "epoch": 0.25, + "learning_rate": 1.748089449033361e-05, + "loss": 1.3618, + "step": 10564 + }, + { + "epoch": 0.25, + "learning_rate": 1.7480377594954773e-05, + "loss": 0.7606, + "step": 10565 + }, + { + "epoch": 0.25, + "learning_rate": 1.7479860654194244e-05, + "loss": 1.1069, + "step": 10566 + }, + { + "epoch": 0.25, + "learning_rate": 1.7479343668055158e-05, + "loss": 1.3156, + "step": 10567 + }, + { + "epoch": 0.25, + "learning_rate": 1.7478826636540653e-05, + "loss": 1.1309, + "step": 10568 + }, + { + "epoch": 0.25, + "learning_rate": 1.747830955965387e-05, + "loss": 1.0795, + "step": 10569 + }, + { + "epoch": 0.25, + "learning_rate": 1.747779243739794e-05, + "loss": 1.1612, + "step": 10570 + }, + { + "epoch": 0.25, + "learning_rate": 1.7477275269776e-05, + "loss": 1.2824, + "step": 10571 + }, + { + "epoch": 0.25, + "learning_rate": 1.747675805679119e-05, + "loss": 1.1944, + "step": 10572 + }, + { + "epoch": 0.25, + "learning_rate": 1.747624079844665e-05, + "loss": 1.0152, + "step": 10573 + }, + { + "epoch": 0.25, + "learning_rate": 1.7475723494745516e-05, + "loss": 1.0171, + "step": 10574 + }, + { + "epoch": 0.25, + "learning_rate": 1.7475206145690925e-05, + "loss": 0.8242, + "step": 10575 + }, + { + "epoch": 0.25, + "learning_rate": 1.7474688751286014e-05, + "loss": 1.0884, + "step": 10576 + }, + { + "epoch": 0.25, + "learning_rate": 1.7474171311533927e-05, + "loss": 1.1225, + "step": 10577 + }, + { + "epoch": 0.25, + "learning_rate": 1.74736538264378e-05, + "loss": 0.9866, + "step": 10578 + }, + { + "epoch": 0.25, + "learning_rate": 1.747313629600077e-05, + "loss": 1.2199, + "step": 10579 + }, + { + "epoch": 0.25, + "learning_rate": 1.7472618720225988e-05, + "loss": 0.9405, + "step": 10580 + }, + { + "epoch": 0.25, + "learning_rate": 1.7472101099116578e-05, + "loss": 1.0562, + "step": 10581 + }, + { + "epoch": 0.25, + "learning_rate": 1.7471583432675692e-05, + "loss": 1.1499, + "step": 10582 + }, + { + "epoch": 0.25, + "learning_rate": 1.7471065720906465e-05, + "loss": 1.2888, + "step": 10583 + }, + { + "epoch": 0.25, + "learning_rate": 1.747054796381204e-05, + "loss": 1.0628, + "step": 10584 + }, + { + "epoch": 0.25, + "learning_rate": 1.7470030161395558e-05, + "loss": 0.9173, + "step": 10585 + }, + { + "epoch": 0.25, + "learning_rate": 1.746951231366016e-05, + "loss": 1.0942, + "step": 10586 + }, + { + "epoch": 0.25, + "learning_rate": 1.7468994420608988e-05, + "loss": 1.3029, + "step": 10587 + }, + { + "epoch": 0.25, + "learning_rate": 1.7468476482245186e-05, + "loss": 1.0504, + "step": 10588 + }, + { + "epoch": 0.25, + "learning_rate": 1.746795849857189e-05, + "loss": 1.1771, + "step": 10589 + }, + { + "epoch": 0.25, + "learning_rate": 1.7467440469592248e-05, + "loss": 1.2272, + "step": 10590 + }, + { + "epoch": 0.25, + "learning_rate": 1.74669223953094e-05, + "loss": 1.2834, + "step": 10591 + }, + { + "epoch": 0.25, + "learning_rate": 1.746640427572649e-05, + "loss": 1.0525, + "step": 10592 + }, + { + "epoch": 0.25, + "learning_rate": 1.7465886110846665e-05, + "loss": 1.2487, + "step": 10593 + }, + { + "epoch": 0.25, + "learning_rate": 1.7465367900673064e-05, + "loss": 1.0089, + "step": 10594 + }, + { + "epoch": 0.25, + "learning_rate": 1.746484964520883e-05, + "loss": 1.3418, + "step": 10595 + }, + { + "epoch": 0.25, + "learning_rate": 1.7464331344457108e-05, + "loss": 1.141, + "step": 10596 + }, + { + "epoch": 0.25, + "learning_rate": 1.7463812998421045e-05, + "loss": 1.1664, + "step": 10597 + }, + { + "epoch": 0.25, + "learning_rate": 1.7463294607103787e-05, + "loss": 1.1331, + "step": 10598 + }, + { + "epoch": 0.25, + "learning_rate": 1.746277617050847e-05, + "loss": 1.2392, + "step": 10599 + }, + { + "epoch": 0.25, + "learning_rate": 1.746225768863825e-05, + "loss": 1.302, + "step": 10600 + }, + { + "epoch": 0.25, + "learning_rate": 1.7461739161496267e-05, + "loss": 1.0145, + "step": 10601 + }, + { + "epoch": 0.25, + "learning_rate": 1.7461220589085668e-05, + "loss": 1.083, + "step": 10602 + }, + { + "epoch": 0.25, + "learning_rate": 1.7460701971409598e-05, + "loss": 0.8253, + "step": 10603 + }, + { + "epoch": 0.26, + "learning_rate": 1.74601833084712e-05, + "loss": 1.1872, + "step": 10604 + }, + { + "epoch": 0.26, + "learning_rate": 1.745966460027363e-05, + "loss": 1.0984, + "step": 10605 + }, + { + "epoch": 0.26, + "learning_rate": 1.745914584682003e-05, + "loss": 1.1144, + "step": 10606 + }, + { + "epoch": 0.26, + "learning_rate": 1.745862704811354e-05, + "loss": 1.0911, + "step": 10607 + }, + { + "epoch": 0.26, + "learning_rate": 1.745810820415732e-05, + "loss": 0.9815, + "step": 10608 + }, + { + "epoch": 0.26, + "learning_rate": 1.745758931495451e-05, + "loss": 1.1404, + "step": 10609 + }, + { + "epoch": 0.26, + "learning_rate": 1.745707038050826e-05, + "loss": 1.2111, + "step": 10610 + }, + { + "epoch": 0.26, + "learning_rate": 1.745655140082172e-05, + "loss": 0.8907, + "step": 10611 + }, + { + "epoch": 0.26, + "learning_rate": 1.7456032375898034e-05, + "loss": 1.2166, + "step": 10612 + }, + { + "epoch": 0.26, + "learning_rate": 1.7455513305740352e-05, + "loss": 1.4834, + "step": 10613 + }, + { + "epoch": 0.26, + "learning_rate": 1.745499419035183e-05, + "loss": 1.2312, + "step": 10614 + }, + { + "epoch": 0.26, + "learning_rate": 1.745447502973561e-05, + "loss": 0.9812, + "step": 10615 + }, + { + "epoch": 0.26, + "learning_rate": 1.7453955823894836e-05, + "loss": 0.9549, + "step": 10616 + }, + { + "epoch": 0.26, + "learning_rate": 1.745343657283267e-05, + "loss": 1.1007, + "step": 10617 + }, + { + "epoch": 0.26, + "learning_rate": 1.7452917276552263e-05, + "loss": 1.1291, + "step": 10618 + }, + { + "epoch": 0.26, + "learning_rate": 1.7452397935056754e-05, + "loss": 1.276, + "step": 10619 + }, + { + "epoch": 0.26, + "learning_rate": 1.74518785483493e-05, + "loss": 0.8589, + "step": 10620 + }, + { + "epoch": 0.26, + "learning_rate": 1.7451359116433052e-05, + "loss": 1.2223, + "step": 10621 + }, + { + "epoch": 0.26, + "learning_rate": 1.7450839639311164e-05, + "loss": 1.1815, + "step": 10622 + }, + { + "epoch": 0.26, + "learning_rate": 1.7450320116986783e-05, + "loss": 1.2994, + "step": 10623 + }, + { + "epoch": 0.26, + "learning_rate": 1.744980054946306e-05, + "loss": 1.3015, + "step": 10624 + }, + { + "epoch": 0.26, + "learning_rate": 1.744928093674315e-05, + "loss": 1.0707, + "step": 10625 + }, + { + "epoch": 0.26, + "learning_rate": 1.7448761278830203e-05, + "loss": 1.1239, + "step": 10626 + }, + { + "epoch": 0.26, + "learning_rate": 1.7448241575727375e-05, + "loss": 1.2422, + "step": 10627 + }, + { + "epoch": 0.26, + "learning_rate": 1.744772182743782e-05, + "loss": 1.084, + "step": 10628 + }, + { + "epoch": 0.26, + "learning_rate": 1.7447202033964686e-05, + "loss": 1.2115, + "step": 10629 + }, + { + "epoch": 0.26, + "learning_rate": 1.7446682195311128e-05, + "loss": 1.061, + "step": 10630 + }, + { + "epoch": 0.26, + "learning_rate": 1.7446162311480305e-05, + "loss": 1.1407, + "step": 10631 + }, + { + "epoch": 0.26, + "learning_rate": 1.7445642382475362e-05, + "loss": 1.2941, + "step": 10632 + }, + { + "epoch": 0.26, + "learning_rate": 1.7445122408299463e-05, + "loss": 1.1469, + "step": 10633 + }, + { + "epoch": 0.26, + "learning_rate": 1.7444602388955752e-05, + "loss": 1.1339, + "step": 10634 + }, + { + "epoch": 0.26, + "learning_rate": 1.744408232444739e-05, + "loss": 0.9345, + "step": 10635 + }, + { + "epoch": 0.26, + "learning_rate": 1.7443562214777533e-05, + "loss": 0.909, + "step": 10636 + }, + { + "epoch": 0.26, + "learning_rate": 1.7443042059949337e-05, + "loss": 1.2141, + "step": 10637 + }, + { + "epoch": 0.26, + "learning_rate": 1.7442521859965956e-05, + "loss": 1.2976, + "step": 10638 + }, + { + "epoch": 0.26, + "learning_rate": 1.744200161483054e-05, + "loss": 1.2553, + "step": 10639 + }, + { + "epoch": 0.26, + "learning_rate": 1.7441481324546254e-05, + "loss": 0.9862, + "step": 10640 + }, + { + "epoch": 0.26, + "learning_rate": 1.744096098911625e-05, + "loss": 1.2094, + "step": 10641 + }, + { + "epoch": 0.26, + "learning_rate": 1.7440440608543688e-05, + "loss": 1.0635, + "step": 10642 + }, + { + "epoch": 0.26, + "learning_rate": 1.743992018283172e-05, + "loss": 0.9664, + "step": 10643 + }, + { + "epoch": 0.26, + "learning_rate": 1.743939971198351e-05, + "loss": 1.2408, + "step": 10644 + }, + { + "epoch": 0.26, + "learning_rate": 1.743887919600221e-05, + "loss": 1.0341, + "step": 10645 + }, + { + "epoch": 0.26, + "learning_rate": 1.7438358634890978e-05, + "loss": 1.1757, + "step": 10646 + }, + { + "epoch": 0.26, + "learning_rate": 1.7437838028652975e-05, + "loss": 1.1736, + "step": 10647 + }, + { + "epoch": 0.26, + "learning_rate": 1.7437317377291358e-05, + "loss": 1.4373, + "step": 10648 + }, + { + "epoch": 0.26, + "learning_rate": 1.743679668080929e-05, + "loss": 1.3215, + "step": 10649 + }, + { + "epoch": 0.26, + "learning_rate": 1.7436275939209918e-05, + "loss": 1.3135, + "step": 10650 + }, + { + "epoch": 0.26, + "learning_rate": 1.743575515249641e-05, + "loss": 1.2648, + "step": 10651 + }, + { + "epoch": 0.26, + "learning_rate": 1.743523432067193e-05, + "loss": 1.2678, + "step": 10652 + }, + { + "epoch": 0.26, + "learning_rate": 1.743471344373963e-05, + "loss": 1.1073, + "step": 10653 + }, + { + "epoch": 0.26, + "learning_rate": 1.7434192521702674e-05, + "loss": 1.1848, + "step": 10654 + }, + { + "epoch": 0.26, + "learning_rate": 1.7433671554564217e-05, + "loss": 1.0587, + "step": 10655 + }, + { + "epoch": 0.26, + "learning_rate": 1.7433150542327425e-05, + "loss": 1.1007, + "step": 10656 + }, + { + "epoch": 0.26, + "learning_rate": 1.7432629484995456e-05, + "loss": 1.0797, + "step": 10657 + }, + { + "epoch": 0.26, + "learning_rate": 1.7432108382571474e-05, + "loss": 1.0762, + "step": 10658 + }, + { + "epoch": 0.26, + "learning_rate": 1.7431587235058635e-05, + "loss": 0.9813, + "step": 10659 + }, + { + "epoch": 0.26, + "learning_rate": 1.743106604246011e-05, + "loss": 1.2098, + "step": 10660 + }, + { + "epoch": 0.26, + "learning_rate": 1.743054480477905e-05, + "loss": 1.2918, + "step": 10661 + }, + { + "epoch": 0.26, + "learning_rate": 1.7430023522018627e-05, + "loss": 1.0937, + "step": 10662 + }, + { + "epoch": 0.26, + "learning_rate": 1.7429502194181993e-05, + "loss": 1.2252, + "step": 10663 + }, + { + "epoch": 0.26, + "learning_rate": 1.742898082127232e-05, + "loss": 1.0121, + "step": 10664 + }, + { + "epoch": 0.26, + "learning_rate": 1.7428459403292768e-05, + "loss": 1.3354, + "step": 10665 + }, + { + "epoch": 0.26, + "learning_rate": 1.74279379402465e-05, + "loss": 0.8188, + "step": 10666 + }, + { + "epoch": 0.26, + "learning_rate": 1.742741643213668e-05, + "loss": 1.0665, + "step": 10667 + }, + { + "epoch": 0.26, + "learning_rate": 1.742689487896647e-05, + "loss": 0.9769, + "step": 10668 + }, + { + "epoch": 0.26, + "learning_rate": 1.742637328073904e-05, + "loss": 1.1488, + "step": 10669 + }, + { + "epoch": 0.26, + "learning_rate": 1.7425851637457546e-05, + "loss": 1.2965, + "step": 10670 + }, + { + "epoch": 0.26, + "learning_rate": 1.7425329949125158e-05, + "loss": 1.2716, + "step": 10671 + }, + { + "epoch": 0.26, + "learning_rate": 1.7424808215745036e-05, + "loss": 1.2723, + "step": 10672 + }, + { + "epoch": 0.26, + "learning_rate": 1.7424286437320353e-05, + "loss": 1.0865, + "step": 10673 + }, + { + "epoch": 0.26, + "learning_rate": 1.742376461385427e-05, + "loss": 1.131, + "step": 10674 + }, + { + "epoch": 0.26, + "learning_rate": 1.7423242745349957e-05, + "loss": 1.3767, + "step": 10675 + }, + { + "epoch": 0.26, + "learning_rate": 1.7422720831810572e-05, + "loss": 0.8882, + "step": 10676 + }, + { + "epoch": 0.26, + "learning_rate": 1.7422198873239285e-05, + "loss": 1.2398, + "step": 10677 + }, + { + "epoch": 0.26, + "learning_rate": 1.7421676869639264e-05, + "loss": 0.6439, + "step": 10678 + }, + { + "epoch": 0.26, + "learning_rate": 1.7421154821013673e-05, + "loss": 1.2518, + "step": 10679 + }, + { + "epoch": 0.26, + "learning_rate": 1.7420632727365686e-05, + "loss": 1.3869, + "step": 10680 + }, + { + "epoch": 0.26, + "learning_rate": 1.7420110588698463e-05, + "loss": 1.0501, + "step": 10681 + }, + { + "epoch": 0.26, + "learning_rate": 1.7419588405015176e-05, + "loss": 1.2616, + "step": 10682 + }, + { + "epoch": 0.26, + "learning_rate": 1.741906617631899e-05, + "loss": 1.0622, + "step": 10683 + }, + { + "epoch": 0.26, + "learning_rate": 1.7418543902613074e-05, + "loss": 0.9838, + "step": 10684 + }, + { + "epoch": 0.26, + "learning_rate": 1.7418021583900596e-05, + "loss": 1.2191, + "step": 10685 + }, + { + "epoch": 0.26, + "learning_rate": 1.741749922018473e-05, + "loss": 1.3097, + "step": 10686 + }, + { + "epoch": 0.26, + "learning_rate": 1.7416976811468636e-05, + "loss": 1.1654, + "step": 10687 + }, + { + "epoch": 0.26, + "learning_rate": 1.741645435775549e-05, + "loss": 1.0932, + "step": 10688 + }, + { + "epoch": 0.26, + "learning_rate": 1.7415931859048464e-05, + "loss": 1.2727, + "step": 10689 + }, + { + "epoch": 0.26, + "learning_rate": 1.741540931535072e-05, + "loss": 1.1321, + "step": 10690 + }, + { + "epoch": 0.26, + "learning_rate": 1.7414886726665435e-05, + "loss": 1.2703, + "step": 10691 + }, + { + "epoch": 0.26, + "learning_rate": 1.7414364092995773e-05, + "loss": 1.1753, + "step": 10692 + }, + { + "epoch": 0.26, + "learning_rate": 1.7413841414344907e-05, + "loss": 1.2601, + "step": 10693 + }, + { + "epoch": 0.26, + "learning_rate": 1.741331869071601e-05, + "loss": 1.1236, + "step": 10694 + }, + { + "epoch": 0.26, + "learning_rate": 1.7412795922112254e-05, + "loss": 1.0673, + "step": 10695 + }, + { + "epoch": 0.26, + "learning_rate": 1.7412273108536805e-05, + "loss": 1.4925, + "step": 10696 + }, + { + "epoch": 0.26, + "learning_rate": 1.741175024999284e-05, + "loss": 1.5211, + "step": 10697 + }, + { + "epoch": 0.26, + "learning_rate": 1.7411227346483534e-05, + "loss": 1.1163, + "step": 10698 + }, + { + "epoch": 0.26, + "learning_rate": 1.741070439801205e-05, + "loss": 1.1823, + "step": 10699 + }, + { + "epoch": 0.26, + "learning_rate": 1.7410181404581567e-05, + "loss": 1.2415, + "step": 10700 + }, + { + "epoch": 0.26, + "learning_rate": 1.740965836619526e-05, + "loss": 1.2565, + "step": 10701 + }, + { + "epoch": 0.26, + "learning_rate": 1.7409135282856297e-05, + "loss": 1.1807, + "step": 10702 + }, + { + "epoch": 0.26, + "learning_rate": 1.7408612154567847e-05, + "loss": 1.0713, + "step": 10703 + }, + { + "epoch": 0.26, + "learning_rate": 1.7408088981333095e-05, + "loss": 1.1082, + "step": 10704 + }, + { + "epoch": 0.26, + "learning_rate": 1.7407565763155208e-05, + "loss": 1.2081, + "step": 10705 + }, + { + "epoch": 0.26, + "learning_rate": 1.7407042500037362e-05, + "loss": 1.1174, + "step": 10706 + }, + { + "epoch": 0.26, + "learning_rate": 1.7406519191982734e-05, + "loss": 0.8366, + "step": 10707 + }, + { + "epoch": 0.26, + "learning_rate": 1.7405995838994495e-05, + "loss": 0.8834, + "step": 10708 + }, + { + "epoch": 0.26, + "learning_rate": 1.7405472441075817e-05, + "loss": 0.9099, + "step": 10709 + }, + { + "epoch": 0.26, + "learning_rate": 1.740494899822988e-05, + "loss": 1.3606, + "step": 10710 + }, + { + "epoch": 0.26, + "learning_rate": 1.7404425510459866e-05, + "loss": 1.1781, + "step": 10711 + }, + { + "epoch": 0.26, + "learning_rate": 1.740390197776894e-05, + "loss": 1.2494, + "step": 10712 + }, + { + "epoch": 0.26, + "learning_rate": 1.7403378400160282e-05, + "loss": 1.316, + "step": 10713 + }, + { + "epoch": 0.26, + "learning_rate": 1.740285477763707e-05, + "loss": 1.1916, + "step": 10714 + }, + { + "epoch": 0.26, + "learning_rate": 1.740233111020248e-05, + "loss": 1.2469, + "step": 10715 + }, + { + "epoch": 0.26, + "learning_rate": 1.7401807397859682e-05, + "loss": 1.0392, + "step": 10716 + }, + { + "epoch": 0.26, + "learning_rate": 1.7401283640611864e-05, + "loss": 1.1883, + "step": 10717 + }, + { + "epoch": 0.26, + "learning_rate": 1.74007598384622e-05, + "loss": 1.1154, + "step": 10718 + }, + { + "epoch": 0.26, + "learning_rate": 1.7400235991413866e-05, + "loss": 1.0609, + "step": 10719 + }, + { + "epoch": 0.26, + "learning_rate": 1.7399712099470042e-05, + "loss": 1.0701, + "step": 10720 + }, + { + "epoch": 0.26, + "learning_rate": 1.7399188162633902e-05, + "loss": 1.4342, + "step": 10721 + }, + { + "epoch": 0.26, + "learning_rate": 1.739866418090863e-05, + "loss": 1.2705, + "step": 10722 + }, + { + "epoch": 0.26, + "learning_rate": 1.7398140154297404e-05, + "loss": 1.1873, + "step": 10723 + }, + { + "epoch": 0.26, + "learning_rate": 1.7397616082803398e-05, + "loss": 1.1013, + "step": 10724 + }, + { + "epoch": 0.26, + "learning_rate": 1.73970919664298e-05, + "loss": 1.1035, + "step": 10725 + }, + { + "epoch": 0.26, + "learning_rate": 1.739656780517978e-05, + "loss": 1.0932, + "step": 10726 + }, + { + "epoch": 0.26, + "learning_rate": 1.7396043599056525e-05, + "loss": 1.2313, + "step": 10727 + }, + { + "epoch": 0.26, + "learning_rate": 1.7395519348063214e-05, + "loss": 1.2099, + "step": 10728 + }, + { + "epoch": 0.26, + "learning_rate": 1.7394995052203025e-05, + "loss": 1.114, + "step": 10729 + }, + { + "epoch": 0.26, + "learning_rate": 1.739447071147914e-05, + "loss": 1.3325, + "step": 10730 + }, + { + "epoch": 0.26, + "learning_rate": 1.7393946325894743e-05, + "loss": 1.2719, + "step": 10731 + }, + { + "epoch": 0.26, + "learning_rate": 1.7393421895453012e-05, + "loss": 1.321, + "step": 10732 + }, + { + "epoch": 0.26, + "learning_rate": 1.739289742015713e-05, + "loss": 1.2816, + "step": 10733 + }, + { + "epoch": 0.26, + "learning_rate": 1.7392372900010277e-05, + "loss": 1.0285, + "step": 10734 + }, + { + "epoch": 0.26, + "learning_rate": 1.739184833501564e-05, + "loss": 1.147, + "step": 10735 + }, + { + "epoch": 0.26, + "learning_rate": 1.7391323725176392e-05, + "loss": 1.095, + "step": 10736 + }, + { + "epoch": 0.26, + "learning_rate": 1.7390799070495723e-05, + "loss": 1.1441, + "step": 10737 + }, + { + "epoch": 0.26, + "learning_rate": 1.7390274370976816e-05, + "loss": 1.3212, + "step": 10738 + }, + { + "epoch": 0.26, + "learning_rate": 1.7389749626622853e-05, + "loss": 1.2467, + "step": 10739 + }, + { + "epoch": 0.26, + "learning_rate": 1.7389224837437018e-05, + "loss": 1.1887, + "step": 10740 + }, + { + "epoch": 0.26, + "learning_rate": 1.7388700003422493e-05, + "loss": 1.2417, + "step": 10741 + }, + { + "epoch": 0.26, + "learning_rate": 1.7388175124582462e-05, + "loss": 1.0429, + "step": 10742 + }, + { + "epoch": 0.26, + "learning_rate": 1.738765020092011e-05, + "loss": 1.0411, + "step": 10743 + }, + { + "epoch": 0.26, + "learning_rate": 1.738712523243862e-05, + "loss": 1.3328, + "step": 10744 + }, + { + "epoch": 0.26, + "learning_rate": 1.7386600219141187e-05, + "loss": 0.9934, + "step": 10745 + }, + { + "epoch": 0.26, + "learning_rate": 1.738607516103098e-05, + "loss": 1.3112, + "step": 10746 + }, + { + "epoch": 0.26, + "learning_rate": 1.7385550058111195e-05, + "loss": 0.6099, + "step": 10747 + }, + { + "epoch": 0.26, + "learning_rate": 1.7385024910385014e-05, + "loss": 1.1053, + "step": 10748 + }, + { + "epoch": 0.26, + "learning_rate": 1.7384499717855625e-05, + "loss": 1.0968, + "step": 10749 + }, + { + "epoch": 0.26, + "learning_rate": 1.7383974480526213e-05, + "loss": 1.1334, + "step": 10750 + }, + { + "epoch": 0.26, + "learning_rate": 1.7383449198399966e-05, + "loss": 1.1288, + "step": 10751 + }, + { + "epoch": 0.26, + "learning_rate": 1.7382923871480065e-05, + "loss": 1.0514, + "step": 10752 + }, + { + "epoch": 0.26, + "learning_rate": 1.7382398499769706e-05, + "loss": 1.2568, + "step": 10753 + }, + { + "epoch": 0.26, + "learning_rate": 1.738187308327207e-05, + "loss": 0.9255, + "step": 10754 + }, + { + "epoch": 0.26, + "learning_rate": 1.7381347621990345e-05, + "loss": 1.3189, + "step": 10755 + }, + { + "epoch": 0.26, + "learning_rate": 1.7380822115927723e-05, + "loss": 1.3127, + "step": 10756 + }, + { + "epoch": 0.26, + "learning_rate": 1.7380296565087388e-05, + "loss": 1.1342, + "step": 10757 + }, + { + "epoch": 0.26, + "learning_rate": 1.737977096947253e-05, + "loss": 1.0665, + "step": 10758 + }, + { + "epoch": 0.26, + "learning_rate": 1.7379245329086335e-05, + "loss": 1.333, + "step": 10759 + }, + { + "epoch": 0.26, + "learning_rate": 1.7378719643932e-05, + "loss": 1.1135, + "step": 10760 + }, + { + "epoch": 0.26, + "learning_rate": 1.73781939140127e-05, + "loss": 1.1356, + "step": 10761 + }, + { + "epoch": 0.26, + "learning_rate": 1.7377668139331638e-05, + "loss": 1.1073, + "step": 10762 + }, + { + "epoch": 0.26, + "learning_rate": 1.7377142319892e-05, + "loss": 1.1342, + "step": 10763 + }, + { + "epoch": 0.26, + "learning_rate": 1.7376616455696972e-05, + "loss": 1.3029, + "step": 10764 + }, + { + "epoch": 0.26, + "learning_rate": 1.737609054674975e-05, + "loss": 1.0936, + "step": 10765 + }, + { + "epoch": 0.26, + "learning_rate": 1.7375564593053518e-05, + "loss": 1.3537, + "step": 10766 + }, + { + "epoch": 0.26, + "learning_rate": 1.7375038594611474e-05, + "loss": 0.9768, + "step": 10767 + }, + { + "epoch": 0.26, + "learning_rate": 1.7374512551426804e-05, + "loss": 0.9919, + "step": 10768 + }, + { + "epoch": 0.26, + "learning_rate": 1.73739864635027e-05, + "loss": 1.1516, + "step": 10769 + }, + { + "epoch": 0.26, + "learning_rate": 1.7373460330842354e-05, + "loss": 1.0942, + "step": 10770 + }, + { + "epoch": 0.26, + "learning_rate": 1.737293415344896e-05, + "loss": 1.2011, + "step": 10771 + }, + { + "epoch": 0.26, + "learning_rate": 1.7372407931325708e-05, + "loss": 1.0607, + "step": 10772 + }, + { + "epoch": 0.26, + "learning_rate": 1.7371881664475792e-05, + "loss": 1.3096, + "step": 10773 + }, + { + "epoch": 0.26, + "learning_rate": 1.7371355352902404e-05, + "loss": 1.1295, + "step": 10774 + }, + { + "epoch": 0.26, + "learning_rate": 1.7370828996608737e-05, + "loss": 0.9405, + "step": 10775 + }, + { + "epoch": 0.26, + "learning_rate": 1.7370302595597984e-05, + "loss": 1.2394, + "step": 10776 + }, + { + "epoch": 0.26, + "learning_rate": 1.7369776149873337e-05, + "loss": 1.0857, + "step": 10777 + }, + { + "epoch": 0.26, + "learning_rate": 1.7369249659437996e-05, + "loss": 1.0885, + "step": 10778 + }, + { + "epoch": 0.26, + "learning_rate": 1.7368723124295148e-05, + "loss": 1.1592, + "step": 10779 + }, + { + "epoch": 0.26, + "learning_rate": 1.7368196544447987e-05, + "loss": 1.0066, + "step": 10780 + }, + { + "epoch": 0.26, + "learning_rate": 1.7367669919899714e-05, + "loss": 1.0993, + "step": 10781 + }, + { + "epoch": 0.26, + "learning_rate": 1.736714325065352e-05, + "loss": 1.263, + "step": 10782 + }, + { + "epoch": 0.26, + "learning_rate": 1.7366616536712603e-05, + "loss": 1.2465, + "step": 10783 + }, + { + "epoch": 0.26, + "learning_rate": 1.7366089778080152e-05, + "loss": 1.0289, + "step": 10784 + }, + { + "epoch": 0.26, + "learning_rate": 1.7365562974759368e-05, + "loss": 1.111, + "step": 10785 + }, + { + "epoch": 0.26, + "learning_rate": 1.7365036126753448e-05, + "loss": 1.0007, + "step": 10786 + }, + { + "epoch": 0.26, + "learning_rate": 1.7364509234065586e-05, + "loss": 1.1408, + "step": 10787 + }, + { + "epoch": 0.26, + "learning_rate": 1.7363982296698978e-05, + "loss": 1.3073, + "step": 10788 + }, + { + "epoch": 0.26, + "learning_rate": 1.736345531465682e-05, + "loss": 0.9873, + "step": 10789 + }, + { + "epoch": 0.26, + "learning_rate": 1.7362928287942314e-05, + "loss": 1.1672, + "step": 10790 + }, + { + "epoch": 0.26, + "learning_rate": 1.736240121655865e-05, + "loss": 0.9952, + "step": 10791 + }, + { + "epoch": 0.26, + "learning_rate": 1.7361874100509034e-05, + "loss": 1.1144, + "step": 10792 + }, + { + "epoch": 0.26, + "learning_rate": 1.7361346939796653e-05, + "loss": 1.2563, + "step": 10793 + }, + { + "epoch": 0.26, + "learning_rate": 1.7360819734424718e-05, + "loss": 1.0188, + "step": 10794 + }, + { + "epoch": 0.26, + "learning_rate": 1.7360292484396416e-05, + "loss": 1.0949, + "step": 10795 + }, + { + "epoch": 0.26, + "learning_rate": 1.7359765189714957e-05, + "loss": 1.3336, + "step": 10796 + }, + { + "epoch": 0.26, + "learning_rate": 1.7359237850383528e-05, + "loss": 1.0774, + "step": 10797 + }, + { + "epoch": 0.26, + "learning_rate": 1.7358710466405337e-05, + "loss": 1.1953, + "step": 10798 + }, + { + "epoch": 0.26, + "learning_rate": 1.7358183037783578e-05, + "loss": 1.223, + "step": 10799 + }, + { + "epoch": 0.26, + "learning_rate": 1.7357655564521457e-05, + "loss": 1.1715, + "step": 10800 + }, + { + "epoch": 0.26, + "learning_rate": 1.7357128046622164e-05, + "loss": 0.9968, + "step": 10801 + }, + { + "epoch": 0.26, + "learning_rate": 1.7356600484088914e-05, + "loss": 1.016, + "step": 10802 + }, + { + "epoch": 0.26, + "learning_rate": 1.7356072876924894e-05, + "loss": 1.223, + "step": 10803 + }, + { + "epoch": 0.26, + "learning_rate": 1.735554522513331e-05, + "loss": 1.0838, + "step": 10804 + }, + { + "epoch": 0.26, + "learning_rate": 1.7355017528717365e-05, + "loss": 1.0558, + "step": 10805 + }, + { + "epoch": 0.26, + "learning_rate": 1.735448978768026e-05, + "loss": 1.1577, + "step": 10806 + }, + { + "epoch": 0.26, + "learning_rate": 1.735396200202519e-05, + "loss": 1.1374, + "step": 10807 + }, + { + "epoch": 0.26, + "learning_rate": 1.7353434171755366e-05, + "loss": 0.8572, + "step": 10808 + }, + { + "epoch": 0.26, + "learning_rate": 1.7352906296873987e-05, + "loss": 1.0342, + "step": 10809 + }, + { + "epoch": 0.26, + "learning_rate": 1.7352378377384254e-05, + "loss": 1.1058, + "step": 10810 + }, + { + "epoch": 0.26, + "learning_rate": 1.735185041328937e-05, + "loss": 1.217, + "step": 10811 + }, + { + "epoch": 0.26, + "learning_rate": 1.735132240459254e-05, + "loss": 0.9958, + "step": 10812 + }, + { + "epoch": 0.26, + "learning_rate": 1.7350794351296964e-05, + "loss": 1.4054, + "step": 10813 + }, + { + "epoch": 0.26, + "learning_rate": 1.7350266253405853e-05, + "loss": 1.2354, + "step": 10814 + }, + { + "epoch": 0.26, + "learning_rate": 1.7349738110922398e-05, + "loss": 1.1139, + "step": 10815 + }, + { + "epoch": 0.26, + "learning_rate": 1.7349209923849818e-05, + "loss": 1.0415, + "step": 10816 + }, + { + "epoch": 0.26, + "learning_rate": 1.7348681692191306e-05, + "loss": 0.9379, + "step": 10817 + }, + { + "epoch": 0.26, + "learning_rate": 1.734815341595007e-05, + "loss": 1.2154, + "step": 10818 + }, + { + "epoch": 0.26, + "learning_rate": 1.7347625095129316e-05, + "loss": 1.2103, + "step": 10819 + }, + { + "epoch": 0.26, + "learning_rate": 1.734709672973225e-05, + "loss": 1.1939, + "step": 10820 + }, + { + "epoch": 0.26, + "learning_rate": 1.7346568319762076e-05, + "loss": 1.125, + "step": 10821 + }, + { + "epoch": 0.26, + "learning_rate": 1.7346039865222e-05, + "loss": 1.2665, + "step": 10822 + }, + { + "epoch": 0.26, + "learning_rate": 1.7345511366115228e-05, + "loss": 1.2288, + "step": 10823 + }, + { + "epoch": 0.26, + "learning_rate": 1.7344982822444968e-05, + "loss": 1.238, + "step": 10824 + }, + { + "epoch": 0.26, + "learning_rate": 1.734445423421442e-05, + "loss": 0.7868, + "step": 10825 + }, + { + "epoch": 0.26, + "learning_rate": 1.73439256014268e-05, + "loss": 1.1238, + "step": 10826 + }, + { + "epoch": 0.26, + "learning_rate": 1.7343396924085308e-05, + "loss": 0.984, + "step": 10827 + }, + { + "epoch": 0.26, + "learning_rate": 1.734286820219316e-05, + "loss": 1.0613, + "step": 10828 + }, + { + "epoch": 0.26, + "learning_rate": 1.7342339435753554e-05, + "loss": 1.4939, + "step": 10829 + }, + { + "epoch": 0.26, + "learning_rate": 1.73418106247697e-05, + "loss": 1.3111, + "step": 10830 + }, + { + "epoch": 0.26, + "learning_rate": 1.734128176924481e-05, + "loss": 0.9693, + "step": 10831 + }, + { + "epoch": 0.26, + "learning_rate": 1.7340752869182087e-05, + "loss": 1.3551, + "step": 10832 + }, + { + "epoch": 0.26, + "learning_rate": 1.7340223924584746e-05, + "loss": 1.1191, + "step": 10833 + }, + { + "epoch": 0.26, + "learning_rate": 1.7339694935455994e-05, + "loss": 1.369, + "step": 10834 + }, + { + "epoch": 0.26, + "learning_rate": 1.7339165901799034e-05, + "loss": 1.2017, + "step": 10835 + }, + { + "epoch": 0.26, + "learning_rate": 1.7338636823617086e-05, + "loss": 1.167, + "step": 10836 + }, + { + "epoch": 0.26, + "learning_rate": 1.7338107700913355e-05, + "loss": 1.205, + "step": 10837 + }, + { + "epoch": 0.26, + "learning_rate": 1.7337578533691047e-05, + "loss": 0.9301, + "step": 10838 + }, + { + "epoch": 0.26, + "learning_rate": 1.733704932195338e-05, + "loss": 1.4042, + "step": 10839 + }, + { + "epoch": 0.26, + "learning_rate": 1.7336520065703557e-05, + "loss": 1.1149, + "step": 10840 + }, + { + "epoch": 0.26, + "learning_rate": 1.7335990764944793e-05, + "loss": 1.1781, + "step": 10841 + }, + { + "epoch": 0.26, + "learning_rate": 1.73354614196803e-05, + "loss": 1.2937, + "step": 10842 + }, + { + "epoch": 0.26, + "learning_rate": 1.7334932029913283e-05, + "loss": 0.9082, + "step": 10843 + }, + { + "epoch": 0.26, + "learning_rate": 1.7334402595646965e-05, + "loss": 1.1082, + "step": 10844 + }, + { + "epoch": 0.26, + "learning_rate": 1.7333873116884546e-05, + "loss": 1.1635, + "step": 10845 + }, + { + "epoch": 0.26, + "learning_rate": 1.733334359362925e-05, + "loss": 1.2338, + "step": 10846 + }, + { + "epoch": 0.26, + "learning_rate": 1.733281402588428e-05, + "loss": 1.12, + "step": 10847 + }, + { + "epoch": 0.26, + "learning_rate": 1.7332284413652852e-05, + "loss": 1.1192, + "step": 10848 + }, + { + "epoch": 0.26, + "learning_rate": 1.733175475693818e-05, + "loss": 0.9627, + "step": 10849 + }, + { + "epoch": 0.26, + "learning_rate": 1.7331225055743478e-05, + "loss": 1.1753, + "step": 10850 + }, + { + "epoch": 0.26, + "learning_rate": 1.7330695310071955e-05, + "loss": 0.919, + "step": 10851 + }, + { + "epoch": 0.26, + "learning_rate": 1.733016551992683e-05, + "loss": 1.0954, + "step": 10852 + }, + { + "epoch": 0.26, + "learning_rate": 1.732963568531131e-05, + "loss": 0.9661, + "step": 10853 + }, + { + "epoch": 0.26, + "learning_rate": 1.732910580622862e-05, + "loss": 1.291, + "step": 10854 + }, + { + "epoch": 0.26, + "learning_rate": 1.7328575882681966e-05, + "loss": 1.1855, + "step": 10855 + }, + { + "epoch": 0.26, + "learning_rate": 1.7328045914674566e-05, + "loss": 1.2511, + "step": 10856 + }, + { + "epoch": 0.26, + "learning_rate": 1.7327515902209636e-05, + "loss": 1.0857, + "step": 10857 + }, + { + "epoch": 0.26, + "learning_rate": 1.732698584529039e-05, + "loss": 1.1856, + "step": 10858 + }, + { + "epoch": 0.26, + "learning_rate": 1.7326455743920046e-05, + "loss": 1.0674, + "step": 10859 + }, + { + "epoch": 0.26, + "learning_rate": 1.7325925598101817e-05, + "loss": 0.8578, + "step": 10860 + }, + { + "epoch": 0.26, + "learning_rate": 1.7325395407838922e-05, + "loss": 1.0252, + "step": 10861 + }, + { + "epoch": 0.26, + "learning_rate": 1.7324865173134576e-05, + "loss": 1.1822, + "step": 10862 + }, + { + "epoch": 0.26, + "learning_rate": 1.7324334893991996e-05, + "loss": 1.1021, + "step": 10863 + }, + { + "epoch": 0.26, + "learning_rate": 1.7323804570414395e-05, + "loss": 1.1612, + "step": 10864 + }, + { + "epoch": 0.26, + "learning_rate": 1.7323274202404997e-05, + "loss": 0.9175, + "step": 10865 + }, + { + "epoch": 0.26, + "learning_rate": 1.7322743789967017e-05, + "loss": 1.2135, + "step": 10866 + }, + { + "epoch": 0.26, + "learning_rate": 1.7322213333103676e-05, + "loss": 1.1275, + "step": 10867 + }, + { + "epoch": 0.26, + "learning_rate": 1.7321682831818184e-05, + "loss": 1.2367, + "step": 10868 + }, + { + "epoch": 0.26, + "learning_rate": 1.7321152286113765e-05, + "loss": 1.0019, + "step": 10869 + }, + { + "epoch": 0.26, + "learning_rate": 1.732062169599364e-05, + "loss": 1.0422, + "step": 10870 + }, + { + "epoch": 0.26, + "learning_rate": 1.7320091061461025e-05, + "loss": 1.1873, + "step": 10871 + }, + { + "epoch": 0.26, + "learning_rate": 1.7319560382519137e-05, + "loss": 1.2328, + "step": 10872 + }, + { + "epoch": 0.26, + "learning_rate": 1.7319029659171198e-05, + "loss": 0.7108, + "step": 10873 + }, + { + "epoch": 0.26, + "learning_rate": 1.7318498891420426e-05, + "loss": 1.039, + "step": 10874 + }, + { + "epoch": 0.26, + "learning_rate": 1.7317968079270048e-05, + "loss": 1.026, + "step": 10875 + }, + { + "epoch": 0.26, + "learning_rate": 1.7317437222723274e-05, + "loss": 0.9804, + "step": 10876 + }, + { + "epoch": 0.26, + "learning_rate": 1.7316906321783334e-05, + "loss": 1.3138, + "step": 10877 + }, + { + "epoch": 0.26, + "learning_rate": 1.731637537645344e-05, + "loss": 1.0529, + "step": 10878 + }, + { + "epoch": 0.26, + "learning_rate": 1.731584438673682e-05, + "loss": 1.1521, + "step": 10879 + }, + { + "epoch": 0.26, + "learning_rate": 1.731531335263669e-05, + "loss": 1.0246, + "step": 10880 + }, + { + "epoch": 0.26, + "learning_rate": 1.7314782274156277e-05, + "loss": 1.0674, + "step": 10881 + }, + { + "epoch": 0.26, + "learning_rate": 1.73142511512988e-05, + "loss": 1.1232, + "step": 10882 + }, + { + "epoch": 0.26, + "learning_rate": 1.7313719984067484e-05, + "loss": 1.2459, + "step": 10883 + }, + { + "epoch": 0.26, + "learning_rate": 1.7313188772465546e-05, + "loss": 1.0656, + "step": 10884 + }, + { + "epoch": 0.26, + "learning_rate": 1.7312657516496216e-05, + "loss": 1.3805, + "step": 10885 + }, + { + "epoch": 0.26, + "learning_rate": 1.7312126216162707e-05, + "loss": 0.7726, + "step": 10886 + }, + { + "epoch": 0.26, + "learning_rate": 1.731159487146825e-05, + "loss": 0.7596, + "step": 10887 + }, + { + "epoch": 0.26, + "learning_rate": 1.7311063482416068e-05, + "loss": 0.9772, + "step": 10888 + }, + { + "epoch": 0.26, + "learning_rate": 1.7310532049009384e-05, + "loss": 1.0889, + "step": 10889 + }, + { + "epoch": 0.26, + "learning_rate": 1.731000057125142e-05, + "loss": 1.3659, + "step": 10890 + }, + { + "epoch": 0.26, + "learning_rate": 1.7309469049145403e-05, + "loss": 1.2689, + "step": 10891 + }, + { + "epoch": 0.26, + "learning_rate": 1.730893748269456e-05, + "loss": 1.1323, + "step": 10892 + }, + { + "epoch": 0.26, + "learning_rate": 1.7308405871902104e-05, + "loss": 1.0646, + "step": 10893 + }, + { + "epoch": 0.26, + "learning_rate": 1.7307874216771276e-05, + "loss": 1.1312, + "step": 10894 + }, + { + "epoch": 0.26, + "learning_rate": 1.730734251730529e-05, + "loss": 1.1161, + "step": 10895 + }, + { + "epoch": 0.26, + "learning_rate": 1.7306810773507376e-05, + "loss": 1.0782, + "step": 10896 + }, + { + "epoch": 0.26, + "learning_rate": 1.7306278985380763e-05, + "loss": 1.1345, + "step": 10897 + }, + { + "epoch": 0.26, + "learning_rate": 1.7305747152928673e-05, + "loss": 0.8375, + "step": 10898 + }, + { + "epoch": 0.26, + "learning_rate": 1.7305215276154333e-05, + "loss": 1.2608, + "step": 10899 + }, + { + "epoch": 0.26, + "learning_rate": 1.7304683355060966e-05, + "loss": 1.1324, + "step": 10900 + }, + { + "epoch": 0.26, + "learning_rate": 1.7304151389651807e-05, + "loss": 1.1722, + "step": 10901 + }, + { + "epoch": 0.26, + "learning_rate": 1.7303619379930083e-05, + "loss": 0.9576, + "step": 10902 + }, + { + "epoch": 0.26, + "learning_rate": 1.730308732589901e-05, + "loss": 0.9456, + "step": 10903 + }, + { + "epoch": 0.26, + "learning_rate": 1.730255522756183e-05, + "loss": 1.2562, + "step": 10904 + }, + { + "epoch": 0.26, + "learning_rate": 1.7302023084921762e-05, + "loss": 1.1795, + "step": 10905 + }, + { + "epoch": 0.26, + "learning_rate": 1.730149089798204e-05, + "loss": 0.9066, + "step": 10906 + }, + { + "epoch": 0.26, + "learning_rate": 1.7300958666745892e-05, + "loss": 1.0813, + "step": 10907 + }, + { + "epoch": 0.26, + "learning_rate": 1.730042639121654e-05, + "loss": 1.2185, + "step": 10908 + }, + { + "epoch": 0.26, + "learning_rate": 1.7299894071397224e-05, + "loss": 1.1459, + "step": 10909 + }, + { + "epoch": 0.26, + "learning_rate": 1.729936170729116e-05, + "loss": 1.1414, + "step": 10910 + }, + { + "epoch": 0.26, + "learning_rate": 1.729882929890159e-05, + "loss": 1.3396, + "step": 10911 + }, + { + "epoch": 0.26, + "learning_rate": 1.7298296846231738e-05, + "loss": 1.3341, + "step": 10912 + }, + { + "epoch": 0.26, + "learning_rate": 1.7297764349284838e-05, + "loss": 1.2986, + "step": 10913 + }, + { + "epoch": 0.26, + "learning_rate": 1.7297231808064118e-05, + "loss": 0.9432, + "step": 10914 + }, + { + "epoch": 0.26, + "learning_rate": 1.729669922257281e-05, + "loss": 1.3111, + "step": 10915 + }, + { + "epoch": 0.26, + "learning_rate": 1.729616659281414e-05, + "loss": 1.2827, + "step": 10916 + }, + { + "epoch": 0.26, + "learning_rate": 1.7295633918791347e-05, + "loss": 1.0259, + "step": 10917 + }, + { + "epoch": 0.26, + "learning_rate": 1.7295101200507656e-05, + "loss": 1.0781, + "step": 10918 + }, + { + "epoch": 0.26, + "learning_rate": 1.7294568437966305e-05, + "loss": 1.0925, + "step": 10919 + }, + { + "epoch": 0.26, + "learning_rate": 1.729403563117052e-05, + "loss": 1.3413, + "step": 10920 + }, + { + "epoch": 0.26, + "learning_rate": 1.729350278012354e-05, + "loss": 0.8996, + "step": 10921 + }, + { + "epoch": 0.26, + "learning_rate": 1.729296988482859e-05, + "loss": 1.0299, + "step": 10922 + }, + { + "epoch": 0.26, + "learning_rate": 1.729243694528891e-05, + "loss": 1.1774, + "step": 10923 + }, + { + "epoch": 0.26, + "learning_rate": 1.729190396150773e-05, + "loss": 1.171, + "step": 10924 + }, + { + "epoch": 0.26, + "learning_rate": 1.7291370933488285e-05, + "loss": 1.2145, + "step": 10925 + }, + { + "epoch": 0.26, + "learning_rate": 1.7290837861233805e-05, + "loss": 1.2028, + "step": 10926 + }, + { + "epoch": 0.26, + "learning_rate": 1.729030474474753e-05, + "loss": 0.9775, + "step": 10927 + }, + { + "epoch": 0.26, + "learning_rate": 1.7289771584032686e-05, + "loss": 0.9643, + "step": 10928 + }, + { + "epoch": 0.26, + "learning_rate": 1.7289238379092514e-05, + "loss": 1.3546, + "step": 10929 + }, + { + "epoch": 0.26, + "learning_rate": 1.728870512993025e-05, + "loss": 0.938, + "step": 10930 + }, + { + "epoch": 0.26, + "learning_rate": 1.7288171836549125e-05, + "loss": 1.1208, + "step": 10931 + }, + { + "epoch": 0.26, + "learning_rate": 1.7287638498952375e-05, + "loss": 1.2795, + "step": 10932 + }, + { + "epoch": 0.26, + "learning_rate": 1.7287105117143237e-05, + "loss": 0.9639, + "step": 10933 + }, + { + "epoch": 0.26, + "learning_rate": 1.7286571691124945e-05, + "loss": 0.8905, + "step": 10934 + }, + { + "epoch": 0.26, + "learning_rate": 1.728603822090074e-05, + "loss": 1.1136, + "step": 10935 + }, + { + "epoch": 0.26, + "learning_rate": 1.728550470647385e-05, + "loss": 1.3758, + "step": 10936 + }, + { + "epoch": 0.26, + "learning_rate": 1.728497114784752e-05, + "loss": 1.1922, + "step": 10937 + }, + { + "epoch": 0.26, + "learning_rate": 1.7284437545024984e-05, + "loss": 1.3281, + "step": 10938 + }, + { + "epoch": 0.26, + "learning_rate": 1.7283903898009478e-05, + "loss": 1.2615, + "step": 10939 + }, + { + "epoch": 0.26, + "learning_rate": 1.728337020680424e-05, + "loss": 1.1114, + "step": 10940 + }, + { + "epoch": 0.26, + "learning_rate": 1.728283647141251e-05, + "loss": 0.9767, + "step": 10941 + }, + { + "epoch": 0.26, + "learning_rate": 1.728230269183752e-05, + "loss": 0.9335, + "step": 10942 + }, + { + "epoch": 0.26, + "learning_rate": 1.7281768868082512e-05, + "loss": 1.1679, + "step": 10943 + }, + { + "epoch": 0.26, + "learning_rate": 1.728123500015073e-05, + "loss": 1.2489, + "step": 10944 + }, + { + "epoch": 0.26, + "learning_rate": 1.7280701088045404e-05, + "loss": 1.0813, + "step": 10945 + }, + { + "epoch": 0.26, + "learning_rate": 1.728016713176978e-05, + "loss": 1.1197, + "step": 10946 + }, + { + "epoch": 0.26, + "learning_rate": 1.727963313132709e-05, + "loss": 1.199, + "step": 10947 + }, + { + "epoch": 0.26, + "learning_rate": 1.7279099086720584e-05, + "loss": 1.1051, + "step": 10948 + }, + { + "epoch": 0.26, + "learning_rate": 1.7278564997953492e-05, + "loss": 1.4145, + "step": 10949 + }, + { + "epoch": 0.26, + "learning_rate": 1.7278030865029062e-05, + "loss": 1.271, + "step": 10950 + }, + { + "epoch": 0.26, + "learning_rate": 1.7277496687950525e-05, + "loss": 1.1659, + "step": 10951 + }, + { + "epoch": 0.26, + "learning_rate": 1.727696246672113e-05, + "loss": 1.0311, + "step": 10952 + }, + { + "epoch": 0.26, + "learning_rate": 1.7276428201344115e-05, + "loss": 1.1869, + "step": 10953 + }, + { + "epoch": 0.26, + "learning_rate": 1.7275893891822724e-05, + "loss": 1.1403, + "step": 10954 + }, + { + "epoch": 0.26, + "learning_rate": 1.7275359538160193e-05, + "loss": 1.1441, + "step": 10955 + }, + { + "epoch": 0.26, + "learning_rate": 1.7274825140359767e-05, + "loss": 0.9053, + "step": 10956 + }, + { + "epoch": 0.26, + "learning_rate": 1.727429069842469e-05, + "loss": 1.132, + "step": 10957 + }, + { + "epoch": 0.26, + "learning_rate": 1.7273756212358204e-05, + "loss": 1.0708, + "step": 10958 + }, + { + "epoch": 0.26, + "learning_rate": 1.7273221682163547e-05, + "loss": 1.1818, + "step": 10959 + }, + { + "epoch": 0.26, + "learning_rate": 1.7272687107843966e-05, + "loss": 0.9478, + "step": 10960 + }, + { + "epoch": 0.26, + "learning_rate": 1.72721524894027e-05, + "loss": 1.115, + "step": 10961 + }, + { + "epoch": 0.26, + "learning_rate": 1.7271617826842998e-05, + "loss": 0.9732, + "step": 10962 + }, + { + "epoch": 0.26, + "learning_rate": 1.72710831201681e-05, + "loss": 1.0795, + "step": 10963 + }, + { + "epoch": 0.26, + "learning_rate": 1.7270548369381254e-05, + "loss": 1.1525, + "step": 10964 + }, + { + "epoch": 0.26, + "learning_rate": 1.72700135744857e-05, + "loss": 1.2795, + "step": 10965 + }, + { + "epoch": 0.26, + "learning_rate": 1.7269478735484682e-05, + "loss": 1.1796, + "step": 10966 + }, + { + "epoch": 0.26, + "learning_rate": 1.7268943852381446e-05, + "loss": 1.3426, + "step": 10967 + }, + { + "epoch": 0.26, + "learning_rate": 1.726840892517924e-05, + "loss": 1.0948, + "step": 10968 + }, + { + "epoch": 0.26, + "learning_rate": 1.7267873953881306e-05, + "loss": 1.0584, + "step": 10969 + }, + { + "epoch": 0.26, + "learning_rate": 1.726733893849089e-05, + "loss": 1.0452, + "step": 10970 + }, + { + "epoch": 0.26, + "learning_rate": 1.7266803879011237e-05, + "loss": 1.1891, + "step": 10971 + }, + { + "epoch": 0.26, + "learning_rate": 1.7266268775445595e-05, + "loss": 0.9744, + "step": 10972 + }, + { + "epoch": 0.26, + "learning_rate": 1.726573362779721e-05, + "loss": 1.1567, + "step": 10973 + }, + { + "epoch": 0.26, + "learning_rate": 1.7265198436069326e-05, + "loss": 1.3023, + "step": 10974 + }, + { + "epoch": 0.26, + "learning_rate": 1.7264663200265192e-05, + "loss": 1.3391, + "step": 10975 + }, + { + "epoch": 0.26, + "learning_rate": 1.7264127920388057e-05, + "loss": 1.1291, + "step": 10976 + }, + { + "epoch": 0.26, + "learning_rate": 1.7263592596441167e-05, + "loss": 1.1219, + "step": 10977 + }, + { + "epoch": 0.26, + "learning_rate": 1.7263057228427764e-05, + "loss": 1.3928, + "step": 10978 + }, + { + "epoch": 0.26, + "learning_rate": 1.7262521816351107e-05, + "loss": 1.0457, + "step": 10979 + }, + { + "epoch": 0.26, + "learning_rate": 1.7261986360214435e-05, + "loss": 1.3334, + "step": 10980 + }, + { + "epoch": 0.26, + "learning_rate": 1.7261450860021002e-05, + "loss": 1.0955, + "step": 10981 + }, + { + "epoch": 0.26, + "learning_rate": 1.7260915315774053e-05, + "loss": 1.2087, + "step": 10982 + }, + { + "epoch": 0.26, + "learning_rate": 1.7260379727476843e-05, + "loss": 1.0385, + "step": 10983 + }, + { + "epoch": 0.26, + "learning_rate": 1.7259844095132612e-05, + "loss": 1.0931, + "step": 10984 + }, + { + "epoch": 0.26, + "learning_rate": 1.7259308418744613e-05, + "loss": 1.2761, + "step": 10985 + }, + { + "epoch": 0.26, + "learning_rate": 1.72587726983161e-05, + "loss": 0.9985, + "step": 10986 + }, + { + "epoch": 0.26, + "learning_rate": 1.7258236933850323e-05, + "loss": 1.0589, + "step": 10987 + }, + { + "epoch": 0.26, + "learning_rate": 1.7257701125350527e-05, + "loss": 1.0553, + "step": 10988 + }, + { + "epoch": 0.26, + "learning_rate": 1.7257165272819967e-05, + "loss": 0.9862, + "step": 10989 + }, + { + "epoch": 0.26, + "learning_rate": 1.725662937626189e-05, + "loss": 1.0506, + "step": 10990 + }, + { + "epoch": 0.26, + "learning_rate": 1.725609343567955e-05, + "loss": 1.4338, + "step": 10991 + }, + { + "epoch": 0.26, + "learning_rate": 1.7255557451076202e-05, + "loss": 1.0529, + "step": 10992 + }, + { + "epoch": 0.26, + "learning_rate": 1.7255021422455088e-05, + "loss": 1.3419, + "step": 10993 + }, + { + "epoch": 0.26, + "learning_rate": 1.725448534981947e-05, + "loss": 1.0133, + "step": 10994 + }, + { + "epoch": 0.26, + "learning_rate": 1.725394923317259e-05, + "loss": 1.0309, + "step": 10995 + }, + { + "epoch": 0.26, + "learning_rate": 1.725341307251771e-05, + "loss": 0.9684, + "step": 10996 + }, + { + "epoch": 0.26, + "learning_rate": 1.725287686785808e-05, + "loss": 1.1708, + "step": 10997 + }, + { + "epoch": 0.26, + "learning_rate": 1.725234061919695e-05, + "loss": 0.9117, + "step": 10998 + }, + { + "epoch": 0.26, + "learning_rate": 1.7251804326537575e-05, + "loss": 1.4707, + "step": 10999 + }, + { + "epoch": 0.26, + "learning_rate": 1.7251267989883212e-05, + "loss": 1.3074, + "step": 11000 + }, + { + "epoch": 0.26, + "learning_rate": 1.725073160923711e-05, + "loss": 1.0741, + "step": 11001 + }, + { + "epoch": 0.26, + "learning_rate": 1.7250195184602522e-05, + "loss": 1.1213, + "step": 11002 + }, + { + "epoch": 0.26, + "learning_rate": 1.7249658715982707e-05, + "loss": 1.2142, + "step": 11003 + }, + { + "epoch": 0.26, + "learning_rate": 1.724912220338092e-05, + "loss": 1.2113, + "step": 11004 + }, + { + "epoch": 0.26, + "learning_rate": 1.724858564680041e-05, + "loss": 0.9798, + "step": 11005 + }, + { + "epoch": 0.26, + "learning_rate": 1.724804904624444e-05, + "loss": 0.9798, + "step": 11006 + }, + { + "epoch": 0.26, + "learning_rate": 1.7247512401716258e-05, + "loss": 1.2824, + "step": 11007 + }, + { + "epoch": 0.26, + "learning_rate": 1.7246975713219124e-05, + "loss": 0.8708, + "step": 11008 + }, + { + "epoch": 0.26, + "learning_rate": 1.7246438980756294e-05, + "loss": 1.204, + "step": 11009 + }, + { + "epoch": 0.26, + "learning_rate": 1.724590220433102e-05, + "loss": 1.3491, + "step": 11010 + }, + { + "epoch": 0.26, + "learning_rate": 1.7245365383946567e-05, + "loss": 1.1014, + "step": 11011 + }, + { + "epoch": 0.26, + "learning_rate": 1.7244828519606183e-05, + "loss": 1.2651, + "step": 11012 + }, + { + "epoch": 0.26, + "learning_rate": 1.7244291611313126e-05, + "loss": 1.1942, + "step": 11013 + }, + { + "epoch": 0.26, + "learning_rate": 1.724375465907066e-05, + "loss": 1.0363, + "step": 11014 + }, + { + "epoch": 0.26, + "learning_rate": 1.7243217662882036e-05, + "loss": 1.1704, + "step": 11015 + }, + { + "epoch": 0.26, + "learning_rate": 1.7242680622750514e-05, + "loss": 1.3166, + "step": 11016 + }, + { + "epoch": 0.26, + "learning_rate": 1.7242143538679354e-05, + "loss": 1.0774, + "step": 11017 + }, + { + "epoch": 0.26, + "learning_rate": 1.7241606410671813e-05, + "loss": 1.086, + "step": 11018 + }, + { + "epoch": 0.26, + "learning_rate": 1.7241069238731146e-05, + "loss": 1.0343, + "step": 11019 + }, + { + "epoch": 0.27, + "learning_rate": 1.724053202286062e-05, + "loss": 0.8735, + "step": 11020 + }, + { + "epoch": 0.27, + "learning_rate": 1.7239994763063482e-05, + "loss": 1.22, + "step": 11021 + }, + { + "epoch": 0.27, + "learning_rate": 1.723945745934301e-05, + "loss": 1.0676, + "step": 11022 + }, + { + "epoch": 0.27, + "learning_rate": 1.7238920111702443e-05, + "loss": 1.2246, + "step": 11023 + }, + { + "epoch": 0.27, + "learning_rate": 1.7238382720145052e-05, + "loss": 1.0704, + "step": 11024 + }, + { + "epoch": 0.27, + "learning_rate": 1.72378452846741e-05, + "loss": 0.9479, + "step": 11025 + }, + { + "epoch": 0.27, + "learning_rate": 1.7237307805292838e-05, + "loss": 0.9463, + "step": 11026 + }, + { + "epoch": 0.27, + "learning_rate": 1.723677028200453e-05, + "loss": 1.0435, + "step": 11027 + }, + { + "epoch": 0.27, + "learning_rate": 1.7236232714812443e-05, + "loss": 1.1301, + "step": 11028 + }, + { + "epoch": 0.27, + "learning_rate": 1.7235695103719833e-05, + "loss": 1.1631, + "step": 11029 + }, + { + "epoch": 0.27, + "learning_rate": 1.7235157448729967e-05, + "loss": 1.0623, + "step": 11030 + }, + { + "epoch": 0.27, + "learning_rate": 1.7234619749846096e-05, + "loss": 1.2135, + "step": 11031 + }, + { + "epoch": 0.27, + "learning_rate": 1.7234082007071494e-05, + "loss": 1.1723, + "step": 11032 + }, + { + "epoch": 0.27, + "learning_rate": 1.7233544220409415e-05, + "loss": 1.0463, + "step": 11033 + }, + { + "epoch": 0.27, + "learning_rate": 1.7233006389863125e-05, + "loss": 1.1682, + "step": 11034 + }, + { + "epoch": 0.27, + "learning_rate": 1.7232468515435887e-05, + "loss": 1.1323, + "step": 11035 + }, + { + "epoch": 0.27, + "learning_rate": 1.7231930597130962e-05, + "loss": 1.1946, + "step": 11036 + }, + { + "epoch": 0.27, + "learning_rate": 1.723139263495162e-05, + "loss": 1.0676, + "step": 11037 + }, + { + "epoch": 0.27, + "learning_rate": 1.7230854628901115e-05, + "loss": 1.0474, + "step": 11038 + }, + { + "epoch": 0.27, + "learning_rate": 1.7230316578982717e-05, + "loss": 1.1952, + "step": 11039 + }, + { + "epoch": 0.27, + "learning_rate": 1.722977848519969e-05, + "loss": 1.278, + "step": 11040 + }, + { + "epoch": 0.27, + "learning_rate": 1.7229240347555295e-05, + "loss": 1.1687, + "step": 11041 + }, + { + "epoch": 0.27, + "learning_rate": 1.72287021660528e-05, + "loss": 1.2177, + "step": 11042 + }, + { + "epoch": 0.27, + "learning_rate": 1.7228163940695473e-05, + "loss": 1.1989, + "step": 11043 + }, + { + "epoch": 0.27, + "learning_rate": 1.722762567148657e-05, + "loss": 0.8145, + "step": 11044 + }, + { + "epoch": 0.27, + "learning_rate": 1.7227087358429365e-05, + "loss": 0.9762, + "step": 11045 + }, + { + "epoch": 0.27, + "learning_rate": 1.722654900152712e-05, + "loss": 1.1505, + "step": 11046 + }, + { + "epoch": 0.27, + "learning_rate": 1.7226010600783107e-05, + "loss": 1.2707, + "step": 11047 + }, + { + "epoch": 0.27, + "learning_rate": 1.722547215620058e-05, + "loss": 1.0769, + "step": 11048 + }, + { + "epoch": 0.27, + "learning_rate": 1.7224933667782817e-05, + "loss": 1.0273, + "step": 11049 + }, + { + "epoch": 0.27, + "learning_rate": 1.722439513553308e-05, + "loss": 1.0366, + "step": 11050 + }, + { + "epoch": 0.27, + "learning_rate": 1.7223856559454636e-05, + "loss": 1.2523, + "step": 11051 + }, + { + "epoch": 0.27, + "learning_rate": 1.7223317939550756e-05, + "loss": 1.2261, + "step": 11052 + }, + { + "epoch": 0.27, + "learning_rate": 1.7222779275824702e-05, + "loss": 1.2625, + "step": 11053 + }, + { + "epoch": 0.27, + "learning_rate": 1.7222240568279746e-05, + "loss": 1.1171, + "step": 11054 + }, + { + "epoch": 0.27, + "learning_rate": 1.7221701816919155e-05, + "loss": 0.8867, + "step": 11055 + }, + { + "epoch": 0.27, + "learning_rate": 1.7221163021746198e-05, + "loss": 0.9373, + "step": 11056 + }, + { + "epoch": 0.27, + "learning_rate": 1.7220624182764145e-05, + "loss": 1.1072, + "step": 11057 + }, + { + "epoch": 0.27, + "learning_rate": 1.722008529997626e-05, + "loss": 1.1883, + "step": 11058 + }, + { + "epoch": 0.27, + "learning_rate": 1.7219546373385824e-05, + "loss": 1.0884, + "step": 11059 + }, + { + "epoch": 0.27, + "learning_rate": 1.7219007402996092e-05, + "loss": 1.1685, + "step": 11060 + }, + { + "epoch": 0.27, + "learning_rate": 1.721846838881034e-05, + "loss": 1.1332, + "step": 11061 + }, + { + "epoch": 0.27, + "learning_rate": 1.7217929330831834e-05, + "loss": 1.2623, + "step": 11062 + }, + { + "epoch": 0.27, + "learning_rate": 1.7217390229063857e-05, + "loss": 1.1939, + "step": 11063 + }, + { + "epoch": 0.27, + "learning_rate": 1.7216851083509663e-05, + "loss": 1.1663, + "step": 11064 + }, + { + "epoch": 0.27, + "learning_rate": 1.7216311894172536e-05, + "loss": 1.0105, + "step": 11065 + }, + { + "epoch": 0.27, + "learning_rate": 1.7215772661055743e-05, + "loss": 1.2622, + "step": 11066 + }, + { + "epoch": 0.27, + "learning_rate": 1.721523338416255e-05, + "loss": 1.4949, + "step": 11067 + }, + { + "epoch": 0.27, + "learning_rate": 1.721469406349624e-05, + "loss": 0.9483, + "step": 11068 + }, + { + "epoch": 0.27, + "learning_rate": 1.7214154699060072e-05, + "loss": 1.0114, + "step": 11069 + }, + { + "epoch": 0.27, + "learning_rate": 1.7213615290857326e-05, + "loss": 1.2639, + "step": 11070 + }, + { + "epoch": 0.27, + "learning_rate": 1.7213075838891275e-05, + "loss": 1.0076, + "step": 11071 + }, + { + "epoch": 0.27, + "learning_rate": 1.7212536343165187e-05, + "loss": 1.0378, + "step": 11072 + }, + { + "epoch": 0.27, + "learning_rate": 1.721199680368234e-05, + "loss": 1.2589, + "step": 11073 + }, + { + "epoch": 0.27, + "learning_rate": 1.7211457220446e-05, + "loss": 1.0356, + "step": 11074 + }, + { + "epoch": 0.27, + "learning_rate": 1.721091759345945e-05, + "loss": 1.0729, + "step": 11075 + }, + { + "epoch": 0.27, + "learning_rate": 1.7210377922725957e-05, + "loss": 1.0301, + "step": 11076 + }, + { + "epoch": 0.27, + "learning_rate": 1.7209838208248796e-05, + "loss": 1.1241, + "step": 11077 + }, + { + "epoch": 0.27, + "learning_rate": 1.7209298450031247e-05, + "loss": 0.6825, + "step": 11078 + }, + { + "epoch": 0.27, + "learning_rate": 1.7208758648076575e-05, + "loss": 0.7248, + "step": 11079 + }, + { + "epoch": 0.27, + "learning_rate": 1.7208218802388065e-05, + "loss": 0.8499, + "step": 11080 + }, + { + "epoch": 0.27, + "learning_rate": 1.720767891296898e-05, + "loss": 1.1869, + "step": 11081 + }, + { + "epoch": 0.27, + "learning_rate": 1.7207138979822608e-05, + "loss": 1.2928, + "step": 11082 + }, + { + "epoch": 0.27, + "learning_rate": 1.7206599002952218e-05, + "loss": 0.9868, + "step": 11083 + }, + { + "epoch": 0.27, + "learning_rate": 1.7206058982361087e-05, + "loss": 0.9667, + "step": 11084 + }, + { + "epoch": 0.27, + "learning_rate": 1.7205518918052488e-05, + "loss": 1.45, + "step": 11085 + }, + { + "epoch": 0.27, + "learning_rate": 1.7204978810029705e-05, + "loss": 0.8577, + "step": 11086 + }, + { + "epoch": 0.27, + "learning_rate": 1.7204438658296007e-05, + "loss": 1.0621, + "step": 11087 + }, + { + "epoch": 0.27, + "learning_rate": 1.7203898462854677e-05, + "loss": 1.1633, + "step": 11088 + }, + { + "epoch": 0.27, + "learning_rate": 1.7203358223708984e-05, + "loss": 1.272, + "step": 11089 + }, + { + "epoch": 0.27, + "learning_rate": 1.7202817940862216e-05, + "loss": 1.3211, + "step": 11090 + }, + { + "epoch": 0.27, + "learning_rate": 1.7202277614317648e-05, + "loss": 1.0207, + "step": 11091 + }, + { + "epoch": 0.27, + "learning_rate": 1.720173724407855e-05, + "loss": 1.1159, + "step": 11092 + }, + { + "epoch": 0.27, + "learning_rate": 1.720119683014821e-05, + "loss": 1.1798, + "step": 11093 + }, + { + "epoch": 0.27, + "learning_rate": 1.72006563725299e-05, + "loss": 1.1323, + "step": 11094 + }, + { + "epoch": 0.27, + "learning_rate": 1.7200115871226904e-05, + "loss": 1.1128, + "step": 11095 + }, + { + "epoch": 0.27, + "learning_rate": 1.7199575326242497e-05, + "loss": 0.9861, + "step": 11096 + }, + { + "epoch": 0.27, + "learning_rate": 1.7199034737579962e-05, + "loss": 0.903, + "step": 11097 + }, + { + "epoch": 0.27, + "learning_rate": 1.7198494105242575e-05, + "loss": 0.907, + "step": 11098 + }, + { + "epoch": 0.27, + "learning_rate": 1.7197953429233612e-05, + "loss": 0.9975, + "step": 11099 + }, + { + "epoch": 0.27, + "learning_rate": 1.7197412709556368e-05, + "loss": 1.1604, + "step": 11100 + }, + { + "epoch": 0.27, + "learning_rate": 1.7196871946214107e-05, + "loss": 1.0112, + "step": 11101 + }, + { + "epoch": 0.27, + "learning_rate": 1.7196331139210124e-05, + "loss": 1.1474, + "step": 11102 + }, + { + "epoch": 0.27, + "learning_rate": 1.7195790288547684e-05, + "loss": 1.0792, + "step": 11103 + }, + { + "epoch": 0.27, + "learning_rate": 1.719524939423008e-05, + "loss": 1.2274, + "step": 11104 + }, + { + "epoch": 0.27, + "learning_rate": 1.7194708456260594e-05, + "loss": 0.9277, + "step": 11105 + }, + { + "epoch": 0.27, + "learning_rate": 1.71941674746425e-05, + "loss": 1.1516, + "step": 11106 + }, + { + "epoch": 0.27, + "learning_rate": 1.719362644937909e-05, + "loss": 1.0097, + "step": 11107 + }, + { + "epoch": 0.27, + "learning_rate": 1.7193085380473633e-05, + "loss": 0.9767, + "step": 11108 + }, + { + "epoch": 0.27, + "learning_rate": 1.7192544267929422e-05, + "loss": 1.1689, + "step": 11109 + }, + { + "epoch": 0.27, + "learning_rate": 1.7192003111749737e-05, + "loss": 1.176, + "step": 11110 + }, + { + "epoch": 0.27, + "learning_rate": 1.719146191193786e-05, + "loss": 1.2558, + "step": 11111 + }, + { + "epoch": 0.27, + "learning_rate": 1.7190920668497074e-05, + "loss": 1.5032, + "step": 11112 + }, + { + "epoch": 0.27, + "learning_rate": 1.7190379381430666e-05, + "loss": 1.021, + "step": 11113 + }, + { + "epoch": 0.27, + "learning_rate": 1.718983805074192e-05, + "loss": 1.1481, + "step": 11114 + }, + { + "epoch": 0.27, + "learning_rate": 1.7189296676434112e-05, + "loss": 1.0072, + "step": 11115 + }, + { + "epoch": 0.27, + "learning_rate": 1.7188755258510536e-05, + "loss": 1.1287, + "step": 11116 + }, + { + "epoch": 0.27, + "learning_rate": 1.718821379697447e-05, + "loss": 1.0745, + "step": 11117 + }, + { + "epoch": 0.27, + "learning_rate": 1.7187672291829205e-05, + "loss": 0.8893, + "step": 11118 + }, + { + "epoch": 0.27, + "learning_rate": 1.718713074307802e-05, + "loss": 1.1935, + "step": 11119 + }, + { + "epoch": 0.27, + "learning_rate": 1.7186589150724203e-05, + "loss": 0.9702, + "step": 11120 + }, + { + "epoch": 0.27, + "learning_rate": 1.7186047514771043e-05, + "loss": 1.0589, + "step": 11121 + }, + { + "epoch": 0.27, + "learning_rate": 1.7185505835221822e-05, + "loss": 1.05, + "step": 11122 + }, + { + "epoch": 0.27, + "learning_rate": 1.7184964112079824e-05, + "loss": 1.3279, + "step": 11123 + }, + { + "epoch": 0.27, + "learning_rate": 1.7184422345348344e-05, + "loss": 1.0964, + "step": 11124 + }, + { + "epoch": 0.27, + "learning_rate": 1.718388053503066e-05, + "loss": 1.2799, + "step": 11125 + }, + { + "epoch": 0.27, + "learning_rate": 1.7183338681130065e-05, + "loss": 1.0265, + "step": 11126 + }, + { + "epoch": 0.27, + "learning_rate": 1.7182796783649838e-05, + "loss": 1.0998, + "step": 11127 + }, + { + "epoch": 0.27, + "learning_rate": 1.718225484259328e-05, + "loss": 0.899, + "step": 11128 + }, + { + "epoch": 0.27, + "learning_rate": 1.7181712857963665e-05, + "loss": 1.15, + "step": 11129 + }, + { + "epoch": 0.27, + "learning_rate": 1.7181170829764292e-05, + "loss": 1.0081, + "step": 11130 + }, + { + "epoch": 0.27, + "learning_rate": 1.718062875799844e-05, + "loss": 1.127, + "step": 11131 + }, + { + "epoch": 0.27, + "learning_rate": 1.7180086642669406e-05, + "loss": 1.0238, + "step": 11132 + }, + { + "epoch": 0.27, + "learning_rate": 1.7179544483780473e-05, + "loss": 1.0811, + "step": 11133 + }, + { + "epoch": 0.27, + "learning_rate": 1.7179002281334934e-05, + "loss": 1.4092, + "step": 11134 + }, + { + "epoch": 0.27, + "learning_rate": 1.717846003533608e-05, + "loss": 1.0562, + "step": 11135 + }, + { + "epoch": 0.27, + "learning_rate": 1.717791774578719e-05, + "loss": 1.0019, + "step": 11136 + }, + { + "epoch": 0.27, + "learning_rate": 1.717737541269157e-05, + "loss": 1.2166, + "step": 11137 + }, + { + "epoch": 0.27, + "learning_rate": 1.7176833036052495e-05, + "loss": 0.9022, + "step": 11138 + }, + { + "epoch": 0.27, + "learning_rate": 1.717629061587326e-05, + "loss": 0.9962, + "step": 11139 + }, + { + "epoch": 0.27, + "learning_rate": 1.7175748152157163e-05, + "loss": 1.0244, + "step": 11140 + }, + { + "epoch": 0.27, + "learning_rate": 1.717520564490749e-05, + "loss": 1.0655, + "step": 11141 + }, + { + "epoch": 0.27, + "learning_rate": 1.717466309412753e-05, + "loss": 1.1785, + "step": 11142 + }, + { + "epoch": 0.27, + "learning_rate": 1.7174120499820576e-05, + "loss": 1.1407, + "step": 11143 + }, + { + "epoch": 0.27, + "learning_rate": 1.7173577861989924e-05, + "loss": 1.2239, + "step": 11144 + }, + { + "epoch": 0.27, + "learning_rate": 1.717303518063886e-05, + "loss": 1.2116, + "step": 11145 + }, + { + "epoch": 0.27, + "learning_rate": 1.717249245577068e-05, + "loss": 0.8016, + "step": 11146 + }, + { + "epoch": 0.27, + "learning_rate": 1.717194968738867e-05, + "loss": 1.2391, + "step": 11147 + }, + { + "epoch": 0.27, + "learning_rate": 1.7171406875496134e-05, + "loss": 1.265, + "step": 11148 + }, + { + "epoch": 0.27, + "learning_rate": 1.7170864020096355e-05, + "loss": 1.1292, + "step": 11149 + }, + { + "epoch": 0.27, + "learning_rate": 1.7170321121192635e-05, + "loss": 1.0952, + "step": 11150 + }, + { + "epoch": 0.27, + "learning_rate": 1.716977817878826e-05, + "loss": 1.0866, + "step": 11151 + }, + { + "epoch": 0.27, + "learning_rate": 1.716923519288653e-05, + "loss": 1.3226, + "step": 11152 + }, + { + "epoch": 0.27, + "learning_rate": 1.7168692163490734e-05, + "loss": 0.9319, + "step": 11153 + }, + { + "epoch": 0.27, + "learning_rate": 1.716814909060417e-05, + "loss": 0.9851, + "step": 11154 + }, + { + "epoch": 0.27, + "learning_rate": 1.716760597423013e-05, + "loss": 1.0934, + "step": 11155 + }, + { + "epoch": 0.27, + "learning_rate": 1.716706281437192e-05, + "loss": 1.0952, + "step": 11156 + }, + { + "epoch": 0.27, + "learning_rate": 1.716651961103281e-05, + "loss": 1.2447, + "step": 11157 + }, + { + "epoch": 0.27, + "learning_rate": 1.7165976364216123e-05, + "loss": 1.2583, + "step": 11158 + }, + { + "epoch": 0.27, + "learning_rate": 1.716543307392514e-05, + "loss": 0.8887, + "step": 11159 + }, + { + "epoch": 0.27, + "learning_rate": 1.7164889740163158e-05, + "loss": 1.2101, + "step": 11160 + }, + { + "epoch": 0.27, + "learning_rate": 1.7164346362933476e-05, + "loss": 1.4574, + "step": 11161 + }, + { + "epoch": 0.27, + "learning_rate": 1.7163802942239388e-05, + "loss": 1.1582, + "step": 11162 + }, + { + "epoch": 0.27, + "learning_rate": 1.7163259478084196e-05, + "loss": 1.1923, + "step": 11163 + }, + { + "epoch": 0.27, + "learning_rate": 1.716271597047119e-05, + "loss": 1.1613, + "step": 11164 + }, + { + "epoch": 0.27, + "learning_rate": 1.7162172419403675e-05, + "loss": 1.1358, + "step": 11165 + }, + { + "epoch": 0.27, + "learning_rate": 1.716162882488494e-05, + "loss": 1.1504, + "step": 11166 + }, + { + "epoch": 0.27, + "learning_rate": 1.716108518691829e-05, + "loss": 0.9321, + "step": 11167 + }, + { + "epoch": 0.27, + "learning_rate": 1.716054150550702e-05, + "loss": 0.8968, + "step": 11168 + }, + { + "epoch": 0.27, + "learning_rate": 1.715999778065443e-05, + "loss": 0.957, + "step": 11169 + }, + { + "epoch": 0.27, + "learning_rate": 1.715945401236382e-05, + "loss": 1.0317, + "step": 11170 + }, + { + "epoch": 0.27, + "learning_rate": 1.715891020063848e-05, + "loss": 1.2532, + "step": 11171 + }, + { + "epoch": 0.27, + "learning_rate": 1.7158366345481722e-05, + "loss": 1.1322, + "step": 11172 + }, + { + "epoch": 0.27, + "learning_rate": 1.7157822446896837e-05, + "loss": 1.0471, + "step": 11173 + }, + { + "epoch": 0.27, + "learning_rate": 1.7157278504887126e-05, + "loss": 1.2957, + "step": 11174 + }, + { + "epoch": 0.27, + "learning_rate": 1.715673451945589e-05, + "loss": 1.0835, + "step": 11175 + }, + { + "epoch": 0.27, + "learning_rate": 1.715619049060643e-05, + "loss": 1.1337, + "step": 11176 + }, + { + "epoch": 0.27, + "learning_rate": 1.7155646418342042e-05, + "loss": 1.061, + "step": 11177 + }, + { + "epoch": 0.27, + "learning_rate": 1.7155102302666034e-05, + "loss": 1.007, + "step": 11178 + }, + { + "epoch": 0.27, + "learning_rate": 1.7154558143581702e-05, + "loss": 1.3041, + "step": 11179 + }, + { + "epoch": 0.27, + "learning_rate": 1.7154013941092352e-05, + "loss": 0.896, + "step": 11180 + }, + { + "epoch": 0.27, + "learning_rate": 1.7153469695201278e-05, + "loss": 0.9869, + "step": 11181 + }, + { + "epoch": 0.27, + "learning_rate": 1.7152925405911786e-05, + "loss": 0.8637, + "step": 11182 + }, + { + "epoch": 0.27, + "learning_rate": 1.7152381073227177e-05, + "loss": 1.0359, + "step": 11183 + }, + { + "epoch": 0.27, + "learning_rate": 1.7151836697150758e-05, + "loss": 1.0953, + "step": 11184 + }, + { + "epoch": 0.27, + "learning_rate": 1.7151292277685828e-05, + "loss": 1.1088, + "step": 11185 + }, + { + "epoch": 0.27, + "learning_rate": 1.7150747814835687e-05, + "loss": 1.1214, + "step": 11186 + }, + { + "epoch": 0.27, + "learning_rate": 1.7150203308603643e-05, + "loss": 1.0571, + "step": 11187 + }, + { + "epoch": 0.27, + "learning_rate": 1.7149658758992995e-05, + "loss": 1.1932, + "step": 11188 + }, + { + "epoch": 0.27, + "learning_rate": 1.714911416600705e-05, + "loss": 1.1703, + "step": 11189 + }, + { + "epoch": 0.27, + "learning_rate": 1.714856952964911e-05, + "loss": 1.0277, + "step": 11190 + }, + { + "epoch": 0.27, + "learning_rate": 1.7148024849922483e-05, + "loss": 1.1464, + "step": 11191 + }, + { + "epoch": 0.27, + "learning_rate": 1.7147480126830468e-05, + "loss": 1.2149, + "step": 11192 + }, + { + "epoch": 0.27, + "learning_rate": 1.7146935360376377e-05, + "loss": 1.3095, + "step": 11193 + }, + { + "epoch": 0.27, + "learning_rate": 1.7146390550563506e-05, + "loss": 1.0876, + "step": 11194 + }, + { + "epoch": 0.27, + "learning_rate": 1.7145845697395166e-05, + "loss": 1.0222, + "step": 11195 + }, + { + "epoch": 0.27, + "learning_rate": 1.714530080087466e-05, + "loss": 1.0013, + "step": 11196 + }, + { + "epoch": 0.27, + "learning_rate": 1.7144755861005294e-05, + "loss": 0.9547, + "step": 11197 + }, + { + "epoch": 0.27, + "learning_rate": 1.7144210877790372e-05, + "loss": 0.946, + "step": 11198 + }, + { + "epoch": 0.27, + "learning_rate": 1.7143665851233208e-05, + "loss": 1.229, + "step": 11199 + }, + { + "epoch": 0.27, + "learning_rate": 1.7143120781337102e-05, + "loss": 1.2474, + "step": 11200 + }, + { + "epoch": 0.27, + "learning_rate": 1.7142575668105364e-05, + "loss": 1.1909, + "step": 11201 + }, + { + "epoch": 0.27, + "learning_rate": 1.7142030511541298e-05, + "loss": 1.0588, + "step": 11202 + }, + { + "epoch": 0.27, + "learning_rate": 1.7141485311648213e-05, + "loss": 1.1452, + "step": 11203 + }, + { + "epoch": 0.27, + "learning_rate": 1.7140940068429413e-05, + "loss": 0.9932, + "step": 11204 + }, + { + "epoch": 0.27, + "learning_rate": 1.7140394781888216e-05, + "loss": 1.2108, + "step": 11205 + }, + { + "epoch": 0.27, + "learning_rate": 1.7139849452027917e-05, + "loss": 1.4563, + "step": 11206 + }, + { + "epoch": 0.27, + "learning_rate": 1.7139304078851837e-05, + "loss": 1.2003, + "step": 11207 + }, + { + "epoch": 0.27, + "learning_rate": 1.713875866236327e-05, + "loss": 1.0641, + "step": 11208 + }, + { + "epoch": 0.27, + "learning_rate": 1.713821320256554e-05, + "loss": 1.3361, + "step": 11209 + }, + { + "epoch": 0.27, + "learning_rate": 1.713766769946195e-05, + "loss": 0.9735, + "step": 11210 + }, + { + "epoch": 0.27, + "learning_rate": 1.7137122153055807e-05, + "loss": 1.0408, + "step": 11211 + }, + { + "epoch": 0.27, + "learning_rate": 1.713657656335042e-05, + "loss": 1.076, + "step": 11212 + }, + { + "epoch": 0.27, + "learning_rate": 1.7136030930349106e-05, + "loss": 1.0469, + "step": 11213 + }, + { + "epoch": 0.27, + "learning_rate": 1.7135485254055167e-05, + "loss": 1.2151, + "step": 11214 + }, + { + "epoch": 0.27, + "learning_rate": 1.713493953447192e-05, + "loss": 1.1486, + "step": 11215 + }, + { + "epoch": 0.27, + "learning_rate": 1.7134393771602672e-05, + "loss": 0.9062, + "step": 11216 + }, + { + "epoch": 0.27, + "learning_rate": 1.7133847965450735e-05, + "loss": 1.0308, + "step": 11217 + }, + { + "epoch": 0.27, + "learning_rate": 1.713330211601942e-05, + "loss": 1.3805, + "step": 11218 + }, + { + "epoch": 0.27, + "learning_rate": 1.713275622331204e-05, + "loss": 1.1856, + "step": 11219 + }, + { + "epoch": 0.27, + "learning_rate": 1.7132210287331905e-05, + "loss": 0.9663, + "step": 11220 + }, + { + "epoch": 0.27, + "learning_rate": 1.713166430808233e-05, + "loss": 1.34, + "step": 11221 + }, + { + "epoch": 0.27, + "learning_rate": 1.7131118285566623e-05, + "loss": 1.1097, + "step": 11222 + }, + { + "epoch": 0.27, + "learning_rate": 1.7130572219788097e-05, + "loss": 1.1908, + "step": 11223 + }, + { + "epoch": 0.27, + "learning_rate": 1.713002611075007e-05, + "loss": 1.1783, + "step": 11224 + }, + { + "epoch": 0.27, + "learning_rate": 1.712947995845585e-05, + "loss": 1.1836, + "step": 11225 + }, + { + "epoch": 0.27, + "learning_rate": 1.7128933762908754e-05, + "loss": 1.1888, + "step": 11226 + }, + { + "epoch": 0.27, + "learning_rate": 1.712838752411209e-05, + "loss": 1.3581, + "step": 11227 + }, + { + "epoch": 0.27, + "learning_rate": 1.712784124206918e-05, + "loss": 1.1375, + "step": 11228 + }, + { + "epoch": 0.27, + "learning_rate": 1.7127294916783327e-05, + "loss": 1.3014, + "step": 11229 + }, + { + "epoch": 0.27, + "learning_rate": 1.7126748548257858e-05, + "loss": 1.2472, + "step": 11230 + }, + { + "epoch": 0.27, + "learning_rate": 1.712620213649608e-05, + "loss": 0.9099, + "step": 11231 + }, + { + "epoch": 0.27, + "learning_rate": 1.712565568150131e-05, + "loss": 1.195, + "step": 11232 + }, + { + "epoch": 0.27, + "learning_rate": 1.7125109183276863e-05, + "loss": 0.9637, + "step": 11233 + }, + { + "epoch": 0.27, + "learning_rate": 1.7124562641826058e-05, + "loss": 0.9003, + "step": 11234 + }, + { + "epoch": 0.27, + "learning_rate": 1.71240160571522e-05, + "loss": 1.1837, + "step": 11235 + }, + { + "epoch": 0.27, + "learning_rate": 1.7123469429258617e-05, + "loss": 1.3143, + "step": 11236 + }, + { + "epoch": 0.27, + "learning_rate": 1.7122922758148622e-05, + "loss": 1.0624, + "step": 11237 + }, + { + "epoch": 0.27, + "learning_rate": 1.7122376043825527e-05, + "loss": 0.6363, + "step": 11238 + }, + { + "epoch": 0.27, + "learning_rate": 1.7121829286292652e-05, + "loss": 1.0448, + "step": 11239 + }, + { + "epoch": 0.27, + "learning_rate": 1.7121282485553315e-05, + "loss": 1.2101, + "step": 11240 + }, + { + "epoch": 0.27, + "learning_rate": 1.712073564161083e-05, + "loss": 0.8326, + "step": 11241 + }, + { + "epoch": 0.27, + "learning_rate": 1.7120188754468522e-05, + "loss": 1.1668, + "step": 11242 + }, + { + "epoch": 0.27, + "learning_rate": 1.7119641824129698e-05, + "loss": 1.0707, + "step": 11243 + }, + { + "epoch": 0.27, + "learning_rate": 1.7119094850597682e-05, + "loss": 1.2215, + "step": 11244 + }, + { + "epoch": 0.27, + "learning_rate": 1.7118547833875795e-05, + "loss": 1.1763, + "step": 11245 + }, + { + "epoch": 0.27, + "learning_rate": 1.711800077396735e-05, + "loss": 1.0314, + "step": 11246 + }, + { + "epoch": 0.27, + "learning_rate": 1.7117453670875673e-05, + "loss": 1.0482, + "step": 11247 + }, + { + "epoch": 0.27, + "learning_rate": 1.7116906524604074e-05, + "loss": 1.0491, + "step": 11248 + }, + { + "epoch": 0.27, + "learning_rate": 1.711635933515588e-05, + "loss": 1.1762, + "step": 11249 + }, + { + "epoch": 0.27, + "learning_rate": 1.7115812102534405e-05, + "loss": 1.2417, + "step": 11250 + }, + { + "epoch": 0.27, + "learning_rate": 1.7115264826742974e-05, + "loss": 1.178, + "step": 11251 + }, + { + "epoch": 0.27, + "learning_rate": 1.7114717507784906e-05, + "loss": 1.1105, + "step": 11252 + }, + { + "epoch": 0.27, + "learning_rate": 1.7114170145663518e-05, + "loss": 0.992, + "step": 11253 + }, + { + "epoch": 0.27, + "learning_rate": 1.7113622740382133e-05, + "loss": 1.2265, + "step": 11254 + }, + { + "epoch": 0.27, + "learning_rate": 1.7113075291944073e-05, + "loss": 1.2195, + "step": 11255 + }, + { + "epoch": 0.27, + "learning_rate": 1.711252780035266e-05, + "loss": 1.2027, + "step": 11256 + }, + { + "epoch": 0.27, + "learning_rate": 1.711198026561121e-05, + "loss": 1.2748, + "step": 11257 + }, + { + "epoch": 0.27, + "learning_rate": 1.711143268772305e-05, + "loss": 1.0775, + "step": 11258 + }, + { + "epoch": 0.27, + "learning_rate": 1.71108850666915e-05, + "loss": 1.4114, + "step": 11259 + }, + { + "epoch": 0.27, + "learning_rate": 1.7110337402519885e-05, + "loss": 0.984, + "step": 11260 + }, + { + "epoch": 0.27, + "learning_rate": 1.7109789695211524e-05, + "loss": 1.0411, + "step": 11261 + }, + { + "epoch": 0.27, + "learning_rate": 1.7109241944769742e-05, + "loss": 1.1075, + "step": 11262 + }, + { + "epoch": 0.27, + "learning_rate": 1.710869415119786e-05, + "loss": 1.0589, + "step": 11263 + }, + { + "epoch": 0.27, + "learning_rate": 1.7108146314499205e-05, + "loss": 1.0327, + "step": 11264 + }, + { + "epoch": 0.27, + "learning_rate": 1.7107598434677095e-05, + "loss": 1.3945, + "step": 11265 + }, + { + "epoch": 0.27, + "learning_rate": 1.710705051173486e-05, + "loss": 0.9503, + "step": 11266 + }, + { + "epoch": 0.27, + "learning_rate": 1.710650254567582e-05, + "loss": 1.209, + "step": 11267 + }, + { + "epoch": 0.27, + "learning_rate": 1.71059545365033e-05, + "loss": 1.1955, + "step": 11268 + }, + { + "epoch": 0.27, + "learning_rate": 1.7105406484220625e-05, + "loss": 1.1681, + "step": 11269 + }, + { + "epoch": 0.27, + "learning_rate": 1.7104858388831123e-05, + "loss": 1.2018, + "step": 11270 + }, + { + "epoch": 0.27, + "learning_rate": 1.7104310250338113e-05, + "loss": 1.1357, + "step": 11271 + }, + { + "epoch": 0.27, + "learning_rate": 1.7103762068744926e-05, + "loss": 1.2303, + "step": 11272 + }, + { + "epoch": 0.27, + "learning_rate": 1.7103213844054886e-05, + "loss": 1.3555, + "step": 11273 + }, + { + "epoch": 0.27, + "learning_rate": 1.7102665576271316e-05, + "loss": 1.1454, + "step": 11274 + }, + { + "epoch": 0.27, + "learning_rate": 1.7102117265397543e-05, + "loss": 1.1354, + "step": 11275 + }, + { + "epoch": 0.27, + "learning_rate": 1.7101568911436902e-05, + "loss": 0.9688, + "step": 11276 + }, + { + "epoch": 0.27, + "learning_rate": 1.7101020514392706e-05, + "loss": 0.8447, + "step": 11277 + }, + { + "epoch": 0.27, + "learning_rate": 1.710047207426829e-05, + "loss": 1.1881, + "step": 11278 + }, + { + "epoch": 0.27, + "learning_rate": 1.7099923591066985e-05, + "loss": 1.2103, + "step": 11279 + }, + { + "epoch": 0.27, + "learning_rate": 1.7099375064792107e-05, + "loss": 0.9739, + "step": 11280 + }, + { + "epoch": 0.27, + "learning_rate": 1.7098826495446993e-05, + "loss": 1.3366, + "step": 11281 + }, + { + "epoch": 0.27, + "learning_rate": 1.709827788303497e-05, + "loss": 1.1116, + "step": 11282 + }, + { + "epoch": 0.27, + "learning_rate": 1.7097729227559362e-05, + "loss": 1.3066, + "step": 11283 + }, + { + "epoch": 0.27, + "learning_rate": 1.7097180529023502e-05, + "loss": 1.3382, + "step": 11284 + }, + { + "epoch": 0.27, + "learning_rate": 1.7096631787430718e-05, + "loss": 0.9328, + "step": 11285 + }, + { + "epoch": 0.27, + "learning_rate": 1.7096083002784337e-05, + "loss": 1.1363, + "step": 11286 + }, + { + "epoch": 0.27, + "learning_rate": 1.709553417508769e-05, + "loss": 0.8514, + "step": 11287 + }, + { + "epoch": 0.27, + "learning_rate": 1.7094985304344105e-05, + "loss": 1.0227, + "step": 11288 + }, + { + "epoch": 0.27, + "learning_rate": 1.7094436390556916e-05, + "loss": 1.202, + "step": 11289 + }, + { + "epoch": 0.27, + "learning_rate": 1.7093887433729445e-05, + "loss": 1.0907, + "step": 11290 + }, + { + "epoch": 0.27, + "learning_rate": 1.709333843386503e-05, + "loss": 1.2565, + "step": 11291 + }, + { + "epoch": 0.27, + "learning_rate": 1.7092789390967e-05, + "loss": 0.9488, + "step": 11292 + }, + { + "epoch": 0.27, + "learning_rate": 1.7092240305038684e-05, + "loss": 0.8935, + "step": 11293 + }, + { + "epoch": 0.27, + "learning_rate": 1.7091691176083414e-05, + "loss": 1.1196, + "step": 11294 + }, + { + "epoch": 0.27, + "learning_rate": 1.7091142004104523e-05, + "loss": 1.077, + "step": 11295 + }, + { + "epoch": 0.27, + "learning_rate": 1.7090592789105343e-05, + "loss": 1.1728, + "step": 11296 + }, + { + "epoch": 0.27, + "learning_rate": 1.70900435310892e-05, + "loss": 0.987, + "step": 11297 + }, + { + "epoch": 0.27, + "learning_rate": 1.7089494230059434e-05, + "loss": 1.2895, + "step": 11298 + }, + { + "epoch": 0.27, + "learning_rate": 1.7088944886019372e-05, + "loss": 1.104, + "step": 11299 + }, + { + "epoch": 0.27, + "learning_rate": 1.708839549897235e-05, + "loss": 1.1197, + "step": 11300 + }, + { + "epoch": 0.27, + "learning_rate": 1.7087846068921697e-05, + "loss": 1.2221, + "step": 11301 + }, + { + "epoch": 0.27, + "learning_rate": 1.7087296595870752e-05, + "loss": 0.9954, + "step": 11302 + }, + { + "epoch": 0.27, + "learning_rate": 1.708674707982285e-05, + "loss": 1.2944, + "step": 11303 + }, + { + "epoch": 0.27, + "learning_rate": 1.7086197520781312e-05, + "loss": 1.4175, + "step": 11304 + }, + { + "epoch": 0.27, + "learning_rate": 1.7085647918749485e-05, + "loss": 1.2705, + "step": 11305 + }, + { + "epoch": 0.27, + "learning_rate": 1.7085098273730696e-05, + "loss": 1.0878, + "step": 11306 + }, + { + "epoch": 0.27, + "learning_rate": 1.7084548585728284e-05, + "loss": 1.2639, + "step": 11307 + }, + { + "epoch": 0.27, + "learning_rate": 1.708399885474558e-05, + "loss": 1.1571, + "step": 11308 + }, + { + "epoch": 0.27, + "learning_rate": 1.7083449080785927e-05, + "loss": 0.9615, + "step": 11309 + }, + { + "epoch": 0.27, + "learning_rate": 1.708289926385265e-05, + "loss": 1.0767, + "step": 11310 + }, + { + "epoch": 0.27, + "learning_rate": 1.708234940394909e-05, + "loss": 1.0544, + "step": 11311 + }, + { + "epoch": 0.27, + "learning_rate": 1.7081799501078583e-05, + "loss": 1.4561, + "step": 11312 + }, + { + "epoch": 0.27, + "learning_rate": 1.708124955524446e-05, + "loss": 0.9923, + "step": 11313 + }, + { + "epoch": 0.27, + "learning_rate": 1.7080699566450066e-05, + "loss": 1.1239, + "step": 11314 + }, + { + "epoch": 0.27, + "learning_rate": 1.708014953469873e-05, + "loss": 1.3571, + "step": 11315 + }, + { + "epoch": 0.27, + "learning_rate": 1.7079599459993797e-05, + "loss": 1.2213, + "step": 11316 + }, + { + "epoch": 0.27, + "learning_rate": 1.7079049342338595e-05, + "loss": 0.9766, + "step": 11317 + }, + { + "epoch": 0.27, + "learning_rate": 1.707849918173647e-05, + "loss": 1.1837, + "step": 11318 + }, + { + "epoch": 0.27, + "learning_rate": 1.707794897819075e-05, + "loss": 1.3075, + "step": 11319 + }, + { + "epoch": 0.27, + "learning_rate": 1.7077398731704778e-05, + "loss": 0.9781, + "step": 11320 + }, + { + "epoch": 0.27, + "learning_rate": 1.7076848442281896e-05, + "loss": 1.089, + "step": 11321 + }, + { + "epoch": 0.27, + "learning_rate": 1.707629810992544e-05, + "loss": 1.2628, + "step": 11322 + }, + { + "epoch": 0.27, + "learning_rate": 1.7075747734638745e-05, + "loss": 1.2173, + "step": 11323 + }, + { + "epoch": 0.27, + "learning_rate": 1.7075197316425157e-05, + "loss": 1.1909, + "step": 11324 + }, + { + "epoch": 0.27, + "learning_rate": 1.707464685528801e-05, + "loss": 0.9754, + "step": 11325 + }, + { + "epoch": 0.27, + "learning_rate": 1.7074096351230643e-05, + "loss": 1.2285, + "step": 11326 + }, + { + "epoch": 0.27, + "learning_rate": 1.7073545804256396e-05, + "loss": 0.8398, + "step": 11327 + }, + { + "epoch": 0.27, + "learning_rate": 1.707299521436861e-05, + "loss": 1.1491, + "step": 11328 + }, + { + "epoch": 0.27, + "learning_rate": 1.707244458157063e-05, + "loss": 1.2724, + "step": 11329 + }, + { + "epoch": 0.27, + "learning_rate": 1.7071893905865788e-05, + "loss": 1.3234, + "step": 11330 + }, + { + "epoch": 0.27, + "learning_rate": 1.707134318725743e-05, + "loss": 1.2197, + "step": 11331 + }, + { + "epoch": 0.27, + "learning_rate": 1.7070792425748897e-05, + "loss": 1.0798, + "step": 11332 + }, + { + "epoch": 0.27, + "learning_rate": 1.707024162134353e-05, + "loss": 1.0388, + "step": 11333 + }, + { + "epoch": 0.27, + "learning_rate": 1.706969077404467e-05, + "loss": 0.9701, + "step": 11334 + }, + { + "epoch": 0.27, + "learning_rate": 1.7069139883855657e-05, + "loss": 1.3205, + "step": 11335 + }, + { + "epoch": 0.27, + "learning_rate": 1.7068588950779838e-05, + "loss": 1.2125, + "step": 11336 + }, + { + "epoch": 0.27, + "learning_rate": 1.706803797482055e-05, + "loss": 1.3501, + "step": 11337 + }, + { + "epoch": 0.27, + "learning_rate": 1.706748695598114e-05, + "loss": 1.0187, + "step": 11338 + }, + { + "epoch": 0.27, + "learning_rate": 1.7066935894264948e-05, + "loss": 1.0726, + "step": 11339 + }, + { + "epoch": 0.27, + "learning_rate": 1.7066384789675318e-05, + "loss": 1.108, + "step": 11340 + }, + { + "epoch": 0.27, + "learning_rate": 1.7065833642215595e-05, + "loss": 1.2345, + "step": 11341 + }, + { + "epoch": 0.27, + "learning_rate": 1.706528245188912e-05, + "loss": 1.2482, + "step": 11342 + }, + { + "epoch": 0.27, + "learning_rate": 1.7064731218699236e-05, + "loss": 1.3893, + "step": 11343 + }, + { + "epoch": 0.27, + "learning_rate": 1.7064179942649293e-05, + "loss": 1.0986, + "step": 11344 + }, + { + "epoch": 0.27, + "learning_rate": 1.706362862374263e-05, + "loss": 1.1319, + "step": 11345 + }, + { + "epoch": 0.27, + "learning_rate": 1.7063077261982594e-05, + "loss": 1.0892, + "step": 11346 + }, + { + "epoch": 0.27, + "learning_rate": 1.7062525857372532e-05, + "loss": 1.0128, + "step": 11347 + }, + { + "epoch": 0.27, + "learning_rate": 1.7061974409915783e-05, + "loss": 1.3093, + "step": 11348 + }, + { + "epoch": 0.27, + "learning_rate": 1.7061422919615698e-05, + "loss": 1.3193, + "step": 11349 + }, + { + "epoch": 0.27, + "learning_rate": 1.7060871386475623e-05, + "loss": 1.0827, + "step": 11350 + }, + { + "epoch": 0.27, + "learning_rate": 1.7060319810498897e-05, + "loss": 1.2391, + "step": 11351 + }, + { + "epoch": 0.27, + "learning_rate": 1.7059768191688877e-05, + "loss": 0.9884, + "step": 11352 + }, + { + "epoch": 0.27, + "learning_rate": 1.70592165300489e-05, + "loss": 1.2947, + "step": 11353 + }, + { + "epoch": 0.27, + "learning_rate": 1.705866482558232e-05, + "loss": 1.0818, + "step": 11354 + }, + { + "epoch": 0.27, + "learning_rate": 1.705811307829248e-05, + "loss": 1.3815, + "step": 11355 + }, + { + "epoch": 0.27, + "learning_rate": 1.7057561288182727e-05, + "loss": 1.4142, + "step": 11356 + }, + { + "epoch": 0.27, + "learning_rate": 1.705700945525641e-05, + "loss": 1.1479, + "step": 11357 + }, + { + "epoch": 0.27, + "learning_rate": 1.7056457579516874e-05, + "loss": 1.0048, + "step": 11358 + }, + { + "epoch": 0.27, + "learning_rate": 1.7055905660967474e-05, + "loss": 1.3957, + "step": 11359 + }, + { + "epoch": 0.27, + "learning_rate": 1.705535369961155e-05, + "loss": 0.9906, + "step": 11360 + }, + { + "epoch": 0.27, + "learning_rate": 1.705480169545246e-05, + "loss": 0.9551, + "step": 11361 + }, + { + "epoch": 0.27, + "learning_rate": 1.705424964849354e-05, + "loss": 0.927, + "step": 11362 + }, + { + "epoch": 0.27, + "learning_rate": 1.7053697558738152e-05, + "loss": 1.111, + "step": 11363 + }, + { + "epoch": 0.27, + "learning_rate": 1.7053145426189643e-05, + "loss": 1.1266, + "step": 11364 + }, + { + "epoch": 0.27, + "learning_rate": 1.7052593250851353e-05, + "loss": 1.1392, + "step": 11365 + }, + { + "epoch": 0.27, + "learning_rate": 1.705204103272664e-05, + "loss": 1.1231, + "step": 11366 + }, + { + "epoch": 0.27, + "learning_rate": 1.7051488771818856e-05, + "loss": 1.0628, + "step": 11367 + }, + { + "epoch": 0.27, + "learning_rate": 1.7050936468131344e-05, + "loss": 1.2015, + "step": 11368 + }, + { + "epoch": 0.27, + "learning_rate": 1.7050384121667462e-05, + "loss": 0.9759, + "step": 11369 + }, + { + "epoch": 0.27, + "learning_rate": 1.704983173243056e-05, + "loss": 0.8081, + "step": 11370 + }, + { + "epoch": 0.27, + "learning_rate": 1.7049279300423984e-05, + "loss": 1.0636, + "step": 11371 + }, + { + "epoch": 0.27, + "learning_rate": 1.7048726825651087e-05, + "loss": 0.9094, + "step": 11372 + }, + { + "epoch": 0.27, + "learning_rate": 1.7048174308115224e-05, + "loss": 1.0631, + "step": 11373 + }, + { + "epoch": 0.27, + "learning_rate": 1.7047621747819744e-05, + "loss": 1.1233, + "step": 11374 + }, + { + "epoch": 0.27, + "learning_rate": 1.7047069144768004e-05, + "loss": 1.1626, + "step": 11375 + }, + { + "epoch": 0.27, + "learning_rate": 1.704651649896335e-05, + "loss": 1.2259, + "step": 11376 + }, + { + "epoch": 0.27, + "learning_rate": 1.7045963810409138e-05, + "loss": 1.1202, + "step": 11377 + }, + { + "epoch": 0.27, + "learning_rate": 1.704541107910872e-05, + "loss": 1.0084, + "step": 11378 + }, + { + "epoch": 0.27, + "learning_rate": 1.7044858305065454e-05, + "loss": 1.1432, + "step": 11379 + }, + { + "epoch": 0.27, + "learning_rate": 1.704430548828269e-05, + "loss": 0.9153, + "step": 11380 + }, + { + "epoch": 0.27, + "learning_rate": 1.7043752628763778e-05, + "loss": 1.0741, + "step": 11381 + }, + { + "epoch": 0.27, + "learning_rate": 1.7043199726512075e-05, + "loss": 1.4048, + "step": 11382 + }, + { + "epoch": 0.27, + "learning_rate": 1.704264678153094e-05, + "loss": 1.0631, + "step": 11383 + }, + { + "epoch": 0.27, + "learning_rate": 1.7042093793823718e-05, + "loss": 1.2558, + "step": 11384 + }, + { + "epoch": 0.27, + "learning_rate": 1.7041540763393775e-05, + "loss": 1.2283, + "step": 11385 + }, + { + "epoch": 0.27, + "learning_rate": 1.7040987690244458e-05, + "loss": 1.2412, + "step": 11386 + }, + { + "epoch": 0.27, + "learning_rate": 1.7040434574379127e-05, + "loss": 1.0207, + "step": 11387 + }, + { + "epoch": 0.27, + "learning_rate": 1.703988141580113e-05, + "loss": 1.2461, + "step": 11388 + }, + { + "epoch": 0.27, + "learning_rate": 1.7039328214513833e-05, + "loss": 1.1559, + "step": 11389 + }, + { + "epoch": 0.27, + "learning_rate": 1.7038774970520585e-05, + "loss": 0.9697, + "step": 11390 + }, + { + "epoch": 0.27, + "learning_rate": 1.7038221683824745e-05, + "loss": 1.1288, + "step": 11391 + }, + { + "epoch": 0.27, + "learning_rate": 1.703766835442967e-05, + "loss": 1.3546, + "step": 11392 + }, + { + "epoch": 0.27, + "learning_rate": 1.7037114982338717e-05, + "loss": 1.0679, + "step": 11393 + }, + { + "epoch": 0.27, + "learning_rate": 1.7036561567555245e-05, + "loss": 1.0295, + "step": 11394 + }, + { + "epoch": 0.27, + "learning_rate": 1.7036008110082604e-05, + "loss": 0.9676, + "step": 11395 + }, + { + "epoch": 0.27, + "learning_rate": 1.703545460992416e-05, + "loss": 1.1733, + "step": 11396 + }, + { + "epoch": 0.27, + "learning_rate": 1.7034901067083266e-05, + "loss": 0.9729, + "step": 11397 + }, + { + "epoch": 0.27, + "learning_rate": 1.7034347481563288e-05, + "loss": 1.1491, + "step": 11398 + }, + { + "epoch": 0.27, + "learning_rate": 1.703379385336757e-05, + "loss": 1.4095, + "step": 11399 + }, + { + "epoch": 0.27, + "learning_rate": 1.7033240182499484e-05, + "loss": 1.1721, + "step": 11400 + }, + { + "epoch": 0.27, + "learning_rate": 1.7032686468962382e-05, + "loss": 1.1502, + "step": 11401 + }, + { + "epoch": 0.27, + "learning_rate": 1.7032132712759624e-05, + "loss": 1.0189, + "step": 11402 + }, + { + "epoch": 0.27, + "learning_rate": 1.7031578913894575e-05, + "loss": 1.1725, + "step": 11403 + }, + { + "epoch": 0.27, + "learning_rate": 1.7031025072370592e-05, + "loss": 1.2191, + "step": 11404 + }, + { + "epoch": 0.27, + "learning_rate": 1.703047118819103e-05, + "loss": 1.1609, + "step": 11405 + }, + { + "epoch": 0.27, + "learning_rate": 1.7029917261359254e-05, + "loss": 0.6483, + "step": 11406 + }, + { + "epoch": 0.27, + "learning_rate": 1.702936329187862e-05, + "loss": 1.1144, + "step": 11407 + }, + { + "epoch": 0.27, + "learning_rate": 1.7028809279752495e-05, + "loss": 1.2475, + "step": 11408 + }, + { + "epoch": 0.27, + "learning_rate": 1.7028255224984237e-05, + "loss": 1.2118, + "step": 11409 + }, + { + "epoch": 0.27, + "learning_rate": 1.7027701127577208e-05, + "loss": 1.0818, + "step": 11410 + }, + { + "epoch": 0.27, + "learning_rate": 1.7027146987534772e-05, + "loss": 0.7584, + "step": 11411 + }, + { + "epoch": 0.27, + "learning_rate": 1.7026592804860283e-05, + "loss": 0.8248, + "step": 11412 + }, + { + "epoch": 0.27, + "learning_rate": 1.7026038579557113e-05, + "loss": 1.2403, + "step": 11413 + }, + { + "epoch": 0.27, + "learning_rate": 1.7025484311628616e-05, + "loss": 1.0621, + "step": 11414 + }, + { + "epoch": 0.27, + "learning_rate": 1.702493000107816e-05, + "loss": 1.0384, + "step": 11415 + }, + { + "epoch": 0.27, + "learning_rate": 1.7024375647909106e-05, + "loss": 1.2523, + "step": 11416 + }, + { + "epoch": 0.27, + "learning_rate": 1.7023821252124817e-05, + "loss": 1.3391, + "step": 11417 + }, + { + "epoch": 0.27, + "learning_rate": 1.7023266813728656e-05, + "loss": 0.8555, + "step": 11418 + }, + { + "epoch": 0.27, + "learning_rate": 1.702271233272399e-05, + "loss": 1.4356, + "step": 11419 + }, + { + "epoch": 0.27, + "learning_rate": 1.7022157809114176e-05, + "loss": 1.0565, + "step": 11420 + }, + { + "epoch": 0.27, + "learning_rate": 1.7021603242902586e-05, + "loss": 0.9839, + "step": 11421 + }, + { + "epoch": 0.27, + "learning_rate": 1.7021048634092578e-05, + "loss": 0.9315, + "step": 11422 + }, + { + "epoch": 0.27, + "learning_rate": 1.702049398268752e-05, + "loss": 0.9816, + "step": 11423 + }, + { + "epoch": 0.27, + "learning_rate": 1.7019939288690773e-05, + "loss": 1.1465, + "step": 11424 + }, + { + "epoch": 0.27, + "learning_rate": 1.7019384552105712e-05, + "loss": 1.1172, + "step": 11425 + }, + { + "epoch": 0.27, + "learning_rate": 1.701882977293569e-05, + "loss": 0.9607, + "step": 11426 + }, + { + "epoch": 0.27, + "learning_rate": 1.7018274951184083e-05, + "loss": 0.9559, + "step": 11427 + }, + { + "epoch": 0.27, + "learning_rate": 1.701772008685425e-05, + "loss": 1.048, + "step": 11428 + }, + { + "epoch": 0.27, + "learning_rate": 1.7017165179949562e-05, + "loss": 1.0449, + "step": 11429 + }, + { + "epoch": 0.27, + "learning_rate": 1.7016610230473385e-05, + "loss": 0.9435, + "step": 11430 + }, + { + "epoch": 0.27, + "learning_rate": 1.701605523842908e-05, + "loss": 1.1457, + "step": 11431 + }, + { + "epoch": 0.27, + "learning_rate": 1.701550020382002e-05, + "loss": 0.9787, + "step": 11432 + }, + { + "epoch": 0.27, + "learning_rate": 1.701494512664957e-05, + "loss": 1.0494, + "step": 11433 + }, + { + "epoch": 0.27, + "learning_rate": 1.7014390006921096e-05, + "loss": 1.2052, + "step": 11434 + }, + { + "epoch": 0.27, + "learning_rate": 1.7013834844637974e-05, + "loss": 1.2118, + "step": 11435 + }, + { + "epoch": 0.28, + "learning_rate": 1.7013279639803563e-05, + "loss": 1.1866, + "step": 11436 + }, + { + "epoch": 0.28, + "learning_rate": 1.7012724392421233e-05, + "loss": 0.8975, + "step": 11437 + }, + { + "epoch": 0.28, + "learning_rate": 1.7012169102494352e-05, + "loss": 1.2279, + "step": 11438 + }, + { + "epoch": 0.28, + "learning_rate": 1.7011613770026292e-05, + "loss": 1.2391, + "step": 11439 + }, + { + "epoch": 0.28, + "learning_rate": 1.7011058395020417e-05, + "loss": 1.1962, + "step": 11440 + }, + { + "epoch": 0.28, + "learning_rate": 1.7010502977480102e-05, + "loss": 1.3976, + "step": 11441 + }, + { + "epoch": 0.28, + "learning_rate": 1.7009947517408717e-05, + "loss": 1.1271, + "step": 11442 + }, + { + "epoch": 0.28, + "learning_rate": 1.7009392014809627e-05, + "loss": 1.2006, + "step": 11443 + }, + { + "epoch": 0.28, + "learning_rate": 1.7008836469686207e-05, + "loss": 1.0161, + "step": 11444 + }, + { + "epoch": 0.28, + "learning_rate": 1.700828088204182e-05, + "loss": 0.6497, + "step": 11445 + }, + { + "epoch": 0.28, + "learning_rate": 1.7007725251879844e-05, + "loss": 1.3677, + "step": 11446 + }, + { + "epoch": 0.28, + "learning_rate": 1.7007169579203647e-05, + "loss": 1.1387, + "step": 11447 + }, + { + "epoch": 0.28, + "learning_rate": 1.7006613864016605e-05, + "loss": 0.9512, + "step": 11448 + }, + { + "epoch": 0.28, + "learning_rate": 1.7006058106322077e-05, + "loss": 1.2925, + "step": 11449 + }, + { + "epoch": 0.28, + "learning_rate": 1.7005502306123447e-05, + "loss": 0.8903, + "step": 11450 + }, + { + "epoch": 0.28, + "learning_rate": 1.700494646342408e-05, + "loss": 1.1918, + "step": 11451 + }, + { + "epoch": 0.28, + "learning_rate": 1.7004390578227355e-05, + "loss": 1.2832, + "step": 11452 + }, + { + "epoch": 0.28, + "learning_rate": 1.7003834650536636e-05, + "loss": 1.0142, + "step": 11453 + }, + { + "epoch": 0.28, + "learning_rate": 1.7003278680355298e-05, + "loss": 0.9292, + "step": 11454 + }, + { + "epoch": 0.28, + "learning_rate": 1.7002722667686717e-05, + "loss": 1.2961, + "step": 11455 + }, + { + "epoch": 0.28, + "learning_rate": 1.7002166612534266e-05, + "loss": 1.0146, + "step": 11456 + }, + { + "epoch": 0.28, + "learning_rate": 1.7001610514901317e-05, + "loss": 1.0297, + "step": 11457 + }, + { + "epoch": 0.28, + "learning_rate": 1.7001054374791245e-05, + "loss": 1.3311, + "step": 11458 + }, + { + "epoch": 0.28, + "learning_rate": 1.700049819220742e-05, + "loss": 1.0876, + "step": 11459 + }, + { + "epoch": 0.28, + "learning_rate": 1.699994196715322e-05, + "loss": 1.1954, + "step": 11460 + }, + { + "epoch": 0.28, + "learning_rate": 1.6999385699632017e-05, + "loss": 0.8139, + "step": 11461 + }, + { + "epoch": 0.28, + "learning_rate": 1.699882938964719e-05, + "loss": 1.291, + "step": 11462 + }, + { + "epoch": 0.28, + "learning_rate": 1.699827303720211e-05, + "loss": 1.0654, + "step": 11463 + }, + { + "epoch": 0.28, + "learning_rate": 1.6997716642300156e-05, + "loss": 1.1799, + "step": 11464 + }, + { + "epoch": 0.28, + "learning_rate": 1.6997160204944698e-05, + "loss": 1.1719, + "step": 11465 + }, + { + "epoch": 0.28, + "learning_rate": 1.6996603725139117e-05, + "loss": 1.1991, + "step": 11466 + }, + { + "epoch": 0.28, + "learning_rate": 1.6996047202886786e-05, + "loss": 1.2748, + "step": 11467 + }, + { + "epoch": 0.28, + "learning_rate": 1.699549063819108e-05, + "loss": 1.2438, + "step": 11468 + }, + { + "epoch": 0.28, + "learning_rate": 1.699493403105538e-05, + "loss": 1.424, + "step": 11469 + }, + { + "epoch": 0.28, + "learning_rate": 1.699437738148306e-05, + "loss": 1.1998, + "step": 11470 + }, + { + "epoch": 0.28, + "learning_rate": 1.6993820689477502e-05, + "loss": 0.8646, + "step": 11471 + }, + { + "epoch": 0.28, + "learning_rate": 1.6993263955042073e-05, + "loss": 1.259, + "step": 11472 + }, + { + "epoch": 0.28, + "learning_rate": 1.699270717818016e-05, + "loss": 1.1636, + "step": 11473 + }, + { + "epoch": 0.28, + "learning_rate": 1.6992150358895134e-05, + "loss": 1.1051, + "step": 11474 + }, + { + "epoch": 0.28, + "learning_rate": 1.699159349719038e-05, + "loss": 1.0946, + "step": 11475 + }, + { + "epoch": 0.28, + "learning_rate": 1.699103659306927e-05, + "loss": 1.0573, + "step": 11476 + }, + { + "epoch": 0.28, + "learning_rate": 1.699047964653519e-05, + "loss": 0.9306, + "step": 11477 + }, + { + "epoch": 0.28, + "learning_rate": 1.6989922657591506e-05, + "loss": 1.1222, + "step": 11478 + }, + { + "epoch": 0.28, + "learning_rate": 1.698936562624161e-05, + "loss": 0.9338, + "step": 11479 + }, + { + "epoch": 0.28, + "learning_rate": 1.698880855248888e-05, + "loss": 1.083, + "step": 11480 + }, + { + "epoch": 0.28, + "learning_rate": 1.6988251436336684e-05, + "loss": 1.0573, + "step": 11481 + }, + { + "epoch": 0.28, + "learning_rate": 1.698769427778842e-05, + "loss": 1.0776, + "step": 11482 + }, + { + "epoch": 0.28, + "learning_rate": 1.698713707684745e-05, + "loss": 0.8876, + "step": 11483 + }, + { + "epoch": 0.28, + "learning_rate": 1.6986579833517168e-05, + "loss": 0.9688, + "step": 11484 + }, + { + "epoch": 0.28, + "learning_rate": 1.6986022547800947e-05, + "loss": 0.9688, + "step": 11485 + }, + { + "epoch": 0.28, + "learning_rate": 1.6985465219702172e-05, + "loss": 0.9088, + "step": 11486 + }, + { + "epoch": 0.28, + "learning_rate": 1.6984907849224225e-05, + "loss": 1.225, + "step": 11487 + }, + { + "epoch": 0.28, + "learning_rate": 1.698435043637048e-05, + "loss": 0.9505, + "step": 11488 + }, + { + "epoch": 0.28, + "learning_rate": 1.6983792981144324e-05, + "loss": 1.0183, + "step": 11489 + }, + { + "epoch": 0.28, + "learning_rate": 1.6983235483549144e-05, + "loss": 1.078, + "step": 11490 + }, + { + "epoch": 0.28, + "learning_rate": 1.6982677943588313e-05, + "loss": 0.9228, + "step": 11491 + }, + { + "epoch": 0.28, + "learning_rate": 1.6982120361265216e-05, + "loss": 1.1514, + "step": 11492 + }, + { + "epoch": 0.28, + "learning_rate": 1.698156273658324e-05, + "loss": 1.1965, + "step": 11493 + }, + { + "epoch": 0.28, + "learning_rate": 1.6981005069545765e-05, + "loss": 1.116, + "step": 11494 + }, + { + "epoch": 0.28, + "learning_rate": 1.6980447360156172e-05, + "loss": 1.2856, + "step": 11495 + }, + { + "epoch": 0.28, + "learning_rate": 1.6979889608417847e-05, + "loss": 1.0246, + "step": 11496 + }, + { + "epoch": 0.28, + "learning_rate": 1.6979331814334175e-05, + "loss": 1.1791, + "step": 11497 + }, + { + "epoch": 0.28, + "learning_rate": 1.6978773977908537e-05, + "loss": 1.1772, + "step": 11498 + }, + { + "epoch": 0.28, + "learning_rate": 1.6978216099144322e-05, + "loss": 1.129, + "step": 11499 + }, + { + "epoch": 0.28, + "learning_rate": 1.697765817804491e-05, + "loss": 1.1419, + "step": 11500 + }, + { + "epoch": 0.28, + "learning_rate": 1.6977100214613684e-05, + "loss": 1.2022, + "step": 11501 + }, + { + "epoch": 0.28, + "learning_rate": 1.6976542208854036e-05, + "loss": 1.1061, + "step": 11502 + }, + { + "epoch": 0.28, + "learning_rate": 1.6975984160769345e-05, + "loss": 1.0679, + "step": 11503 + }, + { + "epoch": 0.28, + "learning_rate": 1.6975426070362998e-05, + "loss": 1.1918, + "step": 11504 + }, + { + "epoch": 0.28, + "learning_rate": 1.6974867937638383e-05, + "loss": 1.2699, + "step": 11505 + }, + { + "epoch": 0.28, + "learning_rate": 1.697430976259889e-05, + "loss": 1.2476, + "step": 11506 + }, + { + "epoch": 0.28, + "learning_rate": 1.697375154524789e-05, + "loss": 1.1183, + "step": 11507 + }, + { + "epoch": 0.28, + "learning_rate": 1.6973193285588786e-05, + "loss": 1.2331, + "step": 11508 + }, + { + "epoch": 0.28, + "learning_rate": 1.697263498362496e-05, + "loss": 1.1548, + "step": 11509 + }, + { + "epoch": 0.28, + "learning_rate": 1.6972076639359795e-05, + "loss": 1.2908, + "step": 11510 + }, + { + "epoch": 0.28, + "learning_rate": 1.697151825279668e-05, + "loss": 1.1927, + "step": 11511 + }, + { + "epoch": 0.28, + "learning_rate": 1.6970959823939006e-05, + "loss": 0.8905, + "step": 11512 + }, + { + "epoch": 0.28, + "learning_rate": 1.6970401352790156e-05, + "loss": 1.2326, + "step": 11513 + }, + { + "epoch": 0.28, + "learning_rate": 1.696984283935352e-05, + "loss": 1.0647, + "step": 11514 + }, + { + "epoch": 0.28, + "learning_rate": 1.696928428363249e-05, + "loss": 0.9585, + "step": 11515 + }, + { + "epoch": 0.28, + "learning_rate": 1.696872568563045e-05, + "loss": 1.0506, + "step": 11516 + }, + { + "epoch": 0.28, + "learning_rate": 1.696816704535079e-05, + "loss": 0.7532, + "step": 11517 + }, + { + "epoch": 0.28, + "learning_rate": 1.69676083627969e-05, + "loss": 0.9882, + "step": 11518 + }, + { + "epoch": 0.28, + "learning_rate": 1.6967049637972168e-05, + "loss": 1.2114, + "step": 11519 + }, + { + "epoch": 0.28, + "learning_rate": 1.6966490870879984e-05, + "loss": 1.2021, + "step": 11520 + }, + { + "epoch": 0.28, + "learning_rate": 1.696593206152374e-05, + "loss": 1.2839, + "step": 11521 + }, + { + "epoch": 0.28, + "learning_rate": 1.6965373209906827e-05, + "loss": 1.1629, + "step": 11522 + }, + { + "epoch": 0.28, + "learning_rate": 1.6964814316032633e-05, + "loss": 0.9641, + "step": 11523 + }, + { + "epoch": 0.28, + "learning_rate": 1.6964255379904546e-05, + "loss": 1.0922, + "step": 11524 + }, + { + "epoch": 0.28, + "learning_rate": 1.696369640152596e-05, + "loss": 1.0295, + "step": 11525 + }, + { + "epoch": 0.28, + "learning_rate": 1.6963137380900265e-05, + "loss": 1.1803, + "step": 11526 + }, + { + "epoch": 0.28, + "learning_rate": 1.6962578318030852e-05, + "loss": 1.2493, + "step": 11527 + }, + { + "epoch": 0.28, + "learning_rate": 1.6962019212921118e-05, + "loss": 0.9854, + "step": 11528 + }, + { + "epoch": 0.28, + "learning_rate": 1.6961460065574452e-05, + "loss": 1.256, + "step": 11529 + }, + { + "epoch": 0.28, + "learning_rate": 1.6960900875994243e-05, + "loss": 1.2159, + "step": 11530 + }, + { + "epoch": 0.28, + "learning_rate": 1.6960341644183884e-05, + "loss": 0.984, + "step": 11531 + }, + { + "epoch": 0.28, + "learning_rate": 1.695978237014677e-05, + "loss": 1.008, + "step": 11532 + }, + { + "epoch": 0.28, + "learning_rate": 1.6959223053886296e-05, + "loss": 1.0476, + "step": 11533 + }, + { + "epoch": 0.28, + "learning_rate": 1.695866369540585e-05, + "loss": 0.9939, + "step": 11534 + }, + { + "epoch": 0.28, + "learning_rate": 1.6958104294708828e-05, + "loss": 1.1122, + "step": 11535 + }, + { + "epoch": 0.28, + "learning_rate": 1.6957544851798628e-05, + "loss": 1.1185, + "step": 11536 + }, + { + "epoch": 0.28, + "learning_rate": 1.6956985366678634e-05, + "loss": 1.1823, + "step": 11537 + }, + { + "epoch": 0.28, + "learning_rate": 1.695642583935225e-05, + "loss": 1.0316, + "step": 11538 + }, + { + "epoch": 0.28, + "learning_rate": 1.6955866269822865e-05, + "loss": 1.2693, + "step": 11539 + }, + { + "epoch": 0.28, + "learning_rate": 1.6955306658093876e-05, + "loss": 1.0577, + "step": 11540 + }, + { + "epoch": 0.28, + "learning_rate": 1.6954747004168677e-05, + "loss": 1.2329, + "step": 11541 + }, + { + "epoch": 0.28, + "learning_rate": 1.6954187308050666e-05, + "loss": 1.3636, + "step": 11542 + }, + { + "epoch": 0.28, + "learning_rate": 1.6953627569743234e-05, + "loss": 1.2505, + "step": 11543 + }, + { + "epoch": 0.28, + "learning_rate": 1.695306778924978e-05, + "loss": 0.8889, + "step": 11544 + }, + { + "epoch": 0.28, + "learning_rate": 1.69525079665737e-05, + "loss": 1.1049, + "step": 11545 + }, + { + "epoch": 0.28, + "learning_rate": 1.6951948101718388e-05, + "loss": 0.9085, + "step": 11546 + }, + { + "epoch": 0.28, + "learning_rate": 1.6951388194687242e-05, + "loss": 0.9642, + "step": 11547 + }, + { + "epoch": 0.28, + "learning_rate": 1.695082824548366e-05, + "loss": 1.0413, + "step": 11548 + }, + { + "epoch": 0.28, + "learning_rate": 1.6950268254111038e-05, + "loss": 1.4507, + "step": 11549 + }, + { + "epoch": 0.28, + "learning_rate": 1.694970822057277e-05, + "loss": 1.1778, + "step": 11550 + }, + { + "epoch": 0.28, + "learning_rate": 1.694914814487226e-05, + "loss": 1.1265, + "step": 11551 + }, + { + "epoch": 0.28, + "learning_rate": 1.6948588027012903e-05, + "loss": 0.9198, + "step": 11552 + }, + { + "epoch": 0.28, + "learning_rate": 1.6948027866998098e-05, + "loss": 1.0849, + "step": 11553 + }, + { + "epoch": 0.28, + "learning_rate": 1.694746766483124e-05, + "loss": 0.9263, + "step": 11554 + }, + { + "epoch": 0.28, + "learning_rate": 1.6946907420515732e-05, + "loss": 1.4192, + "step": 11555 + }, + { + "epoch": 0.28, + "learning_rate": 1.6946347134054967e-05, + "loss": 1.1304, + "step": 11556 + }, + { + "epoch": 0.28, + "learning_rate": 1.694578680545235e-05, + "loss": 1.0735, + "step": 11557 + }, + { + "epoch": 0.28, + "learning_rate": 1.6945226434711277e-05, + "loss": 0.998, + "step": 11558 + }, + { + "epoch": 0.28, + "learning_rate": 1.694466602183515e-05, + "loss": 0.9716, + "step": 11559 + }, + { + "epoch": 0.28, + "learning_rate": 1.694410556682737e-05, + "loss": 1.1159, + "step": 11560 + }, + { + "epoch": 0.28, + "learning_rate": 1.694354506969133e-05, + "loss": 1.1146, + "step": 11561 + }, + { + "epoch": 0.28, + "learning_rate": 1.694298453043044e-05, + "loss": 1.2012, + "step": 11562 + }, + { + "epoch": 0.28, + "learning_rate": 1.6942423949048093e-05, + "loss": 1.2597, + "step": 11563 + }, + { + "epoch": 0.28, + "learning_rate": 1.6941863325547694e-05, + "loss": 1.1171, + "step": 11564 + }, + { + "epoch": 0.28, + "learning_rate": 1.6941302659932645e-05, + "loss": 1.0602, + "step": 11565 + }, + { + "epoch": 0.28, + "learning_rate": 1.6940741952206342e-05, + "loss": 0.931, + "step": 11566 + }, + { + "epoch": 0.28, + "learning_rate": 1.694018120237219e-05, + "loss": 1.2064, + "step": 11567 + }, + { + "epoch": 0.28, + "learning_rate": 1.693962041043359e-05, + "loss": 1.1633, + "step": 11568 + }, + { + "epoch": 0.28, + "learning_rate": 1.693905957639395e-05, + "loss": 0.7072, + "step": 11569 + }, + { + "epoch": 0.28, + "learning_rate": 1.6938498700256664e-05, + "loss": 1.1291, + "step": 11570 + }, + { + "epoch": 0.28, + "learning_rate": 1.693793778202514e-05, + "loss": 1.1491, + "step": 11571 + }, + { + "epoch": 0.28, + "learning_rate": 1.693737682170278e-05, + "loss": 1.116, + "step": 11572 + }, + { + "epoch": 0.28, + "learning_rate": 1.6936815819292985e-05, + "loss": 1.2162, + "step": 11573 + }, + { + "epoch": 0.28, + "learning_rate": 1.693625477479916e-05, + "loss": 1.0861, + "step": 11574 + }, + { + "epoch": 0.28, + "learning_rate": 1.693569368822471e-05, + "loss": 1.2675, + "step": 11575 + }, + { + "epoch": 0.28, + "learning_rate": 1.693513255957304e-05, + "loss": 1.0278, + "step": 11576 + }, + { + "epoch": 0.28, + "learning_rate": 1.6934571388847547e-05, + "loss": 1.2175, + "step": 11577 + }, + { + "epoch": 0.28, + "learning_rate": 1.6934010176051642e-05, + "loss": 1.0299, + "step": 11578 + }, + { + "epoch": 0.28, + "learning_rate": 1.6933448921188728e-05, + "loss": 0.9276, + "step": 11579 + }, + { + "epoch": 0.28, + "learning_rate": 1.6932887624262214e-05, + "loss": 1.2724, + "step": 11580 + }, + { + "epoch": 0.28, + "learning_rate": 1.6932326285275497e-05, + "loss": 0.8277, + "step": 11581 + }, + { + "epoch": 0.28, + "learning_rate": 1.693176490423199e-05, + "loss": 1.1043, + "step": 11582 + }, + { + "epoch": 0.28, + "learning_rate": 1.6931203481135093e-05, + "loss": 1.0975, + "step": 11583 + }, + { + "epoch": 0.28, + "learning_rate": 1.6930642015988216e-05, + "loss": 1.0218, + "step": 11584 + }, + { + "epoch": 0.28, + "learning_rate": 1.6930080508794766e-05, + "loss": 0.8943, + "step": 11585 + }, + { + "epoch": 0.28, + "learning_rate": 1.6929518959558144e-05, + "loss": 1.138, + "step": 11586 + }, + { + "epoch": 0.28, + "learning_rate": 1.6928957368281766e-05, + "loss": 1.009, + "step": 11587 + }, + { + "epoch": 0.28, + "learning_rate": 1.6928395734969026e-05, + "loss": 1.0982, + "step": 11588 + }, + { + "epoch": 0.28, + "learning_rate": 1.6927834059623344e-05, + "loss": 1.1158, + "step": 11589 + }, + { + "epoch": 0.28, + "learning_rate": 1.692727234224812e-05, + "loss": 1.2408, + "step": 11590 + }, + { + "epoch": 0.28, + "learning_rate": 1.6926710582846765e-05, + "loss": 1.1255, + "step": 11591 + }, + { + "epoch": 0.28, + "learning_rate": 1.6926148781422685e-05, + "loss": 1.3719, + "step": 11592 + }, + { + "epoch": 0.28, + "learning_rate": 1.6925586937979296e-05, + "loss": 1.1635, + "step": 11593 + }, + { + "epoch": 0.28, + "learning_rate": 1.6925025052519993e-05, + "loss": 1.0161, + "step": 11594 + }, + { + "epoch": 0.28, + "learning_rate": 1.6924463125048195e-05, + "loss": 0.8972, + "step": 11595 + }, + { + "epoch": 0.28, + "learning_rate": 1.6923901155567303e-05, + "loss": 1.0543, + "step": 11596 + }, + { + "epoch": 0.28, + "learning_rate": 1.6923339144080736e-05, + "loss": 1.0631, + "step": 11597 + }, + { + "epoch": 0.28, + "learning_rate": 1.6922777090591895e-05, + "loss": 1.0353, + "step": 11598 + }, + { + "epoch": 0.28, + "learning_rate": 1.6922214995104197e-05, + "loss": 1.0907, + "step": 11599 + }, + { + "epoch": 0.28, + "learning_rate": 1.692165285762105e-05, + "loss": 1.161, + "step": 11600 + }, + { + "epoch": 0.28, + "learning_rate": 1.6921090678145862e-05, + "loss": 1.209, + "step": 11601 + }, + { + "epoch": 0.28, + "learning_rate": 1.6920528456682044e-05, + "loss": 1.0455, + "step": 11602 + }, + { + "epoch": 0.28, + "learning_rate": 1.6919966193233007e-05, + "loss": 1.1239, + "step": 11603 + }, + { + "epoch": 0.28, + "learning_rate": 1.6919403887802165e-05, + "loss": 1.3309, + "step": 11604 + }, + { + "epoch": 0.28, + "learning_rate": 1.6918841540392926e-05, + "loss": 1.1311, + "step": 11605 + }, + { + "epoch": 0.28, + "learning_rate": 1.69182791510087e-05, + "loss": 1.072, + "step": 11606 + }, + { + "epoch": 0.28, + "learning_rate": 1.6917716719652907e-05, + "loss": 1.2648, + "step": 11607 + }, + { + "epoch": 0.28, + "learning_rate": 1.691715424632895e-05, + "loss": 1.2622, + "step": 11608 + }, + { + "epoch": 0.28, + "learning_rate": 1.6916591731040247e-05, + "loss": 1.2188, + "step": 11609 + }, + { + "epoch": 0.28, + "learning_rate": 1.6916029173790207e-05, + "loss": 1.1916, + "step": 11610 + }, + { + "epoch": 0.28, + "learning_rate": 1.6915466574582246e-05, + "loss": 1.0049, + "step": 11611 + }, + { + "epoch": 0.28, + "learning_rate": 1.6914903933419775e-05, + "loss": 1.0814, + "step": 11612 + }, + { + "epoch": 0.28, + "learning_rate": 1.691434125030621e-05, + "loss": 1.0585, + "step": 11613 + }, + { + "epoch": 0.28, + "learning_rate": 1.691377852524496e-05, + "loss": 1.0973, + "step": 11614 + }, + { + "epoch": 0.28, + "learning_rate": 1.6913215758239446e-05, + "loss": 1.1142, + "step": 11615 + }, + { + "epoch": 0.28, + "learning_rate": 1.6912652949293077e-05, + "loss": 1.2527, + "step": 11616 + }, + { + "epoch": 0.28, + "learning_rate": 1.6912090098409267e-05, + "loss": 0.9434, + "step": 11617 + }, + { + "epoch": 0.28, + "learning_rate": 1.691152720559143e-05, + "loss": 1.2696, + "step": 11618 + }, + { + "epoch": 0.28, + "learning_rate": 1.6910964270842986e-05, + "loss": 0.8801, + "step": 11619 + }, + { + "epoch": 0.28, + "learning_rate": 1.691040129416735e-05, + "loss": 1.2073, + "step": 11620 + }, + { + "epoch": 0.28, + "learning_rate": 1.690983827556793e-05, + "loss": 1.1932, + "step": 11621 + }, + { + "epoch": 0.28, + "learning_rate": 1.690927521504815e-05, + "loss": 1.2712, + "step": 11622 + }, + { + "epoch": 0.28, + "learning_rate": 1.690871211261142e-05, + "loss": 1.077, + "step": 11623 + }, + { + "epoch": 0.28, + "learning_rate": 1.6908148968261156e-05, + "loss": 0.8205, + "step": 11624 + }, + { + "epoch": 0.28, + "learning_rate": 1.6907585782000783e-05, + "loss": 1.1341, + "step": 11625 + }, + { + "epoch": 0.28, + "learning_rate": 1.690702255383371e-05, + "loss": 0.9803, + "step": 11626 + }, + { + "epoch": 0.28, + "learning_rate": 1.6906459283763355e-05, + "loss": 0.7635, + "step": 11627 + }, + { + "epoch": 0.28, + "learning_rate": 1.6905895971793135e-05, + "loss": 0.9125, + "step": 11628 + }, + { + "epoch": 0.28, + "learning_rate": 1.6905332617926473e-05, + "loss": 0.9806, + "step": 11629 + }, + { + "epoch": 0.28, + "learning_rate": 1.6904769222166777e-05, + "loss": 1.0221, + "step": 11630 + }, + { + "epoch": 0.28, + "learning_rate": 1.690420578451747e-05, + "loss": 1.0494, + "step": 11631 + }, + { + "epoch": 0.28, + "learning_rate": 1.690364230498197e-05, + "loss": 1.2378, + "step": 11632 + }, + { + "epoch": 0.28, + "learning_rate": 1.6903078783563702e-05, + "loss": 1.1273, + "step": 11633 + }, + { + "epoch": 0.28, + "learning_rate": 1.6902515220266077e-05, + "loss": 1.1957, + "step": 11634 + }, + { + "epoch": 0.28, + "learning_rate": 1.6901951615092515e-05, + "loss": 1.128, + "step": 11635 + }, + { + "epoch": 0.28, + "learning_rate": 1.6901387968046435e-05, + "loss": 1.4099, + "step": 11636 + }, + { + "epoch": 0.28, + "learning_rate": 1.690082427913126e-05, + "loss": 0.986, + "step": 11637 + }, + { + "epoch": 0.28, + "learning_rate": 1.6900260548350403e-05, + "loss": 1.1106, + "step": 11638 + }, + { + "epoch": 0.28, + "learning_rate": 1.689969677570729e-05, + "loss": 1.208, + "step": 11639 + }, + { + "epoch": 0.28, + "learning_rate": 1.6899132961205338e-05, + "loss": 0.857, + "step": 11640 + }, + { + "epoch": 0.28, + "learning_rate": 1.6898569104847974e-05, + "loss": 0.9864, + "step": 11641 + }, + { + "epoch": 0.28, + "learning_rate": 1.6898005206638607e-05, + "loss": 1.1068, + "step": 11642 + }, + { + "epoch": 0.28, + "learning_rate": 1.689744126658067e-05, + "loss": 0.8308, + "step": 11643 + }, + { + "epoch": 0.28, + "learning_rate": 1.689687728467758e-05, + "loss": 1.0001, + "step": 11644 + }, + { + "epoch": 0.28, + "learning_rate": 1.6896313260932757e-05, + "loss": 1.337, + "step": 11645 + }, + { + "epoch": 0.28, + "learning_rate": 1.689574919534962e-05, + "loss": 1.1665, + "step": 11646 + }, + { + "epoch": 0.28, + "learning_rate": 1.6895185087931598e-05, + "loss": 1.3046, + "step": 11647 + }, + { + "epoch": 0.28, + "learning_rate": 1.6894620938682106e-05, + "loss": 1.075, + "step": 11648 + }, + { + "epoch": 0.28, + "learning_rate": 1.6894056747604575e-05, + "loss": 1.0317, + "step": 11649 + }, + { + "epoch": 0.28, + "learning_rate": 1.689349251470242e-05, + "loss": 1.2891, + "step": 11650 + }, + { + "epoch": 0.28, + "learning_rate": 1.689292823997907e-05, + "loss": 1.0064, + "step": 11651 + }, + { + "epoch": 0.28, + "learning_rate": 1.689236392343794e-05, + "loss": 0.868, + "step": 11652 + }, + { + "epoch": 0.28, + "learning_rate": 1.6891799565082466e-05, + "loss": 1.1341, + "step": 11653 + }, + { + "epoch": 0.28, + "learning_rate": 1.6891235164916066e-05, + "loss": 0.9932, + "step": 11654 + }, + { + "epoch": 0.28, + "learning_rate": 1.689067072294216e-05, + "loss": 1.38, + "step": 11655 + }, + { + "epoch": 0.28, + "learning_rate": 1.6890106239164173e-05, + "loss": 1.2031, + "step": 11656 + }, + { + "epoch": 0.28, + "learning_rate": 1.6889541713585536e-05, + "loss": 1.3559, + "step": 11657 + }, + { + "epoch": 0.28, + "learning_rate": 1.6888977146209668e-05, + "loss": 1.2747, + "step": 11658 + }, + { + "epoch": 0.28, + "learning_rate": 1.688841253704e-05, + "loss": 0.9881, + "step": 11659 + }, + { + "epoch": 0.28, + "learning_rate": 1.6887847886079946e-05, + "loss": 0.7995, + "step": 11660 + }, + { + "epoch": 0.28, + "learning_rate": 1.6887283193332943e-05, + "loss": 1.2172, + "step": 11661 + }, + { + "epoch": 0.28, + "learning_rate": 1.6886718458802414e-05, + "loss": 0.975, + "step": 11662 + }, + { + "epoch": 0.28, + "learning_rate": 1.6886153682491785e-05, + "loss": 1.1639, + "step": 11663 + }, + { + "epoch": 0.28, + "learning_rate": 1.688558886440448e-05, + "loss": 1.2136, + "step": 11664 + }, + { + "epoch": 0.28, + "learning_rate": 1.6885024004543923e-05, + "loss": 1.1326, + "step": 11665 + }, + { + "epoch": 0.28, + "learning_rate": 1.688445910291355e-05, + "loss": 1.2293, + "step": 11666 + }, + { + "epoch": 0.28, + "learning_rate": 1.688389415951678e-05, + "loss": 1.1512, + "step": 11667 + }, + { + "epoch": 0.28, + "learning_rate": 1.6883329174357044e-05, + "loss": 1.1587, + "step": 11668 + }, + { + "epoch": 0.28, + "learning_rate": 1.688276414743777e-05, + "loss": 1.2416, + "step": 11669 + }, + { + "epoch": 0.28, + "learning_rate": 1.6882199078762385e-05, + "loss": 1.1088, + "step": 11670 + }, + { + "epoch": 0.28, + "learning_rate": 1.6881633968334314e-05, + "loss": 1.2164, + "step": 11671 + }, + { + "epoch": 0.28, + "learning_rate": 1.6881068816156993e-05, + "loss": 1.014, + "step": 11672 + }, + { + "epoch": 0.28, + "learning_rate": 1.688050362223384e-05, + "loss": 1.2094, + "step": 11673 + }, + { + "epoch": 0.28, + "learning_rate": 1.6879938386568295e-05, + "loss": 1.1296, + "step": 11674 + }, + { + "epoch": 0.28, + "learning_rate": 1.687937310916378e-05, + "loss": 1.0127, + "step": 11675 + }, + { + "epoch": 0.28, + "learning_rate": 1.6878807790023728e-05, + "loss": 1.047, + "step": 11676 + }, + { + "epoch": 0.28, + "learning_rate": 1.6878242429151566e-05, + "loss": 1.1623, + "step": 11677 + }, + { + "epoch": 0.28, + "learning_rate": 1.6877677026550725e-05, + "loss": 1.0683, + "step": 11678 + }, + { + "epoch": 0.28, + "learning_rate": 1.6877111582224635e-05, + "loss": 1.195, + "step": 11679 + }, + { + "epoch": 0.28, + "learning_rate": 1.6876546096176728e-05, + "loss": 1.2858, + "step": 11680 + }, + { + "epoch": 0.28, + "learning_rate": 1.6875980568410432e-05, + "loss": 1.0952, + "step": 11681 + }, + { + "epoch": 0.28, + "learning_rate": 1.687541499892918e-05, + "loss": 1.39, + "step": 11682 + }, + { + "epoch": 0.28, + "learning_rate": 1.68748493877364e-05, + "loss": 1.1042, + "step": 11683 + }, + { + "epoch": 0.28, + "learning_rate": 1.6874283734835528e-05, + "loss": 0.9507, + "step": 11684 + }, + { + "epoch": 0.28, + "learning_rate": 1.6873718040229992e-05, + "loss": 1.3094, + "step": 11685 + }, + { + "epoch": 0.28, + "learning_rate": 1.6873152303923226e-05, + "loss": 1.2576, + "step": 11686 + }, + { + "epoch": 0.28, + "learning_rate": 1.6872586525918664e-05, + "loss": 0.9404, + "step": 11687 + }, + { + "epoch": 0.28, + "learning_rate": 1.6872020706219736e-05, + "loss": 1.1435, + "step": 11688 + }, + { + "epoch": 0.28, + "learning_rate": 1.687145484482987e-05, + "loss": 1.154, + "step": 11689 + }, + { + "epoch": 0.28, + "learning_rate": 1.6870888941752507e-05, + "loss": 1.1047, + "step": 11690 + }, + { + "epoch": 0.28, + "learning_rate": 1.6870322996991078e-05, + "loss": 1.3921, + "step": 11691 + }, + { + "epoch": 0.28, + "learning_rate": 1.6869757010549014e-05, + "loss": 1.0099, + "step": 11692 + }, + { + "epoch": 0.28, + "learning_rate": 1.6869190982429748e-05, + "loss": 1.1021, + "step": 11693 + }, + { + "epoch": 0.28, + "learning_rate": 1.686862491263672e-05, + "loss": 1.1798, + "step": 11694 + }, + { + "epoch": 0.28, + "learning_rate": 1.6868058801173357e-05, + "loss": 1.1877, + "step": 11695 + }, + { + "epoch": 0.28, + "learning_rate": 1.68674926480431e-05, + "loss": 1.1278, + "step": 11696 + }, + { + "epoch": 0.28, + "learning_rate": 1.686692645324938e-05, + "loss": 0.9319, + "step": 11697 + }, + { + "epoch": 0.28, + "learning_rate": 1.686636021679563e-05, + "loss": 1.3228, + "step": 11698 + }, + { + "epoch": 0.28, + "learning_rate": 1.6865793938685288e-05, + "loss": 1.1301, + "step": 11699 + }, + { + "epoch": 0.28, + "learning_rate": 1.686522761892179e-05, + "loss": 1.1419, + "step": 11700 + }, + { + "epoch": 0.28, + "learning_rate": 1.686466125750857e-05, + "loss": 1.1021, + "step": 11701 + }, + { + "epoch": 0.28, + "learning_rate": 1.6864094854449063e-05, + "loss": 0.9034, + "step": 11702 + }, + { + "epoch": 0.28, + "learning_rate": 1.686352840974671e-05, + "loss": 1.0972, + "step": 11703 + }, + { + "epoch": 0.28, + "learning_rate": 1.6862961923404945e-05, + "loss": 0.9567, + "step": 11704 + }, + { + "epoch": 0.28, + "learning_rate": 1.6862395395427204e-05, + "loss": 0.7874, + "step": 11705 + }, + { + "epoch": 0.28, + "learning_rate": 1.6861828825816925e-05, + "loss": 1.131, + "step": 11706 + }, + { + "epoch": 0.28, + "learning_rate": 1.6861262214577542e-05, + "loss": 1.3238, + "step": 11707 + }, + { + "epoch": 0.28, + "learning_rate": 1.6860695561712493e-05, + "loss": 1.2106, + "step": 11708 + }, + { + "epoch": 0.28, + "learning_rate": 1.686012886722522e-05, + "loss": 0.9952, + "step": 11709 + }, + { + "epoch": 0.28, + "learning_rate": 1.6859562131119162e-05, + "loss": 0.9554, + "step": 11710 + }, + { + "epoch": 0.28, + "learning_rate": 1.6858995353397753e-05, + "loss": 1.2579, + "step": 11711 + }, + { + "epoch": 0.28, + "learning_rate": 1.6858428534064432e-05, + "loss": 0.9984, + "step": 11712 + }, + { + "epoch": 0.28, + "learning_rate": 1.6857861673122636e-05, + "loss": 1.3319, + "step": 11713 + }, + { + "epoch": 0.28, + "learning_rate": 1.685729477057581e-05, + "loss": 1.2447, + "step": 11714 + }, + { + "epoch": 0.28, + "learning_rate": 1.6856727826427385e-05, + "loss": 1.1455, + "step": 11715 + }, + { + "epoch": 0.28, + "learning_rate": 1.6856160840680807e-05, + "loss": 1.2402, + "step": 11716 + }, + { + "epoch": 0.28, + "learning_rate": 1.6855593813339514e-05, + "loss": 1.2383, + "step": 11717 + }, + { + "epoch": 0.28, + "learning_rate": 1.6855026744406945e-05, + "loss": 1.2854, + "step": 11718 + }, + { + "epoch": 0.28, + "learning_rate": 1.6854459633886543e-05, + "loss": 1.1771, + "step": 11719 + }, + { + "epoch": 0.28, + "learning_rate": 1.685389248178175e-05, + "loss": 1.0172, + "step": 11720 + }, + { + "epoch": 0.28, + "learning_rate": 1.6853325288095995e-05, + "loss": 0.9633, + "step": 11721 + }, + { + "epoch": 0.28, + "learning_rate": 1.6852758052832736e-05, + "loss": 1.1917, + "step": 11722 + }, + { + "epoch": 0.28, + "learning_rate": 1.6852190775995398e-05, + "loss": 1.1859, + "step": 11723 + }, + { + "epoch": 0.28, + "learning_rate": 1.6851623457587435e-05, + "loss": 1.298, + "step": 11724 + }, + { + "epoch": 0.28, + "learning_rate": 1.6851056097612285e-05, + "loss": 1.1239, + "step": 11725 + }, + { + "epoch": 0.28, + "learning_rate": 1.6850488696073385e-05, + "loss": 1.1614, + "step": 11726 + }, + { + "epoch": 0.28, + "learning_rate": 1.6849921252974183e-05, + "loss": 1.1729, + "step": 11727 + }, + { + "epoch": 0.28, + "learning_rate": 1.684935376831812e-05, + "loss": 1.0636, + "step": 11728 + }, + { + "epoch": 0.28, + "learning_rate": 1.684878624210864e-05, + "loss": 1.264, + "step": 11729 + }, + { + "epoch": 0.28, + "learning_rate": 1.6848218674349182e-05, + "loss": 0.993, + "step": 11730 + }, + { + "epoch": 0.28, + "learning_rate": 1.6847651065043196e-05, + "loss": 1.0387, + "step": 11731 + }, + { + "epoch": 0.28, + "learning_rate": 1.684708341419412e-05, + "loss": 1.143, + "step": 11732 + }, + { + "epoch": 0.28, + "learning_rate": 1.6846515721805397e-05, + "loss": 1.3498, + "step": 11733 + }, + { + "epoch": 0.28, + "learning_rate": 1.684594798788048e-05, + "loss": 1.49, + "step": 11734 + }, + { + "epoch": 0.28, + "learning_rate": 1.68453802124228e-05, + "loss": 1.269, + "step": 11735 + }, + { + "epoch": 0.28, + "learning_rate": 1.684481239543581e-05, + "loss": 1.097, + "step": 11736 + }, + { + "epoch": 0.28, + "learning_rate": 1.6844244536922954e-05, + "loss": 1.1233, + "step": 11737 + }, + { + "epoch": 0.28, + "learning_rate": 1.684367663688768e-05, + "loss": 1.0944, + "step": 11738 + }, + { + "epoch": 0.28, + "learning_rate": 1.6843108695333422e-05, + "loss": 1.1664, + "step": 11739 + }, + { + "epoch": 0.28, + "learning_rate": 1.684254071226364e-05, + "loss": 1.4106, + "step": 11740 + }, + { + "epoch": 0.28, + "learning_rate": 1.684197268768177e-05, + "loss": 1.2536, + "step": 11741 + }, + { + "epoch": 0.28, + "learning_rate": 1.684140462159126e-05, + "loss": 1.1869, + "step": 11742 + }, + { + "epoch": 0.28, + "learning_rate": 1.684083651399556e-05, + "loss": 1.0543, + "step": 11743 + }, + { + "epoch": 0.28, + "learning_rate": 1.6840268364898112e-05, + "loss": 0.944, + "step": 11744 + }, + { + "epoch": 0.28, + "learning_rate": 1.6839700174302366e-05, + "loss": 1.0799, + "step": 11745 + }, + { + "epoch": 0.28, + "learning_rate": 1.6839131942211767e-05, + "loss": 1.0354, + "step": 11746 + }, + { + "epoch": 0.28, + "learning_rate": 1.683856366862976e-05, + "loss": 1.2969, + "step": 11747 + }, + { + "epoch": 0.28, + "learning_rate": 1.6837995353559802e-05, + "loss": 1.3255, + "step": 11748 + }, + { + "epoch": 0.28, + "learning_rate": 1.683742699700533e-05, + "loss": 1.1814, + "step": 11749 + }, + { + "epoch": 0.28, + "learning_rate": 1.6836858598969797e-05, + "loss": 1.0214, + "step": 11750 + }, + { + "epoch": 0.28, + "learning_rate": 1.6836290159456655e-05, + "loss": 1.2911, + "step": 11751 + }, + { + "epoch": 0.28, + "learning_rate": 1.6835721678469345e-05, + "loss": 1.336, + "step": 11752 + }, + { + "epoch": 0.28, + "learning_rate": 1.6835153156011318e-05, + "loss": 1.345, + "step": 11753 + }, + { + "epoch": 0.28, + "learning_rate": 1.6834584592086028e-05, + "loss": 1.3216, + "step": 11754 + }, + { + "epoch": 0.28, + "learning_rate": 1.683401598669692e-05, + "loss": 0.8483, + "step": 11755 + }, + { + "epoch": 0.28, + "learning_rate": 1.6833447339847442e-05, + "loss": 1.0446, + "step": 11756 + }, + { + "epoch": 0.28, + "learning_rate": 1.6832878651541052e-05, + "loss": 1.1445, + "step": 11757 + }, + { + "epoch": 0.28, + "learning_rate": 1.683230992178119e-05, + "loss": 1.1992, + "step": 11758 + }, + { + "epoch": 0.28, + "learning_rate": 1.6831741150571313e-05, + "loss": 1.0379, + "step": 11759 + }, + { + "epoch": 0.28, + "learning_rate": 1.6831172337914866e-05, + "loss": 1.1829, + "step": 11760 + }, + { + "epoch": 0.28, + "learning_rate": 1.6830603483815305e-05, + "loss": 0.8945, + "step": 11761 + }, + { + "epoch": 0.28, + "learning_rate": 1.6830034588276077e-05, + "loss": 0.9271, + "step": 11762 + }, + { + "epoch": 0.28, + "learning_rate": 1.682946565130064e-05, + "loss": 1.1196, + "step": 11763 + }, + { + "epoch": 0.28, + "learning_rate": 1.682889667289244e-05, + "loss": 1.1994, + "step": 11764 + }, + { + "epoch": 0.28, + "learning_rate": 1.682832765305493e-05, + "loss": 1.2138, + "step": 11765 + }, + { + "epoch": 0.28, + "learning_rate": 1.6827758591791563e-05, + "loss": 1.0668, + "step": 11766 + }, + { + "epoch": 0.28, + "learning_rate": 1.682718948910579e-05, + "loss": 1.0357, + "step": 11767 + }, + { + "epoch": 0.28, + "learning_rate": 1.682662034500106e-05, + "loss": 0.9725, + "step": 11768 + }, + { + "epoch": 0.28, + "learning_rate": 1.6826051159480838e-05, + "loss": 0.9743, + "step": 11769 + }, + { + "epoch": 0.28, + "learning_rate": 1.6825481932548565e-05, + "loss": 1.1428, + "step": 11770 + }, + { + "epoch": 0.28, + "learning_rate": 1.6824912664207697e-05, + "loss": 1.261, + "step": 11771 + }, + { + "epoch": 0.28, + "learning_rate": 1.682434335446169e-05, + "loss": 1.2285, + "step": 11772 + }, + { + "epoch": 0.28, + "learning_rate": 1.6823774003313998e-05, + "loss": 0.9723, + "step": 11773 + }, + { + "epoch": 0.28, + "learning_rate": 1.6823204610768075e-05, + "loss": 1.2745, + "step": 11774 + }, + { + "epoch": 0.28, + "learning_rate": 1.6822635176827374e-05, + "loss": 1.27, + "step": 11775 + }, + { + "epoch": 0.28, + "learning_rate": 1.6822065701495347e-05, + "loss": 1.1259, + "step": 11776 + }, + { + "epoch": 0.28, + "learning_rate": 1.6821496184775457e-05, + "loss": 1.1951, + "step": 11777 + }, + { + "epoch": 0.28, + "learning_rate": 1.682092662667115e-05, + "loss": 1.1114, + "step": 11778 + }, + { + "epoch": 0.28, + "learning_rate": 1.6820357027185886e-05, + "loss": 1.09, + "step": 11779 + }, + { + "epoch": 0.28, + "learning_rate": 1.6819787386323124e-05, + "loss": 1.1045, + "step": 11780 + }, + { + "epoch": 0.28, + "learning_rate": 1.681921770408631e-05, + "loss": 1.0642, + "step": 11781 + }, + { + "epoch": 0.28, + "learning_rate": 1.6818647980478913e-05, + "loss": 1.2418, + "step": 11782 + }, + { + "epoch": 0.28, + "learning_rate": 1.681807821550438e-05, + "loss": 1.047, + "step": 11783 + }, + { + "epoch": 0.28, + "learning_rate": 1.681750840916617e-05, + "loss": 1.1279, + "step": 11784 + }, + { + "epoch": 0.28, + "learning_rate": 1.6816938561467738e-05, + "loss": 1.2767, + "step": 11785 + }, + { + "epoch": 0.28, + "learning_rate": 1.6816368672412548e-05, + "loss": 1.1797, + "step": 11786 + }, + { + "epoch": 0.28, + "learning_rate": 1.6815798742004047e-05, + "loss": 1.2055, + "step": 11787 + }, + { + "epoch": 0.28, + "learning_rate": 1.68152287702457e-05, + "loss": 1.265, + "step": 11788 + }, + { + "epoch": 0.28, + "learning_rate": 1.681465875714096e-05, + "loss": 1.0718, + "step": 11789 + }, + { + "epoch": 0.28, + "learning_rate": 1.6814088702693294e-05, + "loss": 1.0025, + "step": 11790 + }, + { + "epoch": 0.28, + "learning_rate": 1.681351860690615e-05, + "loss": 0.961, + "step": 11791 + }, + { + "epoch": 0.28, + "learning_rate": 1.6812948469782996e-05, + "loss": 1.1205, + "step": 11792 + }, + { + "epoch": 0.28, + "learning_rate": 1.6812378291327284e-05, + "loss": 1.3966, + "step": 11793 + }, + { + "epoch": 0.28, + "learning_rate": 1.6811808071542475e-05, + "loss": 1.2203, + "step": 11794 + }, + { + "epoch": 0.28, + "learning_rate": 1.6811237810432027e-05, + "loss": 1.0378, + "step": 11795 + }, + { + "epoch": 0.28, + "learning_rate": 1.6810667507999406e-05, + "loss": 1.1305, + "step": 11796 + }, + { + "epoch": 0.28, + "learning_rate": 1.6810097164248063e-05, + "loss": 1.2682, + "step": 11797 + }, + { + "epoch": 0.28, + "learning_rate": 1.680952677918146e-05, + "loss": 1.1673, + "step": 11798 + }, + { + "epoch": 0.28, + "learning_rate": 1.6808956352803067e-05, + "loss": 0.9583, + "step": 11799 + }, + { + "epoch": 0.28, + "learning_rate": 1.6808385885116333e-05, + "loss": 1.0849, + "step": 11800 + }, + { + "epoch": 0.28, + "learning_rate": 1.6807815376124726e-05, + "loss": 1.0247, + "step": 11801 + }, + { + "epoch": 0.28, + "learning_rate": 1.68072448258317e-05, + "loss": 1.0523, + "step": 11802 + }, + { + "epoch": 0.28, + "learning_rate": 1.6806674234240726e-05, + "loss": 1.2234, + "step": 11803 + }, + { + "epoch": 0.28, + "learning_rate": 1.6806103601355256e-05, + "loss": 1.0954, + "step": 11804 + }, + { + "epoch": 0.28, + "learning_rate": 1.680553292717876e-05, + "loss": 0.8948, + "step": 11805 + }, + { + "epoch": 0.28, + "learning_rate": 1.680496221171469e-05, + "loss": 1.2124, + "step": 11806 + }, + { + "epoch": 0.28, + "learning_rate": 1.6804391454966522e-05, + "loss": 1.3922, + "step": 11807 + }, + { + "epoch": 0.28, + "learning_rate": 1.6803820656937708e-05, + "loss": 0.6984, + "step": 11808 + }, + { + "epoch": 0.28, + "learning_rate": 1.6803249817631715e-05, + "loss": 1.1925, + "step": 11809 + }, + { + "epoch": 0.28, + "learning_rate": 1.6802678937052005e-05, + "loss": 1.1168, + "step": 11810 + }, + { + "epoch": 0.28, + "learning_rate": 1.6802108015202044e-05, + "loss": 1.0936, + "step": 11811 + }, + { + "epoch": 0.28, + "learning_rate": 1.6801537052085292e-05, + "loss": 0.8544, + "step": 11812 + }, + { + "epoch": 0.28, + "learning_rate": 1.6800966047705215e-05, + "loss": 1.3412, + "step": 11813 + }, + { + "epoch": 0.28, + "learning_rate": 1.680039500206528e-05, + "loss": 1.1675, + "step": 11814 + }, + { + "epoch": 0.28, + "learning_rate": 1.6799823915168943e-05, + "loss": 1.0943, + "step": 11815 + }, + { + "epoch": 0.28, + "learning_rate": 1.6799252787019676e-05, + "loss": 1.1385, + "step": 11816 + }, + { + "epoch": 0.28, + "learning_rate": 1.679868161762094e-05, + "loss": 0.9887, + "step": 11817 + }, + { + "epoch": 0.28, + "learning_rate": 1.6798110406976202e-05, + "loss": 0.7603, + "step": 11818 + }, + { + "epoch": 0.28, + "learning_rate": 1.679753915508893e-05, + "loss": 0.968, + "step": 11819 + }, + { + "epoch": 0.28, + "learning_rate": 1.6796967861962582e-05, + "loss": 1.1094, + "step": 11820 + }, + { + "epoch": 0.28, + "learning_rate": 1.6796396527600634e-05, + "loss": 1.2409, + "step": 11821 + }, + { + "epoch": 0.28, + "learning_rate": 1.6795825152006545e-05, + "loss": 1.2351, + "step": 11822 + }, + { + "epoch": 0.28, + "learning_rate": 1.679525373518378e-05, + "loss": 1.1494, + "step": 11823 + }, + { + "epoch": 0.28, + "learning_rate": 1.6794682277135815e-05, + "loss": 0.7948, + "step": 11824 + }, + { + "epoch": 0.28, + "learning_rate": 1.6794110777866107e-05, + "loss": 0.9214, + "step": 11825 + }, + { + "epoch": 0.28, + "learning_rate": 1.679353923737813e-05, + "loss": 1.2819, + "step": 11826 + }, + { + "epoch": 0.28, + "learning_rate": 1.6792967655675343e-05, + "loss": 1.1498, + "step": 11827 + }, + { + "epoch": 0.28, + "learning_rate": 1.679239603276122e-05, + "loss": 1.1389, + "step": 11828 + }, + { + "epoch": 0.28, + "learning_rate": 1.679182436863923e-05, + "loss": 0.8264, + "step": 11829 + }, + { + "epoch": 0.28, + "learning_rate": 1.679125266331284e-05, + "loss": 0.8232, + "step": 11830 + }, + { + "epoch": 0.28, + "learning_rate": 1.6790680916785514e-05, + "loss": 0.9602, + "step": 11831 + }, + { + "epoch": 0.28, + "learning_rate": 1.679010912906073e-05, + "loss": 1.182, + "step": 11832 + }, + { + "epoch": 0.28, + "learning_rate": 1.6789537300141944e-05, + "loss": 0.9865, + "step": 11833 + }, + { + "epoch": 0.28, + "learning_rate": 1.6788965430032636e-05, + "loss": 1.1991, + "step": 11834 + }, + { + "epoch": 0.28, + "learning_rate": 1.6788393518736273e-05, + "loss": 1.2183, + "step": 11835 + }, + { + "epoch": 0.28, + "learning_rate": 1.678782156625632e-05, + "loss": 1.1985, + "step": 11836 + }, + { + "epoch": 0.28, + "learning_rate": 1.6787249572596253e-05, + "loss": 1.1177, + "step": 11837 + }, + { + "epoch": 0.28, + "learning_rate": 1.678667753775954e-05, + "loss": 1.0618, + "step": 11838 + }, + { + "epoch": 0.28, + "learning_rate": 1.678610546174965e-05, + "loss": 1.3962, + "step": 11839 + }, + { + "epoch": 0.28, + "learning_rate": 1.6785533344570053e-05, + "loss": 1.2197, + "step": 11840 + }, + { + "epoch": 0.28, + "learning_rate": 1.678496118622422e-05, + "loss": 1.1948, + "step": 11841 + }, + { + "epoch": 0.28, + "learning_rate": 1.6784388986715625e-05, + "loss": 1.1016, + "step": 11842 + }, + { + "epoch": 0.28, + "learning_rate": 1.678381674604774e-05, + "loss": 1.1421, + "step": 11843 + }, + { + "epoch": 0.28, + "learning_rate": 1.6783244464224033e-05, + "loss": 1.1891, + "step": 11844 + }, + { + "epoch": 0.28, + "learning_rate": 1.6782672141247973e-05, + "loss": 1.0109, + "step": 11845 + }, + { + "epoch": 0.28, + "learning_rate": 1.678209977712304e-05, + "loss": 1.2011, + "step": 11846 + }, + { + "epoch": 0.28, + "learning_rate": 1.67815273718527e-05, + "loss": 1.1123, + "step": 11847 + }, + { + "epoch": 0.28, + "learning_rate": 1.6780954925440434e-05, + "loss": 1.0793, + "step": 11848 + }, + { + "epoch": 0.28, + "learning_rate": 1.6780382437889708e-05, + "loss": 1.0825, + "step": 11849 + }, + { + "epoch": 0.28, + "learning_rate": 1.6779809909203996e-05, + "loss": 1.1537, + "step": 11850 + }, + { + "epoch": 0.29, + "learning_rate": 1.677923733938677e-05, + "loss": 1.0365, + "step": 11851 + }, + { + "epoch": 0.29, + "learning_rate": 1.6778664728441508e-05, + "loss": 1.0931, + "step": 11852 + }, + { + "epoch": 0.29, + "learning_rate": 1.6778092076371678e-05, + "loss": 0.919, + "step": 11853 + }, + { + "epoch": 0.29, + "learning_rate": 1.6777519383180763e-05, + "loss": 0.9854, + "step": 11854 + }, + { + "epoch": 0.29, + "learning_rate": 1.677694664887223e-05, + "loss": 1.0707, + "step": 11855 + }, + { + "epoch": 0.29, + "learning_rate": 1.6776373873449554e-05, + "loss": 1.3481, + "step": 11856 + }, + { + "epoch": 0.29, + "learning_rate": 1.6775801056916212e-05, + "loss": 1.2938, + "step": 11857 + }, + { + "epoch": 0.29, + "learning_rate": 1.677522819927568e-05, + "loss": 1.332, + "step": 11858 + }, + { + "epoch": 0.29, + "learning_rate": 1.677465530053143e-05, + "loss": 1.255, + "step": 11859 + }, + { + "epoch": 0.29, + "learning_rate": 1.6774082360686944e-05, + "loss": 1.2952, + "step": 11860 + }, + { + "epoch": 0.29, + "learning_rate": 1.677350937974569e-05, + "loss": 1.1632, + "step": 11861 + }, + { + "epoch": 0.29, + "learning_rate": 1.677293635771115e-05, + "loss": 1.2131, + "step": 11862 + }, + { + "epoch": 0.29, + "learning_rate": 1.6772363294586795e-05, + "loss": 1.24, + "step": 11863 + }, + { + "epoch": 0.29, + "learning_rate": 1.677179019037611e-05, + "loss": 1.3089, + "step": 11864 + }, + { + "epoch": 0.29, + "learning_rate": 1.677121704508256e-05, + "loss": 0.9042, + "step": 11865 + }, + { + "epoch": 0.29, + "learning_rate": 1.6770643858709635e-05, + "loss": 1.205, + "step": 11866 + }, + { + "epoch": 0.29, + "learning_rate": 1.6770070631260803e-05, + "loss": 0.8914, + "step": 11867 + }, + { + "epoch": 0.29, + "learning_rate": 1.6769497362739543e-05, + "loss": 1.1318, + "step": 11868 + }, + { + "epoch": 0.29, + "learning_rate": 1.676892405314934e-05, + "loss": 0.9285, + "step": 11869 + }, + { + "epoch": 0.29, + "learning_rate": 1.6768350702493663e-05, + "loss": 1.0155, + "step": 11870 + }, + { + "epoch": 0.29, + "learning_rate": 1.6767777310775998e-05, + "loss": 0.9447, + "step": 11871 + }, + { + "epoch": 0.29, + "learning_rate": 1.6767203877999817e-05, + "loss": 1.1699, + "step": 11872 + }, + { + "epoch": 0.29, + "learning_rate": 1.6766630404168603e-05, + "loss": 1.0088, + "step": 11873 + }, + { + "epoch": 0.29, + "learning_rate": 1.6766056889285832e-05, + "loss": 0.9784, + "step": 11874 + }, + { + "epoch": 0.29, + "learning_rate": 1.6765483333354984e-05, + "loss": 1.0778, + "step": 11875 + }, + { + "epoch": 0.29, + "learning_rate": 1.6764909736379545e-05, + "loss": 1.3065, + "step": 11876 + }, + { + "epoch": 0.29, + "learning_rate": 1.676433609836299e-05, + "loss": 1.1237, + "step": 11877 + }, + { + "epoch": 0.29, + "learning_rate": 1.6763762419308793e-05, + "loss": 0.9984, + "step": 11878 + }, + { + "epoch": 0.29, + "learning_rate": 1.6763188699220446e-05, + "loss": 0.7429, + "step": 11879 + }, + { + "epoch": 0.29, + "learning_rate": 1.676261493810142e-05, + "loss": 1.0244, + "step": 11880 + }, + { + "epoch": 0.29, + "learning_rate": 1.67620411359552e-05, + "loss": 1.0556, + "step": 11881 + }, + { + "epoch": 0.29, + "learning_rate": 1.6761467292785268e-05, + "loss": 1.2235, + "step": 11882 + }, + { + "epoch": 0.29, + "learning_rate": 1.6760893408595106e-05, + "loss": 0.9794, + "step": 11883 + }, + { + "epoch": 0.29, + "learning_rate": 1.6760319483388193e-05, + "loss": 1.0145, + "step": 11884 + }, + { + "epoch": 0.29, + "learning_rate": 1.6759745517168012e-05, + "loss": 1.1278, + "step": 11885 + }, + { + "epoch": 0.29, + "learning_rate": 1.6759171509938044e-05, + "loss": 1.0626, + "step": 11886 + }, + { + "epoch": 0.29, + "learning_rate": 1.6758597461701774e-05, + "loss": 1.1812, + "step": 11887 + }, + { + "epoch": 0.29, + "learning_rate": 1.675802337246268e-05, + "loss": 1.1605, + "step": 11888 + }, + { + "epoch": 0.29, + "learning_rate": 1.675744924222425e-05, + "loss": 1.1598, + "step": 11889 + }, + { + "epoch": 0.29, + "learning_rate": 1.6756875070989967e-05, + "loss": 0.7266, + "step": 11890 + }, + { + "epoch": 0.29, + "learning_rate": 1.6756300858763307e-05, + "loss": 1.3502, + "step": 11891 + }, + { + "epoch": 0.29, + "learning_rate": 1.675572660554776e-05, + "loss": 1.0389, + "step": 11892 + }, + { + "epoch": 0.29, + "learning_rate": 1.675515231134681e-05, + "loss": 0.9272, + "step": 11893 + }, + { + "epoch": 0.29, + "learning_rate": 1.6754577976163944e-05, + "loss": 1.0456, + "step": 11894 + }, + { + "epoch": 0.29, + "learning_rate": 1.6754003600002636e-05, + "loss": 1.5585, + "step": 11895 + }, + { + "epoch": 0.29, + "learning_rate": 1.675342918286638e-05, + "loss": 1.2052, + "step": 11896 + }, + { + "epoch": 0.29, + "learning_rate": 1.6752854724758657e-05, + "loss": 1.0596, + "step": 11897 + }, + { + "epoch": 0.29, + "learning_rate": 1.6752280225682954e-05, + "loss": 1.1382, + "step": 11898 + }, + { + "epoch": 0.29, + "learning_rate": 1.6751705685642753e-05, + "loss": 1.2766, + "step": 11899 + }, + { + "epoch": 0.29, + "learning_rate": 1.6751131104641543e-05, + "loss": 1.3179, + "step": 11900 + }, + { + "epoch": 0.29, + "learning_rate": 1.6750556482682807e-05, + "loss": 1.2785, + "step": 11901 + }, + { + "epoch": 0.29, + "learning_rate": 1.6749981819770035e-05, + "loss": 1.0808, + "step": 11902 + }, + { + "epoch": 0.29, + "learning_rate": 1.674940711590671e-05, + "loss": 1.0108, + "step": 11903 + }, + { + "epoch": 0.29, + "learning_rate": 1.674883237109632e-05, + "loss": 1.1268, + "step": 11904 + }, + { + "epoch": 0.29, + "learning_rate": 1.6748257585342348e-05, + "loss": 1.1016, + "step": 11905 + }, + { + "epoch": 0.29, + "learning_rate": 1.6747682758648288e-05, + "loss": 1.1865, + "step": 11906 + }, + { + "epoch": 0.29, + "learning_rate": 1.6747107891017623e-05, + "loss": 0.9072, + "step": 11907 + }, + { + "epoch": 0.29, + "learning_rate": 1.674653298245384e-05, + "loss": 1.1211, + "step": 11908 + }, + { + "epoch": 0.29, + "learning_rate": 1.674595803296043e-05, + "loss": 1.0613, + "step": 11909 + }, + { + "epoch": 0.29, + "learning_rate": 1.674538304254088e-05, + "loss": 1.1757, + "step": 11910 + }, + { + "epoch": 0.29, + "learning_rate": 1.6744808011198678e-05, + "loss": 1.3606, + "step": 11911 + }, + { + "epoch": 0.29, + "learning_rate": 1.674423293893731e-05, + "loss": 1.2219, + "step": 11912 + }, + { + "epoch": 0.29, + "learning_rate": 1.6743657825760266e-05, + "loss": 1.239, + "step": 11913 + }, + { + "epoch": 0.29, + "learning_rate": 1.6743082671671042e-05, + "loss": 1.0324, + "step": 11914 + }, + { + "epoch": 0.29, + "learning_rate": 1.6742507476673114e-05, + "loss": 1.1071, + "step": 11915 + }, + { + "epoch": 0.29, + "learning_rate": 1.6741932240769983e-05, + "loss": 1.1853, + "step": 11916 + }, + { + "epoch": 0.29, + "learning_rate": 1.6741356963965136e-05, + "loss": 1.1596, + "step": 11917 + }, + { + "epoch": 0.29, + "learning_rate": 1.674078164626206e-05, + "loss": 1.0927, + "step": 11918 + }, + { + "epoch": 0.29, + "learning_rate": 1.674020628766425e-05, + "loss": 1.1664, + "step": 11919 + }, + { + "epoch": 0.29, + "learning_rate": 1.6739630888175188e-05, + "loss": 0.9971, + "step": 11920 + }, + { + "epoch": 0.29, + "learning_rate": 1.6739055447798378e-05, + "loss": 0.991, + "step": 11921 + }, + { + "epoch": 0.29, + "learning_rate": 1.6738479966537297e-05, + "loss": 1.2587, + "step": 11922 + }, + { + "epoch": 0.29, + "learning_rate": 1.6737904444395448e-05, + "loss": 0.9447, + "step": 11923 + }, + { + "epoch": 0.29, + "learning_rate": 1.6737328881376315e-05, + "loss": 1.2767, + "step": 11924 + }, + { + "epoch": 0.29, + "learning_rate": 1.6736753277483396e-05, + "loss": 1.209, + "step": 11925 + }, + { + "epoch": 0.29, + "learning_rate": 1.6736177632720175e-05, + "loss": 0.9559, + "step": 11926 + }, + { + "epoch": 0.29, + "learning_rate": 1.673560194709015e-05, + "loss": 1.1398, + "step": 11927 + }, + { + "epoch": 0.29, + "learning_rate": 1.673502622059681e-05, + "loss": 0.7569, + "step": 11928 + }, + { + "epoch": 0.29, + "learning_rate": 1.6734450453243658e-05, + "loss": 1.2624, + "step": 11929 + }, + { + "epoch": 0.29, + "learning_rate": 1.673387464503417e-05, + "loss": 1.432, + "step": 11930 + }, + { + "epoch": 0.29, + "learning_rate": 1.6733298795971855e-05, + "loss": 0.8752, + "step": 11931 + }, + { + "epoch": 0.29, + "learning_rate": 1.6732722906060197e-05, + "loss": 0.9231, + "step": 11932 + }, + { + "epoch": 0.29, + "learning_rate": 1.6732146975302695e-05, + "loss": 1.1715, + "step": 11933 + }, + { + "epoch": 0.29, + "learning_rate": 1.6731571003702837e-05, + "loss": 1.0973, + "step": 11934 + }, + { + "epoch": 0.29, + "learning_rate": 1.6730994991264125e-05, + "loss": 0.9321, + "step": 11935 + }, + { + "epoch": 0.29, + "learning_rate": 1.6730418937990046e-05, + "loss": 1.0791, + "step": 11936 + }, + { + "epoch": 0.29, + "learning_rate": 1.6729842843884097e-05, + "loss": 1.3491, + "step": 11937 + }, + { + "epoch": 0.29, + "learning_rate": 1.6729266708949778e-05, + "loss": 1.0887, + "step": 11938 + }, + { + "epoch": 0.29, + "learning_rate": 1.672869053319058e-05, + "loss": 0.982, + "step": 11939 + }, + { + "epoch": 0.29, + "learning_rate": 1.672811431661e-05, + "loss": 1.1162, + "step": 11940 + }, + { + "epoch": 0.29, + "learning_rate": 1.6727538059211536e-05, + "loss": 1.0222, + "step": 11941 + }, + { + "epoch": 0.29, + "learning_rate": 1.6726961760998673e-05, + "loss": 1.1041, + "step": 11942 + }, + { + "epoch": 0.29, + "learning_rate": 1.6726385421974923e-05, + "loss": 1.1682, + "step": 11943 + }, + { + "epoch": 0.29, + "learning_rate": 1.672580904214377e-05, + "loss": 1.2874, + "step": 11944 + }, + { + "epoch": 0.29, + "learning_rate": 1.672523262150872e-05, + "loss": 1.2005, + "step": 11945 + }, + { + "epoch": 0.29, + "learning_rate": 1.672465616007326e-05, + "loss": 1.1304, + "step": 11946 + }, + { + "epoch": 0.29, + "learning_rate": 1.6724079657840895e-05, + "loss": 1.1127, + "step": 11947 + }, + { + "epoch": 0.29, + "learning_rate": 1.672350311481512e-05, + "loss": 0.8708, + "step": 11948 + }, + { + "epoch": 0.29, + "learning_rate": 1.6722926530999435e-05, + "loss": 1.2002, + "step": 11949 + }, + { + "epoch": 0.29, + "learning_rate": 1.6722349906397334e-05, + "loss": 1.0899, + "step": 11950 + }, + { + "epoch": 0.29, + "learning_rate": 1.6721773241012318e-05, + "loss": 1.1398, + "step": 11951 + }, + { + "epoch": 0.29, + "learning_rate": 1.6721196534847886e-05, + "loss": 1.2577, + "step": 11952 + }, + { + "epoch": 0.29, + "learning_rate": 1.6720619787907533e-05, + "loss": 1.253, + "step": 11953 + }, + { + "epoch": 0.29, + "learning_rate": 1.6720043000194764e-05, + "loss": 1.087, + "step": 11954 + }, + { + "epoch": 0.29, + "learning_rate": 1.6719466171713072e-05, + "loss": 1.1441, + "step": 11955 + }, + { + "epoch": 0.29, + "learning_rate": 1.671888930246596e-05, + "loss": 1.2142, + "step": 11956 + }, + { + "epoch": 0.29, + "learning_rate": 1.6718312392456922e-05, + "loss": 1.0313, + "step": 11957 + }, + { + "epoch": 0.29, + "learning_rate": 1.671773544168947e-05, + "loss": 1.1009, + "step": 11958 + }, + { + "epoch": 0.29, + "learning_rate": 1.6717158450167096e-05, + "loss": 1.0769, + "step": 11959 + }, + { + "epoch": 0.29, + "learning_rate": 1.67165814178933e-05, + "loss": 1.3448, + "step": 11960 + }, + { + "epoch": 0.29, + "learning_rate": 1.6716004344871587e-05, + "loss": 1.3311, + "step": 11961 + }, + { + "epoch": 0.29, + "learning_rate": 1.671542723110545e-05, + "loss": 1.1048, + "step": 11962 + }, + { + "epoch": 0.29, + "learning_rate": 1.67148500765984e-05, + "loss": 1.0153, + "step": 11963 + }, + { + "epoch": 0.29, + "learning_rate": 1.6714272881353933e-05, + "loss": 1.1289, + "step": 11964 + }, + { + "epoch": 0.29, + "learning_rate": 1.6713695645375553e-05, + "loss": 0.9279, + "step": 11965 + }, + { + "epoch": 0.29, + "learning_rate": 1.6713118368666757e-05, + "loss": 1.041, + "step": 11966 + }, + { + "epoch": 0.29, + "learning_rate": 1.671254105123106e-05, + "loss": 1.1372, + "step": 11967 + }, + { + "epoch": 0.29, + "learning_rate": 1.6711963693071943e-05, + "loss": 0.9478, + "step": 11968 + }, + { + "epoch": 0.29, + "learning_rate": 1.671138629419293e-05, + "loss": 1.0051, + "step": 11969 + }, + { + "epoch": 0.29, + "learning_rate": 1.671080885459751e-05, + "loss": 1.1899, + "step": 11970 + }, + { + "epoch": 0.29, + "learning_rate": 1.6710231374289195e-05, + "loss": 1.2726, + "step": 11971 + }, + { + "epoch": 0.29, + "learning_rate": 1.670965385327148e-05, + "loss": 1.3114, + "step": 11972 + }, + { + "epoch": 0.29, + "learning_rate": 1.6709076291547876e-05, + "loss": 1.1753, + "step": 11973 + }, + { + "epoch": 0.29, + "learning_rate": 1.6708498689121885e-05, + "loss": 1.0448, + "step": 11974 + }, + { + "epoch": 0.29, + "learning_rate": 1.670792104599701e-05, + "loss": 1.0545, + "step": 11975 + }, + { + "epoch": 0.29, + "learning_rate": 1.6707343362176753e-05, + "loss": 0.9545, + "step": 11976 + }, + { + "epoch": 0.29, + "learning_rate": 1.6706765637664622e-05, + "loss": 1.1887, + "step": 11977 + }, + { + "epoch": 0.29, + "learning_rate": 1.6706187872464123e-05, + "loss": 1.2875, + "step": 11978 + }, + { + "epoch": 0.29, + "learning_rate": 1.670561006657876e-05, + "loss": 0.8044, + "step": 11979 + }, + { + "epoch": 0.29, + "learning_rate": 1.670503222001204e-05, + "loss": 1.2448, + "step": 11980 + }, + { + "epoch": 0.29, + "learning_rate": 1.6704454332767464e-05, + "loss": 1.0049, + "step": 11981 + }, + { + "epoch": 0.29, + "learning_rate": 1.670387640484854e-05, + "loss": 1.1701, + "step": 11982 + }, + { + "epoch": 0.29, + "learning_rate": 1.6703298436258774e-05, + "loss": 1.1217, + "step": 11983 + }, + { + "epoch": 0.29, + "learning_rate": 1.6702720427001675e-05, + "loss": 1.1687, + "step": 11984 + }, + { + "epoch": 0.29, + "learning_rate": 1.6702142377080747e-05, + "loss": 1.2251, + "step": 11985 + }, + { + "epoch": 0.29, + "learning_rate": 1.67015642864995e-05, + "loss": 1.0092, + "step": 11986 + }, + { + "epoch": 0.29, + "learning_rate": 1.6700986155261437e-05, + "loss": 1.3054, + "step": 11987 + }, + { + "epoch": 0.29, + "learning_rate": 1.6700407983370065e-05, + "loss": 0.8768, + "step": 11988 + }, + { + "epoch": 0.29, + "learning_rate": 1.6699829770828898e-05, + "loss": 1.1863, + "step": 11989 + }, + { + "epoch": 0.29, + "learning_rate": 1.6699251517641435e-05, + "loss": 1.2275, + "step": 11990 + }, + { + "epoch": 0.29, + "learning_rate": 1.6698673223811193e-05, + "loss": 1.0043, + "step": 11991 + }, + { + "epoch": 0.29, + "learning_rate": 1.6698094889341673e-05, + "loss": 0.8371, + "step": 11992 + }, + { + "epoch": 0.29, + "learning_rate": 1.669751651423639e-05, + "loss": 1.3391, + "step": 11993 + }, + { + "epoch": 0.29, + "learning_rate": 1.6696938098498846e-05, + "loss": 1.0171, + "step": 11994 + }, + { + "epoch": 0.29, + "learning_rate": 1.6696359642132556e-05, + "loss": 0.977, + "step": 11995 + }, + { + "epoch": 0.29, + "learning_rate": 1.6695781145141025e-05, + "loss": 0.9647, + "step": 11996 + }, + { + "epoch": 0.29, + "learning_rate": 1.669520260752777e-05, + "loss": 1.1392, + "step": 11997 + }, + { + "epoch": 0.29, + "learning_rate": 1.6694624029296288e-05, + "loss": 0.935, + "step": 11998 + }, + { + "epoch": 0.29, + "learning_rate": 1.6694045410450102e-05, + "loss": 1.2742, + "step": 11999 + }, + { + "epoch": 0.29, + "learning_rate": 1.669346675099272e-05, + "loss": 1.1311, + "step": 12000 + }, + { + "epoch": 0.29, + "learning_rate": 1.669288805092764e-05, + "loss": 1.1768, + "step": 12001 + }, + { + "epoch": 0.29, + "learning_rate": 1.669230931025839e-05, + "loss": 1.2133, + "step": 12002 + }, + { + "epoch": 0.29, + "learning_rate": 1.669173052898847e-05, + "loss": 1.0404, + "step": 12003 + }, + { + "epoch": 0.29, + "learning_rate": 1.6691151707121397e-05, + "loss": 1.2249, + "step": 12004 + }, + { + "epoch": 0.29, + "learning_rate": 1.6690572844660678e-05, + "loss": 0.9026, + "step": 12005 + }, + { + "epoch": 0.29, + "learning_rate": 1.668999394160983e-05, + "loss": 1.0089, + "step": 12006 + }, + { + "epoch": 0.29, + "learning_rate": 1.668941499797236e-05, + "loss": 1.4097, + "step": 12007 + }, + { + "epoch": 0.29, + "learning_rate": 1.6688836013751784e-05, + "loss": 1.1357, + "step": 12008 + }, + { + "epoch": 0.29, + "learning_rate": 1.6688256988951614e-05, + "loss": 0.9456, + "step": 12009 + }, + { + "epoch": 0.29, + "learning_rate": 1.668767792357536e-05, + "loss": 1.0678, + "step": 12010 + }, + { + "epoch": 0.29, + "learning_rate": 1.6687098817626538e-05, + "loss": 1.1007, + "step": 12011 + }, + { + "epoch": 0.29, + "learning_rate": 1.6686519671108662e-05, + "loss": 0.9327, + "step": 12012 + }, + { + "epoch": 0.29, + "learning_rate": 1.6685940484025243e-05, + "loss": 1.2039, + "step": 12013 + }, + { + "epoch": 0.29, + "learning_rate": 1.6685361256379795e-05, + "loss": 1.2538, + "step": 12014 + }, + { + "epoch": 0.29, + "learning_rate": 1.6684781988175834e-05, + "loss": 0.8406, + "step": 12015 + }, + { + "epoch": 0.29, + "learning_rate": 1.668420267941687e-05, + "loss": 0.9985, + "step": 12016 + }, + { + "epoch": 0.29, + "learning_rate": 1.6683623330106426e-05, + "loss": 1.0966, + "step": 12017 + }, + { + "epoch": 0.29, + "learning_rate": 1.6683043940248005e-05, + "loss": 1.0439, + "step": 12018 + }, + { + "epoch": 0.29, + "learning_rate": 1.668246450984513e-05, + "loss": 1.1605, + "step": 12019 + }, + { + "epoch": 0.29, + "learning_rate": 1.668188503890132e-05, + "loss": 1.353, + "step": 12020 + }, + { + "epoch": 0.29, + "learning_rate": 1.668130552742008e-05, + "loss": 0.9843, + "step": 12021 + }, + { + "epoch": 0.29, + "learning_rate": 1.6680725975404934e-05, + "loss": 0.8972, + "step": 12022 + }, + { + "epoch": 0.29, + "learning_rate": 1.668014638285939e-05, + "loss": 1.1989, + "step": 12023 + }, + { + "epoch": 0.29, + "learning_rate": 1.6679566749786974e-05, + "loss": 1.0887, + "step": 12024 + }, + { + "epoch": 0.29, + "learning_rate": 1.6678987076191195e-05, + "loss": 1.0052, + "step": 12025 + }, + { + "epoch": 0.29, + "learning_rate": 1.6678407362075572e-05, + "loss": 1.1774, + "step": 12026 + }, + { + "epoch": 0.29, + "learning_rate": 1.6677827607443622e-05, + "loss": 0.9434, + "step": 12027 + }, + { + "epoch": 0.29, + "learning_rate": 1.6677247812298866e-05, + "loss": 0.8983, + "step": 12028 + }, + { + "epoch": 0.29, + "learning_rate": 1.6676667976644813e-05, + "loss": 1.1405, + "step": 12029 + }, + { + "epoch": 0.29, + "learning_rate": 1.667608810048499e-05, + "loss": 1.0121, + "step": 12030 + }, + { + "epoch": 0.29, + "learning_rate": 1.667550818382291e-05, + "loss": 1.092, + "step": 12031 + }, + { + "epoch": 0.29, + "learning_rate": 1.667492822666209e-05, + "loss": 1.3047, + "step": 12032 + }, + { + "epoch": 0.29, + "learning_rate": 1.6674348229006056e-05, + "loss": 0.928, + "step": 12033 + }, + { + "epoch": 0.29, + "learning_rate": 1.6673768190858317e-05, + "loss": 0.9747, + "step": 12034 + }, + { + "epoch": 0.29, + "learning_rate": 1.6673188112222394e-05, + "loss": 1.1196, + "step": 12035 + }, + { + "epoch": 0.29, + "learning_rate": 1.6672607993101813e-05, + "loss": 0.9205, + "step": 12036 + }, + { + "epoch": 0.29, + "learning_rate": 1.6672027833500088e-05, + "loss": 1.0434, + "step": 12037 + }, + { + "epoch": 0.29, + "learning_rate": 1.6671447633420738e-05, + "loss": 1.1146, + "step": 12038 + }, + { + "epoch": 0.29, + "learning_rate": 1.6670867392867284e-05, + "loss": 1.1664, + "step": 12039 + }, + { + "epoch": 0.29, + "learning_rate": 1.6670287111843247e-05, + "loss": 1.0498, + "step": 12040 + }, + { + "epoch": 0.29, + "learning_rate": 1.6669706790352146e-05, + "loss": 1.0895, + "step": 12041 + }, + { + "epoch": 0.29, + "learning_rate": 1.6669126428397505e-05, + "loss": 1.2762, + "step": 12042 + }, + { + "epoch": 0.29, + "learning_rate": 1.666854602598284e-05, + "loss": 1.0615, + "step": 12043 + }, + { + "epoch": 0.29, + "learning_rate": 1.666796558311168e-05, + "loss": 1.2513, + "step": 12044 + }, + { + "epoch": 0.29, + "learning_rate": 1.6667385099787537e-05, + "loss": 1.3826, + "step": 12045 + }, + { + "epoch": 0.29, + "learning_rate": 1.6666804576013936e-05, + "loss": 1.2228, + "step": 12046 + }, + { + "epoch": 0.29, + "learning_rate": 1.6666224011794404e-05, + "loss": 1.1189, + "step": 12047 + }, + { + "epoch": 0.29, + "learning_rate": 1.6665643407132454e-05, + "loss": 0.8676, + "step": 12048 + }, + { + "epoch": 0.29, + "learning_rate": 1.6665062762031616e-05, + "loss": 1.1596, + "step": 12049 + }, + { + "epoch": 0.29, + "learning_rate": 1.666448207649541e-05, + "loss": 1.2869, + "step": 12050 + }, + { + "epoch": 0.29, + "learning_rate": 1.6663901350527357e-05, + "loss": 0.8888, + "step": 12051 + }, + { + "epoch": 0.29, + "learning_rate": 1.6663320584130988e-05, + "loss": 0.9654, + "step": 12052 + }, + { + "epoch": 0.29, + "learning_rate": 1.6662739777309812e-05, + "loss": 0.9236, + "step": 12053 + }, + { + "epoch": 0.29, + "learning_rate": 1.6662158930067363e-05, + "loss": 1.1439, + "step": 12054 + }, + { + "epoch": 0.29, + "learning_rate": 1.6661578042407165e-05, + "loss": 1.2648, + "step": 12055 + }, + { + "epoch": 0.29, + "learning_rate": 1.666099711433274e-05, + "loss": 1.017, + "step": 12056 + }, + { + "epoch": 0.29, + "learning_rate": 1.6660416145847613e-05, + "loss": 0.8629, + "step": 12057 + }, + { + "epoch": 0.29, + "learning_rate": 1.6659835136955305e-05, + "loss": 1.3371, + "step": 12058 + }, + { + "epoch": 0.29, + "learning_rate": 1.6659254087659347e-05, + "loss": 1.302, + "step": 12059 + }, + { + "epoch": 0.29, + "learning_rate": 1.6658672997963258e-05, + "loss": 0.9964, + "step": 12060 + }, + { + "epoch": 0.29, + "learning_rate": 1.665809186787057e-05, + "loss": 0.97, + "step": 12061 + }, + { + "epoch": 0.29, + "learning_rate": 1.66575106973848e-05, + "loss": 1.1408, + "step": 12062 + }, + { + "epoch": 0.29, + "learning_rate": 1.665692948650948e-05, + "loss": 1.0598, + "step": 12063 + }, + { + "epoch": 0.29, + "learning_rate": 1.6656348235248134e-05, + "loss": 1.0492, + "step": 12064 + }, + { + "epoch": 0.29, + "learning_rate": 1.665576694360429e-05, + "loss": 1.1447, + "step": 12065 + }, + { + "epoch": 0.29, + "learning_rate": 1.6655185611581473e-05, + "loss": 1.2901, + "step": 12066 + }, + { + "epoch": 0.29, + "learning_rate": 1.665460423918321e-05, + "loss": 1.2685, + "step": 12067 + }, + { + "epoch": 0.29, + "learning_rate": 1.665402282641303e-05, + "loss": 0.9922, + "step": 12068 + }, + { + "epoch": 0.29, + "learning_rate": 1.6653441373274458e-05, + "loss": 0.7872, + "step": 12069 + }, + { + "epoch": 0.29, + "learning_rate": 1.6652859879771024e-05, + "loss": 1.0878, + "step": 12070 + }, + { + "epoch": 0.29, + "learning_rate": 1.665227834590625e-05, + "loss": 1.1252, + "step": 12071 + }, + { + "epoch": 0.29, + "learning_rate": 1.665169677168367e-05, + "loss": 1.0521, + "step": 12072 + }, + { + "epoch": 0.29, + "learning_rate": 1.665111515710681e-05, + "loss": 0.8854, + "step": 12073 + }, + { + "epoch": 0.29, + "learning_rate": 1.66505335021792e-05, + "loss": 1.0575, + "step": 12074 + }, + { + "epoch": 0.29, + "learning_rate": 1.6649951806904366e-05, + "loss": 0.9368, + "step": 12075 + }, + { + "epoch": 0.29, + "learning_rate": 1.6649370071285837e-05, + "loss": 1.0778, + "step": 12076 + }, + { + "epoch": 0.29, + "learning_rate": 1.6648788295327144e-05, + "loss": 1.0659, + "step": 12077 + }, + { + "epoch": 0.29, + "learning_rate": 1.664820647903182e-05, + "loss": 1.1386, + "step": 12078 + }, + { + "epoch": 0.29, + "learning_rate": 1.664762462240339e-05, + "loss": 1.2864, + "step": 12079 + }, + { + "epoch": 0.29, + "learning_rate": 1.6647042725445386e-05, + "loss": 0.86, + "step": 12080 + }, + { + "epoch": 0.29, + "learning_rate": 1.6646460788161333e-05, + "loss": 1.0587, + "step": 12081 + }, + { + "epoch": 0.29, + "learning_rate": 1.6645878810554768e-05, + "loss": 1.2696, + "step": 12082 + }, + { + "epoch": 0.29, + "learning_rate": 1.664529679262922e-05, + "loss": 0.9687, + "step": 12083 + }, + { + "epoch": 0.29, + "learning_rate": 1.664471473438822e-05, + "loss": 1.1696, + "step": 12084 + }, + { + "epoch": 0.29, + "learning_rate": 1.6644132635835293e-05, + "loss": 1.0402, + "step": 12085 + }, + { + "epoch": 0.29, + "learning_rate": 1.6643550496973984e-05, + "loss": 0.935, + "step": 12086 + }, + { + "epoch": 0.29, + "learning_rate": 1.6642968317807814e-05, + "loss": 0.9696, + "step": 12087 + }, + { + "epoch": 0.29, + "learning_rate": 1.664238609834032e-05, + "loss": 1.5332, + "step": 12088 + }, + { + "epoch": 0.29, + "learning_rate": 1.6641803838575027e-05, + "loss": 1.3819, + "step": 12089 + }, + { + "epoch": 0.29, + "learning_rate": 1.6641221538515474e-05, + "loss": 0.9006, + "step": 12090 + }, + { + "epoch": 0.29, + "learning_rate": 1.6640639198165192e-05, + "loss": 1.1916, + "step": 12091 + }, + { + "epoch": 0.29, + "learning_rate": 1.6640056817527713e-05, + "loss": 1.2604, + "step": 12092 + }, + { + "epoch": 0.29, + "learning_rate": 1.6639474396606577e-05, + "loss": 1.1828, + "step": 12093 + }, + { + "epoch": 0.29, + "learning_rate": 1.6638891935405307e-05, + "loss": 1.2214, + "step": 12094 + }, + { + "epoch": 0.29, + "learning_rate": 1.663830943392744e-05, + "loss": 1.0444, + "step": 12095 + }, + { + "epoch": 0.29, + "learning_rate": 1.663772689217651e-05, + "loss": 1.1512, + "step": 12096 + }, + { + "epoch": 0.29, + "learning_rate": 1.6637144310156056e-05, + "loss": 1.154, + "step": 12097 + }, + { + "epoch": 0.29, + "learning_rate": 1.6636561687869608e-05, + "loss": 1.1837, + "step": 12098 + }, + { + "epoch": 0.29, + "learning_rate": 1.6635979025320702e-05, + "loss": 1.3253, + "step": 12099 + }, + { + "epoch": 0.29, + "learning_rate": 1.663539632251287e-05, + "loss": 1.0343, + "step": 12100 + }, + { + "epoch": 0.29, + "learning_rate": 1.663481357944965e-05, + "loss": 1.1155, + "step": 12101 + }, + { + "epoch": 0.29, + "learning_rate": 1.6634230796134578e-05, + "loss": 1.064, + "step": 12102 + }, + { + "epoch": 0.29, + "learning_rate": 1.6633647972571187e-05, + "loss": 1.0844, + "step": 12103 + }, + { + "epoch": 0.29, + "learning_rate": 1.663306510876301e-05, + "loss": 1.0075, + "step": 12104 + }, + { + "epoch": 0.29, + "learning_rate": 1.6632482204713594e-05, + "loss": 1.013, + "step": 12105 + }, + { + "epoch": 0.29, + "learning_rate": 1.6631899260426463e-05, + "loss": 1.1359, + "step": 12106 + }, + { + "epoch": 0.29, + "learning_rate": 1.6631316275905165e-05, + "loss": 1.3445, + "step": 12107 + }, + { + "epoch": 0.29, + "learning_rate": 1.6630733251153224e-05, + "loss": 1.4947, + "step": 12108 + }, + { + "epoch": 0.29, + "learning_rate": 1.663015018617419e-05, + "loss": 1.2215, + "step": 12109 + }, + { + "epoch": 0.29, + "learning_rate": 1.662956708097159e-05, + "loss": 1.2946, + "step": 12110 + }, + { + "epoch": 0.29, + "learning_rate": 1.662898393554897e-05, + "loss": 0.9177, + "step": 12111 + }, + { + "epoch": 0.29, + "learning_rate": 1.662840074990986e-05, + "loss": 0.8705, + "step": 12112 + }, + { + "epoch": 0.29, + "learning_rate": 1.6627817524057804e-05, + "loss": 1.2536, + "step": 12113 + }, + { + "epoch": 0.29, + "learning_rate": 1.6627234257996338e-05, + "loss": 1.0278, + "step": 12114 + }, + { + "epoch": 0.29, + "learning_rate": 1.6626650951729e-05, + "loss": 0.9657, + "step": 12115 + }, + { + "epoch": 0.29, + "learning_rate": 1.662606760525933e-05, + "loss": 0.9309, + "step": 12116 + }, + { + "epoch": 0.29, + "learning_rate": 1.6625484218590867e-05, + "loss": 0.8589, + "step": 12117 + }, + { + "epoch": 0.29, + "learning_rate": 1.6624900791727145e-05, + "loss": 1.0645, + "step": 12118 + }, + { + "epoch": 0.29, + "learning_rate": 1.6624317324671712e-05, + "loss": 1.1176, + "step": 12119 + }, + { + "epoch": 0.29, + "learning_rate": 1.6623733817428104e-05, + "loss": 1.3721, + "step": 12120 + }, + { + "epoch": 0.29, + "learning_rate": 1.662315026999986e-05, + "loss": 0.9829, + "step": 12121 + }, + { + "epoch": 0.29, + "learning_rate": 1.662256668239052e-05, + "loss": 1.0092, + "step": 12122 + }, + { + "epoch": 0.29, + "learning_rate": 1.662198305460363e-05, + "loss": 1.0973, + "step": 12123 + }, + { + "epoch": 0.29, + "learning_rate": 1.662139938664272e-05, + "loss": 1.1923, + "step": 12124 + }, + { + "epoch": 0.29, + "learning_rate": 1.662081567851134e-05, + "loss": 1.1642, + "step": 12125 + }, + { + "epoch": 0.29, + "learning_rate": 1.662023193021303e-05, + "loss": 1.1318, + "step": 12126 + }, + { + "epoch": 0.29, + "learning_rate": 1.661964814175133e-05, + "loss": 1.2866, + "step": 12127 + }, + { + "epoch": 0.29, + "learning_rate": 1.661906431312978e-05, + "loss": 1.269, + "step": 12128 + }, + { + "epoch": 0.29, + "learning_rate": 1.6618480444351928e-05, + "loss": 1.0733, + "step": 12129 + }, + { + "epoch": 0.29, + "learning_rate": 1.661789653542131e-05, + "loss": 1.2348, + "step": 12130 + }, + { + "epoch": 0.29, + "learning_rate": 1.661731258634147e-05, + "loss": 1.1406, + "step": 12131 + }, + { + "epoch": 0.29, + "learning_rate": 1.6616728597115947e-05, + "loss": 1.1336, + "step": 12132 + }, + { + "epoch": 0.29, + "learning_rate": 1.6616144567748294e-05, + "loss": 1.0258, + "step": 12133 + }, + { + "epoch": 0.29, + "learning_rate": 1.6615560498242043e-05, + "loss": 1.2096, + "step": 12134 + }, + { + "epoch": 0.29, + "learning_rate": 1.6614976388600746e-05, + "loss": 1.2599, + "step": 12135 + }, + { + "epoch": 0.29, + "learning_rate": 1.6614392238827942e-05, + "loss": 1.0262, + "step": 12136 + }, + { + "epoch": 0.29, + "learning_rate": 1.6613808048927176e-05, + "loss": 1.1577, + "step": 12137 + }, + { + "epoch": 0.29, + "learning_rate": 1.6613223818901992e-05, + "loss": 1.218, + "step": 12138 + }, + { + "epoch": 0.29, + "learning_rate": 1.6612639548755937e-05, + "loss": 1.0776, + "step": 12139 + }, + { + "epoch": 0.29, + "learning_rate": 1.661205523849255e-05, + "loss": 0.9704, + "step": 12140 + }, + { + "epoch": 0.29, + "learning_rate": 1.661147088811538e-05, + "loss": 1.1779, + "step": 12141 + }, + { + "epoch": 0.29, + "learning_rate": 1.6610886497627973e-05, + "loss": 1.0121, + "step": 12142 + }, + { + "epoch": 0.29, + "learning_rate": 1.6610302067033874e-05, + "loss": 1.0438, + "step": 12143 + }, + { + "epoch": 0.29, + "learning_rate": 1.660971759633662e-05, + "loss": 0.9286, + "step": 12144 + }, + { + "epoch": 0.29, + "learning_rate": 1.660913308553977e-05, + "loss": 1.0323, + "step": 12145 + }, + { + "epoch": 0.29, + "learning_rate": 1.660854853464686e-05, + "loss": 1.2024, + "step": 12146 + }, + { + "epoch": 0.29, + "learning_rate": 1.6607963943661446e-05, + "loss": 0.9983, + "step": 12147 + }, + { + "epoch": 0.29, + "learning_rate": 1.6607379312587066e-05, + "loss": 0.6557, + "step": 12148 + }, + { + "epoch": 0.29, + "learning_rate": 1.6606794641427272e-05, + "loss": 1.0962, + "step": 12149 + }, + { + "epoch": 0.29, + "learning_rate": 1.6606209930185604e-05, + "loss": 1.0279, + "step": 12150 + }, + { + "epoch": 0.29, + "learning_rate": 1.6605625178865617e-05, + "loss": 1.0569, + "step": 12151 + }, + { + "epoch": 0.29, + "learning_rate": 1.6605040387470855e-05, + "loss": 1.281, + "step": 12152 + }, + { + "epoch": 0.29, + "learning_rate": 1.6604455556004868e-05, + "loss": 1.1839, + "step": 12153 + }, + { + "epoch": 0.29, + "learning_rate": 1.66038706844712e-05, + "loss": 1.0803, + "step": 12154 + }, + { + "epoch": 0.29, + "learning_rate": 1.6603285772873407e-05, + "loss": 1.0689, + "step": 12155 + }, + { + "epoch": 0.29, + "learning_rate": 1.6602700821215027e-05, + "loss": 0.9412, + "step": 12156 + }, + { + "epoch": 0.29, + "learning_rate": 1.660211582949962e-05, + "loss": 1.1314, + "step": 12157 + }, + { + "epoch": 0.29, + "learning_rate": 1.6601530797730726e-05, + "loss": 0.8386, + "step": 12158 + }, + { + "epoch": 0.29, + "learning_rate": 1.6600945725911897e-05, + "loss": 0.9283, + "step": 12159 + }, + { + "epoch": 0.29, + "learning_rate": 1.660036061404668e-05, + "loss": 1.2635, + "step": 12160 + }, + { + "epoch": 0.29, + "learning_rate": 1.6599775462138633e-05, + "loss": 1.1346, + "step": 12161 + }, + { + "epoch": 0.29, + "learning_rate": 1.6599190270191297e-05, + "loss": 1.3623, + "step": 12162 + }, + { + "epoch": 0.29, + "learning_rate": 1.6598605038208227e-05, + "loss": 1.3461, + "step": 12163 + }, + { + "epoch": 0.29, + "learning_rate": 1.6598019766192975e-05, + "loss": 0.8964, + "step": 12164 + }, + { + "epoch": 0.29, + "learning_rate": 1.6597434454149082e-05, + "loss": 0.6478, + "step": 12165 + }, + { + "epoch": 0.29, + "learning_rate": 1.6596849102080112e-05, + "loss": 1.0052, + "step": 12166 + }, + { + "epoch": 0.29, + "learning_rate": 1.6596263709989607e-05, + "loss": 1.0153, + "step": 12167 + }, + { + "epoch": 0.29, + "learning_rate": 1.6595678277881126e-05, + "loss": 1.1755, + "step": 12168 + }, + { + "epoch": 0.29, + "learning_rate": 1.6595092805758212e-05, + "loss": 1.0772, + "step": 12169 + }, + { + "epoch": 0.29, + "learning_rate": 1.6594507293624426e-05, + "loss": 1.1768, + "step": 12170 + }, + { + "epoch": 0.29, + "learning_rate": 1.659392174148331e-05, + "loss": 1.3063, + "step": 12171 + }, + { + "epoch": 0.29, + "learning_rate": 1.6593336149338425e-05, + "loss": 1.1445, + "step": 12172 + }, + { + "epoch": 0.29, + "learning_rate": 1.6592750517193318e-05, + "loss": 1.0173, + "step": 12173 + }, + { + "epoch": 0.29, + "learning_rate": 1.6592164845051546e-05, + "loss": 1.3871, + "step": 12174 + }, + { + "epoch": 0.29, + "learning_rate": 1.659157913291666e-05, + "loss": 1.216, + "step": 12175 + }, + { + "epoch": 0.29, + "learning_rate": 1.6590993380792215e-05, + "loss": 0.9973, + "step": 12176 + }, + { + "epoch": 0.29, + "learning_rate": 1.659040758868176e-05, + "loss": 1.2558, + "step": 12177 + }, + { + "epoch": 0.29, + "learning_rate": 1.658982175658886e-05, + "loss": 1.0386, + "step": 12178 + }, + { + "epoch": 0.29, + "learning_rate": 1.6589235884517055e-05, + "loss": 1.1264, + "step": 12179 + }, + { + "epoch": 0.29, + "learning_rate": 1.6588649972469907e-05, + "loss": 1.331, + "step": 12180 + }, + { + "epoch": 0.29, + "learning_rate": 1.6588064020450966e-05, + "loss": 0.9361, + "step": 12181 + }, + { + "epoch": 0.29, + "learning_rate": 1.6587478028463797e-05, + "loss": 1.308, + "step": 12182 + }, + { + "epoch": 0.29, + "learning_rate": 1.6586891996511942e-05, + "loss": 1.3591, + "step": 12183 + }, + { + "epoch": 0.29, + "learning_rate": 1.6586305924598967e-05, + "loss": 1.0828, + "step": 12184 + }, + { + "epoch": 0.29, + "learning_rate": 1.658571981272842e-05, + "loss": 1.0513, + "step": 12185 + }, + { + "epoch": 0.29, + "learning_rate": 1.6585133660903862e-05, + "loss": 1.0874, + "step": 12186 + }, + { + "epoch": 0.29, + "learning_rate": 1.6584547469128848e-05, + "loss": 1.0593, + "step": 12187 + }, + { + "epoch": 0.29, + "learning_rate": 1.6583961237406932e-05, + "loss": 1.0924, + "step": 12188 + }, + { + "epoch": 0.29, + "learning_rate": 1.658337496574167e-05, + "loss": 1.1132, + "step": 12189 + }, + { + "epoch": 0.29, + "learning_rate": 1.6582788654136623e-05, + "loss": 1.0616, + "step": 12190 + }, + { + "epoch": 0.29, + "learning_rate": 1.6582202302595345e-05, + "loss": 1.0383, + "step": 12191 + }, + { + "epoch": 0.29, + "learning_rate": 1.6581615911121393e-05, + "loss": 0.9923, + "step": 12192 + }, + { + "epoch": 0.29, + "learning_rate": 1.6581029479718328e-05, + "loss": 1.3494, + "step": 12193 + }, + { + "epoch": 0.29, + "learning_rate": 1.65804430083897e-05, + "loss": 1.2045, + "step": 12194 + }, + { + "epoch": 0.29, + "learning_rate": 1.6579856497139076e-05, + "loss": 1.204, + "step": 12195 + }, + { + "epoch": 0.29, + "learning_rate": 1.657926994597001e-05, + "loss": 1.2973, + "step": 12196 + }, + { + "epoch": 0.29, + "learning_rate": 1.657868335488606e-05, + "loss": 1.2076, + "step": 12197 + }, + { + "epoch": 0.29, + "learning_rate": 1.6578096723890788e-05, + "loss": 1.0253, + "step": 12198 + }, + { + "epoch": 0.29, + "learning_rate": 1.6577510052987748e-05, + "loss": 1.3921, + "step": 12199 + }, + { + "epoch": 0.29, + "learning_rate": 1.65769233421805e-05, + "loss": 1.3174, + "step": 12200 + }, + { + "epoch": 0.29, + "learning_rate": 1.6576336591472605e-05, + "loss": 0.9068, + "step": 12201 + }, + { + "epoch": 0.29, + "learning_rate": 1.6575749800867627e-05, + "loss": 0.656, + "step": 12202 + }, + { + "epoch": 0.29, + "learning_rate": 1.657516297036912e-05, + "loss": 1.0258, + "step": 12203 + }, + { + "epoch": 0.29, + "learning_rate": 1.6574576099980645e-05, + "loss": 1.3229, + "step": 12204 + }, + { + "epoch": 0.29, + "learning_rate": 1.6573989189705765e-05, + "loss": 1.3029, + "step": 12205 + }, + { + "epoch": 0.29, + "learning_rate": 1.6573402239548034e-05, + "loss": 1.2855, + "step": 12206 + }, + { + "epoch": 0.29, + "learning_rate": 1.657281524951102e-05, + "loss": 0.865, + "step": 12207 + }, + { + "epoch": 0.29, + "learning_rate": 1.6572228219598283e-05, + "loss": 1.0586, + "step": 12208 + }, + { + "epoch": 0.29, + "learning_rate": 1.6571641149813384e-05, + "loss": 1.0779, + "step": 12209 + }, + { + "epoch": 0.29, + "learning_rate": 1.6571054040159884e-05, + "loss": 1.0587, + "step": 12210 + }, + { + "epoch": 0.29, + "learning_rate": 1.657046689064134e-05, + "loss": 1.2651, + "step": 12211 + }, + { + "epoch": 0.29, + "learning_rate": 1.6569879701261323e-05, + "loss": 1.007, + "step": 12212 + }, + { + "epoch": 0.29, + "learning_rate": 1.6569292472023393e-05, + "loss": 1.0199, + "step": 12213 + }, + { + "epoch": 0.29, + "learning_rate": 1.656870520293111e-05, + "loss": 1.0111, + "step": 12214 + }, + { + "epoch": 0.29, + "learning_rate": 1.6568117893988037e-05, + "loss": 1.1646, + "step": 12215 + }, + { + "epoch": 0.29, + "learning_rate": 1.6567530545197733e-05, + "loss": 1.2226, + "step": 12216 + }, + { + "epoch": 0.29, + "learning_rate": 1.656694315656377e-05, + "loss": 1.0385, + "step": 12217 + }, + { + "epoch": 0.29, + "learning_rate": 1.656635572808971e-05, + "loss": 1.0783, + "step": 12218 + }, + { + "epoch": 0.29, + "learning_rate": 1.656576825977911e-05, + "loss": 1.4213, + "step": 12219 + }, + { + "epoch": 0.29, + "learning_rate": 1.656518075163554e-05, + "loss": 1.3958, + "step": 12220 + }, + { + "epoch": 0.29, + "learning_rate": 1.656459320366256e-05, + "loss": 1.0198, + "step": 12221 + }, + { + "epoch": 0.29, + "learning_rate": 1.6564005615863744e-05, + "loss": 1.3293, + "step": 12222 + }, + { + "epoch": 0.29, + "learning_rate": 1.6563417988242645e-05, + "loss": 0.9946, + "step": 12223 + }, + { + "epoch": 0.29, + "learning_rate": 1.656283032080283e-05, + "loss": 1.1486, + "step": 12224 + }, + { + "epoch": 0.29, + "learning_rate": 1.6562242613547872e-05, + "loss": 1.2174, + "step": 12225 + }, + { + "epoch": 0.29, + "learning_rate": 1.656165486648133e-05, + "loss": 1.2265, + "step": 12226 + }, + { + "epoch": 0.29, + "learning_rate": 1.656106707960677e-05, + "loss": 0.9005, + "step": 12227 + }, + { + "epoch": 0.29, + "learning_rate": 1.6560479252927763e-05, + "loss": 1.0767, + "step": 12228 + }, + { + "epoch": 0.29, + "learning_rate": 1.655989138644787e-05, + "loss": 1.2291, + "step": 12229 + }, + { + "epoch": 0.29, + "learning_rate": 1.655930348017066e-05, + "loss": 1.3097, + "step": 12230 + }, + { + "epoch": 0.29, + "learning_rate": 1.6558715534099693e-05, + "loss": 1.0497, + "step": 12231 + }, + { + "epoch": 0.29, + "learning_rate": 1.655812754823855e-05, + "loss": 1.0818, + "step": 12232 + }, + { + "epoch": 0.29, + "learning_rate": 1.6557539522590782e-05, + "loss": 1.2005, + "step": 12233 + }, + { + "epoch": 0.29, + "learning_rate": 1.655695145715997e-05, + "loss": 1.0362, + "step": 12234 + }, + { + "epoch": 0.29, + "learning_rate": 1.6556363351949672e-05, + "loss": 1.0091, + "step": 12235 + }, + { + "epoch": 0.29, + "learning_rate": 1.655577520696346e-05, + "loss": 1.1627, + "step": 12236 + }, + { + "epoch": 0.29, + "learning_rate": 1.6555187022204902e-05, + "loss": 1.0767, + "step": 12237 + }, + { + "epoch": 0.29, + "learning_rate": 1.6554598797677568e-05, + "loss": 0.982, + "step": 12238 + }, + { + "epoch": 0.29, + "learning_rate": 1.6554010533385024e-05, + "loss": 1.072, + "step": 12239 + }, + { + "epoch": 0.29, + "learning_rate": 1.6553422229330837e-05, + "loss": 1.0049, + "step": 12240 + }, + { + "epoch": 0.29, + "learning_rate": 1.6552833885518583e-05, + "loss": 1.1151, + "step": 12241 + }, + { + "epoch": 0.29, + "learning_rate": 1.6552245501951823e-05, + "loss": 1.3339, + "step": 12242 + }, + { + "epoch": 0.29, + "learning_rate": 1.6551657078634136e-05, + "loss": 1.0988, + "step": 12243 + }, + { + "epoch": 0.29, + "learning_rate": 1.6551068615569082e-05, + "loss": 1.2243, + "step": 12244 + }, + { + "epoch": 0.29, + "learning_rate": 1.6550480112760237e-05, + "loss": 1.0967, + "step": 12245 + }, + { + "epoch": 0.29, + "learning_rate": 1.6549891570211173e-05, + "loss": 1.2091, + "step": 12246 + }, + { + "epoch": 0.29, + "learning_rate": 1.6549302987925453e-05, + "loss": 1.2711, + "step": 12247 + }, + { + "epoch": 0.29, + "learning_rate": 1.6548714365906654e-05, + "loss": 1.0978, + "step": 12248 + }, + { + "epoch": 0.29, + "learning_rate": 1.6548125704158344e-05, + "loss": 0.9585, + "step": 12249 + }, + { + "epoch": 0.29, + "learning_rate": 1.6547537002684097e-05, + "loss": 1.4268, + "step": 12250 + }, + { + "epoch": 0.29, + "learning_rate": 1.6546948261487484e-05, + "loss": 0.8224, + "step": 12251 + }, + { + "epoch": 0.29, + "learning_rate": 1.6546359480572073e-05, + "loss": 1.2122, + "step": 12252 + }, + { + "epoch": 0.29, + "learning_rate": 1.654577065994144e-05, + "loss": 1.2621, + "step": 12253 + }, + { + "epoch": 0.29, + "learning_rate": 1.6545181799599153e-05, + "loss": 1.0775, + "step": 12254 + }, + { + "epoch": 0.29, + "learning_rate": 1.6544592899548796e-05, + "loss": 1.0627, + "step": 12255 + }, + { + "epoch": 0.29, + "learning_rate": 1.6544003959793925e-05, + "loss": 1.1011, + "step": 12256 + }, + { + "epoch": 0.29, + "learning_rate": 1.6543414980338125e-05, + "loss": 0.8844, + "step": 12257 + }, + { + "epoch": 0.29, + "learning_rate": 1.6542825961184967e-05, + "loss": 1.0633, + "step": 12258 + }, + { + "epoch": 0.29, + "learning_rate": 1.654223690233802e-05, + "loss": 1.1389, + "step": 12259 + }, + { + "epoch": 0.29, + "learning_rate": 1.6541647803800863e-05, + "loss": 1.2821, + "step": 12260 + }, + { + "epoch": 0.29, + "learning_rate": 1.6541058665577065e-05, + "loss": 1.2844, + "step": 12261 + }, + { + "epoch": 0.29, + "learning_rate": 1.6540469487670204e-05, + "loss": 1.1955, + "step": 12262 + }, + { + "epoch": 0.29, + "learning_rate": 1.653988027008385e-05, + "loss": 1.4228, + "step": 12263 + }, + { + "epoch": 0.29, + "learning_rate": 1.6539291012821583e-05, + "loss": 1.0841, + "step": 12264 + }, + { + "epoch": 0.29, + "learning_rate": 1.6538701715886976e-05, + "loss": 0.7556, + "step": 12265 + }, + { + "epoch": 0.29, + "learning_rate": 1.6538112379283603e-05, + "loss": 1.4373, + "step": 12266 + }, + { + "epoch": 0.3, + "learning_rate": 1.6537523003015043e-05, + "loss": 1.0668, + "step": 12267 + }, + { + "epoch": 0.3, + "learning_rate": 1.6536933587084866e-05, + "loss": 0.928, + "step": 12268 + }, + { + "epoch": 0.3, + "learning_rate": 1.6536344131496647e-05, + "loss": 1.3167, + "step": 12269 + }, + { + "epoch": 0.3, + "learning_rate": 1.653575463625397e-05, + "loss": 1.1667, + "step": 12270 + }, + { + "epoch": 0.3, + "learning_rate": 1.653516510136041e-05, + "loss": 1.1622, + "step": 12271 + }, + { + "epoch": 0.3, + "learning_rate": 1.6534575526819535e-05, + "loss": 1.0995, + "step": 12272 + }, + { + "epoch": 0.3, + "learning_rate": 1.6533985912634928e-05, + "loss": 1.067, + "step": 12273 + }, + { + "epoch": 0.3, + "learning_rate": 1.6533396258810167e-05, + "loss": 1.1106, + "step": 12274 + }, + { + "epoch": 0.3, + "learning_rate": 1.6532806565348825e-05, + "loss": 1.399, + "step": 12275 + }, + { + "epoch": 0.3, + "learning_rate": 1.6532216832254487e-05, + "loss": 1.1874, + "step": 12276 + }, + { + "epoch": 0.3, + "learning_rate": 1.6531627059530722e-05, + "loss": 1.1586, + "step": 12277 + }, + { + "epoch": 0.3, + "learning_rate": 1.6531037247181113e-05, + "loss": 1.244, + "step": 12278 + }, + { + "epoch": 0.3, + "learning_rate": 1.6530447395209238e-05, + "loss": 1.2637, + "step": 12279 + }, + { + "epoch": 0.3, + "learning_rate": 1.652985750361867e-05, + "loss": 1.2232, + "step": 12280 + }, + { + "epoch": 0.3, + "learning_rate": 1.6529267572412998e-05, + "loss": 1.323, + "step": 12281 + }, + { + "epoch": 0.3, + "learning_rate": 1.6528677601595796e-05, + "loss": 1.0793, + "step": 12282 + }, + { + "epoch": 0.3, + "learning_rate": 1.652808759117064e-05, + "loss": 0.9607, + "step": 12283 + }, + { + "epoch": 0.3, + "learning_rate": 1.6527497541141112e-05, + "loss": 1.2175, + "step": 12284 + }, + { + "epoch": 0.3, + "learning_rate": 1.6526907451510792e-05, + "loss": 1.3158, + "step": 12285 + }, + { + "epoch": 0.3, + "learning_rate": 1.652631732228326e-05, + "loss": 0.996, + "step": 12286 + }, + { + "epoch": 0.3, + "learning_rate": 1.6525727153462096e-05, + "loss": 1.0851, + "step": 12287 + }, + { + "epoch": 0.3, + "learning_rate": 1.652513694505088e-05, + "loss": 1.0039, + "step": 12288 + }, + { + "epoch": 0.3, + "learning_rate": 1.6524546697053194e-05, + "loss": 1.0735, + "step": 12289 + }, + { + "epoch": 0.3, + "learning_rate": 1.6523956409472616e-05, + "loss": 1.2078, + "step": 12290 + }, + { + "epoch": 0.3, + "learning_rate": 1.652336608231273e-05, + "loss": 1.2203, + "step": 12291 + }, + { + "epoch": 0.3, + "learning_rate": 1.6522775715577115e-05, + "loss": 0.8818, + "step": 12292 + }, + { + "epoch": 0.3, + "learning_rate": 1.6522185309269356e-05, + "loss": 1.1651, + "step": 12293 + }, + { + "epoch": 0.3, + "learning_rate": 1.6521594863393028e-05, + "loss": 1.2694, + "step": 12294 + }, + { + "epoch": 0.3, + "learning_rate": 1.6521004377951726e-05, + "loss": 1.0024, + "step": 12295 + }, + { + "epoch": 0.3, + "learning_rate": 1.652041385294902e-05, + "loss": 1.0377, + "step": 12296 + }, + { + "epoch": 0.3, + "learning_rate": 1.6519823288388498e-05, + "loss": 0.9489, + "step": 12297 + }, + { + "epoch": 0.3, + "learning_rate": 1.6519232684273737e-05, + "loss": 1.0064, + "step": 12298 + }, + { + "epoch": 0.3, + "learning_rate": 1.651864204060833e-05, + "loss": 1.0867, + "step": 12299 + }, + { + "epoch": 0.3, + "learning_rate": 1.6518051357395855e-05, + "loss": 1.1008, + "step": 12300 + }, + { + "epoch": 0.3, + "learning_rate": 1.6517460634639896e-05, + "loss": 1.0556, + "step": 12301 + }, + { + "epoch": 0.3, + "learning_rate": 1.6516869872344033e-05, + "loss": 0.9119, + "step": 12302 + }, + { + "epoch": 0.3, + "learning_rate": 1.6516279070511853e-05, + "loss": 1.3553, + "step": 12303 + }, + { + "epoch": 0.3, + "learning_rate": 1.6515688229146945e-05, + "loss": 1.0889, + "step": 12304 + }, + { + "epoch": 0.3, + "learning_rate": 1.6515097348252886e-05, + "loss": 1.3205, + "step": 12305 + }, + { + "epoch": 0.3, + "learning_rate": 1.6514506427833264e-05, + "loss": 1.1706, + "step": 12306 + }, + { + "epoch": 0.3, + "learning_rate": 1.6513915467891665e-05, + "loss": 1.1043, + "step": 12307 + }, + { + "epoch": 0.3, + "learning_rate": 1.6513324468431672e-05, + "loss": 1.1689, + "step": 12308 + }, + { + "epoch": 0.3, + "learning_rate": 1.651273342945687e-05, + "loss": 0.9907, + "step": 12309 + }, + { + "epoch": 0.3, + "learning_rate": 1.6512142350970852e-05, + "loss": 0.7971, + "step": 12310 + }, + { + "epoch": 0.3, + "learning_rate": 1.651155123297719e-05, + "loss": 0.998, + "step": 12311 + }, + { + "epoch": 0.3, + "learning_rate": 1.6510960075479484e-05, + "loss": 1.0901, + "step": 12312 + }, + { + "epoch": 0.3, + "learning_rate": 1.651036887848131e-05, + "loss": 1.1129, + "step": 12313 + }, + { + "epoch": 0.3, + "learning_rate": 1.6509777641986264e-05, + "loss": 1.193, + "step": 12314 + }, + { + "epoch": 0.3, + "learning_rate": 1.6509186365997928e-05, + "loss": 1.3263, + "step": 12315 + }, + { + "epoch": 0.3, + "learning_rate": 1.6508595050519887e-05, + "loss": 1.1928, + "step": 12316 + }, + { + "epoch": 0.3, + "learning_rate": 1.650800369555573e-05, + "loss": 0.9008, + "step": 12317 + }, + { + "epoch": 0.3, + "learning_rate": 1.6507412301109046e-05, + "loss": 1.0948, + "step": 12318 + }, + { + "epoch": 0.3, + "learning_rate": 1.6506820867183425e-05, + "loss": 1.1308, + "step": 12319 + }, + { + "epoch": 0.3, + "learning_rate": 1.650622939378245e-05, + "loss": 1.1279, + "step": 12320 + }, + { + "epoch": 0.3, + "learning_rate": 1.6505637880909713e-05, + "loss": 1.0865, + "step": 12321 + }, + { + "epoch": 0.3, + "learning_rate": 1.65050463285688e-05, + "loss": 1.0824, + "step": 12322 + }, + { + "epoch": 0.3, + "learning_rate": 1.6504454736763303e-05, + "loss": 1.4667, + "step": 12323 + }, + { + "epoch": 0.3, + "learning_rate": 1.6503863105496803e-05, + "loss": 0.8866, + "step": 12324 + }, + { + "epoch": 0.3, + "learning_rate": 1.6503271434772904e-05, + "loss": 1.3468, + "step": 12325 + }, + { + "epoch": 0.3, + "learning_rate": 1.6502679724595185e-05, + "loss": 0.9903, + "step": 12326 + }, + { + "epoch": 0.3, + "learning_rate": 1.6502087974967232e-05, + "loss": 1.2851, + "step": 12327 + }, + { + "epoch": 0.3, + "learning_rate": 1.6501496185892643e-05, + "loss": 1.0062, + "step": 12328 + }, + { + "epoch": 0.3, + "learning_rate": 1.6500904357375006e-05, + "loss": 1.2098, + "step": 12329 + }, + { + "epoch": 0.3, + "learning_rate": 1.6500312489417912e-05, + "loss": 1.0669, + "step": 12330 + }, + { + "epoch": 0.3, + "learning_rate": 1.6499720582024953e-05, + "loss": 1.2827, + "step": 12331 + }, + { + "epoch": 0.3, + "learning_rate": 1.6499128635199717e-05, + "loss": 0.7728, + "step": 12332 + }, + { + "epoch": 0.3, + "learning_rate": 1.6498536648945795e-05, + "loss": 1.1617, + "step": 12333 + }, + { + "epoch": 0.3, + "learning_rate": 1.649794462326678e-05, + "loss": 1.1284, + "step": 12334 + }, + { + "epoch": 0.3, + "learning_rate": 1.6497352558166263e-05, + "loss": 1.0841, + "step": 12335 + }, + { + "epoch": 0.3, + "learning_rate": 1.6496760453647836e-05, + "loss": 1.0944, + "step": 12336 + }, + { + "epoch": 0.3, + "learning_rate": 1.6496168309715096e-05, + "loss": 1.1227, + "step": 12337 + }, + { + "epoch": 0.3, + "learning_rate": 1.6495576126371626e-05, + "loss": 0.8917, + "step": 12338 + }, + { + "epoch": 0.3, + "learning_rate": 1.6494983903621024e-05, + "loss": 1.2925, + "step": 12339 + }, + { + "epoch": 0.3, + "learning_rate": 1.649439164146688e-05, + "loss": 1.233, + "step": 12340 + }, + { + "epoch": 0.3, + "learning_rate": 1.6493799339912796e-05, + "loss": 1.0139, + "step": 12341 + }, + { + "epoch": 0.3, + "learning_rate": 1.6493206998962353e-05, + "loss": 1.3009, + "step": 12342 + }, + { + "epoch": 0.3, + "learning_rate": 1.6492614618619153e-05, + "loss": 1.2559, + "step": 12343 + }, + { + "epoch": 0.3, + "learning_rate": 1.649202219888679e-05, + "loss": 1.0649, + "step": 12344 + }, + { + "epoch": 0.3, + "learning_rate": 1.649142973976885e-05, + "loss": 1.0183, + "step": 12345 + }, + { + "epoch": 0.3, + "learning_rate": 1.649083724126894e-05, + "loss": 1.2627, + "step": 12346 + }, + { + "epoch": 0.3, + "learning_rate": 1.6490244703390637e-05, + "loss": 1.1455, + "step": 12347 + }, + { + "epoch": 0.3, + "learning_rate": 1.6489652126137552e-05, + "loss": 1.1958, + "step": 12348 + }, + { + "epoch": 0.3, + "learning_rate": 1.6489059509513275e-05, + "loss": 0.9397, + "step": 12349 + }, + { + "epoch": 0.3, + "learning_rate": 1.64884668535214e-05, + "loss": 1.1141, + "step": 12350 + }, + { + "epoch": 0.3, + "learning_rate": 1.648787415816552e-05, + "loss": 1.2879, + "step": 12351 + }, + { + "epoch": 0.3, + "learning_rate": 1.648728142344923e-05, + "loss": 1.4174, + "step": 12352 + }, + { + "epoch": 0.3, + "learning_rate": 1.648668864937614e-05, + "loss": 1.2248, + "step": 12353 + }, + { + "epoch": 0.3, + "learning_rate": 1.648609583594983e-05, + "loss": 1.1227, + "step": 12354 + }, + { + "epoch": 0.3, + "learning_rate": 1.64855029831739e-05, + "loss": 0.8228, + "step": 12355 + }, + { + "epoch": 0.3, + "learning_rate": 1.6484910091051952e-05, + "loss": 1.1327, + "step": 12356 + }, + { + "epoch": 0.3, + "learning_rate": 1.6484317159587577e-05, + "loss": 1.0841, + "step": 12357 + }, + { + "epoch": 0.3, + "learning_rate": 1.648372418878438e-05, + "loss": 1.3465, + "step": 12358 + }, + { + "epoch": 0.3, + "learning_rate": 1.648313117864595e-05, + "loss": 1.1278, + "step": 12359 + }, + { + "epoch": 0.3, + "learning_rate": 1.648253812917589e-05, + "loss": 1.3471, + "step": 12360 + }, + { + "epoch": 0.3, + "learning_rate": 1.6481945040377795e-05, + "loss": 1.2787, + "step": 12361 + }, + { + "epoch": 0.3, + "learning_rate": 1.648135191225526e-05, + "loss": 1.312, + "step": 12362 + }, + { + "epoch": 0.3, + "learning_rate": 1.648075874481189e-05, + "loss": 1.1432, + "step": 12363 + }, + { + "epoch": 0.3, + "learning_rate": 1.6480165538051282e-05, + "loss": 1.2596, + "step": 12364 + }, + { + "epoch": 0.3, + "learning_rate": 1.6479572291977035e-05, + "loss": 1.5321, + "step": 12365 + }, + { + "epoch": 0.3, + "learning_rate": 1.6478979006592747e-05, + "loss": 1.132, + "step": 12366 + }, + { + "epoch": 0.3, + "learning_rate": 1.6478385681902014e-05, + "loss": 1.1149, + "step": 12367 + }, + { + "epoch": 0.3, + "learning_rate": 1.6477792317908445e-05, + "loss": 1.3701, + "step": 12368 + }, + { + "epoch": 0.3, + "learning_rate": 1.647719891461563e-05, + "loss": 1.2926, + "step": 12369 + }, + { + "epoch": 0.3, + "learning_rate": 1.6476605472027173e-05, + "loss": 1.2245, + "step": 12370 + }, + { + "epoch": 0.3, + "learning_rate": 1.647601199014667e-05, + "loss": 0.8637, + "step": 12371 + }, + { + "epoch": 0.3, + "learning_rate": 1.647541846897773e-05, + "loss": 1.0636, + "step": 12372 + }, + { + "epoch": 0.3, + "learning_rate": 1.6474824908523947e-05, + "loss": 1.0462, + "step": 12373 + }, + { + "epoch": 0.3, + "learning_rate": 1.6474231308788928e-05, + "loss": 1.0522, + "step": 12374 + }, + { + "epoch": 0.3, + "learning_rate": 1.6473637669776265e-05, + "loss": 1.1705, + "step": 12375 + }, + { + "epoch": 0.3, + "learning_rate": 1.647304399148957e-05, + "loss": 1.3378, + "step": 12376 + }, + { + "epoch": 0.3, + "learning_rate": 1.6472450273932436e-05, + "loss": 1.0376, + "step": 12377 + }, + { + "epoch": 0.3, + "learning_rate": 1.647185651710847e-05, + "loss": 1.1862, + "step": 12378 + }, + { + "epoch": 0.3, + "learning_rate": 1.6471262721021274e-05, + "loss": 1.0695, + "step": 12379 + }, + { + "epoch": 0.3, + "learning_rate": 1.647066888567445e-05, + "loss": 1.1808, + "step": 12380 + }, + { + "epoch": 0.3, + "learning_rate": 1.6470075011071593e-05, + "loss": 1.2994, + "step": 12381 + }, + { + "epoch": 0.3, + "learning_rate": 1.646948109721632e-05, + "loss": 0.9148, + "step": 12382 + }, + { + "epoch": 0.3, + "learning_rate": 1.6468887144112226e-05, + "loss": 1.0443, + "step": 12383 + }, + { + "epoch": 0.3, + "learning_rate": 1.6468293151762914e-05, + "loss": 1.0865, + "step": 12384 + }, + { + "epoch": 0.3, + "learning_rate": 1.646769912017199e-05, + "loss": 0.8539, + "step": 12385 + }, + { + "epoch": 0.3, + "learning_rate": 1.6467105049343055e-05, + "loss": 1.0529, + "step": 12386 + }, + { + "epoch": 0.3, + "learning_rate": 1.6466510939279714e-05, + "loss": 1.0653, + "step": 12387 + }, + { + "epoch": 0.3, + "learning_rate": 1.6465916789985574e-05, + "loss": 1.081, + "step": 12388 + }, + { + "epoch": 0.3, + "learning_rate": 1.6465322601464238e-05, + "loss": 1.159, + "step": 12389 + }, + { + "epoch": 0.3, + "learning_rate": 1.6464728373719307e-05, + "loss": 1.0722, + "step": 12390 + }, + { + "epoch": 0.3, + "learning_rate": 1.6464134106754395e-05, + "loss": 1.1593, + "step": 12391 + }, + { + "epoch": 0.3, + "learning_rate": 1.6463539800573098e-05, + "loss": 1.1386, + "step": 12392 + }, + { + "epoch": 0.3, + "learning_rate": 1.6462945455179027e-05, + "loss": 0.9732, + "step": 12393 + }, + { + "epoch": 0.3, + "learning_rate": 1.6462351070575787e-05, + "loss": 1.2812, + "step": 12394 + }, + { + "epoch": 0.3, + "learning_rate": 1.646175664676698e-05, + "loss": 1.2029, + "step": 12395 + }, + { + "epoch": 0.3, + "learning_rate": 1.646116218375622e-05, + "loss": 1.0501, + "step": 12396 + }, + { + "epoch": 0.3, + "learning_rate": 1.6460567681547103e-05, + "loss": 1.3312, + "step": 12397 + }, + { + "epoch": 0.3, + "learning_rate": 1.6459973140143245e-05, + "loss": 1.0016, + "step": 12398 + }, + { + "epoch": 0.3, + "learning_rate": 1.645937855954825e-05, + "loss": 1.019, + "step": 12399 + }, + { + "epoch": 0.3, + "learning_rate": 1.645878393976572e-05, + "loss": 0.7737, + "step": 12400 + }, + { + "epoch": 0.3, + "learning_rate": 1.645818928079927e-05, + "loss": 1.1375, + "step": 12401 + }, + { + "epoch": 0.3, + "learning_rate": 1.6457594582652507e-05, + "loss": 1.0452, + "step": 12402 + }, + { + "epoch": 0.3, + "learning_rate": 1.6456999845329035e-05, + "loss": 0.9347, + "step": 12403 + }, + { + "epoch": 0.3, + "learning_rate": 1.6456405068832463e-05, + "loss": 1.0439, + "step": 12404 + }, + { + "epoch": 0.3, + "learning_rate": 1.64558102531664e-05, + "loss": 1.1704, + "step": 12405 + }, + { + "epoch": 0.3, + "learning_rate": 1.6455215398334457e-05, + "loss": 1.1693, + "step": 12406 + }, + { + "epoch": 0.3, + "learning_rate": 1.6454620504340237e-05, + "loss": 0.8874, + "step": 12407 + }, + { + "epoch": 0.3, + "learning_rate": 1.6454025571187354e-05, + "loss": 1.1303, + "step": 12408 + }, + { + "epoch": 0.3, + "learning_rate": 1.645343059887942e-05, + "loss": 1.0067, + "step": 12409 + }, + { + "epoch": 0.3, + "learning_rate": 1.6452835587420034e-05, + "loss": 1.2275, + "step": 12410 + }, + { + "epoch": 0.3, + "learning_rate": 1.6452240536812817e-05, + "loss": 1.1699, + "step": 12411 + }, + { + "epoch": 0.3, + "learning_rate": 1.645164544706137e-05, + "loss": 0.9732, + "step": 12412 + }, + { + "epoch": 0.3, + "learning_rate": 1.645105031816931e-05, + "loss": 1.0569, + "step": 12413 + }, + { + "epoch": 0.3, + "learning_rate": 1.6450455150140247e-05, + "loss": 1.0323, + "step": 12414 + }, + { + "epoch": 0.3, + "learning_rate": 1.6449859942977786e-05, + "loss": 1.1386, + "step": 12415 + }, + { + "epoch": 0.3, + "learning_rate": 1.6449264696685544e-05, + "loss": 0.91, + "step": 12416 + }, + { + "epoch": 0.3, + "learning_rate": 1.644866941126713e-05, + "loss": 0.8899, + "step": 12417 + }, + { + "epoch": 0.3, + "learning_rate": 1.6448074086726157e-05, + "loss": 1.186, + "step": 12418 + }, + { + "epoch": 0.3, + "learning_rate": 1.644747872306623e-05, + "loss": 0.9181, + "step": 12419 + }, + { + "epoch": 0.3, + "learning_rate": 1.644688332029097e-05, + "loss": 1.0407, + "step": 12420 + }, + { + "epoch": 0.3, + "learning_rate": 1.6446287878403988e-05, + "loss": 1.2569, + "step": 12421 + }, + { + "epoch": 0.3, + "learning_rate": 1.644569239740889e-05, + "loss": 1.1472, + "step": 12422 + }, + { + "epoch": 0.3, + "learning_rate": 1.644509687730929e-05, + "loss": 1.0995, + "step": 12423 + }, + { + "epoch": 0.3, + "learning_rate": 1.644450131810881e-05, + "loss": 1.1483, + "step": 12424 + }, + { + "epoch": 0.3, + "learning_rate": 1.6443905719811052e-05, + "loss": 1.3952, + "step": 12425 + }, + { + "epoch": 0.3, + "learning_rate": 1.6443310082419636e-05, + "loss": 1.1526, + "step": 12426 + }, + { + "epoch": 0.3, + "learning_rate": 1.644271440593817e-05, + "loss": 0.9412, + "step": 12427 + }, + { + "epoch": 0.3, + "learning_rate": 1.6442118690370272e-05, + "loss": 1.2952, + "step": 12428 + }, + { + "epoch": 0.3, + "learning_rate": 1.6441522935719555e-05, + "loss": 0.9864, + "step": 12429 + }, + { + "epoch": 0.3, + "learning_rate": 1.6440927141989635e-05, + "loss": 1.0254, + "step": 12430 + }, + { + "epoch": 0.3, + "learning_rate": 1.6440331309184126e-05, + "loss": 1.1045, + "step": 12431 + }, + { + "epoch": 0.3, + "learning_rate": 1.6439735437306638e-05, + "loss": 1.2, + "step": 12432 + }, + { + "epoch": 0.3, + "learning_rate": 1.643913952636079e-05, + "loss": 1.0283, + "step": 12433 + }, + { + "epoch": 0.3, + "learning_rate": 1.6438543576350202e-05, + "loss": 1.1846, + "step": 12434 + }, + { + "epoch": 0.3, + "learning_rate": 1.643794758727848e-05, + "loss": 0.8838, + "step": 12435 + }, + { + "epoch": 0.3, + "learning_rate": 1.6437351559149246e-05, + "loss": 1.2103, + "step": 12436 + }, + { + "epoch": 0.3, + "learning_rate": 1.6436755491966116e-05, + "loss": 1.1672, + "step": 12437 + }, + { + "epoch": 0.3, + "learning_rate": 1.64361593857327e-05, + "loss": 0.9243, + "step": 12438 + }, + { + "epoch": 0.3, + "learning_rate": 1.6435563240452624e-05, + "loss": 1.2864, + "step": 12439 + }, + { + "epoch": 0.3, + "learning_rate": 1.6434967056129496e-05, + "loss": 1.258, + "step": 12440 + }, + { + "epoch": 0.3, + "learning_rate": 1.6434370832766938e-05, + "loss": 1.0771, + "step": 12441 + }, + { + "epoch": 0.3, + "learning_rate": 1.6433774570368565e-05, + "loss": 1.2209, + "step": 12442 + }, + { + "epoch": 0.3, + "learning_rate": 1.6433178268937996e-05, + "loss": 0.9323, + "step": 12443 + }, + { + "epoch": 0.3, + "learning_rate": 1.6432581928478845e-05, + "loss": 0.9811, + "step": 12444 + }, + { + "epoch": 0.3, + "learning_rate": 1.6431985548994735e-05, + "loss": 1.1489, + "step": 12445 + }, + { + "epoch": 0.3, + "learning_rate": 1.643138913048928e-05, + "loss": 1.0501, + "step": 12446 + }, + { + "epoch": 0.3, + "learning_rate": 1.64307926729661e-05, + "loss": 1.2425, + "step": 12447 + }, + { + "epoch": 0.3, + "learning_rate": 1.6430196176428818e-05, + "loss": 1.1467, + "step": 12448 + }, + { + "epoch": 0.3, + "learning_rate": 1.6429599640881042e-05, + "loss": 1.3608, + "step": 12449 + }, + { + "epoch": 0.3, + "learning_rate": 1.64290030663264e-05, + "loss": 1.0718, + "step": 12450 + }, + { + "epoch": 0.3, + "learning_rate": 1.642840645276851e-05, + "loss": 1.1361, + "step": 12451 + }, + { + "epoch": 0.3, + "learning_rate": 1.642780980021099e-05, + "loss": 0.8895, + "step": 12452 + }, + { + "epoch": 0.3, + "learning_rate": 1.6427213108657457e-05, + "loss": 1.2152, + "step": 12453 + }, + { + "epoch": 0.3, + "learning_rate": 1.6426616378111533e-05, + "loss": 1.1552, + "step": 12454 + }, + { + "epoch": 0.3, + "learning_rate": 1.6426019608576842e-05, + "loss": 1.1652, + "step": 12455 + }, + { + "epoch": 0.3, + "learning_rate": 1.6425422800057002e-05, + "loss": 0.9617, + "step": 12456 + }, + { + "epoch": 0.3, + "learning_rate": 1.6424825952555628e-05, + "loss": 1.0589, + "step": 12457 + }, + { + "epoch": 0.3, + "learning_rate": 1.6424229066076353e-05, + "loss": 0.9473, + "step": 12458 + }, + { + "epoch": 0.3, + "learning_rate": 1.6423632140622788e-05, + "loss": 0.9966, + "step": 12459 + }, + { + "epoch": 0.3, + "learning_rate": 1.6423035176198558e-05, + "loss": 1.0873, + "step": 12460 + }, + { + "epoch": 0.3, + "learning_rate": 1.6422438172807282e-05, + "loss": 1.3151, + "step": 12461 + }, + { + "epoch": 0.3, + "learning_rate": 1.642184113045259e-05, + "loss": 1.1575, + "step": 12462 + }, + { + "epoch": 0.3, + "learning_rate": 1.642124404913809e-05, + "loss": 1.2057, + "step": 12463 + }, + { + "epoch": 0.3, + "learning_rate": 1.6420646928867422e-05, + "loss": 1.144, + "step": 12464 + }, + { + "epoch": 0.3, + "learning_rate": 1.642004976964419e-05, + "loss": 1.3369, + "step": 12465 + }, + { + "epoch": 0.3, + "learning_rate": 1.6419452571472035e-05, + "loss": 1.0513, + "step": 12466 + }, + { + "epoch": 0.3, + "learning_rate": 1.6418855334354567e-05, + "loss": 1.0655, + "step": 12467 + }, + { + "epoch": 0.3, + "learning_rate": 1.6418258058295413e-05, + "loss": 1.0558, + "step": 12468 + }, + { + "epoch": 0.3, + "learning_rate": 1.6417660743298202e-05, + "loss": 0.7613, + "step": 12469 + }, + { + "epoch": 0.3, + "learning_rate": 1.6417063389366546e-05, + "loss": 1.2424, + "step": 12470 + }, + { + "epoch": 0.3, + "learning_rate": 1.641646599650408e-05, + "loss": 1.0885, + "step": 12471 + }, + { + "epoch": 0.3, + "learning_rate": 1.6415868564714422e-05, + "loss": 1.0592, + "step": 12472 + }, + { + "epoch": 0.3, + "learning_rate": 1.6415271094001198e-05, + "loss": 1.0027, + "step": 12473 + }, + { + "epoch": 0.3, + "learning_rate": 1.6414673584368034e-05, + "loss": 1.0426, + "step": 12474 + }, + { + "epoch": 0.3, + "learning_rate": 1.6414076035818557e-05, + "loss": 1.1426, + "step": 12475 + }, + { + "epoch": 0.3, + "learning_rate": 1.6413478448356387e-05, + "loss": 1.0304, + "step": 12476 + }, + { + "epoch": 0.3, + "learning_rate": 1.641288082198515e-05, + "loss": 1.1153, + "step": 12477 + }, + { + "epoch": 0.3, + "learning_rate": 1.6412283156708477e-05, + "loss": 1.1114, + "step": 12478 + }, + { + "epoch": 0.3, + "learning_rate": 1.6411685452529988e-05, + "loss": 1.0112, + "step": 12479 + }, + { + "epoch": 0.3, + "learning_rate": 1.6411087709453313e-05, + "loss": 1.1288, + "step": 12480 + }, + { + "epoch": 0.3, + "learning_rate": 1.6410489927482075e-05, + "loss": 1.2868, + "step": 12481 + }, + { + "epoch": 0.3, + "learning_rate": 1.6409892106619905e-05, + "loss": 1.1944, + "step": 12482 + }, + { + "epoch": 0.3, + "learning_rate": 1.6409294246870425e-05, + "loss": 1.0694, + "step": 12483 + }, + { + "epoch": 0.3, + "learning_rate": 1.6408696348237264e-05, + "loss": 1.2326, + "step": 12484 + }, + { + "epoch": 0.3, + "learning_rate": 1.640809841072405e-05, + "loss": 0.9069, + "step": 12485 + }, + { + "epoch": 0.3, + "learning_rate": 1.640750043433441e-05, + "loss": 1.0097, + "step": 12486 + }, + { + "epoch": 0.3, + "learning_rate": 1.6406902419071976e-05, + "loss": 1.0852, + "step": 12487 + }, + { + "epoch": 0.3, + "learning_rate": 1.6406304364940368e-05, + "loss": 1.1998, + "step": 12488 + }, + { + "epoch": 0.3, + "learning_rate": 1.640570627194322e-05, + "loss": 1.0221, + "step": 12489 + }, + { + "epoch": 0.3, + "learning_rate": 1.6405108140084157e-05, + "loss": 1.2353, + "step": 12490 + }, + { + "epoch": 0.3, + "learning_rate": 1.6404509969366812e-05, + "loss": 0.9852, + "step": 12491 + }, + { + "epoch": 0.3, + "learning_rate": 1.640391175979481e-05, + "loss": 1.1854, + "step": 12492 + }, + { + "epoch": 0.3, + "learning_rate": 1.640331351137178e-05, + "loss": 1.1495, + "step": 12493 + }, + { + "epoch": 0.3, + "learning_rate": 1.6402715224101356e-05, + "loss": 1.1711, + "step": 12494 + }, + { + "epoch": 0.3, + "learning_rate": 1.6402116897987163e-05, + "loss": 0.8454, + "step": 12495 + }, + { + "epoch": 0.3, + "learning_rate": 1.640151853303283e-05, + "loss": 1.0209, + "step": 12496 + }, + { + "epoch": 0.3, + "learning_rate": 1.6400920129241994e-05, + "loss": 1.2729, + "step": 12497 + }, + { + "epoch": 0.3, + "learning_rate": 1.640032168661828e-05, + "loss": 1.1828, + "step": 12498 + }, + { + "epoch": 0.3, + "learning_rate": 1.639972320516532e-05, + "loss": 0.9109, + "step": 12499 + }, + { + "epoch": 0.3, + "learning_rate": 1.6399124684886744e-05, + "loss": 1.0506, + "step": 12500 + }, + { + "epoch": 0.3, + "learning_rate": 1.6398526125786185e-05, + "loss": 1.0461, + "step": 12501 + }, + { + "epoch": 0.3, + "learning_rate": 1.6397927527867272e-05, + "loss": 1.1667, + "step": 12502 + }, + { + "epoch": 0.3, + "learning_rate": 1.639732889113364e-05, + "loss": 1.0539, + "step": 12503 + }, + { + "epoch": 0.3, + "learning_rate": 1.6396730215588913e-05, + "loss": 1.2545, + "step": 12504 + }, + { + "epoch": 0.3, + "learning_rate": 1.6396131501236732e-05, + "loss": 1.4228, + "step": 12505 + }, + { + "epoch": 0.3, + "learning_rate": 1.6395532748080725e-05, + "loss": 0.9831, + "step": 12506 + }, + { + "epoch": 0.3, + "learning_rate": 1.6394933956124524e-05, + "loss": 1.3084, + "step": 12507 + }, + { + "epoch": 0.3, + "learning_rate": 1.6394335125371768e-05, + "loss": 1.1936, + "step": 12508 + }, + { + "epoch": 0.3, + "learning_rate": 1.6393736255826082e-05, + "loss": 1.1699, + "step": 12509 + }, + { + "epoch": 0.3, + "learning_rate": 1.6393137347491098e-05, + "loss": 0.8393, + "step": 12510 + }, + { + "epoch": 0.3, + "learning_rate": 1.6392538400370455e-05, + "loss": 1.1547, + "step": 12511 + }, + { + "epoch": 0.3, + "learning_rate": 1.6391939414467788e-05, + "loss": 0.9872, + "step": 12512 + }, + { + "epoch": 0.3, + "learning_rate": 1.6391340389786727e-05, + "loss": 1.0117, + "step": 12513 + }, + { + "epoch": 0.3, + "learning_rate": 1.639074132633091e-05, + "loss": 0.9804, + "step": 12514 + }, + { + "epoch": 0.3, + "learning_rate": 1.639014222410396e-05, + "loss": 1.0838, + "step": 12515 + }, + { + "epoch": 0.3, + "learning_rate": 1.6389543083109528e-05, + "loss": 1.3133, + "step": 12516 + }, + { + "epoch": 0.3, + "learning_rate": 1.638894390335124e-05, + "loss": 1.1312, + "step": 12517 + }, + { + "epoch": 0.3, + "learning_rate": 1.638834468483273e-05, + "loss": 1.2016, + "step": 12518 + }, + { + "epoch": 0.3, + "learning_rate": 1.6387745427557633e-05, + "loss": 1.004, + "step": 12519 + }, + { + "epoch": 0.3, + "learning_rate": 1.6387146131529594e-05, + "loss": 1.2869, + "step": 12520 + }, + { + "epoch": 0.3, + "learning_rate": 1.6386546796752237e-05, + "loss": 1.2821, + "step": 12521 + }, + { + "epoch": 0.3, + "learning_rate": 1.63859474232292e-05, + "loss": 1.1944, + "step": 12522 + }, + { + "epoch": 0.3, + "learning_rate": 1.638534801096413e-05, + "loss": 1.1634, + "step": 12523 + }, + { + "epoch": 0.3, + "learning_rate": 1.6384748559960648e-05, + "loss": 1.0057, + "step": 12524 + }, + { + "epoch": 0.3, + "learning_rate": 1.63841490702224e-05, + "loss": 1.3143, + "step": 12525 + }, + { + "epoch": 0.3, + "learning_rate": 1.6383549541753024e-05, + "loss": 1.1466, + "step": 12526 + }, + { + "epoch": 0.3, + "learning_rate": 1.6382949974556156e-05, + "loss": 1.096, + "step": 12527 + }, + { + "epoch": 0.3, + "learning_rate": 1.6382350368635428e-05, + "loss": 1.2223, + "step": 12528 + }, + { + "epoch": 0.3, + "learning_rate": 1.6381750723994482e-05, + "loss": 1.225, + "step": 12529 + }, + { + "epoch": 0.3, + "learning_rate": 1.6381151040636955e-05, + "loss": 1.408, + "step": 12530 + }, + { + "epoch": 0.3, + "learning_rate": 1.638055131856649e-05, + "loss": 1.1375, + "step": 12531 + }, + { + "epoch": 0.3, + "learning_rate": 1.6379951557786715e-05, + "loss": 1.068, + "step": 12532 + }, + { + "epoch": 0.3, + "learning_rate": 1.6379351758301276e-05, + "loss": 0.973, + "step": 12533 + }, + { + "epoch": 0.3, + "learning_rate": 1.6378751920113814e-05, + "loss": 1.3192, + "step": 12534 + }, + { + "epoch": 0.3, + "learning_rate": 1.637815204322796e-05, + "loss": 1.3239, + "step": 12535 + }, + { + "epoch": 0.3, + "learning_rate": 1.637755212764736e-05, + "loss": 1.2192, + "step": 12536 + }, + { + "epoch": 0.3, + "learning_rate": 1.637695217337565e-05, + "loss": 0.9274, + "step": 12537 + }, + { + "epoch": 0.3, + "learning_rate": 1.6376352180416476e-05, + "loss": 1.0688, + "step": 12538 + }, + { + "epoch": 0.3, + "learning_rate": 1.637575214877347e-05, + "loss": 1.0481, + "step": 12539 + }, + { + "epoch": 0.3, + "learning_rate": 1.637515207845027e-05, + "loss": 1.2234, + "step": 12540 + }, + { + "epoch": 0.3, + "learning_rate": 1.637455196945053e-05, + "loss": 1.3563, + "step": 12541 + }, + { + "epoch": 0.3, + "learning_rate": 1.637395182177788e-05, + "loss": 1.1411, + "step": 12542 + }, + { + "epoch": 0.3, + "learning_rate": 1.637335163543596e-05, + "loss": 1.1542, + "step": 12543 + }, + { + "epoch": 0.3, + "learning_rate": 1.6372751410428418e-05, + "loss": 1.1204, + "step": 12544 + }, + { + "epoch": 0.3, + "learning_rate": 1.6372151146758892e-05, + "loss": 1.0028, + "step": 12545 + }, + { + "epoch": 0.3, + "learning_rate": 1.6371550844431022e-05, + "loss": 0.8149, + "step": 12546 + }, + { + "epoch": 0.3, + "learning_rate": 1.637095050344845e-05, + "loss": 1.2024, + "step": 12547 + }, + { + "epoch": 0.3, + "learning_rate": 1.6370350123814826e-05, + "loss": 1.0838, + "step": 12548 + }, + { + "epoch": 0.3, + "learning_rate": 1.6369749705533784e-05, + "loss": 0.9779, + "step": 12549 + }, + { + "epoch": 0.3, + "learning_rate": 1.6369149248608967e-05, + "loss": 1.1824, + "step": 12550 + }, + { + "epoch": 0.3, + "learning_rate": 1.636854875304402e-05, + "loss": 1.3835, + "step": 12551 + }, + { + "epoch": 0.3, + "learning_rate": 1.6367948218842588e-05, + "loss": 0.8978, + "step": 12552 + }, + { + "epoch": 0.3, + "learning_rate": 1.636734764600831e-05, + "loss": 1.0363, + "step": 12553 + }, + { + "epoch": 0.3, + "learning_rate": 1.6366747034544828e-05, + "loss": 1.2961, + "step": 12554 + }, + { + "epoch": 0.3, + "learning_rate": 1.6366146384455798e-05, + "loss": 1.3421, + "step": 12555 + }, + { + "epoch": 0.3, + "learning_rate": 1.6365545695744847e-05, + "loss": 1.1659, + "step": 12556 + }, + { + "epoch": 0.3, + "learning_rate": 1.6364944968415632e-05, + "loss": 1.0795, + "step": 12557 + }, + { + "epoch": 0.3, + "learning_rate": 1.6364344202471793e-05, + "loss": 1.2414, + "step": 12558 + }, + { + "epoch": 0.3, + "learning_rate": 1.636374339791697e-05, + "loss": 1.1102, + "step": 12559 + }, + { + "epoch": 0.3, + "learning_rate": 1.6363142554754817e-05, + "loss": 1.2211, + "step": 12560 + }, + { + "epoch": 0.3, + "learning_rate": 1.6362541672988978e-05, + "loss": 0.7771, + "step": 12561 + }, + { + "epoch": 0.3, + "learning_rate": 1.636194075262309e-05, + "loss": 1.1762, + "step": 12562 + }, + { + "epoch": 0.3, + "learning_rate": 1.6361339793660804e-05, + "loss": 0.8328, + "step": 12563 + }, + { + "epoch": 0.3, + "learning_rate": 1.6360738796105767e-05, + "loss": 1.1382, + "step": 12564 + }, + { + "epoch": 0.3, + "learning_rate": 1.6360137759961624e-05, + "loss": 1.3048, + "step": 12565 + }, + { + "epoch": 0.3, + "learning_rate": 1.6359536685232022e-05, + "loss": 1.0698, + "step": 12566 + }, + { + "epoch": 0.3, + "learning_rate": 1.6358935571920606e-05, + "loss": 1.164, + "step": 12567 + }, + { + "epoch": 0.3, + "learning_rate": 1.6358334420031023e-05, + "loss": 1.5199, + "step": 12568 + }, + { + "epoch": 0.3, + "learning_rate": 1.6357733229566923e-05, + "loss": 1.1981, + "step": 12569 + }, + { + "epoch": 0.3, + "learning_rate": 1.6357132000531944e-05, + "loss": 1.1707, + "step": 12570 + }, + { + "epoch": 0.3, + "learning_rate": 1.635653073292975e-05, + "loss": 1.3222, + "step": 12571 + }, + { + "epoch": 0.3, + "learning_rate": 1.6355929426763974e-05, + "loss": 1.0015, + "step": 12572 + }, + { + "epoch": 0.3, + "learning_rate": 1.6355328082038267e-05, + "loss": 1.2313, + "step": 12573 + }, + { + "epoch": 0.3, + "learning_rate": 1.6354726698756285e-05, + "loss": 0.9117, + "step": 12574 + }, + { + "epoch": 0.3, + "learning_rate": 1.6354125276921665e-05, + "loss": 1.1165, + "step": 12575 + }, + { + "epoch": 0.3, + "learning_rate": 1.6353523816538065e-05, + "loss": 1.0757, + "step": 12576 + }, + { + "epoch": 0.3, + "learning_rate": 1.635292231760913e-05, + "loss": 1.2907, + "step": 12577 + }, + { + "epoch": 0.3, + "learning_rate": 1.635232078013851e-05, + "loss": 1.2304, + "step": 12578 + }, + { + "epoch": 0.3, + "learning_rate": 1.6351719204129855e-05, + "loss": 1.318, + "step": 12579 + }, + { + "epoch": 0.3, + "learning_rate": 1.635111758958681e-05, + "loss": 1.2113, + "step": 12580 + }, + { + "epoch": 0.3, + "learning_rate": 1.6350515936513033e-05, + "loss": 1.0491, + "step": 12581 + }, + { + "epoch": 0.3, + "learning_rate": 1.634991424491217e-05, + "loss": 0.9691, + "step": 12582 + }, + { + "epoch": 0.3, + "learning_rate": 1.6349312514787864e-05, + "loss": 1.2156, + "step": 12583 + }, + { + "epoch": 0.3, + "learning_rate": 1.634871074614378e-05, + "loss": 1.1864, + "step": 12584 + }, + { + "epoch": 0.3, + "learning_rate": 1.6348108938983553e-05, + "loss": 1.3256, + "step": 12585 + }, + { + "epoch": 0.3, + "learning_rate": 1.6347507093310852e-05, + "loss": 0.8941, + "step": 12586 + }, + { + "epoch": 0.3, + "learning_rate": 1.6346905209129312e-05, + "loss": 1.1689, + "step": 12587 + }, + { + "epoch": 0.3, + "learning_rate": 1.634630328644259e-05, + "loss": 1.2828, + "step": 12588 + }, + { + "epoch": 0.3, + "learning_rate": 1.634570132525434e-05, + "loss": 1.2599, + "step": 12589 + }, + { + "epoch": 0.3, + "learning_rate": 1.6345099325568217e-05, + "loss": 0.9513, + "step": 12590 + }, + { + "epoch": 0.3, + "learning_rate": 1.6344497287387865e-05, + "loss": 0.9668, + "step": 12591 + }, + { + "epoch": 0.3, + "learning_rate": 1.634389521071694e-05, + "loss": 1.2674, + "step": 12592 + }, + { + "epoch": 0.3, + "learning_rate": 1.6343293095559097e-05, + "loss": 1.2485, + "step": 12593 + }, + { + "epoch": 0.3, + "learning_rate": 1.6342690941917986e-05, + "loss": 0.9761, + "step": 12594 + }, + { + "epoch": 0.3, + "learning_rate": 1.634208874979726e-05, + "loss": 0.9739, + "step": 12595 + }, + { + "epoch": 0.3, + "learning_rate": 1.6341486519200572e-05, + "loss": 1.115, + "step": 12596 + }, + { + "epoch": 0.3, + "learning_rate": 1.6340884250131578e-05, + "loss": 1.1621, + "step": 12597 + }, + { + "epoch": 0.3, + "learning_rate": 1.6340281942593932e-05, + "loss": 1.0881, + "step": 12598 + }, + { + "epoch": 0.3, + "learning_rate": 1.6339679596591286e-05, + "loss": 1.3308, + "step": 12599 + }, + { + "epoch": 0.3, + "learning_rate": 1.6339077212127294e-05, + "loss": 1.0031, + "step": 12600 + }, + { + "epoch": 0.3, + "learning_rate": 1.6338474789205613e-05, + "loss": 1.1978, + "step": 12601 + }, + { + "epoch": 0.3, + "learning_rate": 1.6337872327829897e-05, + "loss": 0.993, + "step": 12602 + }, + { + "epoch": 0.3, + "learning_rate": 1.6337269828003802e-05, + "loss": 1.0172, + "step": 12603 + }, + { + "epoch": 0.3, + "learning_rate": 1.633666728973098e-05, + "loss": 1.2099, + "step": 12604 + }, + { + "epoch": 0.3, + "learning_rate": 1.6336064713015085e-05, + "loss": 1.1779, + "step": 12605 + }, + { + "epoch": 0.3, + "learning_rate": 1.633546209785978e-05, + "loss": 1.1762, + "step": 12606 + }, + { + "epoch": 0.3, + "learning_rate": 1.6334859444268713e-05, + "loss": 1.1432, + "step": 12607 + }, + { + "epoch": 0.3, + "learning_rate": 1.6334256752245546e-05, + "loss": 1.3626, + "step": 12608 + }, + { + "epoch": 0.3, + "learning_rate": 1.633365402179393e-05, + "loss": 1.2976, + "step": 12609 + }, + { + "epoch": 0.3, + "learning_rate": 1.633305125291753e-05, + "loss": 0.8142, + "step": 12610 + }, + { + "epoch": 0.3, + "learning_rate": 1.6332448445619996e-05, + "loss": 1.1521, + "step": 12611 + }, + { + "epoch": 0.3, + "learning_rate": 1.6331845599904984e-05, + "loss": 1.0801, + "step": 12612 + }, + { + "epoch": 0.3, + "learning_rate": 1.6331242715776158e-05, + "loss": 1.2574, + "step": 12613 + }, + { + "epoch": 0.3, + "learning_rate": 1.633063979323717e-05, + "loss": 1.0519, + "step": 12614 + }, + { + "epoch": 0.3, + "learning_rate": 1.6330036832291677e-05, + "loss": 1.0271, + "step": 12615 + }, + { + "epoch": 0.3, + "learning_rate": 1.6329433832943344e-05, + "loss": 0.852, + "step": 12616 + }, + { + "epoch": 0.3, + "learning_rate": 1.632883079519582e-05, + "loss": 1.3274, + "step": 12617 + }, + { + "epoch": 0.3, + "learning_rate": 1.632822771905277e-05, + "loss": 1.3186, + "step": 12618 + }, + { + "epoch": 0.3, + "learning_rate": 1.632762460451785e-05, + "loss": 0.8432, + "step": 12619 + }, + { + "epoch": 0.3, + "learning_rate": 1.6327021451594724e-05, + "loss": 1.0539, + "step": 12620 + }, + { + "epoch": 0.3, + "learning_rate": 1.6326418260287044e-05, + "loss": 1.0437, + "step": 12621 + }, + { + "epoch": 0.3, + "learning_rate": 1.6325815030598472e-05, + "loss": 1.2922, + "step": 12622 + }, + { + "epoch": 0.3, + "learning_rate": 1.632521176253267e-05, + "loss": 1.0509, + "step": 12623 + }, + { + "epoch": 0.3, + "learning_rate": 1.632460845609329e-05, + "loss": 1.3762, + "step": 12624 + }, + { + "epoch": 0.3, + "learning_rate": 1.6324005111284004e-05, + "loss": 1.3027, + "step": 12625 + }, + { + "epoch": 0.3, + "learning_rate": 1.6323401728108467e-05, + "loss": 1.2682, + "step": 12626 + }, + { + "epoch": 0.3, + "learning_rate": 1.6322798306570335e-05, + "loss": 1.0603, + "step": 12627 + }, + { + "epoch": 0.3, + "learning_rate": 1.6322194846673278e-05, + "loss": 1.1602, + "step": 12628 + }, + { + "epoch": 0.3, + "learning_rate": 1.6321591348420947e-05, + "loss": 1.033, + "step": 12629 + }, + { + "epoch": 0.3, + "learning_rate": 1.6320987811817013e-05, + "loss": 1.0564, + "step": 12630 + }, + { + "epoch": 0.3, + "learning_rate": 1.6320384236865127e-05, + "loss": 0.9755, + "step": 12631 + }, + { + "epoch": 0.3, + "learning_rate": 1.6319780623568964e-05, + "loss": 1.1244, + "step": 12632 + }, + { + "epoch": 0.3, + "learning_rate": 1.6319176971932172e-05, + "loss": 1.076, + "step": 12633 + }, + { + "epoch": 0.3, + "learning_rate": 1.6318573281958423e-05, + "loss": 1.0792, + "step": 12634 + }, + { + "epoch": 0.3, + "learning_rate": 1.6317969553651374e-05, + "loss": 1.3296, + "step": 12635 + }, + { + "epoch": 0.3, + "learning_rate": 1.631736578701469e-05, + "loss": 1.2805, + "step": 12636 + }, + { + "epoch": 0.3, + "learning_rate": 1.6316761982052035e-05, + "loss": 1.0142, + "step": 12637 + }, + { + "epoch": 0.3, + "learning_rate": 1.631615813876707e-05, + "loss": 1.1519, + "step": 12638 + }, + { + "epoch": 0.3, + "learning_rate": 1.631555425716346e-05, + "loss": 1.3625, + "step": 12639 + }, + { + "epoch": 0.3, + "learning_rate": 1.6314950337244866e-05, + "loss": 1.1893, + "step": 12640 + }, + { + "epoch": 0.3, + "learning_rate": 1.6314346379014958e-05, + "loss": 1.0912, + "step": 12641 + }, + { + "epoch": 0.3, + "learning_rate": 1.631374238247739e-05, + "loss": 1.1803, + "step": 12642 + }, + { + "epoch": 0.3, + "learning_rate": 1.631313834763584e-05, + "loss": 1.2076, + "step": 12643 + }, + { + "epoch": 0.3, + "learning_rate": 1.6312534274493955e-05, + "loss": 1.3956, + "step": 12644 + }, + { + "epoch": 0.3, + "learning_rate": 1.6311930163055415e-05, + "loss": 1.2283, + "step": 12645 + }, + { + "epoch": 0.3, + "learning_rate": 1.6311326013323878e-05, + "loss": 1.1683, + "step": 12646 + }, + { + "epoch": 0.3, + "learning_rate": 1.6310721825303012e-05, + "loss": 1.243, + "step": 12647 + }, + { + "epoch": 0.3, + "learning_rate": 1.6310117598996477e-05, + "loss": 1.0488, + "step": 12648 + }, + { + "epoch": 0.3, + "learning_rate": 1.6309513334407946e-05, + "loss": 1.1902, + "step": 12649 + }, + { + "epoch": 0.3, + "learning_rate": 1.630890903154108e-05, + "loss": 0.9297, + "step": 12650 + }, + { + "epoch": 0.3, + "learning_rate": 1.630830469039955e-05, + "loss": 1.2288, + "step": 12651 + }, + { + "epoch": 0.3, + "learning_rate": 1.6307700310987018e-05, + "loss": 0.7923, + "step": 12652 + }, + { + "epoch": 0.3, + "learning_rate": 1.630709589330715e-05, + "loss": 1.139, + "step": 12653 + }, + { + "epoch": 0.3, + "learning_rate": 1.6306491437363615e-05, + "loss": 0.9194, + "step": 12654 + }, + { + "epoch": 0.3, + "learning_rate": 1.630588694316008e-05, + "loss": 1.204, + "step": 12655 + }, + { + "epoch": 0.3, + "learning_rate": 1.6305282410700217e-05, + "loss": 0.9128, + "step": 12656 + }, + { + "epoch": 0.3, + "learning_rate": 1.6304677839987683e-05, + "loss": 1.2169, + "step": 12657 + }, + { + "epoch": 0.3, + "learning_rate": 1.6304073231026154e-05, + "loss": 1.115, + "step": 12658 + }, + { + "epoch": 0.3, + "learning_rate": 1.6303468583819295e-05, + "loss": 1.2508, + "step": 12659 + }, + { + "epoch": 0.3, + "learning_rate": 1.6302863898370772e-05, + "loss": 0.9135, + "step": 12660 + }, + { + "epoch": 0.3, + "learning_rate": 1.630225917468426e-05, + "loss": 0.9949, + "step": 12661 + }, + { + "epoch": 0.3, + "learning_rate": 1.6301654412763425e-05, + "loss": 1.1136, + "step": 12662 + }, + { + "epoch": 0.3, + "learning_rate": 1.6301049612611933e-05, + "loss": 1.2899, + "step": 12663 + }, + { + "epoch": 0.3, + "learning_rate": 1.630044477423345e-05, + "loss": 1.179, + "step": 12664 + }, + { + "epoch": 0.3, + "learning_rate": 1.629983989763166e-05, + "loss": 1.3734, + "step": 12665 + }, + { + "epoch": 0.3, + "learning_rate": 1.6299234982810216e-05, + "loss": 1.3372, + "step": 12666 + }, + { + "epoch": 0.3, + "learning_rate": 1.6298630029772797e-05, + "loss": 1.203, + "step": 12667 + }, + { + "epoch": 0.3, + "learning_rate": 1.6298025038523073e-05, + "loss": 1.2072, + "step": 12668 + }, + { + "epoch": 0.3, + "learning_rate": 1.6297420009064712e-05, + "loss": 1.3572, + "step": 12669 + }, + { + "epoch": 0.3, + "learning_rate": 1.6296814941401384e-05, + "loss": 1.0498, + "step": 12670 + }, + { + "epoch": 0.3, + "learning_rate": 1.629620983553676e-05, + "loss": 0.9914, + "step": 12671 + }, + { + "epoch": 0.3, + "learning_rate": 1.6295604691474514e-05, + "loss": 1.1925, + "step": 12672 + }, + { + "epoch": 0.3, + "learning_rate": 1.629499950921831e-05, + "loss": 1.1826, + "step": 12673 + }, + { + "epoch": 0.3, + "learning_rate": 1.6294394288771832e-05, + "loss": 1.0351, + "step": 12674 + }, + { + "epoch": 0.3, + "learning_rate": 1.629378903013874e-05, + "loss": 1.1285, + "step": 12675 + }, + { + "epoch": 0.3, + "learning_rate": 1.629318373332271e-05, + "loss": 1.1072, + "step": 12676 + }, + { + "epoch": 0.3, + "learning_rate": 1.6292578398327415e-05, + "loss": 1.2547, + "step": 12677 + }, + { + "epoch": 0.3, + "learning_rate": 1.629197302515653e-05, + "loss": 1.2206, + "step": 12678 + }, + { + "epoch": 0.3, + "learning_rate": 1.6291367613813722e-05, + "loss": 1.3439, + "step": 12679 + }, + { + "epoch": 0.3, + "learning_rate": 1.6290762164302666e-05, + "loss": 1.2846, + "step": 12680 + }, + { + "epoch": 0.3, + "learning_rate": 1.6290156676627037e-05, + "loss": 1.0371, + "step": 12681 + }, + { + "epoch": 0.3, + "learning_rate": 1.628955115079051e-05, + "loss": 1.1953, + "step": 12682 + }, + { + "epoch": 0.31, + "learning_rate": 1.628894558679675e-05, + "loss": 1.1478, + "step": 12683 + }, + { + "epoch": 0.31, + "learning_rate": 1.628833998464944e-05, + "loss": 1.0843, + "step": 12684 + }, + { + "epoch": 0.31, + "learning_rate": 1.6287734344352245e-05, + "loss": 1.0019, + "step": 12685 + }, + { + "epoch": 0.31, + "learning_rate": 1.628712866590885e-05, + "loss": 1.285, + "step": 12686 + }, + { + "epoch": 0.31, + "learning_rate": 1.6286522949322924e-05, + "loss": 1.0651, + "step": 12687 + }, + { + "epoch": 0.31, + "learning_rate": 1.628591719459814e-05, + "loss": 1.3027, + "step": 12688 + }, + { + "epoch": 0.31, + "learning_rate": 1.6285311401738174e-05, + "loss": 0.934, + "step": 12689 + }, + { + "epoch": 0.31, + "learning_rate": 1.6284705570746708e-05, + "loss": 1.2349, + "step": 12690 + }, + { + "epoch": 0.31, + "learning_rate": 1.6284099701627405e-05, + "loss": 1.3482, + "step": 12691 + }, + { + "epoch": 0.31, + "learning_rate": 1.6283493794383953e-05, + "loss": 0.9979, + "step": 12692 + }, + { + "epoch": 0.31, + "learning_rate": 1.6282887849020016e-05, + "loss": 1.042, + "step": 12693 + }, + { + "epoch": 0.31, + "learning_rate": 1.628228186553928e-05, + "loss": 1.0791, + "step": 12694 + }, + { + "epoch": 0.31, + "learning_rate": 1.6281675843945416e-05, + "loss": 1.1081, + "step": 12695 + }, + { + "epoch": 0.31, + "learning_rate": 1.62810697842421e-05, + "loss": 1.0923, + "step": 12696 + }, + { + "epoch": 0.31, + "learning_rate": 1.6280463686433017e-05, + "loss": 1.0084, + "step": 12697 + }, + { + "epoch": 0.31, + "learning_rate": 1.6279857550521833e-05, + "loss": 1.229, + "step": 12698 + }, + { + "epoch": 0.31, + "learning_rate": 1.627925137651223e-05, + "loss": 1.0783, + "step": 12699 + }, + { + "epoch": 0.31, + "learning_rate": 1.6278645164407892e-05, + "loss": 1.0171, + "step": 12700 + }, + { + "epoch": 0.31, + "learning_rate": 1.6278038914212485e-05, + "loss": 1.15, + "step": 12701 + }, + { + "epoch": 0.31, + "learning_rate": 1.6277432625929696e-05, + "loss": 0.9939, + "step": 12702 + }, + { + "epoch": 0.31, + "learning_rate": 1.6276826299563198e-05, + "loss": 1.0773, + "step": 12703 + }, + { + "epoch": 0.31, + "learning_rate": 1.627621993511667e-05, + "loss": 1.0848, + "step": 12704 + }, + { + "epoch": 0.31, + "learning_rate": 1.627561353259379e-05, + "loss": 1.2574, + "step": 12705 + }, + { + "epoch": 0.31, + "learning_rate": 1.6275007091998246e-05, + "loss": 1.2925, + "step": 12706 + }, + { + "epoch": 0.31, + "learning_rate": 1.6274400613333706e-05, + "loss": 1.0284, + "step": 12707 + }, + { + "epoch": 0.31, + "learning_rate": 1.6273794096603854e-05, + "loss": 1.2129, + "step": 12708 + }, + { + "epoch": 0.31, + "learning_rate": 1.6273187541812366e-05, + "loss": 1.2137, + "step": 12709 + }, + { + "epoch": 0.31, + "learning_rate": 1.6272580948962926e-05, + "loss": 0.8672, + "step": 12710 + }, + { + "epoch": 0.31, + "learning_rate": 1.6271974318059213e-05, + "loss": 1.3027, + "step": 12711 + }, + { + "epoch": 0.31, + "learning_rate": 1.6271367649104906e-05, + "loss": 1.2248, + "step": 12712 + }, + { + "epoch": 0.31, + "learning_rate": 1.6270760942103692e-05, + "loss": 1.1671, + "step": 12713 + }, + { + "epoch": 0.31, + "learning_rate": 1.627015419705924e-05, + "loss": 1.1359, + "step": 12714 + }, + { + "epoch": 0.31, + "learning_rate": 1.6269547413975243e-05, + "loss": 1.1811, + "step": 12715 + }, + { + "epoch": 0.31, + "learning_rate": 1.626894059285537e-05, + "loss": 1.1098, + "step": 12716 + }, + { + "epoch": 0.31, + "learning_rate": 1.6268333733703315e-05, + "loss": 0.9329, + "step": 12717 + }, + { + "epoch": 0.31, + "learning_rate": 1.626772683652275e-05, + "loss": 1.1708, + "step": 12718 + }, + { + "epoch": 0.31, + "learning_rate": 1.6267119901317362e-05, + "loss": 1.2332, + "step": 12719 + }, + { + "epoch": 0.31, + "learning_rate": 1.6266512928090833e-05, + "loss": 1.1178, + "step": 12720 + }, + { + "epoch": 0.31, + "learning_rate": 1.6265905916846843e-05, + "loss": 1.065, + "step": 12721 + }, + { + "epoch": 0.31, + "learning_rate": 1.6265298867589074e-05, + "loss": 1.2185, + "step": 12722 + }, + { + "epoch": 0.31, + "learning_rate": 1.6264691780321207e-05, + "loss": 1.0585, + "step": 12723 + }, + { + "epoch": 0.31, + "learning_rate": 1.6264084655046938e-05, + "loss": 0.8622, + "step": 12724 + }, + { + "epoch": 0.31, + "learning_rate": 1.6263477491769936e-05, + "loss": 1.1736, + "step": 12725 + }, + { + "epoch": 0.31, + "learning_rate": 1.6262870290493887e-05, + "loss": 1.0012, + "step": 12726 + }, + { + "epoch": 0.31, + "learning_rate": 1.626226305122248e-05, + "loss": 1.0315, + "step": 12727 + }, + { + "epoch": 0.31, + "learning_rate": 1.6261655773959395e-05, + "loss": 1.1399, + "step": 12728 + }, + { + "epoch": 0.31, + "learning_rate": 1.6261048458708318e-05, + "loss": 1.257, + "step": 12729 + }, + { + "epoch": 0.31, + "learning_rate": 1.6260441105472932e-05, + "loss": 1.1245, + "step": 12730 + }, + { + "epoch": 0.31, + "learning_rate": 1.6259833714256924e-05, + "loss": 1.2133, + "step": 12731 + }, + { + "epoch": 0.31, + "learning_rate": 1.6259226285063974e-05, + "loss": 1.1418, + "step": 12732 + }, + { + "epoch": 0.31, + "learning_rate": 1.6258618817897775e-05, + "loss": 1.0028, + "step": 12733 + }, + { + "epoch": 0.31, + "learning_rate": 1.6258011312762002e-05, + "loss": 1.3049, + "step": 12734 + }, + { + "epoch": 0.31, + "learning_rate": 1.625740376966035e-05, + "loss": 0.9891, + "step": 12735 + }, + { + "epoch": 0.31, + "learning_rate": 1.62567961885965e-05, + "loss": 1.1556, + "step": 12736 + }, + { + "epoch": 0.31, + "learning_rate": 1.6256188569574138e-05, + "loss": 1.0595, + "step": 12737 + }, + { + "epoch": 0.31, + "learning_rate": 1.6255580912596952e-05, + "loss": 1.4916, + "step": 12738 + }, + { + "epoch": 0.31, + "learning_rate": 1.625497321766863e-05, + "loss": 1.0837, + "step": 12739 + }, + { + "epoch": 0.31, + "learning_rate": 1.625436548479286e-05, + "loss": 0.8633, + "step": 12740 + }, + { + "epoch": 0.31, + "learning_rate": 1.625375771397332e-05, + "loss": 1.0066, + "step": 12741 + }, + { + "epoch": 0.31, + "learning_rate": 1.6253149905213702e-05, + "loss": 1.3392, + "step": 12742 + }, + { + "epoch": 0.31, + "learning_rate": 1.62525420585177e-05, + "loss": 0.898, + "step": 12743 + }, + { + "epoch": 0.31, + "learning_rate": 1.6251934173888992e-05, + "loss": 1.1698, + "step": 12744 + }, + { + "epoch": 0.31, + "learning_rate": 1.625132625133127e-05, + "loss": 1.3778, + "step": 12745 + }, + { + "epoch": 0.31, + "learning_rate": 1.6250718290848223e-05, + "loss": 0.9835, + "step": 12746 + }, + { + "epoch": 0.31, + "learning_rate": 1.625011029244354e-05, + "loss": 1.1476, + "step": 12747 + }, + { + "epoch": 0.31, + "learning_rate": 1.6249502256120906e-05, + "loss": 0.8395, + "step": 12748 + }, + { + "epoch": 0.31, + "learning_rate": 1.6248894181884008e-05, + "loss": 0.9236, + "step": 12749 + }, + { + "epoch": 0.31, + "learning_rate": 1.6248286069736544e-05, + "loss": 0.873, + "step": 12750 + }, + { + "epoch": 0.31, + "learning_rate": 1.6247677919682198e-05, + "loss": 1.2325, + "step": 12751 + }, + { + "epoch": 0.31, + "learning_rate": 1.624706973172466e-05, + "loss": 1.1528, + "step": 12752 + }, + { + "epoch": 0.31, + "learning_rate": 1.6246461505867617e-05, + "loss": 1.1765, + "step": 12753 + }, + { + "epoch": 0.31, + "learning_rate": 1.624585324211476e-05, + "loss": 1.2927, + "step": 12754 + }, + { + "epoch": 0.31, + "learning_rate": 1.6245244940469784e-05, + "loss": 1.2814, + "step": 12755 + }, + { + "epoch": 0.31, + "learning_rate": 1.6244636600936376e-05, + "loss": 0.952, + "step": 12756 + }, + { + "epoch": 0.31, + "learning_rate": 1.6244028223518224e-05, + "loss": 1.2423, + "step": 12757 + }, + { + "epoch": 0.31, + "learning_rate": 1.6243419808219024e-05, + "loss": 0.9489, + "step": 12758 + }, + { + "epoch": 0.31, + "learning_rate": 1.6242811355042462e-05, + "loss": 1.2583, + "step": 12759 + }, + { + "epoch": 0.31, + "learning_rate": 1.6242202863992235e-05, + "loss": 1.1542, + "step": 12760 + }, + { + "epoch": 0.31, + "learning_rate": 1.6241594335072028e-05, + "loss": 1.0626, + "step": 12761 + }, + { + "epoch": 0.31, + "learning_rate": 1.624098576828554e-05, + "loss": 1.349, + "step": 12762 + }, + { + "epoch": 0.31, + "learning_rate": 1.6240377163636454e-05, + "loss": 0.9716, + "step": 12763 + }, + { + "epoch": 0.31, + "learning_rate": 1.623976852112847e-05, + "loss": 1.106, + "step": 12764 + }, + { + "epoch": 0.31, + "learning_rate": 1.6239159840765283e-05, + "loss": 1.2199, + "step": 12765 + }, + { + "epoch": 0.31, + "learning_rate": 1.6238551122550573e-05, + "loss": 1.2055, + "step": 12766 + }, + { + "epoch": 0.31, + "learning_rate": 1.6237942366488043e-05, + "loss": 0.9486, + "step": 12767 + }, + { + "epoch": 0.31, + "learning_rate": 1.623733357258139e-05, + "loss": 1.0945, + "step": 12768 + }, + { + "epoch": 0.31, + "learning_rate": 1.6236724740834296e-05, + "loss": 1.2218, + "step": 12769 + }, + { + "epoch": 0.31, + "learning_rate": 1.623611587125046e-05, + "loss": 1.4561, + "step": 12770 + }, + { + "epoch": 0.31, + "learning_rate": 1.6235506963833577e-05, + "loss": 1.1657, + "step": 12771 + }, + { + "epoch": 0.31, + "learning_rate": 1.6234898018587336e-05, + "loss": 1.1009, + "step": 12772 + }, + { + "epoch": 0.31, + "learning_rate": 1.623428903551544e-05, + "loss": 0.9434, + "step": 12773 + }, + { + "epoch": 0.31, + "learning_rate": 1.6233680014621576e-05, + "loss": 1.2388, + "step": 12774 + }, + { + "epoch": 0.31, + "learning_rate": 1.6233070955909444e-05, + "loss": 1.3553, + "step": 12775 + }, + { + "epoch": 0.31, + "learning_rate": 1.6232461859382736e-05, + "loss": 1.1539, + "step": 12776 + }, + { + "epoch": 0.31, + "learning_rate": 1.6231852725045148e-05, + "loss": 1.1933, + "step": 12777 + }, + { + "epoch": 0.31, + "learning_rate": 1.6231243552900372e-05, + "loss": 1.1866, + "step": 12778 + }, + { + "epoch": 0.31, + "learning_rate": 1.623063434295211e-05, + "loss": 1.1584, + "step": 12779 + }, + { + "epoch": 0.31, + "learning_rate": 1.6230025095204058e-05, + "loss": 1.2282, + "step": 12780 + }, + { + "epoch": 0.31, + "learning_rate": 1.6229415809659904e-05, + "loss": 1.2868, + "step": 12781 + }, + { + "epoch": 0.31, + "learning_rate": 1.6228806486323353e-05, + "loss": 0.9732, + "step": 12782 + }, + { + "epoch": 0.31, + "learning_rate": 1.6228197125198097e-05, + "loss": 1.1988, + "step": 12783 + }, + { + "epoch": 0.31, + "learning_rate": 1.6227587726287833e-05, + "loss": 1.1243, + "step": 12784 + }, + { + "epoch": 0.31, + "learning_rate": 1.622697828959626e-05, + "loss": 0.7138, + "step": 12785 + }, + { + "epoch": 0.31, + "learning_rate": 1.6226368815127073e-05, + "loss": 1.097, + "step": 12786 + }, + { + "epoch": 0.31, + "learning_rate": 1.6225759302883973e-05, + "loss": 1.2302, + "step": 12787 + }, + { + "epoch": 0.31, + "learning_rate": 1.622514975287065e-05, + "loss": 1.0875, + "step": 12788 + }, + { + "epoch": 0.31, + "learning_rate": 1.6224540165090817e-05, + "loss": 1.2637, + "step": 12789 + }, + { + "epoch": 0.31, + "learning_rate": 1.6223930539548157e-05, + "loss": 0.8435, + "step": 12790 + }, + { + "epoch": 0.31, + "learning_rate": 1.6223320876246374e-05, + "loss": 1.2083, + "step": 12791 + }, + { + "epoch": 0.31, + "learning_rate": 1.622271117518917e-05, + "loss": 0.6706, + "step": 12792 + }, + { + "epoch": 0.31, + "learning_rate": 1.622210143638024e-05, + "loss": 1.2336, + "step": 12793 + }, + { + "epoch": 0.31, + "learning_rate": 1.6221491659823283e-05, + "loss": 1.2474, + "step": 12794 + }, + { + "epoch": 0.31, + "learning_rate": 1.6220881845522e-05, + "loss": 1.2147, + "step": 12795 + }, + { + "epoch": 0.31, + "learning_rate": 1.622027199348009e-05, + "loss": 0.8393, + "step": 12796 + }, + { + "epoch": 0.31, + "learning_rate": 1.6219662103701254e-05, + "loss": 0.8652, + "step": 12797 + }, + { + "epoch": 0.31, + "learning_rate": 1.6219052176189193e-05, + "loss": 1.2131, + "step": 12798 + }, + { + "epoch": 0.31, + "learning_rate": 1.62184422109476e-05, + "loss": 1.3261, + "step": 12799 + }, + { + "epoch": 0.31, + "learning_rate": 1.6217832207980185e-05, + "loss": 1.1787, + "step": 12800 + }, + { + "epoch": 0.31, + "learning_rate": 1.621722216729064e-05, + "loss": 1.0262, + "step": 12801 + }, + { + "epoch": 0.31, + "learning_rate": 1.6216612088882677e-05, + "loss": 1.1004, + "step": 12802 + }, + { + "epoch": 0.31, + "learning_rate": 1.6216001972759984e-05, + "loss": 1.6341, + "step": 12803 + }, + { + "epoch": 0.31, + "learning_rate": 1.6215391818926272e-05, + "loss": 1.1982, + "step": 12804 + }, + { + "epoch": 0.31, + "learning_rate": 1.621478162738524e-05, + "loss": 1.1044, + "step": 12805 + }, + { + "epoch": 0.31, + "learning_rate": 1.6214171398140592e-05, + "loss": 0.9482, + "step": 12806 + }, + { + "epoch": 0.31, + "learning_rate": 1.6213561131196022e-05, + "loss": 1.2198, + "step": 12807 + }, + { + "epoch": 0.31, + "learning_rate": 1.6212950826555243e-05, + "loss": 0.9642, + "step": 12808 + }, + { + "epoch": 0.31, + "learning_rate": 1.6212340484221953e-05, + "loss": 1.0299, + "step": 12809 + }, + { + "epoch": 0.31, + "learning_rate": 1.6211730104199853e-05, + "loss": 1.2122, + "step": 12810 + }, + { + "epoch": 0.31, + "learning_rate": 1.6211119686492645e-05, + "loss": 0.9833, + "step": 12811 + }, + { + "epoch": 0.31, + "learning_rate": 1.6210509231104037e-05, + "loss": 0.8653, + "step": 12812 + }, + { + "epoch": 0.31, + "learning_rate": 1.6209898738037734e-05, + "loss": 1.0467, + "step": 12813 + }, + { + "epoch": 0.31, + "learning_rate": 1.620928820729743e-05, + "loss": 1.0143, + "step": 12814 + }, + { + "epoch": 0.31, + "learning_rate": 1.6208677638886835e-05, + "loss": 1.0602, + "step": 12815 + }, + { + "epoch": 0.31, + "learning_rate": 1.6208067032809656e-05, + "loss": 0.9822, + "step": 12816 + }, + { + "epoch": 0.31, + "learning_rate": 1.6207456389069595e-05, + "loss": 1.3645, + "step": 12817 + }, + { + "epoch": 0.31, + "learning_rate": 1.6206845707670354e-05, + "loss": 1.2067, + "step": 12818 + }, + { + "epoch": 0.31, + "learning_rate": 1.6206234988615643e-05, + "loss": 0.9745, + "step": 12819 + }, + { + "epoch": 0.31, + "learning_rate": 1.6205624231909163e-05, + "loss": 0.9708, + "step": 12820 + }, + { + "epoch": 0.31, + "learning_rate": 1.620501343755462e-05, + "loss": 1.1476, + "step": 12821 + }, + { + "epoch": 0.31, + "learning_rate": 1.6204402605555718e-05, + "loss": 0.8709, + "step": 12822 + }, + { + "epoch": 0.31, + "learning_rate": 1.6203791735916168e-05, + "loss": 1.1908, + "step": 12823 + }, + { + "epoch": 0.31, + "learning_rate": 1.620318082863967e-05, + "loss": 0.9902, + "step": 12824 + }, + { + "epoch": 0.31, + "learning_rate": 1.6202569883729936e-05, + "loss": 0.8746, + "step": 12825 + }, + { + "epoch": 0.31, + "learning_rate": 1.620195890119067e-05, + "loss": 1.1268, + "step": 12826 + }, + { + "epoch": 0.31, + "learning_rate": 1.6201347881025574e-05, + "loss": 1.189, + "step": 12827 + }, + { + "epoch": 0.31, + "learning_rate": 1.620073682323836e-05, + "loss": 1.0746, + "step": 12828 + }, + { + "epoch": 0.31, + "learning_rate": 1.6200125727832738e-05, + "loss": 1.1839, + "step": 12829 + }, + { + "epoch": 0.31, + "learning_rate": 1.6199514594812405e-05, + "loss": 0.8993, + "step": 12830 + }, + { + "epoch": 0.31, + "learning_rate": 1.619890342418108e-05, + "loss": 0.8168, + "step": 12831 + }, + { + "epoch": 0.31, + "learning_rate": 1.6198292215942463e-05, + "loss": 1.1484, + "step": 12832 + }, + { + "epoch": 0.31, + "learning_rate": 1.619768097010027e-05, + "loss": 1.1845, + "step": 12833 + }, + { + "epoch": 0.31, + "learning_rate": 1.61970696866582e-05, + "loss": 1.2215, + "step": 12834 + }, + { + "epoch": 0.31, + "learning_rate": 1.619645836561997e-05, + "loss": 1.0604, + "step": 12835 + }, + { + "epoch": 0.31, + "learning_rate": 1.619584700698928e-05, + "loss": 1.1095, + "step": 12836 + }, + { + "epoch": 0.31, + "learning_rate": 1.6195235610769847e-05, + "loss": 1.2044, + "step": 12837 + }, + { + "epoch": 0.31, + "learning_rate": 1.619462417696537e-05, + "loss": 1.2716, + "step": 12838 + }, + { + "epoch": 0.31, + "learning_rate": 1.6194012705579574e-05, + "loss": 1.0494, + "step": 12839 + }, + { + "epoch": 0.31, + "learning_rate": 1.6193401196616153e-05, + "loss": 1.1529, + "step": 12840 + }, + { + "epoch": 0.31, + "learning_rate": 1.619278965007883e-05, + "loss": 1.0728, + "step": 12841 + }, + { + "epoch": 0.31, + "learning_rate": 1.6192178065971304e-05, + "loss": 1.0221, + "step": 12842 + }, + { + "epoch": 0.31, + "learning_rate": 1.619156644429729e-05, + "loss": 1.0676, + "step": 12843 + }, + { + "epoch": 0.31, + "learning_rate": 1.6190954785060502e-05, + "loss": 1.0462, + "step": 12844 + }, + { + "epoch": 0.31, + "learning_rate": 1.6190343088264646e-05, + "loss": 1.154, + "step": 12845 + }, + { + "epoch": 0.31, + "learning_rate": 1.6189731353913435e-05, + "loss": 1.0942, + "step": 12846 + }, + { + "epoch": 0.31, + "learning_rate": 1.618911958201058e-05, + "loss": 0.9494, + "step": 12847 + }, + { + "epoch": 0.31, + "learning_rate": 1.618850777255979e-05, + "loss": 1.0094, + "step": 12848 + }, + { + "epoch": 0.31, + "learning_rate": 1.618789592556478e-05, + "loss": 1.045, + "step": 12849 + }, + { + "epoch": 0.31, + "learning_rate": 1.6187284041029262e-05, + "loss": 1.1293, + "step": 12850 + }, + { + "epoch": 0.31, + "learning_rate": 1.6186672118956947e-05, + "loss": 1.2285, + "step": 12851 + }, + { + "epoch": 0.31, + "learning_rate": 1.6186060159351547e-05, + "loss": 1.241, + "step": 12852 + }, + { + "epoch": 0.31, + "learning_rate": 1.6185448162216774e-05, + "loss": 1.0561, + "step": 12853 + }, + { + "epoch": 0.31, + "learning_rate": 1.6184836127556345e-05, + "loss": 1.2335, + "step": 12854 + }, + { + "epoch": 0.31, + "learning_rate": 1.6184224055373968e-05, + "loss": 1.2189, + "step": 12855 + }, + { + "epoch": 0.31, + "learning_rate": 1.6183611945673358e-05, + "loss": 0.876, + "step": 12856 + }, + { + "epoch": 0.31, + "learning_rate": 1.6182999798458228e-05, + "loss": 1.1617, + "step": 12857 + }, + { + "epoch": 0.31, + "learning_rate": 1.6182387613732295e-05, + "loss": 1.3392, + "step": 12858 + }, + { + "epoch": 0.31, + "learning_rate": 1.618177539149927e-05, + "loss": 1.0912, + "step": 12859 + }, + { + "epoch": 0.31, + "learning_rate": 1.6181163131762865e-05, + "loss": 1.0927, + "step": 12860 + }, + { + "epoch": 0.31, + "learning_rate": 1.61805508345268e-05, + "loss": 1.3022, + "step": 12861 + }, + { + "epoch": 0.31, + "learning_rate": 1.6179938499794784e-05, + "loss": 1.2199, + "step": 12862 + }, + { + "epoch": 0.31, + "learning_rate": 1.617932612757054e-05, + "loss": 1.1277, + "step": 12863 + }, + { + "epoch": 0.31, + "learning_rate": 1.6178713717857774e-05, + "loss": 1.1452, + "step": 12864 + }, + { + "epoch": 0.31, + "learning_rate": 1.6178101270660206e-05, + "loss": 1.3771, + "step": 12865 + }, + { + "epoch": 0.31, + "learning_rate": 1.617748878598155e-05, + "loss": 1.0414, + "step": 12866 + }, + { + "epoch": 0.31, + "learning_rate": 1.617687626382552e-05, + "loss": 0.9456, + "step": 12867 + }, + { + "epoch": 0.31, + "learning_rate": 1.6176263704195842e-05, + "loss": 1.2125, + "step": 12868 + }, + { + "epoch": 0.31, + "learning_rate": 1.6175651107096214e-05, + "loss": 0.9304, + "step": 12869 + }, + { + "epoch": 0.31, + "learning_rate": 1.6175038472530375e-05, + "loss": 1.3187, + "step": 12870 + }, + { + "epoch": 0.31, + "learning_rate": 1.6174425800502023e-05, + "loss": 1.0878, + "step": 12871 + }, + { + "epoch": 0.31, + "learning_rate": 1.6173813091014886e-05, + "loss": 1.1658, + "step": 12872 + }, + { + "epoch": 0.31, + "learning_rate": 1.6173200344072674e-05, + "loss": 0.8773, + "step": 12873 + }, + { + "epoch": 0.31, + "learning_rate": 1.6172587559679106e-05, + "loss": 1.1706, + "step": 12874 + }, + { + "epoch": 0.31, + "learning_rate": 1.6171974737837904e-05, + "loss": 1.3048, + "step": 12875 + }, + { + "epoch": 0.31, + "learning_rate": 1.617136187855278e-05, + "loss": 1.0895, + "step": 12876 + }, + { + "epoch": 0.31, + "learning_rate": 1.617074898182746e-05, + "loss": 1.2192, + "step": 12877 + }, + { + "epoch": 0.31, + "learning_rate": 1.6170136047665656e-05, + "loss": 1.3578, + "step": 12878 + }, + { + "epoch": 0.31, + "learning_rate": 1.6169523076071085e-05, + "loss": 1.3015, + "step": 12879 + }, + { + "epoch": 0.31, + "learning_rate": 1.616891006704747e-05, + "loss": 1.2198, + "step": 12880 + }, + { + "epoch": 0.31, + "learning_rate": 1.616829702059853e-05, + "loss": 1.1355, + "step": 12881 + }, + { + "epoch": 0.31, + "learning_rate": 1.616768393672798e-05, + "loss": 0.9995, + "step": 12882 + }, + { + "epoch": 0.31, + "learning_rate": 1.6167070815439548e-05, + "loss": 0.9623, + "step": 12883 + }, + { + "epoch": 0.31, + "learning_rate": 1.6166457656736944e-05, + "loss": 1.0595, + "step": 12884 + }, + { + "epoch": 0.31, + "learning_rate": 1.6165844460623893e-05, + "loss": 1.408, + "step": 12885 + }, + { + "epoch": 0.31, + "learning_rate": 1.6165231227104113e-05, + "loss": 1.2921, + "step": 12886 + }, + { + "epoch": 0.31, + "learning_rate": 1.6164617956181326e-05, + "loss": 1.1887, + "step": 12887 + }, + { + "epoch": 0.31, + "learning_rate": 1.616400464785925e-05, + "loss": 1.1457, + "step": 12888 + }, + { + "epoch": 0.31, + "learning_rate": 1.616339130214161e-05, + "loss": 1.194, + "step": 12889 + }, + { + "epoch": 0.31, + "learning_rate": 1.616277791903212e-05, + "loss": 1.0308, + "step": 12890 + }, + { + "epoch": 0.31, + "learning_rate": 1.6162164498534513e-05, + "loss": 1.2873, + "step": 12891 + }, + { + "epoch": 0.31, + "learning_rate": 1.61615510406525e-05, + "loss": 1.1332, + "step": 12892 + }, + { + "epoch": 0.31, + "learning_rate": 1.6160937545389804e-05, + "loss": 0.936, + "step": 12893 + }, + { + "epoch": 0.31, + "learning_rate": 1.6160324012750154e-05, + "loss": 1.1479, + "step": 12894 + }, + { + "epoch": 0.31, + "learning_rate": 1.6159710442737262e-05, + "loss": 1.0259, + "step": 12895 + }, + { + "epoch": 0.31, + "learning_rate": 1.6159096835354858e-05, + "loss": 1.0099, + "step": 12896 + }, + { + "epoch": 0.31, + "learning_rate": 1.6158483190606663e-05, + "loss": 0.9007, + "step": 12897 + }, + { + "epoch": 0.31, + "learning_rate": 1.6157869508496398e-05, + "loss": 1.1375, + "step": 12898 + }, + { + "epoch": 0.31, + "learning_rate": 1.6157255789027784e-05, + "loss": 1.22, + "step": 12899 + }, + { + "epoch": 0.31, + "learning_rate": 1.6156642032204554e-05, + "loss": 1.1362, + "step": 12900 + }, + { + "epoch": 0.31, + "learning_rate": 1.615602823803042e-05, + "loss": 1.2218, + "step": 12901 + }, + { + "epoch": 0.31, + "learning_rate": 1.6155414406509112e-05, + "loss": 1.1076, + "step": 12902 + }, + { + "epoch": 0.31, + "learning_rate": 1.6154800537644355e-05, + "loss": 1.013, + "step": 12903 + }, + { + "epoch": 0.31, + "learning_rate": 1.615418663143987e-05, + "loss": 0.9179, + "step": 12904 + }, + { + "epoch": 0.31, + "learning_rate": 1.615357268789938e-05, + "loss": 1.1308, + "step": 12905 + }, + { + "epoch": 0.31, + "learning_rate": 1.6152958707026615e-05, + "loss": 1.1597, + "step": 12906 + }, + { + "epoch": 0.31, + "learning_rate": 1.6152344688825294e-05, + "loss": 1.0802, + "step": 12907 + }, + { + "epoch": 0.31, + "learning_rate": 1.6151730633299144e-05, + "loss": 1.1779, + "step": 12908 + }, + { + "epoch": 0.31, + "learning_rate": 1.6151116540451895e-05, + "loss": 1.0513, + "step": 12909 + }, + { + "epoch": 0.31, + "learning_rate": 1.6150502410287265e-05, + "loss": 1.0271, + "step": 12910 + }, + { + "epoch": 0.31, + "learning_rate": 1.6149888242808987e-05, + "loss": 1.2449, + "step": 12911 + }, + { + "epoch": 0.31, + "learning_rate": 1.614927403802078e-05, + "loss": 1.0573, + "step": 12912 + }, + { + "epoch": 0.31, + "learning_rate": 1.6148659795926375e-05, + "loss": 1.045, + "step": 12913 + }, + { + "epoch": 0.31, + "learning_rate": 1.61480455165295e-05, + "loss": 0.9963, + "step": 12914 + }, + { + "epoch": 0.31, + "learning_rate": 1.6147431199833875e-05, + "loss": 1.0599, + "step": 12915 + }, + { + "epoch": 0.31, + "learning_rate": 1.6146816845843235e-05, + "loss": 1.0403, + "step": 12916 + }, + { + "epoch": 0.31, + "learning_rate": 1.61462024545613e-05, + "loss": 0.9344, + "step": 12917 + }, + { + "epoch": 0.31, + "learning_rate": 1.61455880259918e-05, + "loss": 1.0308, + "step": 12918 + }, + { + "epoch": 0.31, + "learning_rate": 1.6144973560138465e-05, + "loss": 0.902, + "step": 12919 + }, + { + "epoch": 0.31, + "learning_rate": 1.6144359057005018e-05, + "loss": 1.1463, + "step": 12920 + }, + { + "epoch": 0.31, + "learning_rate": 1.6143744516595193e-05, + "loss": 0.9189, + "step": 12921 + }, + { + "epoch": 0.31, + "learning_rate": 1.6143129938912712e-05, + "loss": 1.1546, + "step": 12922 + }, + { + "epoch": 0.31, + "learning_rate": 1.614251532396131e-05, + "loss": 0.9554, + "step": 12923 + }, + { + "epoch": 0.31, + "learning_rate": 1.6141900671744712e-05, + "loss": 1.0337, + "step": 12924 + }, + { + "epoch": 0.31, + "learning_rate": 1.6141285982266643e-05, + "loss": 1.0913, + "step": 12925 + }, + { + "epoch": 0.31, + "learning_rate": 1.6140671255530843e-05, + "loss": 1.0337, + "step": 12926 + }, + { + "epoch": 0.31, + "learning_rate": 1.614005649154103e-05, + "loss": 1.2953, + "step": 12927 + }, + { + "epoch": 0.31, + "learning_rate": 1.613944169030094e-05, + "loss": 1.1484, + "step": 12928 + }, + { + "epoch": 0.31, + "learning_rate": 1.6138826851814302e-05, + "loss": 1.0071, + "step": 12929 + }, + { + "epoch": 0.31, + "learning_rate": 1.6138211976084845e-05, + "loss": 1.0784, + "step": 12930 + }, + { + "epoch": 0.31, + "learning_rate": 1.6137597063116295e-05, + "loss": 1.1972, + "step": 12931 + }, + { + "epoch": 0.31, + "learning_rate": 1.6136982112912394e-05, + "loss": 0.9545, + "step": 12932 + }, + { + "epoch": 0.31, + "learning_rate": 1.6136367125476865e-05, + "loss": 0.9965, + "step": 12933 + }, + { + "epoch": 0.31, + "learning_rate": 1.613575210081344e-05, + "loss": 1.0117, + "step": 12934 + }, + { + "epoch": 0.31, + "learning_rate": 1.6135137038925848e-05, + "loss": 0.8963, + "step": 12935 + }, + { + "epoch": 0.31, + "learning_rate": 1.6134521939817824e-05, + "loss": 1.4828, + "step": 12936 + }, + { + "epoch": 0.31, + "learning_rate": 1.6133906803493097e-05, + "loss": 1.2946, + "step": 12937 + }, + { + "epoch": 0.31, + "learning_rate": 1.61332916299554e-05, + "loss": 1.1195, + "step": 12938 + }, + { + "epoch": 0.31, + "learning_rate": 1.6132676419208467e-05, + "loss": 1.2609, + "step": 12939 + }, + { + "epoch": 0.31, + "learning_rate": 1.613206117125603e-05, + "loss": 1.2844, + "step": 12940 + }, + { + "epoch": 0.31, + "learning_rate": 1.6131445886101816e-05, + "loss": 0.7504, + "step": 12941 + }, + { + "epoch": 0.31, + "learning_rate": 1.6130830563749565e-05, + "loss": 0.8546, + "step": 12942 + }, + { + "epoch": 0.31, + "learning_rate": 1.6130215204203006e-05, + "loss": 1.0123, + "step": 12943 + }, + { + "epoch": 0.31, + "learning_rate": 1.6129599807465873e-05, + "loss": 1.0875, + "step": 12944 + }, + { + "epoch": 0.31, + "learning_rate": 1.6128984373541903e-05, + "loss": 1.2259, + "step": 12945 + }, + { + "epoch": 0.31, + "learning_rate": 1.6128368902434823e-05, + "loss": 1.1889, + "step": 12946 + }, + { + "epoch": 0.31, + "learning_rate": 1.612775339414837e-05, + "loss": 0.991, + "step": 12947 + }, + { + "epoch": 0.31, + "learning_rate": 1.612713784868628e-05, + "loss": 1.0864, + "step": 12948 + }, + { + "epoch": 0.31, + "learning_rate": 1.6126522266052285e-05, + "loss": 1.2788, + "step": 12949 + }, + { + "epoch": 0.31, + "learning_rate": 1.6125906646250124e-05, + "loss": 1.0338, + "step": 12950 + }, + { + "epoch": 0.31, + "learning_rate": 1.6125290989283524e-05, + "loss": 1.2181, + "step": 12951 + }, + { + "epoch": 0.31, + "learning_rate": 1.6124675295156227e-05, + "loss": 1.343, + "step": 12952 + }, + { + "epoch": 0.31, + "learning_rate": 1.6124059563871962e-05, + "loss": 0.9572, + "step": 12953 + }, + { + "epoch": 0.31, + "learning_rate": 1.612344379543447e-05, + "loss": 1.1484, + "step": 12954 + }, + { + "epoch": 0.31, + "learning_rate": 1.6122827989847488e-05, + "loss": 1.1137, + "step": 12955 + }, + { + "epoch": 0.31, + "learning_rate": 1.6122212147114745e-05, + "loss": 1.1239, + "step": 12956 + }, + { + "epoch": 0.31, + "learning_rate": 1.612159626723998e-05, + "loss": 1.1445, + "step": 12957 + }, + { + "epoch": 0.31, + "learning_rate": 1.6120980350226933e-05, + "loss": 0.9984, + "step": 12958 + }, + { + "epoch": 0.31, + "learning_rate": 1.6120364396079336e-05, + "loss": 0.8999, + "step": 12959 + }, + { + "epoch": 0.31, + "learning_rate": 1.6119748404800927e-05, + "loss": 1.1782, + "step": 12960 + }, + { + "epoch": 0.31, + "learning_rate": 1.611913237639545e-05, + "loss": 1.1746, + "step": 12961 + }, + { + "epoch": 0.31, + "learning_rate": 1.611851631086663e-05, + "loss": 1.156, + "step": 12962 + }, + { + "epoch": 0.31, + "learning_rate": 1.6117900208218212e-05, + "loss": 1.276, + "step": 12963 + }, + { + "epoch": 0.31, + "learning_rate": 1.6117284068453932e-05, + "loss": 1.3426, + "step": 12964 + }, + { + "epoch": 0.31, + "learning_rate": 1.611666789157753e-05, + "loss": 1.0569, + "step": 12965 + }, + { + "epoch": 0.31, + "learning_rate": 1.6116051677592743e-05, + "loss": 1.053, + "step": 12966 + }, + { + "epoch": 0.31, + "learning_rate": 1.611543542650331e-05, + "loss": 1.099, + "step": 12967 + }, + { + "epoch": 0.31, + "learning_rate": 1.6114819138312964e-05, + "loss": 1.0652, + "step": 12968 + }, + { + "epoch": 0.31, + "learning_rate": 1.6114202813025453e-05, + "loss": 1.1027, + "step": 12969 + }, + { + "epoch": 0.31, + "learning_rate": 1.611358645064451e-05, + "loss": 1.2456, + "step": 12970 + }, + { + "epoch": 0.31, + "learning_rate": 1.6112970051173876e-05, + "loss": 1.0801, + "step": 12971 + }, + { + "epoch": 0.31, + "learning_rate": 1.611235361461729e-05, + "loss": 1.1084, + "step": 12972 + }, + { + "epoch": 0.31, + "learning_rate": 1.6111737140978495e-05, + "loss": 1.2774, + "step": 12973 + }, + { + "epoch": 0.31, + "learning_rate": 1.6111120630261224e-05, + "loss": 1.1196, + "step": 12974 + }, + { + "epoch": 0.31, + "learning_rate": 1.6110504082469227e-05, + "loss": 0.9895, + "step": 12975 + }, + { + "epoch": 0.31, + "learning_rate": 1.6109887497606233e-05, + "loss": 1.0782, + "step": 12976 + }, + { + "epoch": 0.31, + "learning_rate": 1.6109270875675993e-05, + "loss": 1.2549, + "step": 12977 + }, + { + "epoch": 0.31, + "learning_rate": 1.6108654216682244e-05, + "loss": 1.2969, + "step": 12978 + }, + { + "epoch": 0.31, + "learning_rate": 1.6108037520628723e-05, + "loss": 1.1328, + "step": 12979 + }, + { + "epoch": 0.31, + "learning_rate": 1.6107420787519178e-05, + "loss": 1.0493, + "step": 12980 + }, + { + "epoch": 0.31, + "learning_rate": 1.6106804017357346e-05, + "loss": 1.0358, + "step": 12981 + }, + { + "epoch": 0.31, + "learning_rate": 1.610618721014697e-05, + "loss": 1.2345, + "step": 12982 + }, + { + "epoch": 0.31, + "learning_rate": 1.6105570365891794e-05, + "loss": 1.0441, + "step": 12983 + }, + { + "epoch": 0.31, + "learning_rate": 1.610495348459556e-05, + "loss": 1.214, + "step": 12984 + }, + { + "epoch": 0.31, + "learning_rate": 1.6104336566262006e-05, + "loss": 1.3044, + "step": 12985 + }, + { + "epoch": 0.31, + "learning_rate": 1.6103719610894878e-05, + "loss": 1.3418, + "step": 12986 + }, + { + "epoch": 0.31, + "learning_rate": 1.6103102618497922e-05, + "loss": 1.1526, + "step": 12987 + }, + { + "epoch": 0.31, + "learning_rate": 1.6102485589074877e-05, + "loss": 0.7843, + "step": 12988 + }, + { + "epoch": 0.31, + "learning_rate": 1.6101868522629484e-05, + "loss": 1.1507, + "step": 12989 + }, + { + "epoch": 0.31, + "learning_rate": 1.6101251419165493e-05, + "loss": 1.0737, + "step": 12990 + }, + { + "epoch": 0.31, + "learning_rate": 1.6100634278686643e-05, + "loss": 0.9484, + "step": 12991 + }, + { + "epoch": 0.31, + "learning_rate": 1.6100017101196685e-05, + "loss": 0.989, + "step": 12992 + }, + { + "epoch": 0.31, + "learning_rate": 1.6099399886699352e-05, + "loss": 1.1749, + "step": 12993 + }, + { + "epoch": 0.31, + "learning_rate": 1.6098782635198396e-05, + "loss": 0.9378, + "step": 12994 + }, + { + "epoch": 0.31, + "learning_rate": 1.609816534669756e-05, + "loss": 1.1411, + "step": 12995 + }, + { + "epoch": 0.31, + "learning_rate": 1.6097548021200592e-05, + "loss": 0.9747, + "step": 12996 + }, + { + "epoch": 0.31, + "learning_rate": 1.609693065871123e-05, + "loss": 0.9936, + "step": 12997 + }, + { + "epoch": 0.31, + "learning_rate": 1.6096313259233227e-05, + "loss": 0.8452, + "step": 12998 + }, + { + "epoch": 0.31, + "learning_rate": 1.6095695822770324e-05, + "loss": 1.139, + "step": 12999 + }, + { + "epoch": 0.31, + "learning_rate": 1.6095078349326268e-05, + "loss": 1.1958, + "step": 13000 + }, + { + "epoch": 0.31, + "learning_rate": 1.609446083890481e-05, + "loss": 1.1302, + "step": 13001 + }, + { + "epoch": 0.31, + "learning_rate": 1.6093843291509685e-05, + "loss": 0.9617, + "step": 13002 + }, + { + "epoch": 0.31, + "learning_rate": 1.609322570714465e-05, + "loss": 0.7823, + "step": 13003 + }, + { + "epoch": 0.31, + "learning_rate": 1.6092608085813448e-05, + "loss": 1.2197, + "step": 13004 + }, + { + "epoch": 0.31, + "learning_rate": 1.6091990427519823e-05, + "loss": 0.9369, + "step": 13005 + }, + { + "epoch": 0.31, + "learning_rate": 1.6091372732267525e-05, + "loss": 1.2052, + "step": 13006 + }, + { + "epoch": 0.31, + "learning_rate": 1.6090755000060304e-05, + "loss": 1.0749, + "step": 13007 + }, + { + "epoch": 0.31, + "learning_rate": 1.6090137230901905e-05, + "loss": 0.9208, + "step": 13008 + }, + { + "epoch": 0.31, + "learning_rate": 1.6089519424796074e-05, + "loss": 1.1858, + "step": 13009 + }, + { + "epoch": 0.31, + "learning_rate": 1.6088901581746562e-05, + "loss": 0.9933, + "step": 13010 + }, + { + "epoch": 0.31, + "learning_rate": 1.6088283701757113e-05, + "loss": 0.8864, + "step": 13011 + }, + { + "epoch": 0.31, + "learning_rate": 1.6087665784831482e-05, + "loss": 1.2593, + "step": 13012 + }, + { + "epoch": 0.31, + "learning_rate": 1.6087047830973418e-05, + "loss": 0.9544, + "step": 13013 + }, + { + "epoch": 0.31, + "learning_rate": 1.6086429840186663e-05, + "loss": 0.9505, + "step": 13014 + }, + { + "epoch": 0.31, + "learning_rate": 1.608581181247497e-05, + "loss": 1.4325, + "step": 13015 + }, + { + "epoch": 0.31, + "learning_rate": 1.6085193747842086e-05, + "loss": 1.0872, + "step": 13016 + }, + { + "epoch": 0.31, + "learning_rate": 1.6084575646291767e-05, + "loss": 0.7764, + "step": 13017 + }, + { + "epoch": 0.31, + "learning_rate": 1.6083957507827758e-05, + "loss": 1.3404, + "step": 13018 + }, + { + "epoch": 0.31, + "learning_rate": 1.6083339332453807e-05, + "loss": 1.347, + "step": 13019 + }, + { + "epoch": 0.31, + "learning_rate": 1.608272112017367e-05, + "loss": 1.3395, + "step": 13020 + }, + { + "epoch": 0.31, + "learning_rate": 1.6082102870991093e-05, + "loss": 1.1127, + "step": 13021 + }, + { + "epoch": 0.31, + "learning_rate": 1.6081484584909828e-05, + "loss": 1.1008, + "step": 13022 + }, + { + "epoch": 0.31, + "learning_rate": 1.6080866261933627e-05, + "loss": 1.3765, + "step": 13023 + }, + { + "epoch": 0.31, + "learning_rate": 1.608024790206624e-05, + "loss": 1.2078, + "step": 13024 + }, + { + "epoch": 0.31, + "learning_rate": 1.607962950531142e-05, + "loss": 0.9207, + "step": 13025 + }, + { + "epoch": 0.31, + "learning_rate": 1.6079011071672923e-05, + "loss": 1.1247, + "step": 13026 + }, + { + "epoch": 0.31, + "learning_rate": 1.6078392601154485e-05, + "loss": 0.9368, + "step": 13027 + }, + { + "epoch": 0.31, + "learning_rate": 1.6077774093759875e-05, + "loss": 1.1093, + "step": 13028 + }, + { + "epoch": 0.31, + "learning_rate": 1.607715554949284e-05, + "loss": 1.207, + "step": 13029 + }, + { + "epoch": 0.31, + "learning_rate": 1.607653696835713e-05, + "loss": 1.3925, + "step": 13030 + }, + { + "epoch": 0.31, + "learning_rate": 1.60759183503565e-05, + "loss": 1.3028, + "step": 13031 + }, + { + "epoch": 0.31, + "learning_rate": 1.6075299695494703e-05, + "loss": 1.2118, + "step": 13032 + }, + { + "epoch": 0.31, + "learning_rate": 1.607468100377549e-05, + "loss": 1.0311, + "step": 13033 + }, + { + "epoch": 0.31, + "learning_rate": 1.607406227520262e-05, + "loss": 0.9968, + "step": 13034 + }, + { + "epoch": 0.31, + "learning_rate": 1.607344350977984e-05, + "loss": 1.328, + "step": 13035 + }, + { + "epoch": 0.31, + "learning_rate": 1.6072824707510905e-05, + "loss": 0.977, + "step": 13036 + }, + { + "epoch": 0.31, + "learning_rate": 1.6072205868399572e-05, + "loss": 1.3203, + "step": 13037 + }, + { + "epoch": 0.31, + "learning_rate": 1.6071586992449597e-05, + "loss": 1.0133, + "step": 13038 + }, + { + "epoch": 0.31, + "learning_rate": 1.6070968079664728e-05, + "loss": 1.2732, + "step": 13039 + }, + { + "epoch": 0.31, + "learning_rate": 1.6070349130048724e-05, + "loss": 0.9433, + "step": 13040 + }, + { + "epoch": 0.31, + "learning_rate": 1.606973014360534e-05, + "loss": 1.0446, + "step": 13041 + }, + { + "epoch": 0.31, + "learning_rate": 1.6069111120338335e-05, + "loss": 1.3005, + "step": 13042 + }, + { + "epoch": 0.31, + "learning_rate": 1.6068492060251458e-05, + "loss": 0.978, + "step": 13043 + }, + { + "epoch": 0.31, + "learning_rate": 1.6067872963348466e-05, + "loss": 1.0748, + "step": 13044 + }, + { + "epoch": 0.31, + "learning_rate": 1.6067253829633118e-05, + "loss": 1.3012, + "step": 13045 + }, + { + "epoch": 0.31, + "learning_rate": 1.6066634659109167e-05, + "loss": 1.3471, + "step": 13046 + }, + { + "epoch": 0.31, + "learning_rate": 1.606601545178037e-05, + "loss": 0.9332, + "step": 13047 + }, + { + "epoch": 0.31, + "learning_rate": 1.6065396207650482e-05, + "loss": 1.1399, + "step": 13048 + }, + { + "epoch": 0.31, + "learning_rate": 1.6064776926723267e-05, + "loss": 1.1387, + "step": 13049 + }, + { + "epoch": 0.31, + "learning_rate": 1.6064157609002473e-05, + "loss": 1.225, + "step": 13050 + }, + { + "epoch": 0.31, + "learning_rate": 1.6063538254491862e-05, + "loss": 1.1127, + "step": 13051 + }, + { + "epoch": 0.31, + "learning_rate": 1.606291886319519e-05, + "loss": 0.9114, + "step": 13052 + }, + { + "epoch": 0.31, + "learning_rate": 1.6062299435116218e-05, + "loss": 1.1409, + "step": 13053 + }, + { + "epoch": 0.31, + "learning_rate": 1.60616799702587e-05, + "loss": 0.9958, + "step": 13054 + }, + { + "epoch": 0.31, + "learning_rate": 1.6061060468626394e-05, + "loss": 1.274, + "step": 13055 + }, + { + "epoch": 0.31, + "learning_rate": 1.606044093022306e-05, + "loss": 1.3434, + "step": 13056 + }, + { + "epoch": 0.31, + "learning_rate": 1.605982135505246e-05, + "loss": 0.9679, + "step": 13057 + }, + { + "epoch": 0.31, + "learning_rate": 1.6059201743118342e-05, + "loss": 0.9496, + "step": 13058 + }, + { + "epoch": 0.31, + "learning_rate": 1.6058582094424478e-05, + "loss": 1.4135, + "step": 13059 + }, + { + "epoch": 0.31, + "learning_rate": 1.605796240897462e-05, + "loss": 1.0868, + "step": 13060 + }, + { + "epoch": 0.31, + "learning_rate": 1.605734268677253e-05, + "loss": 1.2725, + "step": 13061 + }, + { + "epoch": 0.31, + "learning_rate": 1.6056722927821966e-05, + "loss": 1.2238, + "step": 13062 + }, + { + "epoch": 0.31, + "learning_rate": 1.605610313212669e-05, + "loss": 1.0715, + "step": 13063 + }, + { + "epoch": 0.31, + "learning_rate": 1.6055483299690458e-05, + "loss": 1.2199, + "step": 13064 + }, + { + "epoch": 0.31, + "learning_rate": 1.605486343051703e-05, + "loss": 0.9612, + "step": 13065 + }, + { + "epoch": 0.31, + "learning_rate": 1.6054243524610175e-05, + "loss": 1.4394, + "step": 13066 + }, + { + "epoch": 0.31, + "learning_rate": 1.605362358197365e-05, + "loss": 1.1937, + "step": 13067 + }, + { + "epoch": 0.31, + "learning_rate": 1.605300360261121e-05, + "loss": 1.2114, + "step": 13068 + }, + { + "epoch": 0.31, + "learning_rate": 1.6052383586526622e-05, + "loss": 1.3239, + "step": 13069 + }, + { + "epoch": 0.31, + "learning_rate": 1.605176353372365e-05, + "loss": 0.741, + "step": 13070 + }, + { + "epoch": 0.31, + "learning_rate": 1.605114344420605e-05, + "loss": 1.2119, + "step": 13071 + }, + { + "epoch": 0.31, + "learning_rate": 1.6050523317977586e-05, + "loss": 1.2865, + "step": 13072 + }, + { + "epoch": 0.31, + "learning_rate": 1.604990315504202e-05, + "loss": 0.995, + "step": 13073 + }, + { + "epoch": 0.31, + "learning_rate": 1.6049282955403115e-05, + "loss": 1.2217, + "step": 13074 + }, + { + "epoch": 0.31, + "learning_rate": 1.6048662719064633e-05, + "loss": 0.9041, + "step": 13075 + }, + { + "epoch": 0.31, + "learning_rate": 1.6048042446030336e-05, + "loss": 1.0918, + "step": 13076 + }, + { + "epoch": 0.31, + "learning_rate": 1.6047422136303988e-05, + "loss": 1.3136, + "step": 13077 + }, + { + "epoch": 0.31, + "learning_rate": 1.6046801789889353e-05, + "loss": 1.1514, + "step": 13078 + }, + { + "epoch": 0.31, + "learning_rate": 1.6046181406790192e-05, + "loss": 1.2126, + "step": 13079 + }, + { + "epoch": 0.31, + "learning_rate": 1.6045560987010276e-05, + "loss": 1.0703, + "step": 13080 + }, + { + "epoch": 0.31, + "learning_rate": 1.6044940530553356e-05, + "loss": 1.1567, + "step": 13081 + }, + { + "epoch": 0.31, + "learning_rate": 1.604432003742321e-05, + "loss": 1.1376, + "step": 13082 + }, + { + "epoch": 0.31, + "learning_rate": 1.6043699507623594e-05, + "loss": 1.114, + "step": 13083 + }, + { + "epoch": 0.31, + "learning_rate": 1.6043078941158273e-05, + "loss": 1.0617, + "step": 13084 + }, + { + "epoch": 0.31, + "learning_rate": 1.6042458338031015e-05, + "loss": 1.2628, + "step": 13085 + }, + { + "epoch": 0.31, + "learning_rate": 1.6041837698245583e-05, + "loss": 1.2503, + "step": 13086 + }, + { + "epoch": 0.31, + "learning_rate": 1.6041217021805743e-05, + "loss": 0.9758, + "step": 13087 + }, + { + "epoch": 0.31, + "learning_rate": 1.604059630871526e-05, + "loss": 0.849, + "step": 13088 + }, + { + "epoch": 0.31, + "learning_rate": 1.6039975558977903e-05, + "loss": 0.7227, + "step": 13089 + }, + { + "epoch": 0.31, + "learning_rate": 1.6039354772597432e-05, + "loss": 1.1344, + "step": 13090 + }, + { + "epoch": 0.31, + "learning_rate": 1.6038733949577615e-05, + "loss": 1.0358, + "step": 13091 + }, + { + "epoch": 0.31, + "learning_rate": 1.6038113089922224e-05, + "loss": 0.8655, + "step": 13092 + }, + { + "epoch": 0.31, + "learning_rate": 1.6037492193635017e-05, + "loss": 1.1375, + "step": 13093 + }, + { + "epoch": 0.31, + "learning_rate": 1.6036871260719766e-05, + "loss": 1.2673, + "step": 13094 + }, + { + "epoch": 0.31, + "learning_rate": 1.6036250291180238e-05, + "loss": 0.9693, + "step": 13095 + }, + { + "epoch": 0.31, + "learning_rate": 1.60356292850202e-05, + "loss": 1.0913, + "step": 13096 + }, + { + "epoch": 0.31, + "learning_rate": 1.6035008242243413e-05, + "loss": 1.0915, + "step": 13097 + }, + { + "epoch": 0.31, + "learning_rate": 1.6034387162853656e-05, + "loss": 0.9765, + "step": 13098 + }, + { + "epoch": 0.32, + "learning_rate": 1.603376604685469e-05, + "loss": 1.1515, + "step": 13099 + }, + { + "epoch": 0.32, + "learning_rate": 1.603314489425029e-05, + "loss": 1.2902, + "step": 13100 + }, + { + "epoch": 0.32, + "learning_rate": 1.603252370504421e-05, + "loss": 0.938, + "step": 13101 + }, + { + "epoch": 0.32, + "learning_rate": 1.603190247924023e-05, + "loss": 0.9178, + "step": 13102 + }, + { + "epoch": 0.32, + "learning_rate": 1.603128121684212e-05, + "loss": 1.1906, + "step": 13103 + }, + { + "epoch": 0.32, + "learning_rate": 1.6030659917853643e-05, + "loss": 0.9509, + "step": 13104 + }, + { + "epoch": 0.32, + "learning_rate": 1.603003858227857e-05, + "loss": 1.2287, + "step": 13105 + }, + { + "epoch": 0.32, + "learning_rate": 1.6029417210120674e-05, + "loss": 1.0382, + "step": 13106 + }, + { + "epoch": 0.32, + "learning_rate": 1.6028795801383717e-05, + "loss": 1.1493, + "step": 13107 + }, + { + "epoch": 0.32, + "learning_rate": 1.602817435607148e-05, + "loss": 1.3092, + "step": 13108 + }, + { + "epoch": 0.32, + "learning_rate": 1.6027552874187727e-05, + "loss": 0.9937, + "step": 13109 + }, + { + "epoch": 0.32, + "learning_rate": 1.6026931355736225e-05, + "loss": 1.3247, + "step": 13110 + }, + { + "epoch": 0.32, + "learning_rate": 1.6026309800720747e-05, + "loss": 1.1637, + "step": 13111 + }, + { + "epoch": 0.32, + "learning_rate": 1.6025688209145068e-05, + "loss": 1.3327, + "step": 13112 + }, + { + "epoch": 0.32, + "learning_rate": 1.6025066581012954e-05, + "loss": 1.2788, + "step": 13113 + }, + { + "epoch": 0.32, + "learning_rate": 1.602444491632818e-05, + "loss": 1.254, + "step": 13114 + }, + { + "epoch": 0.32, + "learning_rate": 1.6023823215094517e-05, + "loss": 1.2887, + "step": 13115 + }, + { + "epoch": 0.32, + "learning_rate": 1.602320147731573e-05, + "loss": 1.059, + "step": 13116 + }, + { + "epoch": 0.32, + "learning_rate": 1.60225797029956e-05, + "loss": 1.2005, + "step": 13117 + }, + { + "epoch": 0.32, + "learning_rate": 1.6021957892137895e-05, + "loss": 0.9906, + "step": 13118 + }, + { + "epoch": 0.32, + "learning_rate": 1.6021336044746393e-05, + "loss": 1.1688, + "step": 13119 + }, + { + "epoch": 0.32, + "learning_rate": 1.6020714160824854e-05, + "loss": 1.1505, + "step": 13120 + }, + { + "epoch": 0.32, + "learning_rate": 1.6020092240377062e-05, + "loss": 1.0954, + "step": 13121 + }, + { + "epoch": 0.32, + "learning_rate": 1.6019470283406785e-05, + "loss": 0.9269, + "step": 13122 + }, + { + "epoch": 0.32, + "learning_rate": 1.60188482899178e-05, + "loss": 0.9627, + "step": 13123 + }, + { + "epoch": 0.32, + "learning_rate": 1.6018226259913872e-05, + "loss": 1.0032, + "step": 13124 + }, + { + "epoch": 0.32, + "learning_rate": 1.6017604193398786e-05, + "loss": 1.3716, + "step": 13125 + }, + { + "epoch": 0.32, + "learning_rate": 1.6016982090376308e-05, + "loss": 1.2077, + "step": 13126 + }, + { + "epoch": 0.32, + "learning_rate": 1.601635995085022e-05, + "loss": 1.2119, + "step": 13127 + }, + { + "epoch": 0.32, + "learning_rate": 1.6015737774824287e-05, + "loss": 1.1047, + "step": 13128 + }, + { + "epoch": 0.32, + "learning_rate": 1.601511556230229e-05, + "loss": 1.0543, + "step": 13129 + }, + { + "epoch": 0.32, + "learning_rate": 1.6014493313288e-05, + "loss": 1.0841, + "step": 13130 + }, + { + "epoch": 0.32, + "learning_rate": 1.6013871027785195e-05, + "loss": 1.165, + "step": 13131 + }, + { + "epoch": 0.32, + "learning_rate": 1.601324870579765e-05, + "loss": 1.2831, + "step": 13132 + }, + { + "epoch": 0.32, + "learning_rate": 1.6012626347329137e-05, + "loss": 1.1826, + "step": 13133 + }, + { + "epoch": 0.32, + "learning_rate": 1.6012003952383436e-05, + "loss": 1.3255, + "step": 13134 + }, + { + "epoch": 0.32, + "learning_rate": 1.6011381520964316e-05, + "loss": 1.2287, + "step": 13135 + }, + { + "epoch": 0.32, + "learning_rate": 1.6010759053075563e-05, + "loss": 0.9994, + "step": 13136 + }, + { + "epoch": 0.32, + "learning_rate": 1.601013654872095e-05, + "loss": 0.9353, + "step": 13137 + }, + { + "epoch": 0.32, + "learning_rate": 1.6009514007904252e-05, + "loss": 1.3896, + "step": 13138 + }, + { + "epoch": 0.32, + "learning_rate": 1.6008891430629244e-05, + "loss": 0.8955, + "step": 13139 + }, + { + "epoch": 0.32, + "learning_rate": 1.6008268816899706e-05, + "loss": 1.3119, + "step": 13140 + }, + { + "epoch": 0.32, + "learning_rate": 1.6007646166719416e-05, + "loss": 1.1115, + "step": 13141 + }, + { + "epoch": 0.32, + "learning_rate": 1.6007023480092146e-05, + "loss": 1.0161, + "step": 13142 + }, + { + "epoch": 0.32, + "learning_rate": 1.600640075702168e-05, + "loss": 0.9407, + "step": 13143 + }, + { + "epoch": 0.32, + "learning_rate": 1.600577799751179e-05, + "loss": 1.1259, + "step": 13144 + }, + { + "epoch": 0.32, + "learning_rate": 1.600515520156626e-05, + "loss": 1.2253, + "step": 13145 + }, + { + "epoch": 0.32, + "learning_rate": 1.6004532369188868e-05, + "loss": 1.2035, + "step": 13146 + }, + { + "epoch": 0.32, + "learning_rate": 1.6003909500383388e-05, + "loss": 0.9862, + "step": 13147 + }, + { + "epoch": 0.32, + "learning_rate": 1.6003286595153604e-05, + "loss": 1.0823, + "step": 13148 + }, + { + "epoch": 0.32, + "learning_rate": 1.6002663653503287e-05, + "loss": 0.9242, + "step": 13149 + }, + { + "epoch": 0.32, + "learning_rate": 1.6002040675436225e-05, + "loss": 0.8475, + "step": 13150 + }, + { + "epoch": 0.32, + "learning_rate": 1.6001417660956193e-05, + "loss": 1.1158, + "step": 13151 + }, + { + "epoch": 0.32, + "learning_rate": 1.6000794610066974e-05, + "loss": 1.1659, + "step": 13152 + }, + { + "epoch": 0.32, + "learning_rate": 1.6000171522772344e-05, + "loss": 1.2191, + "step": 13153 + }, + { + "epoch": 0.32, + "learning_rate": 1.5999548399076083e-05, + "loss": 1.1673, + "step": 13154 + }, + { + "epoch": 0.32, + "learning_rate": 1.5998925238981976e-05, + "loss": 0.9642, + "step": 13155 + }, + { + "epoch": 0.32, + "learning_rate": 1.59983020424938e-05, + "loss": 1.1684, + "step": 13156 + }, + { + "epoch": 0.32, + "learning_rate": 1.5997678809615335e-05, + "loss": 1.0401, + "step": 13157 + }, + { + "epoch": 0.32, + "learning_rate": 1.5997055540350364e-05, + "loss": 1.1303, + "step": 13158 + }, + { + "epoch": 0.32, + "learning_rate": 1.5996432234702668e-05, + "loss": 1.1415, + "step": 13159 + }, + { + "epoch": 0.32, + "learning_rate": 1.5995808892676028e-05, + "loss": 1.0701, + "step": 13160 + }, + { + "epoch": 0.32, + "learning_rate": 1.5995185514274226e-05, + "loss": 1.0463, + "step": 13161 + }, + { + "epoch": 0.32, + "learning_rate": 1.5994562099501042e-05, + "loss": 1.288, + "step": 13162 + }, + { + "epoch": 0.32, + "learning_rate": 1.599393864836026e-05, + "loss": 1.2026, + "step": 13163 + }, + { + "epoch": 0.32, + "learning_rate": 1.5993315160855662e-05, + "loss": 1.1237, + "step": 13164 + }, + { + "epoch": 0.32, + "learning_rate": 1.5992691636991032e-05, + "loss": 1.3147, + "step": 13165 + }, + { + "epoch": 0.32, + "learning_rate": 1.599206807677015e-05, + "loss": 1.1548, + "step": 13166 + }, + { + "epoch": 0.32, + "learning_rate": 1.5991444480196806e-05, + "loss": 0.9596, + "step": 13167 + }, + { + "epoch": 0.32, + "learning_rate": 1.599082084727477e-05, + "loss": 1.0804, + "step": 13168 + }, + { + "epoch": 0.32, + "learning_rate": 1.599019717800784e-05, + "loss": 0.9438, + "step": 13169 + }, + { + "epoch": 0.32, + "learning_rate": 1.5989573472399785e-05, + "loss": 1.0913, + "step": 13170 + }, + { + "epoch": 0.32, + "learning_rate": 1.5988949730454403e-05, + "loss": 1.2151, + "step": 13171 + }, + { + "epoch": 0.32, + "learning_rate": 1.5988325952175468e-05, + "loss": 0.9574, + "step": 13172 + }, + { + "epoch": 0.32, + "learning_rate": 1.5987702137566768e-05, + "loss": 1.1187, + "step": 13173 + }, + { + "epoch": 0.32, + "learning_rate": 1.5987078286632088e-05, + "loss": 1.2913, + "step": 13174 + }, + { + "epoch": 0.32, + "learning_rate": 1.5986454399375216e-05, + "loss": 1.1041, + "step": 13175 + }, + { + "epoch": 0.32, + "learning_rate": 1.5985830475799925e-05, + "loss": 1.0631, + "step": 13176 + }, + { + "epoch": 0.32, + "learning_rate": 1.5985206515910018e-05, + "loss": 1.1047, + "step": 13177 + }, + { + "epoch": 0.32, + "learning_rate": 1.5984582519709266e-05, + "loss": 1.0512, + "step": 13178 + }, + { + "epoch": 0.32, + "learning_rate": 1.598395848720146e-05, + "loss": 1.0361, + "step": 13179 + }, + { + "epoch": 0.32, + "learning_rate": 1.5983334418390382e-05, + "loss": 1.1124, + "step": 13180 + }, + { + "epoch": 0.32, + "learning_rate": 1.5982710313279824e-05, + "loss": 0.805, + "step": 13181 + }, + { + "epoch": 0.32, + "learning_rate": 1.5982086171873575e-05, + "loss": 1.0851, + "step": 13182 + }, + { + "epoch": 0.32, + "learning_rate": 1.598146199417541e-05, + "loss": 0.9434, + "step": 13183 + }, + { + "epoch": 0.32, + "learning_rate": 1.5980837780189124e-05, + "loss": 1.2763, + "step": 13184 + }, + { + "epoch": 0.32, + "learning_rate": 1.59802135299185e-05, + "loss": 1.0277, + "step": 13185 + }, + { + "epoch": 0.32, + "learning_rate": 1.597958924336733e-05, + "loss": 1.0804, + "step": 13186 + }, + { + "epoch": 0.32, + "learning_rate": 1.59789649205394e-05, + "loss": 1.3872, + "step": 13187 + }, + { + "epoch": 0.32, + "learning_rate": 1.597834056143849e-05, + "loss": 1.1314, + "step": 13188 + }, + { + "epoch": 0.32, + "learning_rate": 1.59777161660684e-05, + "loss": 1.1552, + "step": 13189 + }, + { + "epoch": 0.32, + "learning_rate": 1.597709173443291e-05, + "loss": 1.012, + "step": 13190 + }, + { + "epoch": 0.32, + "learning_rate": 1.597646726653581e-05, + "loss": 0.9588, + "step": 13191 + }, + { + "epoch": 0.32, + "learning_rate": 1.597584276238089e-05, + "loss": 1.2697, + "step": 13192 + }, + { + "epoch": 0.32, + "learning_rate": 1.5975218221971935e-05, + "loss": 1.102, + "step": 13193 + }, + { + "epoch": 0.32, + "learning_rate": 1.5974593645312737e-05, + "loss": 1.2815, + "step": 13194 + }, + { + "epoch": 0.32, + "learning_rate": 1.5973969032407086e-05, + "loss": 1.1001, + "step": 13195 + }, + { + "epoch": 0.32, + "learning_rate": 1.597334438325877e-05, + "loss": 1.0422, + "step": 13196 + }, + { + "epoch": 0.32, + "learning_rate": 1.5972719697871576e-05, + "loss": 1.104, + "step": 13197 + }, + { + "epoch": 0.32, + "learning_rate": 1.59720949762493e-05, + "loss": 0.9597, + "step": 13198 + }, + { + "epoch": 0.32, + "learning_rate": 1.5971470218395725e-05, + "loss": 1.1432, + "step": 13199 + }, + { + "epoch": 0.32, + "learning_rate": 1.597084542431465e-05, + "loss": 0.9292, + "step": 13200 + }, + { + "epoch": 0.32, + "learning_rate": 1.5970220594009857e-05, + "loss": 1.0768, + "step": 13201 + }, + { + "epoch": 0.32, + "learning_rate": 1.596959572748514e-05, + "loss": 1.3316, + "step": 13202 + }, + { + "epoch": 0.32, + "learning_rate": 1.596897082474429e-05, + "loss": 1.4401, + "step": 13203 + }, + { + "epoch": 0.32, + "learning_rate": 1.59683458857911e-05, + "loss": 1.1001, + "step": 13204 + }, + { + "epoch": 0.32, + "learning_rate": 1.5967720910629358e-05, + "loss": 1.0228, + "step": 13205 + }, + { + "epoch": 0.32, + "learning_rate": 1.5967095899262853e-05, + "loss": 1.147, + "step": 13206 + }, + { + "epoch": 0.32, + "learning_rate": 1.5966470851695384e-05, + "loss": 1.0371, + "step": 13207 + }, + { + "epoch": 0.32, + "learning_rate": 1.5965845767930738e-05, + "loss": 1.1619, + "step": 13208 + }, + { + "epoch": 0.32, + "learning_rate": 1.5965220647972712e-05, + "loss": 1.2055, + "step": 13209 + }, + { + "epoch": 0.32, + "learning_rate": 1.5964595491825093e-05, + "loss": 1.1746, + "step": 13210 + }, + { + "epoch": 0.32, + "learning_rate": 1.5963970299491676e-05, + "loss": 1.3248, + "step": 13211 + }, + { + "epoch": 0.32, + "learning_rate": 1.596334507097626e-05, + "loss": 1.0904, + "step": 13212 + }, + { + "epoch": 0.32, + "learning_rate": 1.5962719806282627e-05, + "loss": 1.3487, + "step": 13213 + }, + { + "epoch": 0.32, + "learning_rate": 1.5962094505414573e-05, + "loss": 0.9524, + "step": 13214 + }, + { + "epoch": 0.32, + "learning_rate": 1.5961469168375898e-05, + "loss": 1.3208, + "step": 13215 + }, + { + "epoch": 0.32, + "learning_rate": 1.596084379517039e-05, + "loss": 0.8725, + "step": 13216 + }, + { + "epoch": 0.32, + "learning_rate": 1.5960218385801843e-05, + "loss": 0.748, + "step": 13217 + }, + { + "epoch": 0.32, + "learning_rate": 1.5959592940274057e-05, + "loss": 0.8028, + "step": 13218 + }, + { + "epoch": 0.32, + "learning_rate": 1.5958967458590824e-05, + "loss": 1.2188, + "step": 13219 + }, + { + "epoch": 0.32, + "learning_rate": 1.595834194075593e-05, + "loss": 1.107, + "step": 13220 + }, + { + "epoch": 0.32, + "learning_rate": 1.5957716386773182e-05, + "loss": 1.2742, + "step": 13221 + }, + { + "epoch": 0.32, + "learning_rate": 1.5957090796646372e-05, + "loss": 1.1994, + "step": 13222 + }, + { + "epoch": 0.32, + "learning_rate": 1.5956465170379288e-05, + "loss": 1.2773, + "step": 13223 + }, + { + "epoch": 0.32, + "learning_rate": 1.5955839507975732e-05, + "loss": 1.2004, + "step": 13224 + }, + { + "epoch": 0.32, + "learning_rate": 1.59552138094395e-05, + "loss": 1.3778, + "step": 13225 + }, + { + "epoch": 0.32, + "learning_rate": 1.5954588074774387e-05, + "loss": 1.0208, + "step": 13226 + }, + { + "epoch": 0.32, + "learning_rate": 1.5953962303984186e-05, + "loss": 1.2054, + "step": 13227 + }, + { + "epoch": 0.32, + "learning_rate": 1.59533364970727e-05, + "loss": 0.9464, + "step": 13228 + }, + { + "epoch": 0.32, + "learning_rate": 1.595271065404372e-05, + "loss": 1.0864, + "step": 13229 + }, + { + "epoch": 0.32, + "learning_rate": 1.5952084774901043e-05, + "loss": 1.1612, + "step": 13230 + }, + { + "epoch": 0.32, + "learning_rate": 1.595145885964847e-05, + "loss": 1.0484, + "step": 13231 + }, + { + "epoch": 0.32, + "learning_rate": 1.5950832908289795e-05, + "loss": 1.2317, + "step": 13232 + }, + { + "epoch": 0.32, + "learning_rate": 1.5950206920828817e-05, + "loss": 0.9705, + "step": 13233 + }, + { + "epoch": 0.32, + "learning_rate": 1.5949580897269332e-05, + "loss": 1.0064, + "step": 13234 + }, + { + "epoch": 0.32, + "learning_rate": 1.5948954837615144e-05, + "loss": 1.0186, + "step": 13235 + }, + { + "epoch": 0.32, + "learning_rate": 1.594832874187004e-05, + "loss": 1.2409, + "step": 13236 + }, + { + "epoch": 0.32, + "learning_rate": 1.5947702610037827e-05, + "loss": 1.0601, + "step": 13237 + }, + { + "epoch": 0.32, + "learning_rate": 1.59470764421223e-05, + "loss": 0.9082, + "step": 13238 + }, + { + "epoch": 0.32, + "learning_rate": 1.594645023812726e-05, + "loss": 1.142, + "step": 13239 + }, + { + "epoch": 0.32, + "learning_rate": 1.5945823998056505e-05, + "loss": 1.2453, + "step": 13240 + }, + { + "epoch": 0.32, + "learning_rate": 1.5945197721913834e-05, + "loss": 1.0978, + "step": 13241 + }, + { + "epoch": 0.32, + "learning_rate": 1.594457140970305e-05, + "loss": 1.4482, + "step": 13242 + }, + { + "epoch": 0.32, + "learning_rate": 1.5943945061427945e-05, + "loss": 1.0834, + "step": 13243 + }, + { + "epoch": 0.32, + "learning_rate": 1.5943318677092327e-05, + "loss": 0.8499, + "step": 13244 + }, + { + "epoch": 0.32, + "learning_rate": 1.594269225669999e-05, + "loss": 0.9505, + "step": 13245 + }, + { + "epoch": 0.32, + "learning_rate": 1.5942065800254738e-05, + "loss": 1.1772, + "step": 13246 + }, + { + "epoch": 0.32, + "learning_rate": 1.594143930776037e-05, + "loss": 0.9697, + "step": 13247 + }, + { + "epoch": 0.32, + "learning_rate": 1.5940812779220687e-05, + "loss": 0.9978, + "step": 13248 + }, + { + "epoch": 0.32, + "learning_rate": 1.594018621463949e-05, + "loss": 1.1076, + "step": 13249 + }, + { + "epoch": 0.32, + "learning_rate": 1.5939559614020584e-05, + "loss": 1.2388, + "step": 13250 + }, + { + "epoch": 0.32, + "learning_rate": 1.5938932977367762e-05, + "loss": 1.0246, + "step": 13251 + }, + { + "epoch": 0.32, + "learning_rate": 1.5938306304684835e-05, + "loss": 1.0344, + "step": 13252 + }, + { + "epoch": 0.32, + "learning_rate": 1.5937679595975594e-05, + "loss": 1.1414, + "step": 13253 + }, + { + "epoch": 0.32, + "learning_rate": 1.5937052851243855e-05, + "loss": 1.1091, + "step": 13254 + }, + { + "epoch": 0.32, + "learning_rate": 1.5936426070493406e-05, + "loss": 1.1675, + "step": 13255 + }, + { + "epoch": 0.32, + "learning_rate": 1.593579925372806e-05, + "loss": 1.0819, + "step": 13256 + }, + { + "epoch": 0.32, + "learning_rate": 1.5935172400951617e-05, + "loss": 1.1413, + "step": 13257 + }, + { + "epoch": 0.32, + "learning_rate": 1.5934545512167878e-05, + "loss": 1.0245, + "step": 13258 + }, + { + "epoch": 0.32, + "learning_rate": 1.5933918587380648e-05, + "loss": 1.256, + "step": 13259 + }, + { + "epoch": 0.32, + "learning_rate": 1.593329162659373e-05, + "loss": 1.2714, + "step": 13260 + }, + { + "epoch": 0.32, + "learning_rate": 1.5932664629810927e-05, + "loss": 1.0753, + "step": 13261 + }, + { + "epoch": 0.32, + "learning_rate": 1.5932037597036042e-05, + "loss": 1.3127, + "step": 13262 + }, + { + "epoch": 0.32, + "learning_rate": 1.5931410528272876e-05, + "loss": 1.017, + "step": 13263 + }, + { + "epoch": 0.32, + "learning_rate": 1.5930783423525246e-05, + "loss": 1.2147, + "step": 13264 + }, + { + "epoch": 0.32, + "learning_rate": 1.5930156282796944e-05, + "loss": 1.2845, + "step": 13265 + }, + { + "epoch": 0.32, + "learning_rate": 1.5929529106091778e-05, + "loss": 0.9802, + "step": 13266 + }, + { + "epoch": 0.32, + "learning_rate": 1.5928901893413553e-05, + "loss": 0.8389, + "step": 13267 + }, + { + "epoch": 0.32, + "learning_rate": 1.5928274644766077e-05, + "loss": 0.9525, + "step": 13268 + }, + { + "epoch": 0.32, + "learning_rate": 1.5927647360153152e-05, + "loss": 0.9116, + "step": 13269 + }, + { + "epoch": 0.32, + "learning_rate": 1.5927020039578587e-05, + "loss": 1.1328, + "step": 13270 + }, + { + "epoch": 0.32, + "learning_rate": 1.592639268304618e-05, + "loss": 0.8701, + "step": 13271 + }, + { + "epoch": 0.32, + "learning_rate": 1.592576529055975e-05, + "loss": 0.9602, + "step": 13272 + }, + { + "epoch": 0.32, + "learning_rate": 1.592513786212309e-05, + "loss": 1.0863, + "step": 13273 + }, + { + "epoch": 0.32, + "learning_rate": 1.5924510397740015e-05, + "loss": 1.2046, + "step": 13274 + }, + { + "epoch": 0.32, + "learning_rate": 1.5923882897414327e-05, + "loss": 0.9238, + "step": 13275 + }, + { + "epoch": 0.32, + "learning_rate": 1.5923255361149835e-05, + "loss": 0.9651, + "step": 13276 + }, + { + "epoch": 0.32, + "learning_rate": 1.5922627788950346e-05, + "loss": 1.0221, + "step": 13277 + }, + { + "epoch": 0.32, + "learning_rate": 1.592200018081967e-05, + "loss": 1.0864, + "step": 13278 + }, + { + "epoch": 0.32, + "learning_rate": 1.5921372536761607e-05, + "loss": 1.0408, + "step": 13279 + }, + { + "epoch": 0.32, + "learning_rate": 1.5920744856779974e-05, + "loss": 1.1254, + "step": 13280 + }, + { + "epoch": 0.32, + "learning_rate": 1.5920117140878574e-05, + "loss": 0.898, + "step": 13281 + }, + { + "epoch": 0.32, + "learning_rate": 1.5919489389061215e-05, + "loss": 1.2673, + "step": 13282 + }, + { + "epoch": 0.32, + "learning_rate": 1.5918861601331706e-05, + "loss": 0.8172, + "step": 13283 + }, + { + "epoch": 0.32, + "learning_rate": 1.5918233777693858e-05, + "loss": 1.2718, + "step": 13284 + }, + { + "epoch": 0.32, + "learning_rate": 1.5917605918151475e-05, + "loss": 1.1535, + "step": 13285 + }, + { + "epoch": 0.32, + "learning_rate": 1.591697802270837e-05, + "loss": 1.1819, + "step": 13286 + }, + { + "epoch": 0.32, + "learning_rate": 1.5916350091368352e-05, + "loss": 0.7207, + "step": 13287 + }, + { + "epoch": 0.32, + "learning_rate": 1.5915722124135227e-05, + "loss": 1.0264, + "step": 13288 + }, + { + "epoch": 0.32, + "learning_rate": 1.5915094121012805e-05, + "loss": 1.1484, + "step": 13289 + }, + { + "epoch": 0.32, + "learning_rate": 1.5914466082004904e-05, + "loss": 1.1629, + "step": 13290 + }, + { + "epoch": 0.32, + "learning_rate": 1.5913838007115325e-05, + "loss": 1.2512, + "step": 13291 + }, + { + "epoch": 0.32, + "learning_rate": 1.5913209896347886e-05, + "loss": 0.7196, + "step": 13292 + }, + { + "epoch": 0.32, + "learning_rate": 1.5912581749706387e-05, + "loss": 1.0574, + "step": 13293 + }, + { + "epoch": 0.32, + "learning_rate": 1.591195356719465e-05, + "loss": 1.2931, + "step": 13294 + }, + { + "epoch": 0.32, + "learning_rate": 1.591132534881648e-05, + "loss": 1.3551, + "step": 13295 + }, + { + "epoch": 0.32, + "learning_rate": 1.5910697094575693e-05, + "loss": 1.1917, + "step": 13296 + }, + { + "epoch": 0.32, + "learning_rate": 1.5910068804476092e-05, + "loss": 1.0262, + "step": 13297 + }, + { + "epoch": 0.32, + "learning_rate": 1.5909440478521495e-05, + "loss": 1.026, + "step": 13298 + }, + { + "epoch": 0.32, + "learning_rate": 1.5908812116715714e-05, + "loss": 0.9722, + "step": 13299 + }, + { + "epoch": 0.32, + "learning_rate": 1.5908183719062557e-05, + "loss": 1.008, + "step": 13300 + }, + { + "epoch": 0.32, + "learning_rate": 1.5907555285565842e-05, + "loss": 1.1616, + "step": 13301 + }, + { + "epoch": 0.32, + "learning_rate": 1.5906926816229374e-05, + "loss": 1.0443, + "step": 13302 + }, + { + "epoch": 0.32, + "learning_rate": 1.5906298311056975e-05, + "loss": 1.2948, + "step": 13303 + }, + { + "epoch": 0.32, + "learning_rate": 1.5905669770052453e-05, + "loss": 1.0523, + "step": 13304 + }, + { + "epoch": 0.32, + "learning_rate": 1.5905041193219623e-05, + "loss": 1.4591, + "step": 13305 + }, + { + "epoch": 0.32, + "learning_rate": 1.5904412580562294e-05, + "loss": 1.0365, + "step": 13306 + }, + { + "epoch": 0.32, + "learning_rate": 1.5903783932084283e-05, + "loss": 0.9783, + "step": 13307 + }, + { + "epoch": 0.32, + "learning_rate": 1.5903155247789403e-05, + "loss": 1.1095, + "step": 13308 + }, + { + "epoch": 0.32, + "learning_rate": 1.5902526527681472e-05, + "loss": 1.0322, + "step": 13309 + }, + { + "epoch": 0.32, + "learning_rate": 1.5901897771764298e-05, + "loss": 1.1372, + "step": 13310 + }, + { + "epoch": 0.32, + "learning_rate": 1.59012689800417e-05, + "loss": 1.0097, + "step": 13311 + }, + { + "epoch": 0.32, + "learning_rate": 1.590064015251749e-05, + "loss": 1.2187, + "step": 13312 + }, + { + "epoch": 0.32, + "learning_rate": 1.5900011289195483e-05, + "loss": 1.1187, + "step": 13313 + }, + { + "epoch": 0.32, + "learning_rate": 1.5899382390079495e-05, + "loss": 0.9654, + "step": 13314 + }, + { + "epoch": 0.32, + "learning_rate": 1.5898753455173347e-05, + "loss": 1.3081, + "step": 13315 + }, + { + "epoch": 0.32, + "learning_rate": 1.5898124484480845e-05, + "loss": 1.1648, + "step": 13316 + }, + { + "epoch": 0.32, + "learning_rate": 1.589749547800581e-05, + "loss": 1.1967, + "step": 13317 + }, + { + "epoch": 0.32, + "learning_rate": 1.5896866435752055e-05, + "loss": 1.3961, + "step": 13318 + }, + { + "epoch": 0.32, + "learning_rate": 1.58962373577234e-05, + "loss": 1.2186, + "step": 13319 + }, + { + "epoch": 0.32, + "learning_rate": 1.5895608243923662e-05, + "loss": 1.3379, + "step": 13320 + }, + { + "epoch": 0.32, + "learning_rate": 1.5894979094356653e-05, + "loss": 1.1274, + "step": 13321 + }, + { + "epoch": 0.32, + "learning_rate": 1.589434990902619e-05, + "loss": 1.2231, + "step": 13322 + }, + { + "epoch": 0.32, + "learning_rate": 1.5893720687936095e-05, + "loss": 1.4246, + "step": 13323 + }, + { + "epoch": 0.32, + "learning_rate": 1.5893091431090186e-05, + "loss": 1.1276, + "step": 13324 + }, + { + "epoch": 0.32, + "learning_rate": 1.5892462138492276e-05, + "loss": 1.3822, + "step": 13325 + }, + { + "epoch": 0.32, + "learning_rate": 1.589183281014618e-05, + "loss": 1.0939, + "step": 13326 + }, + { + "epoch": 0.32, + "learning_rate": 1.589120344605572e-05, + "loss": 1.2399, + "step": 13327 + }, + { + "epoch": 0.32, + "learning_rate": 1.589057404622472e-05, + "loss": 1.1935, + "step": 13328 + }, + { + "epoch": 0.32, + "learning_rate": 1.588994461065699e-05, + "loss": 1.0114, + "step": 13329 + }, + { + "epoch": 0.32, + "learning_rate": 1.5889315139356352e-05, + "loss": 1.2111, + "step": 13330 + }, + { + "epoch": 0.32, + "learning_rate": 1.5888685632326627e-05, + "loss": 1.2242, + "step": 13331 + }, + { + "epoch": 0.32, + "learning_rate": 1.5888056089571628e-05, + "loss": 1.0076, + "step": 13332 + }, + { + "epoch": 0.32, + "learning_rate": 1.5887426511095175e-05, + "loss": 0.8525, + "step": 13333 + }, + { + "epoch": 0.32, + "learning_rate": 1.5886796896901095e-05, + "loss": 1.0741, + "step": 13334 + }, + { + "epoch": 0.32, + "learning_rate": 1.58861672469932e-05, + "loss": 1.3134, + "step": 13335 + }, + { + "epoch": 0.32, + "learning_rate": 1.5885537561375314e-05, + "loss": 1.0363, + "step": 13336 + }, + { + "epoch": 0.32, + "learning_rate": 1.5884907840051255e-05, + "loss": 1.2365, + "step": 13337 + }, + { + "epoch": 0.32, + "learning_rate": 1.5884278083024848e-05, + "loss": 0.9987, + "step": 13338 + }, + { + "epoch": 0.32, + "learning_rate": 1.5883648290299905e-05, + "loss": 1.2718, + "step": 13339 + }, + { + "epoch": 0.32, + "learning_rate": 1.5883018461880255e-05, + "loss": 1.0932, + "step": 13340 + }, + { + "epoch": 0.32, + "learning_rate": 1.5882388597769717e-05, + "loss": 1.4027, + "step": 13341 + }, + { + "epoch": 0.32, + "learning_rate": 1.5881758697972106e-05, + "loss": 1.0335, + "step": 13342 + }, + { + "epoch": 0.32, + "learning_rate": 1.5881128762491252e-05, + "loss": 0.922, + "step": 13343 + }, + { + "epoch": 0.32, + "learning_rate": 1.5880498791330972e-05, + "loss": 1.1749, + "step": 13344 + }, + { + "epoch": 0.32, + "learning_rate": 1.5879868784495088e-05, + "loss": 1.2149, + "step": 13345 + }, + { + "epoch": 0.32, + "learning_rate": 1.5879238741987424e-05, + "loss": 1.2125, + "step": 13346 + }, + { + "epoch": 0.32, + "learning_rate": 1.5878608663811802e-05, + "loss": 1.1694, + "step": 13347 + }, + { + "epoch": 0.32, + "learning_rate": 1.5877978549972044e-05, + "loss": 1.3013, + "step": 13348 + }, + { + "epoch": 0.32, + "learning_rate": 1.5877348400471976e-05, + "loss": 1.2888, + "step": 13349 + }, + { + "epoch": 0.32, + "learning_rate": 1.587671821531541e-05, + "loss": 1.2973, + "step": 13350 + }, + { + "epoch": 0.32, + "learning_rate": 1.587608799450618e-05, + "loss": 1.0289, + "step": 13351 + }, + { + "epoch": 0.32, + "learning_rate": 1.587545773804811e-05, + "loss": 1.1487, + "step": 13352 + }, + { + "epoch": 0.32, + "learning_rate": 1.5874827445945018e-05, + "loss": 1.2872, + "step": 13353 + }, + { + "epoch": 0.32, + "learning_rate": 1.587419711820073e-05, + "loss": 1.0298, + "step": 13354 + }, + { + "epoch": 0.32, + "learning_rate": 1.5873566754819066e-05, + "loss": 1.182, + "step": 13355 + }, + { + "epoch": 0.32, + "learning_rate": 1.5872936355803857e-05, + "loss": 1.0677, + "step": 13356 + }, + { + "epoch": 0.32, + "learning_rate": 1.5872305921158926e-05, + "loss": 0.9373, + "step": 13357 + }, + { + "epoch": 0.32, + "learning_rate": 1.5871675450888093e-05, + "loss": 1.2022, + "step": 13358 + }, + { + "epoch": 0.32, + "learning_rate": 1.587104494499519e-05, + "loss": 0.8122, + "step": 13359 + }, + { + "epoch": 0.32, + "learning_rate": 1.5870414403484034e-05, + "loss": 0.9992, + "step": 13360 + }, + { + "epoch": 0.32, + "learning_rate": 1.5869783826358457e-05, + "loss": 1.0698, + "step": 13361 + }, + { + "epoch": 0.32, + "learning_rate": 1.586915321362228e-05, + "loss": 1.1495, + "step": 13362 + }, + { + "epoch": 0.32, + "learning_rate": 1.5868522565279336e-05, + "loss": 1.2163, + "step": 13363 + }, + { + "epoch": 0.32, + "learning_rate": 1.586789188133344e-05, + "loss": 1.0648, + "step": 13364 + }, + { + "epoch": 0.32, + "learning_rate": 1.586726116178843e-05, + "loss": 1.1463, + "step": 13365 + }, + { + "epoch": 0.32, + "learning_rate": 1.586663040664812e-05, + "loss": 1.0722, + "step": 13366 + }, + { + "epoch": 0.32, + "learning_rate": 1.586599961591635e-05, + "loss": 1.1053, + "step": 13367 + }, + { + "epoch": 0.32, + "learning_rate": 1.5865368789596932e-05, + "loss": 1.2204, + "step": 13368 + }, + { + "epoch": 0.32, + "learning_rate": 1.5864737927693708e-05, + "loss": 1.1003, + "step": 13369 + }, + { + "epoch": 0.32, + "learning_rate": 1.58641070302105e-05, + "loss": 1.2435, + "step": 13370 + }, + { + "epoch": 0.32, + "learning_rate": 1.5863476097151127e-05, + "loss": 1.0928, + "step": 13371 + }, + { + "epoch": 0.32, + "learning_rate": 1.5862845128519428e-05, + "loss": 1.0257, + "step": 13372 + }, + { + "epoch": 0.32, + "learning_rate": 1.5862214124319226e-05, + "loss": 1.3051, + "step": 13373 + }, + { + "epoch": 0.32, + "learning_rate": 1.586158308455435e-05, + "loss": 1.1529, + "step": 13374 + }, + { + "epoch": 0.32, + "learning_rate": 1.5860952009228625e-05, + "loss": 1.0732, + "step": 13375 + }, + { + "epoch": 0.32, + "learning_rate": 1.5860320898345887e-05, + "loss": 0.9864, + "step": 13376 + }, + { + "epoch": 0.32, + "learning_rate": 1.5859689751909958e-05, + "loss": 0.9691, + "step": 13377 + }, + { + "epoch": 0.32, + "learning_rate": 1.585905856992467e-05, + "loss": 1.401, + "step": 13378 + }, + { + "epoch": 0.32, + "learning_rate": 1.5858427352393852e-05, + "loss": 1.2044, + "step": 13379 + }, + { + "epoch": 0.32, + "learning_rate": 1.5857796099321333e-05, + "loss": 0.9406, + "step": 13380 + }, + { + "epoch": 0.32, + "learning_rate": 1.5857164810710944e-05, + "loss": 1.0896, + "step": 13381 + }, + { + "epoch": 0.32, + "learning_rate": 1.5856533486566513e-05, + "loss": 1.129, + "step": 13382 + }, + { + "epoch": 0.32, + "learning_rate": 1.5855902126891872e-05, + "loss": 1.0518, + "step": 13383 + }, + { + "epoch": 0.32, + "learning_rate": 1.585527073169085e-05, + "loss": 1.2872, + "step": 13384 + }, + { + "epoch": 0.32, + "learning_rate": 1.5854639300967275e-05, + "loss": 1.1281, + "step": 13385 + }, + { + "epoch": 0.32, + "learning_rate": 1.5854007834724986e-05, + "loss": 1.13, + "step": 13386 + }, + { + "epoch": 0.32, + "learning_rate": 1.5853376332967802e-05, + "loss": 1.1283, + "step": 13387 + }, + { + "epoch": 0.32, + "learning_rate": 1.5852744795699564e-05, + "loss": 0.8619, + "step": 13388 + }, + { + "epoch": 0.32, + "learning_rate": 1.58521132229241e-05, + "loss": 1.0073, + "step": 13389 + }, + { + "epoch": 0.32, + "learning_rate": 1.585148161464524e-05, + "loss": 1.3169, + "step": 13390 + }, + { + "epoch": 0.32, + "learning_rate": 1.585084997086682e-05, + "loss": 1.1893, + "step": 13391 + }, + { + "epoch": 0.32, + "learning_rate": 1.5850218291592666e-05, + "loss": 1.171, + "step": 13392 + }, + { + "epoch": 0.32, + "learning_rate": 1.5849586576826618e-05, + "loss": 1.0572, + "step": 13393 + }, + { + "epoch": 0.32, + "learning_rate": 1.5848954826572498e-05, + "loss": 1.1094, + "step": 13394 + }, + { + "epoch": 0.32, + "learning_rate": 1.584832304083415e-05, + "loss": 1.0378, + "step": 13395 + }, + { + "epoch": 0.32, + "learning_rate": 1.58476912196154e-05, + "loss": 1.0777, + "step": 13396 + }, + { + "epoch": 0.32, + "learning_rate": 1.5847059362920086e-05, + "loss": 1.2339, + "step": 13397 + }, + { + "epoch": 0.32, + "learning_rate": 1.5846427470752033e-05, + "loss": 1.2445, + "step": 13398 + }, + { + "epoch": 0.32, + "learning_rate": 1.5845795543115083e-05, + "loss": 0.8247, + "step": 13399 + }, + { + "epoch": 0.32, + "learning_rate": 1.5845163580013062e-05, + "loss": 1.0641, + "step": 13400 + }, + { + "epoch": 0.32, + "learning_rate": 1.5844531581449814e-05, + "loss": 1.0377, + "step": 13401 + }, + { + "epoch": 0.32, + "learning_rate": 1.5843899547429164e-05, + "loss": 1.1344, + "step": 13402 + }, + { + "epoch": 0.32, + "learning_rate": 1.5843267477954953e-05, + "loss": 0.9993, + "step": 13403 + }, + { + "epoch": 0.32, + "learning_rate": 1.584263537303101e-05, + "loss": 1.3812, + "step": 13404 + }, + { + "epoch": 0.32, + "learning_rate": 1.5842003232661174e-05, + "loss": 1.127, + "step": 13405 + }, + { + "epoch": 0.32, + "learning_rate": 1.5841371056849278e-05, + "loss": 1.518, + "step": 13406 + }, + { + "epoch": 0.32, + "learning_rate": 1.5840738845599157e-05, + "loss": 1.0703, + "step": 13407 + }, + { + "epoch": 0.32, + "learning_rate": 1.584010659891465e-05, + "loss": 1.2886, + "step": 13408 + }, + { + "epoch": 0.32, + "learning_rate": 1.5839474316799587e-05, + "loss": 1.1027, + "step": 13409 + }, + { + "epoch": 0.32, + "learning_rate": 1.583884199925781e-05, + "loss": 1.3678, + "step": 13410 + }, + { + "epoch": 0.32, + "learning_rate": 1.5838209646293147e-05, + "loss": 1.0697, + "step": 13411 + }, + { + "epoch": 0.32, + "learning_rate": 1.5837577257909443e-05, + "loss": 1.1998, + "step": 13412 + }, + { + "epoch": 0.32, + "learning_rate": 1.583694483411053e-05, + "loss": 1.0894, + "step": 13413 + }, + { + "epoch": 0.32, + "learning_rate": 1.5836312374900244e-05, + "loss": 1.207, + "step": 13414 + }, + { + "epoch": 0.32, + "learning_rate": 1.5835679880282427e-05, + "loss": 1.1608, + "step": 13415 + }, + { + "epoch": 0.32, + "learning_rate": 1.583504735026091e-05, + "loss": 1.0145, + "step": 13416 + }, + { + "epoch": 0.32, + "learning_rate": 1.5834414784839533e-05, + "loss": 1.1088, + "step": 13417 + }, + { + "epoch": 0.32, + "learning_rate": 1.5833782184022134e-05, + "loss": 1.1743, + "step": 13418 + }, + { + "epoch": 0.32, + "learning_rate": 1.5833149547812555e-05, + "loss": 1.2035, + "step": 13419 + }, + { + "epoch": 0.32, + "learning_rate": 1.5832516876214627e-05, + "loss": 1.1699, + "step": 13420 + }, + { + "epoch": 0.32, + "learning_rate": 1.583188416923219e-05, + "loss": 1.0373, + "step": 13421 + }, + { + "epoch": 0.32, + "learning_rate": 1.5831251426869082e-05, + "loss": 1.0072, + "step": 13422 + }, + { + "epoch": 0.32, + "learning_rate": 1.5830618649129147e-05, + "loss": 1.2735, + "step": 13423 + }, + { + "epoch": 0.32, + "learning_rate": 1.5829985836016218e-05, + "loss": 0.8777, + "step": 13424 + }, + { + "epoch": 0.32, + "learning_rate": 1.5829352987534138e-05, + "loss": 1.2993, + "step": 13425 + }, + { + "epoch": 0.32, + "learning_rate": 1.582872010368674e-05, + "loss": 1.2411, + "step": 13426 + }, + { + "epoch": 0.32, + "learning_rate": 1.582808718447787e-05, + "loss": 0.8163, + "step": 13427 + }, + { + "epoch": 0.32, + "learning_rate": 1.582745422991137e-05, + "loss": 1.0423, + "step": 13428 + }, + { + "epoch": 0.32, + "learning_rate": 1.582682123999107e-05, + "loss": 1.2715, + "step": 13429 + }, + { + "epoch": 0.32, + "learning_rate": 1.582618821472082e-05, + "loss": 1.0551, + "step": 13430 + }, + { + "epoch": 0.32, + "learning_rate": 1.5825555154104458e-05, + "loss": 1.2854, + "step": 13431 + }, + { + "epoch": 0.32, + "learning_rate": 1.582492205814582e-05, + "loss": 1.1872, + "step": 13432 + }, + { + "epoch": 0.32, + "learning_rate": 1.582428892684875e-05, + "loss": 1.011, + "step": 13433 + }, + { + "epoch": 0.32, + "learning_rate": 1.582365576021709e-05, + "loss": 1.2394, + "step": 13434 + }, + { + "epoch": 0.32, + "learning_rate": 1.582302255825468e-05, + "loss": 1.1819, + "step": 13435 + }, + { + "epoch": 0.32, + "learning_rate": 1.5822389320965364e-05, + "loss": 1.166, + "step": 13436 + }, + { + "epoch": 0.32, + "learning_rate": 1.5821756048352975e-05, + "loss": 0.9241, + "step": 13437 + }, + { + "epoch": 0.32, + "learning_rate": 1.5821122740421366e-05, + "loss": 0.9917, + "step": 13438 + }, + { + "epoch": 0.32, + "learning_rate": 1.5820489397174378e-05, + "loss": 0.8302, + "step": 13439 + }, + { + "epoch": 0.32, + "learning_rate": 1.5819856018615843e-05, + "loss": 0.8707, + "step": 13440 + }, + { + "epoch": 0.32, + "learning_rate": 1.5819222604749614e-05, + "loss": 1.3592, + "step": 13441 + }, + { + "epoch": 0.32, + "learning_rate": 1.581858915557953e-05, + "loss": 1.1374, + "step": 13442 + }, + { + "epoch": 0.32, + "learning_rate": 1.5817955671109432e-05, + "loss": 0.9491, + "step": 13443 + }, + { + "epoch": 0.32, + "learning_rate": 1.581732215134317e-05, + "loss": 0.997, + "step": 13444 + }, + { + "epoch": 0.32, + "learning_rate": 1.581668859628458e-05, + "loss": 1.0811, + "step": 13445 + }, + { + "epoch": 0.32, + "learning_rate": 1.581605500593751e-05, + "loss": 1.2458, + "step": 13446 + }, + { + "epoch": 0.32, + "learning_rate": 1.58154213803058e-05, + "loss": 1.1238, + "step": 13447 + }, + { + "epoch": 0.32, + "learning_rate": 1.5814787719393296e-05, + "loss": 1.0014, + "step": 13448 + }, + { + "epoch": 0.32, + "learning_rate": 1.5814154023203843e-05, + "loss": 1.1016, + "step": 13449 + }, + { + "epoch": 0.32, + "learning_rate": 1.581352029174129e-05, + "loss": 0.9585, + "step": 13450 + }, + { + "epoch": 0.32, + "learning_rate": 1.581288652500947e-05, + "loss": 0.9731, + "step": 13451 + }, + { + "epoch": 0.32, + "learning_rate": 1.581225272301224e-05, + "loss": 1.1228, + "step": 13452 + }, + { + "epoch": 0.32, + "learning_rate": 1.5811618885753438e-05, + "loss": 1.1134, + "step": 13453 + }, + { + "epoch": 0.32, + "learning_rate": 1.581098501323691e-05, + "loss": 1.3596, + "step": 13454 + }, + { + "epoch": 0.32, + "learning_rate": 1.58103511054665e-05, + "loss": 1.0638, + "step": 13455 + }, + { + "epoch": 0.32, + "learning_rate": 1.5809717162446064e-05, + "loss": 1.1891, + "step": 13456 + }, + { + "epoch": 0.32, + "learning_rate": 1.580908318417943e-05, + "loss": 1.2662, + "step": 13457 + }, + { + "epoch": 0.32, + "learning_rate": 1.5808449170670464e-05, + "loss": 1.1721, + "step": 13458 + }, + { + "epoch": 0.32, + "learning_rate": 1.5807815121923002e-05, + "loss": 1.0865, + "step": 13459 + }, + { + "epoch": 0.32, + "learning_rate": 1.580718103794089e-05, + "loss": 1.1242, + "step": 13460 + }, + { + "epoch": 0.32, + "learning_rate": 1.5806546918727975e-05, + "loss": 1.2168, + "step": 13461 + }, + { + "epoch": 0.32, + "learning_rate": 1.5805912764288108e-05, + "loss": 1.049, + "step": 13462 + }, + { + "epoch": 0.32, + "learning_rate": 1.5805278574625135e-05, + "loss": 1.063, + "step": 13463 + }, + { + "epoch": 0.32, + "learning_rate": 1.58046443497429e-05, + "loss": 0.9965, + "step": 13464 + }, + { + "epoch": 0.32, + "learning_rate": 1.580401008964525e-05, + "loss": 1.0822, + "step": 13465 + }, + { + "epoch": 0.32, + "learning_rate": 1.5803375794336043e-05, + "loss": 1.0726, + "step": 13466 + }, + { + "epoch": 0.32, + "learning_rate": 1.5802741463819117e-05, + "loss": 1.2086, + "step": 13467 + }, + { + "epoch": 0.32, + "learning_rate": 1.580210709809832e-05, + "loss": 1.298, + "step": 13468 + }, + { + "epoch": 0.32, + "learning_rate": 1.580147269717751e-05, + "loss": 1.0286, + "step": 13469 + }, + { + "epoch": 0.32, + "learning_rate": 1.5800838261060528e-05, + "loss": 1.0841, + "step": 13470 + }, + { + "epoch": 0.32, + "learning_rate": 1.580020378975122e-05, + "loss": 1.2497, + "step": 13471 + }, + { + "epoch": 0.32, + "learning_rate": 1.5799569283253444e-05, + "loss": 1.2051, + "step": 13472 + }, + { + "epoch": 0.32, + "learning_rate": 1.5798934741571045e-05, + "loss": 1.2827, + "step": 13473 + }, + { + "epoch": 0.32, + "learning_rate": 1.5798300164707875e-05, + "loss": 1.016, + "step": 13474 + }, + { + "epoch": 0.32, + "learning_rate": 1.579766555266778e-05, + "loss": 1.1942, + "step": 13475 + }, + { + "epoch": 0.32, + "learning_rate": 1.579703090545461e-05, + "loss": 1.1682, + "step": 13476 + }, + { + "epoch": 0.32, + "learning_rate": 1.5796396223072218e-05, + "loss": 1.0576, + "step": 13477 + }, + { + "epoch": 0.32, + "learning_rate": 1.579576150552445e-05, + "loss": 1.3853, + "step": 13478 + }, + { + "epoch": 0.32, + "learning_rate": 1.5795126752815167e-05, + "loss": 1.2187, + "step": 13479 + }, + { + "epoch": 0.32, + "learning_rate": 1.579449196494821e-05, + "loss": 1.0859, + "step": 13480 + }, + { + "epoch": 0.32, + "learning_rate": 1.579385714192743e-05, + "loss": 0.9563, + "step": 13481 + }, + { + "epoch": 0.32, + "learning_rate": 1.5793222283756688e-05, + "loss": 1.0609, + "step": 13482 + }, + { + "epoch": 0.32, + "learning_rate": 1.5792587390439823e-05, + "loss": 1.1461, + "step": 13483 + }, + { + "epoch": 0.32, + "learning_rate": 1.5791952461980696e-05, + "loss": 1.1481, + "step": 13484 + }, + { + "epoch": 0.32, + "learning_rate": 1.5791317498383157e-05, + "loss": 1.5529, + "step": 13485 + }, + { + "epoch": 0.32, + "learning_rate": 1.5790682499651056e-05, + "loss": 1.2441, + "step": 13486 + }, + { + "epoch": 0.32, + "learning_rate": 1.5790047465788244e-05, + "loss": 0.9398, + "step": 13487 + }, + { + "epoch": 0.32, + "learning_rate": 1.578941239679858e-05, + "loss": 1.3436, + "step": 13488 + }, + { + "epoch": 0.32, + "learning_rate": 1.578877729268591e-05, + "loss": 1.0009, + "step": 13489 + }, + { + "epoch": 0.32, + "learning_rate": 1.5788142153454086e-05, + "loss": 1.0211, + "step": 13490 + }, + { + "epoch": 0.32, + "learning_rate": 1.5787506979106974e-05, + "loss": 1.0964, + "step": 13491 + }, + { + "epoch": 0.32, + "learning_rate": 1.578687176964841e-05, + "loss": 1.1987, + "step": 13492 + }, + { + "epoch": 0.32, + "learning_rate": 1.5786236525082263e-05, + "loss": 1.0738, + "step": 13493 + }, + { + "epoch": 0.32, + "learning_rate": 1.5785601245412373e-05, + "loss": 1.2008, + "step": 13494 + }, + { + "epoch": 0.32, + "learning_rate": 1.5784965930642604e-05, + "loss": 0.9901, + "step": 13495 + }, + { + "epoch": 0.32, + "learning_rate": 1.578433058077681e-05, + "loss": 1.0625, + "step": 13496 + }, + { + "epoch": 0.32, + "learning_rate": 1.578369519581884e-05, + "loss": 0.7842, + "step": 13497 + }, + { + "epoch": 0.32, + "learning_rate": 1.5783059775772553e-05, + "loss": 1.1701, + "step": 13498 + }, + { + "epoch": 0.32, + "learning_rate": 1.5782424320641804e-05, + "loss": 1.3819, + "step": 13499 + }, + { + "epoch": 0.32, + "learning_rate": 1.5781788830430443e-05, + "loss": 1.1301, + "step": 13500 + }, + { + "epoch": 0.32, + "learning_rate": 1.578115330514233e-05, + "loss": 0.9212, + "step": 13501 + }, + { + "epoch": 0.32, + "learning_rate": 1.5780517744781324e-05, + "loss": 1.2676, + "step": 13502 + }, + { + "epoch": 0.32, + "learning_rate": 1.577988214935127e-05, + "loss": 1.3019, + "step": 13503 + }, + { + "epoch": 0.32, + "learning_rate": 1.5779246518856038e-05, + "loss": 1.2839, + "step": 13504 + }, + { + "epoch": 0.32, + "learning_rate": 1.5778610853299472e-05, + "loss": 1.1697, + "step": 13505 + }, + { + "epoch": 0.32, + "learning_rate": 1.5777975152685433e-05, + "loss": 1.2339, + "step": 13506 + }, + { + "epoch": 0.32, + "learning_rate": 1.5777339417017778e-05, + "loss": 1.1713, + "step": 13507 + }, + { + "epoch": 0.32, + "learning_rate": 1.5776703646300365e-05, + "loss": 0.9698, + "step": 13508 + }, + { + "epoch": 0.32, + "learning_rate": 1.5776067840537048e-05, + "loss": 1.2555, + "step": 13509 + }, + { + "epoch": 0.32, + "learning_rate": 1.5775431999731686e-05, + "loss": 1.0785, + "step": 13510 + }, + { + "epoch": 0.32, + "learning_rate": 1.577479612388814e-05, + "loss": 1.2002, + "step": 13511 + }, + { + "epoch": 0.32, + "learning_rate": 1.577416021301026e-05, + "loss": 1.207, + "step": 13512 + }, + { + "epoch": 0.32, + "learning_rate": 1.5773524267101912e-05, + "loss": 0.9578, + "step": 13513 + }, + { + "epoch": 0.32, + "learning_rate": 1.5772888286166947e-05, + "loss": 1.202, + "step": 13514 + }, + { + "epoch": 0.33, + "learning_rate": 1.577225227020923e-05, + "loss": 1.0256, + "step": 13515 + }, + { + "epoch": 0.33, + "learning_rate": 1.5771616219232612e-05, + "loss": 1.0383, + "step": 13516 + }, + { + "epoch": 0.33, + "learning_rate": 1.577098013324096e-05, + "loss": 1.134, + "step": 13517 + }, + { + "epoch": 0.33, + "learning_rate": 1.5770344012238123e-05, + "loss": 1.0458, + "step": 13518 + }, + { + "epoch": 0.33, + "learning_rate": 1.5769707856227973e-05, + "loss": 1.1109, + "step": 13519 + }, + { + "epoch": 0.33, + "learning_rate": 1.5769071665214356e-05, + "loss": 1.1787, + "step": 13520 + }, + { + "epoch": 0.33, + "learning_rate": 1.5768435439201147e-05, + "loss": 1.2248, + "step": 13521 + }, + { + "epoch": 0.33, + "learning_rate": 1.576779917819219e-05, + "loss": 0.9727, + "step": 13522 + }, + { + "epoch": 0.33, + "learning_rate": 1.5767162882191353e-05, + "loss": 1.0448, + "step": 13523 + }, + { + "epoch": 0.33, + "learning_rate": 1.5766526551202496e-05, + "loss": 1.4336, + "step": 13524 + }, + { + "epoch": 0.33, + "learning_rate": 1.576589018522948e-05, + "loss": 1.1598, + "step": 13525 + }, + { + "epoch": 0.33, + "learning_rate": 1.5765253784276162e-05, + "loss": 1.0917, + "step": 13526 + }, + { + "epoch": 0.33, + "learning_rate": 1.576461734834641e-05, + "loss": 1.1413, + "step": 13527 + }, + { + "epoch": 0.33, + "learning_rate": 1.5763980877444075e-05, + "loss": 1.2221, + "step": 13528 + }, + { + "epoch": 0.33, + "learning_rate": 1.5763344371573028e-05, + "loss": 0.8957, + "step": 13529 + }, + { + "epoch": 0.33, + "learning_rate": 1.5762707830737124e-05, + "loss": 0.9907, + "step": 13530 + }, + { + "epoch": 0.33, + "learning_rate": 1.5762071254940224e-05, + "loss": 0.891, + "step": 13531 + }, + { + "epoch": 0.33, + "learning_rate": 1.57614346441862e-05, + "loss": 0.9906, + "step": 13532 + }, + { + "epoch": 0.33, + "learning_rate": 1.5760797998478903e-05, + "loss": 1.1803, + "step": 13533 + }, + { + "epoch": 0.33, + "learning_rate": 1.57601613178222e-05, + "loss": 0.9823, + "step": 13534 + }, + { + "epoch": 0.33, + "learning_rate": 1.5759524602219954e-05, + "loss": 1.1729, + "step": 13535 + }, + { + "epoch": 0.33, + "learning_rate": 1.5758887851676026e-05, + "loss": 0.8645, + "step": 13536 + }, + { + "epoch": 0.33, + "learning_rate": 1.575825106619428e-05, + "loss": 1.0512, + "step": 13537 + }, + { + "epoch": 0.33, + "learning_rate": 1.5757614245778578e-05, + "loss": 0.7462, + "step": 13538 + }, + { + "epoch": 0.33, + "learning_rate": 1.575697739043279e-05, + "loss": 1.2331, + "step": 13539 + }, + { + "epoch": 0.33, + "learning_rate": 1.5756340500160768e-05, + "loss": 1.2626, + "step": 13540 + }, + { + "epoch": 0.33, + "learning_rate": 1.5755703574966382e-05, + "loss": 0.8323, + "step": 13541 + }, + { + "epoch": 0.33, + "learning_rate": 1.57550666148535e-05, + "loss": 1.1594, + "step": 13542 + }, + { + "epoch": 0.33, + "learning_rate": 1.575442961982598e-05, + "loss": 1.2892, + "step": 13543 + }, + { + "epoch": 0.33, + "learning_rate": 1.575379258988769e-05, + "loss": 1.1643, + "step": 13544 + }, + { + "epoch": 0.33, + "learning_rate": 1.5753155525042497e-05, + "loss": 1.0847, + "step": 13545 + }, + { + "epoch": 0.33, + "learning_rate": 1.5752518425294258e-05, + "loss": 1.2604, + "step": 13546 + }, + { + "epoch": 0.33, + "learning_rate": 1.5751881290646846e-05, + "loss": 1.0494, + "step": 13547 + }, + { + "epoch": 0.33, + "learning_rate": 1.5751244121104122e-05, + "loss": 1.3128, + "step": 13548 + }, + { + "epoch": 0.33, + "learning_rate": 1.5750606916669953e-05, + "loss": 0.9298, + "step": 13549 + }, + { + "epoch": 0.33, + "learning_rate": 1.57499696773482e-05, + "loss": 0.934, + "step": 13550 + }, + { + "epoch": 0.33, + "learning_rate": 1.5749332403142742e-05, + "loss": 1.4043, + "step": 13551 + }, + { + "epoch": 0.33, + "learning_rate": 1.5748695094057432e-05, + "loss": 1.3848, + "step": 13552 + }, + { + "epoch": 0.33, + "learning_rate": 1.574805775009614e-05, + "loss": 1.1893, + "step": 13553 + }, + { + "epoch": 0.33, + "learning_rate": 1.5747420371262732e-05, + "loss": 1.05, + "step": 13554 + }, + { + "epoch": 0.33, + "learning_rate": 1.574678295756108e-05, + "loss": 1.1592, + "step": 13555 + }, + { + "epoch": 0.33, + "learning_rate": 1.574614550899505e-05, + "loss": 1.1937, + "step": 13556 + }, + { + "epoch": 0.33, + "learning_rate": 1.57455080255685e-05, + "loss": 1.0842, + "step": 13557 + }, + { + "epoch": 0.33, + "learning_rate": 1.5744870507285308e-05, + "loss": 1.2442, + "step": 13558 + }, + { + "epoch": 0.33, + "learning_rate": 1.574423295414934e-05, + "loss": 1.1955, + "step": 13559 + }, + { + "epoch": 0.33, + "learning_rate": 1.5743595366164458e-05, + "loss": 1.1755, + "step": 13560 + }, + { + "epoch": 0.33, + "learning_rate": 1.5742957743334536e-05, + "loss": 1.2285, + "step": 13561 + }, + { + "epoch": 0.33, + "learning_rate": 1.574232008566344e-05, + "loss": 1.2057, + "step": 13562 + }, + { + "epoch": 0.33, + "learning_rate": 1.5741682393155038e-05, + "loss": 0.9781, + "step": 13563 + }, + { + "epoch": 0.33, + "learning_rate": 1.57410446658132e-05, + "loss": 0.7829, + "step": 13564 + }, + { + "epoch": 0.33, + "learning_rate": 1.5740406903641795e-05, + "loss": 1.2998, + "step": 13565 + }, + { + "epoch": 0.33, + "learning_rate": 1.573976910664469e-05, + "loss": 1.3301, + "step": 13566 + }, + { + "epoch": 0.33, + "learning_rate": 1.5739131274825753e-05, + "loss": 1.3059, + "step": 13567 + }, + { + "epoch": 0.33, + "learning_rate": 1.5738493408188863e-05, + "loss": 0.8751, + "step": 13568 + }, + { + "epoch": 0.33, + "learning_rate": 1.573785550673788e-05, + "loss": 1.1092, + "step": 13569 + }, + { + "epoch": 0.33, + "learning_rate": 1.573721757047668e-05, + "loss": 1.1738, + "step": 13570 + }, + { + "epoch": 0.33, + "learning_rate": 1.5736579599409127e-05, + "loss": 1.2134, + "step": 13571 + }, + { + "epoch": 0.33, + "learning_rate": 1.573594159353909e-05, + "loss": 1.1701, + "step": 13572 + }, + { + "epoch": 0.33, + "learning_rate": 1.5735303552870453e-05, + "loss": 1.1738, + "step": 13573 + }, + { + "epoch": 0.33, + "learning_rate": 1.5734665477407076e-05, + "loss": 0.9474, + "step": 13574 + }, + { + "epoch": 0.33, + "learning_rate": 1.5734027367152834e-05, + "loss": 1.4678, + "step": 13575 + }, + { + "epoch": 0.33, + "learning_rate": 1.5733389222111592e-05, + "loss": 1.0726, + "step": 13576 + }, + { + "epoch": 0.33, + "learning_rate": 1.573275104228723e-05, + "loss": 1.2311, + "step": 13577 + }, + { + "epoch": 0.33, + "learning_rate": 1.5732112827683615e-05, + "loss": 0.9946, + "step": 13578 + }, + { + "epoch": 0.33, + "learning_rate": 1.5731474578304622e-05, + "loss": 0.9717, + "step": 13579 + }, + { + "epoch": 0.33, + "learning_rate": 1.5730836294154116e-05, + "loss": 1.1282, + "step": 13580 + }, + { + "epoch": 0.33, + "learning_rate": 1.5730197975235977e-05, + "loss": 0.9699, + "step": 13581 + }, + { + "epoch": 0.33, + "learning_rate": 1.5729559621554075e-05, + "loss": 1.1202, + "step": 13582 + }, + { + "epoch": 0.33, + "learning_rate": 1.5728921233112278e-05, + "loss": 0.8022, + "step": 13583 + }, + { + "epoch": 0.33, + "learning_rate": 1.572828280991447e-05, + "loss": 0.8264, + "step": 13584 + }, + { + "epoch": 0.33, + "learning_rate": 1.5727644351964516e-05, + "loss": 0.9641, + "step": 13585 + }, + { + "epoch": 0.33, + "learning_rate": 1.572700585926629e-05, + "loss": 1.3052, + "step": 13586 + }, + { + "epoch": 0.33, + "learning_rate": 1.5726367331823665e-05, + "loss": 1.0821, + "step": 13587 + }, + { + "epoch": 0.33, + "learning_rate": 1.5725728769640517e-05, + "loss": 1.1273, + "step": 13588 + }, + { + "epoch": 0.33, + "learning_rate": 1.572509017272072e-05, + "loss": 1.0598, + "step": 13589 + }, + { + "epoch": 0.33, + "learning_rate": 1.5724451541068147e-05, + "loss": 1.034, + "step": 13590 + }, + { + "epoch": 0.33, + "learning_rate": 1.5723812874686675e-05, + "loss": 1.1531, + "step": 13591 + }, + { + "epoch": 0.33, + "learning_rate": 1.5723174173580176e-05, + "loss": 1.2276, + "step": 13592 + }, + { + "epoch": 0.33, + "learning_rate": 1.5722535437752526e-05, + "loss": 0.8779, + "step": 13593 + }, + { + "epoch": 0.33, + "learning_rate": 1.5721896667207593e-05, + "loss": 1.0703, + "step": 13594 + }, + { + "epoch": 0.33, + "learning_rate": 1.5721257861949266e-05, + "loss": 1.0647, + "step": 13595 + }, + { + "epoch": 0.33, + "learning_rate": 1.572061902198141e-05, + "loss": 1.2547, + "step": 13596 + }, + { + "epoch": 0.33, + "learning_rate": 1.5719980147307907e-05, + "loss": 0.9678, + "step": 13597 + }, + { + "epoch": 0.33, + "learning_rate": 1.5719341237932628e-05, + "loss": 1.1844, + "step": 13598 + }, + { + "epoch": 0.33, + "learning_rate": 1.571870229385945e-05, + "loss": 1.1269, + "step": 13599 + }, + { + "epoch": 0.33, + "learning_rate": 1.571806331509225e-05, + "loss": 1.0747, + "step": 13600 + }, + { + "epoch": 0.33, + "learning_rate": 1.5717424301634906e-05, + "loss": 1.1293, + "step": 13601 + }, + { + "epoch": 0.33, + "learning_rate": 1.5716785253491297e-05, + "loss": 1.2688, + "step": 13602 + }, + { + "epoch": 0.33, + "learning_rate": 1.5716146170665293e-05, + "loss": 1.1599, + "step": 13603 + }, + { + "epoch": 0.33, + "learning_rate": 1.5715507053160775e-05, + "loss": 1.2363, + "step": 13604 + }, + { + "epoch": 0.33, + "learning_rate": 1.5714867900981618e-05, + "loss": 1.17, + "step": 13605 + }, + { + "epoch": 0.33, + "learning_rate": 1.5714228714131702e-05, + "loss": 1.1378, + "step": 13606 + }, + { + "epoch": 0.33, + "learning_rate": 1.571358949261491e-05, + "loss": 0.8228, + "step": 13607 + }, + { + "epoch": 0.33, + "learning_rate": 1.571295023643511e-05, + "loss": 0.9991, + "step": 13608 + }, + { + "epoch": 0.33, + "learning_rate": 1.5712310945596184e-05, + "loss": 1.1283, + "step": 13609 + }, + { + "epoch": 0.33, + "learning_rate": 1.5711671620102014e-05, + "loss": 1.1351, + "step": 13610 + }, + { + "epoch": 0.33, + "learning_rate": 1.571103225995647e-05, + "loss": 1.1915, + "step": 13611 + }, + { + "epoch": 0.33, + "learning_rate": 1.5710392865163443e-05, + "loss": 1.2058, + "step": 13612 + }, + { + "epoch": 0.33, + "learning_rate": 1.5709753435726804e-05, + "loss": 1.3905, + "step": 13613 + }, + { + "epoch": 0.33, + "learning_rate": 1.5709113971650434e-05, + "loss": 1.0027, + "step": 13614 + }, + { + "epoch": 0.33, + "learning_rate": 1.570847447293821e-05, + "loss": 1.332, + "step": 13615 + }, + { + "epoch": 0.33, + "learning_rate": 1.5707834939594016e-05, + "loss": 1.2117, + "step": 13616 + }, + { + "epoch": 0.33, + "learning_rate": 1.570719537162173e-05, + "loss": 1.157, + "step": 13617 + }, + { + "epoch": 0.33, + "learning_rate": 1.570655576902523e-05, + "loss": 1.1957, + "step": 13618 + }, + { + "epoch": 0.33, + "learning_rate": 1.57059161318084e-05, + "loss": 0.8473, + "step": 13619 + }, + { + "epoch": 0.33, + "learning_rate": 1.5705276459975124e-05, + "loss": 1.4123, + "step": 13620 + }, + { + "epoch": 0.33, + "learning_rate": 1.570463675352927e-05, + "loss": 1.2296, + "step": 13621 + }, + { + "epoch": 0.33, + "learning_rate": 1.570399701247473e-05, + "loss": 1.1871, + "step": 13622 + }, + { + "epoch": 0.33, + "learning_rate": 1.570335723681538e-05, + "loss": 0.9068, + "step": 13623 + }, + { + "epoch": 0.33, + "learning_rate": 1.5702717426555104e-05, + "loss": 1.2999, + "step": 13624 + }, + { + "epoch": 0.33, + "learning_rate": 1.570207758169778e-05, + "loss": 0.768, + "step": 13625 + }, + { + "epoch": 0.33, + "learning_rate": 1.5701437702247298e-05, + "loss": 1.1272, + "step": 13626 + }, + { + "epoch": 0.33, + "learning_rate": 1.570079778820753e-05, + "loss": 1.1856, + "step": 13627 + }, + { + "epoch": 0.33, + "learning_rate": 1.570015783958237e-05, + "loss": 1.2719, + "step": 13628 + }, + { + "epoch": 0.33, + "learning_rate": 1.5699517856375684e-05, + "loss": 1.1043, + "step": 13629 + }, + { + "epoch": 0.33, + "learning_rate": 1.569887783859137e-05, + "loss": 1.1065, + "step": 13630 + }, + { + "epoch": 0.33, + "learning_rate": 1.5698237786233305e-05, + "loss": 0.8729, + "step": 13631 + }, + { + "epoch": 0.33, + "learning_rate": 1.569759769930537e-05, + "loss": 0.8108, + "step": 13632 + }, + { + "epoch": 0.33, + "learning_rate": 1.5696957577811444e-05, + "loss": 1.1922, + "step": 13633 + }, + { + "epoch": 0.33, + "learning_rate": 1.569631742175542e-05, + "loss": 1.1989, + "step": 13634 + }, + { + "epoch": 0.33, + "learning_rate": 1.5695677231141183e-05, + "loss": 1.187, + "step": 13635 + }, + { + "epoch": 0.33, + "learning_rate": 1.5695037005972608e-05, + "loss": 1.2343, + "step": 13636 + }, + { + "epoch": 0.33, + "learning_rate": 1.5694396746253583e-05, + "loss": 1.0961, + "step": 13637 + }, + { + "epoch": 0.33, + "learning_rate": 1.5693756451987994e-05, + "loss": 1.1214, + "step": 13638 + }, + { + "epoch": 0.33, + "learning_rate": 1.569311612317972e-05, + "loss": 1.139, + "step": 13639 + }, + { + "epoch": 0.33, + "learning_rate": 1.5692475759832655e-05, + "loss": 0.9198, + "step": 13640 + }, + { + "epoch": 0.33, + "learning_rate": 1.5691835361950673e-05, + "loss": 1.1336, + "step": 13641 + }, + { + "epoch": 0.33, + "learning_rate": 1.569119492953767e-05, + "loss": 1.3063, + "step": 13642 + }, + { + "epoch": 0.33, + "learning_rate": 1.5690554462597523e-05, + "loss": 1.1192, + "step": 13643 + }, + { + "epoch": 0.33, + "learning_rate": 1.568991396113412e-05, + "loss": 0.9567, + "step": 13644 + }, + { + "epoch": 0.33, + "learning_rate": 1.568927342515135e-05, + "loss": 1.2438, + "step": 13645 + }, + { + "epoch": 0.33, + "learning_rate": 1.5688632854653095e-05, + "loss": 1.0159, + "step": 13646 + }, + { + "epoch": 0.33, + "learning_rate": 1.568799224964324e-05, + "loss": 1.2107, + "step": 13647 + }, + { + "epoch": 0.33, + "learning_rate": 1.5687351610125676e-05, + "loss": 1.1489, + "step": 13648 + }, + { + "epoch": 0.33, + "learning_rate": 1.568671093610429e-05, + "loss": 1.2456, + "step": 13649 + }, + { + "epoch": 0.33, + "learning_rate": 1.5686070227582964e-05, + "loss": 1.1746, + "step": 13650 + }, + { + "epoch": 0.33, + "learning_rate": 1.5685429484565587e-05, + "loss": 0.9434, + "step": 13651 + }, + { + "epoch": 0.33, + "learning_rate": 1.5684788707056046e-05, + "loss": 1.0881, + "step": 13652 + }, + { + "epoch": 0.33, + "learning_rate": 1.568414789505823e-05, + "loss": 1.144, + "step": 13653 + }, + { + "epoch": 0.33, + "learning_rate": 1.5683507048576027e-05, + "loss": 1.1709, + "step": 13654 + }, + { + "epoch": 0.33, + "learning_rate": 1.568286616761332e-05, + "loss": 0.802, + "step": 13655 + }, + { + "epoch": 0.33, + "learning_rate": 1.5682225252174003e-05, + "loss": 1.1665, + "step": 13656 + }, + { + "epoch": 0.33, + "learning_rate": 1.5681584302261964e-05, + "loss": 1.0316, + "step": 13657 + }, + { + "epoch": 0.33, + "learning_rate": 1.568094331788109e-05, + "loss": 1.1593, + "step": 13658 + }, + { + "epoch": 0.33, + "learning_rate": 1.5680302299035263e-05, + "loss": 0.936, + "step": 13659 + }, + { + "epoch": 0.33, + "learning_rate": 1.5679661245728383e-05, + "loss": 1.1536, + "step": 13660 + }, + { + "epoch": 0.33, + "learning_rate": 1.5679020157964335e-05, + "loss": 1.2081, + "step": 13661 + }, + { + "epoch": 0.33, + "learning_rate": 1.5678379035747e-05, + "loss": 0.7719, + "step": 13662 + }, + { + "epoch": 0.33, + "learning_rate": 1.567773787908028e-05, + "loss": 1.1135, + "step": 13663 + }, + { + "epoch": 0.33, + "learning_rate": 1.5677096687968063e-05, + "loss": 1.0853, + "step": 13664 + }, + { + "epoch": 0.33, + "learning_rate": 1.567645546241423e-05, + "loss": 1.5617, + "step": 13665 + }, + { + "epoch": 0.33, + "learning_rate": 1.567581420242268e-05, + "loss": 1.0809, + "step": 13666 + }, + { + "epoch": 0.33, + "learning_rate": 1.5675172907997302e-05, + "loss": 0.8392, + "step": 13667 + }, + { + "epoch": 0.33, + "learning_rate": 1.567453157914198e-05, + "loss": 1.1974, + "step": 13668 + }, + { + "epoch": 0.33, + "learning_rate": 1.5673890215860616e-05, + "loss": 1.3026, + "step": 13669 + }, + { + "epoch": 0.33, + "learning_rate": 1.5673248818157088e-05, + "loss": 0.9446, + "step": 13670 + }, + { + "epoch": 0.33, + "learning_rate": 1.56726073860353e-05, + "loss": 1.0531, + "step": 13671 + }, + { + "epoch": 0.33, + "learning_rate": 1.5671965919499133e-05, + "loss": 1.1178, + "step": 13672 + }, + { + "epoch": 0.33, + "learning_rate": 1.5671324418552484e-05, + "loss": 1.0778, + "step": 13673 + }, + { + "epoch": 0.33, + "learning_rate": 1.5670682883199245e-05, + "loss": 1.1158, + "step": 13674 + }, + { + "epoch": 0.33, + "learning_rate": 1.5670041313443305e-05, + "loss": 1.0909, + "step": 13675 + }, + { + "epoch": 0.33, + "learning_rate": 1.5669399709288558e-05, + "loss": 1.268, + "step": 13676 + }, + { + "epoch": 0.33, + "learning_rate": 1.5668758070738896e-05, + "loss": 1.029, + "step": 13677 + }, + { + "epoch": 0.33, + "learning_rate": 1.5668116397798213e-05, + "loss": 1.4961, + "step": 13678 + }, + { + "epoch": 0.33, + "learning_rate": 1.5667474690470404e-05, + "loss": 1.2693, + "step": 13679 + }, + { + "epoch": 0.33, + "learning_rate": 1.5666832948759352e-05, + "loss": 1.1079, + "step": 13680 + }, + { + "epoch": 0.33, + "learning_rate": 1.5666191172668965e-05, + "loss": 1.2309, + "step": 13681 + }, + { + "epoch": 0.33, + "learning_rate": 1.5665549362203125e-05, + "loss": 0.9459, + "step": 13682 + }, + { + "epoch": 0.33, + "learning_rate": 1.566490751736573e-05, + "loss": 1.0589, + "step": 13683 + }, + { + "epoch": 0.33, + "learning_rate": 1.5664265638160673e-05, + "loss": 0.9187, + "step": 13684 + }, + { + "epoch": 0.33, + "learning_rate": 1.566362372459185e-05, + "loss": 1.1138, + "step": 13685 + }, + { + "epoch": 0.33, + "learning_rate": 1.5662981776663155e-05, + "loss": 1.1609, + "step": 13686 + }, + { + "epoch": 0.33, + "learning_rate": 1.566233979437848e-05, + "loss": 1.1782, + "step": 13687 + }, + { + "epoch": 0.33, + "learning_rate": 1.566169777774172e-05, + "loss": 1.1398, + "step": 13688 + }, + { + "epoch": 0.33, + "learning_rate": 1.5661055726756774e-05, + "loss": 1.0382, + "step": 13689 + }, + { + "epoch": 0.33, + "learning_rate": 1.5660413641427533e-05, + "loss": 1.0825, + "step": 13690 + }, + { + "epoch": 0.33, + "learning_rate": 1.5659771521757895e-05, + "loss": 1.0065, + "step": 13691 + }, + { + "epoch": 0.33, + "learning_rate": 1.5659129367751754e-05, + "loss": 1.0538, + "step": 13692 + }, + { + "epoch": 0.33, + "learning_rate": 1.5658487179413003e-05, + "loss": 1.1709, + "step": 13693 + }, + { + "epoch": 0.33, + "learning_rate": 1.5657844956745545e-05, + "loss": 0.7981, + "step": 13694 + }, + { + "epoch": 0.33, + "learning_rate": 1.5657202699753273e-05, + "loss": 1.1679, + "step": 13695 + }, + { + "epoch": 0.33, + "learning_rate": 1.565656040844008e-05, + "loss": 1.1911, + "step": 13696 + }, + { + "epoch": 0.33, + "learning_rate": 1.5655918082809867e-05, + "loss": 1.2803, + "step": 13697 + }, + { + "epoch": 0.33, + "learning_rate": 1.565527572286653e-05, + "loss": 1.2006, + "step": 13698 + }, + { + "epoch": 0.33, + "learning_rate": 1.5654633328613965e-05, + "loss": 1.0966, + "step": 13699 + }, + { + "epoch": 0.33, + "learning_rate": 1.565399090005607e-05, + "loss": 1.2299, + "step": 13700 + }, + { + "epoch": 0.33, + "learning_rate": 1.565334843719674e-05, + "loss": 1.3165, + "step": 13701 + }, + { + "epoch": 0.33, + "learning_rate": 1.5652705940039876e-05, + "loss": 1.1016, + "step": 13702 + }, + { + "epoch": 0.33, + "learning_rate": 1.5652063408589376e-05, + "loss": 1.3876, + "step": 13703 + }, + { + "epoch": 0.33, + "learning_rate": 1.5651420842849135e-05, + "loss": 1.0201, + "step": 13704 + }, + { + "epoch": 0.33, + "learning_rate": 1.5650778242823054e-05, + "loss": 1.1206, + "step": 13705 + }, + { + "epoch": 0.33, + "learning_rate": 1.565013560851503e-05, + "loss": 0.9747, + "step": 13706 + }, + { + "epoch": 0.33, + "learning_rate": 1.564949293992896e-05, + "loss": 0.9308, + "step": 13707 + }, + { + "epoch": 0.33, + "learning_rate": 1.5648850237068745e-05, + "loss": 1.1459, + "step": 13708 + }, + { + "epoch": 0.33, + "learning_rate": 1.564820749993829e-05, + "loss": 0.9091, + "step": 13709 + }, + { + "epoch": 0.33, + "learning_rate": 1.5647564728541485e-05, + "loss": 1.0647, + "step": 13710 + }, + { + "epoch": 0.33, + "learning_rate": 1.5646921922882232e-05, + "loss": 1.213, + "step": 13711 + }, + { + "epoch": 0.33, + "learning_rate": 1.5646279082964437e-05, + "loss": 1.2345, + "step": 13712 + }, + { + "epoch": 0.33, + "learning_rate": 1.564563620879199e-05, + "loss": 1.0817, + "step": 13713 + }, + { + "epoch": 0.33, + "learning_rate": 1.5644993300368797e-05, + "loss": 1.0535, + "step": 13714 + }, + { + "epoch": 0.33, + "learning_rate": 1.564435035769876e-05, + "loss": 0.9939, + "step": 13715 + }, + { + "epoch": 0.33, + "learning_rate": 1.5643707380785772e-05, + "loss": 1.236, + "step": 13716 + }, + { + "epoch": 0.33, + "learning_rate": 1.5643064369633744e-05, + "loss": 0.9415, + "step": 13717 + }, + { + "epoch": 0.33, + "learning_rate": 1.564242132424657e-05, + "loss": 1.3592, + "step": 13718 + }, + { + "epoch": 0.33, + "learning_rate": 1.5641778244628153e-05, + "loss": 1.0064, + "step": 13719 + }, + { + "epoch": 0.33, + "learning_rate": 1.5641135130782393e-05, + "loss": 1.3533, + "step": 13720 + }, + { + "epoch": 0.33, + "learning_rate": 1.5640491982713194e-05, + "loss": 1.1418, + "step": 13721 + }, + { + "epoch": 0.33, + "learning_rate": 1.5639848800424458e-05, + "loss": 1.092, + "step": 13722 + }, + { + "epoch": 0.33, + "learning_rate": 1.5639205583920083e-05, + "loss": 1.1131, + "step": 13723 + }, + { + "epoch": 0.33, + "learning_rate": 1.5638562333203977e-05, + "loss": 1.0918, + "step": 13724 + }, + { + "epoch": 0.33, + "learning_rate": 1.5637919048280038e-05, + "loss": 1.246, + "step": 13725 + }, + { + "epoch": 0.33, + "learning_rate": 1.563727572915217e-05, + "loss": 1.0332, + "step": 13726 + }, + { + "epoch": 0.33, + "learning_rate": 1.5636632375824276e-05, + "loss": 0.9715, + "step": 13727 + }, + { + "epoch": 0.33, + "learning_rate": 1.5635988988300263e-05, + "loss": 1.2331, + "step": 13728 + }, + { + "epoch": 0.33, + "learning_rate": 1.5635345566584027e-05, + "loss": 1.2227, + "step": 13729 + }, + { + "epoch": 0.33, + "learning_rate": 1.5634702110679474e-05, + "loss": 1.0514, + "step": 13730 + }, + { + "epoch": 0.33, + "learning_rate": 1.5634058620590512e-05, + "loss": 1.3481, + "step": 13731 + }, + { + "epoch": 0.33, + "learning_rate": 1.5633415096321042e-05, + "loss": 0.9982, + "step": 13732 + }, + { + "epoch": 0.33, + "learning_rate": 1.5632771537874966e-05, + "loss": 0.9548, + "step": 13733 + }, + { + "epoch": 0.33, + "learning_rate": 1.5632127945256185e-05, + "loss": 1.1611, + "step": 13734 + }, + { + "epoch": 0.33, + "learning_rate": 1.5631484318468614e-05, + "loss": 1.1311, + "step": 13735 + }, + { + "epoch": 0.33, + "learning_rate": 1.5630840657516155e-05, + "loss": 1.1431, + "step": 13736 + }, + { + "epoch": 0.33, + "learning_rate": 1.56301969624027e-05, + "loss": 1.1675, + "step": 13737 + }, + { + "epoch": 0.33, + "learning_rate": 1.5629553233132173e-05, + "loss": 0.9815, + "step": 13738 + }, + { + "epoch": 0.33, + "learning_rate": 1.562890946970847e-05, + "loss": 1.0889, + "step": 13739 + }, + { + "epoch": 0.33, + "learning_rate": 1.5628265672135496e-05, + "loss": 1.0716, + "step": 13740 + }, + { + "epoch": 0.33, + "learning_rate": 1.5627621840417156e-05, + "loss": 1.429, + "step": 13741 + }, + { + "epoch": 0.33, + "learning_rate": 1.562697797455736e-05, + "loss": 1.0157, + "step": 13742 + }, + { + "epoch": 0.33, + "learning_rate": 1.5626334074560014e-05, + "loss": 1.2269, + "step": 13743 + }, + { + "epoch": 0.33, + "learning_rate": 1.562569014042902e-05, + "loss": 1.087, + "step": 13744 + }, + { + "epoch": 0.33, + "learning_rate": 1.5625046172168286e-05, + "loss": 1.1125, + "step": 13745 + }, + { + "epoch": 0.33, + "learning_rate": 1.562440216978172e-05, + "loss": 1.0541, + "step": 13746 + }, + { + "epoch": 0.33, + "learning_rate": 1.5623758133273234e-05, + "loss": 1.1284, + "step": 13747 + }, + { + "epoch": 0.33, + "learning_rate": 1.5623114062646724e-05, + "loss": 1.1736, + "step": 13748 + }, + { + "epoch": 0.33, + "learning_rate": 1.5622469957906108e-05, + "loss": 1.3803, + "step": 13749 + }, + { + "epoch": 0.33, + "learning_rate": 1.5621825819055288e-05, + "loss": 1.0991, + "step": 13750 + }, + { + "epoch": 0.33, + "learning_rate": 1.5621181646098172e-05, + "loss": 1.0385, + "step": 13751 + }, + { + "epoch": 0.33, + "learning_rate": 1.562053743903867e-05, + "loss": 1.0098, + "step": 13752 + }, + { + "epoch": 0.33, + "learning_rate": 1.5619893197880686e-05, + "loss": 1.0098, + "step": 13753 + }, + { + "epoch": 0.33, + "learning_rate": 1.561924892262814e-05, + "loss": 0.8888, + "step": 13754 + }, + { + "epoch": 0.33, + "learning_rate": 1.5618604613284926e-05, + "loss": 0.8986, + "step": 13755 + }, + { + "epoch": 0.33, + "learning_rate": 1.561796026985496e-05, + "loss": 1.174, + "step": 13756 + }, + { + "epoch": 0.33, + "learning_rate": 1.561731589234215e-05, + "loss": 1.2642, + "step": 13757 + }, + { + "epoch": 0.33, + "learning_rate": 1.561667148075041e-05, + "loss": 1.4091, + "step": 13758 + }, + { + "epoch": 0.33, + "learning_rate": 1.5616027035083642e-05, + "loss": 1.0969, + "step": 13759 + }, + { + "epoch": 0.33, + "learning_rate": 1.5615382555345758e-05, + "loss": 1.1519, + "step": 13760 + }, + { + "epoch": 0.33, + "learning_rate": 1.561473804154067e-05, + "loss": 0.9608, + "step": 13761 + }, + { + "epoch": 0.33, + "learning_rate": 1.561409349367229e-05, + "loss": 0.9485, + "step": 13762 + }, + { + "epoch": 0.33, + "learning_rate": 1.561344891174452e-05, + "loss": 1.104, + "step": 13763 + }, + { + "epoch": 0.33, + "learning_rate": 1.5612804295761277e-05, + "loss": 1.2343, + "step": 13764 + }, + { + "epoch": 0.33, + "learning_rate": 1.5612159645726474e-05, + "loss": 1.0239, + "step": 13765 + }, + { + "epoch": 0.33, + "learning_rate": 1.5611514961644016e-05, + "loss": 1.1435, + "step": 13766 + }, + { + "epoch": 0.33, + "learning_rate": 1.5610870243517816e-05, + "loss": 0.9291, + "step": 13767 + }, + { + "epoch": 0.33, + "learning_rate": 1.561022549135179e-05, + "loss": 1.2017, + "step": 13768 + }, + { + "epoch": 0.33, + "learning_rate": 1.5609580705149842e-05, + "loss": 1.2994, + "step": 13769 + }, + { + "epoch": 0.33, + "learning_rate": 1.5608935884915887e-05, + "loss": 1.1085, + "step": 13770 + }, + { + "epoch": 0.33, + "learning_rate": 1.5608291030653837e-05, + "loss": 1.0261, + "step": 13771 + }, + { + "epoch": 0.33, + "learning_rate": 1.5607646142367606e-05, + "loss": 1.1262, + "step": 13772 + }, + { + "epoch": 0.33, + "learning_rate": 1.5607001220061105e-05, + "loss": 1.3519, + "step": 13773 + }, + { + "epoch": 0.33, + "learning_rate": 1.5606356263738246e-05, + "loss": 1.1832, + "step": 13774 + }, + { + "epoch": 0.33, + "learning_rate": 1.5605711273402942e-05, + "loss": 1.1055, + "step": 13775 + }, + { + "epoch": 0.33, + "learning_rate": 1.5605066249059106e-05, + "loss": 1.1372, + "step": 13776 + }, + { + "epoch": 0.33, + "learning_rate": 1.5604421190710654e-05, + "loss": 1.1883, + "step": 13777 + }, + { + "epoch": 0.33, + "learning_rate": 1.5603776098361493e-05, + "loss": 1.1909, + "step": 13778 + }, + { + "epoch": 0.33, + "learning_rate": 1.560313097201554e-05, + "loss": 1.1806, + "step": 13779 + }, + { + "epoch": 0.33, + "learning_rate": 1.5602485811676712e-05, + "loss": 1.3334, + "step": 13780 + }, + { + "epoch": 0.33, + "learning_rate": 1.560184061734892e-05, + "loss": 1.1327, + "step": 13781 + }, + { + "epoch": 0.33, + "learning_rate": 1.560119538903608e-05, + "loss": 1.1984, + "step": 13782 + }, + { + "epoch": 0.33, + "learning_rate": 1.5600550126742103e-05, + "loss": 1.1298, + "step": 13783 + }, + { + "epoch": 0.33, + "learning_rate": 1.5599904830470906e-05, + "loss": 1.1089, + "step": 13784 + }, + { + "epoch": 0.33, + "learning_rate": 1.5599259500226406e-05, + "loss": 1.3702, + "step": 13785 + }, + { + "epoch": 0.33, + "learning_rate": 1.5598614136012514e-05, + "loss": 1.2273, + "step": 13786 + }, + { + "epoch": 0.33, + "learning_rate": 1.5597968737833146e-05, + "loss": 0.9949, + "step": 13787 + }, + { + "epoch": 0.33, + "learning_rate": 1.559732330569222e-05, + "loss": 1.1067, + "step": 13788 + }, + { + "epoch": 0.33, + "learning_rate": 1.5596677839593646e-05, + "loss": 1.239, + "step": 13789 + }, + { + "epoch": 0.33, + "learning_rate": 1.5596032339541345e-05, + "loss": 1.1465, + "step": 13790 + }, + { + "epoch": 0.33, + "learning_rate": 1.5595386805539234e-05, + "loss": 0.943, + "step": 13791 + }, + { + "epoch": 0.33, + "learning_rate": 1.5594741237591228e-05, + "loss": 1.0823, + "step": 13792 + }, + { + "epoch": 0.33, + "learning_rate": 1.5594095635701242e-05, + "loss": 1.0428, + "step": 13793 + }, + { + "epoch": 0.33, + "learning_rate": 1.5593449999873195e-05, + "loss": 1.3144, + "step": 13794 + }, + { + "epoch": 0.33, + "learning_rate": 1.5592804330110998e-05, + "loss": 0.7463, + "step": 13795 + }, + { + "epoch": 0.33, + "learning_rate": 1.5592158626418576e-05, + "loss": 1.0132, + "step": 13796 + }, + { + "epoch": 0.33, + "learning_rate": 1.5591512888799845e-05, + "loss": 1.1378, + "step": 13797 + }, + { + "epoch": 0.33, + "learning_rate": 1.5590867117258718e-05, + "loss": 1.0588, + "step": 13798 + }, + { + "epoch": 0.33, + "learning_rate": 1.5590221311799116e-05, + "loss": 1.0495, + "step": 13799 + }, + { + "epoch": 0.33, + "learning_rate": 1.5589575472424954e-05, + "loss": 1.2361, + "step": 13800 + }, + { + "epoch": 0.33, + "learning_rate": 1.5588929599140155e-05, + "loss": 1.119, + "step": 13801 + }, + { + "epoch": 0.33, + "learning_rate": 1.5588283691948633e-05, + "loss": 1.2628, + "step": 13802 + }, + { + "epoch": 0.33, + "learning_rate": 1.558763775085431e-05, + "loss": 1.0264, + "step": 13803 + }, + { + "epoch": 0.33, + "learning_rate": 1.5586991775861103e-05, + "loss": 1.2034, + "step": 13804 + }, + { + "epoch": 0.33, + "learning_rate": 1.558634576697293e-05, + "loss": 1.1214, + "step": 13805 + }, + { + "epoch": 0.33, + "learning_rate": 1.558569972419371e-05, + "loss": 1.0227, + "step": 13806 + }, + { + "epoch": 0.33, + "learning_rate": 1.5585053647527366e-05, + "loss": 1.1217, + "step": 13807 + }, + { + "epoch": 0.33, + "learning_rate": 1.5584407536977816e-05, + "loss": 1.2491, + "step": 13808 + }, + { + "epoch": 0.33, + "learning_rate": 1.5583761392548974e-05, + "loss": 0.9422, + "step": 13809 + }, + { + "epoch": 0.33, + "learning_rate": 1.558311521424477e-05, + "loss": 0.9659, + "step": 13810 + }, + { + "epoch": 0.33, + "learning_rate": 1.558246900206912e-05, + "loss": 1.4976, + "step": 13811 + }, + { + "epoch": 0.33, + "learning_rate": 1.558182275602594e-05, + "loss": 0.9663, + "step": 13812 + }, + { + "epoch": 0.33, + "learning_rate": 1.5581176476119154e-05, + "loss": 1.1133, + "step": 13813 + }, + { + "epoch": 0.33, + "learning_rate": 1.5580530162352685e-05, + "loss": 1.1777, + "step": 13814 + }, + { + "epoch": 0.33, + "learning_rate": 1.557988381473045e-05, + "loss": 1.3164, + "step": 13815 + }, + { + "epoch": 0.33, + "learning_rate": 1.557923743325638e-05, + "loss": 1.2661, + "step": 13816 + }, + { + "epoch": 0.33, + "learning_rate": 1.557859101793438e-05, + "loss": 1.1058, + "step": 13817 + }, + { + "epoch": 0.33, + "learning_rate": 1.5577944568768383e-05, + "loss": 1.429, + "step": 13818 + }, + { + "epoch": 0.33, + "learning_rate": 1.5577298085762308e-05, + "loss": 1.0886, + "step": 13819 + }, + { + "epoch": 0.33, + "learning_rate": 1.557665156892008e-05, + "loss": 1.2933, + "step": 13820 + }, + { + "epoch": 0.33, + "learning_rate": 1.5576005018245613e-05, + "loss": 1.0364, + "step": 13821 + }, + { + "epoch": 0.33, + "learning_rate": 1.557535843374284e-05, + "loss": 1.3697, + "step": 13822 + }, + { + "epoch": 0.33, + "learning_rate": 1.5574711815415676e-05, + "loss": 1.1671, + "step": 13823 + }, + { + "epoch": 0.33, + "learning_rate": 1.557406516326805e-05, + "loss": 1.2048, + "step": 13824 + }, + { + "epoch": 0.33, + "learning_rate": 1.5573418477303877e-05, + "loss": 1.0326, + "step": 13825 + }, + { + "epoch": 0.33, + "learning_rate": 1.557277175752709e-05, + "loss": 1.2967, + "step": 13826 + }, + { + "epoch": 0.33, + "learning_rate": 1.5572125003941608e-05, + "loss": 0.9508, + "step": 13827 + }, + { + "epoch": 0.33, + "learning_rate": 1.557147821655135e-05, + "loss": 0.9286, + "step": 13828 + }, + { + "epoch": 0.33, + "learning_rate": 1.5570831395360246e-05, + "loss": 1.1202, + "step": 13829 + }, + { + "epoch": 0.33, + "learning_rate": 1.557018454037222e-05, + "loss": 1.0314, + "step": 13830 + }, + { + "epoch": 0.33, + "learning_rate": 1.5569537651591193e-05, + "loss": 0.9061, + "step": 13831 + }, + { + "epoch": 0.33, + "learning_rate": 1.556889072902109e-05, + "loss": 1.1979, + "step": 13832 + }, + { + "epoch": 0.33, + "learning_rate": 1.5568243772665835e-05, + "loss": 1.0862, + "step": 13833 + }, + { + "epoch": 0.33, + "learning_rate": 1.556759678252936e-05, + "loss": 1.1268, + "step": 13834 + }, + { + "epoch": 0.33, + "learning_rate": 1.556694975861558e-05, + "loss": 1.0475, + "step": 13835 + }, + { + "epoch": 0.33, + "learning_rate": 1.556630270092843e-05, + "loss": 1.101, + "step": 13836 + }, + { + "epoch": 0.33, + "learning_rate": 1.556565560947183e-05, + "loss": 1.1706, + "step": 13837 + }, + { + "epoch": 0.33, + "learning_rate": 1.55650084842497e-05, + "loss": 0.7836, + "step": 13838 + }, + { + "epoch": 0.33, + "learning_rate": 1.556436132526598e-05, + "loss": 0.9764, + "step": 13839 + }, + { + "epoch": 0.33, + "learning_rate": 1.5563714132524586e-05, + "loss": 1.1581, + "step": 13840 + }, + { + "epoch": 0.33, + "learning_rate": 1.5563066906029446e-05, + "loss": 1.1602, + "step": 13841 + }, + { + "epoch": 0.33, + "learning_rate": 1.556241964578449e-05, + "loss": 1.1184, + "step": 13842 + }, + { + "epoch": 0.33, + "learning_rate": 1.556177235179364e-05, + "loss": 1.1287, + "step": 13843 + }, + { + "epoch": 0.33, + "learning_rate": 1.556112502406083e-05, + "loss": 1.1575, + "step": 13844 + }, + { + "epoch": 0.33, + "learning_rate": 1.5560477662589976e-05, + "loss": 0.9955, + "step": 13845 + }, + { + "epoch": 0.33, + "learning_rate": 1.5559830267385017e-05, + "loss": 0.9582, + "step": 13846 + }, + { + "epoch": 0.33, + "learning_rate": 1.5559182838449873e-05, + "loss": 1.2532, + "step": 13847 + }, + { + "epoch": 0.33, + "learning_rate": 1.5558535375788474e-05, + "loss": 1.2995, + "step": 13848 + }, + { + "epoch": 0.33, + "learning_rate": 1.555788787940475e-05, + "loss": 1.1359, + "step": 13849 + }, + { + "epoch": 0.33, + "learning_rate": 1.5557240349302625e-05, + "loss": 1.418, + "step": 13850 + }, + { + "epoch": 0.33, + "learning_rate": 1.5556592785486033e-05, + "loss": 1.2518, + "step": 13851 + }, + { + "epoch": 0.33, + "learning_rate": 1.5555945187958896e-05, + "loss": 1.1859, + "step": 13852 + }, + { + "epoch": 0.33, + "learning_rate": 1.555529755672515e-05, + "loss": 1.3356, + "step": 13853 + }, + { + "epoch": 0.33, + "learning_rate": 1.5554649891788717e-05, + "loss": 1.1871, + "step": 13854 + }, + { + "epoch": 0.33, + "learning_rate": 1.5554002193153533e-05, + "loss": 1.078, + "step": 13855 + }, + { + "epoch": 0.33, + "learning_rate": 1.555335446082352e-05, + "loss": 1.0403, + "step": 13856 + }, + { + "epoch": 0.33, + "learning_rate": 1.5552706694802618e-05, + "loss": 1.1534, + "step": 13857 + }, + { + "epoch": 0.33, + "learning_rate": 1.5552058895094747e-05, + "loss": 1.4951, + "step": 13858 + }, + { + "epoch": 0.33, + "learning_rate": 1.555141106170384e-05, + "loss": 0.8452, + "step": 13859 + }, + { + "epoch": 0.33, + "learning_rate": 1.555076319463383e-05, + "loss": 1.1755, + "step": 13860 + }, + { + "epoch": 0.33, + "learning_rate": 1.555011529388864e-05, + "loss": 1.1954, + "step": 13861 + }, + { + "epoch": 0.33, + "learning_rate": 1.554946735947221e-05, + "loss": 1.0714, + "step": 13862 + }, + { + "epoch": 0.33, + "learning_rate": 1.5548819391388468e-05, + "loss": 1.14, + "step": 13863 + }, + { + "epoch": 0.33, + "learning_rate": 1.554817138964134e-05, + "loss": 1.2334, + "step": 13864 + }, + { + "epoch": 0.33, + "learning_rate": 1.5547523354234765e-05, + "loss": 1.1689, + "step": 13865 + }, + { + "epoch": 0.33, + "learning_rate": 1.554687528517267e-05, + "loss": 1.2902, + "step": 13866 + }, + { + "epoch": 0.33, + "learning_rate": 1.5546227182458985e-05, + "loss": 0.9984, + "step": 13867 + }, + { + "epoch": 0.33, + "learning_rate": 1.5545579046097645e-05, + "loss": 1.3488, + "step": 13868 + }, + { + "epoch": 0.33, + "learning_rate": 1.5544930876092584e-05, + "loss": 1.016, + "step": 13869 + }, + { + "epoch": 0.33, + "learning_rate": 1.554428267244773e-05, + "loss": 0.8631, + "step": 13870 + }, + { + "epoch": 0.33, + "learning_rate": 1.5543634435167016e-05, + "loss": 1.1772, + "step": 13871 + }, + { + "epoch": 0.33, + "learning_rate": 1.5542986164254378e-05, + "loss": 1.0517, + "step": 13872 + }, + { + "epoch": 0.33, + "learning_rate": 1.5542337859713745e-05, + "loss": 1.1758, + "step": 13873 + }, + { + "epoch": 0.33, + "learning_rate": 1.5541689521549052e-05, + "loss": 1.1526, + "step": 13874 + }, + { + "epoch": 0.33, + "learning_rate": 1.5541041149764232e-05, + "loss": 1.2163, + "step": 13875 + }, + { + "epoch": 0.33, + "learning_rate": 1.554039274436322e-05, + "loss": 1.0065, + "step": 13876 + }, + { + "epoch": 0.33, + "learning_rate": 1.5539744305349946e-05, + "loss": 1.4031, + "step": 13877 + }, + { + "epoch": 0.33, + "learning_rate": 1.5539095832728344e-05, + "loss": 1.2085, + "step": 13878 + }, + { + "epoch": 0.33, + "learning_rate": 1.553844732650236e-05, + "loss": 1.1074, + "step": 13879 + }, + { + "epoch": 0.33, + "learning_rate": 1.553779878667591e-05, + "loss": 0.9202, + "step": 13880 + }, + { + "epoch": 0.33, + "learning_rate": 1.553715021325294e-05, + "loss": 0.8714, + "step": 13881 + }, + { + "epoch": 0.33, + "learning_rate": 1.5536501606237383e-05, + "loss": 1.1968, + "step": 13882 + }, + { + "epoch": 0.33, + "learning_rate": 1.5535852965633173e-05, + "loss": 1.3072, + "step": 13883 + }, + { + "epoch": 0.33, + "learning_rate": 1.5535204291444245e-05, + "loss": 1.1109, + "step": 13884 + }, + { + "epoch": 0.33, + "learning_rate": 1.5534555583674536e-05, + "loss": 0.807, + "step": 13885 + }, + { + "epoch": 0.33, + "learning_rate": 1.5533906842327978e-05, + "loss": 1.1872, + "step": 13886 + }, + { + "epoch": 0.33, + "learning_rate": 1.553325806740851e-05, + "loss": 1.1569, + "step": 13887 + }, + { + "epoch": 0.33, + "learning_rate": 1.5532609258920065e-05, + "loss": 1.0112, + "step": 13888 + }, + { + "epoch": 0.33, + "learning_rate": 1.553196041686658e-05, + "loss": 1.287, + "step": 13889 + }, + { + "epoch": 0.33, + "learning_rate": 1.5531311541251995e-05, + "loss": 1.3491, + "step": 13890 + }, + { + "epoch": 0.33, + "learning_rate": 1.5530662632080244e-05, + "loss": 1.2618, + "step": 13891 + }, + { + "epoch": 0.33, + "learning_rate": 1.5530013689355265e-05, + "loss": 1.1345, + "step": 13892 + }, + { + "epoch": 0.33, + "learning_rate": 1.552936471308099e-05, + "loss": 0.8298, + "step": 13893 + }, + { + "epoch": 0.33, + "learning_rate": 1.552871570326136e-05, + "loss": 1.0993, + "step": 13894 + }, + { + "epoch": 0.33, + "learning_rate": 1.5528066659900315e-05, + "loss": 0.9683, + "step": 13895 + }, + { + "epoch": 0.33, + "learning_rate": 1.552741758300179e-05, + "loss": 1.3842, + "step": 13896 + }, + { + "epoch": 0.33, + "learning_rate": 1.552676847256972e-05, + "loss": 1.0451, + "step": 13897 + }, + { + "epoch": 0.33, + "learning_rate": 1.5526119328608044e-05, + "loss": 1.3017, + "step": 13898 + }, + { + "epoch": 0.33, + "learning_rate": 1.5525470151120707e-05, + "loss": 1.221, + "step": 13899 + }, + { + "epoch": 0.33, + "learning_rate": 1.552482094011164e-05, + "loss": 1.1998, + "step": 13900 + }, + { + "epoch": 0.33, + "learning_rate": 1.5524171695584785e-05, + "loss": 0.928, + "step": 13901 + }, + { + "epoch": 0.33, + "learning_rate": 1.5523522417544076e-05, + "loss": 1.0633, + "step": 13902 + }, + { + "epoch": 0.33, + "learning_rate": 1.552287310599346e-05, + "loss": 1.3549, + "step": 13903 + }, + { + "epoch": 0.33, + "learning_rate": 1.552222376093687e-05, + "loss": 1.3205, + "step": 13904 + }, + { + "epoch": 0.33, + "learning_rate": 1.5521574382378248e-05, + "loss": 1.0628, + "step": 13905 + }, + { + "epoch": 0.33, + "learning_rate": 1.552092497032153e-05, + "loss": 1.4511, + "step": 13906 + }, + { + "epoch": 0.33, + "learning_rate": 1.5520275524770663e-05, + "loss": 1.0061, + "step": 13907 + }, + { + "epoch": 0.33, + "learning_rate": 1.551962604572958e-05, + "loss": 1.2775, + "step": 13908 + }, + { + "epoch": 0.33, + "learning_rate": 1.551897653320223e-05, + "loss": 1.0583, + "step": 13909 + }, + { + "epoch": 0.33, + "learning_rate": 1.5518326987192537e-05, + "loss": 1.1169, + "step": 13910 + }, + { + "epoch": 0.33, + "learning_rate": 1.551767740770446e-05, + "loss": 1.3366, + "step": 13911 + }, + { + "epoch": 0.33, + "learning_rate": 1.551702779474193e-05, + "loss": 1.2206, + "step": 13912 + }, + { + "epoch": 0.33, + "learning_rate": 1.5516378148308892e-05, + "loss": 1.4234, + "step": 13913 + }, + { + "epoch": 0.33, + "learning_rate": 1.5515728468409282e-05, + "loss": 1.1732, + "step": 13914 + }, + { + "epoch": 0.33, + "learning_rate": 1.5515078755047046e-05, + "loss": 1.1937, + "step": 13915 + }, + { + "epoch": 0.33, + "learning_rate": 1.5514429008226124e-05, + "loss": 1.2497, + "step": 13916 + }, + { + "epoch": 0.33, + "learning_rate": 1.5513779227950457e-05, + "loss": 1.0984, + "step": 13917 + }, + { + "epoch": 0.33, + "learning_rate": 1.5513129414223992e-05, + "loss": 1.252, + "step": 13918 + }, + { + "epoch": 0.33, + "learning_rate": 1.5512479567050666e-05, + "loss": 1.0677, + "step": 13919 + }, + { + "epoch": 0.33, + "learning_rate": 1.5511829686434422e-05, + "loss": 0.991, + "step": 13920 + }, + { + "epoch": 0.33, + "learning_rate": 1.5511179772379205e-05, + "loss": 1.0716, + "step": 13921 + }, + { + "epoch": 0.33, + "learning_rate": 1.5510529824888957e-05, + "loss": 1.2959, + "step": 13922 + }, + { + "epoch": 0.33, + "learning_rate": 1.5509879843967618e-05, + "loss": 1.0104, + "step": 13923 + }, + { + "epoch": 0.33, + "learning_rate": 1.5509229829619136e-05, + "loss": 0.9714, + "step": 13924 + }, + { + "epoch": 0.33, + "learning_rate": 1.550857978184745e-05, + "loss": 1.1466, + "step": 13925 + }, + { + "epoch": 0.33, + "learning_rate": 1.5507929700656513e-05, + "loss": 0.923, + "step": 13926 + }, + { + "epoch": 0.33, + "learning_rate": 1.5507279586050254e-05, + "loss": 0.9198, + "step": 13927 + }, + { + "epoch": 0.33, + "learning_rate": 1.550662943803263e-05, + "loss": 0.9717, + "step": 13928 + }, + { + "epoch": 0.33, + "learning_rate": 1.550597925660758e-05, + "loss": 1.1579, + "step": 13929 + }, + { + "epoch": 0.34, + "learning_rate": 1.5505329041779048e-05, + "loss": 0.9758, + "step": 13930 + }, + { + "epoch": 0.34, + "learning_rate": 1.550467879355098e-05, + "loss": 0.87, + "step": 13931 + }, + { + "epoch": 0.34, + "learning_rate": 1.550402851192732e-05, + "loss": 1.1291, + "step": 13932 + }, + { + "epoch": 0.34, + "learning_rate": 1.5503378196912014e-05, + "loss": 1.0701, + "step": 13933 + }, + { + "epoch": 0.34, + "learning_rate": 1.5502727848509008e-05, + "loss": 1.0988, + "step": 13934 + }, + { + "epoch": 0.34, + "learning_rate": 1.5502077466722244e-05, + "loss": 1.1949, + "step": 13935 + }, + { + "epoch": 0.34, + "learning_rate": 1.550142705155567e-05, + "loss": 0.8946, + "step": 13936 + }, + { + "epoch": 0.34, + "learning_rate": 1.5500776603013235e-05, + "loss": 0.8718, + "step": 13937 + }, + { + "epoch": 0.34, + "learning_rate": 1.5500126121098883e-05, + "loss": 0.9303, + "step": 13938 + }, + { + "epoch": 0.34, + "learning_rate": 1.5499475605816556e-05, + "loss": 0.9805, + "step": 13939 + }, + { + "epoch": 0.34, + "learning_rate": 1.5498825057170207e-05, + "loss": 1.0761, + "step": 13940 + }, + { + "epoch": 0.34, + "learning_rate": 1.5498174475163775e-05, + "loss": 1.1922, + "step": 13941 + }, + { + "epoch": 0.34, + "learning_rate": 1.5497523859801215e-05, + "loss": 1.1531, + "step": 13942 + }, + { + "epoch": 0.34, + "learning_rate": 1.5496873211086472e-05, + "loss": 1.1173, + "step": 13943 + }, + { + "epoch": 0.34, + "learning_rate": 1.549622252902349e-05, + "loss": 1.3204, + "step": 13944 + }, + { + "epoch": 0.34, + "learning_rate": 1.549557181361622e-05, + "loss": 0.9968, + "step": 13945 + }, + { + "epoch": 0.34, + "learning_rate": 1.549492106486861e-05, + "loss": 0.9408, + "step": 13946 + }, + { + "epoch": 0.34, + "learning_rate": 1.5494270282784602e-05, + "loss": 0.9258, + "step": 13947 + }, + { + "epoch": 0.34, + "learning_rate": 1.549361946736815e-05, + "loss": 1.0384, + "step": 13948 + }, + { + "epoch": 0.34, + "learning_rate": 1.5492968618623203e-05, + "loss": 1.265, + "step": 13949 + }, + { + "epoch": 0.34, + "learning_rate": 1.549231773655371e-05, + "loss": 1.2828, + "step": 13950 + }, + { + "epoch": 0.34, + "learning_rate": 1.549166682116361e-05, + "loss": 0.8628, + "step": 13951 + }, + { + "epoch": 0.34, + "learning_rate": 1.5491015872456864e-05, + "loss": 1.4415, + "step": 13952 + }, + { + "epoch": 0.34, + "learning_rate": 1.5490364890437414e-05, + "loss": 0.7211, + "step": 13953 + }, + { + "epoch": 0.34, + "learning_rate": 1.548971387510921e-05, + "loss": 1.205, + "step": 13954 + }, + { + "epoch": 0.34, + "learning_rate": 1.5489062826476206e-05, + "loss": 0.6738, + "step": 13955 + }, + { + "epoch": 0.34, + "learning_rate": 1.548841174454235e-05, + "loss": 1.0201, + "step": 13956 + }, + { + "epoch": 0.34, + "learning_rate": 1.5487760629311587e-05, + "loss": 1.1595, + "step": 13957 + }, + { + "epoch": 0.34, + "learning_rate": 1.5487109480787875e-05, + "loss": 1.0221, + "step": 13958 + }, + { + "epoch": 0.34, + "learning_rate": 1.548645829897516e-05, + "loss": 1.095, + "step": 13959 + }, + { + "epoch": 0.34, + "learning_rate": 1.5485807083877393e-05, + "loss": 1.1231, + "step": 13960 + }, + { + "epoch": 0.34, + "learning_rate": 1.5485155835498522e-05, + "loss": 1.049, + "step": 13961 + }, + { + "epoch": 0.34, + "learning_rate": 1.5484504553842503e-05, + "loss": 0.8794, + "step": 13962 + }, + { + "epoch": 0.34, + "learning_rate": 1.5483853238913284e-05, + "loss": 1.1923, + "step": 13963 + }, + { + "epoch": 0.34, + "learning_rate": 1.5483201890714816e-05, + "loss": 1.1467, + "step": 13964 + }, + { + "epoch": 0.34, + "learning_rate": 1.5482550509251054e-05, + "loss": 1.3067, + "step": 13965 + }, + { + "epoch": 0.34, + "learning_rate": 1.5481899094525945e-05, + "loss": 1.1559, + "step": 13966 + }, + { + "epoch": 0.34, + "learning_rate": 1.5481247646543446e-05, + "loss": 1.5115, + "step": 13967 + }, + { + "epoch": 0.34, + "learning_rate": 1.548059616530751e-05, + "loss": 0.8928, + "step": 13968 + }, + { + "epoch": 0.34, + "learning_rate": 1.547994465082208e-05, + "loss": 1.1438, + "step": 13969 + }, + { + "epoch": 0.34, + "learning_rate": 1.5479293103091116e-05, + "loss": 0.9292, + "step": 13970 + }, + { + "epoch": 0.34, + "learning_rate": 1.5478641522118573e-05, + "loss": 1.1053, + "step": 13971 + }, + { + "epoch": 0.34, + "learning_rate": 1.54779899079084e-05, + "loss": 1.1136, + "step": 13972 + }, + { + "epoch": 0.34, + "learning_rate": 1.5477338260464548e-05, + "loss": 1.0975, + "step": 13973 + }, + { + "epoch": 0.34, + "learning_rate": 1.5476686579790975e-05, + "loss": 0.9248, + "step": 13974 + }, + { + "epoch": 0.34, + "learning_rate": 1.547603486589163e-05, + "loss": 1.2148, + "step": 13975 + }, + { + "epoch": 0.34, + "learning_rate": 1.5475383118770473e-05, + "loss": 1.2692, + "step": 13976 + }, + { + "epoch": 0.34, + "learning_rate": 1.5474731338431453e-05, + "loss": 1.215, + "step": 13977 + }, + { + "epoch": 0.34, + "learning_rate": 1.5474079524878525e-05, + "loss": 1.1029, + "step": 13978 + }, + { + "epoch": 0.34, + "learning_rate": 1.5473427678115645e-05, + "loss": 0.9732, + "step": 13979 + }, + { + "epoch": 0.34, + "learning_rate": 1.5472775798146766e-05, + "loss": 1.0586, + "step": 13980 + }, + { + "epoch": 0.34, + "learning_rate": 1.5472123884975846e-05, + "loss": 1.2363, + "step": 13981 + }, + { + "epoch": 0.34, + "learning_rate": 1.5471471938606834e-05, + "loss": 1.3218, + "step": 13982 + }, + { + "epoch": 0.34, + "learning_rate": 1.547081995904369e-05, + "loss": 1.2228, + "step": 13983 + }, + { + "epoch": 0.34, + "learning_rate": 1.547016794629037e-05, + "loss": 1.2357, + "step": 13984 + }, + { + "epoch": 0.34, + "learning_rate": 1.5469515900350828e-05, + "loss": 1.0062, + "step": 13985 + }, + { + "epoch": 0.34, + "learning_rate": 1.5468863821229016e-05, + "loss": 1.1672, + "step": 13986 + }, + { + "epoch": 0.34, + "learning_rate": 1.5468211708928893e-05, + "loss": 0.9934, + "step": 13987 + }, + { + "epoch": 0.34, + "learning_rate": 1.546755956345442e-05, + "loss": 1.3775, + "step": 13988 + }, + { + "epoch": 0.34, + "learning_rate": 1.5466907384809546e-05, + "loss": 1.0849, + "step": 13989 + }, + { + "epoch": 0.34, + "learning_rate": 1.546625517299823e-05, + "loss": 1.1692, + "step": 13990 + }, + { + "epoch": 0.34, + "learning_rate": 1.5465602928024434e-05, + "loss": 1.2366, + "step": 13991 + }, + { + "epoch": 0.34, + "learning_rate": 1.546495064989211e-05, + "loss": 1.1587, + "step": 13992 + }, + { + "epoch": 0.34, + "learning_rate": 1.5464298338605214e-05, + "loss": 1.1108, + "step": 13993 + }, + { + "epoch": 0.34, + "learning_rate": 1.5463645994167705e-05, + "loss": 1.1202, + "step": 13994 + }, + { + "epoch": 0.34, + "learning_rate": 1.5462993616583538e-05, + "loss": 1.1514, + "step": 13995 + }, + { + "epoch": 0.34, + "learning_rate": 1.546234120585668e-05, + "loss": 1.0383, + "step": 13996 + }, + { + "epoch": 0.34, + "learning_rate": 1.546168876199108e-05, + "loss": 1.1889, + "step": 13997 + }, + { + "epoch": 0.34, + "learning_rate": 1.5461036284990698e-05, + "loss": 0.9078, + "step": 13998 + }, + { + "epoch": 0.34, + "learning_rate": 1.5460383774859495e-05, + "loss": 1.0873, + "step": 13999 + }, + { + "epoch": 0.34, + "learning_rate": 1.5459731231601425e-05, + "loss": 1.1521, + "step": 14000 + }, + { + "epoch": 0.34, + "learning_rate": 1.5459078655220455e-05, + "loss": 1.0355, + "step": 14001 + }, + { + "epoch": 0.34, + "learning_rate": 1.5458426045720533e-05, + "loss": 0.9243, + "step": 14002 + }, + { + "epoch": 0.34, + "learning_rate": 1.5457773403105628e-05, + "loss": 0.8103, + "step": 14003 + }, + { + "epoch": 0.34, + "learning_rate": 1.5457120727379694e-05, + "loss": 0.9232, + "step": 14004 + }, + { + "epoch": 0.34, + "learning_rate": 1.545646801854669e-05, + "loss": 1.1908, + "step": 14005 + }, + { + "epoch": 0.34, + "learning_rate": 1.5455815276610586e-05, + "loss": 1.2024, + "step": 14006 + }, + { + "epoch": 0.34, + "learning_rate": 1.5455162501575324e-05, + "loss": 1.1067, + "step": 14007 + }, + { + "epoch": 0.34, + "learning_rate": 1.5454509693444878e-05, + "loss": 1.2721, + "step": 14008 + }, + { + "epoch": 0.34, + "learning_rate": 1.5453856852223207e-05, + "loss": 1.2621, + "step": 14009 + }, + { + "epoch": 0.34, + "learning_rate": 1.5453203977914267e-05, + "loss": 1.0388, + "step": 14010 + }, + { + "epoch": 0.34, + "learning_rate": 1.5452551070522023e-05, + "loss": 1.1262, + "step": 14011 + }, + { + "epoch": 0.34, + "learning_rate": 1.545189813005043e-05, + "loss": 1.2771, + "step": 14012 + }, + { + "epoch": 0.34, + "learning_rate": 1.5451245156503457e-05, + "loss": 1.1587, + "step": 14013 + }, + { + "epoch": 0.34, + "learning_rate": 1.545059214988506e-05, + "loss": 1.2347, + "step": 14014 + }, + { + "epoch": 0.34, + "learning_rate": 1.5449939110199206e-05, + "loss": 1.1887, + "step": 14015 + }, + { + "epoch": 0.34, + "learning_rate": 1.544928603744985e-05, + "loss": 1.1744, + "step": 14016 + }, + { + "epoch": 0.34, + "learning_rate": 1.5448632931640958e-05, + "loss": 1.1296, + "step": 14017 + }, + { + "epoch": 0.34, + "learning_rate": 1.5447979792776493e-05, + "loss": 1.2708, + "step": 14018 + }, + { + "epoch": 0.34, + "learning_rate": 1.5447326620860413e-05, + "loss": 1.1068, + "step": 14019 + }, + { + "epoch": 0.34, + "learning_rate": 1.5446673415896686e-05, + "loss": 1.3169, + "step": 14020 + }, + { + "epoch": 0.34, + "learning_rate": 1.5446020177889274e-05, + "loss": 0.956, + "step": 14021 + }, + { + "epoch": 0.34, + "learning_rate": 1.5445366906842137e-05, + "loss": 1.0667, + "step": 14022 + }, + { + "epoch": 0.34, + "learning_rate": 1.544471360275924e-05, + "loss": 1.1334, + "step": 14023 + }, + { + "epoch": 0.34, + "learning_rate": 1.5444060265644546e-05, + "loss": 0.9238, + "step": 14024 + }, + { + "epoch": 0.34, + "learning_rate": 1.5443406895502017e-05, + "loss": 1.0585, + "step": 14025 + }, + { + "epoch": 0.34, + "learning_rate": 1.544275349233562e-05, + "loss": 0.7408, + "step": 14026 + }, + { + "epoch": 0.34, + "learning_rate": 1.544210005614932e-05, + "loss": 1.1876, + "step": 14027 + }, + { + "epoch": 0.34, + "learning_rate": 1.5441446586947076e-05, + "loss": 1.1623, + "step": 14028 + }, + { + "epoch": 0.34, + "learning_rate": 1.5440793084732857e-05, + "loss": 1.0237, + "step": 14029 + }, + { + "epoch": 0.34, + "learning_rate": 1.5440139549510624e-05, + "loss": 0.9154, + "step": 14030 + }, + { + "epoch": 0.34, + "learning_rate": 1.5439485981284348e-05, + "loss": 1.2185, + "step": 14031 + }, + { + "epoch": 0.34, + "learning_rate": 1.5438832380057987e-05, + "loss": 1.0558, + "step": 14032 + }, + { + "epoch": 0.34, + "learning_rate": 1.543817874583551e-05, + "loss": 1.1175, + "step": 14033 + }, + { + "epoch": 0.34, + "learning_rate": 1.543752507862088e-05, + "loss": 1.1719, + "step": 14034 + }, + { + "epoch": 0.34, + "learning_rate": 1.5436871378418066e-05, + "loss": 1.0603, + "step": 14035 + }, + { + "epoch": 0.34, + "learning_rate": 1.5436217645231037e-05, + "loss": 1.1203, + "step": 14036 + }, + { + "epoch": 0.34, + "learning_rate": 1.5435563879063746e-05, + "loss": 1.1305, + "step": 14037 + }, + { + "epoch": 0.34, + "learning_rate": 1.5434910079920172e-05, + "loss": 1.1216, + "step": 14038 + }, + { + "epoch": 0.34, + "learning_rate": 1.5434256247804275e-05, + "loss": 1.1632, + "step": 14039 + }, + { + "epoch": 0.34, + "learning_rate": 1.5433602382720022e-05, + "loss": 1.034, + "step": 14040 + }, + { + "epoch": 0.34, + "learning_rate": 1.5432948484671386e-05, + "loss": 1.0177, + "step": 14041 + }, + { + "epoch": 0.34, + "learning_rate": 1.5432294553662325e-05, + "loss": 1.0905, + "step": 14042 + }, + { + "epoch": 0.34, + "learning_rate": 1.5431640589696812e-05, + "loss": 0.7599, + "step": 14043 + }, + { + "epoch": 0.34, + "learning_rate": 1.5430986592778817e-05, + "loss": 1.057, + "step": 14044 + }, + { + "epoch": 0.34, + "learning_rate": 1.54303325629123e-05, + "loss": 1.328, + "step": 14045 + }, + { + "epoch": 0.34, + "learning_rate": 1.5429678500101234e-05, + "loss": 1.2217, + "step": 14046 + }, + { + "epoch": 0.34, + "learning_rate": 1.5429024404349585e-05, + "loss": 0.9568, + "step": 14047 + }, + { + "epoch": 0.34, + "learning_rate": 1.542837027566132e-05, + "loss": 1.2468, + "step": 14048 + }, + { + "epoch": 0.34, + "learning_rate": 1.542771611404041e-05, + "loss": 0.9697, + "step": 14049 + }, + { + "epoch": 0.34, + "learning_rate": 1.5427061919490826e-05, + "loss": 1.0617, + "step": 14050 + }, + { + "epoch": 0.34, + "learning_rate": 1.542640769201653e-05, + "loss": 1.0875, + "step": 14051 + }, + { + "epoch": 0.34, + "learning_rate": 1.54257534316215e-05, + "loss": 1.1312, + "step": 14052 + }, + { + "epoch": 0.34, + "learning_rate": 1.5425099138309696e-05, + "loss": 0.8156, + "step": 14053 + }, + { + "epoch": 0.34, + "learning_rate": 1.5424444812085094e-05, + "loss": 1.4935, + "step": 14054 + }, + { + "epoch": 0.34, + "learning_rate": 1.542379045295166e-05, + "loss": 1.0559, + "step": 14055 + }, + { + "epoch": 0.34, + "learning_rate": 1.5423136060913365e-05, + "loss": 1.2513, + "step": 14056 + }, + { + "epoch": 0.34, + "learning_rate": 1.542248163597418e-05, + "loss": 1.2969, + "step": 14057 + }, + { + "epoch": 0.34, + "learning_rate": 1.5421827178138072e-05, + "loss": 1.0515, + "step": 14058 + }, + { + "epoch": 0.34, + "learning_rate": 1.542117268740902e-05, + "loss": 1.2606, + "step": 14059 + }, + { + "epoch": 0.34, + "learning_rate": 1.542051816379098e-05, + "loss": 1.085, + "step": 14060 + }, + { + "epoch": 0.34, + "learning_rate": 1.5419863607287935e-05, + "loss": 1.2166, + "step": 14061 + }, + { + "epoch": 0.34, + "learning_rate": 1.5419209017903855e-05, + "loss": 1.354, + "step": 14062 + }, + { + "epoch": 0.34, + "learning_rate": 1.5418554395642706e-05, + "loss": 1.3198, + "step": 14063 + }, + { + "epoch": 0.34, + "learning_rate": 1.541789974050846e-05, + "loss": 1.1008, + "step": 14064 + }, + { + "epoch": 0.34, + "learning_rate": 1.5417245052505094e-05, + "loss": 1.3692, + "step": 14065 + }, + { + "epoch": 0.34, + "learning_rate": 1.5416590331636574e-05, + "loss": 0.8999, + "step": 14066 + }, + { + "epoch": 0.34, + "learning_rate": 1.5415935577906878e-05, + "loss": 0.9097, + "step": 14067 + }, + { + "epoch": 0.34, + "learning_rate": 1.541528079131997e-05, + "loss": 0.9058, + "step": 14068 + }, + { + "epoch": 0.34, + "learning_rate": 1.5414625971879832e-05, + "loss": 0.7205, + "step": 14069 + }, + { + "epoch": 0.34, + "learning_rate": 1.541397111959043e-05, + "loss": 1.241, + "step": 14070 + }, + { + "epoch": 0.34, + "learning_rate": 1.541331623445574e-05, + "loss": 0.9314, + "step": 14071 + }, + { + "epoch": 0.34, + "learning_rate": 1.5412661316479728e-05, + "loss": 1.4122, + "step": 14072 + }, + { + "epoch": 0.34, + "learning_rate": 1.5412006365666377e-05, + "loss": 0.9864, + "step": 14073 + }, + { + "epoch": 0.34, + "learning_rate": 1.5411351382019657e-05, + "loss": 1.114, + "step": 14074 + }, + { + "epoch": 0.34, + "learning_rate": 1.541069636554354e-05, + "loss": 0.8963, + "step": 14075 + }, + { + "epoch": 0.34, + "learning_rate": 1.5410041316242003e-05, + "loss": 1.3644, + "step": 14076 + }, + { + "epoch": 0.34, + "learning_rate": 1.5409386234119016e-05, + "loss": 1.1334, + "step": 14077 + }, + { + "epoch": 0.34, + "learning_rate": 1.5408731119178554e-05, + "loss": 0.9107, + "step": 14078 + }, + { + "epoch": 0.34, + "learning_rate": 1.540807597142459e-05, + "loss": 1.1192, + "step": 14079 + }, + { + "epoch": 0.34, + "learning_rate": 1.5407420790861106e-05, + "loss": 1.0125, + "step": 14080 + }, + { + "epoch": 0.34, + "learning_rate": 1.5406765577492075e-05, + "loss": 1.1664, + "step": 14081 + }, + { + "epoch": 0.34, + "learning_rate": 1.5406110331321463e-05, + "loss": 1.1201, + "step": 14082 + }, + { + "epoch": 0.34, + "learning_rate": 1.540545505235325e-05, + "loss": 1.3389, + "step": 14083 + }, + { + "epoch": 0.34, + "learning_rate": 1.5404799740591415e-05, + "loss": 0.8723, + "step": 14084 + }, + { + "epoch": 0.34, + "learning_rate": 1.5404144396039932e-05, + "loss": 1.2462, + "step": 14085 + }, + { + "epoch": 0.34, + "learning_rate": 1.5403489018702775e-05, + "loss": 1.1879, + "step": 14086 + }, + { + "epoch": 0.34, + "learning_rate": 1.540283360858392e-05, + "loss": 1.2091, + "step": 14087 + }, + { + "epoch": 0.34, + "learning_rate": 1.5402178165687352e-05, + "loss": 1.1933, + "step": 14088 + }, + { + "epoch": 0.34, + "learning_rate": 1.5401522690017032e-05, + "loss": 1.2195, + "step": 14089 + }, + { + "epoch": 0.34, + "learning_rate": 1.5400867181576945e-05, + "loss": 0.9579, + "step": 14090 + }, + { + "epoch": 0.34, + "learning_rate": 1.540021164037107e-05, + "loss": 1.1848, + "step": 14091 + }, + { + "epoch": 0.34, + "learning_rate": 1.5399556066403382e-05, + "loss": 0.9346, + "step": 14092 + }, + { + "epoch": 0.34, + "learning_rate": 1.539890045967785e-05, + "loss": 1.141, + "step": 14093 + }, + { + "epoch": 0.34, + "learning_rate": 1.5398244820198467e-05, + "loss": 0.9566, + "step": 14094 + }, + { + "epoch": 0.34, + "learning_rate": 1.5397589147969198e-05, + "loss": 1.211, + "step": 14095 + }, + { + "epoch": 0.34, + "learning_rate": 1.5396933442994025e-05, + "loss": 1.225, + "step": 14096 + }, + { + "epoch": 0.34, + "learning_rate": 1.5396277705276926e-05, + "loss": 1.0175, + "step": 14097 + }, + { + "epoch": 0.34, + "learning_rate": 1.5395621934821884e-05, + "loss": 1.187, + "step": 14098 + }, + { + "epoch": 0.34, + "learning_rate": 1.5394966131632872e-05, + "loss": 1.2477, + "step": 14099 + }, + { + "epoch": 0.34, + "learning_rate": 1.5394310295713862e-05, + "loss": 1.3237, + "step": 14100 + }, + { + "epoch": 0.34, + "learning_rate": 1.539365442706885e-05, + "loss": 1.1472, + "step": 14101 + }, + { + "epoch": 0.34, + "learning_rate": 1.53929985257018e-05, + "loss": 1.019, + "step": 14102 + }, + { + "epoch": 0.34, + "learning_rate": 1.5392342591616698e-05, + "loss": 0.9436, + "step": 14103 + }, + { + "epoch": 0.34, + "learning_rate": 1.5391686624817518e-05, + "loss": 1.2449, + "step": 14104 + }, + { + "epoch": 0.34, + "learning_rate": 1.539103062530825e-05, + "loss": 1.1009, + "step": 14105 + }, + { + "epoch": 0.34, + "learning_rate": 1.5390374593092862e-05, + "loss": 1.1143, + "step": 14106 + }, + { + "epoch": 0.34, + "learning_rate": 1.5389718528175345e-05, + "loss": 1.1106, + "step": 14107 + }, + { + "epoch": 0.34, + "learning_rate": 1.538906243055967e-05, + "loss": 1.0958, + "step": 14108 + }, + { + "epoch": 0.34, + "learning_rate": 1.5388406300249815e-05, + "loss": 1.0304, + "step": 14109 + }, + { + "epoch": 0.34, + "learning_rate": 1.5387750137249775e-05, + "loss": 1.1579, + "step": 14110 + }, + { + "epoch": 0.34, + "learning_rate": 1.538709394156352e-05, + "loss": 0.9843, + "step": 14111 + }, + { + "epoch": 0.34, + "learning_rate": 1.538643771319503e-05, + "loss": 1.0422, + "step": 14112 + }, + { + "epoch": 0.34, + "learning_rate": 1.538578145214829e-05, + "loss": 1.2707, + "step": 14113 + }, + { + "epoch": 0.34, + "learning_rate": 1.5385125158427285e-05, + "loss": 0.9554, + "step": 14114 + }, + { + "epoch": 0.34, + "learning_rate": 1.538446883203599e-05, + "loss": 1.2066, + "step": 14115 + }, + { + "epoch": 0.34, + "learning_rate": 1.5383812472978386e-05, + "loss": 1.2098, + "step": 14116 + }, + { + "epoch": 0.34, + "learning_rate": 1.538315608125846e-05, + "loss": 1.2711, + "step": 14117 + }, + { + "epoch": 0.34, + "learning_rate": 1.5382499656880198e-05, + "loss": 1.1958, + "step": 14118 + }, + { + "epoch": 0.34, + "learning_rate": 1.538184319984757e-05, + "loss": 1.1937, + "step": 14119 + }, + { + "epoch": 0.34, + "learning_rate": 1.5381186710164565e-05, + "loss": 0.9943, + "step": 14120 + }, + { + "epoch": 0.34, + "learning_rate": 1.538053018783517e-05, + "loss": 1.1125, + "step": 14121 + }, + { + "epoch": 0.34, + "learning_rate": 1.537987363286336e-05, + "loss": 1.0737, + "step": 14122 + }, + { + "epoch": 0.34, + "learning_rate": 1.5379217045253128e-05, + "loss": 0.8508, + "step": 14123 + }, + { + "epoch": 0.34, + "learning_rate": 1.5378560425008444e-05, + "loss": 1.3145, + "step": 14124 + }, + { + "epoch": 0.34, + "learning_rate": 1.5377903772133302e-05, + "loss": 1.1387, + "step": 14125 + }, + { + "epoch": 0.34, + "learning_rate": 1.537724708663168e-05, + "loss": 1.0849, + "step": 14126 + }, + { + "epoch": 0.34, + "learning_rate": 1.5376590368507573e-05, + "loss": 1.1821, + "step": 14127 + }, + { + "epoch": 0.34, + "learning_rate": 1.537593361776495e-05, + "loss": 0.8707, + "step": 14128 + }, + { + "epoch": 0.34, + "learning_rate": 1.5375276834407805e-05, + "loss": 1.2957, + "step": 14129 + }, + { + "epoch": 0.34, + "learning_rate": 1.537462001844012e-05, + "loss": 1.0334, + "step": 14130 + }, + { + "epoch": 0.34, + "learning_rate": 1.5373963169865875e-05, + "loss": 1.1108, + "step": 14131 + }, + { + "epoch": 0.34, + "learning_rate": 1.5373306288689063e-05, + "loss": 1.248, + "step": 14132 + }, + { + "epoch": 0.34, + "learning_rate": 1.5372649374913665e-05, + "loss": 1.0675, + "step": 14133 + }, + { + "epoch": 0.34, + "learning_rate": 1.5371992428543668e-05, + "loss": 1.1132, + "step": 14134 + }, + { + "epoch": 0.34, + "learning_rate": 1.5371335449583054e-05, + "loss": 1.1353, + "step": 14135 + }, + { + "epoch": 0.34, + "learning_rate": 1.5370678438035813e-05, + "loss": 1.2662, + "step": 14136 + }, + { + "epoch": 0.34, + "learning_rate": 1.537002139390593e-05, + "loss": 1.1663, + "step": 14137 + }, + { + "epoch": 0.34, + "learning_rate": 1.536936431719739e-05, + "loss": 0.9543, + "step": 14138 + }, + { + "epoch": 0.34, + "learning_rate": 1.536870720791418e-05, + "loss": 1.2983, + "step": 14139 + }, + { + "epoch": 0.34, + "learning_rate": 1.5368050066060285e-05, + "loss": 1.0797, + "step": 14140 + }, + { + "epoch": 0.34, + "learning_rate": 1.536739289163969e-05, + "loss": 1.025, + "step": 14141 + }, + { + "epoch": 0.34, + "learning_rate": 1.536673568465639e-05, + "loss": 1.2172, + "step": 14142 + }, + { + "epoch": 0.34, + "learning_rate": 1.536607844511437e-05, + "loss": 1.3075, + "step": 14143 + }, + { + "epoch": 0.34, + "learning_rate": 1.5365421173017604e-05, + "loss": 1.1196, + "step": 14144 + }, + { + "epoch": 0.34, + "learning_rate": 1.5364763868370095e-05, + "loss": 0.9576, + "step": 14145 + }, + { + "epoch": 0.34, + "learning_rate": 1.536410653117583e-05, + "loss": 1.1991, + "step": 14146 + }, + { + "epoch": 0.34, + "learning_rate": 1.5363449161438788e-05, + "loss": 1.1269, + "step": 14147 + }, + { + "epoch": 0.34, + "learning_rate": 1.536279175916296e-05, + "loss": 1.2813, + "step": 14148 + }, + { + "epoch": 0.34, + "learning_rate": 1.5362134324352337e-05, + "loss": 1.041, + "step": 14149 + }, + { + "epoch": 0.34, + "learning_rate": 1.5361476857010907e-05, + "loss": 1.239, + "step": 14150 + }, + { + "epoch": 0.34, + "learning_rate": 1.536081935714266e-05, + "loss": 1.0375, + "step": 14151 + }, + { + "epoch": 0.34, + "learning_rate": 1.5360161824751575e-05, + "loss": 0.9703, + "step": 14152 + }, + { + "epoch": 0.34, + "learning_rate": 1.5359504259841656e-05, + "loss": 1.1877, + "step": 14153 + }, + { + "epoch": 0.34, + "learning_rate": 1.5358846662416884e-05, + "loss": 0.8379, + "step": 14154 + }, + { + "epoch": 0.34, + "learning_rate": 1.5358189032481252e-05, + "loss": 1.1534, + "step": 14155 + }, + { + "epoch": 0.34, + "learning_rate": 1.535753137003874e-05, + "loss": 1.1877, + "step": 14156 + }, + { + "epoch": 0.34, + "learning_rate": 1.5356873675093355e-05, + "loss": 1.0539, + "step": 14157 + }, + { + "epoch": 0.34, + "learning_rate": 1.5356215947649072e-05, + "loss": 1.3668, + "step": 14158 + }, + { + "epoch": 0.34, + "learning_rate": 1.5355558187709888e-05, + "loss": 0.872, + "step": 14159 + }, + { + "epoch": 0.34, + "learning_rate": 1.535490039527979e-05, + "loss": 1.1366, + "step": 14160 + }, + { + "epoch": 0.34, + "learning_rate": 1.5354242570362772e-05, + "loss": 1.1536, + "step": 14161 + }, + { + "epoch": 0.34, + "learning_rate": 1.5353584712962825e-05, + "loss": 1.0053, + "step": 14162 + }, + { + "epoch": 0.34, + "learning_rate": 1.535292682308394e-05, + "loss": 0.9571, + "step": 14163 + }, + { + "epoch": 0.34, + "learning_rate": 1.5352268900730104e-05, + "loss": 1.1584, + "step": 14164 + }, + { + "epoch": 0.34, + "learning_rate": 1.5351610945905312e-05, + "loss": 0.9876, + "step": 14165 + }, + { + "epoch": 0.34, + "learning_rate": 1.5350952958613558e-05, + "loss": 1.1232, + "step": 14166 + }, + { + "epoch": 0.34, + "learning_rate": 1.5350294938858828e-05, + "loss": 1.1726, + "step": 14167 + }, + { + "epoch": 0.34, + "learning_rate": 1.5349636886645115e-05, + "loss": 0.9575, + "step": 14168 + }, + { + "epoch": 0.34, + "learning_rate": 1.5348978801976417e-05, + "loss": 1.1405, + "step": 14169 + }, + { + "epoch": 0.34, + "learning_rate": 1.534832068485672e-05, + "loss": 1.4522, + "step": 14170 + }, + { + "epoch": 0.34, + "learning_rate": 1.5347662535290025e-05, + "loss": 1.2551, + "step": 14171 + }, + { + "epoch": 0.34, + "learning_rate": 1.5347004353280313e-05, + "loss": 1.2042, + "step": 14172 + }, + { + "epoch": 0.34, + "learning_rate": 1.534634613883159e-05, + "loss": 1.077, + "step": 14173 + }, + { + "epoch": 0.34, + "learning_rate": 1.5345687891947837e-05, + "loss": 1.1936, + "step": 14174 + }, + { + "epoch": 0.34, + "learning_rate": 1.5345029612633055e-05, + "loss": 1.0831, + "step": 14175 + }, + { + "epoch": 0.34, + "learning_rate": 1.534437130089123e-05, + "loss": 1.0025, + "step": 14176 + }, + { + "epoch": 0.34, + "learning_rate": 1.534371295672637e-05, + "loss": 1.1213, + "step": 14177 + }, + { + "epoch": 0.34, + "learning_rate": 1.5343054580142457e-05, + "loss": 0.7851, + "step": 14178 + }, + { + "epoch": 0.34, + "learning_rate": 1.5342396171143488e-05, + "loss": 1.1799, + "step": 14179 + }, + { + "epoch": 0.34, + "learning_rate": 1.5341737729733453e-05, + "loss": 1.2484, + "step": 14180 + }, + { + "epoch": 0.34, + "learning_rate": 1.534107925591636e-05, + "loss": 1.0276, + "step": 14181 + }, + { + "epoch": 0.34, + "learning_rate": 1.5340420749696194e-05, + "loss": 0.8598, + "step": 14182 + }, + { + "epoch": 0.34, + "learning_rate": 1.533976221107695e-05, + "loss": 1.2938, + "step": 14183 + }, + { + "epoch": 0.34, + "learning_rate": 1.5339103640062624e-05, + "loss": 1.2509, + "step": 14184 + }, + { + "epoch": 0.34, + "learning_rate": 1.5338445036657213e-05, + "loss": 1.3995, + "step": 14185 + }, + { + "epoch": 0.34, + "learning_rate": 1.5337786400864708e-05, + "loss": 1.0957, + "step": 14186 + }, + { + "epoch": 0.34, + "learning_rate": 1.5337127732689113e-05, + "loss": 1.0686, + "step": 14187 + }, + { + "epoch": 0.34, + "learning_rate": 1.533646903213442e-05, + "loss": 0.957, + "step": 14188 + }, + { + "epoch": 0.34, + "learning_rate": 1.533581029920462e-05, + "loss": 1.2302, + "step": 14189 + }, + { + "epoch": 0.34, + "learning_rate": 1.533515153390372e-05, + "loss": 0.8978, + "step": 14190 + }, + { + "epoch": 0.34, + "learning_rate": 1.5334492736235703e-05, + "loss": 1.1847, + "step": 14191 + }, + { + "epoch": 0.34, + "learning_rate": 1.533383390620458e-05, + "loss": 1.2332, + "step": 14192 + }, + { + "epoch": 0.34, + "learning_rate": 1.533317504381434e-05, + "loss": 1.0144, + "step": 14193 + }, + { + "epoch": 0.34, + "learning_rate": 1.533251614906898e-05, + "loss": 1.0598, + "step": 14194 + }, + { + "epoch": 0.34, + "learning_rate": 1.5331857221972496e-05, + "loss": 1.1692, + "step": 14195 + }, + { + "epoch": 0.34, + "learning_rate": 1.5331198262528895e-05, + "loss": 1.23, + "step": 14196 + }, + { + "epoch": 0.34, + "learning_rate": 1.5330539270742164e-05, + "loss": 1.1421, + "step": 14197 + }, + { + "epoch": 0.34, + "learning_rate": 1.5329880246616308e-05, + "loss": 0.6737, + "step": 14198 + }, + { + "epoch": 0.34, + "learning_rate": 1.5329221190155315e-05, + "loss": 1.049, + "step": 14199 + }, + { + "epoch": 0.34, + "learning_rate": 1.5328562101363196e-05, + "loss": 0.9105, + "step": 14200 + }, + { + "epoch": 0.34, + "learning_rate": 1.5327902980243943e-05, + "loss": 1.2751, + "step": 14201 + }, + { + "epoch": 0.34, + "learning_rate": 1.5327243826801557e-05, + "loss": 1.1944, + "step": 14202 + }, + { + "epoch": 0.34, + "learning_rate": 1.5326584641040033e-05, + "loss": 1.2935, + "step": 14203 + }, + { + "epoch": 0.34, + "learning_rate": 1.5325925422963378e-05, + "loss": 1.0201, + "step": 14204 + }, + { + "epoch": 0.34, + "learning_rate": 1.5325266172575584e-05, + "loss": 1.1224, + "step": 14205 + }, + { + "epoch": 0.34, + "learning_rate": 1.532460688988065e-05, + "loss": 0.9291, + "step": 14206 + }, + { + "epoch": 0.34, + "learning_rate": 1.5323947574882582e-05, + "loss": 1.1542, + "step": 14207 + }, + { + "epoch": 0.34, + "learning_rate": 1.5323288227585373e-05, + "loss": 1.4637, + "step": 14208 + }, + { + "epoch": 0.34, + "learning_rate": 1.5322628847993026e-05, + "loss": 1.2431, + "step": 14209 + }, + { + "epoch": 0.34, + "learning_rate": 1.5321969436109545e-05, + "loss": 1.2478, + "step": 14210 + }, + { + "epoch": 0.34, + "learning_rate": 1.5321309991938926e-05, + "loss": 1.2019, + "step": 14211 + }, + { + "epoch": 0.34, + "learning_rate": 1.532065051548517e-05, + "loss": 1.2575, + "step": 14212 + }, + { + "epoch": 0.34, + "learning_rate": 1.531999100675228e-05, + "loss": 1.0968, + "step": 14213 + }, + { + "epoch": 0.34, + "learning_rate": 1.5319331465744257e-05, + "loss": 0.9446, + "step": 14214 + }, + { + "epoch": 0.34, + "learning_rate": 1.5318671892465098e-05, + "loss": 1.1269, + "step": 14215 + }, + { + "epoch": 0.34, + "learning_rate": 1.531801228691881e-05, + "loss": 1.2015, + "step": 14216 + }, + { + "epoch": 0.34, + "learning_rate": 1.5317352649109393e-05, + "loss": 1.0186, + "step": 14217 + }, + { + "epoch": 0.34, + "learning_rate": 1.5316692979040847e-05, + "loss": 1.0516, + "step": 14218 + }, + { + "epoch": 0.34, + "learning_rate": 1.5316033276717174e-05, + "loss": 1.0771, + "step": 14219 + }, + { + "epoch": 0.34, + "learning_rate": 1.531537354214238e-05, + "loss": 1.1767, + "step": 14220 + }, + { + "epoch": 0.34, + "learning_rate": 1.531471377532046e-05, + "loss": 1.1074, + "step": 14221 + }, + { + "epoch": 0.34, + "learning_rate": 1.5314053976255427e-05, + "loss": 1.3194, + "step": 14222 + }, + { + "epoch": 0.34, + "learning_rate": 1.5313394144951276e-05, + "loss": 1.086, + "step": 14223 + }, + { + "epoch": 0.34, + "learning_rate": 1.5312734281412016e-05, + "loss": 0.9735, + "step": 14224 + }, + { + "epoch": 0.34, + "learning_rate": 1.5312074385641644e-05, + "loss": 1.0152, + "step": 14225 + }, + { + "epoch": 0.34, + "learning_rate": 1.5311414457644165e-05, + "loss": 1.1161, + "step": 14226 + }, + { + "epoch": 0.34, + "learning_rate": 1.5310754497423585e-05, + "loss": 1.1007, + "step": 14227 + }, + { + "epoch": 0.34, + "learning_rate": 1.5310094504983907e-05, + "loss": 1.0949, + "step": 14228 + }, + { + "epoch": 0.34, + "learning_rate": 1.5309434480329134e-05, + "loss": 1.2228, + "step": 14229 + }, + { + "epoch": 0.34, + "learning_rate": 1.530877442346327e-05, + "loss": 1.2682, + "step": 14230 + }, + { + "epoch": 0.34, + "learning_rate": 1.5308114334390324e-05, + "loss": 1.2826, + "step": 14231 + }, + { + "epoch": 0.34, + "learning_rate": 1.5307454213114292e-05, + "loss": 1.1055, + "step": 14232 + }, + { + "epoch": 0.34, + "learning_rate": 1.5306794059639187e-05, + "loss": 1.1423, + "step": 14233 + }, + { + "epoch": 0.34, + "learning_rate": 1.5306133873969008e-05, + "loss": 0.9543, + "step": 14234 + }, + { + "epoch": 0.34, + "learning_rate": 1.5305473656107764e-05, + "loss": 1.3012, + "step": 14235 + }, + { + "epoch": 0.34, + "learning_rate": 1.5304813406059462e-05, + "loss": 1.152, + "step": 14236 + }, + { + "epoch": 0.34, + "learning_rate": 1.5304153123828102e-05, + "loss": 1.2002, + "step": 14237 + }, + { + "epoch": 0.34, + "learning_rate": 1.5303492809417693e-05, + "loss": 1.2357, + "step": 14238 + }, + { + "epoch": 0.34, + "learning_rate": 1.530283246283224e-05, + "loss": 0.9965, + "step": 14239 + }, + { + "epoch": 0.34, + "learning_rate": 1.5302172084075752e-05, + "loss": 1.1836, + "step": 14240 + }, + { + "epoch": 0.34, + "learning_rate": 1.530151167315223e-05, + "loss": 1.1666, + "step": 14241 + }, + { + "epoch": 0.34, + "learning_rate": 1.530085123006569e-05, + "loss": 1.2076, + "step": 14242 + }, + { + "epoch": 0.34, + "learning_rate": 1.5300190754820125e-05, + "loss": 0.9891, + "step": 14243 + }, + { + "epoch": 0.34, + "learning_rate": 1.529953024741955e-05, + "loss": 1.3062, + "step": 14244 + }, + { + "epoch": 0.34, + "learning_rate": 1.5298869707867977e-05, + "loss": 0.8854, + "step": 14245 + }, + { + "epoch": 0.34, + "learning_rate": 1.5298209136169403e-05, + "loss": 1.0742, + "step": 14246 + }, + { + "epoch": 0.34, + "learning_rate": 1.5297548532327842e-05, + "loss": 1.3117, + "step": 14247 + }, + { + "epoch": 0.34, + "learning_rate": 1.5296887896347297e-05, + "loss": 1.0247, + "step": 14248 + }, + { + "epoch": 0.34, + "learning_rate": 1.5296227228231786e-05, + "loss": 1.218, + "step": 14249 + }, + { + "epoch": 0.34, + "learning_rate": 1.5295566527985306e-05, + "loss": 0.9509, + "step": 14250 + }, + { + "epoch": 0.34, + "learning_rate": 1.5294905795611866e-05, + "loss": 0.876, + "step": 14251 + }, + { + "epoch": 0.34, + "learning_rate": 1.5294245031115484e-05, + "loss": 0.9993, + "step": 14252 + }, + { + "epoch": 0.34, + "learning_rate": 1.5293584234500162e-05, + "loss": 1.0453, + "step": 14253 + }, + { + "epoch": 0.34, + "learning_rate": 1.5292923405769907e-05, + "loss": 1.2829, + "step": 14254 + }, + { + "epoch": 0.34, + "learning_rate": 1.529226254492873e-05, + "loss": 0.9647, + "step": 14255 + }, + { + "epoch": 0.34, + "learning_rate": 1.5291601651980644e-05, + "loss": 0.7931, + "step": 14256 + }, + { + "epoch": 0.34, + "learning_rate": 1.5290940726929654e-05, + "loss": 1.0441, + "step": 14257 + }, + { + "epoch": 0.34, + "learning_rate": 1.529027976977977e-05, + "loss": 1.0098, + "step": 14258 + }, + { + "epoch": 0.34, + "learning_rate": 1.5289618780535006e-05, + "loss": 1.0076, + "step": 14259 + }, + { + "epoch": 0.34, + "learning_rate": 1.5288957759199366e-05, + "loss": 1.2408, + "step": 14260 + }, + { + "epoch": 0.34, + "learning_rate": 1.5288296705776864e-05, + "loss": 1.146, + "step": 14261 + }, + { + "epoch": 0.34, + "learning_rate": 1.5287635620271512e-05, + "loss": 1.2839, + "step": 14262 + }, + { + "epoch": 0.34, + "learning_rate": 1.5286974502687315e-05, + "loss": 1.1435, + "step": 14263 + }, + { + "epoch": 0.34, + "learning_rate": 1.528631335302829e-05, + "loss": 0.7473, + "step": 14264 + }, + { + "epoch": 0.34, + "learning_rate": 1.5285652171298445e-05, + "loss": 0.9964, + "step": 14265 + }, + { + "epoch": 0.34, + "learning_rate": 1.5284990957501788e-05, + "loss": 0.9954, + "step": 14266 + }, + { + "epoch": 0.34, + "learning_rate": 1.5284329711642335e-05, + "loss": 1.0512, + "step": 14267 + }, + { + "epoch": 0.34, + "learning_rate": 1.5283668433724098e-05, + "loss": 1.191, + "step": 14268 + }, + { + "epoch": 0.34, + "learning_rate": 1.5283007123751092e-05, + "loss": 1.1974, + "step": 14269 + }, + { + "epoch": 0.34, + "learning_rate": 1.5282345781727318e-05, + "loss": 1.2413, + "step": 14270 + }, + { + "epoch": 0.34, + "learning_rate": 1.5281684407656797e-05, + "loss": 1.0443, + "step": 14271 + }, + { + "epoch": 0.34, + "learning_rate": 1.528102300154354e-05, + "loss": 1.0647, + "step": 14272 + }, + { + "epoch": 0.34, + "learning_rate": 1.5280361563391558e-05, + "loss": 0.7595, + "step": 14273 + }, + { + "epoch": 0.34, + "learning_rate": 1.527970009320486e-05, + "loss": 1.5141, + "step": 14274 + }, + { + "epoch": 0.34, + "learning_rate": 1.527903859098747e-05, + "loss": 1.1937, + "step": 14275 + }, + { + "epoch": 0.34, + "learning_rate": 1.527837705674339e-05, + "loss": 1.2258, + "step": 14276 + }, + { + "epoch": 0.34, + "learning_rate": 1.5277715490476643e-05, + "loss": 1.2635, + "step": 14277 + }, + { + "epoch": 0.34, + "learning_rate": 1.5277053892191233e-05, + "loss": 0.9095, + "step": 14278 + }, + { + "epoch": 0.34, + "learning_rate": 1.527639226189118e-05, + "loss": 1.1221, + "step": 14279 + }, + { + "epoch": 0.34, + "learning_rate": 1.5275730599580497e-05, + "loss": 1.1738, + "step": 14280 + }, + { + "epoch": 0.34, + "learning_rate": 1.5275068905263194e-05, + "loss": 1.1759, + "step": 14281 + }, + { + "epoch": 0.34, + "learning_rate": 1.527440717894329e-05, + "loss": 0.9117, + "step": 14282 + }, + { + "epoch": 0.34, + "learning_rate": 1.52737454206248e-05, + "loss": 0.8689, + "step": 14283 + }, + { + "epoch": 0.34, + "learning_rate": 1.5273083630311736e-05, + "loss": 1.2375, + "step": 14284 + }, + { + "epoch": 0.34, + "learning_rate": 1.5272421808008117e-05, + "loss": 1.0017, + "step": 14285 + }, + { + "epoch": 0.34, + "learning_rate": 1.527175995371795e-05, + "loss": 1.3252, + "step": 14286 + }, + { + "epoch": 0.34, + "learning_rate": 1.5271098067445257e-05, + "loss": 1.2812, + "step": 14287 + }, + { + "epoch": 0.34, + "learning_rate": 1.5270436149194053e-05, + "loss": 0.929, + "step": 14288 + }, + { + "epoch": 0.34, + "learning_rate": 1.5269774198968354e-05, + "loss": 0.9355, + "step": 14289 + }, + { + "epoch": 0.34, + "learning_rate": 1.526911221677217e-05, + "loss": 1.0047, + "step": 14290 + }, + { + "epoch": 0.34, + "learning_rate": 1.526845020260952e-05, + "loss": 1.0618, + "step": 14291 + }, + { + "epoch": 0.34, + "learning_rate": 1.5267788156484428e-05, + "loss": 1.1495, + "step": 14292 + }, + { + "epoch": 0.34, + "learning_rate": 1.52671260784009e-05, + "loss": 0.9112, + "step": 14293 + }, + { + "epoch": 0.34, + "learning_rate": 1.526646396836296e-05, + "loss": 1.1436, + "step": 14294 + }, + { + "epoch": 0.34, + "learning_rate": 1.526580182637462e-05, + "loss": 1.0297, + "step": 14295 + }, + { + "epoch": 0.34, + "learning_rate": 1.52651396524399e-05, + "loss": 1.1316, + "step": 14296 + }, + { + "epoch": 0.34, + "learning_rate": 1.5264477446562817e-05, + "loss": 1.0529, + "step": 14297 + }, + { + "epoch": 0.34, + "learning_rate": 1.526381520874739e-05, + "loss": 1.1187, + "step": 14298 + }, + { + "epoch": 0.34, + "learning_rate": 1.5263152938997627e-05, + "loss": 1.0255, + "step": 14299 + }, + { + "epoch": 0.34, + "learning_rate": 1.5262490637317555e-05, + "loss": 1.0986, + "step": 14300 + }, + { + "epoch": 0.34, + "learning_rate": 1.5261828303711195e-05, + "loss": 0.9053, + "step": 14301 + }, + { + "epoch": 0.34, + "learning_rate": 1.5261165938182555e-05, + "loss": 0.997, + "step": 14302 + }, + { + "epoch": 0.34, + "learning_rate": 1.526050354073566e-05, + "loss": 1.0368, + "step": 14303 + }, + { + "epoch": 0.34, + "learning_rate": 1.525984111137453e-05, + "loss": 1.2039, + "step": 14304 + }, + { + "epoch": 0.34, + "learning_rate": 1.525917865010318e-05, + "loss": 1.3474, + "step": 14305 + }, + { + "epoch": 0.34, + "learning_rate": 1.5258516156925627e-05, + "loss": 1.1696, + "step": 14306 + }, + { + "epoch": 0.34, + "learning_rate": 1.5257853631845899e-05, + "loss": 0.9144, + "step": 14307 + }, + { + "epoch": 0.34, + "learning_rate": 1.5257191074868004e-05, + "loss": 1.1721, + "step": 14308 + }, + { + "epoch": 0.34, + "learning_rate": 1.5256528485995974e-05, + "loss": 0.8334, + "step": 14309 + }, + { + "epoch": 0.34, + "learning_rate": 1.5255865865233817e-05, + "loss": 1.2221, + "step": 14310 + }, + { + "epoch": 0.34, + "learning_rate": 1.5255203212585558e-05, + "loss": 0.917, + "step": 14311 + }, + { + "epoch": 0.34, + "learning_rate": 1.5254540528055219e-05, + "loss": 1.0501, + "step": 14312 + }, + { + "epoch": 0.34, + "learning_rate": 1.5253877811646818e-05, + "loss": 1.0022, + "step": 14313 + }, + { + "epoch": 0.34, + "learning_rate": 1.5253215063364378e-05, + "loss": 1.0728, + "step": 14314 + }, + { + "epoch": 0.34, + "learning_rate": 1.5252552283211914e-05, + "loss": 1.0827, + "step": 14315 + }, + { + "epoch": 0.34, + "learning_rate": 1.5251889471193453e-05, + "loss": 1.2872, + "step": 14316 + }, + { + "epoch": 0.34, + "learning_rate": 1.5251226627313016e-05, + "loss": 0.7596, + "step": 14317 + }, + { + "epoch": 0.34, + "learning_rate": 1.525056375157462e-05, + "loss": 1.0834, + "step": 14318 + }, + { + "epoch": 0.34, + "learning_rate": 1.5249900843982288e-05, + "loss": 1.2872, + "step": 14319 + }, + { + "epoch": 0.34, + "learning_rate": 1.5249237904540041e-05, + "loss": 1.103, + "step": 14320 + }, + { + "epoch": 0.34, + "learning_rate": 1.5248574933251905e-05, + "loss": 0.9104, + "step": 14321 + }, + { + "epoch": 0.34, + "learning_rate": 1.5247911930121899e-05, + "loss": 1.0941, + "step": 14322 + }, + { + "epoch": 0.34, + "learning_rate": 1.5247248895154046e-05, + "loss": 1.2453, + "step": 14323 + }, + { + "epoch": 0.34, + "learning_rate": 1.5246585828352367e-05, + "loss": 1.2777, + "step": 14324 + }, + { + "epoch": 0.34, + "learning_rate": 1.5245922729720887e-05, + "loss": 1.1612, + "step": 14325 + }, + { + "epoch": 0.34, + "learning_rate": 1.5245259599263627e-05, + "loss": 1.1208, + "step": 14326 + }, + { + "epoch": 0.34, + "learning_rate": 1.5244596436984611e-05, + "loss": 0.9646, + "step": 14327 + }, + { + "epoch": 0.34, + "learning_rate": 1.5243933242887863e-05, + "loss": 1.2238, + "step": 14328 + }, + { + "epoch": 0.34, + "learning_rate": 1.5243270016977405e-05, + "loss": 1.2041, + "step": 14329 + }, + { + "epoch": 0.34, + "learning_rate": 1.524260675925726e-05, + "loss": 0.9727, + "step": 14330 + }, + { + "epoch": 0.34, + "learning_rate": 1.5241943469731453e-05, + "loss": 1.0409, + "step": 14331 + }, + { + "epoch": 0.34, + "learning_rate": 1.5241280148404009e-05, + "loss": 1.2459, + "step": 14332 + }, + { + "epoch": 0.34, + "learning_rate": 1.5240616795278951e-05, + "loss": 1.0494, + "step": 14333 + }, + { + "epoch": 0.34, + "learning_rate": 1.5239953410360302e-05, + "loss": 1.0737, + "step": 14334 + }, + { + "epoch": 0.34, + "learning_rate": 1.523928999365209e-05, + "loss": 1.176, + "step": 14335 + }, + { + "epoch": 0.34, + "learning_rate": 1.523862654515834e-05, + "loss": 1.2062, + "step": 14336 + }, + { + "epoch": 0.34, + "learning_rate": 1.5237963064883072e-05, + "loss": 0.948, + "step": 14337 + }, + { + "epoch": 0.34, + "learning_rate": 1.5237299552830313e-05, + "loss": 1.2, + "step": 14338 + }, + { + "epoch": 0.34, + "learning_rate": 1.5236636009004092e-05, + "loss": 1.4403, + "step": 14339 + }, + { + "epoch": 0.34, + "learning_rate": 1.523597243340843e-05, + "loss": 1.2346, + "step": 14340 + }, + { + "epoch": 0.34, + "learning_rate": 1.5235308826047357e-05, + "loss": 1.1193, + "step": 14341 + }, + { + "epoch": 0.34, + "learning_rate": 1.5234645186924892e-05, + "loss": 0.8738, + "step": 14342 + }, + { + "epoch": 0.34, + "learning_rate": 1.523398151604507e-05, + "loss": 1.0555, + "step": 14343 + }, + { + "epoch": 0.34, + "learning_rate": 1.5233317813411916e-05, + "loss": 1.1939, + "step": 14344 + }, + { + "epoch": 0.34, + "learning_rate": 1.523265407902945e-05, + "loss": 0.9658, + "step": 14345 + }, + { + "epoch": 0.35, + "learning_rate": 1.5231990312901701e-05, + "loss": 0.9661, + "step": 14346 + }, + { + "epoch": 0.35, + "learning_rate": 1.52313265150327e-05, + "loss": 1.0665, + "step": 14347 + }, + { + "epoch": 0.35, + "learning_rate": 1.5230662685426472e-05, + "loss": 1.069, + "step": 14348 + }, + { + "epoch": 0.35, + "learning_rate": 1.5229998824087044e-05, + "loss": 1.3199, + "step": 14349 + }, + { + "epoch": 0.35, + "learning_rate": 1.522933493101844e-05, + "loss": 0.8965, + "step": 14350 + }, + { + "epoch": 0.35, + "learning_rate": 1.5228671006224694e-05, + "loss": 1.0743, + "step": 14351 + }, + { + "epoch": 0.35, + "learning_rate": 1.5228007049709829e-05, + "loss": 0.9646, + "step": 14352 + }, + { + "epoch": 0.35, + "learning_rate": 1.5227343061477875e-05, + "loss": 1.0782, + "step": 14353 + }, + { + "epoch": 0.35, + "learning_rate": 1.5226679041532863e-05, + "loss": 1.0577, + "step": 14354 + }, + { + "epoch": 0.35, + "learning_rate": 1.5226014989878817e-05, + "loss": 1.2046, + "step": 14355 + }, + { + "epoch": 0.35, + "learning_rate": 1.5225350906519766e-05, + "loss": 1.1139, + "step": 14356 + }, + { + "epoch": 0.35, + "learning_rate": 1.5224686791459743e-05, + "loss": 0.9962, + "step": 14357 + }, + { + "epoch": 0.35, + "learning_rate": 1.5224022644702769e-05, + "loss": 0.8232, + "step": 14358 + }, + { + "epoch": 0.35, + "learning_rate": 1.5223358466252881e-05, + "loss": 0.9488, + "step": 14359 + }, + { + "epoch": 0.35, + "learning_rate": 1.522269425611411e-05, + "loss": 1.2097, + "step": 14360 + }, + { + "epoch": 0.35, + "learning_rate": 1.5222030014290477e-05, + "loss": 1.264, + "step": 14361 + }, + { + "epoch": 0.35, + "learning_rate": 1.5221365740786016e-05, + "loss": 1.0283, + "step": 14362 + }, + { + "epoch": 0.35, + "learning_rate": 1.522070143560476e-05, + "loss": 1.007, + "step": 14363 + }, + { + "epoch": 0.35, + "learning_rate": 1.5220037098750733e-05, + "loss": 1.2437, + "step": 14364 + }, + { + "epoch": 0.35, + "learning_rate": 1.521937273022797e-05, + "loss": 1.1852, + "step": 14365 + }, + { + "epoch": 0.35, + "learning_rate": 1.5218708330040502e-05, + "loss": 1.0941, + "step": 14366 + }, + { + "epoch": 0.35, + "learning_rate": 1.5218043898192357e-05, + "loss": 1.3141, + "step": 14367 + }, + { + "epoch": 0.35, + "learning_rate": 1.5217379434687566e-05, + "loss": 1.2029, + "step": 14368 + }, + { + "epoch": 0.35, + "learning_rate": 1.5216714939530161e-05, + "loss": 1.2719, + "step": 14369 + }, + { + "epoch": 0.35, + "learning_rate": 1.5216050412724178e-05, + "loss": 0.9687, + "step": 14370 + }, + { + "epoch": 0.35, + "learning_rate": 1.5215385854273638e-05, + "loss": 0.8876, + "step": 14371 + }, + { + "epoch": 0.35, + "learning_rate": 1.5214721264182584e-05, + "loss": 1.0894, + "step": 14372 + }, + { + "epoch": 0.35, + "learning_rate": 1.5214056642455038e-05, + "loss": 1.3105, + "step": 14373 + }, + { + "epoch": 0.35, + "learning_rate": 1.5213391989095036e-05, + "loss": 1.0636, + "step": 14374 + }, + { + "epoch": 0.35, + "learning_rate": 1.5212727304106615e-05, + "loss": 1.2324, + "step": 14375 + }, + { + "epoch": 0.35, + "learning_rate": 1.5212062587493801e-05, + "loss": 1.0949, + "step": 14376 + }, + { + "epoch": 0.35, + "learning_rate": 1.5211397839260634e-05, + "loss": 0.9963, + "step": 14377 + }, + { + "epoch": 0.35, + "learning_rate": 1.5210733059411135e-05, + "loss": 1.2305, + "step": 14378 + }, + { + "epoch": 0.35, + "learning_rate": 1.521006824794935e-05, + "loss": 1.0997, + "step": 14379 + }, + { + "epoch": 0.35, + "learning_rate": 1.5209403404879305e-05, + "loss": 1.1975, + "step": 14380 + }, + { + "epoch": 0.35, + "learning_rate": 1.5208738530205033e-05, + "loss": 0.9975, + "step": 14381 + }, + { + "epoch": 0.35, + "learning_rate": 1.520807362393057e-05, + "loss": 1.167, + "step": 14382 + }, + { + "epoch": 0.35, + "learning_rate": 1.520740868605995e-05, + "loss": 1.2735, + "step": 14383 + }, + { + "epoch": 0.35, + "learning_rate": 1.5206743716597208e-05, + "loss": 1.131, + "step": 14384 + }, + { + "epoch": 0.35, + "learning_rate": 1.5206078715546374e-05, + "loss": 1.1438, + "step": 14385 + }, + { + "epoch": 0.35, + "learning_rate": 1.5205413682911485e-05, + "loss": 1.019, + "step": 14386 + }, + { + "epoch": 0.35, + "learning_rate": 1.5204748618696576e-05, + "loss": 1.1615, + "step": 14387 + }, + { + "epoch": 0.35, + "learning_rate": 1.5204083522905684e-05, + "loss": 1.1644, + "step": 14388 + }, + { + "epoch": 0.35, + "learning_rate": 1.520341839554284e-05, + "loss": 1.0103, + "step": 14389 + }, + { + "epoch": 0.35, + "learning_rate": 1.5202753236612078e-05, + "loss": 1.1505, + "step": 14390 + }, + { + "epoch": 0.35, + "learning_rate": 1.5202088046117437e-05, + "loss": 0.9805, + "step": 14391 + }, + { + "epoch": 0.35, + "learning_rate": 1.5201422824062952e-05, + "loss": 1.0614, + "step": 14392 + }, + { + "epoch": 0.35, + "learning_rate": 1.5200757570452657e-05, + "loss": 0.9562, + "step": 14393 + }, + { + "epoch": 0.35, + "learning_rate": 1.5200092285290594e-05, + "loss": 1.1301, + "step": 14394 + }, + { + "epoch": 0.35, + "learning_rate": 1.5199426968580788e-05, + "loss": 1.358, + "step": 14395 + }, + { + "epoch": 0.35, + "learning_rate": 1.5198761620327287e-05, + "loss": 1.0673, + "step": 14396 + }, + { + "epoch": 0.35, + "learning_rate": 1.5198096240534116e-05, + "loss": 1.1015, + "step": 14397 + }, + { + "epoch": 0.35, + "learning_rate": 1.5197430829205321e-05, + "loss": 1.1146, + "step": 14398 + }, + { + "epoch": 0.35, + "learning_rate": 1.5196765386344937e-05, + "loss": 0.9376, + "step": 14399 + }, + { + "epoch": 0.35, + "learning_rate": 1.5196099911956998e-05, + "loss": 0.5651, + "step": 14400 + }, + { + "epoch": 0.35, + "learning_rate": 1.5195434406045543e-05, + "loss": 0.8439, + "step": 14401 + }, + { + "epoch": 0.35, + "learning_rate": 1.519476886861461e-05, + "loss": 1.2378, + "step": 14402 + }, + { + "epoch": 0.35, + "learning_rate": 1.5194103299668236e-05, + "loss": 1.1475, + "step": 14403 + }, + { + "epoch": 0.35, + "learning_rate": 1.519343769921046e-05, + "loss": 1.2272, + "step": 14404 + }, + { + "epoch": 0.35, + "learning_rate": 1.5192772067245317e-05, + "loss": 1.2697, + "step": 14405 + }, + { + "epoch": 0.35, + "learning_rate": 1.519210640377685e-05, + "loss": 1.2495, + "step": 14406 + }, + { + "epoch": 0.35, + "learning_rate": 1.519144070880909e-05, + "loss": 1.1705, + "step": 14407 + }, + { + "epoch": 0.35, + "learning_rate": 1.5190774982346085e-05, + "loss": 1.1445, + "step": 14408 + }, + { + "epoch": 0.35, + "learning_rate": 1.5190109224391867e-05, + "loss": 1.1831, + "step": 14409 + }, + { + "epoch": 0.35, + "learning_rate": 1.518944343495048e-05, + "loss": 1.1538, + "step": 14410 + }, + { + "epoch": 0.35, + "learning_rate": 1.5188777614025955e-05, + "loss": 1.0933, + "step": 14411 + }, + { + "epoch": 0.35, + "learning_rate": 1.5188111761622342e-05, + "loss": 1.166, + "step": 14412 + }, + { + "epoch": 0.35, + "learning_rate": 1.5187445877743671e-05, + "loss": 0.9111, + "step": 14413 + }, + { + "epoch": 0.35, + "learning_rate": 1.5186779962393989e-05, + "loss": 1.2796, + "step": 14414 + }, + { + "epoch": 0.35, + "learning_rate": 1.5186114015577334e-05, + "loss": 1.1222, + "step": 14415 + }, + { + "epoch": 0.35, + "learning_rate": 1.5185448037297742e-05, + "loss": 0.958, + "step": 14416 + }, + { + "epoch": 0.35, + "learning_rate": 1.5184782027559255e-05, + "loss": 1.2259, + "step": 14417 + }, + { + "epoch": 0.35, + "learning_rate": 1.518411598636592e-05, + "loss": 0.9717, + "step": 14418 + }, + { + "epoch": 0.35, + "learning_rate": 1.5183449913721769e-05, + "loss": 1.4982, + "step": 14419 + }, + { + "epoch": 0.35, + "learning_rate": 1.5182783809630847e-05, + "loss": 1.3768, + "step": 14420 + }, + { + "epoch": 0.35, + "learning_rate": 1.5182117674097196e-05, + "loss": 1.0992, + "step": 14421 + }, + { + "epoch": 0.35, + "learning_rate": 1.5181451507124856e-05, + "loss": 0.9508, + "step": 14422 + }, + { + "epoch": 0.35, + "learning_rate": 1.5180785308717867e-05, + "loss": 1.2126, + "step": 14423 + }, + { + "epoch": 0.35, + "learning_rate": 1.5180119078880276e-05, + "loss": 1.0693, + "step": 14424 + }, + { + "epoch": 0.35, + "learning_rate": 1.5179452817616116e-05, + "loss": 1.2279, + "step": 14425 + }, + { + "epoch": 0.35, + "learning_rate": 1.5178786524929436e-05, + "loss": 1.1555, + "step": 14426 + }, + { + "epoch": 0.35, + "learning_rate": 1.5178120200824276e-05, + "loss": 0.9624, + "step": 14427 + }, + { + "epoch": 0.35, + "learning_rate": 1.5177453845304681e-05, + "loss": 0.941, + "step": 14428 + }, + { + "epoch": 0.35, + "learning_rate": 1.517678745837469e-05, + "loss": 1.1459, + "step": 14429 + }, + { + "epoch": 0.35, + "learning_rate": 1.5176121040038342e-05, + "loss": 1.0434, + "step": 14430 + }, + { + "epoch": 0.35, + "learning_rate": 1.5175454590299692e-05, + "loss": 1.1174, + "step": 14431 + }, + { + "epoch": 0.35, + "learning_rate": 1.5174788109162775e-05, + "loss": 1.0901, + "step": 14432 + }, + { + "epoch": 0.35, + "learning_rate": 1.5174121596631634e-05, + "loss": 1.2903, + "step": 14433 + }, + { + "epoch": 0.35, + "learning_rate": 1.5173455052710315e-05, + "loss": 1.1872, + "step": 14434 + }, + { + "epoch": 0.35, + "learning_rate": 1.5172788477402862e-05, + "loss": 1.3192, + "step": 14435 + }, + { + "epoch": 0.35, + "learning_rate": 1.5172121870713316e-05, + "loss": 1.312, + "step": 14436 + }, + { + "epoch": 0.35, + "learning_rate": 1.5171455232645724e-05, + "loss": 1.1304, + "step": 14437 + }, + { + "epoch": 0.35, + "learning_rate": 1.517078856320413e-05, + "loss": 1.3586, + "step": 14438 + }, + { + "epoch": 0.35, + "learning_rate": 1.5170121862392577e-05, + "loss": 1.1619, + "step": 14439 + }, + { + "epoch": 0.35, + "learning_rate": 1.5169455130215111e-05, + "loss": 0.9754, + "step": 14440 + }, + { + "epoch": 0.35, + "learning_rate": 1.5168788366675777e-05, + "loss": 1.135, + "step": 14441 + }, + { + "epoch": 0.35, + "learning_rate": 1.5168121571778622e-05, + "loss": 1.1449, + "step": 14442 + }, + { + "epoch": 0.35, + "learning_rate": 1.5167454745527686e-05, + "loss": 1.1282, + "step": 14443 + }, + { + "epoch": 0.35, + "learning_rate": 1.5166787887927017e-05, + "loss": 0.9957, + "step": 14444 + }, + { + "epoch": 0.35, + "learning_rate": 1.5166120998980665e-05, + "loss": 1.2446, + "step": 14445 + }, + { + "epoch": 0.35, + "learning_rate": 1.516545407869267e-05, + "loss": 1.1977, + "step": 14446 + }, + { + "epoch": 0.35, + "learning_rate": 1.516478712706708e-05, + "loss": 1.1016, + "step": 14447 + }, + { + "epoch": 0.35, + "learning_rate": 1.5164120144107939e-05, + "loss": 1.1212, + "step": 14448 + }, + { + "epoch": 0.35, + "learning_rate": 1.51634531298193e-05, + "loss": 0.9529, + "step": 14449 + }, + { + "epoch": 0.35, + "learning_rate": 1.5162786084205204e-05, + "loss": 1.0699, + "step": 14450 + }, + { + "epoch": 0.35, + "learning_rate": 1.5162119007269698e-05, + "loss": 1.0658, + "step": 14451 + }, + { + "epoch": 0.35, + "learning_rate": 1.5161451899016828e-05, + "loss": 1.1032, + "step": 14452 + }, + { + "epoch": 0.35, + "learning_rate": 1.5160784759450647e-05, + "loss": 1.1545, + "step": 14453 + }, + { + "epoch": 0.35, + "learning_rate": 1.5160117588575198e-05, + "loss": 0.945, + "step": 14454 + }, + { + "epoch": 0.35, + "learning_rate": 1.5159450386394529e-05, + "loss": 1.0532, + "step": 14455 + }, + { + "epoch": 0.35, + "learning_rate": 1.5158783152912688e-05, + "loss": 1.2337, + "step": 14456 + }, + { + "epoch": 0.35, + "learning_rate": 1.5158115888133723e-05, + "loss": 0.9617, + "step": 14457 + }, + { + "epoch": 0.35, + "learning_rate": 1.5157448592061683e-05, + "loss": 1.3153, + "step": 14458 + }, + { + "epoch": 0.35, + "learning_rate": 1.5156781264700615e-05, + "loss": 1.2512, + "step": 14459 + }, + { + "epoch": 0.35, + "learning_rate": 1.5156113906054568e-05, + "loss": 1.017, + "step": 14460 + }, + { + "epoch": 0.35, + "learning_rate": 1.515544651612759e-05, + "loss": 1.1077, + "step": 14461 + }, + { + "epoch": 0.35, + "learning_rate": 1.5154779094923733e-05, + "loss": 1.0366, + "step": 14462 + }, + { + "epoch": 0.35, + "learning_rate": 1.515411164244704e-05, + "loss": 0.9183, + "step": 14463 + }, + { + "epoch": 0.35, + "learning_rate": 1.5153444158701567e-05, + "loss": 1.1873, + "step": 14464 + }, + { + "epoch": 0.35, + "learning_rate": 1.5152776643691358e-05, + "loss": 1.1195, + "step": 14465 + }, + { + "epoch": 0.35, + "learning_rate": 1.515210909742047e-05, + "loss": 1.1139, + "step": 14466 + }, + { + "epoch": 0.35, + "learning_rate": 1.5151441519892942e-05, + "loss": 1.2277, + "step": 14467 + }, + { + "epoch": 0.35, + "learning_rate": 1.5150773911112833e-05, + "loss": 0.934, + "step": 14468 + }, + { + "epoch": 0.35, + "learning_rate": 1.5150106271084188e-05, + "loss": 1.2718, + "step": 14469 + }, + { + "epoch": 0.35, + "learning_rate": 1.5149438599811062e-05, + "loss": 1.0655, + "step": 14470 + }, + { + "epoch": 0.35, + "learning_rate": 1.5148770897297503e-05, + "loss": 1.0755, + "step": 14471 + }, + { + "epoch": 0.35, + "learning_rate": 1.5148103163547562e-05, + "loss": 1.1974, + "step": 14472 + }, + { + "epoch": 0.35, + "learning_rate": 1.514743539856529e-05, + "loss": 1.3738, + "step": 14473 + }, + { + "epoch": 0.35, + "learning_rate": 1.5146767602354738e-05, + "loss": 1.1075, + "step": 14474 + }, + { + "epoch": 0.35, + "learning_rate": 1.5146099774919955e-05, + "loss": 1.1351, + "step": 14475 + }, + { + "epoch": 0.35, + "learning_rate": 1.5145431916265e-05, + "loss": 1.0684, + "step": 14476 + }, + { + "epoch": 0.35, + "learning_rate": 1.5144764026393917e-05, + "loss": 1.0816, + "step": 14477 + }, + { + "epoch": 0.35, + "learning_rate": 1.5144096105310761e-05, + "loss": 1.1942, + "step": 14478 + }, + { + "epoch": 0.35, + "learning_rate": 1.5143428153019583e-05, + "loss": 1.1859, + "step": 14479 + }, + { + "epoch": 0.35, + "learning_rate": 1.5142760169524438e-05, + "loss": 1.3316, + "step": 14480 + }, + { + "epoch": 0.35, + "learning_rate": 1.5142092154829376e-05, + "loss": 1.0972, + "step": 14481 + }, + { + "epoch": 0.35, + "learning_rate": 1.514142410893845e-05, + "loss": 1.0827, + "step": 14482 + }, + { + "epoch": 0.35, + "learning_rate": 1.5140756031855715e-05, + "loss": 1.1262, + "step": 14483 + }, + { + "epoch": 0.35, + "learning_rate": 1.5140087923585219e-05, + "loss": 1.0604, + "step": 14484 + }, + { + "epoch": 0.35, + "learning_rate": 1.5139419784131022e-05, + "loss": 1.4596, + "step": 14485 + }, + { + "epoch": 0.35, + "learning_rate": 1.513875161349717e-05, + "loss": 1.3129, + "step": 14486 + }, + { + "epoch": 0.35, + "learning_rate": 1.5138083411687725e-05, + "loss": 1.287, + "step": 14487 + }, + { + "epoch": 0.35, + "learning_rate": 1.5137415178706733e-05, + "loss": 1.1876, + "step": 14488 + }, + { + "epoch": 0.35, + "learning_rate": 1.5136746914558256e-05, + "loss": 1.1279, + "step": 14489 + }, + { + "epoch": 0.35, + "learning_rate": 1.5136078619246339e-05, + "loss": 1.1173, + "step": 14490 + }, + { + "epoch": 0.35, + "learning_rate": 1.5135410292775046e-05, + "loss": 0.9753, + "step": 14491 + }, + { + "epoch": 0.35, + "learning_rate": 1.513474193514842e-05, + "loss": 1.0445, + "step": 14492 + }, + { + "epoch": 0.35, + "learning_rate": 1.5134073546370526e-05, + "loss": 1.1142, + "step": 14493 + }, + { + "epoch": 0.35, + "learning_rate": 1.5133405126445417e-05, + "loss": 0.9922, + "step": 14494 + }, + { + "epoch": 0.35, + "learning_rate": 1.5132736675377145e-05, + "loss": 0.909, + "step": 14495 + }, + { + "epoch": 0.35, + "learning_rate": 1.5132068193169765e-05, + "loss": 1.2154, + "step": 14496 + }, + { + "epoch": 0.35, + "learning_rate": 1.5131399679827338e-05, + "loss": 1.0596, + "step": 14497 + }, + { + "epoch": 0.35, + "learning_rate": 1.5130731135353914e-05, + "loss": 1.3106, + "step": 14498 + }, + { + "epoch": 0.35, + "learning_rate": 1.5130062559753553e-05, + "loss": 1.1683, + "step": 14499 + }, + { + "epoch": 0.35, + "learning_rate": 1.5129393953030307e-05, + "loss": 1.2393, + "step": 14500 + }, + { + "epoch": 0.35, + "learning_rate": 1.5128725315188235e-05, + "loss": 1.187, + "step": 14501 + }, + { + "epoch": 0.35, + "learning_rate": 1.5128056646231393e-05, + "loss": 0.7896, + "step": 14502 + }, + { + "epoch": 0.35, + "learning_rate": 1.5127387946163837e-05, + "loss": 1.2749, + "step": 14503 + }, + { + "epoch": 0.35, + "learning_rate": 1.5126719214989622e-05, + "loss": 1.0603, + "step": 14504 + }, + { + "epoch": 0.35, + "learning_rate": 1.5126050452712812e-05, + "loss": 1.0689, + "step": 14505 + }, + { + "epoch": 0.35, + "learning_rate": 1.5125381659337458e-05, + "loss": 1.0885, + "step": 14506 + }, + { + "epoch": 0.35, + "learning_rate": 1.5124712834867618e-05, + "loss": 1.0726, + "step": 14507 + }, + { + "epoch": 0.35, + "learning_rate": 1.5124043979307353e-05, + "loss": 1.2068, + "step": 14508 + }, + { + "epoch": 0.35, + "learning_rate": 1.5123375092660717e-05, + "loss": 1.0033, + "step": 14509 + }, + { + "epoch": 0.35, + "learning_rate": 1.512270617493177e-05, + "loss": 1.0796, + "step": 14510 + }, + { + "epoch": 0.35, + "learning_rate": 1.512203722612457e-05, + "loss": 1.2897, + "step": 14511 + }, + { + "epoch": 0.35, + "learning_rate": 1.512136824624317e-05, + "loss": 1.1752, + "step": 14512 + }, + { + "epoch": 0.35, + "learning_rate": 1.5120699235291637e-05, + "loss": 1.2024, + "step": 14513 + }, + { + "epoch": 0.35, + "learning_rate": 1.512003019327403e-05, + "loss": 1.2191, + "step": 14514 + }, + { + "epoch": 0.35, + "learning_rate": 1.51193611201944e-05, + "loss": 1.331, + "step": 14515 + }, + { + "epoch": 0.35, + "learning_rate": 1.511869201605681e-05, + "loss": 1.2772, + "step": 14516 + }, + { + "epoch": 0.35, + "learning_rate": 1.511802288086532e-05, + "loss": 1.0429, + "step": 14517 + }, + { + "epoch": 0.35, + "learning_rate": 1.511735371462399e-05, + "loss": 1.13, + "step": 14518 + }, + { + "epoch": 0.35, + "learning_rate": 1.5116684517336877e-05, + "loss": 1.3466, + "step": 14519 + }, + { + "epoch": 0.35, + "learning_rate": 1.5116015289008043e-05, + "loss": 0.9333, + "step": 14520 + }, + { + "epoch": 0.35, + "learning_rate": 1.5115346029641549e-05, + "loss": 1.0208, + "step": 14521 + }, + { + "epoch": 0.35, + "learning_rate": 1.5114676739241454e-05, + "loss": 1.1627, + "step": 14522 + }, + { + "epoch": 0.35, + "learning_rate": 1.5114007417811817e-05, + "loss": 1.2562, + "step": 14523 + }, + { + "epoch": 0.35, + "learning_rate": 1.51133380653567e-05, + "loss": 1.2931, + "step": 14524 + }, + { + "epoch": 0.35, + "learning_rate": 1.5112668681880163e-05, + "loss": 1.1717, + "step": 14525 + }, + { + "epoch": 0.35, + "learning_rate": 1.5111999267386268e-05, + "loss": 0.8596, + "step": 14526 + }, + { + "epoch": 0.35, + "learning_rate": 1.5111329821879077e-05, + "loss": 1.3582, + "step": 14527 + }, + { + "epoch": 0.35, + "learning_rate": 1.5110660345362651e-05, + "loss": 1.1894, + "step": 14528 + }, + { + "epoch": 0.35, + "learning_rate": 1.510999083784105e-05, + "loss": 1.3547, + "step": 14529 + }, + { + "epoch": 0.35, + "learning_rate": 1.5109321299318333e-05, + "loss": 1.0927, + "step": 14530 + }, + { + "epoch": 0.35, + "learning_rate": 1.5108651729798569e-05, + "loss": 0.8822, + "step": 14531 + }, + { + "epoch": 0.35, + "learning_rate": 1.5107982129285816e-05, + "loss": 1.3395, + "step": 14532 + }, + { + "epoch": 0.35, + "learning_rate": 1.5107312497784139e-05, + "loss": 1.3723, + "step": 14533 + }, + { + "epoch": 0.35, + "learning_rate": 1.5106642835297596e-05, + "loss": 1.1774, + "step": 14534 + }, + { + "epoch": 0.35, + "learning_rate": 1.510597314183025e-05, + "loss": 1.173, + "step": 14535 + }, + { + "epoch": 0.35, + "learning_rate": 1.5105303417386168e-05, + "loss": 1.2483, + "step": 14536 + }, + { + "epoch": 0.35, + "learning_rate": 1.5104633661969414e-05, + "loss": 1.1742, + "step": 14537 + }, + { + "epoch": 0.35, + "learning_rate": 1.5103963875584046e-05, + "loss": 0.846, + "step": 14538 + }, + { + "epoch": 0.35, + "learning_rate": 1.5103294058234129e-05, + "loss": 1.1238, + "step": 14539 + }, + { + "epoch": 0.35, + "learning_rate": 1.5102624209923723e-05, + "loss": 1.3297, + "step": 14540 + }, + { + "epoch": 0.35, + "learning_rate": 1.5101954330656903e-05, + "loss": 1.1944, + "step": 14541 + }, + { + "epoch": 0.35, + "learning_rate": 1.5101284420437722e-05, + "loss": 1.21, + "step": 14542 + }, + { + "epoch": 0.35, + "learning_rate": 1.5100614479270253e-05, + "loss": 1.2365, + "step": 14543 + }, + { + "epoch": 0.35, + "learning_rate": 1.509994450715855e-05, + "loss": 0.8822, + "step": 14544 + }, + { + "epoch": 0.35, + "learning_rate": 1.5099274504106687e-05, + "loss": 1.1438, + "step": 14545 + }, + { + "epoch": 0.35, + "learning_rate": 1.5098604470118724e-05, + "loss": 1.3434, + "step": 14546 + }, + { + "epoch": 0.35, + "learning_rate": 1.5097934405198728e-05, + "loss": 1.0261, + "step": 14547 + }, + { + "epoch": 0.35, + "learning_rate": 1.5097264309350761e-05, + "loss": 1.3717, + "step": 14548 + }, + { + "epoch": 0.35, + "learning_rate": 1.5096594182578893e-05, + "loss": 1.144, + "step": 14549 + }, + { + "epoch": 0.35, + "learning_rate": 1.5095924024887185e-05, + "loss": 1.0336, + "step": 14550 + }, + { + "epoch": 0.35, + "learning_rate": 1.5095253836279704e-05, + "loss": 1.0232, + "step": 14551 + }, + { + "epoch": 0.35, + "learning_rate": 1.5094583616760518e-05, + "loss": 0.9799, + "step": 14552 + }, + { + "epoch": 0.35, + "learning_rate": 1.5093913366333693e-05, + "loss": 1.1959, + "step": 14553 + }, + { + "epoch": 0.35, + "learning_rate": 1.509324308500329e-05, + "loss": 0.9778, + "step": 14554 + }, + { + "epoch": 0.35, + "learning_rate": 1.5092572772773381e-05, + "loss": 1.2072, + "step": 14555 + }, + { + "epoch": 0.35, + "learning_rate": 1.5091902429648029e-05, + "loss": 1.1883, + "step": 14556 + }, + { + "epoch": 0.35, + "learning_rate": 1.5091232055631306e-05, + "loss": 0.918, + "step": 14557 + }, + { + "epoch": 0.35, + "learning_rate": 1.5090561650727274e-05, + "loss": 1.0075, + "step": 14558 + }, + { + "epoch": 0.35, + "learning_rate": 1.5089891214940002e-05, + "loss": 0.8987, + "step": 14559 + }, + { + "epoch": 0.35, + "learning_rate": 1.5089220748273557e-05, + "loss": 1.1082, + "step": 14560 + }, + { + "epoch": 0.35, + "learning_rate": 1.5088550250732006e-05, + "loss": 1.2962, + "step": 14561 + }, + { + "epoch": 0.35, + "learning_rate": 1.508787972231942e-05, + "loss": 0.947, + "step": 14562 + }, + { + "epoch": 0.35, + "learning_rate": 1.508720916303986e-05, + "loss": 1.1184, + "step": 14563 + }, + { + "epoch": 0.35, + "learning_rate": 1.5086538572897404e-05, + "loss": 1.3852, + "step": 14564 + }, + { + "epoch": 0.35, + "learning_rate": 1.5085867951896111e-05, + "loss": 1.0861, + "step": 14565 + }, + { + "epoch": 0.35, + "learning_rate": 1.5085197300040055e-05, + "loss": 1.3447, + "step": 14566 + }, + { + "epoch": 0.35, + "learning_rate": 1.5084526617333301e-05, + "loss": 1.1989, + "step": 14567 + }, + { + "epoch": 0.35, + "learning_rate": 1.5083855903779925e-05, + "loss": 1.2044, + "step": 14568 + }, + { + "epoch": 0.35, + "learning_rate": 1.5083185159383988e-05, + "loss": 1.2171, + "step": 14569 + }, + { + "epoch": 0.35, + "learning_rate": 1.508251438414956e-05, + "loss": 1.2512, + "step": 14570 + }, + { + "epoch": 0.35, + "learning_rate": 1.5081843578080714e-05, + "loss": 1.0005, + "step": 14571 + }, + { + "epoch": 0.35, + "learning_rate": 1.5081172741181522e-05, + "loss": 1.174, + "step": 14572 + }, + { + "epoch": 0.35, + "learning_rate": 1.5080501873456049e-05, + "loss": 0.9734, + "step": 14573 + }, + { + "epoch": 0.35, + "learning_rate": 1.5079830974908362e-05, + "loss": 1.3616, + "step": 14574 + }, + { + "epoch": 0.35, + "learning_rate": 1.5079160045542538e-05, + "loss": 1.1295, + "step": 14575 + }, + { + "epoch": 0.35, + "learning_rate": 1.5078489085362643e-05, + "loss": 0.911, + "step": 14576 + }, + { + "epoch": 0.35, + "learning_rate": 1.5077818094372751e-05, + "loss": 1.4453, + "step": 14577 + }, + { + "epoch": 0.35, + "learning_rate": 1.5077147072576932e-05, + "loss": 0.7927, + "step": 14578 + }, + { + "epoch": 0.35, + "learning_rate": 1.5076476019979254e-05, + "loss": 1.0719, + "step": 14579 + }, + { + "epoch": 0.35, + "learning_rate": 1.5075804936583791e-05, + "loss": 1.0251, + "step": 14580 + }, + { + "epoch": 0.35, + "learning_rate": 1.5075133822394615e-05, + "loss": 0.9595, + "step": 14581 + }, + { + "epoch": 0.35, + "learning_rate": 1.5074462677415793e-05, + "loss": 1.2206, + "step": 14582 + }, + { + "epoch": 0.35, + "learning_rate": 1.5073791501651399e-05, + "loss": 1.1475, + "step": 14583 + }, + { + "epoch": 0.35, + "learning_rate": 1.5073120295105507e-05, + "loss": 1.1307, + "step": 14584 + }, + { + "epoch": 0.35, + "learning_rate": 1.5072449057782188e-05, + "loss": 1.077, + "step": 14585 + }, + { + "epoch": 0.35, + "learning_rate": 1.5071777789685514e-05, + "loss": 1.2155, + "step": 14586 + }, + { + "epoch": 0.35, + "learning_rate": 1.5071106490819554e-05, + "loss": 1.362, + "step": 14587 + }, + { + "epoch": 0.35, + "learning_rate": 1.5070435161188386e-05, + "loss": 0.9997, + "step": 14588 + }, + { + "epoch": 0.35, + "learning_rate": 1.5069763800796081e-05, + "loss": 1.3891, + "step": 14589 + }, + { + "epoch": 0.35, + "learning_rate": 1.5069092409646713e-05, + "loss": 1.0988, + "step": 14590 + }, + { + "epoch": 0.35, + "learning_rate": 1.5068420987744348e-05, + "loss": 1.1147, + "step": 14591 + }, + { + "epoch": 0.35, + "learning_rate": 1.5067749535093068e-05, + "loss": 0.9501, + "step": 14592 + }, + { + "epoch": 0.35, + "learning_rate": 1.5067078051696947e-05, + "loss": 1.1669, + "step": 14593 + }, + { + "epoch": 0.35, + "learning_rate": 1.5066406537560053e-05, + "loss": 0.8919, + "step": 14594 + }, + { + "epoch": 0.35, + "learning_rate": 1.5065734992686461e-05, + "loss": 1.0715, + "step": 14595 + }, + { + "epoch": 0.35, + "learning_rate": 1.5065063417080246e-05, + "loss": 1.1305, + "step": 14596 + }, + { + "epoch": 0.35, + "learning_rate": 1.5064391810745483e-05, + "loss": 1.0326, + "step": 14597 + }, + { + "epoch": 0.35, + "learning_rate": 1.5063720173686246e-05, + "loss": 1.2578, + "step": 14598 + }, + { + "epoch": 0.35, + "learning_rate": 1.506304850590661e-05, + "loss": 1.0657, + "step": 14599 + }, + { + "epoch": 0.35, + "learning_rate": 1.506237680741065e-05, + "loss": 1.1956, + "step": 14600 + }, + { + "epoch": 0.35, + "learning_rate": 1.5061705078202443e-05, + "loss": 1.3319, + "step": 14601 + }, + { + "epoch": 0.35, + "learning_rate": 1.506103331828606e-05, + "loss": 0.9834, + "step": 14602 + }, + { + "epoch": 0.35, + "learning_rate": 1.5060361527665577e-05, + "loss": 1.086, + "step": 14603 + }, + { + "epoch": 0.35, + "learning_rate": 1.5059689706345072e-05, + "loss": 1.0494, + "step": 14604 + }, + { + "epoch": 0.35, + "learning_rate": 1.5059017854328619e-05, + "loss": 1.1893, + "step": 14605 + }, + { + "epoch": 0.35, + "learning_rate": 1.5058345971620294e-05, + "loss": 1.2993, + "step": 14606 + }, + { + "epoch": 0.35, + "learning_rate": 1.5057674058224174e-05, + "loss": 1.1165, + "step": 14607 + }, + { + "epoch": 0.35, + "learning_rate": 1.5057002114144336e-05, + "loss": 1.1201, + "step": 14608 + }, + { + "epoch": 0.35, + "learning_rate": 1.5056330139384856e-05, + "loss": 1.2898, + "step": 14609 + }, + { + "epoch": 0.35, + "learning_rate": 1.5055658133949811e-05, + "loss": 0.9841, + "step": 14610 + }, + { + "epoch": 0.35, + "learning_rate": 1.5054986097843274e-05, + "loss": 1.1658, + "step": 14611 + }, + { + "epoch": 0.35, + "learning_rate": 1.505431403106933e-05, + "loss": 0.9232, + "step": 14612 + }, + { + "epoch": 0.35, + "learning_rate": 1.5053641933632049e-05, + "loss": 1.2686, + "step": 14613 + }, + { + "epoch": 0.35, + "learning_rate": 1.5052969805535509e-05, + "loss": 0.813, + "step": 14614 + }, + { + "epoch": 0.35, + "learning_rate": 1.5052297646783793e-05, + "loss": 0.9192, + "step": 14615 + }, + { + "epoch": 0.35, + "learning_rate": 1.5051625457380975e-05, + "loss": 1.2614, + "step": 14616 + }, + { + "epoch": 0.35, + "learning_rate": 1.5050953237331134e-05, + "loss": 1.0581, + "step": 14617 + }, + { + "epoch": 0.35, + "learning_rate": 1.5050280986638345e-05, + "loss": 1.1684, + "step": 14618 + }, + { + "epoch": 0.35, + "learning_rate": 1.504960870530669e-05, + "loss": 1.0079, + "step": 14619 + }, + { + "epoch": 0.35, + "learning_rate": 1.504893639334025e-05, + "loss": 1.3041, + "step": 14620 + }, + { + "epoch": 0.35, + "learning_rate": 1.5048264050743097e-05, + "loss": 1.2411, + "step": 14621 + }, + { + "epoch": 0.35, + "learning_rate": 1.5047591677519315e-05, + "loss": 1.4163, + "step": 14622 + }, + { + "epoch": 0.35, + "learning_rate": 1.5046919273672983e-05, + "loss": 1.4346, + "step": 14623 + }, + { + "epoch": 0.35, + "learning_rate": 1.5046246839208176e-05, + "loss": 1.3082, + "step": 14624 + }, + { + "epoch": 0.35, + "learning_rate": 1.5045574374128977e-05, + "loss": 1.3736, + "step": 14625 + }, + { + "epoch": 0.35, + "learning_rate": 1.5044901878439467e-05, + "loss": 1.3535, + "step": 14626 + }, + { + "epoch": 0.35, + "learning_rate": 1.5044229352143719e-05, + "loss": 1.0179, + "step": 14627 + }, + { + "epoch": 0.35, + "learning_rate": 1.5043556795245825e-05, + "loss": 1.2634, + "step": 14628 + }, + { + "epoch": 0.35, + "learning_rate": 1.5042884207749852e-05, + "loss": 1.0342, + "step": 14629 + }, + { + "epoch": 0.35, + "learning_rate": 1.5042211589659892e-05, + "loss": 1.1337, + "step": 14630 + }, + { + "epoch": 0.35, + "learning_rate": 1.5041538940980016e-05, + "loss": 1.0564, + "step": 14631 + }, + { + "epoch": 0.35, + "learning_rate": 1.5040866261714311e-05, + "loss": 1.2426, + "step": 14632 + }, + { + "epoch": 0.35, + "learning_rate": 1.5040193551866856e-05, + "loss": 1.2074, + "step": 14633 + }, + { + "epoch": 0.35, + "learning_rate": 1.5039520811441733e-05, + "loss": 1.1322, + "step": 14634 + }, + { + "epoch": 0.35, + "learning_rate": 1.503884804044302e-05, + "loss": 1.0704, + "step": 14635 + }, + { + "epoch": 0.35, + "learning_rate": 1.5038175238874803e-05, + "loss": 1.0277, + "step": 14636 + }, + { + "epoch": 0.35, + "learning_rate": 1.5037502406741162e-05, + "loss": 1.0996, + "step": 14637 + }, + { + "epoch": 0.35, + "learning_rate": 1.5036829544046178e-05, + "loss": 1.2045, + "step": 14638 + }, + { + "epoch": 0.35, + "learning_rate": 1.5036156650793934e-05, + "loss": 1.1065, + "step": 14639 + }, + { + "epoch": 0.35, + "learning_rate": 1.5035483726988513e-05, + "loss": 1.2138, + "step": 14640 + }, + { + "epoch": 0.35, + "learning_rate": 1.5034810772633996e-05, + "loss": 1.0965, + "step": 14641 + }, + { + "epoch": 0.35, + "learning_rate": 1.5034137787734468e-05, + "loss": 1.0726, + "step": 14642 + }, + { + "epoch": 0.35, + "learning_rate": 1.5033464772294007e-05, + "loss": 1.1895, + "step": 14643 + }, + { + "epoch": 0.35, + "learning_rate": 1.50327917263167e-05, + "loss": 0.8724, + "step": 14644 + }, + { + "epoch": 0.35, + "learning_rate": 1.503211864980663e-05, + "loss": 1.3289, + "step": 14645 + }, + { + "epoch": 0.35, + "learning_rate": 1.5031445542767884e-05, + "loss": 1.3507, + "step": 14646 + }, + { + "epoch": 0.35, + "learning_rate": 1.5030772405204535e-05, + "loss": 1.2679, + "step": 14647 + }, + { + "epoch": 0.35, + "learning_rate": 1.5030099237120674e-05, + "loss": 1.2268, + "step": 14648 + }, + { + "epoch": 0.35, + "learning_rate": 1.5029426038520387e-05, + "loss": 0.9656, + "step": 14649 + }, + { + "epoch": 0.35, + "learning_rate": 1.5028752809407752e-05, + "loss": 1.0454, + "step": 14650 + }, + { + "epoch": 0.35, + "learning_rate": 1.5028079549786858e-05, + "loss": 1.1742, + "step": 14651 + }, + { + "epoch": 0.35, + "learning_rate": 1.5027406259661791e-05, + "loss": 1.1087, + "step": 14652 + }, + { + "epoch": 0.35, + "learning_rate": 1.502673293903663e-05, + "loss": 1.405, + "step": 14653 + }, + { + "epoch": 0.35, + "learning_rate": 1.5026059587915463e-05, + "loss": 1.1607, + "step": 14654 + }, + { + "epoch": 0.35, + "learning_rate": 1.5025386206302374e-05, + "loss": 1.0234, + "step": 14655 + }, + { + "epoch": 0.35, + "learning_rate": 1.5024712794201449e-05, + "loss": 1.0456, + "step": 14656 + }, + { + "epoch": 0.35, + "learning_rate": 1.5024039351616772e-05, + "loss": 1.1982, + "step": 14657 + }, + { + "epoch": 0.35, + "learning_rate": 1.5023365878552434e-05, + "loss": 1.0481, + "step": 14658 + }, + { + "epoch": 0.35, + "learning_rate": 1.5022692375012513e-05, + "loss": 0.8571, + "step": 14659 + }, + { + "epoch": 0.35, + "learning_rate": 1.50220188410011e-05, + "loss": 1.0718, + "step": 14660 + }, + { + "epoch": 0.35, + "learning_rate": 1.502134527652228e-05, + "loss": 1.0123, + "step": 14661 + }, + { + "epoch": 0.35, + "learning_rate": 1.502067168158014e-05, + "loss": 1.0687, + "step": 14662 + }, + { + "epoch": 0.35, + "learning_rate": 1.5019998056178765e-05, + "loss": 1.1931, + "step": 14663 + }, + { + "epoch": 0.35, + "learning_rate": 1.5019324400322244e-05, + "loss": 1.1264, + "step": 14664 + }, + { + "epoch": 0.35, + "learning_rate": 1.5018650714014662e-05, + "loss": 1.072, + "step": 14665 + }, + { + "epoch": 0.35, + "learning_rate": 1.5017976997260104e-05, + "loss": 1.1262, + "step": 14666 + }, + { + "epoch": 0.35, + "learning_rate": 1.5017303250062662e-05, + "loss": 0.6665, + "step": 14667 + }, + { + "epoch": 0.35, + "learning_rate": 1.5016629472426423e-05, + "loss": 1.1068, + "step": 14668 + }, + { + "epoch": 0.35, + "learning_rate": 1.5015955664355471e-05, + "loss": 0.9583, + "step": 14669 + }, + { + "epoch": 0.35, + "learning_rate": 1.5015281825853896e-05, + "loss": 0.8649, + "step": 14670 + }, + { + "epoch": 0.35, + "learning_rate": 1.5014607956925785e-05, + "loss": 0.8008, + "step": 14671 + }, + { + "epoch": 0.35, + "learning_rate": 1.5013934057575228e-05, + "loss": 1.2247, + "step": 14672 + }, + { + "epoch": 0.35, + "learning_rate": 1.5013260127806312e-05, + "loss": 1.1889, + "step": 14673 + }, + { + "epoch": 0.35, + "learning_rate": 1.5012586167623124e-05, + "loss": 0.9885, + "step": 14674 + }, + { + "epoch": 0.35, + "learning_rate": 1.5011912177029758e-05, + "loss": 0.9704, + "step": 14675 + }, + { + "epoch": 0.35, + "learning_rate": 1.50112381560303e-05, + "loss": 1.3976, + "step": 14676 + }, + { + "epoch": 0.35, + "learning_rate": 1.5010564104628838e-05, + "loss": 1.377, + "step": 14677 + }, + { + "epoch": 0.35, + "learning_rate": 1.500989002282946e-05, + "loss": 0.9509, + "step": 14678 + }, + { + "epoch": 0.35, + "learning_rate": 1.5009215910636261e-05, + "loss": 1.1404, + "step": 14679 + }, + { + "epoch": 0.35, + "learning_rate": 1.5008541768053324e-05, + "loss": 1.0514, + "step": 14680 + }, + { + "epoch": 0.35, + "learning_rate": 1.5007867595084746e-05, + "loss": 1.1033, + "step": 14681 + }, + { + "epoch": 0.35, + "learning_rate": 1.5007193391734607e-05, + "loss": 1.063, + "step": 14682 + }, + { + "epoch": 0.35, + "learning_rate": 1.500651915800701e-05, + "loss": 0.9794, + "step": 14683 + }, + { + "epoch": 0.35, + "learning_rate": 1.5005844893906036e-05, + "loss": 1.1546, + "step": 14684 + }, + { + "epoch": 0.35, + "learning_rate": 1.5005170599435779e-05, + "loss": 1.0279, + "step": 14685 + }, + { + "epoch": 0.35, + "learning_rate": 1.5004496274600327e-05, + "loss": 1.1103, + "step": 14686 + }, + { + "epoch": 0.35, + "learning_rate": 1.5003821919403775e-05, + "loss": 0.8611, + "step": 14687 + }, + { + "epoch": 0.35, + "learning_rate": 1.5003147533850215e-05, + "loss": 1.0635, + "step": 14688 + }, + { + "epoch": 0.35, + "learning_rate": 1.5002473117943734e-05, + "loss": 1.0282, + "step": 14689 + }, + { + "epoch": 0.35, + "learning_rate": 1.5001798671688424e-05, + "loss": 1.1877, + "step": 14690 + }, + { + "epoch": 0.35, + "learning_rate": 1.5001124195088376e-05, + "loss": 1.1659, + "step": 14691 + }, + { + "epoch": 0.35, + "learning_rate": 1.500044968814769e-05, + "loss": 0.83, + "step": 14692 + }, + { + "epoch": 0.35, + "learning_rate": 1.4999775150870446e-05, + "loss": 1.1809, + "step": 14693 + }, + { + "epoch": 0.35, + "learning_rate": 1.4999100583260743e-05, + "loss": 1.013, + "step": 14694 + }, + { + "epoch": 0.35, + "learning_rate": 1.4998425985322675e-05, + "loss": 1.1718, + "step": 14695 + }, + { + "epoch": 0.35, + "learning_rate": 1.4997751357060334e-05, + "loss": 1.1395, + "step": 14696 + }, + { + "epoch": 0.35, + "learning_rate": 1.4997076698477805e-05, + "loss": 1.3139, + "step": 14697 + }, + { + "epoch": 0.35, + "learning_rate": 1.4996402009579192e-05, + "loss": 1.1228, + "step": 14698 + }, + { + "epoch": 0.35, + "learning_rate": 1.4995727290368583e-05, + "loss": 1.0341, + "step": 14699 + }, + { + "epoch": 0.35, + "learning_rate": 1.499505254085007e-05, + "loss": 1.2001, + "step": 14700 + }, + { + "epoch": 0.35, + "learning_rate": 1.4994377761027745e-05, + "loss": 1.0204, + "step": 14701 + }, + { + "epoch": 0.35, + "learning_rate": 1.4993702950905711e-05, + "loss": 1.2171, + "step": 14702 + }, + { + "epoch": 0.35, + "learning_rate": 1.4993028110488052e-05, + "loss": 1.0076, + "step": 14703 + }, + { + "epoch": 0.35, + "learning_rate": 1.499235323977887e-05, + "loss": 1.1604, + "step": 14704 + }, + { + "epoch": 0.35, + "learning_rate": 1.499167833878225e-05, + "loss": 1.3843, + "step": 14705 + }, + { + "epoch": 0.35, + "learning_rate": 1.4991003407502293e-05, + "loss": 1.4471, + "step": 14706 + }, + { + "epoch": 0.35, + "learning_rate": 1.4990328445943092e-05, + "loss": 1.2211, + "step": 14707 + }, + { + "epoch": 0.35, + "learning_rate": 1.4989653454108744e-05, + "loss": 1.0173, + "step": 14708 + }, + { + "epoch": 0.35, + "learning_rate": 1.498897843200334e-05, + "loss": 1.2067, + "step": 14709 + }, + { + "epoch": 0.35, + "learning_rate": 1.498830337963098e-05, + "loss": 1.1919, + "step": 14710 + }, + { + "epoch": 0.35, + "learning_rate": 1.4987628296995755e-05, + "loss": 1.1638, + "step": 14711 + }, + { + "epoch": 0.35, + "learning_rate": 1.4986953184101763e-05, + "loss": 1.049, + "step": 14712 + }, + { + "epoch": 0.35, + "learning_rate": 1.49862780409531e-05, + "loss": 1.2186, + "step": 14713 + }, + { + "epoch": 0.35, + "learning_rate": 1.4985602867553857e-05, + "loss": 1.0333, + "step": 14714 + }, + { + "epoch": 0.35, + "learning_rate": 1.4984927663908137e-05, + "loss": 1.1693, + "step": 14715 + }, + { + "epoch": 0.35, + "learning_rate": 1.4984252430020033e-05, + "loss": 1.1404, + "step": 14716 + }, + { + "epoch": 0.35, + "learning_rate": 1.4983577165893643e-05, + "loss": 1.1173, + "step": 14717 + }, + { + "epoch": 0.35, + "learning_rate": 1.498290187153306e-05, + "loss": 1.1286, + "step": 14718 + }, + { + "epoch": 0.35, + "learning_rate": 1.4982226546942386e-05, + "loss": 1.1349, + "step": 14719 + }, + { + "epoch": 0.35, + "learning_rate": 1.4981551192125714e-05, + "loss": 0.9253, + "step": 14720 + }, + { + "epoch": 0.35, + "learning_rate": 1.4980875807087145e-05, + "loss": 1.205, + "step": 14721 + }, + { + "epoch": 0.35, + "learning_rate": 1.4980200391830771e-05, + "loss": 1.0715, + "step": 14722 + }, + { + "epoch": 0.35, + "learning_rate": 1.4979524946360693e-05, + "loss": 1.3518, + "step": 14723 + }, + { + "epoch": 0.35, + "learning_rate": 1.497884947068101e-05, + "loss": 1.365, + "step": 14724 + }, + { + "epoch": 0.35, + "learning_rate": 1.4978173964795817e-05, + "loss": 1.0975, + "step": 14725 + }, + { + "epoch": 0.35, + "learning_rate": 1.4977498428709214e-05, + "loss": 1.0325, + "step": 14726 + }, + { + "epoch": 0.35, + "learning_rate": 1.4976822862425298e-05, + "loss": 1.2192, + "step": 14727 + }, + { + "epoch": 0.35, + "learning_rate": 1.4976147265948168e-05, + "loss": 1.0619, + "step": 14728 + }, + { + "epoch": 0.35, + "learning_rate": 1.4975471639281927e-05, + "loss": 1.0473, + "step": 14729 + }, + { + "epoch": 0.35, + "learning_rate": 1.4974795982430665e-05, + "loss": 1.0924, + "step": 14730 + }, + { + "epoch": 0.35, + "learning_rate": 1.4974120295398489e-05, + "loss": 0.9509, + "step": 14731 + }, + { + "epoch": 0.35, + "learning_rate": 1.4973444578189492e-05, + "loss": 1.1153, + "step": 14732 + }, + { + "epoch": 0.35, + "learning_rate": 1.4972768830807783e-05, + "loss": 1.2255, + "step": 14733 + }, + { + "epoch": 0.35, + "learning_rate": 1.4972093053257447e-05, + "loss": 1.1008, + "step": 14734 + }, + { + "epoch": 0.35, + "learning_rate": 1.4971417245542597e-05, + "loss": 1.15, + "step": 14735 + }, + { + "epoch": 0.35, + "learning_rate": 1.4970741407667326e-05, + "loss": 1.1121, + "step": 14736 + }, + { + "epoch": 0.35, + "learning_rate": 1.4970065539635736e-05, + "loss": 1.2418, + "step": 14737 + }, + { + "epoch": 0.35, + "learning_rate": 1.4969389641451928e-05, + "loss": 0.9555, + "step": 14738 + }, + { + "epoch": 0.35, + "learning_rate": 1.4968713713120005e-05, + "loss": 1.2073, + "step": 14739 + }, + { + "epoch": 0.35, + "learning_rate": 1.4968037754644061e-05, + "loss": 1.0752, + "step": 14740 + }, + { + "epoch": 0.35, + "learning_rate": 1.4967361766028198e-05, + "loss": 1.1053, + "step": 14741 + }, + { + "epoch": 0.35, + "learning_rate": 1.4966685747276524e-05, + "loss": 1.3764, + "step": 14742 + }, + { + "epoch": 0.35, + "learning_rate": 1.4966009698393136e-05, + "loss": 1.457, + "step": 14743 + }, + { + "epoch": 0.35, + "learning_rate": 1.4965333619382134e-05, + "loss": 1.1136, + "step": 14744 + }, + { + "epoch": 0.35, + "learning_rate": 1.4964657510247618e-05, + "loss": 1.0196, + "step": 14745 + }, + { + "epoch": 0.35, + "learning_rate": 1.4963981370993696e-05, + "loss": 0.7432, + "step": 14746 + }, + { + "epoch": 0.35, + "learning_rate": 1.4963305201624467e-05, + "loss": 1.3889, + "step": 14747 + }, + { + "epoch": 0.35, + "learning_rate": 1.496262900214403e-05, + "loss": 1.2384, + "step": 14748 + }, + { + "epoch": 0.35, + "learning_rate": 1.4961952772556492e-05, + "loss": 1.1603, + "step": 14749 + }, + { + "epoch": 0.35, + "learning_rate": 1.4961276512865954e-05, + "loss": 0.7732, + "step": 14750 + }, + { + "epoch": 0.35, + "learning_rate": 1.4960600223076518e-05, + "loss": 1.1325, + "step": 14751 + }, + { + "epoch": 0.35, + "learning_rate": 1.4959923903192289e-05, + "loss": 1.063, + "step": 14752 + }, + { + "epoch": 0.35, + "learning_rate": 1.4959247553217365e-05, + "loss": 1.313, + "step": 14753 + }, + { + "epoch": 0.35, + "learning_rate": 1.4958571173155854e-05, + "loss": 1.2566, + "step": 14754 + }, + { + "epoch": 0.35, + "learning_rate": 1.4957894763011858e-05, + "loss": 1.2808, + "step": 14755 + }, + { + "epoch": 0.35, + "learning_rate": 1.4957218322789481e-05, + "loss": 1.2247, + "step": 14756 + }, + { + "epoch": 0.35, + "learning_rate": 1.4956541852492824e-05, + "loss": 1.1346, + "step": 14757 + }, + { + "epoch": 0.35, + "learning_rate": 1.4955865352125995e-05, + "loss": 1.1049, + "step": 14758 + }, + { + "epoch": 0.35, + "learning_rate": 1.4955188821693099e-05, + "loss": 1.0802, + "step": 14759 + }, + { + "epoch": 0.35, + "learning_rate": 1.4954512261198235e-05, + "loss": 0.9344, + "step": 14760 + }, + { + "epoch": 0.35, + "learning_rate": 1.4953835670645512e-05, + "loss": 1.166, + "step": 14761 + }, + { + "epoch": 0.36, + "learning_rate": 1.4953159050039033e-05, + "loss": 0.9547, + "step": 14762 + }, + { + "epoch": 0.36, + "learning_rate": 1.4952482399382904e-05, + "loss": 0.977, + "step": 14763 + }, + { + "epoch": 0.36, + "learning_rate": 1.4951805718681227e-05, + "loss": 1.0176, + "step": 14764 + }, + { + "epoch": 0.36, + "learning_rate": 1.495112900793811e-05, + "loss": 0.6322, + "step": 14765 + }, + { + "epoch": 0.36, + "learning_rate": 1.4950452267157658e-05, + "loss": 1.1315, + "step": 14766 + }, + { + "epoch": 0.36, + "learning_rate": 1.4949775496343977e-05, + "loss": 0.8963, + "step": 14767 + }, + { + "epoch": 0.36, + "learning_rate": 1.4949098695501172e-05, + "loss": 0.9888, + "step": 14768 + }, + { + "epoch": 0.36, + "learning_rate": 1.494842186463335e-05, + "loss": 1.2447, + "step": 14769 + }, + { + "epoch": 0.36, + "learning_rate": 1.4947745003744616e-05, + "loss": 1.1144, + "step": 14770 + }, + { + "epoch": 0.36, + "learning_rate": 1.4947068112839077e-05, + "loss": 1.2646, + "step": 14771 + }, + { + "epoch": 0.36, + "learning_rate": 1.4946391191920839e-05, + "loss": 1.1988, + "step": 14772 + }, + { + "epoch": 0.36, + "learning_rate": 1.4945714240994007e-05, + "loss": 1.1, + "step": 14773 + }, + { + "epoch": 0.36, + "learning_rate": 1.4945037260062694e-05, + "loss": 1.1663, + "step": 14774 + }, + { + "epoch": 0.36, + "learning_rate": 1.4944360249131002e-05, + "loss": 0.7245, + "step": 14775 + }, + { + "epoch": 0.36, + "learning_rate": 1.4943683208203039e-05, + "loss": 1.1116, + "step": 14776 + }, + { + "epoch": 0.36, + "learning_rate": 1.4943006137282911e-05, + "loss": 0.6957, + "step": 14777 + }, + { + "epoch": 0.36, + "learning_rate": 1.4942329036374729e-05, + "loss": 1.0082, + "step": 14778 + }, + { + "epoch": 0.36, + "learning_rate": 1.4941651905482602e-05, + "loss": 0.982, + "step": 14779 + }, + { + "epoch": 0.36, + "learning_rate": 1.494097474461063e-05, + "loss": 0.8668, + "step": 14780 + }, + { + "epoch": 0.36, + "learning_rate": 1.4940297553762927e-05, + "loss": 1.2077, + "step": 14781 + }, + { + "epoch": 0.36, + "learning_rate": 1.4939620332943605e-05, + "loss": 1.1784, + "step": 14782 + }, + { + "epoch": 0.36, + "learning_rate": 1.4938943082156768e-05, + "loss": 1.0145, + "step": 14783 + }, + { + "epoch": 0.36, + "learning_rate": 1.493826580140652e-05, + "loss": 0.9111, + "step": 14784 + }, + { + "epoch": 0.36, + "learning_rate": 1.4937588490696979e-05, + "loss": 1.1511, + "step": 14785 + }, + { + "epoch": 0.36, + "learning_rate": 1.493691115003225e-05, + "loss": 1.0247, + "step": 14786 + }, + { + "epoch": 0.36, + "learning_rate": 1.493623377941644e-05, + "loss": 1.2408, + "step": 14787 + }, + { + "epoch": 0.36, + "learning_rate": 1.4935556378853662e-05, + "loss": 1.03, + "step": 14788 + }, + { + "epoch": 0.36, + "learning_rate": 1.4934878948348023e-05, + "loss": 1.0662, + "step": 14789 + }, + { + "epoch": 0.36, + "learning_rate": 1.4934201487903635e-05, + "loss": 1.1082, + "step": 14790 + }, + { + "epoch": 0.36, + "learning_rate": 1.4933523997524608e-05, + "loss": 1.2427, + "step": 14791 + }, + { + "epoch": 0.36, + "learning_rate": 1.493284647721505e-05, + "loss": 0.8312, + "step": 14792 + }, + { + "epoch": 0.36, + "learning_rate": 1.4932168926979074e-05, + "loss": 0.9338, + "step": 14793 + }, + { + "epoch": 0.36, + "learning_rate": 1.4931491346820787e-05, + "loss": 1.0964, + "step": 14794 + }, + { + "epoch": 0.36, + "learning_rate": 1.4930813736744307e-05, + "loss": 0.8632, + "step": 14795 + }, + { + "epoch": 0.36, + "learning_rate": 1.4930136096753736e-05, + "loss": 1.2669, + "step": 14796 + }, + { + "epoch": 0.36, + "learning_rate": 1.4929458426853187e-05, + "loss": 1.1808, + "step": 14797 + }, + { + "epoch": 0.36, + "learning_rate": 1.4928780727046777e-05, + "loss": 1.0278, + "step": 14798 + }, + { + "epoch": 0.36, + "learning_rate": 1.4928102997338614e-05, + "loss": 1.23, + "step": 14799 + }, + { + "epoch": 0.36, + "learning_rate": 1.492742523773281e-05, + "loss": 1.4164, + "step": 14800 + }, + { + "epoch": 0.36, + "learning_rate": 1.4926747448233472e-05, + "loss": 0.9299, + "step": 14801 + }, + { + "epoch": 0.36, + "learning_rate": 1.4926069628844717e-05, + "loss": 1.2277, + "step": 14802 + }, + { + "epoch": 0.36, + "learning_rate": 1.4925391779570657e-05, + "loss": 0.6957, + "step": 14803 + }, + { + "epoch": 0.36, + "learning_rate": 1.4924713900415407e-05, + "loss": 1.1618, + "step": 14804 + }, + { + "epoch": 0.36, + "learning_rate": 1.492403599138307e-05, + "loss": 0.9918, + "step": 14805 + }, + { + "epoch": 0.36, + "learning_rate": 1.4923358052477767e-05, + "loss": 1.0767, + "step": 14806 + }, + { + "epoch": 0.36, + "learning_rate": 1.4922680083703612e-05, + "loss": 1.0959, + "step": 14807 + }, + { + "epoch": 0.36, + "learning_rate": 1.4922002085064711e-05, + "loss": 1.136, + "step": 14808 + }, + { + "epoch": 0.36, + "learning_rate": 1.492132405656518e-05, + "loss": 0.9638, + "step": 14809 + }, + { + "epoch": 0.36, + "learning_rate": 1.4920645998209138e-05, + "loss": 1.001, + "step": 14810 + }, + { + "epoch": 0.36, + "learning_rate": 1.4919967910000693e-05, + "loss": 1.2683, + "step": 14811 + }, + { + "epoch": 0.36, + "learning_rate": 1.4919289791943958e-05, + "loss": 1.2962, + "step": 14812 + }, + { + "epoch": 0.36, + "learning_rate": 1.4918611644043048e-05, + "loss": 1.1466, + "step": 14813 + }, + { + "epoch": 0.36, + "learning_rate": 1.4917933466302082e-05, + "loss": 1.0784, + "step": 14814 + }, + { + "epoch": 0.36, + "learning_rate": 1.4917255258725169e-05, + "loss": 1.4503, + "step": 14815 + }, + { + "epoch": 0.36, + "learning_rate": 1.4916577021316425e-05, + "loss": 1.0657, + "step": 14816 + }, + { + "epoch": 0.36, + "learning_rate": 1.4915898754079963e-05, + "loss": 0.9579, + "step": 14817 + }, + { + "epoch": 0.36, + "learning_rate": 1.4915220457019902e-05, + "loss": 1.2828, + "step": 14818 + }, + { + "epoch": 0.36, + "learning_rate": 1.4914542130140352e-05, + "loss": 1.2966, + "step": 14819 + }, + { + "epoch": 0.36, + "learning_rate": 1.4913863773445432e-05, + "loss": 1.29, + "step": 14820 + }, + { + "epoch": 0.36, + "learning_rate": 1.4913185386939255e-05, + "loss": 0.9349, + "step": 14821 + }, + { + "epoch": 0.36, + "learning_rate": 1.491250697062594e-05, + "loss": 1.1336, + "step": 14822 + }, + { + "epoch": 0.36, + "learning_rate": 1.49118285245096e-05, + "loss": 1.0159, + "step": 14823 + }, + { + "epoch": 0.36, + "learning_rate": 1.491115004859435e-05, + "loss": 0.914, + "step": 14824 + }, + { + "epoch": 0.36, + "learning_rate": 1.491047154288431e-05, + "loss": 1.0794, + "step": 14825 + }, + { + "epoch": 0.36, + "learning_rate": 1.4909793007383595e-05, + "loss": 1.1221, + "step": 14826 + }, + { + "epoch": 0.36, + "learning_rate": 1.490911444209632e-05, + "loss": 0.9616, + "step": 14827 + }, + { + "epoch": 0.36, + "learning_rate": 1.4908435847026601e-05, + "loss": 1.0982, + "step": 14828 + }, + { + "epoch": 0.36, + "learning_rate": 1.4907757222178555e-05, + "loss": 1.1668, + "step": 14829 + }, + { + "epoch": 0.36, + "learning_rate": 1.4907078567556303e-05, + "loss": 0.8673, + "step": 14830 + }, + { + "epoch": 0.36, + "learning_rate": 1.490639988316396e-05, + "loss": 1.1298, + "step": 14831 + }, + { + "epoch": 0.36, + "learning_rate": 1.490572116900564e-05, + "loss": 1.0195, + "step": 14832 + }, + { + "epoch": 0.36, + "learning_rate": 1.4905042425085467e-05, + "loss": 1.3559, + "step": 14833 + }, + { + "epoch": 0.36, + "learning_rate": 1.4904363651407554e-05, + "loss": 1.141, + "step": 14834 + }, + { + "epoch": 0.36, + "learning_rate": 1.4903684847976019e-05, + "loss": 1.2265, + "step": 14835 + }, + { + "epoch": 0.36, + "learning_rate": 1.4903006014794983e-05, + "loss": 1.1764, + "step": 14836 + }, + { + "epoch": 0.36, + "learning_rate": 1.4902327151868562e-05, + "loss": 1.2067, + "step": 14837 + }, + { + "epoch": 0.36, + "learning_rate": 1.4901648259200875e-05, + "loss": 0.9948, + "step": 14838 + }, + { + "epoch": 0.36, + "learning_rate": 1.4900969336796043e-05, + "loss": 1.0347, + "step": 14839 + }, + { + "epoch": 0.36, + "learning_rate": 1.4900290384658181e-05, + "loss": 1.0582, + "step": 14840 + }, + { + "epoch": 0.36, + "learning_rate": 1.489961140279141e-05, + "loss": 1.191, + "step": 14841 + }, + { + "epoch": 0.36, + "learning_rate": 1.4898932391199847e-05, + "loss": 1.2873, + "step": 14842 + }, + { + "epoch": 0.36, + "learning_rate": 1.4898253349887617e-05, + "loss": 1.1989, + "step": 14843 + }, + { + "epoch": 0.36, + "learning_rate": 1.4897574278858833e-05, + "loss": 0.9314, + "step": 14844 + }, + { + "epoch": 0.36, + "learning_rate": 1.4896895178117621e-05, + "loss": 1.0396, + "step": 14845 + }, + { + "epoch": 0.36, + "learning_rate": 1.4896216047668097e-05, + "loss": 1.121, + "step": 14846 + }, + { + "epoch": 0.36, + "learning_rate": 1.4895536887514382e-05, + "loss": 1.2612, + "step": 14847 + }, + { + "epoch": 0.36, + "learning_rate": 1.4894857697660592e-05, + "loss": 0.9731, + "step": 14848 + }, + { + "epoch": 0.36, + "learning_rate": 1.4894178478110856e-05, + "loss": 1.1653, + "step": 14849 + }, + { + "epoch": 0.36, + "learning_rate": 1.489349922886929e-05, + "loss": 1.0039, + "step": 14850 + }, + { + "epoch": 0.36, + "learning_rate": 1.4892819949940016e-05, + "loss": 1.2541, + "step": 14851 + }, + { + "epoch": 0.36, + "learning_rate": 1.4892140641327151e-05, + "loss": 1.2681, + "step": 14852 + }, + { + "epoch": 0.36, + "learning_rate": 1.4891461303034822e-05, + "loss": 1.0303, + "step": 14853 + }, + { + "epoch": 0.36, + "learning_rate": 1.489078193506715e-05, + "loss": 1.1774, + "step": 14854 + }, + { + "epoch": 0.36, + "learning_rate": 1.489010253742825e-05, + "loss": 1.0893, + "step": 14855 + }, + { + "epoch": 0.36, + "learning_rate": 1.4889423110122248e-05, + "loss": 1.119, + "step": 14856 + }, + { + "epoch": 0.36, + "learning_rate": 1.4888743653153268e-05, + "loss": 1.4431, + "step": 14857 + }, + { + "epoch": 0.36, + "learning_rate": 1.4888064166525431e-05, + "loss": 1.0593, + "step": 14858 + }, + { + "epoch": 0.36, + "learning_rate": 1.4887384650242857e-05, + "loss": 1.0161, + "step": 14859 + }, + { + "epoch": 0.36, + "learning_rate": 1.488670510430967e-05, + "loss": 1.1602, + "step": 14860 + }, + { + "epoch": 0.36, + "learning_rate": 1.4886025528729993e-05, + "loss": 1.1514, + "step": 14861 + }, + { + "epoch": 0.36, + "learning_rate": 1.4885345923507948e-05, + "loss": 0.7997, + "step": 14862 + }, + { + "epoch": 0.36, + "learning_rate": 1.488466628864766e-05, + "loss": 1.2359, + "step": 14863 + }, + { + "epoch": 0.36, + "learning_rate": 1.4883986624153246e-05, + "loss": 1.0868, + "step": 14864 + }, + { + "epoch": 0.36, + "learning_rate": 1.4883306930028839e-05, + "loss": 1.0101, + "step": 14865 + }, + { + "epoch": 0.36, + "learning_rate": 1.4882627206278555e-05, + "loss": 1.2092, + "step": 14866 + }, + { + "epoch": 0.36, + "learning_rate": 1.488194745290652e-05, + "loss": 1.1734, + "step": 14867 + }, + { + "epoch": 0.36, + "learning_rate": 1.488126766991686e-05, + "loss": 1.3693, + "step": 14868 + }, + { + "epoch": 0.36, + "learning_rate": 1.4880587857313695e-05, + "loss": 1.05, + "step": 14869 + }, + { + "epoch": 0.36, + "learning_rate": 1.4879908015101152e-05, + "loss": 1.1508, + "step": 14870 + }, + { + "epoch": 0.36, + "learning_rate": 1.4879228143283354e-05, + "loss": 0.9781, + "step": 14871 + }, + { + "epoch": 0.36, + "learning_rate": 1.4878548241864428e-05, + "loss": 1.2553, + "step": 14872 + }, + { + "epoch": 0.36, + "learning_rate": 1.4877868310848497e-05, + "loss": 1.0172, + "step": 14873 + }, + { + "epoch": 0.36, + "learning_rate": 1.4877188350239687e-05, + "loss": 1.0375, + "step": 14874 + }, + { + "epoch": 0.36, + "learning_rate": 1.4876508360042121e-05, + "loss": 1.0631, + "step": 14875 + }, + { + "epoch": 0.36, + "learning_rate": 1.4875828340259925e-05, + "loss": 1.1858, + "step": 14876 + }, + { + "epoch": 0.36, + "learning_rate": 1.4875148290897227e-05, + "loss": 1.1379, + "step": 14877 + }, + { + "epoch": 0.36, + "learning_rate": 1.4874468211958152e-05, + "loss": 1.2266, + "step": 14878 + }, + { + "epoch": 0.36, + "learning_rate": 1.4873788103446826e-05, + "loss": 1.0197, + "step": 14879 + }, + { + "epoch": 0.36, + "learning_rate": 1.487310796536737e-05, + "loss": 1.3103, + "step": 14880 + }, + { + "epoch": 0.36, + "learning_rate": 1.4872427797723918e-05, + "loss": 1.3221, + "step": 14881 + }, + { + "epoch": 0.36, + "learning_rate": 1.4871747600520592e-05, + "loss": 1.3111, + "step": 14882 + }, + { + "epoch": 0.36, + "learning_rate": 1.4871067373761516e-05, + "loss": 0.8385, + "step": 14883 + }, + { + "epoch": 0.36, + "learning_rate": 1.4870387117450822e-05, + "loss": 1.0326, + "step": 14884 + }, + { + "epoch": 0.36, + "learning_rate": 1.4869706831592633e-05, + "loss": 1.1436, + "step": 14885 + }, + { + "epoch": 0.36, + "learning_rate": 1.4869026516191082e-05, + "loss": 1.1995, + "step": 14886 + }, + { + "epoch": 0.36, + "learning_rate": 1.4868346171250293e-05, + "loss": 1.1711, + "step": 14887 + }, + { + "epoch": 0.36, + "learning_rate": 1.4867665796774389e-05, + "loss": 1.1557, + "step": 14888 + }, + { + "epoch": 0.36, + "learning_rate": 1.4866985392767501e-05, + "loss": 1.299, + "step": 14889 + }, + { + "epoch": 0.36, + "learning_rate": 1.4866304959233762e-05, + "loss": 1.0868, + "step": 14890 + }, + { + "epoch": 0.36, + "learning_rate": 1.486562449617729e-05, + "loss": 0.9906, + "step": 14891 + }, + { + "epoch": 0.36, + "learning_rate": 1.4864944003602222e-05, + "loss": 1.0786, + "step": 14892 + }, + { + "epoch": 0.36, + "learning_rate": 1.4864263481512682e-05, + "loss": 1.1769, + "step": 14893 + }, + { + "epoch": 0.36, + "learning_rate": 1.48635829299128e-05, + "loss": 1.1596, + "step": 14894 + }, + { + "epoch": 0.36, + "learning_rate": 1.4862902348806704e-05, + "loss": 1.0818, + "step": 14895 + }, + { + "epoch": 0.36, + "learning_rate": 1.4862221738198523e-05, + "loss": 1.3308, + "step": 14896 + }, + { + "epoch": 0.36, + "learning_rate": 1.4861541098092389e-05, + "loss": 1.2447, + "step": 14897 + }, + { + "epoch": 0.36, + "learning_rate": 1.4860860428492427e-05, + "loss": 1.3788, + "step": 14898 + }, + { + "epoch": 0.36, + "learning_rate": 1.4860179729402768e-05, + "loss": 1.0105, + "step": 14899 + }, + { + "epoch": 0.36, + "learning_rate": 1.4859499000827539e-05, + "loss": 1.3327, + "step": 14900 + }, + { + "epoch": 0.36, + "learning_rate": 1.4858818242770876e-05, + "loss": 1.0596, + "step": 14901 + }, + { + "epoch": 0.36, + "learning_rate": 1.4858137455236906e-05, + "loss": 0.9532, + "step": 14902 + }, + { + "epoch": 0.36, + "learning_rate": 1.4857456638229757e-05, + "loss": 1.1828, + "step": 14903 + }, + { + "epoch": 0.36, + "learning_rate": 1.4856775791753561e-05, + "loss": 1.0156, + "step": 14904 + }, + { + "epoch": 0.36, + "learning_rate": 1.485609491581245e-05, + "loss": 1.1665, + "step": 14905 + }, + { + "epoch": 0.36, + "learning_rate": 1.4855414010410552e-05, + "loss": 0.9698, + "step": 14906 + }, + { + "epoch": 0.36, + "learning_rate": 1.4854733075551999e-05, + "loss": 1.2557, + "step": 14907 + }, + { + "epoch": 0.36, + "learning_rate": 1.4854052111240924e-05, + "loss": 0.9589, + "step": 14908 + }, + { + "epoch": 0.36, + "learning_rate": 1.4853371117481457e-05, + "loss": 0.9523, + "step": 14909 + }, + { + "epoch": 0.36, + "learning_rate": 1.4852690094277727e-05, + "loss": 1.1982, + "step": 14910 + }, + { + "epoch": 0.36, + "learning_rate": 1.4852009041633867e-05, + "loss": 1.1081, + "step": 14911 + }, + { + "epoch": 0.36, + "learning_rate": 1.4851327959554012e-05, + "loss": 0.9917, + "step": 14912 + }, + { + "epoch": 0.36, + "learning_rate": 1.485064684804229e-05, + "loss": 1.4253, + "step": 14913 + }, + { + "epoch": 0.36, + "learning_rate": 1.4849965707102836e-05, + "loss": 1.0772, + "step": 14914 + }, + { + "epoch": 0.36, + "learning_rate": 1.4849284536739776e-05, + "loss": 1.3649, + "step": 14915 + }, + { + "epoch": 0.36, + "learning_rate": 1.4848603336957252e-05, + "loss": 1.212, + "step": 14916 + }, + { + "epoch": 0.36, + "learning_rate": 1.4847922107759392e-05, + "loss": 0.9546, + "step": 14917 + }, + { + "epoch": 0.36, + "learning_rate": 1.4847240849150328e-05, + "loss": 0.6611, + "step": 14918 + }, + { + "epoch": 0.36, + "learning_rate": 1.4846559561134193e-05, + "loss": 0.9007, + "step": 14919 + }, + { + "epoch": 0.36, + "learning_rate": 1.4845878243715122e-05, + "loss": 1.0435, + "step": 14920 + }, + { + "epoch": 0.36, + "learning_rate": 1.4845196896897247e-05, + "loss": 1.1655, + "step": 14921 + }, + { + "epoch": 0.36, + "learning_rate": 1.4844515520684703e-05, + "loss": 0.9821, + "step": 14922 + }, + { + "epoch": 0.36, + "learning_rate": 1.4843834115081621e-05, + "loss": 0.9533, + "step": 14923 + }, + { + "epoch": 0.36, + "learning_rate": 1.484315268009214e-05, + "loss": 0.9723, + "step": 14924 + }, + { + "epoch": 0.36, + "learning_rate": 1.4842471215720389e-05, + "loss": 1.0405, + "step": 14925 + }, + { + "epoch": 0.36, + "learning_rate": 1.4841789721970503e-05, + "loss": 1.0369, + "step": 14926 + }, + { + "epoch": 0.36, + "learning_rate": 1.4841108198846618e-05, + "loss": 1.0632, + "step": 14927 + }, + { + "epoch": 0.36, + "learning_rate": 1.484042664635287e-05, + "loss": 0.8875, + "step": 14928 + }, + { + "epoch": 0.36, + "learning_rate": 1.4839745064493391e-05, + "loss": 0.7997, + "step": 14929 + }, + { + "epoch": 0.36, + "learning_rate": 1.4839063453272318e-05, + "loss": 1.0808, + "step": 14930 + }, + { + "epoch": 0.36, + "learning_rate": 1.4838381812693782e-05, + "loss": 1.1752, + "step": 14931 + }, + { + "epoch": 0.36, + "learning_rate": 1.4837700142761923e-05, + "loss": 1.2256, + "step": 14932 + }, + { + "epoch": 0.36, + "learning_rate": 1.4837018443480877e-05, + "loss": 1.286, + "step": 14933 + }, + { + "epoch": 0.36, + "learning_rate": 1.4836336714854778e-05, + "loss": 1.1017, + "step": 14934 + }, + { + "epoch": 0.36, + "learning_rate": 1.4835654956887759e-05, + "loss": 1.0202, + "step": 14935 + }, + { + "epoch": 0.36, + "learning_rate": 1.4834973169583962e-05, + "loss": 1.1563, + "step": 14936 + }, + { + "epoch": 0.36, + "learning_rate": 1.4834291352947518e-05, + "loss": 1.1737, + "step": 14937 + }, + { + "epoch": 0.36, + "learning_rate": 1.4833609506982568e-05, + "loss": 1.1299, + "step": 14938 + }, + { + "epoch": 0.36, + "learning_rate": 1.4832927631693242e-05, + "loss": 1.1241, + "step": 14939 + }, + { + "epoch": 0.36, + "learning_rate": 1.4832245727083683e-05, + "loss": 1.1563, + "step": 14940 + }, + { + "epoch": 0.36, + "learning_rate": 1.4831563793158026e-05, + "loss": 1.2516, + "step": 14941 + }, + { + "epoch": 0.36, + "learning_rate": 1.4830881829920407e-05, + "loss": 0.7956, + "step": 14942 + }, + { + "epoch": 0.36, + "learning_rate": 1.4830199837374964e-05, + "loss": 1.3575, + "step": 14943 + }, + { + "epoch": 0.36, + "learning_rate": 1.4829517815525836e-05, + "loss": 1.3533, + "step": 14944 + }, + { + "epoch": 0.36, + "learning_rate": 1.4828835764377158e-05, + "loss": 0.991, + "step": 14945 + }, + { + "epoch": 0.36, + "learning_rate": 1.4828153683933072e-05, + "loss": 0.9602, + "step": 14946 + }, + { + "epoch": 0.36, + "learning_rate": 1.482747157419771e-05, + "loss": 1.0254, + "step": 14947 + }, + { + "epoch": 0.36, + "learning_rate": 1.4826789435175215e-05, + "loss": 1.1201, + "step": 14948 + }, + { + "epoch": 0.36, + "learning_rate": 1.4826107266869723e-05, + "loss": 1.3006, + "step": 14949 + }, + { + "epoch": 0.36, + "learning_rate": 1.4825425069285374e-05, + "loss": 1.1588, + "step": 14950 + }, + { + "epoch": 0.36, + "learning_rate": 1.4824742842426305e-05, + "loss": 1.1031, + "step": 14951 + }, + { + "epoch": 0.36, + "learning_rate": 1.4824060586296658e-05, + "loss": 1.0396, + "step": 14952 + }, + { + "epoch": 0.36, + "learning_rate": 1.4823378300900571e-05, + "loss": 0.9834, + "step": 14953 + }, + { + "epoch": 0.36, + "learning_rate": 1.4822695986242178e-05, + "loss": 0.9209, + "step": 14954 + }, + { + "epoch": 0.36, + "learning_rate": 1.4822013642325625e-05, + "loss": 1.1652, + "step": 14955 + }, + { + "epoch": 0.36, + "learning_rate": 1.482133126915505e-05, + "loss": 1.2262, + "step": 14956 + }, + { + "epoch": 0.36, + "learning_rate": 1.4820648866734591e-05, + "loss": 1.0594, + "step": 14957 + }, + { + "epoch": 0.36, + "learning_rate": 1.4819966435068391e-05, + "loss": 1.2073, + "step": 14958 + }, + { + "epoch": 0.36, + "learning_rate": 1.4819283974160588e-05, + "loss": 1.1412, + "step": 14959 + }, + { + "epoch": 0.36, + "learning_rate": 1.4818601484015323e-05, + "loss": 0.9304, + "step": 14960 + }, + { + "epoch": 0.36, + "learning_rate": 1.4817918964636735e-05, + "loss": 1.1692, + "step": 14961 + }, + { + "epoch": 0.36, + "learning_rate": 1.4817236416028966e-05, + "loss": 1.2738, + "step": 14962 + }, + { + "epoch": 0.36, + "learning_rate": 1.4816553838196157e-05, + "loss": 1.2194, + "step": 14963 + }, + { + "epoch": 0.36, + "learning_rate": 1.481587123114245e-05, + "loss": 0.9904, + "step": 14964 + }, + { + "epoch": 0.36, + "learning_rate": 1.4815188594871983e-05, + "loss": 1.1118, + "step": 14965 + }, + { + "epoch": 0.36, + "learning_rate": 1.48145059293889e-05, + "loss": 1.0891, + "step": 14966 + }, + { + "epoch": 0.36, + "learning_rate": 1.4813823234697342e-05, + "loss": 1.0545, + "step": 14967 + }, + { + "epoch": 0.36, + "learning_rate": 1.4813140510801452e-05, + "loss": 0.8844, + "step": 14968 + }, + { + "epoch": 0.36, + "learning_rate": 1.481245775770537e-05, + "loss": 1.0396, + "step": 14969 + }, + { + "epoch": 0.36, + "learning_rate": 1.4811774975413238e-05, + "loss": 0.857, + "step": 14970 + }, + { + "epoch": 0.36, + "learning_rate": 1.4811092163929198e-05, + "loss": 1.1302, + "step": 14971 + }, + { + "epoch": 0.36, + "learning_rate": 1.4810409323257396e-05, + "loss": 1.0562, + "step": 14972 + }, + { + "epoch": 0.36, + "learning_rate": 1.4809726453401972e-05, + "loss": 0.9736, + "step": 14973 + }, + { + "epoch": 0.36, + "learning_rate": 1.4809043554367066e-05, + "loss": 1.0483, + "step": 14974 + }, + { + "epoch": 0.36, + "learning_rate": 1.4808360626156825e-05, + "loss": 1.0524, + "step": 14975 + }, + { + "epoch": 0.36, + "learning_rate": 1.4807677668775393e-05, + "loss": 1.1639, + "step": 14976 + }, + { + "epoch": 0.36, + "learning_rate": 1.4806994682226911e-05, + "loss": 1.1321, + "step": 14977 + }, + { + "epoch": 0.36, + "learning_rate": 1.4806311666515523e-05, + "loss": 1.0297, + "step": 14978 + }, + { + "epoch": 0.36, + "learning_rate": 1.4805628621645372e-05, + "loss": 1.215, + "step": 14979 + }, + { + "epoch": 0.36, + "learning_rate": 1.4804945547620601e-05, + "loss": 1.1475, + "step": 14980 + }, + { + "epoch": 0.36, + "learning_rate": 1.480426244444536e-05, + "loss": 1.0288, + "step": 14981 + }, + { + "epoch": 0.36, + "learning_rate": 1.4803579312123783e-05, + "loss": 1.3268, + "step": 14982 + }, + { + "epoch": 0.36, + "learning_rate": 1.4802896150660022e-05, + "loss": 0.8857, + "step": 14983 + }, + { + "epoch": 0.36, + "learning_rate": 1.4802212960058222e-05, + "loss": 1.0299, + "step": 14984 + }, + { + "epoch": 0.36, + "learning_rate": 1.4801529740322525e-05, + "loss": 1.211, + "step": 14985 + }, + { + "epoch": 0.36, + "learning_rate": 1.4800846491457078e-05, + "loss": 0.9994, + "step": 14986 + }, + { + "epoch": 0.36, + "learning_rate": 1.480016321346602e-05, + "loss": 0.9178, + "step": 14987 + }, + { + "epoch": 0.36, + "learning_rate": 1.4799479906353503e-05, + "loss": 1.0497, + "step": 14988 + }, + { + "epoch": 0.36, + "learning_rate": 1.479879657012367e-05, + "loss": 1.0842, + "step": 14989 + }, + { + "epoch": 0.36, + "learning_rate": 1.4798113204780668e-05, + "loss": 1.126, + "step": 14990 + }, + { + "epoch": 0.36, + "learning_rate": 1.4797429810328639e-05, + "loss": 0.914, + "step": 14991 + }, + { + "epoch": 0.36, + "learning_rate": 1.4796746386771735e-05, + "loss": 1.1509, + "step": 14992 + }, + { + "epoch": 0.36, + "learning_rate": 1.4796062934114099e-05, + "loss": 1.194, + "step": 14993 + }, + { + "epoch": 0.36, + "learning_rate": 1.4795379452359875e-05, + "loss": 1.4088, + "step": 14994 + }, + { + "epoch": 0.36, + "learning_rate": 1.4794695941513213e-05, + "loss": 1.1043, + "step": 14995 + }, + { + "epoch": 0.36, + "learning_rate": 1.479401240157826e-05, + "loss": 1.0918, + "step": 14996 + }, + { + "epoch": 0.36, + "learning_rate": 1.479332883255916e-05, + "loss": 1.1243, + "step": 14997 + }, + { + "epoch": 0.36, + "learning_rate": 1.479264523446006e-05, + "loss": 1.2937, + "step": 14998 + }, + { + "epoch": 0.36, + "learning_rate": 1.479196160728511e-05, + "loss": 0.8998, + "step": 14999 + }, + { + "epoch": 0.36, + "learning_rate": 1.4791277951038456e-05, + "loss": 1.2295, + "step": 15000 + }, + { + "epoch": 0.36, + "learning_rate": 1.4790594265724245e-05, + "loss": 1.2252, + "step": 15001 + }, + { + "epoch": 0.36, + "learning_rate": 1.4789910551346625e-05, + "loss": 0.9622, + "step": 15002 + }, + { + "epoch": 0.36, + "learning_rate": 1.4789226807909747e-05, + "loss": 1.2963, + "step": 15003 + }, + { + "epoch": 0.36, + "learning_rate": 1.4788543035417754e-05, + "loss": 1.2302, + "step": 15004 + }, + { + "epoch": 0.36, + "learning_rate": 1.4787859233874798e-05, + "loss": 1.1139, + "step": 15005 + }, + { + "epoch": 0.36, + "learning_rate": 1.4787175403285025e-05, + "loss": 1.304, + "step": 15006 + }, + { + "epoch": 0.36, + "learning_rate": 1.4786491543652587e-05, + "loss": 1.1968, + "step": 15007 + }, + { + "epoch": 0.36, + "learning_rate": 1.4785807654981627e-05, + "loss": 0.9992, + "step": 15008 + }, + { + "epoch": 0.36, + "learning_rate": 1.4785123737276301e-05, + "loss": 1.2431, + "step": 15009 + }, + { + "epoch": 0.36, + "learning_rate": 1.4784439790540753e-05, + "loss": 1.1517, + "step": 15010 + }, + { + "epoch": 0.36, + "learning_rate": 1.4783755814779135e-05, + "loss": 1.108, + "step": 15011 + }, + { + "epoch": 0.36, + "learning_rate": 1.4783071809995598e-05, + "loss": 1.0034, + "step": 15012 + }, + { + "epoch": 0.36, + "learning_rate": 1.4782387776194286e-05, + "loss": 0.8573, + "step": 15013 + }, + { + "epoch": 0.36, + "learning_rate": 1.4781703713379352e-05, + "loss": 1.1701, + "step": 15014 + }, + { + "epoch": 0.36, + "learning_rate": 1.4781019621554948e-05, + "loss": 1.0107, + "step": 15015 + }, + { + "epoch": 0.36, + "learning_rate": 1.4780335500725223e-05, + "loss": 1.2481, + "step": 15016 + }, + { + "epoch": 0.36, + "learning_rate": 1.4779651350894326e-05, + "loss": 1.1739, + "step": 15017 + }, + { + "epoch": 0.36, + "learning_rate": 1.4778967172066407e-05, + "loss": 1.0005, + "step": 15018 + }, + { + "epoch": 0.36, + "learning_rate": 1.477828296424562e-05, + "loss": 1.1428, + "step": 15019 + }, + { + "epoch": 0.36, + "learning_rate": 1.4777598727436115e-05, + "loss": 0.9887, + "step": 15020 + }, + { + "epoch": 0.36, + "learning_rate": 1.477691446164204e-05, + "loss": 1.1797, + "step": 15021 + }, + { + "epoch": 0.36, + "learning_rate": 1.4776230166867549e-05, + "loss": 1.2094, + "step": 15022 + }, + { + "epoch": 0.36, + "learning_rate": 1.4775545843116795e-05, + "loss": 1.1035, + "step": 15023 + }, + { + "epoch": 0.36, + "learning_rate": 1.4774861490393926e-05, + "loss": 1.012, + "step": 15024 + }, + { + "epoch": 0.36, + "learning_rate": 1.4774177108703096e-05, + "loss": 1.1145, + "step": 15025 + }, + { + "epoch": 0.36, + "learning_rate": 1.4773492698048454e-05, + "loss": 0.9559, + "step": 15026 + }, + { + "epoch": 0.36, + "learning_rate": 1.4772808258434159e-05, + "loss": 1.2412, + "step": 15027 + }, + { + "epoch": 0.36, + "learning_rate": 1.4772123789864356e-05, + "loss": 1.216, + "step": 15028 + }, + { + "epoch": 0.36, + "learning_rate": 1.4771439292343202e-05, + "loss": 1.2163, + "step": 15029 + }, + { + "epoch": 0.36, + "learning_rate": 1.4770754765874846e-05, + "loss": 1.1572, + "step": 15030 + }, + { + "epoch": 0.36, + "learning_rate": 1.4770070210463446e-05, + "loss": 1.0309, + "step": 15031 + }, + { + "epoch": 0.36, + "learning_rate": 1.4769385626113151e-05, + "loss": 1.1553, + "step": 15032 + }, + { + "epoch": 0.36, + "learning_rate": 1.4768701012828115e-05, + "loss": 1.0461, + "step": 15033 + }, + { + "epoch": 0.36, + "learning_rate": 1.4768016370612488e-05, + "loss": 0.8958, + "step": 15034 + }, + { + "epoch": 0.36, + "learning_rate": 1.4767331699470431e-05, + "loss": 1.1068, + "step": 15035 + }, + { + "epoch": 0.36, + "learning_rate": 1.4766646999406097e-05, + "loss": 1.2851, + "step": 15036 + }, + { + "epoch": 0.36, + "learning_rate": 1.476596227042363e-05, + "loss": 0.9596, + "step": 15037 + }, + { + "epoch": 0.36, + "learning_rate": 1.4765277512527195e-05, + "loss": 0.9669, + "step": 15038 + }, + { + "epoch": 0.36, + "learning_rate": 1.4764592725720942e-05, + "loss": 0.9912, + "step": 15039 + }, + { + "epoch": 0.36, + "learning_rate": 1.476390791000902e-05, + "loss": 1.2173, + "step": 15040 + }, + { + "epoch": 0.36, + "learning_rate": 1.4763223065395598e-05, + "loss": 1.2504, + "step": 15041 + }, + { + "epoch": 0.36, + "learning_rate": 1.4762538191884815e-05, + "loss": 1.3001, + "step": 15042 + }, + { + "epoch": 0.36, + "learning_rate": 1.4761853289480836e-05, + "loss": 0.994, + "step": 15043 + }, + { + "epoch": 0.36, + "learning_rate": 1.4761168358187817e-05, + "loss": 1.0932, + "step": 15044 + }, + { + "epoch": 0.36, + "learning_rate": 1.4760483398009903e-05, + "loss": 1.1708, + "step": 15045 + }, + { + "epoch": 0.36, + "learning_rate": 1.4759798408951258e-05, + "loss": 1.2817, + "step": 15046 + }, + { + "epoch": 0.36, + "learning_rate": 1.4759113391016038e-05, + "loss": 0.9727, + "step": 15047 + }, + { + "epoch": 0.36, + "learning_rate": 1.4758428344208395e-05, + "loss": 1.2324, + "step": 15048 + }, + { + "epoch": 0.36, + "learning_rate": 1.4757743268532487e-05, + "loss": 1.3596, + "step": 15049 + }, + { + "epoch": 0.36, + "learning_rate": 1.4757058163992466e-05, + "loss": 1.2332, + "step": 15050 + }, + { + "epoch": 0.36, + "learning_rate": 1.4756373030592497e-05, + "loss": 1.0959, + "step": 15051 + }, + { + "epoch": 0.36, + "learning_rate": 1.475568786833673e-05, + "loss": 1.1012, + "step": 15052 + }, + { + "epoch": 0.36, + "learning_rate": 1.4755002677229322e-05, + "loss": 1.0271, + "step": 15053 + }, + { + "epoch": 0.36, + "learning_rate": 1.4754317457274432e-05, + "loss": 0.9997, + "step": 15054 + }, + { + "epoch": 0.36, + "learning_rate": 1.4753632208476216e-05, + "loss": 1.2976, + "step": 15055 + }, + { + "epoch": 0.36, + "learning_rate": 1.4752946930838834e-05, + "loss": 0.8713, + "step": 15056 + }, + { + "epoch": 0.36, + "learning_rate": 1.475226162436644e-05, + "loss": 1.0883, + "step": 15057 + }, + { + "epoch": 0.36, + "learning_rate": 1.475157628906319e-05, + "loss": 1.2307, + "step": 15058 + }, + { + "epoch": 0.36, + "learning_rate": 1.4750890924933245e-05, + "loss": 0.6909, + "step": 15059 + }, + { + "epoch": 0.36, + "learning_rate": 1.4750205531980764e-05, + "loss": 1.2127, + "step": 15060 + }, + { + "epoch": 0.36, + "learning_rate": 1.4749520110209903e-05, + "loss": 1.134, + "step": 15061 + }, + { + "epoch": 0.36, + "learning_rate": 1.4748834659624817e-05, + "loss": 0.9937, + "step": 15062 + }, + { + "epoch": 0.36, + "learning_rate": 1.474814918022967e-05, + "loss": 1.3327, + "step": 15063 + }, + { + "epoch": 0.36, + "learning_rate": 1.4747463672028621e-05, + "loss": 1.0189, + "step": 15064 + }, + { + "epoch": 0.36, + "learning_rate": 1.4746778135025828e-05, + "loss": 1.4043, + "step": 15065 + }, + { + "epoch": 0.36, + "learning_rate": 1.4746092569225444e-05, + "loss": 0.945, + "step": 15066 + }, + { + "epoch": 0.36, + "learning_rate": 1.4745406974631635e-05, + "loss": 1.1354, + "step": 15067 + }, + { + "epoch": 0.36, + "learning_rate": 1.4744721351248558e-05, + "loss": 1.1076, + "step": 15068 + }, + { + "epoch": 0.36, + "learning_rate": 1.4744035699080375e-05, + "loss": 1.0064, + "step": 15069 + }, + { + "epoch": 0.36, + "learning_rate": 1.4743350018131237e-05, + "loss": 0.9936, + "step": 15070 + }, + { + "epoch": 0.36, + "learning_rate": 1.4742664308405315e-05, + "loss": 0.9154, + "step": 15071 + }, + { + "epoch": 0.36, + "learning_rate": 1.4741978569906762e-05, + "loss": 0.7978, + "step": 15072 + }, + { + "epoch": 0.36, + "learning_rate": 1.4741292802639744e-05, + "loss": 1.3542, + "step": 15073 + }, + { + "epoch": 0.36, + "learning_rate": 1.4740607006608412e-05, + "loss": 1.3143, + "step": 15074 + }, + { + "epoch": 0.36, + "learning_rate": 1.4739921181816937e-05, + "loss": 1.1896, + "step": 15075 + }, + { + "epoch": 0.36, + "learning_rate": 1.4739235328269474e-05, + "loss": 1.2181, + "step": 15076 + }, + { + "epoch": 0.36, + "learning_rate": 1.4738549445970186e-05, + "loss": 1.0696, + "step": 15077 + }, + { + "epoch": 0.36, + "learning_rate": 1.473786353492323e-05, + "loss": 1.1348, + "step": 15078 + }, + { + "epoch": 0.36, + "learning_rate": 1.4737177595132773e-05, + "loss": 1.0943, + "step": 15079 + }, + { + "epoch": 0.36, + "learning_rate": 1.473649162660297e-05, + "loss": 1.1919, + "step": 15080 + }, + { + "epoch": 0.36, + "learning_rate": 1.473580562933799e-05, + "loss": 1.2236, + "step": 15081 + }, + { + "epoch": 0.36, + "learning_rate": 1.4735119603341991e-05, + "loss": 0.9422, + "step": 15082 + }, + { + "epoch": 0.36, + "learning_rate": 1.4734433548619133e-05, + "loss": 1.1207, + "step": 15083 + }, + { + "epoch": 0.36, + "learning_rate": 1.4733747465173584e-05, + "loss": 1.2689, + "step": 15084 + }, + { + "epoch": 0.36, + "learning_rate": 1.47330613530095e-05, + "loss": 1.2099, + "step": 15085 + }, + { + "epoch": 0.36, + "learning_rate": 1.4732375212131045e-05, + "loss": 1.1711, + "step": 15086 + }, + { + "epoch": 0.36, + "learning_rate": 1.4731689042542385e-05, + "loss": 1.2635, + "step": 15087 + }, + { + "epoch": 0.36, + "learning_rate": 1.473100284424768e-05, + "loss": 1.1346, + "step": 15088 + }, + { + "epoch": 0.36, + "learning_rate": 1.4730316617251092e-05, + "loss": 1.0432, + "step": 15089 + }, + { + "epoch": 0.36, + "learning_rate": 1.4729630361556789e-05, + "loss": 1.1466, + "step": 15090 + }, + { + "epoch": 0.36, + "learning_rate": 1.4728944077168927e-05, + "loss": 1.0508, + "step": 15091 + }, + { + "epoch": 0.36, + "learning_rate": 1.4728257764091676e-05, + "loss": 0.9442, + "step": 15092 + }, + { + "epoch": 0.36, + "learning_rate": 1.4727571422329196e-05, + "loss": 1.2813, + "step": 15093 + }, + { + "epoch": 0.36, + "learning_rate": 1.4726885051885654e-05, + "loss": 0.9626, + "step": 15094 + }, + { + "epoch": 0.36, + "learning_rate": 1.4726198652765212e-05, + "loss": 0.9976, + "step": 15095 + }, + { + "epoch": 0.36, + "learning_rate": 1.4725512224972033e-05, + "loss": 1.1648, + "step": 15096 + }, + { + "epoch": 0.36, + "learning_rate": 1.4724825768510283e-05, + "loss": 1.0643, + "step": 15097 + }, + { + "epoch": 0.36, + "learning_rate": 1.4724139283384127e-05, + "loss": 1.0291, + "step": 15098 + }, + { + "epoch": 0.36, + "learning_rate": 1.472345276959773e-05, + "loss": 1.0571, + "step": 15099 + }, + { + "epoch": 0.36, + "learning_rate": 1.4722766227155256e-05, + "loss": 1.0104, + "step": 15100 + }, + { + "epoch": 0.36, + "learning_rate": 1.4722079656060869e-05, + "loss": 0.9884, + "step": 15101 + }, + { + "epoch": 0.36, + "learning_rate": 1.4721393056318736e-05, + "loss": 1.1846, + "step": 15102 + }, + { + "epoch": 0.36, + "learning_rate": 1.4720706427933023e-05, + "loss": 0.898, + "step": 15103 + }, + { + "epoch": 0.36, + "learning_rate": 1.4720019770907892e-05, + "loss": 1.1773, + "step": 15104 + }, + { + "epoch": 0.36, + "learning_rate": 1.4719333085247512e-05, + "loss": 1.0581, + "step": 15105 + }, + { + "epoch": 0.36, + "learning_rate": 1.4718646370956049e-05, + "loss": 1.302, + "step": 15106 + }, + { + "epoch": 0.36, + "learning_rate": 1.4717959628037668e-05, + "loss": 1.2804, + "step": 15107 + }, + { + "epoch": 0.36, + "learning_rate": 1.4717272856496537e-05, + "loss": 1.1091, + "step": 15108 + }, + { + "epoch": 0.36, + "learning_rate": 1.4716586056336818e-05, + "loss": 1.3108, + "step": 15109 + }, + { + "epoch": 0.36, + "learning_rate": 1.4715899227562682e-05, + "loss": 1.1623, + "step": 15110 + }, + { + "epoch": 0.36, + "learning_rate": 1.4715212370178296e-05, + "loss": 1.2123, + "step": 15111 + }, + { + "epoch": 0.36, + "learning_rate": 1.4714525484187825e-05, + "loss": 0.8877, + "step": 15112 + }, + { + "epoch": 0.36, + "learning_rate": 1.4713838569595435e-05, + "loss": 1.433, + "step": 15113 + }, + { + "epoch": 0.36, + "learning_rate": 1.4713151626405297e-05, + "loss": 0.8699, + "step": 15114 + }, + { + "epoch": 0.36, + "learning_rate": 1.4712464654621577e-05, + "loss": 1.1595, + "step": 15115 + }, + { + "epoch": 0.36, + "learning_rate": 1.4711777654248442e-05, + "loss": 1.2442, + "step": 15116 + }, + { + "epoch": 0.36, + "learning_rate": 1.4711090625290059e-05, + "loss": 0.6276, + "step": 15117 + }, + { + "epoch": 0.36, + "learning_rate": 1.47104035677506e-05, + "loss": 1.1082, + "step": 15118 + }, + { + "epoch": 0.36, + "learning_rate": 1.470971648163423e-05, + "loss": 0.861, + "step": 15119 + }, + { + "epoch": 0.36, + "learning_rate": 1.4709029366945115e-05, + "loss": 1.0307, + "step": 15120 + }, + { + "epoch": 0.36, + "learning_rate": 1.4708342223687428e-05, + "loss": 1.1559, + "step": 15121 + }, + { + "epoch": 0.36, + "learning_rate": 1.4707655051865337e-05, + "loss": 1.061, + "step": 15122 + }, + { + "epoch": 0.36, + "learning_rate": 1.4706967851483009e-05, + "loss": 1.0262, + "step": 15123 + }, + { + "epoch": 0.36, + "learning_rate": 1.4706280622544615e-05, + "loss": 1.0159, + "step": 15124 + }, + { + "epoch": 0.36, + "learning_rate": 1.4705593365054322e-05, + "loss": 1.2974, + "step": 15125 + }, + { + "epoch": 0.36, + "learning_rate": 1.4704906079016304e-05, + "loss": 1.1454, + "step": 15126 + }, + { + "epoch": 0.36, + "learning_rate": 1.4704218764434726e-05, + "loss": 1.1712, + "step": 15127 + }, + { + "epoch": 0.36, + "learning_rate": 1.470353142131376e-05, + "loss": 1.0059, + "step": 15128 + }, + { + "epoch": 0.36, + "learning_rate": 1.470284404965757e-05, + "loss": 1.2536, + "step": 15129 + }, + { + "epoch": 0.36, + "learning_rate": 1.4702156649470336e-05, + "loss": 0.945, + "step": 15130 + }, + { + "epoch": 0.36, + "learning_rate": 1.4701469220756224e-05, + "loss": 1.0858, + "step": 15131 + }, + { + "epoch": 0.36, + "learning_rate": 1.4700781763519403e-05, + "loss": 1.0014, + "step": 15132 + }, + { + "epoch": 0.36, + "learning_rate": 1.4700094277764044e-05, + "loss": 1.2336, + "step": 15133 + }, + { + "epoch": 0.36, + "learning_rate": 1.469940676349432e-05, + "loss": 1.0032, + "step": 15134 + }, + { + "epoch": 0.36, + "learning_rate": 1.4698719220714401e-05, + "loss": 0.7849, + "step": 15135 + }, + { + "epoch": 0.36, + "learning_rate": 1.4698031649428457e-05, + "loss": 1.4368, + "step": 15136 + }, + { + "epoch": 0.36, + "learning_rate": 1.4697344049640658e-05, + "loss": 0.8399, + "step": 15137 + }, + { + "epoch": 0.36, + "learning_rate": 1.469665642135518e-05, + "loss": 1.0861, + "step": 15138 + }, + { + "epoch": 0.36, + "learning_rate": 1.4695968764576191e-05, + "loss": 1.3865, + "step": 15139 + }, + { + "epoch": 0.36, + "learning_rate": 1.4695281079307865e-05, + "loss": 1.1713, + "step": 15140 + }, + { + "epoch": 0.36, + "learning_rate": 1.469459336555437e-05, + "loss": 1.0725, + "step": 15141 + }, + { + "epoch": 0.36, + "learning_rate": 1.4693905623319886e-05, + "loss": 1.1811, + "step": 15142 + }, + { + "epoch": 0.36, + "learning_rate": 1.4693217852608579e-05, + "loss": 1.1598, + "step": 15143 + }, + { + "epoch": 0.36, + "learning_rate": 1.4692530053424624e-05, + "loss": 1.0969, + "step": 15144 + }, + { + "epoch": 0.36, + "learning_rate": 1.4691842225772189e-05, + "loss": 1.228, + "step": 15145 + }, + { + "epoch": 0.36, + "learning_rate": 1.4691154369655454e-05, + "loss": 0.9583, + "step": 15146 + }, + { + "epoch": 0.36, + "learning_rate": 1.4690466485078589e-05, + "loss": 0.9888, + "step": 15147 + }, + { + "epoch": 0.36, + "learning_rate": 1.4689778572045767e-05, + "loss": 1.2152, + "step": 15148 + }, + { + "epoch": 0.36, + "learning_rate": 1.468909063056116e-05, + "loss": 1.2708, + "step": 15149 + }, + { + "epoch": 0.36, + "learning_rate": 1.4688402660628945e-05, + "loss": 0.6305, + "step": 15150 + }, + { + "epoch": 0.36, + "learning_rate": 1.4687714662253293e-05, + "loss": 0.9776, + "step": 15151 + }, + { + "epoch": 0.36, + "learning_rate": 1.4687026635438377e-05, + "loss": 1.1274, + "step": 15152 + }, + { + "epoch": 0.36, + "learning_rate": 1.4686338580188374e-05, + "loss": 1.0661, + "step": 15153 + }, + { + "epoch": 0.36, + "learning_rate": 1.4685650496507458e-05, + "loss": 1.2671, + "step": 15154 + }, + { + "epoch": 0.36, + "learning_rate": 1.4684962384399803e-05, + "loss": 1.1478, + "step": 15155 + }, + { + "epoch": 0.36, + "learning_rate": 1.4684274243869583e-05, + "loss": 1.1058, + "step": 15156 + }, + { + "epoch": 0.36, + "learning_rate": 1.468358607492097e-05, + "loss": 1.0217, + "step": 15157 + }, + { + "epoch": 0.36, + "learning_rate": 1.4682897877558144e-05, + "loss": 0.9332, + "step": 15158 + }, + { + "epoch": 0.36, + "learning_rate": 1.468220965178528e-05, + "loss": 0.8347, + "step": 15159 + }, + { + "epoch": 0.36, + "learning_rate": 1.4681521397606551e-05, + "loss": 0.9217, + "step": 15160 + }, + { + "epoch": 0.36, + "learning_rate": 1.468083311502613e-05, + "loss": 1.0757, + "step": 15161 + }, + { + "epoch": 0.36, + "learning_rate": 1.4680144804048198e-05, + "loss": 1.255, + "step": 15162 + }, + { + "epoch": 0.36, + "learning_rate": 1.4679456464676927e-05, + "loss": 1.1022, + "step": 15163 + }, + { + "epoch": 0.36, + "learning_rate": 1.4678768096916493e-05, + "loss": 1.2107, + "step": 15164 + }, + { + "epoch": 0.36, + "learning_rate": 1.4678079700771076e-05, + "loss": 1.2926, + "step": 15165 + }, + { + "epoch": 0.36, + "learning_rate": 1.4677391276244847e-05, + "loss": 1.2082, + "step": 15166 + }, + { + "epoch": 0.36, + "learning_rate": 1.4676702823341988e-05, + "loss": 1.2432, + "step": 15167 + }, + { + "epoch": 0.36, + "learning_rate": 1.4676014342066673e-05, + "loss": 1.0865, + "step": 15168 + }, + { + "epoch": 0.36, + "learning_rate": 1.4675325832423077e-05, + "loss": 1.2022, + "step": 15169 + }, + { + "epoch": 0.36, + "learning_rate": 1.467463729441538e-05, + "loss": 1.2023, + "step": 15170 + }, + { + "epoch": 0.36, + "learning_rate": 1.467394872804776e-05, + "loss": 1.1994, + "step": 15171 + }, + { + "epoch": 0.36, + "learning_rate": 1.4673260133324388e-05, + "loss": 0.9745, + "step": 15172 + }, + { + "epoch": 0.36, + "learning_rate": 1.4672571510249449e-05, + "loss": 1.1698, + "step": 15173 + }, + { + "epoch": 0.36, + "learning_rate": 1.4671882858827117e-05, + "loss": 1.0691, + "step": 15174 + }, + { + "epoch": 0.36, + "learning_rate": 1.4671194179061572e-05, + "loss": 1.23, + "step": 15175 + }, + { + "epoch": 0.36, + "learning_rate": 1.4670505470956988e-05, + "loss": 1.4584, + "step": 15176 + }, + { + "epoch": 0.36, + "learning_rate": 1.4669816734517548e-05, + "loss": 1.2818, + "step": 15177 + }, + { + "epoch": 0.37, + "learning_rate": 1.4669127969747427e-05, + "loss": 1.1629, + "step": 15178 + }, + { + "epoch": 0.37, + "learning_rate": 1.4668439176650808e-05, + "loss": 1.1579, + "step": 15179 + }, + { + "epoch": 0.37, + "learning_rate": 1.4667750355231863e-05, + "loss": 1.0863, + "step": 15180 + }, + { + "epoch": 0.37, + "learning_rate": 1.4667061505494776e-05, + "loss": 1.1162, + "step": 15181 + }, + { + "epoch": 0.37, + "learning_rate": 1.4666372627443724e-05, + "loss": 1.2568, + "step": 15182 + }, + { + "epoch": 0.37, + "learning_rate": 1.4665683721082886e-05, + "loss": 1.0866, + "step": 15183 + }, + { + "epoch": 0.37, + "learning_rate": 1.4664994786416442e-05, + "loss": 1.187, + "step": 15184 + }, + { + "epoch": 0.37, + "learning_rate": 1.4664305823448575e-05, + "loss": 1.1142, + "step": 15185 + }, + { + "epoch": 0.37, + "learning_rate": 1.466361683218346e-05, + "loss": 1.014, + "step": 15186 + }, + { + "epoch": 0.37, + "learning_rate": 1.4662927812625278e-05, + "loss": 1.2253, + "step": 15187 + }, + { + "epoch": 0.37, + "learning_rate": 1.4662238764778209e-05, + "loss": 1.125, + "step": 15188 + }, + { + "epoch": 0.37, + "learning_rate": 1.4661549688646437e-05, + "loss": 1.2538, + "step": 15189 + }, + { + "epoch": 0.37, + "learning_rate": 1.4660860584234138e-05, + "loss": 0.8502, + "step": 15190 + }, + { + "epoch": 0.37, + "learning_rate": 1.4660171451545493e-05, + "loss": 1.1888, + "step": 15191 + }, + { + "epoch": 0.37, + "learning_rate": 1.4659482290584683e-05, + "loss": 1.1147, + "step": 15192 + }, + { + "epoch": 0.37, + "learning_rate": 1.4658793101355895e-05, + "loss": 1.3888, + "step": 15193 + }, + { + "epoch": 0.37, + "learning_rate": 1.4658103883863301e-05, + "loss": 1.1545, + "step": 15194 + }, + { + "epoch": 0.37, + "learning_rate": 1.4657414638111088e-05, + "loss": 1.049, + "step": 15195 + }, + { + "epoch": 0.37, + "learning_rate": 1.4656725364103434e-05, + "loss": 1.0854, + "step": 15196 + }, + { + "epoch": 0.37, + "learning_rate": 1.4656036061844523e-05, + "loss": 1.2622, + "step": 15197 + }, + { + "epoch": 0.37, + "learning_rate": 1.4655346731338537e-05, + "loss": 1.0882, + "step": 15198 + }, + { + "epoch": 0.37, + "learning_rate": 1.465465737258966e-05, + "loss": 1.2319, + "step": 15199 + }, + { + "epoch": 0.37, + "learning_rate": 1.4653967985602066e-05, + "loss": 1.0994, + "step": 15200 + }, + { + "epoch": 0.37, + "learning_rate": 1.4653278570379948e-05, + "loss": 1.226, + "step": 15201 + }, + { + "epoch": 0.37, + "learning_rate": 1.4652589126927481e-05, + "loss": 1.0562, + "step": 15202 + }, + { + "epoch": 0.37, + "learning_rate": 1.4651899655248852e-05, + "loss": 1.4651, + "step": 15203 + }, + { + "epoch": 0.37, + "learning_rate": 1.4651210155348237e-05, + "loss": 1.0903, + "step": 15204 + }, + { + "epoch": 0.37, + "learning_rate": 1.465052062722983e-05, + "loss": 0.7409, + "step": 15205 + }, + { + "epoch": 0.37, + "learning_rate": 1.4649831070897804e-05, + "loss": 1.1303, + "step": 15206 + }, + { + "epoch": 0.37, + "learning_rate": 1.464914148635635e-05, + "loss": 1.1144, + "step": 15207 + }, + { + "epoch": 0.37, + "learning_rate": 1.464845187360964e-05, + "loss": 1.1381, + "step": 15208 + }, + { + "epoch": 0.37, + "learning_rate": 1.4647762232661875e-05, + "loss": 1.3162, + "step": 15209 + }, + { + "epoch": 0.37, + "learning_rate": 1.4647072563517228e-05, + "loss": 1.0858, + "step": 15210 + }, + { + "epoch": 0.37, + "learning_rate": 1.4646382866179885e-05, + "loss": 1.2115, + "step": 15211 + }, + { + "epoch": 0.37, + "learning_rate": 1.4645693140654028e-05, + "loss": 0.9373, + "step": 15212 + }, + { + "epoch": 0.37, + "learning_rate": 1.4645003386943843e-05, + "loss": 1.0509, + "step": 15213 + }, + { + "epoch": 0.37, + "learning_rate": 1.4644313605053517e-05, + "loss": 1.002, + "step": 15214 + }, + { + "epoch": 0.37, + "learning_rate": 1.4643623794987234e-05, + "loss": 1.1174, + "step": 15215 + }, + { + "epoch": 0.37, + "learning_rate": 1.4642933956749174e-05, + "loss": 1.1194, + "step": 15216 + }, + { + "epoch": 0.37, + "learning_rate": 1.4642244090343527e-05, + "loss": 1.2837, + "step": 15217 + }, + { + "epoch": 0.37, + "learning_rate": 1.464155419577448e-05, + "loss": 1.0318, + "step": 15218 + }, + { + "epoch": 0.37, + "learning_rate": 1.4640864273046211e-05, + "loss": 1.0226, + "step": 15219 + }, + { + "epoch": 0.37, + "learning_rate": 1.4640174322162913e-05, + "loss": 0.8541, + "step": 15220 + }, + { + "epoch": 0.37, + "learning_rate": 1.4639484343128767e-05, + "loss": 0.6876, + "step": 15221 + }, + { + "epoch": 0.37, + "learning_rate": 1.4638794335947963e-05, + "loss": 1.4128, + "step": 15222 + }, + { + "epoch": 0.37, + "learning_rate": 1.4638104300624686e-05, + "loss": 1.1104, + "step": 15223 + }, + { + "epoch": 0.37, + "learning_rate": 1.4637414237163116e-05, + "loss": 1.0292, + "step": 15224 + }, + { + "epoch": 0.37, + "learning_rate": 1.4636724145567447e-05, + "loss": 1.0995, + "step": 15225 + }, + { + "epoch": 0.37, + "learning_rate": 1.4636034025841865e-05, + "loss": 0.8853, + "step": 15226 + }, + { + "epoch": 0.37, + "learning_rate": 1.4635343877990553e-05, + "loss": 1.1066, + "step": 15227 + }, + { + "epoch": 0.37, + "learning_rate": 1.46346537020177e-05, + "loss": 0.5406, + "step": 15228 + }, + { + "epoch": 0.37, + "learning_rate": 1.4633963497927494e-05, + "loss": 0.9352, + "step": 15229 + }, + { + "epoch": 0.37, + "learning_rate": 1.4633273265724121e-05, + "loss": 1.4136, + "step": 15230 + }, + { + "epoch": 0.37, + "learning_rate": 1.4632583005411771e-05, + "loss": 1.1441, + "step": 15231 + }, + { + "epoch": 0.37, + "learning_rate": 1.4631892716994624e-05, + "loss": 1.0191, + "step": 15232 + }, + { + "epoch": 0.37, + "learning_rate": 1.4631202400476878e-05, + "loss": 1.0886, + "step": 15233 + }, + { + "epoch": 0.37, + "learning_rate": 1.4630512055862718e-05, + "loss": 0.9233, + "step": 15234 + }, + { + "epoch": 0.37, + "learning_rate": 1.4629821683156327e-05, + "loss": 1.0428, + "step": 15235 + }, + { + "epoch": 0.37, + "learning_rate": 1.4629131282361897e-05, + "loss": 0.9165, + "step": 15236 + }, + { + "epoch": 0.37, + "learning_rate": 1.4628440853483616e-05, + "loss": 1.1859, + "step": 15237 + }, + { + "epoch": 0.37, + "learning_rate": 1.4627750396525675e-05, + "loss": 1.1595, + "step": 15238 + }, + { + "epoch": 0.37, + "learning_rate": 1.462705991149226e-05, + "loss": 1.405, + "step": 15239 + }, + { + "epoch": 0.37, + "learning_rate": 1.462636939838756e-05, + "loss": 1.0208, + "step": 15240 + }, + { + "epoch": 0.37, + "learning_rate": 1.4625678857215767e-05, + "loss": 1.2643, + "step": 15241 + }, + { + "epoch": 0.37, + "learning_rate": 1.4624988287981068e-05, + "loss": 1.0254, + "step": 15242 + }, + { + "epoch": 0.37, + "learning_rate": 1.4624297690687652e-05, + "loss": 1.2025, + "step": 15243 + }, + { + "epoch": 0.37, + "learning_rate": 1.4623607065339709e-05, + "loss": 0.9308, + "step": 15244 + }, + { + "epoch": 0.37, + "learning_rate": 1.462291641194143e-05, + "loss": 1.3841, + "step": 15245 + }, + { + "epoch": 0.37, + "learning_rate": 1.4622225730497006e-05, + "loss": 1.2799, + "step": 15246 + }, + { + "epoch": 0.37, + "learning_rate": 1.4621535021010625e-05, + "loss": 0.923, + "step": 15247 + }, + { + "epoch": 0.37, + "learning_rate": 1.4620844283486477e-05, + "loss": 1.2978, + "step": 15248 + }, + { + "epoch": 0.37, + "learning_rate": 1.4620153517928754e-05, + "loss": 1.0312, + "step": 15249 + }, + { + "epoch": 0.37, + "learning_rate": 1.4619462724341647e-05, + "loss": 0.8279, + "step": 15250 + }, + { + "epoch": 0.37, + "learning_rate": 1.4618771902729343e-05, + "loss": 1.1321, + "step": 15251 + }, + { + "epoch": 0.37, + "learning_rate": 1.461808105309604e-05, + "loss": 1.3907, + "step": 15252 + }, + { + "epoch": 0.37, + "learning_rate": 1.4617390175445927e-05, + "loss": 0.8235, + "step": 15253 + }, + { + "epoch": 0.37, + "learning_rate": 1.461669926978319e-05, + "loss": 1.3598, + "step": 15254 + }, + { + "epoch": 0.37, + "learning_rate": 1.4616008336112027e-05, + "loss": 1.1999, + "step": 15255 + }, + { + "epoch": 0.37, + "learning_rate": 1.4615317374436624e-05, + "loss": 1.0356, + "step": 15256 + }, + { + "epoch": 0.37, + "learning_rate": 1.461462638476118e-05, + "loss": 0.8712, + "step": 15257 + }, + { + "epoch": 0.37, + "learning_rate": 1.4613935367089881e-05, + "loss": 1.2786, + "step": 15258 + }, + { + "epoch": 0.37, + "learning_rate": 1.4613244321426921e-05, + "loss": 1.3821, + "step": 15259 + }, + { + "epoch": 0.37, + "learning_rate": 1.4612553247776493e-05, + "loss": 1.3097, + "step": 15260 + }, + { + "epoch": 0.37, + "learning_rate": 1.461186214614279e-05, + "loss": 0.9112, + "step": 15261 + }, + { + "epoch": 0.37, + "learning_rate": 1.4611171016530006e-05, + "loss": 1.3568, + "step": 15262 + }, + { + "epoch": 0.37, + "learning_rate": 1.4610479858942328e-05, + "loss": 1.0149, + "step": 15263 + }, + { + "epoch": 0.37, + "learning_rate": 1.4609788673383955e-05, + "loss": 1.0176, + "step": 15264 + }, + { + "epoch": 0.37, + "learning_rate": 1.4609097459859078e-05, + "loss": 1.1292, + "step": 15265 + }, + { + "epoch": 0.37, + "learning_rate": 1.4608406218371894e-05, + "loss": 1.0682, + "step": 15266 + }, + { + "epoch": 0.37, + "learning_rate": 1.4607714948926588e-05, + "loss": 1.0301, + "step": 15267 + }, + { + "epoch": 0.37, + "learning_rate": 1.4607023651527364e-05, + "loss": 1.0316, + "step": 15268 + }, + { + "epoch": 0.37, + "learning_rate": 1.4606332326178407e-05, + "loss": 1.1187, + "step": 15269 + }, + { + "epoch": 0.37, + "learning_rate": 1.4605640972883919e-05, + "loss": 0.8685, + "step": 15270 + }, + { + "epoch": 0.37, + "learning_rate": 1.4604949591648087e-05, + "loss": 1.1682, + "step": 15271 + }, + { + "epoch": 0.37, + "learning_rate": 1.4604258182475112e-05, + "loss": 1.295, + "step": 15272 + }, + { + "epoch": 0.37, + "learning_rate": 1.4603566745369187e-05, + "loss": 1.1417, + "step": 15273 + }, + { + "epoch": 0.37, + "learning_rate": 1.46028752803345e-05, + "loss": 1.0642, + "step": 15274 + }, + { + "epoch": 0.37, + "learning_rate": 1.4602183787375255e-05, + "loss": 1.2639, + "step": 15275 + }, + { + "epoch": 0.37, + "learning_rate": 1.4601492266495643e-05, + "loss": 1.1416, + "step": 15276 + }, + { + "epoch": 0.37, + "learning_rate": 1.4600800717699862e-05, + "loss": 0.9947, + "step": 15277 + }, + { + "epoch": 0.37, + "learning_rate": 1.4600109140992102e-05, + "loss": 1.0983, + "step": 15278 + }, + { + "epoch": 0.37, + "learning_rate": 1.459941753637656e-05, + "loss": 1.3342, + "step": 15279 + }, + { + "epoch": 0.37, + "learning_rate": 1.4598725903857438e-05, + "loss": 1.2827, + "step": 15280 + }, + { + "epoch": 0.37, + "learning_rate": 1.4598034243438925e-05, + "loss": 1.0171, + "step": 15281 + }, + { + "epoch": 0.37, + "learning_rate": 1.4597342555125222e-05, + "loss": 1.3856, + "step": 15282 + }, + { + "epoch": 0.37, + "learning_rate": 1.4596650838920518e-05, + "loss": 1.0486, + "step": 15283 + }, + { + "epoch": 0.37, + "learning_rate": 1.4595959094829021e-05, + "loss": 1.0879, + "step": 15284 + }, + { + "epoch": 0.37, + "learning_rate": 1.4595267322854918e-05, + "loss": 1.149, + "step": 15285 + }, + { + "epoch": 0.37, + "learning_rate": 1.459457552300241e-05, + "loss": 1.4459, + "step": 15286 + }, + { + "epoch": 0.37, + "learning_rate": 1.4593883695275693e-05, + "loss": 1.1278, + "step": 15287 + }, + { + "epoch": 0.37, + "learning_rate": 1.4593191839678965e-05, + "loss": 1.2108, + "step": 15288 + }, + { + "epoch": 0.37, + "learning_rate": 1.459249995621642e-05, + "loss": 1.0347, + "step": 15289 + }, + { + "epoch": 0.37, + "learning_rate": 1.4591808044892257e-05, + "loss": 1.0255, + "step": 15290 + }, + { + "epoch": 0.37, + "learning_rate": 1.4591116105710676e-05, + "loss": 1.1016, + "step": 15291 + }, + { + "epoch": 0.37, + "learning_rate": 1.4590424138675874e-05, + "loss": 1.0401, + "step": 15292 + }, + { + "epoch": 0.37, + "learning_rate": 1.458973214379205e-05, + "loss": 1.0309, + "step": 15293 + }, + { + "epoch": 0.37, + "learning_rate": 1.45890401210634e-05, + "loss": 1.1362, + "step": 15294 + }, + { + "epoch": 0.37, + "learning_rate": 1.458834807049412e-05, + "loss": 1.1308, + "step": 15295 + }, + { + "epoch": 0.37, + "learning_rate": 1.4587655992088415e-05, + "loss": 0.8783, + "step": 15296 + }, + { + "epoch": 0.37, + "learning_rate": 1.458696388585048e-05, + "loss": 1.1589, + "step": 15297 + }, + { + "epoch": 0.37, + "learning_rate": 1.4586271751784513e-05, + "loss": 1.0061, + "step": 15298 + }, + { + "epoch": 0.37, + "learning_rate": 1.4585579589894714e-05, + "loss": 0.9066, + "step": 15299 + }, + { + "epoch": 0.37, + "learning_rate": 1.458488740018528e-05, + "loss": 1.1491, + "step": 15300 + }, + { + "epoch": 0.37, + "learning_rate": 1.4584195182660417e-05, + "loss": 1.3375, + "step": 15301 + }, + { + "epoch": 0.37, + "learning_rate": 1.4583502937324319e-05, + "loss": 1.0686, + "step": 15302 + }, + { + "epoch": 0.37, + "learning_rate": 1.4582810664181185e-05, + "loss": 0.9818, + "step": 15303 + }, + { + "epoch": 0.37, + "learning_rate": 1.4582118363235215e-05, + "loss": 1.2997, + "step": 15304 + }, + { + "epoch": 0.37, + "learning_rate": 1.4581426034490616e-05, + "loss": 0.7962, + "step": 15305 + }, + { + "epoch": 0.37, + "learning_rate": 1.4580733677951579e-05, + "loss": 0.8183, + "step": 15306 + }, + { + "epoch": 0.37, + "learning_rate": 1.4580041293622309e-05, + "loss": 1.2587, + "step": 15307 + }, + { + "epoch": 0.37, + "learning_rate": 1.4579348881507008e-05, + "loss": 1.2356, + "step": 15308 + }, + { + "epoch": 0.37, + "learning_rate": 1.4578656441609871e-05, + "loss": 1.2334, + "step": 15309 + }, + { + "epoch": 0.37, + "learning_rate": 1.4577963973935105e-05, + "loss": 1.2449, + "step": 15310 + }, + { + "epoch": 0.37, + "learning_rate": 1.4577271478486905e-05, + "loss": 1.0056, + "step": 15311 + }, + { + "epoch": 0.37, + "learning_rate": 1.4576578955269479e-05, + "loss": 0.9343, + "step": 15312 + }, + { + "epoch": 0.37, + "learning_rate": 1.4575886404287025e-05, + "loss": 1.1247, + "step": 15313 + }, + { + "epoch": 0.37, + "learning_rate": 1.4575193825543743e-05, + "loss": 1.0498, + "step": 15314 + }, + { + "epoch": 0.37, + "learning_rate": 1.4574501219043835e-05, + "loss": 0.879, + "step": 15315 + }, + { + "epoch": 0.37, + "learning_rate": 1.4573808584791507e-05, + "loss": 1.1365, + "step": 15316 + }, + { + "epoch": 0.37, + "learning_rate": 1.4573115922790955e-05, + "loss": 1.0688, + "step": 15317 + }, + { + "epoch": 0.37, + "learning_rate": 1.4572423233046386e-05, + "loss": 1.0272, + "step": 15318 + }, + { + "epoch": 0.37, + "learning_rate": 1.4571730515562001e-05, + "loss": 1.197, + "step": 15319 + }, + { + "epoch": 0.37, + "learning_rate": 1.4571037770342003e-05, + "loss": 0.8478, + "step": 15320 + }, + { + "epoch": 0.37, + "learning_rate": 1.4570344997390593e-05, + "loss": 1.1316, + "step": 15321 + }, + { + "epoch": 0.37, + "learning_rate": 1.4569652196711976e-05, + "loss": 1.176, + "step": 15322 + }, + { + "epoch": 0.37, + "learning_rate": 1.4568959368310351e-05, + "loss": 1.0761, + "step": 15323 + }, + { + "epoch": 0.37, + "learning_rate": 1.4568266512189926e-05, + "loss": 1.3721, + "step": 15324 + }, + { + "epoch": 0.37, + "learning_rate": 1.4567573628354902e-05, + "loss": 1.2272, + "step": 15325 + }, + { + "epoch": 0.37, + "learning_rate": 1.4566880716809487e-05, + "loss": 1.1854, + "step": 15326 + }, + { + "epoch": 0.37, + "learning_rate": 1.4566187777557876e-05, + "loss": 0.9162, + "step": 15327 + }, + { + "epoch": 0.37, + "learning_rate": 1.456549481060428e-05, + "loss": 1.0996, + "step": 15328 + }, + { + "epoch": 0.37, + "learning_rate": 1.45648018159529e-05, + "loss": 0.9135, + "step": 15329 + }, + { + "epoch": 0.37, + "learning_rate": 1.4564108793607944e-05, + "loss": 1.355, + "step": 15330 + }, + { + "epoch": 0.37, + "learning_rate": 1.4563415743573608e-05, + "loss": 1.0994, + "step": 15331 + }, + { + "epoch": 0.37, + "learning_rate": 1.4562722665854105e-05, + "loss": 0.9856, + "step": 15332 + }, + { + "epoch": 0.37, + "learning_rate": 1.4562029560453639e-05, + "loss": 1.3565, + "step": 15333 + }, + { + "epoch": 0.37, + "learning_rate": 1.4561336427376409e-05, + "loss": 1.1969, + "step": 15334 + }, + { + "epoch": 0.37, + "learning_rate": 1.4560643266626624e-05, + "loss": 0.8923, + "step": 15335 + }, + { + "epoch": 0.37, + "learning_rate": 1.4559950078208491e-05, + "loss": 0.9977, + "step": 15336 + }, + { + "epoch": 0.37, + "learning_rate": 1.4559256862126215e-05, + "loss": 1.2754, + "step": 15337 + }, + { + "epoch": 0.37, + "learning_rate": 1.4558563618383996e-05, + "loss": 1.0053, + "step": 15338 + }, + { + "epoch": 0.37, + "learning_rate": 1.4557870346986045e-05, + "loss": 1.2986, + "step": 15339 + }, + { + "epoch": 0.37, + "learning_rate": 1.4557177047936568e-05, + "loss": 1.2617, + "step": 15340 + }, + { + "epoch": 0.37, + "learning_rate": 1.4556483721239771e-05, + "loss": 1.1936, + "step": 15341 + }, + { + "epoch": 0.37, + "learning_rate": 1.4555790366899856e-05, + "loss": 0.9211, + "step": 15342 + }, + { + "epoch": 0.37, + "learning_rate": 1.4555096984921034e-05, + "loss": 0.9902, + "step": 15343 + }, + { + "epoch": 0.37, + "learning_rate": 1.455440357530751e-05, + "loss": 1.1599, + "step": 15344 + }, + { + "epoch": 0.37, + "learning_rate": 1.455371013806349e-05, + "loss": 1.3499, + "step": 15345 + }, + { + "epoch": 0.37, + "learning_rate": 1.455301667319318e-05, + "loss": 1.278, + "step": 15346 + }, + { + "epoch": 0.37, + "learning_rate": 1.4552323180700792e-05, + "loss": 0.7085, + "step": 15347 + }, + { + "epoch": 0.37, + "learning_rate": 1.4551629660590529e-05, + "loss": 0.8953, + "step": 15348 + }, + { + "epoch": 0.37, + "learning_rate": 1.45509361128666e-05, + "loss": 1.1295, + "step": 15349 + }, + { + "epoch": 0.37, + "learning_rate": 1.455024253753321e-05, + "loss": 1.2213, + "step": 15350 + }, + { + "epoch": 0.37, + "learning_rate": 1.4549548934594571e-05, + "loss": 0.7577, + "step": 15351 + }, + { + "epoch": 0.37, + "learning_rate": 1.4548855304054888e-05, + "loss": 1.2465, + "step": 15352 + }, + { + "epoch": 0.37, + "learning_rate": 1.4548161645918371e-05, + "loss": 1.1934, + "step": 15353 + }, + { + "epoch": 0.37, + "learning_rate": 1.4547467960189223e-05, + "loss": 1.155, + "step": 15354 + }, + { + "epoch": 0.37, + "learning_rate": 1.454677424687166e-05, + "loss": 1.1137, + "step": 15355 + }, + { + "epoch": 0.37, + "learning_rate": 1.4546080505969888e-05, + "loss": 1.0745, + "step": 15356 + }, + { + "epoch": 0.37, + "learning_rate": 1.4545386737488114e-05, + "loss": 1.0982, + "step": 15357 + }, + { + "epoch": 0.37, + "learning_rate": 1.4544692941430546e-05, + "loss": 0.9268, + "step": 15358 + }, + { + "epoch": 0.37, + "learning_rate": 1.4543999117801396e-05, + "loss": 1.191, + "step": 15359 + }, + { + "epoch": 0.37, + "learning_rate": 1.4543305266604873e-05, + "loss": 0.8794, + "step": 15360 + }, + { + "epoch": 0.37, + "learning_rate": 1.4542611387845184e-05, + "loss": 1.2708, + "step": 15361 + }, + { + "epoch": 0.37, + "learning_rate": 1.454191748152654e-05, + "loss": 1.4429, + "step": 15362 + }, + { + "epoch": 0.37, + "learning_rate": 1.4541223547653152e-05, + "loss": 1.0765, + "step": 15363 + }, + { + "epoch": 0.37, + "learning_rate": 1.4540529586229228e-05, + "loss": 1.0461, + "step": 15364 + }, + { + "epoch": 0.37, + "learning_rate": 1.4539835597258982e-05, + "loss": 1.1287, + "step": 15365 + }, + { + "epoch": 0.37, + "learning_rate": 1.4539141580746617e-05, + "loss": 1.1469, + "step": 15366 + }, + { + "epoch": 0.37, + "learning_rate": 1.453844753669635e-05, + "loss": 0.9669, + "step": 15367 + }, + { + "epoch": 0.37, + "learning_rate": 1.453775346511239e-05, + "loss": 1.4165, + "step": 15368 + }, + { + "epoch": 0.37, + "learning_rate": 1.4537059365998947e-05, + "loss": 1.162, + "step": 15369 + }, + { + "epoch": 0.37, + "learning_rate": 1.4536365239360231e-05, + "loss": 1.0726, + "step": 15370 + }, + { + "epoch": 0.37, + "learning_rate": 1.4535671085200453e-05, + "loss": 1.1978, + "step": 15371 + }, + { + "epoch": 0.37, + "learning_rate": 1.4534976903523828e-05, + "loss": 1.1631, + "step": 15372 + }, + { + "epoch": 0.37, + "learning_rate": 1.4534282694334562e-05, + "loss": 1.1371, + "step": 15373 + }, + { + "epoch": 0.37, + "learning_rate": 1.4533588457636871e-05, + "loss": 1.05, + "step": 15374 + }, + { + "epoch": 0.37, + "learning_rate": 1.4532894193434964e-05, + "loss": 1.2532, + "step": 15375 + }, + { + "epoch": 0.37, + "learning_rate": 1.4532199901733054e-05, + "loss": 1.0883, + "step": 15376 + }, + { + "epoch": 0.37, + "learning_rate": 1.4531505582535354e-05, + "loss": 1.0301, + "step": 15377 + }, + { + "epoch": 0.37, + "learning_rate": 1.4530811235846077e-05, + "loss": 1.268, + "step": 15378 + }, + { + "epoch": 0.37, + "learning_rate": 1.4530116861669431e-05, + "loss": 1.0939, + "step": 15379 + }, + { + "epoch": 0.37, + "learning_rate": 1.4529422460009635e-05, + "loss": 1.1722, + "step": 15380 + }, + { + "epoch": 0.37, + "learning_rate": 1.4528728030870897e-05, + "loss": 1.2974, + "step": 15381 + }, + { + "epoch": 0.37, + "learning_rate": 1.4528033574257432e-05, + "loss": 0.9179, + "step": 15382 + }, + { + "epoch": 0.37, + "learning_rate": 1.4527339090173451e-05, + "loss": 1.3546, + "step": 15383 + }, + { + "epoch": 0.37, + "learning_rate": 1.452664457862317e-05, + "loss": 1.2134, + "step": 15384 + }, + { + "epoch": 0.37, + "learning_rate": 1.45259500396108e-05, + "loss": 1.4507, + "step": 15385 + }, + { + "epoch": 0.37, + "learning_rate": 1.4525255473140554e-05, + "loss": 1.2533, + "step": 15386 + }, + { + "epoch": 0.37, + "learning_rate": 1.4524560879216649e-05, + "loss": 1.1434, + "step": 15387 + }, + { + "epoch": 0.37, + "learning_rate": 1.4523866257843299e-05, + "loss": 1.2491, + "step": 15388 + }, + { + "epoch": 0.37, + "learning_rate": 1.4523171609024715e-05, + "loss": 1.0643, + "step": 15389 + }, + { + "epoch": 0.37, + "learning_rate": 1.4522476932765112e-05, + "loss": 1.0856, + "step": 15390 + }, + { + "epoch": 0.37, + "learning_rate": 1.4521782229068706e-05, + "loss": 1.115, + "step": 15391 + }, + { + "epoch": 0.37, + "learning_rate": 1.4521087497939712e-05, + "loss": 0.8406, + "step": 15392 + }, + { + "epoch": 0.37, + "learning_rate": 1.4520392739382343e-05, + "loss": 1.0733, + "step": 15393 + }, + { + "epoch": 0.37, + "learning_rate": 1.4519697953400813e-05, + "loss": 1.2305, + "step": 15394 + }, + { + "epoch": 0.37, + "learning_rate": 1.451900313999934e-05, + "loss": 0.9416, + "step": 15395 + }, + { + "epoch": 0.37, + "learning_rate": 1.4518308299182138e-05, + "loss": 1.1131, + "step": 15396 + }, + { + "epoch": 0.37, + "learning_rate": 1.4517613430953421e-05, + "loss": 1.0738, + "step": 15397 + }, + { + "epoch": 0.37, + "learning_rate": 1.4516918535317405e-05, + "loss": 1.297, + "step": 15398 + }, + { + "epoch": 0.37, + "learning_rate": 1.4516223612278308e-05, + "loss": 0.911, + "step": 15399 + }, + { + "epoch": 0.37, + "learning_rate": 1.4515528661840347e-05, + "loss": 0.9859, + "step": 15400 + }, + { + "epoch": 0.37, + "learning_rate": 1.4514833684007734e-05, + "loss": 0.9966, + "step": 15401 + }, + { + "epoch": 0.37, + "learning_rate": 1.4514138678784687e-05, + "loss": 0.7046, + "step": 15402 + }, + { + "epoch": 0.37, + "learning_rate": 1.4513443646175423e-05, + "loss": 0.8148, + "step": 15403 + }, + { + "epoch": 0.37, + "learning_rate": 1.4512748586184156e-05, + "loss": 1.2282, + "step": 15404 + }, + { + "epoch": 0.37, + "learning_rate": 1.4512053498815107e-05, + "loss": 0.7914, + "step": 15405 + }, + { + "epoch": 0.37, + "learning_rate": 1.4511358384072489e-05, + "loss": 1.1711, + "step": 15406 + }, + { + "epoch": 0.37, + "learning_rate": 1.4510663241960522e-05, + "loss": 0.9794, + "step": 15407 + }, + { + "epoch": 0.37, + "learning_rate": 1.450996807248342e-05, + "loss": 0.9334, + "step": 15408 + }, + { + "epoch": 0.37, + "learning_rate": 1.4509272875645405e-05, + "loss": 0.9452, + "step": 15409 + }, + { + "epoch": 0.37, + "learning_rate": 1.450857765145069e-05, + "loss": 1.1033, + "step": 15410 + }, + { + "epoch": 0.37, + "learning_rate": 1.4507882399903497e-05, + "loss": 1.2178, + "step": 15411 + }, + { + "epoch": 0.37, + "learning_rate": 1.4507187121008042e-05, + "loss": 1.142, + "step": 15412 + }, + { + "epoch": 0.37, + "learning_rate": 1.4506491814768542e-05, + "loss": 1.1221, + "step": 15413 + }, + { + "epoch": 0.37, + "learning_rate": 1.4505796481189214e-05, + "loss": 1.2476, + "step": 15414 + }, + { + "epoch": 0.37, + "learning_rate": 1.4505101120274281e-05, + "loss": 1.0196, + "step": 15415 + }, + { + "epoch": 0.37, + "learning_rate": 1.450440573202796e-05, + "loss": 1.2466, + "step": 15416 + }, + { + "epoch": 0.37, + "learning_rate": 1.4503710316454467e-05, + "loss": 1.3406, + "step": 15417 + }, + { + "epoch": 0.37, + "learning_rate": 1.4503014873558023e-05, + "loss": 1.0234, + "step": 15418 + }, + { + "epoch": 0.37, + "learning_rate": 1.4502319403342848e-05, + "loss": 1.3092, + "step": 15419 + }, + { + "epoch": 0.37, + "learning_rate": 1.4501623905813161e-05, + "loss": 1.0221, + "step": 15420 + }, + { + "epoch": 0.37, + "learning_rate": 1.4500928380973179e-05, + "loss": 1.1779, + "step": 15421 + }, + { + "epoch": 0.37, + "learning_rate": 1.450023282882712e-05, + "loss": 1.1081, + "step": 15422 + }, + { + "epoch": 0.37, + "learning_rate": 1.4499537249379212e-05, + "loss": 1.061, + "step": 15423 + }, + { + "epoch": 0.37, + "learning_rate": 1.449884164263367e-05, + "loss": 1.2424, + "step": 15424 + }, + { + "epoch": 0.37, + "learning_rate": 1.449814600859471e-05, + "loss": 1.3199, + "step": 15425 + }, + { + "epoch": 0.37, + "learning_rate": 1.4497450347266559e-05, + "loss": 1.1669, + "step": 15426 + }, + { + "epoch": 0.37, + "learning_rate": 1.4496754658653434e-05, + "loss": 1.2961, + "step": 15427 + }, + { + "epoch": 0.37, + "learning_rate": 1.4496058942759556e-05, + "loss": 1.0966, + "step": 15428 + }, + { + "epoch": 0.37, + "learning_rate": 1.4495363199589145e-05, + "loss": 1.2516, + "step": 15429 + }, + { + "epoch": 0.37, + "learning_rate": 1.4494667429146426e-05, + "loss": 1.0782, + "step": 15430 + }, + { + "epoch": 0.37, + "learning_rate": 1.4493971631435613e-05, + "loss": 1.2134, + "step": 15431 + }, + { + "epoch": 0.37, + "learning_rate": 1.4493275806460934e-05, + "loss": 0.9835, + "step": 15432 + }, + { + "epoch": 0.37, + "learning_rate": 1.4492579954226607e-05, + "loss": 0.868, + "step": 15433 + }, + { + "epoch": 0.37, + "learning_rate": 1.4491884074736853e-05, + "loss": 1.1051, + "step": 15434 + }, + { + "epoch": 0.37, + "learning_rate": 1.4491188167995895e-05, + "loss": 1.1344, + "step": 15435 + }, + { + "epoch": 0.37, + "learning_rate": 1.4490492234007955e-05, + "loss": 1.1528, + "step": 15436 + }, + { + "epoch": 0.37, + "learning_rate": 1.4489796272777254e-05, + "loss": 1.0423, + "step": 15437 + }, + { + "epoch": 0.37, + "learning_rate": 1.4489100284308018e-05, + "loss": 0.9668, + "step": 15438 + }, + { + "epoch": 0.37, + "learning_rate": 1.4488404268604465e-05, + "loss": 1.1055, + "step": 15439 + }, + { + "epoch": 0.37, + "learning_rate": 1.448770822567082e-05, + "loss": 0.9857, + "step": 15440 + }, + { + "epoch": 0.37, + "learning_rate": 1.4487012155511302e-05, + "loss": 0.8885, + "step": 15441 + }, + { + "epoch": 0.37, + "learning_rate": 1.4486316058130139e-05, + "loss": 0.7736, + "step": 15442 + }, + { + "epoch": 0.37, + "learning_rate": 1.4485619933531551e-05, + "loss": 0.956, + "step": 15443 + }, + { + "epoch": 0.37, + "learning_rate": 1.4484923781719762e-05, + "loss": 1.0988, + "step": 15444 + }, + { + "epoch": 0.37, + "learning_rate": 1.4484227602698995e-05, + "loss": 1.1352, + "step": 15445 + }, + { + "epoch": 0.37, + "learning_rate": 1.4483531396473474e-05, + "loss": 1.2791, + "step": 15446 + }, + { + "epoch": 0.37, + "learning_rate": 1.4482835163047422e-05, + "loss": 0.9298, + "step": 15447 + }, + { + "epoch": 0.37, + "learning_rate": 1.4482138902425065e-05, + "loss": 0.9646, + "step": 15448 + }, + { + "epoch": 0.37, + "learning_rate": 1.4481442614610623e-05, + "loss": 1.0708, + "step": 15449 + }, + { + "epoch": 0.37, + "learning_rate": 1.4480746299608326e-05, + "loss": 1.149, + "step": 15450 + }, + { + "epoch": 0.37, + "learning_rate": 1.4480049957422395e-05, + "loss": 1.0759, + "step": 15451 + }, + { + "epoch": 0.37, + "learning_rate": 1.4479353588057052e-05, + "loss": 1.1808, + "step": 15452 + }, + { + "epoch": 0.37, + "learning_rate": 1.4478657191516526e-05, + "loss": 1.0471, + "step": 15453 + }, + { + "epoch": 0.37, + "learning_rate": 1.447796076780504e-05, + "loss": 1.2949, + "step": 15454 + }, + { + "epoch": 0.37, + "learning_rate": 1.4477264316926821e-05, + "loss": 1.0479, + "step": 15455 + }, + { + "epoch": 0.37, + "learning_rate": 1.4476567838886089e-05, + "loss": 1.0816, + "step": 15456 + }, + { + "epoch": 0.37, + "learning_rate": 1.4475871333687073e-05, + "loss": 1.3472, + "step": 15457 + }, + { + "epoch": 0.37, + "learning_rate": 1.4475174801334e-05, + "loss": 1.1083, + "step": 15458 + }, + { + "epoch": 0.37, + "learning_rate": 1.4474478241831093e-05, + "loss": 1.1463, + "step": 15459 + }, + { + "epoch": 0.37, + "learning_rate": 1.4473781655182582e-05, + "loss": 1.0289, + "step": 15460 + }, + { + "epoch": 0.37, + "learning_rate": 1.4473085041392687e-05, + "loss": 1.132, + "step": 15461 + }, + { + "epoch": 0.37, + "learning_rate": 1.4472388400465639e-05, + "loss": 0.8471, + "step": 15462 + }, + { + "epoch": 0.37, + "learning_rate": 1.4471691732405662e-05, + "loss": 1.0695, + "step": 15463 + }, + { + "epoch": 0.37, + "learning_rate": 1.4470995037216983e-05, + "loss": 0.9883, + "step": 15464 + }, + { + "epoch": 0.37, + "learning_rate": 1.4470298314903826e-05, + "loss": 0.9395, + "step": 15465 + }, + { + "epoch": 0.37, + "learning_rate": 1.4469601565470424e-05, + "loss": 1.2076, + "step": 15466 + }, + { + "epoch": 0.37, + "learning_rate": 1.4468904788921e-05, + "loss": 1.1589, + "step": 15467 + }, + { + "epoch": 0.37, + "learning_rate": 1.4468207985259784e-05, + "loss": 1.2687, + "step": 15468 + }, + { + "epoch": 0.37, + "learning_rate": 1.4467511154490997e-05, + "loss": 1.2585, + "step": 15469 + }, + { + "epoch": 0.37, + "learning_rate": 1.4466814296618875e-05, + "loss": 1.0079, + "step": 15470 + }, + { + "epoch": 0.37, + "learning_rate": 1.4466117411647638e-05, + "loss": 1.0478, + "step": 15471 + }, + { + "epoch": 0.37, + "learning_rate": 1.4465420499581516e-05, + "loss": 1.0209, + "step": 15472 + }, + { + "epoch": 0.37, + "learning_rate": 1.446472356042474e-05, + "loss": 0.6769, + "step": 15473 + }, + { + "epoch": 0.37, + "learning_rate": 1.4464026594181537e-05, + "loss": 1.105, + "step": 15474 + }, + { + "epoch": 0.37, + "learning_rate": 1.4463329600856135e-05, + "loss": 1.2132, + "step": 15475 + }, + { + "epoch": 0.37, + "learning_rate": 1.4462632580452763e-05, + "loss": 0.9084, + "step": 15476 + }, + { + "epoch": 0.37, + "learning_rate": 1.4461935532975643e-05, + "loss": 0.9927, + "step": 15477 + }, + { + "epoch": 0.37, + "learning_rate": 1.4461238458429017e-05, + "loss": 1.359, + "step": 15478 + }, + { + "epoch": 0.37, + "learning_rate": 1.4460541356817102e-05, + "loss": 1.3166, + "step": 15479 + }, + { + "epoch": 0.37, + "learning_rate": 1.4459844228144134e-05, + "loss": 1.1638, + "step": 15480 + }, + { + "epoch": 0.37, + "learning_rate": 1.4459147072414336e-05, + "loss": 1.0653, + "step": 15481 + }, + { + "epoch": 0.37, + "learning_rate": 1.4458449889631946e-05, + "loss": 1.3572, + "step": 15482 + }, + { + "epoch": 0.37, + "learning_rate": 1.4457752679801189e-05, + "loss": 1.2211, + "step": 15483 + }, + { + "epoch": 0.37, + "learning_rate": 1.4457055442926296e-05, + "loss": 1.1248, + "step": 15484 + }, + { + "epoch": 0.37, + "learning_rate": 1.4456358179011492e-05, + "loss": 1.1127, + "step": 15485 + }, + { + "epoch": 0.37, + "learning_rate": 1.4455660888061011e-05, + "loss": 1.2365, + "step": 15486 + }, + { + "epoch": 0.37, + "learning_rate": 1.4454963570079088e-05, + "loss": 0.9587, + "step": 15487 + }, + { + "epoch": 0.37, + "learning_rate": 1.4454266225069946e-05, + "loss": 0.8974, + "step": 15488 + }, + { + "epoch": 0.37, + "learning_rate": 1.4453568853037819e-05, + "loss": 0.9176, + "step": 15489 + }, + { + "epoch": 0.37, + "learning_rate": 1.4452871453986936e-05, + "loss": 1.0173, + "step": 15490 + }, + { + "epoch": 0.37, + "learning_rate": 1.4452174027921532e-05, + "loss": 1.0869, + "step": 15491 + }, + { + "epoch": 0.37, + "learning_rate": 1.4451476574845833e-05, + "loss": 1.1617, + "step": 15492 + }, + { + "epoch": 0.37, + "learning_rate": 1.4450779094764074e-05, + "loss": 1.0736, + "step": 15493 + }, + { + "epoch": 0.37, + "learning_rate": 1.4450081587680485e-05, + "loss": 0.8976, + "step": 15494 + }, + { + "epoch": 0.37, + "learning_rate": 1.4449384053599298e-05, + "loss": 1.2212, + "step": 15495 + }, + { + "epoch": 0.37, + "learning_rate": 1.4448686492524744e-05, + "loss": 1.3099, + "step": 15496 + }, + { + "epoch": 0.37, + "learning_rate": 1.4447988904461054e-05, + "loss": 1.0687, + "step": 15497 + }, + { + "epoch": 0.37, + "learning_rate": 1.4447291289412464e-05, + "loss": 1.1424, + "step": 15498 + }, + { + "epoch": 0.37, + "learning_rate": 1.4446593647383203e-05, + "loss": 1.2357, + "step": 15499 + }, + { + "epoch": 0.37, + "learning_rate": 1.4445895978377504e-05, + "loss": 1.4559, + "step": 15500 + }, + { + "epoch": 0.37, + "learning_rate": 1.44451982823996e-05, + "loss": 1.0274, + "step": 15501 + }, + { + "epoch": 0.37, + "learning_rate": 1.4444500559453722e-05, + "loss": 1.1731, + "step": 15502 + }, + { + "epoch": 0.37, + "learning_rate": 1.4443802809544107e-05, + "loss": 1.1268, + "step": 15503 + }, + { + "epoch": 0.37, + "learning_rate": 1.4443105032674985e-05, + "loss": 0.9149, + "step": 15504 + }, + { + "epoch": 0.37, + "learning_rate": 1.4442407228850587e-05, + "loss": 1.1883, + "step": 15505 + }, + { + "epoch": 0.37, + "learning_rate": 1.4441709398075152e-05, + "loss": 0.9264, + "step": 15506 + }, + { + "epoch": 0.37, + "learning_rate": 1.4441011540352911e-05, + "loss": 1.3047, + "step": 15507 + }, + { + "epoch": 0.37, + "learning_rate": 1.4440313655688096e-05, + "loss": 1.4049, + "step": 15508 + }, + { + "epoch": 0.37, + "learning_rate": 1.4439615744084945e-05, + "loss": 1.0741, + "step": 15509 + }, + { + "epoch": 0.37, + "learning_rate": 1.4438917805547687e-05, + "loss": 1.175, + "step": 15510 + }, + { + "epoch": 0.37, + "learning_rate": 1.443821984008056e-05, + "loss": 1.309, + "step": 15511 + }, + { + "epoch": 0.37, + "learning_rate": 1.4437521847687796e-05, + "loss": 1.2678, + "step": 15512 + }, + { + "epoch": 0.37, + "learning_rate": 1.4436823828373635e-05, + "loss": 1.1061, + "step": 15513 + }, + { + "epoch": 0.37, + "learning_rate": 1.4436125782142303e-05, + "loss": 1.0579, + "step": 15514 + }, + { + "epoch": 0.37, + "learning_rate": 1.443542770899804e-05, + "loss": 1.251, + "step": 15515 + }, + { + "epoch": 0.37, + "learning_rate": 1.443472960894508e-05, + "loss": 1.0948, + "step": 15516 + }, + { + "epoch": 0.37, + "learning_rate": 1.443403148198766e-05, + "loss": 1.0754, + "step": 15517 + }, + { + "epoch": 0.37, + "learning_rate": 1.4433333328130015e-05, + "loss": 1.0633, + "step": 15518 + }, + { + "epoch": 0.37, + "learning_rate": 1.4432635147376378e-05, + "loss": 1.2841, + "step": 15519 + }, + { + "epoch": 0.37, + "learning_rate": 1.4431936939730984e-05, + "loss": 1.3042, + "step": 15520 + }, + { + "epoch": 0.37, + "learning_rate": 1.4431238705198074e-05, + "loss": 1.1042, + "step": 15521 + }, + { + "epoch": 0.37, + "learning_rate": 1.443054044378188e-05, + "loss": 1.4331, + "step": 15522 + }, + { + "epoch": 0.37, + "learning_rate": 1.442984215548664e-05, + "loss": 1.1636, + "step": 15523 + }, + { + "epoch": 0.37, + "learning_rate": 1.4429143840316586e-05, + "loss": 1.2006, + "step": 15524 + }, + { + "epoch": 0.37, + "learning_rate": 1.4428445498275963e-05, + "loss": 1.1443, + "step": 15525 + }, + { + "epoch": 0.37, + "learning_rate": 1.4427747129369002e-05, + "loss": 1.228, + "step": 15526 + }, + { + "epoch": 0.37, + "learning_rate": 1.442704873359994e-05, + "loss": 1.0007, + "step": 15527 + }, + { + "epoch": 0.37, + "learning_rate": 1.4426350310973011e-05, + "loss": 1.1359, + "step": 15528 + }, + { + "epoch": 0.37, + "learning_rate": 1.442565186149246e-05, + "loss": 0.9885, + "step": 15529 + }, + { + "epoch": 0.37, + "learning_rate": 1.4424953385162522e-05, + "loss": 1.0216, + "step": 15530 + }, + { + "epoch": 0.37, + "learning_rate": 1.442425488198743e-05, + "loss": 0.9937, + "step": 15531 + }, + { + "epoch": 0.37, + "learning_rate": 1.4423556351971422e-05, + "loss": 0.9903, + "step": 15532 + }, + { + "epoch": 0.37, + "learning_rate": 1.442285779511874e-05, + "loss": 1.291, + "step": 15533 + }, + { + "epoch": 0.37, + "learning_rate": 1.4422159211433622e-05, + "loss": 1.3735, + "step": 15534 + }, + { + "epoch": 0.37, + "learning_rate": 1.4421460600920302e-05, + "loss": 0.852, + "step": 15535 + }, + { + "epoch": 0.37, + "learning_rate": 1.4420761963583018e-05, + "loss": 1.1308, + "step": 15536 + }, + { + "epoch": 0.37, + "learning_rate": 1.4420063299426015e-05, + "loss": 1.1788, + "step": 15537 + }, + { + "epoch": 0.37, + "learning_rate": 1.4419364608453528e-05, + "loss": 1.2041, + "step": 15538 + }, + { + "epoch": 0.37, + "learning_rate": 1.4418665890669793e-05, + "loss": 1.1933, + "step": 15539 + }, + { + "epoch": 0.37, + "learning_rate": 1.4417967146079053e-05, + "loss": 1.193, + "step": 15540 + }, + { + "epoch": 0.37, + "learning_rate": 1.4417268374685542e-05, + "loss": 1.1996, + "step": 15541 + }, + { + "epoch": 0.37, + "learning_rate": 1.4416569576493506e-05, + "loss": 1.2539, + "step": 15542 + }, + { + "epoch": 0.37, + "learning_rate": 1.4415870751507182e-05, + "loss": 0.903, + "step": 15543 + }, + { + "epoch": 0.37, + "learning_rate": 1.4415171899730808e-05, + "loss": 1.0861, + "step": 15544 + }, + { + "epoch": 0.37, + "learning_rate": 1.4414473021168622e-05, + "loss": 0.9996, + "step": 15545 + }, + { + "epoch": 0.37, + "learning_rate": 1.441377411582487e-05, + "loss": 1.2419, + "step": 15546 + }, + { + "epoch": 0.37, + "learning_rate": 1.4413075183703788e-05, + "loss": 1.1592, + "step": 15547 + }, + { + "epoch": 0.37, + "learning_rate": 1.4412376224809614e-05, + "loss": 1.1527, + "step": 15548 + }, + { + "epoch": 0.37, + "learning_rate": 1.4411677239146595e-05, + "loss": 1.1352, + "step": 15549 + }, + { + "epoch": 0.37, + "learning_rate": 1.4410978226718966e-05, + "loss": 1.2262, + "step": 15550 + }, + { + "epoch": 0.37, + "learning_rate": 1.4410279187530971e-05, + "loss": 0.9775, + "step": 15551 + }, + { + "epoch": 0.37, + "learning_rate": 1.4409580121586848e-05, + "loss": 1.3937, + "step": 15552 + }, + { + "epoch": 0.37, + "learning_rate": 1.440888102889084e-05, + "loss": 1.0026, + "step": 15553 + }, + { + "epoch": 0.37, + "learning_rate": 1.440818190944719e-05, + "loss": 0.9484, + "step": 15554 + }, + { + "epoch": 0.37, + "learning_rate": 1.4407482763260135e-05, + "loss": 1.2694, + "step": 15555 + }, + { + "epoch": 0.37, + "learning_rate": 1.4406783590333916e-05, + "loss": 1.3208, + "step": 15556 + }, + { + "epoch": 0.37, + "learning_rate": 1.4406084390672783e-05, + "loss": 1.0102, + "step": 15557 + }, + { + "epoch": 0.37, + "learning_rate": 1.440538516428097e-05, + "loss": 1.0195, + "step": 15558 + }, + { + "epoch": 0.37, + "learning_rate": 1.4404685911162722e-05, + "loss": 0.992, + "step": 15559 + }, + { + "epoch": 0.37, + "learning_rate": 1.4403986631322278e-05, + "loss": 1.0665, + "step": 15560 + }, + { + "epoch": 0.37, + "learning_rate": 1.4403287324763886e-05, + "loss": 1.0218, + "step": 15561 + }, + { + "epoch": 0.37, + "learning_rate": 1.4402587991491785e-05, + "loss": 1.1577, + "step": 15562 + }, + { + "epoch": 0.37, + "learning_rate": 1.4401888631510219e-05, + "loss": 1.1764, + "step": 15563 + }, + { + "epoch": 0.37, + "learning_rate": 1.4401189244823425e-05, + "loss": 1.0602, + "step": 15564 + }, + { + "epoch": 0.37, + "learning_rate": 1.4400489831435658e-05, + "loss": 1.1816, + "step": 15565 + }, + { + "epoch": 0.37, + "learning_rate": 1.439979039135115e-05, + "loss": 1.0593, + "step": 15566 + }, + { + "epoch": 0.37, + "learning_rate": 1.439909092457415e-05, + "loss": 1.0534, + "step": 15567 + }, + { + "epoch": 0.37, + "learning_rate": 1.4398391431108897e-05, + "loss": 0.906, + "step": 15568 + }, + { + "epoch": 0.37, + "learning_rate": 1.4397691910959642e-05, + "loss": 1.1738, + "step": 15569 + }, + { + "epoch": 0.37, + "learning_rate": 1.4396992364130624e-05, + "loss": 0.9375, + "step": 15570 + }, + { + "epoch": 0.37, + "learning_rate": 1.4396292790626087e-05, + "loss": 0.9193, + "step": 15571 + }, + { + "epoch": 0.37, + "learning_rate": 1.4395593190450275e-05, + "loss": 1.2401, + "step": 15572 + }, + { + "epoch": 0.37, + "learning_rate": 1.4394893563607433e-05, + "loss": 1.1762, + "step": 15573 + }, + { + "epoch": 0.37, + "learning_rate": 1.4394193910101805e-05, + "loss": 1.0348, + "step": 15574 + }, + { + "epoch": 0.37, + "learning_rate": 1.439349422993764e-05, + "loss": 1.1933, + "step": 15575 + }, + { + "epoch": 0.37, + "learning_rate": 1.4392794523119174e-05, + "loss": 1.1364, + "step": 15576 + }, + { + "epoch": 0.37, + "learning_rate": 1.4392094789650659e-05, + "loss": 1.0506, + "step": 15577 + }, + { + "epoch": 0.37, + "learning_rate": 1.439139502953634e-05, + "loss": 1.3966, + "step": 15578 + }, + { + "epoch": 0.37, + "learning_rate": 1.4390695242780457e-05, + "loss": 1.0571, + "step": 15579 + }, + { + "epoch": 0.37, + "learning_rate": 1.4389995429387258e-05, + "loss": 1.3481, + "step": 15580 + }, + { + "epoch": 0.37, + "learning_rate": 1.4389295589360992e-05, + "loss": 0.9856, + "step": 15581 + }, + { + "epoch": 0.37, + "learning_rate": 1.43885957227059e-05, + "loss": 1.0882, + "step": 15582 + }, + { + "epoch": 0.37, + "learning_rate": 1.4387895829426232e-05, + "loss": 1.3659, + "step": 15583 + }, + { + "epoch": 0.37, + "learning_rate": 1.438719590952623e-05, + "loss": 1.0157, + "step": 15584 + }, + { + "epoch": 0.37, + "learning_rate": 1.4386495963010145e-05, + "loss": 1.0396, + "step": 15585 + }, + { + "epoch": 0.37, + "learning_rate": 1.4385795989882221e-05, + "loss": 1.2852, + "step": 15586 + }, + { + "epoch": 0.37, + "learning_rate": 1.4385095990146702e-05, + "loss": 1.2006, + "step": 15587 + }, + { + "epoch": 0.37, + "learning_rate": 1.4384395963807835e-05, + "loss": 1.0578, + "step": 15588 + }, + { + "epoch": 0.37, + "learning_rate": 1.4383695910869872e-05, + "loss": 1.2586, + "step": 15589 + }, + { + "epoch": 0.37, + "learning_rate": 1.4382995831337058e-05, + "loss": 1.0864, + "step": 15590 + }, + { + "epoch": 0.37, + "learning_rate": 1.438229572521364e-05, + "loss": 1.0663, + "step": 15591 + }, + { + "epoch": 0.37, + "learning_rate": 1.438159559250386e-05, + "loss": 1.2094, + "step": 15592 + }, + { + "epoch": 0.37, + "learning_rate": 1.4380895433211975e-05, + "loss": 0.732, + "step": 15593 + }, + { + "epoch": 0.38, + "learning_rate": 1.4380195247342226e-05, + "loss": 1.337, + "step": 15594 + }, + { + "epoch": 0.38, + "learning_rate": 1.437949503489886e-05, + "loss": 1.2363, + "step": 15595 + }, + { + "epoch": 0.38, + "learning_rate": 1.437879479588613e-05, + "loss": 1.079, + "step": 15596 + }, + { + "epoch": 0.38, + "learning_rate": 1.4378094530308283e-05, + "loss": 1.0281, + "step": 15597 + }, + { + "epoch": 0.38, + "learning_rate": 1.4377394238169567e-05, + "loss": 0.827, + "step": 15598 + }, + { + "epoch": 0.38, + "learning_rate": 1.4376693919474225e-05, + "loss": 1.1396, + "step": 15599 + }, + { + "epoch": 0.38, + "learning_rate": 1.4375993574226515e-05, + "loss": 0.9245, + "step": 15600 + }, + { + "epoch": 0.38, + "learning_rate": 1.437529320243068e-05, + "loss": 1.0913, + "step": 15601 + }, + { + "epoch": 0.38, + "learning_rate": 1.437459280409097e-05, + "loss": 0.9347, + "step": 15602 + }, + { + "epoch": 0.38, + "learning_rate": 1.4373892379211634e-05, + "loss": 1.3393, + "step": 15603 + }, + { + "epoch": 0.38, + "learning_rate": 1.4373191927796924e-05, + "loss": 1.1645, + "step": 15604 + }, + { + "epoch": 0.38, + "learning_rate": 1.4372491449851084e-05, + "loss": 1.2263, + "step": 15605 + }, + { + "epoch": 0.38, + "learning_rate": 1.4371790945378368e-05, + "loss": 1.0056, + "step": 15606 + }, + { + "epoch": 0.38, + "learning_rate": 1.4371090414383024e-05, + "loss": 1.0836, + "step": 15607 + }, + { + "epoch": 0.38, + "learning_rate": 1.4370389856869304e-05, + "loss": 1.0647, + "step": 15608 + }, + { + "epoch": 0.38, + "learning_rate": 1.4369689272841458e-05, + "loss": 1.1321, + "step": 15609 + }, + { + "epoch": 0.38, + "learning_rate": 1.4368988662303733e-05, + "loss": 1.1889, + "step": 15610 + }, + { + "epoch": 0.38, + "learning_rate": 1.4368288025260379e-05, + "loss": 1.3151, + "step": 15611 + }, + { + "epoch": 0.38, + "learning_rate": 1.4367587361715654e-05, + "loss": 1.0408, + "step": 15612 + }, + { + "epoch": 0.38, + "learning_rate": 1.4366886671673801e-05, + "loss": 1.2437, + "step": 15613 + }, + { + "epoch": 0.38, + "learning_rate": 1.4366185955139077e-05, + "loss": 1.1515, + "step": 15614 + }, + { + "epoch": 0.38, + "learning_rate": 1.4365485212115724e-05, + "loss": 0.9842, + "step": 15615 + }, + { + "epoch": 0.38, + "learning_rate": 1.4364784442608001e-05, + "loss": 0.9126, + "step": 15616 + }, + { + "epoch": 0.38, + "learning_rate": 1.436408364662016e-05, + "loss": 1.022, + "step": 15617 + }, + { + "epoch": 0.38, + "learning_rate": 1.4363382824156449e-05, + "loss": 1.2133, + "step": 15618 + }, + { + "epoch": 0.38, + "learning_rate": 1.436268197522112e-05, + "loss": 1.0097, + "step": 15619 + }, + { + "epoch": 0.38, + "learning_rate": 1.4361981099818425e-05, + "loss": 1.1208, + "step": 15620 + }, + { + "epoch": 0.38, + "learning_rate": 1.4361280197952619e-05, + "loss": 0.8788, + "step": 15621 + }, + { + "epoch": 0.38, + "learning_rate": 1.436057926962795e-05, + "loss": 1.3341, + "step": 15622 + }, + { + "epoch": 0.38, + "learning_rate": 1.4359878314848671e-05, + "loss": 0.9304, + "step": 15623 + }, + { + "epoch": 0.38, + "learning_rate": 1.4359177333619036e-05, + "loss": 1.0832, + "step": 15624 + }, + { + "epoch": 0.38, + "learning_rate": 1.43584763259433e-05, + "loss": 1.2374, + "step": 15625 + }, + { + "epoch": 0.38, + "learning_rate": 1.4357775291825713e-05, + "loss": 1.0581, + "step": 15626 + }, + { + "epoch": 0.38, + "learning_rate": 1.4357074231270529e-05, + "loss": 1.1452, + "step": 15627 + }, + { + "epoch": 0.38, + "learning_rate": 1.4356373144281996e-05, + "loss": 1.1093, + "step": 15628 + }, + { + "epoch": 0.38, + "learning_rate": 1.4355672030864377e-05, + "loss": 1.3395, + "step": 15629 + }, + { + "epoch": 0.38, + "learning_rate": 1.4354970891021918e-05, + "loss": 1.1358, + "step": 15630 + }, + { + "epoch": 0.38, + "learning_rate": 1.4354269724758875e-05, + "loss": 1.0542, + "step": 15631 + }, + { + "epoch": 0.38, + "learning_rate": 1.43535685320795e-05, + "loss": 1.1889, + "step": 15632 + }, + { + "epoch": 0.38, + "learning_rate": 1.4352867312988053e-05, + "loss": 1.0649, + "step": 15633 + }, + { + "epoch": 0.38, + "learning_rate": 1.435216606748878e-05, + "loss": 1.1727, + "step": 15634 + }, + { + "epoch": 0.38, + "learning_rate": 1.4351464795585941e-05, + "loss": 1.2066, + "step": 15635 + }, + { + "epoch": 0.38, + "learning_rate": 1.4350763497283789e-05, + "loss": 1.2109, + "step": 15636 + }, + { + "epoch": 0.38, + "learning_rate": 1.4350062172586576e-05, + "loss": 1.0834, + "step": 15637 + }, + { + "epoch": 0.38, + "learning_rate": 1.4349360821498562e-05, + "loss": 0.9749, + "step": 15638 + }, + { + "epoch": 0.38, + "learning_rate": 1.4348659444023995e-05, + "loss": 0.9732, + "step": 15639 + }, + { + "epoch": 0.38, + "learning_rate": 1.4347958040167138e-05, + "loss": 1.0455, + "step": 15640 + }, + { + "epoch": 0.38, + "learning_rate": 1.434725660993224e-05, + "loss": 1.1047, + "step": 15641 + }, + { + "epoch": 0.38, + "learning_rate": 1.4346555153323561e-05, + "loss": 1.328, + "step": 15642 + }, + { + "epoch": 0.38, + "learning_rate": 1.4345853670345351e-05, + "loss": 1.2433, + "step": 15643 + }, + { + "epoch": 0.38, + "learning_rate": 1.4345152161001871e-05, + "loss": 1.007, + "step": 15644 + }, + { + "epoch": 0.38, + "learning_rate": 1.4344450625297375e-05, + "loss": 1.247, + "step": 15645 + }, + { + "epoch": 0.38, + "learning_rate": 1.4343749063236119e-05, + "loss": 1.2878, + "step": 15646 + }, + { + "epoch": 0.38, + "learning_rate": 1.4343047474822359e-05, + "loss": 1.2323, + "step": 15647 + }, + { + "epoch": 0.38, + "learning_rate": 1.434234586006035e-05, + "loss": 1.155, + "step": 15648 + }, + { + "epoch": 0.38, + "learning_rate": 1.4341644218954351e-05, + "loss": 1.0823, + "step": 15649 + }, + { + "epoch": 0.38, + "learning_rate": 1.4340942551508619e-05, + "loss": 1.1141, + "step": 15650 + }, + { + "epoch": 0.38, + "learning_rate": 1.4340240857727405e-05, + "loss": 0.8866, + "step": 15651 + }, + { + "epoch": 0.38, + "learning_rate": 1.4339539137614975e-05, + "loss": 1.1039, + "step": 15652 + }, + { + "epoch": 0.38, + "learning_rate": 1.4338837391175582e-05, + "loss": 1.1087, + "step": 15653 + }, + { + "epoch": 0.38, + "learning_rate": 1.4338135618413484e-05, + "loss": 0.7103, + "step": 15654 + }, + { + "epoch": 0.38, + "learning_rate": 1.4337433819332933e-05, + "loss": 1.1336, + "step": 15655 + }, + { + "epoch": 0.38, + "learning_rate": 1.4336731993938196e-05, + "loss": 1.111, + "step": 15656 + }, + { + "epoch": 0.38, + "learning_rate": 1.4336030142233524e-05, + "loss": 1.1503, + "step": 15657 + }, + { + "epoch": 0.38, + "learning_rate": 1.4335328264223176e-05, + "loss": 1.1857, + "step": 15658 + }, + { + "epoch": 0.38, + "learning_rate": 1.4334626359911412e-05, + "loss": 1.0613, + "step": 15659 + }, + { + "epoch": 0.38, + "learning_rate": 1.4333924429302489e-05, + "loss": 1.3239, + "step": 15660 + }, + { + "epoch": 0.38, + "learning_rate": 1.4333222472400668e-05, + "loss": 1.0274, + "step": 15661 + }, + { + "epoch": 0.38, + "learning_rate": 1.4332520489210203e-05, + "loss": 1.1185, + "step": 15662 + }, + { + "epoch": 0.38, + "learning_rate": 1.4331818479735356e-05, + "loss": 1.1165, + "step": 15663 + }, + { + "epoch": 0.38, + "learning_rate": 1.4331116443980384e-05, + "loss": 1.2598, + "step": 15664 + }, + { + "epoch": 0.38, + "learning_rate": 1.4330414381949549e-05, + "loss": 0.9721, + "step": 15665 + }, + { + "epoch": 0.38, + "learning_rate": 1.432971229364711e-05, + "loss": 1.089, + "step": 15666 + }, + { + "epoch": 0.38, + "learning_rate": 1.432901017907732e-05, + "loss": 1.3441, + "step": 15667 + }, + { + "epoch": 0.38, + "learning_rate": 1.4328308038244447e-05, + "loss": 1.1305, + "step": 15668 + }, + { + "epoch": 0.38, + "learning_rate": 1.4327605871152748e-05, + "loss": 0.9655, + "step": 15669 + }, + { + "epoch": 0.38, + "learning_rate": 1.4326903677806481e-05, + "loss": 1.1188, + "step": 15670 + }, + { + "epoch": 0.38, + "learning_rate": 1.4326201458209906e-05, + "loss": 1.2412, + "step": 15671 + }, + { + "epoch": 0.38, + "learning_rate": 1.4325499212367287e-05, + "loss": 0.9918, + "step": 15672 + }, + { + "epoch": 0.38, + "learning_rate": 1.4324796940282878e-05, + "loss": 0.8279, + "step": 15673 + }, + { + "epoch": 0.38, + "learning_rate": 1.4324094641960949e-05, + "loss": 1.1556, + "step": 15674 + }, + { + "epoch": 0.38, + "learning_rate": 1.4323392317405746e-05, + "loss": 1.1393, + "step": 15675 + }, + { + "epoch": 0.38, + "learning_rate": 1.4322689966621545e-05, + "loss": 1.0078, + "step": 15676 + }, + { + "epoch": 0.38, + "learning_rate": 1.43219875896126e-05, + "loss": 1.2598, + "step": 15677 + }, + { + "epoch": 0.38, + "learning_rate": 1.4321285186383173e-05, + "loss": 1.1045, + "step": 15678 + }, + { + "epoch": 0.38, + "learning_rate": 1.4320582756937524e-05, + "loss": 1.165, + "step": 15679 + }, + { + "epoch": 0.38, + "learning_rate": 1.4319880301279918e-05, + "loss": 1.3392, + "step": 15680 + }, + { + "epoch": 0.38, + "learning_rate": 1.4319177819414613e-05, + "loss": 1.0927, + "step": 15681 + }, + { + "epoch": 0.38, + "learning_rate": 1.4318475311345869e-05, + "loss": 0.8365, + "step": 15682 + }, + { + "epoch": 0.38, + "learning_rate": 1.4317772777077953e-05, + "loss": 1.159, + "step": 15683 + }, + { + "epoch": 0.38, + "learning_rate": 1.4317070216615127e-05, + "loss": 1.0187, + "step": 15684 + }, + { + "epoch": 0.38, + "learning_rate": 1.4316367629961651e-05, + "loss": 1.1573, + "step": 15685 + }, + { + "epoch": 0.38, + "learning_rate": 1.4315665017121787e-05, + "loss": 0.9239, + "step": 15686 + }, + { + "epoch": 0.38, + "learning_rate": 1.4314962378099799e-05, + "loss": 1.251, + "step": 15687 + }, + { + "epoch": 0.38, + "learning_rate": 1.431425971289995e-05, + "loss": 0.9532, + "step": 15688 + }, + { + "epoch": 0.38, + "learning_rate": 1.43135570215265e-05, + "loss": 0.8496, + "step": 15689 + }, + { + "epoch": 0.38, + "learning_rate": 1.4312854303983715e-05, + "loss": 1.4379, + "step": 15690 + }, + { + "epoch": 0.38, + "learning_rate": 1.4312151560275858e-05, + "loss": 1.0234, + "step": 15691 + }, + { + "epoch": 0.38, + "learning_rate": 1.4311448790407192e-05, + "loss": 1.2585, + "step": 15692 + }, + { + "epoch": 0.38, + "learning_rate": 1.431074599438198e-05, + "loss": 0.897, + "step": 15693 + }, + { + "epoch": 0.38, + "learning_rate": 1.4310043172204484e-05, + "loss": 1.026, + "step": 15694 + }, + { + "epoch": 0.38, + "learning_rate": 1.4309340323878971e-05, + "loss": 1.0161, + "step": 15695 + }, + { + "epoch": 0.38, + "learning_rate": 1.4308637449409705e-05, + "loss": 1.2752, + "step": 15696 + }, + { + "epoch": 0.38, + "learning_rate": 1.430793454880095e-05, + "loss": 1.0709, + "step": 15697 + }, + { + "epoch": 0.38, + "learning_rate": 1.4307231622056965e-05, + "loss": 1.0546, + "step": 15698 + }, + { + "epoch": 0.38, + "learning_rate": 1.4306528669182023e-05, + "loss": 1.1021, + "step": 15699 + }, + { + "epoch": 0.38, + "learning_rate": 1.4305825690180383e-05, + "loss": 1.4135, + "step": 15700 + }, + { + "epoch": 0.38, + "learning_rate": 1.4305122685056314e-05, + "loss": 0.8267, + "step": 15701 + }, + { + "epoch": 0.38, + "learning_rate": 1.4304419653814075e-05, + "loss": 1.1907, + "step": 15702 + }, + { + "epoch": 0.38, + "learning_rate": 1.4303716596457934e-05, + "loss": 0.8802, + "step": 15703 + }, + { + "epoch": 0.38, + "learning_rate": 1.430301351299216e-05, + "loss": 1.4237, + "step": 15704 + }, + { + "epoch": 0.38, + "learning_rate": 1.4302310403421011e-05, + "loss": 1.0206, + "step": 15705 + }, + { + "epoch": 0.38, + "learning_rate": 1.4301607267748759e-05, + "loss": 0.994, + "step": 15706 + }, + { + "epoch": 0.38, + "learning_rate": 1.4300904105979664e-05, + "loss": 1.0947, + "step": 15707 + }, + { + "epoch": 0.38, + "learning_rate": 1.4300200918118001e-05, + "loss": 1.0531, + "step": 15708 + }, + { + "epoch": 0.38, + "learning_rate": 1.4299497704168028e-05, + "loss": 0.816, + "step": 15709 + }, + { + "epoch": 0.38, + "learning_rate": 1.4298794464134015e-05, + "loss": 1.0572, + "step": 15710 + }, + { + "epoch": 0.38, + "learning_rate": 1.4298091198020225e-05, + "loss": 1.2713, + "step": 15711 + }, + { + "epoch": 0.38, + "learning_rate": 1.4297387905830928e-05, + "loss": 1.166, + "step": 15712 + }, + { + "epoch": 0.38, + "learning_rate": 1.429668458757039e-05, + "loss": 1.0796, + "step": 15713 + }, + { + "epoch": 0.38, + "learning_rate": 1.4295981243242876e-05, + "loss": 1.0562, + "step": 15714 + }, + { + "epoch": 0.38, + "learning_rate": 1.4295277872852653e-05, + "loss": 1.097, + "step": 15715 + }, + { + "epoch": 0.38, + "learning_rate": 1.4294574476403994e-05, + "loss": 1.1153, + "step": 15716 + }, + { + "epoch": 0.38, + "learning_rate": 1.4293871053901158e-05, + "loss": 1.0174, + "step": 15717 + }, + { + "epoch": 0.38, + "learning_rate": 1.4293167605348417e-05, + "loss": 1.0977, + "step": 15718 + }, + { + "epoch": 0.38, + "learning_rate": 1.4292464130750036e-05, + "loss": 1.0954, + "step": 15719 + }, + { + "epoch": 0.38, + "learning_rate": 1.429176063011029e-05, + "loss": 1.3271, + "step": 15720 + }, + { + "epoch": 0.38, + "learning_rate": 1.4291057103433436e-05, + "loss": 1.2965, + "step": 15721 + }, + { + "epoch": 0.38, + "learning_rate": 1.429035355072375e-05, + "loss": 0.9927, + "step": 15722 + }, + { + "epoch": 0.38, + "learning_rate": 1.42896499719855e-05, + "loss": 0.7336, + "step": 15723 + }, + { + "epoch": 0.38, + "learning_rate": 1.428894636722295e-05, + "loss": 1.0722, + "step": 15724 + }, + { + "epoch": 0.38, + "learning_rate": 1.4288242736440375e-05, + "loss": 0.9636, + "step": 15725 + }, + { + "epoch": 0.38, + "learning_rate": 1.4287539079642037e-05, + "loss": 1.0528, + "step": 15726 + }, + { + "epoch": 0.38, + "learning_rate": 1.4286835396832208e-05, + "loss": 1.153, + "step": 15727 + }, + { + "epoch": 0.38, + "learning_rate": 1.428613168801516e-05, + "loss": 1.0413, + "step": 15728 + }, + { + "epoch": 0.38, + "learning_rate": 1.4285427953195155e-05, + "loss": 0.9548, + "step": 15729 + }, + { + "epoch": 0.38, + "learning_rate": 1.4284724192376467e-05, + "loss": 1.1406, + "step": 15730 + }, + { + "epoch": 0.38, + "learning_rate": 1.4284020405563365e-05, + "loss": 1.3226, + "step": 15731 + }, + { + "epoch": 0.38, + "learning_rate": 1.428331659276012e-05, + "loss": 1.0984, + "step": 15732 + }, + { + "epoch": 0.38, + "learning_rate": 1.4282612753971003e-05, + "loss": 1.0699, + "step": 15733 + }, + { + "epoch": 0.38, + "learning_rate": 1.4281908889200278e-05, + "loss": 1.0583, + "step": 15734 + }, + { + "epoch": 0.38, + "learning_rate": 1.4281204998452224e-05, + "loss": 0.8972, + "step": 15735 + }, + { + "epoch": 0.38, + "learning_rate": 1.4280501081731101e-05, + "loss": 0.7773, + "step": 15736 + }, + { + "epoch": 0.38, + "learning_rate": 1.4279797139041187e-05, + "loss": 0.7975, + "step": 15737 + }, + { + "epoch": 0.38, + "learning_rate": 1.4279093170386748e-05, + "loss": 0.9963, + "step": 15738 + }, + { + "epoch": 0.38, + "learning_rate": 1.4278389175772061e-05, + "loss": 0.8799, + "step": 15739 + }, + { + "epoch": 0.38, + "learning_rate": 1.4277685155201392e-05, + "loss": 1.1795, + "step": 15740 + }, + { + "epoch": 0.38, + "learning_rate": 1.4276981108679013e-05, + "loss": 1.0568, + "step": 15741 + }, + { + "epoch": 0.38, + "learning_rate": 1.4276277036209194e-05, + "loss": 1.0012, + "step": 15742 + }, + { + "epoch": 0.38, + "learning_rate": 1.4275572937796212e-05, + "loss": 0.995, + "step": 15743 + }, + { + "epoch": 0.38, + "learning_rate": 1.4274868813444332e-05, + "loss": 1.0807, + "step": 15744 + }, + { + "epoch": 0.38, + "learning_rate": 1.427416466315783e-05, + "loss": 1.1618, + "step": 15745 + }, + { + "epoch": 0.38, + "learning_rate": 1.4273460486940973e-05, + "loss": 0.9875, + "step": 15746 + }, + { + "epoch": 0.38, + "learning_rate": 1.4272756284798037e-05, + "loss": 0.9707, + "step": 15747 + }, + { + "epoch": 0.38, + "learning_rate": 1.4272052056733297e-05, + "loss": 1.2672, + "step": 15748 + }, + { + "epoch": 0.38, + "learning_rate": 1.427134780275102e-05, + "loss": 1.0801, + "step": 15749 + }, + { + "epoch": 0.38, + "learning_rate": 1.427064352285548e-05, + "loss": 1.0816, + "step": 15750 + }, + { + "epoch": 0.38, + "learning_rate": 1.426993921705095e-05, + "loss": 1.1073, + "step": 15751 + }, + { + "epoch": 0.38, + "learning_rate": 1.4269234885341705e-05, + "loss": 0.9984, + "step": 15752 + }, + { + "epoch": 0.38, + "learning_rate": 1.4268530527732017e-05, + "loss": 1.2093, + "step": 15753 + }, + { + "epoch": 0.38, + "learning_rate": 1.4267826144226154e-05, + "loss": 0.9782, + "step": 15754 + }, + { + "epoch": 0.38, + "learning_rate": 1.4267121734828395e-05, + "loss": 1.0182, + "step": 15755 + }, + { + "epoch": 0.38, + "learning_rate": 1.4266417299543016e-05, + "loss": 1.0244, + "step": 15756 + }, + { + "epoch": 0.38, + "learning_rate": 1.4265712838374284e-05, + "loss": 1.1673, + "step": 15757 + }, + { + "epoch": 0.38, + "learning_rate": 1.4265008351326475e-05, + "loss": 0.8371, + "step": 15758 + }, + { + "epoch": 0.38, + "learning_rate": 1.4264303838403863e-05, + "loss": 1.0748, + "step": 15759 + }, + { + "epoch": 0.38, + "learning_rate": 1.4263599299610725e-05, + "loss": 1.1018, + "step": 15760 + }, + { + "epoch": 0.38, + "learning_rate": 1.426289473495133e-05, + "loss": 1.12, + "step": 15761 + }, + { + "epoch": 0.38, + "learning_rate": 1.4262190144429958e-05, + "loss": 1.3, + "step": 15762 + }, + { + "epoch": 0.38, + "learning_rate": 1.426148552805088e-05, + "loss": 0.9002, + "step": 15763 + }, + { + "epoch": 0.38, + "learning_rate": 1.426078088581837e-05, + "loss": 0.947, + "step": 15764 + }, + { + "epoch": 0.38, + "learning_rate": 1.4260076217736707e-05, + "loss": 1.2081, + "step": 15765 + }, + { + "epoch": 0.38, + "learning_rate": 1.4259371523810162e-05, + "loss": 1.182, + "step": 15766 + }, + { + "epoch": 0.38, + "learning_rate": 1.4258666804043013e-05, + "loss": 1.0062, + "step": 15767 + }, + { + "epoch": 0.38, + "learning_rate": 1.4257962058439536e-05, + "loss": 1.1792, + "step": 15768 + }, + { + "epoch": 0.38, + "learning_rate": 1.4257257287004001e-05, + "loss": 0.987, + "step": 15769 + }, + { + "epoch": 0.38, + "learning_rate": 1.425655248974069e-05, + "loss": 0.9721, + "step": 15770 + }, + { + "epoch": 0.38, + "learning_rate": 1.4255847666653875e-05, + "loss": 1.3185, + "step": 15771 + }, + { + "epoch": 0.38, + "learning_rate": 1.4255142817747833e-05, + "loss": 1.0237, + "step": 15772 + }, + { + "epoch": 0.38, + "learning_rate": 1.425443794302684e-05, + "loss": 1.0454, + "step": 15773 + }, + { + "epoch": 0.38, + "learning_rate": 1.4253733042495173e-05, + "loss": 1.1184, + "step": 15774 + }, + { + "epoch": 0.38, + "learning_rate": 1.425302811615711e-05, + "loss": 1.1544, + "step": 15775 + }, + { + "epoch": 0.38, + "learning_rate": 1.4252323164016924e-05, + "loss": 1.0379, + "step": 15776 + }, + { + "epoch": 0.38, + "learning_rate": 1.4251618186078893e-05, + "loss": 1.2368, + "step": 15777 + }, + { + "epoch": 0.38, + "learning_rate": 1.4250913182347294e-05, + "loss": 0.9301, + "step": 15778 + }, + { + "epoch": 0.38, + "learning_rate": 1.4250208152826409e-05, + "loss": 0.9253, + "step": 15779 + }, + { + "epoch": 0.38, + "learning_rate": 1.4249503097520507e-05, + "loss": 1.263, + "step": 15780 + }, + { + "epoch": 0.38, + "learning_rate": 1.4248798016433868e-05, + "loss": 1.0926, + "step": 15781 + }, + { + "epoch": 0.38, + "learning_rate": 1.4248092909570774e-05, + "loss": 1.0724, + "step": 15782 + }, + { + "epoch": 0.38, + "learning_rate": 1.42473877769355e-05, + "loss": 1.2129, + "step": 15783 + }, + { + "epoch": 0.38, + "learning_rate": 1.4246682618532321e-05, + "loss": 1.0024, + "step": 15784 + }, + { + "epoch": 0.38, + "learning_rate": 1.4245977434365518e-05, + "loss": 1.2457, + "step": 15785 + }, + { + "epoch": 0.38, + "learning_rate": 1.4245272224439371e-05, + "loss": 1.536, + "step": 15786 + }, + { + "epoch": 0.38, + "learning_rate": 1.4244566988758152e-05, + "loss": 0.8037, + "step": 15787 + }, + { + "epoch": 0.38, + "learning_rate": 1.4243861727326147e-05, + "loss": 1.0325, + "step": 15788 + }, + { + "epoch": 0.38, + "learning_rate": 1.424315644014763e-05, + "loss": 0.798, + "step": 15789 + }, + { + "epoch": 0.38, + "learning_rate": 1.4242451127226877e-05, + "loss": 1.0397, + "step": 15790 + }, + { + "epoch": 0.38, + "learning_rate": 1.4241745788568176e-05, + "loss": 1.1887, + "step": 15791 + }, + { + "epoch": 0.38, + "learning_rate": 1.42410404241758e-05, + "loss": 1.1301, + "step": 15792 + }, + { + "epoch": 0.38, + "learning_rate": 1.4240335034054029e-05, + "loss": 1.3304, + "step": 15793 + }, + { + "epoch": 0.38, + "learning_rate": 1.423962961820714e-05, + "loss": 1.2515, + "step": 15794 + }, + { + "epoch": 0.38, + "learning_rate": 1.423892417663942e-05, + "loss": 1.3046, + "step": 15795 + }, + { + "epoch": 0.38, + "learning_rate": 1.4238218709355141e-05, + "loss": 1.0913, + "step": 15796 + }, + { + "epoch": 0.38, + "learning_rate": 1.4237513216358587e-05, + "loss": 0.9391, + "step": 15797 + }, + { + "epoch": 0.38, + "learning_rate": 1.4236807697654037e-05, + "loss": 1.1681, + "step": 15798 + }, + { + "epoch": 0.38, + "learning_rate": 1.423610215324577e-05, + "loss": 0.7642, + "step": 15799 + }, + { + "epoch": 0.38, + "learning_rate": 1.423539658313807e-05, + "loss": 1.1179, + "step": 15800 + }, + { + "epoch": 0.38, + "learning_rate": 1.4234690987335214e-05, + "loss": 0.9677, + "step": 15801 + }, + { + "epoch": 0.38, + "learning_rate": 1.4233985365841482e-05, + "loss": 1.1793, + "step": 15802 + }, + { + "epoch": 0.38, + "learning_rate": 1.423327971866116e-05, + "loss": 1.0752, + "step": 15803 + }, + { + "epoch": 0.38, + "learning_rate": 1.4232574045798525e-05, + "loss": 1.0027, + "step": 15804 + }, + { + "epoch": 0.38, + "learning_rate": 1.4231868347257856e-05, + "loss": 1.0998, + "step": 15805 + }, + { + "epoch": 0.38, + "learning_rate": 1.4231162623043439e-05, + "loss": 1.3359, + "step": 15806 + }, + { + "epoch": 0.38, + "learning_rate": 1.4230456873159555e-05, + "loss": 1.2665, + "step": 15807 + }, + { + "epoch": 0.38, + "learning_rate": 1.4229751097610483e-05, + "loss": 0.9716, + "step": 15808 + }, + { + "epoch": 0.38, + "learning_rate": 1.4229045296400503e-05, + "loss": 1.1142, + "step": 15809 + }, + { + "epoch": 0.38, + "learning_rate": 1.4228339469533902e-05, + "loss": 0.9742, + "step": 15810 + }, + { + "epoch": 0.38, + "learning_rate": 1.4227633617014961e-05, + "loss": 1.3281, + "step": 15811 + }, + { + "epoch": 0.38, + "learning_rate": 1.4226927738847964e-05, + "loss": 1.2644, + "step": 15812 + }, + { + "epoch": 0.38, + "learning_rate": 1.4226221835037186e-05, + "loss": 1.1476, + "step": 15813 + }, + { + "epoch": 0.38, + "learning_rate": 1.4225515905586918e-05, + "loss": 1.1287, + "step": 15814 + }, + { + "epoch": 0.38, + "learning_rate": 1.4224809950501435e-05, + "loss": 1.067, + "step": 15815 + }, + { + "epoch": 0.38, + "learning_rate": 1.4224103969785026e-05, + "loss": 0.8338, + "step": 15816 + }, + { + "epoch": 0.38, + "learning_rate": 1.422339796344197e-05, + "loss": 1.258, + "step": 15817 + }, + { + "epoch": 0.38, + "learning_rate": 1.4222691931476552e-05, + "loss": 1.0784, + "step": 15818 + }, + { + "epoch": 0.38, + "learning_rate": 1.4221985873893061e-05, + "loss": 1.1018, + "step": 15819 + }, + { + "epoch": 0.38, + "learning_rate": 1.4221279790695768e-05, + "loss": 1.2696, + "step": 15820 + }, + { + "epoch": 0.38, + "learning_rate": 1.4220573681888965e-05, + "loss": 0.9059, + "step": 15821 + }, + { + "epoch": 0.38, + "learning_rate": 1.4219867547476936e-05, + "loss": 0.9491, + "step": 15822 + }, + { + "epoch": 0.38, + "learning_rate": 1.4219161387463962e-05, + "loss": 1.284, + "step": 15823 + }, + { + "epoch": 0.38, + "learning_rate": 1.421845520185433e-05, + "loss": 1.1369, + "step": 15824 + }, + { + "epoch": 0.38, + "learning_rate": 1.4217748990652317e-05, + "loss": 1.1432, + "step": 15825 + }, + { + "epoch": 0.38, + "learning_rate": 1.421704275386222e-05, + "loss": 1.0573, + "step": 15826 + }, + { + "epoch": 0.38, + "learning_rate": 1.4216336491488315e-05, + "loss": 1.1518, + "step": 15827 + }, + { + "epoch": 0.38, + "learning_rate": 1.4215630203534886e-05, + "loss": 1.4594, + "step": 15828 + }, + { + "epoch": 0.38, + "learning_rate": 1.4214923890006218e-05, + "loss": 0.8669, + "step": 15829 + }, + { + "epoch": 0.38, + "learning_rate": 1.4214217550906602e-05, + "loss": 1.297, + "step": 15830 + }, + { + "epoch": 0.38, + "learning_rate": 1.4213511186240318e-05, + "loss": 1.0681, + "step": 15831 + }, + { + "epoch": 0.38, + "learning_rate": 1.4212804796011655e-05, + "loss": 1.1968, + "step": 15832 + }, + { + "epoch": 0.38, + "learning_rate": 1.4212098380224893e-05, + "loss": 1.2644, + "step": 15833 + }, + { + "epoch": 0.38, + "learning_rate": 1.4211391938884322e-05, + "loss": 1.0981, + "step": 15834 + }, + { + "epoch": 0.38, + "learning_rate": 1.4210685471994227e-05, + "loss": 1.2463, + "step": 15835 + }, + { + "epoch": 0.38, + "learning_rate": 1.4209978979558893e-05, + "loss": 1.2041, + "step": 15836 + }, + { + "epoch": 0.38, + "learning_rate": 1.4209272461582607e-05, + "loss": 1.1194, + "step": 15837 + }, + { + "epoch": 0.38, + "learning_rate": 1.4208565918069654e-05, + "loss": 1.1967, + "step": 15838 + }, + { + "epoch": 0.38, + "learning_rate": 1.4207859349024325e-05, + "loss": 1.2006, + "step": 15839 + }, + { + "epoch": 0.38, + "learning_rate": 1.4207152754450902e-05, + "loss": 1.2391, + "step": 15840 + }, + { + "epoch": 0.38, + "learning_rate": 1.4206446134353668e-05, + "loss": 1.001, + "step": 15841 + }, + { + "epoch": 0.38, + "learning_rate": 1.420573948873692e-05, + "loss": 0.843, + "step": 15842 + }, + { + "epoch": 0.38, + "learning_rate": 1.4205032817604939e-05, + "loss": 1.2067, + "step": 15843 + }, + { + "epoch": 0.38, + "learning_rate": 1.4204326120962014e-05, + "loss": 1.2262, + "step": 15844 + }, + { + "epoch": 0.38, + "learning_rate": 1.4203619398812427e-05, + "loss": 1.0317, + "step": 15845 + }, + { + "epoch": 0.38, + "learning_rate": 1.4202912651160471e-05, + "loss": 1.1313, + "step": 15846 + }, + { + "epoch": 0.38, + "learning_rate": 1.4202205878010436e-05, + "loss": 1.1199, + "step": 15847 + }, + { + "epoch": 0.38, + "learning_rate": 1.4201499079366604e-05, + "loss": 1.4505, + "step": 15848 + }, + { + "epoch": 0.38, + "learning_rate": 1.4200792255233264e-05, + "loss": 0.9883, + "step": 15849 + }, + { + "epoch": 0.38, + "learning_rate": 1.420008540561471e-05, + "loss": 1.1958, + "step": 15850 + }, + { + "epoch": 0.38, + "learning_rate": 1.4199378530515222e-05, + "loss": 1.0577, + "step": 15851 + }, + { + "epoch": 0.38, + "learning_rate": 1.4198671629939091e-05, + "loss": 1.1336, + "step": 15852 + }, + { + "epoch": 0.38, + "learning_rate": 1.419796470389061e-05, + "loss": 1.1406, + "step": 15853 + }, + { + "epoch": 0.38, + "learning_rate": 1.4197257752374062e-05, + "loss": 1.0451, + "step": 15854 + }, + { + "epoch": 0.38, + "learning_rate": 1.4196550775393742e-05, + "loss": 0.8859, + "step": 15855 + }, + { + "epoch": 0.38, + "learning_rate": 1.4195843772953932e-05, + "loss": 1.1072, + "step": 15856 + }, + { + "epoch": 0.38, + "learning_rate": 1.4195136745058926e-05, + "loss": 1.0946, + "step": 15857 + }, + { + "epoch": 0.38, + "learning_rate": 1.4194429691713014e-05, + "loss": 1.1801, + "step": 15858 + }, + { + "epoch": 0.38, + "learning_rate": 1.4193722612920482e-05, + "loss": 1.1136, + "step": 15859 + }, + { + "epoch": 0.38, + "learning_rate": 1.4193015508685618e-05, + "loss": 1.2612, + "step": 15860 + }, + { + "epoch": 0.38, + "learning_rate": 1.419230837901272e-05, + "loss": 0.762, + "step": 15861 + }, + { + "epoch": 0.38, + "learning_rate": 1.4191601223906073e-05, + "loss": 1.1655, + "step": 15862 + }, + { + "epoch": 0.38, + "learning_rate": 1.4190894043369964e-05, + "loss": 1.2119, + "step": 15863 + }, + { + "epoch": 0.38, + "learning_rate": 1.4190186837408688e-05, + "loss": 1.1285, + "step": 15864 + }, + { + "epoch": 0.38, + "learning_rate": 1.4189479606026537e-05, + "loss": 1.1179, + "step": 15865 + }, + { + "epoch": 0.38, + "learning_rate": 1.4188772349227795e-05, + "loss": 1.2495, + "step": 15866 + }, + { + "epoch": 0.38, + "learning_rate": 1.4188065067016759e-05, + "loss": 1.0222, + "step": 15867 + }, + { + "epoch": 0.38, + "learning_rate": 1.4187357759397716e-05, + "loss": 1.1932, + "step": 15868 + }, + { + "epoch": 0.38, + "learning_rate": 1.4186650426374957e-05, + "loss": 0.9605, + "step": 15869 + }, + { + "epoch": 0.38, + "learning_rate": 1.4185943067952776e-05, + "loss": 1.3421, + "step": 15870 + }, + { + "epoch": 0.38, + "learning_rate": 1.4185235684135466e-05, + "loss": 1.01, + "step": 15871 + }, + { + "epoch": 0.38, + "learning_rate": 1.4184528274927309e-05, + "loss": 1.2883, + "step": 15872 + }, + { + "epoch": 0.38, + "learning_rate": 1.4183820840332608e-05, + "loss": 1.0921, + "step": 15873 + }, + { + "epoch": 0.38, + "learning_rate": 1.4183113380355646e-05, + "loss": 0.9474, + "step": 15874 + }, + { + "epoch": 0.38, + "learning_rate": 1.4182405895000723e-05, + "loss": 1.2143, + "step": 15875 + }, + { + "epoch": 0.38, + "learning_rate": 1.4181698384272127e-05, + "loss": 0.9704, + "step": 15876 + }, + { + "epoch": 0.38, + "learning_rate": 1.4180990848174148e-05, + "loss": 1.1637, + "step": 15877 + }, + { + "epoch": 0.38, + "learning_rate": 1.4180283286711083e-05, + "loss": 1.109, + "step": 15878 + }, + { + "epoch": 0.38, + "learning_rate": 1.4179575699887223e-05, + "loss": 1.161, + "step": 15879 + }, + { + "epoch": 0.38, + "learning_rate": 1.4178868087706858e-05, + "loss": 1.0712, + "step": 15880 + }, + { + "epoch": 0.38, + "learning_rate": 1.4178160450174282e-05, + "loss": 1.1742, + "step": 15881 + }, + { + "epoch": 0.38, + "learning_rate": 1.4177452787293795e-05, + "loss": 1.1644, + "step": 15882 + }, + { + "epoch": 0.38, + "learning_rate": 1.4176745099069682e-05, + "loss": 1.1264, + "step": 15883 + }, + { + "epoch": 0.38, + "learning_rate": 1.4176037385506238e-05, + "loss": 1.2091, + "step": 15884 + }, + { + "epoch": 0.38, + "learning_rate": 1.4175329646607756e-05, + "loss": 1.0478, + "step": 15885 + }, + { + "epoch": 0.38, + "learning_rate": 1.4174621882378535e-05, + "loss": 0.9672, + "step": 15886 + }, + { + "epoch": 0.38, + "learning_rate": 1.4173914092822863e-05, + "loss": 1.1688, + "step": 15887 + }, + { + "epoch": 0.38, + "learning_rate": 1.4173206277945035e-05, + "loss": 1.2477, + "step": 15888 + }, + { + "epoch": 0.38, + "learning_rate": 1.4172498437749348e-05, + "loss": 1.2096, + "step": 15889 + }, + { + "epoch": 0.38, + "learning_rate": 1.4171790572240092e-05, + "loss": 1.0755, + "step": 15890 + }, + { + "epoch": 0.38, + "learning_rate": 1.4171082681421567e-05, + "loss": 0.6321, + "step": 15891 + }, + { + "epoch": 0.38, + "learning_rate": 1.417037476529806e-05, + "loss": 1.0287, + "step": 15892 + }, + { + "epoch": 0.38, + "learning_rate": 1.4169666823873873e-05, + "loss": 1.1307, + "step": 15893 + }, + { + "epoch": 0.38, + "learning_rate": 1.4168958857153298e-05, + "loss": 1.1635, + "step": 15894 + }, + { + "epoch": 0.38, + "learning_rate": 1.416825086514063e-05, + "loss": 0.9766, + "step": 15895 + }, + { + "epoch": 0.38, + "learning_rate": 1.4167542847840166e-05, + "loss": 1.0719, + "step": 15896 + }, + { + "epoch": 0.38, + "learning_rate": 1.4166834805256197e-05, + "loss": 1.0084, + "step": 15897 + }, + { + "epoch": 0.38, + "learning_rate": 1.4166126737393024e-05, + "loss": 0.9286, + "step": 15898 + }, + { + "epoch": 0.38, + "learning_rate": 1.416541864425494e-05, + "loss": 1.151, + "step": 15899 + }, + { + "epoch": 0.38, + "learning_rate": 1.4164710525846238e-05, + "loss": 0.9568, + "step": 15900 + }, + { + "epoch": 0.38, + "learning_rate": 1.416400238217122e-05, + "loss": 1.3194, + "step": 15901 + }, + { + "epoch": 0.38, + "learning_rate": 1.4163294213234177e-05, + "loss": 1.013, + "step": 15902 + }, + { + "epoch": 0.38, + "learning_rate": 1.416258601903941e-05, + "loss": 1.1888, + "step": 15903 + }, + { + "epoch": 0.38, + "learning_rate": 1.4161877799591209e-05, + "loss": 1.3088, + "step": 15904 + }, + { + "epoch": 0.38, + "learning_rate": 1.4161169554893875e-05, + "loss": 0.892, + "step": 15905 + }, + { + "epoch": 0.38, + "learning_rate": 1.4160461284951705e-05, + "loss": 1.1068, + "step": 15906 + }, + { + "epoch": 0.38, + "learning_rate": 1.4159752989768998e-05, + "loss": 0.9346, + "step": 15907 + }, + { + "epoch": 0.38, + "learning_rate": 1.4159044669350041e-05, + "loss": 1.0107, + "step": 15908 + }, + { + "epoch": 0.38, + "learning_rate": 1.4158336323699145e-05, + "loss": 1.1692, + "step": 15909 + }, + { + "epoch": 0.38, + "learning_rate": 1.4157627952820598e-05, + "loss": 1.004, + "step": 15910 + }, + { + "epoch": 0.38, + "learning_rate": 1.41569195567187e-05, + "loss": 1.0, + "step": 15911 + }, + { + "epoch": 0.38, + "learning_rate": 1.4156211135397748e-05, + "loss": 1.2382, + "step": 15912 + }, + { + "epoch": 0.38, + "learning_rate": 1.4155502688862042e-05, + "loss": 1.0873, + "step": 15913 + }, + { + "epoch": 0.38, + "learning_rate": 1.4154794217115878e-05, + "loss": 1.143, + "step": 15914 + }, + { + "epoch": 0.38, + "learning_rate": 1.4154085720163556e-05, + "loss": 0.9094, + "step": 15915 + }, + { + "epoch": 0.38, + "learning_rate": 1.415337719800937e-05, + "loss": 1.06, + "step": 15916 + }, + { + "epoch": 0.38, + "learning_rate": 1.4152668650657623e-05, + "loss": 1.148, + "step": 15917 + }, + { + "epoch": 0.38, + "learning_rate": 1.4151960078112614e-05, + "loss": 1.2113, + "step": 15918 + }, + { + "epoch": 0.38, + "learning_rate": 1.4151251480378638e-05, + "loss": 1.1406, + "step": 15919 + }, + { + "epoch": 0.38, + "learning_rate": 1.4150542857459995e-05, + "loss": 1.1603, + "step": 15920 + }, + { + "epoch": 0.38, + "learning_rate": 1.4149834209360987e-05, + "loss": 0.9632, + "step": 15921 + }, + { + "epoch": 0.38, + "learning_rate": 1.4149125536085909e-05, + "loss": 1.1075, + "step": 15922 + }, + { + "epoch": 0.38, + "learning_rate": 1.4148416837639063e-05, + "loss": 0.94, + "step": 15923 + }, + { + "epoch": 0.38, + "learning_rate": 1.4147708114024747e-05, + "loss": 1.1433, + "step": 15924 + }, + { + "epoch": 0.38, + "learning_rate": 1.4146999365247262e-05, + "loss": 1.179, + "step": 15925 + }, + { + "epoch": 0.38, + "learning_rate": 1.414629059131091e-05, + "loss": 1.0006, + "step": 15926 + }, + { + "epoch": 0.38, + "learning_rate": 1.4145581792219985e-05, + "loss": 1.4103, + "step": 15927 + }, + { + "epoch": 0.38, + "learning_rate": 1.4144872967978789e-05, + "loss": 0.9548, + "step": 15928 + }, + { + "epoch": 0.38, + "learning_rate": 1.4144164118591625e-05, + "loss": 0.9456, + "step": 15929 + }, + { + "epoch": 0.38, + "learning_rate": 1.4143455244062796e-05, + "loss": 0.9246, + "step": 15930 + }, + { + "epoch": 0.38, + "learning_rate": 1.4142746344396593e-05, + "loss": 1.2255, + "step": 15931 + }, + { + "epoch": 0.38, + "learning_rate": 1.4142037419597326e-05, + "loss": 1.308, + "step": 15932 + }, + { + "epoch": 0.38, + "learning_rate": 1.4141328469669291e-05, + "loss": 1.2909, + "step": 15933 + }, + { + "epoch": 0.38, + "learning_rate": 1.414061949461679e-05, + "loss": 0.9708, + "step": 15934 + }, + { + "epoch": 0.38, + "learning_rate": 1.4139910494444127e-05, + "loss": 0.9581, + "step": 15935 + }, + { + "epoch": 0.38, + "learning_rate": 1.4139201469155596e-05, + "loss": 1.0678, + "step": 15936 + }, + { + "epoch": 0.38, + "learning_rate": 1.4138492418755506e-05, + "loss": 1.0854, + "step": 15937 + }, + { + "epoch": 0.38, + "learning_rate": 1.4137783343248158e-05, + "loss": 1.0997, + "step": 15938 + }, + { + "epoch": 0.38, + "learning_rate": 1.4137074242637847e-05, + "loss": 1.274, + "step": 15939 + }, + { + "epoch": 0.38, + "learning_rate": 1.4136365116928883e-05, + "loss": 1.2278, + "step": 15940 + }, + { + "epoch": 0.38, + "learning_rate": 1.4135655966125566e-05, + "loss": 1.2055, + "step": 15941 + }, + { + "epoch": 0.38, + "learning_rate": 1.4134946790232193e-05, + "loss": 0.9462, + "step": 15942 + }, + { + "epoch": 0.38, + "learning_rate": 1.4134237589253073e-05, + "loss": 1.0135, + "step": 15943 + }, + { + "epoch": 0.38, + "learning_rate": 1.4133528363192506e-05, + "loss": 1.0661, + "step": 15944 + }, + { + "epoch": 0.38, + "learning_rate": 1.4132819112054795e-05, + "loss": 1.0587, + "step": 15945 + }, + { + "epoch": 0.38, + "learning_rate": 1.4132109835844241e-05, + "loss": 1.0631, + "step": 15946 + }, + { + "epoch": 0.38, + "learning_rate": 1.4131400534565148e-05, + "loss": 1.0357, + "step": 15947 + }, + { + "epoch": 0.38, + "learning_rate": 1.413069120822182e-05, + "loss": 1.3198, + "step": 15948 + }, + { + "epoch": 0.38, + "learning_rate": 1.4129981856818561e-05, + "loss": 1.3497, + "step": 15949 + }, + { + "epoch": 0.38, + "learning_rate": 1.4129272480359674e-05, + "loss": 1.199, + "step": 15950 + }, + { + "epoch": 0.38, + "learning_rate": 1.4128563078849463e-05, + "loss": 1.3128, + "step": 15951 + }, + { + "epoch": 0.38, + "learning_rate": 1.412785365229223e-05, + "loss": 1.0386, + "step": 15952 + }, + { + "epoch": 0.38, + "learning_rate": 1.4127144200692278e-05, + "loss": 1.162, + "step": 15953 + }, + { + "epoch": 0.38, + "learning_rate": 1.4126434724053915e-05, + "loss": 0.9222, + "step": 15954 + }, + { + "epoch": 0.38, + "learning_rate": 1.412572522238144e-05, + "loss": 1.1791, + "step": 15955 + }, + { + "epoch": 0.38, + "learning_rate": 1.4125015695679163e-05, + "loss": 1.2188, + "step": 15956 + }, + { + "epoch": 0.38, + "learning_rate": 1.4124306143951385e-05, + "loss": 1.078, + "step": 15957 + }, + { + "epoch": 0.38, + "learning_rate": 1.4123596567202413e-05, + "loss": 1.2153, + "step": 15958 + }, + { + "epoch": 0.38, + "learning_rate": 1.412288696543655e-05, + "loss": 1.187, + "step": 15959 + }, + { + "epoch": 0.38, + "learning_rate": 1.4122177338658101e-05, + "loss": 1.197, + "step": 15960 + }, + { + "epoch": 0.38, + "learning_rate": 1.4121467686871373e-05, + "loss": 0.8655, + "step": 15961 + }, + { + "epoch": 0.38, + "learning_rate": 1.412075801008067e-05, + "loss": 1.0257, + "step": 15962 + }, + { + "epoch": 0.38, + "learning_rate": 1.4120048308290297e-05, + "loss": 0.9165, + "step": 15963 + }, + { + "epoch": 0.38, + "learning_rate": 1.4119338581504559e-05, + "loss": 0.8367, + "step": 15964 + }, + { + "epoch": 0.38, + "learning_rate": 1.4118628829727763e-05, + "loss": 1.0417, + "step": 15965 + }, + { + "epoch": 0.38, + "learning_rate": 1.4117919052964216e-05, + "loss": 1.0963, + "step": 15966 + }, + { + "epoch": 0.38, + "learning_rate": 1.4117209251218223e-05, + "loss": 0.821, + "step": 15967 + }, + { + "epoch": 0.38, + "learning_rate": 1.4116499424494085e-05, + "loss": 1.004, + "step": 15968 + }, + { + "epoch": 0.38, + "learning_rate": 1.411578957279612e-05, + "loss": 1.1334, + "step": 15969 + }, + { + "epoch": 0.38, + "learning_rate": 1.4115079696128624e-05, + "loss": 1.2223, + "step": 15970 + }, + { + "epoch": 0.38, + "learning_rate": 1.411436979449591e-05, + "loss": 1.1593, + "step": 15971 + }, + { + "epoch": 0.38, + "learning_rate": 1.4113659867902278e-05, + "loss": 1.0778, + "step": 15972 + }, + { + "epoch": 0.38, + "learning_rate": 1.411294991635204e-05, + "loss": 0.7866, + "step": 15973 + }, + { + "epoch": 0.38, + "learning_rate": 1.4112239939849507e-05, + "loss": 1.2513, + "step": 15974 + }, + { + "epoch": 0.38, + "learning_rate": 1.411152993839898e-05, + "loss": 1.0898, + "step": 15975 + }, + { + "epoch": 0.38, + "learning_rate": 1.4110819912004763e-05, + "loss": 1.2255, + "step": 15976 + }, + { + "epoch": 0.38, + "learning_rate": 1.4110109860671172e-05, + "loss": 1.3171, + "step": 15977 + }, + { + "epoch": 0.38, + "learning_rate": 1.4109399784402513e-05, + "loss": 1.1703, + "step": 15978 + }, + { + "epoch": 0.38, + "learning_rate": 1.4108689683203088e-05, + "loss": 1.1002, + "step": 15979 + }, + { + "epoch": 0.38, + "learning_rate": 1.4107979557077212e-05, + "loss": 1.091, + "step": 15980 + }, + { + "epoch": 0.38, + "learning_rate": 1.4107269406029192e-05, + "loss": 0.8813, + "step": 15981 + }, + { + "epoch": 0.38, + "learning_rate": 1.410655923006333e-05, + "loss": 1.2468, + "step": 15982 + }, + { + "epoch": 0.38, + "learning_rate": 1.410584902918394e-05, + "loss": 1.0835, + "step": 15983 + }, + { + "epoch": 0.38, + "learning_rate": 1.4105138803395333e-05, + "loss": 1.1057, + "step": 15984 + }, + { + "epoch": 0.38, + "learning_rate": 1.4104428552701812e-05, + "loss": 1.1044, + "step": 15985 + }, + { + "epoch": 0.38, + "learning_rate": 1.410371827710769e-05, + "loss": 1.2082, + "step": 15986 + }, + { + "epoch": 0.38, + "learning_rate": 1.4103007976617271e-05, + "loss": 1.1617, + "step": 15987 + }, + { + "epoch": 0.38, + "learning_rate": 1.410229765123487e-05, + "loss": 1.0368, + "step": 15988 + }, + { + "epoch": 0.38, + "learning_rate": 1.4101587300964794e-05, + "loss": 1.0282, + "step": 15989 + }, + { + "epoch": 0.38, + "learning_rate": 1.4100876925811355e-05, + "loss": 0.9694, + "step": 15990 + }, + { + "epoch": 0.38, + "learning_rate": 1.4100166525778856e-05, + "loss": 0.7629, + "step": 15991 + }, + { + "epoch": 0.38, + "learning_rate": 1.4099456100871613e-05, + "loss": 1.1926, + "step": 15992 + }, + { + "epoch": 0.38, + "learning_rate": 1.4098745651093934e-05, + "loss": 1.2231, + "step": 15993 + }, + { + "epoch": 0.38, + "learning_rate": 1.4098035176450131e-05, + "loss": 1.168, + "step": 15994 + }, + { + "epoch": 0.38, + "learning_rate": 1.4097324676944508e-05, + "loss": 1.2334, + "step": 15995 + }, + { + "epoch": 0.38, + "learning_rate": 1.4096614152581384e-05, + "loss": 1.2228, + "step": 15996 + }, + { + "epoch": 0.38, + "learning_rate": 1.4095903603365067e-05, + "loss": 1.1175, + "step": 15997 + }, + { + "epoch": 0.38, + "learning_rate": 1.4095193029299863e-05, + "loss": 1.164, + "step": 15998 + }, + { + "epoch": 0.38, + "learning_rate": 1.4094482430390084e-05, + "loss": 1.1877, + "step": 15999 + }, + { + "epoch": 0.38, + "learning_rate": 1.4093771806640046e-05, + "loss": 0.9961, + "step": 16000 + }, + { + "epoch": 0.38, + "learning_rate": 1.4093061158054058e-05, + "loss": 1.2989, + "step": 16001 + }, + { + "epoch": 0.38, + "learning_rate": 1.4092350484636432e-05, + "loss": 1.13, + "step": 16002 + }, + { + "epoch": 0.38, + "learning_rate": 1.4091639786391474e-05, + "loss": 1.1272, + "step": 16003 + }, + { + "epoch": 0.38, + "learning_rate": 1.4090929063323502e-05, + "loss": 1.1194, + "step": 16004 + }, + { + "epoch": 0.38, + "learning_rate": 1.4090218315436825e-05, + "loss": 1.0905, + "step": 16005 + }, + { + "epoch": 0.38, + "learning_rate": 1.4089507542735758e-05, + "loss": 1.1423, + "step": 16006 + }, + { + "epoch": 0.38, + "learning_rate": 1.4088796745224607e-05, + "loss": 1.2422, + "step": 16007 + }, + { + "epoch": 0.38, + "learning_rate": 1.4088085922907691e-05, + "loss": 1.0888, + "step": 16008 + }, + { + "epoch": 0.38, + "learning_rate": 1.4087375075789318e-05, + "loss": 1.1736, + "step": 16009 + }, + { + "epoch": 0.39, + "learning_rate": 1.4086664203873803e-05, + "loss": 0.9681, + "step": 16010 + }, + { + "epoch": 0.39, + "learning_rate": 1.4085953307165455e-05, + "loss": 1.2032, + "step": 16011 + }, + { + "epoch": 0.39, + "learning_rate": 1.4085242385668592e-05, + "loss": 0.9009, + "step": 16012 + }, + { + "epoch": 0.39, + "learning_rate": 1.4084531439387522e-05, + "loss": 1.2844, + "step": 16013 + }, + { + "epoch": 0.39, + "learning_rate": 1.4083820468326562e-05, + "loss": 1.1734, + "step": 16014 + }, + { + "epoch": 0.39, + "learning_rate": 1.4083109472490022e-05, + "loss": 1.2232, + "step": 16015 + }, + { + "epoch": 0.39, + "learning_rate": 1.4082398451882218e-05, + "loss": 1.1722, + "step": 16016 + }, + { + "epoch": 0.39, + "learning_rate": 1.4081687406507465e-05, + "loss": 0.931, + "step": 16017 + }, + { + "epoch": 0.39, + "learning_rate": 1.4080976336370074e-05, + "loss": 0.8107, + "step": 16018 + }, + { + "epoch": 0.39, + "learning_rate": 1.4080265241474356e-05, + "loss": 1.12, + "step": 16019 + }, + { + "epoch": 0.39, + "learning_rate": 1.4079554121824633e-05, + "loss": 1.0097, + "step": 16020 + }, + { + "epoch": 0.39, + "learning_rate": 1.407884297742521e-05, + "loss": 1.0694, + "step": 16021 + }, + { + "epoch": 0.39, + "learning_rate": 1.4078131808280409e-05, + "loss": 1.1203, + "step": 16022 + }, + { + "epoch": 0.39, + "learning_rate": 1.4077420614394542e-05, + "loss": 1.1532, + "step": 16023 + }, + { + "epoch": 0.39, + "learning_rate": 1.4076709395771921e-05, + "loss": 1.0866, + "step": 16024 + }, + { + "epoch": 0.39, + "learning_rate": 1.4075998152416865e-05, + "loss": 0.9793, + "step": 16025 + }, + { + "epoch": 0.39, + "learning_rate": 1.4075286884333687e-05, + "loss": 1.2213, + "step": 16026 + }, + { + "epoch": 0.39, + "learning_rate": 1.40745755915267e-05, + "loss": 1.2684, + "step": 16027 + }, + { + "epoch": 0.39, + "learning_rate": 1.4073864274000224e-05, + "loss": 0.9757, + "step": 16028 + }, + { + "epoch": 0.39, + "learning_rate": 1.4073152931758571e-05, + "loss": 1.3453, + "step": 16029 + }, + { + "epoch": 0.39, + "learning_rate": 1.4072441564806055e-05, + "loss": 1.0864, + "step": 16030 + }, + { + "epoch": 0.39, + "learning_rate": 1.4071730173146995e-05, + "loss": 1.0817, + "step": 16031 + }, + { + "epoch": 0.39, + "learning_rate": 1.4071018756785709e-05, + "loss": 0.9374, + "step": 16032 + }, + { + "epoch": 0.39, + "learning_rate": 1.4070307315726509e-05, + "loss": 1.1214, + "step": 16033 + }, + { + "epoch": 0.39, + "learning_rate": 1.4069595849973708e-05, + "loss": 1.0298, + "step": 16034 + }, + { + "epoch": 0.39, + "learning_rate": 1.4068884359531626e-05, + "loss": 1.1821, + "step": 16035 + }, + { + "epoch": 0.39, + "learning_rate": 1.4068172844404584e-05, + "loss": 0.9465, + "step": 16036 + }, + { + "epoch": 0.39, + "learning_rate": 1.4067461304596892e-05, + "loss": 0.9962, + "step": 16037 + }, + { + "epoch": 0.39, + "learning_rate": 1.4066749740112869e-05, + "loss": 1.0688, + "step": 16038 + }, + { + "epoch": 0.39, + "learning_rate": 1.406603815095683e-05, + "loss": 1.1925, + "step": 16039 + }, + { + "epoch": 0.39, + "learning_rate": 1.4065326537133094e-05, + "loss": 1.2581, + "step": 16040 + }, + { + "epoch": 0.39, + "learning_rate": 1.406461489864598e-05, + "loss": 1.2689, + "step": 16041 + }, + { + "epoch": 0.39, + "learning_rate": 1.4063903235499807e-05, + "loss": 1.1776, + "step": 16042 + }, + { + "epoch": 0.39, + "learning_rate": 1.4063191547698883e-05, + "loss": 1.0321, + "step": 16043 + }, + { + "epoch": 0.39, + "learning_rate": 1.4062479835247533e-05, + "loss": 1.0426, + "step": 16044 + }, + { + "epoch": 0.39, + "learning_rate": 1.4061768098150077e-05, + "loss": 1.1879, + "step": 16045 + }, + { + "epoch": 0.39, + "learning_rate": 1.4061056336410825e-05, + "loss": 1.1553, + "step": 16046 + }, + { + "epoch": 0.39, + "learning_rate": 1.4060344550034102e-05, + "loss": 1.1464, + "step": 16047 + }, + { + "epoch": 0.39, + "learning_rate": 1.4059632739024223e-05, + "loss": 0.8916, + "step": 16048 + }, + { + "epoch": 0.39, + "learning_rate": 1.4058920903385507e-05, + "loss": 1.2221, + "step": 16049 + }, + { + "epoch": 0.39, + "learning_rate": 1.4058209043122273e-05, + "loss": 1.2545, + "step": 16050 + }, + { + "epoch": 0.39, + "learning_rate": 1.4057497158238837e-05, + "loss": 1.2351, + "step": 16051 + }, + { + "epoch": 0.39, + "learning_rate": 1.4056785248739524e-05, + "loss": 0.9463, + "step": 16052 + }, + { + "epoch": 0.39, + "learning_rate": 1.4056073314628647e-05, + "loss": 1.2023, + "step": 16053 + }, + { + "epoch": 0.39, + "learning_rate": 1.4055361355910528e-05, + "loss": 1.2607, + "step": 16054 + }, + { + "epoch": 0.39, + "learning_rate": 1.4054649372589482e-05, + "loss": 0.8835, + "step": 16055 + }, + { + "epoch": 0.39, + "learning_rate": 1.4053937364669836e-05, + "loss": 1.1485, + "step": 16056 + }, + { + "epoch": 0.39, + "learning_rate": 1.4053225332155904e-05, + "loss": 1.3112, + "step": 16057 + }, + { + "epoch": 0.39, + "learning_rate": 1.4052513275052008e-05, + "loss": 1.1216, + "step": 16058 + }, + { + "epoch": 0.39, + "learning_rate": 1.4051801193362466e-05, + "loss": 0.8939, + "step": 16059 + }, + { + "epoch": 0.39, + "learning_rate": 1.40510890870916e-05, + "loss": 1.0954, + "step": 16060 + }, + { + "epoch": 0.39, + "learning_rate": 1.405037695624373e-05, + "loss": 0.8962, + "step": 16061 + }, + { + "epoch": 0.39, + "learning_rate": 1.4049664800823175e-05, + "loss": 1.3629, + "step": 16062 + }, + { + "epoch": 0.39, + "learning_rate": 1.4048952620834255e-05, + "loss": 1.1575, + "step": 16063 + }, + { + "epoch": 0.39, + "learning_rate": 1.4048240416281296e-05, + "loss": 1.1716, + "step": 16064 + }, + { + "epoch": 0.39, + "learning_rate": 1.4047528187168612e-05, + "loss": 0.9674, + "step": 16065 + }, + { + "epoch": 0.39, + "learning_rate": 1.4046815933500524e-05, + "loss": 0.7951, + "step": 16066 + }, + { + "epoch": 0.39, + "learning_rate": 1.4046103655281358e-05, + "loss": 0.9138, + "step": 16067 + }, + { + "epoch": 0.39, + "learning_rate": 1.4045391352515432e-05, + "loss": 1.2987, + "step": 16068 + }, + { + "epoch": 0.39, + "learning_rate": 1.404467902520707e-05, + "loss": 1.1481, + "step": 16069 + }, + { + "epoch": 0.39, + "learning_rate": 1.404396667336059e-05, + "loss": 1.034, + "step": 16070 + }, + { + "epoch": 0.39, + "learning_rate": 1.4043254296980316e-05, + "loss": 1.2561, + "step": 16071 + }, + { + "epoch": 0.39, + "learning_rate": 1.404254189607057e-05, + "loss": 1.2364, + "step": 16072 + }, + { + "epoch": 0.39, + "learning_rate": 1.4041829470635672e-05, + "loss": 1.2504, + "step": 16073 + }, + { + "epoch": 0.39, + "learning_rate": 1.4041117020679943e-05, + "loss": 1.1021, + "step": 16074 + }, + { + "epoch": 0.39, + "learning_rate": 1.4040404546207713e-05, + "loss": 1.0445, + "step": 16075 + }, + { + "epoch": 0.39, + "learning_rate": 1.4039692047223297e-05, + "loss": 1.1345, + "step": 16076 + }, + { + "epoch": 0.39, + "learning_rate": 1.4038979523731018e-05, + "loss": 1.0106, + "step": 16077 + }, + { + "epoch": 0.39, + "learning_rate": 1.40382669757352e-05, + "loss": 1.0941, + "step": 16078 + }, + { + "epoch": 0.39, + "learning_rate": 1.4037554403240169e-05, + "loss": 1.3074, + "step": 16079 + }, + { + "epoch": 0.39, + "learning_rate": 1.4036841806250242e-05, + "loss": 1.2711, + "step": 16080 + }, + { + "epoch": 0.39, + "learning_rate": 1.4036129184769746e-05, + "loss": 0.9381, + "step": 16081 + }, + { + "epoch": 0.39, + "learning_rate": 1.4035416538803004e-05, + "loss": 1.1073, + "step": 16082 + }, + { + "epoch": 0.39, + "learning_rate": 1.403470386835434e-05, + "loss": 1.0109, + "step": 16083 + }, + { + "epoch": 0.39, + "learning_rate": 1.4033991173428077e-05, + "loss": 1.1347, + "step": 16084 + }, + { + "epoch": 0.39, + "learning_rate": 1.4033278454028536e-05, + "loss": 0.9856, + "step": 16085 + }, + { + "epoch": 0.39, + "learning_rate": 1.4032565710160044e-05, + "loss": 1.2348, + "step": 16086 + }, + { + "epoch": 0.39, + "learning_rate": 1.4031852941826924e-05, + "loss": 1.0464, + "step": 16087 + }, + { + "epoch": 0.39, + "learning_rate": 1.4031140149033502e-05, + "loss": 0.9214, + "step": 16088 + }, + { + "epoch": 0.39, + "learning_rate": 1.4030427331784102e-05, + "loss": 1.0664, + "step": 16089 + }, + { + "epoch": 0.39, + "learning_rate": 1.4029714490083043e-05, + "loss": 1.1887, + "step": 16090 + }, + { + "epoch": 0.39, + "learning_rate": 1.4029001623934657e-05, + "loss": 0.9958, + "step": 16091 + }, + { + "epoch": 0.39, + "learning_rate": 1.4028288733343267e-05, + "loss": 1.0857, + "step": 16092 + }, + { + "epoch": 0.39, + "learning_rate": 1.4027575818313195e-05, + "loss": 0.9208, + "step": 16093 + }, + { + "epoch": 0.39, + "learning_rate": 1.4026862878848767e-05, + "loss": 1.0318, + "step": 16094 + }, + { + "epoch": 0.39, + "learning_rate": 1.4026149914954312e-05, + "loss": 0.9525, + "step": 16095 + }, + { + "epoch": 0.39, + "learning_rate": 1.402543692663415e-05, + "loss": 1.099, + "step": 16096 + }, + { + "epoch": 0.39, + "learning_rate": 1.4024723913892611e-05, + "loss": 0.9655, + "step": 16097 + }, + { + "epoch": 0.39, + "learning_rate": 1.4024010876734016e-05, + "loss": 1.4405, + "step": 16098 + }, + { + "epoch": 0.39, + "learning_rate": 1.4023297815162696e-05, + "loss": 1.0152, + "step": 16099 + }, + { + "epoch": 0.39, + "learning_rate": 1.4022584729182973e-05, + "loss": 1.0651, + "step": 16100 + }, + { + "epoch": 0.39, + "learning_rate": 1.4021871618799177e-05, + "loss": 1.161, + "step": 16101 + }, + { + "epoch": 0.39, + "learning_rate": 1.4021158484015626e-05, + "loss": 1.3334, + "step": 16102 + }, + { + "epoch": 0.39, + "learning_rate": 1.4020445324836658e-05, + "loss": 1.0786, + "step": 16103 + }, + { + "epoch": 0.39, + "learning_rate": 1.4019732141266593e-05, + "loss": 1.4973, + "step": 16104 + }, + { + "epoch": 0.39, + "learning_rate": 1.4019018933309758e-05, + "loss": 0.9411, + "step": 16105 + }, + { + "epoch": 0.39, + "learning_rate": 1.4018305700970479e-05, + "loss": 1.2625, + "step": 16106 + }, + { + "epoch": 0.39, + "learning_rate": 1.4017592444253088e-05, + "loss": 0.8002, + "step": 16107 + }, + { + "epoch": 0.39, + "learning_rate": 1.4016879163161906e-05, + "loss": 1.1223, + "step": 16108 + }, + { + "epoch": 0.39, + "learning_rate": 1.401616585770126e-05, + "loss": 1.1461, + "step": 16109 + }, + { + "epoch": 0.39, + "learning_rate": 1.4015452527875487e-05, + "loss": 1.2028, + "step": 16110 + }, + { + "epoch": 0.39, + "learning_rate": 1.4014739173688903e-05, + "loss": 1.0628, + "step": 16111 + }, + { + "epoch": 0.39, + "learning_rate": 1.4014025795145845e-05, + "loss": 1.0213, + "step": 16112 + }, + { + "epoch": 0.39, + "learning_rate": 1.401331239225063e-05, + "loss": 1.0522, + "step": 16113 + }, + { + "epoch": 0.39, + "learning_rate": 1.4012598965007598e-05, + "loss": 1.0982, + "step": 16114 + }, + { + "epoch": 0.39, + "learning_rate": 1.401188551342107e-05, + "loss": 1.2709, + "step": 16115 + }, + { + "epoch": 0.39, + "learning_rate": 1.401117203749538e-05, + "loss": 1.0251, + "step": 16116 + }, + { + "epoch": 0.39, + "learning_rate": 1.4010458537234847e-05, + "loss": 1.0111, + "step": 16117 + }, + { + "epoch": 0.39, + "learning_rate": 1.4009745012643809e-05, + "loss": 0.9596, + "step": 16118 + }, + { + "epoch": 0.39, + "learning_rate": 1.400903146372659e-05, + "loss": 1.3188, + "step": 16119 + }, + { + "epoch": 0.39, + "learning_rate": 1.4008317890487522e-05, + "loss": 0.9839, + "step": 16120 + }, + { + "epoch": 0.39, + "learning_rate": 1.4007604292930928e-05, + "loss": 1.2069, + "step": 16121 + }, + { + "epoch": 0.39, + "learning_rate": 1.4006890671061143e-05, + "loss": 1.066, + "step": 16122 + }, + { + "epoch": 0.39, + "learning_rate": 1.40061770248825e-05, + "loss": 1.1823, + "step": 16123 + }, + { + "epoch": 0.39, + "learning_rate": 1.4005463354399316e-05, + "loss": 1.2668, + "step": 16124 + }, + { + "epoch": 0.39, + "learning_rate": 1.400474965961593e-05, + "loss": 1.0601, + "step": 16125 + }, + { + "epoch": 0.39, + "learning_rate": 1.4004035940536671e-05, + "loss": 0.9051, + "step": 16126 + }, + { + "epoch": 0.39, + "learning_rate": 1.4003322197165867e-05, + "loss": 1.2007, + "step": 16127 + }, + { + "epoch": 0.39, + "learning_rate": 1.4002608429507852e-05, + "loss": 1.2392, + "step": 16128 + }, + { + "epoch": 0.39, + "learning_rate": 1.4001894637566948e-05, + "loss": 1.1088, + "step": 16129 + }, + { + "epoch": 0.39, + "learning_rate": 1.4001180821347492e-05, + "loss": 1.0404, + "step": 16130 + }, + { + "epoch": 0.39, + "learning_rate": 1.4000466980853815e-05, + "loss": 1.2507, + "step": 16131 + }, + { + "epoch": 0.39, + "learning_rate": 1.3999753116090244e-05, + "loss": 1.1541, + "step": 16132 + }, + { + "epoch": 0.39, + "learning_rate": 1.3999039227061115e-05, + "loss": 1.0087, + "step": 16133 + }, + { + "epoch": 0.39, + "learning_rate": 1.3998325313770751e-05, + "loss": 1.1806, + "step": 16134 + }, + { + "epoch": 0.39, + "learning_rate": 1.399761137622349e-05, + "loss": 1.2629, + "step": 16135 + }, + { + "epoch": 0.39, + "learning_rate": 1.3996897414423663e-05, + "loss": 1.1589, + "step": 16136 + }, + { + "epoch": 0.39, + "learning_rate": 1.3996183428375595e-05, + "loss": 1.1673, + "step": 16137 + }, + { + "epoch": 0.39, + "learning_rate": 1.3995469418083625e-05, + "loss": 1.0668, + "step": 16138 + }, + { + "epoch": 0.39, + "learning_rate": 1.3994755383552084e-05, + "loss": 1.2008, + "step": 16139 + }, + { + "epoch": 0.39, + "learning_rate": 1.3994041324785298e-05, + "loss": 0.8271, + "step": 16140 + }, + { + "epoch": 0.39, + "learning_rate": 1.3993327241787604e-05, + "loss": 0.943, + "step": 16141 + }, + { + "epoch": 0.39, + "learning_rate": 1.3992613134563331e-05, + "loss": 1.2569, + "step": 16142 + }, + { + "epoch": 0.39, + "learning_rate": 1.3991899003116815e-05, + "loss": 1.1749, + "step": 16143 + }, + { + "epoch": 0.39, + "learning_rate": 1.3991184847452388e-05, + "loss": 0.6838, + "step": 16144 + }, + { + "epoch": 0.39, + "learning_rate": 1.3990470667574381e-05, + "loss": 1.2886, + "step": 16145 + }, + { + "epoch": 0.39, + "learning_rate": 1.3989756463487126e-05, + "loss": 1.1193, + "step": 16146 + }, + { + "epoch": 0.39, + "learning_rate": 1.3989042235194957e-05, + "loss": 0.9642, + "step": 16147 + }, + { + "epoch": 0.39, + "learning_rate": 1.3988327982702208e-05, + "loss": 0.8729, + "step": 16148 + }, + { + "epoch": 0.39, + "learning_rate": 1.3987613706013207e-05, + "loss": 1.2353, + "step": 16149 + }, + { + "epoch": 0.39, + "learning_rate": 1.3986899405132297e-05, + "loss": 1.171, + "step": 16150 + }, + { + "epoch": 0.39, + "learning_rate": 1.3986185080063806e-05, + "loss": 1.0691, + "step": 16151 + }, + { + "epoch": 0.39, + "learning_rate": 1.3985470730812065e-05, + "loss": 1.1648, + "step": 16152 + }, + { + "epoch": 0.39, + "learning_rate": 1.398475635738141e-05, + "loss": 1.1477, + "step": 16153 + }, + { + "epoch": 0.39, + "learning_rate": 1.3984041959776177e-05, + "loss": 1.1876, + "step": 16154 + }, + { + "epoch": 0.39, + "learning_rate": 1.3983327538000699e-05, + "loss": 1.1767, + "step": 16155 + }, + { + "epoch": 0.39, + "learning_rate": 1.398261309205931e-05, + "loss": 0.9315, + "step": 16156 + }, + { + "epoch": 0.39, + "learning_rate": 1.398189862195634e-05, + "loss": 1.2155, + "step": 16157 + }, + { + "epoch": 0.39, + "learning_rate": 1.3981184127696133e-05, + "loss": 1.2065, + "step": 16158 + }, + { + "epoch": 0.39, + "learning_rate": 1.3980469609283016e-05, + "loss": 1.3561, + "step": 16159 + }, + { + "epoch": 0.39, + "learning_rate": 1.3979755066721327e-05, + "loss": 1.1048, + "step": 16160 + }, + { + "epoch": 0.39, + "learning_rate": 1.39790405000154e-05, + "loss": 0.9798, + "step": 16161 + }, + { + "epoch": 0.39, + "learning_rate": 1.3978325909169568e-05, + "loss": 1.3651, + "step": 16162 + }, + { + "epoch": 0.39, + "learning_rate": 1.3977611294188171e-05, + "loss": 1.2261, + "step": 16163 + }, + { + "epoch": 0.39, + "learning_rate": 1.3976896655075543e-05, + "loss": 1.1899, + "step": 16164 + }, + { + "epoch": 0.39, + "learning_rate": 1.3976181991836013e-05, + "loss": 1.1513, + "step": 16165 + }, + { + "epoch": 0.39, + "learning_rate": 1.3975467304473926e-05, + "loss": 1.0702, + "step": 16166 + }, + { + "epoch": 0.39, + "learning_rate": 1.3974752592993615e-05, + "loss": 1.2505, + "step": 16167 + }, + { + "epoch": 0.39, + "learning_rate": 1.3974037857399414e-05, + "loss": 1.0821, + "step": 16168 + }, + { + "epoch": 0.39, + "learning_rate": 1.3973323097695657e-05, + "loss": 0.9767, + "step": 16169 + }, + { + "epoch": 0.39, + "learning_rate": 1.397260831388669e-05, + "loss": 1.0528, + "step": 16170 + }, + { + "epoch": 0.39, + "learning_rate": 1.3971893505976837e-05, + "loss": 1.1144, + "step": 16171 + }, + { + "epoch": 0.39, + "learning_rate": 1.3971178673970442e-05, + "loss": 1.0868, + "step": 16172 + }, + { + "epoch": 0.39, + "learning_rate": 1.3970463817871837e-05, + "loss": 1.1912, + "step": 16173 + }, + { + "epoch": 0.39, + "learning_rate": 1.3969748937685368e-05, + "loss": 1.205, + "step": 16174 + }, + { + "epoch": 0.39, + "learning_rate": 1.3969034033415363e-05, + "loss": 1.2951, + "step": 16175 + }, + { + "epoch": 0.39, + "learning_rate": 1.3968319105066162e-05, + "loss": 1.2164, + "step": 16176 + }, + { + "epoch": 0.39, + "learning_rate": 1.39676041526421e-05, + "loss": 0.916, + "step": 16177 + }, + { + "epoch": 0.39, + "learning_rate": 1.396688917614752e-05, + "loss": 1.1417, + "step": 16178 + }, + { + "epoch": 0.39, + "learning_rate": 1.3966174175586758e-05, + "loss": 1.2301, + "step": 16179 + }, + { + "epoch": 0.39, + "learning_rate": 1.3965459150964146e-05, + "loss": 1.2598, + "step": 16180 + }, + { + "epoch": 0.39, + "learning_rate": 1.3964744102284027e-05, + "loss": 1.0553, + "step": 16181 + }, + { + "epoch": 0.39, + "learning_rate": 1.3964029029550741e-05, + "loss": 1.0309, + "step": 16182 + }, + { + "epoch": 0.39, + "learning_rate": 1.396331393276862e-05, + "loss": 1.0028, + "step": 16183 + }, + { + "epoch": 0.39, + "learning_rate": 1.3962598811942007e-05, + "loss": 1.204, + "step": 16184 + }, + { + "epoch": 0.39, + "learning_rate": 1.396188366707524e-05, + "loss": 1.0679, + "step": 16185 + }, + { + "epoch": 0.39, + "learning_rate": 1.3961168498172655e-05, + "loss": 1.1032, + "step": 16186 + }, + { + "epoch": 0.39, + "learning_rate": 1.3960453305238595e-05, + "loss": 0.9754, + "step": 16187 + }, + { + "epoch": 0.39, + "learning_rate": 1.3959738088277393e-05, + "loss": 0.9743, + "step": 16188 + }, + { + "epoch": 0.39, + "learning_rate": 1.3959022847293393e-05, + "loss": 0.82, + "step": 16189 + }, + { + "epoch": 0.39, + "learning_rate": 1.3958307582290932e-05, + "loss": 1.0905, + "step": 16190 + }, + { + "epoch": 0.39, + "learning_rate": 1.3957592293274351e-05, + "loss": 1.2977, + "step": 16191 + }, + { + "epoch": 0.39, + "learning_rate": 1.3956876980247988e-05, + "loss": 1.205, + "step": 16192 + }, + { + "epoch": 0.39, + "learning_rate": 1.3956161643216182e-05, + "loss": 1.2018, + "step": 16193 + }, + { + "epoch": 0.39, + "learning_rate": 1.3955446282183274e-05, + "loss": 0.995, + "step": 16194 + }, + { + "epoch": 0.39, + "learning_rate": 1.3954730897153605e-05, + "loss": 1.1317, + "step": 16195 + }, + { + "epoch": 0.39, + "learning_rate": 1.3954015488131511e-05, + "loss": 1.1404, + "step": 16196 + }, + { + "epoch": 0.39, + "learning_rate": 1.3953300055121339e-05, + "loss": 1.3568, + "step": 16197 + }, + { + "epoch": 0.39, + "learning_rate": 1.3952584598127425e-05, + "loss": 1.206, + "step": 16198 + }, + { + "epoch": 0.39, + "learning_rate": 1.395186911715411e-05, + "loss": 1.3779, + "step": 16199 + }, + { + "epoch": 0.39, + "learning_rate": 1.3951153612205732e-05, + "loss": 0.932, + "step": 16200 + }, + { + "epoch": 0.39, + "learning_rate": 1.3950438083286636e-05, + "loss": 1.2657, + "step": 16201 + }, + { + "epoch": 0.39, + "learning_rate": 1.3949722530401162e-05, + "loss": 1.0974, + "step": 16202 + }, + { + "epoch": 0.39, + "learning_rate": 1.394900695355365e-05, + "loss": 1.2226, + "step": 16203 + }, + { + "epoch": 0.39, + "learning_rate": 1.394829135274844e-05, + "loss": 1.0752, + "step": 16204 + }, + { + "epoch": 0.39, + "learning_rate": 1.3947575727989876e-05, + "loss": 1.0271, + "step": 16205 + }, + { + "epoch": 0.39, + "learning_rate": 1.39468600792823e-05, + "loss": 1.1724, + "step": 16206 + }, + { + "epoch": 0.39, + "learning_rate": 1.394614440663005e-05, + "loss": 0.9696, + "step": 16207 + }, + { + "epoch": 0.39, + "learning_rate": 1.3945428710037472e-05, + "loss": 1.183, + "step": 16208 + }, + { + "epoch": 0.39, + "learning_rate": 1.3944712989508903e-05, + "loss": 1.1316, + "step": 16209 + }, + { + "epoch": 0.39, + "learning_rate": 1.394399724504869e-05, + "loss": 1.1881, + "step": 16210 + }, + { + "epoch": 0.39, + "learning_rate": 1.3943281476661176e-05, + "loss": 1.1182, + "step": 16211 + }, + { + "epoch": 0.39, + "learning_rate": 1.39425656843507e-05, + "loss": 1.1966, + "step": 16212 + }, + { + "epoch": 0.39, + "learning_rate": 1.3941849868121603e-05, + "loss": 1.0869, + "step": 16213 + }, + { + "epoch": 0.39, + "learning_rate": 1.394113402797823e-05, + "loss": 1.0105, + "step": 16214 + }, + { + "epoch": 0.39, + "learning_rate": 1.3940418163924925e-05, + "loss": 1.0921, + "step": 16215 + }, + { + "epoch": 0.39, + "learning_rate": 1.3939702275966031e-05, + "loss": 1.1347, + "step": 16216 + }, + { + "epoch": 0.39, + "learning_rate": 1.393898636410589e-05, + "loss": 0.9161, + "step": 16217 + }, + { + "epoch": 0.39, + "learning_rate": 1.3938270428348842e-05, + "loss": 1.311, + "step": 16218 + }, + { + "epoch": 0.39, + "learning_rate": 1.3937554468699235e-05, + "loss": 1.3392, + "step": 16219 + }, + { + "epoch": 0.39, + "learning_rate": 1.3936838485161413e-05, + "loss": 0.8353, + "step": 16220 + }, + { + "epoch": 0.39, + "learning_rate": 1.3936122477739715e-05, + "loss": 0.9451, + "step": 16221 + }, + { + "epoch": 0.39, + "learning_rate": 1.3935406446438491e-05, + "loss": 0.9364, + "step": 16222 + }, + { + "epoch": 0.39, + "learning_rate": 1.393469039126208e-05, + "loss": 1.0151, + "step": 16223 + }, + { + "epoch": 0.39, + "learning_rate": 1.3933974312214828e-05, + "loss": 1.0592, + "step": 16224 + }, + { + "epoch": 0.39, + "learning_rate": 1.393325820930108e-05, + "loss": 1.1735, + "step": 16225 + }, + { + "epoch": 0.39, + "learning_rate": 1.3932542082525177e-05, + "loss": 1.2286, + "step": 16226 + }, + { + "epoch": 0.39, + "learning_rate": 1.393182593189147e-05, + "loss": 1.1936, + "step": 16227 + }, + { + "epoch": 0.39, + "learning_rate": 1.3931109757404299e-05, + "loss": 1.3713, + "step": 16228 + }, + { + "epoch": 0.39, + "learning_rate": 1.3930393559068007e-05, + "loss": 1.3674, + "step": 16229 + }, + { + "epoch": 0.39, + "learning_rate": 1.3929677336886945e-05, + "loss": 1.1853, + "step": 16230 + }, + { + "epoch": 0.39, + "learning_rate": 1.3928961090865451e-05, + "loss": 1.0883, + "step": 16231 + }, + { + "epoch": 0.39, + "learning_rate": 1.3928244821007879e-05, + "loss": 1.1911, + "step": 16232 + }, + { + "epoch": 0.39, + "learning_rate": 1.3927528527318566e-05, + "loss": 1.2473, + "step": 16233 + }, + { + "epoch": 0.39, + "learning_rate": 1.3926812209801863e-05, + "loss": 1.1491, + "step": 16234 + }, + { + "epoch": 0.39, + "learning_rate": 1.3926095868462115e-05, + "loss": 0.8875, + "step": 16235 + }, + { + "epoch": 0.39, + "learning_rate": 1.3925379503303662e-05, + "loss": 0.8465, + "step": 16236 + }, + { + "epoch": 0.39, + "learning_rate": 1.3924663114330858e-05, + "loss": 1.2969, + "step": 16237 + }, + { + "epoch": 0.39, + "learning_rate": 1.3923946701548049e-05, + "loss": 1.1029, + "step": 16238 + }, + { + "epoch": 0.39, + "learning_rate": 1.3923230264959574e-05, + "loss": 1.31, + "step": 16239 + }, + { + "epoch": 0.39, + "learning_rate": 1.3922513804569784e-05, + "loss": 1.3189, + "step": 16240 + }, + { + "epoch": 0.39, + "learning_rate": 1.3921797320383027e-05, + "loss": 1.1389, + "step": 16241 + }, + { + "epoch": 0.39, + "learning_rate": 1.3921080812403648e-05, + "loss": 1.2588, + "step": 16242 + }, + { + "epoch": 0.39, + "learning_rate": 1.3920364280635996e-05, + "loss": 0.9292, + "step": 16243 + }, + { + "epoch": 0.39, + "learning_rate": 1.3919647725084411e-05, + "loss": 0.9042, + "step": 16244 + }, + { + "epoch": 0.39, + "learning_rate": 1.391893114575325e-05, + "loss": 1.0006, + "step": 16245 + }, + { + "epoch": 0.39, + "learning_rate": 1.3918214542646853e-05, + "loss": 1.2789, + "step": 16246 + }, + { + "epoch": 0.39, + "learning_rate": 1.3917497915769573e-05, + "loss": 0.8185, + "step": 16247 + }, + { + "epoch": 0.39, + "learning_rate": 1.391678126512575e-05, + "loss": 1.2286, + "step": 16248 + }, + { + "epoch": 0.39, + "learning_rate": 1.3916064590719738e-05, + "loss": 1.1229, + "step": 16249 + }, + { + "epoch": 0.39, + "learning_rate": 1.3915347892555887e-05, + "loss": 1.233, + "step": 16250 + }, + { + "epoch": 0.39, + "learning_rate": 1.3914631170638537e-05, + "loss": 1.2302, + "step": 16251 + }, + { + "epoch": 0.39, + "learning_rate": 1.391391442497204e-05, + "loss": 0.998, + "step": 16252 + }, + { + "epoch": 0.39, + "learning_rate": 1.3913197655560746e-05, + "loss": 1.0018, + "step": 16253 + }, + { + "epoch": 0.39, + "learning_rate": 1.3912480862409004e-05, + "loss": 1.2478, + "step": 16254 + }, + { + "epoch": 0.39, + "learning_rate": 1.391176404552116e-05, + "loss": 1.1894, + "step": 16255 + }, + { + "epoch": 0.39, + "learning_rate": 1.391104720490156e-05, + "loss": 1.0456, + "step": 16256 + }, + { + "epoch": 0.39, + "learning_rate": 1.391033034055456e-05, + "loss": 1.013, + "step": 16257 + }, + { + "epoch": 0.39, + "learning_rate": 1.3909613452484508e-05, + "loss": 0.9898, + "step": 16258 + }, + { + "epoch": 0.39, + "learning_rate": 1.3908896540695746e-05, + "loss": 1.3069, + "step": 16259 + }, + { + "epoch": 0.39, + "learning_rate": 1.3908179605192631e-05, + "loss": 0.9025, + "step": 16260 + }, + { + "epoch": 0.39, + "learning_rate": 1.3907462645979507e-05, + "loss": 1.0468, + "step": 16261 + }, + { + "epoch": 0.39, + "learning_rate": 1.390674566306073e-05, + "loss": 1.1764, + "step": 16262 + }, + { + "epoch": 0.39, + "learning_rate": 1.3906028656440644e-05, + "loss": 1.137, + "step": 16263 + }, + { + "epoch": 0.39, + "learning_rate": 1.39053116261236e-05, + "loss": 0.7378, + "step": 16264 + }, + { + "epoch": 0.39, + "learning_rate": 1.390459457211395e-05, + "loss": 1.0721, + "step": 16265 + }, + { + "epoch": 0.39, + "learning_rate": 1.3903877494416043e-05, + "loss": 1.1933, + "step": 16266 + }, + { + "epoch": 0.39, + "learning_rate": 1.390316039303423e-05, + "loss": 1.3226, + "step": 16267 + }, + { + "epoch": 0.39, + "learning_rate": 1.3902443267972858e-05, + "loss": 1.1523, + "step": 16268 + }, + { + "epoch": 0.39, + "learning_rate": 1.3901726119236285e-05, + "loss": 1.1216, + "step": 16269 + }, + { + "epoch": 0.39, + "learning_rate": 1.3901008946828856e-05, + "loss": 1.0614, + "step": 16270 + }, + { + "epoch": 0.39, + "learning_rate": 1.3900291750754922e-05, + "loss": 1.2019, + "step": 16271 + }, + { + "epoch": 0.39, + "learning_rate": 1.3899574531018836e-05, + "loss": 0.9339, + "step": 16272 + }, + { + "epoch": 0.39, + "learning_rate": 1.389885728762495e-05, + "loss": 1.2238, + "step": 16273 + }, + { + "epoch": 0.39, + "learning_rate": 1.3898140020577611e-05, + "loss": 1.175, + "step": 16274 + }, + { + "epoch": 0.39, + "learning_rate": 1.3897422729881176e-05, + "loss": 0.8628, + "step": 16275 + }, + { + "epoch": 0.39, + "learning_rate": 1.389670541553999e-05, + "loss": 1.1467, + "step": 16276 + }, + { + "epoch": 0.39, + "learning_rate": 1.3895988077558414e-05, + "loss": 1.3175, + "step": 16277 + }, + { + "epoch": 0.39, + "learning_rate": 1.3895270715940792e-05, + "loss": 1.1082, + "step": 16278 + }, + { + "epoch": 0.39, + "learning_rate": 1.389455333069148e-05, + "loss": 0.9071, + "step": 16279 + }, + { + "epoch": 0.39, + "learning_rate": 1.3893835921814826e-05, + "loss": 1.1744, + "step": 16280 + }, + { + "epoch": 0.39, + "learning_rate": 1.3893118489315189e-05, + "loss": 1.05, + "step": 16281 + }, + { + "epoch": 0.39, + "learning_rate": 1.3892401033196916e-05, + "loss": 0.8795, + "step": 16282 + }, + { + "epoch": 0.39, + "learning_rate": 1.3891683553464362e-05, + "loss": 1.0172, + "step": 16283 + }, + { + "epoch": 0.39, + "learning_rate": 1.3890966050121879e-05, + "loss": 1.191, + "step": 16284 + }, + { + "epoch": 0.39, + "learning_rate": 1.3890248523173821e-05, + "loss": 1.3208, + "step": 16285 + }, + { + "epoch": 0.39, + "learning_rate": 1.388953097262454e-05, + "loss": 0.7308, + "step": 16286 + }, + { + "epoch": 0.39, + "learning_rate": 1.3888813398478387e-05, + "loss": 1.1701, + "step": 16287 + }, + { + "epoch": 0.39, + "learning_rate": 1.388809580073972e-05, + "loss": 1.2463, + "step": 16288 + }, + { + "epoch": 0.39, + "learning_rate": 1.3887378179412891e-05, + "loss": 1.0891, + "step": 16289 + }, + { + "epoch": 0.39, + "learning_rate": 1.388666053450225e-05, + "loss": 1.2372, + "step": 16290 + }, + { + "epoch": 0.39, + "learning_rate": 1.3885942866012157e-05, + "loss": 0.9419, + "step": 16291 + }, + { + "epoch": 0.39, + "learning_rate": 1.3885225173946961e-05, + "loss": 1.0504, + "step": 16292 + }, + { + "epoch": 0.39, + "learning_rate": 1.3884507458311018e-05, + "loss": 1.1458, + "step": 16293 + }, + { + "epoch": 0.39, + "learning_rate": 1.3883789719108682e-05, + "loss": 1.0979, + "step": 16294 + }, + { + "epoch": 0.39, + "learning_rate": 1.388307195634431e-05, + "loss": 1.3839, + "step": 16295 + }, + { + "epoch": 0.39, + "learning_rate": 1.3882354170022247e-05, + "loss": 1.3595, + "step": 16296 + }, + { + "epoch": 0.39, + "learning_rate": 1.388163636014686e-05, + "loss": 1.1683, + "step": 16297 + }, + { + "epoch": 0.39, + "learning_rate": 1.3880918526722497e-05, + "loss": 1.18, + "step": 16298 + }, + { + "epoch": 0.39, + "learning_rate": 1.3880200669753517e-05, + "loss": 1.147, + "step": 16299 + }, + { + "epoch": 0.39, + "learning_rate": 1.3879482789244267e-05, + "loss": 1.2388, + "step": 16300 + }, + { + "epoch": 0.39, + "learning_rate": 1.387876488519911e-05, + "loss": 1.1976, + "step": 16301 + }, + { + "epoch": 0.39, + "learning_rate": 1.38780469576224e-05, + "loss": 0.8989, + "step": 16302 + }, + { + "epoch": 0.39, + "learning_rate": 1.3877329006518489e-05, + "loss": 1.1216, + "step": 16303 + }, + { + "epoch": 0.39, + "learning_rate": 1.3876611031891733e-05, + "loss": 1.1949, + "step": 16304 + }, + { + "epoch": 0.39, + "learning_rate": 1.3875893033746495e-05, + "loss": 0.9361, + "step": 16305 + }, + { + "epoch": 0.39, + "learning_rate": 1.3875175012087123e-05, + "loss": 1.0779, + "step": 16306 + }, + { + "epoch": 0.39, + "learning_rate": 1.3874456966917978e-05, + "loss": 1.0982, + "step": 16307 + }, + { + "epoch": 0.39, + "learning_rate": 1.3873738898243407e-05, + "loss": 1.0033, + "step": 16308 + }, + { + "epoch": 0.39, + "learning_rate": 1.387302080606778e-05, + "loss": 0.9702, + "step": 16309 + }, + { + "epoch": 0.39, + "learning_rate": 1.3872302690395445e-05, + "loss": 1.2257, + "step": 16310 + }, + { + "epoch": 0.39, + "learning_rate": 1.387158455123076e-05, + "loss": 0.6843, + "step": 16311 + }, + { + "epoch": 0.39, + "learning_rate": 1.387086638857808e-05, + "loss": 0.9038, + "step": 16312 + }, + { + "epoch": 0.39, + "learning_rate": 1.3870148202441768e-05, + "loss": 1.3404, + "step": 16313 + }, + { + "epoch": 0.39, + "learning_rate": 1.3869429992826175e-05, + "loss": 1.1342, + "step": 16314 + }, + { + "epoch": 0.39, + "learning_rate": 1.3868711759735663e-05, + "loss": 1.1587, + "step": 16315 + }, + { + "epoch": 0.39, + "learning_rate": 1.3867993503174584e-05, + "loss": 1.1259, + "step": 16316 + }, + { + "epoch": 0.39, + "learning_rate": 1.3867275223147298e-05, + "loss": 1.0344, + "step": 16317 + }, + { + "epoch": 0.39, + "learning_rate": 1.3866556919658165e-05, + "loss": 1.1553, + "step": 16318 + }, + { + "epoch": 0.39, + "learning_rate": 1.3865838592711538e-05, + "loss": 0.9094, + "step": 16319 + }, + { + "epoch": 0.39, + "learning_rate": 1.3865120242311778e-05, + "loss": 1.0418, + "step": 16320 + }, + { + "epoch": 0.39, + "learning_rate": 1.3864401868463247e-05, + "loss": 1.2325, + "step": 16321 + }, + { + "epoch": 0.39, + "learning_rate": 1.3863683471170296e-05, + "loss": 0.8444, + "step": 16322 + }, + { + "epoch": 0.39, + "learning_rate": 1.3862965050437282e-05, + "loss": 1.3186, + "step": 16323 + }, + { + "epoch": 0.39, + "learning_rate": 1.3862246606268572e-05, + "loss": 1.1249, + "step": 16324 + }, + { + "epoch": 0.39, + "learning_rate": 1.3861528138668519e-05, + "loss": 1.1486, + "step": 16325 + }, + { + "epoch": 0.39, + "learning_rate": 1.3860809647641484e-05, + "loss": 1.1231, + "step": 16326 + }, + { + "epoch": 0.39, + "learning_rate": 1.3860091133191825e-05, + "loss": 1.2488, + "step": 16327 + }, + { + "epoch": 0.39, + "learning_rate": 1.3859372595323902e-05, + "loss": 0.9326, + "step": 16328 + }, + { + "epoch": 0.39, + "learning_rate": 1.3858654034042069e-05, + "loss": 1.2573, + "step": 16329 + }, + { + "epoch": 0.39, + "learning_rate": 1.3857935449350695e-05, + "loss": 1.1538, + "step": 16330 + }, + { + "epoch": 0.39, + "learning_rate": 1.3857216841254128e-05, + "loss": 1.324, + "step": 16331 + }, + { + "epoch": 0.39, + "learning_rate": 1.3856498209756738e-05, + "loss": 0.9498, + "step": 16332 + }, + { + "epoch": 0.39, + "learning_rate": 1.385577955486288e-05, + "loss": 1.2624, + "step": 16333 + }, + { + "epoch": 0.39, + "learning_rate": 1.3855060876576915e-05, + "loss": 1.3925, + "step": 16334 + }, + { + "epoch": 0.39, + "learning_rate": 1.3854342174903202e-05, + "loss": 0.9214, + "step": 16335 + }, + { + "epoch": 0.39, + "learning_rate": 1.3853623449846102e-05, + "loss": 1.1572, + "step": 16336 + }, + { + "epoch": 0.39, + "learning_rate": 1.3852904701409972e-05, + "loss": 1.2793, + "step": 16337 + }, + { + "epoch": 0.39, + "learning_rate": 1.3852185929599178e-05, + "loss": 0.955, + "step": 16338 + }, + { + "epoch": 0.39, + "learning_rate": 1.3851467134418077e-05, + "loss": 1.0566, + "step": 16339 + }, + { + "epoch": 0.39, + "learning_rate": 1.3850748315871031e-05, + "loss": 1.1507, + "step": 16340 + }, + { + "epoch": 0.39, + "learning_rate": 1.3850029473962402e-05, + "loss": 1.0845, + "step": 16341 + }, + { + "epoch": 0.39, + "learning_rate": 1.3849310608696552e-05, + "loss": 1.1438, + "step": 16342 + }, + { + "epoch": 0.39, + "learning_rate": 1.3848591720077833e-05, + "loss": 1.0075, + "step": 16343 + }, + { + "epoch": 0.39, + "learning_rate": 1.3847872808110618e-05, + "loss": 1.2075, + "step": 16344 + }, + { + "epoch": 0.39, + "learning_rate": 1.3847153872799264e-05, + "loss": 0.9865, + "step": 16345 + }, + { + "epoch": 0.39, + "learning_rate": 1.384643491414813e-05, + "loss": 1.2178, + "step": 16346 + }, + { + "epoch": 0.39, + "learning_rate": 1.3845715932161581e-05, + "loss": 1.1901, + "step": 16347 + }, + { + "epoch": 0.39, + "learning_rate": 1.3844996926843979e-05, + "loss": 1.126, + "step": 16348 + }, + { + "epoch": 0.39, + "learning_rate": 1.3844277898199684e-05, + "loss": 0.9243, + "step": 16349 + }, + { + "epoch": 0.39, + "learning_rate": 1.3843558846233061e-05, + "loss": 0.9675, + "step": 16350 + }, + { + "epoch": 0.39, + "learning_rate": 1.3842839770948466e-05, + "loss": 1.0909, + "step": 16351 + }, + { + "epoch": 0.39, + "learning_rate": 1.384212067235027e-05, + "loss": 0.9605, + "step": 16352 + }, + { + "epoch": 0.39, + "learning_rate": 1.3841401550442833e-05, + "loss": 1.3927, + "step": 16353 + }, + { + "epoch": 0.39, + "learning_rate": 1.3840682405230515e-05, + "loss": 1.1527, + "step": 16354 + }, + { + "epoch": 0.39, + "learning_rate": 1.3839963236717675e-05, + "loss": 1.2338, + "step": 16355 + }, + { + "epoch": 0.39, + "learning_rate": 1.3839244044908689e-05, + "loss": 1.1526, + "step": 16356 + }, + { + "epoch": 0.39, + "learning_rate": 1.3838524829807906e-05, + "loss": 0.992, + "step": 16357 + }, + { + "epoch": 0.39, + "learning_rate": 1.38378055914197e-05, + "loss": 1.103, + "step": 16358 + }, + { + "epoch": 0.39, + "learning_rate": 1.3837086329748429e-05, + "loss": 1.0807, + "step": 16359 + }, + { + "epoch": 0.39, + "learning_rate": 1.3836367044798458e-05, + "loss": 1.2356, + "step": 16360 + }, + { + "epoch": 0.39, + "learning_rate": 1.3835647736574149e-05, + "loss": 0.7847, + "step": 16361 + }, + { + "epoch": 0.39, + "learning_rate": 1.383492840507987e-05, + "loss": 1.1058, + "step": 16362 + }, + { + "epoch": 0.39, + "learning_rate": 1.383420905031998e-05, + "loss": 1.1988, + "step": 16363 + }, + { + "epoch": 0.39, + "learning_rate": 1.3833489672298845e-05, + "loss": 0.9254, + "step": 16364 + }, + { + "epoch": 0.39, + "learning_rate": 1.3832770271020835e-05, + "loss": 1.3857, + "step": 16365 + }, + { + "epoch": 0.39, + "learning_rate": 1.3832050846490303e-05, + "loss": 1.0588, + "step": 16366 + }, + { + "epoch": 0.39, + "learning_rate": 1.3831331398711623e-05, + "loss": 1.1743, + "step": 16367 + }, + { + "epoch": 0.39, + "learning_rate": 1.3830611927689157e-05, + "loss": 0.9357, + "step": 16368 + }, + { + "epoch": 0.39, + "learning_rate": 1.3829892433427268e-05, + "loss": 0.8761, + "step": 16369 + }, + { + "epoch": 0.39, + "learning_rate": 1.3829172915930322e-05, + "loss": 1.2132, + "step": 16370 + }, + { + "epoch": 0.39, + "learning_rate": 1.3828453375202686e-05, + "loss": 1.1094, + "step": 16371 + }, + { + "epoch": 0.39, + "learning_rate": 1.3827733811248724e-05, + "loss": 0.8783, + "step": 16372 + }, + { + "epoch": 0.39, + "learning_rate": 1.3827014224072801e-05, + "loss": 1.0413, + "step": 16373 + }, + { + "epoch": 0.39, + "learning_rate": 1.382629461367928e-05, + "loss": 1.071, + "step": 16374 + }, + { + "epoch": 0.39, + "learning_rate": 1.3825574980072532e-05, + "loss": 1.1491, + "step": 16375 + }, + { + "epoch": 0.39, + "learning_rate": 1.382485532325692e-05, + "loss": 1.2505, + "step": 16376 + }, + { + "epoch": 0.39, + "learning_rate": 1.3824135643236813e-05, + "loss": 1.095, + "step": 16377 + }, + { + "epoch": 0.39, + "learning_rate": 1.3823415940016573e-05, + "loss": 0.9128, + "step": 16378 + }, + { + "epoch": 0.39, + "learning_rate": 1.3822696213600565e-05, + "loss": 1.2074, + "step": 16379 + }, + { + "epoch": 0.39, + "learning_rate": 1.3821976463993161e-05, + "loss": 0.9541, + "step": 16380 + }, + { + "epoch": 0.39, + "learning_rate": 1.3821256691198725e-05, + "loss": 0.9118, + "step": 16381 + }, + { + "epoch": 0.39, + "learning_rate": 1.3820536895221625e-05, + "loss": 1.1069, + "step": 16382 + }, + { + "epoch": 0.39, + "learning_rate": 1.3819817076066223e-05, + "loss": 1.1417, + "step": 16383 + }, + { + "epoch": 0.39, + "learning_rate": 1.3819097233736888e-05, + "loss": 0.9332, + "step": 16384 + }, + { + "epoch": 0.39, + "learning_rate": 1.3818377368237994e-05, + "loss": 1.2973, + "step": 16385 + }, + { + "epoch": 0.39, + "learning_rate": 1.38176574795739e-05, + "loss": 1.2338, + "step": 16386 + }, + { + "epoch": 0.39, + "learning_rate": 1.3816937567748974e-05, + "loss": 1.3003, + "step": 16387 + }, + { + "epoch": 0.39, + "learning_rate": 1.3816217632767588e-05, + "loss": 1.0361, + "step": 16388 + }, + { + "epoch": 0.39, + "learning_rate": 1.3815497674634106e-05, + "loss": 1.1306, + "step": 16389 + }, + { + "epoch": 0.39, + "learning_rate": 1.3814777693352899e-05, + "loss": 1.1928, + "step": 16390 + }, + { + "epoch": 0.39, + "learning_rate": 1.3814057688928328e-05, + "loss": 1.1053, + "step": 16391 + }, + { + "epoch": 0.39, + "learning_rate": 1.3813337661364772e-05, + "loss": 0.8912, + "step": 16392 + }, + { + "epoch": 0.39, + "learning_rate": 1.381261761066659e-05, + "loss": 1.0083, + "step": 16393 + }, + { + "epoch": 0.39, + "learning_rate": 1.3811897536838155e-05, + "loss": 0.9696, + "step": 16394 + }, + { + "epoch": 0.39, + "learning_rate": 1.3811177439883833e-05, + "loss": 1.0426, + "step": 16395 + }, + { + "epoch": 0.39, + "learning_rate": 1.3810457319807996e-05, + "loss": 1.3034, + "step": 16396 + }, + { + "epoch": 0.39, + "learning_rate": 1.380973717661501e-05, + "loss": 0.86, + "step": 16397 + }, + { + "epoch": 0.39, + "learning_rate": 1.3809017010309244e-05, + "loss": 1.2945, + "step": 16398 + }, + { + "epoch": 0.39, + "learning_rate": 1.3808296820895066e-05, + "loss": 1.0303, + "step": 16399 + }, + { + "epoch": 0.39, + "learning_rate": 1.3807576608376851e-05, + "loss": 1.1599, + "step": 16400 + }, + { + "epoch": 0.39, + "learning_rate": 1.3806856372758964e-05, + "loss": 1.0805, + "step": 16401 + }, + { + "epoch": 0.39, + "learning_rate": 1.3806136114045771e-05, + "loss": 1.2663, + "step": 16402 + }, + { + "epoch": 0.39, + "learning_rate": 1.3805415832241647e-05, + "loss": 1.0253, + "step": 16403 + }, + { + "epoch": 0.39, + "learning_rate": 1.3804695527350963e-05, + "loss": 1.0497, + "step": 16404 + }, + { + "epoch": 0.39, + "learning_rate": 1.3803975199378083e-05, + "loss": 1.1408, + "step": 16405 + }, + { + "epoch": 0.39, + "learning_rate": 1.380325484832738e-05, + "loss": 1.0032, + "step": 16406 + }, + { + "epoch": 0.39, + "learning_rate": 1.3802534474203223e-05, + "loss": 0.9776, + "step": 16407 + }, + { + "epoch": 0.39, + "learning_rate": 1.3801814077009988e-05, + "loss": 0.9689, + "step": 16408 + }, + { + "epoch": 0.39, + "learning_rate": 1.380109365675204e-05, + "loss": 1.0991, + "step": 16409 + }, + { + "epoch": 0.39, + "learning_rate": 1.3800373213433748e-05, + "loss": 0.8017, + "step": 16410 + }, + { + "epoch": 0.39, + "learning_rate": 1.3799652747059491e-05, + "loss": 1.2484, + "step": 16411 + }, + { + "epoch": 0.39, + "learning_rate": 1.3798932257633631e-05, + "loss": 1.2805, + "step": 16412 + }, + { + "epoch": 0.39, + "learning_rate": 1.3798211745160543e-05, + "loss": 1.0864, + "step": 16413 + }, + { + "epoch": 0.39, + "learning_rate": 1.3797491209644598e-05, + "loss": 1.0799, + "step": 16414 + }, + { + "epoch": 0.39, + "learning_rate": 1.3796770651090165e-05, + "loss": 1.0101, + "step": 16415 + }, + { + "epoch": 0.39, + "learning_rate": 1.3796050069501621e-05, + "loss": 1.1373, + "step": 16416 + }, + { + "epoch": 0.39, + "learning_rate": 1.3795329464883334e-05, + "loss": 1.191, + "step": 16417 + }, + { + "epoch": 0.39, + "learning_rate": 1.3794608837239671e-05, + "loss": 1.0005, + "step": 16418 + }, + { + "epoch": 0.39, + "learning_rate": 1.3793888186575013e-05, + "loss": 1.2806, + "step": 16419 + }, + { + "epoch": 0.39, + "learning_rate": 1.3793167512893728e-05, + "loss": 1.0264, + "step": 16420 + }, + { + "epoch": 0.39, + "learning_rate": 1.3792446816200187e-05, + "loss": 0.9537, + "step": 16421 + }, + { + "epoch": 0.39, + "learning_rate": 1.379172609649876e-05, + "loss": 1.3571, + "step": 16422 + }, + { + "epoch": 0.39, + "learning_rate": 1.3791005353793827e-05, + "loss": 1.1794, + "step": 16423 + }, + { + "epoch": 0.39, + "learning_rate": 1.3790284588089754e-05, + "loss": 1.3959, + "step": 16424 + }, + { + "epoch": 0.4, + "learning_rate": 1.378956379939092e-05, + "loss": 1.2796, + "step": 16425 + }, + { + "epoch": 0.4, + "learning_rate": 1.378884298770169e-05, + "loss": 1.0058, + "step": 16426 + }, + { + "epoch": 0.4, + "learning_rate": 1.378812215302644e-05, + "loss": 1.1269, + "step": 16427 + }, + { + "epoch": 0.4, + "learning_rate": 1.3787401295369547e-05, + "loss": 0.8661, + "step": 16428 + }, + { + "epoch": 0.4, + "learning_rate": 1.3786680414735378e-05, + "loss": 1.0963, + "step": 16429 + }, + { + "epoch": 0.4, + "learning_rate": 1.3785959511128311e-05, + "loss": 0.8468, + "step": 16430 + }, + { + "epoch": 0.4, + "learning_rate": 1.3785238584552718e-05, + "loss": 1.1272, + "step": 16431 + }, + { + "epoch": 0.4, + "learning_rate": 1.3784517635012974e-05, + "loss": 1.1248, + "step": 16432 + }, + { + "epoch": 0.4, + "learning_rate": 1.3783796662513451e-05, + "loss": 1.0368, + "step": 16433 + }, + { + "epoch": 0.4, + "learning_rate": 1.378307566705852e-05, + "loss": 0.9365, + "step": 16434 + }, + { + "epoch": 0.4, + "learning_rate": 1.3782354648652562e-05, + "loss": 1.0143, + "step": 16435 + }, + { + "epoch": 0.4, + "learning_rate": 1.3781633607299952e-05, + "loss": 0.661, + "step": 16436 + }, + { + "epoch": 0.4, + "learning_rate": 1.3780912543005057e-05, + "loss": 1.1407, + "step": 16437 + }, + { + "epoch": 0.4, + "learning_rate": 1.3780191455772252e-05, + "loss": 1.0733, + "step": 16438 + }, + { + "epoch": 0.4, + "learning_rate": 1.3779470345605916e-05, + "loss": 1.0285, + "step": 16439 + }, + { + "epoch": 0.4, + "learning_rate": 1.3778749212510426e-05, + "loss": 1.3297, + "step": 16440 + }, + { + "epoch": 0.4, + "learning_rate": 1.3778028056490151e-05, + "loss": 1.1952, + "step": 16441 + }, + { + "epoch": 0.4, + "learning_rate": 1.3777306877549465e-05, + "loss": 0.7504, + "step": 16442 + }, + { + "epoch": 0.4, + "learning_rate": 1.3776585675692751e-05, + "loss": 1.1432, + "step": 16443 + }, + { + "epoch": 0.4, + "learning_rate": 1.3775864450924379e-05, + "loss": 1.2769, + "step": 16444 + }, + { + "epoch": 0.4, + "learning_rate": 1.3775143203248726e-05, + "loss": 1.1632, + "step": 16445 + }, + { + "epoch": 0.4, + "learning_rate": 1.3774421932670166e-05, + "loss": 1.1378, + "step": 16446 + }, + { + "epoch": 0.4, + "learning_rate": 1.3773700639193078e-05, + "loss": 1.2286, + "step": 16447 + }, + { + "epoch": 0.4, + "learning_rate": 1.3772979322821833e-05, + "loss": 1.0254, + "step": 16448 + }, + { + "epoch": 0.4, + "learning_rate": 1.3772257983560812e-05, + "loss": 1.0224, + "step": 16449 + }, + { + "epoch": 0.4, + "learning_rate": 1.3771536621414387e-05, + "loss": 0.9606, + "step": 16450 + }, + { + "epoch": 0.4, + "learning_rate": 1.377081523638694e-05, + "loss": 1.1833, + "step": 16451 + }, + { + "epoch": 0.4, + "learning_rate": 1.377009382848284e-05, + "loss": 1.1428, + "step": 16452 + }, + { + "epoch": 0.4, + "learning_rate": 1.3769372397706469e-05, + "loss": 1.1468, + "step": 16453 + }, + { + "epoch": 0.4, + "learning_rate": 1.3768650944062203e-05, + "loss": 0.9097, + "step": 16454 + }, + { + "epoch": 0.4, + "learning_rate": 1.3767929467554418e-05, + "loss": 1.181, + "step": 16455 + }, + { + "epoch": 0.4, + "learning_rate": 1.376720796818749e-05, + "loss": 0.9552, + "step": 16456 + }, + { + "epoch": 0.4, + "learning_rate": 1.3766486445965797e-05, + "loss": 0.8488, + "step": 16457 + }, + { + "epoch": 0.4, + "learning_rate": 1.3765764900893717e-05, + "loss": 1.1181, + "step": 16458 + }, + { + "epoch": 0.4, + "learning_rate": 1.3765043332975627e-05, + "loss": 1.0791, + "step": 16459 + }, + { + "epoch": 0.4, + "learning_rate": 1.3764321742215907e-05, + "loss": 1.135, + "step": 16460 + }, + { + "epoch": 0.4, + "learning_rate": 1.3763600128618932e-05, + "loss": 1.0353, + "step": 16461 + }, + { + "epoch": 0.4, + "learning_rate": 1.3762878492189078e-05, + "loss": 1.1281, + "step": 16462 + }, + { + "epoch": 0.4, + "learning_rate": 1.3762156832930726e-05, + "loss": 1.2156, + "step": 16463 + }, + { + "epoch": 0.4, + "learning_rate": 1.3761435150848254e-05, + "loss": 1.0032, + "step": 16464 + }, + { + "epoch": 0.4, + "learning_rate": 1.376071344594604e-05, + "loss": 0.9917, + "step": 16465 + }, + { + "epoch": 0.4, + "learning_rate": 1.3759991718228461e-05, + "loss": 1.0789, + "step": 16466 + }, + { + "epoch": 0.4, + "learning_rate": 1.37592699676999e-05, + "loss": 1.2695, + "step": 16467 + }, + { + "epoch": 0.4, + "learning_rate": 1.3758548194364729e-05, + "loss": 1.2962, + "step": 16468 + }, + { + "epoch": 0.4, + "learning_rate": 1.3757826398227333e-05, + "loss": 0.9585, + "step": 16469 + }, + { + "epoch": 0.4, + "learning_rate": 1.3757104579292082e-05, + "loss": 1.1968, + "step": 16470 + }, + { + "epoch": 0.4, + "learning_rate": 1.3756382737563368e-05, + "loss": 0.9938, + "step": 16471 + }, + { + "epoch": 0.4, + "learning_rate": 1.375566087304556e-05, + "loss": 0.7841, + "step": 16472 + }, + { + "epoch": 0.4, + "learning_rate": 1.3754938985743042e-05, + "loss": 1.1262, + "step": 16473 + }, + { + "epoch": 0.4, + "learning_rate": 1.375421707566019e-05, + "loss": 1.0242, + "step": 16474 + }, + { + "epoch": 0.4, + "learning_rate": 1.3753495142801389e-05, + "loss": 1.2375, + "step": 16475 + }, + { + "epoch": 0.4, + "learning_rate": 1.3752773187171016e-05, + "loss": 0.8473, + "step": 16476 + }, + { + "epoch": 0.4, + "learning_rate": 1.375205120877345e-05, + "loss": 0.8093, + "step": 16477 + }, + { + "epoch": 0.4, + "learning_rate": 1.375132920761307e-05, + "loss": 0.9887, + "step": 16478 + }, + { + "epoch": 0.4, + "learning_rate": 1.375060718369426e-05, + "loss": 1.1411, + "step": 16479 + }, + { + "epoch": 0.4, + "learning_rate": 1.3749885137021399e-05, + "loss": 1.0382, + "step": 16480 + }, + { + "epoch": 0.4, + "learning_rate": 1.3749163067598868e-05, + "loss": 1.2426, + "step": 16481 + }, + { + "epoch": 0.4, + "learning_rate": 1.3748440975431041e-05, + "loss": 0.8492, + "step": 16482 + }, + { + "epoch": 0.4, + "learning_rate": 1.3747718860522308e-05, + "loss": 1.1643, + "step": 16483 + }, + { + "epoch": 0.4, + "learning_rate": 1.3746996722877048e-05, + "loss": 1.1052, + "step": 16484 + }, + { + "epoch": 0.4, + "learning_rate": 1.374627456249964e-05, + "loss": 1.1741, + "step": 16485 + }, + { + "epoch": 0.4, + "learning_rate": 1.374555237939446e-05, + "loss": 1.0191, + "step": 16486 + }, + { + "epoch": 0.4, + "learning_rate": 1.37448301735659e-05, + "loss": 1.2757, + "step": 16487 + }, + { + "epoch": 0.4, + "learning_rate": 1.3744107945018334e-05, + "loss": 1.305, + "step": 16488 + }, + { + "epoch": 0.4, + "learning_rate": 1.3743385693756146e-05, + "loss": 0.9548, + "step": 16489 + }, + { + "epoch": 0.4, + "learning_rate": 1.3742663419783717e-05, + "loss": 1.1085, + "step": 16490 + }, + { + "epoch": 0.4, + "learning_rate": 1.3741941123105431e-05, + "loss": 0.926, + "step": 16491 + }, + { + "epoch": 0.4, + "learning_rate": 1.374121880372567e-05, + "loss": 0.9938, + "step": 16492 + }, + { + "epoch": 0.4, + "learning_rate": 1.3740496461648808e-05, + "loss": 0.9135, + "step": 16493 + }, + { + "epoch": 0.4, + "learning_rate": 1.373977409687924e-05, + "loss": 1.0447, + "step": 16494 + }, + { + "epoch": 0.4, + "learning_rate": 1.373905170942134e-05, + "loss": 1.0265, + "step": 16495 + }, + { + "epoch": 0.4, + "learning_rate": 1.3738329299279491e-05, + "loss": 0.9742, + "step": 16496 + }, + { + "epoch": 0.4, + "learning_rate": 1.373760686645808e-05, + "loss": 1.1775, + "step": 16497 + }, + { + "epoch": 0.4, + "learning_rate": 1.3736884410961485e-05, + "loss": 0.9901, + "step": 16498 + }, + { + "epoch": 0.4, + "learning_rate": 1.3736161932794094e-05, + "loss": 1.0993, + "step": 16499 + }, + { + "epoch": 0.4, + "learning_rate": 1.3735439431960287e-05, + "loss": 1.1127, + "step": 16500 + }, + { + "epoch": 0.4, + "learning_rate": 1.3734716908464445e-05, + "loss": 0.9007, + "step": 16501 + }, + { + "epoch": 0.4, + "learning_rate": 1.3733994362310957e-05, + "loss": 1.1599, + "step": 16502 + }, + { + "epoch": 0.4, + "learning_rate": 1.3733271793504202e-05, + "loss": 1.025, + "step": 16503 + }, + { + "epoch": 0.4, + "learning_rate": 1.3732549202048566e-05, + "loss": 1.3071, + "step": 16504 + }, + { + "epoch": 0.4, + "learning_rate": 1.3731826587948429e-05, + "loss": 0.9583, + "step": 16505 + }, + { + "epoch": 0.4, + "learning_rate": 1.373110395120818e-05, + "loss": 1.2396, + "step": 16506 + }, + { + "epoch": 0.4, + "learning_rate": 1.3730381291832203e-05, + "loss": 1.0528, + "step": 16507 + }, + { + "epoch": 0.4, + "learning_rate": 1.3729658609824879e-05, + "loss": 1.1175, + "step": 16508 + }, + { + "epoch": 0.4, + "learning_rate": 1.3728935905190591e-05, + "loss": 0.9853, + "step": 16509 + }, + { + "epoch": 0.4, + "learning_rate": 1.3728213177933727e-05, + "loss": 1.0935, + "step": 16510 + }, + { + "epoch": 0.4, + "learning_rate": 1.3727490428058672e-05, + "loss": 1.0539, + "step": 16511 + }, + { + "epoch": 0.4, + "learning_rate": 1.372676765556981e-05, + "loss": 1.0289, + "step": 16512 + }, + { + "epoch": 0.4, + "learning_rate": 1.3726044860471522e-05, + "loss": 1.2206, + "step": 16513 + }, + { + "epoch": 0.4, + "learning_rate": 1.3725322042768199e-05, + "loss": 1.0586, + "step": 16514 + }, + { + "epoch": 0.4, + "learning_rate": 1.3724599202464222e-05, + "loss": 1.242, + "step": 16515 + }, + { + "epoch": 0.4, + "learning_rate": 1.3723876339563979e-05, + "loss": 1.2642, + "step": 16516 + }, + { + "epoch": 0.4, + "learning_rate": 1.3723153454071851e-05, + "loss": 1.338, + "step": 16517 + }, + { + "epoch": 0.4, + "learning_rate": 1.3722430545992232e-05, + "loss": 1.0808, + "step": 16518 + }, + { + "epoch": 0.4, + "learning_rate": 1.3721707615329499e-05, + "loss": 1.224, + "step": 16519 + }, + { + "epoch": 0.4, + "learning_rate": 1.3720984662088042e-05, + "loss": 1.1787, + "step": 16520 + }, + { + "epoch": 0.4, + "learning_rate": 1.3720261686272245e-05, + "loss": 0.9439, + "step": 16521 + }, + { + "epoch": 0.4, + "learning_rate": 1.3719538687886495e-05, + "loss": 1.0306, + "step": 16522 + }, + { + "epoch": 0.4, + "learning_rate": 1.3718815666935182e-05, + "loss": 1.1412, + "step": 16523 + }, + { + "epoch": 0.4, + "learning_rate": 1.3718092623422688e-05, + "loss": 1.1556, + "step": 16524 + }, + { + "epoch": 0.4, + "learning_rate": 1.3717369557353397e-05, + "loss": 1.208, + "step": 16525 + }, + { + "epoch": 0.4, + "learning_rate": 1.3716646468731703e-05, + "loss": 1.0781, + "step": 16526 + }, + { + "epoch": 0.4, + "learning_rate": 1.3715923357561989e-05, + "loss": 1.1793, + "step": 16527 + }, + { + "epoch": 0.4, + "learning_rate": 1.3715200223848642e-05, + "loss": 1.0053, + "step": 16528 + }, + { + "epoch": 0.4, + "learning_rate": 1.3714477067596046e-05, + "loss": 1.0669, + "step": 16529 + }, + { + "epoch": 0.4, + "learning_rate": 1.3713753888808594e-05, + "loss": 0.9639, + "step": 16530 + }, + { + "epoch": 0.4, + "learning_rate": 1.371303068749067e-05, + "loss": 1.0271, + "step": 16531 + }, + { + "epoch": 0.4, + "learning_rate": 1.3712307463646664e-05, + "loss": 1.0271, + "step": 16532 + }, + { + "epoch": 0.4, + "learning_rate": 1.371158421728096e-05, + "loss": 1.0113, + "step": 16533 + }, + { + "epoch": 0.4, + "learning_rate": 1.3710860948397948e-05, + "loss": 1.1284, + "step": 16534 + }, + { + "epoch": 0.4, + "learning_rate": 1.3710137657002016e-05, + "loss": 1.2511, + "step": 16535 + }, + { + "epoch": 0.4, + "learning_rate": 1.370941434309755e-05, + "loss": 1.095, + "step": 16536 + }, + { + "epoch": 0.4, + "learning_rate": 1.3708691006688941e-05, + "loss": 0.9315, + "step": 16537 + }, + { + "epoch": 0.4, + "learning_rate": 1.3707967647780576e-05, + "loss": 1.0422, + "step": 16538 + }, + { + "epoch": 0.4, + "learning_rate": 1.3707244266376843e-05, + "loss": 0.9694, + "step": 16539 + }, + { + "epoch": 0.4, + "learning_rate": 1.3706520862482128e-05, + "loss": 1.0039, + "step": 16540 + }, + { + "epoch": 0.4, + "learning_rate": 1.3705797436100825e-05, + "loss": 1.1176, + "step": 16541 + }, + { + "epoch": 0.4, + "learning_rate": 1.3705073987237323e-05, + "loss": 1.1614, + "step": 16542 + }, + { + "epoch": 0.4, + "learning_rate": 1.3704350515896008e-05, + "loss": 1.2391, + "step": 16543 + }, + { + "epoch": 0.4, + "learning_rate": 1.3703627022081269e-05, + "loss": 1.0576, + "step": 16544 + }, + { + "epoch": 0.4, + "learning_rate": 1.3702903505797493e-05, + "loss": 0.7466, + "step": 16545 + }, + { + "epoch": 0.4, + "learning_rate": 1.3702179967049076e-05, + "loss": 1.0092, + "step": 16546 + }, + { + "epoch": 0.4, + "learning_rate": 1.3701456405840403e-05, + "loss": 0.9967, + "step": 16547 + }, + { + "epoch": 0.4, + "learning_rate": 1.3700732822175866e-05, + "loss": 1.2148, + "step": 16548 + }, + { + "epoch": 0.4, + "learning_rate": 1.370000921605985e-05, + "loss": 1.1974, + "step": 16549 + }, + { + "epoch": 0.4, + "learning_rate": 1.3699285587496752e-05, + "loss": 0.8062, + "step": 16550 + }, + { + "epoch": 0.4, + "learning_rate": 1.3698561936490958e-05, + "loss": 0.7694, + "step": 16551 + }, + { + "epoch": 0.4, + "learning_rate": 1.3697838263046856e-05, + "loss": 0.9478, + "step": 16552 + }, + { + "epoch": 0.4, + "learning_rate": 1.3697114567168841e-05, + "loss": 1.0258, + "step": 16553 + }, + { + "epoch": 0.4, + "learning_rate": 1.3696390848861301e-05, + "loss": 1.2616, + "step": 16554 + }, + { + "epoch": 0.4, + "learning_rate": 1.3695667108128628e-05, + "loss": 1.2406, + "step": 16555 + }, + { + "epoch": 0.4, + "learning_rate": 1.3694943344975214e-05, + "loss": 1.1176, + "step": 16556 + }, + { + "epoch": 0.4, + "learning_rate": 1.3694219559405441e-05, + "loss": 0.9937, + "step": 16557 + }, + { + "epoch": 0.4, + "learning_rate": 1.3693495751423714e-05, + "loss": 1.1658, + "step": 16558 + }, + { + "epoch": 0.4, + "learning_rate": 1.3692771921034414e-05, + "loss": 0.9359, + "step": 16559 + }, + { + "epoch": 0.4, + "learning_rate": 1.3692048068241936e-05, + "loss": 1.0605, + "step": 16560 + }, + { + "epoch": 0.4, + "learning_rate": 1.3691324193050668e-05, + "loss": 1.0274, + "step": 16561 + }, + { + "epoch": 0.4, + "learning_rate": 1.3690600295465006e-05, + "loss": 1.0499, + "step": 16562 + }, + { + "epoch": 0.4, + "learning_rate": 1.3689876375489341e-05, + "loss": 1.1626, + "step": 16563 + }, + { + "epoch": 0.4, + "learning_rate": 1.3689152433128066e-05, + "loss": 1.1437, + "step": 16564 + }, + { + "epoch": 0.4, + "learning_rate": 1.3688428468385566e-05, + "loss": 1.1079, + "step": 16565 + }, + { + "epoch": 0.4, + "learning_rate": 1.3687704481266242e-05, + "loss": 1.1531, + "step": 16566 + }, + { + "epoch": 0.4, + "learning_rate": 1.368698047177448e-05, + "loss": 0.9383, + "step": 16567 + }, + { + "epoch": 0.4, + "learning_rate": 1.3686256439914678e-05, + "loss": 0.927, + "step": 16568 + }, + { + "epoch": 0.4, + "learning_rate": 1.368553238569122e-05, + "loss": 0.9848, + "step": 16569 + }, + { + "epoch": 0.4, + "learning_rate": 1.3684808309108508e-05, + "loss": 1.2643, + "step": 16570 + }, + { + "epoch": 0.4, + "learning_rate": 1.3684084210170929e-05, + "loss": 1.367, + "step": 16571 + }, + { + "epoch": 0.4, + "learning_rate": 1.368336008888288e-05, + "loss": 1.1333, + "step": 16572 + }, + { + "epoch": 0.4, + "learning_rate": 1.3682635945248748e-05, + "loss": 0.8982, + "step": 16573 + }, + { + "epoch": 0.4, + "learning_rate": 1.368191177927293e-05, + "loss": 1.0036, + "step": 16574 + }, + { + "epoch": 0.4, + "learning_rate": 1.3681187590959823e-05, + "loss": 1.0304, + "step": 16575 + }, + { + "epoch": 0.4, + "learning_rate": 1.3680463380313818e-05, + "loss": 1.162, + "step": 16576 + }, + { + "epoch": 0.4, + "learning_rate": 1.3679739147339301e-05, + "loss": 1.0655, + "step": 16577 + }, + { + "epoch": 0.4, + "learning_rate": 1.3679014892040675e-05, + "loss": 0.7518, + "step": 16578 + }, + { + "epoch": 0.4, + "learning_rate": 1.3678290614422331e-05, + "loss": 1.0217, + "step": 16579 + }, + { + "epoch": 0.4, + "learning_rate": 1.3677566314488664e-05, + "loss": 1.1955, + "step": 16580 + }, + { + "epoch": 0.4, + "learning_rate": 1.3676841992244066e-05, + "loss": 1.0421, + "step": 16581 + }, + { + "epoch": 0.4, + "learning_rate": 1.3676117647692934e-05, + "loss": 1.1811, + "step": 16582 + }, + { + "epoch": 0.4, + "learning_rate": 1.3675393280839661e-05, + "loss": 1.1218, + "step": 16583 + }, + { + "epoch": 0.4, + "learning_rate": 1.3674668891688638e-05, + "loss": 1.1378, + "step": 16584 + }, + { + "epoch": 0.4, + "learning_rate": 1.3673944480244266e-05, + "loss": 1.295, + "step": 16585 + }, + { + "epoch": 0.4, + "learning_rate": 1.3673220046510938e-05, + "loss": 0.9191, + "step": 16586 + }, + { + "epoch": 0.4, + "learning_rate": 1.3672495590493047e-05, + "loss": 1.1264, + "step": 16587 + }, + { + "epoch": 0.4, + "learning_rate": 1.3671771112194986e-05, + "loss": 1.1602, + "step": 16588 + }, + { + "epoch": 0.4, + "learning_rate": 1.3671046611621157e-05, + "loss": 1.1509, + "step": 16589 + }, + { + "epoch": 0.4, + "learning_rate": 1.3670322088775952e-05, + "loss": 1.0432, + "step": 16590 + }, + { + "epoch": 0.4, + "learning_rate": 1.3669597543663764e-05, + "loss": 1.0916, + "step": 16591 + }, + { + "epoch": 0.4, + "learning_rate": 1.3668872976288991e-05, + "loss": 0.968, + "step": 16592 + }, + { + "epoch": 0.4, + "learning_rate": 1.3668148386656031e-05, + "loss": 1.3056, + "step": 16593 + }, + { + "epoch": 0.4, + "learning_rate": 1.3667423774769275e-05, + "loss": 1.1686, + "step": 16594 + }, + { + "epoch": 0.4, + "learning_rate": 1.3666699140633122e-05, + "loss": 1.1375, + "step": 16595 + }, + { + "epoch": 0.4, + "learning_rate": 1.3665974484251966e-05, + "loss": 0.9919, + "step": 16596 + }, + { + "epoch": 0.4, + "learning_rate": 1.3665249805630209e-05, + "loss": 1.0082, + "step": 16597 + }, + { + "epoch": 0.4, + "learning_rate": 1.3664525104772244e-05, + "loss": 0.9832, + "step": 16598 + }, + { + "epoch": 0.4, + "learning_rate": 1.3663800381682465e-05, + "loss": 1.4304, + "step": 16599 + }, + { + "epoch": 0.4, + "learning_rate": 1.3663075636365268e-05, + "loss": 1.1751, + "step": 16600 + }, + { + "epoch": 0.4, + "learning_rate": 1.3662350868825057e-05, + "loss": 1.1033, + "step": 16601 + }, + { + "epoch": 0.4, + "learning_rate": 1.3661626079066224e-05, + "loss": 1.1605, + "step": 16602 + }, + { + "epoch": 0.4, + "learning_rate": 1.3660901267093167e-05, + "loss": 1.1198, + "step": 16603 + }, + { + "epoch": 0.4, + "learning_rate": 1.3660176432910279e-05, + "loss": 1.0334, + "step": 16604 + }, + { + "epoch": 0.4, + "learning_rate": 1.3659451576521967e-05, + "loss": 0.8785, + "step": 16605 + }, + { + "epoch": 0.4, + "learning_rate": 1.365872669793262e-05, + "loss": 1.1202, + "step": 16606 + }, + { + "epoch": 0.4, + "learning_rate": 1.365800179714664e-05, + "loss": 1.1358, + "step": 16607 + }, + { + "epoch": 0.4, + "learning_rate": 1.3657276874168422e-05, + "loss": 1.1077, + "step": 16608 + }, + { + "epoch": 0.4, + "learning_rate": 1.3656551929002367e-05, + "loss": 1.0744, + "step": 16609 + }, + { + "epoch": 0.4, + "learning_rate": 1.3655826961652871e-05, + "loss": 1.3876, + "step": 16610 + }, + { + "epoch": 0.4, + "learning_rate": 1.3655101972124333e-05, + "loss": 1.0942, + "step": 16611 + }, + { + "epoch": 0.4, + "learning_rate": 1.365437696042115e-05, + "loss": 1.1215, + "step": 16612 + }, + { + "epoch": 0.4, + "learning_rate": 1.3653651926547722e-05, + "loss": 1.3177, + "step": 16613 + }, + { + "epoch": 0.4, + "learning_rate": 1.3652926870508449e-05, + "loss": 1.0878, + "step": 16614 + }, + { + "epoch": 0.4, + "learning_rate": 1.3652201792307726e-05, + "loss": 1.2496, + "step": 16615 + }, + { + "epoch": 0.4, + "learning_rate": 1.3651476691949955e-05, + "loss": 0.8887, + "step": 16616 + }, + { + "epoch": 0.4, + "learning_rate": 1.3650751569439531e-05, + "loss": 1.2008, + "step": 16617 + }, + { + "epoch": 0.4, + "learning_rate": 1.365002642478086e-05, + "loss": 1.1354, + "step": 16618 + }, + { + "epoch": 0.4, + "learning_rate": 1.3649301257978334e-05, + "loss": 1.0848, + "step": 16619 + }, + { + "epoch": 0.4, + "learning_rate": 1.3648576069036356e-05, + "loss": 1.1257, + "step": 16620 + }, + { + "epoch": 0.4, + "learning_rate": 1.3647850857959327e-05, + "loss": 1.2092, + "step": 16621 + }, + { + "epoch": 0.4, + "learning_rate": 1.3647125624751643e-05, + "loss": 1.2411, + "step": 16622 + }, + { + "epoch": 0.4, + "learning_rate": 1.3646400369417704e-05, + "loss": 0.9169, + "step": 16623 + }, + { + "epoch": 0.4, + "learning_rate": 1.3645675091961914e-05, + "loss": 1.0093, + "step": 16624 + }, + { + "epoch": 0.4, + "learning_rate": 1.3644949792388671e-05, + "loss": 1.0429, + "step": 16625 + }, + { + "epoch": 0.4, + "learning_rate": 1.3644224470702376e-05, + "loss": 1.1065, + "step": 16626 + }, + { + "epoch": 0.4, + "learning_rate": 1.3643499126907428e-05, + "loss": 1.1406, + "step": 16627 + }, + { + "epoch": 0.4, + "learning_rate": 1.3642773761008226e-05, + "loss": 1.2608, + "step": 16628 + }, + { + "epoch": 0.4, + "learning_rate": 1.3642048373009172e-05, + "loss": 1.5286, + "step": 16629 + }, + { + "epoch": 0.4, + "learning_rate": 1.3641322962914669e-05, + "loss": 1.0524, + "step": 16630 + }, + { + "epoch": 0.4, + "learning_rate": 1.3640597530729116e-05, + "loss": 1.0728, + "step": 16631 + }, + { + "epoch": 0.4, + "learning_rate": 1.3639872076456912e-05, + "loss": 0.818, + "step": 16632 + }, + { + "epoch": 0.4, + "learning_rate": 1.3639146600102464e-05, + "loss": 1.3268, + "step": 16633 + }, + { + "epoch": 0.4, + "learning_rate": 1.3638421101670166e-05, + "loss": 0.887, + "step": 16634 + }, + { + "epoch": 0.4, + "learning_rate": 1.3637695581164426e-05, + "loss": 1.003, + "step": 16635 + }, + { + "epoch": 0.4, + "learning_rate": 1.3636970038589642e-05, + "loss": 1.0367, + "step": 16636 + }, + { + "epoch": 0.4, + "learning_rate": 1.3636244473950213e-05, + "loss": 1.2358, + "step": 16637 + }, + { + "epoch": 0.4, + "learning_rate": 1.3635518887250547e-05, + "loss": 1.0863, + "step": 16638 + }, + { + "epoch": 0.4, + "learning_rate": 1.3634793278495044e-05, + "loss": 1.13, + "step": 16639 + }, + { + "epoch": 0.4, + "learning_rate": 1.36340676476881e-05, + "loss": 1.2838, + "step": 16640 + }, + { + "epoch": 0.4, + "learning_rate": 1.3633341994834128e-05, + "loss": 1.3961, + "step": 16641 + }, + { + "epoch": 0.4, + "learning_rate": 1.3632616319937522e-05, + "loss": 1.0137, + "step": 16642 + }, + { + "epoch": 0.4, + "learning_rate": 1.3631890623002689e-05, + "loss": 0.8927, + "step": 16643 + }, + { + "epoch": 0.4, + "learning_rate": 1.3631164904034027e-05, + "loss": 1.1773, + "step": 16644 + }, + { + "epoch": 0.4, + "learning_rate": 1.3630439163035944e-05, + "loss": 0.9524, + "step": 16645 + }, + { + "epoch": 0.4, + "learning_rate": 1.362971340001284e-05, + "loss": 1.3605, + "step": 16646 + }, + { + "epoch": 0.4, + "learning_rate": 1.362898761496912e-05, + "loss": 0.8275, + "step": 16647 + }, + { + "epoch": 0.4, + "learning_rate": 1.3628261807909183e-05, + "loss": 1.0949, + "step": 16648 + }, + { + "epoch": 0.4, + "learning_rate": 1.3627535978837435e-05, + "loss": 1.1929, + "step": 16649 + }, + { + "epoch": 0.4, + "learning_rate": 1.3626810127758283e-05, + "loss": 1.0393, + "step": 16650 + }, + { + "epoch": 0.4, + "learning_rate": 1.3626084254676123e-05, + "loss": 1.1807, + "step": 16651 + }, + { + "epoch": 0.4, + "learning_rate": 1.3625358359595364e-05, + "loss": 1.1613, + "step": 16652 + }, + { + "epoch": 0.4, + "learning_rate": 1.3624632442520409e-05, + "loss": 1.1971, + "step": 16653 + }, + { + "epoch": 0.4, + "learning_rate": 1.3623906503455664e-05, + "loss": 1.2771, + "step": 16654 + }, + { + "epoch": 0.4, + "learning_rate": 1.3623180542405527e-05, + "loss": 0.9404, + "step": 16655 + }, + { + "epoch": 0.4, + "learning_rate": 1.3622454559374405e-05, + "loss": 1.2879, + "step": 16656 + }, + { + "epoch": 0.4, + "learning_rate": 1.3621728554366706e-05, + "loss": 1.157, + "step": 16657 + }, + { + "epoch": 0.4, + "learning_rate": 1.3621002527386832e-05, + "loss": 1.2713, + "step": 16658 + }, + { + "epoch": 0.4, + "learning_rate": 1.3620276478439188e-05, + "loss": 1.1441, + "step": 16659 + }, + { + "epoch": 0.4, + "learning_rate": 1.3619550407528175e-05, + "loss": 1.2523, + "step": 16660 + }, + { + "epoch": 0.4, + "learning_rate": 1.3618824314658201e-05, + "loss": 1.1526, + "step": 16661 + }, + { + "epoch": 0.4, + "learning_rate": 1.3618098199833675e-05, + "loss": 1.3174, + "step": 16662 + }, + { + "epoch": 0.4, + "learning_rate": 1.3617372063058994e-05, + "loss": 1.0617, + "step": 16663 + }, + { + "epoch": 0.4, + "learning_rate": 1.361664590433857e-05, + "loss": 1.4219, + "step": 16664 + }, + { + "epoch": 0.4, + "learning_rate": 1.3615919723676805e-05, + "loss": 1.2497, + "step": 16665 + }, + { + "epoch": 0.4, + "learning_rate": 1.3615193521078107e-05, + "loss": 1.0025, + "step": 16666 + }, + { + "epoch": 0.4, + "learning_rate": 1.3614467296546878e-05, + "loss": 1.1506, + "step": 16667 + }, + { + "epoch": 0.4, + "learning_rate": 1.3613741050087528e-05, + "loss": 1.1423, + "step": 16668 + }, + { + "epoch": 0.4, + "learning_rate": 1.361301478170446e-05, + "loss": 1.2372, + "step": 16669 + }, + { + "epoch": 0.4, + "learning_rate": 1.3612288491402084e-05, + "loss": 1.0733, + "step": 16670 + }, + { + "epoch": 0.4, + "learning_rate": 1.3611562179184798e-05, + "loss": 1.2557, + "step": 16671 + }, + { + "epoch": 0.4, + "learning_rate": 1.3610835845057019e-05, + "loss": 1.2061, + "step": 16672 + }, + { + "epoch": 0.4, + "learning_rate": 1.3610109489023144e-05, + "loss": 0.9736, + "step": 16673 + }, + { + "epoch": 0.4, + "learning_rate": 1.3609383111087589e-05, + "loss": 1.033, + "step": 16674 + }, + { + "epoch": 0.4, + "learning_rate": 1.3608656711254749e-05, + "loss": 1.2112, + "step": 16675 + }, + { + "epoch": 0.4, + "learning_rate": 1.3607930289529044e-05, + "loss": 0.9666, + "step": 16676 + }, + { + "epoch": 0.4, + "learning_rate": 1.360720384591487e-05, + "loss": 1.1999, + "step": 16677 + }, + { + "epoch": 0.4, + "learning_rate": 1.360647738041664e-05, + "loss": 0.7887, + "step": 16678 + }, + { + "epoch": 0.4, + "learning_rate": 1.360575089303876e-05, + "loss": 0.9736, + "step": 16679 + }, + { + "epoch": 0.4, + "learning_rate": 1.360502438378564e-05, + "loss": 1.1857, + "step": 16680 + }, + { + "epoch": 0.4, + "learning_rate": 1.360429785266168e-05, + "loss": 1.0871, + "step": 16681 + }, + { + "epoch": 0.4, + "learning_rate": 1.3603571299671297e-05, + "loss": 1.1834, + "step": 16682 + }, + { + "epoch": 0.4, + "learning_rate": 1.3602844724818893e-05, + "loss": 0.838, + "step": 16683 + }, + { + "epoch": 0.4, + "learning_rate": 1.3602118128108877e-05, + "loss": 1.1115, + "step": 16684 + }, + { + "epoch": 0.4, + "learning_rate": 1.360139150954566e-05, + "loss": 0.9248, + "step": 16685 + }, + { + "epoch": 0.4, + "learning_rate": 1.3600664869133646e-05, + "loss": 0.8501, + "step": 16686 + }, + { + "epoch": 0.4, + "learning_rate": 1.3599938206877244e-05, + "loss": 1.3581, + "step": 16687 + }, + { + "epoch": 0.4, + "learning_rate": 1.3599211522780865e-05, + "loss": 1.0032, + "step": 16688 + }, + { + "epoch": 0.4, + "learning_rate": 1.3598484816848916e-05, + "loss": 1.039, + "step": 16689 + }, + { + "epoch": 0.4, + "learning_rate": 1.3597758089085807e-05, + "loss": 0.9085, + "step": 16690 + }, + { + "epoch": 0.4, + "learning_rate": 1.3597031339495944e-05, + "loss": 1.2832, + "step": 16691 + }, + { + "epoch": 0.4, + "learning_rate": 1.3596304568083739e-05, + "loss": 1.0554, + "step": 16692 + }, + { + "epoch": 0.4, + "learning_rate": 1.3595577774853602e-05, + "loss": 1.0974, + "step": 16693 + }, + { + "epoch": 0.4, + "learning_rate": 1.3594850959809937e-05, + "loss": 1.0622, + "step": 16694 + }, + { + "epoch": 0.4, + "learning_rate": 1.3594124122957157e-05, + "loss": 0.7621, + "step": 16695 + }, + { + "epoch": 0.4, + "learning_rate": 1.3593397264299673e-05, + "loss": 1.1283, + "step": 16696 + }, + { + "epoch": 0.4, + "learning_rate": 1.3592670383841893e-05, + "loss": 1.086, + "step": 16697 + }, + { + "epoch": 0.4, + "learning_rate": 1.3591943481588228e-05, + "loss": 1.2778, + "step": 16698 + }, + { + "epoch": 0.4, + "learning_rate": 1.3591216557543083e-05, + "loss": 1.2448, + "step": 16699 + }, + { + "epoch": 0.4, + "learning_rate": 1.3590489611710874e-05, + "loss": 1.0531, + "step": 16700 + }, + { + "epoch": 0.4, + "learning_rate": 1.3589762644096009e-05, + "loss": 1.2199, + "step": 16701 + }, + { + "epoch": 0.4, + "learning_rate": 1.35890356547029e-05, + "loss": 0.9619, + "step": 16702 + }, + { + "epoch": 0.4, + "learning_rate": 1.3588308643535952e-05, + "loss": 1.0835, + "step": 16703 + }, + { + "epoch": 0.4, + "learning_rate": 1.3587581610599583e-05, + "loss": 1.2246, + "step": 16704 + }, + { + "epoch": 0.4, + "learning_rate": 1.3586854555898199e-05, + "loss": 0.7466, + "step": 16705 + }, + { + "epoch": 0.4, + "learning_rate": 1.3586127479436212e-05, + "loss": 1.0053, + "step": 16706 + }, + { + "epoch": 0.4, + "learning_rate": 1.358540038121803e-05, + "loss": 1.0281, + "step": 16707 + }, + { + "epoch": 0.4, + "learning_rate": 1.3584673261248071e-05, + "loss": 1.1584, + "step": 16708 + }, + { + "epoch": 0.4, + "learning_rate": 1.3583946119530741e-05, + "loss": 1.3721, + "step": 16709 + }, + { + "epoch": 0.4, + "learning_rate": 1.3583218956070457e-05, + "loss": 0.9443, + "step": 16710 + }, + { + "epoch": 0.4, + "learning_rate": 1.3582491770871622e-05, + "loss": 0.9304, + "step": 16711 + }, + { + "epoch": 0.4, + "learning_rate": 1.3581764563938654e-05, + "loss": 1.2581, + "step": 16712 + }, + { + "epoch": 0.4, + "learning_rate": 1.3581037335275961e-05, + "loss": 1.2821, + "step": 16713 + }, + { + "epoch": 0.4, + "learning_rate": 1.358031008488796e-05, + "loss": 1.3436, + "step": 16714 + }, + { + "epoch": 0.4, + "learning_rate": 1.3579582812779055e-05, + "loss": 1.1137, + "step": 16715 + }, + { + "epoch": 0.4, + "learning_rate": 1.3578855518953666e-05, + "loss": 1.1236, + "step": 16716 + }, + { + "epoch": 0.4, + "learning_rate": 1.3578128203416204e-05, + "loss": 1.2041, + "step": 16717 + }, + { + "epoch": 0.4, + "learning_rate": 1.357740086617108e-05, + "loss": 1.2707, + "step": 16718 + }, + { + "epoch": 0.4, + "learning_rate": 1.3576673507222702e-05, + "loss": 0.906, + "step": 16719 + }, + { + "epoch": 0.4, + "learning_rate": 1.357594612657549e-05, + "loss": 0.9996, + "step": 16720 + }, + { + "epoch": 0.4, + "learning_rate": 1.3575218724233854e-05, + "loss": 1.2512, + "step": 16721 + }, + { + "epoch": 0.4, + "learning_rate": 1.3574491300202206e-05, + "loss": 1.3195, + "step": 16722 + }, + { + "epoch": 0.4, + "learning_rate": 1.3573763854484961e-05, + "loss": 1.043, + "step": 16723 + }, + { + "epoch": 0.4, + "learning_rate": 1.357303638708653e-05, + "loss": 0.8529, + "step": 16724 + }, + { + "epoch": 0.4, + "learning_rate": 1.3572308898011328e-05, + "loss": 1.0529, + "step": 16725 + }, + { + "epoch": 0.4, + "learning_rate": 1.357158138726377e-05, + "loss": 0.8664, + "step": 16726 + }, + { + "epoch": 0.4, + "learning_rate": 1.3570853854848265e-05, + "loss": 1.1797, + "step": 16727 + }, + { + "epoch": 0.4, + "learning_rate": 1.3570126300769233e-05, + "loss": 0.9112, + "step": 16728 + }, + { + "epoch": 0.4, + "learning_rate": 1.3569398725031082e-05, + "loss": 1.1801, + "step": 16729 + }, + { + "epoch": 0.4, + "learning_rate": 1.3568671127638231e-05, + "loss": 1.3748, + "step": 16730 + }, + { + "epoch": 0.4, + "learning_rate": 1.3567943508595087e-05, + "loss": 0.9638, + "step": 16731 + }, + { + "epoch": 0.4, + "learning_rate": 1.3567215867906072e-05, + "loss": 0.9914, + "step": 16732 + }, + { + "epoch": 0.4, + "learning_rate": 1.3566488205575596e-05, + "loss": 1.2135, + "step": 16733 + }, + { + "epoch": 0.4, + "learning_rate": 1.3565760521608078e-05, + "loss": 1.1177, + "step": 16734 + }, + { + "epoch": 0.4, + "learning_rate": 1.3565032816007927e-05, + "loss": 1.1071, + "step": 16735 + }, + { + "epoch": 0.4, + "learning_rate": 1.3564305088779561e-05, + "loss": 1.2441, + "step": 16736 + }, + { + "epoch": 0.4, + "learning_rate": 1.3563577339927397e-05, + "loss": 0.8542, + "step": 16737 + }, + { + "epoch": 0.4, + "learning_rate": 1.3562849569455843e-05, + "loss": 1.0916, + "step": 16738 + }, + { + "epoch": 0.4, + "learning_rate": 1.356212177736932e-05, + "loss": 0.9783, + "step": 16739 + }, + { + "epoch": 0.4, + "learning_rate": 1.3561393963672243e-05, + "loss": 1.3066, + "step": 16740 + }, + { + "epoch": 0.4, + "learning_rate": 1.3560666128369026e-05, + "loss": 1.2639, + "step": 16741 + }, + { + "epoch": 0.4, + "learning_rate": 1.3559938271464086e-05, + "loss": 1.2182, + "step": 16742 + }, + { + "epoch": 0.4, + "learning_rate": 1.3559210392961837e-05, + "loss": 1.1627, + "step": 16743 + }, + { + "epoch": 0.4, + "learning_rate": 1.3558482492866696e-05, + "loss": 1.3477, + "step": 16744 + }, + { + "epoch": 0.4, + "learning_rate": 1.355775457118308e-05, + "loss": 0.9525, + "step": 16745 + }, + { + "epoch": 0.4, + "learning_rate": 1.3557026627915402e-05, + "loss": 1.2854, + "step": 16746 + }, + { + "epoch": 0.4, + "learning_rate": 1.355629866306808e-05, + "loss": 0.9275, + "step": 16747 + }, + { + "epoch": 0.4, + "learning_rate": 1.3555570676645532e-05, + "loss": 1.0215, + "step": 16748 + }, + { + "epoch": 0.4, + "learning_rate": 1.3554842668652173e-05, + "loss": 1.0053, + "step": 16749 + }, + { + "epoch": 0.4, + "learning_rate": 1.3554114639092414e-05, + "loss": 1.1706, + "step": 16750 + }, + { + "epoch": 0.4, + "learning_rate": 1.3553386587970685e-05, + "loss": 1.2745, + "step": 16751 + }, + { + "epoch": 0.4, + "learning_rate": 1.3552658515291392e-05, + "loss": 1.0187, + "step": 16752 + }, + { + "epoch": 0.4, + "learning_rate": 1.3551930421058959e-05, + "loss": 1.1556, + "step": 16753 + }, + { + "epoch": 0.4, + "learning_rate": 1.3551202305277795e-05, + "loss": 1.028, + "step": 16754 + }, + { + "epoch": 0.4, + "learning_rate": 1.3550474167952323e-05, + "loss": 1.0665, + "step": 16755 + }, + { + "epoch": 0.4, + "learning_rate": 1.3549746009086962e-05, + "loss": 1.3761, + "step": 16756 + }, + { + "epoch": 0.4, + "learning_rate": 1.3549017828686124e-05, + "loss": 1.2532, + "step": 16757 + }, + { + "epoch": 0.4, + "learning_rate": 1.354828962675423e-05, + "loss": 1.0793, + "step": 16758 + }, + { + "epoch": 0.4, + "learning_rate": 1.3547561403295697e-05, + "loss": 0.9576, + "step": 16759 + }, + { + "epoch": 0.4, + "learning_rate": 1.3546833158314944e-05, + "loss": 1.1038, + "step": 16760 + }, + { + "epoch": 0.4, + "learning_rate": 1.3546104891816389e-05, + "loss": 0.9886, + "step": 16761 + }, + { + "epoch": 0.4, + "learning_rate": 1.354537660380445e-05, + "loss": 1.2575, + "step": 16762 + }, + { + "epoch": 0.4, + "learning_rate": 1.3544648294283544e-05, + "loss": 1.0778, + "step": 16763 + }, + { + "epoch": 0.4, + "learning_rate": 1.3543919963258092e-05, + "loss": 1.1454, + "step": 16764 + }, + { + "epoch": 0.4, + "learning_rate": 1.3543191610732512e-05, + "loss": 1.0126, + "step": 16765 + }, + { + "epoch": 0.4, + "learning_rate": 1.354246323671122e-05, + "loss": 1.158, + "step": 16766 + }, + { + "epoch": 0.4, + "learning_rate": 1.3541734841198638e-05, + "loss": 1.0344, + "step": 16767 + }, + { + "epoch": 0.4, + "learning_rate": 1.3541006424199182e-05, + "loss": 1.1799, + "step": 16768 + }, + { + "epoch": 0.4, + "learning_rate": 1.3540277985717274e-05, + "loss": 1.0265, + "step": 16769 + }, + { + "epoch": 0.4, + "learning_rate": 1.353954952575733e-05, + "loss": 1.0347, + "step": 16770 + }, + { + "epoch": 0.4, + "learning_rate": 1.3538821044323773e-05, + "loss": 1.0999, + "step": 16771 + }, + { + "epoch": 0.4, + "learning_rate": 1.3538092541421022e-05, + "loss": 1.2163, + "step": 16772 + }, + { + "epoch": 0.4, + "learning_rate": 1.3537364017053496e-05, + "loss": 1.1664, + "step": 16773 + }, + { + "epoch": 0.4, + "learning_rate": 1.3536635471225612e-05, + "loss": 0.9231, + "step": 16774 + }, + { + "epoch": 0.4, + "learning_rate": 1.3535906903941796e-05, + "loss": 0.9674, + "step": 16775 + }, + { + "epoch": 0.4, + "learning_rate": 1.3535178315206462e-05, + "loss": 0.9753, + "step": 16776 + }, + { + "epoch": 0.4, + "learning_rate": 1.3534449705024036e-05, + "loss": 1.2349, + "step": 16777 + }, + { + "epoch": 0.4, + "learning_rate": 1.3533721073398932e-05, + "loss": 1.2172, + "step": 16778 + }, + { + "epoch": 0.4, + "learning_rate": 1.3532992420335574e-05, + "loss": 0.9411, + "step": 16779 + }, + { + "epoch": 0.4, + "learning_rate": 1.3532263745838381e-05, + "loss": 0.9095, + "step": 16780 + }, + { + "epoch": 0.4, + "learning_rate": 1.3531535049911777e-05, + "loss": 1.0253, + "step": 16781 + }, + { + "epoch": 0.4, + "learning_rate": 1.3530806332560177e-05, + "loss": 1.129, + "step": 16782 + }, + { + "epoch": 0.4, + "learning_rate": 1.3530077593788009e-05, + "loss": 1.236, + "step": 16783 + }, + { + "epoch": 0.4, + "learning_rate": 1.352934883359969e-05, + "loss": 1.0375, + "step": 16784 + }, + { + "epoch": 0.4, + "learning_rate": 1.3528620051999642e-05, + "loss": 1.1223, + "step": 16785 + }, + { + "epoch": 0.4, + "learning_rate": 1.3527891248992284e-05, + "loss": 1.2525, + "step": 16786 + }, + { + "epoch": 0.4, + "learning_rate": 1.3527162424582039e-05, + "loss": 1.1073, + "step": 16787 + }, + { + "epoch": 0.4, + "learning_rate": 1.3526433578773332e-05, + "loss": 1.1275, + "step": 16788 + }, + { + "epoch": 0.4, + "learning_rate": 1.3525704711570582e-05, + "loss": 0.9993, + "step": 16789 + }, + { + "epoch": 0.4, + "learning_rate": 1.3524975822978206e-05, + "loss": 1.1435, + "step": 16790 + }, + { + "epoch": 0.4, + "learning_rate": 1.3524246913000636e-05, + "loss": 0.9571, + "step": 16791 + }, + { + "epoch": 0.4, + "learning_rate": 1.3523517981642287e-05, + "loss": 0.6133, + "step": 16792 + }, + { + "epoch": 0.4, + "learning_rate": 1.3522789028907583e-05, + "loss": 1.1404, + "step": 16793 + }, + { + "epoch": 0.4, + "learning_rate": 1.3522060054800947e-05, + "loss": 1.0, + "step": 16794 + }, + { + "epoch": 0.4, + "learning_rate": 1.3521331059326803e-05, + "loss": 1.2868, + "step": 16795 + }, + { + "epoch": 0.4, + "learning_rate": 1.352060204248957e-05, + "loss": 1.0911, + "step": 16796 + }, + { + "epoch": 0.4, + "learning_rate": 1.3519873004293673e-05, + "loss": 1.0775, + "step": 16797 + }, + { + "epoch": 0.4, + "learning_rate": 1.351914394474353e-05, + "loss": 1.2443, + "step": 16798 + }, + { + "epoch": 0.4, + "learning_rate": 1.3518414863843574e-05, + "loss": 1.3371, + "step": 16799 + }, + { + "epoch": 0.4, + "learning_rate": 1.3517685761598223e-05, + "loss": 0.979, + "step": 16800 + }, + { + "epoch": 0.4, + "learning_rate": 1.3516956638011898e-05, + "loss": 1.1109, + "step": 16801 + }, + { + "epoch": 0.4, + "learning_rate": 1.3516227493089022e-05, + "loss": 1.1525, + "step": 16802 + }, + { + "epoch": 0.4, + "learning_rate": 1.3515498326834025e-05, + "loss": 0.9498, + "step": 16803 + }, + { + "epoch": 0.4, + "learning_rate": 1.3514769139251325e-05, + "loss": 1.1821, + "step": 16804 + }, + { + "epoch": 0.4, + "learning_rate": 1.3514039930345346e-05, + "loss": 0.8999, + "step": 16805 + }, + { + "epoch": 0.4, + "learning_rate": 1.3513310700120514e-05, + "loss": 0.9535, + "step": 16806 + }, + { + "epoch": 0.4, + "learning_rate": 1.3512581448581252e-05, + "loss": 1.1151, + "step": 16807 + }, + { + "epoch": 0.4, + "learning_rate": 1.3511852175731988e-05, + "loss": 1.0256, + "step": 16808 + }, + { + "epoch": 0.4, + "learning_rate": 1.3511122881577138e-05, + "loss": 1.3074, + "step": 16809 + }, + { + "epoch": 0.4, + "learning_rate": 1.3510393566121134e-05, + "loss": 1.3898, + "step": 16810 + }, + { + "epoch": 0.4, + "learning_rate": 1.3509664229368396e-05, + "loss": 1.3531, + "step": 16811 + }, + { + "epoch": 0.4, + "learning_rate": 1.3508934871323353e-05, + "loss": 1.18, + "step": 16812 + }, + { + "epoch": 0.4, + "learning_rate": 1.3508205491990427e-05, + "loss": 1.0842, + "step": 16813 + }, + { + "epoch": 0.4, + "learning_rate": 1.3507476091374042e-05, + "loss": 1.1424, + "step": 16814 + }, + { + "epoch": 0.4, + "learning_rate": 1.3506746669478626e-05, + "loss": 0.9222, + "step": 16815 + }, + { + "epoch": 0.4, + "learning_rate": 1.3506017226308603e-05, + "loss": 1.3122, + "step": 16816 + }, + { + "epoch": 0.4, + "learning_rate": 1.3505287761868398e-05, + "loss": 0.9093, + "step": 16817 + }, + { + "epoch": 0.4, + "learning_rate": 1.3504558276162434e-05, + "loss": 0.9989, + "step": 16818 + }, + { + "epoch": 0.4, + "learning_rate": 1.3503828769195142e-05, + "loss": 1.0788, + "step": 16819 + }, + { + "epoch": 0.4, + "learning_rate": 1.3503099240970945e-05, + "loss": 0.8928, + "step": 16820 + }, + { + "epoch": 0.4, + "learning_rate": 1.3502369691494267e-05, + "loss": 0.9695, + "step": 16821 + }, + { + "epoch": 0.4, + "learning_rate": 1.3501640120769537e-05, + "loss": 1.2183, + "step": 16822 + }, + { + "epoch": 0.4, + "learning_rate": 1.350091052880118e-05, + "loss": 0.9849, + "step": 16823 + }, + { + "epoch": 0.4, + "learning_rate": 1.3500180915593624e-05, + "loss": 0.9582, + "step": 16824 + }, + { + "epoch": 0.4, + "learning_rate": 1.3499451281151292e-05, + "loss": 1.0671, + "step": 16825 + }, + { + "epoch": 0.4, + "learning_rate": 1.3498721625478611e-05, + "loss": 1.0967, + "step": 16826 + }, + { + "epoch": 0.4, + "learning_rate": 1.3497991948580011e-05, + "loss": 0.8804, + "step": 16827 + }, + { + "epoch": 0.4, + "learning_rate": 1.3497262250459917e-05, + "loss": 1.1712, + "step": 16828 + }, + { + "epoch": 0.4, + "learning_rate": 1.3496532531122754e-05, + "loss": 1.2885, + "step": 16829 + }, + { + "epoch": 0.4, + "learning_rate": 1.3495802790572949e-05, + "loss": 0.9593, + "step": 16830 + }, + { + "epoch": 0.4, + "learning_rate": 1.3495073028814933e-05, + "loss": 1.0393, + "step": 16831 + }, + { + "epoch": 0.4, + "learning_rate": 1.3494343245853132e-05, + "loss": 1.1823, + "step": 16832 + }, + { + "epoch": 0.4, + "learning_rate": 1.3493613441691972e-05, + "loss": 1.0243, + "step": 16833 + }, + { + "epoch": 0.4, + "learning_rate": 1.349288361633588e-05, + "loss": 1.2629, + "step": 16834 + }, + { + "epoch": 0.4, + "learning_rate": 1.3492153769789283e-05, + "loss": 1.0819, + "step": 16835 + }, + { + "epoch": 0.4, + "learning_rate": 1.3491423902056613e-05, + "loss": 1.1146, + "step": 16836 + }, + { + "epoch": 0.4, + "learning_rate": 1.3490694013142294e-05, + "loss": 1.0273, + "step": 16837 + }, + { + "epoch": 0.4, + "learning_rate": 1.3489964103050756e-05, + "loss": 1.0497, + "step": 16838 + }, + { + "epoch": 0.4, + "learning_rate": 1.3489234171786428e-05, + "loss": 0.9588, + "step": 16839 + }, + { + "epoch": 0.4, + "learning_rate": 1.3488504219353736e-05, + "loss": 1.1404, + "step": 16840 + }, + { + "epoch": 0.41, + "learning_rate": 1.3487774245757108e-05, + "loss": 1.0567, + "step": 16841 + }, + { + "epoch": 0.41, + "learning_rate": 1.3487044251000975e-05, + "loss": 1.0126, + "step": 16842 + }, + { + "epoch": 0.41, + "learning_rate": 1.3486314235089766e-05, + "loss": 1.1687, + "step": 16843 + }, + { + "epoch": 0.41, + "learning_rate": 1.3485584198027909e-05, + "loss": 1.1078, + "step": 16844 + }, + { + "epoch": 0.41, + "learning_rate": 1.348485413981983e-05, + "loss": 0.9891, + "step": 16845 + }, + { + "epoch": 0.41, + "learning_rate": 1.348412406046996e-05, + "loss": 1.1662, + "step": 16846 + }, + { + "epoch": 0.41, + "learning_rate": 1.3483393959982732e-05, + "loss": 1.0909, + "step": 16847 + }, + { + "epoch": 0.41, + "learning_rate": 1.3482663838362573e-05, + "loss": 1.0897, + "step": 16848 + }, + { + "epoch": 0.41, + "learning_rate": 1.3481933695613906e-05, + "loss": 1.0306, + "step": 16849 + }, + { + "epoch": 0.41, + "learning_rate": 1.348120353174117e-05, + "loss": 1.0642, + "step": 16850 + }, + { + "epoch": 0.41, + "learning_rate": 1.3480473346748792e-05, + "loss": 0.7603, + "step": 16851 + }, + { + "epoch": 0.41, + "learning_rate": 1.34797431406412e-05, + "loss": 1.1151, + "step": 16852 + }, + { + "epoch": 0.41, + "learning_rate": 1.3479012913422823e-05, + "loss": 1.1343, + "step": 16853 + }, + { + "epoch": 0.41, + "learning_rate": 1.3478282665098095e-05, + "loss": 1.2649, + "step": 16854 + }, + { + "epoch": 0.41, + "learning_rate": 1.3477552395671444e-05, + "loss": 1.0298, + "step": 16855 + }, + { + "epoch": 0.41, + "learning_rate": 1.34768221051473e-05, + "loss": 0.9635, + "step": 16856 + }, + { + "epoch": 0.41, + "learning_rate": 1.3476091793530092e-05, + "loss": 1.4065, + "step": 16857 + }, + { + "epoch": 0.41, + "learning_rate": 1.3475361460824256e-05, + "loss": 1.3106, + "step": 16858 + }, + { + "epoch": 0.41, + "learning_rate": 1.347463110703422e-05, + "loss": 0.854, + "step": 16859 + }, + { + "epoch": 0.41, + "learning_rate": 1.3473900732164414e-05, + "loss": 1.3094, + "step": 16860 + }, + { + "epoch": 0.41, + "learning_rate": 1.3473170336219265e-05, + "loss": 0.8807, + "step": 16861 + }, + { + "epoch": 0.41, + "learning_rate": 1.3472439919203212e-05, + "loss": 1.2329, + "step": 16862 + }, + { + "epoch": 0.41, + "learning_rate": 1.3471709481120685e-05, + "loss": 0.9504, + "step": 16863 + }, + { + "epoch": 0.41, + "learning_rate": 1.3470979021976112e-05, + "loss": 1.1081, + "step": 16864 + }, + { + "epoch": 0.41, + "learning_rate": 1.3470248541773923e-05, + "loss": 1.01, + "step": 16865 + }, + { + "epoch": 0.41, + "learning_rate": 1.3469518040518552e-05, + "loss": 1.3098, + "step": 16866 + }, + { + "epoch": 0.41, + "learning_rate": 1.3468787518214435e-05, + "loss": 0.969, + "step": 16867 + }, + { + "epoch": 0.41, + "learning_rate": 1.3468056974865998e-05, + "loss": 1.2471, + "step": 16868 + }, + { + "epoch": 0.41, + "learning_rate": 1.3467326410477672e-05, + "loss": 1.0853, + "step": 16869 + }, + { + "epoch": 0.41, + "learning_rate": 1.3466595825053896e-05, + "loss": 0.911, + "step": 16870 + }, + { + "epoch": 0.41, + "learning_rate": 1.3465865218599099e-05, + "loss": 1.118, + "step": 16871 + }, + { + "epoch": 0.41, + "learning_rate": 1.3465134591117713e-05, + "loss": 1.0933, + "step": 16872 + }, + { + "epoch": 0.41, + "learning_rate": 1.3464403942614169e-05, + "loss": 1.0218, + "step": 16873 + }, + { + "epoch": 0.41, + "learning_rate": 1.34636732730929e-05, + "loss": 0.988, + "step": 16874 + }, + { + "epoch": 0.41, + "learning_rate": 1.346294258255834e-05, + "loss": 1.272, + "step": 16875 + }, + { + "epoch": 0.41, + "learning_rate": 1.3462211871014923e-05, + "loss": 1.1055, + "step": 16876 + }, + { + "epoch": 0.41, + "learning_rate": 1.346148113846708e-05, + "loss": 0.96, + "step": 16877 + }, + { + "epoch": 0.41, + "learning_rate": 1.3460750384919247e-05, + "loss": 1.0745, + "step": 16878 + }, + { + "epoch": 0.41, + "learning_rate": 1.3460019610375854e-05, + "loss": 1.0157, + "step": 16879 + }, + { + "epoch": 0.41, + "learning_rate": 1.3459288814841339e-05, + "loss": 0.9831, + "step": 16880 + }, + { + "epoch": 0.41, + "learning_rate": 1.3458557998320127e-05, + "loss": 0.9821, + "step": 16881 + }, + { + "epoch": 0.41, + "learning_rate": 1.3457827160816661e-05, + "loss": 1.0698, + "step": 16882 + }, + { + "epoch": 0.41, + "learning_rate": 1.345709630233537e-05, + "loss": 0.9994, + "step": 16883 + }, + { + "epoch": 0.41, + "learning_rate": 1.345636542288069e-05, + "loss": 1.1023, + "step": 16884 + }, + { + "epoch": 0.41, + "learning_rate": 1.345563452245705e-05, + "loss": 0.9538, + "step": 16885 + }, + { + "epoch": 0.41, + "learning_rate": 1.3454903601068891e-05, + "loss": 1.1961, + "step": 16886 + }, + { + "epoch": 0.41, + "learning_rate": 1.3454172658720648e-05, + "loss": 0.6671, + "step": 16887 + }, + { + "epoch": 0.41, + "learning_rate": 1.3453441695416746e-05, + "loss": 1.1744, + "step": 16888 + }, + { + "epoch": 0.41, + "learning_rate": 1.3452710711161627e-05, + "loss": 1.0258, + "step": 16889 + }, + { + "epoch": 0.41, + "learning_rate": 1.3451979705959729e-05, + "loss": 0.8704, + "step": 16890 + }, + { + "epoch": 0.41, + "learning_rate": 1.3451248679815476e-05, + "loss": 1.106, + "step": 16891 + }, + { + "epoch": 0.41, + "learning_rate": 1.3450517632733313e-05, + "loss": 1.006, + "step": 16892 + }, + { + "epoch": 0.41, + "learning_rate": 1.3449786564717668e-05, + "loss": 1.1878, + "step": 16893 + }, + { + "epoch": 0.41, + "learning_rate": 1.344905547577298e-05, + "loss": 1.0265, + "step": 16894 + }, + { + "epoch": 0.41, + "learning_rate": 1.3448324365903687e-05, + "loss": 1.3631, + "step": 16895 + }, + { + "epoch": 0.41, + "learning_rate": 1.3447593235114217e-05, + "loss": 1.3443, + "step": 16896 + }, + { + "epoch": 0.41, + "learning_rate": 1.344686208340901e-05, + "loss": 1.1636, + "step": 16897 + }, + { + "epoch": 0.41, + "learning_rate": 1.3446130910792506e-05, + "loss": 0.9411, + "step": 16898 + }, + { + "epoch": 0.41, + "learning_rate": 1.3445399717269132e-05, + "loss": 0.8868, + "step": 16899 + }, + { + "epoch": 0.41, + "learning_rate": 1.344466850284333e-05, + "loss": 1.305, + "step": 16900 + }, + { + "epoch": 0.41, + "learning_rate": 1.3443937267519534e-05, + "loss": 1.1843, + "step": 16901 + }, + { + "epoch": 0.41, + "learning_rate": 1.3443206011302181e-05, + "loss": 0.8565, + "step": 16902 + }, + { + "epoch": 0.41, + "learning_rate": 1.3442474734195707e-05, + "loss": 1.3433, + "step": 16903 + }, + { + "epoch": 0.41, + "learning_rate": 1.3441743436204549e-05, + "loss": 0.9695, + "step": 16904 + }, + { + "epoch": 0.41, + "learning_rate": 1.344101211733314e-05, + "loss": 1.0691, + "step": 16905 + }, + { + "epoch": 0.41, + "learning_rate": 1.3440280777585923e-05, + "loss": 1.413, + "step": 16906 + }, + { + "epoch": 0.41, + "learning_rate": 1.3439549416967333e-05, + "loss": 1.1817, + "step": 16907 + }, + { + "epoch": 0.41, + "learning_rate": 1.3438818035481802e-05, + "loss": 1.0186, + "step": 16908 + }, + { + "epoch": 0.41, + "learning_rate": 1.3438086633133773e-05, + "loss": 1.0444, + "step": 16909 + }, + { + "epoch": 0.41, + "learning_rate": 1.343735520992768e-05, + "loss": 1.1131, + "step": 16910 + }, + { + "epoch": 0.41, + "learning_rate": 1.3436623765867962e-05, + "loss": 0.9907, + "step": 16911 + }, + { + "epoch": 0.41, + "learning_rate": 1.3435892300959057e-05, + "loss": 0.8499, + "step": 16912 + }, + { + "epoch": 0.41, + "learning_rate": 1.34351608152054e-05, + "loss": 1.0144, + "step": 16913 + }, + { + "epoch": 0.41, + "learning_rate": 1.3434429308611432e-05, + "loss": 1.1003, + "step": 16914 + }, + { + "epoch": 0.41, + "learning_rate": 1.3433697781181587e-05, + "loss": 1.1752, + "step": 16915 + }, + { + "epoch": 0.41, + "learning_rate": 1.3432966232920308e-05, + "loss": 1.2445, + "step": 16916 + }, + { + "epoch": 0.41, + "learning_rate": 1.3432234663832026e-05, + "loss": 1.1195, + "step": 16917 + }, + { + "epoch": 0.41, + "learning_rate": 1.3431503073921189e-05, + "loss": 0.8893, + "step": 16918 + }, + { + "epoch": 0.41, + "learning_rate": 1.3430771463192226e-05, + "loss": 1.1603, + "step": 16919 + }, + { + "epoch": 0.41, + "learning_rate": 1.3430039831649579e-05, + "loss": 1.0962, + "step": 16920 + }, + { + "epoch": 0.41, + "learning_rate": 1.342930817929769e-05, + "loss": 1.2219, + "step": 16921 + }, + { + "epoch": 0.41, + "learning_rate": 1.3428576506140995e-05, + "loss": 0.8423, + "step": 16922 + }, + { + "epoch": 0.41, + "learning_rate": 1.3427844812183932e-05, + "loss": 0.8151, + "step": 16923 + }, + { + "epoch": 0.41, + "learning_rate": 1.342711309743094e-05, + "loss": 1.2663, + "step": 16924 + }, + { + "epoch": 0.41, + "learning_rate": 1.3426381361886458e-05, + "loss": 1.12, + "step": 16925 + }, + { + "epoch": 0.41, + "learning_rate": 1.342564960555493e-05, + "loss": 1.112, + "step": 16926 + }, + { + "epoch": 0.41, + "learning_rate": 1.3424917828440788e-05, + "loss": 1.2012, + "step": 16927 + }, + { + "epoch": 0.41, + "learning_rate": 1.3424186030548476e-05, + "loss": 1.3142, + "step": 16928 + }, + { + "epoch": 0.41, + "learning_rate": 1.3423454211882434e-05, + "loss": 1.2483, + "step": 16929 + }, + { + "epoch": 0.41, + "learning_rate": 1.3422722372447098e-05, + "loss": 1.0217, + "step": 16930 + }, + { + "epoch": 0.41, + "learning_rate": 1.3421990512246914e-05, + "loss": 1.2268, + "step": 16931 + }, + { + "epoch": 0.41, + "learning_rate": 1.3421258631286315e-05, + "loss": 1.2568, + "step": 16932 + }, + { + "epoch": 0.41, + "learning_rate": 1.3420526729569748e-05, + "loss": 1.1469, + "step": 16933 + }, + { + "epoch": 0.41, + "learning_rate": 1.3419794807101647e-05, + "loss": 1.0538, + "step": 16934 + }, + { + "epoch": 0.41, + "learning_rate": 1.3419062863886456e-05, + "loss": 1.1242, + "step": 16935 + }, + { + "epoch": 0.41, + "learning_rate": 1.3418330899928615e-05, + "loss": 0.8539, + "step": 16936 + }, + { + "epoch": 0.41, + "learning_rate": 1.3417598915232564e-05, + "loss": 1.2277, + "step": 16937 + }, + { + "epoch": 0.41, + "learning_rate": 1.3416866909802746e-05, + "loss": 1.2442, + "step": 16938 + }, + { + "epoch": 0.41, + "learning_rate": 1.3416134883643598e-05, + "loss": 1.0676, + "step": 16939 + }, + { + "epoch": 0.41, + "learning_rate": 1.3415402836759563e-05, + "loss": 1.0892, + "step": 16940 + }, + { + "epoch": 0.41, + "learning_rate": 1.3414670769155085e-05, + "loss": 1.011, + "step": 16941 + }, + { + "epoch": 0.41, + "learning_rate": 1.3413938680834601e-05, + "loss": 0.9834, + "step": 16942 + }, + { + "epoch": 0.41, + "learning_rate": 1.3413206571802558e-05, + "loss": 1.159, + "step": 16943 + }, + { + "epoch": 0.41, + "learning_rate": 1.3412474442063386e-05, + "loss": 0.8101, + "step": 16944 + }, + { + "epoch": 0.41, + "learning_rate": 1.3411742291621538e-05, + "loss": 1.0428, + "step": 16945 + }, + { + "epoch": 0.41, + "learning_rate": 1.3411010120481453e-05, + "loss": 0.9598, + "step": 16946 + }, + { + "epoch": 0.41, + "learning_rate": 1.341027792864757e-05, + "loss": 1.0899, + "step": 16947 + }, + { + "epoch": 0.41, + "learning_rate": 1.3409545716124335e-05, + "loss": 0.9991, + "step": 16948 + }, + { + "epoch": 0.41, + "learning_rate": 1.3408813482916187e-05, + "loss": 1.1062, + "step": 16949 + }, + { + "epoch": 0.41, + "learning_rate": 1.340808122902757e-05, + "loss": 0.9816, + "step": 16950 + }, + { + "epoch": 0.41, + "learning_rate": 1.3407348954462926e-05, + "loss": 1.1573, + "step": 16951 + }, + { + "epoch": 0.41, + "learning_rate": 1.3406616659226695e-05, + "loss": 1.0624, + "step": 16952 + }, + { + "epoch": 0.41, + "learning_rate": 1.3405884343323324e-05, + "loss": 1.1793, + "step": 16953 + }, + { + "epoch": 0.41, + "learning_rate": 1.3405152006757253e-05, + "loss": 1.046, + "step": 16954 + }, + { + "epoch": 0.41, + "learning_rate": 1.3404419649532926e-05, + "loss": 1.021, + "step": 16955 + }, + { + "epoch": 0.41, + "learning_rate": 1.3403687271654784e-05, + "loss": 1.1343, + "step": 16956 + }, + { + "epoch": 0.41, + "learning_rate": 1.3402954873127274e-05, + "loss": 1.2199, + "step": 16957 + }, + { + "epoch": 0.41, + "learning_rate": 1.3402222453954837e-05, + "loss": 1.0458, + "step": 16958 + }, + { + "epoch": 0.41, + "learning_rate": 1.3401490014141918e-05, + "loss": 1.302, + "step": 16959 + }, + { + "epoch": 0.41, + "learning_rate": 1.3400757553692955e-05, + "loss": 1.1739, + "step": 16960 + }, + { + "epoch": 0.41, + "learning_rate": 1.34000250726124e-05, + "loss": 1.0185, + "step": 16961 + }, + { + "epoch": 0.41, + "learning_rate": 1.3399292570904691e-05, + "loss": 0.9838, + "step": 16962 + }, + { + "epoch": 0.41, + "learning_rate": 1.3398560048574275e-05, + "loss": 1.0571, + "step": 16963 + }, + { + "epoch": 0.41, + "learning_rate": 1.339782750562559e-05, + "loss": 1.1043, + "step": 16964 + }, + { + "epoch": 0.41, + "learning_rate": 1.339709494206309e-05, + "loss": 0.583, + "step": 16965 + }, + { + "epoch": 0.41, + "learning_rate": 1.3396362357891212e-05, + "loss": 1.2702, + "step": 16966 + }, + { + "epoch": 0.41, + "learning_rate": 1.3395629753114404e-05, + "loss": 1.3242, + "step": 16967 + }, + { + "epoch": 0.41, + "learning_rate": 1.3394897127737105e-05, + "loss": 1.2343, + "step": 16968 + }, + { + "epoch": 0.41, + "learning_rate": 1.3394164481763768e-05, + "loss": 1.1478, + "step": 16969 + }, + { + "epoch": 0.41, + "learning_rate": 1.3393431815198831e-05, + "loss": 1.0623, + "step": 16970 + }, + { + "epoch": 0.41, + "learning_rate": 1.3392699128046742e-05, + "loss": 1.1465, + "step": 16971 + }, + { + "epoch": 0.41, + "learning_rate": 1.3391966420311943e-05, + "loss": 1.1809, + "step": 16972 + }, + { + "epoch": 0.41, + "learning_rate": 1.3391233691998886e-05, + "loss": 1.0306, + "step": 16973 + }, + { + "epoch": 0.41, + "learning_rate": 1.339050094311201e-05, + "loss": 0.9277, + "step": 16974 + }, + { + "epoch": 0.41, + "learning_rate": 1.3389768173655763e-05, + "loss": 0.9068, + "step": 16975 + }, + { + "epoch": 0.41, + "learning_rate": 1.3389035383634587e-05, + "loss": 1.241, + "step": 16976 + }, + { + "epoch": 0.41, + "learning_rate": 1.3388302573052933e-05, + "loss": 1.0656, + "step": 16977 + }, + { + "epoch": 0.41, + "learning_rate": 1.3387569741915244e-05, + "loss": 0.871, + "step": 16978 + }, + { + "epoch": 0.41, + "learning_rate": 1.3386836890225966e-05, + "loss": 0.8364, + "step": 16979 + }, + { + "epoch": 0.41, + "learning_rate": 1.3386104017989543e-05, + "loss": 1.1916, + "step": 16980 + }, + { + "epoch": 0.41, + "learning_rate": 1.3385371125210428e-05, + "loss": 1.0799, + "step": 16981 + }, + { + "epoch": 0.41, + "learning_rate": 1.338463821189306e-05, + "loss": 1.1299, + "step": 16982 + }, + { + "epoch": 0.41, + "learning_rate": 1.3383905278041888e-05, + "loss": 1.016, + "step": 16983 + }, + { + "epoch": 0.41, + "learning_rate": 1.3383172323661358e-05, + "loss": 1.0554, + "step": 16984 + }, + { + "epoch": 0.41, + "learning_rate": 1.3382439348755918e-05, + "loss": 1.2565, + "step": 16985 + }, + { + "epoch": 0.41, + "learning_rate": 1.3381706353330015e-05, + "loss": 1.2876, + "step": 16986 + }, + { + "epoch": 0.41, + "learning_rate": 1.3380973337388094e-05, + "loss": 1.0329, + "step": 16987 + }, + { + "epoch": 0.41, + "learning_rate": 1.3380240300934601e-05, + "loss": 1.0885, + "step": 16988 + }, + { + "epoch": 0.41, + "learning_rate": 1.3379507243973988e-05, + "loss": 1.1738, + "step": 16989 + }, + { + "epoch": 0.41, + "learning_rate": 1.33787741665107e-05, + "loss": 1.1497, + "step": 16990 + }, + { + "epoch": 0.41, + "learning_rate": 1.3378041068549184e-05, + "loss": 1.1137, + "step": 16991 + }, + { + "epoch": 0.41, + "learning_rate": 1.3377307950093883e-05, + "loss": 1.027, + "step": 16992 + }, + { + "epoch": 0.41, + "learning_rate": 1.3376574811149254e-05, + "loss": 1.085, + "step": 16993 + }, + { + "epoch": 0.41, + "learning_rate": 1.3375841651719739e-05, + "loss": 1.0121, + "step": 16994 + }, + { + "epoch": 0.41, + "learning_rate": 1.3375108471809786e-05, + "loss": 1.1046, + "step": 16995 + }, + { + "epoch": 0.41, + "learning_rate": 1.3374375271423841e-05, + "loss": 1.0454, + "step": 16996 + }, + { + "epoch": 0.41, + "learning_rate": 1.3373642050566358e-05, + "loss": 1.0672, + "step": 16997 + }, + { + "epoch": 0.41, + "learning_rate": 1.3372908809241782e-05, + "loss": 1.1295, + "step": 16998 + }, + { + "epoch": 0.41, + "learning_rate": 1.3372175547454562e-05, + "loss": 1.1162, + "step": 16999 + }, + { + "epoch": 0.41, + "learning_rate": 1.3371442265209145e-05, + "loss": 1.1338, + "step": 17000 + }, + { + "epoch": 0.41, + "learning_rate": 1.3370708962509981e-05, + "loss": 0.9049, + "step": 17001 + }, + { + "epoch": 0.41, + "learning_rate": 1.3369975639361518e-05, + "loss": 1.0533, + "step": 17002 + }, + { + "epoch": 0.41, + "learning_rate": 1.3369242295768209e-05, + "loss": 1.2224, + "step": 17003 + }, + { + "epoch": 0.41, + "learning_rate": 1.3368508931734495e-05, + "loss": 1.1446, + "step": 17004 + }, + { + "epoch": 0.41, + "learning_rate": 1.3367775547264832e-05, + "loss": 1.1068, + "step": 17005 + }, + { + "epoch": 0.41, + "learning_rate": 1.3367042142363666e-05, + "loss": 1.2131, + "step": 17006 + }, + { + "epoch": 0.41, + "learning_rate": 1.3366308717035446e-05, + "loss": 1.3846, + "step": 17007 + }, + { + "epoch": 0.41, + "learning_rate": 1.3365575271284624e-05, + "loss": 1.0381, + "step": 17008 + }, + { + "epoch": 0.41, + "learning_rate": 1.336484180511565e-05, + "loss": 1.0081, + "step": 17009 + }, + { + "epoch": 0.41, + "learning_rate": 1.336410831853297e-05, + "loss": 1.0544, + "step": 17010 + }, + { + "epoch": 0.41, + "learning_rate": 1.3363374811541037e-05, + "loss": 0.9808, + "step": 17011 + }, + { + "epoch": 0.41, + "learning_rate": 1.33626412841443e-05, + "loss": 1.016, + "step": 17012 + }, + { + "epoch": 0.41, + "learning_rate": 1.336190773634721e-05, + "loss": 1.2141, + "step": 17013 + }, + { + "epoch": 0.41, + "learning_rate": 1.3361174168154216e-05, + "loss": 1.066, + "step": 17014 + }, + { + "epoch": 0.41, + "learning_rate": 1.336044057956977e-05, + "loss": 1.0109, + "step": 17015 + }, + { + "epoch": 0.41, + "learning_rate": 1.3359706970598319e-05, + "loss": 0.9055, + "step": 17016 + }, + { + "epoch": 0.41, + "learning_rate": 1.3358973341244319e-05, + "loss": 1.1966, + "step": 17017 + }, + { + "epoch": 0.41, + "learning_rate": 1.3358239691512216e-05, + "loss": 0.9433, + "step": 17018 + }, + { + "epoch": 0.41, + "learning_rate": 1.335750602140646e-05, + "loss": 1.0029, + "step": 17019 + }, + { + "epoch": 0.41, + "learning_rate": 1.335677233093151e-05, + "loss": 1.199, + "step": 17020 + }, + { + "epoch": 0.41, + "learning_rate": 1.3356038620091809e-05, + "loss": 1.2526, + "step": 17021 + }, + { + "epoch": 0.41, + "learning_rate": 1.3355304888891813e-05, + "loss": 1.1348, + "step": 17022 + }, + { + "epoch": 0.41, + "learning_rate": 1.3354571137335967e-05, + "loss": 0.9322, + "step": 17023 + }, + { + "epoch": 0.41, + "learning_rate": 1.3353837365428732e-05, + "loss": 1.3025, + "step": 17024 + }, + { + "epoch": 0.41, + "learning_rate": 1.3353103573174553e-05, + "loss": 1.0422, + "step": 17025 + }, + { + "epoch": 0.41, + "learning_rate": 1.3352369760577883e-05, + "loss": 1.2937, + "step": 17026 + }, + { + "epoch": 0.41, + "learning_rate": 1.3351635927643174e-05, + "loss": 0.9828, + "step": 17027 + }, + { + "epoch": 0.41, + "learning_rate": 1.3350902074374878e-05, + "loss": 1.0928, + "step": 17028 + }, + { + "epoch": 0.41, + "learning_rate": 1.335016820077745e-05, + "loss": 1.1069, + "step": 17029 + }, + { + "epoch": 0.41, + "learning_rate": 1.3349434306855335e-05, + "loss": 1.2582, + "step": 17030 + }, + { + "epoch": 0.41, + "learning_rate": 1.3348700392612991e-05, + "loss": 0.9471, + "step": 17031 + }, + { + "epoch": 0.41, + "learning_rate": 1.334796645805487e-05, + "loss": 1.2372, + "step": 17032 + }, + { + "epoch": 0.41, + "learning_rate": 1.3347232503185425e-05, + "loss": 1.1216, + "step": 17033 + }, + { + "epoch": 0.41, + "learning_rate": 1.3346498528009109e-05, + "loss": 1.26, + "step": 17034 + }, + { + "epoch": 0.41, + "learning_rate": 1.334576453253037e-05, + "loss": 1.3065, + "step": 17035 + }, + { + "epoch": 0.41, + "learning_rate": 1.3345030516753665e-05, + "loss": 0.9941, + "step": 17036 + }, + { + "epoch": 0.41, + "learning_rate": 1.334429648068345e-05, + "loss": 1.113, + "step": 17037 + }, + { + "epoch": 0.41, + "learning_rate": 1.3343562424324172e-05, + "loss": 1.004, + "step": 17038 + }, + { + "epoch": 0.41, + "learning_rate": 1.3342828347680288e-05, + "loss": 1.0346, + "step": 17039 + }, + { + "epoch": 0.41, + "learning_rate": 1.3342094250756248e-05, + "loss": 1.1411, + "step": 17040 + }, + { + "epoch": 0.41, + "learning_rate": 1.334136013355651e-05, + "loss": 0.9315, + "step": 17041 + }, + { + "epoch": 0.41, + "learning_rate": 1.334062599608553e-05, + "loss": 1.0548, + "step": 17042 + }, + { + "epoch": 0.41, + "learning_rate": 1.3339891838347754e-05, + "loss": 1.1358, + "step": 17043 + }, + { + "epoch": 0.41, + "learning_rate": 1.333915766034764e-05, + "loss": 0.9928, + "step": 17044 + }, + { + "epoch": 0.41, + "learning_rate": 1.3338423462089643e-05, + "loss": 0.9179, + "step": 17045 + }, + { + "epoch": 0.41, + "learning_rate": 1.3337689243578216e-05, + "loss": 1.0467, + "step": 17046 + }, + { + "epoch": 0.41, + "learning_rate": 1.3336955004817811e-05, + "loss": 1.0669, + "step": 17047 + }, + { + "epoch": 0.41, + "learning_rate": 1.3336220745812889e-05, + "loss": 0.9542, + "step": 17048 + }, + { + "epoch": 0.41, + "learning_rate": 1.3335486466567897e-05, + "loss": 1.0317, + "step": 17049 + }, + { + "epoch": 0.41, + "learning_rate": 1.3334752167087295e-05, + "loss": 0.9309, + "step": 17050 + }, + { + "epoch": 0.41, + "learning_rate": 1.3334017847375536e-05, + "loss": 0.8277, + "step": 17051 + }, + { + "epoch": 0.41, + "learning_rate": 1.3333283507437074e-05, + "loss": 1.176, + "step": 17052 + }, + { + "epoch": 0.41, + "learning_rate": 1.3332549147276365e-05, + "loss": 0.9785, + "step": 17053 + }, + { + "epoch": 0.41, + "learning_rate": 1.3331814766897865e-05, + "loss": 1.1983, + "step": 17054 + }, + { + "epoch": 0.41, + "learning_rate": 1.3331080366306029e-05, + "loss": 0.8686, + "step": 17055 + }, + { + "epoch": 0.41, + "learning_rate": 1.333034594550531e-05, + "loss": 1.3352, + "step": 17056 + }, + { + "epoch": 0.41, + "learning_rate": 1.3329611504500166e-05, + "loss": 1.0877, + "step": 17057 + }, + { + "epoch": 0.41, + "learning_rate": 1.3328877043295055e-05, + "loss": 0.9109, + "step": 17058 + }, + { + "epoch": 0.41, + "learning_rate": 1.3328142561894425e-05, + "loss": 0.9175, + "step": 17059 + }, + { + "epoch": 0.41, + "learning_rate": 1.332740806030274e-05, + "loss": 1.0644, + "step": 17060 + }, + { + "epoch": 0.41, + "learning_rate": 1.3326673538524454e-05, + "loss": 1.2736, + "step": 17061 + }, + { + "epoch": 0.41, + "learning_rate": 1.332593899656402e-05, + "loss": 0.9028, + "step": 17062 + }, + { + "epoch": 0.41, + "learning_rate": 1.3325204434425895e-05, + "loss": 1.1626, + "step": 17063 + }, + { + "epoch": 0.41, + "learning_rate": 1.3324469852114541e-05, + "loss": 1.0602, + "step": 17064 + }, + { + "epoch": 0.41, + "learning_rate": 1.332373524963441e-05, + "loss": 1.0664, + "step": 17065 + }, + { + "epoch": 0.41, + "learning_rate": 1.3323000626989957e-05, + "loss": 1.2459, + "step": 17066 + }, + { + "epoch": 0.41, + "learning_rate": 1.332226598418564e-05, + "loss": 0.9786, + "step": 17067 + }, + { + "epoch": 0.41, + "learning_rate": 1.3321531321225918e-05, + "loss": 1.0709, + "step": 17068 + }, + { + "epoch": 0.41, + "learning_rate": 1.3320796638115248e-05, + "loss": 0.8589, + "step": 17069 + }, + { + "epoch": 0.41, + "learning_rate": 1.3320061934858083e-05, + "loss": 0.9904, + "step": 17070 + }, + { + "epoch": 0.41, + "learning_rate": 1.3319327211458884e-05, + "loss": 1.148, + "step": 17071 + }, + { + "epoch": 0.41, + "learning_rate": 1.331859246792211e-05, + "loss": 1.0027, + "step": 17072 + }, + { + "epoch": 0.41, + "learning_rate": 1.3317857704252212e-05, + "loss": 0.8111, + "step": 17073 + }, + { + "epoch": 0.41, + "learning_rate": 1.3317122920453655e-05, + "loss": 1.1365, + "step": 17074 + }, + { + "epoch": 0.41, + "learning_rate": 1.3316388116530892e-05, + "loss": 0.9227, + "step": 17075 + }, + { + "epoch": 0.41, + "learning_rate": 1.3315653292488385e-05, + "loss": 1.1632, + "step": 17076 + }, + { + "epoch": 0.41, + "learning_rate": 1.3314918448330588e-05, + "loss": 1.0546, + "step": 17077 + }, + { + "epoch": 0.41, + "learning_rate": 1.3314183584061959e-05, + "loss": 1.1208, + "step": 17078 + }, + { + "epoch": 0.41, + "learning_rate": 1.3313448699686956e-05, + "loss": 1.3201, + "step": 17079 + }, + { + "epoch": 0.41, + "learning_rate": 1.3312713795210041e-05, + "loss": 0.9336, + "step": 17080 + }, + { + "epoch": 0.41, + "learning_rate": 1.3311978870635673e-05, + "loss": 1.1829, + "step": 17081 + }, + { + "epoch": 0.41, + "learning_rate": 1.3311243925968307e-05, + "loss": 0.9636, + "step": 17082 + }, + { + "epoch": 0.41, + "learning_rate": 1.33105089612124e-05, + "loss": 1.0173, + "step": 17083 + }, + { + "epoch": 0.41, + "learning_rate": 1.3309773976372417e-05, + "loss": 1.1017, + "step": 17084 + }, + { + "epoch": 0.41, + "learning_rate": 1.3309038971452815e-05, + "loss": 1.269, + "step": 17085 + }, + { + "epoch": 0.41, + "learning_rate": 1.330830394645805e-05, + "loss": 0.9593, + "step": 17086 + }, + { + "epoch": 0.41, + "learning_rate": 1.330756890139258e-05, + "loss": 0.9716, + "step": 17087 + }, + { + "epoch": 0.41, + "learning_rate": 1.3306833836260871e-05, + "loss": 1.1071, + "step": 17088 + }, + { + "epoch": 0.41, + "learning_rate": 1.330609875106738e-05, + "loss": 1.1664, + "step": 17089 + }, + { + "epoch": 0.41, + "learning_rate": 1.3305363645816564e-05, + "loss": 1.0759, + "step": 17090 + }, + { + "epoch": 0.41, + "learning_rate": 1.3304628520512884e-05, + "loss": 1.0035, + "step": 17091 + }, + { + "epoch": 0.41, + "learning_rate": 1.3303893375160798e-05, + "loss": 0.8577, + "step": 17092 + }, + { + "epoch": 0.41, + "learning_rate": 1.3303158209764772e-05, + "loss": 1.3161, + "step": 17093 + }, + { + "epoch": 0.41, + "learning_rate": 1.3302423024329259e-05, + "loss": 1.3914, + "step": 17094 + }, + { + "epoch": 0.41, + "learning_rate": 1.3301687818858723e-05, + "loss": 1.1417, + "step": 17095 + }, + { + "epoch": 0.41, + "learning_rate": 1.3300952593357625e-05, + "loss": 1.318, + "step": 17096 + }, + { + "epoch": 0.41, + "learning_rate": 1.3300217347830424e-05, + "loss": 1.158, + "step": 17097 + }, + { + "epoch": 0.41, + "learning_rate": 1.3299482082281578e-05, + "loss": 0.9696, + "step": 17098 + }, + { + "epoch": 0.41, + "learning_rate": 1.3298746796715554e-05, + "loss": 1.0967, + "step": 17099 + }, + { + "epoch": 0.41, + "learning_rate": 1.3298011491136808e-05, + "loss": 1.1227, + "step": 17100 + }, + { + "epoch": 0.41, + "learning_rate": 1.3297276165549801e-05, + "loss": 1.2602, + "step": 17101 + }, + { + "epoch": 0.41, + "learning_rate": 1.3296540819958994e-05, + "loss": 1.1006, + "step": 17102 + }, + { + "epoch": 0.41, + "learning_rate": 1.329580545436885e-05, + "loss": 1.1003, + "step": 17103 + }, + { + "epoch": 0.41, + "learning_rate": 1.3295070068783831e-05, + "loss": 1.2264, + "step": 17104 + }, + { + "epoch": 0.41, + "learning_rate": 1.3294334663208397e-05, + "loss": 1.1468, + "step": 17105 + }, + { + "epoch": 0.41, + "learning_rate": 1.3293599237647006e-05, + "loss": 1.0611, + "step": 17106 + }, + { + "epoch": 0.41, + "learning_rate": 1.3292863792104128e-05, + "loss": 1.0213, + "step": 17107 + }, + { + "epoch": 0.41, + "learning_rate": 1.3292128326584218e-05, + "loss": 0.7907, + "step": 17108 + }, + { + "epoch": 0.41, + "learning_rate": 1.3291392841091739e-05, + "loss": 1.1462, + "step": 17109 + }, + { + "epoch": 0.41, + "learning_rate": 1.3290657335631153e-05, + "loss": 1.5067, + "step": 17110 + }, + { + "epoch": 0.41, + "learning_rate": 1.3289921810206925e-05, + "loss": 1.1004, + "step": 17111 + }, + { + "epoch": 0.41, + "learning_rate": 1.3289186264823514e-05, + "loss": 1.1812, + "step": 17112 + }, + { + "epoch": 0.41, + "learning_rate": 1.3288450699485385e-05, + "loss": 1.2989, + "step": 17113 + }, + { + "epoch": 0.41, + "learning_rate": 1.3287715114196998e-05, + "loss": 1.2168, + "step": 17114 + }, + { + "epoch": 0.41, + "learning_rate": 1.3286979508962814e-05, + "loss": 0.999, + "step": 17115 + }, + { + "epoch": 0.41, + "learning_rate": 1.3286243883787302e-05, + "loss": 0.913, + "step": 17116 + }, + { + "epoch": 0.41, + "learning_rate": 1.3285508238674923e-05, + "loss": 0.9877, + "step": 17117 + }, + { + "epoch": 0.41, + "learning_rate": 1.3284772573630133e-05, + "loss": 0.9246, + "step": 17118 + }, + { + "epoch": 0.41, + "learning_rate": 1.3284036888657404e-05, + "loss": 0.9747, + "step": 17119 + }, + { + "epoch": 0.41, + "learning_rate": 1.3283301183761194e-05, + "loss": 1.0155, + "step": 17120 + }, + { + "epoch": 0.41, + "learning_rate": 1.3282565458945971e-05, + "loss": 1.0395, + "step": 17121 + }, + { + "epoch": 0.41, + "learning_rate": 1.3281829714216192e-05, + "loss": 1.0429, + "step": 17122 + }, + { + "epoch": 0.41, + "learning_rate": 1.3281093949576324e-05, + "loss": 1.3255, + "step": 17123 + }, + { + "epoch": 0.41, + "learning_rate": 1.3280358165030833e-05, + "loss": 0.9721, + "step": 17124 + }, + { + "epoch": 0.41, + "learning_rate": 1.327962236058418e-05, + "loss": 1.1675, + "step": 17125 + }, + { + "epoch": 0.41, + "learning_rate": 1.3278886536240829e-05, + "loss": 1.2443, + "step": 17126 + }, + { + "epoch": 0.41, + "learning_rate": 1.3278150692005243e-05, + "loss": 1.0249, + "step": 17127 + }, + { + "epoch": 0.41, + "learning_rate": 1.327741482788189e-05, + "loss": 1.127, + "step": 17128 + }, + { + "epoch": 0.41, + "learning_rate": 1.3276678943875235e-05, + "loss": 1.0255, + "step": 17129 + }, + { + "epoch": 0.41, + "learning_rate": 1.3275943039989735e-05, + "loss": 1.1023, + "step": 17130 + }, + { + "epoch": 0.41, + "learning_rate": 1.327520711622986e-05, + "loss": 1.1016, + "step": 17131 + }, + { + "epoch": 0.41, + "learning_rate": 1.3274471172600073e-05, + "loss": 0.9938, + "step": 17132 + }, + { + "epoch": 0.41, + "learning_rate": 1.3273735209104844e-05, + "loss": 1.0417, + "step": 17133 + }, + { + "epoch": 0.41, + "learning_rate": 1.3272999225748628e-05, + "loss": 1.0967, + "step": 17134 + }, + { + "epoch": 0.41, + "learning_rate": 1.32722632225359e-05, + "loss": 1.1453, + "step": 17135 + }, + { + "epoch": 0.41, + "learning_rate": 1.3271527199471116e-05, + "loss": 1.2326, + "step": 17136 + }, + { + "epoch": 0.41, + "learning_rate": 1.3270791156558752e-05, + "loss": 1.1084, + "step": 17137 + }, + { + "epoch": 0.41, + "learning_rate": 1.3270055093803263e-05, + "loss": 0.7875, + "step": 17138 + }, + { + "epoch": 0.41, + "learning_rate": 1.3269319011209118e-05, + "loss": 1.3401, + "step": 17139 + }, + { + "epoch": 0.41, + "learning_rate": 1.3268582908780788e-05, + "loss": 1.2506, + "step": 17140 + }, + { + "epoch": 0.41, + "learning_rate": 1.326784678652273e-05, + "loss": 1.0593, + "step": 17141 + }, + { + "epoch": 0.41, + "learning_rate": 1.3267110644439418e-05, + "loss": 1.0442, + "step": 17142 + }, + { + "epoch": 0.41, + "learning_rate": 1.3266374482535311e-05, + "loss": 1.1047, + "step": 17143 + }, + { + "epoch": 0.41, + "learning_rate": 1.326563830081488e-05, + "loss": 0.9627, + "step": 17144 + }, + { + "epoch": 0.41, + "learning_rate": 1.3264902099282591e-05, + "loss": 0.899, + "step": 17145 + }, + { + "epoch": 0.41, + "learning_rate": 1.3264165877942906e-05, + "loss": 1.0706, + "step": 17146 + }, + { + "epoch": 0.41, + "learning_rate": 1.3263429636800296e-05, + "loss": 0.9962, + "step": 17147 + }, + { + "epoch": 0.41, + "learning_rate": 1.3262693375859225e-05, + "loss": 0.9873, + "step": 17148 + }, + { + "epoch": 0.41, + "learning_rate": 1.3261957095124163e-05, + "loss": 1.2412, + "step": 17149 + }, + { + "epoch": 0.41, + "learning_rate": 1.3261220794599569e-05, + "loss": 1.0821, + "step": 17150 + }, + { + "epoch": 0.41, + "learning_rate": 1.3260484474289921e-05, + "loss": 1.0331, + "step": 17151 + }, + { + "epoch": 0.41, + "learning_rate": 1.325974813419968e-05, + "loss": 0.9239, + "step": 17152 + }, + { + "epoch": 0.41, + "learning_rate": 1.3259011774333312e-05, + "loss": 1.1721, + "step": 17153 + }, + { + "epoch": 0.41, + "learning_rate": 1.3258275394695286e-05, + "loss": 1.2981, + "step": 17154 + }, + { + "epoch": 0.41, + "learning_rate": 1.3257538995290071e-05, + "loss": 1.1256, + "step": 17155 + }, + { + "epoch": 0.41, + "learning_rate": 1.3256802576122134e-05, + "loss": 1.0439, + "step": 17156 + }, + { + "epoch": 0.41, + "learning_rate": 1.325606613719594e-05, + "loss": 1.1344, + "step": 17157 + }, + { + "epoch": 0.41, + "learning_rate": 1.3255329678515959e-05, + "loss": 1.103, + "step": 17158 + }, + { + "epoch": 0.41, + "learning_rate": 1.3254593200086659e-05, + "loss": 1.2915, + "step": 17159 + }, + { + "epoch": 0.41, + "learning_rate": 1.3253856701912507e-05, + "loss": 1.214, + "step": 17160 + }, + { + "epoch": 0.41, + "learning_rate": 1.3253120183997972e-05, + "loss": 1.1399, + "step": 17161 + }, + { + "epoch": 0.41, + "learning_rate": 1.325238364634752e-05, + "loss": 0.9391, + "step": 17162 + }, + { + "epoch": 0.41, + "learning_rate": 1.3251647088965624e-05, + "loss": 1.0022, + "step": 17163 + }, + { + "epoch": 0.41, + "learning_rate": 1.325091051185675e-05, + "loss": 1.1667, + "step": 17164 + }, + { + "epoch": 0.41, + "learning_rate": 1.3250173915025366e-05, + "loss": 0.8364, + "step": 17165 + }, + { + "epoch": 0.41, + "learning_rate": 1.324943729847594e-05, + "loss": 0.5693, + "step": 17166 + }, + { + "epoch": 0.41, + "learning_rate": 1.3248700662212944e-05, + "loss": 1.2264, + "step": 17167 + }, + { + "epoch": 0.41, + "learning_rate": 1.3247964006240845e-05, + "loss": 1.3426, + "step": 17168 + }, + { + "epoch": 0.41, + "learning_rate": 1.3247227330564113e-05, + "loss": 1.1272, + "step": 17169 + }, + { + "epoch": 0.41, + "learning_rate": 1.3246490635187214e-05, + "loss": 1.2677, + "step": 17170 + }, + { + "epoch": 0.41, + "learning_rate": 1.324575392011462e-05, + "loss": 1.369, + "step": 17171 + }, + { + "epoch": 0.41, + "learning_rate": 1.3245017185350803e-05, + "loss": 1.0107, + "step": 17172 + }, + { + "epoch": 0.41, + "learning_rate": 1.3244280430900229e-05, + "loss": 1.1588, + "step": 17173 + }, + { + "epoch": 0.41, + "learning_rate": 1.3243543656767367e-05, + "loss": 1.0025, + "step": 17174 + }, + { + "epoch": 0.41, + "learning_rate": 1.3242806862956692e-05, + "loss": 0.9028, + "step": 17175 + }, + { + "epoch": 0.41, + "learning_rate": 1.3242070049472671e-05, + "loss": 1.2368, + "step": 17176 + }, + { + "epoch": 0.41, + "learning_rate": 1.324133321631977e-05, + "loss": 0.903, + "step": 17177 + }, + { + "epoch": 0.41, + "learning_rate": 1.3240596363502462e-05, + "loss": 1.1252, + "step": 17178 + }, + { + "epoch": 0.41, + "learning_rate": 1.3239859491025223e-05, + "loss": 0.8919, + "step": 17179 + }, + { + "epoch": 0.41, + "learning_rate": 1.3239122598892516e-05, + "loss": 0.6719, + "step": 17180 + }, + { + "epoch": 0.41, + "learning_rate": 1.3238385687108813e-05, + "loss": 1.1184, + "step": 17181 + }, + { + "epoch": 0.41, + "learning_rate": 1.3237648755678585e-05, + "loss": 1.092, + "step": 17182 + }, + { + "epoch": 0.41, + "learning_rate": 1.3236911804606307e-05, + "loss": 0.8994, + "step": 17183 + }, + { + "epoch": 0.41, + "learning_rate": 1.3236174833896445e-05, + "loss": 1.2146, + "step": 17184 + }, + { + "epoch": 0.41, + "learning_rate": 1.323543784355347e-05, + "loss": 1.0519, + "step": 17185 + }, + { + "epoch": 0.41, + "learning_rate": 1.3234700833581855e-05, + "loss": 0.7492, + "step": 17186 + }, + { + "epoch": 0.41, + "learning_rate": 1.3233963803986072e-05, + "loss": 1.2205, + "step": 17187 + }, + { + "epoch": 0.41, + "learning_rate": 1.3233226754770592e-05, + "loss": 0.9484, + "step": 17188 + }, + { + "epoch": 0.41, + "learning_rate": 1.3232489685939882e-05, + "loss": 1.2831, + "step": 17189 + }, + { + "epoch": 0.41, + "learning_rate": 1.3231752597498421e-05, + "loss": 1.2041, + "step": 17190 + }, + { + "epoch": 0.41, + "learning_rate": 1.3231015489450676e-05, + "loss": 1.1078, + "step": 17191 + }, + { + "epoch": 0.41, + "learning_rate": 1.323027836180112e-05, + "loss": 1.1752, + "step": 17192 + }, + { + "epoch": 0.41, + "learning_rate": 1.3229541214554225e-05, + "loss": 0.8911, + "step": 17193 + }, + { + "epoch": 0.41, + "learning_rate": 1.3228804047714462e-05, + "loss": 0.9164, + "step": 17194 + }, + { + "epoch": 0.41, + "learning_rate": 1.3228066861286308e-05, + "loss": 0.9952, + "step": 17195 + }, + { + "epoch": 0.41, + "learning_rate": 1.322732965527423e-05, + "loss": 1.2145, + "step": 17196 + }, + { + "epoch": 0.41, + "learning_rate": 1.3226592429682698e-05, + "loss": 1.0227, + "step": 17197 + }, + { + "epoch": 0.41, + "learning_rate": 1.3225855184516194e-05, + "loss": 1.3639, + "step": 17198 + }, + { + "epoch": 0.41, + "learning_rate": 1.3225117919779184e-05, + "loss": 1.0317, + "step": 17199 + }, + { + "epoch": 0.41, + "learning_rate": 1.3224380635476142e-05, + "loss": 1.1404, + "step": 17200 + }, + { + "epoch": 0.41, + "learning_rate": 1.3223643331611538e-05, + "loss": 1.0631, + "step": 17201 + }, + { + "epoch": 0.41, + "learning_rate": 1.3222906008189851e-05, + "loss": 1.1527, + "step": 17202 + }, + { + "epoch": 0.41, + "learning_rate": 1.3222168665215554e-05, + "loss": 1.154, + "step": 17203 + }, + { + "epoch": 0.41, + "learning_rate": 1.3221431302693116e-05, + "loss": 1.1114, + "step": 17204 + }, + { + "epoch": 0.41, + "learning_rate": 1.3220693920627008e-05, + "loss": 1.152, + "step": 17205 + }, + { + "epoch": 0.41, + "learning_rate": 1.321995651902171e-05, + "loss": 1.0316, + "step": 17206 + }, + { + "epoch": 0.41, + "learning_rate": 1.3219219097881693e-05, + "loss": 1.1159, + "step": 17207 + }, + { + "epoch": 0.41, + "learning_rate": 1.3218481657211431e-05, + "loss": 1.4267, + "step": 17208 + }, + { + "epoch": 0.41, + "learning_rate": 1.3217744197015399e-05, + "loss": 1.1296, + "step": 17209 + }, + { + "epoch": 0.41, + "learning_rate": 1.3217006717298066e-05, + "loss": 0.9951, + "step": 17210 + }, + { + "epoch": 0.41, + "learning_rate": 1.3216269218063915e-05, + "loss": 1.0974, + "step": 17211 + }, + { + "epoch": 0.41, + "learning_rate": 1.3215531699317412e-05, + "loss": 1.1076, + "step": 17212 + }, + { + "epoch": 0.41, + "learning_rate": 1.3214794161063038e-05, + "loss": 1.2849, + "step": 17213 + }, + { + "epoch": 0.41, + "learning_rate": 1.3214056603305258e-05, + "loss": 1.069, + "step": 17214 + }, + { + "epoch": 0.41, + "learning_rate": 1.3213319026048557e-05, + "loss": 1.1271, + "step": 17215 + }, + { + "epoch": 0.41, + "learning_rate": 1.3212581429297407e-05, + "loss": 1.0237, + "step": 17216 + }, + { + "epoch": 0.41, + "learning_rate": 1.3211843813056276e-05, + "loss": 1.1293, + "step": 17217 + }, + { + "epoch": 0.41, + "learning_rate": 1.3211106177329647e-05, + "loss": 1.3016, + "step": 17218 + }, + { + "epoch": 0.41, + "learning_rate": 1.3210368522121992e-05, + "loss": 1.1644, + "step": 17219 + }, + { + "epoch": 0.41, + "learning_rate": 1.3209630847437787e-05, + "loss": 1.2766, + "step": 17220 + }, + { + "epoch": 0.41, + "learning_rate": 1.3208893153281504e-05, + "loss": 0.9713, + "step": 17221 + }, + { + "epoch": 0.41, + "learning_rate": 1.3208155439657624e-05, + "loss": 0.9207, + "step": 17222 + }, + { + "epoch": 0.41, + "learning_rate": 1.3207417706570617e-05, + "loss": 1.1854, + "step": 17223 + }, + { + "epoch": 0.41, + "learning_rate": 1.3206679954024963e-05, + "loss": 0.8527, + "step": 17224 + }, + { + "epoch": 0.41, + "learning_rate": 1.3205942182025136e-05, + "loss": 1.0658, + "step": 17225 + }, + { + "epoch": 0.41, + "learning_rate": 1.320520439057561e-05, + "loss": 1.2648, + "step": 17226 + }, + { + "epoch": 0.41, + "learning_rate": 1.3204466579680866e-05, + "loss": 1.2575, + "step": 17227 + }, + { + "epoch": 0.41, + "learning_rate": 1.3203728749345375e-05, + "loss": 1.1684, + "step": 17228 + }, + { + "epoch": 0.41, + "learning_rate": 1.3202990899573613e-05, + "loss": 1.0015, + "step": 17229 + }, + { + "epoch": 0.41, + "learning_rate": 1.3202253030370062e-05, + "loss": 0.9405, + "step": 17230 + }, + { + "epoch": 0.41, + "learning_rate": 1.3201515141739193e-05, + "loss": 1.073, + "step": 17231 + }, + { + "epoch": 0.41, + "learning_rate": 1.3200777233685488e-05, + "loss": 1.1686, + "step": 17232 + }, + { + "epoch": 0.41, + "learning_rate": 1.3200039306213417e-05, + "loss": 0.7792, + "step": 17233 + }, + { + "epoch": 0.41, + "learning_rate": 1.3199301359327462e-05, + "loss": 1.1821, + "step": 17234 + }, + { + "epoch": 0.41, + "learning_rate": 1.3198563393032098e-05, + "loss": 0.9368, + "step": 17235 + }, + { + "epoch": 0.41, + "learning_rate": 1.3197825407331801e-05, + "loss": 1.2418, + "step": 17236 + }, + { + "epoch": 0.41, + "learning_rate": 1.319708740223105e-05, + "loss": 1.2368, + "step": 17237 + }, + { + "epoch": 0.41, + "learning_rate": 1.3196349377734321e-05, + "loss": 1.1238, + "step": 17238 + }, + { + "epoch": 0.41, + "learning_rate": 1.3195611333846094e-05, + "loss": 1.2385, + "step": 17239 + }, + { + "epoch": 0.41, + "learning_rate": 1.3194873270570845e-05, + "loss": 0.7684, + "step": 17240 + }, + { + "epoch": 0.41, + "learning_rate": 1.3194135187913048e-05, + "loss": 1.0386, + "step": 17241 + }, + { + "epoch": 0.41, + "learning_rate": 1.3193397085877186e-05, + "loss": 0.9116, + "step": 17242 + }, + { + "epoch": 0.41, + "learning_rate": 1.3192658964467736e-05, + "loss": 1.1839, + "step": 17243 + }, + { + "epoch": 0.41, + "learning_rate": 1.3191920823689178e-05, + "loss": 1.2845, + "step": 17244 + }, + { + "epoch": 0.41, + "learning_rate": 1.319118266354598e-05, + "loss": 1.1143, + "step": 17245 + }, + { + "epoch": 0.41, + "learning_rate": 1.3190444484042631e-05, + "loss": 1.0938, + "step": 17246 + }, + { + "epoch": 0.41, + "learning_rate": 1.3189706285183608e-05, + "loss": 1.1892, + "step": 17247 + }, + { + "epoch": 0.41, + "learning_rate": 1.3188968066973385e-05, + "loss": 1.1907, + "step": 17248 + }, + { + "epoch": 0.41, + "learning_rate": 1.3188229829416439e-05, + "loss": 1.2204, + "step": 17249 + }, + { + "epoch": 0.41, + "learning_rate": 1.3187491572517257e-05, + "loss": 1.2752, + "step": 17250 + }, + { + "epoch": 0.41, + "learning_rate": 1.3186753296280314e-05, + "loss": 1.2413, + "step": 17251 + }, + { + "epoch": 0.41, + "learning_rate": 1.3186015000710086e-05, + "loss": 1.0318, + "step": 17252 + }, + { + "epoch": 0.41, + "learning_rate": 1.3185276685811054e-05, + "loss": 1.1622, + "step": 17253 + }, + { + "epoch": 0.41, + "learning_rate": 1.31845383515877e-05, + "loss": 1.0644, + "step": 17254 + }, + { + "epoch": 0.41, + "learning_rate": 1.31837999980445e-05, + "loss": 1.1472, + "step": 17255 + }, + { + "epoch": 0.41, + "learning_rate": 1.3183061625185936e-05, + "loss": 1.2633, + "step": 17256 + }, + { + "epoch": 0.42, + "learning_rate": 1.318232323301648e-05, + "loss": 1.0186, + "step": 17257 + }, + { + "epoch": 0.42, + "learning_rate": 1.3181584821540622e-05, + "loss": 1.1684, + "step": 17258 + }, + { + "epoch": 0.42, + "learning_rate": 1.3180846390762838e-05, + "loss": 1.2952, + "step": 17259 + }, + { + "epoch": 0.42, + "learning_rate": 1.3180107940687607e-05, + "loss": 1.2447, + "step": 17260 + }, + { + "epoch": 0.42, + "learning_rate": 1.3179369471319405e-05, + "loss": 1.0032, + "step": 17261 + }, + { + "epoch": 0.42, + "learning_rate": 1.317863098266272e-05, + "loss": 0.9203, + "step": 17262 + }, + { + "epoch": 0.42, + "learning_rate": 1.3177892474722027e-05, + "loss": 0.9445, + "step": 17263 + }, + { + "epoch": 0.42, + "learning_rate": 1.3177153947501808e-05, + "loss": 1.0696, + "step": 17264 + }, + { + "epoch": 0.42, + "learning_rate": 1.3176415401006543e-05, + "loss": 1.1844, + "step": 17265 + }, + { + "epoch": 0.42, + "learning_rate": 1.3175676835240716e-05, + "loss": 1.1667, + "step": 17266 + }, + { + "epoch": 0.42, + "learning_rate": 1.3174938250208802e-05, + "loss": 1.0075, + "step": 17267 + }, + { + "epoch": 0.42, + "learning_rate": 1.3174199645915283e-05, + "loss": 1.1627, + "step": 17268 + }, + { + "epoch": 0.42, + "learning_rate": 1.3173461022364645e-05, + "loss": 1.2273, + "step": 17269 + }, + { + "epoch": 0.42, + "learning_rate": 1.3172722379561363e-05, + "loss": 0.9845, + "step": 17270 + }, + { + "epoch": 0.42, + "learning_rate": 1.3171983717509923e-05, + "loss": 1.2201, + "step": 17271 + }, + { + "epoch": 0.42, + "learning_rate": 1.3171245036214799e-05, + "loss": 0.9714, + "step": 17272 + }, + { + "epoch": 0.42, + "learning_rate": 1.317050633568048e-05, + "loss": 0.8178, + "step": 17273 + }, + { + "epoch": 0.42, + "learning_rate": 1.3169767615911446e-05, + "loss": 1.0129, + "step": 17274 + }, + { + "epoch": 0.42, + "learning_rate": 1.3169028876912177e-05, + "loss": 1.1593, + "step": 17275 + }, + { + "epoch": 0.42, + "learning_rate": 1.3168290118687153e-05, + "loss": 1.1038, + "step": 17276 + }, + { + "epoch": 0.42, + "learning_rate": 1.3167551341240858e-05, + "loss": 1.2852, + "step": 17277 + }, + { + "epoch": 0.42, + "learning_rate": 1.3166812544577779e-05, + "loss": 1.1903, + "step": 17278 + }, + { + "epoch": 0.42, + "learning_rate": 1.3166073728702389e-05, + "loss": 1.0435, + "step": 17279 + }, + { + "epoch": 0.42, + "learning_rate": 1.3165334893619175e-05, + "loss": 1.1367, + "step": 17280 + }, + { + "epoch": 0.42, + "learning_rate": 1.3164596039332617e-05, + "loss": 1.0789, + "step": 17281 + }, + { + "epoch": 0.42, + "learning_rate": 1.3163857165847205e-05, + "loss": 1.2019, + "step": 17282 + }, + { + "epoch": 0.42, + "learning_rate": 1.3163118273167413e-05, + "loss": 0.9245, + "step": 17283 + }, + { + "epoch": 0.42, + "learning_rate": 1.3162379361297721e-05, + "loss": 0.9809, + "step": 17284 + }, + { + "epoch": 0.42, + "learning_rate": 1.3161640430242624e-05, + "loss": 0.8924, + "step": 17285 + }, + { + "epoch": 0.42, + "learning_rate": 1.3160901480006596e-05, + "loss": 1.3057, + "step": 17286 + }, + { + "epoch": 0.42, + "learning_rate": 1.3160162510594123e-05, + "loss": 1.1283, + "step": 17287 + }, + { + "epoch": 0.42, + "learning_rate": 1.3159423522009684e-05, + "loss": 1.195, + "step": 17288 + }, + { + "epoch": 0.42, + "learning_rate": 1.315868451425777e-05, + "loss": 1.0168, + "step": 17289 + }, + { + "epoch": 0.42, + "learning_rate": 1.3157945487342855e-05, + "loss": 1.0507, + "step": 17290 + }, + { + "epoch": 0.42, + "learning_rate": 1.3157206441269432e-05, + "loss": 1.1266, + "step": 17291 + }, + { + "epoch": 0.42, + "learning_rate": 1.315646737604198e-05, + "loss": 1.043, + "step": 17292 + }, + { + "epoch": 0.42, + "learning_rate": 1.315572829166498e-05, + "loss": 1.2382, + "step": 17293 + }, + { + "epoch": 0.42, + "learning_rate": 1.3154989188142921e-05, + "loss": 1.2073, + "step": 17294 + }, + { + "epoch": 0.42, + "learning_rate": 1.3154250065480282e-05, + "loss": 1.3042, + "step": 17295 + }, + { + "epoch": 0.42, + "learning_rate": 1.3153510923681554e-05, + "loss": 0.9772, + "step": 17296 + }, + { + "epoch": 0.42, + "learning_rate": 1.3152771762751214e-05, + "loss": 1.2144, + "step": 17297 + }, + { + "epoch": 0.42, + "learning_rate": 1.315203258269375e-05, + "loss": 0.946, + "step": 17298 + }, + { + "epoch": 0.42, + "learning_rate": 1.3151293383513645e-05, + "loss": 1.1392, + "step": 17299 + }, + { + "epoch": 0.42, + "learning_rate": 1.3150554165215387e-05, + "loss": 1.0126, + "step": 17300 + }, + { + "epoch": 0.42, + "learning_rate": 1.3149814927803455e-05, + "loss": 0.982, + "step": 17301 + }, + { + "epoch": 0.42, + "learning_rate": 1.314907567128234e-05, + "loss": 1.2757, + "step": 17302 + }, + { + "epoch": 0.42, + "learning_rate": 1.314833639565652e-05, + "loss": 0.9708, + "step": 17303 + }, + { + "epoch": 0.42, + "learning_rate": 1.3147597100930482e-05, + "loss": 1.0544, + "step": 17304 + }, + { + "epoch": 0.42, + "learning_rate": 1.3146857787108717e-05, + "loss": 1.1933, + "step": 17305 + }, + { + "epoch": 0.42, + "learning_rate": 1.3146118454195706e-05, + "loss": 1.197, + "step": 17306 + }, + { + "epoch": 0.42, + "learning_rate": 1.3145379102195935e-05, + "loss": 0.9345, + "step": 17307 + }, + { + "epoch": 0.42, + "learning_rate": 1.3144639731113884e-05, + "loss": 0.9029, + "step": 17308 + }, + { + "epoch": 0.42, + "learning_rate": 1.3143900340954048e-05, + "loss": 1.0219, + "step": 17309 + }, + { + "epoch": 0.42, + "learning_rate": 1.3143160931720906e-05, + "loss": 1.1819, + "step": 17310 + }, + { + "epoch": 0.42, + "learning_rate": 1.3142421503418947e-05, + "loss": 1.119, + "step": 17311 + }, + { + "epoch": 0.42, + "learning_rate": 1.3141682056052653e-05, + "loss": 1.1178, + "step": 17312 + }, + { + "epoch": 0.42, + "learning_rate": 1.3140942589626516e-05, + "loss": 1.0686, + "step": 17313 + }, + { + "epoch": 0.42, + "learning_rate": 1.314020310414502e-05, + "loss": 0.963, + "step": 17314 + }, + { + "epoch": 0.42, + "learning_rate": 1.3139463599612647e-05, + "loss": 0.8521, + "step": 17315 + }, + { + "epoch": 0.42, + "learning_rate": 1.3138724076033888e-05, + "loss": 1.1153, + "step": 17316 + }, + { + "epoch": 0.42, + "learning_rate": 1.313798453341323e-05, + "loss": 0.9646, + "step": 17317 + }, + { + "epoch": 0.42, + "learning_rate": 1.3137244971755156e-05, + "loss": 1.2303, + "step": 17318 + }, + { + "epoch": 0.42, + "learning_rate": 1.3136505391064155e-05, + "loss": 1.2108, + "step": 17319 + }, + { + "epoch": 0.42, + "learning_rate": 1.3135765791344715e-05, + "loss": 1.2281, + "step": 17320 + }, + { + "epoch": 0.42, + "learning_rate": 1.313502617260132e-05, + "loss": 1.1588, + "step": 17321 + }, + { + "epoch": 0.42, + "learning_rate": 1.3134286534838459e-05, + "loss": 1.1314, + "step": 17322 + }, + { + "epoch": 0.42, + "learning_rate": 1.313354687806062e-05, + "loss": 1.0351, + "step": 17323 + }, + { + "epoch": 0.42, + "learning_rate": 1.3132807202272286e-05, + "loss": 1.224, + "step": 17324 + }, + { + "epoch": 0.42, + "learning_rate": 1.3132067507477951e-05, + "loss": 1.2644, + "step": 17325 + }, + { + "epoch": 0.42, + "learning_rate": 1.3131327793682099e-05, + "loss": 1.0051, + "step": 17326 + }, + { + "epoch": 0.42, + "learning_rate": 1.3130588060889216e-05, + "loss": 1.0411, + "step": 17327 + }, + { + "epoch": 0.42, + "learning_rate": 1.312984830910379e-05, + "loss": 1.0785, + "step": 17328 + }, + { + "epoch": 0.42, + "learning_rate": 1.3129108538330315e-05, + "loss": 1.1621, + "step": 17329 + }, + { + "epoch": 0.42, + "learning_rate": 1.3128368748573272e-05, + "loss": 1.2731, + "step": 17330 + }, + { + "epoch": 0.42, + "learning_rate": 1.3127628939837151e-05, + "loss": 0.9989, + "step": 17331 + }, + { + "epoch": 0.42, + "learning_rate": 1.3126889112126442e-05, + "loss": 1.1576, + "step": 17332 + }, + { + "epoch": 0.42, + "learning_rate": 1.3126149265445632e-05, + "loss": 1.168, + "step": 17333 + }, + { + "epoch": 0.42, + "learning_rate": 1.312540939979921e-05, + "loss": 0.9376, + "step": 17334 + }, + { + "epoch": 0.42, + "learning_rate": 1.3124669515191664e-05, + "loss": 1.1397, + "step": 17335 + }, + { + "epoch": 0.42, + "learning_rate": 1.3123929611627481e-05, + "loss": 1.0566, + "step": 17336 + }, + { + "epoch": 0.42, + "learning_rate": 1.3123189689111155e-05, + "loss": 0.5314, + "step": 17337 + }, + { + "epoch": 0.42, + "learning_rate": 1.3122449747647172e-05, + "loss": 0.8878, + "step": 17338 + }, + { + "epoch": 0.42, + "learning_rate": 1.312170978724002e-05, + "loss": 1.0763, + "step": 17339 + }, + { + "epoch": 0.42, + "learning_rate": 1.3120969807894186e-05, + "loss": 1.0182, + "step": 17340 + }, + { + "epoch": 0.42, + "learning_rate": 1.3120229809614167e-05, + "loss": 1.0747, + "step": 17341 + }, + { + "epoch": 0.42, + "learning_rate": 1.3119489792404445e-05, + "loss": 1.1778, + "step": 17342 + }, + { + "epoch": 0.42, + "learning_rate": 1.3118749756269514e-05, + "loss": 1.1966, + "step": 17343 + }, + { + "epoch": 0.42, + "learning_rate": 1.3118009701213859e-05, + "loss": 1.1738, + "step": 17344 + }, + { + "epoch": 0.42, + "learning_rate": 1.3117269627241973e-05, + "loss": 1.2175, + "step": 17345 + }, + { + "epoch": 0.42, + "learning_rate": 1.3116529534358346e-05, + "loss": 1.2226, + "step": 17346 + }, + { + "epoch": 0.42, + "learning_rate": 1.3115789422567469e-05, + "loss": 1.1856, + "step": 17347 + }, + { + "epoch": 0.42, + "learning_rate": 1.3115049291873826e-05, + "loss": 1.1694, + "step": 17348 + }, + { + "epoch": 0.42, + "learning_rate": 1.3114309142281917e-05, + "loss": 0.8467, + "step": 17349 + }, + { + "epoch": 0.42, + "learning_rate": 1.3113568973796224e-05, + "loss": 1.0121, + "step": 17350 + }, + { + "epoch": 0.42, + "learning_rate": 1.311282878642124e-05, + "loss": 0.8909, + "step": 17351 + }, + { + "epoch": 0.42, + "learning_rate": 1.311208858016146e-05, + "loss": 0.816, + "step": 17352 + }, + { + "epoch": 0.42, + "learning_rate": 1.3111348355021367e-05, + "loss": 1.014, + "step": 17353 + }, + { + "epoch": 0.42, + "learning_rate": 1.3110608111005456e-05, + "loss": 0.898, + "step": 17354 + }, + { + "epoch": 0.42, + "learning_rate": 1.3109867848118216e-05, + "loss": 1.0745, + "step": 17355 + }, + { + "epoch": 0.42, + "learning_rate": 1.3109127566364142e-05, + "loss": 1.1391, + "step": 17356 + }, + { + "epoch": 0.42, + "learning_rate": 1.310838726574772e-05, + "loss": 1.0588, + "step": 17357 + }, + { + "epoch": 0.42, + "learning_rate": 1.3107646946273444e-05, + "loss": 1.116, + "step": 17358 + }, + { + "epoch": 0.42, + "learning_rate": 1.3106906607945807e-05, + "loss": 0.9322, + "step": 17359 + }, + { + "epoch": 0.42, + "learning_rate": 1.3106166250769297e-05, + "loss": 0.9526, + "step": 17360 + }, + { + "epoch": 0.42, + "learning_rate": 1.3105425874748407e-05, + "loss": 1.0334, + "step": 17361 + }, + { + "epoch": 0.42, + "learning_rate": 1.3104685479887628e-05, + "loss": 1.0146, + "step": 17362 + }, + { + "epoch": 0.42, + "learning_rate": 1.310394506619145e-05, + "loss": 1.3373, + "step": 17363 + }, + { + "epoch": 0.42, + "learning_rate": 1.3103204633664373e-05, + "loss": 0.863, + "step": 17364 + }, + { + "epoch": 0.42, + "learning_rate": 1.310246418231088e-05, + "loss": 1.0104, + "step": 17365 + }, + { + "epoch": 0.42, + "learning_rate": 1.3101723712135468e-05, + "loss": 1.0849, + "step": 17366 + }, + { + "epoch": 0.42, + "learning_rate": 1.3100983223142625e-05, + "loss": 1.1401, + "step": 17367 + }, + { + "epoch": 0.42, + "learning_rate": 1.3100242715336851e-05, + "loss": 1.1132, + "step": 17368 + }, + { + "epoch": 0.42, + "learning_rate": 1.309950218872263e-05, + "loss": 1.054, + "step": 17369 + }, + { + "epoch": 0.42, + "learning_rate": 1.3098761643304461e-05, + "loss": 1.1168, + "step": 17370 + }, + { + "epoch": 0.42, + "learning_rate": 1.3098021079086829e-05, + "loss": 1.0147, + "step": 17371 + }, + { + "epoch": 0.42, + "learning_rate": 1.3097280496074237e-05, + "loss": 1.2312, + "step": 17372 + }, + { + "epoch": 0.42, + "learning_rate": 1.3096539894271169e-05, + "loss": 1.0842, + "step": 17373 + }, + { + "epoch": 0.42, + "learning_rate": 1.3095799273682124e-05, + "loss": 1.0791, + "step": 17374 + }, + { + "epoch": 0.42, + "learning_rate": 1.3095058634311593e-05, + "loss": 1.0581, + "step": 17375 + }, + { + "epoch": 0.42, + "learning_rate": 1.3094317976164066e-05, + "loss": 1.0548, + "step": 17376 + }, + { + "epoch": 0.42, + "learning_rate": 1.3093577299244045e-05, + "loss": 0.9191, + "step": 17377 + }, + { + "epoch": 0.42, + "learning_rate": 1.3092836603556016e-05, + "loss": 0.9157, + "step": 17378 + }, + { + "epoch": 0.42, + "learning_rate": 1.3092095889104473e-05, + "loss": 1.1253, + "step": 17379 + }, + { + "epoch": 0.42, + "learning_rate": 1.3091355155893912e-05, + "loss": 1.2753, + "step": 17380 + }, + { + "epoch": 0.42, + "learning_rate": 1.3090614403928828e-05, + "loss": 1.1356, + "step": 17381 + }, + { + "epoch": 0.42, + "learning_rate": 1.3089873633213714e-05, + "loss": 1.111, + "step": 17382 + }, + { + "epoch": 0.42, + "learning_rate": 1.3089132843753064e-05, + "loss": 1.1438, + "step": 17383 + }, + { + "epoch": 0.42, + "learning_rate": 1.3088392035551367e-05, + "loss": 1.1739, + "step": 17384 + }, + { + "epoch": 0.42, + "learning_rate": 1.3087651208613123e-05, + "loss": 1.1849, + "step": 17385 + }, + { + "epoch": 0.42, + "learning_rate": 1.3086910362942829e-05, + "loss": 1.1408, + "step": 17386 + }, + { + "epoch": 0.42, + "learning_rate": 1.3086169498544975e-05, + "loss": 1.0685, + "step": 17387 + }, + { + "epoch": 0.42, + "learning_rate": 1.3085428615424055e-05, + "loss": 1.2092, + "step": 17388 + }, + { + "epoch": 0.42, + "learning_rate": 1.3084687713584565e-05, + "loss": 0.9271, + "step": 17389 + }, + { + "epoch": 0.42, + "learning_rate": 1.3083946793031003e-05, + "loss": 1.1971, + "step": 17390 + }, + { + "epoch": 0.42, + "learning_rate": 1.3083205853767858e-05, + "loss": 1.2665, + "step": 17391 + }, + { + "epoch": 0.42, + "learning_rate": 1.308246489579963e-05, + "loss": 1.043, + "step": 17392 + }, + { + "epoch": 0.42, + "learning_rate": 1.3081723919130813e-05, + "loss": 0.957, + "step": 17393 + }, + { + "epoch": 0.42, + "learning_rate": 1.3080982923765902e-05, + "loss": 1.2075, + "step": 17394 + }, + { + "epoch": 0.42, + "learning_rate": 1.3080241909709388e-05, + "loss": 1.2098, + "step": 17395 + }, + { + "epoch": 0.42, + "learning_rate": 1.3079500876965776e-05, + "loss": 1.2785, + "step": 17396 + }, + { + "epoch": 0.42, + "learning_rate": 1.3078759825539556e-05, + "loss": 1.2914, + "step": 17397 + }, + { + "epoch": 0.42, + "learning_rate": 1.3078018755435223e-05, + "loss": 1.1699, + "step": 17398 + }, + { + "epoch": 0.42, + "learning_rate": 1.3077277666657272e-05, + "loss": 1.188, + "step": 17399 + }, + { + "epoch": 0.42, + "learning_rate": 1.3076536559210204e-05, + "loss": 1.3675, + "step": 17400 + }, + { + "epoch": 0.42, + "learning_rate": 1.3075795433098513e-05, + "loss": 0.9611, + "step": 17401 + }, + { + "epoch": 0.42, + "learning_rate": 1.3075054288326694e-05, + "loss": 1.1337, + "step": 17402 + }, + { + "epoch": 0.42, + "learning_rate": 1.307431312489924e-05, + "loss": 0.9395, + "step": 17403 + }, + { + "epoch": 0.42, + "learning_rate": 1.3073571942820658e-05, + "loss": 0.9046, + "step": 17404 + }, + { + "epoch": 0.42, + "learning_rate": 1.3072830742095433e-05, + "loss": 1.3093, + "step": 17405 + }, + { + "epoch": 0.42, + "learning_rate": 1.3072089522728069e-05, + "loss": 1.0757, + "step": 17406 + }, + { + "epoch": 0.42, + "learning_rate": 1.3071348284723057e-05, + "loss": 0.9429, + "step": 17407 + }, + { + "epoch": 0.42, + "learning_rate": 1.3070607028084903e-05, + "loss": 1.2477, + "step": 17408 + }, + { + "epoch": 0.42, + "learning_rate": 1.3069865752818097e-05, + "loss": 1.203, + "step": 17409 + }, + { + "epoch": 0.42, + "learning_rate": 1.3069124458927136e-05, + "loss": 1.0685, + "step": 17410 + }, + { + "epoch": 0.42, + "learning_rate": 1.3068383146416516e-05, + "loss": 1.1454, + "step": 17411 + }, + { + "epoch": 0.42, + "learning_rate": 1.3067641815290742e-05, + "loss": 1.1768, + "step": 17412 + }, + { + "epoch": 0.42, + "learning_rate": 1.3066900465554304e-05, + "loss": 1.3161, + "step": 17413 + }, + { + "epoch": 0.42, + "learning_rate": 1.3066159097211703e-05, + "loss": 1.1588, + "step": 17414 + }, + { + "epoch": 0.42, + "learning_rate": 1.3065417710267437e-05, + "loss": 1.013, + "step": 17415 + }, + { + "epoch": 0.42, + "learning_rate": 1.3064676304726001e-05, + "loss": 1.2626, + "step": 17416 + }, + { + "epoch": 0.42, + "learning_rate": 1.3063934880591895e-05, + "loss": 0.9803, + "step": 17417 + }, + { + "epoch": 0.42, + "learning_rate": 1.306319343786962e-05, + "loss": 1.231, + "step": 17418 + }, + { + "epoch": 0.42, + "learning_rate": 1.3062451976563667e-05, + "loss": 1.1174, + "step": 17419 + }, + { + "epoch": 0.42, + "learning_rate": 1.306171049667854e-05, + "loss": 1.2661, + "step": 17420 + }, + { + "epoch": 0.42, + "learning_rate": 1.3060968998218733e-05, + "loss": 0.8865, + "step": 17421 + }, + { + "epoch": 0.42, + "learning_rate": 1.306022748118875e-05, + "loss": 1.2038, + "step": 17422 + }, + { + "epoch": 0.42, + "learning_rate": 1.3059485945593084e-05, + "loss": 1.0272, + "step": 17423 + }, + { + "epoch": 0.42, + "learning_rate": 1.305874439143624e-05, + "loss": 0.9782, + "step": 17424 + }, + { + "epoch": 0.42, + "learning_rate": 1.305800281872271e-05, + "loss": 1.3669, + "step": 17425 + }, + { + "epoch": 0.42, + "learning_rate": 1.3057261227457e-05, + "loss": 1.2801, + "step": 17426 + }, + { + "epoch": 0.42, + "learning_rate": 1.3056519617643599e-05, + "loss": 1.0239, + "step": 17427 + }, + { + "epoch": 0.42, + "learning_rate": 1.3055777989287018e-05, + "loss": 1.1461, + "step": 17428 + }, + { + "epoch": 0.42, + "learning_rate": 1.3055036342391749e-05, + "loss": 0.9544, + "step": 17429 + }, + { + "epoch": 0.42, + "learning_rate": 1.3054294676962294e-05, + "loss": 1.1009, + "step": 17430 + }, + { + "epoch": 0.42, + "learning_rate": 1.3053552993003148e-05, + "loss": 1.0148, + "step": 17431 + }, + { + "epoch": 0.42, + "learning_rate": 1.305281129051882e-05, + "loss": 1.0196, + "step": 17432 + }, + { + "epoch": 0.42, + "learning_rate": 1.3052069569513798e-05, + "loss": 1.2892, + "step": 17433 + }, + { + "epoch": 0.42, + "learning_rate": 1.3051327829992593e-05, + "loss": 1.0314, + "step": 17434 + }, + { + "epoch": 0.42, + "learning_rate": 1.3050586071959694e-05, + "loss": 0.8157, + "step": 17435 + }, + { + "epoch": 0.42, + "learning_rate": 1.3049844295419614e-05, + "loss": 1.1659, + "step": 17436 + }, + { + "epoch": 0.42, + "learning_rate": 1.304910250037684e-05, + "loss": 1.0308, + "step": 17437 + }, + { + "epoch": 0.42, + "learning_rate": 1.3048360686835882e-05, + "loss": 1.1326, + "step": 17438 + }, + { + "epoch": 0.42, + "learning_rate": 1.3047618854801233e-05, + "loss": 1.3542, + "step": 17439 + }, + { + "epoch": 0.42, + "learning_rate": 1.30468770042774e-05, + "loss": 1.1739, + "step": 17440 + }, + { + "epoch": 0.42, + "learning_rate": 1.3046135135268883e-05, + "loss": 0.8866, + "step": 17441 + }, + { + "epoch": 0.42, + "learning_rate": 1.3045393247780176e-05, + "loss": 1.1707, + "step": 17442 + }, + { + "epoch": 0.42, + "learning_rate": 1.3044651341815785e-05, + "loss": 1.1613, + "step": 17443 + }, + { + "epoch": 0.42, + "learning_rate": 1.3043909417380213e-05, + "loss": 1.0328, + "step": 17444 + }, + { + "epoch": 0.42, + "learning_rate": 1.3043167474477957e-05, + "loss": 1.1515, + "step": 17445 + }, + { + "epoch": 0.42, + "learning_rate": 1.304242551311352e-05, + "loss": 1.2475, + "step": 17446 + }, + { + "epoch": 0.42, + "learning_rate": 1.3041683533291404e-05, + "loss": 1.0238, + "step": 17447 + }, + { + "epoch": 0.42, + "learning_rate": 1.304094153501611e-05, + "loss": 0.9793, + "step": 17448 + }, + { + "epoch": 0.42, + "learning_rate": 1.3040199518292139e-05, + "loss": 0.915, + "step": 17449 + }, + { + "epoch": 0.42, + "learning_rate": 1.303945748312399e-05, + "loss": 1.1735, + "step": 17450 + }, + { + "epoch": 0.42, + "learning_rate": 1.303871542951617e-05, + "loss": 1.1732, + "step": 17451 + }, + { + "epoch": 0.42, + "learning_rate": 1.3037973357473175e-05, + "loss": 1.2199, + "step": 17452 + }, + { + "epoch": 0.42, + "learning_rate": 1.3037231266999512e-05, + "loss": 1.1191, + "step": 17453 + }, + { + "epoch": 0.42, + "learning_rate": 1.3036489158099682e-05, + "loss": 1.1476, + "step": 17454 + }, + { + "epoch": 0.42, + "learning_rate": 1.3035747030778185e-05, + "loss": 1.1945, + "step": 17455 + }, + { + "epoch": 0.42, + "learning_rate": 1.3035004885039526e-05, + "loss": 0.821, + "step": 17456 + }, + { + "epoch": 0.42, + "learning_rate": 1.3034262720888207e-05, + "loss": 1.1419, + "step": 17457 + }, + { + "epoch": 0.42, + "learning_rate": 1.3033520538328728e-05, + "loss": 1.0077, + "step": 17458 + }, + { + "epoch": 0.42, + "learning_rate": 1.3032778337365593e-05, + "loss": 1.0541, + "step": 17459 + }, + { + "epoch": 0.42, + "learning_rate": 1.303203611800331e-05, + "loss": 0.9554, + "step": 17460 + }, + { + "epoch": 0.42, + "learning_rate": 1.3031293880246372e-05, + "loss": 1.2124, + "step": 17461 + }, + { + "epoch": 0.42, + "learning_rate": 1.3030551624099288e-05, + "loss": 1.0756, + "step": 17462 + }, + { + "epoch": 0.42, + "learning_rate": 1.302980934956656e-05, + "loss": 1.0365, + "step": 17463 + }, + { + "epoch": 0.42, + "learning_rate": 1.302906705665269e-05, + "loss": 1.0106, + "step": 17464 + }, + { + "epoch": 0.42, + "learning_rate": 1.3028324745362186e-05, + "loss": 0.9801, + "step": 17465 + }, + { + "epoch": 0.42, + "learning_rate": 1.3027582415699547e-05, + "loss": 1.1868, + "step": 17466 + }, + { + "epoch": 0.42, + "learning_rate": 1.3026840067669277e-05, + "loss": 1.2359, + "step": 17467 + }, + { + "epoch": 0.42, + "learning_rate": 1.3026097701275881e-05, + "loss": 1.2376, + "step": 17468 + }, + { + "epoch": 0.42, + "learning_rate": 1.3025355316523862e-05, + "loss": 0.742, + "step": 17469 + }, + { + "epoch": 0.42, + "learning_rate": 1.3024612913417725e-05, + "loss": 1.0519, + "step": 17470 + }, + { + "epoch": 0.42, + "learning_rate": 1.3023870491961969e-05, + "loss": 1.0079, + "step": 17471 + }, + { + "epoch": 0.42, + "learning_rate": 1.3023128052161104e-05, + "loss": 1.1728, + "step": 17472 + }, + { + "epoch": 0.42, + "learning_rate": 1.3022385594019633e-05, + "loss": 1.1674, + "step": 17473 + }, + { + "epoch": 0.42, + "learning_rate": 1.302164311754206e-05, + "loss": 1.2384, + "step": 17474 + }, + { + "epoch": 0.42, + "learning_rate": 1.3020900622732886e-05, + "loss": 1.2129, + "step": 17475 + }, + { + "epoch": 0.42, + "learning_rate": 1.3020158109596622e-05, + "loss": 1.1163, + "step": 17476 + }, + { + "epoch": 0.42, + "learning_rate": 1.3019415578137768e-05, + "loss": 1.1668, + "step": 17477 + }, + { + "epoch": 0.42, + "learning_rate": 1.3018673028360829e-05, + "loss": 0.8046, + "step": 17478 + }, + { + "epoch": 0.42, + "learning_rate": 1.3017930460270312e-05, + "loss": 1.1959, + "step": 17479 + }, + { + "epoch": 0.42, + "learning_rate": 1.3017187873870721e-05, + "loss": 1.1346, + "step": 17480 + }, + { + "epoch": 0.42, + "learning_rate": 1.3016445269166558e-05, + "loss": 0.8796, + "step": 17481 + }, + { + "epoch": 0.42, + "learning_rate": 1.3015702646162333e-05, + "loss": 1.1048, + "step": 17482 + }, + { + "epoch": 0.42, + "learning_rate": 1.3014960004862549e-05, + "loss": 0.8859, + "step": 17483 + }, + { + "epoch": 0.42, + "learning_rate": 1.3014217345271713e-05, + "loss": 0.9263, + "step": 17484 + }, + { + "epoch": 0.42, + "learning_rate": 1.3013474667394327e-05, + "loss": 1.2405, + "step": 17485 + }, + { + "epoch": 0.42, + "learning_rate": 1.3012731971234903e-05, + "loss": 1.0903, + "step": 17486 + }, + { + "epoch": 0.42, + "learning_rate": 1.3011989256797938e-05, + "loss": 1.2767, + "step": 17487 + }, + { + "epoch": 0.42, + "learning_rate": 1.3011246524087946e-05, + "loss": 1.3224, + "step": 17488 + }, + { + "epoch": 0.42, + "learning_rate": 1.301050377310943e-05, + "loss": 1.2412, + "step": 17489 + }, + { + "epoch": 0.42, + "learning_rate": 1.3009761003866892e-05, + "loss": 0.8892, + "step": 17490 + }, + { + "epoch": 0.42, + "learning_rate": 1.3009018216364844e-05, + "loss": 1.2685, + "step": 17491 + }, + { + "epoch": 0.42, + "learning_rate": 1.3008275410607792e-05, + "loss": 0.9685, + "step": 17492 + }, + { + "epoch": 0.42, + "learning_rate": 1.300753258660024e-05, + "loss": 1.1519, + "step": 17493 + }, + { + "epoch": 0.42, + "learning_rate": 1.3006789744346693e-05, + "loss": 0.8634, + "step": 17494 + }, + { + "epoch": 0.42, + "learning_rate": 1.3006046883851662e-05, + "loss": 1.2596, + "step": 17495 + }, + { + "epoch": 0.42, + "learning_rate": 1.300530400511965e-05, + "loss": 0.8088, + "step": 17496 + }, + { + "epoch": 0.42, + "learning_rate": 1.3004561108155166e-05, + "loss": 1.137, + "step": 17497 + }, + { + "epoch": 0.42, + "learning_rate": 1.3003818192962716e-05, + "loss": 0.9767, + "step": 17498 + }, + { + "epoch": 0.42, + "learning_rate": 1.3003075259546808e-05, + "loss": 1.04, + "step": 17499 + }, + { + "epoch": 0.42, + "learning_rate": 1.3002332307911952e-05, + "loss": 1.2512, + "step": 17500 + }, + { + "epoch": 0.42, + "learning_rate": 1.300158933806265e-05, + "loss": 1.0227, + "step": 17501 + }, + { + "epoch": 0.42, + "learning_rate": 1.300084635000341e-05, + "loss": 1.204, + "step": 17502 + }, + { + "epoch": 0.42, + "learning_rate": 1.300010334373874e-05, + "loss": 0.9572, + "step": 17503 + }, + { + "epoch": 0.42, + "learning_rate": 1.299936031927315e-05, + "loss": 1.1728, + "step": 17504 + }, + { + "epoch": 0.42, + "learning_rate": 1.299861727661115e-05, + "loss": 1.0987, + "step": 17505 + }, + { + "epoch": 0.42, + "learning_rate": 1.2997874215757237e-05, + "loss": 1.2788, + "step": 17506 + }, + { + "epoch": 0.42, + "learning_rate": 1.2997131136715933e-05, + "loss": 0.9228, + "step": 17507 + }, + { + "epoch": 0.42, + "learning_rate": 1.2996388039491736e-05, + "loss": 1.1515, + "step": 17508 + }, + { + "epoch": 0.42, + "learning_rate": 1.299564492408916e-05, + "loss": 0.8904, + "step": 17509 + }, + { + "epoch": 0.42, + "learning_rate": 1.2994901790512708e-05, + "loss": 1.137, + "step": 17510 + }, + { + "epoch": 0.42, + "learning_rate": 1.2994158638766892e-05, + "loss": 1.1643, + "step": 17511 + }, + { + "epoch": 0.42, + "learning_rate": 1.299341546885622e-05, + "loss": 1.2271, + "step": 17512 + }, + { + "epoch": 0.42, + "learning_rate": 1.2992672280785202e-05, + "loss": 1.0793, + "step": 17513 + }, + { + "epoch": 0.42, + "learning_rate": 1.2991929074558343e-05, + "loss": 1.366, + "step": 17514 + }, + { + "epoch": 0.42, + "learning_rate": 1.2991185850180155e-05, + "loss": 1.0705, + "step": 17515 + }, + { + "epoch": 0.42, + "learning_rate": 1.2990442607655146e-05, + "loss": 1.0793, + "step": 17516 + }, + { + "epoch": 0.42, + "learning_rate": 1.2989699346987825e-05, + "loss": 1.275, + "step": 17517 + }, + { + "epoch": 0.42, + "learning_rate": 1.2988956068182702e-05, + "loss": 1.2267, + "step": 17518 + }, + { + "epoch": 0.42, + "learning_rate": 1.2988212771244285e-05, + "loss": 1.2081, + "step": 17519 + }, + { + "epoch": 0.42, + "learning_rate": 1.2987469456177084e-05, + "loss": 1.0718, + "step": 17520 + }, + { + "epoch": 0.42, + "learning_rate": 1.2986726122985606e-05, + "loss": 1.3033, + "step": 17521 + }, + { + "epoch": 0.42, + "learning_rate": 1.2985982771674369e-05, + "loss": 1.1809, + "step": 17522 + }, + { + "epoch": 0.42, + "learning_rate": 1.2985239402247873e-05, + "loss": 1.1568, + "step": 17523 + }, + { + "epoch": 0.42, + "learning_rate": 1.2984496014710633e-05, + "loss": 1.1526, + "step": 17524 + }, + { + "epoch": 0.42, + "learning_rate": 1.2983752609067155e-05, + "loss": 1.0966, + "step": 17525 + }, + { + "epoch": 0.42, + "learning_rate": 1.2983009185321953e-05, + "loss": 1.2811, + "step": 17526 + }, + { + "epoch": 0.42, + "learning_rate": 1.2982265743479538e-05, + "loss": 1.1066, + "step": 17527 + }, + { + "epoch": 0.42, + "learning_rate": 1.2981522283544416e-05, + "loss": 0.9163, + "step": 17528 + }, + { + "epoch": 0.42, + "learning_rate": 1.2980778805521098e-05, + "loss": 1.1257, + "step": 17529 + }, + { + "epoch": 0.42, + "learning_rate": 1.29800353094141e-05, + "loss": 1.2257, + "step": 17530 + }, + { + "epoch": 0.42, + "learning_rate": 1.2979291795227928e-05, + "loss": 1.102, + "step": 17531 + }, + { + "epoch": 0.42, + "learning_rate": 1.2978548262967095e-05, + "loss": 0.7808, + "step": 17532 + }, + { + "epoch": 0.42, + "learning_rate": 1.2977804712636104e-05, + "loss": 1.1309, + "step": 17533 + }, + { + "epoch": 0.42, + "learning_rate": 1.2977061144239479e-05, + "loss": 0.9476, + "step": 17534 + }, + { + "epoch": 0.42, + "learning_rate": 1.297631755778172e-05, + "loss": 1.0951, + "step": 17535 + }, + { + "epoch": 0.42, + "learning_rate": 1.2975573953267345e-05, + "loss": 0.9329, + "step": 17536 + }, + { + "epoch": 0.42, + "learning_rate": 1.297483033070086e-05, + "loss": 1.1258, + "step": 17537 + }, + { + "epoch": 0.42, + "learning_rate": 1.297408669008678e-05, + "loss": 1.0403, + "step": 17538 + }, + { + "epoch": 0.42, + "learning_rate": 1.2973343031429618e-05, + "loss": 1.2998, + "step": 17539 + }, + { + "epoch": 0.42, + "learning_rate": 1.2972599354733879e-05, + "loss": 1.1881, + "step": 17540 + }, + { + "epoch": 0.42, + "learning_rate": 1.2971855660004082e-05, + "loss": 1.0752, + "step": 17541 + }, + { + "epoch": 0.42, + "learning_rate": 1.2971111947244735e-05, + "loss": 1.1623, + "step": 17542 + }, + { + "epoch": 0.42, + "learning_rate": 1.2970368216460352e-05, + "loss": 1.2748, + "step": 17543 + }, + { + "epoch": 0.42, + "learning_rate": 1.2969624467655441e-05, + "loss": 1.0344, + "step": 17544 + }, + { + "epoch": 0.42, + "learning_rate": 1.296888070083452e-05, + "loss": 0.9818, + "step": 17545 + }, + { + "epoch": 0.42, + "learning_rate": 1.2968136916002094e-05, + "loss": 1.2813, + "step": 17546 + }, + { + "epoch": 0.42, + "learning_rate": 1.2967393113162683e-05, + "loss": 1.2062, + "step": 17547 + }, + { + "epoch": 0.42, + "learning_rate": 1.2966649292320797e-05, + "loss": 1.231, + "step": 17548 + }, + { + "epoch": 0.42, + "learning_rate": 1.2965905453480944e-05, + "loss": 1.2169, + "step": 17549 + }, + { + "epoch": 0.42, + "learning_rate": 1.2965161596647639e-05, + "loss": 1.1835, + "step": 17550 + }, + { + "epoch": 0.42, + "learning_rate": 1.29644177218254e-05, + "loss": 1.3075, + "step": 17551 + }, + { + "epoch": 0.42, + "learning_rate": 1.2963673829018735e-05, + "loss": 1.1202, + "step": 17552 + }, + { + "epoch": 0.42, + "learning_rate": 1.2962929918232159e-05, + "loss": 1.2846, + "step": 17553 + }, + { + "epoch": 0.42, + "learning_rate": 1.2962185989470182e-05, + "loss": 1.2326, + "step": 17554 + }, + { + "epoch": 0.42, + "learning_rate": 1.296144204273732e-05, + "loss": 1.0817, + "step": 17555 + }, + { + "epoch": 0.42, + "learning_rate": 1.2960698078038085e-05, + "loss": 1.1324, + "step": 17556 + }, + { + "epoch": 0.42, + "learning_rate": 1.2959954095376994e-05, + "loss": 1.0774, + "step": 17557 + }, + { + "epoch": 0.42, + "learning_rate": 1.2959210094758554e-05, + "loss": 1.1863, + "step": 17558 + }, + { + "epoch": 0.42, + "learning_rate": 1.2958466076187283e-05, + "loss": 1.1123, + "step": 17559 + }, + { + "epoch": 0.42, + "learning_rate": 1.2957722039667698e-05, + "loss": 0.9873, + "step": 17560 + }, + { + "epoch": 0.42, + "learning_rate": 1.2956977985204304e-05, + "loss": 1.0547, + "step": 17561 + }, + { + "epoch": 0.42, + "learning_rate": 1.2956233912801623e-05, + "loss": 0.9067, + "step": 17562 + }, + { + "epoch": 0.42, + "learning_rate": 1.2955489822464167e-05, + "loss": 1.0852, + "step": 17563 + }, + { + "epoch": 0.42, + "learning_rate": 1.2954745714196448e-05, + "loss": 1.1349, + "step": 17564 + }, + { + "epoch": 0.42, + "learning_rate": 1.295400158800298e-05, + "loss": 1.1239, + "step": 17565 + }, + { + "epoch": 0.42, + "learning_rate": 1.2953257443888281e-05, + "loss": 0.9985, + "step": 17566 + }, + { + "epoch": 0.42, + "learning_rate": 1.2952513281856865e-05, + "loss": 1.2718, + "step": 17567 + }, + { + "epoch": 0.42, + "learning_rate": 1.2951769101913245e-05, + "loss": 1.224, + "step": 17568 + }, + { + "epoch": 0.42, + "learning_rate": 1.2951024904061933e-05, + "loss": 1.1414, + "step": 17569 + }, + { + "epoch": 0.42, + "learning_rate": 1.295028068830745e-05, + "loss": 0.9811, + "step": 17570 + }, + { + "epoch": 0.42, + "learning_rate": 1.2949536454654308e-05, + "loss": 0.6253, + "step": 17571 + }, + { + "epoch": 0.42, + "learning_rate": 1.2948792203107022e-05, + "loss": 1.0449, + "step": 17572 + }, + { + "epoch": 0.42, + "learning_rate": 1.2948047933670105e-05, + "loss": 1.2332, + "step": 17573 + }, + { + "epoch": 0.42, + "learning_rate": 1.2947303646348078e-05, + "loss": 1.3627, + "step": 17574 + }, + { + "epoch": 0.42, + "learning_rate": 1.294655934114545e-05, + "loss": 1.1299, + "step": 17575 + }, + { + "epoch": 0.42, + "learning_rate": 1.2945815018066743e-05, + "loss": 1.153, + "step": 17576 + }, + { + "epoch": 0.42, + "learning_rate": 1.2945070677116465e-05, + "loss": 1.3433, + "step": 17577 + }, + { + "epoch": 0.42, + "learning_rate": 1.2944326318299139e-05, + "loss": 0.899, + "step": 17578 + }, + { + "epoch": 0.42, + "learning_rate": 1.2943581941619279e-05, + "loss": 1.1536, + "step": 17579 + }, + { + "epoch": 0.42, + "learning_rate": 1.2942837547081396e-05, + "loss": 1.3154, + "step": 17580 + }, + { + "epoch": 0.42, + "learning_rate": 1.294209313469001e-05, + "loss": 1.1468, + "step": 17581 + }, + { + "epoch": 0.42, + "learning_rate": 1.2941348704449639e-05, + "loss": 0.8952, + "step": 17582 + }, + { + "epoch": 0.42, + "learning_rate": 1.2940604256364796e-05, + "loss": 1.2448, + "step": 17583 + }, + { + "epoch": 0.42, + "learning_rate": 1.2939859790440002e-05, + "loss": 1.1183, + "step": 17584 + }, + { + "epoch": 0.42, + "learning_rate": 1.2939115306679767e-05, + "loss": 1.3545, + "step": 17585 + }, + { + "epoch": 0.42, + "learning_rate": 1.2938370805088608e-05, + "loss": 1.2304, + "step": 17586 + }, + { + "epoch": 0.42, + "learning_rate": 1.2937626285671049e-05, + "loss": 1.0182, + "step": 17587 + }, + { + "epoch": 0.42, + "learning_rate": 1.2936881748431601e-05, + "loss": 1.1065, + "step": 17588 + }, + { + "epoch": 0.42, + "learning_rate": 1.2936137193374782e-05, + "loss": 1.0399, + "step": 17589 + }, + { + "epoch": 0.42, + "learning_rate": 1.293539262050511e-05, + "loss": 1.0634, + "step": 17590 + }, + { + "epoch": 0.42, + "learning_rate": 1.29346480298271e-05, + "loss": 1.161, + "step": 17591 + }, + { + "epoch": 0.42, + "learning_rate": 1.2933903421345272e-05, + "loss": 1.1532, + "step": 17592 + }, + { + "epoch": 0.42, + "learning_rate": 1.2933158795064142e-05, + "loss": 0.9469, + "step": 17593 + }, + { + "epoch": 0.42, + "learning_rate": 1.2932414150988226e-05, + "loss": 1.2823, + "step": 17594 + }, + { + "epoch": 0.42, + "learning_rate": 1.2931669489122046e-05, + "loss": 1.0313, + "step": 17595 + }, + { + "epoch": 0.42, + "learning_rate": 1.2930924809470115e-05, + "loss": 1.0825, + "step": 17596 + }, + { + "epoch": 0.42, + "learning_rate": 1.2930180112036952e-05, + "loss": 1.155, + "step": 17597 + }, + { + "epoch": 0.42, + "learning_rate": 1.2929435396827076e-05, + "loss": 0.9352, + "step": 17598 + }, + { + "epoch": 0.42, + "learning_rate": 1.2928690663845004e-05, + "loss": 1.1718, + "step": 17599 + }, + { + "epoch": 0.42, + "learning_rate": 1.2927945913095256e-05, + "loss": 1.1582, + "step": 17600 + }, + { + "epoch": 0.42, + "learning_rate": 1.2927201144582346e-05, + "loss": 1.0987, + "step": 17601 + }, + { + "epoch": 0.42, + "learning_rate": 1.2926456358310797e-05, + "loss": 1.1399, + "step": 17602 + }, + { + "epoch": 0.42, + "learning_rate": 1.2925711554285127e-05, + "loss": 1.2035, + "step": 17603 + }, + { + "epoch": 0.42, + "learning_rate": 1.2924966732509854e-05, + "loss": 1.2336, + "step": 17604 + }, + { + "epoch": 0.42, + "learning_rate": 1.2924221892989492e-05, + "loss": 1.1456, + "step": 17605 + }, + { + "epoch": 0.42, + "learning_rate": 1.2923477035728564e-05, + "loss": 0.9338, + "step": 17606 + }, + { + "epoch": 0.42, + "learning_rate": 1.2922732160731593e-05, + "loss": 1.2006, + "step": 17607 + }, + { + "epoch": 0.42, + "learning_rate": 1.2921987268003087e-05, + "loss": 0.9844, + "step": 17608 + }, + { + "epoch": 0.42, + "learning_rate": 1.2921242357547575e-05, + "loss": 1.3119, + "step": 17609 + }, + { + "epoch": 0.42, + "learning_rate": 1.2920497429369574e-05, + "loss": 1.293, + "step": 17610 + }, + { + "epoch": 0.42, + "learning_rate": 1.2919752483473599e-05, + "loss": 1.0173, + "step": 17611 + }, + { + "epoch": 0.42, + "learning_rate": 1.2919007519864173e-05, + "loss": 1.2941, + "step": 17612 + }, + { + "epoch": 0.42, + "learning_rate": 1.2918262538545819e-05, + "loss": 0.9936, + "step": 17613 + }, + { + "epoch": 0.42, + "learning_rate": 1.2917517539523048e-05, + "loss": 0.7974, + "step": 17614 + }, + { + "epoch": 0.42, + "learning_rate": 1.2916772522800387e-05, + "loss": 1.1626, + "step": 17615 + }, + { + "epoch": 0.42, + "learning_rate": 1.291602748838235e-05, + "loss": 1.1728, + "step": 17616 + }, + { + "epoch": 0.42, + "learning_rate": 1.2915282436273465e-05, + "loss": 1.0698, + "step": 17617 + }, + { + "epoch": 0.42, + "learning_rate": 1.2914537366478243e-05, + "loss": 1.1104, + "step": 17618 + }, + { + "epoch": 0.42, + "learning_rate": 1.291379227900121e-05, + "loss": 1.1104, + "step": 17619 + }, + { + "epoch": 0.42, + "learning_rate": 1.2913047173846884e-05, + "loss": 1.0661, + "step": 17620 + }, + { + "epoch": 0.42, + "learning_rate": 1.2912302051019787e-05, + "loss": 1.1754, + "step": 17621 + }, + { + "epoch": 0.42, + "learning_rate": 1.2911556910524436e-05, + "loss": 1.1363, + "step": 17622 + }, + { + "epoch": 0.42, + "learning_rate": 1.2910811752365357e-05, + "loss": 0.8824, + "step": 17623 + }, + { + "epoch": 0.42, + "learning_rate": 1.2910066576547067e-05, + "loss": 0.8439, + "step": 17624 + }, + { + "epoch": 0.42, + "learning_rate": 1.2909321383074085e-05, + "loss": 1.1491, + "step": 17625 + }, + { + "epoch": 0.42, + "learning_rate": 1.2908576171950937e-05, + "loss": 1.0907, + "step": 17626 + }, + { + "epoch": 0.42, + "learning_rate": 1.290783094318214e-05, + "loss": 1.1176, + "step": 17627 + }, + { + "epoch": 0.42, + "learning_rate": 1.2907085696772219e-05, + "loss": 1.0544, + "step": 17628 + }, + { + "epoch": 0.42, + "learning_rate": 1.290634043272569e-05, + "loss": 0.9636, + "step": 17629 + }, + { + "epoch": 0.42, + "learning_rate": 1.290559515104708e-05, + "loss": 1.2026, + "step": 17630 + }, + { + "epoch": 0.42, + "learning_rate": 1.2904849851740907e-05, + "loss": 1.0487, + "step": 17631 + }, + { + "epoch": 0.42, + "learning_rate": 1.2904104534811692e-05, + "loss": 1.1245, + "step": 17632 + }, + { + "epoch": 0.42, + "learning_rate": 1.2903359200263958e-05, + "loss": 1.0311, + "step": 17633 + }, + { + "epoch": 0.42, + "learning_rate": 1.2902613848102228e-05, + "loss": 1.0261, + "step": 17634 + }, + { + "epoch": 0.42, + "learning_rate": 1.2901868478331023e-05, + "loss": 1.2157, + "step": 17635 + }, + { + "epoch": 0.42, + "learning_rate": 1.2901123090954864e-05, + "loss": 0.9842, + "step": 17636 + }, + { + "epoch": 0.42, + "learning_rate": 1.2900377685978272e-05, + "loss": 1.1234, + "step": 17637 + }, + { + "epoch": 0.42, + "learning_rate": 1.289963226340577e-05, + "loss": 0.9702, + "step": 17638 + }, + { + "epoch": 0.42, + "learning_rate": 1.2898886823241883e-05, + "loss": 0.9592, + "step": 17639 + }, + { + "epoch": 0.42, + "learning_rate": 1.2898141365491132e-05, + "loss": 1.0361, + "step": 17640 + }, + { + "epoch": 0.42, + "learning_rate": 1.2897395890158038e-05, + "loss": 0.787, + "step": 17641 + }, + { + "epoch": 0.42, + "learning_rate": 1.2896650397247125e-05, + "loss": 1.0058, + "step": 17642 + }, + { + "epoch": 0.42, + "learning_rate": 1.2895904886762915e-05, + "loss": 1.1932, + "step": 17643 + }, + { + "epoch": 0.42, + "learning_rate": 1.2895159358709933e-05, + "loss": 1.2768, + "step": 17644 + }, + { + "epoch": 0.42, + "learning_rate": 1.2894413813092697e-05, + "loss": 1.0326, + "step": 17645 + }, + { + "epoch": 0.42, + "learning_rate": 1.2893668249915737e-05, + "loss": 0.8678, + "step": 17646 + }, + { + "epoch": 0.42, + "learning_rate": 1.2892922669183569e-05, + "loss": 1.0957, + "step": 17647 + }, + { + "epoch": 0.42, + "learning_rate": 1.289217707090072e-05, + "loss": 1.245, + "step": 17648 + }, + { + "epoch": 0.42, + "learning_rate": 1.2891431455071715e-05, + "loss": 1.3061, + "step": 17649 + }, + { + "epoch": 0.42, + "learning_rate": 1.2890685821701074e-05, + "loss": 1.022, + "step": 17650 + }, + { + "epoch": 0.42, + "learning_rate": 1.2889940170793322e-05, + "loss": 1.0522, + "step": 17651 + }, + { + "epoch": 0.42, + "learning_rate": 1.2889194502352981e-05, + "loss": 1.0321, + "step": 17652 + }, + { + "epoch": 0.42, + "learning_rate": 1.288844881638458e-05, + "loss": 0.8867, + "step": 17653 + }, + { + "epoch": 0.42, + "learning_rate": 1.2887703112892638e-05, + "loss": 1.1615, + "step": 17654 + }, + { + "epoch": 0.42, + "learning_rate": 1.2886957391881683e-05, + "loss": 1.0379, + "step": 17655 + }, + { + "epoch": 0.42, + "learning_rate": 1.288621165335623e-05, + "loss": 1.0957, + "step": 17656 + }, + { + "epoch": 0.42, + "learning_rate": 1.2885465897320817e-05, + "loss": 1.2042, + "step": 17657 + }, + { + "epoch": 0.42, + "learning_rate": 1.2884720123779957e-05, + "loss": 1.153, + "step": 17658 + }, + { + "epoch": 0.42, + "learning_rate": 1.288397433273818e-05, + "loss": 1.3699, + "step": 17659 + }, + { + "epoch": 0.42, + "learning_rate": 1.2883228524200006e-05, + "loss": 0.9003, + "step": 17660 + }, + { + "epoch": 0.42, + "learning_rate": 1.2882482698169966e-05, + "loss": 1.3686, + "step": 17661 + }, + { + "epoch": 0.42, + "learning_rate": 1.2881736854652582e-05, + "loss": 1.1645, + "step": 17662 + }, + { + "epoch": 0.42, + "learning_rate": 1.2880990993652379e-05, + "loss": 0.9924, + "step": 17663 + }, + { + "epoch": 0.42, + "learning_rate": 1.2880245115173877e-05, + "loss": 0.971, + "step": 17664 + }, + { + "epoch": 0.42, + "learning_rate": 1.2879499219221607e-05, + "loss": 1.2055, + "step": 17665 + }, + { + "epoch": 0.42, + "learning_rate": 1.2878753305800094e-05, + "loss": 1.0575, + "step": 17666 + }, + { + "epoch": 0.42, + "learning_rate": 1.287800737491386e-05, + "loss": 1.3194, + "step": 17667 + }, + { + "epoch": 0.42, + "learning_rate": 1.2877261426567433e-05, + "loss": 1.1482, + "step": 17668 + }, + { + "epoch": 0.42, + "learning_rate": 1.287651546076534e-05, + "loss": 0.9813, + "step": 17669 + }, + { + "epoch": 0.42, + "learning_rate": 1.2875769477512101e-05, + "loss": 1.1589, + "step": 17670 + }, + { + "epoch": 0.42, + "learning_rate": 1.2875023476812247e-05, + "loss": 1.0991, + "step": 17671 + }, + { + "epoch": 0.42, + "learning_rate": 1.2874277458670301e-05, + "loss": 1.1134, + "step": 17672 + }, + { + "epoch": 0.43, + "learning_rate": 1.287353142309079e-05, + "loss": 1.2906, + "step": 17673 + }, + { + "epoch": 0.43, + "learning_rate": 1.287278537007824e-05, + "loss": 1.3205, + "step": 17674 + }, + { + "epoch": 0.43, + "learning_rate": 1.2872039299637178e-05, + "loss": 1.0137, + "step": 17675 + }, + { + "epoch": 0.43, + "learning_rate": 1.2871293211772124e-05, + "loss": 1.1885, + "step": 17676 + }, + { + "epoch": 0.43, + "learning_rate": 1.2870547106487613e-05, + "loss": 1.1877, + "step": 17677 + }, + { + "epoch": 0.43, + "learning_rate": 1.286980098378817e-05, + "loss": 1.2007, + "step": 17678 + }, + { + "epoch": 0.43, + "learning_rate": 1.2869054843678317e-05, + "loss": 1.0831, + "step": 17679 + }, + { + "epoch": 0.43, + "learning_rate": 1.2868308686162583e-05, + "loss": 1.2942, + "step": 17680 + }, + { + "epoch": 0.43, + "learning_rate": 1.2867562511245497e-05, + "loss": 1.2364, + "step": 17681 + }, + { + "epoch": 0.43, + "learning_rate": 1.2866816318931582e-05, + "loss": 1.0118, + "step": 17682 + }, + { + "epoch": 0.43, + "learning_rate": 1.2866070109225368e-05, + "loss": 1.1381, + "step": 17683 + }, + { + "epoch": 0.43, + "learning_rate": 1.2865323882131377e-05, + "loss": 1.1979, + "step": 17684 + }, + { + "epoch": 0.43, + "learning_rate": 1.2864577637654144e-05, + "loss": 1.1605, + "step": 17685 + }, + { + "epoch": 0.43, + "learning_rate": 1.286383137579819e-05, + "loss": 1.0735, + "step": 17686 + }, + { + "epoch": 0.43, + "learning_rate": 1.2863085096568047e-05, + "loss": 1.0396, + "step": 17687 + }, + { + "epoch": 0.43, + "learning_rate": 1.2862338799968237e-05, + "loss": 1.2483, + "step": 17688 + }, + { + "epoch": 0.43, + "learning_rate": 1.2861592486003294e-05, + "loss": 1.1789, + "step": 17689 + }, + { + "epoch": 0.43, + "learning_rate": 1.286084615467774e-05, + "loss": 1.0533, + "step": 17690 + }, + { + "epoch": 0.43, + "learning_rate": 1.2860099805996106e-05, + "loss": 1.0066, + "step": 17691 + }, + { + "epoch": 0.43, + "learning_rate": 1.2859353439962918e-05, + "loss": 1.0282, + "step": 17692 + }, + { + "epoch": 0.43, + "learning_rate": 1.2858607056582709e-05, + "loss": 1.131, + "step": 17693 + }, + { + "epoch": 0.43, + "learning_rate": 1.285786065586e-05, + "loss": 1.0996, + "step": 17694 + }, + { + "epoch": 0.43, + "learning_rate": 1.2857114237799322e-05, + "loss": 1.1774, + "step": 17695 + }, + { + "epoch": 0.43, + "learning_rate": 1.2856367802405205e-05, + "loss": 0.9246, + "step": 17696 + }, + { + "epoch": 0.43, + "learning_rate": 1.2855621349682179e-05, + "loss": 1.0826, + "step": 17697 + }, + { + "epoch": 0.43, + "learning_rate": 1.285487487963477e-05, + "loss": 1.1698, + "step": 17698 + }, + { + "epoch": 0.43, + "learning_rate": 1.2854128392267501e-05, + "loss": 0.9803, + "step": 17699 + }, + { + "epoch": 0.43, + "learning_rate": 1.285338188758491e-05, + "loss": 1.1148, + "step": 17700 + }, + { + "epoch": 0.43, + "learning_rate": 1.2852635365591524e-05, + "loss": 0.9847, + "step": 17701 + }, + { + "epoch": 0.43, + "learning_rate": 1.285188882629187e-05, + "loss": 0.9398, + "step": 17702 + }, + { + "epoch": 0.43, + "learning_rate": 1.2851142269690474e-05, + "loss": 1.318, + "step": 17703 + }, + { + "epoch": 0.43, + "learning_rate": 1.2850395695791868e-05, + "loss": 0.5867, + "step": 17704 + }, + { + "epoch": 0.43, + "learning_rate": 1.2849649104600585e-05, + "loss": 1.1805, + "step": 17705 + }, + { + "epoch": 0.43, + "learning_rate": 1.2848902496121152e-05, + "loss": 1.1023, + "step": 17706 + }, + { + "epoch": 0.43, + "learning_rate": 1.2848155870358094e-05, + "loss": 1.2766, + "step": 17707 + }, + { + "epoch": 0.43, + "learning_rate": 1.2847409227315943e-05, + "loss": 1.2074, + "step": 17708 + }, + { + "epoch": 0.43, + "learning_rate": 1.2846662566999233e-05, + "loss": 1.3214, + "step": 17709 + }, + { + "epoch": 0.43, + "learning_rate": 1.2845915889412492e-05, + "loss": 1.0295, + "step": 17710 + }, + { + "epoch": 0.43, + "learning_rate": 1.2845169194560248e-05, + "loss": 1.2308, + "step": 17711 + }, + { + "epoch": 0.43, + "learning_rate": 1.284442248244703e-05, + "loss": 1.2073, + "step": 17712 + }, + { + "epoch": 0.43, + "learning_rate": 1.2843675753077371e-05, + "loss": 0.9681, + "step": 17713 + }, + { + "epoch": 0.43, + "learning_rate": 1.2842929006455802e-05, + "loss": 1.0606, + "step": 17714 + }, + { + "epoch": 0.43, + "learning_rate": 1.284218224258685e-05, + "loss": 1.1109, + "step": 17715 + }, + { + "epoch": 0.43, + "learning_rate": 1.2841435461475045e-05, + "loss": 0.9204, + "step": 17716 + }, + { + "epoch": 0.43, + "learning_rate": 1.2840688663124921e-05, + "loss": 1.1098, + "step": 17717 + }, + { + "epoch": 0.43, + "learning_rate": 1.2839941847541007e-05, + "loss": 1.1575, + "step": 17718 + }, + { + "epoch": 0.43, + "learning_rate": 1.2839195014727835e-05, + "loss": 1.1463, + "step": 17719 + }, + { + "epoch": 0.43, + "learning_rate": 1.2838448164689932e-05, + "loss": 0.9726, + "step": 17720 + }, + { + "epoch": 0.43, + "learning_rate": 1.2837701297431833e-05, + "loss": 0.9151, + "step": 17721 + }, + { + "epoch": 0.43, + "learning_rate": 1.2836954412958067e-05, + "loss": 1.1476, + "step": 17722 + }, + { + "epoch": 0.43, + "learning_rate": 1.2836207511273166e-05, + "loss": 1.0838, + "step": 17723 + }, + { + "epoch": 0.43, + "learning_rate": 1.2835460592381662e-05, + "loss": 0.9652, + "step": 17724 + }, + { + "epoch": 0.43, + "learning_rate": 1.2834713656288085e-05, + "loss": 1.0382, + "step": 17725 + }, + { + "epoch": 0.43, + "learning_rate": 1.2833966702996968e-05, + "loss": 1.2467, + "step": 17726 + }, + { + "epoch": 0.43, + "learning_rate": 1.283321973251284e-05, + "loss": 0.9844, + "step": 17727 + }, + { + "epoch": 0.43, + "learning_rate": 1.2832472744840234e-05, + "loss": 1.1964, + "step": 17728 + }, + { + "epoch": 0.43, + "learning_rate": 1.2831725739983685e-05, + "loss": 1.3022, + "step": 17729 + }, + { + "epoch": 0.43, + "learning_rate": 1.283097871794772e-05, + "loss": 0.798, + "step": 17730 + }, + { + "epoch": 0.43, + "learning_rate": 1.2830231678736873e-05, + "loss": 0.8765, + "step": 17731 + }, + { + "epoch": 0.43, + "learning_rate": 1.2829484622355675e-05, + "loss": 0.7638, + "step": 17732 + }, + { + "epoch": 0.43, + "learning_rate": 1.2828737548808661e-05, + "loss": 1.1911, + "step": 17733 + }, + { + "epoch": 0.43, + "learning_rate": 1.282799045810036e-05, + "loss": 1.1128, + "step": 17734 + }, + { + "epoch": 0.43, + "learning_rate": 1.2827243350235307e-05, + "loss": 1.1203, + "step": 17735 + }, + { + "epoch": 0.43, + "learning_rate": 1.2826496225218035e-05, + "loss": 1.3163, + "step": 17736 + }, + { + "epoch": 0.43, + "learning_rate": 1.2825749083053074e-05, + "loss": 1.0274, + "step": 17737 + }, + { + "epoch": 0.43, + "learning_rate": 1.282500192374496e-05, + "loss": 1.1912, + "step": 17738 + }, + { + "epoch": 0.43, + "learning_rate": 1.2824254747298219e-05, + "loss": 1.2115, + "step": 17739 + }, + { + "epoch": 0.43, + "learning_rate": 1.2823507553717394e-05, + "loss": 0.7621, + "step": 17740 + }, + { + "epoch": 0.43, + "learning_rate": 1.2822760343007011e-05, + "loss": 1.1523, + "step": 17741 + }, + { + "epoch": 0.43, + "learning_rate": 1.2822013115171603e-05, + "loss": 1.0366, + "step": 17742 + }, + { + "epoch": 0.43, + "learning_rate": 1.2821265870215707e-05, + "loss": 1.1522, + "step": 17743 + }, + { + "epoch": 0.43, + "learning_rate": 1.2820518608143855e-05, + "loss": 1.246, + "step": 17744 + }, + { + "epoch": 0.43, + "learning_rate": 1.281977132896058e-05, + "loss": 0.9423, + "step": 17745 + }, + { + "epoch": 0.43, + "learning_rate": 1.2819024032670416e-05, + "loss": 1.0624, + "step": 17746 + }, + { + "epoch": 0.43, + "learning_rate": 1.2818276719277893e-05, + "loss": 1.1865, + "step": 17747 + }, + { + "epoch": 0.43, + "learning_rate": 1.2817529388787548e-05, + "loss": 1.0885, + "step": 17748 + }, + { + "epoch": 0.43, + "learning_rate": 1.281678204120392e-05, + "loss": 1.0515, + "step": 17749 + }, + { + "epoch": 0.43, + "learning_rate": 1.2816034676531534e-05, + "loss": 1.0583, + "step": 17750 + }, + { + "epoch": 0.43, + "learning_rate": 1.281528729477493e-05, + "loss": 0.9711, + "step": 17751 + }, + { + "epoch": 0.43, + "learning_rate": 1.2814539895938638e-05, + "loss": 1.0305, + "step": 17752 + }, + { + "epoch": 0.43, + "learning_rate": 1.2813792480027196e-05, + "loss": 0.9307, + "step": 17753 + }, + { + "epoch": 0.43, + "learning_rate": 1.2813045047045139e-05, + "loss": 1.271, + "step": 17754 + }, + { + "epoch": 0.43, + "learning_rate": 1.2812297596996995e-05, + "loss": 0.9994, + "step": 17755 + }, + { + "epoch": 0.43, + "learning_rate": 1.2811550129887304e-05, + "loss": 1.2631, + "step": 17756 + }, + { + "epoch": 0.43, + "learning_rate": 1.2810802645720602e-05, + "loss": 1.043, + "step": 17757 + }, + { + "epoch": 0.43, + "learning_rate": 1.2810055144501422e-05, + "loss": 1.2358, + "step": 17758 + }, + { + "epoch": 0.43, + "learning_rate": 1.2809307626234293e-05, + "loss": 1.0105, + "step": 17759 + }, + { + "epoch": 0.43, + "learning_rate": 1.280856009092376e-05, + "loss": 0.9933, + "step": 17760 + }, + { + "epoch": 0.43, + "learning_rate": 1.2807812538574352e-05, + "loss": 1.0334, + "step": 17761 + }, + { + "epoch": 0.43, + "learning_rate": 1.2807064969190605e-05, + "loss": 0.9316, + "step": 17762 + }, + { + "epoch": 0.43, + "learning_rate": 1.2806317382777056e-05, + "loss": 1.1778, + "step": 17763 + }, + { + "epoch": 0.43, + "learning_rate": 1.280556977933824e-05, + "loss": 1.0845, + "step": 17764 + }, + { + "epoch": 0.43, + "learning_rate": 1.2804822158878693e-05, + "loss": 1.4412, + "step": 17765 + }, + { + "epoch": 0.43, + "learning_rate": 1.2804074521402949e-05, + "loss": 1.3497, + "step": 17766 + }, + { + "epoch": 0.43, + "learning_rate": 1.2803326866915542e-05, + "loss": 0.8568, + "step": 17767 + }, + { + "epoch": 0.43, + "learning_rate": 1.2802579195421014e-05, + "loss": 1.0863, + "step": 17768 + }, + { + "epoch": 0.43, + "learning_rate": 1.2801831506923894e-05, + "loss": 0.9649, + "step": 17769 + }, + { + "epoch": 0.43, + "learning_rate": 1.2801083801428724e-05, + "loss": 1.2618, + "step": 17770 + }, + { + "epoch": 0.43, + "learning_rate": 1.2800336078940034e-05, + "loss": 1.2305, + "step": 17771 + }, + { + "epoch": 0.43, + "learning_rate": 1.2799588339462368e-05, + "loss": 1.1927, + "step": 17772 + }, + { + "epoch": 0.43, + "learning_rate": 1.2798840583000256e-05, + "loss": 1.0779, + "step": 17773 + }, + { + "epoch": 0.43, + "learning_rate": 1.2798092809558238e-05, + "loss": 1.0271, + "step": 17774 + }, + { + "epoch": 0.43, + "learning_rate": 1.2797345019140846e-05, + "loss": 1.1584, + "step": 17775 + }, + { + "epoch": 0.43, + "learning_rate": 1.279659721175262e-05, + "loss": 1.2228, + "step": 17776 + }, + { + "epoch": 0.43, + "learning_rate": 1.2795849387398099e-05, + "loss": 1.2413, + "step": 17777 + }, + { + "epoch": 0.43, + "learning_rate": 1.2795101546081814e-05, + "loss": 1.2448, + "step": 17778 + }, + { + "epoch": 0.43, + "learning_rate": 1.2794353687808308e-05, + "loss": 1.1378, + "step": 17779 + }, + { + "epoch": 0.43, + "learning_rate": 1.2793605812582114e-05, + "loss": 1.1249, + "step": 17780 + }, + { + "epoch": 0.43, + "learning_rate": 1.2792857920407772e-05, + "loss": 0.8679, + "step": 17781 + }, + { + "epoch": 0.43, + "learning_rate": 1.2792110011289817e-05, + "loss": 1.1461, + "step": 17782 + }, + { + "epoch": 0.43, + "learning_rate": 1.279136208523279e-05, + "loss": 0.9373, + "step": 17783 + }, + { + "epoch": 0.43, + "learning_rate": 1.2790614142241224e-05, + "loss": 0.97, + "step": 17784 + }, + { + "epoch": 0.43, + "learning_rate": 1.2789866182319659e-05, + "loss": 1.4637, + "step": 17785 + }, + { + "epoch": 0.43, + "learning_rate": 1.2789118205472627e-05, + "loss": 1.1335, + "step": 17786 + }, + { + "epoch": 0.43, + "learning_rate": 1.2788370211704678e-05, + "loss": 1.1402, + "step": 17787 + }, + { + "epoch": 0.43, + "learning_rate": 1.278762220102034e-05, + "loss": 1.2985, + "step": 17788 + }, + { + "epoch": 0.43, + "learning_rate": 1.2786874173424153e-05, + "loss": 1.0392, + "step": 17789 + }, + { + "epoch": 0.43, + "learning_rate": 1.2786126128920655e-05, + "loss": 1.0375, + "step": 17790 + }, + { + "epoch": 0.43, + "learning_rate": 1.2785378067514387e-05, + "loss": 1.3739, + "step": 17791 + }, + { + "epoch": 0.43, + "learning_rate": 1.2784629989209884e-05, + "loss": 1.1526, + "step": 17792 + }, + { + "epoch": 0.43, + "learning_rate": 1.2783881894011688e-05, + "loss": 1.3936, + "step": 17793 + }, + { + "epoch": 0.43, + "learning_rate": 1.2783133781924335e-05, + "loss": 0.8875, + "step": 17794 + }, + { + "epoch": 0.43, + "learning_rate": 1.278238565295236e-05, + "loss": 1.2677, + "step": 17795 + }, + { + "epoch": 0.43, + "learning_rate": 1.278163750710031e-05, + "loss": 0.9903, + "step": 17796 + }, + { + "epoch": 0.43, + "learning_rate": 1.278088934437272e-05, + "loss": 1.1397, + "step": 17797 + }, + { + "epoch": 0.43, + "learning_rate": 1.2780141164774128e-05, + "loss": 1.0922, + "step": 17798 + }, + { + "epoch": 0.43, + "learning_rate": 1.277939296830907e-05, + "loss": 1.2448, + "step": 17799 + }, + { + "epoch": 0.43, + "learning_rate": 1.2778644754982091e-05, + "loss": 1.2376, + "step": 17800 + }, + { + "epoch": 0.43, + "learning_rate": 1.2777896524797729e-05, + "loss": 1.2068, + "step": 17801 + }, + { + "epoch": 0.43, + "learning_rate": 1.2777148277760521e-05, + "loss": 1.0959, + "step": 17802 + }, + { + "epoch": 0.43, + "learning_rate": 1.2776400013875006e-05, + "loss": 1.1384, + "step": 17803 + }, + { + "epoch": 0.43, + "learning_rate": 1.2775651733145727e-05, + "loss": 1.1468, + "step": 17804 + }, + { + "epoch": 0.43, + "learning_rate": 1.2774903435577222e-05, + "loss": 1.2569, + "step": 17805 + }, + { + "epoch": 0.43, + "learning_rate": 1.2774155121174032e-05, + "loss": 1.0754, + "step": 17806 + }, + { + "epoch": 0.43, + "learning_rate": 1.2773406789940692e-05, + "loss": 1.0948, + "step": 17807 + }, + { + "epoch": 0.43, + "learning_rate": 1.2772658441881746e-05, + "loss": 1.1512, + "step": 17808 + }, + { + "epoch": 0.43, + "learning_rate": 1.2771910077001737e-05, + "loss": 0.9512, + "step": 17809 + }, + { + "epoch": 0.43, + "learning_rate": 1.2771161695305198e-05, + "loss": 1.1143, + "step": 17810 + }, + { + "epoch": 0.43, + "learning_rate": 1.2770413296796673e-05, + "loss": 1.0433, + "step": 17811 + }, + { + "epoch": 0.43, + "learning_rate": 1.2769664881480704e-05, + "loss": 1.017, + "step": 17812 + }, + { + "epoch": 0.43, + "learning_rate": 1.2768916449361828e-05, + "loss": 1.0413, + "step": 17813 + }, + { + "epoch": 0.43, + "learning_rate": 1.2768168000444587e-05, + "loss": 1.16, + "step": 17814 + }, + { + "epoch": 0.43, + "learning_rate": 1.2767419534733524e-05, + "loss": 1.0227, + "step": 17815 + }, + { + "epoch": 0.43, + "learning_rate": 1.2766671052233174e-05, + "loss": 1.1902, + "step": 17816 + }, + { + "epoch": 0.43, + "learning_rate": 1.2765922552948085e-05, + "loss": 1.1041, + "step": 17817 + }, + { + "epoch": 0.43, + "learning_rate": 1.2765174036882791e-05, + "loss": 0.7851, + "step": 17818 + }, + { + "epoch": 0.43, + "learning_rate": 1.276442550404184e-05, + "loss": 1.0623, + "step": 17819 + }, + { + "epoch": 0.43, + "learning_rate": 1.2763676954429766e-05, + "loss": 1.12, + "step": 17820 + }, + { + "epoch": 0.43, + "learning_rate": 1.2762928388051119e-05, + "loss": 1.1946, + "step": 17821 + }, + { + "epoch": 0.43, + "learning_rate": 1.2762179804910431e-05, + "loss": 1.0165, + "step": 17822 + }, + { + "epoch": 0.43, + "learning_rate": 1.2761431205012249e-05, + "loss": 0.9768, + "step": 17823 + }, + { + "epoch": 0.43, + "learning_rate": 1.2760682588361115e-05, + "loss": 1.1832, + "step": 17824 + }, + { + "epoch": 0.43, + "learning_rate": 1.275993395496157e-05, + "loss": 1.0866, + "step": 17825 + }, + { + "epoch": 0.43, + "learning_rate": 1.2759185304818149e-05, + "loss": 1.3311, + "step": 17826 + }, + { + "epoch": 0.43, + "learning_rate": 1.2758436637935405e-05, + "loss": 1.0568, + "step": 17827 + }, + { + "epoch": 0.43, + "learning_rate": 1.2757687954317874e-05, + "loss": 1.2283, + "step": 17828 + }, + { + "epoch": 0.43, + "learning_rate": 1.2756939253970098e-05, + "loss": 1.2491, + "step": 17829 + }, + { + "epoch": 0.43, + "learning_rate": 1.275619053689662e-05, + "loss": 1.0727, + "step": 17830 + }, + { + "epoch": 0.43, + "learning_rate": 1.2755441803101982e-05, + "loss": 1.2098, + "step": 17831 + }, + { + "epoch": 0.43, + "learning_rate": 1.2754693052590728e-05, + "loss": 0.7534, + "step": 17832 + }, + { + "epoch": 0.43, + "learning_rate": 1.27539442853674e-05, + "loss": 1.0789, + "step": 17833 + }, + { + "epoch": 0.43, + "learning_rate": 1.2753195501436535e-05, + "loss": 1.1074, + "step": 17834 + }, + { + "epoch": 0.43, + "learning_rate": 1.2752446700802684e-05, + "loss": 1.0816, + "step": 17835 + }, + { + "epoch": 0.43, + "learning_rate": 1.2751697883470386e-05, + "loss": 1.0072, + "step": 17836 + }, + { + "epoch": 0.43, + "learning_rate": 1.2750949049444185e-05, + "loss": 1.3757, + "step": 17837 + }, + { + "epoch": 0.43, + "learning_rate": 1.2750200198728622e-05, + "loss": 1.1851, + "step": 17838 + }, + { + "epoch": 0.43, + "learning_rate": 1.2749451331328242e-05, + "loss": 1.0584, + "step": 17839 + }, + { + "epoch": 0.43, + "learning_rate": 1.2748702447247587e-05, + "loss": 1.0196, + "step": 17840 + }, + { + "epoch": 0.43, + "learning_rate": 1.27479535464912e-05, + "loss": 0.8836, + "step": 17841 + }, + { + "epoch": 0.43, + "learning_rate": 1.2747204629063626e-05, + "loss": 1.1261, + "step": 17842 + }, + { + "epoch": 0.43, + "learning_rate": 1.274645569496941e-05, + "loss": 1.1679, + "step": 17843 + }, + { + "epoch": 0.43, + "learning_rate": 1.2745706744213089e-05, + "loss": 1.0267, + "step": 17844 + }, + { + "epoch": 0.43, + "learning_rate": 1.2744957776799215e-05, + "loss": 1.0557, + "step": 17845 + }, + { + "epoch": 0.43, + "learning_rate": 1.2744208792732324e-05, + "loss": 0.8363, + "step": 17846 + }, + { + "epoch": 0.43, + "learning_rate": 1.2743459792016965e-05, + "loss": 1.0637, + "step": 17847 + }, + { + "epoch": 0.43, + "learning_rate": 1.2742710774657683e-05, + "loss": 1.1756, + "step": 17848 + }, + { + "epoch": 0.43, + "learning_rate": 1.2741961740659019e-05, + "loss": 1.1036, + "step": 17849 + }, + { + "epoch": 0.43, + "learning_rate": 1.2741212690025514e-05, + "loss": 1.1933, + "step": 17850 + }, + { + "epoch": 0.43, + "learning_rate": 1.2740463622761721e-05, + "loss": 1.0746, + "step": 17851 + }, + { + "epoch": 0.43, + "learning_rate": 1.2739714538872179e-05, + "loss": 1.3592, + "step": 17852 + }, + { + "epoch": 0.43, + "learning_rate": 1.2738965438361435e-05, + "loss": 1.1946, + "step": 17853 + }, + { + "epoch": 0.43, + "learning_rate": 1.2738216321234026e-05, + "loss": 1.3341, + "step": 17854 + }, + { + "epoch": 0.43, + "learning_rate": 1.2737467187494508e-05, + "loss": 1.0998, + "step": 17855 + }, + { + "epoch": 0.43, + "learning_rate": 1.273671803714742e-05, + "loss": 1.2258, + "step": 17856 + }, + { + "epoch": 0.43, + "learning_rate": 1.2735968870197307e-05, + "loss": 1.0739, + "step": 17857 + }, + { + "epoch": 0.43, + "learning_rate": 1.273521968664871e-05, + "loss": 1.2313, + "step": 17858 + }, + { + "epoch": 0.43, + "learning_rate": 1.2734470486506183e-05, + "loss": 0.8174, + "step": 17859 + }, + { + "epoch": 0.43, + "learning_rate": 1.2733721269774267e-05, + "loss": 1.2702, + "step": 17860 + }, + { + "epoch": 0.43, + "learning_rate": 1.2732972036457504e-05, + "loss": 1.23, + "step": 17861 + }, + { + "epoch": 0.43, + "learning_rate": 1.2732222786560446e-05, + "loss": 1.1059, + "step": 17862 + }, + { + "epoch": 0.43, + "learning_rate": 1.2731473520087635e-05, + "loss": 0.9921, + "step": 17863 + }, + { + "epoch": 0.43, + "learning_rate": 1.2730724237043614e-05, + "loss": 0.9856, + "step": 17864 + }, + { + "epoch": 0.43, + "learning_rate": 1.2729974937432932e-05, + "loss": 1.1477, + "step": 17865 + }, + { + "epoch": 0.43, + "learning_rate": 1.2729225621260134e-05, + "loss": 1.003, + "step": 17866 + }, + { + "epoch": 0.43, + "learning_rate": 1.2728476288529767e-05, + "loss": 1.4693, + "step": 17867 + }, + { + "epoch": 0.43, + "learning_rate": 1.2727726939246376e-05, + "loss": 1.2789, + "step": 17868 + }, + { + "epoch": 0.43, + "learning_rate": 1.2726977573414506e-05, + "loss": 1.2314, + "step": 17869 + }, + { + "epoch": 0.43, + "learning_rate": 1.2726228191038705e-05, + "loss": 1.079, + "step": 17870 + }, + { + "epoch": 0.43, + "learning_rate": 1.272547879212352e-05, + "loss": 1.1604, + "step": 17871 + }, + { + "epoch": 0.43, + "learning_rate": 1.2724729376673496e-05, + "loss": 1.2235, + "step": 17872 + }, + { + "epoch": 0.43, + "learning_rate": 1.272397994469318e-05, + "loss": 0.8182, + "step": 17873 + }, + { + "epoch": 0.43, + "learning_rate": 1.2723230496187116e-05, + "loss": 0.8952, + "step": 17874 + }, + { + "epoch": 0.43, + "learning_rate": 1.2722481031159854e-05, + "loss": 1.2424, + "step": 17875 + }, + { + "epoch": 0.43, + "learning_rate": 1.2721731549615943e-05, + "loss": 1.1672, + "step": 17876 + }, + { + "epoch": 0.43, + "learning_rate": 1.2720982051559925e-05, + "loss": 1.2427, + "step": 17877 + }, + { + "epoch": 0.43, + "learning_rate": 1.2720232536996347e-05, + "loss": 0.9708, + "step": 17878 + }, + { + "epoch": 0.43, + "learning_rate": 1.271948300592976e-05, + "loss": 0.9918, + "step": 17879 + }, + { + "epoch": 0.43, + "learning_rate": 1.271873345836471e-05, + "loss": 1.0637, + "step": 17880 + }, + { + "epoch": 0.43, + "learning_rate": 1.271798389430574e-05, + "loss": 1.3372, + "step": 17881 + }, + { + "epoch": 0.43, + "learning_rate": 1.2717234313757404e-05, + "loss": 0.9608, + "step": 17882 + }, + { + "epoch": 0.43, + "learning_rate": 1.2716484716724247e-05, + "loss": 0.8353, + "step": 17883 + }, + { + "epoch": 0.43, + "learning_rate": 1.2715735103210816e-05, + "loss": 1.0573, + "step": 17884 + }, + { + "epoch": 0.43, + "learning_rate": 1.271498547322166e-05, + "loss": 0.8846, + "step": 17885 + }, + { + "epoch": 0.43, + "learning_rate": 1.2714235826761321e-05, + "loss": 1.1458, + "step": 17886 + }, + { + "epoch": 0.43, + "learning_rate": 1.2713486163834354e-05, + "loss": 1.1562, + "step": 17887 + }, + { + "epoch": 0.43, + "learning_rate": 1.2712736484445306e-05, + "loss": 1.0102, + "step": 17888 + }, + { + "epoch": 0.43, + "learning_rate": 1.2711986788598724e-05, + "loss": 1.1663, + "step": 17889 + }, + { + "epoch": 0.43, + "learning_rate": 1.2711237076299155e-05, + "loss": 1.0368, + "step": 17890 + }, + { + "epoch": 0.43, + "learning_rate": 1.2710487347551148e-05, + "loss": 0.9538, + "step": 17891 + }, + { + "epoch": 0.43, + "learning_rate": 1.2709737602359253e-05, + "loss": 1.2312, + "step": 17892 + }, + { + "epoch": 0.43, + "learning_rate": 1.2708987840728018e-05, + "loss": 1.3266, + "step": 17893 + }, + { + "epoch": 0.43, + "learning_rate": 1.2708238062661988e-05, + "loss": 1.0393, + "step": 17894 + }, + { + "epoch": 0.43, + "learning_rate": 1.2707488268165715e-05, + "loss": 1.1288, + "step": 17895 + }, + { + "epoch": 0.43, + "learning_rate": 1.270673845724375e-05, + "loss": 1.1148, + "step": 17896 + }, + { + "epoch": 0.43, + "learning_rate": 1.2705988629900641e-05, + "loss": 1.169, + "step": 17897 + }, + { + "epoch": 0.43, + "learning_rate": 1.270523878614093e-05, + "loss": 1.0426, + "step": 17898 + }, + { + "epoch": 0.43, + "learning_rate": 1.2704488925969175e-05, + "loss": 0.7915, + "step": 17899 + }, + { + "epoch": 0.43, + "learning_rate": 1.270373904938992e-05, + "loss": 0.8898, + "step": 17900 + }, + { + "epoch": 0.43, + "learning_rate": 1.2702989156407719e-05, + "loss": 1.2194, + "step": 17901 + }, + { + "epoch": 0.43, + "learning_rate": 1.2702239247027115e-05, + "loss": 1.1568, + "step": 17902 + }, + { + "epoch": 0.43, + "learning_rate": 1.2701489321252665e-05, + "loss": 1.1319, + "step": 17903 + }, + { + "epoch": 0.43, + "learning_rate": 1.2700739379088913e-05, + "loss": 0.8555, + "step": 17904 + }, + { + "epoch": 0.43, + "learning_rate": 1.2699989420540408e-05, + "loss": 1.1637, + "step": 17905 + }, + { + "epoch": 0.43, + "learning_rate": 1.2699239445611706e-05, + "loss": 1.191, + "step": 17906 + }, + { + "epoch": 0.43, + "learning_rate": 1.2698489454307352e-05, + "loss": 1.2327, + "step": 17907 + }, + { + "epoch": 0.43, + "learning_rate": 1.2697739446631898e-05, + "loss": 1.1117, + "step": 17908 + }, + { + "epoch": 0.43, + "learning_rate": 1.269698942258989e-05, + "loss": 1.3624, + "step": 17909 + }, + { + "epoch": 0.43, + "learning_rate": 1.2696239382185884e-05, + "loss": 1.1452, + "step": 17910 + }, + { + "epoch": 0.43, + "learning_rate": 1.2695489325424428e-05, + "loss": 1.0916, + "step": 17911 + }, + { + "epoch": 0.43, + "learning_rate": 1.269473925231007e-05, + "loss": 0.8384, + "step": 17912 + }, + { + "epoch": 0.43, + "learning_rate": 1.2693989162847365e-05, + "loss": 1.3032, + "step": 17913 + }, + { + "epoch": 0.43, + "learning_rate": 1.2693239057040863e-05, + "loss": 1.062, + "step": 17914 + }, + { + "epoch": 0.43, + "learning_rate": 1.269248893489511e-05, + "loss": 1.144, + "step": 17915 + }, + { + "epoch": 0.43, + "learning_rate": 1.2691738796414663e-05, + "loss": 1.2518, + "step": 17916 + }, + { + "epoch": 0.43, + "learning_rate": 1.2690988641604066e-05, + "loss": 1.1279, + "step": 17917 + }, + { + "epoch": 0.43, + "learning_rate": 1.2690238470467877e-05, + "loss": 0.7508, + "step": 17918 + }, + { + "epoch": 0.43, + "learning_rate": 1.2689488283010643e-05, + "loss": 1.2306, + "step": 17919 + }, + { + "epoch": 0.43, + "learning_rate": 1.2688738079236917e-05, + "loss": 1.0607, + "step": 17920 + }, + { + "epoch": 0.43, + "learning_rate": 1.2687987859151245e-05, + "loss": 1.2839, + "step": 17921 + }, + { + "epoch": 0.43, + "learning_rate": 1.2687237622758188e-05, + "loss": 0.9571, + "step": 17922 + }, + { + "epoch": 0.43, + "learning_rate": 1.2686487370062292e-05, + "loss": 0.8671, + "step": 17923 + }, + { + "epoch": 0.43, + "learning_rate": 1.2685737101068108e-05, + "loss": 1.086, + "step": 17924 + }, + { + "epoch": 0.43, + "learning_rate": 1.2684986815780187e-05, + "loss": 1.2168, + "step": 17925 + }, + { + "epoch": 0.43, + "learning_rate": 1.2684236514203085e-05, + "loss": 1.077, + "step": 17926 + }, + { + "epoch": 0.43, + "learning_rate": 1.2683486196341351e-05, + "loss": 1.1556, + "step": 17927 + }, + { + "epoch": 0.43, + "learning_rate": 1.2682735862199539e-05, + "loss": 1.305, + "step": 17928 + }, + { + "epoch": 0.43, + "learning_rate": 1.2681985511782196e-05, + "loss": 1.1914, + "step": 17929 + }, + { + "epoch": 0.43, + "learning_rate": 1.2681235145093878e-05, + "loss": 1.1203, + "step": 17930 + }, + { + "epoch": 0.43, + "learning_rate": 1.268048476213914e-05, + "loss": 0.9652, + "step": 17931 + }, + { + "epoch": 0.43, + "learning_rate": 1.267973436292253e-05, + "loss": 0.965, + "step": 17932 + }, + { + "epoch": 0.43, + "learning_rate": 1.26789839474486e-05, + "loss": 1.1595, + "step": 17933 + }, + { + "epoch": 0.43, + "learning_rate": 1.2678233515721907e-05, + "loss": 1.1224, + "step": 17934 + }, + { + "epoch": 0.43, + "learning_rate": 1.2677483067747e-05, + "loss": 0.92, + "step": 17935 + }, + { + "epoch": 0.43, + "learning_rate": 1.2676732603528435e-05, + "loss": 1.3033, + "step": 17936 + }, + { + "epoch": 0.43, + "learning_rate": 1.2675982123070761e-05, + "loss": 1.0429, + "step": 17937 + }, + { + "epoch": 0.43, + "learning_rate": 1.2675231626378532e-05, + "loss": 1.0732, + "step": 17938 + }, + { + "epoch": 0.43, + "learning_rate": 1.2674481113456304e-05, + "loss": 1.0895, + "step": 17939 + }, + { + "epoch": 0.43, + "learning_rate": 1.2673730584308629e-05, + "loss": 1.3507, + "step": 17940 + }, + { + "epoch": 0.43, + "learning_rate": 1.2672980038940055e-05, + "loss": 0.9688, + "step": 17941 + }, + { + "epoch": 0.43, + "learning_rate": 1.2672229477355144e-05, + "loss": 0.956, + "step": 17942 + }, + { + "epoch": 0.43, + "learning_rate": 1.2671478899558445e-05, + "loss": 1.1278, + "step": 17943 + }, + { + "epoch": 0.43, + "learning_rate": 1.2670728305554509e-05, + "loss": 1.156, + "step": 17944 + }, + { + "epoch": 0.43, + "learning_rate": 1.2669977695347894e-05, + "loss": 1.0157, + "step": 17945 + }, + { + "epoch": 0.43, + "learning_rate": 1.2669227068943154e-05, + "loss": 0.9494, + "step": 17946 + }, + { + "epoch": 0.43, + "learning_rate": 1.266847642634484e-05, + "loss": 1.1949, + "step": 17947 + }, + { + "epoch": 0.43, + "learning_rate": 1.2667725767557508e-05, + "loss": 0.9651, + "step": 17948 + }, + { + "epoch": 0.43, + "learning_rate": 1.266697509258571e-05, + "loss": 1.0843, + "step": 17949 + }, + { + "epoch": 0.43, + "learning_rate": 1.2666224401434003e-05, + "loss": 1.1974, + "step": 17950 + }, + { + "epoch": 0.43, + "learning_rate": 1.2665473694106938e-05, + "loss": 1.1186, + "step": 17951 + }, + { + "epoch": 0.43, + "learning_rate": 1.2664722970609069e-05, + "loss": 1.2811, + "step": 17952 + }, + { + "epoch": 0.43, + "learning_rate": 1.2663972230944957e-05, + "loss": 1.192, + "step": 17953 + }, + { + "epoch": 0.43, + "learning_rate": 1.2663221475119149e-05, + "loss": 1.1587, + "step": 17954 + }, + { + "epoch": 0.43, + "learning_rate": 1.2662470703136204e-05, + "loss": 1.1492, + "step": 17955 + }, + { + "epoch": 0.43, + "learning_rate": 1.2661719915000675e-05, + "loss": 1.2231, + "step": 17956 + }, + { + "epoch": 0.43, + "learning_rate": 1.2660969110717116e-05, + "loss": 1.0626, + "step": 17957 + }, + { + "epoch": 0.43, + "learning_rate": 1.2660218290290085e-05, + "loss": 1.1588, + "step": 17958 + }, + { + "epoch": 0.43, + "learning_rate": 1.2659467453724136e-05, + "loss": 1.3353, + "step": 17959 + }, + { + "epoch": 0.43, + "learning_rate": 1.2658716601023824e-05, + "loss": 0.9616, + "step": 17960 + }, + { + "epoch": 0.43, + "learning_rate": 1.2657965732193702e-05, + "loss": 1.134, + "step": 17961 + }, + { + "epoch": 0.43, + "learning_rate": 1.2657214847238329e-05, + "loss": 1.2088, + "step": 17962 + }, + { + "epoch": 0.43, + "learning_rate": 1.2656463946162257e-05, + "loss": 0.7719, + "step": 17963 + }, + { + "epoch": 0.43, + "learning_rate": 1.2655713028970045e-05, + "loss": 1.0854, + "step": 17964 + }, + { + "epoch": 0.43, + "learning_rate": 1.2654962095666243e-05, + "loss": 0.8867, + "step": 17965 + }, + { + "epoch": 0.43, + "learning_rate": 1.2654211146255412e-05, + "loss": 1.1091, + "step": 17966 + }, + { + "epoch": 0.43, + "learning_rate": 1.2653460180742107e-05, + "loss": 1.0324, + "step": 17967 + }, + { + "epoch": 0.43, + "learning_rate": 1.2652709199130884e-05, + "loss": 0.9066, + "step": 17968 + }, + { + "epoch": 0.43, + "learning_rate": 1.2651958201426297e-05, + "loss": 0.938, + "step": 17969 + }, + { + "epoch": 0.43, + "learning_rate": 1.2651207187632903e-05, + "loss": 1.1343, + "step": 17970 + }, + { + "epoch": 0.43, + "learning_rate": 1.265045615775526e-05, + "loss": 0.9126, + "step": 17971 + }, + { + "epoch": 0.43, + "learning_rate": 1.2649705111797923e-05, + "loss": 1.1324, + "step": 17972 + }, + { + "epoch": 0.43, + "learning_rate": 1.2648954049765445e-05, + "loss": 0.8556, + "step": 17973 + }, + { + "epoch": 0.43, + "learning_rate": 1.2648202971662391e-05, + "loss": 1.1281, + "step": 17974 + }, + { + "epoch": 0.43, + "learning_rate": 1.2647451877493309e-05, + "loss": 0.8546, + "step": 17975 + }, + { + "epoch": 0.43, + "learning_rate": 1.2646700767262761e-05, + "loss": 1.087, + "step": 17976 + }, + { + "epoch": 0.43, + "learning_rate": 1.2645949640975299e-05, + "loss": 1.0999, + "step": 17977 + }, + { + "epoch": 0.43, + "learning_rate": 1.2645198498635486e-05, + "loss": 1.1799, + "step": 17978 + }, + { + "epoch": 0.43, + "learning_rate": 1.2644447340247874e-05, + "loss": 1.1209, + "step": 17979 + }, + { + "epoch": 0.43, + "learning_rate": 1.2643696165817026e-05, + "loss": 1.2381, + "step": 17980 + }, + { + "epoch": 0.43, + "learning_rate": 1.264294497534749e-05, + "loss": 1.281, + "step": 17981 + }, + { + "epoch": 0.43, + "learning_rate": 1.264219376884383e-05, + "loss": 1.1903, + "step": 17982 + }, + { + "epoch": 0.43, + "learning_rate": 1.2641442546310603e-05, + "loss": 0.9584, + "step": 17983 + }, + { + "epoch": 0.43, + "learning_rate": 1.2640691307752367e-05, + "loss": 0.9586, + "step": 17984 + }, + { + "epoch": 0.43, + "learning_rate": 1.2639940053173674e-05, + "loss": 1.3635, + "step": 17985 + }, + { + "epoch": 0.43, + "learning_rate": 1.2639188782579086e-05, + "loss": 1.2113, + "step": 17986 + }, + { + "epoch": 0.43, + "learning_rate": 1.2638437495973163e-05, + "loss": 1.2475, + "step": 17987 + }, + { + "epoch": 0.43, + "learning_rate": 1.2637686193360459e-05, + "loss": 1.2648, + "step": 17988 + }, + { + "epoch": 0.43, + "learning_rate": 1.2636934874745531e-05, + "loss": 0.9582, + "step": 17989 + }, + { + "epoch": 0.43, + "learning_rate": 1.2636183540132945e-05, + "loss": 1.1173, + "step": 17990 + }, + { + "epoch": 0.43, + "learning_rate": 1.2635432189527247e-05, + "loss": 0.909, + "step": 17991 + }, + { + "epoch": 0.43, + "learning_rate": 1.2634680822933005e-05, + "loss": 0.9967, + "step": 17992 + }, + { + "epoch": 0.43, + "learning_rate": 1.2633929440354776e-05, + "loss": 1.1013, + "step": 17993 + }, + { + "epoch": 0.43, + "learning_rate": 1.2633178041797116e-05, + "loss": 1.3328, + "step": 17994 + }, + { + "epoch": 0.43, + "learning_rate": 1.2632426627264583e-05, + "loss": 0.995, + "step": 17995 + }, + { + "epoch": 0.43, + "learning_rate": 1.2631675196761735e-05, + "loss": 1.0109, + "step": 17996 + }, + { + "epoch": 0.43, + "learning_rate": 1.2630923750293134e-05, + "loss": 1.1399, + "step": 17997 + }, + { + "epoch": 0.43, + "learning_rate": 1.263017228786334e-05, + "loss": 1.3277, + "step": 17998 + }, + { + "epoch": 0.43, + "learning_rate": 1.2629420809476907e-05, + "loss": 1.2163, + "step": 17999 + }, + { + "epoch": 0.43, + "learning_rate": 1.2628669315138395e-05, + "loss": 1.1436, + "step": 18000 + }, + { + "epoch": 0.43, + "learning_rate": 1.2627917804852368e-05, + "loss": 1.215, + "step": 18001 + }, + { + "epoch": 0.43, + "learning_rate": 1.2627166278623379e-05, + "loss": 1.2512, + "step": 18002 + }, + { + "epoch": 0.43, + "learning_rate": 1.2626414736455993e-05, + "loss": 0.9314, + "step": 18003 + }, + { + "epoch": 0.43, + "learning_rate": 1.2625663178354763e-05, + "loss": 0.9019, + "step": 18004 + }, + { + "epoch": 0.43, + "learning_rate": 1.2624911604324255e-05, + "loss": 1.0369, + "step": 18005 + }, + { + "epoch": 0.43, + "learning_rate": 1.2624160014369025e-05, + "loss": 1.2364, + "step": 18006 + }, + { + "epoch": 0.43, + "learning_rate": 1.2623408408493633e-05, + "loss": 1.3496, + "step": 18007 + }, + { + "epoch": 0.43, + "learning_rate": 1.2622656786702638e-05, + "loss": 0.935, + "step": 18008 + }, + { + "epoch": 0.43, + "learning_rate": 1.2621905149000604e-05, + "loss": 1.103, + "step": 18009 + }, + { + "epoch": 0.43, + "learning_rate": 1.2621153495392086e-05, + "loss": 0.8508, + "step": 18010 + }, + { + "epoch": 0.43, + "learning_rate": 1.2620401825881647e-05, + "loss": 1.0011, + "step": 18011 + }, + { + "epoch": 0.43, + "learning_rate": 1.2619650140473845e-05, + "loss": 1.1131, + "step": 18012 + }, + { + "epoch": 0.43, + "learning_rate": 1.2618898439173244e-05, + "loss": 1.158, + "step": 18013 + }, + { + "epoch": 0.43, + "learning_rate": 1.26181467219844e-05, + "loss": 1.1454, + "step": 18014 + }, + { + "epoch": 0.43, + "learning_rate": 1.261739498891188e-05, + "loss": 1.2501, + "step": 18015 + }, + { + "epoch": 0.43, + "learning_rate": 1.2616643239960236e-05, + "loss": 0.9981, + "step": 18016 + }, + { + "epoch": 0.43, + "learning_rate": 1.2615891475134034e-05, + "loss": 1.0103, + "step": 18017 + }, + { + "epoch": 0.43, + "learning_rate": 1.2615139694437835e-05, + "loss": 1.1609, + "step": 18018 + }, + { + "epoch": 0.43, + "learning_rate": 1.2614387897876197e-05, + "loss": 1.0939, + "step": 18019 + }, + { + "epoch": 0.43, + "learning_rate": 1.2613636085453682e-05, + "loss": 1.1341, + "step": 18020 + }, + { + "epoch": 0.43, + "learning_rate": 1.2612884257174853e-05, + "loss": 1.1622, + "step": 18021 + }, + { + "epoch": 0.43, + "learning_rate": 1.261213241304427e-05, + "loss": 1.0791, + "step": 18022 + }, + { + "epoch": 0.43, + "learning_rate": 1.2611380553066495e-05, + "loss": 1.2822, + "step": 18023 + }, + { + "epoch": 0.43, + "learning_rate": 1.2610628677246083e-05, + "loss": 1.0819, + "step": 18024 + }, + { + "epoch": 0.43, + "learning_rate": 1.2609876785587607e-05, + "loss": 1.0623, + "step": 18025 + }, + { + "epoch": 0.43, + "learning_rate": 1.260912487809562e-05, + "loss": 1.0715, + "step": 18026 + }, + { + "epoch": 0.43, + "learning_rate": 1.2608372954774687e-05, + "loss": 1.1259, + "step": 18027 + }, + { + "epoch": 0.43, + "learning_rate": 1.2607621015629367e-05, + "loss": 1.0974, + "step": 18028 + }, + { + "epoch": 0.43, + "learning_rate": 1.2606869060664227e-05, + "loss": 1.206, + "step": 18029 + }, + { + "epoch": 0.43, + "learning_rate": 1.2606117089883822e-05, + "loss": 0.9358, + "step": 18030 + }, + { + "epoch": 0.43, + "learning_rate": 1.2605365103292719e-05, + "loss": 1.0368, + "step": 18031 + }, + { + "epoch": 0.43, + "learning_rate": 1.2604613100895478e-05, + "loss": 1.2967, + "step": 18032 + }, + { + "epoch": 0.43, + "learning_rate": 1.2603861082696662e-05, + "loss": 1.1088, + "step": 18033 + }, + { + "epoch": 0.43, + "learning_rate": 1.2603109048700835e-05, + "loss": 0.9781, + "step": 18034 + }, + { + "epoch": 0.43, + "learning_rate": 1.2602356998912558e-05, + "loss": 1.2829, + "step": 18035 + }, + { + "epoch": 0.43, + "learning_rate": 1.260160493333639e-05, + "loss": 0.9512, + "step": 18036 + }, + { + "epoch": 0.43, + "learning_rate": 1.26008528519769e-05, + "loss": 0.9843, + "step": 18037 + }, + { + "epoch": 0.43, + "learning_rate": 1.2600100754838647e-05, + "loss": 1.0971, + "step": 18038 + }, + { + "epoch": 0.43, + "learning_rate": 1.259934864192619e-05, + "loss": 1.1163, + "step": 18039 + }, + { + "epoch": 0.43, + "learning_rate": 1.2598596513244099e-05, + "loss": 0.9896, + "step": 18040 + }, + { + "epoch": 0.43, + "learning_rate": 1.2597844368796936e-05, + "loss": 0.9092, + "step": 18041 + }, + { + "epoch": 0.43, + "learning_rate": 1.2597092208589261e-05, + "loss": 1.1117, + "step": 18042 + }, + { + "epoch": 0.43, + "learning_rate": 1.2596340032625638e-05, + "loss": 1.1306, + "step": 18043 + }, + { + "epoch": 0.43, + "learning_rate": 1.259558784091063e-05, + "loss": 1.0753, + "step": 18044 + }, + { + "epoch": 0.43, + "learning_rate": 1.2594835633448803e-05, + "loss": 1.1284, + "step": 18045 + }, + { + "epoch": 0.43, + "learning_rate": 1.259408341024472e-05, + "loss": 1.1128, + "step": 18046 + }, + { + "epoch": 0.43, + "learning_rate": 1.259333117130294e-05, + "loss": 1.411, + "step": 18047 + }, + { + "epoch": 0.43, + "learning_rate": 1.259257891662803e-05, + "loss": 1.0184, + "step": 18048 + }, + { + "epoch": 0.43, + "learning_rate": 1.2591826646224554e-05, + "loss": 1.1032, + "step": 18049 + }, + { + "epoch": 0.43, + "learning_rate": 1.2591074360097078e-05, + "loss": 1.1326, + "step": 18050 + }, + { + "epoch": 0.43, + "learning_rate": 1.2590322058250162e-05, + "loss": 1.0389, + "step": 18051 + }, + { + "epoch": 0.43, + "learning_rate": 1.2589569740688368e-05, + "loss": 1.0846, + "step": 18052 + }, + { + "epoch": 0.43, + "learning_rate": 1.2588817407416265e-05, + "loss": 1.1192, + "step": 18053 + }, + { + "epoch": 0.43, + "learning_rate": 1.2588065058438419e-05, + "loss": 1.0756, + "step": 18054 + }, + { + "epoch": 0.43, + "learning_rate": 1.2587312693759388e-05, + "loss": 1.1013, + "step": 18055 + }, + { + "epoch": 0.43, + "learning_rate": 1.2586560313383738e-05, + "loss": 1.0447, + "step": 18056 + }, + { + "epoch": 0.43, + "learning_rate": 1.2585807917316038e-05, + "loss": 1.3228, + "step": 18057 + }, + { + "epoch": 0.43, + "learning_rate": 1.2585055505560848e-05, + "loss": 1.0629, + "step": 18058 + }, + { + "epoch": 0.43, + "learning_rate": 1.2584303078122737e-05, + "loss": 0.8623, + "step": 18059 + }, + { + "epoch": 0.43, + "learning_rate": 1.2583550635006263e-05, + "loss": 1.0503, + "step": 18060 + }, + { + "epoch": 0.43, + "learning_rate": 1.2582798176215998e-05, + "loss": 1.1378, + "step": 18061 + }, + { + "epoch": 0.43, + "learning_rate": 1.2582045701756503e-05, + "loss": 0.9233, + "step": 18062 + }, + { + "epoch": 0.43, + "learning_rate": 1.2581293211632344e-05, + "loss": 1.2712, + "step": 18063 + }, + { + "epoch": 0.43, + "learning_rate": 1.2580540705848085e-05, + "loss": 1.0935, + "step": 18064 + }, + { + "epoch": 0.43, + "learning_rate": 1.2579788184408295e-05, + "loss": 1.2796, + "step": 18065 + }, + { + "epoch": 0.43, + "learning_rate": 1.2579035647317537e-05, + "loss": 1.0735, + "step": 18066 + }, + { + "epoch": 0.43, + "learning_rate": 1.2578283094580374e-05, + "loss": 1.0249, + "step": 18067 + }, + { + "epoch": 0.43, + "learning_rate": 1.2577530526201373e-05, + "loss": 1.2209, + "step": 18068 + }, + { + "epoch": 0.43, + "learning_rate": 1.2576777942185103e-05, + "loss": 0.7953, + "step": 18069 + }, + { + "epoch": 0.43, + "learning_rate": 1.2576025342536129e-05, + "loss": 0.9514, + "step": 18070 + }, + { + "epoch": 0.43, + "learning_rate": 1.2575272727259012e-05, + "loss": 0.9985, + "step": 18071 + }, + { + "epoch": 0.43, + "learning_rate": 1.257452009635832e-05, + "loss": 0.97, + "step": 18072 + }, + { + "epoch": 0.43, + "learning_rate": 1.2573767449838623e-05, + "loss": 1.0176, + "step": 18073 + }, + { + "epoch": 0.43, + "learning_rate": 1.2573014787704484e-05, + "loss": 0.9128, + "step": 18074 + }, + { + "epoch": 0.43, + "learning_rate": 1.257226210996047e-05, + "loss": 1.1025, + "step": 18075 + }, + { + "epoch": 0.43, + "learning_rate": 1.2571509416611145e-05, + "loss": 1.1588, + "step": 18076 + }, + { + "epoch": 0.43, + "learning_rate": 1.2570756707661078e-05, + "loss": 1.1505, + "step": 18077 + }, + { + "epoch": 0.43, + "learning_rate": 1.2570003983114836e-05, + "loss": 0.9663, + "step": 18078 + }, + { + "epoch": 0.43, + "learning_rate": 1.256925124297698e-05, + "loss": 1.2712, + "step": 18079 + }, + { + "epoch": 0.43, + "learning_rate": 1.2568498487252086e-05, + "loss": 1.3559, + "step": 18080 + }, + { + "epoch": 0.43, + "learning_rate": 1.2567745715944715e-05, + "loss": 1.1218, + "step": 18081 + }, + { + "epoch": 0.43, + "learning_rate": 1.2566992929059433e-05, + "loss": 1.2844, + "step": 18082 + }, + { + "epoch": 0.43, + "learning_rate": 1.2566240126600808e-05, + "loss": 1.258, + "step": 18083 + }, + { + "epoch": 0.43, + "learning_rate": 1.256548730857341e-05, + "loss": 1.0235, + "step": 18084 + }, + { + "epoch": 0.43, + "learning_rate": 1.2564734474981804e-05, + "loss": 0.9449, + "step": 18085 + }, + { + "epoch": 0.43, + "learning_rate": 1.2563981625830557e-05, + "loss": 1.2592, + "step": 18086 + }, + { + "epoch": 0.43, + "learning_rate": 1.2563228761124233e-05, + "loss": 1.0452, + "step": 18087 + }, + { + "epoch": 0.43, + "learning_rate": 1.2562475880867406e-05, + "loss": 1.0017, + "step": 18088 + }, + { + "epoch": 0.44, + "learning_rate": 1.2561722985064642e-05, + "loss": 1.2355, + "step": 18089 + }, + { + "epoch": 0.44, + "learning_rate": 1.2560970073720506e-05, + "loss": 1.2397, + "step": 18090 + }, + { + "epoch": 0.44, + "learning_rate": 1.2560217146839567e-05, + "loss": 1.3048, + "step": 18091 + }, + { + "epoch": 0.44, + "learning_rate": 1.255946420442639e-05, + "loss": 0.9195, + "step": 18092 + }, + { + "epoch": 0.44, + "learning_rate": 1.2558711246485552e-05, + "loss": 1.0077, + "step": 18093 + }, + { + "epoch": 0.44, + "learning_rate": 1.2557958273021608e-05, + "loss": 0.8693, + "step": 18094 + }, + { + "epoch": 0.44, + "learning_rate": 1.2557205284039135e-05, + "loss": 1.2414, + "step": 18095 + }, + { + "epoch": 0.44, + "learning_rate": 1.2556452279542702e-05, + "loss": 1.0717, + "step": 18096 + }, + { + "epoch": 0.44, + "learning_rate": 1.2555699259536871e-05, + "loss": 1.0123, + "step": 18097 + }, + { + "epoch": 0.44, + "learning_rate": 1.2554946224026215e-05, + "loss": 0.9989, + "step": 18098 + }, + { + "epoch": 0.44, + "learning_rate": 1.2554193173015299e-05, + "loss": 1.2521, + "step": 18099 + }, + { + "epoch": 0.44, + "learning_rate": 1.2553440106508696e-05, + "loss": 1.1048, + "step": 18100 + }, + { + "epoch": 0.44, + "learning_rate": 1.2552687024510972e-05, + "loss": 1.1426, + "step": 18101 + }, + { + "epoch": 0.44, + "learning_rate": 1.2551933927026698e-05, + "loss": 1.3084, + "step": 18102 + }, + { + "epoch": 0.44, + "learning_rate": 1.2551180814060438e-05, + "loss": 1.05, + "step": 18103 + }, + { + "epoch": 0.44, + "learning_rate": 1.2550427685616767e-05, + "loss": 1.0705, + "step": 18104 + }, + { + "epoch": 0.44, + "learning_rate": 1.254967454170025e-05, + "loss": 1.1142, + "step": 18105 + }, + { + "epoch": 0.44, + "learning_rate": 1.2548921382315458e-05, + "loss": 0.9661, + "step": 18106 + }, + { + "epoch": 0.44, + "learning_rate": 1.2548168207466955e-05, + "loss": 1.1032, + "step": 18107 + }, + { + "epoch": 0.44, + "learning_rate": 1.254741501715932e-05, + "loss": 1.1254, + "step": 18108 + }, + { + "epoch": 0.44, + "learning_rate": 1.2546661811397115e-05, + "loss": 1.3569, + "step": 18109 + }, + { + "epoch": 0.44, + "learning_rate": 1.2545908590184914e-05, + "loss": 1.1637, + "step": 18110 + }, + { + "epoch": 0.44, + "learning_rate": 1.2545155353527281e-05, + "loss": 0.995, + "step": 18111 + }, + { + "epoch": 0.44, + "learning_rate": 1.2544402101428792e-05, + "loss": 0.981, + "step": 18112 + }, + { + "epoch": 0.44, + "learning_rate": 1.2543648833894014e-05, + "loss": 0.903, + "step": 18113 + }, + { + "epoch": 0.44, + "learning_rate": 1.2542895550927516e-05, + "loss": 1.1408, + "step": 18114 + }, + { + "epoch": 0.44, + "learning_rate": 1.2542142252533867e-05, + "loss": 1.0108, + "step": 18115 + }, + { + "epoch": 0.44, + "learning_rate": 1.254138893871764e-05, + "loss": 1.2049, + "step": 18116 + }, + { + "epoch": 0.44, + "learning_rate": 1.2540635609483407e-05, + "loss": 1.1634, + "step": 18117 + }, + { + "epoch": 0.44, + "learning_rate": 1.2539882264835733e-05, + "loss": 0.9107, + "step": 18118 + }, + { + "epoch": 0.44, + "learning_rate": 1.2539128904779188e-05, + "loss": 0.9694, + "step": 18119 + }, + { + "epoch": 0.44, + "learning_rate": 1.253837552931835e-05, + "loss": 1.2228, + "step": 18120 + }, + { + "epoch": 0.44, + "learning_rate": 1.2537622138457781e-05, + "loss": 1.1126, + "step": 18121 + }, + { + "epoch": 0.44, + "learning_rate": 1.2536868732202057e-05, + "loss": 0.9489, + "step": 18122 + }, + { + "epoch": 0.44, + "learning_rate": 1.2536115310555746e-05, + "loss": 1.1394, + "step": 18123 + }, + { + "epoch": 0.44, + "learning_rate": 1.253536187352342e-05, + "loss": 1.2189, + "step": 18124 + }, + { + "epoch": 0.44, + "learning_rate": 1.253460842110965e-05, + "loss": 0.7754, + "step": 18125 + }, + { + "epoch": 0.44, + "learning_rate": 1.2533854953319009e-05, + "loss": 1.0963, + "step": 18126 + }, + { + "epoch": 0.44, + "learning_rate": 1.2533101470156064e-05, + "loss": 0.9513, + "step": 18127 + }, + { + "epoch": 0.44, + "learning_rate": 1.2532347971625389e-05, + "loss": 1.0914, + "step": 18128 + }, + { + "epoch": 0.44, + "learning_rate": 1.2531594457731555e-05, + "loss": 0.9216, + "step": 18129 + }, + { + "epoch": 0.44, + "learning_rate": 1.2530840928479132e-05, + "loss": 1.0197, + "step": 18130 + }, + { + "epoch": 0.44, + "learning_rate": 1.253008738387269e-05, + "loss": 1.0877, + "step": 18131 + }, + { + "epoch": 0.44, + "learning_rate": 1.2529333823916807e-05, + "loss": 1.1127, + "step": 18132 + }, + { + "epoch": 0.44, + "learning_rate": 1.2528580248616051e-05, + "loss": 1.0285, + "step": 18133 + }, + { + "epoch": 0.44, + "learning_rate": 1.2527826657974994e-05, + "loss": 1.1204, + "step": 18134 + }, + { + "epoch": 0.44, + "learning_rate": 1.2527073051998203e-05, + "loss": 1.0409, + "step": 18135 + }, + { + "epoch": 0.44, + "learning_rate": 1.2526319430690254e-05, + "loss": 1.0961, + "step": 18136 + }, + { + "epoch": 0.44, + "learning_rate": 1.2525565794055722e-05, + "loss": 0.9738, + "step": 18137 + }, + { + "epoch": 0.44, + "learning_rate": 1.2524812142099176e-05, + "loss": 1.2026, + "step": 18138 + }, + { + "epoch": 0.44, + "learning_rate": 1.2524058474825187e-05, + "loss": 1.1425, + "step": 18139 + }, + { + "epoch": 0.44, + "learning_rate": 1.2523304792238332e-05, + "loss": 1.3765, + "step": 18140 + }, + { + "epoch": 0.44, + "learning_rate": 1.2522551094343179e-05, + "loss": 1.1591, + "step": 18141 + }, + { + "epoch": 0.44, + "learning_rate": 1.25217973811443e-05, + "loss": 1.0699, + "step": 18142 + }, + { + "epoch": 0.44, + "learning_rate": 1.2521043652646272e-05, + "loss": 1.1652, + "step": 18143 + }, + { + "epoch": 0.44, + "learning_rate": 1.2520289908853665e-05, + "loss": 1.1557, + "step": 18144 + }, + { + "epoch": 0.44, + "learning_rate": 1.251953614977105e-05, + "loss": 0.9186, + "step": 18145 + }, + { + "epoch": 0.44, + "learning_rate": 1.2518782375403005e-05, + "loss": 1.1879, + "step": 18146 + }, + { + "epoch": 0.44, + "learning_rate": 1.2518028585754094e-05, + "loss": 1.0008, + "step": 18147 + }, + { + "epoch": 0.44, + "learning_rate": 1.2517274780828901e-05, + "loss": 1.1826, + "step": 18148 + }, + { + "epoch": 0.44, + "learning_rate": 1.2516520960631993e-05, + "loss": 0.9239, + "step": 18149 + }, + { + "epoch": 0.44, + "learning_rate": 1.2515767125167944e-05, + "loss": 1.0032, + "step": 18150 + }, + { + "epoch": 0.44, + "learning_rate": 1.2515013274441326e-05, + "loss": 1.0165, + "step": 18151 + }, + { + "epoch": 0.44, + "learning_rate": 1.2514259408456715e-05, + "loss": 1.0633, + "step": 18152 + }, + { + "epoch": 0.44, + "learning_rate": 1.2513505527218684e-05, + "loss": 0.7899, + "step": 18153 + }, + { + "epoch": 0.44, + "learning_rate": 1.2512751630731808e-05, + "loss": 1.0377, + "step": 18154 + }, + { + "epoch": 0.44, + "learning_rate": 1.2511997719000654e-05, + "loss": 1.1485, + "step": 18155 + }, + { + "epoch": 0.44, + "learning_rate": 1.2511243792029803e-05, + "loss": 1.0085, + "step": 18156 + }, + { + "epoch": 0.44, + "learning_rate": 1.2510489849823827e-05, + "loss": 1.1407, + "step": 18157 + }, + { + "epoch": 0.44, + "learning_rate": 1.2509735892387303e-05, + "loss": 0.8303, + "step": 18158 + }, + { + "epoch": 0.44, + "learning_rate": 1.2508981919724796e-05, + "loss": 1.0985, + "step": 18159 + }, + { + "epoch": 0.44, + "learning_rate": 1.2508227931840888e-05, + "loss": 1.1374, + "step": 18160 + }, + { + "epoch": 0.44, + "learning_rate": 1.2507473928740152e-05, + "loss": 1.1138, + "step": 18161 + }, + { + "epoch": 0.44, + "learning_rate": 1.250671991042716e-05, + "loss": 0.9939, + "step": 18162 + }, + { + "epoch": 0.44, + "learning_rate": 1.250596587690649e-05, + "loss": 1.0446, + "step": 18163 + }, + { + "epoch": 0.44, + "learning_rate": 1.2505211828182712e-05, + "loss": 1.2123, + "step": 18164 + }, + { + "epoch": 0.44, + "learning_rate": 1.2504457764260408e-05, + "loss": 1.2118, + "step": 18165 + }, + { + "epoch": 0.44, + "learning_rate": 1.2503703685144141e-05, + "loss": 1.0676, + "step": 18166 + }, + { + "epoch": 0.44, + "learning_rate": 1.2502949590838496e-05, + "loss": 0.9164, + "step": 18167 + }, + { + "epoch": 0.44, + "learning_rate": 1.2502195481348045e-05, + "loss": 1.1239, + "step": 18168 + }, + { + "epoch": 0.44, + "learning_rate": 1.2501441356677364e-05, + "loss": 1.2127, + "step": 18169 + }, + { + "epoch": 0.44, + "learning_rate": 1.2500687216831022e-05, + "loss": 1.1609, + "step": 18170 + }, + { + "epoch": 0.44, + "learning_rate": 1.2499933061813602e-05, + "loss": 1.1222, + "step": 18171 + }, + { + "epoch": 0.44, + "learning_rate": 1.2499178891629674e-05, + "loss": 1.2509, + "step": 18172 + }, + { + "epoch": 0.44, + "learning_rate": 1.249842470628382e-05, + "loss": 1.1316, + "step": 18173 + }, + { + "epoch": 0.44, + "learning_rate": 1.2497670505780605e-05, + "loss": 1.2987, + "step": 18174 + }, + { + "epoch": 0.44, + "learning_rate": 1.2496916290124616e-05, + "loss": 1.138, + "step": 18175 + }, + { + "epoch": 0.44, + "learning_rate": 1.249616205932042e-05, + "loss": 1.1002, + "step": 18176 + }, + { + "epoch": 0.44, + "learning_rate": 1.2495407813372598e-05, + "loss": 0.7906, + "step": 18177 + }, + { + "epoch": 0.44, + "learning_rate": 1.2494653552285721e-05, + "loss": 1.3013, + "step": 18178 + }, + { + "epoch": 0.44, + "learning_rate": 1.249389927606437e-05, + "loss": 0.6892, + "step": 18179 + }, + { + "epoch": 0.44, + "learning_rate": 1.2493144984713119e-05, + "loss": 1.1488, + "step": 18180 + }, + { + "epoch": 0.44, + "learning_rate": 1.2492390678236542e-05, + "loss": 1.0533, + "step": 18181 + }, + { + "epoch": 0.44, + "learning_rate": 1.2491636356639217e-05, + "loss": 0.8816, + "step": 18182 + }, + { + "epoch": 0.44, + "learning_rate": 1.249088201992572e-05, + "loss": 0.9838, + "step": 18183 + }, + { + "epoch": 0.44, + "learning_rate": 1.2490127668100632e-05, + "loss": 0.9769, + "step": 18184 + }, + { + "epoch": 0.44, + "learning_rate": 1.248937330116852e-05, + "loss": 0.8371, + "step": 18185 + }, + { + "epoch": 0.44, + "learning_rate": 1.2488618919133966e-05, + "loss": 1.3046, + "step": 18186 + }, + { + "epoch": 0.44, + "learning_rate": 1.248786452200155e-05, + "loss": 1.0394, + "step": 18187 + }, + { + "epoch": 0.44, + "learning_rate": 1.2487110109775841e-05, + "loss": 0.9657, + "step": 18188 + }, + { + "epoch": 0.44, + "learning_rate": 1.2486355682461423e-05, + "loss": 1.0052, + "step": 18189 + }, + { + "epoch": 0.44, + "learning_rate": 1.2485601240062868e-05, + "loss": 0.9123, + "step": 18190 + }, + { + "epoch": 0.44, + "learning_rate": 1.2484846782584757e-05, + "loss": 1.1498, + "step": 18191 + }, + { + "epoch": 0.44, + "learning_rate": 1.2484092310031663e-05, + "loss": 0.9013, + "step": 18192 + }, + { + "epoch": 0.44, + "learning_rate": 1.2483337822408169e-05, + "loss": 1.1917, + "step": 18193 + }, + { + "epoch": 0.44, + "learning_rate": 1.2482583319718842e-05, + "loss": 1.2954, + "step": 18194 + }, + { + "epoch": 0.44, + "learning_rate": 1.248182880196827e-05, + "loss": 1.0287, + "step": 18195 + }, + { + "epoch": 0.44, + "learning_rate": 1.2481074269161026e-05, + "loss": 1.135, + "step": 18196 + }, + { + "epoch": 0.44, + "learning_rate": 1.248031972130169e-05, + "loss": 1.1311, + "step": 18197 + }, + { + "epoch": 0.44, + "learning_rate": 1.2479565158394833e-05, + "loss": 0.9933, + "step": 18198 + }, + { + "epoch": 0.44, + "learning_rate": 1.247881058044504e-05, + "loss": 0.9974, + "step": 18199 + }, + { + "epoch": 0.44, + "learning_rate": 1.2478055987456887e-05, + "loss": 1.158, + "step": 18200 + }, + { + "epoch": 0.44, + "learning_rate": 1.2477301379434952e-05, + "loss": 1.015, + "step": 18201 + }, + { + "epoch": 0.44, + "learning_rate": 1.2476546756383809e-05, + "loss": 1.3435, + "step": 18202 + }, + { + "epoch": 0.44, + "learning_rate": 1.2475792118308041e-05, + "loss": 1.1638, + "step": 18203 + }, + { + "epoch": 0.44, + "learning_rate": 1.2475037465212226e-05, + "loss": 1.1583, + "step": 18204 + }, + { + "epoch": 0.44, + "learning_rate": 1.247428279710094e-05, + "loss": 1.0597, + "step": 18205 + }, + { + "epoch": 0.44, + "learning_rate": 1.2473528113978761e-05, + "loss": 1.1349, + "step": 18206 + }, + { + "epoch": 0.44, + "learning_rate": 1.2472773415850268e-05, + "loss": 1.1215, + "step": 18207 + }, + { + "epoch": 0.44, + "learning_rate": 1.2472018702720042e-05, + "loss": 1.2111, + "step": 18208 + }, + { + "epoch": 0.44, + "learning_rate": 1.247126397459266e-05, + "loss": 1.0004, + "step": 18209 + }, + { + "epoch": 0.44, + "learning_rate": 1.2470509231472698e-05, + "loss": 0.9739, + "step": 18210 + }, + { + "epoch": 0.44, + "learning_rate": 1.2469754473364742e-05, + "loss": 1.1548, + "step": 18211 + }, + { + "epoch": 0.44, + "learning_rate": 1.2468999700273365e-05, + "loss": 1.391, + "step": 18212 + }, + { + "epoch": 0.44, + "learning_rate": 1.2468244912203146e-05, + "loss": 1.1576, + "step": 18213 + }, + { + "epoch": 0.44, + "learning_rate": 1.2467490109158667e-05, + "loss": 1.1084, + "step": 18214 + }, + { + "epoch": 0.44, + "learning_rate": 1.2466735291144505e-05, + "loss": 1.0205, + "step": 18215 + }, + { + "epoch": 0.44, + "learning_rate": 1.2465980458165243e-05, + "loss": 1.0217, + "step": 18216 + }, + { + "epoch": 0.44, + "learning_rate": 1.2465225610225455e-05, + "loss": 1.1041, + "step": 18217 + }, + { + "epoch": 0.44, + "learning_rate": 1.2464470747329723e-05, + "loss": 1.0525, + "step": 18218 + }, + { + "epoch": 0.44, + "learning_rate": 1.2463715869482629e-05, + "loss": 1.2328, + "step": 18219 + }, + { + "epoch": 0.44, + "learning_rate": 1.246296097668875e-05, + "loss": 0.9414, + "step": 18220 + }, + { + "epoch": 0.44, + "learning_rate": 1.2462206068952663e-05, + "loss": 1.2464, + "step": 18221 + }, + { + "epoch": 0.44, + "learning_rate": 1.2461451146278951e-05, + "loss": 1.3174, + "step": 18222 + }, + { + "epoch": 0.44, + "learning_rate": 1.2460696208672196e-05, + "loss": 1.0857, + "step": 18223 + }, + { + "epoch": 0.44, + "learning_rate": 1.2459941256136976e-05, + "loss": 1.1411, + "step": 18224 + }, + { + "epoch": 0.44, + "learning_rate": 1.245918628867787e-05, + "loss": 1.2362, + "step": 18225 + }, + { + "epoch": 0.44, + "learning_rate": 1.2458431306299459e-05, + "loss": 1.2314, + "step": 18226 + }, + { + "epoch": 0.44, + "learning_rate": 1.2457676309006325e-05, + "loss": 1.1183, + "step": 18227 + }, + { + "epoch": 0.44, + "learning_rate": 1.2456921296803047e-05, + "loss": 1.0558, + "step": 18228 + }, + { + "epoch": 0.44, + "learning_rate": 1.2456166269694206e-05, + "loss": 1.0726, + "step": 18229 + }, + { + "epoch": 0.44, + "learning_rate": 1.2455411227684377e-05, + "loss": 1.1538, + "step": 18230 + }, + { + "epoch": 0.44, + "learning_rate": 1.245465617077815e-05, + "loss": 1.174, + "step": 18231 + }, + { + "epoch": 0.44, + "learning_rate": 1.2453901098980102e-05, + "loss": 1.2978, + "step": 18232 + }, + { + "epoch": 0.44, + "learning_rate": 1.2453146012294812e-05, + "loss": 1.388, + "step": 18233 + }, + { + "epoch": 0.44, + "learning_rate": 1.2452390910726859e-05, + "loss": 1.101, + "step": 18234 + }, + { + "epoch": 0.44, + "learning_rate": 1.2451635794280832e-05, + "loss": 1.0313, + "step": 18235 + }, + { + "epoch": 0.44, + "learning_rate": 1.2450880662961307e-05, + "loss": 1.2344, + "step": 18236 + }, + { + "epoch": 0.44, + "learning_rate": 1.2450125516772861e-05, + "loss": 1.1281, + "step": 18237 + }, + { + "epoch": 0.44, + "learning_rate": 1.2449370355720083e-05, + "loss": 0.8128, + "step": 18238 + }, + { + "epoch": 0.44, + "learning_rate": 1.244861517980755e-05, + "loss": 1.0003, + "step": 18239 + }, + { + "epoch": 0.44, + "learning_rate": 1.2447859989039848e-05, + "loss": 1.061, + "step": 18240 + }, + { + "epoch": 0.44, + "learning_rate": 1.2447104783421553e-05, + "loss": 0.9838, + "step": 18241 + }, + { + "epoch": 0.44, + "learning_rate": 1.2446349562957247e-05, + "loss": 1.1027, + "step": 18242 + }, + { + "epoch": 0.44, + "learning_rate": 1.2445594327651516e-05, + "loss": 0.9842, + "step": 18243 + }, + { + "epoch": 0.44, + "learning_rate": 1.2444839077508939e-05, + "loss": 1.0717, + "step": 18244 + }, + { + "epoch": 0.44, + "learning_rate": 1.2444083812534098e-05, + "loss": 1.1007, + "step": 18245 + }, + { + "epoch": 0.44, + "learning_rate": 1.2443328532731574e-05, + "loss": 1.1841, + "step": 18246 + }, + { + "epoch": 0.44, + "learning_rate": 1.2442573238105953e-05, + "loss": 1.1541, + "step": 18247 + }, + { + "epoch": 0.44, + "learning_rate": 1.2441817928661814e-05, + "loss": 1.225, + "step": 18248 + }, + { + "epoch": 0.44, + "learning_rate": 1.2441062604403737e-05, + "loss": 1.0642, + "step": 18249 + }, + { + "epoch": 0.44, + "learning_rate": 1.244030726533631e-05, + "loss": 1.24, + "step": 18250 + }, + { + "epoch": 0.44, + "learning_rate": 1.2439551911464114e-05, + "loss": 1.1009, + "step": 18251 + }, + { + "epoch": 0.44, + "learning_rate": 1.2438796542791729e-05, + "loss": 0.9845, + "step": 18252 + }, + { + "epoch": 0.44, + "learning_rate": 1.2438041159323739e-05, + "loss": 1.115, + "step": 18253 + }, + { + "epoch": 0.44, + "learning_rate": 1.2437285761064728e-05, + "loss": 1.148, + "step": 18254 + }, + { + "epoch": 0.44, + "learning_rate": 1.2436530348019276e-05, + "loss": 1.1694, + "step": 18255 + }, + { + "epoch": 0.44, + "learning_rate": 1.2435774920191968e-05, + "loss": 1.323, + "step": 18256 + }, + { + "epoch": 0.44, + "learning_rate": 1.2435019477587385e-05, + "loss": 1.0964, + "step": 18257 + }, + { + "epoch": 0.44, + "learning_rate": 1.2434264020210113e-05, + "loss": 1.2955, + "step": 18258 + }, + { + "epoch": 0.44, + "learning_rate": 1.2433508548064734e-05, + "loss": 0.9345, + "step": 18259 + }, + { + "epoch": 0.44, + "learning_rate": 1.2432753061155833e-05, + "loss": 1.1181, + "step": 18260 + }, + { + "epoch": 0.44, + "learning_rate": 1.2431997559487987e-05, + "loss": 1.0823, + "step": 18261 + }, + { + "epoch": 0.44, + "learning_rate": 1.2431242043065788e-05, + "loss": 0.9683, + "step": 18262 + }, + { + "epoch": 0.44, + "learning_rate": 1.2430486511893813e-05, + "loss": 0.9104, + "step": 18263 + }, + { + "epoch": 0.44, + "learning_rate": 1.2429730965976647e-05, + "loss": 1.213, + "step": 18264 + }, + { + "epoch": 0.44, + "learning_rate": 1.2428975405318874e-05, + "loss": 0.9284, + "step": 18265 + }, + { + "epoch": 0.44, + "learning_rate": 1.2428219829925083e-05, + "loss": 1.0644, + "step": 18266 + }, + { + "epoch": 0.44, + "learning_rate": 1.2427464239799851e-05, + "loss": 1.0463, + "step": 18267 + }, + { + "epoch": 0.44, + "learning_rate": 1.2426708634947767e-05, + "loss": 0.9884, + "step": 18268 + }, + { + "epoch": 0.44, + "learning_rate": 1.2425953015373406e-05, + "loss": 1.017, + "step": 18269 + }, + { + "epoch": 0.44, + "learning_rate": 1.2425197381081364e-05, + "loss": 0.8028, + "step": 18270 + }, + { + "epoch": 0.44, + "learning_rate": 1.242444173207622e-05, + "loss": 0.8265, + "step": 18271 + }, + { + "epoch": 0.44, + "learning_rate": 1.2423686068362556e-05, + "loss": 0.8369, + "step": 18272 + }, + { + "epoch": 0.44, + "learning_rate": 1.2422930389944961e-05, + "loss": 1.1674, + "step": 18273 + }, + { + "epoch": 0.44, + "learning_rate": 1.2422174696828015e-05, + "loss": 0.7092, + "step": 18274 + }, + { + "epoch": 0.44, + "learning_rate": 1.2421418989016307e-05, + "loss": 1.0191, + "step": 18275 + }, + { + "epoch": 0.44, + "learning_rate": 1.2420663266514419e-05, + "loss": 0.92, + "step": 18276 + }, + { + "epoch": 0.44, + "learning_rate": 1.2419907529326934e-05, + "loss": 0.9298, + "step": 18277 + }, + { + "epoch": 0.44, + "learning_rate": 1.241915177745844e-05, + "loss": 0.8332, + "step": 18278 + }, + { + "epoch": 0.44, + "learning_rate": 1.2418396010913525e-05, + "loss": 1.276, + "step": 18279 + }, + { + "epoch": 0.44, + "learning_rate": 1.2417640229696766e-05, + "loss": 0.9545, + "step": 18280 + }, + { + "epoch": 0.44, + "learning_rate": 1.2416884433812753e-05, + "loss": 0.9106, + "step": 18281 + }, + { + "epoch": 0.44, + "learning_rate": 1.241612862326607e-05, + "loss": 1.1395, + "step": 18282 + }, + { + "epoch": 0.44, + "learning_rate": 1.2415372798061307e-05, + "loss": 1.3163, + "step": 18283 + }, + { + "epoch": 0.44, + "learning_rate": 1.2414616958203044e-05, + "loss": 1.219, + "step": 18284 + }, + { + "epoch": 0.44, + "learning_rate": 1.2413861103695864e-05, + "loss": 1.1288, + "step": 18285 + }, + { + "epoch": 0.44, + "learning_rate": 1.2413105234544358e-05, + "loss": 1.1007, + "step": 18286 + }, + { + "epoch": 0.44, + "learning_rate": 1.2412349350753112e-05, + "loss": 1.1047, + "step": 18287 + }, + { + "epoch": 0.44, + "learning_rate": 1.2411593452326711e-05, + "loss": 1.1157, + "step": 18288 + }, + { + "epoch": 0.44, + "learning_rate": 1.2410837539269736e-05, + "loss": 1.0109, + "step": 18289 + }, + { + "epoch": 0.44, + "learning_rate": 1.2410081611586775e-05, + "loss": 1.0507, + "step": 18290 + }, + { + "epoch": 0.44, + "learning_rate": 1.2409325669282421e-05, + "loss": 1.1961, + "step": 18291 + }, + { + "epoch": 0.44, + "learning_rate": 1.2408569712361253e-05, + "loss": 1.0995, + "step": 18292 + }, + { + "epoch": 0.44, + "learning_rate": 1.2407813740827859e-05, + "loss": 0.9776, + "step": 18293 + }, + { + "epoch": 0.44, + "learning_rate": 1.2407057754686824e-05, + "loss": 1.0943, + "step": 18294 + }, + { + "epoch": 0.44, + "learning_rate": 1.240630175394274e-05, + "loss": 1.0952, + "step": 18295 + }, + { + "epoch": 0.44, + "learning_rate": 1.2405545738600187e-05, + "loss": 1.331, + "step": 18296 + }, + { + "epoch": 0.44, + "learning_rate": 1.240478970866375e-05, + "loss": 1.0397, + "step": 18297 + }, + { + "epoch": 0.44, + "learning_rate": 1.2404033664138024e-05, + "loss": 1.2101, + "step": 18298 + }, + { + "epoch": 0.44, + "learning_rate": 1.2403277605027593e-05, + "loss": 0.8275, + "step": 18299 + }, + { + "epoch": 0.44, + "learning_rate": 1.240252153133704e-05, + "loss": 1.2267, + "step": 18300 + }, + { + "epoch": 0.44, + "learning_rate": 1.240176544307095e-05, + "loss": 1.1482, + "step": 18301 + }, + { + "epoch": 0.44, + "learning_rate": 1.2401009340233917e-05, + "loss": 1.3544, + "step": 18302 + }, + { + "epoch": 0.44, + "learning_rate": 1.2400253222830526e-05, + "loss": 1.0416, + "step": 18303 + }, + { + "epoch": 0.44, + "learning_rate": 1.2399497090865365e-05, + "loss": 1.0366, + "step": 18304 + }, + { + "epoch": 0.44, + "learning_rate": 1.2398740944343013e-05, + "loss": 1.1536, + "step": 18305 + }, + { + "epoch": 0.44, + "learning_rate": 1.2397984783268068e-05, + "loss": 0.9972, + "step": 18306 + }, + { + "epoch": 0.44, + "learning_rate": 1.2397228607645116e-05, + "loss": 1.1819, + "step": 18307 + }, + { + "epoch": 0.44, + "learning_rate": 1.239647241747874e-05, + "loss": 1.2647, + "step": 18308 + }, + { + "epoch": 0.44, + "learning_rate": 1.2395716212773525e-05, + "loss": 0.9448, + "step": 18309 + }, + { + "epoch": 0.44, + "learning_rate": 1.2394959993534069e-05, + "loss": 1.0145, + "step": 18310 + }, + { + "epoch": 0.44, + "learning_rate": 1.2394203759764954e-05, + "loss": 0.976, + "step": 18311 + }, + { + "epoch": 0.44, + "learning_rate": 1.2393447511470765e-05, + "loss": 1.092, + "step": 18312 + }, + { + "epoch": 0.44, + "learning_rate": 1.2392691248656093e-05, + "loss": 0.9381, + "step": 18313 + }, + { + "epoch": 0.44, + "learning_rate": 1.2391934971325527e-05, + "loss": 1.2526, + "step": 18314 + }, + { + "epoch": 0.44, + "learning_rate": 1.2391178679483654e-05, + "loss": 1.2558, + "step": 18315 + }, + { + "epoch": 0.44, + "learning_rate": 1.2390422373135063e-05, + "loss": 1.1486, + "step": 18316 + }, + { + "epoch": 0.44, + "learning_rate": 1.238966605228434e-05, + "loss": 1.131, + "step": 18317 + }, + { + "epoch": 0.44, + "learning_rate": 1.2388909716936077e-05, + "loss": 1.3045, + "step": 18318 + }, + { + "epoch": 0.44, + "learning_rate": 1.238815336709486e-05, + "loss": 1.2391, + "step": 18319 + }, + { + "epoch": 0.44, + "learning_rate": 1.2387397002765277e-05, + "loss": 1.0291, + "step": 18320 + }, + { + "epoch": 0.44, + "learning_rate": 1.238664062395192e-05, + "loss": 1.1555, + "step": 18321 + }, + { + "epoch": 0.44, + "learning_rate": 1.2385884230659375e-05, + "loss": 1.3662, + "step": 18322 + }, + { + "epoch": 0.44, + "learning_rate": 1.238512782289223e-05, + "loss": 1.3101, + "step": 18323 + }, + { + "epoch": 0.44, + "learning_rate": 1.238437140065508e-05, + "loss": 1.1489, + "step": 18324 + }, + { + "epoch": 0.44, + "learning_rate": 1.2383614963952504e-05, + "loss": 1.0643, + "step": 18325 + }, + { + "epoch": 0.44, + "learning_rate": 1.23828585127891e-05, + "loss": 1.1279, + "step": 18326 + }, + { + "epoch": 0.44, + "learning_rate": 1.2382102047169452e-05, + "loss": 1.0775, + "step": 18327 + }, + { + "epoch": 0.44, + "learning_rate": 1.2381345567098156e-05, + "loss": 1.0598, + "step": 18328 + }, + { + "epoch": 0.44, + "learning_rate": 1.238058907257979e-05, + "loss": 1.1377, + "step": 18329 + }, + { + "epoch": 0.44, + "learning_rate": 1.2379832563618952e-05, + "loss": 1.2995, + "step": 18330 + }, + { + "epoch": 0.44, + "learning_rate": 1.2379076040220232e-05, + "loss": 0.8502, + "step": 18331 + }, + { + "epoch": 0.44, + "learning_rate": 1.2378319502388216e-05, + "loss": 0.9034, + "step": 18332 + }, + { + "epoch": 0.44, + "learning_rate": 1.2377562950127493e-05, + "loss": 0.9257, + "step": 18333 + }, + { + "epoch": 0.44, + "learning_rate": 1.2376806383442657e-05, + "loss": 1.2187, + "step": 18334 + }, + { + "epoch": 0.44, + "learning_rate": 1.2376049802338297e-05, + "loss": 1.0788, + "step": 18335 + }, + { + "epoch": 0.44, + "learning_rate": 1.2375293206818997e-05, + "loss": 0.7425, + "step": 18336 + }, + { + "epoch": 0.44, + "learning_rate": 1.2374536596889354e-05, + "loss": 1.2016, + "step": 18337 + }, + { + "epoch": 0.44, + "learning_rate": 1.237377997255396e-05, + "loss": 1.1294, + "step": 18338 + }, + { + "epoch": 0.44, + "learning_rate": 1.2373023333817398e-05, + "loss": 0.9906, + "step": 18339 + }, + { + "epoch": 0.44, + "learning_rate": 1.2372266680684259e-05, + "loss": 1.0912, + "step": 18340 + }, + { + "epoch": 0.44, + "learning_rate": 1.237151001315914e-05, + "loss": 1.2946, + "step": 18341 + }, + { + "epoch": 0.44, + "learning_rate": 1.2370753331246628e-05, + "loss": 1.347, + "step": 18342 + }, + { + "epoch": 0.44, + "learning_rate": 1.236999663495131e-05, + "loss": 1.1466, + "step": 18343 + }, + { + "epoch": 0.44, + "learning_rate": 1.236923992427778e-05, + "loss": 1.2335, + "step": 18344 + }, + { + "epoch": 0.44, + "learning_rate": 1.2368483199230632e-05, + "loss": 0.8599, + "step": 18345 + }, + { + "epoch": 0.44, + "learning_rate": 1.236772645981445e-05, + "loss": 1.1617, + "step": 18346 + }, + { + "epoch": 0.44, + "learning_rate": 1.2366969706033831e-05, + "loss": 1.2223, + "step": 18347 + }, + { + "epoch": 0.44, + "learning_rate": 1.236621293789336e-05, + "loss": 1.26, + "step": 18348 + }, + { + "epoch": 0.44, + "learning_rate": 1.2365456155397637e-05, + "loss": 1.0477, + "step": 18349 + }, + { + "epoch": 0.44, + "learning_rate": 1.2364699358551245e-05, + "loss": 1.2805, + "step": 18350 + }, + { + "epoch": 0.44, + "learning_rate": 1.2363942547358778e-05, + "loss": 1.1457, + "step": 18351 + }, + { + "epoch": 0.44, + "learning_rate": 1.2363185721824826e-05, + "loss": 1.0271, + "step": 18352 + }, + { + "epoch": 0.44, + "learning_rate": 1.2362428881953985e-05, + "loss": 1.0577, + "step": 18353 + }, + { + "epoch": 0.44, + "learning_rate": 1.2361672027750845e-05, + "loss": 1.0768, + "step": 18354 + }, + { + "epoch": 0.44, + "learning_rate": 1.2360915159219993e-05, + "loss": 1.1606, + "step": 18355 + }, + { + "epoch": 0.44, + "learning_rate": 1.2360158276366023e-05, + "loss": 0.9506, + "step": 18356 + }, + { + "epoch": 0.44, + "learning_rate": 1.2359401379193532e-05, + "loss": 1.2104, + "step": 18357 + }, + { + "epoch": 0.44, + "learning_rate": 1.2358644467707107e-05, + "loss": 0.9417, + "step": 18358 + }, + { + "epoch": 0.44, + "learning_rate": 1.2357887541911337e-05, + "loss": 1.1764, + "step": 18359 + }, + { + "epoch": 0.44, + "learning_rate": 1.235713060181082e-05, + "loss": 1.2326, + "step": 18360 + }, + { + "epoch": 0.44, + "learning_rate": 1.2356373647410147e-05, + "loss": 1.1873, + "step": 18361 + }, + { + "epoch": 0.44, + "learning_rate": 1.2355616678713909e-05, + "loss": 1.1799, + "step": 18362 + }, + { + "epoch": 0.44, + "learning_rate": 1.23548596957267e-05, + "loss": 1.0616, + "step": 18363 + }, + { + "epoch": 0.44, + "learning_rate": 1.2354102698453107e-05, + "loss": 0.9486, + "step": 18364 + }, + { + "epoch": 0.44, + "learning_rate": 1.235334568689773e-05, + "loss": 1.0867, + "step": 18365 + }, + { + "epoch": 0.44, + "learning_rate": 1.235258866106516e-05, + "loss": 1.2139, + "step": 18366 + }, + { + "epoch": 0.44, + "learning_rate": 1.2351831620959985e-05, + "loss": 1.0066, + "step": 18367 + }, + { + "epoch": 0.44, + "learning_rate": 1.2351074566586798e-05, + "loss": 1.2307, + "step": 18368 + }, + { + "epoch": 0.44, + "learning_rate": 1.23503174979502e-05, + "loss": 0.9539, + "step": 18369 + }, + { + "epoch": 0.44, + "learning_rate": 1.2349560415054775e-05, + "loss": 1.0005, + "step": 18370 + }, + { + "epoch": 0.44, + "learning_rate": 1.2348803317905122e-05, + "loss": 1.1402, + "step": 18371 + }, + { + "epoch": 0.44, + "learning_rate": 1.234804620650583e-05, + "loss": 0.9313, + "step": 18372 + }, + { + "epoch": 0.44, + "learning_rate": 1.2347289080861495e-05, + "loss": 0.8184, + "step": 18373 + }, + { + "epoch": 0.44, + "learning_rate": 1.2346531940976707e-05, + "loss": 0.7537, + "step": 18374 + }, + { + "epoch": 0.44, + "learning_rate": 1.2345774786856064e-05, + "loss": 1.1358, + "step": 18375 + }, + { + "epoch": 0.44, + "learning_rate": 1.2345017618504155e-05, + "loss": 1.0295, + "step": 18376 + }, + { + "epoch": 0.44, + "learning_rate": 1.2344260435925576e-05, + "loss": 1.2009, + "step": 18377 + }, + { + "epoch": 0.44, + "learning_rate": 1.2343503239124926e-05, + "loss": 1.2807, + "step": 18378 + }, + { + "epoch": 0.44, + "learning_rate": 1.2342746028106786e-05, + "loss": 1.1377, + "step": 18379 + }, + { + "epoch": 0.44, + "learning_rate": 1.234198880287576e-05, + "loss": 0.8778, + "step": 18380 + }, + { + "epoch": 0.44, + "learning_rate": 1.2341231563436438e-05, + "loss": 1.276, + "step": 18381 + }, + { + "epoch": 0.44, + "learning_rate": 1.2340474309793415e-05, + "loss": 0.9868, + "step": 18382 + }, + { + "epoch": 0.44, + "learning_rate": 1.2339717041951287e-05, + "loss": 1.0752, + "step": 18383 + }, + { + "epoch": 0.44, + "learning_rate": 1.2338959759914644e-05, + "loss": 1.0357, + "step": 18384 + }, + { + "epoch": 0.44, + "learning_rate": 1.2338202463688082e-05, + "loss": 0.9125, + "step": 18385 + }, + { + "epoch": 0.44, + "learning_rate": 1.2337445153276197e-05, + "loss": 0.8772, + "step": 18386 + }, + { + "epoch": 0.44, + "learning_rate": 1.2336687828683582e-05, + "loss": 0.956, + "step": 18387 + }, + { + "epoch": 0.44, + "learning_rate": 1.233593048991483e-05, + "loss": 1.2466, + "step": 18388 + }, + { + "epoch": 0.44, + "learning_rate": 1.2335173136974536e-05, + "loss": 1.0126, + "step": 18389 + }, + { + "epoch": 0.44, + "learning_rate": 1.23344157698673e-05, + "loss": 1.3606, + "step": 18390 + }, + { + "epoch": 0.44, + "learning_rate": 1.2333658388597712e-05, + "loss": 1.0166, + "step": 18391 + }, + { + "epoch": 0.44, + "learning_rate": 1.2332900993170363e-05, + "loss": 1.1063, + "step": 18392 + }, + { + "epoch": 0.44, + "learning_rate": 1.2332143583589855e-05, + "loss": 1.1715, + "step": 18393 + }, + { + "epoch": 0.44, + "learning_rate": 1.233138615986078e-05, + "loss": 1.0952, + "step": 18394 + }, + { + "epoch": 0.44, + "learning_rate": 1.2330628721987737e-05, + "loss": 1.3686, + "step": 18395 + }, + { + "epoch": 0.44, + "learning_rate": 1.2329871269975314e-05, + "loss": 0.7907, + "step": 18396 + }, + { + "epoch": 0.44, + "learning_rate": 1.232911380382811e-05, + "loss": 1.3791, + "step": 18397 + }, + { + "epoch": 0.44, + "learning_rate": 1.2328356323550723e-05, + "loss": 1.0238, + "step": 18398 + }, + { + "epoch": 0.44, + "learning_rate": 1.2327598829147745e-05, + "loss": 0.8851, + "step": 18399 + }, + { + "epoch": 0.44, + "learning_rate": 1.2326841320623769e-05, + "loss": 0.9443, + "step": 18400 + }, + { + "epoch": 0.44, + "learning_rate": 1.2326083797983398e-05, + "loss": 1.1984, + "step": 18401 + }, + { + "epoch": 0.44, + "learning_rate": 1.2325326261231224e-05, + "loss": 1.234, + "step": 18402 + }, + { + "epoch": 0.44, + "learning_rate": 1.232456871037184e-05, + "loss": 1.2808, + "step": 18403 + }, + { + "epoch": 0.44, + "learning_rate": 1.2323811145409843e-05, + "loss": 0.8841, + "step": 18404 + }, + { + "epoch": 0.44, + "learning_rate": 1.2323053566349834e-05, + "loss": 1.0606, + "step": 18405 + }, + { + "epoch": 0.44, + "learning_rate": 1.2322295973196403e-05, + "loss": 1.1623, + "step": 18406 + }, + { + "epoch": 0.44, + "learning_rate": 1.232153836595415e-05, + "loss": 1.0306, + "step": 18407 + }, + { + "epoch": 0.44, + "learning_rate": 1.232078074462767e-05, + "loss": 1.2889, + "step": 18408 + }, + { + "epoch": 0.44, + "learning_rate": 1.2320023109221556e-05, + "loss": 1.0702, + "step": 18409 + }, + { + "epoch": 0.44, + "learning_rate": 1.2319265459740412e-05, + "loss": 1.0962, + "step": 18410 + }, + { + "epoch": 0.44, + "learning_rate": 1.231850779618883e-05, + "loss": 1.2885, + "step": 18411 + }, + { + "epoch": 0.44, + "learning_rate": 1.2317750118571403e-05, + "loss": 0.8955, + "step": 18412 + }, + { + "epoch": 0.44, + "learning_rate": 1.2316992426892734e-05, + "loss": 1.0353, + "step": 18413 + }, + { + "epoch": 0.44, + "learning_rate": 1.2316234721157416e-05, + "loss": 1.0004, + "step": 18414 + }, + { + "epoch": 0.44, + "learning_rate": 1.2315477001370047e-05, + "loss": 1.3867, + "step": 18415 + }, + { + "epoch": 0.44, + "learning_rate": 1.231471926753522e-05, + "loss": 1.2168, + "step": 18416 + }, + { + "epoch": 0.44, + "learning_rate": 1.231396151965754e-05, + "loss": 0.8615, + "step": 18417 + }, + { + "epoch": 0.44, + "learning_rate": 1.2313203757741598e-05, + "loss": 1.211, + "step": 18418 + }, + { + "epoch": 0.44, + "learning_rate": 1.2312445981791993e-05, + "loss": 0.9494, + "step": 18419 + }, + { + "epoch": 0.44, + "learning_rate": 1.2311688191813323e-05, + "loss": 1.0451, + "step": 18420 + }, + { + "epoch": 0.44, + "learning_rate": 1.2310930387810186e-05, + "loss": 0.9061, + "step": 18421 + }, + { + "epoch": 0.44, + "learning_rate": 1.2310172569787176e-05, + "loss": 1.1439, + "step": 18422 + }, + { + "epoch": 0.44, + "learning_rate": 1.230941473774889e-05, + "loss": 1.0084, + "step": 18423 + }, + { + "epoch": 0.44, + "learning_rate": 1.2308656891699931e-05, + "loss": 1.0602, + "step": 18424 + }, + { + "epoch": 0.44, + "learning_rate": 1.2307899031644894e-05, + "loss": 1.0874, + "step": 18425 + }, + { + "epoch": 0.44, + "learning_rate": 1.2307141157588376e-05, + "loss": 0.8556, + "step": 18426 + }, + { + "epoch": 0.44, + "learning_rate": 1.2306383269534972e-05, + "loss": 0.9413, + "step": 18427 + }, + { + "epoch": 0.44, + "learning_rate": 1.2305625367489288e-05, + "loss": 1.1813, + "step": 18428 + }, + { + "epoch": 0.44, + "learning_rate": 1.2304867451455916e-05, + "loss": 1.0132, + "step": 18429 + }, + { + "epoch": 0.44, + "learning_rate": 1.2304109521439453e-05, + "loss": 1.0729, + "step": 18430 + }, + { + "epoch": 0.44, + "learning_rate": 1.23033515774445e-05, + "loss": 1.2226, + "step": 18431 + }, + { + "epoch": 0.44, + "learning_rate": 1.2302593619475656e-05, + "loss": 1.1104, + "step": 18432 + }, + { + "epoch": 0.44, + "learning_rate": 1.2301835647537519e-05, + "loss": 0.9224, + "step": 18433 + }, + { + "epoch": 0.44, + "learning_rate": 1.2301077661634686e-05, + "loss": 1.3226, + "step": 18434 + }, + { + "epoch": 0.44, + "learning_rate": 1.2300319661771753e-05, + "loss": 0.9869, + "step": 18435 + }, + { + "epoch": 0.44, + "learning_rate": 1.2299561647953327e-05, + "loss": 0.84, + "step": 18436 + }, + { + "epoch": 0.44, + "learning_rate": 1.2298803620183997e-05, + "loss": 1.1448, + "step": 18437 + }, + { + "epoch": 0.44, + "learning_rate": 1.229804557846837e-05, + "loss": 1.2442, + "step": 18438 + }, + { + "epoch": 0.44, + "learning_rate": 1.2297287522811035e-05, + "loss": 1.2071, + "step": 18439 + }, + { + "epoch": 0.44, + "learning_rate": 1.22965294532166e-05, + "loss": 1.0671, + "step": 18440 + }, + { + "epoch": 0.44, + "learning_rate": 1.2295771369689663e-05, + "loss": 1.1158, + "step": 18441 + }, + { + "epoch": 0.44, + "learning_rate": 1.2295013272234821e-05, + "loss": 1.1798, + "step": 18442 + }, + { + "epoch": 0.44, + "learning_rate": 1.229425516085667e-05, + "loss": 1.171, + "step": 18443 + }, + { + "epoch": 0.44, + "learning_rate": 1.2293497035559816e-05, + "loss": 1.1232, + "step": 18444 + }, + { + "epoch": 0.44, + "learning_rate": 1.2292738896348854e-05, + "loss": 1.1675, + "step": 18445 + }, + { + "epoch": 0.44, + "learning_rate": 1.2291980743228384e-05, + "loss": 0.8969, + "step": 18446 + }, + { + "epoch": 0.44, + "learning_rate": 1.2291222576203004e-05, + "loss": 0.9062, + "step": 18447 + }, + { + "epoch": 0.44, + "learning_rate": 1.229046439527732e-05, + "loss": 0.9299, + "step": 18448 + }, + { + "epoch": 0.44, + "learning_rate": 1.2289706200455924e-05, + "loss": 1.1076, + "step": 18449 + }, + { + "epoch": 0.44, + "learning_rate": 1.228894799174342e-05, + "loss": 0.968, + "step": 18450 + }, + { + "epoch": 0.44, + "learning_rate": 1.2288189769144406e-05, + "loss": 1.2015, + "step": 18451 + }, + { + "epoch": 0.44, + "learning_rate": 1.2287431532663483e-05, + "loss": 1.1401, + "step": 18452 + }, + { + "epoch": 0.44, + "learning_rate": 1.2286673282305253e-05, + "loss": 1.2695, + "step": 18453 + }, + { + "epoch": 0.44, + "learning_rate": 1.2285915018074312e-05, + "loss": 0.9417, + "step": 18454 + }, + { + "epoch": 0.44, + "learning_rate": 1.228515673997526e-05, + "loss": 1.058, + "step": 18455 + }, + { + "epoch": 0.44, + "learning_rate": 1.2284398448012704e-05, + "loss": 0.9239, + "step": 18456 + }, + { + "epoch": 0.44, + "learning_rate": 1.2283640142191238e-05, + "loss": 1.1164, + "step": 18457 + }, + { + "epoch": 0.44, + "learning_rate": 1.2282881822515464e-05, + "loss": 1.1471, + "step": 18458 + }, + { + "epoch": 0.44, + "learning_rate": 1.2282123488989983e-05, + "loss": 1.0684, + "step": 18459 + }, + { + "epoch": 0.44, + "learning_rate": 1.2281365141619397e-05, + "loss": 1.21, + "step": 18460 + }, + { + "epoch": 0.44, + "learning_rate": 1.2280606780408304e-05, + "loss": 1.0916, + "step": 18461 + }, + { + "epoch": 0.44, + "learning_rate": 1.2279848405361307e-05, + "loss": 1.0022, + "step": 18462 + }, + { + "epoch": 0.44, + "learning_rate": 1.2279090016483004e-05, + "loss": 0.9748, + "step": 18463 + }, + { + "epoch": 0.44, + "learning_rate": 1.2278331613778e-05, + "loss": 1.3252, + "step": 18464 + }, + { + "epoch": 0.44, + "learning_rate": 1.2277573197250894e-05, + "loss": 1.1922, + "step": 18465 + }, + { + "epoch": 0.44, + "learning_rate": 1.227681476690629e-05, + "loss": 1.1332, + "step": 18466 + }, + { + "epoch": 0.44, + "learning_rate": 1.227605632274878e-05, + "loss": 1.076, + "step": 18467 + }, + { + "epoch": 0.44, + "learning_rate": 1.2275297864782974e-05, + "loss": 1.0643, + "step": 18468 + }, + { + "epoch": 0.44, + "learning_rate": 1.2274539393013471e-05, + "loss": 1.1248, + "step": 18469 + }, + { + "epoch": 0.44, + "learning_rate": 1.2273780907444875e-05, + "loss": 0.8024, + "step": 18470 + }, + { + "epoch": 0.44, + "learning_rate": 1.2273022408081783e-05, + "loss": 1.0515, + "step": 18471 + }, + { + "epoch": 0.44, + "learning_rate": 1.2272263894928798e-05, + "loss": 1.0658, + "step": 18472 + }, + { + "epoch": 0.44, + "learning_rate": 1.2271505367990524e-05, + "loss": 1.2025, + "step": 18473 + }, + { + "epoch": 0.44, + "learning_rate": 1.2270746827271562e-05, + "loss": 0.9645, + "step": 18474 + }, + { + "epoch": 0.44, + "learning_rate": 1.226998827277651e-05, + "loss": 0.9711, + "step": 18475 + }, + { + "epoch": 0.44, + "learning_rate": 1.2269229704509974e-05, + "loss": 1.0436, + "step": 18476 + }, + { + "epoch": 0.44, + "learning_rate": 1.2268471122476558e-05, + "loss": 0.9958, + "step": 18477 + }, + { + "epoch": 0.44, + "learning_rate": 1.2267712526680859e-05, + "loss": 1.0273, + "step": 18478 + }, + { + "epoch": 0.44, + "learning_rate": 1.226695391712748e-05, + "loss": 1.09, + "step": 18479 + }, + { + "epoch": 0.44, + "learning_rate": 1.2266195293821027e-05, + "loss": 1.1533, + "step": 18480 + }, + { + "epoch": 0.44, + "learning_rate": 1.2265436656766101e-05, + "loss": 1.2239, + "step": 18481 + }, + { + "epoch": 0.44, + "learning_rate": 1.2264678005967303e-05, + "loss": 1.1333, + "step": 18482 + }, + { + "epoch": 0.44, + "learning_rate": 1.2263919341429233e-05, + "loss": 1.0677, + "step": 18483 + }, + { + "epoch": 0.44, + "learning_rate": 1.2263160663156501e-05, + "loss": 1.1624, + "step": 18484 + }, + { + "epoch": 0.44, + "learning_rate": 1.2262401971153704e-05, + "loss": 1.062, + "step": 18485 + }, + { + "epoch": 0.44, + "learning_rate": 1.2261643265425447e-05, + "loss": 1.4648, + "step": 18486 + }, + { + "epoch": 0.44, + "learning_rate": 1.226088454597633e-05, + "loss": 0.9608, + "step": 18487 + }, + { + "epoch": 0.44, + "learning_rate": 1.2260125812810961e-05, + "loss": 1.3645, + "step": 18488 + }, + { + "epoch": 0.44, + "learning_rate": 1.225936706593394e-05, + "loss": 0.8016, + "step": 18489 + }, + { + "epoch": 0.44, + "learning_rate": 1.225860830534987e-05, + "loss": 1.2066, + "step": 18490 + }, + { + "epoch": 0.44, + "learning_rate": 1.2257849531063353e-05, + "loss": 0.9655, + "step": 18491 + }, + { + "epoch": 0.44, + "learning_rate": 1.2257090743078997e-05, + "loss": 1.1034, + "step": 18492 + }, + { + "epoch": 0.44, + "learning_rate": 1.2256331941401401e-05, + "loss": 1.309, + "step": 18493 + }, + { + "epoch": 0.44, + "learning_rate": 1.225557312603517e-05, + "loss": 1.279, + "step": 18494 + }, + { + "epoch": 0.44, + "learning_rate": 1.2254814296984905e-05, + "loss": 1.0953, + "step": 18495 + }, + { + "epoch": 0.44, + "learning_rate": 1.2254055454255213e-05, + "loss": 0.8887, + "step": 18496 + }, + { + "epoch": 0.44, + "learning_rate": 1.2253296597850698e-05, + "loss": 1.1339, + "step": 18497 + }, + { + "epoch": 0.44, + "learning_rate": 1.2252537727775965e-05, + "loss": 1.2517, + "step": 18498 + }, + { + "epoch": 0.44, + "learning_rate": 1.2251778844035611e-05, + "loss": 1.0958, + "step": 18499 + }, + { + "epoch": 0.44, + "learning_rate": 1.2251019946634246e-05, + "loss": 1.0715, + "step": 18500 + }, + { + "epoch": 0.44, + "learning_rate": 1.2250261035576472e-05, + "loss": 1.1933, + "step": 18501 + }, + { + "epoch": 0.44, + "learning_rate": 1.2249502110866895e-05, + "loss": 0.8426, + "step": 18502 + }, + { + "epoch": 0.44, + "learning_rate": 1.2248743172510116e-05, + "loss": 0.8749, + "step": 18503 + }, + { + "epoch": 0.45, + "learning_rate": 1.2247984220510742e-05, + "loss": 0.9193, + "step": 18504 + }, + { + "epoch": 0.45, + "learning_rate": 1.2247225254873378e-05, + "loss": 0.9992, + "step": 18505 + }, + { + "epoch": 0.45, + "learning_rate": 1.2246466275602624e-05, + "loss": 0.9273, + "step": 18506 + }, + { + "epoch": 0.45, + "learning_rate": 1.224570728270309e-05, + "loss": 1.1743, + "step": 18507 + }, + { + "epoch": 0.45, + "learning_rate": 1.2244948276179378e-05, + "loss": 1.1225, + "step": 18508 + }, + { + "epoch": 0.45, + "learning_rate": 1.2244189256036092e-05, + "loss": 1.099, + "step": 18509 + }, + { + "epoch": 0.45, + "learning_rate": 1.2243430222277835e-05, + "loss": 0.8315, + "step": 18510 + }, + { + "epoch": 0.45, + "learning_rate": 1.2242671174909218e-05, + "loss": 1.0092, + "step": 18511 + }, + { + "epoch": 0.45, + "learning_rate": 1.2241912113934842e-05, + "loss": 0.9475, + "step": 18512 + }, + { + "epoch": 0.45, + "learning_rate": 1.2241153039359312e-05, + "loss": 1.0926, + "step": 18513 + }, + { + "epoch": 0.45, + "learning_rate": 1.2240393951187232e-05, + "loss": 1.2706, + "step": 18514 + }, + { + "epoch": 0.45, + "learning_rate": 1.2239634849423213e-05, + "loss": 1.1185, + "step": 18515 + }, + { + "epoch": 0.45, + "learning_rate": 1.2238875734071853e-05, + "loss": 1.2058, + "step": 18516 + }, + { + "epoch": 0.45, + "learning_rate": 1.2238116605137764e-05, + "loss": 0.9063, + "step": 18517 + }, + { + "epoch": 0.45, + "learning_rate": 1.2237357462625543e-05, + "loss": 0.9725, + "step": 18518 + }, + { + "epoch": 0.45, + "learning_rate": 1.2236598306539804e-05, + "loss": 1.1931, + "step": 18519 + }, + { + "epoch": 0.45, + "learning_rate": 1.2235839136885148e-05, + "loss": 1.0475, + "step": 18520 + }, + { + "epoch": 0.45, + "learning_rate": 1.2235079953666183e-05, + "loss": 1.136, + "step": 18521 + }, + { + "epoch": 0.45, + "learning_rate": 1.2234320756887512e-05, + "loss": 1.2073, + "step": 18522 + }, + { + "epoch": 0.45, + "learning_rate": 1.2233561546553743e-05, + "loss": 1.3313, + "step": 18523 + }, + { + "epoch": 0.45, + "learning_rate": 1.2232802322669481e-05, + "loss": 1.0516, + "step": 18524 + }, + { + "epoch": 0.45, + "learning_rate": 1.2232043085239336e-05, + "loss": 1.0426, + "step": 18525 + }, + { + "epoch": 0.45, + "learning_rate": 1.2231283834267904e-05, + "loss": 0.9254, + "step": 18526 + }, + { + "epoch": 0.45, + "learning_rate": 1.2230524569759804e-05, + "loss": 1.0124, + "step": 18527 + }, + { + "epoch": 0.45, + "learning_rate": 1.2229765291719633e-05, + "loss": 1.0127, + "step": 18528 + }, + { + "epoch": 0.45, + "learning_rate": 1.2229006000152002e-05, + "loss": 1.3777, + "step": 18529 + }, + { + "epoch": 0.45, + "learning_rate": 1.2228246695061511e-05, + "loss": 0.8562, + "step": 18530 + }, + { + "epoch": 0.45, + "learning_rate": 1.2227487376452777e-05, + "loss": 1.2555, + "step": 18531 + }, + { + "epoch": 0.45, + "learning_rate": 1.22267280443304e-05, + "loss": 1.1717, + "step": 18532 + }, + { + "epoch": 0.45, + "learning_rate": 1.222596869869899e-05, + "loss": 1.2539, + "step": 18533 + }, + { + "epoch": 0.45, + "learning_rate": 1.2225209339563144e-05, + "loss": 1.2073, + "step": 18534 + }, + { + "epoch": 0.45, + "learning_rate": 1.2224449966927483e-05, + "loss": 1.4501, + "step": 18535 + }, + { + "epoch": 0.45, + "learning_rate": 1.2223690580796606e-05, + "loss": 1.4245, + "step": 18536 + }, + { + "epoch": 0.45, + "learning_rate": 1.2222931181175122e-05, + "loss": 1.0676, + "step": 18537 + }, + { + "epoch": 0.45, + "learning_rate": 1.2222171768067632e-05, + "loss": 1.0996, + "step": 18538 + }, + { + "epoch": 0.45, + "learning_rate": 1.2221412341478753e-05, + "loss": 1.0347, + "step": 18539 + }, + { + "epoch": 0.45, + "learning_rate": 1.2220652901413086e-05, + "loss": 1.0777, + "step": 18540 + }, + { + "epoch": 0.45, + "learning_rate": 1.2219893447875241e-05, + "loss": 1.2509, + "step": 18541 + }, + { + "epoch": 0.45, + "learning_rate": 1.2219133980869828e-05, + "loss": 1.2452, + "step": 18542 + }, + { + "epoch": 0.45, + "learning_rate": 1.2218374500401446e-05, + "loss": 1.3024, + "step": 18543 + }, + { + "epoch": 0.45, + "learning_rate": 1.221761500647471e-05, + "loss": 1.0612, + "step": 18544 + }, + { + "epoch": 0.45, + "learning_rate": 1.2216855499094227e-05, + "loss": 1.213, + "step": 18545 + }, + { + "epoch": 0.45, + "learning_rate": 1.22160959782646e-05, + "loss": 1.2078, + "step": 18546 + }, + { + "epoch": 0.45, + "learning_rate": 1.221533644399044e-05, + "loss": 1.0513, + "step": 18547 + }, + { + "epoch": 0.45, + "learning_rate": 1.2214576896276357e-05, + "loss": 1.3614, + "step": 18548 + }, + { + "epoch": 0.45, + "learning_rate": 1.2213817335126956e-05, + "loss": 1.2571, + "step": 18549 + }, + { + "epoch": 0.45, + "learning_rate": 1.2213057760546845e-05, + "loss": 0.9655, + "step": 18550 + }, + { + "epoch": 0.45, + "learning_rate": 1.2212298172540635e-05, + "loss": 1.241, + "step": 18551 + }, + { + "epoch": 0.45, + "learning_rate": 1.2211538571112933e-05, + "loss": 1.0853, + "step": 18552 + }, + { + "epoch": 0.45, + "learning_rate": 1.2210778956268345e-05, + "loss": 1.0466, + "step": 18553 + }, + { + "epoch": 0.45, + "learning_rate": 1.221001932801148e-05, + "loss": 1.2465, + "step": 18554 + }, + { + "epoch": 0.45, + "learning_rate": 1.220925968634695e-05, + "loss": 0.9562, + "step": 18555 + }, + { + "epoch": 0.45, + "learning_rate": 1.220850003127936e-05, + "loss": 1.2782, + "step": 18556 + }, + { + "epoch": 0.45, + "learning_rate": 1.220774036281332e-05, + "loss": 1.2541, + "step": 18557 + }, + { + "epoch": 0.45, + "learning_rate": 1.2206980680953438e-05, + "loss": 0.9348, + "step": 18558 + }, + { + "epoch": 0.45, + "learning_rate": 1.2206220985704323e-05, + "loss": 0.9659, + "step": 18559 + }, + { + "epoch": 0.45, + "learning_rate": 1.2205461277070584e-05, + "loss": 1.2289, + "step": 18560 + }, + { + "epoch": 0.45, + "learning_rate": 1.2204701555056833e-05, + "loss": 1.0935, + "step": 18561 + }, + { + "epoch": 0.45, + "learning_rate": 1.2203941819667673e-05, + "loss": 1.1141, + "step": 18562 + }, + { + "epoch": 0.45, + "learning_rate": 1.2203182070907718e-05, + "loss": 1.0375, + "step": 18563 + }, + { + "epoch": 0.45, + "learning_rate": 1.2202422308781574e-05, + "loss": 1.0912, + "step": 18564 + }, + { + "epoch": 0.45, + "learning_rate": 1.2201662533293854e-05, + "loss": 0.896, + "step": 18565 + }, + { + "epoch": 0.45, + "learning_rate": 1.2200902744449164e-05, + "loss": 1.1757, + "step": 18566 + }, + { + "epoch": 0.45, + "learning_rate": 1.2200142942252116e-05, + "loss": 1.1737, + "step": 18567 + }, + { + "epoch": 0.45, + "learning_rate": 1.2199383126707315e-05, + "loss": 0.9411, + "step": 18568 + }, + { + "epoch": 0.45, + "learning_rate": 1.2198623297819378e-05, + "loss": 0.83, + "step": 18569 + }, + { + "epoch": 0.45, + "learning_rate": 1.2197863455592906e-05, + "loss": 1.0933, + "step": 18570 + }, + { + "epoch": 0.45, + "learning_rate": 1.2197103600032518e-05, + "loss": 1.3016, + "step": 18571 + }, + { + "epoch": 0.45, + "learning_rate": 1.2196343731142819e-05, + "loss": 1.0456, + "step": 18572 + }, + { + "epoch": 0.45, + "learning_rate": 1.2195583848928418e-05, + "loss": 0.6398, + "step": 18573 + }, + { + "epoch": 0.45, + "learning_rate": 1.2194823953393924e-05, + "loss": 0.9144, + "step": 18574 + }, + { + "epoch": 0.45, + "learning_rate": 1.2194064044543952e-05, + "loss": 1.2627, + "step": 18575 + }, + { + "epoch": 0.45, + "learning_rate": 1.219330412238311e-05, + "loss": 1.2791, + "step": 18576 + }, + { + "epoch": 0.45, + "learning_rate": 1.2192544186916007e-05, + "loss": 1.227, + "step": 18577 + }, + { + "epoch": 0.45, + "learning_rate": 1.219178423814725e-05, + "loss": 1.1608, + "step": 18578 + }, + { + "epoch": 0.45, + "learning_rate": 1.2191024276081458e-05, + "loss": 1.0432, + "step": 18579 + }, + { + "epoch": 0.45, + "learning_rate": 1.2190264300723238e-05, + "loss": 1.0715, + "step": 18580 + }, + { + "epoch": 0.45, + "learning_rate": 1.2189504312077199e-05, + "loss": 1.0468, + "step": 18581 + }, + { + "epoch": 0.45, + "learning_rate": 1.2188744310147949e-05, + "loss": 1.1768, + "step": 18582 + }, + { + "epoch": 0.45, + "learning_rate": 1.2187984294940105e-05, + "loss": 1.053, + "step": 18583 + }, + { + "epoch": 0.45, + "learning_rate": 1.2187224266458276e-05, + "loss": 1.2874, + "step": 18584 + }, + { + "epoch": 0.45, + "learning_rate": 1.2186464224707071e-05, + "loss": 1.0664, + "step": 18585 + }, + { + "epoch": 0.45, + "learning_rate": 1.2185704169691098e-05, + "loss": 1.0044, + "step": 18586 + }, + { + "epoch": 0.45, + "learning_rate": 1.2184944101414975e-05, + "loss": 0.9557, + "step": 18587 + }, + { + "epoch": 0.45, + "learning_rate": 1.2184184019883311e-05, + "loss": 1.0812, + "step": 18588 + }, + { + "epoch": 0.45, + "learning_rate": 1.2183423925100717e-05, + "loss": 1.0797, + "step": 18589 + }, + { + "epoch": 0.45, + "learning_rate": 1.2182663817071801e-05, + "loss": 1.1592, + "step": 18590 + }, + { + "epoch": 0.45, + "learning_rate": 1.2181903695801177e-05, + "loss": 1.2198, + "step": 18591 + }, + { + "epoch": 0.45, + "learning_rate": 1.2181143561293459e-05, + "loss": 1.3439, + "step": 18592 + }, + { + "epoch": 0.45, + "learning_rate": 1.2180383413553252e-05, + "loss": 0.997, + "step": 18593 + }, + { + "epoch": 0.45, + "learning_rate": 1.2179623252585175e-05, + "loss": 1.193, + "step": 18594 + }, + { + "epoch": 0.45, + "learning_rate": 1.2178863078393834e-05, + "loss": 0.9668, + "step": 18595 + }, + { + "epoch": 0.45, + "learning_rate": 1.2178102890983847e-05, + "loss": 1.0912, + "step": 18596 + }, + { + "epoch": 0.45, + "learning_rate": 1.2177342690359817e-05, + "loss": 1.3017, + "step": 18597 + }, + { + "epoch": 0.45, + "learning_rate": 1.2176582476526366e-05, + "loss": 1.1007, + "step": 18598 + }, + { + "epoch": 0.45, + "learning_rate": 1.2175822249488098e-05, + "loss": 1.2698, + "step": 18599 + }, + { + "epoch": 0.45, + "learning_rate": 1.2175062009249631e-05, + "loss": 0.8266, + "step": 18600 + }, + { + "epoch": 0.45, + "learning_rate": 1.2174301755815572e-05, + "loss": 0.9863, + "step": 18601 + }, + { + "epoch": 0.45, + "learning_rate": 1.2173541489190535e-05, + "loss": 0.9521, + "step": 18602 + }, + { + "epoch": 0.45, + "learning_rate": 1.2172781209379134e-05, + "loss": 1.1216, + "step": 18603 + }, + { + "epoch": 0.45, + "learning_rate": 1.2172020916385983e-05, + "loss": 1.1379, + "step": 18604 + }, + { + "epoch": 0.45, + "learning_rate": 1.2171260610215687e-05, + "loss": 0.7604, + "step": 18605 + }, + { + "epoch": 0.45, + "learning_rate": 1.2170500290872869e-05, + "loss": 1.1516, + "step": 18606 + }, + { + "epoch": 0.45, + "learning_rate": 1.2169739958362133e-05, + "loss": 1.2252, + "step": 18607 + }, + { + "epoch": 0.45, + "learning_rate": 1.2168979612688097e-05, + "loss": 1.1474, + "step": 18608 + }, + { + "epoch": 0.45, + "learning_rate": 1.2168219253855366e-05, + "loss": 1.2349, + "step": 18609 + }, + { + "epoch": 0.45, + "learning_rate": 1.2167458881868566e-05, + "loss": 0.9692, + "step": 18610 + }, + { + "epoch": 0.45, + "learning_rate": 1.21666984967323e-05, + "loss": 1.1104, + "step": 18611 + }, + { + "epoch": 0.45, + "learning_rate": 1.2165938098451184e-05, + "loss": 1.2785, + "step": 18612 + }, + { + "epoch": 0.45, + "learning_rate": 1.216517768702983e-05, + "loss": 1.1838, + "step": 18613 + }, + { + "epoch": 0.45, + "learning_rate": 1.2164417262472854e-05, + "loss": 1.1038, + "step": 18614 + }, + { + "epoch": 0.45, + "learning_rate": 1.2163656824784866e-05, + "loss": 1.2376, + "step": 18615 + }, + { + "epoch": 0.45, + "learning_rate": 1.2162896373970483e-05, + "loss": 0.7953, + "step": 18616 + }, + { + "epoch": 0.45, + "learning_rate": 1.2162135910034313e-05, + "loss": 0.7878, + "step": 18617 + }, + { + "epoch": 0.45, + "learning_rate": 1.2161375432980976e-05, + "loss": 0.8555, + "step": 18618 + }, + { + "epoch": 0.45, + "learning_rate": 1.2160614942815084e-05, + "loss": 1.1192, + "step": 18619 + }, + { + "epoch": 0.45, + "learning_rate": 1.2159854439541245e-05, + "loss": 0.9463, + "step": 18620 + }, + { + "epoch": 0.45, + "learning_rate": 1.2159093923164078e-05, + "loss": 1.3127, + "step": 18621 + }, + { + "epoch": 0.45, + "learning_rate": 1.21583333936882e-05, + "loss": 1.3091, + "step": 18622 + }, + { + "epoch": 0.45, + "learning_rate": 1.2157572851118218e-05, + "loss": 1.0637, + "step": 18623 + }, + { + "epoch": 0.45, + "learning_rate": 1.215681229545875e-05, + "loss": 1.1676, + "step": 18624 + }, + { + "epoch": 0.45, + "learning_rate": 1.2156051726714407e-05, + "loss": 0.974, + "step": 18625 + }, + { + "epoch": 0.45, + "learning_rate": 1.2155291144889804e-05, + "loss": 1.0029, + "step": 18626 + }, + { + "epoch": 0.45, + "learning_rate": 1.215453054998956e-05, + "loss": 0.9716, + "step": 18627 + }, + { + "epoch": 0.45, + "learning_rate": 1.2153769942018286e-05, + "loss": 1.1016, + "step": 18628 + }, + { + "epoch": 0.45, + "learning_rate": 1.2153009320980596e-05, + "loss": 0.819, + "step": 18629 + }, + { + "epoch": 0.45, + "learning_rate": 1.2152248686881102e-05, + "loss": 1.0405, + "step": 18630 + }, + { + "epoch": 0.45, + "learning_rate": 1.2151488039724427e-05, + "loss": 1.2252, + "step": 18631 + }, + { + "epoch": 0.45, + "learning_rate": 1.2150727379515176e-05, + "loss": 0.9713, + "step": 18632 + }, + { + "epoch": 0.45, + "learning_rate": 1.2149966706257968e-05, + "loss": 1.1521, + "step": 18633 + }, + { + "epoch": 0.45, + "learning_rate": 1.2149206019957418e-05, + "loss": 0.8812, + "step": 18634 + }, + { + "epoch": 0.45, + "learning_rate": 1.2148445320618142e-05, + "loss": 1.2039, + "step": 18635 + }, + { + "epoch": 0.45, + "learning_rate": 1.2147684608244754e-05, + "loss": 1.0349, + "step": 18636 + }, + { + "epoch": 0.45, + "learning_rate": 1.2146923882841865e-05, + "loss": 1.2404, + "step": 18637 + }, + { + "epoch": 0.45, + "learning_rate": 1.2146163144414097e-05, + "loss": 1.14, + "step": 18638 + }, + { + "epoch": 0.45, + "learning_rate": 1.2145402392966061e-05, + "loss": 1.1609, + "step": 18639 + }, + { + "epoch": 0.45, + "learning_rate": 1.2144641628502373e-05, + "loss": 0.8791, + "step": 18640 + }, + { + "epoch": 0.45, + "learning_rate": 1.2143880851027649e-05, + "loss": 0.7422, + "step": 18641 + }, + { + "epoch": 0.45, + "learning_rate": 1.2143120060546505e-05, + "loss": 1.0611, + "step": 18642 + }, + { + "epoch": 0.45, + "learning_rate": 1.2142359257063556e-05, + "loss": 1.1014, + "step": 18643 + }, + { + "epoch": 0.45, + "learning_rate": 1.2141598440583417e-05, + "loss": 1.1111, + "step": 18644 + }, + { + "epoch": 0.45, + "learning_rate": 1.2140837611110702e-05, + "loss": 1.2174, + "step": 18645 + }, + { + "epoch": 0.45, + "learning_rate": 1.2140076768650031e-05, + "loss": 1.0529, + "step": 18646 + }, + { + "epoch": 0.45, + "learning_rate": 1.2139315913206016e-05, + "loss": 1.0585, + "step": 18647 + }, + { + "epoch": 0.45, + "learning_rate": 1.2138555044783279e-05, + "loss": 1.2496, + "step": 18648 + }, + { + "epoch": 0.45, + "learning_rate": 1.2137794163386424e-05, + "loss": 1.2957, + "step": 18649 + }, + { + "epoch": 0.45, + "learning_rate": 1.213703326902008e-05, + "loss": 1.2946, + "step": 18650 + }, + { + "epoch": 0.45, + "learning_rate": 1.2136272361688856e-05, + "loss": 1.1694, + "step": 18651 + }, + { + "epoch": 0.45, + "learning_rate": 1.2135511441397369e-05, + "loss": 1.0557, + "step": 18652 + }, + { + "epoch": 0.45, + "learning_rate": 1.2134750508150236e-05, + "loss": 1.091, + "step": 18653 + }, + { + "epoch": 0.45, + "learning_rate": 1.2133989561952077e-05, + "loss": 1.2446, + "step": 18654 + }, + { + "epoch": 0.45, + "learning_rate": 1.2133228602807505e-05, + "loss": 0.8403, + "step": 18655 + }, + { + "epoch": 0.45, + "learning_rate": 1.2132467630721137e-05, + "loss": 1.1687, + "step": 18656 + }, + { + "epoch": 0.45, + "learning_rate": 1.2131706645697587e-05, + "loss": 1.3237, + "step": 18657 + }, + { + "epoch": 0.45, + "learning_rate": 1.2130945647741474e-05, + "loss": 1.0907, + "step": 18658 + }, + { + "epoch": 0.45, + "learning_rate": 1.2130184636857418e-05, + "loss": 1.3644, + "step": 18659 + }, + { + "epoch": 0.45, + "learning_rate": 1.212942361305003e-05, + "loss": 1.165, + "step": 18660 + }, + { + "epoch": 0.45, + "learning_rate": 1.212866257632393e-05, + "loss": 1.1667, + "step": 18661 + }, + { + "epoch": 0.45, + "learning_rate": 1.2127901526683737e-05, + "loss": 0.6942, + "step": 18662 + }, + { + "epoch": 0.45, + "learning_rate": 1.2127140464134066e-05, + "loss": 1.0292, + "step": 18663 + }, + { + "epoch": 0.45, + "learning_rate": 1.2126379388679532e-05, + "loss": 0.9862, + "step": 18664 + }, + { + "epoch": 0.45, + "learning_rate": 1.2125618300324754e-05, + "loss": 1.0822, + "step": 18665 + }, + { + "epoch": 0.45, + "learning_rate": 1.2124857199074354e-05, + "loss": 1.1806, + "step": 18666 + }, + { + "epoch": 0.45, + "learning_rate": 1.2124096084932941e-05, + "loss": 1.1639, + "step": 18667 + }, + { + "epoch": 0.45, + "learning_rate": 1.212333495790514e-05, + "loss": 1.0706, + "step": 18668 + }, + { + "epoch": 0.45, + "learning_rate": 1.212257381799556e-05, + "loss": 1.2598, + "step": 18669 + }, + { + "epoch": 0.45, + "learning_rate": 1.2121812665208827e-05, + "loss": 1.0954, + "step": 18670 + }, + { + "epoch": 0.45, + "learning_rate": 1.2121051499549557e-05, + "loss": 1.1128, + "step": 18671 + }, + { + "epoch": 0.45, + "learning_rate": 1.2120290321022366e-05, + "loss": 0.8093, + "step": 18672 + }, + { + "epoch": 0.45, + "learning_rate": 1.2119529129631871e-05, + "loss": 1.0796, + "step": 18673 + }, + { + "epoch": 0.45, + "learning_rate": 1.211876792538269e-05, + "loss": 0.9825, + "step": 18674 + }, + { + "epoch": 0.45, + "learning_rate": 1.2118006708279446e-05, + "loss": 1.2364, + "step": 18675 + }, + { + "epoch": 0.45, + "learning_rate": 1.2117245478326752e-05, + "loss": 1.0241, + "step": 18676 + }, + { + "epoch": 0.45, + "learning_rate": 1.2116484235529227e-05, + "loss": 0.9705, + "step": 18677 + }, + { + "epoch": 0.45, + "learning_rate": 1.2115722979891488e-05, + "loss": 1.1015, + "step": 18678 + }, + { + "epoch": 0.45, + "learning_rate": 1.2114961711418161e-05, + "loss": 1.178, + "step": 18679 + }, + { + "epoch": 0.45, + "learning_rate": 1.2114200430113854e-05, + "loss": 1.3981, + "step": 18680 + }, + { + "epoch": 0.45, + "learning_rate": 1.2113439135983191e-05, + "loss": 0.9668, + "step": 18681 + }, + { + "epoch": 0.45, + "learning_rate": 1.2112677829030792e-05, + "loss": 1.1881, + "step": 18682 + }, + { + "epoch": 0.45, + "learning_rate": 1.2111916509261272e-05, + "loss": 0.9128, + "step": 18683 + }, + { + "epoch": 0.45, + "learning_rate": 1.211115517667925e-05, + "loss": 0.9655, + "step": 18684 + }, + { + "epoch": 0.45, + "learning_rate": 1.2110393831289348e-05, + "loss": 1.0737, + "step": 18685 + }, + { + "epoch": 0.45, + "learning_rate": 1.2109632473096183e-05, + "loss": 1.0884, + "step": 18686 + }, + { + "epoch": 0.45, + "learning_rate": 1.2108871102104373e-05, + "loss": 1.0013, + "step": 18687 + }, + { + "epoch": 0.45, + "learning_rate": 1.2108109718318537e-05, + "loss": 0.914, + "step": 18688 + }, + { + "epoch": 0.45, + "learning_rate": 1.21073483217433e-05, + "loss": 1.0392, + "step": 18689 + }, + { + "epoch": 0.45, + "learning_rate": 1.210658691238327e-05, + "loss": 0.966, + "step": 18690 + }, + { + "epoch": 0.45, + "learning_rate": 1.210582549024308e-05, + "loss": 1.0196, + "step": 18691 + }, + { + "epoch": 0.45, + "learning_rate": 1.2105064055327335e-05, + "loss": 1.18, + "step": 18692 + }, + { + "epoch": 0.45, + "learning_rate": 1.2104302607640666e-05, + "loss": 1.1192, + "step": 18693 + }, + { + "epoch": 0.45, + "learning_rate": 1.2103541147187684e-05, + "loss": 1.1703, + "step": 18694 + }, + { + "epoch": 0.45, + "learning_rate": 1.2102779673973018e-05, + "loss": 1.0279, + "step": 18695 + }, + { + "epoch": 0.45, + "learning_rate": 1.2102018188001278e-05, + "loss": 1.1938, + "step": 18696 + }, + { + "epoch": 0.45, + "learning_rate": 1.210125668927709e-05, + "loss": 1.3011, + "step": 18697 + }, + { + "epoch": 0.45, + "learning_rate": 1.2100495177805072e-05, + "loss": 0.7427, + "step": 18698 + }, + { + "epoch": 0.45, + "learning_rate": 1.2099733653589845e-05, + "loss": 1.0521, + "step": 18699 + }, + { + "epoch": 0.45, + "learning_rate": 1.2098972116636022e-05, + "loss": 1.1492, + "step": 18700 + }, + { + "epoch": 0.45, + "learning_rate": 1.2098210566948235e-05, + "loss": 0.8663, + "step": 18701 + }, + { + "epoch": 0.45, + "learning_rate": 1.2097449004531097e-05, + "loss": 1.0202, + "step": 18702 + }, + { + "epoch": 0.45, + "learning_rate": 1.209668742938923e-05, + "loss": 0.9407, + "step": 18703 + }, + { + "epoch": 0.45, + "learning_rate": 1.2095925841527249e-05, + "loss": 1.3157, + "step": 18704 + }, + { + "epoch": 0.45, + "learning_rate": 1.2095164240949781e-05, + "loss": 1.0651, + "step": 18705 + }, + { + "epoch": 0.45, + "learning_rate": 1.2094402627661447e-05, + "loss": 1.2543, + "step": 18706 + }, + { + "epoch": 0.45, + "learning_rate": 1.209364100166686e-05, + "loss": 0.9776, + "step": 18707 + }, + { + "epoch": 0.45, + "learning_rate": 1.2092879362970652e-05, + "loss": 0.9929, + "step": 18708 + }, + { + "epoch": 0.45, + "learning_rate": 1.209211771157743e-05, + "loss": 1.2564, + "step": 18709 + }, + { + "epoch": 0.45, + "learning_rate": 1.2091356047491827e-05, + "loss": 1.2644, + "step": 18710 + }, + { + "epoch": 0.45, + "learning_rate": 1.2090594370718458e-05, + "loss": 1.2142, + "step": 18711 + }, + { + "epoch": 0.45, + "learning_rate": 1.2089832681261943e-05, + "loss": 1.1901, + "step": 18712 + }, + { + "epoch": 0.45, + "learning_rate": 1.2089070979126904e-05, + "loss": 1.1969, + "step": 18713 + }, + { + "epoch": 0.45, + "learning_rate": 1.2088309264317965e-05, + "loss": 0.9878, + "step": 18714 + }, + { + "epoch": 0.45, + "learning_rate": 1.2087547536839745e-05, + "loss": 1.1553, + "step": 18715 + }, + { + "epoch": 0.45, + "learning_rate": 1.2086785796696866e-05, + "loss": 1.309, + "step": 18716 + }, + { + "epoch": 0.45, + "learning_rate": 1.2086024043893945e-05, + "loss": 1.0459, + "step": 18717 + }, + { + "epoch": 0.45, + "learning_rate": 1.208526227843561e-05, + "loss": 1.122, + "step": 18718 + }, + { + "epoch": 0.45, + "learning_rate": 1.208450050032648e-05, + "loss": 0.8743, + "step": 18719 + }, + { + "epoch": 0.45, + "learning_rate": 1.2083738709571172e-05, + "loss": 0.8593, + "step": 18720 + }, + { + "epoch": 0.45, + "learning_rate": 1.2082976906174313e-05, + "loss": 1.1872, + "step": 18721 + }, + { + "epoch": 0.45, + "learning_rate": 1.2082215090140523e-05, + "loss": 1.1643, + "step": 18722 + }, + { + "epoch": 0.45, + "learning_rate": 1.2081453261474425e-05, + "loss": 1.0176, + "step": 18723 + }, + { + "epoch": 0.45, + "learning_rate": 1.2080691420180637e-05, + "loss": 1.0981, + "step": 18724 + }, + { + "epoch": 0.45, + "learning_rate": 1.2079929566263784e-05, + "loss": 0.9393, + "step": 18725 + }, + { + "epoch": 0.45, + "learning_rate": 1.207916769972849e-05, + "loss": 1.0315, + "step": 18726 + }, + { + "epoch": 0.45, + "learning_rate": 1.2078405820579373e-05, + "loss": 1.0432, + "step": 18727 + }, + { + "epoch": 0.45, + "learning_rate": 1.2077643928821056e-05, + "loss": 0.8885, + "step": 18728 + }, + { + "epoch": 0.45, + "learning_rate": 1.2076882024458163e-05, + "loss": 1.0073, + "step": 18729 + }, + { + "epoch": 0.45, + "learning_rate": 1.2076120107495317e-05, + "loss": 1.1253, + "step": 18730 + }, + { + "epoch": 0.45, + "learning_rate": 1.2075358177937138e-05, + "loss": 1.0921, + "step": 18731 + }, + { + "epoch": 0.45, + "learning_rate": 1.2074596235788245e-05, + "loss": 0.9525, + "step": 18732 + }, + { + "epoch": 0.45, + "learning_rate": 1.207383428105327e-05, + "loss": 1.0372, + "step": 18733 + }, + { + "epoch": 0.45, + "learning_rate": 1.2073072313736826e-05, + "loss": 0.8058, + "step": 18734 + }, + { + "epoch": 0.45, + "learning_rate": 1.2072310333843544e-05, + "loss": 1.1676, + "step": 18735 + }, + { + "epoch": 0.45, + "learning_rate": 1.207154834137804e-05, + "loss": 1.0331, + "step": 18736 + }, + { + "epoch": 0.45, + "learning_rate": 1.207078633634494e-05, + "loss": 0.9755, + "step": 18737 + }, + { + "epoch": 0.45, + "learning_rate": 1.2070024318748866e-05, + "loss": 1.173, + "step": 18738 + }, + { + "epoch": 0.45, + "learning_rate": 1.2069262288594443e-05, + "loss": 0.9573, + "step": 18739 + }, + { + "epoch": 0.45, + "learning_rate": 1.2068500245886288e-05, + "loss": 1.0273, + "step": 18740 + }, + { + "epoch": 0.45, + "learning_rate": 1.2067738190629032e-05, + "loss": 0.8843, + "step": 18741 + }, + { + "epoch": 0.45, + "learning_rate": 1.2066976122827296e-05, + "loss": 1.1641, + "step": 18742 + }, + { + "epoch": 0.45, + "learning_rate": 1.20662140424857e-05, + "loss": 1.0142, + "step": 18743 + }, + { + "epoch": 0.45, + "learning_rate": 1.2065451949608868e-05, + "loss": 1.1653, + "step": 18744 + }, + { + "epoch": 0.45, + "learning_rate": 1.206468984420143e-05, + "loss": 1.0752, + "step": 18745 + }, + { + "epoch": 0.45, + "learning_rate": 1.2063927726268001e-05, + "loss": 1.1931, + "step": 18746 + }, + { + "epoch": 0.45, + "learning_rate": 1.2063165595813206e-05, + "loss": 1.0377, + "step": 18747 + }, + { + "epoch": 0.45, + "learning_rate": 1.2062403452841674e-05, + "loss": 1.1587, + "step": 18748 + }, + { + "epoch": 0.45, + "learning_rate": 1.2061641297358021e-05, + "loss": 1.2533, + "step": 18749 + }, + { + "epoch": 0.45, + "learning_rate": 1.206087912936688e-05, + "loss": 1.0394, + "step": 18750 + }, + { + "epoch": 0.45, + "learning_rate": 1.206011694887287e-05, + "loss": 1.3254, + "step": 18751 + }, + { + "epoch": 0.45, + "learning_rate": 1.205935475588061e-05, + "loss": 1.0508, + "step": 18752 + }, + { + "epoch": 0.45, + "learning_rate": 1.2058592550394735e-05, + "loss": 0.9922, + "step": 18753 + }, + { + "epoch": 0.45, + "learning_rate": 1.205783033241986e-05, + "loss": 1.2365, + "step": 18754 + }, + { + "epoch": 0.45, + "learning_rate": 1.2057068101960615e-05, + "loss": 0.9923, + "step": 18755 + }, + { + "epoch": 0.45, + "learning_rate": 1.2056305859021615e-05, + "loss": 1.0274, + "step": 18756 + }, + { + "epoch": 0.45, + "learning_rate": 1.2055543603607499e-05, + "loss": 0.7581, + "step": 18757 + }, + { + "epoch": 0.45, + "learning_rate": 1.2054781335722879e-05, + "loss": 1.2523, + "step": 18758 + }, + { + "epoch": 0.45, + "learning_rate": 1.2054019055372386e-05, + "loss": 1.0165, + "step": 18759 + }, + { + "epoch": 0.45, + "learning_rate": 1.2053256762560638e-05, + "loss": 1.196, + "step": 18760 + }, + { + "epoch": 0.45, + "learning_rate": 1.205249445729227e-05, + "loss": 1.2969, + "step": 18761 + }, + { + "epoch": 0.45, + "learning_rate": 1.20517321395719e-05, + "loss": 0.9777, + "step": 18762 + }, + { + "epoch": 0.45, + "learning_rate": 1.205096980940415e-05, + "loss": 0.9733, + "step": 18763 + }, + { + "epoch": 0.45, + "learning_rate": 1.2050207466793651e-05, + "loss": 0.9911, + "step": 18764 + }, + { + "epoch": 0.45, + "learning_rate": 1.2049445111745027e-05, + "loss": 1.3089, + "step": 18765 + }, + { + "epoch": 0.45, + "learning_rate": 1.20486827442629e-05, + "loss": 0.7899, + "step": 18766 + }, + { + "epoch": 0.45, + "learning_rate": 1.2047920364351895e-05, + "loss": 1.1965, + "step": 18767 + }, + { + "epoch": 0.45, + "learning_rate": 1.2047157972016641e-05, + "loss": 1.0313, + "step": 18768 + }, + { + "epoch": 0.45, + "learning_rate": 1.2046395567261761e-05, + "loss": 1.0728, + "step": 18769 + }, + { + "epoch": 0.45, + "learning_rate": 1.204563315009188e-05, + "loss": 1.0204, + "step": 18770 + }, + { + "epoch": 0.45, + "learning_rate": 1.2044870720511624e-05, + "loss": 0.643, + "step": 18771 + }, + { + "epoch": 0.45, + "learning_rate": 1.204410827852562e-05, + "loss": 1.1194, + "step": 18772 + }, + { + "epoch": 0.45, + "learning_rate": 1.2043345824138489e-05, + "loss": 0.9304, + "step": 18773 + }, + { + "epoch": 0.45, + "learning_rate": 1.204258335735486e-05, + "loss": 1.2923, + "step": 18774 + }, + { + "epoch": 0.45, + "learning_rate": 1.2041820878179359e-05, + "loss": 1.1462, + "step": 18775 + }, + { + "epoch": 0.45, + "learning_rate": 1.204105838661661e-05, + "loss": 1.0048, + "step": 18776 + }, + { + "epoch": 0.45, + "learning_rate": 1.2040295882671242e-05, + "loss": 1.0103, + "step": 18777 + }, + { + "epoch": 0.45, + "learning_rate": 1.2039533366347878e-05, + "loss": 1.1848, + "step": 18778 + }, + { + "epoch": 0.45, + "learning_rate": 1.2038770837651143e-05, + "loss": 1.0709, + "step": 18779 + }, + { + "epoch": 0.45, + "learning_rate": 1.2038008296585667e-05, + "loss": 1.2013, + "step": 18780 + }, + { + "epoch": 0.45, + "learning_rate": 1.2037245743156074e-05, + "loss": 0.9764, + "step": 18781 + }, + { + "epoch": 0.45, + "learning_rate": 1.2036483177366989e-05, + "loss": 0.7531, + "step": 18782 + }, + { + "epoch": 0.45, + "learning_rate": 1.203572059922304e-05, + "loss": 0.9922, + "step": 18783 + }, + { + "epoch": 0.45, + "learning_rate": 1.2034958008728852e-05, + "loss": 1.1855, + "step": 18784 + }, + { + "epoch": 0.45, + "learning_rate": 1.2034195405889053e-05, + "loss": 0.9952, + "step": 18785 + }, + { + "epoch": 0.45, + "learning_rate": 1.203343279070827e-05, + "loss": 1.0106, + "step": 18786 + }, + { + "epoch": 0.45, + "learning_rate": 1.2032670163191127e-05, + "loss": 1.3053, + "step": 18787 + }, + { + "epoch": 0.45, + "learning_rate": 1.2031907523342254e-05, + "loss": 1.1873, + "step": 18788 + }, + { + "epoch": 0.45, + "learning_rate": 1.2031144871166276e-05, + "loss": 1.0538, + "step": 18789 + }, + { + "epoch": 0.45, + "learning_rate": 1.2030382206667816e-05, + "loss": 0.9358, + "step": 18790 + }, + { + "epoch": 0.45, + "learning_rate": 1.2029619529851509e-05, + "loss": 1.2788, + "step": 18791 + }, + { + "epoch": 0.45, + "learning_rate": 1.2028856840721975e-05, + "loss": 1.0838, + "step": 18792 + }, + { + "epoch": 0.45, + "learning_rate": 1.2028094139283846e-05, + "loss": 1.3435, + "step": 18793 + }, + { + "epoch": 0.45, + "learning_rate": 1.2027331425541745e-05, + "loss": 1.4864, + "step": 18794 + }, + { + "epoch": 0.45, + "learning_rate": 1.2026568699500304e-05, + "loss": 1.0522, + "step": 18795 + }, + { + "epoch": 0.45, + "learning_rate": 1.2025805961164143e-05, + "loss": 1.2446, + "step": 18796 + }, + { + "epoch": 0.45, + "learning_rate": 1.2025043210537896e-05, + "loss": 1.1644, + "step": 18797 + }, + { + "epoch": 0.45, + "learning_rate": 1.202428044762619e-05, + "loss": 1.2631, + "step": 18798 + }, + { + "epoch": 0.45, + "learning_rate": 1.202351767243365e-05, + "loss": 1.3121, + "step": 18799 + }, + { + "epoch": 0.45, + "learning_rate": 1.2022754884964902e-05, + "loss": 1.0312, + "step": 18800 + }, + { + "epoch": 0.45, + "learning_rate": 1.2021992085224578e-05, + "loss": 0.9608, + "step": 18801 + }, + { + "epoch": 0.45, + "learning_rate": 1.2021229273217303e-05, + "loss": 1.0469, + "step": 18802 + }, + { + "epoch": 0.45, + "learning_rate": 1.2020466448947703e-05, + "loss": 1.0659, + "step": 18803 + }, + { + "epoch": 0.45, + "learning_rate": 1.2019703612420412e-05, + "loss": 1.0245, + "step": 18804 + }, + { + "epoch": 0.45, + "learning_rate": 1.2018940763640052e-05, + "loss": 0.6183, + "step": 18805 + }, + { + "epoch": 0.45, + "learning_rate": 1.2018177902611256e-05, + "loss": 1.0451, + "step": 18806 + }, + { + "epoch": 0.45, + "learning_rate": 1.2017415029338647e-05, + "loss": 1.2334, + "step": 18807 + }, + { + "epoch": 0.45, + "learning_rate": 1.2016652143826855e-05, + "loss": 1.2711, + "step": 18808 + }, + { + "epoch": 0.45, + "learning_rate": 1.201588924608051e-05, + "loss": 1.1136, + "step": 18809 + }, + { + "epoch": 0.45, + "learning_rate": 1.201512633610424e-05, + "loss": 0.9768, + "step": 18810 + }, + { + "epoch": 0.45, + "learning_rate": 1.201436341390267e-05, + "loss": 1.0636, + "step": 18811 + }, + { + "epoch": 0.45, + "learning_rate": 1.2013600479480431e-05, + "loss": 0.9554, + "step": 18812 + }, + { + "epoch": 0.45, + "learning_rate": 1.2012837532842155e-05, + "loss": 1.0749, + "step": 18813 + }, + { + "epoch": 0.45, + "learning_rate": 1.2012074573992465e-05, + "loss": 1.1149, + "step": 18814 + }, + { + "epoch": 0.45, + "learning_rate": 1.2011311602935992e-05, + "loss": 1.0392, + "step": 18815 + }, + { + "epoch": 0.45, + "learning_rate": 1.2010548619677365e-05, + "loss": 1.0704, + "step": 18816 + }, + { + "epoch": 0.45, + "learning_rate": 1.200978562422121e-05, + "loss": 1.1152, + "step": 18817 + }, + { + "epoch": 0.45, + "learning_rate": 1.2009022616572162e-05, + "loss": 1.2403, + "step": 18818 + }, + { + "epoch": 0.45, + "learning_rate": 1.2008259596734843e-05, + "loss": 0.9409, + "step": 18819 + }, + { + "epoch": 0.45, + "learning_rate": 1.2007496564713888e-05, + "loss": 1.1368, + "step": 18820 + }, + { + "epoch": 0.45, + "learning_rate": 1.2006733520513923e-05, + "loss": 1.0595, + "step": 18821 + }, + { + "epoch": 0.45, + "learning_rate": 1.2005970464139579e-05, + "loss": 1.2116, + "step": 18822 + }, + { + "epoch": 0.45, + "learning_rate": 1.200520739559548e-05, + "loss": 1.163, + "step": 18823 + }, + { + "epoch": 0.45, + "learning_rate": 1.2004444314886262e-05, + "loss": 1.1062, + "step": 18824 + }, + { + "epoch": 0.45, + "learning_rate": 1.2003681222016553e-05, + "loss": 0.851, + "step": 18825 + }, + { + "epoch": 0.45, + "learning_rate": 1.2002918116990979e-05, + "loss": 0.9907, + "step": 18826 + }, + { + "epoch": 0.45, + "learning_rate": 1.2002154999814171e-05, + "loss": 1.1315, + "step": 18827 + }, + { + "epoch": 0.45, + "learning_rate": 1.2001391870490761e-05, + "loss": 1.2476, + "step": 18828 + }, + { + "epoch": 0.45, + "learning_rate": 1.2000628729025379e-05, + "loss": 1.1712, + "step": 18829 + }, + { + "epoch": 0.45, + "learning_rate": 1.1999865575422652e-05, + "loss": 1.0573, + "step": 18830 + }, + { + "epoch": 0.45, + "learning_rate": 1.199910240968721e-05, + "loss": 1.1761, + "step": 18831 + }, + { + "epoch": 0.45, + "learning_rate": 1.1998339231823683e-05, + "loss": 1.2452, + "step": 18832 + }, + { + "epoch": 0.45, + "learning_rate": 1.1997576041836705e-05, + "loss": 1.0726, + "step": 18833 + }, + { + "epoch": 0.45, + "learning_rate": 1.1996812839730901e-05, + "loss": 1.2054, + "step": 18834 + }, + { + "epoch": 0.45, + "learning_rate": 1.1996049625510901e-05, + "loss": 1.2413, + "step": 18835 + }, + { + "epoch": 0.45, + "learning_rate": 1.1995286399181341e-05, + "loss": 1.0726, + "step": 18836 + }, + { + "epoch": 0.45, + "learning_rate": 1.1994523160746848e-05, + "loss": 0.9776, + "step": 18837 + }, + { + "epoch": 0.45, + "learning_rate": 1.199375991021205e-05, + "loss": 0.7607, + "step": 18838 + }, + { + "epoch": 0.45, + "learning_rate": 1.199299664758158e-05, + "loss": 1.2803, + "step": 18839 + }, + { + "epoch": 0.45, + "learning_rate": 1.1992233372860068e-05, + "loss": 1.1988, + "step": 18840 + }, + { + "epoch": 0.45, + "learning_rate": 1.1991470086052145e-05, + "loss": 1.1559, + "step": 18841 + }, + { + "epoch": 0.45, + "learning_rate": 1.1990706787162444e-05, + "loss": 1.1552, + "step": 18842 + }, + { + "epoch": 0.45, + "learning_rate": 1.1989943476195588e-05, + "loss": 1.1124, + "step": 18843 + }, + { + "epoch": 0.45, + "learning_rate": 1.1989180153156217e-05, + "loss": 1.1392, + "step": 18844 + }, + { + "epoch": 0.45, + "learning_rate": 1.1988416818048958e-05, + "loss": 1.4965, + "step": 18845 + }, + { + "epoch": 0.45, + "learning_rate": 1.1987653470878441e-05, + "loss": 1.3134, + "step": 18846 + }, + { + "epoch": 0.45, + "learning_rate": 1.1986890111649296e-05, + "loss": 1.063, + "step": 18847 + }, + { + "epoch": 0.45, + "learning_rate": 1.1986126740366157e-05, + "loss": 1.1299, + "step": 18848 + }, + { + "epoch": 0.45, + "learning_rate": 1.1985363357033656e-05, + "loss": 1.1739, + "step": 18849 + }, + { + "epoch": 0.45, + "learning_rate": 1.198459996165642e-05, + "loss": 1.0216, + "step": 18850 + }, + { + "epoch": 0.45, + "learning_rate": 1.1983836554239086e-05, + "loss": 1.2489, + "step": 18851 + }, + { + "epoch": 0.45, + "learning_rate": 1.1983073134786281e-05, + "loss": 0.9777, + "step": 18852 + }, + { + "epoch": 0.45, + "learning_rate": 1.1982309703302637e-05, + "loss": 1.0244, + "step": 18853 + }, + { + "epoch": 0.45, + "learning_rate": 1.1981546259792785e-05, + "loss": 0.9113, + "step": 18854 + }, + { + "epoch": 0.45, + "learning_rate": 1.198078280426136e-05, + "loss": 1.0991, + "step": 18855 + }, + { + "epoch": 0.45, + "learning_rate": 1.1980019336712992e-05, + "loss": 1.0156, + "step": 18856 + }, + { + "epoch": 0.45, + "learning_rate": 1.197925585715231e-05, + "loss": 1.1068, + "step": 18857 + }, + { + "epoch": 0.45, + "learning_rate": 1.1978492365583949e-05, + "loss": 1.0448, + "step": 18858 + }, + { + "epoch": 0.45, + "learning_rate": 1.197772886201254e-05, + "loss": 1.1918, + "step": 18859 + }, + { + "epoch": 0.45, + "learning_rate": 1.1976965346442716e-05, + "loss": 1.085, + "step": 18860 + }, + { + "epoch": 0.45, + "learning_rate": 1.197620181887911e-05, + "loss": 0.9565, + "step": 18861 + }, + { + "epoch": 0.45, + "learning_rate": 1.1975438279326351e-05, + "loss": 0.8497, + "step": 18862 + }, + { + "epoch": 0.45, + "learning_rate": 1.197467472778907e-05, + "loss": 1.3038, + "step": 18863 + }, + { + "epoch": 0.45, + "learning_rate": 1.1973911164271905e-05, + "loss": 0.9122, + "step": 18864 + }, + { + "epoch": 0.45, + "learning_rate": 1.1973147588779485e-05, + "loss": 0.8342, + "step": 18865 + }, + { + "epoch": 0.45, + "learning_rate": 1.1972384001316442e-05, + "loss": 1.1007, + "step": 18866 + }, + { + "epoch": 0.45, + "learning_rate": 1.197162040188741e-05, + "loss": 1.1197, + "step": 18867 + }, + { + "epoch": 0.45, + "learning_rate": 1.197085679049702e-05, + "loss": 0.9484, + "step": 18868 + }, + { + "epoch": 0.45, + "learning_rate": 1.1970093167149906e-05, + "loss": 0.7263, + "step": 18869 + }, + { + "epoch": 0.45, + "learning_rate": 1.1969329531850697e-05, + "loss": 0.972, + "step": 18870 + }, + { + "epoch": 0.45, + "learning_rate": 1.1968565884604032e-05, + "loss": 1.0081, + "step": 18871 + }, + { + "epoch": 0.45, + "learning_rate": 1.1967802225414542e-05, + "loss": 1.2067, + "step": 18872 + }, + { + "epoch": 0.45, + "learning_rate": 1.1967038554286858e-05, + "loss": 1.2335, + "step": 18873 + }, + { + "epoch": 0.45, + "learning_rate": 1.1966274871225612e-05, + "loss": 1.4026, + "step": 18874 + }, + { + "epoch": 0.45, + "learning_rate": 1.1965511176235438e-05, + "loss": 1.1326, + "step": 18875 + }, + { + "epoch": 0.45, + "learning_rate": 1.1964747469320972e-05, + "loss": 1.2067, + "step": 18876 + }, + { + "epoch": 0.45, + "learning_rate": 1.1963983750486846e-05, + "loss": 1.0768, + "step": 18877 + }, + { + "epoch": 0.45, + "learning_rate": 1.1963220019737691e-05, + "loss": 0.8772, + "step": 18878 + }, + { + "epoch": 0.45, + "learning_rate": 1.196245627707814e-05, + "loss": 1.0797, + "step": 18879 + }, + { + "epoch": 0.45, + "learning_rate": 1.1961692522512834e-05, + "loss": 1.1889, + "step": 18880 + }, + { + "epoch": 0.45, + "learning_rate": 1.1960928756046397e-05, + "loss": 1.1565, + "step": 18881 + }, + { + "epoch": 0.45, + "learning_rate": 1.1960164977683469e-05, + "loss": 1.0777, + "step": 18882 + }, + { + "epoch": 0.45, + "learning_rate": 1.1959401187428676e-05, + "loss": 1.1708, + "step": 18883 + }, + { + "epoch": 0.45, + "learning_rate": 1.1958637385286662e-05, + "loss": 1.2536, + "step": 18884 + }, + { + "epoch": 0.45, + "learning_rate": 1.1957873571262053e-05, + "loss": 1.1621, + "step": 18885 + }, + { + "epoch": 0.45, + "learning_rate": 1.1957109745359487e-05, + "loss": 1.1716, + "step": 18886 + }, + { + "epoch": 0.45, + "learning_rate": 1.1956345907583595e-05, + "loss": 1.2508, + "step": 18887 + }, + { + "epoch": 0.45, + "learning_rate": 1.1955582057939014e-05, + "loss": 1.1921, + "step": 18888 + }, + { + "epoch": 0.45, + "learning_rate": 1.1954818196430378e-05, + "loss": 1.118, + "step": 18889 + }, + { + "epoch": 0.45, + "learning_rate": 1.1954054323062317e-05, + "loss": 1.2588, + "step": 18890 + }, + { + "epoch": 0.45, + "learning_rate": 1.195329043783947e-05, + "loss": 1.146, + "step": 18891 + }, + { + "epoch": 0.45, + "learning_rate": 1.1952526540766466e-05, + "loss": 0.9625, + "step": 18892 + }, + { + "epoch": 0.45, + "learning_rate": 1.1951762631847948e-05, + "loss": 0.9853, + "step": 18893 + }, + { + "epoch": 0.45, + "learning_rate": 1.195099871108854e-05, + "loss": 1.1378, + "step": 18894 + }, + { + "epoch": 0.45, + "learning_rate": 1.1950234778492884e-05, + "loss": 1.2101, + "step": 18895 + }, + { + "epoch": 0.45, + "learning_rate": 1.1949470834065612e-05, + "loss": 1.1088, + "step": 18896 + }, + { + "epoch": 0.45, + "learning_rate": 1.1948706877811361e-05, + "loss": 1.0331, + "step": 18897 + }, + { + "epoch": 0.45, + "learning_rate": 1.194794290973476e-05, + "loss": 1.2545, + "step": 18898 + }, + { + "epoch": 0.45, + "learning_rate": 1.1947178929840451e-05, + "loss": 1.5034, + "step": 18899 + }, + { + "epoch": 0.45, + "learning_rate": 1.1946414938133065e-05, + "loss": 0.988, + "step": 18900 + }, + { + "epoch": 0.45, + "learning_rate": 1.1945650934617236e-05, + "loss": 1.0689, + "step": 18901 + }, + { + "epoch": 0.45, + "learning_rate": 1.19448869192976e-05, + "loss": 1.2823, + "step": 18902 + }, + { + "epoch": 0.45, + "learning_rate": 1.1944122892178794e-05, + "loss": 1.2031, + "step": 18903 + }, + { + "epoch": 0.45, + "learning_rate": 1.1943358853265453e-05, + "loss": 1.1215, + "step": 18904 + }, + { + "epoch": 0.45, + "learning_rate": 1.1942594802562208e-05, + "loss": 0.9711, + "step": 18905 + }, + { + "epoch": 0.45, + "learning_rate": 1.1941830740073696e-05, + "loss": 1.0331, + "step": 18906 + }, + { + "epoch": 0.45, + "learning_rate": 1.1941066665804556e-05, + "loss": 0.9059, + "step": 18907 + }, + { + "epoch": 0.45, + "learning_rate": 1.1940302579759424e-05, + "loss": 1.1151, + "step": 18908 + }, + { + "epoch": 0.45, + "learning_rate": 1.193953848194293e-05, + "loss": 0.8975, + "step": 18909 + }, + { + "epoch": 0.45, + "learning_rate": 1.193877437235971e-05, + "loss": 1.1325, + "step": 18910 + }, + { + "epoch": 0.45, + "learning_rate": 1.1938010251014405e-05, + "loss": 1.1545, + "step": 18911 + }, + { + "epoch": 0.45, + "learning_rate": 1.1937246117911647e-05, + "loss": 0.596, + "step": 18912 + }, + { + "epoch": 0.45, + "learning_rate": 1.193648197305607e-05, + "loss": 0.937, + "step": 18913 + }, + { + "epoch": 0.45, + "learning_rate": 1.1935717816452315e-05, + "loss": 0.931, + "step": 18914 + }, + { + "epoch": 0.45, + "learning_rate": 1.1934953648105015e-05, + "loss": 1.0351, + "step": 18915 + }, + { + "epoch": 0.45, + "learning_rate": 1.1934189468018807e-05, + "loss": 1.0834, + "step": 18916 + }, + { + "epoch": 0.45, + "learning_rate": 1.1933425276198325e-05, + "loss": 1.1741, + "step": 18917 + }, + { + "epoch": 0.45, + "learning_rate": 1.1932661072648207e-05, + "loss": 0.6647, + "step": 18918 + }, + { + "epoch": 0.45, + "learning_rate": 1.193189685737309e-05, + "loss": 1.1689, + "step": 18919 + }, + { + "epoch": 0.46, + "learning_rate": 1.1931132630377607e-05, + "loss": 1.223, + "step": 18920 + }, + { + "epoch": 0.46, + "learning_rate": 1.1930368391666398e-05, + "loss": 1.1755, + "step": 18921 + }, + { + "epoch": 0.46, + "learning_rate": 1.1929604141244097e-05, + "loss": 1.225, + "step": 18922 + }, + { + "epoch": 0.46, + "learning_rate": 1.1928839879115342e-05, + "loss": 1.272, + "step": 18923 + }, + { + "epoch": 0.46, + "learning_rate": 1.192807560528477e-05, + "loss": 0.9671, + "step": 18924 + }, + { + "epoch": 0.46, + "learning_rate": 1.1927311319757018e-05, + "loss": 1.1789, + "step": 18925 + }, + { + "epoch": 0.46, + "learning_rate": 1.1926547022536716e-05, + "loss": 1.1509, + "step": 18926 + }, + { + "epoch": 0.46, + "learning_rate": 1.1925782713628512e-05, + "loss": 1.1162, + "step": 18927 + }, + { + "epoch": 0.46, + "learning_rate": 1.1925018393037036e-05, + "loss": 1.1523, + "step": 18928 + }, + { + "epoch": 0.46, + "learning_rate": 1.1924254060766926e-05, + "loss": 1.0041, + "step": 18929 + }, + { + "epoch": 0.46, + "learning_rate": 1.1923489716822818e-05, + "loss": 1.0423, + "step": 18930 + }, + { + "epoch": 0.46, + "learning_rate": 1.192272536120935e-05, + "loss": 1.0795, + "step": 18931 + }, + { + "epoch": 0.46, + "learning_rate": 1.1921960993931162e-05, + "loss": 0.8356, + "step": 18932 + }, + { + "epoch": 0.46, + "learning_rate": 1.1921196614992885e-05, + "loss": 1.2113, + "step": 18933 + }, + { + "epoch": 0.46, + "learning_rate": 1.1920432224399163e-05, + "loss": 1.0617, + "step": 18934 + }, + { + "epoch": 0.46, + "learning_rate": 1.1919667822154631e-05, + "loss": 1.1877, + "step": 18935 + }, + { + "epoch": 0.46, + "learning_rate": 1.1918903408263924e-05, + "loss": 1.2955, + "step": 18936 + }, + { + "epoch": 0.46, + "learning_rate": 1.1918138982731682e-05, + "loss": 1.0903, + "step": 18937 + }, + { + "epoch": 0.46, + "learning_rate": 1.1917374545562542e-05, + "loss": 1.114, + "step": 18938 + }, + { + "epoch": 0.46, + "learning_rate": 1.1916610096761144e-05, + "loss": 0.9529, + "step": 18939 + }, + { + "epoch": 0.46, + "learning_rate": 1.191584563633212e-05, + "loss": 1.0741, + "step": 18940 + }, + { + "epoch": 0.46, + "learning_rate": 1.191508116428011e-05, + "loss": 0.9582, + "step": 18941 + }, + { + "epoch": 0.46, + "learning_rate": 1.1914316680609756e-05, + "loss": 1.0437, + "step": 18942 + }, + { + "epoch": 0.46, + "learning_rate": 1.1913552185325692e-05, + "loss": 1.2297, + "step": 18943 + }, + { + "epoch": 0.46, + "learning_rate": 1.1912787678432559e-05, + "loss": 0.943, + "step": 18944 + }, + { + "epoch": 0.46, + "learning_rate": 1.1912023159934989e-05, + "loss": 1.3285, + "step": 18945 + }, + { + "epoch": 0.46, + "learning_rate": 1.1911258629837627e-05, + "loss": 1.0212, + "step": 18946 + }, + { + "epoch": 0.46, + "learning_rate": 1.1910494088145109e-05, + "loss": 1.0007, + "step": 18947 + }, + { + "epoch": 0.46, + "learning_rate": 1.1909729534862071e-05, + "loss": 0.8742, + "step": 18948 + }, + { + "epoch": 0.46, + "learning_rate": 1.1908964969993153e-05, + "loss": 1.1593, + "step": 18949 + }, + { + "epoch": 0.46, + "learning_rate": 1.1908200393542995e-05, + "loss": 0.7543, + "step": 18950 + }, + { + "epoch": 0.46, + "learning_rate": 1.1907435805516232e-05, + "loss": 1.2366, + "step": 18951 + }, + { + "epoch": 0.46, + "learning_rate": 1.1906671205917509e-05, + "loss": 1.1513, + "step": 18952 + }, + { + "epoch": 0.46, + "learning_rate": 1.1905906594751455e-05, + "loss": 1.1256, + "step": 18953 + }, + { + "epoch": 0.46, + "learning_rate": 1.1905141972022716e-05, + "loss": 0.9774, + "step": 18954 + }, + { + "epoch": 0.46, + "learning_rate": 1.1904377337735929e-05, + "loss": 1.2388, + "step": 18955 + }, + { + "epoch": 0.46, + "learning_rate": 1.1903612691895734e-05, + "loss": 1.1594, + "step": 18956 + }, + { + "epoch": 0.46, + "learning_rate": 1.1902848034506764e-05, + "loss": 1.0819, + "step": 18957 + }, + { + "epoch": 0.46, + "learning_rate": 1.1902083365573665e-05, + "loss": 1.1039, + "step": 18958 + }, + { + "epoch": 0.46, + "learning_rate": 1.1901318685101075e-05, + "loss": 1.1793, + "step": 18959 + }, + { + "epoch": 0.46, + "learning_rate": 1.190055399309363e-05, + "loss": 0.9416, + "step": 18960 + }, + { + "epoch": 0.46, + "learning_rate": 1.1899789289555974e-05, + "loss": 1.1473, + "step": 18961 + }, + { + "epoch": 0.46, + "learning_rate": 1.1899024574492739e-05, + "loss": 1.1003, + "step": 18962 + }, + { + "epoch": 0.46, + "learning_rate": 1.1898259847908573e-05, + "loss": 1.2614, + "step": 18963 + }, + { + "epoch": 0.46, + "learning_rate": 1.1897495109808108e-05, + "loss": 1.1157, + "step": 18964 + }, + { + "epoch": 0.46, + "learning_rate": 1.189673036019599e-05, + "loss": 0.9541, + "step": 18965 + }, + { + "epoch": 0.46, + "learning_rate": 1.1895965599076849e-05, + "loss": 1.2756, + "step": 18966 + }, + { + "epoch": 0.46, + "learning_rate": 1.1895200826455335e-05, + "loss": 0.6919, + "step": 18967 + }, + { + "epoch": 0.46, + "learning_rate": 1.1894436042336082e-05, + "loss": 0.8875, + "step": 18968 + }, + { + "epoch": 0.46, + "learning_rate": 1.1893671246723733e-05, + "loss": 0.9617, + "step": 18969 + }, + { + "epoch": 0.46, + "learning_rate": 1.1892906439622923e-05, + "loss": 1.0986, + "step": 18970 + }, + { + "epoch": 0.46, + "learning_rate": 1.1892141621038295e-05, + "loss": 1.078, + "step": 18971 + }, + { + "epoch": 0.46, + "learning_rate": 1.1891376790974491e-05, + "loss": 1.2901, + "step": 18972 + }, + { + "epoch": 0.46, + "learning_rate": 1.1890611949436148e-05, + "loss": 1.2203, + "step": 18973 + }, + { + "epoch": 0.46, + "learning_rate": 1.1889847096427907e-05, + "loss": 1.0425, + "step": 18974 + }, + { + "epoch": 0.46, + "learning_rate": 1.188908223195441e-05, + "loss": 1.1294, + "step": 18975 + }, + { + "epoch": 0.46, + "learning_rate": 1.1888317356020293e-05, + "loss": 1.1536, + "step": 18976 + }, + { + "epoch": 0.46, + "learning_rate": 1.1887552468630199e-05, + "loss": 0.9883, + "step": 18977 + }, + { + "epoch": 0.46, + "learning_rate": 1.188678756978877e-05, + "loss": 1.0826, + "step": 18978 + }, + { + "epoch": 0.46, + "learning_rate": 1.1886022659500642e-05, + "loss": 0.9885, + "step": 18979 + }, + { + "epoch": 0.46, + "learning_rate": 1.188525773777046e-05, + "loss": 1.1747, + "step": 18980 + }, + { + "epoch": 0.46, + "learning_rate": 1.188449280460286e-05, + "loss": 1.0461, + "step": 18981 + }, + { + "epoch": 0.46, + "learning_rate": 1.1883727860002487e-05, + "loss": 1.1145, + "step": 18982 + }, + { + "epoch": 0.46, + "learning_rate": 1.1882962903973982e-05, + "loss": 1.0529, + "step": 18983 + }, + { + "epoch": 0.46, + "learning_rate": 1.1882197936521985e-05, + "loss": 1.0973, + "step": 18984 + }, + { + "epoch": 0.46, + "learning_rate": 1.188143295765113e-05, + "loss": 1.0773, + "step": 18985 + }, + { + "epoch": 0.46, + "learning_rate": 1.1880667967366069e-05, + "loss": 1.0602, + "step": 18986 + }, + { + "epoch": 0.46, + "learning_rate": 1.1879902965671436e-05, + "loss": 1.3271, + "step": 18987 + }, + { + "epoch": 0.46, + "learning_rate": 1.1879137952571876e-05, + "loss": 0.9862, + "step": 18988 + }, + { + "epoch": 0.46, + "learning_rate": 1.1878372928072025e-05, + "loss": 1.056, + "step": 18989 + }, + { + "epoch": 0.46, + "learning_rate": 1.1877607892176527e-05, + "loss": 0.9131, + "step": 18990 + }, + { + "epoch": 0.46, + "learning_rate": 1.1876842844890026e-05, + "loss": 1.2757, + "step": 18991 + }, + { + "epoch": 0.46, + "learning_rate": 1.1876077786217159e-05, + "loss": 1.0905, + "step": 18992 + }, + { + "epoch": 0.46, + "learning_rate": 1.187531271616257e-05, + "loss": 1.1519, + "step": 18993 + }, + { + "epoch": 0.46, + "learning_rate": 1.18745476347309e-05, + "loss": 1.2199, + "step": 18994 + }, + { + "epoch": 0.46, + "learning_rate": 1.1873782541926791e-05, + "loss": 1.2187, + "step": 18995 + }, + { + "epoch": 0.46, + "learning_rate": 1.1873017437754884e-05, + "loss": 1.2153, + "step": 18996 + }, + { + "epoch": 0.46, + "learning_rate": 1.1872252322219819e-05, + "loss": 1.1571, + "step": 18997 + }, + { + "epoch": 0.46, + "learning_rate": 1.187148719532624e-05, + "loss": 1.1598, + "step": 18998 + }, + { + "epoch": 0.46, + "learning_rate": 1.1870722057078789e-05, + "loss": 1.1609, + "step": 18999 + }, + { + "epoch": 0.46, + "learning_rate": 1.186995690748211e-05, + "loss": 1.1548, + "step": 19000 + }, + { + "epoch": 0.46, + "learning_rate": 1.1869191746540838e-05, + "loss": 1.1456, + "step": 19001 + }, + { + "epoch": 0.46, + "learning_rate": 1.1868426574259621e-05, + "loss": 1.173, + "step": 19002 + }, + { + "epoch": 0.46, + "learning_rate": 1.18676613906431e-05, + "loss": 0.8576, + "step": 19003 + }, + { + "epoch": 0.46, + "learning_rate": 1.1866896195695916e-05, + "loss": 0.9731, + "step": 19004 + }, + { + "epoch": 0.46, + "learning_rate": 1.1866130989422711e-05, + "loss": 1.2356, + "step": 19005 + }, + { + "epoch": 0.46, + "learning_rate": 1.186536577182813e-05, + "loss": 1.267, + "step": 19006 + }, + { + "epoch": 0.46, + "learning_rate": 1.1864600542916813e-05, + "loss": 1.2373, + "step": 19007 + }, + { + "epoch": 0.46, + "learning_rate": 1.1863835302693404e-05, + "loss": 1.2108, + "step": 19008 + }, + { + "epoch": 0.46, + "learning_rate": 1.1863070051162541e-05, + "loss": 0.9966, + "step": 19009 + }, + { + "epoch": 0.46, + "learning_rate": 1.1862304788328876e-05, + "loss": 0.9286, + "step": 19010 + }, + { + "epoch": 0.46, + "learning_rate": 1.1861539514197043e-05, + "loss": 1.305, + "step": 19011 + }, + { + "epoch": 0.46, + "learning_rate": 1.1860774228771687e-05, + "loss": 1.1114, + "step": 19012 + }, + { + "epoch": 0.46, + "learning_rate": 1.1860008932057449e-05, + "loss": 1.1827, + "step": 19013 + }, + { + "epoch": 0.46, + "learning_rate": 1.1859243624058979e-05, + "loss": 1.046, + "step": 19014 + }, + { + "epoch": 0.46, + "learning_rate": 1.1858478304780915e-05, + "loss": 1.0218, + "step": 19015 + }, + { + "epoch": 0.46, + "learning_rate": 1.1857712974227897e-05, + "loss": 1.1059, + "step": 19016 + }, + { + "epoch": 0.46, + "learning_rate": 1.1856947632404572e-05, + "loss": 1.1043, + "step": 19017 + }, + { + "epoch": 0.46, + "learning_rate": 1.1856182279315585e-05, + "loss": 1.0548, + "step": 19018 + }, + { + "epoch": 0.46, + "learning_rate": 1.1855416914965575e-05, + "loss": 0.9043, + "step": 19019 + }, + { + "epoch": 0.46, + "learning_rate": 1.1854651539359186e-05, + "loss": 1.1721, + "step": 19020 + }, + { + "epoch": 0.46, + "learning_rate": 1.1853886152501064e-05, + "loss": 1.2671, + "step": 19021 + }, + { + "epoch": 0.46, + "learning_rate": 1.185312075439585e-05, + "loss": 1.0878, + "step": 19022 + }, + { + "epoch": 0.46, + "learning_rate": 1.185235534504819e-05, + "loss": 1.2471, + "step": 19023 + }, + { + "epoch": 0.46, + "learning_rate": 1.1851589924462723e-05, + "loss": 0.948, + "step": 19024 + }, + { + "epoch": 0.46, + "learning_rate": 1.1850824492644097e-05, + "loss": 1.2039, + "step": 19025 + }, + { + "epoch": 0.46, + "learning_rate": 1.1850059049596954e-05, + "loss": 1.185, + "step": 19026 + }, + { + "epoch": 0.46, + "learning_rate": 1.1849293595325939e-05, + "loss": 1.1984, + "step": 19027 + }, + { + "epoch": 0.46, + "learning_rate": 1.1848528129835692e-05, + "loss": 1.1316, + "step": 19028 + }, + { + "epoch": 0.46, + "learning_rate": 1.1847762653130861e-05, + "loss": 1.0843, + "step": 19029 + }, + { + "epoch": 0.46, + "learning_rate": 1.1846997165216089e-05, + "loss": 1.1312, + "step": 19030 + }, + { + "epoch": 0.46, + "learning_rate": 1.184623166609602e-05, + "loss": 0.9657, + "step": 19031 + }, + { + "epoch": 0.46, + "learning_rate": 1.1845466155775296e-05, + "loss": 0.9221, + "step": 19032 + }, + { + "epoch": 0.46, + "learning_rate": 1.1844700634258566e-05, + "loss": 0.9297, + "step": 19033 + }, + { + "epoch": 0.46, + "learning_rate": 1.184393510155047e-05, + "loss": 0.8632, + "step": 19034 + }, + { + "epoch": 0.46, + "learning_rate": 1.1843169557655651e-05, + "loss": 1.115, + "step": 19035 + }, + { + "epoch": 0.46, + "learning_rate": 1.1842404002578758e-05, + "loss": 1.1236, + "step": 19036 + }, + { + "epoch": 0.46, + "learning_rate": 1.1841638436324433e-05, + "loss": 0.8085, + "step": 19037 + }, + { + "epoch": 0.46, + "learning_rate": 1.1840872858897321e-05, + "loss": 1.0864, + "step": 19038 + }, + { + "epoch": 0.46, + "learning_rate": 1.1840107270302065e-05, + "loss": 1.2761, + "step": 19039 + }, + { + "epoch": 0.46, + "learning_rate": 1.1839341670543312e-05, + "loss": 1.0869, + "step": 19040 + }, + { + "epoch": 0.46, + "learning_rate": 1.1838576059625703e-05, + "loss": 1.4328, + "step": 19041 + }, + { + "epoch": 0.46, + "learning_rate": 1.183781043755389e-05, + "loss": 1.0996, + "step": 19042 + }, + { + "epoch": 0.46, + "learning_rate": 1.1837044804332511e-05, + "loss": 0.7531, + "step": 19043 + }, + { + "epoch": 0.46, + "learning_rate": 1.1836279159966216e-05, + "loss": 1.0834, + "step": 19044 + }, + { + "epoch": 0.46, + "learning_rate": 1.1835513504459644e-05, + "loss": 0.9087, + "step": 19045 + }, + { + "epoch": 0.46, + "learning_rate": 1.1834747837817445e-05, + "loss": 0.9082, + "step": 19046 + }, + { + "epoch": 0.46, + "learning_rate": 1.1833982160044261e-05, + "loss": 1.2139, + "step": 19047 + }, + { + "epoch": 0.46, + "learning_rate": 1.1833216471144739e-05, + "loss": 1.1203, + "step": 19048 + }, + { + "epoch": 0.46, + "learning_rate": 1.1832450771123525e-05, + "loss": 0.9944, + "step": 19049 + }, + { + "epoch": 0.46, + "learning_rate": 1.1831685059985263e-05, + "loss": 1.1525, + "step": 19050 + }, + { + "epoch": 0.46, + "learning_rate": 1.1830919337734598e-05, + "loss": 0.9682, + "step": 19051 + }, + { + "epoch": 0.46, + "learning_rate": 1.1830153604376177e-05, + "loss": 1.2054, + "step": 19052 + }, + { + "epoch": 0.46, + "learning_rate": 1.1829387859914643e-05, + "loss": 1.005, + "step": 19053 + }, + { + "epoch": 0.46, + "learning_rate": 1.1828622104354643e-05, + "loss": 1.0231, + "step": 19054 + }, + { + "epoch": 0.46, + "learning_rate": 1.1827856337700825e-05, + "loss": 1.1893, + "step": 19055 + }, + { + "epoch": 0.46, + "learning_rate": 1.1827090559957832e-05, + "loss": 1.0454, + "step": 19056 + }, + { + "epoch": 0.46, + "learning_rate": 1.182632477113031e-05, + "loss": 1.1718, + "step": 19057 + }, + { + "epoch": 0.46, + "learning_rate": 1.1825558971222903e-05, + "loss": 1.2036, + "step": 19058 + }, + { + "epoch": 0.46, + "learning_rate": 1.182479316024026e-05, + "loss": 0.9262, + "step": 19059 + }, + { + "epoch": 0.46, + "learning_rate": 1.1824027338187028e-05, + "loss": 0.9884, + "step": 19060 + }, + { + "epoch": 0.46, + "learning_rate": 1.182326150506785e-05, + "loss": 1.1625, + "step": 19061 + }, + { + "epoch": 0.46, + "learning_rate": 1.1822495660887374e-05, + "loss": 1.1975, + "step": 19062 + }, + { + "epoch": 0.46, + "learning_rate": 1.1821729805650245e-05, + "loss": 0.9013, + "step": 19063 + }, + { + "epoch": 0.46, + "learning_rate": 1.1820963939361107e-05, + "loss": 0.9052, + "step": 19064 + }, + { + "epoch": 0.46, + "learning_rate": 1.1820198062024613e-05, + "loss": 1.2701, + "step": 19065 + }, + { + "epoch": 0.46, + "learning_rate": 1.1819432173645405e-05, + "loss": 1.184, + "step": 19066 + }, + { + "epoch": 0.46, + "learning_rate": 1.1818666274228129e-05, + "loss": 1.096, + "step": 19067 + }, + { + "epoch": 0.46, + "learning_rate": 1.1817900363777433e-05, + "loss": 0.9659, + "step": 19068 + }, + { + "epoch": 0.46, + "learning_rate": 1.1817134442297961e-05, + "loss": 1.1892, + "step": 19069 + }, + { + "epoch": 0.46, + "learning_rate": 1.1816368509794365e-05, + "loss": 0.9905, + "step": 19070 + }, + { + "epoch": 0.46, + "learning_rate": 1.181560256627129e-05, + "loss": 1.124, + "step": 19071 + }, + { + "epoch": 0.46, + "learning_rate": 1.1814836611733375e-05, + "loss": 1.1487, + "step": 19072 + }, + { + "epoch": 0.46, + "learning_rate": 1.1814070646185276e-05, + "loss": 1.0768, + "step": 19073 + }, + { + "epoch": 0.46, + "learning_rate": 1.181330466963164e-05, + "loss": 1.1578, + "step": 19074 + }, + { + "epoch": 0.46, + "learning_rate": 1.1812538682077107e-05, + "loss": 0.9538, + "step": 19075 + }, + { + "epoch": 0.46, + "learning_rate": 1.181177268352633e-05, + "loss": 1.1848, + "step": 19076 + }, + { + "epoch": 0.46, + "learning_rate": 1.1811006673983956e-05, + "loss": 0.8963, + "step": 19077 + }, + { + "epoch": 0.46, + "learning_rate": 1.1810240653454628e-05, + "loss": 1.0532, + "step": 19078 + }, + { + "epoch": 0.46, + "learning_rate": 1.1809474621942996e-05, + "loss": 1.1909, + "step": 19079 + }, + { + "epoch": 0.46, + "learning_rate": 1.1808708579453707e-05, + "loss": 1.2128, + "step": 19080 + }, + { + "epoch": 0.46, + "learning_rate": 1.180794252599141e-05, + "loss": 1.1351, + "step": 19081 + }, + { + "epoch": 0.46, + "learning_rate": 1.180717646156075e-05, + "loss": 1.023, + "step": 19082 + }, + { + "epoch": 0.46, + "learning_rate": 1.1806410386166376e-05, + "loss": 1.0229, + "step": 19083 + }, + { + "epoch": 0.46, + "learning_rate": 1.1805644299812932e-05, + "loss": 1.5092, + "step": 19084 + }, + { + "epoch": 0.46, + "learning_rate": 1.1804878202505072e-05, + "loss": 1.3297, + "step": 19085 + }, + { + "epoch": 0.46, + "learning_rate": 1.1804112094247439e-05, + "loss": 0.9003, + "step": 19086 + }, + { + "epoch": 0.46, + "learning_rate": 1.1803345975044685e-05, + "loss": 0.9969, + "step": 19087 + }, + { + "epoch": 0.46, + "learning_rate": 1.1802579844901452e-05, + "loss": 1.1974, + "step": 19088 + }, + { + "epoch": 0.46, + "learning_rate": 1.1801813703822393e-05, + "loss": 1.1735, + "step": 19089 + }, + { + "epoch": 0.46, + "learning_rate": 1.1801047551812153e-05, + "loss": 1.0804, + "step": 19090 + }, + { + "epoch": 0.46, + "learning_rate": 1.1800281388875382e-05, + "loss": 1.2287, + "step": 19091 + }, + { + "epoch": 0.46, + "learning_rate": 1.1799515215016726e-05, + "loss": 0.9538, + "step": 19092 + }, + { + "epoch": 0.46, + "learning_rate": 1.1798749030240835e-05, + "loss": 1.0231, + "step": 19093 + }, + { + "epoch": 0.46, + "learning_rate": 1.1797982834552358e-05, + "loss": 0.9788, + "step": 19094 + }, + { + "epoch": 0.46, + "learning_rate": 1.179721662795594e-05, + "loss": 1.0697, + "step": 19095 + }, + { + "epoch": 0.46, + "learning_rate": 1.179645041045623e-05, + "loss": 0.7386, + "step": 19096 + }, + { + "epoch": 0.46, + "learning_rate": 1.1795684182057883e-05, + "loss": 1.0623, + "step": 19097 + }, + { + "epoch": 0.46, + "learning_rate": 1.179491794276554e-05, + "loss": 0.9089, + "step": 19098 + }, + { + "epoch": 0.46, + "learning_rate": 1.1794151692583852e-05, + "loss": 1.0484, + "step": 19099 + }, + { + "epoch": 0.46, + "learning_rate": 1.1793385431517466e-05, + "loss": 1.0273, + "step": 19100 + }, + { + "epoch": 0.46, + "learning_rate": 1.1792619159571034e-05, + "loss": 1.2456, + "step": 19101 + }, + { + "epoch": 0.46, + "learning_rate": 1.1791852876749204e-05, + "loss": 1.1252, + "step": 19102 + }, + { + "epoch": 0.46, + "learning_rate": 1.1791086583056623e-05, + "loss": 1.0644, + "step": 19103 + }, + { + "epoch": 0.46, + "learning_rate": 1.1790320278497939e-05, + "loss": 0.8905, + "step": 19104 + }, + { + "epoch": 0.46, + "learning_rate": 1.1789553963077807e-05, + "loss": 1.1777, + "step": 19105 + }, + { + "epoch": 0.46, + "learning_rate": 1.178878763680087e-05, + "loss": 1.2038, + "step": 19106 + }, + { + "epoch": 0.46, + "learning_rate": 1.1788021299671777e-05, + "loss": 0.9601, + "step": 19107 + }, + { + "epoch": 0.46, + "learning_rate": 1.1787254951695181e-05, + "loss": 0.8773, + "step": 19108 + }, + { + "epoch": 0.46, + "learning_rate": 1.1786488592875733e-05, + "loss": 0.8973, + "step": 19109 + }, + { + "epoch": 0.46, + "learning_rate": 1.1785722223218077e-05, + "loss": 0.9507, + "step": 19110 + }, + { + "epoch": 0.46, + "learning_rate": 1.1784955842726862e-05, + "loss": 1.1121, + "step": 19111 + }, + { + "epoch": 0.46, + "learning_rate": 1.178418945140674e-05, + "loss": 1.0149, + "step": 19112 + }, + { + "epoch": 0.46, + "learning_rate": 1.1783423049262363e-05, + "loss": 1.0571, + "step": 19113 + }, + { + "epoch": 0.46, + "learning_rate": 1.1782656636298378e-05, + "loss": 1.0151, + "step": 19114 + }, + { + "epoch": 0.46, + "learning_rate": 1.1781890212519429e-05, + "loss": 1.1443, + "step": 19115 + }, + { + "epoch": 0.46, + "learning_rate": 1.1781123777930178e-05, + "loss": 1.1061, + "step": 19116 + }, + { + "epoch": 0.46, + "learning_rate": 1.1780357332535264e-05, + "loss": 1.2273, + "step": 19117 + }, + { + "epoch": 0.46, + "learning_rate": 1.1779590876339341e-05, + "loss": 1.0523, + "step": 19118 + }, + { + "epoch": 0.46, + "learning_rate": 1.1778824409347058e-05, + "loss": 0.7929, + "step": 19119 + }, + { + "epoch": 0.46, + "learning_rate": 1.177805793156307e-05, + "loss": 1.1373, + "step": 19120 + }, + { + "epoch": 0.46, + "learning_rate": 1.1777291442992018e-05, + "loss": 1.1289, + "step": 19121 + }, + { + "epoch": 0.46, + "learning_rate": 1.1776524943638561e-05, + "loss": 1.1613, + "step": 19122 + }, + { + "epoch": 0.46, + "learning_rate": 1.177575843350734e-05, + "loss": 1.3331, + "step": 19123 + }, + { + "epoch": 0.46, + "learning_rate": 1.1774991912603013e-05, + "loss": 1.1476, + "step": 19124 + }, + { + "epoch": 0.46, + "learning_rate": 1.1774225380930229e-05, + "loss": 1.0268, + "step": 19125 + }, + { + "epoch": 0.46, + "learning_rate": 1.1773458838493634e-05, + "loss": 1.1323, + "step": 19126 + }, + { + "epoch": 0.46, + "learning_rate": 1.1772692285297884e-05, + "loss": 1.1795, + "step": 19127 + }, + { + "epoch": 0.46, + "learning_rate": 1.1771925721347625e-05, + "loss": 1.2437, + "step": 19128 + }, + { + "epoch": 0.46, + "learning_rate": 1.177115914664751e-05, + "loss": 1.0546, + "step": 19129 + }, + { + "epoch": 0.46, + "learning_rate": 1.177039256120219e-05, + "loss": 1.2022, + "step": 19130 + }, + { + "epoch": 0.46, + "learning_rate": 1.1769625965016316e-05, + "loss": 1.0733, + "step": 19131 + }, + { + "epoch": 0.46, + "learning_rate": 1.1768859358094533e-05, + "loss": 1.0871, + "step": 19132 + }, + { + "epoch": 0.46, + "learning_rate": 1.1768092740441501e-05, + "loss": 1.0261, + "step": 19133 + }, + { + "epoch": 0.46, + "learning_rate": 1.1767326112061864e-05, + "loss": 1.2381, + "step": 19134 + }, + { + "epoch": 0.46, + "learning_rate": 1.1766559472960275e-05, + "loss": 1.0618, + "step": 19135 + }, + { + "epoch": 0.46, + "learning_rate": 1.1765792823141385e-05, + "loss": 0.9764, + "step": 19136 + }, + { + "epoch": 0.46, + "learning_rate": 1.1765026162609847e-05, + "loss": 0.9064, + "step": 19137 + }, + { + "epoch": 0.46, + "learning_rate": 1.1764259491370309e-05, + "loss": 0.9973, + "step": 19138 + }, + { + "epoch": 0.46, + "learning_rate": 1.1763492809427425e-05, + "loss": 1.1673, + "step": 19139 + }, + { + "epoch": 0.46, + "learning_rate": 1.1762726116785843e-05, + "loss": 1.3049, + "step": 19140 + }, + { + "epoch": 0.46, + "learning_rate": 1.1761959413450217e-05, + "loss": 1.0343, + "step": 19141 + }, + { + "epoch": 0.46, + "learning_rate": 1.17611926994252e-05, + "loss": 0.9507, + "step": 19142 + }, + { + "epoch": 0.46, + "learning_rate": 1.1760425974715438e-05, + "loss": 0.7755, + "step": 19143 + }, + { + "epoch": 0.46, + "learning_rate": 1.1759659239325585e-05, + "loss": 1.125, + "step": 19144 + }, + { + "epoch": 0.46, + "learning_rate": 1.1758892493260294e-05, + "loss": 1.2563, + "step": 19145 + }, + { + "epoch": 0.46, + "learning_rate": 1.1758125736524218e-05, + "loss": 1.2901, + "step": 19146 + }, + { + "epoch": 0.46, + "learning_rate": 1.1757358969122007e-05, + "loss": 1.021, + "step": 19147 + }, + { + "epoch": 0.46, + "learning_rate": 1.1756592191058308e-05, + "loss": 1.3085, + "step": 19148 + }, + { + "epoch": 0.46, + "learning_rate": 1.1755825402337782e-05, + "loss": 0.8549, + "step": 19149 + }, + { + "epoch": 0.46, + "learning_rate": 1.1755058602965074e-05, + "loss": 1.3544, + "step": 19150 + }, + { + "epoch": 0.46, + "learning_rate": 1.1754291792944839e-05, + "loss": 1.1511, + "step": 19151 + }, + { + "epoch": 0.46, + "learning_rate": 1.175352497228173e-05, + "loss": 1.2067, + "step": 19152 + }, + { + "epoch": 0.46, + "learning_rate": 1.1752758140980396e-05, + "loss": 1.082, + "step": 19153 + }, + { + "epoch": 0.46, + "learning_rate": 1.1751991299045491e-05, + "loss": 1.1175, + "step": 19154 + }, + { + "epoch": 0.46, + "learning_rate": 1.1751224446481664e-05, + "loss": 1.0889, + "step": 19155 + }, + { + "epoch": 0.46, + "learning_rate": 1.1750457583293574e-05, + "loss": 0.9601, + "step": 19156 + }, + { + "epoch": 0.46, + "learning_rate": 1.1749690709485869e-05, + "loss": 1.1877, + "step": 19157 + }, + { + "epoch": 0.46, + "learning_rate": 1.1748923825063201e-05, + "loss": 0.8972, + "step": 19158 + }, + { + "epoch": 0.46, + "learning_rate": 1.1748156930030221e-05, + "loss": 1.1047, + "step": 19159 + }, + { + "epoch": 0.46, + "learning_rate": 1.1747390024391587e-05, + "loss": 0.9848, + "step": 19160 + }, + { + "epoch": 0.46, + "learning_rate": 1.1746623108151948e-05, + "loss": 1.0017, + "step": 19161 + }, + { + "epoch": 0.46, + "learning_rate": 1.174585618131596e-05, + "loss": 1.2385, + "step": 19162 + }, + { + "epoch": 0.46, + "learning_rate": 1.174508924388827e-05, + "loss": 1.1816, + "step": 19163 + }, + { + "epoch": 0.46, + "learning_rate": 1.1744322295873535e-05, + "loss": 1.0713, + "step": 19164 + }, + { + "epoch": 0.46, + "learning_rate": 1.1743555337276406e-05, + "loss": 1.161, + "step": 19165 + }, + { + "epoch": 0.46, + "learning_rate": 1.1742788368101541e-05, + "loss": 0.9652, + "step": 19166 + }, + { + "epoch": 0.46, + "learning_rate": 1.1742021388353584e-05, + "loss": 1.3273, + "step": 19167 + }, + { + "epoch": 0.46, + "learning_rate": 1.1741254398037194e-05, + "loss": 1.0918, + "step": 19168 + }, + { + "epoch": 0.46, + "learning_rate": 1.1740487397157025e-05, + "loss": 1.1106, + "step": 19169 + }, + { + "epoch": 0.46, + "learning_rate": 1.1739720385717728e-05, + "loss": 0.8729, + "step": 19170 + }, + { + "epoch": 0.46, + "learning_rate": 1.1738953363723956e-05, + "loss": 1.1142, + "step": 19171 + }, + { + "epoch": 0.46, + "learning_rate": 1.173818633118036e-05, + "loss": 1.2105, + "step": 19172 + }, + { + "epoch": 0.46, + "learning_rate": 1.17374192880916e-05, + "loss": 1.2048, + "step": 19173 + }, + { + "epoch": 0.46, + "learning_rate": 1.1736652234462327e-05, + "loss": 1.2612, + "step": 19174 + }, + { + "epoch": 0.46, + "learning_rate": 1.173588517029719e-05, + "loss": 1.1759, + "step": 19175 + }, + { + "epoch": 0.46, + "learning_rate": 1.1735118095600848e-05, + "loss": 1.3114, + "step": 19176 + }, + { + "epoch": 0.46, + "learning_rate": 1.1734351010377951e-05, + "loss": 1.2863, + "step": 19177 + }, + { + "epoch": 0.46, + "learning_rate": 1.1733583914633157e-05, + "loss": 1.0542, + "step": 19178 + }, + { + "epoch": 0.46, + "learning_rate": 1.1732816808371115e-05, + "loss": 1.1224, + "step": 19179 + }, + { + "epoch": 0.46, + "learning_rate": 1.173204969159648e-05, + "loss": 1.0356, + "step": 19180 + }, + { + "epoch": 0.46, + "learning_rate": 1.173128256431391e-05, + "loss": 1.2707, + "step": 19181 + }, + { + "epoch": 0.46, + "learning_rate": 1.1730515426528055e-05, + "loss": 1.3191, + "step": 19182 + }, + { + "epoch": 0.46, + "learning_rate": 1.1729748278243569e-05, + "loss": 1.2059, + "step": 19183 + }, + { + "epoch": 0.46, + "learning_rate": 1.1728981119465109e-05, + "loss": 1.1902, + "step": 19184 + }, + { + "epoch": 0.46, + "learning_rate": 1.1728213950197325e-05, + "loss": 1.206, + "step": 19185 + }, + { + "epoch": 0.46, + "learning_rate": 1.1727446770444875e-05, + "loss": 1.0858, + "step": 19186 + }, + { + "epoch": 0.46, + "learning_rate": 1.1726679580212407e-05, + "loss": 0.9772, + "step": 19187 + }, + { + "epoch": 0.46, + "learning_rate": 1.1725912379504588e-05, + "loss": 1.0353, + "step": 19188 + }, + { + "epoch": 0.46, + "learning_rate": 1.172514516832606e-05, + "loss": 0.8521, + "step": 19189 + }, + { + "epoch": 0.46, + "learning_rate": 1.1724377946681482e-05, + "loss": 1.0822, + "step": 19190 + }, + { + "epoch": 0.46, + "learning_rate": 1.1723610714575508e-05, + "loss": 1.1765, + "step": 19191 + }, + { + "epoch": 0.46, + "learning_rate": 1.1722843472012794e-05, + "loss": 1.2078, + "step": 19192 + }, + { + "epoch": 0.46, + "learning_rate": 1.1722076218997995e-05, + "loss": 1.2802, + "step": 19193 + }, + { + "epoch": 0.46, + "learning_rate": 1.1721308955535765e-05, + "loss": 1.1252, + "step": 19194 + }, + { + "epoch": 0.46, + "learning_rate": 1.1720541681630756e-05, + "loss": 1.3287, + "step": 19195 + }, + { + "epoch": 0.46, + "learning_rate": 1.1719774397287627e-05, + "loss": 1.1962, + "step": 19196 + }, + { + "epoch": 0.46, + "learning_rate": 1.1719007102511032e-05, + "loss": 1.0213, + "step": 19197 + }, + { + "epoch": 0.46, + "learning_rate": 1.1718239797305622e-05, + "loss": 1.0446, + "step": 19198 + }, + { + "epoch": 0.46, + "learning_rate": 1.1717472481676057e-05, + "loss": 1.085, + "step": 19199 + }, + { + "epoch": 0.46, + "learning_rate": 1.1716705155626993e-05, + "loss": 1.1144, + "step": 19200 + }, + { + "epoch": 0.46, + "learning_rate": 1.1715937819163082e-05, + "loss": 1.2058, + "step": 19201 + }, + { + "epoch": 0.46, + "learning_rate": 1.1715170472288975e-05, + "loss": 1.2084, + "step": 19202 + }, + { + "epoch": 0.46, + "learning_rate": 1.1714403115009336e-05, + "loss": 0.9952, + "step": 19203 + }, + { + "epoch": 0.46, + "learning_rate": 1.1713635747328817e-05, + "loss": 0.97, + "step": 19204 + }, + { + "epoch": 0.46, + "learning_rate": 1.1712868369252074e-05, + "loss": 1.0415, + "step": 19205 + }, + { + "epoch": 0.46, + "learning_rate": 1.1712100980783759e-05, + "loss": 1.0955, + "step": 19206 + }, + { + "epoch": 0.46, + "learning_rate": 1.1711333581928527e-05, + "loss": 1.1005, + "step": 19207 + }, + { + "epoch": 0.46, + "learning_rate": 1.1710566172691042e-05, + "loss": 1.0087, + "step": 19208 + }, + { + "epoch": 0.46, + "learning_rate": 1.1709798753075954e-05, + "loss": 1.1904, + "step": 19209 + }, + { + "epoch": 0.46, + "learning_rate": 1.1709031323087918e-05, + "loss": 0.9807, + "step": 19210 + }, + { + "epoch": 0.46, + "learning_rate": 1.170826388273159e-05, + "loss": 1.1775, + "step": 19211 + }, + { + "epoch": 0.46, + "learning_rate": 1.1707496432011626e-05, + "loss": 0.9963, + "step": 19212 + }, + { + "epoch": 0.46, + "learning_rate": 1.1706728970932686e-05, + "loss": 1.116, + "step": 19213 + }, + { + "epoch": 0.46, + "learning_rate": 1.1705961499499421e-05, + "loss": 0.8165, + "step": 19214 + }, + { + "epoch": 0.46, + "learning_rate": 1.1705194017716488e-05, + "loss": 1.0352, + "step": 19215 + }, + { + "epoch": 0.46, + "learning_rate": 1.1704426525588545e-05, + "loss": 1.3148, + "step": 19216 + }, + { + "epoch": 0.46, + "learning_rate": 1.1703659023120248e-05, + "loss": 1.0561, + "step": 19217 + }, + { + "epoch": 0.46, + "learning_rate": 1.1702891510316253e-05, + "loss": 1.0933, + "step": 19218 + }, + { + "epoch": 0.46, + "learning_rate": 1.1702123987181213e-05, + "loss": 1.0586, + "step": 19219 + }, + { + "epoch": 0.46, + "learning_rate": 1.1701356453719787e-05, + "loss": 1.1468, + "step": 19220 + }, + { + "epoch": 0.46, + "learning_rate": 1.1700588909936634e-05, + "loss": 0.9656, + "step": 19221 + }, + { + "epoch": 0.46, + "learning_rate": 1.169982135583641e-05, + "loss": 1.1106, + "step": 19222 + }, + { + "epoch": 0.46, + "learning_rate": 1.1699053791423765e-05, + "loss": 1.0239, + "step": 19223 + }, + { + "epoch": 0.46, + "learning_rate": 1.1698286216703362e-05, + "loss": 1.2664, + "step": 19224 + }, + { + "epoch": 0.46, + "learning_rate": 1.1697518631679855e-05, + "loss": 1.2967, + "step": 19225 + }, + { + "epoch": 0.46, + "learning_rate": 1.1696751036357906e-05, + "loss": 1.1042, + "step": 19226 + }, + { + "epoch": 0.46, + "learning_rate": 1.1695983430742162e-05, + "loss": 1.0413, + "step": 19227 + }, + { + "epoch": 0.46, + "learning_rate": 1.1695215814837288e-05, + "loss": 1.4597, + "step": 19228 + }, + { + "epoch": 0.46, + "learning_rate": 1.169444818864794e-05, + "loss": 0.9944, + "step": 19229 + }, + { + "epoch": 0.46, + "learning_rate": 1.169368055217877e-05, + "loss": 1.0406, + "step": 19230 + }, + { + "epoch": 0.46, + "learning_rate": 1.169291290543444e-05, + "loss": 1.3125, + "step": 19231 + }, + { + "epoch": 0.46, + "learning_rate": 1.1692145248419606e-05, + "loss": 1.0381, + "step": 19232 + }, + { + "epoch": 0.46, + "learning_rate": 1.1691377581138926e-05, + "loss": 1.1136, + "step": 19233 + }, + { + "epoch": 0.46, + "learning_rate": 1.1690609903597053e-05, + "loss": 1.1932, + "step": 19234 + }, + { + "epoch": 0.46, + "learning_rate": 1.1689842215798648e-05, + "loss": 0.9275, + "step": 19235 + }, + { + "epoch": 0.46, + "learning_rate": 1.168907451774837e-05, + "loss": 1.1888, + "step": 19236 + }, + { + "epoch": 0.46, + "learning_rate": 1.1688306809450874e-05, + "loss": 1.025, + "step": 19237 + }, + { + "epoch": 0.46, + "learning_rate": 1.1687539090910816e-05, + "loss": 1.1217, + "step": 19238 + }, + { + "epoch": 0.46, + "learning_rate": 1.1686771362132857e-05, + "loss": 1.0524, + "step": 19239 + }, + { + "epoch": 0.46, + "learning_rate": 1.168600362312165e-05, + "loss": 1.0601, + "step": 19240 + }, + { + "epoch": 0.46, + "learning_rate": 1.1685235873881859e-05, + "loss": 1.177, + "step": 19241 + }, + { + "epoch": 0.46, + "learning_rate": 1.1684468114418134e-05, + "loss": 1.0391, + "step": 19242 + }, + { + "epoch": 0.46, + "learning_rate": 1.168370034473514e-05, + "loss": 1.1637, + "step": 19243 + }, + { + "epoch": 0.46, + "learning_rate": 1.1682932564837533e-05, + "loss": 0.9155, + "step": 19244 + }, + { + "epoch": 0.46, + "learning_rate": 1.1682164774729969e-05, + "loss": 1.1385, + "step": 19245 + }, + { + "epoch": 0.46, + "learning_rate": 1.1681396974417108e-05, + "loss": 1.2775, + "step": 19246 + }, + { + "epoch": 0.46, + "learning_rate": 1.1680629163903605e-05, + "loss": 0.989, + "step": 19247 + }, + { + "epoch": 0.46, + "learning_rate": 1.1679861343194121e-05, + "loss": 0.7923, + "step": 19248 + }, + { + "epoch": 0.46, + "learning_rate": 1.1679093512293313e-05, + "loss": 1.1792, + "step": 19249 + }, + { + "epoch": 0.46, + "learning_rate": 1.1678325671205841e-05, + "loss": 1.0312, + "step": 19250 + }, + { + "epoch": 0.46, + "learning_rate": 1.167755781993636e-05, + "loss": 0.9385, + "step": 19251 + }, + { + "epoch": 0.46, + "learning_rate": 1.1676789958489532e-05, + "loss": 1.0543, + "step": 19252 + }, + { + "epoch": 0.46, + "learning_rate": 1.1676022086870013e-05, + "loss": 0.7429, + "step": 19253 + }, + { + "epoch": 0.46, + "learning_rate": 1.1675254205082463e-05, + "loss": 0.9755, + "step": 19254 + }, + { + "epoch": 0.46, + "learning_rate": 1.167448631313154e-05, + "loss": 1.2918, + "step": 19255 + }, + { + "epoch": 0.46, + "learning_rate": 1.16737184110219e-05, + "loss": 1.1348, + "step": 19256 + }, + { + "epoch": 0.46, + "learning_rate": 1.1672950498758207e-05, + "loss": 1.1557, + "step": 19257 + }, + { + "epoch": 0.46, + "learning_rate": 1.1672182576345113e-05, + "loss": 1.0425, + "step": 19258 + }, + { + "epoch": 0.46, + "learning_rate": 1.1671414643787285e-05, + "loss": 1.4763, + "step": 19259 + }, + { + "epoch": 0.46, + "learning_rate": 1.1670646701089374e-05, + "loss": 0.9731, + "step": 19260 + }, + { + "epoch": 0.46, + "learning_rate": 1.1669878748256046e-05, + "loss": 1.1082, + "step": 19261 + }, + { + "epoch": 0.46, + "learning_rate": 1.166911078529195e-05, + "loss": 1.0009, + "step": 19262 + }, + { + "epoch": 0.46, + "learning_rate": 1.1668342812201758e-05, + "loss": 1.1431, + "step": 19263 + }, + { + "epoch": 0.46, + "learning_rate": 1.1667574828990121e-05, + "loss": 1.0447, + "step": 19264 + }, + { + "epoch": 0.46, + "learning_rate": 1.1666806835661699e-05, + "loss": 1.0141, + "step": 19265 + }, + { + "epoch": 0.46, + "learning_rate": 1.1666038832221152e-05, + "loss": 1.0246, + "step": 19266 + }, + { + "epoch": 0.46, + "learning_rate": 1.1665270818673139e-05, + "loss": 1.2669, + "step": 19267 + }, + { + "epoch": 0.46, + "learning_rate": 1.1664502795022321e-05, + "loss": 1.1711, + "step": 19268 + }, + { + "epoch": 0.46, + "learning_rate": 1.1663734761273356e-05, + "loss": 1.3068, + "step": 19269 + }, + { + "epoch": 0.46, + "learning_rate": 1.16629667174309e-05, + "loss": 1.14, + "step": 19270 + }, + { + "epoch": 0.46, + "learning_rate": 1.166219866349962e-05, + "loss": 1.0507, + "step": 19271 + }, + { + "epoch": 0.46, + "learning_rate": 1.1661430599484169e-05, + "loss": 1.1534, + "step": 19272 + }, + { + "epoch": 0.46, + "learning_rate": 1.1660662525389212e-05, + "loss": 0.8059, + "step": 19273 + }, + { + "epoch": 0.46, + "learning_rate": 1.1659894441219403e-05, + "loss": 0.9106, + "step": 19274 + }, + { + "epoch": 0.46, + "learning_rate": 1.1659126346979408e-05, + "loss": 1.1332, + "step": 19275 + }, + { + "epoch": 0.46, + "learning_rate": 1.1658358242673882e-05, + "loss": 1.1955, + "step": 19276 + }, + { + "epoch": 0.46, + "learning_rate": 1.1657590128307486e-05, + "loss": 0.9149, + "step": 19277 + }, + { + "epoch": 0.46, + "learning_rate": 1.1656822003884883e-05, + "loss": 1.3013, + "step": 19278 + }, + { + "epoch": 0.46, + "learning_rate": 1.165605386941073e-05, + "loss": 1.0302, + "step": 19279 + }, + { + "epoch": 0.46, + "learning_rate": 1.1655285724889687e-05, + "loss": 1.1907, + "step": 19280 + }, + { + "epoch": 0.46, + "learning_rate": 1.1654517570326411e-05, + "loss": 1.2239, + "step": 19281 + }, + { + "epoch": 0.46, + "learning_rate": 1.1653749405725572e-05, + "loss": 1.0376, + "step": 19282 + }, + { + "epoch": 0.46, + "learning_rate": 1.1652981231091823e-05, + "loss": 0.8745, + "step": 19283 + }, + { + "epoch": 0.46, + "learning_rate": 1.1652213046429824e-05, + "loss": 1.098, + "step": 19284 + }, + { + "epoch": 0.46, + "learning_rate": 1.1651444851744236e-05, + "loss": 1.1147, + "step": 19285 + }, + { + "epoch": 0.46, + "learning_rate": 1.1650676647039723e-05, + "loss": 1.2348, + "step": 19286 + }, + { + "epoch": 0.46, + "learning_rate": 1.1649908432320944e-05, + "loss": 1.0989, + "step": 19287 + }, + { + "epoch": 0.46, + "learning_rate": 1.1649140207592557e-05, + "loss": 1.1049, + "step": 19288 + }, + { + "epoch": 0.46, + "learning_rate": 1.1648371972859223e-05, + "loss": 0.9815, + "step": 19289 + }, + { + "epoch": 0.46, + "learning_rate": 1.1647603728125603e-05, + "loss": 1.1531, + "step": 19290 + }, + { + "epoch": 0.46, + "learning_rate": 1.1646835473396361e-05, + "loss": 1.3077, + "step": 19291 + }, + { + "epoch": 0.46, + "learning_rate": 1.1646067208676157e-05, + "loss": 1.0501, + "step": 19292 + }, + { + "epoch": 0.46, + "learning_rate": 1.1645298933969649e-05, + "loss": 1.0426, + "step": 19293 + }, + { + "epoch": 0.46, + "learning_rate": 1.1644530649281498e-05, + "loss": 0.9722, + "step": 19294 + }, + { + "epoch": 0.46, + "learning_rate": 1.1643762354616369e-05, + "loss": 1.1499, + "step": 19295 + }, + { + "epoch": 0.46, + "learning_rate": 1.164299404997892e-05, + "loss": 1.2492, + "step": 19296 + }, + { + "epoch": 0.46, + "learning_rate": 1.1642225735373812e-05, + "loss": 1.1323, + "step": 19297 + }, + { + "epoch": 0.46, + "learning_rate": 1.1641457410805703e-05, + "loss": 1.1294, + "step": 19298 + }, + { + "epoch": 0.46, + "learning_rate": 1.1640689076279264e-05, + "loss": 1.1348, + "step": 19299 + }, + { + "epoch": 0.46, + "learning_rate": 1.1639920731799146e-05, + "loss": 1.2054, + "step": 19300 + }, + { + "epoch": 0.46, + "learning_rate": 1.1639152377370018e-05, + "loss": 1.1687, + "step": 19301 + }, + { + "epoch": 0.46, + "learning_rate": 1.1638384012996535e-05, + "loss": 0.9784, + "step": 19302 + }, + { + "epoch": 0.46, + "learning_rate": 1.1637615638683364e-05, + "loss": 1.1327, + "step": 19303 + }, + { + "epoch": 0.46, + "learning_rate": 1.1636847254435162e-05, + "loss": 0.8129, + "step": 19304 + }, + { + "epoch": 0.46, + "learning_rate": 1.1636078860256594e-05, + "loss": 1.1878, + "step": 19305 + }, + { + "epoch": 0.46, + "learning_rate": 1.163531045615232e-05, + "loss": 0.9997, + "step": 19306 + }, + { + "epoch": 0.46, + "learning_rate": 1.1634542042127002e-05, + "loss": 1.1272, + "step": 19307 + }, + { + "epoch": 0.46, + "learning_rate": 1.1633773618185302e-05, + "loss": 0.9346, + "step": 19308 + }, + { + "epoch": 0.46, + "learning_rate": 1.1633005184331885e-05, + "loss": 1.0086, + "step": 19309 + }, + { + "epoch": 0.46, + "learning_rate": 1.1632236740571403e-05, + "loss": 1.1585, + "step": 19310 + }, + { + "epoch": 0.46, + "learning_rate": 1.1631468286908529e-05, + "loss": 1.1824, + "step": 19311 + }, + { + "epoch": 0.46, + "learning_rate": 1.1630699823347922e-05, + "loss": 1.1202, + "step": 19312 + }, + { + "epoch": 0.46, + "learning_rate": 1.1629931349894238e-05, + "loss": 1.0634, + "step": 19313 + }, + { + "epoch": 0.46, + "learning_rate": 1.1629162866552145e-05, + "loss": 0.8031, + "step": 19314 + }, + { + "epoch": 0.46, + "learning_rate": 1.1628394373326308e-05, + "loss": 1.0141, + "step": 19315 + }, + { + "epoch": 0.46, + "learning_rate": 1.1627625870221384e-05, + "loss": 1.0656, + "step": 19316 + }, + { + "epoch": 0.46, + "learning_rate": 1.1626857357242034e-05, + "loss": 0.9557, + "step": 19317 + }, + { + "epoch": 0.46, + "learning_rate": 1.1626088834392926e-05, + "loss": 1.0348, + "step": 19318 + }, + { + "epoch": 0.46, + "learning_rate": 1.1625320301678717e-05, + "loss": 1.0824, + "step": 19319 + }, + { + "epoch": 0.46, + "learning_rate": 1.1624551759104075e-05, + "loss": 1.2504, + "step": 19320 + }, + { + "epoch": 0.46, + "learning_rate": 1.1623783206673656e-05, + "loss": 1.143, + "step": 19321 + }, + { + "epoch": 0.46, + "learning_rate": 1.1623014644392128e-05, + "loss": 1.2081, + "step": 19322 + }, + { + "epoch": 0.46, + "learning_rate": 1.1622246072264151e-05, + "loss": 1.0097, + "step": 19323 + }, + { + "epoch": 0.46, + "learning_rate": 1.162147749029439e-05, + "loss": 1.2316, + "step": 19324 + }, + { + "epoch": 0.46, + "learning_rate": 1.1620708898487508e-05, + "loss": 1.0127, + "step": 19325 + }, + { + "epoch": 0.46, + "learning_rate": 1.1619940296848163e-05, + "loss": 0.9215, + "step": 19326 + }, + { + "epoch": 0.46, + "learning_rate": 1.1619171685381024e-05, + "loss": 0.6445, + "step": 19327 + }, + { + "epoch": 0.46, + "learning_rate": 1.1618403064090748e-05, + "loss": 1.009, + "step": 19328 + }, + { + "epoch": 0.46, + "learning_rate": 1.1617634432982004e-05, + "loss": 0.8986, + "step": 19329 + }, + { + "epoch": 0.46, + "learning_rate": 1.161686579205945e-05, + "loss": 0.7943, + "step": 19330 + }, + { + "epoch": 0.46, + "learning_rate": 1.1616097141327756e-05, + "loss": 1.0916, + "step": 19331 + }, + { + "epoch": 0.46, + "learning_rate": 1.1615328480791576e-05, + "loss": 1.1327, + "step": 19332 + }, + { + "epoch": 0.46, + "learning_rate": 1.161455981045558e-05, + "loss": 1.0669, + "step": 19333 + }, + { + "epoch": 0.46, + "learning_rate": 1.161379113032443e-05, + "loss": 0.9607, + "step": 19334 + }, + { + "epoch": 0.46, + "learning_rate": 1.1613022440402788e-05, + "loss": 1.0771, + "step": 19335 + }, + { + "epoch": 0.47, + "learning_rate": 1.1612253740695318e-05, + "loss": 1.0462, + "step": 19336 + }, + { + "epoch": 0.47, + "learning_rate": 1.1611485031206684e-05, + "loss": 1.1045, + "step": 19337 + }, + { + "epoch": 0.47, + "learning_rate": 1.1610716311941551e-05, + "loss": 1.1475, + "step": 19338 + }, + { + "epoch": 0.47, + "learning_rate": 1.160994758290458e-05, + "loss": 1.2538, + "step": 19339 + }, + { + "epoch": 0.47, + "learning_rate": 1.1609178844100435e-05, + "loss": 0.744, + "step": 19340 + }, + { + "epoch": 0.47, + "learning_rate": 1.160841009553378e-05, + "loss": 1.2863, + "step": 19341 + }, + { + "epoch": 0.47, + "learning_rate": 1.1607641337209279e-05, + "loss": 0.9187, + "step": 19342 + }, + { + "epoch": 0.47, + "learning_rate": 1.1606872569131597e-05, + "loss": 1.0804, + "step": 19343 + }, + { + "epoch": 0.47, + "learning_rate": 1.1606103791305398e-05, + "loss": 1.0931, + "step": 19344 + }, + { + "epoch": 0.47, + "learning_rate": 1.1605335003735344e-05, + "loss": 1.194, + "step": 19345 + }, + { + "epoch": 0.47, + "learning_rate": 1.16045662064261e-05, + "loss": 0.9975, + "step": 19346 + }, + { + "epoch": 0.47, + "learning_rate": 1.1603797399382332e-05, + "loss": 1.19, + "step": 19347 + }, + { + "epoch": 0.47, + "learning_rate": 1.1603028582608703e-05, + "loss": 1.1274, + "step": 19348 + }, + { + "epoch": 0.47, + "learning_rate": 1.1602259756109871e-05, + "loss": 1.0079, + "step": 19349 + }, + { + "epoch": 0.47, + "learning_rate": 1.160149091989051e-05, + "loss": 1.0874, + "step": 19350 + }, + { + "epoch": 0.47, + "learning_rate": 1.1600722073955281e-05, + "loss": 1.1287, + "step": 19351 + }, + { + "epoch": 0.47, + "learning_rate": 1.1599953218308848e-05, + "loss": 0.9421, + "step": 19352 + }, + { + "epoch": 0.47, + "learning_rate": 1.159918435295587e-05, + "loss": 0.9051, + "step": 19353 + }, + { + "epoch": 0.47, + "learning_rate": 1.1598415477901022e-05, + "loss": 0.9357, + "step": 19354 + }, + { + "epoch": 0.47, + "learning_rate": 1.1597646593148963e-05, + "loss": 0.8496, + "step": 19355 + }, + { + "epoch": 0.47, + "learning_rate": 1.1596877698704357e-05, + "loss": 0.8199, + "step": 19356 + }, + { + "epoch": 0.47, + "learning_rate": 1.1596108794571868e-05, + "loss": 1.0172, + "step": 19357 + }, + { + "epoch": 0.47, + "learning_rate": 1.1595339880756162e-05, + "loss": 1.1409, + "step": 19358 + }, + { + "epoch": 0.47, + "learning_rate": 1.1594570957261909e-05, + "loss": 1.0206, + "step": 19359 + }, + { + "epoch": 0.47, + "learning_rate": 1.1593802024093765e-05, + "loss": 0.9233, + "step": 19360 + }, + { + "epoch": 0.47, + "learning_rate": 1.1593033081256397e-05, + "loss": 1.2662, + "step": 19361 + }, + { + "epoch": 0.47, + "learning_rate": 1.1592264128754475e-05, + "loss": 0.6166, + "step": 19362 + }, + { + "epoch": 0.47, + "learning_rate": 1.1591495166592663e-05, + "loss": 1.1578, + "step": 19363 + }, + { + "epoch": 0.47, + "learning_rate": 1.1590726194775621e-05, + "loss": 0.8687, + "step": 19364 + }, + { + "epoch": 0.47, + "learning_rate": 1.1589957213308019e-05, + "loss": 0.8587, + "step": 19365 + }, + { + "epoch": 0.47, + "learning_rate": 1.1589188222194521e-05, + "loss": 1.1179, + "step": 19366 + }, + { + "epoch": 0.47, + "learning_rate": 1.158841922143979e-05, + "loss": 1.136, + "step": 19367 + }, + { + "epoch": 0.47, + "learning_rate": 1.1587650211048493e-05, + "loss": 1.0364, + "step": 19368 + }, + { + "epoch": 0.47, + "learning_rate": 1.1586881191025297e-05, + "loss": 1.0338, + "step": 19369 + }, + { + "epoch": 0.47, + "learning_rate": 1.1586112161374867e-05, + "loss": 0.9544, + "step": 19370 + }, + { + "epoch": 0.47, + "learning_rate": 1.1585343122101866e-05, + "loss": 1.3695, + "step": 19371 + }, + { + "epoch": 0.47, + "learning_rate": 1.1584574073210961e-05, + "loss": 1.0638, + "step": 19372 + }, + { + "epoch": 0.47, + "learning_rate": 1.1583805014706817e-05, + "loss": 1.2911, + "step": 19373 + }, + { + "epoch": 0.47, + "learning_rate": 1.1583035946594101e-05, + "loss": 1.1711, + "step": 19374 + }, + { + "epoch": 0.47, + "learning_rate": 1.1582266868877481e-05, + "loss": 1.0934, + "step": 19375 + }, + { + "epoch": 0.47, + "learning_rate": 1.1581497781561616e-05, + "loss": 1.2038, + "step": 19376 + }, + { + "epoch": 0.47, + "learning_rate": 1.1580728684651179e-05, + "loss": 1.0548, + "step": 19377 + }, + { + "epoch": 0.47, + "learning_rate": 1.1579959578150833e-05, + "loss": 1.3319, + "step": 19378 + }, + { + "epoch": 0.47, + "learning_rate": 1.1579190462065241e-05, + "loss": 0.7937, + "step": 19379 + }, + { + "epoch": 0.47, + "learning_rate": 1.1578421336399076e-05, + "loss": 1.1776, + "step": 19380 + }, + { + "epoch": 0.47, + "learning_rate": 1.1577652201156994e-05, + "loss": 1.0266, + "step": 19381 + }, + { + "epoch": 0.47, + "learning_rate": 1.1576883056343672e-05, + "loss": 1.0765, + "step": 19382 + }, + { + "epoch": 0.47, + "learning_rate": 1.1576113901963772e-05, + "loss": 1.2791, + "step": 19383 + }, + { + "epoch": 0.47, + "learning_rate": 1.1575344738021958e-05, + "loss": 1.0296, + "step": 19384 + }, + { + "epoch": 0.47, + "learning_rate": 1.1574575564522896e-05, + "loss": 0.7451, + "step": 19385 + }, + { + "epoch": 0.47, + "learning_rate": 1.1573806381471258e-05, + "loss": 1.1162, + "step": 19386 + }, + { + "epoch": 0.47, + "learning_rate": 1.1573037188871705e-05, + "loss": 1.0331, + "step": 19387 + }, + { + "epoch": 0.47, + "learning_rate": 1.1572267986728906e-05, + "loss": 0.9898, + "step": 19388 + }, + { + "epoch": 0.47, + "learning_rate": 1.1571498775047527e-05, + "loss": 0.9956, + "step": 19389 + }, + { + "epoch": 0.47, + "learning_rate": 1.1570729553832234e-05, + "loss": 1.2517, + "step": 19390 + }, + { + "epoch": 0.47, + "learning_rate": 1.1569960323087694e-05, + "loss": 0.9997, + "step": 19391 + }, + { + "epoch": 0.47, + "learning_rate": 1.1569191082818577e-05, + "loss": 1.2281, + "step": 19392 + }, + { + "epoch": 0.47, + "learning_rate": 1.1568421833029541e-05, + "loss": 0.9647, + "step": 19393 + }, + { + "epoch": 0.47, + "learning_rate": 1.1567652573725263e-05, + "loss": 1.2303, + "step": 19394 + }, + { + "epoch": 0.47, + "learning_rate": 1.1566883304910405e-05, + "loss": 1.2646, + "step": 19395 + }, + { + "epoch": 0.47, + "learning_rate": 1.1566114026589634e-05, + "loss": 1.0745, + "step": 19396 + }, + { + "epoch": 0.47, + "learning_rate": 1.1565344738767616e-05, + "loss": 1.0111, + "step": 19397 + }, + { + "epoch": 0.47, + "learning_rate": 1.1564575441449024e-05, + "loss": 1.1821, + "step": 19398 + }, + { + "epoch": 0.47, + "learning_rate": 1.1563806134638519e-05, + "loss": 0.9492, + "step": 19399 + }, + { + "epoch": 0.47, + "learning_rate": 1.156303681834077e-05, + "loss": 0.9542, + "step": 19400 + }, + { + "epoch": 0.47, + "learning_rate": 1.1562267492560439e-05, + "loss": 1.0142, + "step": 19401 + }, + { + "epoch": 0.47, + "learning_rate": 1.1561498157302205e-05, + "loss": 0.8977, + "step": 19402 + }, + { + "epoch": 0.47, + "learning_rate": 1.1560728812570727e-05, + "loss": 1.3055, + "step": 19403 + }, + { + "epoch": 0.47, + "learning_rate": 1.1559959458370674e-05, + "loss": 1.4488, + "step": 19404 + }, + { + "epoch": 0.47, + "learning_rate": 1.1559190094706714e-05, + "loss": 0.9077, + "step": 19405 + }, + { + "epoch": 0.47, + "learning_rate": 1.1558420721583514e-05, + "loss": 1.0032, + "step": 19406 + }, + { + "epoch": 0.47, + "learning_rate": 1.1557651339005744e-05, + "loss": 1.0341, + "step": 19407 + }, + { + "epoch": 0.47, + "learning_rate": 1.1556881946978067e-05, + "loss": 1.1788, + "step": 19408 + }, + { + "epoch": 0.47, + "learning_rate": 1.1556112545505156e-05, + "loss": 1.0921, + "step": 19409 + }, + { + "epoch": 0.47, + "learning_rate": 1.1555343134591674e-05, + "loss": 1.2525, + "step": 19410 + }, + { + "epoch": 0.47, + "learning_rate": 1.1554573714242295e-05, + "loss": 1.1591, + "step": 19411 + }, + { + "epoch": 0.47, + "learning_rate": 1.1553804284461679e-05, + "loss": 1.154, + "step": 19412 + }, + { + "epoch": 0.47, + "learning_rate": 1.1553034845254499e-05, + "loss": 0.8845, + "step": 19413 + }, + { + "epoch": 0.47, + "learning_rate": 1.1552265396625421e-05, + "loss": 1.0295, + "step": 19414 + }, + { + "epoch": 0.47, + "learning_rate": 1.1551495938579115e-05, + "loss": 1.2671, + "step": 19415 + }, + { + "epoch": 0.47, + "learning_rate": 1.1550726471120246e-05, + "loss": 0.9352, + "step": 19416 + }, + { + "epoch": 0.47, + "learning_rate": 1.1549956994253488e-05, + "loss": 1.3076, + "step": 19417 + }, + { + "epoch": 0.47, + "learning_rate": 1.1549187507983504e-05, + "loss": 1.1078, + "step": 19418 + }, + { + "epoch": 0.47, + "learning_rate": 1.1548418012314963e-05, + "loss": 1.0307, + "step": 19419 + }, + { + "epoch": 0.47, + "learning_rate": 1.1547648507252534e-05, + "loss": 0.9452, + "step": 19420 + }, + { + "epoch": 0.47, + "learning_rate": 1.1546878992800884e-05, + "loss": 0.8138, + "step": 19421 + }, + { + "epoch": 0.47, + "learning_rate": 1.1546109468964688e-05, + "loss": 1.1466, + "step": 19422 + }, + { + "epoch": 0.47, + "learning_rate": 1.1545339935748608e-05, + "loss": 0.956, + "step": 19423 + }, + { + "epoch": 0.47, + "learning_rate": 1.1544570393157311e-05, + "loss": 0.8739, + "step": 19424 + }, + { + "epoch": 0.47, + "learning_rate": 1.154380084119547e-05, + "loss": 1.092, + "step": 19425 + }, + { + "epoch": 0.47, + "learning_rate": 1.1543031279867754e-05, + "loss": 1.1036, + "step": 19426 + }, + { + "epoch": 0.47, + "learning_rate": 1.1542261709178828e-05, + "loss": 1.0622, + "step": 19427 + }, + { + "epoch": 0.47, + "learning_rate": 1.1541492129133362e-05, + "loss": 1.167, + "step": 19428 + }, + { + "epoch": 0.47, + "learning_rate": 1.154072253973603e-05, + "loss": 1.1397, + "step": 19429 + }, + { + "epoch": 0.47, + "learning_rate": 1.1539952940991495e-05, + "loss": 1.12, + "step": 19430 + }, + { + "epoch": 0.47, + "learning_rate": 1.1539183332904426e-05, + "loss": 1.0625, + "step": 19431 + }, + { + "epoch": 0.47, + "learning_rate": 1.1538413715479495e-05, + "loss": 1.1588, + "step": 19432 + }, + { + "epoch": 0.47, + "learning_rate": 1.153764408872137e-05, + "loss": 1.0405, + "step": 19433 + }, + { + "epoch": 0.47, + "learning_rate": 1.153687445263472e-05, + "loss": 1.0956, + "step": 19434 + }, + { + "epoch": 0.47, + "learning_rate": 1.1536104807224215e-05, + "loss": 1.2226, + "step": 19435 + }, + { + "epoch": 0.47, + "learning_rate": 1.153533515249452e-05, + "loss": 1.0441, + "step": 19436 + }, + { + "epoch": 0.47, + "learning_rate": 1.1534565488450311e-05, + "loss": 1.1424, + "step": 19437 + }, + { + "epoch": 0.47, + "learning_rate": 1.1533795815096254e-05, + "loss": 1.0916, + "step": 19438 + }, + { + "epoch": 0.47, + "learning_rate": 1.1533026132437017e-05, + "loss": 1.2316, + "step": 19439 + }, + { + "epoch": 0.47, + "learning_rate": 1.1532256440477273e-05, + "loss": 1.1159, + "step": 19440 + }, + { + "epoch": 0.47, + "learning_rate": 1.1531486739221689e-05, + "loss": 1.1516, + "step": 19441 + }, + { + "epoch": 0.47, + "learning_rate": 1.1530717028674934e-05, + "loss": 0.9576, + "step": 19442 + }, + { + "epoch": 0.47, + "learning_rate": 1.152994730884168e-05, + "loss": 1.1469, + "step": 19443 + }, + { + "epoch": 0.47, + "learning_rate": 1.1529177579726593e-05, + "loss": 1.1994, + "step": 19444 + }, + { + "epoch": 0.47, + "learning_rate": 1.1528407841334349e-05, + "loss": 1.0899, + "step": 19445 + }, + { + "epoch": 0.47, + "learning_rate": 1.1527638093669613e-05, + "loss": 0.9737, + "step": 19446 + }, + { + "epoch": 0.47, + "learning_rate": 1.1526868336737056e-05, + "loss": 0.8845, + "step": 19447 + }, + { + "epoch": 0.47, + "learning_rate": 1.1526098570541347e-05, + "loss": 0.9562, + "step": 19448 + }, + { + "epoch": 0.47, + "learning_rate": 1.1525328795087155e-05, + "loss": 1.1497, + "step": 19449 + }, + { + "epoch": 0.47, + "learning_rate": 1.1524559010379156e-05, + "loss": 0.9117, + "step": 19450 + }, + { + "epoch": 0.47, + "learning_rate": 1.1523789216422015e-05, + "loss": 0.7663, + "step": 19451 + }, + { + "epoch": 0.47, + "learning_rate": 1.1523019413220402e-05, + "loss": 1.0378, + "step": 19452 + }, + { + "epoch": 0.47, + "learning_rate": 1.1522249600778989e-05, + "loss": 0.9473, + "step": 19453 + }, + { + "epoch": 0.47, + "learning_rate": 1.1521479779102447e-05, + "loss": 1.0969, + "step": 19454 + }, + { + "epoch": 0.47, + "learning_rate": 1.1520709948195443e-05, + "loss": 1.0368, + "step": 19455 + }, + { + "epoch": 0.47, + "learning_rate": 1.151994010806265e-05, + "loss": 1.0414, + "step": 19456 + }, + { + "epoch": 0.47, + "learning_rate": 1.1519170258708735e-05, + "loss": 1.0605, + "step": 19457 + }, + { + "epoch": 0.47, + "learning_rate": 1.1518400400138376e-05, + "loss": 0.9896, + "step": 19458 + }, + { + "epoch": 0.47, + "learning_rate": 1.1517630532356238e-05, + "loss": 1.199, + "step": 19459 + }, + { + "epoch": 0.47, + "learning_rate": 1.1516860655366993e-05, + "loss": 1.2386, + "step": 19460 + }, + { + "epoch": 0.47, + "learning_rate": 1.151609076917531e-05, + "loss": 1.2261, + "step": 19461 + }, + { + "epoch": 0.47, + "learning_rate": 1.1515320873785862e-05, + "loss": 1.0885, + "step": 19462 + }, + { + "epoch": 0.47, + "learning_rate": 1.1514550969203318e-05, + "loss": 1.0416, + "step": 19463 + }, + { + "epoch": 0.47, + "learning_rate": 1.151378105543235e-05, + "loss": 0.9922, + "step": 19464 + }, + { + "epoch": 0.47, + "learning_rate": 1.1513011132477626e-05, + "loss": 1.1243, + "step": 19465 + }, + { + "epoch": 0.47, + "learning_rate": 1.1512241200343821e-05, + "loss": 1.1926, + "step": 19466 + }, + { + "epoch": 0.47, + "learning_rate": 1.1511471259035607e-05, + "loss": 1.0111, + "step": 19467 + }, + { + "epoch": 0.47, + "learning_rate": 1.1510701308557647e-05, + "loss": 1.2003, + "step": 19468 + }, + { + "epoch": 0.47, + "learning_rate": 1.1509931348914622e-05, + "loss": 1.1659, + "step": 19469 + }, + { + "epoch": 0.47, + "learning_rate": 1.1509161380111198e-05, + "loss": 1.158, + "step": 19470 + }, + { + "epoch": 0.47, + "learning_rate": 1.1508391402152046e-05, + "loss": 1.1522, + "step": 19471 + }, + { + "epoch": 0.47, + "learning_rate": 1.1507621415041837e-05, + "loss": 1.2301, + "step": 19472 + }, + { + "epoch": 0.47, + "learning_rate": 1.1506851418785243e-05, + "loss": 1.0651, + "step": 19473 + }, + { + "epoch": 0.47, + "learning_rate": 1.1506081413386938e-05, + "loss": 1.1732, + "step": 19474 + }, + { + "epoch": 0.47, + "learning_rate": 1.1505311398851592e-05, + "loss": 1.0198, + "step": 19475 + }, + { + "epoch": 0.47, + "learning_rate": 1.1504541375183873e-05, + "loss": 1.0956, + "step": 19476 + }, + { + "epoch": 0.47, + "learning_rate": 1.1503771342388456e-05, + "loss": 1.1746, + "step": 19477 + }, + { + "epoch": 0.47, + "learning_rate": 1.1503001300470016e-05, + "loss": 1.3878, + "step": 19478 + }, + { + "epoch": 0.47, + "learning_rate": 1.1502231249433218e-05, + "loss": 1.0196, + "step": 19479 + }, + { + "epoch": 0.47, + "learning_rate": 1.1501461189282734e-05, + "loss": 1.1247, + "step": 19480 + }, + { + "epoch": 0.47, + "learning_rate": 1.1500691120023239e-05, + "loss": 1.1593, + "step": 19481 + }, + { + "epoch": 0.47, + "learning_rate": 1.1499921041659407e-05, + "loss": 0.8946, + "step": 19482 + }, + { + "epoch": 0.47, + "learning_rate": 1.1499150954195905e-05, + "loss": 1.0935, + "step": 19483 + }, + { + "epoch": 0.47, + "learning_rate": 1.1498380857637405e-05, + "loss": 0.9382, + "step": 19484 + }, + { + "epoch": 0.47, + "learning_rate": 1.1497610751988584e-05, + "loss": 1.3957, + "step": 19485 + }, + { + "epoch": 0.47, + "learning_rate": 1.149684063725411e-05, + "loss": 1.0787, + "step": 19486 + }, + { + "epoch": 0.47, + "learning_rate": 1.1496070513438655e-05, + "loss": 1.0874, + "step": 19487 + }, + { + "epoch": 0.47, + "learning_rate": 1.149530038054689e-05, + "loss": 1.1062, + "step": 19488 + }, + { + "epoch": 0.47, + "learning_rate": 1.1494530238583493e-05, + "loss": 1.1481, + "step": 19489 + }, + { + "epoch": 0.47, + "learning_rate": 1.1493760087553133e-05, + "loss": 0.9611, + "step": 19490 + }, + { + "epoch": 0.47, + "learning_rate": 1.1492989927460475e-05, + "loss": 1.1469, + "step": 19491 + }, + { + "epoch": 0.47, + "learning_rate": 1.1492219758310205e-05, + "loss": 0.907, + "step": 19492 + }, + { + "epoch": 0.47, + "learning_rate": 1.1491449580106988e-05, + "loss": 1.0142, + "step": 19493 + }, + { + "epoch": 0.47, + "learning_rate": 1.1490679392855498e-05, + "loss": 1.0886, + "step": 19494 + }, + { + "epoch": 0.47, + "learning_rate": 1.1489909196560402e-05, + "loss": 1.1637, + "step": 19495 + }, + { + "epoch": 0.47, + "learning_rate": 1.148913899122638e-05, + "loss": 1.238, + "step": 19496 + }, + { + "epoch": 0.47, + "learning_rate": 1.1488368776858105e-05, + "loss": 1.1617, + "step": 19497 + }, + { + "epoch": 0.47, + "learning_rate": 1.1487598553460245e-05, + "loss": 1.2117, + "step": 19498 + }, + { + "epoch": 0.47, + "learning_rate": 1.148682832103747e-05, + "loss": 1.1974, + "step": 19499 + }, + { + "epoch": 0.47, + "learning_rate": 1.1486058079594461e-05, + "loss": 0.9606, + "step": 19500 + }, + { + "epoch": 0.47, + "learning_rate": 1.148528782913589e-05, + "loss": 1.1118, + "step": 19501 + }, + { + "epoch": 0.47, + "learning_rate": 1.1484517569666425e-05, + "loss": 1.1218, + "step": 19502 + }, + { + "epoch": 0.47, + "learning_rate": 1.1483747301190737e-05, + "loss": 1.2141, + "step": 19503 + }, + { + "epoch": 0.47, + "learning_rate": 1.1482977023713506e-05, + "loss": 1.0494, + "step": 19504 + }, + { + "epoch": 0.47, + "learning_rate": 1.1482206737239405e-05, + "loss": 1.1931, + "step": 19505 + }, + { + "epoch": 0.47, + "learning_rate": 1.1481436441773102e-05, + "loss": 1.1135, + "step": 19506 + }, + { + "epoch": 0.47, + "learning_rate": 1.1480666137319269e-05, + "loss": 0.9961, + "step": 19507 + }, + { + "epoch": 0.47, + "learning_rate": 1.1479895823882588e-05, + "loss": 1.2807, + "step": 19508 + }, + { + "epoch": 0.47, + "learning_rate": 1.1479125501467727e-05, + "loss": 1.2644, + "step": 19509 + }, + { + "epoch": 0.47, + "learning_rate": 1.1478355170079358e-05, + "loss": 1.0255, + "step": 19510 + }, + { + "epoch": 0.47, + "learning_rate": 1.1477584829722155e-05, + "loss": 1.1292, + "step": 19511 + }, + { + "epoch": 0.47, + "learning_rate": 1.1476814480400792e-05, + "loss": 1.2481, + "step": 19512 + }, + { + "epoch": 0.47, + "learning_rate": 1.1476044122119945e-05, + "loss": 1.076, + "step": 19513 + }, + { + "epoch": 0.47, + "learning_rate": 1.1475273754884287e-05, + "loss": 1.1219, + "step": 19514 + }, + { + "epoch": 0.47, + "learning_rate": 1.1474503378698485e-05, + "loss": 1.0214, + "step": 19515 + }, + { + "epoch": 0.47, + "learning_rate": 1.1473732993567222e-05, + "loss": 1.2939, + "step": 19516 + }, + { + "epoch": 0.47, + "learning_rate": 1.1472962599495166e-05, + "loss": 1.062, + "step": 19517 + }, + { + "epoch": 0.47, + "learning_rate": 1.1472192196486993e-05, + "loss": 1.171, + "step": 19518 + }, + { + "epoch": 0.47, + "learning_rate": 1.1471421784547375e-05, + "loss": 1.4549, + "step": 19519 + }, + { + "epoch": 0.47, + "learning_rate": 1.1470651363680987e-05, + "loss": 1.1271, + "step": 19520 + }, + { + "epoch": 0.47, + "learning_rate": 1.1469880933892505e-05, + "loss": 1.2251, + "step": 19521 + }, + { + "epoch": 0.47, + "learning_rate": 1.1469110495186602e-05, + "loss": 1.0338, + "step": 19522 + }, + { + "epoch": 0.47, + "learning_rate": 1.1468340047567947e-05, + "loss": 1.1146, + "step": 19523 + }, + { + "epoch": 0.47, + "learning_rate": 1.146756959104122e-05, + "loss": 0.8284, + "step": 19524 + }, + { + "epoch": 0.47, + "learning_rate": 1.1466799125611094e-05, + "loss": 1.0491, + "step": 19525 + }, + { + "epoch": 0.47, + "learning_rate": 1.1466028651282244e-05, + "loss": 0.8839, + "step": 19526 + }, + { + "epoch": 0.47, + "learning_rate": 1.1465258168059341e-05, + "loss": 1.104, + "step": 19527 + }, + { + "epoch": 0.47, + "learning_rate": 1.1464487675947062e-05, + "loss": 1.1401, + "step": 19528 + }, + { + "epoch": 0.47, + "learning_rate": 1.1463717174950081e-05, + "loss": 1.0007, + "step": 19529 + }, + { + "epoch": 0.47, + "learning_rate": 1.1462946665073073e-05, + "loss": 1.2349, + "step": 19530 + }, + { + "epoch": 0.47, + "learning_rate": 1.146217614632071e-05, + "loss": 0.8512, + "step": 19531 + }, + { + "epoch": 0.47, + "learning_rate": 1.146140561869767e-05, + "loss": 0.9869, + "step": 19532 + }, + { + "epoch": 0.47, + "learning_rate": 1.1460635082208624e-05, + "loss": 0.9149, + "step": 19533 + }, + { + "epoch": 0.47, + "learning_rate": 1.1459864536858247e-05, + "loss": 1.0596, + "step": 19534 + }, + { + "epoch": 0.47, + "learning_rate": 1.1459093982651219e-05, + "loss": 1.249, + "step": 19535 + }, + { + "epoch": 0.47, + "learning_rate": 1.145832341959221e-05, + "loss": 1.0224, + "step": 19536 + }, + { + "epoch": 0.47, + "learning_rate": 1.1457552847685895e-05, + "loss": 1.0126, + "step": 19537 + }, + { + "epoch": 0.47, + "learning_rate": 1.145678226693695e-05, + "loss": 1.1388, + "step": 19538 + }, + { + "epoch": 0.47, + "learning_rate": 1.1456011677350052e-05, + "loss": 0.9822, + "step": 19539 + }, + { + "epoch": 0.47, + "learning_rate": 1.1455241078929868e-05, + "loss": 1.0382, + "step": 19540 + }, + { + "epoch": 0.47, + "learning_rate": 1.1454470471681083e-05, + "loss": 0.7413, + "step": 19541 + }, + { + "epoch": 0.47, + "learning_rate": 1.1453699855608367e-05, + "loss": 1.1551, + "step": 19542 + }, + { + "epoch": 0.47, + "learning_rate": 1.1452929230716394e-05, + "loss": 1.31, + "step": 19543 + }, + { + "epoch": 0.47, + "learning_rate": 1.1452158597009843e-05, + "loss": 1.1275, + "step": 19544 + }, + { + "epoch": 0.47, + "learning_rate": 1.145138795449339e-05, + "loss": 1.3455, + "step": 19545 + }, + { + "epoch": 0.47, + "learning_rate": 1.1450617303171704e-05, + "loss": 1.125, + "step": 19546 + }, + { + "epoch": 0.47, + "learning_rate": 1.1449846643049462e-05, + "loss": 1.0641, + "step": 19547 + }, + { + "epoch": 0.47, + "learning_rate": 1.1449075974131347e-05, + "loss": 1.1245, + "step": 19548 + }, + { + "epoch": 0.47, + "learning_rate": 1.1448305296422026e-05, + "loss": 1.2241, + "step": 19549 + }, + { + "epoch": 0.47, + "learning_rate": 1.1447534609926179e-05, + "loss": 1.0441, + "step": 19550 + }, + { + "epoch": 0.47, + "learning_rate": 1.1446763914648478e-05, + "loss": 1.0851, + "step": 19551 + }, + { + "epoch": 0.47, + "learning_rate": 1.14459932105936e-05, + "loss": 1.0925, + "step": 19552 + }, + { + "epoch": 0.47, + "learning_rate": 1.1445222497766223e-05, + "loss": 1.1401, + "step": 19553 + }, + { + "epoch": 0.47, + "learning_rate": 1.1444451776171021e-05, + "loss": 1.4354, + "step": 19554 + }, + { + "epoch": 0.47, + "learning_rate": 1.144368104581267e-05, + "loss": 1.1689, + "step": 19555 + }, + { + "epoch": 0.47, + "learning_rate": 1.1442910306695845e-05, + "loss": 1.1048, + "step": 19556 + }, + { + "epoch": 0.47, + "learning_rate": 1.1442139558825223e-05, + "loss": 0.9401, + "step": 19557 + }, + { + "epoch": 0.47, + "learning_rate": 1.144136880220548e-05, + "loss": 1.1128, + "step": 19558 + }, + { + "epoch": 0.47, + "learning_rate": 1.1440598036841288e-05, + "loss": 0.9217, + "step": 19559 + }, + { + "epoch": 0.47, + "learning_rate": 1.143982726273733e-05, + "loss": 1.0023, + "step": 19560 + }, + { + "epoch": 0.47, + "learning_rate": 1.1439056479898278e-05, + "loss": 0.9413, + "step": 19561 + }, + { + "epoch": 0.47, + "learning_rate": 1.1438285688328811e-05, + "loss": 1.0422, + "step": 19562 + }, + { + "epoch": 0.47, + "learning_rate": 1.14375148880336e-05, + "loss": 1.1863, + "step": 19563 + }, + { + "epoch": 0.47, + "learning_rate": 1.1436744079017326e-05, + "loss": 1.0598, + "step": 19564 + }, + { + "epoch": 0.47, + "learning_rate": 1.1435973261284663e-05, + "loss": 1.1857, + "step": 19565 + }, + { + "epoch": 0.47, + "learning_rate": 1.1435202434840287e-05, + "loss": 1.2792, + "step": 19566 + }, + { + "epoch": 0.47, + "learning_rate": 1.1434431599688876e-05, + "loss": 1.0264, + "step": 19567 + }, + { + "epoch": 0.47, + "learning_rate": 1.1433660755835105e-05, + "loss": 1.1569, + "step": 19568 + }, + { + "epoch": 0.47, + "learning_rate": 1.1432889903283653e-05, + "loss": 1.015, + "step": 19569 + }, + { + "epoch": 0.47, + "learning_rate": 1.1432119042039196e-05, + "loss": 1.2306, + "step": 19570 + }, + { + "epoch": 0.47, + "learning_rate": 1.1431348172106407e-05, + "loss": 1.0094, + "step": 19571 + }, + { + "epoch": 0.47, + "learning_rate": 1.1430577293489966e-05, + "loss": 0.7931, + "step": 19572 + }, + { + "epoch": 0.47, + "learning_rate": 1.142980640619455e-05, + "loss": 1.3015, + "step": 19573 + }, + { + "epoch": 0.47, + "learning_rate": 1.1429035510224833e-05, + "loss": 1.0291, + "step": 19574 + }, + { + "epoch": 0.47, + "learning_rate": 1.1428264605585494e-05, + "loss": 1.2336, + "step": 19575 + }, + { + "epoch": 0.47, + "learning_rate": 1.1427493692281213e-05, + "loss": 1.0518, + "step": 19576 + }, + { + "epoch": 0.47, + "learning_rate": 1.142672277031666e-05, + "loss": 1.0352, + "step": 19577 + }, + { + "epoch": 0.47, + "learning_rate": 1.1425951839696517e-05, + "loss": 1.1153, + "step": 19578 + }, + { + "epoch": 0.47, + "learning_rate": 1.142518090042546e-05, + "loss": 1.2918, + "step": 19579 + }, + { + "epoch": 0.47, + "learning_rate": 1.1424409952508165e-05, + "loss": 1.1104, + "step": 19580 + }, + { + "epoch": 0.47, + "learning_rate": 1.1423638995949309e-05, + "loss": 1.0769, + "step": 19581 + }, + { + "epoch": 0.47, + "learning_rate": 1.1422868030753569e-05, + "loss": 1.212, + "step": 19582 + }, + { + "epoch": 0.47, + "learning_rate": 1.1422097056925627e-05, + "loss": 1.3131, + "step": 19583 + }, + { + "epoch": 0.47, + "learning_rate": 1.1421326074470156e-05, + "loss": 1.0913, + "step": 19584 + }, + { + "epoch": 0.47, + "learning_rate": 1.1420555083391833e-05, + "loss": 1.1985, + "step": 19585 + }, + { + "epoch": 0.47, + "learning_rate": 1.1419784083695334e-05, + "loss": 1.242, + "step": 19586 + }, + { + "epoch": 0.47, + "learning_rate": 1.1419013075385342e-05, + "loss": 1.3205, + "step": 19587 + }, + { + "epoch": 0.47, + "learning_rate": 1.141824205846653e-05, + "loss": 0.8006, + "step": 19588 + }, + { + "epoch": 0.47, + "learning_rate": 1.1417471032943579e-05, + "loss": 1.1017, + "step": 19589 + }, + { + "epoch": 0.47, + "learning_rate": 1.141669999882116e-05, + "loss": 1.131, + "step": 19590 + }, + { + "epoch": 0.47, + "learning_rate": 1.141592895610396e-05, + "loss": 1.1375, + "step": 19591 + }, + { + "epoch": 0.47, + "learning_rate": 1.1415157904796652e-05, + "loss": 1.1512, + "step": 19592 + }, + { + "epoch": 0.47, + "learning_rate": 1.1414386844903913e-05, + "loss": 1.0799, + "step": 19593 + }, + { + "epoch": 0.47, + "learning_rate": 1.141361577643042e-05, + "loss": 0.9981, + "step": 19594 + }, + { + "epoch": 0.47, + "learning_rate": 1.1412844699380856e-05, + "loss": 1.0745, + "step": 19595 + }, + { + "epoch": 0.47, + "learning_rate": 1.1412073613759894e-05, + "loss": 1.1623, + "step": 19596 + }, + { + "epoch": 0.47, + "learning_rate": 1.1411302519572215e-05, + "loss": 1.1785, + "step": 19597 + }, + { + "epoch": 0.47, + "learning_rate": 1.1410531416822491e-05, + "loss": 1.1233, + "step": 19598 + }, + { + "epoch": 0.47, + "learning_rate": 1.1409760305515407e-05, + "loss": 0.8732, + "step": 19599 + }, + { + "epoch": 0.47, + "learning_rate": 1.140898918565564e-05, + "loss": 1.0717, + "step": 19600 + }, + { + "epoch": 0.47, + "learning_rate": 1.1408218057247866e-05, + "loss": 0.8904, + "step": 19601 + }, + { + "epoch": 0.47, + "learning_rate": 1.1407446920296763e-05, + "loss": 1.2103, + "step": 19602 + }, + { + "epoch": 0.47, + "learning_rate": 1.1406675774807012e-05, + "loss": 0.8387, + "step": 19603 + }, + { + "epoch": 0.47, + "learning_rate": 1.1405904620783291e-05, + "loss": 1.0759, + "step": 19604 + }, + { + "epoch": 0.47, + "learning_rate": 1.1405133458230275e-05, + "loss": 1.0542, + "step": 19605 + }, + { + "epoch": 0.47, + "learning_rate": 1.1404362287152646e-05, + "loss": 0.968, + "step": 19606 + }, + { + "epoch": 0.47, + "learning_rate": 1.140359110755508e-05, + "loss": 1.0307, + "step": 19607 + }, + { + "epoch": 0.47, + "learning_rate": 1.140281991944226e-05, + "loss": 0.9384, + "step": 19608 + }, + { + "epoch": 0.47, + "learning_rate": 1.1402048722818862e-05, + "loss": 1.0361, + "step": 19609 + }, + { + "epoch": 0.47, + "learning_rate": 1.140127751768956e-05, + "loss": 1.2444, + "step": 19610 + }, + { + "epoch": 0.47, + "learning_rate": 1.1400506304059038e-05, + "loss": 0.9171, + "step": 19611 + }, + { + "epoch": 0.47, + "learning_rate": 1.1399735081931975e-05, + "loss": 0.9248, + "step": 19612 + }, + { + "epoch": 0.47, + "learning_rate": 1.139896385131305e-05, + "loss": 1.4324, + "step": 19613 + }, + { + "epoch": 0.47, + "learning_rate": 1.1398192612206936e-05, + "loss": 1.1533, + "step": 19614 + }, + { + "epoch": 0.47, + "learning_rate": 1.139742136461832e-05, + "loss": 1.0695, + "step": 19615 + }, + { + "epoch": 0.47, + "learning_rate": 1.1396650108551874e-05, + "loss": 0.7516, + "step": 19616 + }, + { + "epoch": 0.47, + "learning_rate": 1.1395878844012284e-05, + "loss": 1.0519, + "step": 19617 + }, + { + "epoch": 0.47, + "learning_rate": 1.1395107571004223e-05, + "loss": 1.1641, + "step": 19618 + }, + { + "epoch": 0.47, + "learning_rate": 1.1394336289532372e-05, + "loss": 1.0601, + "step": 19619 + }, + { + "epoch": 0.47, + "learning_rate": 1.1393564999601415e-05, + "loss": 0.8713, + "step": 19620 + }, + { + "epoch": 0.47, + "learning_rate": 1.139279370121602e-05, + "loss": 1.1172, + "step": 19621 + }, + { + "epoch": 0.47, + "learning_rate": 1.1392022394380877e-05, + "loss": 1.1135, + "step": 19622 + }, + { + "epoch": 0.47, + "learning_rate": 1.1391251079100663e-05, + "loss": 0.9562, + "step": 19623 + }, + { + "epoch": 0.47, + "learning_rate": 1.1390479755380053e-05, + "loss": 1.1526, + "step": 19624 + }, + { + "epoch": 0.47, + "learning_rate": 1.1389708423223731e-05, + "loss": 1.3084, + "step": 19625 + }, + { + "epoch": 0.47, + "learning_rate": 1.1388937082636373e-05, + "loss": 0.88, + "step": 19626 + }, + { + "epoch": 0.47, + "learning_rate": 1.1388165733622662e-05, + "loss": 1.0628, + "step": 19627 + }, + { + "epoch": 0.47, + "learning_rate": 1.1387394376187277e-05, + "loss": 1.1308, + "step": 19628 + }, + { + "epoch": 0.47, + "learning_rate": 1.1386623010334898e-05, + "loss": 1.1727, + "step": 19629 + }, + { + "epoch": 0.47, + "learning_rate": 1.13858516360702e-05, + "loss": 1.2254, + "step": 19630 + }, + { + "epoch": 0.47, + "learning_rate": 1.1385080253397867e-05, + "loss": 1.0365, + "step": 19631 + }, + { + "epoch": 0.47, + "learning_rate": 1.1384308862322579e-05, + "loss": 1.1808, + "step": 19632 + }, + { + "epoch": 0.47, + "learning_rate": 1.1383537462849014e-05, + "loss": 1.1838, + "step": 19633 + }, + { + "epoch": 0.47, + "learning_rate": 1.1382766054981851e-05, + "loss": 1.0195, + "step": 19634 + }, + { + "epoch": 0.47, + "learning_rate": 1.1381994638725776e-05, + "loss": 1.0888, + "step": 19635 + }, + { + "epoch": 0.47, + "learning_rate": 1.138122321408546e-05, + "loss": 0.8637, + "step": 19636 + }, + { + "epoch": 0.47, + "learning_rate": 1.1380451781065591e-05, + "loss": 1.2597, + "step": 19637 + }, + { + "epoch": 0.47, + "learning_rate": 1.1379680339670843e-05, + "loss": 1.0924, + "step": 19638 + }, + { + "epoch": 0.47, + "learning_rate": 1.13789088899059e-05, + "loss": 1.262, + "step": 19639 + }, + { + "epoch": 0.47, + "learning_rate": 1.137813743177544e-05, + "loss": 0.9345, + "step": 19640 + }, + { + "epoch": 0.47, + "learning_rate": 1.1377365965284145e-05, + "loss": 1.0427, + "step": 19641 + }, + { + "epoch": 0.47, + "learning_rate": 1.1376594490436697e-05, + "loss": 0.8199, + "step": 19642 + }, + { + "epoch": 0.47, + "learning_rate": 1.1375823007237771e-05, + "loss": 1.1744, + "step": 19643 + }, + { + "epoch": 0.47, + "learning_rate": 1.1375051515692052e-05, + "loss": 1.1779, + "step": 19644 + }, + { + "epoch": 0.47, + "learning_rate": 1.137428001580422e-05, + "loss": 0.9805, + "step": 19645 + }, + { + "epoch": 0.47, + "learning_rate": 1.1373508507578951e-05, + "loss": 0.8211, + "step": 19646 + }, + { + "epoch": 0.47, + "learning_rate": 1.1372736991020932e-05, + "loss": 0.9686, + "step": 19647 + }, + { + "epoch": 0.47, + "learning_rate": 1.137196546613484e-05, + "loss": 1.2333, + "step": 19648 + }, + { + "epoch": 0.47, + "learning_rate": 1.1371193932925356e-05, + "loss": 1.0158, + "step": 19649 + }, + { + "epoch": 0.47, + "learning_rate": 1.1370422391397158e-05, + "loss": 1.2825, + "step": 19650 + }, + { + "epoch": 0.47, + "learning_rate": 1.1369650841554933e-05, + "loss": 0.9829, + "step": 19651 + }, + { + "epoch": 0.47, + "learning_rate": 1.136887928340336e-05, + "loss": 1.0304, + "step": 19652 + }, + { + "epoch": 0.47, + "learning_rate": 1.1368107716947116e-05, + "loss": 1.0766, + "step": 19653 + }, + { + "epoch": 0.47, + "learning_rate": 1.1367336142190881e-05, + "loss": 0.9987, + "step": 19654 + }, + { + "epoch": 0.47, + "learning_rate": 1.1366564559139346e-05, + "loss": 0.9946, + "step": 19655 + }, + { + "epoch": 0.47, + "learning_rate": 1.1365792967797181e-05, + "loss": 0.7124, + "step": 19656 + }, + { + "epoch": 0.47, + "learning_rate": 1.1365021368169074e-05, + "loss": 1.1693, + "step": 19657 + }, + { + "epoch": 0.47, + "learning_rate": 1.1364249760259698e-05, + "loss": 0.752, + "step": 19658 + }, + { + "epoch": 0.47, + "learning_rate": 1.1363478144073745e-05, + "loss": 1.1035, + "step": 19659 + }, + { + "epoch": 0.47, + "learning_rate": 1.1362706519615889e-05, + "loss": 0.9282, + "step": 19660 + }, + { + "epoch": 0.47, + "learning_rate": 1.1361934886890813e-05, + "loss": 1.0042, + "step": 19661 + }, + { + "epoch": 0.47, + "learning_rate": 1.1361163245903196e-05, + "loss": 1.0891, + "step": 19662 + }, + { + "epoch": 0.47, + "learning_rate": 1.1360391596657727e-05, + "loss": 1.0663, + "step": 19663 + }, + { + "epoch": 0.47, + "learning_rate": 1.1359619939159079e-05, + "loss": 1.2996, + "step": 19664 + }, + { + "epoch": 0.47, + "learning_rate": 1.1358848273411937e-05, + "loss": 1.2639, + "step": 19665 + }, + { + "epoch": 0.47, + "learning_rate": 1.1358076599420981e-05, + "loss": 1.0665, + "step": 19666 + }, + { + "epoch": 0.47, + "learning_rate": 1.1357304917190897e-05, + "loss": 1.1984, + "step": 19667 + }, + { + "epoch": 0.47, + "learning_rate": 1.1356533226726363e-05, + "loss": 1.179, + "step": 19668 + }, + { + "epoch": 0.47, + "learning_rate": 1.1355761528032056e-05, + "loss": 1.3049, + "step": 19669 + }, + { + "epoch": 0.47, + "learning_rate": 1.1354989821112667e-05, + "loss": 0.867, + "step": 19670 + }, + { + "epoch": 0.47, + "learning_rate": 1.1354218105972871e-05, + "loss": 0.8412, + "step": 19671 + }, + { + "epoch": 0.47, + "learning_rate": 1.1353446382617357e-05, + "loss": 1.0418, + "step": 19672 + }, + { + "epoch": 0.47, + "learning_rate": 1.1352674651050797e-05, + "loss": 0.9271, + "step": 19673 + }, + { + "epoch": 0.47, + "learning_rate": 1.1351902911277878e-05, + "loss": 1.3067, + "step": 19674 + }, + { + "epoch": 0.47, + "learning_rate": 1.1351131163303285e-05, + "loss": 1.1833, + "step": 19675 + }, + { + "epoch": 0.47, + "learning_rate": 1.1350359407131699e-05, + "loss": 1.0298, + "step": 19676 + }, + { + "epoch": 0.47, + "learning_rate": 1.1349587642767794e-05, + "loss": 1.1706, + "step": 19677 + }, + { + "epoch": 0.47, + "learning_rate": 1.1348815870216263e-05, + "loss": 1.2998, + "step": 19678 + }, + { + "epoch": 0.47, + "learning_rate": 1.1348044089481783e-05, + "loss": 1.2064, + "step": 19679 + }, + { + "epoch": 0.47, + "learning_rate": 1.1347272300569034e-05, + "loss": 1.0551, + "step": 19680 + }, + { + "epoch": 0.47, + "learning_rate": 1.13465005034827e-05, + "loss": 1.24, + "step": 19681 + }, + { + "epoch": 0.47, + "learning_rate": 1.1345728698227467e-05, + "loss": 1.0181, + "step": 19682 + }, + { + "epoch": 0.47, + "learning_rate": 1.1344956884808017e-05, + "loss": 1.2237, + "step": 19683 + }, + { + "epoch": 0.47, + "learning_rate": 1.1344185063229025e-05, + "loss": 1.2061, + "step": 19684 + }, + { + "epoch": 0.47, + "learning_rate": 1.134341323349518e-05, + "loss": 1.147, + "step": 19685 + }, + { + "epoch": 0.47, + "learning_rate": 1.1342641395611162e-05, + "loss": 0.949, + "step": 19686 + }, + { + "epoch": 0.47, + "learning_rate": 1.1341869549581657e-05, + "loss": 1.1747, + "step": 19687 + }, + { + "epoch": 0.47, + "learning_rate": 1.1341097695411345e-05, + "loss": 1.0057, + "step": 19688 + }, + { + "epoch": 0.47, + "learning_rate": 1.1340325833104904e-05, + "loss": 1.0231, + "step": 19689 + }, + { + "epoch": 0.47, + "learning_rate": 1.1339553962667024e-05, + "loss": 1.0594, + "step": 19690 + }, + { + "epoch": 0.47, + "learning_rate": 1.1338782084102389e-05, + "loss": 0.7326, + "step": 19691 + }, + { + "epoch": 0.47, + "learning_rate": 1.1338010197415675e-05, + "loss": 1.0869, + "step": 19692 + }, + { + "epoch": 0.47, + "learning_rate": 1.1337238302611566e-05, + "loss": 1.2789, + "step": 19693 + }, + { + "epoch": 0.47, + "learning_rate": 1.133646639969475e-05, + "loss": 0.9454, + "step": 19694 + }, + { + "epoch": 0.47, + "learning_rate": 1.1335694488669903e-05, + "loss": 1.2121, + "step": 19695 + }, + { + "epoch": 0.47, + "learning_rate": 1.1334922569541715e-05, + "loss": 1.0135, + "step": 19696 + }, + { + "epoch": 0.47, + "learning_rate": 1.1334150642314864e-05, + "loss": 1.0966, + "step": 19697 + }, + { + "epoch": 0.47, + "learning_rate": 1.1333378706994036e-05, + "loss": 1.1591, + "step": 19698 + }, + { + "epoch": 0.47, + "learning_rate": 1.1332606763583914e-05, + "loss": 1.0976, + "step": 19699 + }, + { + "epoch": 0.47, + "learning_rate": 1.133183481208918e-05, + "loss": 0.9427, + "step": 19700 + }, + { + "epoch": 0.47, + "learning_rate": 1.1331062852514514e-05, + "loss": 1.0711, + "step": 19701 + }, + { + "epoch": 0.47, + "learning_rate": 1.1330290884864605e-05, + "loss": 1.0932, + "step": 19702 + }, + { + "epoch": 0.47, + "learning_rate": 1.1329518909144135e-05, + "loss": 0.9521, + "step": 19703 + }, + { + "epoch": 0.47, + "learning_rate": 1.1328746925357785e-05, + "loss": 1.0097, + "step": 19704 + }, + { + "epoch": 0.47, + "learning_rate": 1.132797493351024e-05, + "loss": 1.2242, + "step": 19705 + }, + { + "epoch": 0.47, + "learning_rate": 1.1327202933606185e-05, + "loss": 1.1696, + "step": 19706 + }, + { + "epoch": 0.47, + "learning_rate": 1.1326430925650298e-05, + "loss": 0.883, + "step": 19707 + }, + { + "epoch": 0.47, + "learning_rate": 1.132565890964727e-05, + "loss": 1.0743, + "step": 19708 + }, + { + "epoch": 0.47, + "learning_rate": 1.132488688560178e-05, + "loss": 1.2588, + "step": 19709 + }, + { + "epoch": 0.47, + "learning_rate": 1.1324114853518515e-05, + "loss": 1.0, + "step": 19710 + }, + { + "epoch": 0.47, + "learning_rate": 1.1323342813402157e-05, + "loss": 1.207, + "step": 19711 + }, + { + "epoch": 0.47, + "learning_rate": 1.1322570765257387e-05, + "loss": 0.939, + "step": 19712 + }, + { + "epoch": 0.47, + "learning_rate": 1.1321798709088889e-05, + "loss": 1.2383, + "step": 19713 + }, + { + "epoch": 0.47, + "learning_rate": 1.1321026644901353e-05, + "loss": 1.1583, + "step": 19714 + }, + { + "epoch": 0.47, + "learning_rate": 1.1320254572699458e-05, + "loss": 0.9033, + "step": 19715 + }, + { + "epoch": 0.47, + "learning_rate": 1.1319482492487893e-05, + "loss": 1.0767, + "step": 19716 + }, + { + "epoch": 0.47, + "learning_rate": 1.131871040427133e-05, + "loss": 1.1679, + "step": 19717 + }, + { + "epoch": 0.47, + "learning_rate": 1.1317938308054468e-05, + "loss": 0.8733, + "step": 19718 + }, + { + "epoch": 0.47, + "learning_rate": 1.131716620384198e-05, + "loss": 1.188, + "step": 19719 + }, + { + "epoch": 0.47, + "learning_rate": 1.1316394091638559e-05, + "loss": 1.0323, + "step": 19720 + }, + { + "epoch": 0.47, + "learning_rate": 1.1315621971448878e-05, + "loss": 0.5517, + "step": 19721 + }, + { + "epoch": 0.47, + "learning_rate": 1.1314849843277633e-05, + "loss": 1.2328, + "step": 19722 + }, + { + "epoch": 0.47, + "learning_rate": 1.1314077707129502e-05, + "loss": 1.0767, + "step": 19723 + }, + { + "epoch": 0.47, + "learning_rate": 1.1313305563009172e-05, + "loss": 1.0617, + "step": 19724 + }, + { + "epoch": 0.47, + "learning_rate": 1.1312533410921325e-05, + "loss": 1.0851, + "step": 19725 + }, + { + "epoch": 0.47, + "learning_rate": 1.1311761250870645e-05, + "loss": 0.996, + "step": 19726 + }, + { + "epoch": 0.47, + "learning_rate": 1.131098908286182e-05, + "loss": 1.1367, + "step": 19727 + }, + { + "epoch": 0.47, + "learning_rate": 1.1310216906899534e-05, + "loss": 1.2393, + "step": 19728 + }, + { + "epoch": 0.47, + "learning_rate": 1.1309444722988464e-05, + "loss": 1.1918, + "step": 19729 + }, + { + "epoch": 0.47, + "learning_rate": 1.1308672531133306e-05, + "loss": 1.0471, + "step": 19730 + }, + { + "epoch": 0.47, + "learning_rate": 1.1307900331338742e-05, + "loss": 0.987, + "step": 19731 + }, + { + "epoch": 0.47, + "learning_rate": 1.130712812360945e-05, + "loss": 1.0095, + "step": 19732 + }, + { + "epoch": 0.47, + "learning_rate": 1.130635590795012e-05, + "loss": 1.3307, + "step": 19733 + }, + { + "epoch": 0.47, + "learning_rate": 1.1305583684365436e-05, + "loss": 1.2557, + "step": 19734 + }, + { + "epoch": 0.47, + "learning_rate": 1.1304811452860083e-05, + "loss": 1.1818, + "step": 19735 + }, + { + "epoch": 0.47, + "learning_rate": 1.1304039213438747e-05, + "loss": 1.0602, + "step": 19736 + }, + { + "epoch": 0.47, + "learning_rate": 1.1303266966106107e-05, + "loss": 1.1603, + "step": 19737 + }, + { + "epoch": 0.47, + "learning_rate": 1.1302494710866859e-05, + "loss": 1.0864, + "step": 19738 + }, + { + "epoch": 0.47, + "learning_rate": 1.1301722447725679e-05, + "loss": 1.0371, + "step": 19739 + }, + { + "epoch": 0.47, + "learning_rate": 1.1300950176687255e-05, + "loss": 1.0629, + "step": 19740 + }, + { + "epoch": 0.47, + "learning_rate": 1.1300177897756273e-05, + "loss": 1.1679, + "step": 19741 + }, + { + "epoch": 0.47, + "learning_rate": 1.1299405610937417e-05, + "loss": 1.2673, + "step": 19742 + }, + { + "epoch": 0.47, + "learning_rate": 1.1298633316235374e-05, + "loss": 0.8611, + "step": 19743 + }, + { + "epoch": 0.47, + "learning_rate": 1.1297861013654827e-05, + "loss": 0.7921, + "step": 19744 + }, + { + "epoch": 0.47, + "learning_rate": 1.129708870320046e-05, + "loss": 1.0868, + "step": 19745 + }, + { + "epoch": 0.47, + "learning_rate": 1.1296316384876966e-05, + "loss": 1.101, + "step": 19746 + }, + { + "epoch": 0.47, + "learning_rate": 1.1295544058689023e-05, + "loss": 1.2567, + "step": 19747 + }, + { + "epoch": 0.47, + "learning_rate": 1.1294771724641318e-05, + "loss": 1.1246, + "step": 19748 + }, + { + "epoch": 0.47, + "learning_rate": 1.1293999382738538e-05, + "loss": 1.3536, + "step": 19749 + }, + { + "epoch": 0.47, + "learning_rate": 1.129322703298537e-05, + "loss": 1.112, + "step": 19750 + }, + { + "epoch": 0.47, + "learning_rate": 1.1292454675386498e-05, + "loss": 0.9169, + "step": 19751 + }, + { + "epoch": 0.48, + "learning_rate": 1.1291682309946604e-05, + "loss": 1.0704, + "step": 19752 + }, + { + "epoch": 0.48, + "learning_rate": 1.1290909936670379e-05, + "loss": 1.0757, + "step": 19753 + }, + { + "epoch": 0.48, + "learning_rate": 1.1290137555562507e-05, + "loss": 1.3323, + "step": 19754 + }, + { + "epoch": 0.48, + "learning_rate": 1.1289365166627674e-05, + "loss": 0.8707, + "step": 19755 + }, + { + "epoch": 0.48, + "learning_rate": 1.1288592769870565e-05, + "loss": 1.1665, + "step": 19756 + }, + { + "epoch": 0.48, + "learning_rate": 1.1287820365295868e-05, + "loss": 1.0233, + "step": 19757 + }, + { + "epoch": 0.48, + "learning_rate": 1.1287047952908268e-05, + "loss": 0.9385, + "step": 19758 + }, + { + "epoch": 0.48, + "learning_rate": 1.1286275532712452e-05, + "loss": 1.1587, + "step": 19759 + }, + { + "epoch": 0.48, + "learning_rate": 1.12855031047131e-05, + "loss": 1.0727, + "step": 19760 + }, + { + "epoch": 0.48, + "learning_rate": 1.1284730668914908e-05, + "loss": 1.1923, + "step": 19761 + }, + { + "epoch": 0.48, + "learning_rate": 1.1283958225322556e-05, + "loss": 1.0537, + "step": 19762 + }, + { + "epoch": 0.48, + "learning_rate": 1.128318577394073e-05, + "loss": 1.0577, + "step": 19763 + }, + { + "epoch": 0.48, + "learning_rate": 1.1282413314774117e-05, + "loss": 1.0169, + "step": 19764 + }, + { + "epoch": 0.48, + "learning_rate": 1.1281640847827407e-05, + "loss": 1.006, + "step": 19765 + }, + { + "epoch": 0.48, + "learning_rate": 1.1280868373105282e-05, + "loss": 1.0354, + "step": 19766 + }, + { + "epoch": 0.48, + "learning_rate": 1.128009589061243e-05, + "loss": 1.2824, + "step": 19767 + }, + { + "epoch": 0.48, + "learning_rate": 1.1279323400353536e-05, + "loss": 1.0588, + "step": 19768 + }, + { + "epoch": 0.48, + "learning_rate": 1.127855090233329e-05, + "loss": 1.1492, + "step": 19769 + }, + { + "epoch": 0.48, + "learning_rate": 1.1277778396556376e-05, + "loss": 1.0172, + "step": 19770 + }, + { + "epoch": 0.48, + "learning_rate": 1.1277005883027481e-05, + "loss": 1.2314, + "step": 19771 + }, + { + "epoch": 0.48, + "learning_rate": 1.127623336175129e-05, + "loss": 0.9709, + "step": 19772 + }, + { + "epoch": 0.48, + "learning_rate": 1.1275460832732491e-05, + "loss": 1.1775, + "step": 19773 + }, + { + "epoch": 0.48, + "learning_rate": 1.1274688295975775e-05, + "loss": 0.9954, + "step": 19774 + }, + { + "epoch": 0.48, + "learning_rate": 1.1273915751485821e-05, + "loss": 1.2087, + "step": 19775 + }, + { + "epoch": 0.48, + "learning_rate": 1.1273143199267323e-05, + "loss": 0.9019, + "step": 19776 + }, + { + "epoch": 0.48, + "learning_rate": 1.1272370639324963e-05, + "loss": 0.84, + "step": 19777 + }, + { + "epoch": 0.48, + "learning_rate": 1.127159807166343e-05, + "loss": 0.9295, + "step": 19778 + }, + { + "epoch": 0.48, + "learning_rate": 1.127082549628741e-05, + "loss": 1.0464, + "step": 19779 + }, + { + "epoch": 0.48, + "learning_rate": 1.127005291320159e-05, + "loss": 0.9648, + "step": 19780 + }, + { + "epoch": 0.48, + "learning_rate": 1.1269280322410658e-05, + "loss": 0.9568, + "step": 19781 + }, + { + "epoch": 0.48, + "learning_rate": 1.1268507723919302e-05, + "loss": 0.8944, + "step": 19782 + }, + { + "epoch": 0.48, + "learning_rate": 1.1267735117732207e-05, + "loss": 0.9259, + "step": 19783 + }, + { + "epoch": 0.48, + "learning_rate": 1.1266962503854062e-05, + "loss": 1.1863, + "step": 19784 + }, + { + "epoch": 0.48, + "learning_rate": 1.1266189882289552e-05, + "loss": 1.08, + "step": 19785 + }, + { + "epoch": 0.48, + "learning_rate": 1.1265417253043368e-05, + "loss": 0.9485, + "step": 19786 + }, + { + "epoch": 0.48, + "learning_rate": 1.1264644616120195e-05, + "loss": 1.2853, + "step": 19787 + }, + { + "epoch": 0.48, + "learning_rate": 1.1263871971524718e-05, + "loss": 0.9592, + "step": 19788 + }, + { + "epoch": 0.48, + "learning_rate": 1.126309931926163e-05, + "loss": 0.968, + "step": 19789 + }, + { + "epoch": 0.48, + "learning_rate": 1.1262326659335611e-05, + "loss": 1.1056, + "step": 19790 + }, + { + "epoch": 0.48, + "learning_rate": 1.126155399175136e-05, + "loss": 0.8468, + "step": 19791 + }, + { + "epoch": 0.48, + "learning_rate": 1.1260781316513551e-05, + "loss": 1.1123, + "step": 19792 + }, + { + "epoch": 0.48, + "learning_rate": 1.1260008633626882e-05, + "loss": 1.3105, + "step": 19793 + }, + { + "epoch": 0.48, + "learning_rate": 1.1259235943096036e-05, + "loss": 0.982, + "step": 19794 + }, + { + "epoch": 0.48, + "learning_rate": 1.1258463244925704e-05, + "loss": 1.127, + "step": 19795 + }, + { + "epoch": 0.48, + "learning_rate": 1.1257690539120567e-05, + "loss": 0.8949, + "step": 19796 + }, + { + "epoch": 0.48, + "learning_rate": 1.125691782568532e-05, + "loss": 1.3001, + "step": 19797 + }, + { + "epoch": 0.48, + "learning_rate": 1.125614510462465e-05, + "loss": 0.8346, + "step": 19798 + }, + { + "epoch": 0.48, + "learning_rate": 1.1255372375943242e-05, + "loss": 1.0544, + "step": 19799 + }, + { + "epoch": 0.48, + "learning_rate": 1.1254599639645783e-05, + "loss": 1.201, + "step": 19800 + }, + { + "epoch": 0.48, + "learning_rate": 1.1253826895736964e-05, + "loss": 0.9815, + "step": 19801 + }, + { + "epoch": 0.48, + "learning_rate": 1.1253054144221474e-05, + "loss": 1.0781, + "step": 19802 + }, + { + "epoch": 0.48, + "learning_rate": 1.1252281385103997e-05, + "loss": 1.1681, + "step": 19803 + }, + { + "epoch": 0.48, + "learning_rate": 1.1251508618389221e-05, + "loss": 1.0346, + "step": 19804 + }, + { + "epoch": 0.48, + "learning_rate": 1.1250735844081842e-05, + "loss": 0.9912, + "step": 19805 + }, + { + "epoch": 0.48, + "learning_rate": 1.1249963062186541e-05, + "loss": 0.8699, + "step": 19806 + }, + { + "epoch": 0.48, + "learning_rate": 1.1249190272708009e-05, + "loss": 1.0905, + "step": 19807 + }, + { + "epoch": 0.48, + "learning_rate": 1.124841747565093e-05, + "loss": 1.0644, + "step": 19808 + }, + { + "epoch": 0.48, + "learning_rate": 1.124764467102e-05, + "loss": 1.2011, + "step": 19809 + }, + { + "epoch": 0.48, + "learning_rate": 1.1246871858819902e-05, + "loss": 1.1484, + "step": 19810 + }, + { + "epoch": 0.48, + "learning_rate": 1.1246099039055323e-05, + "loss": 1.1874, + "step": 19811 + }, + { + "epoch": 0.48, + "learning_rate": 1.1245326211730956e-05, + "loss": 1.0811, + "step": 19812 + }, + { + "epoch": 0.48, + "learning_rate": 1.1244553376851488e-05, + "loss": 1.2106, + "step": 19813 + }, + { + "epoch": 0.48, + "learning_rate": 1.1243780534421607e-05, + "loss": 0.8841, + "step": 19814 + }, + { + "epoch": 0.48, + "learning_rate": 1.1243007684446003e-05, + "loss": 0.9202, + "step": 19815 + }, + { + "epoch": 0.48, + "learning_rate": 1.124223482692936e-05, + "loss": 1.0777, + "step": 19816 + }, + { + "epoch": 0.48, + "learning_rate": 1.1241461961876377e-05, + "loss": 0.8625, + "step": 19817 + }, + { + "epoch": 0.48, + "learning_rate": 1.1240689089291732e-05, + "loss": 1.0904, + "step": 19818 + }, + { + "epoch": 0.48, + "learning_rate": 1.123991620918012e-05, + "loss": 1.0949, + "step": 19819 + }, + { + "epoch": 0.48, + "learning_rate": 1.1239143321546226e-05, + "loss": 1.3042, + "step": 19820 + }, + { + "epoch": 0.48, + "learning_rate": 1.1238370426394743e-05, + "loss": 1.0156, + "step": 19821 + }, + { + "epoch": 0.48, + "learning_rate": 1.1237597523730357e-05, + "loss": 0.9669, + "step": 19822 + }, + { + "epoch": 0.48, + "learning_rate": 1.1236824613557758e-05, + "loss": 1.0949, + "step": 19823 + }, + { + "epoch": 0.48, + "learning_rate": 1.1236051695881634e-05, + "loss": 1.1039, + "step": 19824 + }, + { + "epoch": 0.48, + "learning_rate": 1.1235278770706675e-05, + "loss": 1.029, + "step": 19825 + }, + { + "epoch": 0.48, + "learning_rate": 1.1234505838037574e-05, + "loss": 1.1211, + "step": 19826 + }, + { + "epoch": 0.48, + "learning_rate": 1.1233732897879013e-05, + "loss": 0.9054, + "step": 19827 + }, + { + "epoch": 0.48, + "learning_rate": 1.1232959950235682e-05, + "loss": 1.121, + "step": 19828 + }, + { + "epoch": 0.48, + "learning_rate": 1.1232186995112278e-05, + "loss": 1.3603, + "step": 19829 + }, + { + "epoch": 0.48, + "learning_rate": 1.1231414032513484e-05, + "loss": 1.0744, + "step": 19830 + }, + { + "epoch": 0.48, + "learning_rate": 1.1230641062443988e-05, + "loss": 1.0651, + "step": 19831 + }, + { + "epoch": 0.48, + "learning_rate": 1.1229868084908487e-05, + "loss": 0.8623, + "step": 19832 + }, + { + "epoch": 0.48, + "learning_rate": 1.1229095099911661e-05, + "loss": 1.0336, + "step": 19833 + }, + { + "epoch": 0.48, + "learning_rate": 1.1228322107458207e-05, + "loss": 1.1141, + "step": 19834 + }, + { + "epoch": 0.48, + "learning_rate": 1.1227549107552811e-05, + "loss": 1.1183, + "step": 19835 + }, + { + "epoch": 0.48, + "learning_rate": 1.1226776100200162e-05, + "loss": 1.1905, + "step": 19836 + }, + { + "epoch": 0.48, + "learning_rate": 1.1226003085404954e-05, + "loss": 1.2126, + "step": 19837 + }, + { + "epoch": 0.48, + "learning_rate": 1.1225230063171872e-05, + "loss": 1.2458, + "step": 19838 + }, + { + "epoch": 0.48, + "learning_rate": 1.1224457033505604e-05, + "loss": 1.0954, + "step": 19839 + }, + { + "epoch": 0.48, + "learning_rate": 1.1223683996410847e-05, + "loss": 1.3128, + "step": 19840 + }, + { + "epoch": 0.48, + "learning_rate": 1.122291095189229e-05, + "loss": 1.1897, + "step": 19841 + }, + { + "epoch": 0.48, + "learning_rate": 1.1222137899954615e-05, + "loss": 1.0889, + "step": 19842 + }, + { + "epoch": 0.48, + "learning_rate": 1.1221364840602518e-05, + "loss": 1.1497, + "step": 19843 + }, + { + "epoch": 0.48, + "learning_rate": 1.1220591773840687e-05, + "loss": 1.1432, + "step": 19844 + }, + { + "epoch": 0.48, + "learning_rate": 1.1219818699673814e-05, + "loss": 0.9608, + "step": 19845 + }, + { + "epoch": 0.48, + "learning_rate": 1.121904561810659e-05, + "loss": 1.0947, + "step": 19846 + }, + { + "epoch": 0.48, + "learning_rate": 1.1218272529143699e-05, + "loss": 1.0285, + "step": 19847 + }, + { + "epoch": 0.48, + "learning_rate": 1.1217499432789836e-05, + "loss": 1.0937, + "step": 19848 + }, + { + "epoch": 0.48, + "learning_rate": 1.1216726329049693e-05, + "loss": 1.3138, + "step": 19849 + }, + { + "epoch": 0.48, + "learning_rate": 1.1215953217927956e-05, + "loss": 0.9434, + "step": 19850 + }, + { + "epoch": 0.48, + "learning_rate": 1.1215180099429314e-05, + "loss": 1.0155, + "step": 19851 + }, + { + "epoch": 0.48, + "learning_rate": 1.1214406973558462e-05, + "loss": 0.876, + "step": 19852 + }, + { + "epoch": 0.48, + "learning_rate": 1.121363384032009e-05, + "loss": 1.2667, + "step": 19853 + }, + { + "epoch": 0.48, + "learning_rate": 1.1212860699718888e-05, + "loss": 0.9836, + "step": 19854 + }, + { + "epoch": 0.48, + "learning_rate": 1.121208755175954e-05, + "loss": 1.1044, + "step": 19855 + }, + { + "epoch": 0.48, + "learning_rate": 1.1211314396446746e-05, + "loss": 1.115, + "step": 19856 + }, + { + "epoch": 0.48, + "learning_rate": 1.1210541233785192e-05, + "loss": 1.1305, + "step": 19857 + }, + { + "epoch": 0.48, + "learning_rate": 1.1209768063779572e-05, + "loss": 1.1498, + "step": 19858 + }, + { + "epoch": 0.48, + "learning_rate": 1.1208994886434568e-05, + "loss": 0.9801, + "step": 19859 + }, + { + "epoch": 0.48, + "learning_rate": 1.120822170175488e-05, + "loss": 1.3115, + "step": 19860 + }, + { + "epoch": 0.48, + "learning_rate": 1.1207448509745193e-05, + "loss": 1.0681, + "step": 19861 + }, + { + "epoch": 0.48, + "learning_rate": 1.1206675310410201e-05, + "loss": 1.0434, + "step": 19862 + }, + { + "epoch": 0.48, + "learning_rate": 1.1205902103754593e-05, + "loss": 1.252, + "step": 19863 + }, + { + "epoch": 0.48, + "learning_rate": 1.1205128889783063e-05, + "loss": 1.1202, + "step": 19864 + }, + { + "epoch": 0.48, + "learning_rate": 1.1204355668500297e-05, + "loss": 1.2063, + "step": 19865 + }, + { + "epoch": 0.48, + "learning_rate": 1.120358243991099e-05, + "loss": 1.1107, + "step": 19866 + }, + { + "epoch": 0.48, + "learning_rate": 1.120280920401983e-05, + "loss": 1.1006, + "step": 19867 + }, + { + "epoch": 0.48, + "learning_rate": 1.1202035960831512e-05, + "loss": 0.8932, + "step": 19868 + }, + { + "epoch": 0.48, + "learning_rate": 1.1201262710350722e-05, + "loss": 1.1294, + "step": 19869 + }, + { + "epoch": 0.48, + "learning_rate": 1.1200489452582156e-05, + "loss": 0.7769, + "step": 19870 + }, + { + "epoch": 0.48, + "learning_rate": 1.1199716187530498e-05, + "loss": 1.2812, + "step": 19871 + }, + { + "epoch": 0.48, + "learning_rate": 1.119894291520045e-05, + "loss": 1.1198, + "step": 19872 + }, + { + "epoch": 0.48, + "learning_rate": 1.1198169635596694e-05, + "loss": 0.7912, + "step": 19873 + }, + { + "epoch": 0.48, + "learning_rate": 1.1197396348723923e-05, + "loss": 1.1525, + "step": 19874 + }, + { + "epoch": 0.48, + "learning_rate": 1.1196623054586833e-05, + "loss": 1.0809, + "step": 19875 + }, + { + "epoch": 0.48, + "learning_rate": 1.119584975319011e-05, + "loss": 1.2977, + "step": 19876 + }, + { + "epoch": 0.48, + "learning_rate": 1.119507644453845e-05, + "loss": 1.1228, + "step": 19877 + }, + { + "epoch": 0.48, + "learning_rate": 1.1194303128636543e-05, + "loss": 1.1162, + "step": 19878 + }, + { + "epoch": 0.48, + "learning_rate": 1.1193529805489078e-05, + "loss": 1.0548, + "step": 19879 + }, + { + "epoch": 0.48, + "learning_rate": 1.119275647510075e-05, + "loss": 0.8415, + "step": 19880 + }, + { + "epoch": 0.48, + "learning_rate": 1.1191983137476248e-05, + "loss": 1.4243, + "step": 19881 + }, + { + "epoch": 0.48, + "learning_rate": 1.1191209792620266e-05, + "loss": 0.9026, + "step": 19882 + }, + { + "epoch": 0.48, + "learning_rate": 1.1190436440537491e-05, + "loss": 1.3724, + "step": 19883 + }, + { + "epoch": 0.48, + "learning_rate": 1.1189663081232622e-05, + "loss": 0.9677, + "step": 19884 + }, + { + "epoch": 0.48, + "learning_rate": 1.1188889714710345e-05, + "loss": 1.141, + "step": 19885 + }, + { + "epoch": 0.48, + "learning_rate": 1.1188116340975355e-05, + "loss": 1.1749, + "step": 19886 + }, + { + "epoch": 0.48, + "learning_rate": 1.1187342960032342e-05, + "loss": 0.9849, + "step": 19887 + }, + { + "epoch": 0.48, + "learning_rate": 1.1186569571885996e-05, + "loss": 1.1697, + "step": 19888 + }, + { + "epoch": 0.48, + "learning_rate": 1.1185796176541015e-05, + "loss": 1.2059, + "step": 19889 + }, + { + "epoch": 0.48, + "learning_rate": 1.1185022774002087e-05, + "loss": 0.9584, + "step": 19890 + }, + { + "epoch": 0.48, + "learning_rate": 1.1184249364273903e-05, + "loss": 0.8563, + "step": 19891 + }, + { + "epoch": 0.48, + "learning_rate": 1.1183475947361158e-05, + "loss": 1.1264, + "step": 19892 + }, + { + "epoch": 0.48, + "learning_rate": 1.1182702523268543e-05, + "loss": 1.1423, + "step": 19893 + }, + { + "epoch": 0.48, + "learning_rate": 1.118192909200075e-05, + "loss": 1.147, + "step": 19894 + }, + { + "epoch": 0.48, + "learning_rate": 1.1181155653562468e-05, + "loss": 1.0246, + "step": 19895 + }, + { + "epoch": 0.48, + "learning_rate": 1.1180382207958396e-05, + "loss": 1.2644, + "step": 19896 + }, + { + "epoch": 0.48, + "learning_rate": 1.1179608755193221e-05, + "loss": 1.4656, + "step": 19897 + }, + { + "epoch": 0.48, + "learning_rate": 1.117883529527164e-05, + "loss": 0.8851, + "step": 19898 + }, + { + "epoch": 0.48, + "learning_rate": 1.1178061828198337e-05, + "loss": 1.3006, + "step": 19899 + }, + { + "epoch": 0.48, + "learning_rate": 1.1177288353978016e-05, + "loss": 1.2193, + "step": 19900 + }, + { + "epoch": 0.48, + "learning_rate": 1.117651487261536e-05, + "loss": 1.1548, + "step": 19901 + }, + { + "epoch": 0.48, + "learning_rate": 1.1175741384115067e-05, + "loss": 0.9879, + "step": 19902 + }, + { + "epoch": 0.48, + "learning_rate": 1.1174967888481826e-05, + "loss": 1.1311, + "step": 19903 + }, + { + "epoch": 0.48, + "learning_rate": 1.117419438572033e-05, + "loss": 1.1468, + "step": 19904 + }, + { + "epoch": 0.48, + "learning_rate": 1.1173420875835274e-05, + "loss": 1.0488, + "step": 19905 + }, + { + "epoch": 0.48, + "learning_rate": 1.1172647358831351e-05, + "loss": 1.3212, + "step": 19906 + }, + { + "epoch": 0.48, + "learning_rate": 1.117187383471325e-05, + "loss": 1.0008, + "step": 19907 + }, + { + "epoch": 0.48, + "learning_rate": 1.117110030348567e-05, + "loss": 1.0594, + "step": 19908 + }, + { + "epoch": 0.48, + "learning_rate": 1.1170326765153296e-05, + "loss": 1.0719, + "step": 19909 + }, + { + "epoch": 0.48, + "learning_rate": 1.1169553219720828e-05, + "loss": 1.0185, + "step": 19910 + }, + { + "epoch": 0.48, + "learning_rate": 1.1168779667192951e-05, + "loss": 1.14, + "step": 19911 + }, + { + "epoch": 0.48, + "learning_rate": 1.1168006107574365e-05, + "loss": 0.8924, + "step": 19912 + }, + { + "epoch": 0.48, + "learning_rate": 1.1167232540869764e-05, + "loss": 1.0287, + "step": 19913 + }, + { + "epoch": 0.48, + "learning_rate": 1.1166458967083836e-05, + "loss": 0.7861, + "step": 19914 + }, + { + "epoch": 0.48, + "learning_rate": 1.1165685386221272e-05, + "loss": 1.0193, + "step": 19915 + }, + { + "epoch": 0.48, + "learning_rate": 1.1164911798286771e-05, + "loss": 1.1846, + "step": 19916 + }, + { + "epoch": 0.48, + "learning_rate": 1.1164138203285027e-05, + "loss": 1.061, + "step": 19917 + }, + { + "epoch": 0.48, + "learning_rate": 1.1163364601220725e-05, + "loss": 1.3141, + "step": 19918 + }, + { + "epoch": 0.48, + "learning_rate": 1.1162590992098568e-05, + "loss": 1.0317, + "step": 19919 + }, + { + "epoch": 0.48, + "learning_rate": 1.1161817375923243e-05, + "loss": 1.0203, + "step": 19920 + }, + { + "epoch": 0.48, + "learning_rate": 1.1161043752699448e-05, + "loss": 1.0441, + "step": 19921 + }, + { + "epoch": 0.48, + "learning_rate": 1.1160270122431869e-05, + "loss": 1.2191, + "step": 19922 + }, + { + "epoch": 0.48, + "learning_rate": 1.1159496485125208e-05, + "loss": 0.937, + "step": 19923 + }, + { + "epoch": 0.48, + "learning_rate": 1.1158722840784154e-05, + "loss": 1.0842, + "step": 19924 + }, + { + "epoch": 0.48, + "learning_rate": 1.11579491894134e-05, + "loss": 1.3172, + "step": 19925 + }, + { + "epoch": 0.48, + "learning_rate": 1.1157175531017639e-05, + "loss": 1.1327, + "step": 19926 + }, + { + "epoch": 0.48, + "learning_rate": 1.115640186560157e-05, + "loss": 1.1711, + "step": 19927 + }, + { + "epoch": 0.48, + "learning_rate": 1.115562819316988e-05, + "loss": 1.0102, + "step": 19928 + }, + { + "epoch": 0.48, + "learning_rate": 1.1154854513727266e-05, + "loss": 1.3188, + "step": 19929 + }, + { + "epoch": 0.48, + "learning_rate": 1.1154080827278422e-05, + "loss": 0.9741, + "step": 19930 + }, + { + "epoch": 0.48, + "learning_rate": 1.1153307133828042e-05, + "loss": 1.0367, + "step": 19931 + }, + { + "epoch": 0.48, + "learning_rate": 1.1152533433380817e-05, + "loss": 1.1822, + "step": 19932 + }, + { + "epoch": 0.48, + "learning_rate": 1.1151759725941444e-05, + "loss": 1.1376, + "step": 19933 + }, + { + "epoch": 0.48, + "learning_rate": 1.1150986011514616e-05, + "loss": 1.2518, + "step": 19934 + }, + { + "epoch": 0.48, + "learning_rate": 1.1150212290105025e-05, + "loss": 1.1828, + "step": 19935 + }, + { + "epoch": 0.48, + "learning_rate": 1.1149438561717367e-05, + "loss": 1.063, + "step": 19936 + }, + { + "epoch": 0.48, + "learning_rate": 1.1148664826356336e-05, + "loss": 1.0407, + "step": 19937 + }, + { + "epoch": 0.48, + "learning_rate": 1.1147891084026625e-05, + "loss": 1.0065, + "step": 19938 + }, + { + "epoch": 0.48, + "learning_rate": 1.1147117334732932e-05, + "loss": 1.1576, + "step": 19939 + }, + { + "epoch": 0.48, + "learning_rate": 1.1146343578479943e-05, + "loss": 0.9687, + "step": 19940 + }, + { + "epoch": 0.48, + "learning_rate": 1.1145569815272362e-05, + "loss": 0.8855, + "step": 19941 + }, + { + "epoch": 0.48, + "learning_rate": 1.1144796045114873e-05, + "loss": 1.1395, + "step": 19942 + }, + { + "epoch": 0.48, + "learning_rate": 1.114402226801218e-05, + "loss": 1.0782, + "step": 19943 + }, + { + "epoch": 0.48, + "learning_rate": 1.1143248483968971e-05, + "loss": 1.1927, + "step": 19944 + }, + { + "epoch": 0.48, + "learning_rate": 1.1142474692989944e-05, + "loss": 1.03, + "step": 19945 + }, + { + "epoch": 0.48, + "learning_rate": 1.114170089507979e-05, + "loss": 1.2391, + "step": 19946 + }, + { + "epoch": 0.48, + "learning_rate": 1.1140927090243206e-05, + "loss": 1.1736, + "step": 19947 + }, + { + "epoch": 0.48, + "learning_rate": 1.1140153278484887e-05, + "loss": 1.0659, + "step": 19948 + }, + { + "epoch": 0.48, + "learning_rate": 1.1139379459809524e-05, + "loss": 1.1261, + "step": 19949 + }, + { + "epoch": 0.48, + "learning_rate": 1.1138605634221815e-05, + "loss": 1.1184, + "step": 19950 + }, + { + "epoch": 0.48, + "learning_rate": 1.1137831801726453e-05, + "loss": 1.1943, + "step": 19951 + }, + { + "epoch": 0.48, + "learning_rate": 1.1137057962328136e-05, + "loss": 0.9019, + "step": 19952 + }, + { + "epoch": 0.48, + "learning_rate": 1.113628411603155e-05, + "loss": 0.8436, + "step": 19953 + }, + { + "epoch": 0.48, + "learning_rate": 1.1135510262841397e-05, + "loss": 1.2176, + "step": 19954 + }, + { + "epoch": 0.48, + "learning_rate": 1.1134736402762375e-05, + "loss": 1.1112, + "step": 19955 + }, + { + "epoch": 0.48, + "learning_rate": 1.113396253579917e-05, + "loss": 1.1446, + "step": 19956 + }, + { + "epoch": 0.48, + "learning_rate": 1.1133188661956483e-05, + "loss": 1.3162, + "step": 19957 + }, + { + "epoch": 0.48, + "learning_rate": 1.1132414781239003e-05, + "loss": 0.8391, + "step": 19958 + }, + { + "epoch": 0.48, + "learning_rate": 1.1131640893651431e-05, + "loss": 1.1225, + "step": 19959 + }, + { + "epoch": 0.48, + "learning_rate": 1.1130866999198461e-05, + "loss": 0.8388, + "step": 19960 + }, + { + "epoch": 0.48, + "learning_rate": 1.1130093097884789e-05, + "loss": 1.2473, + "step": 19961 + }, + { + "epoch": 0.48, + "learning_rate": 1.1129319189715105e-05, + "loss": 0.9134, + "step": 19962 + }, + { + "epoch": 0.48, + "learning_rate": 1.1128545274694106e-05, + "loss": 1.0532, + "step": 19963 + }, + { + "epoch": 0.48, + "learning_rate": 1.1127771352826491e-05, + "loss": 0.9358, + "step": 19964 + }, + { + "epoch": 0.48, + "learning_rate": 1.1126997424116953e-05, + "loss": 1.1029, + "step": 19965 + }, + { + "epoch": 0.48, + "learning_rate": 1.1126223488570183e-05, + "loss": 0.763, + "step": 19966 + }, + { + "epoch": 0.48, + "learning_rate": 1.1125449546190882e-05, + "loss": 0.9798, + "step": 19967 + }, + { + "epoch": 0.48, + "learning_rate": 1.1124675596983744e-05, + "loss": 0.9681, + "step": 19968 + }, + { + "epoch": 0.48, + "learning_rate": 1.1123901640953465e-05, + "loss": 1.1379, + "step": 19969 + }, + { + "epoch": 0.48, + "learning_rate": 1.1123127678104734e-05, + "loss": 1.0064, + "step": 19970 + }, + { + "epoch": 0.48, + "learning_rate": 1.1122353708442257e-05, + "loss": 1.1191, + "step": 19971 + }, + { + "epoch": 0.48, + "learning_rate": 1.1121579731970722e-05, + "loss": 1.3283, + "step": 19972 + }, + { + "epoch": 0.48, + "learning_rate": 1.1120805748694827e-05, + "loss": 1.0139, + "step": 19973 + }, + { + "epoch": 0.48, + "learning_rate": 1.1120031758619265e-05, + "loss": 1.3045, + "step": 19974 + }, + { + "epoch": 0.48, + "learning_rate": 1.1119257761748736e-05, + "loss": 1.0541, + "step": 19975 + }, + { + "epoch": 0.48, + "learning_rate": 1.1118483758087934e-05, + "loss": 0.9936, + "step": 19976 + }, + { + "epoch": 0.48, + "learning_rate": 1.1117709747641554e-05, + "loss": 0.4431, + "step": 19977 + }, + { + "epoch": 0.48, + "learning_rate": 1.1116935730414289e-05, + "loss": 1.1719, + "step": 19978 + }, + { + "epoch": 0.48, + "learning_rate": 1.111616170641084e-05, + "loss": 1.1322, + "step": 19979 + }, + { + "epoch": 0.48, + "learning_rate": 1.11153876756359e-05, + "loss": 1.0421, + "step": 19980 + }, + { + "epoch": 0.48, + "learning_rate": 1.1114613638094167e-05, + "loss": 1.0856, + "step": 19981 + }, + { + "epoch": 0.48, + "learning_rate": 1.1113839593790332e-05, + "loss": 0.8751, + "step": 19982 + }, + { + "epoch": 0.48, + "learning_rate": 1.1113065542729096e-05, + "loss": 0.877, + "step": 19983 + }, + { + "epoch": 0.48, + "learning_rate": 1.1112291484915153e-05, + "loss": 1.2877, + "step": 19984 + }, + { + "epoch": 0.48, + "learning_rate": 1.1111517420353199e-05, + "loss": 1.1706, + "step": 19985 + }, + { + "epoch": 0.48, + "learning_rate": 1.111074334904793e-05, + "loss": 1.0381, + "step": 19986 + }, + { + "epoch": 0.48, + "learning_rate": 1.1109969271004042e-05, + "loss": 0.9909, + "step": 19987 + }, + { + "epoch": 0.48, + "learning_rate": 1.1109195186226234e-05, + "loss": 0.9593, + "step": 19988 + }, + { + "epoch": 0.48, + "learning_rate": 1.1108421094719198e-05, + "loss": 1.0739, + "step": 19989 + }, + { + "epoch": 0.48, + "learning_rate": 1.1107646996487631e-05, + "loss": 0.9847, + "step": 19990 + }, + { + "epoch": 0.48, + "learning_rate": 1.1106872891536232e-05, + "loss": 0.7727, + "step": 19991 + }, + { + "epoch": 0.48, + "learning_rate": 1.1106098779869694e-05, + "loss": 1.0518, + "step": 19992 + }, + { + "epoch": 0.48, + "learning_rate": 1.1105324661492715e-05, + "loss": 1.0067, + "step": 19993 + }, + { + "epoch": 0.48, + "learning_rate": 1.110455053640999e-05, + "loss": 1.1656, + "step": 19994 + }, + { + "epoch": 0.48, + "learning_rate": 1.1103776404626219e-05, + "loss": 0.7733, + "step": 19995 + }, + { + "epoch": 0.48, + "learning_rate": 1.1103002266146096e-05, + "loss": 1.0875, + "step": 19996 + }, + { + "epoch": 0.48, + "learning_rate": 1.1102228120974318e-05, + "loss": 1.1573, + "step": 19997 + }, + { + "epoch": 0.48, + "learning_rate": 1.1101453969115579e-05, + "loss": 1.0995, + "step": 19998 + }, + { + "epoch": 0.48, + "learning_rate": 1.110067981057458e-05, + "loss": 0.9907, + "step": 19999 + }, + { + "epoch": 0.48, + "learning_rate": 1.1099905645356016e-05, + "loss": 0.9759, + "step": 20000 + }, + { + "epoch": 0.48, + "learning_rate": 1.1099131473464582e-05, + "loss": 1.023, + "step": 20001 + }, + { + "epoch": 0.48, + "learning_rate": 1.1098357294904976e-05, + "loss": 1.276, + "step": 20002 + }, + { + "epoch": 0.48, + "learning_rate": 1.1097583109681894e-05, + "loss": 1.0196, + "step": 20003 + }, + { + "epoch": 0.48, + "learning_rate": 1.1096808917800035e-05, + "loss": 1.3398, + "step": 20004 + }, + { + "epoch": 0.48, + "learning_rate": 1.109603471926409e-05, + "loss": 0.8264, + "step": 20005 + }, + { + "epoch": 0.48, + "learning_rate": 1.1095260514078766e-05, + "loss": 1.2063, + "step": 20006 + }, + { + "epoch": 0.48, + "learning_rate": 1.109448630224875e-05, + "loss": 1.2259, + "step": 20007 + }, + { + "epoch": 0.48, + "learning_rate": 1.1093712083778748e-05, + "loss": 1.2893, + "step": 20008 + }, + { + "epoch": 0.48, + "learning_rate": 1.1092937858673445e-05, + "loss": 0.9344, + "step": 20009 + }, + { + "epoch": 0.48, + "learning_rate": 1.1092163626937549e-05, + "loss": 1.0229, + "step": 20010 + }, + { + "epoch": 0.48, + "learning_rate": 1.1091389388575753e-05, + "loss": 1.0517, + "step": 20011 + }, + { + "epoch": 0.48, + "learning_rate": 1.1090615143592755e-05, + "loss": 1.1572, + "step": 20012 + }, + { + "epoch": 0.48, + "learning_rate": 1.1089840891993249e-05, + "loss": 0.9529, + "step": 20013 + }, + { + "epoch": 0.48, + "learning_rate": 1.1089066633781937e-05, + "loss": 0.9604, + "step": 20014 + }, + { + "epoch": 0.48, + "learning_rate": 1.1088292368963513e-05, + "loss": 1.0092, + "step": 20015 + }, + { + "epoch": 0.48, + "learning_rate": 1.1087518097542677e-05, + "loss": 1.1497, + "step": 20016 + }, + { + "epoch": 0.48, + "learning_rate": 1.1086743819524122e-05, + "loss": 1.1653, + "step": 20017 + }, + { + "epoch": 0.48, + "learning_rate": 1.1085969534912548e-05, + "loss": 1.3889, + "step": 20018 + }, + { + "epoch": 0.48, + "learning_rate": 1.1085195243712655e-05, + "loss": 0.9458, + "step": 20019 + }, + { + "epoch": 0.48, + "learning_rate": 1.1084420945929136e-05, + "loss": 1.14, + "step": 20020 + }, + { + "epoch": 0.48, + "learning_rate": 1.1083646641566687e-05, + "loss": 1.1402, + "step": 20021 + }, + { + "epoch": 0.48, + "learning_rate": 1.1082872330630013e-05, + "loss": 1.1968, + "step": 20022 + }, + { + "epoch": 0.48, + "learning_rate": 1.1082098013123806e-05, + "loss": 1.2157, + "step": 20023 + }, + { + "epoch": 0.48, + "learning_rate": 1.1081323689052766e-05, + "loss": 1.09, + "step": 20024 + }, + { + "epoch": 0.48, + "learning_rate": 1.1080549358421586e-05, + "loss": 1.0202, + "step": 20025 + }, + { + "epoch": 0.48, + "learning_rate": 1.1079775021234969e-05, + "loss": 1.0603, + "step": 20026 + }, + { + "epoch": 0.48, + "learning_rate": 1.1079000677497614e-05, + "loss": 1.1216, + "step": 20027 + }, + { + "epoch": 0.48, + "learning_rate": 1.1078226327214212e-05, + "loss": 1.1799, + "step": 20028 + }, + { + "epoch": 0.48, + "learning_rate": 1.1077451970389463e-05, + "loss": 1.3994, + "step": 20029 + }, + { + "epoch": 0.48, + "learning_rate": 1.107667760702807e-05, + "loss": 1.0948, + "step": 20030 + }, + { + "epoch": 0.48, + "learning_rate": 1.1075903237134726e-05, + "loss": 1.1915, + "step": 20031 + }, + { + "epoch": 0.48, + "learning_rate": 1.1075128860714132e-05, + "loss": 0.8828, + "step": 20032 + }, + { + "epoch": 0.48, + "learning_rate": 1.1074354477770981e-05, + "loss": 0.8959, + "step": 20033 + }, + { + "epoch": 0.48, + "learning_rate": 1.1073580088309977e-05, + "loss": 1.1783, + "step": 20034 + }, + { + "epoch": 0.48, + "learning_rate": 1.1072805692335812e-05, + "loss": 0.939, + "step": 20035 + }, + { + "epoch": 0.48, + "learning_rate": 1.1072031289853189e-05, + "loss": 0.9851, + "step": 20036 + }, + { + "epoch": 0.48, + "learning_rate": 1.1071256880866802e-05, + "loss": 1.0641, + "step": 20037 + }, + { + "epoch": 0.48, + "learning_rate": 1.1070482465381356e-05, + "loss": 0.9343, + "step": 20038 + }, + { + "epoch": 0.48, + "learning_rate": 1.1069708043401542e-05, + "loss": 1.1967, + "step": 20039 + }, + { + "epoch": 0.48, + "learning_rate": 1.1068933614932061e-05, + "loss": 0.9673, + "step": 20040 + }, + { + "epoch": 0.48, + "learning_rate": 1.1068159179977611e-05, + "loss": 1.0311, + "step": 20041 + }, + { + "epoch": 0.48, + "learning_rate": 1.1067384738542888e-05, + "loss": 1.265, + "step": 20042 + }, + { + "epoch": 0.48, + "learning_rate": 1.1066610290632597e-05, + "loss": 1.0529, + "step": 20043 + }, + { + "epoch": 0.48, + "learning_rate": 1.106583583625143e-05, + "loss": 1.0407, + "step": 20044 + }, + { + "epoch": 0.48, + "learning_rate": 1.106506137540409e-05, + "loss": 1.1284, + "step": 20045 + }, + { + "epoch": 0.48, + "learning_rate": 1.106428690809527e-05, + "loss": 1.2595, + "step": 20046 + }, + { + "epoch": 0.48, + "learning_rate": 1.1063512434329675e-05, + "loss": 1.0309, + "step": 20047 + }, + { + "epoch": 0.48, + "learning_rate": 1.1062737954112e-05, + "loss": 0.9588, + "step": 20048 + }, + { + "epoch": 0.48, + "learning_rate": 1.1061963467446938e-05, + "loss": 1.2505, + "step": 20049 + }, + { + "epoch": 0.48, + "learning_rate": 1.1061188974339201e-05, + "loss": 1.3559, + "step": 20050 + }, + { + "epoch": 0.48, + "learning_rate": 1.1060414474793477e-05, + "loss": 0.9339, + "step": 20051 + }, + { + "epoch": 0.48, + "learning_rate": 1.1059639968814469e-05, + "loss": 1.2921, + "step": 20052 + }, + { + "epoch": 0.48, + "learning_rate": 1.105886545640687e-05, + "loss": 1.0977, + "step": 20053 + }, + { + "epoch": 0.48, + "learning_rate": 1.1058090937575388e-05, + "loss": 1.1795, + "step": 20054 + }, + { + "epoch": 0.48, + "learning_rate": 1.1057316412324718e-05, + "loss": 1.0634, + "step": 20055 + }, + { + "epoch": 0.48, + "learning_rate": 1.1056541880659554e-05, + "loss": 0.894, + "step": 20056 + }, + { + "epoch": 0.48, + "learning_rate": 1.10557673425846e-05, + "loss": 1.3524, + "step": 20057 + }, + { + "epoch": 0.48, + "learning_rate": 1.1054992798104553e-05, + "loss": 1.1781, + "step": 20058 + }, + { + "epoch": 0.48, + "learning_rate": 1.1054218247224115e-05, + "loss": 0.9996, + "step": 20059 + }, + { + "epoch": 0.48, + "learning_rate": 1.1053443689947984e-05, + "loss": 1.1029, + "step": 20060 + }, + { + "epoch": 0.48, + "learning_rate": 1.1052669126280854e-05, + "loss": 1.0824, + "step": 20061 + }, + { + "epoch": 0.48, + "learning_rate": 1.1051894556227431e-05, + "loss": 0.8178, + "step": 20062 + }, + { + "epoch": 0.48, + "learning_rate": 1.1051119979792411e-05, + "loss": 1.0066, + "step": 20063 + }, + { + "epoch": 0.48, + "learning_rate": 1.1050345396980493e-05, + "loss": 1.2054, + "step": 20064 + }, + { + "epoch": 0.48, + "learning_rate": 1.1049570807796375e-05, + "loss": 0.9703, + "step": 20065 + }, + { + "epoch": 0.48, + "learning_rate": 1.1048796212244761e-05, + "loss": 0.7455, + "step": 20066 + }, + { + "epoch": 0.48, + "learning_rate": 1.1048021610330345e-05, + "loss": 0.7636, + "step": 20067 + }, + { + "epoch": 0.48, + "learning_rate": 1.1047247002057828e-05, + "loss": 1.1597, + "step": 20068 + }, + { + "epoch": 0.48, + "learning_rate": 1.104647238743191e-05, + "loss": 1.1572, + "step": 20069 + }, + { + "epoch": 0.48, + "learning_rate": 1.104569776645729e-05, + "loss": 0.8791, + "step": 20070 + }, + { + "epoch": 0.48, + "learning_rate": 1.104492313913867e-05, + "loss": 1.0866, + "step": 20071 + }, + { + "epoch": 0.48, + "learning_rate": 1.1044148505480744e-05, + "loss": 1.1081, + "step": 20072 + }, + { + "epoch": 0.48, + "learning_rate": 1.1043373865488216e-05, + "loss": 0.9285, + "step": 20073 + }, + { + "epoch": 0.48, + "learning_rate": 1.1042599219165784e-05, + "loss": 0.8862, + "step": 20074 + }, + { + "epoch": 0.48, + "learning_rate": 1.1041824566518146e-05, + "loss": 1.0733, + "step": 20075 + }, + { + "epoch": 0.48, + "learning_rate": 1.1041049907550008e-05, + "loss": 1.1801, + "step": 20076 + }, + { + "epoch": 0.48, + "learning_rate": 1.1040275242266061e-05, + "loss": 1.3299, + "step": 20077 + }, + { + "epoch": 0.48, + "learning_rate": 1.103950057067101e-05, + "loss": 1.3307, + "step": 20078 + }, + { + "epoch": 0.48, + "learning_rate": 1.1038725892769553e-05, + "loss": 1.0732, + "step": 20079 + }, + { + "epoch": 0.48, + "learning_rate": 1.1037951208566392e-05, + "loss": 1.0343, + "step": 20080 + }, + { + "epoch": 0.48, + "learning_rate": 1.1037176518066222e-05, + "loss": 1.0846, + "step": 20081 + }, + { + "epoch": 0.48, + "learning_rate": 1.103640182127375e-05, + "loss": 1.2644, + "step": 20082 + }, + { + "epoch": 0.48, + "learning_rate": 1.103562711819367e-05, + "loss": 0.8519, + "step": 20083 + }, + { + "epoch": 0.48, + "learning_rate": 1.1034852408830683e-05, + "loss": 1.2198, + "step": 20084 + }, + { + "epoch": 0.48, + "learning_rate": 1.103407769318949e-05, + "loss": 1.0576, + "step": 20085 + }, + { + "epoch": 0.48, + "learning_rate": 1.1033302971274793e-05, + "loss": 1.0794, + "step": 20086 + }, + { + "epoch": 0.48, + "learning_rate": 1.1032528243091288e-05, + "loss": 1.0124, + "step": 20087 + }, + { + "epoch": 0.48, + "learning_rate": 1.103175350864368e-05, + "loss": 0.9965, + "step": 20088 + }, + { + "epoch": 0.48, + "learning_rate": 1.103097876793666e-05, + "loss": 1.0154, + "step": 20089 + }, + { + "epoch": 0.48, + "learning_rate": 1.1030204020974939e-05, + "loss": 0.7159, + "step": 20090 + }, + { + "epoch": 0.48, + "learning_rate": 1.1029429267763212e-05, + "loss": 0.9704, + "step": 20091 + }, + { + "epoch": 0.48, + "learning_rate": 1.1028654508306178e-05, + "loss": 1.2316, + "step": 20092 + }, + { + "epoch": 0.48, + "learning_rate": 1.1027879742608541e-05, + "loss": 1.0525, + "step": 20093 + }, + { + "epoch": 0.48, + "learning_rate": 1.1027104970674998e-05, + "loss": 1.1331, + "step": 20094 + }, + { + "epoch": 0.48, + "learning_rate": 1.1026330192510254e-05, + "loss": 1.2916, + "step": 20095 + }, + { + "epoch": 0.48, + "learning_rate": 1.1025555408119e-05, + "loss": 1.0817, + "step": 20096 + }, + { + "epoch": 0.48, + "learning_rate": 1.1024780617505948e-05, + "loss": 1.1809, + "step": 20097 + }, + { + "epoch": 0.48, + "learning_rate": 1.1024005820675791e-05, + "loss": 1.2925, + "step": 20098 + }, + { + "epoch": 0.48, + "learning_rate": 1.1023231017633232e-05, + "loss": 1.0919, + "step": 20099 + }, + { + "epoch": 0.48, + "learning_rate": 1.102245620838297e-05, + "loss": 0.738, + "step": 20100 + }, + { + "epoch": 0.48, + "learning_rate": 1.1021681392929707e-05, + "loss": 0.903, + "step": 20101 + }, + { + "epoch": 0.48, + "learning_rate": 1.1020906571278145e-05, + "loss": 1.1586, + "step": 20102 + }, + { + "epoch": 0.48, + "learning_rate": 1.1020131743432982e-05, + "loss": 1.0946, + "step": 20103 + }, + { + "epoch": 0.48, + "learning_rate": 1.1019356909398917e-05, + "loss": 1.3035, + "step": 20104 + }, + { + "epoch": 0.48, + "learning_rate": 1.1018582069180656e-05, + "loss": 0.9751, + "step": 20105 + }, + { + "epoch": 0.48, + "learning_rate": 1.10178072227829e-05, + "loss": 1.2484, + "step": 20106 + }, + { + "epoch": 0.48, + "learning_rate": 1.1017032370210343e-05, + "loss": 1.1657, + "step": 20107 + }, + { + "epoch": 0.48, + "learning_rate": 1.101625751146769e-05, + "loss": 0.999, + "step": 20108 + }, + { + "epoch": 0.48, + "learning_rate": 1.1015482646559641e-05, + "loss": 0.8612, + "step": 20109 + }, + { + "epoch": 0.48, + "learning_rate": 1.10147077754909e-05, + "loss": 1.0958, + "step": 20110 + }, + { + "epoch": 0.48, + "learning_rate": 1.1013932898266166e-05, + "loss": 1.1433, + "step": 20111 + }, + { + "epoch": 0.48, + "learning_rate": 1.1013158014890138e-05, + "loss": 1.2257, + "step": 20112 + }, + { + "epoch": 0.48, + "learning_rate": 1.1012383125367519e-05, + "loss": 1.0181, + "step": 20113 + }, + { + "epoch": 0.48, + "learning_rate": 1.1011608229703011e-05, + "loss": 1.195, + "step": 20114 + }, + { + "epoch": 0.48, + "learning_rate": 1.101083332790131e-05, + "loss": 1.0001, + "step": 20115 + }, + { + "epoch": 0.48, + "learning_rate": 1.1010058419967123e-05, + "loss": 0.8662, + "step": 20116 + }, + { + "epoch": 0.48, + "learning_rate": 1.100928350590515e-05, + "loss": 1.016, + "step": 20117 + }, + { + "epoch": 0.48, + "learning_rate": 1.1008508585720093e-05, + "loss": 1.1474, + "step": 20118 + }, + { + "epoch": 0.48, + "learning_rate": 1.1007733659416649e-05, + "loss": 1.3305, + "step": 20119 + }, + { + "epoch": 0.48, + "learning_rate": 1.1006958726999522e-05, + "loss": 1.1838, + "step": 20120 + }, + { + "epoch": 0.48, + "learning_rate": 1.1006183788473412e-05, + "loss": 1.1033, + "step": 20121 + }, + { + "epoch": 0.48, + "learning_rate": 1.1005408843843026e-05, + "loss": 0.9119, + "step": 20122 + }, + { + "epoch": 0.48, + "learning_rate": 1.1004633893113059e-05, + "loss": 1.154, + "step": 20123 + }, + { + "epoch": 0.48, + "learning_rate": 1.1003858936288212e-05, + "loss": 0.9483, + "step": 20124 + }, + { + "epoch": 0.48, + "learning_rate": 1.1003083973373188e-05, + "loss": 1.0675, + "step": 20125 + }, + { + "epoch": 0.48, + "learning_rate": 1.1002309004372693e-05, + "loss": 1.2217, + "step": 20126 + }, + { + "epoch": 0.48, + "learning_rate": 1.1001534029291425e-05, + "loss": 1.1301, + "step": 20127 + }, + { + "epoch": 0.48, + "learning_rate": 1.1000759048134085e-05, + "loss": 1.266, + "step": 20128 + }, + { + "epoch": 0.48, + "learning_rate": 1.0999984060905375e-05, + "loss": 1.0507, + "step": 20129 + }, + { + "epoch": 0.48, + "learning_rate": 1.0999209067609998e-05, + "loss": 0.8863, + "step": 20130 + }, + { + "epoch": 0.48, + "learning_rate": 1.0998434068252652e-05, + "loss": 0.906, + "step": 20131 + }, + { + "epoch": 0.48, + "learning_rate": 1.0997659062838041e-05, + "loss": 1.0362, + "step": 20132 + }, + { + "epoch": 0.48, + "learning_rate": 1.099688405137087e-05, + "loss": 1.0483, + "step": 20133 + }, + { + "epoch": 0.48, + "learning_rate": 1.099610903385584e-05, + "loss": 1.2704, + "step": 20134 + }, + { + "epoch": 0.48, + "learning_rate": 1.0995334010297646e-05, + "loss": 1.086, + "step": 20135 + }, + { + "epoch": 0.48, + "learning_rate": 1.0994558980700997e-05, + "loss": 1.2473, + "step": 20136 + }, + { + "epoch": 0.48, + "learning_rate": 1.0993783945070594e-05, + "loss": 1.1102, + "step": 20137 + }, + { + "epoch": 0.48, + "learning_rate": 1.0993008903411135e-05, + "loss": 1.0328, + "step": 20138 + }, + { + "epoch": 0.48, + "learning_rate": 1.0992233855727325e-05, + "loss": 1.2706, + "step": 20139 + }, + { + "epoch": 0.48, + "learning_rate": 1.0991458802023864e-05, + "loss": 1.2673, + "step": 20140 + }, + { + "epoch": 0.48, + "learning_rate": 1.099068374230546e-05, + "loss": 0.973, + "step": 20141 + }, + { + "epoch": 0.48, + "learning_rate": 1.0989908676576808e-05, + "loss": 1.1983, + "step": 20142 + }, + { + "epoch": 0.48, + "learning_rate": 1.0989133604842614e-05, + "loss": 1.212, + "step": 20143 + }, + { + "epoch": 0.48, + "learning_rate": 1.0988358527107576e-05, + "loss": 1.1995, + "step": 20144 + }, + { + "epoch": 0.48, + "learning_rate": 1.0987583443376404e-05, + "loss": 1.0195, + "step": 20145 + }, + { + "epoch": 0.48, + "learning_rate": 1.0986808353653792e-05, + "loss": 0.99, + "step": 20146 + }, + { + "epoch": 0.48, + "learning_rate": 1.0986033257944448e-05, + "loss": 1.2044, + "step": 20147 + }, + { + "epoch": 0.48, + "learning_rate": 1.0985258156253069e-05, + "loss": 0.9886, + "step": 20148 + }, + { + "epoch": 0.48, + "learning_rate": 1.0984483048584365e-05, + "loss": 1.0944, + "step": 20149 + }, + { + "epoch": 0.48, + "learning_rate": 1.0983707934943031e-05, + "loss": 1.339, + "step": 20150 + }, + { + "epoch": 0.48, + "learning_rate": 1.0982932815333773e-05, + "loss": 1.217, + "step": 20151 + }, + { + "epoch": 0.48, + "learning_rate": 1.0982157689761293e-05, + "loss": 1.1642, + "step": 20152 + }, + { + "epoch": 0.48, + "learning_rate": 1.0981382558230291e-05, + "loss": 1.0622, + "step": 20153 + }, + { + "epoch": 0.48, + "learning_rate": 1.0980607420745475e-05, + "loss": 1.0891, + "step": 20154 + }, + { + "epoch": 0.48, + "learning_rate": 1.0979832277311543e-05, + "loss": 1.1023, + "step": 20155 + }, + { + "epoch": 0.48, + "learning_rate": 1.0979057127933199e-05, + "loss": 0.9679, + "step": 20156 + }, + { + "epoch": 0.48, + "learning_rate": 1.0978281972615146e-05, + "loss": 1.3794, + "step": 20157 + }, + { + "epoch": 0.48, + "learning_rate": 1.0977506811362089e-05, + "loss": 0.8829, + "step": 20158 + }, + { + "epoch": 0.48, + "learning_rate": 1.0976731644178723e-05, + "loss": 1.2328, + "step": 20159 + }, + { + "epoch": 0.48, + "learning_rate": 1.0975956471069758e-05, + "loss": 1.0843, + "step": 20160 + }, + { + "epoch": 0.48, + "learning_rate": 1.0975181292039895e-05, + "loss": 1.1088, + "step": 20161 + }, + { + "epoch": 0.48, + "learning_rate": 1.0974406107093837e-05, + "loss": 1.179, + "step": 20162 + }, + { + "epoch": 0.48, + "learning_rate": 1.097363091623629e-05, + "loss": 1.0085, + "step": 20163 + }, + { + "epoch": 0.48, + "learning_rate": 1.0972855719471943e-05, + "loss": 1.1258, + "step": 20164 + }, + { + "epoch": 0.48, + "learning_rate": 1.097208051680552e-05, + "loss": 1.1424, + "step": 20165 + }, + { + "epoch": 0.48, + "learning_rate": 1.0971305308241706e-05, + "loss": 1.3302, + "step": 20166 + }, + { + "epoch": 0.48, + "learning_rate": 1.0970530093785214e-05, + "loss": 1.284, + "step": 20167 + }, + { + "epoch": 0.49, + "learning_rate": 1.0969754873440743e-05, + "loss": 1.0378, + "step": 20168 + }, + { + "epoch": 0.49, + "learning_rate": 1.0968979647212998e-05, + "loss": 1.0578, + "step": 20169 + }, + { + "epoch": 0.49, + "learning_rate": 1.0968204415106682e-05, + "loss": 1.1514, + "step": 20170 + }, + { + "epoch": 0.49, + "learning_rate": 1.0967429177126499e-05, + "loss": 0.8276, + "step": 20171 + }, + { + "epoch": 0.49, + "learning_rate": 1.0966653933277147e-05, + "loss": 1.1968, + "step": 20172 + }, + { + "epoch": 0.49, + "learning_rate": 1.0965878683563336e-05, + "loss": 1.182, + "step": 20173 + }, + { + "epoch": 0.49, + "learning_rate": 1.0965103427989766e-05, + "loss": 1.1703, + "step": 20174 + }, + { + "epoch": 0.49, + "learning_rate": 1.0964328166561136e-05, + "loss": 1.0222, + "step": 20175 + }, + { + "epoch": 0.49, + "learning_rate": 1.0963552899282158e-05, + "loss": 1.1019, + "step": 20176 + }, + { + "epoch": 0.49, + "learning_rate": 1.096277762615753e-05, + "loss": 1.0417, + "step": 20177 + }, + { + "epoch": 0.49, + "learning_rate": 1.0962002347191958e-05, + "loss": 1.0089, + "step": 20178 + }, + { + "epoch": 0.49, + "learning_rate": 1.0961227062390143e-05, + "loss": 1.2052, + "step": 20179 + }, + { + "epoch": 0.49, + "learning_rate": 1.0960451771756788e-05, + "loss": 0.983, + "step": 20180 + }, + { + "epoch": 0.49, + "learning_rate": 1.0959676475296601e-05, + "loss": 0.7141, + "step": 20181 + }, + { + "epoch": 0.49, + "learning_rate": 1.0958901173014283e-05, + "loss": 1.062, + "step": 20182 + }, + { + "epoch": 0.49, + "learning_rate": 1.0958125864914532e-05, + "loss": 1.0911, + "step": 20183 + }, + { + "epoch": 0.49, + "learning_rate": 1.0957350551002062e-05, + "loss": 1.017, + "step": 20184 + }, + { + "epoch": 0.49, + "learning_rate": 1.095657523128157e-05, + "loss": 0.9874, + "step": 20185 + }, + { + "epoch": 0.49, + "learning_rate": 1.095579990575776e-05, + "loss": 1.3526, + "step": 20186 + }, + { + "epoch": 0.49, + "learning_rate": 1.0955024574435336e-05, + "loss": 1.1107, + "step": 20187 + }, + { + "epoch": 0.49, + "learning_rate": 1.0954249237319003e-05, + "loss": 1.334, + "step": 20188 + }, + { + "epoch": 0.49, + "learning_rate": 1.0953473894413467e-05, + "loss": 1.3782, + "step": 20189 + }, + { + "epoch": 0.49, + "learning_rate": 1.0952698545723428e-05, + "loss": 1.1897, + "step": 20190 + }, + { + "epoch": 0.49, + "learning_rate": 1.0951923191253589e-05, + "loss": 0.8956, + "step": 20191 + }, + { + "epoch": 0.49, + "learning_rate": 1.0951147831008657e-05, + "loss": 1.0399, + "step": 20192 + }, + { + "epoch": 0.49, + "learning_rate": 1.0950372464993335e-05, + "loss": 1.0198, + "step": 20193 + }, + { + "epoch": 0.49, + "learning_rate": 1.094959709321233e-05, + "loss": 0.9987, + "step": 20194 + }, + { + "epoch": 0.49, + "learning_rate": 1.0948821715670336e-05, + "loss": 1.1058, + "step": 20195 + }, + { + "epoch": 0.49, + "learning_rate": 1.094804633237207e-05, + "loss": 0.9899, + "step": 20196 + }, + { + "epoch": 0.49, + "learning_rate": 1.0947270943322229e-05, + "loss": 1.1435, + "step": 20197 + }, + { + "epoch": 0.49, + "learning_rate": 1.0946495548525515e-05, + "loss": 1.2839, + "step": 20198 + }, + { + "epoch": 0.49, + "learning_rate": 1.0945720147986635e-05, + "loss": 1.1733, + "step": 20199 + }, + { + "epoch": 0.49, + "learning_rate": 1.0944944741710295e-05, + "loss": 1.1832, + "step": 20200 + }, + { + "epoch": 0.49, + "learning_rate": 1.0944169329701198e-05, + "loss": 1.0072, + "step": 20201 + }, + { + "epoch": 0.49, + "learning_rate": 1.0943393911964049e-05, + "loss": 0.7028, + "step": 20202 + }, + { + "epoch": 0.49, + "learning_rate": 1.0942618488503547e-05, + "loss": 1.0543, + "step": 20203 + }, + { + "epoch": 0.49, + "learning_rate": 1.0941843059324402e-05, + "loss": 1.0959, + "step": 20204 + }, + { + "epoch": 0.49, + "learning_rate": 1.0941067624431319e-05, + "loss": 1.1755, + "step": 20205 + }, + { + "epoch": 0.49, + "learning_rate": 1.0940292183828998e-05, + "loss": 1.2632, + "step": 20206 + }, + { + "epoch": 0.49, + "learning_rate": 1.0939516737522144e-05, + "loss": 0.9737, + "step": 20207 + }, + { + "epoch": 0.49, + "learning_rate": 1.0938741285515463e-05, + "loss": 1.115, + "step": 20208 + }, + { + "epoch": 0.49, + "learning_rate": 1.0937965827813661e-05, + "loss": 0.8948, + "step": 20209 + }, + { + "epoch": 0.49, + "learning_rate": 1.0937190364421442e-05, + "loss": 1.2473, + "step": 20210 + }, + { + "epoch": 0.49, + "learning_rate": 1.093641489534351e-05, + "loss": 0.9834, + "step": 20211 + }, + { + "epoch": 0.49, + "learning_rate": 1.0935639420584565e-05, + "loss": 0.7308, + "step": 20212 + }, + { + "epoch": 0.49, + "learning_rate": 1.0934863940149319e-05, + "loss": 1.0666, + "step": 20213 + }, + { + "epoch": 0.49, + "learning_rate": 1.0934088454042474e-05, + "loss": 1.2805, + "step": 20214 + }, + { + "epoch": 0.49, + "learning_rate": 1.0933312962268732e-05, + "loss": 0.9635, + "step": 20215 + }, + { + "epoch": 0.49, + "learning_rate": 1.09325374648328e-05, + "loss": 1.0857, + "step": 20216 + }, + { + "epoch": 0.49, + "learning_rate": 1.0931761961739385e-05, + "loss": 1.0427, + "step": 20217 + }, + { + "epoch": 0.49, + "learning_rate": 1.0930986452993187e-05, + "loss": 1.1179, + "step": 20218 + }, + { + "epoch": 0.49, + "learning_rate": 1.0930210938598911e-05, + "loss": 1.0186, + "step": 20219 + }, + { + "epoch": 0.49, + "learning_rate": 1.0929435418561268e-05, + "loss": 1.2175, + "step": 20220 + }, + { + "epoch": 0.49, + "learning_rate": 1.0928659892884959e-05, + "loss": 1.1516, + "step": 20221 + }, + { + "epoch": 0.49, + "learning_rate": 1.0927884361574688e-05, + "loss": 0.9254, + "step": 20222 + }, + { + "epoch": 0.49, + "learning_rate": 1.0927108824635158e-05, + "loss": 0.933, + "step": 20223 + }, + { + "epoch": 0.49, + "learning_rate": 1.092633328207108e-05, + "loss": 0.954, + "step": 20224 + }, + { + "epoch": 0.49, + "learning_rate": 1.0925557733887155e-05, + "loss": 0.7968, + "step": 20225 + }, + { + "epoch": 0.49, + "learning_rate": 1.092478218008809e-05, + "loss": 1.4009, + "step": 20226 + }, + { + "epoch": 0.49, + "learning_rate": 1.0924006620678587e-05, + "loss": 0.968, + "step": 20227 + }, + { + "epoch": 0.49, + "learning_rate": 1.0923231055663352e-05, + "loss": 1.1471, + "step": 20228 + }, + { + "epoch": 0.49, + "learning_rate": 1.0922455485047096e-05, + "loss": 1.2377, + "step": 20229 + }, + { + "epoch": 0.49, + "learning_rate": 1.0921679908834518e-05, + "loss": 0.9789, + "step": 20230 + }, + { + "epoch": 0.49, + "learning_rate": 1.0920904327030322e-05, + "loss": 1.0324, + "step": 20231 + }, + { + "epoch": 0.49, + "learning_rate": 1.0920128739639218e-05, + "loss": 1.2634, + "step": 20232 + }, + { + "epoch": 0.49, + "learning_rate": 1.091935314666591e-05, + "loss": 0.7718, + "step": 20233 + }, + { + "epoch": 0.49, + "learning_rate": 1.0918577548115101e-05, + "loss": 1.1862, + "step": 20234 + }, + { + "epoch": 0.49, + "learning_rate": 1.09178019439915e-05, + "loss": 0.7455, + "step": 20235 + }, + { + "epoch": 0.49, + "learning_rate": 1.091702633429981e-05, + "loss": 0.9183, + "step": 20236 + }, + { + "epoch": 0.49, + "learning_rate": 1.0916250719044737e-05, + "loss": 0.9639, + "step": 20237 + }, + { + "epoch": 0.49, + "learning_rate": 1.0915475098230986e-05, + "loss": 1.2853, + "step": 20238 + }, + { + "epoch": 0.49, + "learning_rate": 1.0914699471863259e-05, + "loss": 1.1882, + "step": 20239 + }, + { + "epoch": 0.49, + "learning_rate": 1.091392383994627e-05, + "loss": 0.9976, + "step": 20240 + }, + { + "epoch": 0.49, + "learning_rate": 1.091314820248472e-05, + "loss": 1.0954, + "step": 20241 + }, + { + "epoch": 0.49, + "learning_rate": 1.0912372559483314e-05, + "loss": 0.8188, + "step": 20242 + }, + { + "epoch": 0.49, + "learning_rate": 1.0911596910946758e-05, + "loss": 1.0697, + "step": 20243 + }, + { + "epoch": 0.49, + "learning_rate": 1.0910821256879758e-05, + "loss": 0.951, + "step": 20244 + }, + { + "epoch": 0.49, + "learning_rate": 1.091004559728702e-05, + "loss": 1.1208, + "step": 20245 + }, + { + "epoch": 0.49, + "learning_rate": 1.090926993217325e-05, + "loss": 1.3223, + "step": 20246 + }, + { + "epoch": 0.49, + "learning_rate": 1.0908494261543151e-05, + "loss": 1.0356, + "step": 20247 + }, + { + "epoch": 0.49, + "learning_rate": 1.0907718585401432e-05, + "loss": 1.0862, + "step": 20248 + }, + { + "epoch": 0.49, + "learning_rate": 1.09069429037528e-05, + "loss": 0.9876, + "step": 20249 + }, + { + "epoch": 0.49, + "learning_rate": 1.0906167216601958e-05, + "loss": 0.9676, + "step": 20250 + }, + { + "epoch": 0.49, + "learning_rate": 1.090539152395361e-05, + "loss": 1.2516, + "step": 20251 + }, + { + "epoch": 0.49, + "learning_rate": 1.0904615825812466e-05, + "loss": 1.1601, + "step": 20252 + }, + { + "epoch": 0.49, + "learning_rate": 1.0903840122183232e-05, + "loss": 1.0548, + "step": 20253 + }, + { + "epoch": 0.49, + "learning_rate": 1.0903064413070611e-05, + "loss": 1.1238, + "step": 20254 + }, + { + "epoch": 0.49, + "learning_rate": 1.090228869847931e-05, + "loss": 1.0014, + "step": 20255 + }, + { + "epoch": 0.49, + "learning_rate": 1.0901512978414037e-05, + "loss": 1.1011, + "step": 20256 + }, + { + "epoch": 0.49, + "learning_rate": 1.0900737252879496e-05, + "loss": 1.2929, + "step": 20257 + }, + { + "epoch": 0.49, + "learning_rate": 1.0899961521880396e-05, + "loss": 1.1802, + "step": 20258 + }, + { + "epoch": 0.49, + "learning_rate": 1.0899185785421437e-05, + "loss": 1.0604, + "step": 20259 + }, + { + "epoch": 0.49, + "learning_rate": 1.0898410043507335e-05, + "loss": 0.931, + "step": 20260 + }, + { + "epoch": 0.49, + "learning_rate": 1.0897634296142786e-05, + "loss": 1.1441, + "step": 20261 + }, + { + "epoch": 0.49, + "learning_rate": 1.0896858543332502e-05, + "loss": 0.9835, + "step": 20262 + }, + { + "epoch": 0.49, + "learning_rate": 1.0896082785081188e-05, + "loss": 1.1668, + "step": 20263 + }, + { + "epoch": 0.49, + "learning_rate": 1.0895307021393553e-05, + "loss": 1.0179, + "step": 20264 + }, + { + "epoch": 0.49, + "learning_rate": 1.0894531252274298e-05, + "loss": 1.1732, + "step": 20265 + }, + { + "epoch": 0.49, + "learning_rate": 1.0893755477728131e-05, + "loss": 1.2158, + "step": 20266 + }, + { + "epoch": 0.49, + "learning_rate": 1.0892979697759765e-05, + "loss": 0.9783, + "step": 20267 + }, + { + "epoch": 0.49, + "learning_rate": 1.0892203912373898e-05, + "loss": 1.1737, + "step": 20268 + }, + { + "epoch": 0.49, + "learning_rate": 1.089142812157524e-05, + "loss": 1.0778, + "step": 20269 + }, + { + "epoch": 0.49, + "learning_rate": 1.0890652325368495e-05, + "loss": 1.1483, + "step": 20270 + }, + { + "epoch": 0.49, + "learning_rate": 1.0889876523758376e-05, + "loss": 0.9682, + "step": 20271 + }, + { + "epoch": 0.49, + "learning_rate": 1.0889100716749584e-05, + "loss": 1.5236, + "step": 20272 + }, + { + "epoch": 0.49, + "learning_rate": 1.0888324904346825e-05, + "loss": 0.9921, + "step": 20273 + }, + { + "epoch": 0.49, + "learning_rate": 1.088754908655481e-05, + "loss": 1.1747, + "step": 20274 + }, + { + "epoch": 0.49, + "learning_rate": 1.0886773263378242e-05, + "loss": 1.147, + "step": 20275 + }, + { + "epoch": 0.49, + "learning_rate": 1.0885997434821831e-05, + "loss": 1.2907, + "step": 20276 + }, + { + "epoch": 0.49, + "learning_rate": 1.0885221600890282e-05, + "loss": 0.9679, + "step": 20277 + }, + { + "epoch": 0.49, + "learning_rate": 1.0884445761588299e-05, + "loss": 1.0891, + "step": 20278 + }, + { + "epoch": 0.49, + "learning_rate": 1.0883669916920595e-05, + "loss": 0.9946, + "step": 20279 + }, + { + "epoch": 0.49, + "learning_rate": 1.088289406689187e-05, + "loss": 0.8828, + "step": 20280 + }, + { + "epoch": 0.49, + "learning_rate": 1.0882118211506835e-05, + "loss": 1.0324, + "step": 20281 + }, + { + "epoch": 0.49, + "learning_rate": 1.0881342350770196e-05, + "loss": 0.9482, + "step": 20282 + }, + { + "epoch": 0.49, + "learning_rate": 1.0880566484686663e-05, + "loss": 0.8027, + "step": 20283 + }, + { + "epoch": 0.49, + "learning_rate": 1.087979061326094e-05, + "loss": 1.1244, + "step": 20284 + }, + { + "epoch": 0.49, + "learning_rate": 1.0879014736497735e-05, + "loss": 1.024, + "step": 20285 + }, + { + "epoch": 0.49, + "learning_rate": 1.087823885440175e-05, + "loss": 1.0155, + "step": 20286 + }, + { + "epoch": 0.49, + "learning_rate": 1.0877462966977698e-05, + "loss": 1.4319, + "step": 20287 + }, + { + "epoch": 0.49, + "learning_rate": 1.0876687074230288e-05, + "loss": 1.1041, + "step": 20288 + }, + { + "epoch": 0.49, + "learning_rate": 1.0875911176164222e-05, + "loss": 1.2631, + "step": 20289 + }, + { + "epoch": 0.49, + "learning_rate": 1.0875135272784205e-05, + "loss": 1.1511, + "step": 20290 + }, + { + "epoch": 0.49, + "learning_rate": 1.087435936409495e-05, + "loss": 1.1168, + "step": 20291 + }, + { + "epoch": 0.49, + "learning_rate": 1.0873583450101166e-05, + "loss": 1.1269, + "step": 20292 + }, + { + "epoch": 0.49, + "learning_rate": 1.0872807530807555e-05, + "loss": 1.1235, + "step": 20293 + }, + { + "epoch": 0.49, + "learning_rate": 1.0872031606218824e-05, + "loss": 1.5016, + "step": 20294 + }, + { + "epoch": 0.49, + "learning_rate": 1.0871255676339683e-05, + "loss": 1.3685, + "step": 20295 + }, + { + "epoch": 0.49, + "learning_rate": 1.087047974117484e-05, + "loss": 1.0745, + "step": 20296 + }, + { + "epoch": 0.49, + "learning_rate": 1.0869703800729002e-05, + "loss": 1.0198, + "step": 20297 + }, + { + "epoch": 0.49, + "learning_rate": 1.0868927855006874e-05, + "loss": 1.189, + "step": 20298 + }, + { + "epoch": 0.49, + "learning_rate": 1.0868151904013165e-05, + "loss": 0.778, + "step": 20299 + }, + { + "epoch": 0.49, + "learning_rate": 1.0867375947752583e-05, + "loss": 0.7938, + "step": 20300 + }, + { + "epoch": 0.49, + "learning_rate": 1.0866599986229835e-05, + "loss": 0.9372, + "step": 20301 + }, + { + "epoch": 0.49, + "learning_rate": 1.0865824019449627e-05, + "loss": 1.1598, + "step": 20302 + }, + { + "epoch": 0.49, + "learning_rate": 1.086504804741667e-05, + "loss": 1.3388, + "step": 20303 + }, + { + "epoch": 0.49, + "learning_rate": 1.0864272070135674e-05, + "loss": 1.1556, + "step": 20304 + }, + { + "epoch": 0.49, + "learning_rate": 1.0863496087611338e-05, + "loss": 1.0563, + "step": 20305 + }, + { + "epoch": 0.49, + "learning_rate": 1.0862720099848374e-05, + "loss": 1.2626, + "step": 20306 + }, + { + "epoch": 0.49, + "learning_rate": 1.0861944106851492e-05, + "loss": 1.2005, + "step": 20307 + }, + { + "epoch": 0.49, + "learning_rate": 1.0861168108625397e-05, + "loss": 1.0023, + "step": 20308 + }, + { + "epoch": 0.49, + "learning_rate": 1.0860392105174797e-05, + "loss": 1.143, + "step": 20309 + }, + { + "epoch": 0.49, + "learning_rate": 1.0859616096504399e-05, + "loss": 1.1354, + "step": 20310 + }, + { + "epoch": 0.49, + "learning_rate": 1.0858840082618917e-05, + "loss": 1.0998, + "step": 20311 + }, + { + "epoch": 0.49, + "learning_rate": 1.0858064063523053e-05, + "loss": 1.1144, + "step": 20312 + }, + { + "epoch": 0.49, + "learning_rate": 1.0857288039221514e-05, + "loss": 0.9569, + "step": 20313 + }, + { + "epoch": 0.49, + "learning_rate": 1.085651200971901e-05, + "loss": 1.0021, + "step": 20314 + }, + { + "epoch": 0.49, + "learning_rate": 1.085573597502025e-05, + "loss": 1.2135, + "step": 20315 + }, + { + "epoch": 0.49, + "learning_rate": 1.0854959935129944e-05, + "loss": 1.1262, + "step": 20316 + }, + { + "epoch": 0.49, + "learning_rate": 1.0854183890052794e-05, + "loss": 0.9808, + "step": 20317 + }, + { + "epoch": 0.49, + "learning_rate": 1.0853407839793508e-05, + "loss": 1.0585, + "step": 20318 + }, + { + "epoch": 0.49, + "learning_rate": 1.0852631784356803e-05, + "loss": 1.085, + "step": 20319 + }, + { + "epoch": 0.49, + "learning_rate": 1.085185572374738e-05, + "loss": 1.107, + "step": 20320 + }, + { + "epoch": 0.49, + "learning_rate": 1.085107965796995e-05, + "loss": 0.8182, + "step": 20321 + }, + { + "epoch": 0.49, + "learning_rate": 1.0850303587029216e-05, + "loss": 1.1325, + "step": 20322 + }, + { + "epoch": 0.49, + "learning_rate": 1.0849527510929892e-05, + "loss": 0.9108, + "step": 20323 + }, + { + "epoch": 0.49, + "learning_rate": 1.0848751429676685e-05, + "loss": 1.0071, + "step": 20324 + }, + { + "epoch": 0.49, + "learning_rate": 1.0847975343274303e-05, + "loss": 0.9529, + "step": 20325 + }, + { + "epoch": 0.49, + "learning_rate": 1.0847199251727452e-05, + "loss": 1.2058, + "step": 20326 + }, + { + "epoch": 0.49, + "learning_rate": 1.0846423155040844e-05, + "loss": 1.0715, + "step": 20327 + }, + { + "epoch": 0.49, + "learning_rate": 1.0845647053219185e-05, + "loss": 1.1759, + "step": 20328 + }, + { + "epoch": 0.49, + "learning_rate": 1.0844870946267187e-05, + "loss": 0.9287, + "step": 20329 + }, + { + "epoch": 0.49, + "learning_rate": 1.0844094834189553e-05, + "loss": 1.1867, + "step": 20330 + }, + { + "epoch": 0.49, + "learning_rate": 1.0843318716990995e-05, + "loss": 1.2078, + "step": 20331 + }, + { + "epoch": 0.49, + "learning_rate": 1.0842542594676221e-05, + "loss": 0.8198, + "step": 20332 + }, + { + "epoch": 0.49, + "learning_rate": 1.0841766467249939e-05, + "loss": 1.0577, + "step": 20333 + }, + { + "epoch": 0.49, + "learning_rate": 1.0840990334716856e-05, + "loss": 0.9706, + "step": 20334 + }, + { + "epoch": 0.49, + "learning_rate": 1.0840214197081685e-05, + "loss": 1.1172, + "step": 20335 + }, + { + "epoch": 0.49, + "learning_rate": 1.0839438054349131e-05, + "loss": 0.9289, + "step": 20336 + }, + { + "epoch": 0.49, + "learning_rate": 1.0838661906523907e-05, + "loss": 1.0888, + "step": 20337 + }, + { + "epoch": 0.49, + "learning_rate": 1.0837885753610713e-05, + "loss": 1.3821, + "step": 20338 + }, + { + "epoch": 0.49, + "learning_rate": 1.0837109595614266e-05, + "loss": 1.259, + "step": 20339 + }, + { + "epoch": 0.49, + "learning_rate": 1.0836333432539272e-05, + "loss": 1.2453, + "step": 20340 + }, + { + "epoch": 0.49, + "learning_rate": 1.0835557264390439e-05, + "loss": 1.023, + "step": 20341 + }, + { + "epoch": 0.49, + "learning_rate": 1.0834781091172475e-05, + "loss": 1.2573, + "step": 20342 + }, + { + "epoch": 0.49, + "learning_rate": 1.0834004912890092e-05, + "loss": 1.0766, + "step": 20343 + }, + { + "epoch": 0.49, + "learning_rate": 1.0833228729548e-05, + "loss": 1.103, + "step": 20344 + }, + { + "epoch": 0.49, + "learning_rate": 1.0832452541150904e-05, + "loss": 1.1239, + "step": 20345 + }, + { + "epoch": 0.49, + "learning_rate": 1.0831676347703512e-05, + "loss": 1.0436, + "step": 20346 + }, + { + "epoch": 0.49, + "learning_rate": 1.0830900149210537e-05, + "loss": 0.9954, + "step": 20347 + }, + { + "epoch": 0.49, + "learning_rate": 1.0830123945676688e-05, + "loss": 0.9907, + "step": 20348 + }, + { + "epoch": 0.49, + "learning_rate": 1.0829347737106667e-05, + "loss": 1.1364, + "step": 20349 + }, + { + "epoch": 0.49, + "learning_rate": 1.0828571523505192e-05, + "loss": 1.4931, + "step": 20350 + }, + { + "epoch": 0.49, + "learning_rate": 1.082779530487697e-05, + "loss": 0.8331, + "step": 20351 + }, + { + "epoch": 0.49, + "learning_rate": 1.0827019081226706e-05, + "loss": 1.369, + "step": 20352 + }, + { + "epoch": 0.49, + "learning_rate": 1.082624285255911e-05, + "loss": 1.2665, + "step": 20353 + }, + { + "epoch": 0.49, + "learning_rate": 1.0825466618878896e-05, + "loss": 0.9527, + "step": 20354 + }, + { + "epoch": 0.49, + "learning_rate": 1.082469038019077e-05, + "loss": 1.2638, + "step": 20355 + }, + { + "epoch": 0.49, + "learning_rate": 1.082391413649944e-05, + "loss": 1.0294, + "step": 20356 + }, + { + "epoch": 0.49, + "learning_rate": 1.0823137887809616e-05, + "loss": 1.189, + "step": 20357 + }, + { + "epoch": 0.49, + "learning_rate": 1.082236163412601e-05, + "loss": 1.1569, + "step": 20358 + }, + { + "epoch": 0.49, + "learning_rate": 1.082158537545333e-05, + "loss": 1.1134, + "step": 20359 + }, + { + "epoch": 0.49, + "learning_rate": 1.0820809111796285e-05, + "loss": 1.3121, + "step": 20360 + }, + { + "epoch": 0.49, + "learning_rate": 1.0820032843159583e-05, + "loss": 1.1831, + "step": 20361 + }, + { + "epoch": 0.49, + "learning_rate": 1.0819256569547933e-05, + "loss": 1.0781, + "step": 20362 + }, + { + "epoch": 0.49, + "learning_rate": 1.081848029096605e-05, + "loss": 0.8291, + "step": 20363 + }, + { + "epoch": 0.49, + "learning_rate": 1.0817704007418637e-05, + "loss": 1.2847, + "step": 20364 + }, + { + "epoch": 0.49, + "learning_rate": 1.0816927718910403e-05, + "loss": 1.1664, + "step": 20365 + }, + { + "epoch": 0.49, + "learning_rate": 1.0816151425446066e-05, + "loss": 1.3135, + "step": 20366 + }, + { + "epoch": 0.49, + "learning_rate": 1.081537512703033e-05, + "loss": 1.3507, + "step": 20367 + }, + { + "epoch": 0.49, + "learning_rate": 1.0814598823667904e-05, + "loss": 1.2749, + "step": 20368 + }, + { + "epoch": 0.49, + "learning_rate": 1.0813822515363495e-05, + "loss": 1.0625, + "step": 20369 + }, + { + "epoch": 0.49, + "learning_rate": 1.081304620212182e-05, + "loss": 1.0856, + "step": 20370 + }, + { + "epoch": 0.49, + "learning_rate": 1.0812269883947583e-05, + "loss": 1.0824, + "step": 20371 + }, + { + "epoch": 0.49, + "learning_rate": 1.0811493560845499e-05, + "loss": 1.0832, + "step": 20372 + }, + { + "epoch": 0.49, + "learning_rate": 1.081071723282027e-05, + "loss": 1.0481, + "step": 20373 + }, + { + "epoch": 0.49, + "learning_rate": 1.080994089987661e-05, + "loss": 1.0721, + "step": 20374 + }, + { + "epoch": 0.49, + "learning_rate": 1.0809164562019232e-05, + "loss": 0.9504, + "step": 20375 + }, + { + "epoch": 0.49, + "learning_rate": 1.0808388219252844e-05, + "loss": 1.1111, + "step": 20376 + }, + { + "epoch": 0.49, + "learning_rate": 1.0807611871582154e-05, + "loss": 1.1622, + "step": 20377 + }, + { + "epoch": 0.49, + "learning_rate": 1.080683551901187e-05, + "loss": 0.9871, + "step": 20378 + }, + { + "epoch": 0.49, + "learning_rate": 1.0806059161546707e-05, + "loss": 0.9767, + "step": 20379 + }, + { + "epoch": 0.49, + "learning_rate": 1.0805282799191372e-05, + "loss": 0.9606, + "step": 20380 + }, + { + "epoch": 0.49, + "learning_rate": 1.0804506431950577e-05, + "loss": 1.1178, + "step": 20381 + }, + { + "epoch": 0.49, + "learning_rate": 1.0803730059829027e-05, + "loss": 0.9914, + "step": 20382 + }, + { + "epoch": 0.49, + "learning_rate": 1.080295368283144e-05, + "loss": 1.3612, + "step": 20383 + }, + { + "epoch": 0.49, + "learning_rate": 1.0802177300962519e-05, + "loss": 0.9837, + "step": 20384 + }, + { + "epoch": 0.49, + "learning_rate": 1.0801400914226981e-05, + "loss": 0.9162, + "step": 20385 + }, + { + "epoch": 0.49, + "learning_rate": 1.0800624522629529e-05, + "loss": 1.2278, + "step": 20386 + }, + { + "epoch": 0.49, + "learning_rate": 1.0799848126174874e-05, + "loss": 1.1518, + "step": 20387 + }, + { + "epoch": 0.49, + "learning_rate": 1.0799071724867734e-05, + "loss": 1.0349, + "step": 20388 + }, + { + "epoch": 0.49, + "learning_rate": 1.0798295318712808e-05, + "loss": 1.1153, + "step": 20389 + }, + { + "epoch": 0.49, + "learning_rate": 1.0797518907714817e-05, + "loss": 1.1667, + "step": 20390 + }, + { + "epoch": 0.49, + "learning_rate": 1.0796742491878464e-05, + "loss": 0.9983, + "step": 20391 + }, + { + "epoch": 0.49, + "learning_rate": 1.0795966071208464e-05, + "loss": 1.1263, + "step": 20392 + }, + { + "epoch": 0.49, + "learning_rate": 1.079518964570952e-05, + "loss": 1.0298, + "step": 20393 + }, + { + "epoch": 0.49, + "learning_rate": 1.0794413215386352e-05, + "loss": 1.1713, + "step": 20394 + }, + { + "epoch": 0.49, + "learning_rate": 1.0793636780243665e-05, + "loss": 0.8971, + "step": 20395 + }, + { + "epoch": 0.49, + "learning_rate": 1.0792860340286171e-05, + "loss": 0.8014, + "step": 20396 + }, + { + "epoch": 0.49, + "learning_rate": 1.0792083895518577e-05, + "loss": 0.8591, + "step": 20397 + }, + { + "epoch": 0.49, + "learning_rate": 1.0791307445945598e-05, + "loss": 1.2068, + "step": 20398 + }, + { + "epoch": 0.49, + "learning_rate": 1.0790530991571943e-05, + "loss": 1.0429, + "step": 20399 + }, + { + "epoch": 0.49, + "learning_rate": 1.0789754532402324e-05, + "loss": 1.1407, + "step": 20400 + }, + { + "epoch": 0.49, + "learning_rate": 1.0788978068441447e-05, + "loss": 1.0491, + "step": 20401 + }, + { + "epoch": 0.49, + "learning_rate": 1.0788201599694028e-05, + "loss": 1.0939, + "step": 20402 + }, + { + "epoch": 0.49, + "learning_rate": 1.0787425126164774e-05, + "loss": 0.9682, + "step": 20403 + }, + { + "epoch": 0.49, + "learning_rate": 1.0786648647858397e-05, + "loss": 1.0631, + "step": 20404 + }, + { + "epoch": 0.49, + "learning_rate": 1.0785872164779607e-05, + "loss": 1.4097, + "step": 20405 + }, + { + "epoch": 0.49, + "learning_rate": 1.0785095676933118e-05, + "loss": 1.0804, + "step": 20406 + }, + { + "epoch": 0.49, + "learning_rate": 1.0784319184323637e-05, + "loss": 1.2327, + "step": 20407 + }, + { + "epoch": 0.49, + "learning_rate": 1.0783542686955876e-05, + "loss": 1.1544, + "step": 20408 + }, + { + "epoch": 0.49, + "learning_rate": 1.0782766184834544e-05, + "loss": 1.1459, + "step": 20409 + }, + { + "epoch": 0.49, + "learning_rate": 1.0781989677964355e-05, + "loss": 0.7598, + "step": 20410 + }, + { + "epoch": 0.49, + "learning_rate": 1.078121316635002e-05, + "loss": 1.2435, + "step": 20411 + }, + { + "epoch": 0.49, + "learning_rate": 1.0780436649996246e-05, + "loss": 0.9649, + "step": 20412 + }, + { + "epoch": 0.49, + "learning_rate": 1.0779660128907747e-05, + "loss": 0.9443, + "step": 20413 + }, + { + "epoch": 0.49, + "learning_rate": 1.0778883603089233e-05, + "loss": 1.1735, + "step": 20414 + }, + { + "epoch": 0.49, + "learning_rate": 1.0778107072545418e-05, + "loss": 0.9782, + "step": 20415 + }, + { + "epoch": 0.49, + "learning_rate": 1.077733053728101e-05, + "loss": 1.095, + "step": 20416 + }, + { + "epoch": 0.49, + "learning_rate": 1.0776553997300718e-05, + "loss": 0.9397, + "step": 20417 + }, + { + "epoch": 0.49, + "learning_rate": 1.0775777452609257e-05, + "loss": 1.4091, + "step": 20418 + }, + { + "epoch": 0.49, + "learning_rate": 1.0775000903211336e-05, + "loss": 1.083, + "step": 20419 + }, + { + "epoch": 0.49, + "learning_rate": 1.0774224349111668e-05, + "loss": 1.1254, + "step": 20420 + }, + { + "epoch": 0.49, + "learning_rate": 1.0773447790314962e-05, + "loss": 1.0806, + "step": 20421 + }, + { + "epoch": 0.49, + "learning_rate": 1.077267122682593e-05, + "loss": 1.0011, + "step": 20422 + }, + { + "epoch": 0.49, + "learning_rate": 1.0771894658649284e-05, + "loss": 1.1271, + "step": 20423 + }, + { + "epoch": 0.49, + "learning_rate": 1.0771118085789735e-05, + "loss": 1.2168, + "step": 20424 + }, + { + "epoch": 0.49, + "learning_rate": 1.0770341508251991e-05, + "loss": 0.8839, + "step": 20425 + }, + { + "epoch": 0.49, + "learning_rate": 1.076956492604077e-05, + "loss": 0.9691, + "step": 20426 + }, + { + "epoch": 0.49, + "learning_rate": 1.0768788339160779e-05, + "loss": 1.0702, + "step": 20427 + }, + { + "epoch": 0.49, + "learning_rate": 1.0768011747616732e-05, + "loss": 1.1543, + "step": 20428 + }, + { + "epoch": 0.49, + "learning_rate": 1.0767235151413331e-05, + "loss": 0.7533, + "step": 20429 + }, + { + "epoch": 0.49, + "learning_rate": 1.0766458550555301e-05, + "loss": 1.0885, + "step": 20430 + }, + { + "epoch": 0.49, + "learning_rate": 1.0765681945047347e-05, + "loss": 0.6088, + "step": 20431 + }, + { + "epoch": 0.49, + "learning_rate": 1.0764905334894177e-05, + "loss": 1.0453, + "step": 20432 + }, + { + "epoch": 0.49, + "learning_rate": 1.0764128720100509e-05, + "loss": 1.1242, + "step": 20433 + }, + { + "epoch": 0.49, + "learning_rate": 1.0763352100671051e-05, + "loss": 1.2166, + "step": 20434 + }, + { + "epoch": 0.49, + "learning_rate": 1.0762575476610517e-05, + "loss": 1.3584, + "step": 20435 + }, + { + "epoch": 0.49, + "learning_rate": 1.0761798847923615e-05, + "loss": 0.9742, + "step": 20436 + }, + { + "epoch": 0.49, + "learning_rate": 1.076102221461506e-05, + "loss": 1.0855, + "step": 20437 + }, + { + "epoch": 0.49, + "learning_rate": 1.0760245576689564e-05, + "loss": 1.0369, + "step": 20438 + }, + { + "epoch": 0.49, + "learning_rate": 1.0759468934151833e-05, + "loss": 0.9013, + "step": 20439 + }, + { + "epoch": 0.49, + "learning_rate": 1.0758692287006583e-05, + "loss": 1.0539, + "step": 20440 + }, + { + "epoch": 0.49, + "learning_rate": 1.0757915635258527e-05, + "loss": 0.7454, + "step": 20441 + }, + { + "epoch": 0.49, + "learning_rate": 1.0757138978912374e-05, + "loss": 1.1736, + "step": 20442 + }, + { + "epoch": 0.49, + "learning_rate": 1.075636231797284e-05, + "loss": 1.2451, + "step": 20443 + }, + { + "epoch": 0.49, + "learning_rate": 1.0755585652444629e-05, + "loss": 0.6631, + "step": 20444 + }, + { + "epoch": 0.49, + "learning_rate": 1.0754808982332458e-05, + "loss": 1.0493, + "step": 20445 + }, + { + "epoch": 0.49, + "learning_rate": 1.0754032307641043e-05, + "loss": 1.057, + "step": 20446 + }, + { + "epoch": 0.49, + "learning_rate": 1.0753255628375088e-05, + "loss": 1.1152, + "step": 20447 + }, + { + "epoch": 0.49, + "learning_rate": 1.0752478944539308e-05, + "loss": 1.2201, + "step": 20448 + }, + { + "epoch": 0.49, + "learning_rate": 1.0751702256138416e-05, + "loss": 1.0958, + "step": 20449 + }, + { + "epoch": 0.49, + "learning_rate": 1.0750925563177123e-05, + "loss": 1.2787, + "step": 20450 + }, + { + "epoch": 0.49, + "learning_rate": 1.0750148865660144e-05, + "loss": 1.2542, + "step": 20451 + }, + { + "epoch": 0.49, + "learning_rate": 1.0749372163592185e-05, + "loss": 1.0365, + "step": 20452 + }, + { + "epoch": 0.49, + "learning_rate": 1.0748595456977962e-05, + "loss": 1.0479, + "step": 20453 + }, + { + "epoch": 0.49, + "learning_rate": 1.0747818745822186e-05, + "loss": 1.0438, + "step": 20454 + }, + { + "epoch": 0.49, + "learning_rate": 1.074704203012957e-05, + "loss": 1.0856, + "step": 20455 + }, + { + "epoch": 0.49, + "learning_rate": 1.0746265309904827e-05, + "loss": 1.0961, + "step": 20456 + }, + { + "epoch": 0.49, + "learning_rate": 1.0745488585152664e-05, + "loss": 1.0862, + "step": 20457 + }, + { + "epoch": 0.49, + "learning_rate": 1.0744711855877802e-05, + "loss": 1.1348, + "step": 20458 + }, + { + "epoch": 0.49, + "learning_rate": 1.0743935122084945e-05, + "loss": 1.2273, + "step": 20459 + }, + { + "epoch": 0.49, + "learning_rate": 1.0743158383778811e-05, + "loss": 0.9443, + "step": 20460 + }, + { + "epoch": 0.49, + "learning_rate": 1.0742381640964106e-05, + "loss": 1.0961, + "step": 20461 + }, + { + "epoch": 0.49, + "learning_rate": 1.074160489364555e-05, + "loss": 0.9457, + "step": 20462 + }, + { + "epoch": 0.49, + "learning_rate": 1.0740828141827851e-05, + "loss": 1.3564, + "step": 20463 + }, + { + "epoch": 0.49, + "learning_rate": 1.0740051385515721e-05, + "loss": 1.0907, + "step": 20464 + }, + { + "epoch": 0.49, + "learning_rate": 1.0739274624713873e-05, + "loss": 0.8613, + "step": 20465 + }, + { + "epoch": 0.49, + "learning_rate": 1.0738497859427022e-05, + "loss": 0.8968, + "step": 20466 + }, + { + "epoch": 0.49, + "learning_rate": 1.0737721089659875e-05, + "loss": 1.0328, + "step": 20467 + }, + { + "epoch": 0.49, + "learning_rate": 1.073694431541715e-05, + "loss": 1.15, + "step": 20468 + }, + { + "epoch": 0.49, + "learning_rate": 1.0736167536703557e-05, + "loss": 0.8383, + "step": 20469 + }, + { + "epoch": 0.49, + "learning_rate": 1.0735390753523806e-05, + "loss": 0.977, + "step": 20470 + }, + { + "epoch": 0.49, + "learning_rate": 1.0734613965882616e-05, + "loss": 1.0018, + "step": 20471 + }, + { + "epoch": 0.49, + "learning_rate": 1.0733837173784694e-05, + "loss": 1.0183, + "step": 20472 + }, + { + "epoch": 0.49, + "learning_rate": 1.0733060377234754e-05, + "loss": 1.1076, + "step": 20473 + }, + { + "epoch": 0.49, + "learning_rate": 1.073228357623751e-05, + "loss": 1.2911, + "step": 20474 + }, + { + "epoch": 0.49, + "learning_rate": 1.0731506770797675e-05, + "loss": 0.9035, + "step": 20475 + }, + { + "epoch": 0.49, + "learning_rate": 1.0730729960919959e-05, + "loss": 1.1081, + "step": 20476 + }, + { + "epoch": 0.49, + "learning_rate": 1.0729953146609076e-05, + "loss": 0.9477, + "step": 20477 + }, + { + "epoch": 0.49, + "learning_rate": 1.0729176327869738e-05, + "loss": 1.1271, + "step": 20478 + }, + { + "epoch": 0.49, + "learning_rate": 1.0728399504706661e-05, + "loss": 1.2288, + "step": 20479 + }, + { + "epoch": 0.49, + "learning_rate": 1.0727622677124554e-05, + "loss": 1.188, + "step": 20480 + }, + { + "epoch": 0.49, + "learning_rate": 1.0726845845128133e-05, + "loss": 1.1444, + "step": 20481 + }, + { + "epoch": 0.49, + "learning_rate": 1.0726069008722109e-05, + "loss": 0.8724, + "step": 20482 + }, + { + "epoch": 0.49, + "learning_rate": 1.0725292167911194e-05, + "loss": 0.8505, + "step": 20483 + }, + { + "epoch": 0.49, + "learning_rate": 1.0724515322700101e-05, + "loss": 1.0828, + "step": 20484 + }, + { + "epoch": 0.49, + "learning_rate": 1.0723738473093545e-05, + "loss": 1.2171, + "step": 20485 + }, + { + "epoch": 0.49, + "learning_rate": 1.072296161909624e-05, + "loss": 1.1445, + "step": 20486 + }, + { + "epoch": 0.49, + "learning_rate": 1.0722184760712895e-05, + "loss": 0.9505, + "step": 20487 + }, + { + "epoch": 0.49, + "learning_rate": 1.0721407897948224e-05, + "loss": 1.16, + "step": 20488 + }, + { + "epoch": 0.49, + "learning_rate": 1.0720631030806942e-05, + "loss": 1.2404, + "step": 20489 + }, + { + "epoch": 0.49, + "learning_rate": 1.0719854159293761e-05, + "loss": 1.2686, + "step": 20490 + }, + { + "epoch": 0.49, + "learning_rate": 1.0719077283413395e-05, + "loss": 1.017, + "step": 20491 + }, + { + "epoch": 0.49, + "learning_rate": 1.0718300403170555e-05, + "loss": 1.1082, + "step": 20492 + }, + { + "epoch": 0.49, + "learning_rate": 1.0717523518569957e-05, + "loss": 1.0456, + "step": 20493 + }, + { + "epoch": 0.49, + "learning_rate": 1.071674662961631e-05, + "loss": 1.07, + "step": 20494 + }, + { + "epoch": 0.49, + "learning_rate": 1.0715969736314334e-05, + "loss": 1.1711, + "step": 20495 + }, + { + "epoch": 0.49, + "learning_rate": 1.0715192838668732e-05, + "loss": 1.1916, + "step": 20496 + }, + { + "epoch": 0.49, + "learning_rate": 1.0714415936684229e-05, + "loss": 1.3159, + "step": 20497 + }, + { + "epoch": 0.49, + "learning_rate": 1.0713639030365528e-05, + "loss": 1.1387, + "step": 20498 + }, + { + "epoch": 0.49, + "learning_rate": 1.0712862119717351e-05, + "loss": 0.7794, + "step": 20499 + }, + { + "epoch": 0.49, + "learning_rate": 1.0712085204744402e-05, + "loss": 1.0795, + "step": 20500 + }, + { + "epoch": 0.49, + "learning_rate": 1.0711308285451405e-05, + "loss": 1.1423, + "step": 20501 + }, + { + "epoch": 0.49, + "learning_rate": 1.0710531361843066e-05, + "loss": 1.0684, + "step": 20502 + }, + { + "epoch": 0.49, + "learning_rate": 1.07097544339241e-05, + "loss": 0.9793, + "step": 20503 + }, + { + "epoch": 0.49, + "learning_rate": 1.0708977501699218e-05, + "loss": 1.1034, + "step": 20504 + }, + { + "epoch": 0.49, + "learning_rate": 1.070820056517314e-05, + "loss": 1.1156, + "step": 20505 + }, + { + "epoch": 0.49, + "learning_rate": 1.0707423624350577e-05, + "loss": 1.263, + "step": 20506 + }, + { + "epoch": 0.49, + "learning_rate": 1.070664667923624e-05, + "loss": 0.8549, + "step": 20507 + }, + { + "epoch": 0.49, + "learning_rate": 1.0705869729834838e-05, + "loss": 0.9998, + "step": 20508 + }, + { + "epoch": 0.49, + "learning_rate": 1.0705092776151097e-05, + "loss": 0.9932, + "step": 20509 + }, + { + "epoch": 0.49, + "learning_rate": 1.0704315818189723e-05, + "loss": 0.8347, + "step": 20510 + }, + { + "epoch": 0.49, + "learning_rate": 1.070353885595543e-05, + "loss": 0.9777, + "step": 20511 + }, + { + "epoch": 0.49, + "learning_rate": 1.070276188945293e-05, + "loss": 1.0089, + "step": 20512 + }, + { + "epoch": 0.49, + "learning_rate": 1.0701984918686944e-05, + "loss": 1.094, + "step": 20513 + }, + { + "epoch": 0.49, + "learning_rate": 1.0701207943662177e-05, + "loss": 0.9772, + "step": 20514 + }, + { + "epoch": 0.49, + "learning_rate": 1.0700430964383345e-05, + "loss": 0.9635, + "step": 20515 + }, + { + "epoch": 0.49, + "learning_rate": 1.0699653980855164e-05, + "loss": 1.2079, + "step": 20516 + }, + { + "epoch": 0.49, + "learning_rate": 1.069887699308235e-05, + "loss": 1.027, + "step": 20517 + }, + { + "epoch": 0.49, + "learning_rate": 1.0698100001069611e-05, + "loss": 1.0795, + "step": 20518 + }, + { + "epoch": 0.49, + "learning_rate": 1.0697323004821662e-05, + "loss": 1.1756, + "step": 20519 + }, + { + "epoch": 0.49, + "learning_rate": 1.0696546004343222e-05, + "loss": 1.0028, + "step": 20520 + }, + { + "epoch": 0.49, + "learning_rate": 1.0695768999639e-05, + "loss": 0.9577, + "step": 20521 + }, + { + "epoch": 0.49, + "learning_rate": 1.0694991990713712e-05, + "loss": 0.9382, + "step": 20522 + }, + { + "epoch": 0.49, + "learning_rate": 1.0694214977572068e-05, + "loss": 1.1015, + "step": 20523 + }, + { + "epoch": 0.49, + "learning_rate": 1.0693437960218787e-05, + "loss": 1.2556, + "step": 20524 + }, + { + "epoch": 0.49, + "learning_rate": 1.069266093865858e-05, + "loss": 0.9823, + "step": 20525 + }, + { + "epoch": 0.49, + "learning_rate": 1.0691883912896166e-05, + "loss": 1.0161, + "step": 20526 + }, + { + "epoch": 0.49, + "learning_rate": 1.0691106882936248e-05, + "loss": 0.8755, + "step": 20527 + }, + { + "epoch": 0.49, + "learning_rate": 1.0690329848783554e-05, + "loss": 1.0099, + "step": 20528 + }, + { + "epoch": 0.49, + "learning_rate": 1.0689552810442786e-05, + "loss": 1.1269, + "step": 20529 + }, + { + "epoch": 0.49, + "learning_rate": 1.0688775767918667e-05, + "loss": 0.7967, + "step": 20530 + }, + { + "epoch": 0.49, + "learning_rate": 1.0687998721215904e-05, + "loss": 1.0927, + "step": 20531 + }, + { + "epoch": 0.49, + "learning_rate": 1.0687221670339219e-05, + "loss": 0.9468, + "step": 20532 + }, + { + "epoch": 0.49, + "learning_rate": 1.0686444615293319e-05, + "loss": 1.28, + "step": 20533 + }, + { + "epoch": 0.49, + "learning_rate": 1.0685667556082922e-05, + "loss": 0.9618, + "step": 20534 + }, + { + "epoch": 0.49, + "learning_rate": 1.0684890492712739e-05, + "loss": 1.0311, + "step": 20535 + }, + { + "epoch": 0.49, + "learning_rate": 1.0684113425187487e-05, + "loss": 1.0293, + "step": 20536 + }, + { + "epoch": 0.49, + "learning_rate": 1.0683336353511882e-05, + "loss": 1.0139, + "step": 20537 + }, + { + "epoch": 0.49, + "learning_rate": 1.0682559277690636e-05, + "loss": 0.9889, + "step": 20538 + }, + { + "epoch": 0.49, + "learning_rate": 1.0681782197728458e-05, + "loss": 1.2435, + "step": 20539 + }, + { + "epoch": 0.49, + "learning_rate": 1.0681005113630072e-05, + "loss": 1.0665, + "step": 20540 + }, + { + "epoch": 0.49, + "learning_rate": 1.0680228025400189e-05, + "loss": 1.2128, + "step": 20541 + }, + { + "epoch": 0.49, + "learning_rate": 1.0679450933043522e-05, + "loss": 1.3257, + "step": 20542 + }, + { + "epoch": 0.49, + "learning_rate": 1.0678673836564785e-05, + "loss": 0.9379, + "step": 20543 + }, + { + "epoch": 0.49, + "learning_rate": 1.0677896735968694e-05, + "loss": 1.0846, + "step": 20544 + }, + { + "epoch": 0.49, + "learning_rate": 1.0677119631259962e-05, + "loss": 1.2457, + "step": 20545 + }, + { + "epoch": 0.49, + "learning_rate": 1.0676342522443308e-05, + "loss": 0.7752, + "step": 20546 + }, + { + "epoch": 0.49, + "learning_rate": 1.0675565409523441e-05, + "loss": 0.993, + "step": 20547 + }, + { + "epoch": 0.49, + "learning_rate": 1.0674788292505073e-05, + "loss": 1.1377, + "step": 20548 + }, + { + "epoch": 0.49, + "learning_rate": 1.067401117139293e-05, + "loss": 1.1432, + "step": 20549 + }, + { + "epoch": 0.49, + "learning_rate": 1.0673234046191716e-05, + "loss": 1.0986, + "step": 20550 + }, + { + "epoch": 0.49, + "learning_rate": 1.0672456916906153e-05, + "loss": 1.0236, + "step": 20551 + }, + { + "epoch": 0.49, + "learning_rate": 1.0671679783540947e-05, + "loss": 1.2338, + "step": 20552 + }, + { + "epoch": 0.49, + "learning_rate": 1.067090264610082e-05, + "loss": 1.0474, + "step": 20553 + }, + { + "epoch": 0.49, + "learning_rate": 1.0670125504590484e-05, + "loss": 1.032, + "step": 20554 + }, + { + "epoch": 0.49, + "learning_rate": 1.0669348359014658e-05, + "loss": 0.891, + "step": 20555 + }, + { + "epoch": 0.49, + "learning_rate": 1.0668571209378048e-05, + "loss": 1.1452, + "step": 20556 + }, + { + "epoch": 0.49, + "learning_rate": 1.0667794055685376e-05, + "loss": 1.0851, + "step": 20557 + }, + { + "epoch": 0.49, + "learning_rate": 1.0667016897941354e-05, + "loss": 0.9609, + "step": 20558 + }, + { + "epoch": 0.49, + "learning_rate": 1.0666239736150696e-05, + "loss": 1.1134, + "step": 20559 + }, + { + "epoch": 0.49, + "learning_rate": 1.0665462570318121e-05, + "loss": 0.7066, + "step": 20560 + }, + { + "epoch": 0.49, + "learning_rate": 1.066468540044834e-05, + "loss": 1.1863, + "step": 20561 + }, + { + "epoch": 0.49, + "learning_rate": 1.066390822654607e-05, + "loss": 1.2144, + "step": 20562 + }, + { + "epoch": 0.49, + "learning_rate": 1.0663131048616023e-05, + "loss": 1.0812, + "step": 20563 + }, + { + "epoch": 0.49, + "learning_rate": 1.0662353866662915e-05, + "loss": 1.1967, + "step": 20564 + }, + { + "epoch": 0.49, + "learning_rate": 1.0661576680691466e-05, + "loss": 1.1678, + "step": 20565 + }, + { + "epoch": 0.49, + "learning_rate": 1.0660799490706383e-05, + "loss": 1.1835, + "step": 20566 + }, + { + "epoch": 0.49, + "learning_rate": 1.0660022296712385e-05, + "loss": 1.11, + "step": 20567 + }, + { + "epoch": 0.49, + "learning_rate": 1.0659245098714188e-05, + "loss": 1.0723, + "step": 20568 + }, + { + "epoch": 0.49, + "learning_rate": 1.065846789671651e-05, + "loss": 1.1239, + "step": 20569 + }, + { + "epoch": 0.49, + "learning_rate": 1.0657690690724056e-05, + "loss": 0.9951, + "step": 20570 + }, + { + "epoch": 0.49, + "learning_rate": 1.0656913480741549e-05, + "loss": 1.2066, + "step": 20571 + }, + { + "epoch": 0.49, + "learning_rate": 1.0656136266773704e-05, + "loss": 0.8028, + "step": 20572 + }, + { + "epoch": 0.49, + "learning_rate": 1.0655359048825233e-05, + "loss": 1.1399, + "step": 20573 + }, + { + "epoch": 0.49, + "learning_rate": 1.0654581826900855e-05, + "loss": 1.2806, + "step": 20574 + }, + { + "epoch": 0.49, + "learning_rate": 1.065380460100528e-05, + "loss": 1.2608, + "step": 20575 + }, + { + "epoch": 0.49, + "learning_rate": 1.065302737114323e-05, + "loss": 0.8089, + "step": 20576 + }, + { + "epoch": 0.49, + "learning_rate": 1.0652250137319414e-05, + "loss": 1.0608, + "step": 20577 + }, + { + "epoch": 0.49, + "learning_rate": 1.065147289953855e-05, + "loss": 1.1118, + "step": 20578 + }, + { + "epoch": 0.49, + "learning_rate": 1.0650695657805352e-05, + "loss": 1.2436, + "step": 20579 + }, + { + "epoch": 0.49, + "learning_rate": 1.064991841212454e-05, + "loss": 0.5372, + "step": 20580 + }, + { + "epoch": 0.49, + "learning_rate": 1.0649141162500825e-05, + "loss": 1.1099, + "step": 20581 + }, + { + "epoch": 0.49, + "learning_rate": 1.0648363908938921e-05, + "loss": 1.1995, + "step": 20582 + }, + { + "epoch": 0.49, + "learning_rate": 1.0647586651443547e-05, + "loss": 0.9316, + "step": 20583 + }, + { + "epoch": 0.5, + "learning_rate": 1.0646809390019416e-05, + "loss": 1.3074, + "step": 20584 + }, + { + "epoch": 0.5, + "learning_rate": 1.0646032124671246e-05, + "loss": 1.2441, + "step": 20585 + }, + { + "epoch": 0.5, + "learning_rate": 1.064525485540375e-05, + "loss": 1.068, + "step": 20586 + }, + { + "epoch": 0.5, + "learning_rate": 1.0644477582221644e-05, + "loss": 1.0025, + "step": 20587 + }, + { + "epoch": 0.5, + "learning_rate": 1.0643700305129644e-05, + "loss": 1.0096, + "step": 20588 + }, + { + "epoch": 0.5, + "learning_rate": 1.064292302413247e-05, + "loss": 0.8896, + "step": 20589 + }, + { + "epoch": 0.5, + "learning_rate": 1.064214573923483e-05, + "loss": 1.1066, + "step": 20590 + }, + { + "epoch": 0.5, + "learning_rate": 1.064136845044144e-05, + "loss": 1.2012, + "step": 20591 + }, + { + "epoch": 0.5, + "learning_rate": 1.0640591157757023e-05, + "loss": 1.0072, + "step": 20592 + }, + { + "epoch": 0.5, + "learning_rate": 1.063981386118629e-05, + "loss": 0.975, + "step": 20593 + }, + { + "epoch": 0.5, + "learning_rate": 1.0639036560733956e-05, + "loss": 1.3629, + "step": 20594 + }, + { + "epoch": 0.5, + "learning_rate": 1.0638259256404734e-05, + "loss": 0.9791, + "step": 20595 + }, + { + "epoch": 0.5, + "learning_rate": 1.0637481948203348e-05, + "loss": 0.9303, + "step": 20596 + }, + { + "epoch": 0.5, + "learning_rate": 1.0636704636134507e-05, + "loss": 0.8463, + "step": 20597 + }, + { + "epoch": 0.5, + "learning_rate": 1.0635927320202928e-05, + "loss": 0.759, + "step": 20598 + }, + { + "epoch": 0.5, + "learning_rate": 1.0635150000413328e-05, + "loss": 0.9604, + "step": 20599 + }, + { + "epoch": 0.5, + "learning_rate": 1.0634372676770422e-05, + "loss": 1.0633, + "step": 20600 + }, + { + "epoch": 0.5, + "learning_rate": 1.0633595349278927e-05, + "loss": 1.1672, + "step": 20601 + }, + { + "epoch": 0.5, + "learning_rate": 1.0632818017943558e-05, + "loss": 1.044, + "step": 20602 + }, + { + "epoch": 0.5, + "learning_rate": 1.0632040682769029e-05, + "loss": 1.0388, + "step": 20603 + }, + { + "epoch": 0.5, + "learning_rate": 1.0631263343760061e-05, + "loss": 0.9855, + "step": 20604 + }, + { + "epoch": 0.5, + "learning_rate": 1.0630486000921364e-05, + "loss": 1.2451, + "step": 20605 + }, + { + "epoch": 0.5, + "learning_rate": 1.0629708654257658e-05, + "loss": 1.2523, + "step": 20606 + }, + { + "epoch": 0.5, + "learning_rate": 1.062893130377366e-05, + "loss": 1.1157, + "step": 20607 + }, + { + "epoch": 0.5, + "learning_rate": 1.062815394947408e-05, + "loss": 1.006, + "step": 20608 + }, + { + "epoch": 0.5, + "learning_rate": 1.062737659136364e-05, + "loss": 1.1478, + "step": 20609 + }, + { + "epoch": 0.5, + "learning_rate": 1.0626599229447053e-05, + "loss": 1.1344, + "step": 20610 + }, + { + "epoch": 0.5, + "learning_rate": 1.0625821863729036e-05, + "loss": 1.2839, + "step": 20611 + }, + { + "epoch": 0.5, + "learning_rate": 1.0625044494214305e-05, + "loss": 1.2943, + "step": 20612 + }, + { + "epoch": 0.5, + "learning_rate": 1.0624267120907576e-05, + "loss": 0.9613, + "step": 20613 + }, + { + "epoch": 0.5, + "learning_rate": 1.0623489743813564e-05, + "loss": 1.145, + "step": 20614 + }, + { + "epoch": 0.5, + "learning_rate": 1.062271236293699e-05, + "loss": 1.0927, + "step": 20615 + }, + { + "epoch": 0.5, + "learning_rate": 1.0621934978282561e-05, + "loss": 1.0674, + "step": 20616 + }, + { + "epoch": 0.5, + "learning_rate": 1.0621157589855005e-05, + "loss": 1.1917, + "step": 20617 + }, + { + "epoch": 0.5, + "learning_rate": 1.0620380197659026e-05, + "loss": 1.2306, + "step": 20618 + }, + { + "epoch": 0.5, + "learning_rate": 1.0619602801699349e-05, + "loss": 0.9407, + "step": 20619 + }, + { + "epoch": 0.5, + "learning_rate": 1.0618825401980689e-05, + "loss": 1.3145, + "step": 20620 + }, + { + "epoch": 0.5, + "learning_rate": 1.061804799850776e-05, + "loss": 1.1755, + "step": 20621 + }, + { + "epoch": 0.5, + "learning_rate": 1.0617270591285278e-05, + "loss": 0.9014, + "step": 20622 + }, + { + "epoch": 0.5, + "learning_rate": 1.0616493180317961e-05, + "loss": 0.9727, + "step": 20623 + }, + { + "epoch": 0.5, + "learning_rate": 1.0615715765610523e-05, + "loss": 1.1829, + "step": 20624 + }, + { + "epoch": 0.5, + "learning_rate": 1.0614938347167685e-05, + "loss": 0.9133, + "step": 20625 + }, + { + "epoch": 0.5, + "learning_rate": 1.0614160924994161e-05, + "loss": 1.0408, + "step": 20626 + }, + { + "epoch": 0.5, + "learning_rate": 1.0613383499094665e-05, + "loss": 1.0525, + "step": 20627 + }, + { + "epoch": 0.5, + "learning_rate": 1.0612606069473917e-05, + "loss": 1.1255, + "step": 20628 + }, + { + "epoch": 0.5, + "learning_rate": 1.0611828636136631e-05, + "loss": 1.0649, + "step": 20629 + }, + { + "epoch": 0.5, + "learning_rate": 1.0611051199087527e-05, + "loss": 1.0104, + "step": 20630 + }, + { + "epoch": 0.5, + "learning_rate": 1.0610273758331315e-05, + "loss": 1.224, + "step": 20631 + }, + { + "epoch": 0.5, + "learning_rate": 1.060949631387272e-05, + "loss": 1.2071, + "step": 20632 + }, + { + "epoch": 0.5, + "learning_rate": 1.0608718865716453e-05, + "loss": 1.1105, + "step": 20633 + }, + { + "epoch": 0.5, + "learning_rate": 1.0607941413867231e-05, + "loss": 1.0478, + "step": 20634 + }, + { + "epoch": 0.5, + "learning_rate": 1.060716395832977e-05, + "loss": 1.094, + "step": 20635 + }, + { + "epoch": 0.5, + "learning_rate": 1.060638649910879e-05, + "loss": 0.9888, + "step": 20636 + }, + { + "epoch": 0.5, + "learning_rate": 1.0605609036209005e-05, + "loss": 1.1644, + "step": 20637 + }, + { + "epoch": 0.5, + "learning_rate": 1.0604831569635134e-05, + "loss": 1.0421, + "step": 20638 + }, + { + "epoch": 0.5, + "learning_rate": 1.060405409939189e-05, + "loss": 1.343, + "step": 20639 + }, + { + "epoch": 0.5, + "learning_rate": 1.0603276625483994e-05, + "loss": 1.2927, + "step": 20640 + }, + { + "epoch": 0.5, + "learning_rate": 1.0602499147916159e-05, + "loss": 0.7962, + "step": 20641 + }, + { + "epoch": 0.5, + "learning_rate": 1.0601721666693105e-05, + "loss": 1.0464, + "step": 20642 + }, + { + "epoch": 0.5, + "learning_rate": 1.0600944181819543e-05, + "loss": 1.3072, + "step": 20643 + }, + { + "epoch": 0.5, + "learning_rate": 1.06001666933002e-05, + "loss": 1.0791, + "step": 20644 + }, + { + "epoch": 0.5, + "learning_rate": 1.0599389201139783e-05, + "loss": 0.933, + "step": 20645 + }, + { + "epoch": 0.5, + "learning_rate": 1.0598611705343011e-05, + "loss": 1.2714, + "step": 20646 + }, + { + "epoch": 0.5, + "learning_rate": 1.0597834205914607e-05, + "loss": 0.8107, + "step": 20647 + }, + { + "epoch": 0.5, + "learning_rate": 1.0597056702859283e-05, + "loss": 1.0968, + "step": 20648 + }, + { + "epoch": 0.5, + "learning_rate": 1.0596279196181756e-05, + "loss": 1.0681, + "step": 20649 + }, + { + "epoch": 0.5, + "learning_rate": 1.059550168588674e-05, + "loss": 1.157, + "step": 20650 + }, + { + "epoch": 0.5, + "learning_rate": 1.0594724171978958e-05, + "loss": 1.0839, + "step": 20651 + }, + { + "epoch": 0.5, + "learning_rate": 1.0593946654463125e-05, + "loss": 1.0717, + "step": 20652 + }, + { + "epoch": 0.5, + "learning_rate": 1.0593169133343957e-05, + "loss": 1.0317, + "step": 20653 + }, + { + "epoch": 0.5, + "learning_rate": 1.0592391608626169e-05, + "loss": 1.1252, + "step": 20654 + }, + { + "epoch": 0.5, + "learning_rate": 1.0591614080314484e-05, + "loss": 1.2361, + "step": 20655 + }, + { + "epoch": 0.5, + "learning_rate": 1.0590836548413615e-05, + "loss": 1.3068, + "step": 20656 + }, + { + "epoch": 0.5, + "learning_rate": 1.0590059012928277e-05, + "loss": 1.3912, + "step": 20657 + }, + { + "epoch": 0.5, + "learning_rate": 1.058928147386319e-05, + "loss": 1.1612, + "step": 20658 + }, + { + "epoch": 0.5, + "learning_rate": 1.0588503931223072e-05, + "loss": 0.8499, + "step": 20659 + }, + { + "epoch": 0.5, + "learning_rate": 1.058772638501264e-05, + "loss": 1.0479, + "step": 20660 + }, + { + "epoch": 0.5, + "learning_rate": 1.058694883523661e-05, + "loss": 1.1886, + "step": 20661 + }, + { + "epoch": 0.5, + "learning_rate": 1.0586171281899698e-05, + "loss": 1.234, + "step": 20662 + }, + { + "epoch": 0.5, + "learning_rate": 1.0585393725006626e-05, + "loss": 1.1416, + "step": 20663 + }, + { + "epoch": 0.5, + "learning_rate": 1.0584616164562105e-05, + "loss": 1.1102, + "step": 20664 + }, + { + "epoch": 0.5, + "learning_rate": 1.0583838600570855e-05, + "loss": 1.23, + "step": 20665 + }, + { + "epoch": 0.5, + "learning_rate": 1.0583061033037594e-05, + "loss": 0.8945, + "step": 20666 + }, + { + "epoch": 0.5, + "learning_rate": 1.058228346196704e-05, + "loss": 1.0758, + "step": 20667 + }, + { + "epoch": 0.5, + "learning_rate": 1.0581505887363908e-05, + "loss": 1.1046, + "step": 20668 + }, + { + "epoch": 0.5, + "learning_rate": 1.0580728309232917e-05, + "loss": 0.8874, + "step": 20669 + }, + { + "epoch": 0.5, + "learning_rate": 1.0579950727578783e-05, + "loss": 1.2656, + "step": 20670 + }, + { + "epoch": 0.5, + "learning_rate": 1.0579173142406226e-05, + "loss": 1.0803, + "step": 20671 + }, + { + "epoch": 0.5, + "learning_rate": 1.0578395553719962e-05, + "loss": 1.1355, + "step": 20672 + }, + { + "epoch": 0.5, + "learning_rate": 1.0577617961524709e-05, + "loss": 1.1124, + "step": 20673 + }, + { + "epoch": 0.5, + "learning_rate": 1.0576840365825178e-05, + "loss": 1.1829, + "step": 20674 + }, + { + "epoch": 0.5, + "learning_rate": 1.0576062766626098e-05, + "loss": 1.0236, + "step": 20675 + }, + { + "epoch": 0.5, + "learning_rate": 1.0575285163932179e-05, + "loss": 0.9773, + "step": 20676 + }, + { + "epoch": 0.5, + "learning_rate": 1.0574507557748142e-05, + "loss": 0.9843, + "step": 20677 + }, + { + "epoch": 0.5, + "learning_rate": 1.0573729948078699e-05, + "loss": 0.9155, + "step": 20678 + }, + { + "epoch": 0.5, + "learning_rate": 1.0572952334928573e-05, + "loss": 0.9725, + "step": 20679 + }, + { + "epoch": 0.5, + "learning_rate": 1.0572174718302482e-05, + "loss": 0.795, + "step": 20680 + }, + { + "epoch": 0.5, + "learning_rate": 1.0571397098205139e-05, + "loss": 1.1584, + "step": 20681 + }, + { + "epoch": 0.5, + "learning_rate": 1.0570619474641263e-05, + "loss": 0.9846, + "step": 20682 + }, + { + "epoch": 0.5, + "learning_rate": 1.0569841847615576e-05, + "loss": 1.0722, + "step": 20683 + }, + { + "epoch": 0.5, + "learning_rate": 1.0569064217132791e-05, + "loss": 1.2538, + "step": 20684 + }, + { + "epoch": 0.5, + "learning_rate": 1.0568286583197629e-05, + "loss": 1.1193, + "step": 20685 + }, + { + "epoch": 0.5, + "learning_rate": 1.0567508945814804e-05, + "loss": 0.9472, + "step": 20686 + }, + { + "epoch": 0.5, + "learning_rate": 1.0566731304989036e-05, + "loss": 1.0677, + "step": 20687 + }, + { + "epoch": 0.5, + "learning_rate": 1.0565953660725043e-05, + "loss": 0.8891, + "step": 20688 + }, + { + "epoch": 0.5, + "learning_rate": 1.056517601302754e-05, + "loss": 0.965, + "step": 20689 + }, + { + "epoch": 0.5, + "learning_rate": 1.0564398361901252e-05, + "loss": 1.1341, + "step": 20690 + }, + { + "epoch": 0.5, + "learning_rate": 1.0563620707350888e-05, + "loss": 0.9377, + "step": 20691 + }, + { + "epoch": 0.5, + "learning_rate": 1.0562843049381171e-05, + "loss": 1.1661, + "step": 20692 + }, + { + "epoch": 0.5, + "learning_rate": 1.0562065387996815e-05, + "loss": 1.0211, + "step": 20693 + }, + { + "epoch": 0.5, + "learning_rate": 1.0561287723202544e-05, + "loss": 1.027, + "step": 20694 + }, + { + "epoch": 0.5, + "learning_rate": 1.0560510055003073e-05, + "loss": 1.119, + "step": 20695 + }, + { + "epoch": 0.5, + "learning_rate": 1.0559732383403118e-05, + "loss": 1.0214, + "step": 20696 + }, + { + "epoch": 0.5, + "learning_rate": 1.0558954708407395e-05, + "loss": 1.1432, + "step": 20697 + }, + { + "epoch": 0.5, + "learning_rate": 1.0558177030020629e-05, + "loss": 1.1792, + "step": 20698 + }, + { + "epoch": 0.5, + "learning_rate": 1.0557399348247534e-05, + "loss": 0.9522, + "step": 20699 + }, + { + "epoch": 0.5, + "learning_rate": 1.0556621663092828e-05, + "loss": 1.1341, + "step": 20700 + }, + { + "epoch": 0.5, + "learning_rate": 1.0555843974561229e-05, + "loss": 0.9658, + "step": 20701 + }, + { + "epoch": 0.5, + "learning_rate": 1.0555066282657456e-05, + "loss": 0.7324, + "step": 20702 + }, + { + "epoch": 0.5, + "learning_rate": 1.0554288587386226e-05, + "loss": 1.1152, + "step": 20703 + }, + { + "epoch": 0.5, + "learning_rate": 1.0553510888752256e-05, + "loss": 1.0793, + "step": 20704 + }, + { + "epoch": 0.5, + "learning_rate": 1.0552733186760267e-05, + "loss": 1.2356, + "step": 20705 + }, + { + "epoch": 0.5, + "learning_rate": 1.0551955481414976e-05, + "loss": 1.0577, + "step": 20706 + }, + { + "epoch": 0.5, + "learning_rate": 1.0551177772721097e-05, + "loss": 1.1206, + "step": 20707 + }, + { + "epoch": 0.5, + "learning_rate": 1.0550400060683356e-05, + "loss": 1.1182, + "step": 20708 + }, + { + "epoch": 0.5, + "learning_rate": 1.0549622345306469e-05, + "loss": 1.1943, + "step": 20709 + }, + { + "epoch": 0.5, + "learning_rate": 1.0548844626595149e-05, + "loss": 1.1027, + "step": 20710 + }, + { + "epoch": 0.5, + "learning_rate": 1.054806690455412e-05, + "loss": 0.9287, + "step": 20711 + }, + { + "epoch": 0.5, + "learning_rate": 1.0547289179188099e-05, + "loss": 0.8107, + "step": 20712 + }, + { + "epoch": 0.5, + "learning_rate": 1.05465114505018e-05, + "loss": 0.9291, + "step": 20713 + }, + { + "epoch": 0.5, + "learning_rate": 1.0545733718499944e-05, + "loss": 0.967, + "step": 20714 + }, + { + "epoch": 0.5, + "learning_rate": 1.0544955983187253e-05, + "loss": 0.9792, + "step": 20715 + }, + { + "epoch": 0.5, + "learning_rate": 1.0544178244568442e-05, + "loss": 1.1378, + "step": 20716 + }, + { + "epoch": 0.5, + "learning_rate": 1.054340050264823e-05, + "loss": 0.8829, + "step": 20717 + }, + { + "epoch": 0.5, + "learning_rate": 1.054262275743133e-05, + "loss": 1.0922, + "step": 20718 + }, + { + "epoch": 0.5, + "learning_rate": 1.0541845008922472e-05, + "loss": 0.8891, + "step": 20719 + }, + { + "epoch": 0.5, + "learning_rate": 1.0541067257126364e-05, + "loss": 1.0982, + "step": 20720 + }, + { + "epoch": 0.5, + "learning_rate": 1.0540289502047729e-05, + "loss": 1.0686, + "step": 20721 + }, + { + "epoch": 0.5, + "learning_rate": 1.0539511743691283e-05, + "loss": 1.1314, + "step": 20722 + }, + { + "epoch": 0.5, + "learning_rate": 1.053873398206175e-05, + "loss": 1.013, + "step": 20723 + }, + { + "epoch": 0.5, + "learning_rate": 1.0537956217163843e-05, + "loss": 1.0961, + "step": 20724 + }, + { + "epoch": 0.5, + "learning_rate": 1.0537178449002283e-05, + "loss": 0.9344, + "step": 20725 + }, + { + "epoch": 0.5, + "learning_rate": 1.0536400677581783e-05, + "loss": 0.7407, + "step": 20726 + }, + { + "epoch": 0.5, + "learning_rate": 1.0535622902907072e-05, + "loss": 0.9838, + "step": 20727 + }, + { + "epoch": 0.5, + "learning_rate": 1.053484512498286e-05, + "loss": 1.012, + "step": 20728 + }, + { + "epoch": 0.5, + "learning_rate": 1.0534067343813868e-05, + "loss": 0.8583, + "step": 20729 + }, + { + "epoch": 0.5, + "learning_rate": 1.0533289559404815e-05, + "loss": 1.0261, + "step": 20730 + }, + { + "epoch": 0.5, + "learning_rate": 1.0532511771760424e-05, + "loss": 1.1248, + "step": 20731 + }, + { + "epoch": 0.5, + "learning_rate": 1.0531733980885406e-05, + "loss": 0.8872, + "step": 20732 + }, + { + "epoch": 0.5, + "learning_rate": 1.053095618678448e-05, + "loss": 1.1478, + "step": 20733 + }, + { + "epoch": 0.5, + "learning_rate": 1.053017838946237e-05, + "loss": 1.0743, + "step": 20734 + }, + { + "epoch": 0.5, + "learning_rate": 1.0529400588923795e-05, + "loss": 1.3202, + "step": 20735 + }, + { + "epoch": 0.5, + "learning_rate": 1.052862278517347e-05, + "loss": 1.1124, + "step": 20736 + }, + { + "epoch": 0.5, + "learning_rate": 1.0527844978216113e-05, + "loss": 0.9978, + "step": 20737 + }, + { + "epoch": 0.5, + "learning_rate": 1.0527067168056447e-05, + "loss": 1.0764, + "step": 20738 + }, + { + "epoch": 0.5, + "learning_rate": 1.0526289354699188e-05, + "loss": 1.1911, + "step": 20739 + }, + { + "epoch": 0.5, + "learning_rate": 1.0525511538149055e-05, + "loss": 1.041, + "step": 20740 + }, + { + "epoch": 0.5, + "learning_rate": 1.0524733718410765e-05, + "loss": 1.1411, + "step": 20741 + }, + { + "epoch": 0.5, + "learning_rate": 1.0523955895489042e-05, + "loss": 1.2643, + "step": 20742 + }, + { + "epoch": 0.5, + "learning_rate": 1.05231780693886e-05, + "loss": 1.0995, + "step": 20743 + }, + { + "epoch": 0.5, + "learning_rate": 1.0522400240114162e-05, + "loss": 1.2447, + "step": 20744 + }, + { + "epoch": 0.5, + "learning_rate": 1.052162240767044e-05, + "loss": 1.1515, + "step": 20745 + }, + { + "epoch": 0.5, + "learning_rate": 1.052084457206216e-05, + "loss": 1.1704, + "step": 20746 + }, + { + "epoch": 0.5, + "learning_rate": 1.052006673329404e-05, + "loss": 1.2231, + "step": 20747 + }, + { + "epoch": 0.5, + "learning_rate": 1.0519288891370799e-05, + "loss": 1.0966, + "step": 20748 + }, + { + "epoch": 0.5, + "learning_rate": 1.0518511046297148e-05, + "loss": 1.2649, + "step": 20749 + }, + { + "epoch": 0.5, + "learning_rate": 1.0517733198077817e-05, + "loss": 1.103, + "step": 20750 + }, + { + "epoch": 0.5, + "learning_rate": 1.051695534671752e-05, + "loss": 1.1839, + "step": 20751 + }, + { + "epoch": 0.5, + "learning_rate": 1.0516177492220978e-05, + "loss": 1.1688, + "step": 20752 + }, + { + "epoch": 0.5, + "learning_rate": 1.0515399634592905e-05, + "loss": 0.6827, + "step": 20753 + }, + { + "epoch": 0.5, + "learning_rate": 1.0514621773838025e-05, + "loss": 0.8317, + "step": 20754 + }, + { + "epoch": 0.5, + "learning_rate": 1.0513843909961056e-05, + "loss": 1.3617, + "step": 20755 + }, + { + "epoch": 0.5, + "learning_rate": 1.051306604296672e-05, + "loss": 1.19, + "step": 20756 + }, + { + "epoch": 0.5, + "learning_rate": 1.0512288172859727e-05, + "loss": 1.1492, + "step": 20757 + }, + { + "epoch": 0.5, + "learning_rate": 1.0511510299644805e-05, + "loss": 1.2288, + "step": 20758 + }, + { + "epoch": 0.5, + "learning_rate": 1.0510732423326672e-05, + "loss": 0.9706, + "step": 20759 + }, + { + "epoch": 0.5, + "learning_rate": 1.0509954543910044e-05, + "loss": 1.1384, + "step": 20760 + }, + { + "epoch": 0.5, + "learning_rate": 1.0509176661399641e-05, + "loss": 1.1784, + "step": 20761 + }, + { + "epoch": 0.5, + "learning_rate": 1.0508398775800184e-05, + "loss": 1.2062, + "step": 20762 + }, + { + "epoch": 0.5, + "learning_rate": 1.0507620887116392e-05, + "loss": 1.2965, + "step": 20763 + }, + { + "epoch": 0.5, + "learning_rate": 1.050684299535298e-05, + "loss": 1.2003, + "step": 20764 + }, + { + "epoch": 0.5, + "learning_rate": 1.0506065100514673e-05, + "loss": 0.9665, + "step": 20765 + }, + { + "epoch": 0.5, + "learning_rate": 1.0505287202606188e-05, + "loss": 1.0013, + "step": 20766 + }, + { + "epoch": 0.5, + "learning_rate": 1.0504509301632246e-05, + "loss": 0.8323, + "step": 20767 + }, + { + "epoch": 0.5, + "learning_rate": 1.0503731397597566e-05, + "loss": 1.0156, + "step": 20768 + }, + { + "epoch": 0.5, + "learning_rate": 1.050295349050686e-05, + "loss": 0.9606, + "step": 20769 + }, + { + "epoch": 0.5, + "learning_rate": 1.0502175580364857e-05, + "loss": 1.0256, + "step": 20770 + }, + { + "epoch": 0.5, + "learning_rate": 1.0501397667176274e-05, + "loss": 1.0946, + "step": 20771 + }, + { + "epoch": 0.5, + "learning_rate": 1.0500619750945829e-05, + "loss": 1.0053, + "step": 20772 + }, + { + "epoch": 0.5, + "learning_rate": 1.0499841831678238e-05, + "loss": 1.1754, + "step": 20773 + }, + { + "epoch": 0.5, + "learning_rate": 1.049906390937823e-05, + "loss": 1.0267, + "step": 20774 + }, + { + "epoch": 0.5, + "learning_rate": 1.0498285984050516e-05, + "loss": 0.945, + "step": 20775 + }, + { + "epoch": 0.5, + "learning_rate": 1.0497508055699814e-05, + "loss": 0.9673, + "step": 20776 + }, + { + "epoch": 0.5, + "learning_rate": 1.0496730124330855e-05, + "loss": 1.1842, + "step": 20777 + }, + { + "epoch": 0.5, + "learning_rate": 1.0495952189948346e-05, + "loss": 0.9822, + "step": 20778 + }, + { + "epoch": 0.5, + "learning_rate": 1.0495174252557015e-05, + "loss": 0.7029, + "step": 20779 + }, + { + "epoch": 0.5, + "learning_rate": 1.0494396312161575e-05, + "loss": 1.0393, + "step": 20780 + }, + { + "epoch": 0.5, + "learning_rate": 1.0493618368766749e-05, + "loss": 1.1745, + "step": 20781 + }, + { + "epoch": 0.5, + "learning_rate": 1.049284042237726e-05, + "loss": 1.0955, + "step": 20782 + }, + { + "epoch": 0.5, + "learning_rate": 1.0492062472997821e-05, + "loss": 1.197, + "step": 20783 + }, + { + "epoch": 0.5, + "learning_rate": 1.0491284520633157e-05, + "loss": 0.9131, + "step": 20784 + }, + { + "epoch": 0.5, + "learning_rate": 1.0490506565287983e-05, + "loss": 1.3097, + "step": 20785 + }, + { + "epoch": 0.5, + "learning_rate": 1.0489728606967022e-05, + "loss": 1.1153, + "step": 20786 + }, + { + "epoch": 0.5, + "learning_rate": 1.0488950645674993e-05, + "loss": 1.3349, + "step": 20787 + }, + { + "epoch": 0.5, + "learning_rate": 1.0488172681416616e-05, + "loss": 1.0899, + "step": 20788 + }, + { + "epoch": 0.5, + "learning_rate": 1.0487394714196608e-05, + "loss": 1.0276, + "step": 20789 + }, + { + "epoch": 0.5, + "learning_rate": 1.0486616744019691e-05, + "loss": 0.9682, + "step": 20790 + }, + { + "epoch": 0.5, + "learning_rate": 1.0485838770890587e-05, + "loss": 1.0044, + "step": 20791 + }, + { + "epoch": 0.5, + "learning_rate": 1.0485060794814013e-05, + "loss": 0.9132, + "step": 20792 + }, + { + "epoch": 0.5, + "learning_rate": 1.0484282815794687e-05, + "loss": 1.0315, + "step": 20793 + }, + { + "epoch": 0.5, + "learning_rate": 1.0483504833837332e-05, + "loss": 1.1048, + "step": 20794 + }, + { + "epoch": 0.5, + "learning_rate": 1.0482726848946669e-05, + "loss": 1.0308, + "step": 20795 + }, + { + "epoch": 0.5, + "learning_rate": 1.0481948861127416e-05, + "loss": 1.2816, + "step": 20796 + }, + { + "epoch": 0.5, + "learning_rate": 1.0481170870384289e-05, + "loss": 1.0652, + "step": 20797 + }, + { + "epoch": 0.5, + "learning_rate": 1.0480392876722015e-05, + "loss": 1.2239, + "step": 20798 + }, + { + "epoch": 0.5, + "learning_rate": 1.047961488014531e-05, + "loss": 0.9891, + "step": 20799 + }, + { + "epoch": 0.5, + "learning_rate": 1.0478836880658894e-05, + "loss": 1.1538, + "step": 20800 + }, + { + "epoch": 0.5, + "learning_rate": 1.0478058878267483e-05, + "loss": 1.0772, + "step": 20801 + }, + { + "epoch": 0.5, + "learning_rate": 1.0477280872975805e-05, + "loss": 0.9162, + "step": 20802 + }, + { + "epoch": 0.5, + "learning_rate": 1.0476502864788578e-05, + "loss": 1.1659, + "step": 20803 + }, + { + "epoch": 0.5, + "learning_rate": 1.0475724853710519e-05, + "loss": 1.0129, + "step": 20804 + }, + { + "epoch": 0.5, + "learning_rate": 1.0474946839746347e-05, + "loss": 1.1006, + "step": 20805 + }, + { + "epoch": 0.5, + "learning_rate": 1.0474168822900785e-05, + "loss": 1.0147, + "step": 20806 + }, + { + "epoch": 0.5, + "learning_rate": 1.0473390803178553e-05, + "loss": 0.9588, + "step": 20807 + }, + { + "epoch": 0.5, + "learning_rate": 1.0472612780584373e-05, + "loss": 0.9914, + "step": 20808 + }, + { + "epoch": 0.5, + "learning_rate": 1.0471834755122959e-05, + "loss": 0.941, + "step": 20809 + }, + { + "epoch": 0.5, + "learning_rate": 1.0471056726799034e-05, + "loss": 1.007, + "step": 20810 + }, + { + "epoch": 0.5, + "learning_rate": 1.0470278695617323e-05, + "loss": 0.7858, + "step": 20811 + }, + { + "epoch": 0.5, + "learning_rate": 1.0469500661582536e-05, + "loss": 1.0153, + "step": 20812 + }, + { + "epoch": 0.5, + "learning_rate": 1.0468722624699401e-05, + "loss": 1.1084, + "step": 20813 + }, + { + "epoch": 0.5, + "learning_rate": 1.0467944584972638e-05, + "loss": 1.2863, + "step": 20814 + }, + { + "epoch": 0.5, + "learning_rate": 1.0467166542406964e-05, + "loss": 1.2371, + "step": 20815 + }, + { + "epoch": 0.5, + "learning_rate": 1.0466388497007098e-05, + "loss": 0.9136, + "step": 20816 + }, + { + "epoch": 0.5, + "learning_rate": 1.0465610448777765e-05, + "loss": 0.8481, + "step": 20817 + }, + { + "epoch": 0.5, + "learning_rate": 1.0464832397723686e-05, + "loss": 1.0222, + "step": 20818 + }, + { + "epoch": 0.5, + "learning_rate": 1.0464054343849575e-05, + "loss": 1.2331, + "step": 20819 + }, + { + "epoch": 0.5, + "learning_rate": 1.0463276287160154e-05, + "loss": 1.2948, + "step": 20820 + }, + { + "epoch": 0.5, + "learning_rate": 1.0462498227660147e-05, + "loss": 1.0206, + "step": 20821 + }, + { + "epoch": 0.5, + "learning_rate": 1.0461720165354272e-05, + "loss": 1.114, + "step": 20822 + }, + { + "epoch": 0.5, + "learning_rate": 1.0460942100247249e-05, + "loss": 1.2374, + "step": 20823 + }, + { + "epoch": 0.5, + "learning_rate": 1.0460164032343797e-05, + "loss": 0.9362, + "step": 20824 + }, + { + "epoch": 0.5, + "learning_rate": 1.045938596164864e-05, + "loss": 1.3638, + "step": 20825 + }, + { + "epoch": 0.5, + "learning_rate": 1.0458607888166496e-05, + "loss": 1.1135, + "step": 20826 + }, + { + "epoch": 0.5, + "learning_rate": 1.0457829811902087e-05, + "loss": 0.9949, + "step": 20827 + }, + { + "epoch": 0.5, + "learning_rate": 1.045705173286013e-05, + "loss": 1.0114, + "step": 20828 + }, + { + "epoch": 0.5, + "learning_rate": 1.0456273651045347e-05, + "loss": 1.1553, + "step": 20829 + }, + { + "epoch": 0.5, + "learning_rate": 1.0455495566462463e-05, + "loss": 1.2792, + "step": 20830 + }, + { + "epoch": 0.5, + "learning_rate": 1.0454717479116192e-05, + "loss": 0.659, + "step": 20831 + }, + { + "epoch": 0.5, + "learning_rate": 1.0453939389011257e-05, + "loss": 1.2725, + "step": 20832 + }, + { + "epoch": 0.5, + "learning_rate": 1.0453161296152378e-05, + "loss": 1.1877, + "step": 20833 + }, + { + "epoch": 0.5, + "learning_rate": 1.0452383200544278e-05, + "loss": 0.9448, + "step": 20834 + }, + { + "epoch": 0.5, + "learning_rate": 1.0451605102191676e-05, + "loss": 1.1926, + "step": 20835 + }, + { + "epoch": 0.5, + "learning_rate": 1.0450827001099288e-05, + "loss": 1.138, + "step": 20836 + }, + { + "epoch": 0.5, + "learning_rate": 1.0450048897271841e-05, + "loss": 1.3073, + "step": 20837 + }, + { + "epoch": 0.5, + "learning_rate": 1.0449270790714055e-05, + "loss": 1.1566, + "step": 20838 + }, + { + "epoch": 0.5, + "learning_rate": 1.0448492681430647e-05, + "loss": 1.3757, + "step": 20839 + }, + { + "epoch": 0.5, + "learning_rate": 1.0447714569426337e-05, + "loss": 1.1635, + "step": 20840 + }, + { + "epoch": 0.5, + "learning_rate": 1.0446936454705851e-05, + "loss": 1.075, + "step": 20841 + }, + { + "epoch": 0.5, + "learning_rate": 1.0446158337273908e-05, + "loss": 0.9988, + "step": 20842 + }, + { + "epoch": 0.5, + "learning_rate": 1.0445380217135226e-05, + "loss": 0.8855, + "step": 20843 + }, + { + "epoch": 0.5, + "learning_rate": 1.0444602094294525e-05, + "loss": 0.864, + "step": 20844 + }, + { + "epoch": 0.5, + "learning_rate": 1.0443823968756528e-05, + "loss": 1.0533, + "step": 20845 + }, + { + "epoch": 0.5, + "learning_rate": 1.0443045840525958e-05, + "loss": 1.2599, + "step": 20846 + }, + { + "epoch": 0.5, + "learning_rate": 1.0442267709607531e-05, + "loss": 1.2919, + "step": 20847 + }, + { + "epoch": 0.5, + "learning_rate": 1.044148957600597e-05, + "loss": 1.0213, + "step": 20848 + }, + { + "epoch": 0.5, + "learning_rate": 1.0440711439725997e-05, + "loss": 1.2155, + "step": 20849 + }, + { + "epoch": 0.5, + "learning_rate": 1.043993330077233e-05, + "loss": 1.1611, + "step": 20850 + }, + { + "epoch": 0.5, + "learning_rate": 1.0439155159149691e-05, + "loss": 1.1547, + "step": 20851 + }, + { + "epoch": 0.5, + "learning_rate": 1.0438377014862801e-05, + "loss": 1.1636, + "step": 20852 + }, + { + "epoch": 0.5, + "learning_rate": 1.0437598867916381e-05, + "loss": 1.1586, + "step": 20853 + }, + { + "epoch": 0.5, + "learning_rate": 1.0436820718315152e-05, + "loss": 1.1034, + "step": 20854 + }, + { + "epoch": 0.5, + "learning_rate": 1.0436042566063836e-05, + "loss": 1.0682, + "step": 20855 + }, + { + "epoch": 0.5, + "learning_rate": 1.0435264411167148e-05, + "loss": 1.0204, + "step": 20856 + }, + { + "epoch": 0.5, + "learning_rate": 1.0434486253629816e-05, + "loss": 1.238, + "step": 20857 + }, + { + "epoch": 0.5, + "learning_rate": 1.0433708093456559e-05, + "loss": 0.9111, + "step": 20858 + }, + { + "epoch": 0.5, + "learning_rate": 1.0432929930652096e-05, + "loss": 1.2086, + "step": 20859 + }, + { + "epoch": 0.5, + "learning_rate": 1.0432151765221148e-05, + "loss": 1.2011, + "step": 20860 + }, + { + "epoch": 0.5, + "learning_rate": 1.0431373597168437e-05, + "loss": 0.9574, + "step": 20861 + }, + { + "epoch": 0.5, + "learning_rate": 1.0430595426498686e-05, + "loss": 1.182, + "step": 20862 + }, + { + "epoch": 0.5, + "learning_rate": 1.042981725321661e-05, + "loss": 1.1504, + "step": 20863 + }, + { + "epoch": 0.5, + "learning_rate": 1.0429039077326937e-05, + "loss": 0.9352, + "step": 20864 + }, + { + "epoch": 0.5, + "learning_rate": 1.0428260898834384e-05, + "loss": 1.146, + "step": 20865 + }, + { + "epoch": 0.5, + "learning_rate": 1.0427482717743673e-05, + "loss": 0.8341, + "step": 20866 + }, + { + "epoch": 0.5, + "learning_rate": 1.0426704534059524e-05, + "loss": 0.9274, + "step": 20867 + }, + { + "epoch": 0.5, + "learning_rate": 1.0425926347786662e-05, + "loss": 1.1552, + "step": 20868 + }, + { + "epoch": 0.5, + "learning_rate": 1.04251481589298e-05, + "loss": 0.9933, + "step": 20869 + }, + { + "epoch": 0.5, + "learning_rate": 1.0424369967493668e-05, + "loss": 1.2259, + "step": 20870 + }, + { + "epoch": 0.5, + "learning_rate": 1.0423591773482981e-05, + "loss": 0.8753, + "step": 20871 + }, + { + "epoch": 0.5, + "learning_rate": 1.0422813576902463e-05, + "loss": 1.4149, + "step": 20872 + }, + { + "epoch": 0.5, + "learning_rate": 1.0422035377756837e-05, + "loss": 1.0645, + "step": 20873 + }, + { + "epoch": 0.5, + "learning_rate": 1.0421257176050818e-05, + "loss": 1.0337, + "step": 20874 + }, + { + "epoch": 0.5, + "learning_rate": 1.0420478971789131e-05, + "loss": 0.9574, + "step": 20875 + }, + { + "epoch": 0.5, + "learning_rate": 1.04197007649765e-05, + "loss": 1.04, + "step": 20876 + }, + { + "epoch": 0.5, + "learning_rate": 1.0418922555617641e-05, + "loss": 1.0819, + "step": 20877 + }, + { + "epoch": 0.5, + "learning_rate": 1.0418144343717278e-05, + "loss": 1.3535, + "step": 20878 + }, + { + "epoch": 0.5, + "learning_rate": 1.0417366129280133e-05, + "loss": 1.3832, + "step": 20879 + }, + { + "epoch": 0.5, + "learning_rate": 1.0416587912310923e-05, + "loss": 0.9771, + "step": 20880 + }, + { + "epoch": 0.5, + "learning_rate": 1.0415809692814374e-05, + "loss": 1.0302, + "step": 20881 + }, + { + "epoch": 0.5, + "learning_rate": 1.0415031470795204e-05, + "loss": 1.1389, + "step": 20882 + }, + { + "epoch": 0.5, + "learning_rate": 1.0414253246258137e-05, + "loss": 1.1129, + "step": 20883 + }, + { + "epoch": 0.5, + "learning_rate": 1.0413475019207891e-05, + "loss": 0.9128, + "step": 20884 + }, + { + "epoch": 0.5, + "learning_rate": 1.0412696789649193e-05, + "loss": 1.1714, + "step": 20885 + }, + { + "epoch": 0.5, + "learning_rate": 1.0411918557586758e-05, + "loss": 1.0281, + "step": 20886 + }, + { + "epoch": 0.5, + "learning_rate": 1.041114032302531e-05, + "loss": 0.7423, + "step": 20887 + }, + { + "epoch": 0.5, + "learning_rate": 1.041036208596957e-05, + "loss": 1.1625, + "step": 20888 + }, + { + "epoch": 0.5, + "learning_rate": 1.0409583846424262e-05, + "loss": 0.9622, + "step": 20889 + }, + { + "epoch": 0.5, + "learning_rate": 1.0408805604394103e-05, + "loss": 1.1577, + "step": 20890 + }, + { + "epoch": 0.5, + "learning_rate": 1.040802735988382e-05, + "loss": 1.0797, + "step": 20891 + }, + { + "epoch": 0.5, + "learning_rate": 1.0407249112898126e-05, + "loss": 0.9599, + "step": 20892 + }, + { + "epoch": 0.5, + "learning_rate": 1.0406470863441748e-05, + "loss": 1.1914, + "step": 20893 + }, + { + "epoch": 0.5, + "learning_rate": 1.0405692611519411e-05, + "loss": 1.0611, + "step": 20894 + }, + { + "epoch": 0.5, + "learning_rate": 1.040491435713583e-05, + "loss": 1.2011, + "step": 20895 + }, + { + "epoch": 0.5, + "learning_rate": 1.0404136100295727e-05, + "loss": 1.1727, + "step": 20896 + }, + { + "epoch": 0.5, + "learning_rate": 1.040335784100383e-05, + "loss": 1.1567, + "step": 20897 + }, + { + "epoch": 0.5, + "learning_rate": 1.0402579579264854e-05, + "loss": 1.1229, + "step": 20898 + }, + { + "epoch": 0.5, + "learning_rate": 1.0401801315083519e-05, + "loss": 1.0888, + "step": 20899 + }, + { + "epoch": 0.5, + "learning_rate": 1.040102304846455e-05, + "loss": 0.7899, + "step": 20900 + }, + { + "epoch": 0.5, + "learning_rate": 1.0400244779412673e-05, + "loss": 1.1744, + "step": 20901 + }, + { + "epoch": 0.5, + "learning_rate": 1.0399466507932603e-05, + "loss": 0.7436, + "step": 20902 + }, + { + "epoch": 0.5, + "learning_rate": 1.0398688234029062e-05, + "loss": 1.036, + "step": 20903 + }, + { + "epoch": 0.5, + "learning_rate": 1.0397909957706776e-05, + "loss": 1.0189, + "step": 20904 + }, + { + "epoch": 0.5, + "learning_rate": 1.0397131678970462e-05, + "loss": 1.0545, + "step": 20905 + }, + { + "epoch": 0.5, + "learning_rate": 1.0396353397824845e-05, + "loss": 1.1294, + "step": 20906 + }, + { + "epoch": 0.5, + "learning_rate": 1.0395575114274644e-05, + "loss": 1.0132, + "step": 20907 + }, + { + "epoch": 0.5, + "learning_rate": 1.0394796828324583e-05, + "loss": 1.0353, + "step": 20908 + }, + { + "epoch": 0.5, + "learning_rate": 1.0394018539979382e-05, + "loss": 1.1084, + "step": 20909 + }, + { + "epoch": 0.5, + "learning_rate": 1.0393240249243763e-05, + "loss": 1.0717, + "step": 20910 + }, + { + "epoch": 0.5, + "learning_rate": 1.0392461956122447e-05, + "loss": 0.9761, + "step": 20911 + }, + { + "epoch": 0.5, + "learning_rate": 1.0391683660620158e-05, + "loss": 1.3616, + "step": 20912 + }, + { + "epoch": 0.5, + "learning_rate": 1.0390905362741615e-05, + "loss": 1.1022, + "step": 20913 + }, + { + "epoch": 0.5, + "learning_rate": 1.0390127062491544e-05, + "loss": 1.0844, + "step": 20914 + }, + { + "epoch": 0.5, + "learning_rate": 1.038934875987466e-05, + "loss": 1.0203, + "step": 20915 + }, + { + "epoch": 0.5, + "learning_rate": 1.038857045489569e-05, + "loss": 1.2291, + "step": 20916 + }, + { + "epoch": 0.5, + "learning_rate": 1.0387792147559355e-05, + "loss": 1.0747, + "step": 20917 + }, + { + "epoch": 0.5, + "learning_rate": 1.0387013837870376e-05, + "loss": 1.2221, + "step": 20918 + }, + { + "epoch": 0.5, + "learning_rate": 1.0386235525833474e-05, + "loss": 1.1357, + "step": 20919 + }, + { + "epoch": 0.5, + "learning_rate": 1.0385457211453374e-05, + "loss": 1.0045, + "step": 20920 + }, + { + "epoch": 0.5, + "learning_rate": 1.0384678894734796e-05, + "loss": 1.1279, + "step": 20921 + }, + { + "epoch": 0.5, + "learning_rate": 1.0383900575682462e-05, + "loss": 1.2215, + "step": 20922 + }, + { + "epoch": 0.5, + "learning_rate": 1.038312225430109e-05, + "loss": 1.1719, + "step": 20923 + }, + { + "epoch": 0.5, + "learning_rate": 1.0382343930595407e-05, + "loss": 1.1792, + "step": 20924 + }, + { + "epoch": 0.5, + "learning_rate": 1.0381565604570134e-05, + "loss": 1.2397, + "step": 20925 + }, + { + "epoch": 0.5, + "learning_rate": 1.0380787276229994e-05, + "loss": 0.7078, + "step": 20926 + }, + { + "epoch": 0.5, + "learning_rate": 1.0380008945579702e-05, + "loss": 1.1806, + "step": 20927 + }, + { + "epoch": 0.5, + "learning_rate": 1.037923061262399e-05, + "loss": 1.0898, + "step": 20928 + }, + { + "epoch": 0.5, + "learning_rate": 1.0378452277367572e-05, + "loss": 1.0515, + "step": 20929 + }, + { + "epoch": 0.5, + "learning_rate": 1.0377673939815176e-05, + "loss": 1.0008, + "step": 20930 + }, + { + "epoch": 0.5, + "learning_rate": 1.0376895599971518e-05, + "loss": 1.1461, + "step": 20931 + }, + { + "epoch": 0.5, + "learning_rate": 1.0376117257841326e-05, + "loss": 0.9818, + "step": 20932 + }, + { + "epoch": 0.5, + "learning_rate": 1.0375338913429318e-05, + "loss": 1.0464, + "step": 20933 + }, + { + "epoch": 0.5, + "learning_rate": 1.0374560566740217e-05, + "loss": 1.0331, + "step": 20934 + }, + { + "epoch": 0.5, + "learning_rate": 1.0373782217778743e-05, + "loss": 1.2009, + "step": 20935 + }, + { + "epoch": 0.5, + "learning_rate": 1.0373003866549622e-05, + "loss": 0.8743, + "step": 20936 + }, + { + "epoch": 0.5, + "learning_rate": 1.0372225513057574e-05, + "loss": 0.9803, + "step": 20937 + }, + { + "epoch": 0.5, + "learning_rate": 1.0371447157307322e-05, + "loss": 0.8192, + "step": 20938 + }, + { + "epoch": 0.5, + "learning_rate": 1.0370668799303587e-05, + "loss": 1.1985, + "step": 20939 + }, + { + "epoch": 0.5, + "learning_rate": 1.0369890439051092e-05, + "loss": 1.233, + "step": 20940 + }, + { + "epoch": 0.5, + "learning_rate": 1.0369112076554556e-05, + "loss": 0.9713, + "step": 20941 + }, + { + "epoch": 0.5, + "learning_rate": 1.036833371181871e-05, + "loss": 1.2483, + "step": 20942 + }, + { + "epoch": 0.5, + "learning_rate": 1.0367555344848264e-05, + "loss": 1.1134, + "step": 20943 + }, + { + "epoch": 0.5, + "learning_rate": 1.0366776975647948e-05, + "loss": 0.8312, + "step": 20944 + }, + { + "epoch": 0.5, + "learning_rate": 1.0365998604222484e-05, + "loss": 1.1671, + "step": 20945 + }, + { + "epoch": 0.5, + "learning_rate": 1.0365220230576592e-05, + "loss": 1.2308, + "step": 20946 + }, + { + "epoch": 0.5, + "learning_rate": 1.0364441854714993e-05, + "loss": 1.2736, + "step": 20947 + }, + { + "epoch": 0.5, + "learning_rate": 1.0363663476642411e-05, + "loss": 1.2607, + "step": 20948 + }, + { + "epoch": 0.5, + "learning_rate": 1.0362885096363572e-05, + "loss": 1.0796, + "step": 20949 + }, + { + "epoch": 0.5, + "learning_rate": 1.0362106713883188e-05, + "loss": 0.9786, + "step": 20950 + }, + { + "epoch": 0.5, + "learning_rate": 1.0361328329205992e-05, + "loss": 1.0831, + "step": 20951 + }, + { + "epoch": 0.5, + "learning_rate": 1.0360549942336701e-05, + "loss": 1.0725, + "step": 20952 + }, + { + "epoch": 0.5, + "learning_rate": 1.035977155328004e-05, + "loss": 0.9536, + "step": 20953 + }, + { + "epoch": 0.5, + "learning_rate": 1.0358993162040728e-05, + "loss": 1.226, + "step": 20954 + }, + { + "epoch": 0.5, + "learning_rate": 1.0358214768623489e-05, + "loss": 0.7656, + "step": 20955 + }, + { + "epoch": 0.5, + "learning_rate": 1.0357436373033048e-05, + "loss": 0.8357, + "step": 20956 + }, + { + "epoch": 0.5, + "learning_rate": 1.0356657975274121e-05, + "loss": 0.7235, + "step": 20957 + }, + { + "epoch": 0.5, + "learning_rate": 1.0355879575351435e-05, + "loss": 1.0823, + "step": 20958 + }, + { + "epoch": 0.5, + "learning_rate": 1.0355101173269708e-05, + "loss": 1.1155, + "step": 20959 + }, + { + "epoch": 0.5, + "learning_rate": 1.0354322769033671e-05, + "loss": 1.3709, + "step": 20960 + }, + { + "epoch": 0.5, + "learning_rate": 1.0353544362648041e-05, + "loss": 1.1154, + "step": 20961 + }, + { + "epoch": 0.5, + "learning_rate": 1.0352765954117539e-05, + "loss": 1.0644, + "step": 20962 + }, + { + "epoch": 0.5, + "learning_rate": 1.0351987543446886e-05, + "loss": 1.2317, + "step": 20963 + }, + { + "epoch": 0.5, + "learning_rate": 1.0351209130640812e-05, + "loss": 0.8979, + "step": 20964 + }, + { + "epoch": 0.5, + "learning_rate": 1.0350430715704032e-05, + "loss": 1.2685, + "step": 20965 + }, + { + "epoch": 0.5, + "learning_rate": 1.0349652298641273e-05, + "loss": 1.0199, + "step": 20966 + }, + { + "epoch": 0.5, + "learning_rate": 1.0348873879457253e-05, + "loss": 1.0426, + "step": 20967 + }, + { + "epoch": 0.5, + "learning_rate": 1.0348095458156699e-05, + "loss": 1.3673, + "step": 20968 + }, + { + "epoch": 0.5, + "learning_rate": 1.0347317034744333e-05, + "loss": 1.2597, + "step": 20969 + }, + { + "epoch": 0.5, + "learning_rate": 1.0346538609224877e-05, + "loss": 1.1375, + "step": 20970 + }, + { + "epoch": 0.5, + "learning_rate": 1.0345760181603048e-05, + "loss": 1.087, + "step": 20971 + }, + { + "epoch": 0.5, + "learning_rate": 1.0344981751883578e-05, + "loss": 1.0333, + "step": 20972 + }, + { + "epoch": 0.5, + "learning_rate": 1.0344203320071182e-05, + "loss": 1.2185, + "step": 20973 + }, + { + "epoch": 0.5, + "learning_rate": 1.0343424886170588e-05, + "loss": 0.8313, + "step": 20974 + }, + { + "epoch": 0.5, + "learning_rate": 1.0342646450186513e-05, + "loss": 1.166, + "step": 20975 + }, + { + "epoch": 0.5, + "learning_rate": 1.0341868012123683e-05, + "loss": 1.1307, + "step": 20976 + }, + { + "epoch": 0.5, + "learning_rate": 1.0341089571986824e-05, + "loss": 1.0939, + "step": 20977 + }, + { + "epoch": 0.5, + "learning_rate": 1.0340311129780652e-05, + "loss": 1.2734, + "step": 20978 + }, + { + "epoch": 0.5, + "learning_rate": 1.033953268550989e-05, + "loss": 1.1416, + "step": 20979 + }, + { + "epoch": 0.5, + "learning_rate": 1.0338754239179267e-05, + "loss": 1.0723, + "step": 20980 + }, + { + "epoch": 0.5, + "learning_rate": 1.03379757907935e-05, + "loss": 1.262, + "step": 20981 + }, + { + "epoch": 0.5, + "learning_rate": 1.0337197340357315e-05, + "loss": 0.9547, + "step": 20982 + }, + { + "epoch": 0.5, + "learning_rate": 1.033641888787543e-05, + "loss": 0.7528, + "step": 20983 + }, + { + "epoch": 0.5, + "learning_rate": 1.0335640433352572e-05, + "loss": 1.2015, + "step": 20984 + }, + { + "epoch": 0.5, + "learning_rate": 1.0334861976793465e-05, + "loss": 1.064, + "step": 20985 + }, + { + "epoch": 0.5, + "learning_rate": 1.0334083518202824e-05, + "loss": 0.963, + "step": 20986 + }, + { + "epoch": 0.5, + "learning_rate": 1.0333305057585381e-05, + "loss": 1.0892, + "step": 20987 + }, + { + "epoch": 0.5, + "learning_rate": 1.0332526594945853e-05, + "loss": 1.0174, + "step": 20988 + }, + { + "epoch": 0.5, + "learning_rate": 1.0331748130288966e-05, + "loss": 1.0701, + "step": 20989 + }, + { + "epoch": 0.5, + "learning_rate": 1.0330969663619438e-05, + "loss": 1.3424, + "step": 20990 + }, + { + "epoch": 0.5, + "learning_rate": 1.0330191194941999e-05, + "loss": 1.1927, + "step": 20991 + }, + { + "epoch": 0.5, + "learning_rate": 1.0329412724261365e-05, + "loss": 1.0963, + "step": 20992 + }, + { + "epoch": 0.5, + "learning_rate": 1.0328634251582264e-05, + "loss": 0.9638, + "step": 20993 + }, + { + "epoch": 0.5, + "learning_rate": 1.0327855776909412e-05, + "loss": 1.0149, + "step": 20994 + }, + { + "epoch": 0.5, + "learning_rate": 1.0327077300247539e-05, + "loss": 0.8245, + "step": 20995 + }, + { + "epoch": 0.5, + "learning_rate": 1.0326298821601363e-05, + "loss": 1.07, + "step": 20996 + }, + { + "epoch": 0.5, + "learning_rate": 1.0325520340975613e-05, + "loss": 1.131, + "step": 20997 + }, + { + "epoch": 0.5, + "learning_rate": 1.0324741858375003e-05, + "loss": 0.9762, + "step": 20998 + }, + { + "epoch": 0.51, + "learning_rate": 1.0323963373804263e-05, + "loss": 1.011, + "step": 20999 + }, + { + "epoch": 0.51, + "learning_rate": 1.0323184887268111e-05, + "loss": 1.0865, + "step": 21000 + }, + { + "epoch": 0.51, + "learning_rate": 1.0322406398771276e-05, + "loss": 0.9758, + "step": 21001 + }, + { + "epoch": 0.51, + "learning_rate": 1.0321627908318475e-05, + "loss": 0.7592, + "step": 21002 + }, + { + "epoch": 0.51, + "learning_rate": 1.0320849415914433e-05, + "loss": 1.2416, + "step": 21003 + }, + { + "epoch": 0.51, + "learning_rate": 1.0320070921563874e-05, + "loss": 1.1615, + "step": 21004 + }, + { + "epoch": 0.51, + "learning_rate": 1.031929242527152e-05, + "loss": 0.9027, + "step": 21005 + }, + { + "epoch": 0.51, + "learning_rate": 1.0318513927042094e-05, + "loss": 0.9421, + "step": 21006 + }, + { + "epoch": 0.51, + "learning_rate": 1.0317735426880318e-05, + "loss": 0.9671, + "step": 21007 + }, + { + "epoch": 0.51, + "learning_rate": 1.0316956924790915e-05, + "loss": 1.2148, + "step": 21008 + }, + { + "epoch": 0.51, + "learning_rate": 1.031617842077861e-05, + "loss": 0.9678, + "step": 21009 + }, + { + "epoch": 0.51, + "learning_rate": 1.0315399914848125e-05, + "loss": 0.9977, + "step": 21010 + }, + { + "epoch": 0.51, + "learning_rate": 1.0314621407004183e-05, + "loss": 1.0008, + "step": 21011 + }, + { + "epoch": 0.51, + "learning_rate": 1.0313842897251509e-05, + "loss": 1.0267, + "step": 21012 + }, + { + "epoch": 0.51, + "learning_rate": 1.0313064385594822e-05, + "loss": 1.1748, + "step": 21013 + }, + { + "epoch": 0.51, + "learning_rate": 1.0312285872038846e-05, + "loss": 1.1465, + "step": 21014 + }, + { + "epoch": 0.51, + "learning_rate": 1.0311507356588306e-05, + "loss": 1.1227, + "step": 21015 + }, + { + "epoch": 0.51, + "learning_rate": 1.0310728839247923e-05, + "loss": 1.0228, + "step": 21016 + }, + { + "epoch": 0.51, + "learning_rate": 1.0309950320022423e-05, + "loss": 1.2345, + "step": 21017 + }, + { + "epoch": 0.51, + "learning_rate": 1.0309171798916526e-05, + "loss": 1.0486, + "step": 21018 + }, + { + "epoch": 0.51, + "learning_rate": 1.0308393275934959e-05, + "loss": 0.8918, + "step": 21019 + }, + { + "epoch": 0.51, + "learning_rate": 1.030761475108244e-05, + "loss": 0.9575, + "step": 21020 + }, + { + "epoch": 0.51, + "learning_rate": 1.0306836224363696e-05, + "loss": 1.0015, + "step": 21021 + }, + { + "epoch": 0.51, + "learning_rate": 1.0306057695783446e-05, + "loss": 0.967, + "step": 21022 + }, + { + "epoch": 0.51, + "learning_rate": 1.0305279165346419e-05, + "loss": 1.2098, + "step": 21023 + }, + { + "epoch": 0.51, + "learning_rate": 1.0304500633057335e-05, + "loss": 1.0345, + "step": 21024 + }, + { + "epoch": 0.51, + "learning_rate": 1.0303722098920917e-05, + "loss": 1.0807, + "step": 21025 + }, + { + "epoch": 0.51, + "learning_rate": 1.0302943562941885e-05, + "loss": 1.0361, + "step": 21026 + }, + { + "epoch": 0.51, + "learning_rate": 1.0302165025124971e-05, + "loss": 1.0715, + "step": 21027 + }, + { + "epoch": 0.51, + "learning_rate": 1.0301386485474888e-05, + "loss": 0.9761, + "step": 21028 + }, + { + "epoch": 0.51, + "learning_rate": 1.0300607943996368e-05, + "loss": 1.0292, + "step": 21029 + }, + { + "epoch": 0.51, + "learning_rate": 1.0299829400694128e-05, + "loss": 1.1446, + "step": 21030 + }, + { + "epoch": 0.51, + "learning_rate": 1.0299050855572894e-05, + "loss": 0.7681, + "step": 21031 + }, + { + "epoch": 0.51, + "learning_rate": 1.0298272308637389e-05, + "loss": 1.0137, + "step": 21032 + }, + { + "epoch": 0.51, + "learning_rate": 1.0297493759892336e-05, + "loss": 1.0973, + "step": 21033 + }, + { + "epoch": 0.51, + "learning_rate": 1.0296715209342458e-05, + "loss": 1.0416, + "step": 21034 + }, + { + "epoch": 0.51, + "learning_rate": 1.029593665699248e-05, + "loss": 1.0656, + "step": 21035 + }, + { + "epoch": 0.51, + "learning_rate": 1.0295158102847124e-05, + "loss": 0.8773, + "step": 21036 + }, + { + "epoch": 0.51, + "learning_rate": 1.029437954691111e-05, + "loss": 0.7685, + "step": 21037 + }, + { + "epoch": 0.51, + "learning_rate": 1.0293600989189168e-05, + "loss": 0.8619, + "step": 21038 + }, + { + "epoch": 0.51, + "learning_rate": 1.0292822429686019e-05, + "loss": 1.0275, + "step": 21039 + }, + { + "epoch": 0.51, + "learning_rate": 1.0292043868406379e-05, + "loss": 1.1125, + "step": 21040 + }, + { + "epoch": 0.51, + "learning_rate": 1.0291265305354983e-05, + "loss": 0.9325, + "step": 21041 + }, + { + "epoch": 0.51, + "learning_rate": 1.0290486740536544e-05, + "loss": 1.1508, + "step": 21042 + }, + { + "epoch": 0.51, + "learning_rate": 1.0289708173955794e-05, + "loss": 1.1003, + "step": 21043 + }, + { + "epoch": 0.51, + "learning_rate": 1.0288929605617454e-05, + "loss": 1.1227, + "step": 21044 + }, + { + "epoch": 0.51, + "learning_rate": 1.0288151035526244e-05, + "loss": 0.9726, + "step": 21045 + }, + { + "epoch": 0.51, + "learning_rate": 1.0287372463686888e-05, + "loss": 1.0968, + "step": 21046 + }, + { + "epoch": 0.51, + "learning_rate": 1.0286593890104113e-05, + "loss": 1.1183, + "step": 21047 + }, + { + "epoch": 0.51, + "learning_rate": 1.028581531478264e-05, + "loss": 0.8855, + "step": 21048 + }, + { + "epoch": 0.51, + "learning_rate": 1.0285036737727193e-05, + "loss": 0.9335, + "step": 21049 + }, + { + "epoch": 0.51, + "learning_rate": 1.0284258158942494e-05, + "loss": 1.1245, + "step": 21050 + }, + { + "epoch": 0.51, + "learning_rate": 1.0283479578433268e-05, + "loss": 1.0055, + "step": 21051 + }, + { + "epoch": 0.51, + "learning_rate": 1.028270099620424e-05, + "loss": 1.098, + "step": 21052 + }, + { + "epoch": 0.51, + "learning_rate": 1.028192241226013e-05, + "loss": 0.986, + "step": 21053 + }, + { + "epoch": 0.51, + "learning_rate": 1.0281143826605662e-05, + "loss": 1.0551, + "step": 21054 + }, + { + "epoch": 0.51, + "learning_rate": 1.0280365239245562e-05, + "loss": 1.1169, + "step": 21055 + }, + { + "epoch": 0.51, + "learning_rate": 1.0279586650184553e-05, + "loss": 1.1618, + "step": 21056 + }, + { + "epoch": 0.51, + "learning_rate": 1.0278808059427356e-05, + "loss": 1.1182, + "step": 21057 + }, + { + "epoch": 0.51, + "learning_rate": 1.0278029466978695e-05, + "loss": 0.8094, + "step": 21058 + }, + { + "epoch": 0.51, + "learning_rate": 1.0277250872843297e-05, + "loss": 1.078, + "step": 21059 + }, + { + "epoch": 0.51, + "learning_rate": 1.0276472277025882e-05, + "loss": 1.1391, + "step": 21060 + }, + { + "epoch": 0.51, + "learning_rate": 1.0275693679531175e-05, + "loss": 0.9975, + "step": 21061 + }, + { + "epoch": 0.51, + "learning_rate": 1.0274915080363896e-05, + "loss": 1.0588, + "step": 21062 + }, + { + "epoch": 0.51, + "learning_rate": 1.0274136479528776e-05, + "loss": 1.2135, + "step": 21063 + }, + { + "epoch": 0.51, + "learning_rate": 1.0273357877030534e-05, + "loss": 1.0121, + "step": 21064 + }, + { + "epoch": 0.51, + "learning_rate": 1.0272579272873895e-05, + "loss": 0.9628, + "step": 21065 + }, + { + "epoch": 0.51, + "learning_rate": 1.0271800667063578e-05, + "loss": 1.0661, + "step": 21066 + }, + { + "epoch": 0.51, + "learning_rate": 1.0271022059604312e-05, + "loss": 1.2987, + "step": 21067 + }, + { + "epoch": 0.51, + "learning_rate": 1.027024345050082e-05, + "loss": 0.8996, + "step": 21068 + }, + { + "epoch": 0.51, + "learning_rate": 1.0269464839757826e-05, + "loss": 1.0061, + "step": 21069 + }, + { + "epoch": 0.51, + "learning_rate": 1.0268686227380047e-05, + "loss": 1.1281, + "step": 21070 + }, + { + "epoch": 0.51, + "learning_rate": 1.0267907613372217e-05, + "loss": 1.3411, + "step": 21071 + }, + { + "epoch": 0.51, + "learning_rate": 1.026712899773905e-05, + "loss": 0.9802, + "step": 21072 + }, + { + "epoch": 0.51, + "learning_rate": 1.0266350380485278e-05, + "loss": 1.0022, + "step": 21073 + }, + { + "epoch": 0.51, + "learning_rate": 1.0265571761615618e-05, + "loss": 0.8952, + "step": 21074 + }, + { + "epoch": 0.51, + "learning_rate": 1.02647931411348e-05, + "loss": 0.9973, + "step": 21075 + }, + { + "epoch": 0.51, + "learning_rate": 1.0264014519047543e-05, + "loss": 1.1695, + "step": 21076 + }, + { + "epoch": 0.51, + "learning_rate": 1.026323589535857e-05, + "loss": 1.0561, + "step": 21077 + }, + { + "epoch": 0.51, + "learning_rate": 1.0262457270072607e-05, + "loss": 1.0807, + "step": 21078 + }, + { + "epoch": 0.51, + "learning_rate": 1.0261678643194379e-05, + "loss": 1.0164, + "step": 21079 + }, + { + "epoch": 0.51, + "learning_rate": 1.0260900014728609e-05, + "loss": 1.1795, + "step": 21080 + }, + { + "epoch": 0.51, + "learning_rate": 1.0260121384680017e-05, + "loss": 1.0911, + "step": 21081 + }, + { + "epoch": 0.51, + "learning_rate": 1.0259342753053332e-05, + "loss": 1.0801, + "step": 21082 + }, + { + "epoch": 0.51, + "learning_rate": 1.0258564119853277e-05, + "loss": 0.7816, + "step": 21083 + }, + { + "epoch": 0.51, + "learning_rate": 1.0257785485084572e-05, + "loss": 0.8605, + "step": 21084 + }, + { + "epoch": 0.51, + "learning_rate": 1.0257006848751943e-05, + "loss": 1.2183, + "step": 21085 + }, + { + "epoch": 0.51, + "learning_rate": 1.0256228210860113e-05, + "loss": 0.9038, + "step": 21086 + }, + { + "epoch": 0.51, + "learning_rate": 1.0255449571413809e-05, + "loss": 1.0717, + "step": 21087 + }, + { + "epoch": 0.51, + "learning_rate": 1.0254670930417752e-05, + "loss": 1.2065, + "step": 21088 + }, + { + "epoch": 0.51, + "learning_rate": 1.0253892287876666e-05, + "loss": 1.0503, + "step": 21089 + }, + { + "epoch": 0.51, + "learning_rate": 1.0253113643795273e-05, + "loss": 0.9659, + "step": 21090 + }, + { + "epoch": 0.51, + "learning_rate": 1.0252334998178304e-05, + "loss": 0.8076, + "step": 21091 + }, + { + "epoch": 0.51, + "learning_rate": 1.0251556351030475e-05, + "loss": 1.3365, + "step": 21092 + }, + { + "epoch": 0.51, + "learning_rate": 1.0250777702356512e-05, + "loss": 1.1628, + "step": 21093 + }, + { + "epoch": 0.51, + "learning_rate": 1.0249999052161142e-05, + "loss": 0.9277, + "step": 21094 + }, + { + "epoch": 0.51, + "learning_rate": 1.0249220400449085e-05, + "loss": 0.8336, + "step": 21095 + }, + { + "epoch": 0.51, + "learning_rate": 1.0248441747225067e-05, + "loss": 0.9941, + "step": 21096 + }, + { + "epoch": 0.51, + "learning_rate": 1.024766309249381e-05, + "loss": 1.0192, + "step": 21097 + }, + { + "epoch": 0.51, + "learning_rate": 1.024688443626004e-05, + "loss": 1.0145, + "step": 21098 + }, + { + "epoch": 0.51, + "learning_rate": 1.024610577852848e-05, + "loss": 1.0272, + "step": 21099 + }, + { + "epoch": 0.51, + "learning_rate": 1.0245327119303856e-05, + "loss": 1.1978, + "step": 21100 + }, + { + "epoch": 0.51, + "learning_rate": 1.0244548458590886e-05, + "loss": 1.2608, + "step": 21101 + }, + { + "epoch": 0.51, + "learning_rate": 1.0243769796394303e-05, + "loss": 1.2055, + "step": 21102 + }, + { + "epoch": 0.51, + "learning_rate": 1.0242991132718823e-05, + "loss": 1.265, + "step": 21103 + }, + { + "epoch": 0.51, + "learning_rate": 1.0242212467569174e-05, + "loss": 0.969, + "step": 21104 + }, + { + "epoch": 0.51, + "learning_rate": 1.0241433800950078e-05, + "loss": 1.1501, + "step": 21105 + }, + { + "epoch": 0.51, + "learning_rate": 1.024065513286626e-05, + "loss": 0.8132, + "step": 21106 + }, + { + "epoch": 0.51, + "learning_rate": 1.0239876463322444e-05, + "loss": 0.9246, + "step": 21107 + }, + { + "epoch": 0.51, + "learning_rate": 1.0239097792323355e-05, + "loss": 0.688, + "step": 21108 + }, + { + "epoch": 0.51, + "learning_rate": 1.0238319119873714e-05, + "loss": 1.1725, + "step": 21109 + }, + { + "epoch": 0.51, + "learning_rate": 1.0237540445978248e-05, + "loss": 1.0272, + "step": 21110 + }, + { + "epoch": 0.51, + "learning_rate": 1.0236761770641682e-05, + "loss": 0.6966, + "step": 21111 + }, + { + "epoch": 0.51, + "learning_rate": 1.0235983093868738e-05, + "loss": 0.785, + "step": 21112 + }, + { + "epoch": 0.51, + "learning_rate": 1.0235204415664135e-05, + "loss": 1.0513, + "step": 21113 + }, + { + "epoch": 0.51, + "learning_rate": 1.0234425736032607e-05, + "loss": 1.0253, + "step": 21114 + }, + { + "epoch": 0.51, + "learning_rate": 1.0233647054978872e-05, + "loss": 0.8954, + "step": 21115 + }, + { + "epoch": 0.51, + "learning_rate": 1.0232868372507656e-05, + "loss": 0.7556, + "step": 21116 + }, + { + "epoch": 0.51, + "learning_rate": 1.023208968862368e-05, + "loss": 1.2528, + "step": 21117 + }, + { + "epoch": 0.51, + "learning_rate": 1.023131100333167e-05, + "loss": 1.0148, + "step": 21118 + }, + { + "epoch": 0.51, + "learning_rate": 1.0230532316636354e-05, + "loss": 1.131, + "step": 21119 + }, + { + "epoch": 0.51, + "learning_rate": 1.022975362854245e-05, + "loss": 1.2603, + "step": 21120 + }, + { + "epoch": 0.51, + "learning_rate": 1.0228974939054685e-05, + "loss": 1.0407, + "step": 21121 + }, + { + "epoch": 0.51, + "learning_rate": 1.0228196248177784e-05, + "loss": 1.2841, + "step": 21122 + }, + { + "epoch": 0.51, + "learning_rate": 1.0227417555916471e-05, + "loss": 1.3219, + "step": 21123 + }, + { + "epoch": 0.51, + "learning_rate": 1.0226638862275466e-05, + "loss": 1.2061, + "step": 21124 + }, + { + "epoch": 0.51, + "learning_rate": 1.0225860167259495e-05, + "loss": 1.0351, + "step": 21125 + }, + { + "epoch": 0.51, + "learning_rate": 1.0225081470873289e-05, + "loss": 0.9792, + "step": 21126 + }, + { + "epoch": 0.51, + "learning_rate": 1.0224302773121564e-05, + "loss": 0.9537, + "step": 21127 + }, + { + "epoch": 0.51, + "learning_rate": 1.0223524074009048e-05, + "loss": 1.2259, + "step": 21128 + }, + { + "epoch": 0.51, + "learning_rate": 1.0222745373540462e-05, + "loss": 1.0418, + "step": 21129 + }, + { + "epoch": 0.51, + "learning_rate": 1.0221966671720534e-05, + "loss": 0.9791, + "step": 21130 + }, + { + "epoch": 0.51, + "learning_rate": 1.0221187968553985e-05, + "loss": 1.0297, + "step": 21131 + }, + { + "epoch": 0.51, + "learning_rate": 1.0220409264045543e-05, + "loss": 1.1948, + "step": 21132 + }, + { + "epoch": 0.51, + "learning_rate": 1.0219630558199927e-05, + "loss": 1.0178, + "step": 21133 + }, + { + "epoch": 0.51, + "learning_rate": 1.0218851851021865e-05, + "loss": 0.8128, + "step": 21134 + }, + { + "epoch": 0.51, + "learning_rate": 1.0218073142516081e-05, + "loss": 0.9001, + "step": 21135 + }, + { + "epoch": 0.51, + "learning_rate": 1.02172944326873e-05, + "loss": 1.0302, + "step": 21136 + }, + { + "epoch": 0.51, + "learning_rate": 1.021651572154024e-05, + "loss": 0.7735, + "step": 21137 + }, + { + "epoch": 0.51, + "learning_rate": 1.0215737009079633e-05, + "loss": 1.0016, + "step": 21138 + }, + { + "epoch": 0.51, + "learning_rate": 1.0214958295310202e-05, + "loss": 1.2501, + "step": 21139 + }, + { + "epoch": 0.51, + "learning_rate": 1.0214179580236667e-05, + "loss": 1.2708, + "step": 21140 + }, + { + "epoch": 0.51, + "learning_rate": 1.0213400863863754e-05, + "loss": 1.0055, + "step": 21141 + }, + { + "epoch": 0.51, + "learning_rate": 1.0212622146196192e-05, + "loss": 0.7674, + "step": 21142 + }, + { + "epoch": 0.51, + "learning_rate": 1.02118434272387e-05, + "loss": 1.0214, + "step": 21143 + }, + { + "epoch": 0.51, + "learning_rate": 1.0211064706996003e-05, + "loss": 1.2218, + "step": 21144 + }, + { + "epoch": 0.51, + "learning_rate": 1.0210285985472824e-05, + "loss": 1.0583, + "step": 21145 + }, + { + "epoch": 0.51, + "learning_rate": 1.0209507262673892e-05, + "loss": 1.0354, + "step": 21146 + }, + { + "epoch": 0.51, + "learning_rate": 1.0208728538603929e-05, + "loss": 1.0822, + "step": 21147 + }, + { + "epoch": 0.51, + "learning_rate": 1.020794981326766e-05, + "loss": 1.071, + "step": 21148 + }, + { + "epoch": 0.51, + "learning_rate": 1.0207171086669806e-05, + "loss": 1.1617, + "step": 21149 + }, + { + "epoch": 0.51, + "learning_rate": 1.0206392358815094e-05, + "loss": 1.2626, + "step": 21150 + }, + { + "epoch": 0.51, + "learning_rate": 1.0205613629708249e-05, + "loss": 1.1329, + "step": 21151 + }, + { + "epoch": 0.51, + "learning_rate": 1.0204834899353992e-05, + "loss": 0.9279, + "step": 21152 + }, + { + "epoch": 0.51, + "learning_rate": 1.0204056167757053e-05, + "loss": 1.1999, + "step": 21153 + }, + { + "epoch": 0.51, + "learning_rate": 1.0203277434922152e-05, + "loss": 1.0023, + "step": 21154 + }, + { + "epoch": 0.51, + "learning_rate": 1.0202498700854016e-05, + "loss": 1.0526, + "step": 21155 + }, + { + "epoch": 0.51, + "learning_rate": 1.0201719965557366e-05, + "loss": 1.3036, + "step": 21156 + }, + { + "epoch": 0.51, + "learning_rate": 1.0200941229036928e-05, + "loss": 1.0602, + "step": 21157 + }, + { + "epoch": 0.51, + "learning_rate": 1.0200162491297428e-05, + "loss": 1.1598, + "step": 21158 + }, + { + "epoch": 0.51, + "learning_rate": 1.0199383752343591e-05, + "loss": 1.2426, + "step": 21159 + }, + { + "epoch": 0.51, + "learning_rate": 1.0198605012180133e-05, + "loss": 0.988, + "step": 21160 + }, + { + "epoch": 0.51, + "learning_rate": 1.0197826270811792e-05, + "loss": 1.1403, + "step": 21161 + }, + { + "epoch": 0.51, + "learning_rate": 1.0197047528243284e-05, + "loss": 1.0144, + "step": 21162 + }, + { + "epoch": 0.51, + "learning_rate": 1.0196268784479335e-05, + "loss": 1.3617, + "step": 21163 + }, + { + "epoch": 0.51, + "learning_rate": 1.0195490039524667e-05, + "loss": 1.1122, + "step": 21164 + }, + { + "epoch": 0.51, + "learning_rate": 1.019471129338401e-05, + "loss": 0.9886, + "step": 21165 + }, + { + "epoch": 0.51, + "learning_rate": 1.0193932546062083e-05, + "loss": 1.1709, + "step": 21166 + }, + { + "epoch": 0.51, + "learning_rate": 1.0193153797563617e-05, + "loss": 1.0276, + "step": 21167 + }, + { + "epoch": 0.51, + "learning_rate": 1.0192375047893327e-05, + "loss": 1.3696, + "step": 21168 + }, + { + "epoch": 0.51, + "learning_rate": 1.0191596297055945e-05, + "loss": 1.2151, + "step": 21169 + }, + { + "epoch": 0.51, + "learning_rate": 1.0190817545056195e-05, + "loss": 1.3254, + "step": 21170 + }, + { + "epoch": 0.51, + "learning_rate": 1.01900387918988e-05, + "loss": 0.9307, + "step": 21171 + }, + { + "epoch": 0.51, + "learning_rate": 1.018926003758848e-05, + "loss": 1.2843, + "step": 21172 + }, + { + "epoch": 0.51, + "learning_rate": 1.0188481282129967e-05, + "loss": 1.045, + "step": 21173 + }, + { + "epoch": 0.51, + "learning_rate": 1.0187702525527983e-05, + "loss": 1.27, + "step": 21174 + }, + { + "epoch": 0.51, + "learning_rate": 1.0186923767787252e-05, + "loss": 1.1047, + "step": 21175 + }, + { + "epoch": 0.51, + "learning_rate": 1.0186145008912496e-05, + "loss": 1.326, + "step": 21176 + }, + { + "epoch": 0.51, + "learning_rate": 1.0185366248908444e-05, + "loss": 1.0944, + "step": 21177 + }, + { + "epoch": 0.51, + "learning_rate": 1.0184587487779818e-05, + "loss": 1.2352, + "step": 21178 + }, + { + "epoch": 0.51, + "learning_rate": 1.0183808725531344e-05, + "loss": 0.9393, + "step": 21179 + }, + { + "epoch": 0.51, + "learning_rate": 1.0183029962167744e-05, + "loss": 1.024, + "step": 21180 + }, + { + "epoch": 0.51, + "learning_rate": 1.0182251197693744e-05, + "loss": 0.9893, + "step": 21181 + }, + { + "epoch": 0.51, + "learning_rate": 1.0181472432114072e-05, + "loss": 1.1824, + "step": 21182 + }, + { + "epoch": 0.51, + "learning_rate": 1.0180693665433447e-05, + "loss": 1.0585, + "step": 21183 + }, + { + "epoch": 0.51, + "learning_rate": 1.0179914897656595e-05, + "loss": 1.0896, + "step": 21184 + }, + { + "epoch": 0.51, + "learning_rate": 1.0179136128788245e-05, + "loss": 0.8508, + "step": 21185 + }, + { + "epoch": 0.51, + "learning_rate": 1.0178357358833117e-05, + "loss": 0.9771, + "step": 21186 + }, + { + "epoch": 0.51, + "learning_rate": 1.0177578587795936e-05, + "loss": 0.8774, + "step": 21187 + }, + { + "epoch": 0.51, + "learning_rate": 1.0176799815681427e-05, + "loss": 1.0617, + "step": 21188 + }, + { + "epoch": 0.51, + "learning_rate": 1.0176021042494314e-05, + "loss": 1.2655, + "step": 21189 + }, + { + "epoch": 0.51, + "learning_rate": 1.0175242268239327e-05, + "loss": 0.9852, + "step": 21190 + }, + { + "epoch": 0.51, + "learning_rate": 1.0174463492921185e-05, + "loss": 0.8561, + "step": 21191 + }, + { + "epoch": 0.51, + "learning_rate": 1.017368471654461e-05, + "loss": 1.1241, + "step": 21192 + }, + { + "epoch": 0.51, + "learning_rate": 1.0172905939114335e-05, + "loss": 1.1968, + "step": 21193 + }, + { + "epoch": 0.51, + "learning_rate": 1.017212716063508e-05, + "loss": 0.9035, + "step": 21194 + }, + { + "epoch": 0.51, + "learning_rate": 1.0171348381111568e-05, + "loss": 1.0848, + "step": 21195 + }, + { + "epoch": 0.51, + "learning_rate": 1.0170569600548525e-05, + "loss": 1.0457, + "step": 21196 + }, + { + "epoch": 0.51, + "learning_rate": 1.016979081895068e-05, + "loss": 1.048, + "step": 21197 + }, + { + "epoch": 0.51, + "learning_rate": 1.0169012036322751e-05, + "loss": 1.0593, + "step": 21198 + }, + { + "epoch": 0.51, + "learning_rate": 1.0168233252669468e-05, + "loss": 1.1572, + "step": 21199 + }, + { + "epoch": 0.51, + "learning_rate": 1.016745446799555e-05, + "loss": 1.0339, + "step": 21200 + }, + { + "epoch": 0.51, + "learning_rate": 1.0166675682305729e-05, + "loss": 1.143, + "step": 21201 + }, + { + "epoch": 0.51, + "learning_rate": 1.0165896895604723e-05, + "loss": 1.1513, + "step": 21202 + }, + { + "epoch": 0.51, + "learning_rate": 1.0165118107897262e-05, + "loss": 1.0796, + "step": 21203 + }, + { + "epoch": 0.51, + "learning_rate": 1.0164339319188066e-05, + "loss": 0.7603, + "step": 21204 + }, + { + "epoch": 0.51, + "learning_rate": 1.0163560529481863e-05, + "loss": 1.1733, + "step": 21205 + }, + { + "epoch": 0.51, + "learning_rate": 1.0162781738783378e-05, + "loss": 1.0203, + "step": 21206 + }, + { + "epoch": 0.51, + "learning_rate": 1.0162002947097331e-05, + "loss": 1.105, + "step": 21207 + }, + { + "epoch": 0.51, + "learning_rate": 1.016122415442845e-05, + "loss": 1.1536, + "step": 21208 + }, + { + "epoch": 0.51, + "learning_rate": 1.0160445360781463e-05, + "loss": 0.8429, + "step": 21209 + }, + { + "epoch": 0.51, + "learning_rate": 1.0159666566161093e-05, + "loss": 1.2033, + "step": 21210 + }, + { + "epoch": 0.51, + "learning_rate": 1.015888777057206e-05, + "loss": 1.1393, + "step": 21211 + }, + { + "epoch": 0.51, + "learning_rate": 1.0158108974019094e-05, + "loss": 1.1385, + "step": 21212 + }, + { + "epoch": 0.51, + "learning_rate": 1.0157330176506918e-05, + "loss": 0.9605, + "step": 21213 + }, + { + "epoch": 0.51, + "learning_rate": 1.0156551378040259e-05, + "loss": 1.0754, + "step": 21214 + }, + { + "epoch": 0.51, + "learning_rate": 1.0155772578623838e-05, + "loss": 1.1262, + "step": 21215 + }, + { + "epoch": 0.51, + "learning_rate": 1.0154993778262379e-05, + "loss": 1.1181, + "step": 21216 + }, + { + "epoch": 0.51, + "learning_rate": 1.015421497696061e-05, + "loss": 0.9672, + "step": 21217 + }, + { + "epoch": 0.51, + "learning_rate": 1.0153436174723256e-05, + "loss": 1.0914, + "step": 21218 + }, + { + "epoch": 0.51, + "learning_rate": 1.0152657371555042e-05, + "loss": 1.1327, + "step": 21219 + }, + { + "epoch": 0.51, + "learning_rate": 1.0151878567460689e-05, + "loss": 1.0149, + "step": 21220 + }, + { + "epoch": 0.51, + "learning_rate": 1.0151099762444928e-05, + "loss": 1.0744, + "step": 21221 + }, + { + "epoch": 0.51, + "learning_rate": 1.0150320956512479e-05, + "loss": 0.9964, + "step": 21222 + }, + { + "epoch": 0.51, + "learning_rate": 1.0149542149668065e-05, + "loss": 1.0391, + "step": 21223 + }, + { + "epoch": 0.51, + "learning_rate": 1.0148763341916417e-05, + "loss": 1.051, + "step": 21224 + }, + { + "epoch": 0.51, + "learning_rate": 1.0147984533262256e-05, + "loss": 0.9439, + "step": 21225 + }, + { + "epoch": 0.51, + "learning_rate": 1.0147205723710307e-05, + "loss": 0.9466, + "step": 21226 + }, + { + "epoch": 0.51, + "learning_rate": 1.0146426913265297e-05, + "loss": 1.1145, + "step": 21227 + }, + { + "epoch": 0.51, + "learning_rate": 1.0145648101931948e-05, + "loss": 1.0254, + "step": 21228 + }, + { + "epoch": 0.51, + "learning_rate": 1.0144869289714987e-05, + "loss": 0.8719, + "step": 21229 + }, + { + "epoch": 0.51, + "learning_rate": 1.014409047661914e-05, + "loss": 0.9806, + "step": 21230 + }, + { + "epoch": 0.51, + "learning_rate": 1.0143311662649126e-05, + "loss": 0.9535, + "step": 21231 + }, + { + "epoch": 0.51, + "learning_rate": 1.0142532847809674e-05, + "loss": 1.0612, + "step": 21232 + }, + { + "epoch": 0.51, + "learning_rate": 1.014175403210551e-05, + "loss": 0.8583, + "step": 21233 + }, + { + "epoch": 0.51, + "learning_rate": 1.014097521554136e-05, + "loss": 0.9176, + "step": 21234 + }, + { + "epoch": 0.51, + "learning_rate": 1.0140196398121947e-05, + "loss": 1.004, + "step": 21235 + }, + { + "epoch": 0.51, + "learning_rate": 1.013941757985199e-05, + "loss": 1.267, + "step": 21236 + }, + { + "epoch": 0.51, + "learning_rate": 1.0138638760736223e-05, + "loss": 1.1725, + "step": 21237 + }, + { + "epoch": 0.51, + "learning_rate": 1.0137859940779367e-05, + "loss": 1.4363, + "step": 21238 + }, + { + "epoch": 0.51, + "learning_rate": 1.0137081119986147e-05, + "loss": 0.9773, + "step": 21239 + }, + { + "epoch": 0.51, + "learning_rate": 1.0136302298361286e-05, + "loss": 1.2119, + "step": 21240 + }, + { + "epoch": 0.51, + "learning_rate": 1.0135523475909515e-05, + "loss": 1.2866, + "step": 21241 + }, + { + "epoch": 0.51, + "learning_rate": 1.0134744652635554e-05, + "loss": 1.117, + "step": 21242 + }, + { + "epoch": 0.51, + "learning_rate": 1.0133965828544128e-05, + "loss": 0.8584, + "step": 21243 + }, + { + "epoch": 0.51, + "learning_rate": 1.0133187003639959e-05, + "loss": 1.145, + "step": 21244 + }, + { + "epoch": 0.51, + "learning_rate": 1.0132408177927782e-05, + "loss": 0.9904, + "step": 21245 + }, + { + "epoch": 0.51, + "learning_rate": 1.0131629351412314e-05, + "loss": 0.9756, + "step": 21246 + }, + { + "epoch": 0.51, + "learning_rate": 1.0130850524098277e-05, + "loss": 1.1015, + "step": 21247 + }, + { + "epoch": 0.51, + "learning_rate": 1.0130071695990404e-05, + "loss": 0.9374, + "step": 21248 + }, + { + "epoch": 0.51, + "learning_rate": 1.012929286709342e-05, + "loss": 0.9704, + "step": 21249 + }, + { + "epoch": 0.51, + "learning_rate": 1.0128514037412042e-05, + "loss": 0.8681, + "step": 21250 + }, + { + "epoch": 0.51, + "learning_rate": 1.0127735206951002e-05, + "loss": 0.9265, + "step": 21251 + }, + { + "epoch": 0.51, + "learning_rate": 1.012695637571502e-05, + "loss": 0.9145, + "step": 21252 + }, + { + "epoch": 0.51, + "learning_rate": 1.0126177543708826e-05, + "loss": 1.1392, + "step": 21253 + }, + { + "epoch": 0.51, + "learning_rate": 1.0125398710937142e-05, + "loss": 1.0496, + "step": 21254 + }, + { + "epoch": 0.51, + "learning_rate": 1.0124619877404694e-05, + "loss": 1.1659, + "step": 21255 + }, + { + "epoch": 0.51, + "learning_rate": 1.0123841043116205e-05, + "loss": 0.667, + "step": 21256 + }, + { + "epoch": 0.51, + "learning_rate": 1.0123062208076402e-05, + "loss": 0.8633, + "step": 21257 + }, + { + "epoch": 0.51, + "learning_rate": 1.0122283372290011e-05, + "loss": 1.209, + "step": 21258 + }, + { + "epoch": 0.51, + "learning_rate": 1.0121504535761753e-05, + "loss": 0.9958, + "step": 21259 + }, + { + "epoch": 0.51, + "learning_rate": 1.0120725698496356e-05, + "loss": 1.0397, + "step": 21260 + }, + { + "epoch": 0.51, + "learning_rate": 1.0119946860498549e-05, + "loss": 0.9104, + "step": 21261 + }, + { + "epoch": 0.51, + "learning_rate": 1.0119168021773047e-05, + "loss": 1.1733, + "step": 21262 + }, + { + "epoch": 0.51, + "learning_rate": 1.0118389182324582e-05, + "loss": 1.2337, + "step": 21263 + }, + { + "epoch": 0.51, + "learning_rate": 1.011761034215788e-05, + "loss": 1.016, + "step": 21264 + }, + { + "epoch": 0.51, + "learning_rate": 1.0116831501277662e-05, + "loss": 1.0313, + "step": 21265 + }, + { + "epoch": 0.51, + "learning_rate": 1.0116052659688655e-05, + "loss": 1.2131, + "step": 21266 + }, + { + "epoch": 0.51, + "learning_rate": 1.0115273817395584e-05, + "loss": 1.3289, + "step": 21267 + }, + { + "epoch": 0.51, + "learning_rate": 1.0114494974403173e-05, + "loss": 0.9892, + "step": 21268 + }, + { + "epoch": 0.51, + "learning_rate": 1.0113716130716149e-05, + "loss": 1.2292, + "step": 21269 + }, + { + "epoch": 0.51, + "learning_rate": 1.0112937286339234e-05, + "loss": 0.9764, + "step": 21270 + }, + { + "epoch": 0.51, + "learning_rate": 1.0112158441277158e-05, + "loss": 1.1406, + "step": 21271 + }, + { + "epoch": 0.51, + "learning_rate": 1.011137959553464e-05, + "loss": 0.9846, + "step": 21272 + }, + { + "epoch": 0.51, + "learning_rate": 1.011060074911641e-05, + "loss": 1.2137, + "step": 21273 + }, + { + "epoch": 0.51, + "learning_rate": 1.0109821902027193e-05, + "loss": 0.954, + "step": 21274 + }, + { + "epoch": 0.51, + "learning_rate": 1.010904305427171e-05, + "loss": 1.1393, + "step": 21275 + }, + { + "epoch": 0.51, + "learning_rate": 1.0108264205854688e-05, + "loss": 1.1533, + "step": 21276 + }, + { + "epoch": 0.51, + "learning_rate": 1.0107485356780855e-05, + "loss": 1.1419, + "step": 21277 + }, + { + "epoch": 0.51, + "learning_rate": 1.0106706507054933e-05, + "loss": 1.1045, + "step": 21278 + }, + { + "epoch": 0.51, + "learning_rate": 1.0105927656681645e-05, + "loss": 1.0103, + "step": 21279 + }, + { + "epoch": 0.51, + "learning_rate": 1.010514880566572e-05, + "loss": 1.2068, + "step": 21280 + }, + { + "epoch": 0.51, + "learning_rate": 1.0104369954011883e-05, + "loss": 1.1213, + "step": 21281 + }, + { + "epoch": 0.51, + "learning_rate": 1.0103591101724859e-05, + "loss": 0.9949, + "step": 21282 + }, + { + "epoch": 0.51, + "learning_rate": 1.0102812248809368e-05, + "loss": 1.0325, + "step": 21283 + }, + { + "epoch": 0.51, + "learning_rate": 1.0102033395270144e-05, + "loss": 0.9317, + "step": 21284 + }, + { + "epoch": 0.51, + "learning_rate": 1.0101254541111904e-05, + "loss": 1.0454, + "step": 21285 + }, + { + "epoch": 0.51, + "learning_rate": 1.010047568633938e-05, + "loss": 1.1812, + "step": 21286 + }, + { + "epoch": 0.51, + "learning_rate": 1.009969683095729e-05, + "loss": 0.9157, + "step": 21287 + }, + { + "epoch": 0.51, + "learning_rate": 1.0098917974970363e-05, + "loss": 1.1626, + "step": 21288 + }, + { + "epoch": 0.51, + "learning_rate": 1.0098139118383328e-05, + "loss": 1.2411, + "step": 21289 + }, + { + "epoch": 0.51, + "learning_rate": 1.0097360261200902e-05, + "loss": 1.2897, + "step": 21290 + }, + { + "epoch": 0.51, + "learning_rate": 1.0096581403427812e-05, + "loss": 1.3492, + "step": 21291 + }, + { + "epoch": 0.51, + "learning_rate": 1.0095802545068791e-05, + "loss": 0.8848, + "step": 21292 + }, + { + "epoch": 0.51, + "learning_rate": 1.0095023686128556e-05, + "loss": 1.0224, + "step": 21293 + }, + { + "epoch": 0.51, + "learning_rate": 1.0094244826611836e-05, + "loss": 1.0397, + "step": 21294 + }, + { + "epoch": 0.51, + "learning_rate": 1.0093465966523353e-05, + "loss": 1.1997, + "step": 21295 + }, + { + "epoch": 0.51, + "learning_rate": 1.0092687105867833e-05, + "loss": 0.9851, + "step": 21296 + }, + { + "epoch": 0.51, + "learning_rate": 1.0091908244650006e-05, + "loss": 0.945, + "step": 21297 + }, + { + "epoch": 0.51, + "learning_rate": 1.0091129382874591e-05, + "loss": 1.124, + "step": 21298 + }, + { + "epoch": 0.51, + "learning_rate": 1.0090350520546314e-05, + "loss": 1.2853, + "step": 21299 + }, + { + "epoch": 0.51, + "learning_rate": 1.0089571657669904e-05, + "loss": 1.2288, + "step": 21300 + }, + { + "epoch": 0.51, + "learning_rate": 1.0088792794250085e-05, + "loss": 1.2532, + "step": 21301 + }, + { + "epoch": 0.51, + "learning_rate": 1.0088013930291578e-05, + "loss": 1.18, + "step": 21302 + }, + { + "epoch": 0.51, + "learning_rate": 1.008723506579911e-05, + "loss": 1.1706, + "step": 21303 + }, + { + "epoch": 0.51, + "learning_rate": 1.0086456200777412e-05, + "loss": 1.2761, + "step": 21304 + }, + { + "epoch": 0.51, + "learning_rate": 1.0085677335231203e-05, + "loss": 1.1309, + "step": 21305 + }, + { + "epoch": 0.51, + "learning_rate": 1.008489846916521e-05, + "loss": 0.8277, + "step": 21306 + }, + { + "epoch": 0.51, + "learning_rate": 1.0084119602584154e-05, + "loss": 0.9191, + "step": 21307 + }, + { + "epoch": 0.51, + "learning_rate": 1.0083340735492769e-05, + "loss": 1.0685, + "step": 21308 + }, + { + "epoch": 0.51, + "learning_rate": 1.0082561867895773e-05, + "loss": 1.1474, + "step": 21309 + }, + { + "epoch": 0.51, + "learning_rate": 1.0081782999797895e-05, + "loss": 0.6426, + "step": 21310 + }, + { + "epoch": 0.51, + "learning_rate": 1.0081004131203855e-05, + "loss": 1.0966, + "step": 21311 + }, + { + "epoch": 0.51, + "learning_rate": 1.0080225262118386e-05, + "loss": 1.0306, + "step": 21312 + }, + { + "epoch": 0.51, + "learning_rate": 1.0079446392546208e-05, + "loss": 1.1825, + "step": 21313 + }, + { + "epoch": 0.51, + "learning_rate": 1.007866752249205e-05, + "loss": 1.172, + "step": 21314 + }, + { + "epoch": 0.51, + "learning_rate": 1.007788865196063e-05, + "loss": 1.0302, + "step": 21315 + }, + { + "epoch": 0.51, + "learning_rate": 1.007710978095668e-05, + "loss": 1.194, + "step": 21316 + }, + { + "epoch": 0.51, + "learning_rate": 1.0076330909484924e-05, + "loss": 1.2257, + "step": 21317 + }, + { + "epoch": 0.51, + "learning_rate": 1.0075552037550084e-05, + "loss": 1.1853, + "step": 21318 + }, + { + "epoch": 0.51, + "learning_rate": 1.0074773165156888e-05, + "loss": 1.1423, + "step": 21319 + }, + { + "epoch": 0.51, + "learning_rate": 1.0073994292310063e-05, + "loss": 1.1115, + "step": 21320 + }, + { + "epoch": 0.51, + "learning_rate": 1.007321541901433e-05, + "loss": 1.2703, + "step": 21321 + }, + { + "epoch": 0.51, + "learning_rate": 1.0072436545274418e-05, + "loss": 1.1822, + "step": 21322 + }, + { + "epoch": 0.51, + "learning_rate": 1.0071657671095047e-05, + "loss": 1.2316, + "step": 21323 + }, + { + "epoch": 0.51, + "learning_rate": 1.0070878796480949e-05, + "loss": 1.0846, + "step": 21324 + }, + { + "epoch": 0.51, + "learning_rate": 1.0070099921436847e-05, + "loss": 1.1786, + "step": 21325 + }, + { + "epoch": 0.51, + "learning_rate": 1.0069321045967464e-05, + "loss": 1.0042, + "step": 21326 + }, + { + "epoch": 0.51, + "learning_rate": 1.0068542170077524e-05, + "loss": 0.9025, + "step": 21327 + }, + { + "epoch": 0.51, + "learning_rate": 1.0067763293771757e-05, + "loss": 1.1782, + "step": 21328 + }, + { + "epoch": 0.51, + "learning_rate": 1.0066984417054888e-05, + "loss": 0.9128, + "step": 21329 + }, + { + "epoch": 0.51, + "learning_rate": 1.0066205539931637e-05, + "loss": 1.0677, + "step": 21330 + }, + { + "epoch": 0.51, + "learning_rate": 1.0065426662406734e-05, + "loss": 0.8308, + "step": 21331 + }, + { + "epoch": 0.51, + "learning_rate": 1.0064647784484903e-05, + "loss": 1.0638, + "step": 21332 + }, + { + "epoch": 0.51, + "learning_rate": 1.0063868906170868e-05, + "loss": 0.9986, + "step": 21333 + }, + { + "epoch": 0.51, + "learning_rate": 1.0063090027469356e-05, + "loss": 1.3306, + "step": 21334 + }, + { + "epoch": 0.51, + "learning_rate": 1.006231114838509e-05, + "loss": 1.0906, + "step": 21335 + }, + { + "epoch": 0.51, + "learning_rate": 1.00615322689228e-05, + "loss": 1.1405, + "step": 21336 + }, + { + "epoch": 0.51, + "learning_rate": 1.0060753389087208e-05, + "loss": 1.1803, + "step": 21337 + }, + { + "epoch": 0.51, + "learning_rate": 1.0059974508883037e-05, + "loss": 1.1069, + "step": 21338 + }, + { + "epoch": 0.51, + "learning_rate": 1.0059195628315017e-05, + "loss": 1.1019, + "step": 21339 + }, + { + "epoch": 0.51, + "learning_rate": 1.0058416747387872e-05, + "loss": 0.7464, + "step": 21340 + }, + { + "epoch": 0.51, + "learning_rate": 1.0057637866106325e-05, + "loss": 1.0657, + "step": 21341 + }, + { + "epoch": 0.51, + "learning_rate": 1.00568589844751e-05, + "loss": 1.0507, + "step": 21342 + }, + { + "epoch": 0.51, + "learning_rate": 1.0056080102498925e-05, + "loss": 1.2042, + "step": 21343 + }, + { + "epoch": 0.51, + "learning_rate": 1.005530122018253e-05, + "loss": 1.1121, + "step": 21344 + }, + { + "epoch": 0.51, + "learning_rate": 1.0054522337530632e-05, + "loss": 1.1722, + "step": 21345 + }, + { + "epoch": 0.51, + "learning_rate": 1.005374345454796e-05, + "loss": 1.1114, + "step": 21346 + }, + { + "epoch": 0.51, + "learning_rate": 1.0052964571239238e-05, + "loss": 1.1444, + "step": 21347 + }, + { + "epoch": 0.51, + "learning_rate": 1.0052185687609197e-05, + "loss": 0.9542, + "step": 21348 + }, + { + "epoch": 0.51, + "learning_rate": 1.0051406803662556e-05, + "loss": 1.1616, + "step": 21349 + }, + { + "epoch": 0.51, + "learning_rate": 1.0050627919404037e-05, + "loss": 1.0425, + "step": 21350 + }, + { + "epoch": 0.51, + "learning_rate": 1.0049849034838374e-05, + "loss": 0.8462, + "step": 21351 + }, + { + "epoch": 0.51, + "learning_rate": 1.004907014997029e-05, + "loss": 1.1688, + "step": 21352 + }, + { + "epoch": 0.51, + "learning_rate": 1.0048291264804509e-05, + "loss": 1.0261, + "step": 21353 + }, + { + "epoch": 0.51, + "learning_rate": 1.0047512379345755e-05, + "loss": 1.1095, + "step": 21354 + }, + { + "epoch": 0.51, + "learning_rate": 1.0046733493598753e-05, + "loss": 0.8718, + "step": 21355 + }, + { + "epoch": 0.51, + "learning_rate": 1.0045954607568231e-05, + "loss": 0.924, + "step": 21356 + }, + { + "epoch": 0.51, + "learning_rate": 1.0045175721258915e-05, + "loss": 1.0496, + "step": 21357 + }, + { + "epoch": 0.51, + "learning_rate": 1.0044396834675525e-05, + "loss": 1.1291, + "step": 21358 + }, + { + "epoch": 0.51, + "learning_rate": 1.0043617947822793e-05, + "loss": 1.2922, + "step": 21359 + }, + { + "epoch": 0.51, + "learning_rate": 1.004283906070544e-05, + "loss": 0.9871, + "step": 21360 + }, + { + "epoch": 0.51, + "learning_rate": 1.0042060173328195e-05, + "loss": 1.3701, + "step": 21361 + }, + { + "epoch": 0.51, + "learning_rate": 1.0041281285695776e-05, + "loss": 1.1172, + "step": 21362 + }, + { + "epoch": 0.51, + "learning_rate": 1.0040502397812917e-05, + "loss": 1.0567, + "step": 21363 + }, + { + "epoch": 0.51, + "learning_rate": 1.0039723509684338e-05, + "loss": 0.9307, + "step": 21364 + }, + { + "epoch": 0.51, + "learning_rate": 1.0038944621314766e-05, + "loss": 0.7172, + "step": 21365 + }, + { + "epoch": 0.51, + "learning_rate": 1.0038165732708925e-05, + "loss": 0.8711, + "step": 21366 + }, + { + "epoch": 0.51, + "learning_rate": 1.0037386843871542e-05, + "loss": 1.2483, + "step": 21367 + }, + { + "epoch": 0.51, + "learning_rate": 1.0036607954807343e-05, + "loss": 0.9108, + "step": 21368 + }, + { + "epoch": 0.51, + "learning_rate": 1.0035829065521052e-05, + "loss": 1.222, + "step": 21369 + }, + { + "epoch": 0.51, + "learning_rate": 1.0035050176017392e-05, + "loss": 1.2148, + "step": 21370 + }, + { + "epoch": 0.51, + "learning_rate": 1.0034271286301094e-05, + "loss": 1.0883, + "step": 21371 + }, + { + "epoch": 0.51, + "learning_rate": 1.0033492396376879e-05, + "loss": 1.0854, + "step": 21372 + }, + { + "epoch": 0.51, + "learning_rate": 1.0032713506249473e-05, + "loss": 1.2201, + "step": 21373 + }, + { + "epoch": 0.51, + "learning_rate": 1.00319346159236e-05, + "loss": 0.983, + "step": 21374 + }, + { + "epoch": 0.51, + "learning_rate": 1.003115572540399e-05, + "loss": 1.1118, + "step": 21375 + }, + { + "epoch": 0.51, + "learning_rate": 1.0030376834695366e-05, + "loss": 0.9436, + "step": 21376 + }, + { + "epoch": 0.51, + "learning_rate": 1.0029597943802451e-05, + "loss": 1.2069, + "step": 21377 + }, + { + "epoch": 0.51, + "learning_rate": 1.0028819052729973e-05, + "loss": 1.2328, + "step": 21378 + }, + { + "epoch": 0.51, + "learning_rate": 1.0028040161482658e-05, + "loss": 0.9774, + "step": 21379 + }, + { + "epoch": 0.51, + "learning_rate": 1.0027261270065227e-05, + "loss": 1.1354, + "step": 21380 + }, + { + "epoch": 0.51, + "learning_rate": 1.0026482378482412e-05, + "loss": 0.6908, + "step": 21381 + }, + { + "epoch": 0.51, + "learning_rate": 1.002570348673893e-05, + "loss": 1.2386, + "step": 21382 + }, + { + "epoch": 0.51, + "learning_rate": 1.0024924594839514e-05, + "loss": 1.2517, + "step": 21383 + }, + { + "epoch": 0.51, + "learning_rate": 1.0024145702788887e-05, + "loss": 0.9987, + "step": 21384 + }, + { + "epoch": 0.51, + "learning_rate": 1.0023366810591772e-05, + "loss": 1.0785, + "step": 21385 + }, + { + "epoch": 0.51, + "learning_rate": 1.0022587918252894e-05, + "loss": 0.9908, + "step": 21386 + }, + { + "epoch": 0.51, + "learning_rate": 1.0021809025776985e-05, + "loss": 0.9409, + "step": 21387 + }, + { + "epoch": 0.51, + "learning_rate": 1.0021030133168765e-05, + "loss": 1.1665, + "step": 21388 + }, + { + "epoch": 0.51, + "learning_rate": 1.0020251240432959e-05, + "loss": 1.2369, + "step": 21389 + }, + { + "epoch": 0.51, + "learning_rate": 1.0019472347574293e-05, + "loss": 1.0445, + "step": 21390 + }, + { + "epoch": 0.51, + "learning_rate": 1.0018693454597492e-05, + "loss": 0.9913, + "step": 21391 + }, + { + "epoch": 0.51, + "learning_rate": 1.0017914561507285e-05, + "loss": 1.017, + "step": 21392 + }, + { + "epoch": 0.51, + "learning_rate": 1.0017135668308392e-05, + "loss": 0.9341, + "step": 21393 + }, + { + "epoch": 0.51, + "learning_rate": 1.0016356775005542e-05, + "loss": 0.9195, + "step": 21394 + }, + { + "epoch": 0.51, + "learning_rate": 1.001557788160346e-05, + "loss": 0.9757, + "step": 21395 + }, + { + "epoch": 0.51, + "learning_rate": 1.0014798988106871e-05, + "loss": 1.4087, + "step": 21396 + }, + { + "epoch": 0.51, + "learning_rate": 1.00140200945205e-05, + "loss": 1.0186, + "step": 21397 + }, + { + "epoch": 0.51, + "learning_rate": 1.0013241200849069e-05, + "loss": 1.1033, + "step": 21398 + }, + { + "epoch": 0.51, + "learning_rate": 1.001246230709731e-05, + "loss": 0.9873, + "step": 21399 + }, + { + "epoch": 0.51, + "learning_rate": 1.0011683413269946e-05, + "loss": 1.3786, + "step": 21400 + }, + { + "epoch": 0.51, + "learning_rate": 1.0010904519371698e-05, + "loss": 1.0793, + "step": 21401 + }, + { + "epoch": 0.51, + "learning_rate": 1.0010125625407296e-05, + "loss": 1.1609, + "step": 21402 + }, + { + "epoch": 0.51, + "learning_rate": 1.0009346731381466e-05, + "loss": 1.2488, + "step": 21403 + }, + { + "epoch": 0.51, + "learning_rate": 1.000856783729893e-05, + "loss": 0.9818, + "step": 21404 + }, + { + "epoch": 0.51, + "learning_rate": 1.0007788943164415e-05, + "loss": 1.1409, + "step": 21405 + }, + { + "epoch": 0.51, + "learning_rate": 1.0007010048982646e-05, + "loss": 1.1691, + "step": 21406 + }, + { + "epoch": 0.51, + "learning_rate": 1.000623115475835e-05, + "loss": 1.0571, + "step": 21407 + }, + { + "epoch": 0.51, + "learning_rate": 1.0005452260496251e-05, + "loss": 1.1482, + "step": 21408 + }, + { + "epoch": 0.51, + "learning_rate": 1.0004673366201075e-05, + "loss": 1.2882, + "step": 21409 + }, + { + "epoch": 0.51, + "learning_rate": 1.0003894471877545e-05, + "loss": 1.0725, + "step": 21410 + }, + { + "epoch": 0.51, + "learning_rate": 1.0003115577530388e-05, + "loss": 0.8922, + "step": 21411 + }, + { + "epoch": 0.51, + "learning_rate": 1.0002336683164331e-05, + "loss": 1.3469, + "step": 21412 + }, + { + "epoch": 0.51, + "learning_rate": 1.0001557788784097e-05, + "loss": 1.0297, + "step": 21413 + }, + { + "epoch": 0.51, + "learning_rate": 1.0000778894394412e-05, + "loss": 0.9852, + "step": 21414 + }, + { + "epoch": 0.52, + "learning_rate": 1e-05, + "loss": 1.1453, + "step": 21415 + }, + { + "epoch": 0.52, + "learning_rate": 9.999221105605593e-06, + "loss": 1.0404, + "step": 21416 + }, + { + "epoch": 0.52, + "learning_rate": 9.998442211215906e-06, + "loss": 1.0653, + "step": 21417 + }, + { + "epoch": 0.52, + "learning_rate": 9.997663316835674e-06, + "loss": 1.1842, + "step": 21418 + }, + { + "epoch": 0.52, + "learning_rate": 9.996884422469614e-06, + "loss": 0.9388, + "step": 21419 + }, + { + "epoch": 0.52, + "learning_rate": 9.99610552812246e-06, + "loss": 1.0386, + "step": 21420 + }, + { + "epoch": 0.52, + "learning_rate": 9.995326633798928e-06, + "loss": 1.3168, + "step": 21421 + }, + { + "epoch": 0.52, + "learning_rate": 9.994547739503752e-06, + "loss": 1.0335, + "step": 21422 + }, + { + "epoch": 0.52, + "learning_rate": 9.993768845241652e-06, + "loss": 1.2319, + "step": 21423 + }, + { + "epoch": 0.52, + "learning_rate": 9.992989951017357e-06, + "loss": 0.9733, + "step": 21424 + }, + { + "epoch": 0.52, + "learning_rate": 9.992211056835586e-06, + "loss": 1.0843, + "step": 21425 + }, + { + "epoch": 0.52, + "learning_rate": 9.991432162701073e-06, + "loss": 1.2156, + "step": 21426 + }, + { + "epoch": 0.52, + "learning_rate": 9.990653268618538e-06, + "loss": 1.0475, + "step": 21427 + }, + { + "epoch": 0.52, + "learning_rate": 9.989874374592707e-06, + "loss": 1.0744, + "step": 21428 + }, + { + "epoch": 0.52, + "learning_rate": 9.989095480628304e-06, + "loss": 1.1093, + "step": 21429 + }, + { + "epoch": 0.52, + "learning_rate": 9.988316586730057e-06, + "loss": 0.838, + "step": 21430 + }, + { + "epoch": 0.52, + "learning_rate": 9.987537692902692e-06, + "loss": 1.1553, + "step": 21431 + }, + { + "epoch": 0.52, + "learning_rate": 9.986758799150934e-06, + "loss": 1.0482, + "step": 21432 + }, + { + "epoch": 0.52, + "learning_rate": 9.985979905479503e-06, + "loss": 1.1812, + "step": 21433 + }, + { + "epoch": 0.52, + "learning_rate": 9.985201011893132e-06, + "loss": 1.1067, + "step": 21434 + }, + { + "epoch": 0.52, + "learning_rate": 9.984422118396543e-06, + "loss": 1.1644, + "step": 21435 + }, + { + "epoch": 0.52, + "learning_rate": 9.98364322499446e-06, + "loss": 1.0453, + "step": 21436 + }, + { + "epoch": 0.52, + "learning_rate": 9.98286433169161e-06, + "loss": 0.9997, + "step": 21437 + }, + { + "epoch": 0.52, + "learning_rate": 9.982085438492719e-06, + "loss": 1.155, + "step": 21438 + }, + { + "epoch": 0.52, + "learning_rate": 9.98130654540251e-06, + "loss": 1.0795, + "step": 21439 + }, + { + "epoch": 0.52, + "learning_rate": 9.98052765242571e-06, + "loss": 1.0095, + "step": 21440 + }, + { + "epoch": 0.52, + "learning_rate": 9.979748759567043e-06, + "loss": 1.169, + "step": 21441 + }, + { + "epoch": 0.52, + "learning_rate": 9.978969866831237e-06, + "loss": 1.1724, + "step": 21442 + }, + { + "epoch": 0.52, + "learning_rate": 9.978190974223018e-06, + "loss": 1.4061, + "step": 21443 + }, + { + "epoch": 0.52, + "learning_rate": 9.977412081747107e-06, + "loss": 1.0502, + "step": 21444 + }, + { + "epoch": 0.52, + "learning_rate": 9.97663318940823e-06, + "loss": 0.9641, + "step": 21445 + }, + { + "epoch": 0.52, + "learning_rate": 9.975854297211116e-06, + "loss": 1.0613, + "step": 21446 + }, + { + "epoch": 0.52, + "learning_rate": 9.975075405160487e-06, + "loss": 1.1158, + "step": 21447 + }, + { + "epoch": 0.52, + "learning_rate": 9.974296513261072e-06, + "loss": 0.896, + "step": 21448 + }, + { + "epoch": 0.52, + "learning_rate": 9.973517621517593e-06, + "loss": 1.2289, + "step": 21449 + }, + { + "epoch": 0.52, + "learning_rate": 9.972738729934775e-06, + "loss": 1.0554, + "step": 21450 + }, + { + "epoch": 0.52, + "learning_rate": 9.971959838517347e-06, + "loss": 0.9225, + "step": 21451 + }, + { + "epoch": 0.52, + "learning_rate": 9.97118094727003e-06, + "loss": 1.0608, + "step": 21452 + }, + { + "epoch": 0.52, + "learning_rate": 9.970402056197552e-06, + "loss": 1.0969, + "step": 21453 + }, + { + "epoch": 0.52, + "learning_rate": 9.969623165304639e-06, + "loss": 0.9729, + "step": 21454 + }, + { + "epoch": 0.52, + "learning_rate": 9.968844274596014e-06, + "loss": 1.091, + "step": 21455 + }, + { + "epoch": 0.52, + "learning_rate": 9.968065384076403e-06, + "loss": 1.1101, + "step": 21456 + }, + { + "epoch": 0.52, + "learning_rate": 9.967286493750532e-06, + "loss": 1.4156, + "step": 21457 + }, + { + "epoch": 0.52, + "learning_rate": 9.966507603623125e-06, + "loss": 1.0295, + "step": 21458 + }, + { + "epoch": 0.52, + "learning_rate": 9.965728713698911e-06, + "loss": 1.1789, + "step": 21459 + }, + { + "epoch": 0.52, + "learning_rate": 9.964949823982613e-06, + "loss": 1.1285, + "step": 21460 + }, + { + "epoch": 0.52, + "learning_rate": 9.964170934478953e-06, + "loss": 1.2454, + "step": 21461 + }, + { + "epoch": 0.52, + "learning_rate": 9.963392045192662e-06, + "loss": 0.956, + "step": 21462 + }, + { + "epoch": 0.52, + "learning_rate": 9.962613156128463e-06, + "loss": 1.1203, + "step": 21463 + }, + { + "epoch": 0.52, + "learning_rate": 9.961834267291079e-06, + "loss": 0.9072, + "step": 21464 + }, + { + "epoch": 0.52, + "learning_rate": 9.96105537868524e-06, + "loss": 1.1524, + "step": 21465 + }, + { + "epoch": 0.52, + "learning_rate": 9.960276490315668e-06, + "loss": 0.9538, + "step": 21466 + }, + { + "epoch": 0.52, + "learning_rate": 9.95949760218709e-06, + "loss": 1.2432, + "step": 21467 + }, + { + "epoch": 0.52, + "learning_rate": 9.958718714304227e-06, + "loss": 0.9909, + "step": 21468 + }, + { + "epoch": 0.52, + "learning_rate": 9.95793982667181e-06, + "loss": 1.2263, + "step": 21469 + }, + { + "epoch": 0.52, + "learning_rate": 9.957160939294565e-06, + "loss": 1.1934, + "step": 21470 + }, + { + "epoch": 0.52, + "learning_rate": 9.956382052177212e-06, + "loss": 1.1331, + "step": 21471 + }, + { + "epoch": 0.52, + "learning_rate": 9.955603165324478e-06, + "loss": 1.1263, + "step": 21472 + }, + { + "epoch": 0.52, + "learning_rate": 9.95482427874109e-06, + "loss": 1.2239, + "step": 21473 + }, + { + "epoch": 0.52, + "learning_rate": 9.954045392431772e-06, + "loss": 1.0394, + "step": 21474 + }, + { + "epoch": 0.52, + "learning_rate": 9.953266506401252e-06, + "loss": 1.2826, + "step": 21475 + }, + { + "epoch": 0.52, + "learning_rate": 9.95248762065425e-06, + "loss": 1.1822, + "step": 21476 + }, + { + "epoch": 0.52, + "learning_rate": 9.951708735195496e-06, + "loss": 1.1895, + "step": 21477 + }, + { + "epoch": 0.52, + "learning_rate": 9.950929850029715e-06, + "loss": 1.0259, + "step": 21478 + }, + { + "epoch": 0.52, + "learning_rate": 9.95015096516163e-06, + "loss": 1.293, + "step": 21479 + }, + { + "epoch": 0.52, + "learning_rate": 9.949372080595964e-06, + "loss": 0.9618, + "step": 21480 + }, + { + "epoch": 0.52, + "learning_rate": 9.94859319633745e-06, + "loss": 1.0945, + "step": 21481 + }, + { + "epoch": 0.52, + "learning_rate": 9.94781431239081e-06, + "loss": 1.1723, + "step": 21482 + }, + { + "epoch": 0.52, + "learning_rate": 9.947035428760765e-06, + "loss": 1.0823, + "step": 21483 + }, + { + "epoch": 0.52, + "learning_rate": 9.946256545452043e-06, + "loss": 1.2988, + "step": 21484 + }, + { + "epoch": 0.52, + "learning_rate": 9.945477662469373e-06, + "loss": 1.2802, + "step": 21485 + }, + { + "epoch": 0.52, + "learning_rate": 9.944698779817475e-06, + "loss": 1.2957, + "step": 21486 + }, + { + "epoch": 0.52, + "learning_rate": 9.943919897501078e-06, + "loss": 1.1347, + "step": 21487 + }, + { + "epoch": 0.52, + "learning_rate": 9.943141015524905e-06, + "loss": 1.2033, + "step": 21488 + }, + { + "epoch": 0.52, + "learning_rate": 9.942362133893682e-06, + "loss": 1.0396, + "step": 21489 + }, + { + "epoch": 0.52, + "learning_rate": 9.941583252612135e-06, + "loss": 1.2414, + "step": 21490 + }, + { + "epoch": 0.52, + "learning_rate": 9.940804371684988e-06, + "loss": 1.2027, + "step": 21491 + }, + { + "epoch": 0.52, + "learning_rate": 9.940025491116965e-06, + "loss": 1.0457, + "step": 21492 + }, + { + "epoch": 0.52, + "learning_rate": 9.939246610912797e-06, + "loss": 1.1895, + "step": 21493 + }, + { + "epoch": 0.52, + "learning_rate": 9.938467731077204e-06, + "loss": 0.878, + "step": 21494 + }, + { + "epoch": 0.52, + "learning_rate": 9.937688851614913e-06, + "loss": 0.9438, + "step": 21495 + }, + { + "epoch": 0.52, + "learning_rate": 9.936909972530647e-06, + "loss": 1.1846, + "step": 21496 + }, + { + "epoch": 0.52, + "learning_rate": 9.936131093829136e-06, + "loss": 1.0908, + "step": 21497 + }, + { + "epoch": 0.52, + "learning_rate": 9.9353522155151e-06, + "loss": 1.1895, + "step": 21498 + }, + { + "epoch": 0.52, + "learning_rate": 9.934573337593268e-06, + "loss": 0.8229, + "step": 21499 + }, + { + "epoch": 0.52, + "learning_rate": 9.933794460068364e-06, + "loss": 1.1362, + "step": 21500 + }, + { + "epoch": 0.52, + "learning_rate": 9.933015582945112e-06, + "loss": 1.0776, + "step": 21501 + }, + { + "epoch": 0.52, + "learning_rate": 9.932236706228243e-06, + "loss": 1.1518, + "step": 21502 + }, + { + "epoch": 0.52, + "learning_rate": 9.931457829922476e-06, + "loss": 1.0459, + "step": 21503 + }, + { + "epoch": 0.52, + "learning_rate": 9.930678954032539e-06, + "loss": 1.1177, + "step": 21504 + }, + { + "epoch": 0.52, + "learning_rate": 9.929900078563154e-06, + "loss": 1.1346, + "step": 21505 + }, + { + "epoch": 0.52, + "learning_rate": 9.92912120351905e-06, + "loss": 1.1129, + "step": 21506 + }, + { + "epoch": 0.52, + "learning_rate": 9.928342328904951e-06, + "loss": 1.0804, + "step": 21507 + }, + { + "epoch": 0.52, + "learning_rate": 9.927563454725584e-06, + "loss": 1.0308, + "step": 21508 + }, + { + "epoch": 0.52, + "learning_rate": 9.92678458098567e-06, + "loss": 1.166, + "step": 21509 + }, + { + "epoch": 0.52, + "learning_rate": 9.926005707689937e-06, + "loss": 1.2006, + "step": 21510 + }, + { + "epoch": 0.52, + "learning_rate": 9.925226834843112e-06, + "loss": 1.1629, + "step": 21511 + }, + { + "epoch": 0.52, + "learning_rate": 9.924447962449917e-06, + "loss": 0.9484, + "step": 21512 + }, + { + "epoch": 0.52, + "learning_rate": 9.923669090515077e-06, + "loss": 1.0599, + "step": 21513 + }, + { + "epoch": 0.52, + "learning_rate": 9.92289021904332e-06, + "loss": 1.1681, + "step": 21514 + }, + { + "epoch": 0.52, + "learning_rate": 9.922111348039371e-06, + "loss": 1.1036, + "step": 21515 + }, + { + "epoch": 0.52, + "learning_rate": 9.921332477507955e-06, + "loss": 1.0533, + "step": 21516 + }, + { + "epoch": 0.52, + "learning_rate": 9.920553607453792e-06, + "loss": 1.0926, + "step": 21517 + }, + { + "epoch": 0.52, + "learning_rate": 9.919774737881615e-06, + "loss": 1.2489, + "step": 21518 + }, + { + "epoch": 0.52, + "learning_rate": 9.918995868796145e-06, + "loss": 1.043, + "step": 21519 + }, + { + "epoch": 0.52, + "learning_rate": 9.918217000202108e-06, + "loss": 0.8201, + "step": 21520 + }, + { + "epoch": 0.52, + "learning_rate": 9.917438132104227e-06, + "loss": 1.1571, + "step": 21521 + }, + { + "epoch": 0.52, + "learning_rate": 9.916659264507233e-06, + "loss": 0.9811, + "step": 21522 + }, + { + "epoch": 0.52, + "learning_rate": 9.915880397415848e-06, + "loss": 0.9423, + "step": 21523 + }, + { + "epoch": 0.52, + "learning_rate": 9.915101530834795e-06, + "loss": 1.1725, + "step": 21524 + }, + { + "epoch": 0.52, + "learning_rate": 9.914322664768799e-06, + "loss": 1.2212, + "step": 21525 + }, + { + "epoch": 0.52, + "learning_rate": 9.91354379922259e-06, + "loss": 0.8837, + "step": 21526 + }, + { + "epoch": 0.52, + "learning_rate": 9.912764934200889e-06, + "loss": 0.8723, + "step": 21527 + }, + { + "epoch": 0.52, + "learning_rate": 9.911986069708424e-06, + "loss": 1.1014, + "step": 21528 + }, + { + "epoch": 0.52, + "learning_rate": 9.911207205749917e-06, + "loss": 1.0867, + "step": 21529 + }, + { + "epoch": 0.52, + "learning_rate": 9.910428342330096e-06, + "loss": 1.1533, + "step": 21530 + }, + { + "epoch": 0.52, + "learning_rate": 9.909649479453686e-06, + "loss": 0.8803, + "step": 21531 + }, + { + "epoch": 0.52, + "learning_rate": 9.90887061712541e-06, + "loss": 0.8223, + "step": 21532 + }, + { + "epoch": 0.52, + "learning_rate": 9.908091755349994e-06, + "loss": 0.9151, + "step": 21533 + }, + { + "epoch": 0.52, + "learning_rate": 9.907312894132165e-06, + "loss": 1.2095, + "step": 21534 + }, + { + "epoch": 0.52, + "learning_rate": 9.906534033476648e-06, + "loss": 1.1203, + "step": 21535 + }, + { + "epoch": 0.52, + "learning_rate": 9.905755173388167e-06, + "loss": 1.2309, + "step": 21536 + }, + { + "epoch": 0.52, + "learning_rate": 9.904976313871444e-06, + "loss": 1.002, + "step": 21537 + }, + { + "epoch": 0.52, + "learning_rate": 9.90419745493121e-06, + "loss": 1.1656, + "step": 21538 + }, + { + "epoch": 0.52, + "learning_rate": 9.903418596572188e-06, + "loss": 0.9461, + "step": 21539 + }, + { + "epoch": 0.52, + "learning_rate": 9.902639738799102e-06, + "loss": 1.14, + "step": 21540 + }, + { + "epoch": 0.52, + "learning_rate": 9.901860881616675e-06, + "loss": 1.1319, + "step": 21541 + }, + { + "epoch": 0.52, + "learning_rate": 9.901082025029639e-06, + "loss": 1.1407, + "step": 21542 + }, + { + "epoch": 0.52, + "learning_rate": 9.900303169042713e-06, + "loss": 1.1188, + "step": 21543 + }, + { + "epoch": 0.52, + "learning_rate": 9.899524313660623e-06, + "loss": 1.2146, + "step": 21544 + }, + { + "epoch": 0.52, + "learning_rate": 9.898745458888097e-06, + "loss": 1.0954, + "step": 21545 + }, + { + "epoch": 0.52, + "learning_rate": 9.89796660472986e-06, + "loss": 1.1294, + "step": 21546 + }, + { + "epoch": 0.52, + "learning_rate": 9.897187751190634e-06, + "loss": 1.0246, + "step": 21547 + }, + { + "epoch": 0.52, + "learning_rate": 9.896408898275143e-06, + "loss": 0.9932, + "step": 21548 + }, + { + "epoch": 0.52, + "learning_rate": 9.89563004598812e-06, + "loss": 1.0592, + "step": 21549 + }, + { + "epoch": 0.52, + "learning_rate": 9.894851194334283e-06, + "loss": 0.9843, + "step": 21550 + }, + { + "epoch": 0.52, + "learning_rate": 9.894072343318358e-06, + "loss": 0.9409, + "step": 21551 + }, + { + "epoch": 0.52, + "learning_rate": 9.89329349294507e-06, + "loss": 0.9658, + "step": 21552 + }, + { + "epoch": 0.52, + "learning_rate": 9.892514643219148e-06, + "loss": 1.0212, + "step": 21553 + }, + { + "epoch": 0.52, + "learning_rate": 9.891735794145314e-06, + "loss": 0.9182, + "step": 21554 + }, + { + "epoch": 0.52, + "learning_rate": 9.890956945728293e-06, + "loss": 1.1652, + "step": 21555 + }, + { + "epoch": 0.52, + "learning_rate": 9.89017809797281e-06, + "loss": 0.9974, + "step": 21556 + }, + { + "epoch": 0.52, + "learning_rate": 9.889399250883591e-06, + "loss": 1.1826, + "step": 21557 + }, + { + "epoch": 0.52, + "learning_rate": 9.888620404465363e-06, + "loss": 1.1747, + "step": 21558 + }, + { + "epoch": 0.52, + "learning_rate": 9.887841558722847e-06, + "loss": 1.1345, + "step": 21559 + }, + { + "epoch": 0.52, + "learning_rate": 9.887062713660767e-06, + "loss": 1.3562, + "step": 21560 + }, + { + "epoch": 0.52, + "learning_rate": 9.886283869283853e-06, + "loss": 0.9984, + "step": 21561 + }, + { + "epoch": 0.52, + "learning_rate": 9.88550502559683e-06, + "loss": 0.844, + "step": 21562 + }, + { + "epoch": 0.52, + "learning_rate": 9.88472618260442e-06, + "loss": 1.2926, + "step": 21563 + }, + { + "epoch": 0.52, + "learning_rate": 9.883947340311347e-06, + "loss": 0.9398, + "step": 21564 + }, + { + "epoch": 0.52, + "learning_rate": 9.883168498722342e-06, + "loss": 0.882, + "step": 21565 + }, + { + "epoch": 0.52, + "learning_rate": 9.882389657842125e-06, + "loss": 1.3655, + "step": 21566 + }, + { + "epoch": 0.52, + "learning_rate": 9.881610817675422e-06, + "loss": 1.2296, + "step": 21567 + }, + { + "epoch": 0.52, + "learning_rate": 9.880831978226955e-06, + "loss": 0.9306, + "step": 21568 + }, + { + "epoch": 0.52, + "learning_rate": 9.880053139501456e-06, + "loss": 1.2132, + "step": 21569 + }, + { + "epoch": 0.52, + "learning_rate": 9.879274301503645e-06, + "loss": 1.024, + "step": 21570 + }, + { + "epoch": 0.52, + "learning_rate": 9.87849546423825e-06, + "loss": 1.1461, + "step": 21571 + }, + { + "epoch": 0.52, + "learning_rate": 9.877716627709992e-06, + "loss": 1.0792, + "step": 21572 + }, + { + "epoch": 0.52, + "learning_rate": 9.8769377919236e-06, + "loss": 1.2712, + "step": 21573 + }, + { + "epoch": 0.52, + "learning_rate": 9.876158956883799e-06, + "loss": 1.0006, + "step": 21574 + }, + { + "epoch": 0.52, + "learning_rate": 9.875380122595311e-06, + "loss": 1.1656, + "step": 21575 + }, + { + "epoch": 0.52, + "learning_rate": 9.87460128906286e-06, + "loss": 1.111, + "step": 21576 + }, + { + "epoch": 0.52, + "learning_rate": 9.873822456291177e-06, + "loss": 1.1673, + "step": 21577 + }, + { + "epoch": 0.52, + "learning_rate": 9.873043624284983e-06, + "loss": 1.0667, + "step": 21578 + }, + { + "epoch": 0.52, + "learning_rate": 9.872264793049002e-06, + "loss": 1.0586, + "step": 21579 + }, + { + "epoch": 0.52, + "learning_rate": 9.87148596258796e-06, + "loss": 1.0534, + "step": 21580 + }, + { + "epoch": 0.52, + "learning_rate": 9.870707132906584e-06, + "loss": 0.8601, + "step": 21581 + }, + { + "epoch": 0.52, + "learning_rate": 9.869928304009598e-06, + "loss": 1.0422, + "step": 21582 + }, + { + "epoch": 0.52, + "learning_rate": 9.869149475901725e-06, + "loss": 1.0737, + "step": 21583 + }, + { + "epoch": 0.52, + "learning_rate": 9.868370648587691e-06, + "loss": 0.9266, + "step": 21584 + }, + { + "epoch": 0.52, + "learning_rate": 9.867591822072222e-06, + "loss": 1.2542, + "step": 21585 + }, + { + "epoch": 0.52, + "learning_rate": 9.866812996360043e-06, + "loss": 1.1975, + "step": 21586 + }, + { + "epoch": 0.52, + "learning_rate": 9.866034171455875e-06, + "loss": 1.2937, + "step": 21587 + }, + { + "epoch": 0.52, + "learning_rate": 9.865255347364451e-06, + "loss": 0.998, + "step": 21588 + }, + { + "epoch": 0.52, + "learning_rate": 9.864476524090489e-06, + "loss": 1.3117, + "step": 21589 + }, + { + "epoch": 0.52, + "learning_rate": 9.863697701638715e-06, + "loss": 1.0185, + "step": 21590 + }, + { + "epoch": 0.52, + "learning_rate": 9.862918880013855e-06, + "loss": 1.0763, + "step": 21591 + }, + { + "epoch": 0.52, + "learning_rate": 9.862140059220637e-06, + "loss": 0.8906, + "step": 21592 + }, + { + "epoch": 0.52, + "learning_rate": 9.86136123926378e-06, + "loss": 1.071, + "step": 21593 + }, + { + "epoch": 0.52, + "learning_rate": 9.860582420148012e-06, + "loss": 1.1441, + "step": 21594 + }, + { + "epoch": 0.52, + "learning_rate": 9.859803601878058e-06, + "loss": 1.1535, + "step": 21595 + }, + { + "epoch": 0.52, + "learning_rate": 9.859024784458643e-06, + "loss": 1.0006, + "step": 21596 + }, + { + "epoch": 0.52, + "learning_rate": 9.85824596789449e-06, + "loss": 1.2833, + "step": 21597 + }, + { + "epoch": 0.52, + "learning_rate": 9.857467152190328e-06, + "loss": 0.8718, + "step": 21598 + }, + { + "epoch": 0.52, + "learning_rate": 9.856688337350875e-06, + "loss": 1.1923, + "step": 21599 + }, + { + "epoch": 0.52, + "learning_rate": 9.855909523380865e-06, + "loss": 0.9651, + "step": 21600 + }, + { + "epoch": 0.52, + "learning_rate": 9.855130710285016e-06, + "loss": 1.0122, + "step": 21601 + }, + { + "epoch": 0.52, + "learning_rate": 9.854351898068056e-06, + "loss": 1.0684, + "step": 21602 + }, + { + "epoch": 0.52, + "learning_rate": 9.853573086734706e-06, + "loss": 1.2061, + "step": 21603 + }, + { + "epoch": 0.52, + "learning_rate": 9.852794276289694e-06, + "loss": 1.1416, + "step": 21604 + }, + { + "epoch": 0.52, + "learning_rate": 9.852015466737747e-06, + "loss": 0.9689, + "step": 21605 + }, + { + "epoch": 0.52, + "learning_rate": 9.851236658083586e-06, + "loss": 1.1143, + "step": 21606 + }, + { + "epoch": 0.52, + "learning_rate": 9.850457850331936e-06, + "loss": 0.9953, + "step": 21607 + }, + { + "epoch": 0.52, + "learning_rate": 9.849679043487526e-06, + "loss": 0.858, + "step": 21608 + }, + { + "epoch": 0.52, + "learning_rate": 9.848900237555076e-06, + "loss": 0.76, + "step": 21609 + }, + { + "epoch": 0.52, + "learning_rate": 9.848121432539313e-06, + "loss": 1.0955, + "step": 21610 + }, + { + "epoch": 0.52, + "learning_rate": 9.84734262844496e-06, + "loss": 1.0543, + "step": 21611 + }, + { + "epoch": 0.52, + "learning_rate": 9.846563825276745e-06, + "loss": 1.2283, + "step": 21612 + }, + { + "epoch": 0.52, + "learning_rate": 9.845785023039393e-06, + "loss": 0.9238, + "step": 21613 + }, + { + "epoch": 0.52, + "learning_rate": 9.845006221737624e-06, + "loss": 1.059, + "step": 21614 + }, + { + "epoch": 0.52, + "learning_rate": 9.844227421376166e-06, + "loss": 0.7567, + "step": 21615 + }, + { + "epoch": 0.52, + "learning_rate": 9.843448621959745e-06, + "loss": 1.0915, + "step": 21616 + }, + { + "epoch": 0.52, + "learning_rate": 9.842669823493083e-06, + "loss": 0.9725, + "step": 21617 + }, + { + "epoch": 0.52, + "learning_rate": 9.841891025980908e-06, + "loss": 0.8005, + "step": 21618 + }, + { + "epoch": 0.52, + "learning_rate": 9.841112229427941e-06, + "loss": 1.2689, + "step": 21619 + }, + { + "epoch": 0.52, + "learning_rate": 9.84033343383891e-06, + "loss": 0.9411, + "step": 21620 + }, + { + "epoch": 0.52, + "learning_rate": 9.839554639218539e-06, + "loss": 1.0631, + "step": 21621 + }, + { + "epoch": 0.52, + "learning_rate": 9.838775845571553e-06, + "loss": 1.101, + "step": 21622 + }, + { + "epoch": 0.52, + "learning_rate": 9.83799705290267e-06, + "loss": 0.9038, + "step": 21623 + }, + { + "epoch": 0.52, + "learning_rate": 9.837218261216628e-06, + "loss": 0.9975, + "step": 21624 + }, + { + "epoch": 0.52, + "learning_rate": 9.836439470518142e-06, + "loss": 1.1324, + "step": 21625 + }, + { + "epoch": 0.52, + "learning_rate": 9.835660680811939e-06, + "loss": 0.9571, + "step": 21626 + }, + { + "epoch": 0.52, + "learning_rate": 9.834881892102743e-06, + "loss": 1.0608, + "step": 21627 + }, + { + "epoch": 0.52, + "learning_rate": 9.83410310439528e-06, + "loss": 1.0301, + "step": 21628 + }, + { + "epoch": 0.52, + "learning_rate": 9.833324317694278e-06, + "loss": 0.9609, + "step": 21629 + }, + { + "epoch": 0.52, + "learning_rate": 9.832545532004454e-06, + "loss": 1.2519, + "step": 21630 + }, + { + "epoch": 0.52, + "learning_rate": 9.831766747330537e-06, + "loss": 1.0933, + "step": 21631 + }, + { + "epoch": 0.52, + "learning_rate": 9.830987963677254e-06, + "loss": 1.3431, + "step": 21632 + }, + { + "epoch": 0.52, + "learning_rate": 9.830209181049326e-06, + "loss": 1.2278, + "step": 21633 + }, + { + "epoch": 0.52, + "learning_rate": 9.829430399451478e-06, + "loss": 0.977, + "step": 21634 + }, + { + "epoch": 0.52, + "learning_rate": 9.828651618888437e-06, + "loss": 1.3451, + "step": 21635 + }, + { + "epoch": 0.52, + "learning_rate": 9.827872839364925e-06, + "loss": 1.1672, + "step": 21636 + }, + { + "epoch": 0.52, + "learning_rate": 9.82709406088567e-06, + "loss": 1.2388, + "step": 21637 + }, + { + "epoch": 0.52, + "learning_rate": 9.826315283455392e-06, + "loss": 1.0782, + "step": 21638 + }, + { + "epoch": 0.52, + "learning_rate": 9.825536507078821e-06, + "loss": 1.0107, + "step": 21639 + }, + { + "epoch": 0.52, + "learning_rate": 9.824757731760678e-06, + "loss": 1.0672, + "step": 21640 + }, + { + "epoch": 0.52, + "learning_rate": 9.82397895750569e-06, + "loss": 1.219, + "step": 21641 + }, + { + "epoch": 0.52, + "learning_rate": 9.823200184318577e-06, + "loss": 1.1078, + "step": 21642 + }, + { + "epoch": 0.52, + "learning_rate": 9.82242141220407e-06, + "loss": 1.2347, + "step": 21643 + }, + { + "epoch": 0.52, + "learning_rate": 9.82164264116689e-06, + "loss": 1.1354, + "step": 21644 + }, + { + "epoch": 0.52, + "learning_rate": 9.820863871211762e-06, + "loss": 1.1001, + "step": 21645 + }, + { + "epoch": 0.52, + "learning_rate": 9.820085102343408e-06, + "loss": 0.9209, + "step": 21646 + }, + { + "epoch": 0.52, + "learning_rate": 9.819306334566558e-06, + "loss": 1.1162, + "step": 21647 + }, + { + "epoch": 0.52, + "learning_rate": 9.818527567885933e-06, + "loss": 0.6469, + "step": 21648 + }, + { + "epoch": 0.52, + "learning_rate": 9.817748802306259e-06, + "loss": 0.8999, + "step": 21649 + }, + { + "epoch": 0.52, + "learning_rate": 9.81697003783226e-06, + "loss": 1.1833, + "step": 21650 + }, + { + "epoch": 0.52, + "learning_rate": 9.816191274468661e-06, + "loss": 1.1894, + "step": 21651 + }, + { + "epoch": 0.52, + "learning_rate": 9.815412512220187e-06, + "loss": 1.0688, + "step": 21652 + }, + { + "epoch": 0.52, + "learning_rate": 9.814633751091561e-06, + "loss": 1.2505, + "step": 21653 + }, + { + "epoch": 0.52, + "learning_rate": 9.813854991087507e-06, + "loss": 0.9642, + "step": 21654 + }, + { + "epoch": 0.52, + "learning_rate": 9.813076232212753e-06, + "loss": 1.0999, + "step": 21655 + }, + { + "epoch": 0.52, + "learning_rate": 9.812297474472022e-06, + "loss": 0.9126, + "step": 21656 + }, + { + "epoch": 0.52, + "learning_rate": 9.811518717870037e-06, + "loss": 0.9213, + "step": 21657 + }, + { + "epoch": 0.52, + "learning_rate": 9.810739962411522e-06, + "loss": 1.1293, + "step": 21658 + }, + { + "epoch": 0.52, + "learning_rate": 9.809961208101205e-06, + "loss": 1.2257, + "step": 21659 + }, + { + "epoch": 0.52, + "learning_rate": 9.80918245494381e-06, + "loss": 1.3806, + "step": 21660 + }, + { + "epoch": 0.52, + "learning_rate": 9.808403702944058e-06, + "loss": 0.9516, + "step": 21661 + }, + { + "epoch": 0.52, + "learning_rate": 9.807624952106676e-06, + "loss": 0.8049, + "step": 21662 + }, + { + "epoch": 0.52, + "learning_rate": 9.806846202436388e-06, + "loss": 0.9182, + "step": 21663 + }, + { + "epoch": 0.52, + "learning_rate": 9.80606745393792e-06, + "loss": 1.0324, + "step": 21664 + }, + { + "epoch": 0.52, + "learning_rate": 9.805288706615996e-06, + "loss": 1.2054, + "step": 21665 + }, + { + "epoch": 0.52, + "learning_rate": 9.804509960475332e-06, + "loss": 0.8115, + "step": 21666 + }, + { + "epoch": 0.52, + "learning_rate": 9.803731215520666e-06, + "loss": 1.1227, + "step": 21667 + }, + { + "epoch": 0.52, + "learning_rate": 9.802952471756716e-06, + "loss": 1.0563, + "step": 21668 + }, + { + "epoch": 0.52, + "learning_rate": 9.802173729188208e-06, + "loss": 0.9366, + "step": 21669 + }, + { + "epoch": 0.52, + "learning_rate": 9.801394987819865e-06, + "loss": 1.0205, + "step": 21670 + }, + { + "epoch": 0.52, + "learning_rate": 9.80061624765641e-06, + "loss": 1.4656, + "step": 21671 + }, + { + "epoch": 0.52, + "learning_rate": 9.799837508702574e-06, + "loss": 1.1585, + "step": 21672 + }, + { + "epoch": 0.52, + "learning_rate": 9.799058770963074e-06, + "loss": 1.3458, + "step": 21673 + }, + { + "epoch": 0.52, + "learning_rate": 9.798280034442637e-06, + "loss": 1.0856, + "step": 21674 + }, + { + "epoch": 0.52, + "learning_rate": 9.797501299145986e-06, + "loss": 1.3366, + "step": 21675 + }, + { + "epoch": 0.52, + "learning_rate": 9.79672256507785e-06, + "loss": 1.0795, + "step": 21676 + }, + { + "epoch": 0.52, + "learning_rate": 9.795943832242949e-06, + "loss": 1.2561, + "step": 21677 + }, + { + "epoch": 0.52, + "learning_rate": 9.79516510064601e-06, + "loss": 1.0126, + "step": 21678 + }, + { + "epoch": 0.52, + "learning_rate": 9.794386370291753e-06, + "loss": 1.1108, + "step": 21679 + }, + { + "epoch": 0.52, + "learning_rate": 9.793607641184908e-06, + "loss": 1.1197, + "step": 21680 + }, + { + "epoch": 0.52, + "learning_rate": 9.792828913330196e-06, + "loss": 1.2062, + "step": 21681 + }, + { + "epoch": 0.52, + "learning_rate": 9.792050186732344e-06, + "loss": 1.1888, + "step": 21682 + }, + { + "epoch": 0.52, + "learning_rate": 9.791271461396071e-06, + "loss": 1.2438, + "step": 21683 + }, + { + "epoch": 0.52, + "learning_rate": 9.790492737326108e-06, + "loss": 0.9063, + "step": 21684 + }, + { + "epoch": 0.52, + "learning_rate": 9.789714014527176e-06, + "loss": 1.1695, + "step": 21685 + }, + { + "epoch": 0.52, + "learning_rate": 9.788935293004e-06, + "loss": 1.2305, + "step": 21686 + }, + { + "epoch": 0.52, + "learning_rate": 9.7881565727613e-06, + "loss": 1.2183, + "step": 21687 + }, + { + "epoch": 0.52, + "learning_rate": 9.78737785380381e-06, + "loss": 1.0665, + "step": 21688 + }, + { + "epoch": 0.52, + "learning_rate": 9.786599136136246e-06, + "loss": 1.0383, + "step": 21689 + }, + { + "epoch": 0.52, + "learning_rate": 9.785820419763335e-06, + "loss": 0.8568, + "step": 21690 + }, + { + "epoch": 0.52, + "learning_rate": 9.7850417046898e-06, + "loss": 1.06, + "step": 21691 + }, + { + "epoch": 0.52, + "learning_rate": 9.784262990920367e-06, + "loss": 1.2722, + "step": 21692 + }, + { + "epoch": 0.52, + "learning_rate": 9.78348427845976e-06, + "loss": 1.2044, + "step": 21693 + }, + { + "epoch": 0.52, + "learning_rate": 9.782705567312704e-06, + "loss": 0.9326, + "step": 21694 + }, + { + "epoch": 0.52, + "learning_rate": 9.78192685748392e-06, + "loss": 1.2118, + "step": 21695 + }, + { + "epoch": 0.52, + "learning_rate": 9.781148148978135e-06, + "loss": 1.1632, + "step": 21696 + }, + { + "epoch": 0.52, + "learning_rate": 9.780369441800075e-06, + "loss": 1.1668, + "step": 21697 + }, + { + "epoch": 0.52, + "learning_rate": 9.77959073595446e-06, + "loss": 1.0438, + "step": 21698 + }, + { + "epoch": 0.52, + "learning_rate": 9.778812031446015e-06, + "loss": 1.241, + "step": 21699 + }, + { + "epoch": 0.52, + "learning_rate": 9.778033328279468e-06, + "loss": 1.0691, + "step": 21700 + }, + { + "epoch": 0.52, + "learning_rate": 9.77725462645954e-06, + "loss": 0.8954, + "step": 21701 + }, + { + "epoch": 0.52, + "learning_rate": 9.776475925990955e-06, + "loss": 1.186, + "step": 21702 + }, + { + "epoch": 0.52, + "learning_rate": 9.775697226878436e-06, + "loss": 1.2124, + "step": 21703 + }, + { + "epoch": 0.52, + "learning_rate": 9.774918529126713e-06, + "loss": 1.2555, + "step": 21704 + }, + { + "epoch": 0.52, + "learning_rate": 9.774139832740505e-06, + "loss": 1.1174, + "step": 21705 + }, + { + "epoch": 0.52, + "learning_rate": 9.773361137724537e-06, + "loss": 1.201, + "step": 21706 + }, + { + "epoch": 0.52, + "learning_rate": 9.772582444083532e-06, + "loss": 1.1316, + "step": 21707 + }, + { + "epoch": 0.52, + "learning_rate": 9.771803751822219e-06, + "loss": 0.9723, + "step": 21708 + }, + { + "epoch": 0.52, + "learning_rate": 9.771025060945317e-06, + "loss": 1.0649, + "step": 21709 + }, + { + "epoch": 0.52, + "learning_rate": 9.770246371457553e-06, + "loss": 1.4523, + "step": 21710 + }, + { + "epoch": 0.52, + "learning_rate": 9.769467683363649e-06, + "loss": 1.1898, + "step": 21711 + }, + { + "epoch": 0.52, + "learning_rate": 9.768688996668332e-06, + "loss": 1.1644, + "step": 21712 + }, + { + "epoch": 0.52, + "learning_rate": 9.767910311376324e-06, + "loss": 1.1486, + "step": 21713 + }, + { + "epoch": 0.52, + "learning_rate": 9.767131627492348e-06, + "loss": 1.1046, + "step": 21714 + }, + { + "epoch": 0.52, + "learning_rate": 9.766352945021131e-06, + "loss": 1.2276, + "step": 21715 + }, + { + "epoch": 0.52, + "learning_rate": 9.765574263967397e-06, + "loss": 1.134, + "step": 21716 + }, + { + "epoch": 0.52, + "learning_rate": 9.764795584335868e-06, + "loss": 1.0389, + "step": 21717 + }, + { + "epoch": 0.52, + "learning_rate": 9.764016906131267e-06, + "loss": 1.1515, + "step": 21718 + }, + { + "epoch": 0.52, + "learning_rate": 9.76323822935832e-06, + "loss": 1.262, + "step": 21719 + }, + { + "epoch": 0.52, + "learning_rate": 9.762459554021753e-06, + "loss": 0.8833, + "step": 21720 + }, + { + "epoch": 0.52, + "learning_rate": 9.76168088012629e-06, + "loss": 1.2727, + "step": 21721 + }, + { + "epoch": 0.52, + "learning_rate": 9.760902207676646e-06, + "loss": 1.0837, + "step": 21722 + }, + { + "epoch": 0.52, + "learning_rate": 9.760123536677558e-06, + "loss": 1.1092, + "step": 21723 + }, + { + "epoch": 0.52, + "learning_rate": 9.759344867133742e-06, + "loss": 1.297, + "step": 21724 + }, + { + "epoch": 0.52, + "learning_rate": 9.758566199049926e-06, + "loss": 1.2451, + "step": 21725 + }, + { + "epoch": 0.52, + "learning_rate": 9.757787532430827e-06, + "loss": 0.9744, + "step": 21726 + }, + { + "epoch": 0.52, + "learning_rate": 9.757008867281179e-06, + "loss": 1.1876, + "step": 21727 + }, + { + "epoch": 0.52, + "learning_rate": 9.7562302036057e-06, + "loss": 0.9971, + "step": 21728 + }, + { + "epoch": 0.52, + "learning_rate": 9.755451541409116e-06, + "loss": 1.314, + "step": 21729 + }, + { + "epoch": 0.52, + "learning_rate": 9.754672880696147e-06, + "loss": 1.3016, + "step": 21730 + }, + { + "epoch": 0.52, + "learning_rate": 9.753894221471523e-06, + "loss": 1.1374, + "step": 21731 + }, + { + "epoch": 0.52, + "learning_rate": 9.753115563739963e-06, + "loss": 1.0543, + "step": 21732 + }, + { + "epoch": 0.52, + "learning_rate": 9.752336907506194e-06, + "loss": 1.1326, + "step": 21733 + }, + { + "epoch": 0.52, + "learning_rate": 9.751558252774937e-06, + "loss": 1.0189, + "step": 21734 + }, + { + "epoch": 0.52, + "learning_rate": 9.750779599550919e-06, + "loss": 0.9598, + "step": 21735 + }, + { + "epoch": 0.52, + "learning_rate": 9.750000947838862e-06, + "loss": 1.2314, + "step": 21736 + }, + { + "epoch": 0.52, + "learning_rate": 9.749222297643491e-06, + "loss": 1.0882, + "step": 21737 + }, + { + "epoch": 0.52, + "learning_rate": 9.748443648969528e-06, + "loss": 1.0616, + "step": 21738 + }, + { + "epoch": 0.52, + "learning_rate": 9.7476650018217e-06, + "loss": 1.3164, + "step": 21739 + }, + { + "epoch": 0.52, + "learning_rate": 9.746886356204729e-06, + "loss": 0.9193, + "step": 21740 + }, + { + "epoch": 0.52, + "learning_rate": 9.746107712123339e-06, + "loss": 1.0198, + "step": 21741 + }, + { + "epoch": 0.52, + "learning_rate": 9.74532906958225e-06, + "loss": 1.0145, + "step": 21742 + }, + { + "epoch": 0.52, + "learning_rate": 9.744550428586194e-06, + "loss": 0.9799, + "step": 21743 + }, + { + "epoch": 0.52, + "learning_rate": 9.743771789139889e-06, + "loss": 1.1668, + "step": 21744 + }, + { + "epoch": 0.52, + "learning_rate": 9.742993151248062e-06, + "loss": 1.114, + "step": 21745 + }, + { + "epoch": 0.52, + "learning_rate": 9.742214514915431e-06, + "loss": 0.9774, + "step": 21746 + }, + { + "epoch": 0.52, + "learning_rate": 9.741435880146727e-06, + "loss": 1.0099, + "step": 21747 + }, + { + "epoch": 0.52, + "learning_rate": 9.74065724694667e-06, + "loss": 1.1396, + "step": 21748 + }, + { + "epoch": 0.52, + "learning_rate": 9.739878615319986e-06, + "loss": 1.0648, + "step": 21749 + }, + { + "epoch": 0.52, + "learning_rate": 9.739099985271394e-06, + "loss": 1.3206, + "step": 21750 + }, + { + "epoch": 0.52, + "learning_rate": 9.738321356805623e-06, + "loss": 1.387, + "step": 21751 + }, + { + "epoch": 0.52, + "learning_rate": 9.737542729927394e-06, + "loss": 0.6781, + "step": 21752 + }, + { + "epoch": 0.52, + "learning_rate": 9.736764104641433e-06, + "loss": 1.1106, + "step": 21753 + }, + { + "epoch": 0.52, + "learning_rate": 9.73598548095246e-06, + "loss": 0.7767, + "step": 21754 + }, + { + "epoch": 0.52, + "learning_rate": 9.735206858865204e-06, + "loss": 1.0046, + "step": 21755 + }, + { + "epoch": 0.52, + "learning_rate": 9.734428238384385e-06, + "loss": 0.8947, + "step": 21756 + }, + { + "epoch": 0.52, + "learning_rate": 9.733649619514726e-06, + "loss": 0.8439, + "step": 21757 + }, + { + "epoch": 0.52, + "learning_rate": 9.732871002260951e-06, + "loss": 0.8409, + "step": 21758 + }, + { + "epoch": 0.52, + "learning_rate": 9.732092386627787e-06, + "loss": 1.0748, + "step": 21759 + }, + { + "epoch": 0.52, + "learning_rate": 9.731313772619956e-06, + "loss": 1.1605, + "step": 21760 + }, + { + "epoch": 0.52, + "learning_rate": 9.730535160242178e-06, + "loss": 1.1848, + "step": 21761 + }, + { + "epoch": 0.52, + "learning_rate": 9.729756549499181e-06, + "loss": 0.9515, + "step": 21762 + }, + { + "epoch": 0.52, + "learning_rate": 9.72897794039569e-06, + "loss": 0.8971, + "step": 21763 + }, + { + "epoch": 0.52, + "learning_rate": 9.728199332936425e-06, + "loss": 0.7849, + "step": 21764 + }, + { + "epoch": 0.52, + "learning_rate": 9.727420727126109e-06, + "loss": 1.0759, + "step": 21765 + }, + { + "epoch": 0.52, + "learning_rate": 9.72664212296947e-06, + "loss": 1.0005, + "step": 21766 + }, + { + "epoch": 0.52, + "learning_rate": 9.725863520471226e-06, + "loss": 1.0739, + "step": 21767 + }, + { + "epoch": 0.52, + "learning_rate": 9.725084919636106e-06, + "loss": 1.2034, + "step": 21768 + }, + { + "epoch": 0.52, + "learning_rate": 9.724306320468828e-06, + "loss": 1.1807, + "step": 21769 + }, + { + "epoch": 0.52, + "learning_rate": 9.723527722974121e-06, + "loss": 1.2387, + "step": 21770 + }, + { + "epoch": 0.52, + "learning_rate": 9.722749127156706e-06, + "loss": 1.1963, + "step": 21771 + }, + { + "epoch": 0.52, + "learning_rate": 9.721970533021307e-06, + "loss": 1.0955, + "step": 21772 + }, + { + "epoch": 0.52, + "learning_rate": 9.721191940572647e-06, + "loss": 1.2245, + "step": 21773 + }, + { + "epoch": 0.52, + "learning_rate": 9.72041334981545e-06, + "loss": 1.231, + "step": 21774 + }, + { + "epoch": 0.52, + "learning_rate": 9.719634760754441e-06, + "loss": 0.8889, + "step": 21775 + }, + { + "epoch": 0.52, + "learning_rate": 9.718856173394342e-06, + "loss": 1.1649, + "step": 21776 + }, + { + "epoch": 0.52, + "learning_rate": 9.718077587739872e-06, + "loss": 1.0044, + "step": 21777 + }, + { + "epoch": 0.52, + "learning_rate": 9.717299003795763e-06, + "loss": 0.8832, + "step": 21778 + }, + { + "epoch": 0.52, + "learning_rate": 9.716520421566733e-06, + "loss": 1.2836, + "step": 21779 + }, + { + "epoch": 0.52, + "learning_rate": 9.715741841057509e-06, + "loss": 1.2749, + "step": 21780 + }, + { + "epoch": 0.52, + "learning_rate": 9.714963262272808e-06, + "loss": 1.0955, + "step": 21781 + }, + { + "epoch": 0.52, + "learning_rate": 9.714184685217361e-06, + "loss": 1.1101, + "step": 21782 + }, + { + "epoch": 0.52, + "learning_rate": 9.713406109895888e-06, + "loss": 1.0205, + "step": 21783 + }, + { + "epoch": 0.52, + "learning_rate": 9.712627536313113e-06, + "loss": 1.1936, + "step": 21784 + }, + { + "epoch": 0.52, + "learning_rate": 9.711848964473757e-06, + "loss": 1.1944, + "step": 21785 + }, + { + "epoch": 0.52, + "learning_rate": 9.711070394382549e-06, + "loss": 1.2536, + "step": 21786 + }, + { + "epoch": 0.52, + "learning_rate": 9.710291826044207e-06, + "loss": 1.1003, + "step": 21787 + }, + { + "epoch": 0.52, + "learning_rate": 9.709513259463457e-06, + "loss": 1.0596, + "step": 21788 + }, + { + "epoch": 0.52, + "learning_rate": 9.70873469464502e-06, + "loss": 0.8967, + "step": 21789 + }, + { + "epoch": 0.52, + "learning_rate": 9.707956131593623e-06, + "loss": 0.9981, + "step": 21790 + }, + { + "epoch": 0.52, + "learning_rate": 9.707177570313988e-06, + "loss": 1.2933, + "step": 21791 + }, + { + "epoch": 0.52, + "learning_rate": 9.706399010810837e-06, + "loss": 1.4271, + "step": 21792 + }, + { + "epoch": 0.52, + "learning_rate": 9.705620453088891e-06, + "loss": 1.0274, + "step": 21793 + }, + { + "epoch": 0.52, + "learning_rate": 9.70484189715288e-06, + "loss": 0.9073, + "step": 21794 + }, + { + "epoch": 0.52, + "learning_rate": 9.704063343007525e-06, + "loss": 1.0618, + "step": 21795 + }, + { + "epoch": 0.52, + "learning_rate": 9.703284790657547e-06, + "loss": 1.0938, + "step": 21796 + }, + { + "epoch": 0.52, + "learning_rate": 9.702506240107668e-06, + "loss": 1.2397, + "step": 21797 + }, + { + "epoch": 0.52, + "learning_rate": 9.701727691362614e-06, + "loss": 1.1476, + "step": 21798 + }, + { + "epoch": 0.52, + "learning_rate": 9.700949144427111e-06, + "loss": 1.052, + "step": 21799 + }, + { + "epoch": 0.52, + "learning_rate": 9.700170599305877e-06, + "loss": 1.1924, + "step": 21800 + }, + { + "epoch": 0.52, + "learning_rate": 9.699392056003637e-06, + "loss": 0.9416, + "step": 21801 + }, + { + "epoch": 0.52, + "learning_rate": 9.698613514525117e-06, + "loss": 1.1056, + "step": 21802 + }, + { + "epoch": 0.52, + "learning_rate": 9.697834974875036e-06, + "loss": 1.2195, + "step": 21803 + }, + { + "epoch": 0.52, + "learning_rate": 9.69705643705812e-06, + "loss": 1.1353, + "step": 21804 + }, + { + "epoch": 0.52, + "learning_rate": 9.696277901079088e-06, + "loss": 1.1557, + "step": 21805 + }, + { + "epoch": 0.52, + "learning_rate": 9.695499366942672e-06, + "loss": 1.0991, + "step": 21806 + }, + { + "epoch": 0.52, + "learning_rate": 9.694720834653586e-06, + "loss": 0.9878, + "step": 21807 + }, + { + "epoch": 0.52, + "learning_rate": 9.693942304216557e-06, + "loss": 0.7847, + "step": 21808 + }, + { + "epoch": 0.52, + "learning_rate": 9.693163775636309e-06, + "loss": 1.0494, + "step": 21809 + }, + { + "epoch": 0.52, + "learning_rate": 9.692385248917565e-06, + "loss": 1.2075, + "step": 21810 + }, + { + "epoch": 0.52, + "learning_rate": 9.691606724065046e-06, + "loss": 1.1906, + "step": 21811 + }, + { + "epoch": 0.52, + "learning_rate": 9.690828201083476e-06, + "loss": 1.0553, + "step": 21812 + }, + { + "epoch": 0.52, + "learning_rate": 9.690049679977582e-06, + "loss": 0.7713, + "step": 21813 + }, + { + "epoch": 0.52, + "learning_rate": 9.68927116075208e-06, + "loss": 0.8725, + "step": 21814 + }, + { + "epoch": 0.52, + "learning_rate": 9.6884926434117e-06, + "loss": 1.2589, + "step": 21815 + }, + { + "epoch": 0.52, + "learning_rate": 9.687714127961158e-06, + "loss": 1.1584, + "step": 21816 + }, + { + "epoch": 0.52, + "learning_rate": 9.686935614405183e-06, + "loss": 1.141, + "step": 21817 + }, + { + "epoch": 0.52, + "learning_rate": 9.686157102748496e-06, + "loss": 1.1973, + "step": 21818 + }, + { + "epoch": 0.52, + "learning_rate": 9.685378592995822e-06, + "loss": 1.0653, + "step": 21819 + }, + { + "epoch": 0.52, + "learning_rate": 9.684600085151877e-06, + "loss": 1.1606, + "step": 21820 + }, + { + "epoch": 0.52, + "learning_rate": 9.683821579221393e-06, + "loss": 1.1342, + "step": 21821 + }, + { + "epoch": 0.52, + "learning_rate": 9.683043075209088e-06, + "loss": 1.2019, + "step": 21822 + }, + { + "epoch": 0.52, + "learning_rate": 9.682264573119687e-06, + "loss": 1.1101, + "step": 21823 + }, + { + "epoch": 0.52, + "learning_rate": 9.681486072957911e-06, + "loss": 1.1901, + "step": 21824 + }, + { + "epoch": 0.52, + "learning_rate": 9.680707574728485e-06, + "loss": 1.1206, + "step": 21825 + }, + { + "epoch": 0.52, + "learning_rate": 9.679929078436131e-06, + "loss": 1.1703, + "step": 21826 + }, + { + "epoch": 0.52, + "learning_rate": 9.679150584085572e-06, + "loss": 1.2317, + "step": 21827 + }, + { + "epoch": 0.52, + "learning_rate": 9.678372091681529e-06, + "loss": 1.0328, + "step": 21828 + }, + { + "epoch": 0.52, + "learning_rate": 9.677593601228729e-06, + "loss": 0.8403, + "step": 21829 + }, + { + "epoch": 0.52, + "learning_rate": 9.676815112731892e-06, + "loss": 1.2515, + "step": 21830 + }, + { + "epoch": 0.53, + "learning_rate": 9.676036626195742e-06, + "loss": 1.0027, + "step": 21831 + }, + { + "epoch": 0.53, + "learning_rate": 9.675258141624999e-06, + "loss": 1.0077, + "step": 21832 + }, + { + "epoch": 0.53, + "learning_rate": 9.67447965902439e-06, + "loss": 1.1604, + "step": 21833 + }, + { + "epoch": 0.53, + "learning_rate": 9.673701178398635e-06, + "loss": 1.124, + "step": 21834 + }, + { + "epoch": 0.53, + "learning_rate": 9.672922699752461e-06, + "loss": 1.265, + "step": 21835 + }, + { + "epoch": 0.53, + "learning_rate": 9.67214422309059e-06, + "loss": 0.9388, + "step": 21836 + }, + { + "epoch": 0.53, + "learning_rate": 9.67136574841774e-06, + "loss": 1.1612, + "step": 21837 + }, + { + "epoch": 0.53, + "learning_rate": 9.670587275738635e-06, + "loss": 1.3237, + "step": 21838 + }, + { + "epoch": 0.53, + "learning_rate": 9.669808805058003e-06, + "loss": 1.1904, + "step": 21839 + }, + { + "epoch": 0.53, + "learning_rate": 9.669030336380562e-06, + "loss": 1.1903, + "step": 21840 + }, + { + "epoch": 0.53, + "learning_rate": 9.668251869711034e-06, + "loss": 1.2891, + "step": 21841 + }, + { + "epoch": 0.53, + "learning_rate": 9.667473405054147e-06, + "loss": 1.1652, + "step": 21842 + }, + { + "epoch": 0.53, + "learning_rate": 9.666694942414619e-06, + "loss": 0.8987, + "step": 21843 + }, + { + "epoch": 0.53, + "learning_rate": 9.665916481797176e-06, + "loss": 1.0595, + "step": 21844 + }, + { + "epoch": 0.53, + "learning_rate": 9.665138023206537e-06, + "loss": 1.1463, + "step": 21845 + }, + { + "epoch": 0.53, + "learning_rate": 9.664359566647428e-06, + "loss": 1.1649, + "step": 21846 + }, + { + "epoch": 0.53, + "learning_rate": 9.66358111212457e-06, + "loss": 1.0163, + "step": 21847 + }, + { + "epoch": 0.53, + "learning_rate": 9.662802659642688e-06, + "loss": 1.0667, + "step": 21848 + }, + { + "epoch": 0.53, + "learning_rate": 9.6620242092065e-06, + "loss": 1.3038, + "step": 21849 + }, + { + "epoch": 0.53, + "learning_rate": 9.661245760820734e-06, + "loss": 0.8467, + "step": 21850 + }, + { + "epoch": 0.53, + "learning_rate": 9.66046731449011e-06, + "loss": 1.2956, + "step": 21851 + }, + { + "epoch": 0.53, + "learning_rate": 9.659688870219352e-06, + "loss": 0.9303, + "step": 21852 + }, + { + "epoch": 0.53, + "learning_rate": 9.658910428013177e-06, + "loss": 1.2076, + "step": 21853 + }, + { + "epoch": 0.53, + "learning_rate": 9.658131987876317e-06, + "loss": 1.0253, + "step": 21854 + }, + { + "epoch": 0.53, + "learning_rate": 9.657353549813487e-06, + "loss": 1.0227, + "step": 21855 + }, + { + "epoch": 0.53, + "learning_rate": 9.656575113829415e-06, + "loss": 1.1426, + "step": 21856 + }, + { + "epoch": 0.53, + "learning_rate": 9.655796679928818e-06, + "loss": 0.8505, + "step": 21857 + }, + { + "epoch": 0.53, + "learning_rate": 9.655018248116424e-06, + "loss": 0.7923, + "step": 21858 + }, + { + "epoch": 0.53, + "learning_rate": 9.654239818396952e-06, + "loss": 1.1225, + "step": 21859 + }, + { + "epoch": 0.53, + "learning_rate": 9.653461390775127e-06, + "loss": 1.0464, + "step": 21860 + }, + { + "epoch": 0.53, + "learning_rate": 9.652682965255667e-06, + "loss": 1.0266, + "step": 21861 + }, + { + "epoch": 0.53, + "learning_rate": 9.651904541843301e-06, + "loss": 1.0033, + "step": 21862 + }, + { + "epoch": 0.53, + "learning_rate": 9.651126120542747e-06, + "loss": 0.9578, + "step": 21863 + }, + { + "epoch": 0.53, + "learning_rate": 9.65034770135873e-06, + "loss": 0.8831, + "step": 21864 + }, + { + "epoch": 0.53, + "learning_rate": 9.649569284295968e-06, + "loss": 0.9745, + "step": 21865 + }, + { + "epoch": 0.53, + "learning_rate": 9.64879086935919e-06, + "loss": 0.8405, + "step": 21866 + }, + { + "epoch": 0.53, + "learning_rate": 9.648012456553114e-06, + "loss": 1.2052, + "step": 21867 + }, + { + "epoch": 0.53, + "learning_rate": 9.647234045882465e-06, + "loss": 1.0323, + "step": 21868 + }, + { + "epoch": 0.53, + "learning_rate": 9.64645563735196e-06, + "loss": 1.1752, + "step": 21869 + }, + { + "epoch": 0.53, + "learning_rate": 9.64567723096633e-06, + "loss": 1.2212, + "step": 21870 + }, + { + "epoch": 0.53, + "learning_rate": 9.644898826730292e-06, + "loss": 1.0539, + "step": 21871 + }, + { + "epoch": 0.53, + "learning_rate": 9.644120424648568e-06, + "loss": 1.1992, + "step": 21872 + }, + { + "epoch": 0.53, + "learning_rate": 9.64334202472588e-06, + "loss": 0.8693, + "step": 21873 + }, + { + "epoch": 0.53, + "learning_rate": 9.642563626966956e-06, + "loss": 1.1908, + "step": 21874 + }, + { + "epoch": 0.53, + "learning_rate": 9.641785231376513e-06, + "loss": 0.9192, + "step": 21875 + }, + { + "epoch": 0.53, + "learning_rate": 9.641006837959274e-06, + "loss": 1.1323, + "step": 21876 + }, + { + "epoch": 0.53, + "learning_rate": 9.640228446719962e-06, + "loss": 0.9894, + "step": 21877 + }, + { + "epoch": 0.53, + "learning_rate": 9.6394500576633e-06, + "loss": 1.0833, + "step": 21878 + }, + { + "epoch": 0.53, + "learning_rate": 9.63867167079401e-06, + "loss": 1.1065, + "step": 21879 + }, + { + "epoch": 0.53, + "learning_rate": 9.637893286116814e-06, + "loss": 1.0273, + "step": 21880 + }, + { + "epoch": 0.53, + "learning_rate": 9.637114903636432e-06, + "loss": 1.2889, + "step": 21881 + }, + { + "epoch": 0.53, + "learning_rate": 9.636336523357592e-06, + "loss": 1.3373, + "step": 21882 + }, + { + "epoch": 0.53, + "learning_rate": 9.63555814528501e-06, + "loss": 1.0808, + "step": 21883 + }, + { + "epoch": 0.53, + "learning_rate": 9.634779769423412e-06, + "loss": 1.0473, + "step": 21884 + }, + { + "epoch": 0.53, + "learning_rate": 9.63400139577752e-06, + "loss": 0.9676, + "step": 21885 + }, + { + "epoch": 0.53, + "learning_rate": 9.633223024352054e-06, + "loss": 0.9115, + "step": 21886 + }, + { + "epoch": 0.53, + "learning_rate": 9.63244465515174e-06, + "loss": 1.1422, + "step": 21887 + }, + { + "epoch": 0.53, + "learning_rate": 9.631666288181293e-06, + "loss": 1.1765, + "step": 21888 + }, + { + "epoch": 0.53, + "learning_rate": 9.630887923445445e-06, + "loss": 1.168, + "step": 21889 + }, + { + "epoch": 0.53, + "learning_rate": 9.630109560948912e-06, + "loss": 1.1905, + "step": 21890 + }, + { + "epoch": 0.53, + "learning_rate": 9.629331200696416e-06, + "loss": 1.0866, + "step": 21891 + }, + { + "epoch": 0.53, + "learning_rate": 9.62855284269268e-06, + "loss": 1.2868, + "step": 21892 + }, + { + "epoch": 0.53, + "learning_rate": 9.627774486942428e-06, + "loss": 1.0118, + "step": 21893 + }, + { + "epoch": 0.53, + "learning_rate": 9.626996133450382e-06, + "loss": 0.9377, + "step": 21894 + }, + { + "epoch": 0.53, + "learning_rate": 9.626217782221262e-06, + "loss": 0.8379, + "step": 21895 + }, + { + "epoch": 0.53, + "learning_rate": 9.625439433259786e-06, + "loss": 1.0504, + "step": 21896 + }, + { + "epoch": 0.53, + "learning_rate": 9.624661086570685e-06, + "loss": 1.098, + "step": 21897 + }, + { + "epoch": 0.53, + "learning_rate": 9.623882742158678e-06, + "loss": 0.9589, + "step": 21898 + }, + { + "epoch": 0.53, + "learning_rate": 9.623104400028484e-06, + "loss": 1.0226, + "step": 21899 + }, + { + "epoch": 0.53, + "learning_rate": 9.622326060184827e-06, + "loss": 1.0532, + "step": 21900 + }, + { + "epoch": 0.53, + "learning_rate": 9.62154772263243e-06, + "loss": 1.0566, + "step": 21901 + }, + { + "epoch": 0.53, + "learning_rate": 9.620769387376013e-06, + "loss": 0.9013, + "step": 21902 + }, + { + "epoch": 0.53, + "learning_rate": 9.6199910544203e-06, + "loss": 1.1126, + "step": 21903 + }, + { + "epoch": 0.53, + "learning_rate": 9.61921272377001e-06, + "loss": 1.1318, + "step": 21904 + }, + { + "epoch": 0.53, + "learning_rate": 9.618434395429868e-06, + "loss": 1.1385, + "step": 21905 + }, + { + "epoch": 0.53, + "learning_rate": 9.617656069404595e-06, + "loss": 1.1912, + "step": 21906 + }, + { + "epoch": 0.53, + "learning_rate": 9.616877745698914e-06, + "loss": 1.2672, + "step": 21907 + }, + { + "epoch": 0.53, + "learning_rate": 9.616099424317541e-06, + "loss": 0.9738, + "step": 21908 + }, + { + "epoch": 0.53, + "learning_rate": 9.615321105265208e-06, + "loss": 1.0297, + "step": 21909 + }, + { + "epoch": 0.53, + "learning_rate": 9.614542788546627e-06, + "loss": 1.2469, + "step": 21910 + }, + { + "epoch": 0.53, + "learning_rate": 9.613764474166527e-06, + "loss": 1.107, + "step": 21911 + }, + { + "epoch": 0.53, + "learning_rate": 9.612986162129625e-06, + "loss": 0.9362, + "step": 21912 + }, + { + "epoch": 0.53, + "learning_rate": 9.612207852440646e-06, + "loss": 1.0125, + "step": 21913 + }, + { + "epoch": 0.53, + "learning_rate": 9.611429545104313e-06, + "loss": 1.1057, + "step": 21914 + }, + { + "epoch": 0.53, + "learning_rate": 9.610651240125343e-06, + "loss": 1.1579, + "step": 21915 + }, + { + "epoch": 0.53, + "learning_rate": 9.609872937508459e-06, + "loss": 0.8736, + "step": 21916 + }, + { + "epoch": 0.53, + "learning_rate": 9.609094637258387e-06, + "loss": 0.9714, + "step": 21917 + }, + { + "epoch": 0.53, + "learning_rate": 9.608316339379847e-06, + "loss": 1.2381, + "step": 21918 + }, + { + "epoch": 0.53, + "learning_rate": 9.607538043877557e-06, + "loss": 0.9737, + "step": 21919 + }, + { + "epoch": 0.53, + "learning_rate": 9.606759750756239e-06, + "loss": 0.9278, + "step": 21920 + }, + { + "epoch": 0.53, + "learning_rate": 9.605981460020621e-06, + "loss": 0.8404, + "step": 21921 + }, + { + "epoch": 0.53, + "learning_rate": 9.60520317167542e-06, + "loss": 1.0457, + "step": 21922 + }, + { + "epoch": 0.53, + "learning_rate": 9.60442488572536e-06, + "loss": 0.9301, + "step": 21923 + }, + { + "epoch": 0.53, + "learning_rate": 9.603646602175158e-06, + "loss": 1.3101, + "step": 21924 + }, + { + "epoch": 0.53, + "learning_rate": 9.60286832102954e-06, + "loss": 0.7369, + "step": 21925 + }, + { + "epoch": 0.53, + "learning_rate": 9.602090042293228e-06, + "loss": 0.9952, + "step": 21926 + }, + { + "epoch": 0.53, + "learning_rate": 9.60131176597094e-06, + "loss": 0.8946, + "step": 21927 + }, + { + "epoch": 0.53, + "learning_rate": 9.600533492067398e-06, + "loss": 1.1145, + "step": 21928 + }, + { + "epoch": 0.53, + "learning_rate": 9.59975522058733e-06, + "loss": 0.7499, + "step": 21929 + }, + { + "epoch": 0.53, + "learning_rate": 9.598976951535451e-06, + "loss": 1.2764, + "step": 21930 + }, + { + "epoch": 0.53, + "learning_rate": 9.598198684916483e-06, + "loss": 1.1746, + "step": 21931 + }, + { + "epoch": 0.53, + "learning_rate": 9.597420420735152e-06, + "loss": 1.0335, + "step": 21932 + }, + { + "epoch": 0.53, + "learning_rate": 9.596642158996174e-06, + "loss": 1.0815, + "step": 21933 + }, + { + "epoch": 0.53, + "learning_rate": 9.595863899704276e-06, + "loss": 1.1173, + "step": 21934 + }, + { + "epoch": 0.53, + "learning_rate": 9.595085642864173e-06, + "loss": 1.0166, + "step": 21935 + }, + { + "epoch": 0.53, + "learning_rate": 9.594307388480592e-06, + "loss": 1.1159, + "step": 21936 + }, + { + "epoch": 0.53, + "learning_rate": 9.593529136558254e-06, + "loss": 0.9685, + "step": 21937 + }, + { + "epoch": 0.53, + "learning_rate": 9.592750887101877e-06, + "loss": 1.122, + "step": 21938 + }, + { + "epoch": 0.53, + "learning_rate": 9.591972640116184e-06, + "loss": 1.0864, + "step": 21939 + }, + { + "epoch": 0.53, + "learning_rate": 9.591194395605899e-06, + "loss": 1.2104, + "step": 21940 + }, + { + "epoch": 0.53, + "learning_rate": 9.590416153575741e-06, + "loss": 0.9292, + "step": 21941 + }, + { + "epoch": 0.53, + "learning_rate": 9.589637914030433e-06, + "loss": 0.8169, + "step": 21942 + }, + { + "epoch": 0.53, + "learning_rate": 9.588859676974691e-06, + "loss": 1.2982, + "step": 21943 + }, + { + "epoch": 0.53, + "learning_rate": 9.588081442413245e-06, + "loss": 1.1672, + "step": 21944 + }, + { + "epoch": 0.53, + "learning_rate": 9.58730321035081e-06, + "loss": 1.1551, + "step": 21945 + }, + { + "epoch": 0.53, + "learning_rate": 9.58652498079211e-06, + "loss": 1.1029, + "step": 21946 + }, + { + "epoch": 0.53, + "learning_rate": 9.585746753741866e-06, + "loss": 0.8874, + "step": 21947 + }, + { + "epoch": 0.53, + "learning_rate": 9.5849685292048e-06, + "loss": 1.0157, + "step": 21948 + }, + { + "epoch": 0.53, + "learning_rate": 9.58419030718563e-06, + "loss": 1.0387, + "step": 21949 + }, + { + "epoch": 0.53, + "learning_rate": 9.58341208768908e-06, + "loss": 1.1748, + "step": 21950 + }, + { + "epoch": 0.53, + "learning_rate": 9.58263387071987e-06, + "loss": 0.9578, + "step": 21951 + }, + { + "epoch": 0.53, + "learning_rate": 9.581855656282724e-06, + "loss": 0.8751, + "step": 21952 + }, + { + "epoch": 0.53, + "learning_rate": 9.581077444382362e-06, + "loss": 1.0875, + "step": 21953 + }, + { + "epoch": 0.53, + "learning_rate": 9.580299235023505e-06, + "loss": 1.3028, + "step": 21954 + }, + { + "epoch": 0.53, + "learning_rate": 9.57952102821087e-06, + "loss": 1.2589, + "step": 21955 + }, + { + "epoch": 0.53, + "learning_rate": 9.578742823949185e-06, + "loss": 1.0845, + "step": 21956 + }, + { + "epoch": 0.53, + "learning_rate": 9.577964622243168e-06, + "loss": 1.0858, + "step": 21957 + }, + { + "epoch": 0.53, + "learning_rate": 9.57718642309754e-06, + "loss": 1.0666, + "step": 21958 + }, + { + "epoch": 0.53, + "learning_rate": 9.57640822651702e-06, + "loss": 0.9684, + "step": 21959 + }, + { + "epoch": 0.53, + "learning_rate": 9.575630032506335e-06, + "loss": 1.115, + "step": 21960 + }, + { + "epoch": 0.53, + "learning_rate": 9.574851841070203e-06, + "loss": 1.1198, + "step": 21961 + }, + { + "epoch": 0.53, + "learning_rate": 9.574073652213345e-06, + "loss": 1.0173, + "step": 21962 + }, + { + "epoch": 0.53, + "learning_rate": 9.573295465940478e-06, + "loss": 1.1486, + "step": 21963 + }, + { + "epoch": 0.53, + "learning_rate": 9.57251728225633e-06, + "loss": 1.0988, + "step": 21964 + }, + { + "epoch": 0.53, + "learning_rate": 9.57173910116562e-06, + "loss": 1.0108, + "step": 21965 + }, + { + "epoch": 0.53, + "learning_rate": 9.570960922673068e-06, + "loss": 1.1401, + "step": 21966 + }, + { + "epoch": 0.53, + "learning_rate": 9.570182746783392e-06, + "loss": 1.1511, + "step": 21967 + }, + { + "epoch": 0.53, + "learning_rate": 9.56940457350132e-06, + "loss": 0.95, + "step": 21968 + }, + { + "epoch": 0.53, + "learning_rate": 9.568626402831568e-06, + "loss": 1.1182, + "step": 21969 + }, + { + "epoch": 0.53, + "learning_rate": 9.567848234778857e-06, + "loss": 0.7423, + "step": 21970 + }, + { + "epoch": 0.53, + "learning_rate": 9.567070069347909e-06, + "loss": 1.1992, + "step": 21971 + }, + { + "epoch": 0.53, + "learning_rate": 9.566291906543446e-06, + "loss": 1.0348, + "step": 21972 + }, + { + "epoch": 0.53, + "learning_rate": 9.565513746370189e-06, + "loss": 0.7595, + "step": 21973 + }, + { + "epoch": 0.53, + "learning_rate": 9.564735588832857e-06, + "loss": 1.0352, + "step": 21974 + }, + { + "epoch": 0.53, + "learning_rate": 9.563957433936169e-06, + "loss": 0.6744, + "step": 21975 + }, + { + "epoch": 0.53, + "learning_rate": 9.563179281684853e-06, + "loss": 1.097, + "step": 21976 + }, + { + "epoch": 0.53, + "learning_rate": 9.562401132083624e-06, + "loss": 1.0334, + "step": 21977 + }, + { + "epoch": 0.53, + "learning_rate": 9.561622985137202e-06, + "loss": 0.8568, + "step": 21978 + }, + { + "epoch": 0.53, + "learning_rate": 9.560844840850314e-06, + "loss": 1.1543, + "step": 21979 + }, + { + "epoch": 0.53, + "learning_rate": 9.560066699227675e-06, + "loss": 0.9289, + "step": 21980 + }, + { + "epoch": 0.53, + "learning_rate": 9.55928856027401e-06, + "loss": 1.0246, + "step": 21981 + }, + { + "epoch": 0.53, + "learning_rate": 9.558510423994033e-06, + "loss": 0.8601, + "step": 21982 + }, + { + "epoch": 0.53, + "learning_rate": 9.557732290392474e-06, + "loss": 1.2552, + "step": 21983 + }, + { + "epoch": 0.53, + "learning_rate": 9.556954159474048e-06, + "loss": 0.7017, + "step": 21984 + }, + { + "epoch": 0.53, + "learning_rate": 9.556176031243475e-06, + "loss": 1.1198, + "step": 21985 + }, + { + "epoch": 0.53, + "learning_rate": 9.55539790570548e-06, + "loss": 1.0977, + "step": 21986 + }, + { + "epoch": 0.53, + "learning_rate": 9.55461978286478e-06, + "loss": 1.2185, + "step": 21987 + }, + { + "epoch": 0.53, + "learning_rate": 9.553841662726098e-06, + "loss": 0.7018, + "step": 21988 + }, + { + "epoch": 0.53, + "learning_rate": 9.553063545294154e-06, + "loss": 0.9865, + "step": 21989 + }, + { + "epoch": 0.53, + "learning_rate": 9.552285430573665e-06, + "loss": 1.1849, + "step": 21990 + }, + { + "epoch": 0.53, + "learning_rate": 9.551507318569358e-06, + "loss": 0.79, + "step": 21991 + }, + { + "epoch": 0.53, + "learning_rate": 9.55072920928595e-06, + "loss": 0.9587, + "step": 21992 + }, + { + "epoch": 0.53, + "learning_rate": 9.549951102728164e-06, + "loss": 1.0972, + "step": 21993 + }, + { + "epoch": 0.53, + "learning_rate": 9.549172998900715e-06, + "loss": 1.1703, + "step": 21994 + }, + { + "epoch": 0.53, + "learning_rate": 9.548394897808329e-06, + "loss": 0.809, + "step": 21995 + }, + { + "epoch": 0.53, + "learning_rate": 9.547616799455727e-06, + "loss": 0.9779, + "step": 21996 + }, + { + "epoch": 0.53, + "learning_rate": 9.546838703847627e-06, + "loss": 0.9685, + "step": 21997 + }, + { + "epoch": 0.53, + "learning_rate": 9.546060610988747e-06, + "loss": 1.0334, + "step": 21998 + }, + { + "epoch": 0.53, + "learning_rate": 9.54528252088381e-06, + "loss": 1.2355, + "step": 21999 + }, + { + "epoch": 0.53, + "learning_rate": 9.544504433537537e-06, + "loss": 1.0285, + "step": 22000 + }, + { + "epoch": 0.53, + "learning_rate": 9.543726348954651e-06, + "loss": 0.9606, + "step": 22001 + }, + { + "epoch": 0.53, + "learning_rate": 9.542948267139871e-06, + "loss": 1.0044, + "step": 22002 + }, + { + "epoch": 0.53, + "learning_rate": 9.542170188097916e-06, + "loss": 1.1988, + "step": 22003 + }, + { + "epoch": 0.53, + "learning_rate": 9.541392111833504e-06, + "loss": 1.0713, + "step": 22004 + }, + { + "epoch": 0.53, + "learning_rate": 9.54061403835136e-06, + "loss": 1.1396, + "step": 22005 + }, + { + "epoch": 0.53, + "learning_rate": 9.539835967656204e-06, + "loss": 1.1849, + "step": 22006 + }, + { + "epoch": 0.53, + "learning_rate": 9.539057899752755e-06, + "loss": 1.2987, + "step": 22007 + }, + { + "epoch": 0.53, + "learning_rate": 9.538279834645728e-06, + "loss": 0.9837, + "step": 22008 + }, + { + "epoch": 0.53, + "learning_rate": 9.537501772339853e-06, + "loss": 1.2079, + "step": 22009 + }, + { + "epoch": 0.53, + "learning_rate": 9.536723712839848e-06, + "loss": 0.9362, + "step": 22010 + }, + { + "epoch": 0.53, + "learning_rate": 9.535945656150426e-06, + "loss": 1.0259, + "step": 22011 + }, + { + "epoch": 0.53, + "learning_rate": 9.535167602276315e-06, + "loss": 1.1508, + "step": 22012 + }, + { + "epoch": 0.53, + "learning_rate": 9.534389551222235e-06, + "loss": 1.1026, + "step": 22013 + }, + { + "epoch": 0.53, + "learning_rate": 9.533611502992902e-06, + "loss": 1.2343, + "step": 22014 + }, + { + "epoch": 0.53, + "learning_rate": 9.532833457593038e-06, + "loss": 0.8727, + "step": 22015 + }, + { + "epoch": 0.53, + "learning_rate": 9.532055415027364e-06, + "loss": 1.0557, + "step": 22016 + }, + { + "epoch": 0.53, + "learning_rate": 9.531277375300599e-06, + "loss": 1.0837, + "step": 22017 + }, + { + "epoch": 0.53, + "learning_rate": 9.530499338417466e-06, + "loss": 0.7595, + "step": 22018 + }, + { + "epoch": 0.53, + "learning_rate": 9.529721304382679e-06, + "loss": 0.9938, + "step": 22019 + }, + { + "epoch": 0.53, + "learning_rate": 9.528943273200965e-06, + "loss": 1.1235, + "step": 22020 + }, + { + "epoch": 0.53, + "learning_rate": 9.528165244877043e-06, + "loss": 1.1246, + "step": 22021 + }, + { + "epoch": 0.53, + "learning_rate": 9.52738721941563e-06, + "loss": 1.4124, + "step": 22022 + }, + { + "epoch": 0.53, + "learning_rate": 9.526609196821445e-06, + "loss": 1.1176, + "step": 22023 + }, + { + "epoch": 0.53, + "learning_rate": 9.525831177099215e-06, + "loss": 1.1588, + "step": 22024 + }, + { + "epoch": 0.53, + "learning_rate": 9.525053160253654e-06, + "loss": 1.1231, + "step": 22025 + }, + { + "epoch": 0.53, + "learning_rate": 9.524275146289484e-06, + "loss": 0.9803, + "step": 22026 + }, + { + "epoch": 0.53, + "learning_rate": 9.523497135211422e-06, + "loss": 0.92, + "step": 22027 + }, + { + "epoch": 0.53, + "learning_rate": 9.522719127024195e-06, + "loss": 1.2771, + "step": 22028 + }, + { + "epoch": 0.53, + "learning_rate": 9.521941121732516e-06, + "loss": 1.1187, + "step": 22029 + }, + { + "epoch": 0.53, + "learning_rate": 9.52116311934111e-06, + "loss": 1.3236, + "step": 22030 + }, + { + "epoch": 0.53, + "learning_rate": 9.520385119854691e-06, + "loss": 1.1107, + "step": 22031 + }, + { + "epoch": 0.53, + "learning_rate": 9.519607123277987e-06, + "loss": 1.0575, + "step": 22032 + }, + { + "epoch": 0.53, + "learning_rate": 9.518829129615711e-06, + "loss": 1.3702, + "step": 22033 + }, + { + "epoch": 0.53, + "learning_rate": 9.518051138872587e-06, + "loss": 0.8938, + "step": 22034 + }, + { + "epoch": 0.53, + "learning_rate": 9.517273151053331e-06, + "loss": 1.3185, + "step": 22035 + }, + { + "epoch": 0.53, + "learning_rate": 9.516495166162667e-06, + "loss": 0.9732, + "step": 22036 + }, + { + "epoch": 0.53, + "learning_rate": 9.515717184205313e-06, + "loss": 1.0287, + "step": 22037 + }, + { + "epoch": 0.53, + "learning_rate": 9.51493920518599e-06, + "loss": 1.1404, + "step": 22038 + }, + { + "epoch": 0.53, + "learning_rate": 9.514161229109413e-06, + "loss": 0.8579, + "step": 22039 + }, + { + "epoch": 0.53, + "learning_rate": 9.51338325598031e-06, + "loss": 0.9131, + "step": 22040 + }, + { + "epoch": 0.53, + "learning_rate": 9.512605285803395e-06, + "loss": 1.0927, + "step": 22041 + }, + { + "epoch": 0.53, + "learning_rate": 9.51182731858339e-06, + "loss": 1.2311, + "step": 22042 + }, + { + "epoch": 0.53, + "learning_rate": 9.51104935432501e-06, + "loss": 1.1027, + "step": 22043 + }, + { + "epoch": 0.53, + "learning_rate": 9.510271393032981e-06, + "loss": 1.0742, + "step": 22044 + }, + { + "epoch": 0.53, + "learning_rate": 9.50949343471202e-06, + "loss": 1.1374, + "step": 22045 + }, + { + "epoch": 0.53, + "learning_rate": 9.508715479366848e-06, + "loss": 0.9728, + "step": 22046 + }, + { + "epoch": 0.53, + "learning_rate": 9.50793752700218e-06, + "loss": 1.2371, + "step": 22047 + }, + { + "epoch": 0.53, + "learning_rate": 9.507159577622743e-06, + "loss": 1.0065, + "step": 22048 + }, + { + "epoch": 0.53, + "learning_rate": 9.506381631233253e-06, + "loss": 1.1367, + "step": 22049 + }, + { + "epoch": 0.53, + "learning_rate": 9.50560368783843e-06, + "loss": 0.82, + "step": 22050 + }, + { + "epoch": 0.53, + "learning_rate": 9.504825747442988e-06, + "loss": 1.0182, + "step": 22051 + }, + { + "epoch": 0.53, + "learning_rate": 9.504047810051656e-06, + "loss": 1.0316, + "step": 22052 + }, + { + "epoch": 0.53, + "learning_rate": 9.50326987566915e-06, + "loss": 1.0431, + "step": 22053 + }, + { + "epoch": 0.53, + "learning_rate": 9.502491944300188e-06, + "loss": 0.9884, + "step": 22054 + }, + { + "epoch": 0.53, + "learning_rate": 9.501714015949487e-06, + "loss": 0.8025, + "step": 22055 + }, + { + "epoch": 0.53, + "learning_rate": 9.500936090621774e-06, + "loss": 0.6782, + "step": 22056 + }, + { + "epoch": 0.53, + "learning_rate": 9.500158168321763e-06, + "loss": 1.1843, + "step": 22057 + }, + { + "epoch": 0.53, + "learning_rate": 9.499380249054174e-06, + "loss": 1.2522, + "step": 22058 + }, + { + "epoch": 0.53, + "learning_rate": 9.498602332823729e-06, + "loss": 1.0538, + "step": 22059 + }, + { + "epoch": 0.53, + "learning_rate": 9.497824419635145e-06, + "loss": 0.8259, + "step": 22060 + }, + { + "epoch": 0.53, + "learning_rate": 9.497046509493142e-06, + "loss": 0.8405, + "step": 22061 + }, + { + "epoch": 0.53, + "learning_rate": 9.49626860240244e-06, + "loss": 1.0984, + "step": 22062 + }, + { + "epoch": 0.53, + "learning_rate": 9.495490698367757e-06, + "loss": 0.8527, + "step": 22063 + }, + { + "epoch": 0.53, + "learning_rate": 9.494712797393813e-06, + "loss": 1.0035, + "step": 22064 + }, + { + "epoch": 0.53, + "learning_rate": 9.493934899485329e-06, + "loss": 1.1551, + "step": 22065 + }, + { + "epoch": 0.53, + "learning_rate": 9.49315700464702e-06, + "loss": 1.1978, + "step": 22066 + }, + { + "epoch": 0.53, + "learning_rate": 9.492379112883612e-06, + "loss": 0.7063, + "step": 22067 + }, + { + "epoch": 0.53, + "learning_rate": 9.49160122419982e-06, + "loss": 1.0367, + "step": 22068 + }, + { + "epoch": 0.53, + "learning_rate": 9.490823338600362e-06, + "loss": 1.1532, + "step": 22069 + }, + { + "epoch": 0.53, + "learning_rate": 9.490045456089958e-06, + "loss": 1.1261, + "step": 22070 + }, + { + "epoch": 0.53, + "learning_rate": 9.489267576673331e-06, + "loss": 1.1732, + "step": 22071 + }, + { + "epoch": 0.53, + "learning_rate": 9.488489700355197e-06, + "loss": 1.1025, + "step": 22072 + }, + { + "epoch": 0.53, + "learning_rate": 9.487711827140276e-06, + "loss": 1.17, + "step": 22073 + }, + { + "epoch": 0.53, + "learning_rate": 9.486933957033283e-06, + "loss": 1.1676, + "step": 22074 + }, + { + "epoch": 0.53, + "learning_rate": 9.486156090038945e-06, + "loss": 1.1822, + "step": 22075 + }, + { + "epoch": 0.53, + "learning_rate": 9.485378226161978e-06, + "loss": 1.0361, + "step": 22076 + }, + { + "epoch": 0.53, + "learning_rate": 9.484600365407099e-06, + "loss": 0.9261, + "step": 22077 + }, + { + "epoch": 0.53, + "learning_rate": 9.483822507779026e-06, + "loss": 1.2155, + "step": 22078 + }, + { + "epoch": 0.53, + "learning_rate": 9.483044653282483e-06, + "loss": 1.0203, + "step": 22079 + }, + { + "epoch": 0.53, + "learning_rate": 9.482266801922186e-06, + "loss": 1.0208, + "step": 22080 + }, + { + "epoch": 0.53, + "learning_rate": 9.481488953702855e-06, + "loss": 0.9263, + "step": 22081 + }, + { + "epoch": 0.53, + "learning_rate": 9.480711108629206e-06, + "loss": 0.9239, + "step": 22082 + }, + { + "epoch": 0.53, + "learning_rate": 9.479933266705963e-06, + "loss": 1.207, + "step": 22083 + }, + { + "epoch": 0.53, + "learning_rate": 9.479155427937843e-06, + "loss": 1.0978, + "step": 22084 + }, + { + "epoch": 0.53, + "learning_rate": 9.478377592329563e-06, + "loss": 1.0141, + "step": 22085 + }, + { + "epoch": 0.53, + "learning_rate": 9.477599759885842e-06, + "loss": 1.1503, + "step": 22086 + }, + { + "epoch": 0.53, + "learning_rate": 9.476821930611402e-06, + "loss": 0.8783, + "step": 22087 + }, + { + "epoch": 0.53, + "learning_rate": 9.476044104510961e-06, + "loss": 1.074, + "step": 22088 + }, + { + "epoch": 0.53, + "learning_rate": 9.475266281589238e-06, + "loss": 1.3384, + "step": 22089 + }, + { + "epoch": 0.53, + "learning_rate": 9.474488461850948e-06, + "loss": 1.2282, + "step": 22090 + }, + { + "epoch": 0.53, + "learning_rate": 9.473710645300815e-06, + "loss": 0.9116, + "step": 22091 + }, + { + "epoch": 0.53, + "learning_rate": 9.472932831943556e-06, + "loss": 0.926, + "step": 22092 + }, + { + "epoch": 0.53, + "learning_rate": 9.472155021783889e-06, + "loss": 1.1863, + "step": 22093 + }, + { + "epoch": 0.53, + "learning_rate": 9.471377214826533e-06, + "loss": 1.1426, + "step": 22094 + }, + { + "epoch": 0.53, + "learning_rate": 9.470599411076207e-06, + "loss": 1.0071, + "step": 22095 + }, + { + "epoch": 0.53, + "learning_rate": 9.469821610537631e-06, + "loss": 1.1503, + "step": 22096 + }, + { + "epoch": 0.53, + "learning_rate": 9.469043813215522e-06, + "loss": 1.0793, + "step": 22097 + }, + { + "epoch": 0.53, + "learning_rate": 9.468266019114599e-06, + "loss": 1.2135, + "step": 22098 + }, + { + "epoch": 0.53, + "learning_rate": 9.46748822823958e-06, + "loss": 1.3915, + "step": 22099 + }, + { + "epoch": 0.53, + "learning_rate": 9.466710440595186e-06, + "loss": 1.2533, + "step": 22100 + }, + { + "epoch": 0.53, + "learning_rate": 9.465932656186136e-06, + "loss": 1.0676, + "step": 22101 + }, + { + "epoch": 0.53, + "learning_rate": 9.465154875017142e-06, + "loss": 1.3444, + "step": 22102 + }, + { + "epoch": 0.53, + "learning_rate": 9.464377097092931e-06, + "loss": 1.1598, + "step": 22103 + }, + { + "epoch": 0.53, + "learning_rate": 9.463599322418218e-06, + "loss": 1.0057, + "step": 22104 + }, + { + "epoch": 0.53, + "learning_rate": 9.46282155099772e-06, + "loss": 0.9869, + "step": 22105 + }, + { + "epoch": 0.53, + "learning_rate": 9.46204378283616e-06, + "loss": 0.9157, + "step": 22106 + }, + { + "epoch": 0.53, + "learning_rate": 9.461266017938253e-06, + "loss": 0.8263, + "step": 22107 + }, + { + "epoch": 0.53, + "learning_rate": 9.460488256308718e-06, + "loss": 1.0153, + "step": 22108 + }, + { + "epoch": 0.53, + "learning_rate": 9.459710497952273e-06, + "loss": 1.0262, + "step": 22109 + }, + { + "epoch": 0.53, + "learning_rate": 9.458932742873639e-06, + "loss": 1.2723, + "step": 22110 + }, + { + "epoch": 0.53, + "learning_rate": 9.458154991077532e-06, + "loss": 1.0075, + "step": 22111 + }, + { + "epoch": 0.53, + "learning_rate": 9.457377242568671e-06, + "loss": 1.0969, + "step": 22112 + }, + { + "epoch": 0.53, + "learning_rate": 9.456599497351774e-06, + "loss": 0.8222, + "step": 22113 + }, + { + "epoch": 0.53, + "learning_rate": 9.455821755431561e-06, + "loss": 1.1404, + "step": 22114 + }, + { + "epoch": 0.53, + "learning_rate": 9.45504401681275e-06, + "loss": 1.194, + "step": 22115 + }, + { + "epoch": 0.53, + "learning_rate": 9.454266281500059e-06, + "loss": 1.1857, + "step": 22116 + }, + { + "epoch": 0.53, + "learning_rate": 9.453488549498202e-06, + "loss": 1.1649, + "step": 22117 + }, + { + "epoch": 0.53, + "learning_rate": 9.452710820811905e-06, + "loss": 1.0419, + "step": 22118 + }, + { + "epoch": 0.53, + "learning_rate": 9.451933095445884e-06, + "loss": 1.0116, + "step": 22119 + }, + { + "epoch": 0.53, + "learning_rate": 9.451155373404855e-06, + "loss": 1.0798, + "step": 22120 + }, + { + "epoch": 0.53, + "learning_rate": 9.450377654693534e-06, + "loss": 0.9825, + "step": 22121 + }, + { + "epoch": 0.53, + "learning_rate": 9.449599939316645e-06, + "loss": 0.9681, + "step": 22122 + }, + { + "epoch": 0.53, + "learning_rate": 9.448822227278905e-06, + "loss": 1.0178, + "step": 22123 + }, + { + "epoch": 0.53, + "learning_rate": 9.44804451858503e-06, + "loss": 1.1602, + "step": 22124 + }, + { + "epoch": 0.53, + "learning_rate": 9.447266813239737e-06, + "loss": 1.092, + "step": 22125 + }, + { + "epoch": 0.53, + "learning_rate": 9.446489111247749e-06, + "loss": 0.9078, + "step": 22126 + }, + { + "epoch": 0.53, + "learning_rate": 9.44571141261378e-06, + "loss": 1.084, + "step": 22127 + }, + { + "epoch": 0.53, + "learning_rate": 9.44493371734255e-06, + "loss": 1.2119, + "step": 22128 + }, + { + "epoch": 0.53, + "learning_rate": 9.444156025438776e-06, + "loss": 0.9825, + "step": 22129 + }, + { + "epoch": 0.53, + "learning_rate": 9.443378336907177e-06, + "loss": 1.0303, + "step": 22130 + }, + { + "epoch": 0.53, + "learning_rate": 9.442600651752471e-06, + "loss": 1.1223, + "step": 22131 + }, + { + "epoch": 0.53, + "learning_rate": 9.441822969979377e-06, + "loss": 0.7428, + "step": 22132 + }, + { + "epoch": 0.53, + "learning_rate": 9.441045291592608e-06, + "loss": 1.1525, + "step": 22133 + }, + { + "epoch": 0.53, + "learning_rate": 9.440267616596888e-06, + "loss": 1.1209, + "step": 22134 + }, + { + "epoch": 0.53, + "learning_rate": 9.439489944996932e-06, + "loss": 1.0979, + "step": 22135 + }, + { + "epoch": 0.53, + "learning_rate": 9.438712276797461e-06, + "loss": 1.0555, + "step": 22136 + }, + { + "epoch": 0.53, + "learning_rate": 9.437934612003186e-06, + "loss": 1.019, + "step": 22137 + }, + { + "epoch": 0.53, + "learning_rate": 9.437156950618834e-06, + "loss": 1.2846, + "step": 22138 + }, + { + "epoch": 0.53, + "learning_rate": 9.436379292649117e-06, + "loss": 1.0117, + "step": 22139 + }, + { + "epoch": 0.53, + "learning_rate": 9.435601638098755e-06, + "loss": 1.0311, + "step": 22140 + }, + { + "epoch": 0.53, + "learning_rate": 9.434823986972463e-06, + "loss": 1.009, + "step": 22141 + }, + { + "epoch": 0.53, + "learning_rate": 9.434046339274962e-06, + "loss": 1.0126, + "step": 22142 + }, + { + "epoch": 0.53, + "learning_rate": 9.433268695010968e-06, + "loss": 1.0754, + "step": 22143 + }, + { + "epoch": 0.53, + "learning_rate": 9.432491054185203e-06, + "loss": 1.3567, + "step": 22144 + }, + { + "epoch": 0.53, + "learning_rate": 9.431713416802376e-06, + "loss": 1.0679, + "step": 22145 + }, + { + "epoch": 0.53, + "learning_rate": 9.430935782867214e-06, + "loss": 0.9393, + "step": 22146 + }, + { + "epoch": 0.53, + "learning_rate": 9.430158152384429e-06, + "loss": 1.154, + "step": 22147 + }, + { + "epoch": 0.53, + "learning_rate": 9.429380525358738e-06, + "loss": 1.2588, + "step": 22148 + }, + { + "epoch": 0.53, + "learning_rate": 9.428602901794866e-06, + "loss": 0.9625, + "step": 22149 + }, + { + "epoch": 0.53, + "learning_rate": 9.427825281697523e-06, + "loss": 1.1258, + "step": 22150 + }, + { + "epoch": 0.53, + "learning_rate": 9.427047665071432e-06, + "loss": 1.1233, + "step": 22151 + }, + { + "epoch": 0.53, + "learning_rate": 9.426270051921305e-06, + "loss": 1.1795, + "step": 22152 + }, + { + "epoch": 0.53, + "learning_rate": 9.425492442251863e-06, + "loss": 1.0306, + "step": 22153 + }, + { + "epoch": 0.53, + "learning_rate": 9.424714836067826e-06, + "loss": 0.9602, + "step": 22154 + }, + { + "epoch": 0.53, + "learning_rate": 9.423937233373907e-06, + "loss": 0.772, + "step": 22155 + }, + { + "epoch": 0.53, + "learning_rate": 9.423159634174824e-06, + "loss": 1.3459, + "step": 22156 + }, + { + "epoch": 0.53, + "learning_rate": 9.422382038475296e-06, + "loss": 1.0209, + "step": 22157 + }, + { + "epoch": 0.53, + "learning_rate": 9.421604446280043e-06, + "loss": 1.1312, + "step": 22158 + }, + { + "epoch": 0.53, + "learning_rate": 9.420826857593779e-06, + "loss": 1.124, + "step": 22159 + }, + { + "epoch": 0.53, + "learning_rate": 9.420049272421219e-06, + "loss": 1.227, + "step": 22160 + }, + { + "epoch": 0.53, + "learning_rate": 9.419271690767087e-06, + "loss": 1.1576, + "step": 22161 + }, + { + "epoch": 0.53, + "learning_rate": 9.418494112636097e-06, + "loss": 0.7896, + "step": 22162 + }, + { + "epoch": 0.53, + "learning_rate": 9.417716538032967e-06, + "loss": 1.0582, + "step": 22163 + }, + { + "epoch": 0.53, + "learning_rate": 9.41693896696241e-06, + "loss": 1.124, + "step": 22164 + }, + { + "epoch": 0.53, + "learning_rate": 9.416161399429148e-06, + "loss": 1.0631, + "step": 22165 + }, + { + "epoch": 0.53, + "learning_rate": 9.415383835437896e-06, + "loss": 1.1702, + "step": 22166 + }, + { + "epoch": 0.53, + "learning_rate": 9.414606274993376e-06, + "loss": 1.2342, + "step": 22167 + }, + { + "epoch": 0.53, + "learning_rate": 9.413828718100304e-06, + "loss": 1.1279, + "step": 22168 + }, + { + "epoch": 0.53, + "learning_rate": 9.413051164763391e-06, + "loss": 1.0918, + "step": 22169 + }, + { + "epoch": 0.53, + "learning_rate": 9.412273614987359e-06, + "loss": 0.93, + "step": 22170 + }, + { + "epoch": 0.53, + "learning_rate": 9.411496068776928e-06, + "loss": 1.1922, + "step": 22171 + }, + { + "epoch": 0.53, + "learning_rate": 9.410718526136811e-06, + "loss": 1.0035, + "step": 22172 + }, + { + "epoch": 0.53, + "learning_rate": 9.409940987071726e-06, + "loss": 1.016, + "step": 22173 + }, + { + "epoch": 0.53, + "learning_rate": 9.409163451586387e-06, + "loss": 0.9075, + "step": 22174 + }, + { + "epoch": 0.53, + "learning_rate": 9.408385919685518e-06, + "loss": 1.3998, + "step": 22175 + }, + { + "epoch": 0.53, + "learning_rate": 9.407608391373831e-06, + "loss": 1.0139, + "step": 22176 + }, + { + "epoch": 0.53, + "learning_rate": 9.406830866656046e-06, + "loss": 1.1371, + "step": 22177 + }, + { + "epoch": 0.53, + "learning_rate": 9.406053345536875e-06, + "loss": 1.19, + "step": 22178 + }, + { + "epoch": 0.53, + "learning_rate": 9.405275828021042e-06, + "loss": 1.0393, + "step": 22179 + }, + { + "epoch": 0.53, + "learning_rate": 9.40449831411326e-06, + "loss": 1.0106, + "step": 22180 + }, + { + "epoch": 0.53, + "learning_rate": 9.403720803818245e-06, + "loss": 1.0971, + "step": 22181 + }, + { + "epoch": 0.53, + "learning_rate": 9.402943297140718e-06, + "loss": 1.179, + "step": 22182 + }, + { + "epoch": 0.53, + "learning_rate": 9.402165794085394e-06, + "loss": 0.9914, + "step": 22183 + }, + { + "epoch": 0.53, + "learning_rate": 9.401388294656987e-06, + "loss": 1.1751, + "step": 22184 + }, + { + "epoch": 0.53, + "learning_rate": 9.400610798860217e-06, + "loss": 0.9443, + "step": 22185 + }, + { + "epoch": 0.53, + "learning_rate": 9.399833306699801e-06, + "loss": 1.0294, + "step": 22186 + }, + { + "epoch": 0.53, + "learning_rate": 9.399055818180455e-06, + "loss": 1.0058, + "step": 22187 + }, + { + "epoch": 0.53, + "learning_rate": 9.398278333306898e-06, + "loss": 1.0646, + "step": 22188 + }, + { + "epoch": 0.53, + "learning_rate": 9.397500852083841e-06, + "loss": 0.9949, + "step": 22189 + }, + { + "epoch": 0.53, + "learning_rate": 9.396723374516006e-06, + "loss": 1.0622, + "step": 22190 + }, + { + "epoch": 0.53, + "learning_rate": 9.39594590060811e-06, + "loss": 1.2474, + "step": 22191 + }, + { + "epoch": 0.53, + "learning_rate": 9.395168430364869e-06, + "loss": 1.1745, + "step": 22192 + }, + { + "epoch": 0.53, + "learning_rate": 9.394390963790995e-06, + "loss": 0.8369, + "step": 22193 + }, + { + "epoch": 0.53, + "learning_rate": 9.39361350089121e-06, + "loss": 1.1408, + "step": 22194 + }, + { + "epoch": 0.53, + "learning_rate": 9.39283604167023e-06, + "loss": 1.193, + "step": 22195 + }, + { + "epoch": 0.53, + "learning_rate": 9.392058586132772e-06, + "loss": 0.9391, + "step": 22196 + }, + { + "epoch": 0.53, + "learning_rate": 9.391281134283549e-06, + "loss": 1.5042, + "step": 22197 + }, + { + "epoch": 0.53, + "learning_rate": 9.390503686127281e-06, + "loss": 0.7159, + "step": 22198 + }, + { + "epoch": 0.53, + "learning_rate": 9.389726241668685e-06, + "loss": 0.9412, + "step": 22199 + }, + { + "epoch": 0.53, + "learning_rate": 9.388948800912476e-06, + "loss": 0.9199, + "step": 22200 + }, + { + "epoch": 0.53, + "learning_rate": 9.388171363863369e-06, + "loss": 0.8523, + "step": 22201 + }, + { + "epoch": 0.53, + "learning_rate": 9.387393930526084e-06, + "loss": 1.1635, + "step": 22202 + }, + { + "epoch": 0.53, + "learning_rate": 9.386616500905337e-06, + "loss": 1.1166, + "step": 22203 + }, + { + "epoch": 0.53, + "learning_rate": 9.385839075005842e-06, + "loss": 0.8853, + "step": 22204 + }, + { + "epoch": 0.53, + "learning_rate": 9.385061652832315e-06, + "loss": 0.9191, + "step": 22205 + }, + { + "epoch": 0.53, + "learning_rate": 9.384284234389479e-06, + "loss": 0.8146, + "step": 22206 + }, + { + "epoch": 0.53, + "learning_rate": 9.383506819682042e-06, + "loss": 1.044, + "step": 22207 + }, + { + "epoch": 0.53, + "learning_rate": 9.382729408714725e-06, + "loss": 1.0808, + "step": 22208 + }, + { + "epoch": 0.53, + "learning_rate": 9.381952001492243e-06, + "loss": 1.2007, + "step": 22209 + }, + { + "epoch": 0.53, + "learning_rate": 9.381174598019313e-06, + "loss": 1.0547, + "step": 22210 + }, + { + "epoch": 0.53, + "learning_rate": 9.380397198300653e-06, + "loss": 1.2032, + "step": 22211 + }, + { + "epoch": 0.53, + "learning_rate": 9.379619802340976e-06, + "loss": 0.8273, + "step": 22212 + }, + { + "epoch": 0.53, + "learning_rate": 9.378842410144999e-06, + "loss": 0.9767, + "step": 22213 + }, + { + "epoch": 0.53, + "learning_rate": 9.37806502171744e-06, + "loss": 0.9718, + "step": 22214 + }, + { + "epoch": 0.53, + "learning_rate": 9.377287637063016e-06, + "loss": 1.0858, + "step": 22215 + }, + { + "epoch": 0.53, + "learning_rate": 9.37651025618644e-06, + "loss": 0.983, + "step": 22216 + }, + { + "epoch": 0.53, + "learning_rate": 9.375732879092426e-06, + "loss": 1.1699, + "step": 22217 + }, + { + "epoch": 0.53, + "learning_rate": 9.374955505785698e-06, + "loss": 0.8003, + "step": 22218 + }, + { + "epoch": 0.53, + "learning_rate": 9.374178136270968e-06, + "loss": 0.9684, + "step": 22219 + }, + { + "epoch": 0.53, + "learning_rate": 9.37340077055295e-06, + "loss": 0.8933, + "step": 22220 + }, + { + "epoch": 0.53, + "learning_rate": 9.372623408636362e-06, + "loss": 1.0333, + "step": 22221 + }, + { + "epoch": 0.53, + "learning_rate": 9.371846050525922e-06, + "loss": 0.9939, + "step": 22222 + }, + { + "epoch": 0.53, + "learning_rate": 9.371068696226344e-06, + "loss": 1.0517, + "step": 22223 + }, + { + "epoch": 0.53, + "learning_rate": 9.370291345742343e-06, + "loss": 0.9552, + "step": 22224 + }, + { + "epoch": 0.53, + "learning_rate": 9.369513999078637e-06, + "loss": 1.0874, + "step": 22225 + }, + { + "epoch": 0.53, + "learning_rate": 9.368736656239942e-06, + "loss": 0.9095, + "step": 22226 + }, + { + "epoch": 0.53, + "learning_rate": 9.367959317230973e-06, + "loss": 1.0754, + "step": 22227 + }, + { + "epoch": 0.53, + "learning_rate": 9.367181982056444e-06, + "loss": 1.2837, + "step": 22228 + }, + { + "epoch": 0.53, + "learning_rate": 9.366404650721074e-06, + "loss": 1.155, + "step": 22229 + }, + { + "epoch": 0.53, + "learning_rate": 9.36562732322958e-06, + "loss": 1.03, + "step": 22230 + }, + { + "epoch": 0.53, + "learning_rate": 9.364849999586676e-06, + "loss": 1.2444, + "step": 22231 + }, + { + "epoch": 0.53, + "learning_rate": 9.364072679797074e-06, + "loss": 1.0549, + "step": 22232 + }, + { + "epoch": 0.53, + "learning_rate": 9.363295363865497e-06, + "loss": 1.2077, + "step": 22233 + }, + { + "epoch": 0.53, + "learning_rate": 9.362518051796656e-06, + "loss": 1.1474, + "step": 22234 + }, + { + "epoch": 0.53, + "learning_rate": 9.361740743595269e-06, + "loss": 1.1418, + "step": 22235 + }, + { + "epoch": 0.53, + "learning_rate": 9.360963439266048e-06, + "loss": 0.8588, + "step": 22236 + }, + { + "epoch": 0.53, + "learning_rate": 9.360186138813713e-06, + "loss": 1.0637, + "step": 22237 + }, + { + "epoch": 0.53, + "learning_rate": 9.35940884224298e-06, + "loss": 1.1866, + "step": 22238 + }, + { + "epoch": 0.53, + "learning_rate": 9.358631549558561e-06, + "loss": 1.093, + "step": 22239 + }, + { + "epoch": 0.53, + "learning_rate": 9.357854260765173e-06, + "loss": 0.9647, + "step": 22240 + }, + { + "epoch": 0.53, + "learning_rate": 9.357076975867534e-06, + "loss": 1.0712, + "step": 22241 + }, + { + "epoch": 0.53, + "learning_rate": 9.356299694870357e-06, + "loss": 0.9479, + "step": 22242 + }, + { + "epoch": 0.53, + "learning_rate": 9.355522417778358e-06, + "loss": 1.0841, + "step": 22243 + }, + { + "epoch": 0.53, + "learning_rate": 9.354745144596253e-06, + "loss": 1.0464, + "step": 22244 + }, + { + "epoch": 0.53, + "learning_rate": 9.353967875328757e-06, + "loss": 0.8272, + "step": 22245 + }, + { + "epoch": 0.53, + "learning_rate": 9.353190609980587e-06, + "loss": 1.2026, + "step": 22246 + }, + { + "epoch": 0.54, + "learning_rate": 9.352413348556458e-06, + "loss": 0.9052, + "step": 22247 + }, + { + "epoch": 0.54, + "learning_rate": 9.351636091061082e-06, + "loss": 1.1224, + "step": 22248 + }, + { + "epoch": 0.54, + "learning_rate": 9.350858837499179e-06, + "loss": 0.8774, + "step": 22249 + }, + { + "epoch": 0.54, + "learning_rate": 9.350081587875463e-06, + "loss": 0.9623, + "step": 22250 + }, + { + "epoch": 0.54, + "learning_rate": 9.34930434219465e-06, + "loss": 0.8634, + "step": 22251 + }, + { + "epoch": 0.54, + "learning_rate": 9.348527100461453e-06, + "loss": 1.1051, + "step": 22252 + }, + { + "epoch": 0.54, + "learning_rate": 9.347749862680588e-06, + "loss": 0.8827, + "step": 22253 + }, + { + "epoch": 0.54, + "learning_rate": 9.346972628856775e-06, + "loss": 0.8491, + "step": 22254 + }, + { + "epoch": 0.54, + "learning_rate": 9.346195398994722e-06, + "loss": 0.9298, + "step": 22255 + }, + { + "epoch": 0.54, + "learning_rate": 9.345418173099146e-06, + "loss": 1.1899, + "step": 22256 + }, + { + "epoch": 0.54, + "learning_rate": 9.344640951174769e-06, + "loss": 1.0497, + "step": 22257 + }, + { + "epoch": 0.54, + "learning_rate": 9.343863733226298e-06, + "loss": 0.8576, + "step": 22258 + }, + { + "epoch": 0.54, + "learning_rate": 9.343086519258453e-06, + "loss": 1.0552, + "step": 22259 + }, + { + "epoch": 0.54, + "learning_rate": 9.342309309275946e-06, + "loss": 1.0173, + "step": 22260 + }, + { + "epoch": 0.54, + "learning_rate": 9.341532103283496e-06, + "loss": 0.7006, + "step": 22261 + }, + { + "epoch": 0.54, + "learning_rate": 9.340754901285814e-06, + "loss": 1.1246, + "step": 22262 + }, + { + "epoch": 0.54, + "learning_rate": 9.339977703287616e-06, + "loss": 1.2763, + "step": 22263 + }, + { + "epoch": 0.54, + "learning_rate": 9.339200509293618e-06, + "loss": 1.0227, + "step": 22264 + }, + { + "epoch": 0.54, + "learning_rate": 9.338423319308537e-06, + "loss": 1.2003, + "step": 22265 + }, + { + "epoch": 0.54, + "learning_rate": 9.337646133337087e-06, + "loss": 1.2947, + "step": 22266 + }, + { + "epoch": 0.54, + "learning_rate": 9.336868951383982e-06, + "loss": 1.1602, + "step": 22267 + }, + { + "epoch": 0.54, + "learning_rate": 9.336091773453933e-06, + "loss": 0.8923, + "step": 22268 + }, + { + "epoch": 0.54, + "learning_rate": 9.335314599551664e-06, + "loss": 0.8997, + "step": 22269 + }, + { + "epoch": 0.54, + "learning_rate": 9.334537429681882e-06, + "loss": 0.8288, + "step": 22270 + }, + { + "epoch": 0.54, + "learning_rate": 9.333760263849307e-06, + "loss": 1.265, + "step": 22271 + }, + { + "epoch": 0.54, + "learning_rate": 9.332983102058648e-06, + "loss": 1.2714, + "step": 22272 + }, + { + "epoch": 0.54, + "learning_rate": 9.332205944314628e-06, + "loss": 1.114, + "step": 22273 + }, + { + "epoch": 0.54, + "learning_rate": 9.331428790621954e-06, + "loss": 1.1009, + "step": 22274 + }, + { + "epoch": 0.54, + "learning_rate": 9.330651640985345e-06, + "loss": 1.1718, + "step": 22275 + }, + { + "epoch": 0.54, + "learning_rate": 9.329874495409517e-06, + "loss": 1.0499, + "step": 22276 + }, + { + "epoch": 0.54, + "learning_rate": 9.329097353899183e-06, + "loss": 1.2702, + "step": 22277 + }, + { + "epoch": 0.54, + "learning_rate": 9.328320216459056e-06, + "loss": 1.1785, + "step": 22278 + }, + { + "epoch": 0.54, + "learning_rate": 9.32754308309385e-06, + "loss": 1.1675, + "step": 22279 + }, + { + "epoch": 0.54, + "learning_rate": 9.326765953808285e-06, + "loss": 1.1196, + "step": 22280 + }, + { + "epoch": 0.54, + "learning_rate": 9.325988828607073e-06, + "loss": 1.1503, + "step": 22281 + }, + { + "epoch": 0.54, + "learning_rate": 9.325211707494928e-06, + "loss": 1.2115, + "step": 22282 + }, + { + "epoch": 0.54, + "learning_rate": 9.324434590476562e-06, + "loss": 0.9169, + "step": 22283 + }, + { + "epoch": 0.54, + "learning_rate": 9.323657477556695e-06, + "loss": 1.038, + "step": 22284 + }, + { + "epoch": 0.54, + "learning_rate": 9.32288036874004e-06, + "loss": 1.1002, + "step": 22285 + }, + { + "epoch": 0.54, + "learning_rate": 9.32210326403131e-06, + "loss": 1.0836, + "step": 22286 + }, + { + "epoch": 0.54, + "learning_rate": 9.321326163435217e-06, + "loss": 0.9492, + "step": 22287 + }, + { + "epoch": 0.54, + "learning_rate": 9.320549066956481e-06, + "loss": 1.1144, + "step": 22288 + }, + { + "epoch": 0.54, + "learning_rate": 9.319771974599813e-06, + "loss": 1.1377, + "step": 22289 + }, + { + "epoch": 0.54, + "learning_rate": 9.31899488636993e-06, + "loss": 0.8063, + "step": 22290 + }, + { + "epoch": 0.54, + "learning_rate": 9.318217802271544e-06, + "loss": 1.0357, + "step": 22291 + }, + { + "epoch": 0.54, + "learning_rate": 9.31744072230937e-06, + "loss": 1.128, + "step": 22292 + }, + { + "epoch": 0.54, + "learning_rate": 9.316663646488123e-06, + "loss": 1.2096, + "step": 22293 + }, + { + "epoch": 0.54, + "learning_rate": 9.315886574812518e-06, + "loss": 1.0486, + "step": 22294 + }, + { + "epoch": 0.54, + "learning_rate": 9.315109507287264e-06, + "loss": 1.1029, + "step": 22295 + }, + { + "epoch": 0.54, + "learning_rate": 9.314332443917083e-06, + "loss": 0.9086, + "step": 22296 + }, + { + "epoch": 0.54, + "learning_rate": 9.313555384706686e-06, + "loss": 0.8981, + "step": 22297 + }, + { + "epoch": 0.54, + "learning_rate": 9.312778329660788e-06, + "loss": 1.0625, + "step": 22298 + }, + { + "epoch": 0.54, + "learning_rate": 9.312001278784099e-06, + "loss": 1.069, + "step": 22299 + }, + { + "epoch": 0.54, + "learning_rate": 9.311224232081338e-06, + "loss": 1.1996, + "step": 22300 + }, + { + "epoch": 0.54, + "learning_rate": 9.310447189557219e-06, + "loss": 0.9895, + "step": 22301 + }, + { + "epoch": 0.54, + "learning_rate": 9.309670151216451e-06, + "loss": 0.9747, + "step": 22302 + }, + { + "epoch": 0.54, + "learning_rate": 9.308893117063754e-06, + "loss": 0.9477, + "step": 22303 + }, + { + "epoch": 0.54, + "learning_rate": 9.30811608710384e-06, + "loss": 1.12, + "step": 22304 + }, + { + "epoch": 0.54, + "learning_rate": 9.307339061341425e-06, + "loss": 1.1885, + "step": 22305 + }, + { + "epoch": 0.54, + "learning_rate": 9.306562039781217e-06, + "loss": 1.2447, + "step": 22306 + }, + { + "epoch": 0.54, + "learning_rate": 9.305785022427935e-06, + "loss": 0.9569, + "step": 22307 + }, + { + "epoch": 0.54, + "learning_rate": 9.305008009286293e-06, + "loss": 0.8879, + "step": 22308 + }, + { + "epoch": 0.54, + "learning_rate": 9.304231000361005e-06, + "loss": 0.8447, + "step": 22309 + }, + { + "epoch": 0.54, + "learning_rate": 9.303453995656783e-06, + "loss": 0.9581, + "step": 22310 + }, + { + "epoch": 0.54, + "learning_rate": 9.30267699517834e-06, + "loss": 1.131, + "step": 22311 + }, + { + "epoch": 0.54, + "learning_rate": 9.301899998930392e-06, + "loss": 0.9543, + "step": 22312 + }, + { + "epoch": 0.54, + "learning_rate": 9.301123006917655e-06, + "loss": 1.3922, + "step": 22313 + }, + { + "epoch": 0.54, + "learning_rate": 9.30034601914484e-06, + "loss": 1.1874, + "step": 22314 + }, + { + "epoch": 0.54, + "learning_rate": 9.299569035616658e-06, + "loss": 1.2203, + "step": 22315 + }, + { + "epoch": 0.54, + "learning_rate": 9.298792056337828e-06, + "loss": 1.0547, + "step": 22316 + }, + { + "epoch": 0.54, + "learning_rate": 9.298015081313063e-06, + "loss": 0.9805, + "step": 22317 + }, + { + "epoch": 0.54, + "learning_rate": 9.297238110547075e-06, + "loss": 0.7418, + "step": 22318 + }, + { + "epoch": 0.54, + "learning_rate": 9.296461144044575e-06, + "loss": 1.0223, + "step": 22319 + }, + { + "epoch": 0.54, + "learning_rate": 9.295684181810282e-06, + "loss": 0.8828, + "step": 22320 + }, + { + "epoch": 0.54, + "learning_rate": 9.294907223848908e-06, + "loss": 1.2749, + "step": 22321 + }, + { + "epoch": 0.54, + "learning_rate": 9.294130270165163e-06, + "loss": 0.8397, + "step": 22322 + }, + { + "epoch": 0.54, + "learning_rate": 9.293353320763767e-06, + "loss": 0.9492, + "step": 22323 + }, + { + "epoch": 0.54, + "learning_rate": 9.29257637564943e-06, + "loss": 0.9558, + "step": 22324 + }, + { + "epoch": 0.54, + "learning_rate": 9.291799434826864e-06, + "loss": 1.0094, + "step": 22325 + }, + { + "epoch": 0.54, + "learning_rate": 9.291022498300784e-06, + "loss": 1.256, + "step": 22326 + }, + { + "epoch": 0.54, + "learning_rate": 9.290245566075904e-06, + "loss": 0.9911, + "step": 22327 + }, + { + "epoch": 0.54, + "learning_rate": 9.28946863815694e-06, + "loss": 1.3372, + "step": 22328 + }, + { + "epoch": 0.54, + "learning_rate": 9.2886917145486e-06, + "loss": 1.1976, + "step": 22329 + }, + { + "epoch": 0.54, + "learning_rate": 9.2879147952556e-06, + "loss": 1.0661, + "step": 22330 + }, + { + "epoch": 0.54, + "learning_rate": 9.287137880282655e-06, + "loss": 1.0236, + "step": 22331 + }, + { + "epoch": 0.54, + "learning_rate": 9.286360969634472e-06, + "loss": 1.09, + "step": 22332 + }, + { + "epoch": 0.54, + "learning_rate": 9.285584063315773e-06, + "loss": 1.1521, + "step": 22333 + }, + { + "epoch": 0.54, + "learning_rate": 9.284807161331269e-06, + "loss": 1.1282, + "step": 22334 + }, + { + "epoch": 0.54, + "learning_rate": 9.284030263685671e-06, + "loss": 1.224, + "step": 22335 + }, + { + "epoch": 0.54, + "learning_rate": 9.28325337038369e-06, + "loss": 0.7494, + "step": 22336 + }, + { + "epoch": 0.54, + "learning_rate": 9.282476481430045e-06, + "loss": 0.8988, + "step": 22337 + }, + { + "epoch": 0.54, + "learning_rate": 9.281699596829447e-06, + "loss": 0.8958, + "step": 22338 + }, + { + "epoch": 0.54, + "learning_rate": 9.280922716586607e-06, + "loss": 1.0517, + "step": 22339 + }, + { + "epoch": 0.54, + "learning_rate": 9.280145840706239e-06, + "loss": 1.0893, + "step": 22340 + }, + { + "epoch": 0.54, + "learning_rate": 9.27936896919306e-06, + "loss": 1.239, + "step": 22341 + }, + { + "epoch": 0.54, + "learning_rate": 9.278592102051777e-06, + "loss": 1.0771, + "step": 22342 + }, + { + "epoch": 0.54, + "learning_rate": 9.277815239287106e-06, + "loss": 1.0426, + "step": 22343 + }, + { + "epoch": 0.54, + "learning_rate": 9.27703838090376e-06, + "loss": 1.1738, + "step": 22344 + }, + { + "epoch": 0.54, + "learning_rate": 9.276261526906455e-06, + "loss": 1.2325, + "step": 22345 + }, + { + "epoch": 0.54, + "learning_rate": 9.2754846772999e-06, + "loss": 0.7462, + "step": 22346 + }, + { + "epoch": 0.54, + "learning_rate": 9.274707832088808e-06, + "loss": 1.0216, + "step": 22347 + }, + { + "epoch": 0.54, + "learning_rate": 9.273930991277893e-06, + "loss": 1.152, + "step": 22348 + }, + { + "epoch": 0.54, + "learning_rate": 9.273154154871867e-06, + "loss": 1.2965, + "step": 22349 + }, + { + "epoch": 0.54, + "learning_rate": 9.272377322875446e-06, + "loss": 1.2082, + "step": 22350 + }, + { + "epoch": 0.54, + "learning_rate": 9.27160049529334e-06, + "loss": 1.0238, + "step": 22351 + }, + { + "epoch": 0.54, + "learning_rate": 9.270823672130262e-06, + "loss": 1.0313, + "step": 22352 + }, + { + "epoch": 0.54, + "learning_rate": 9.270046853390924e-06, + "loss": 1.172, + "step": 22353 + }, + { + "epoch": 0.54, + "learning_rate": 9.269270039080043e-06, + "loss": 1.0013, + "step": 22354 + }, + { + "epoch": 0.54, + "learning_rate": 9.268493229202325e-06, + "loss": 1.1778, + "step": 22355 + }, + { + "epoch": 0.54, + "learning_rate": 9.267716423762489e-06, + "loss": 1.141, + "step": 22356 + }, + { + "epoch": 0.54, + "learning_rate": 9.266939622765246e-06, + "loss": 1.0342, + "step": 22357 + }, + { + "epoch": 0.54, + "learning_rate": 9.266162826215307e-06, + "loss": 1.1001, + "step": 22358 + }, + { + "epoch": 0.54, + "learning_rate": 9.265386034117384e-06, + "loss": 1.2333, + "step": 22359 + }, + { + "epoch": 0.54, + "learning_rate": 9.264609246476194e-06, + "loss": 1.2779, + "step": 22360 + }, + { + "epoch": 0.54, + "learning_rate": 9.263832463296444e-06, + "loss": 1.0839, + "step": 22361 + }, + { + "epoch": 0.54, + "learning_rate": 9.263055684582853e-06, + "loss": 1.0986, + "step": 22362 + }, + { + "epoch": 0.54, + "learning_rate": 9.262278910340125e-06, + "loss": 1.0336, + "step": 22363 + }, + { + "epoch": 0.54, + "learning_rate": 9.26150214057298e-06, + "loss": 1.054, + "step": 22364 + }, + { + "epoch": 0.54, + "learning_rate": 9.260725375286129e-06, + "loss": 0.9352, + "step": 22365 + }, + { + "epoch": 0.54, + "learning_rate": 9.25994861448428e-06, + "loss": 1.166, + "step": 22366 + }, + { + "epoch": 0.54, + "learning_rate": 9.25917185817215e-06, + "loss": 0.8914, + "step": 22367 + }, + { + "epoch": 0.54, + "learning_rate": 9.25839510635445e-06, + "loss": 1.005, + "step": 22368 + }, + { + "epoch": 0.54, + "learning_rate": 9.257618359035894e-06, + "loss": 1.2289, + "step": 22369 + }, + { + "epoch": 0.54, + "learning_rate": 9.256841616221192e-06, + "loss": 1.0486, + "step": 22370 + }, + { + "epoch": 0.54, + "learning_rate": 9.256064877915056e-06, + "loss": 0.9104, + "step": 22371 + }, + { + "epoch": 0.54, + "learning_rate": 9.2552881441222e-06, + "loss": 1.0458, + "step": 22372 + }, + { + "epoch": 0.54, + "learning_rate": 9.254511414847336e-06, + "loss": 0.9772, + "step": 22373 + }, + { + "epoch": 0.54, + "learning_rate": 9.253734690095178e-06, + "loss": 0.8993, + "step": 22374 + }, + { + "epoch": 0.54, + "learning_rate": 9.252957969870433e-06, + "loss": 1.1842, + "step": 22375 + }, + { + "epoch": 0.54, + "learning_rate": 9.252181254177817e-06, + "loss": 1.2562, + "step": 22376 + }, + { + "epoch": 0.54, + "learning_rate": 9.251404543022042e-06, + "loss": 1.1475, + "step": 22377 + }, + { + "epoch": 0.54, + "learning_rate": 9.25062783640782e-06, + "loss": 1.265, + "step": 22378 + }, + { + "epoch": 0.54, + "learning_rate": 9.24985113433986e-06, + "loss": 1.3173, + "step": 22379 + }, + { + "epoch": 0.54, + "learning_rate": 9.249074436822878e-06, + "loss": 1.0479, + "step": 22380 + }, + { + "epoch": 0.54, + "learning_rate": 9.248297743861586e-06, + "loss": 1.1896, + "step": 22381 + }, + { + "epoch": 0.54, + "learning_rate": 9.247521055460695e-06, + "loss": 1.2053, + "step": 22382 + }, + { + "epoch": 0.54, + "learning_rate": 9.246744371624914e-06, + "loss": 1.1193, + "step": 22383 + }, + { + "epoch": 0.54, + "learning_rate": 9.24596769235896e-06, + "loss": 0.9931, + "step": 22384 + }, + { + "epoch": 0.54, + "learning_rate": 9.245191017667543e-06, + "loss": 1.3173, + "step": 22385 + }, + { + "epoch": 0.54, + "learning_rate": 9.244414347555375e-06, + "loss": 0.9769, + "step": 22386 + }, + { + "epoch": 0.54, + "learning_rate": 9.243637682027165e-06, + "loss": 1.159, + "step": 22387 + }, + { + "epoch": 0.54, + "learning_rate": 9.242861021087628e-06, + "loss": 0.9935, + "step": 22388 + }, + { + "epoch": 0.54, + "learning_rate": 9.242084364741476e-06, + "loss": 0.9904, + "step": 22389 + }, + { + "epoch": 0.54, + "learning_rate": 9.24130771299342e-06, + "loss": 1.0522, + "step": 22390 + }, + { + "epoch": 0.54, + "learning_rate": 9.240531065848169e-06, + "loss": 0.9912, + "step": 22391 + }, + { + "epoch": 0.54, + "learning_rate": 9.239754423310441e-06, + "loss": 1.0961, + "step": 22392 + }, + { + "epoch": 0.54, + "learning_rate": 9.238977785384942e-06, + "loss": 1.2014, + "step": 22393 + }, + { + "epoch": 0.54, + "learning_rate": 9.238201152076389e-06, + "loss": 1.1801, + "step": 22394 + }, + { + "epoch": 0.54, + "learning_rate": 9.237424523389484e-06, + "loss": 0.8326, + "step": 22395 + }, + { + "epoch": 0.54, + "learning_rate": 9.23664789932895e-06, + "loss": 0.9843, + "step": 22396 + }, + { + "epoch": 0.54, + "learning_rate": 9.235871279899493e-06, + "loss": 1.0647, + "step": 22397 + }, + { + "epoch": 0.54, + "learning_rate": 9.235094665105826e-06, + "loss": 1.0767, + "step": 22398 + }, + { + "epoch": 0.54, + "learning_rate": 9.234318054952657e-06, + "loss": 0.9237, + "step": 22399 + }, + { + "epoch": 0.54, + "learning_rate": 9.233541449444702e-06, + "loss": 1.1156, + "step": 22400 + }, + { + "epoch": 0.54, + "learning_rate": 9.23276484858667e-06, + "loss": 0.9919, + "step": 22401 + }, + { + "epoch": 0.54, + "learning_rate": 9.231988252383273e-06, + "loss": 1.1507, + "step": 22402 + }, + { + "epoch": 0.54, + "learning_rate": 9.231211660839223e-06, + "loss": 1.0927, + "step": 22403 + }, + { + "epoch": 0.54, + "learning_rate": 9.230435073959232e-06, + "loss": 1.2571, + "step": 22404 + }, + { + "epoch": 0.54, + "learning_rate": 9.22965849174801e-06, + "loss": 1.2828, + "step": 22405 + }, + { + "epoch": 0.54, + "learning_rate": 9.228881914210267e-06, + "loss": 1.2885, + "step": 22406 + }, + { + "epoch": 0.54, + "learning_rate": 9.228105341350718e-06, + "loss": 1.1219, + "step": 22407 + }, + { + "epoch": 0.54, + "learning_rate": 9.227328773174072e-06, + "loss": 1.1854, + "step": 22408 + }, + { + "epoch": 0.54, + "learning_rate": 9.226552209685042e-06, + "loss": 1.1293, + "step": 22409 + }, + { + "epoch": 0.54, + "learning_rate": 9.225775650888334e-06, + "loss": 0.9111, + "step": 22410 + }, + { + "epoch": 0.54, + "learning_rate": 9.224999096788665e-06, + "loss": 1.2124, + "step": 22411 + }, + { + "epoch": 0.54, + "learning_rate": 9.224222547390747e-06, + "loss": 1.0071, + "step": 22412 + }, + { + "epoch": 0.54, + "learning_rate": 9.223446002699285e-06, + "loss": 1.0483, + "step": 22413 + }, + { + "epoch": 0.54, + "learning_rate": 9.222669462718992e-06, + "loss": 1.1103, + "step": 22414 + }, + { + "epoch": 0.54, + "learning_rate": 9.221892927454584e-06, + "loss": 0.8675, + "step": 22415 + }, + { + "epoch": 0.54, + "learning_rate": 9.221116396910768e-06, + "loss": 1.0879, + "step": 22416 + }, + { + "epoch": 0.54, + "learning_rate": 9.220339871092255e-06, + "loss": 0.9584, + "step": 22417 + }, + { + "epoch": 0.54, + "learning_rate": 9.219563350003756e-06, + "loss": 0.9005, + "step": 22418 + }, + { + "epoch": 0.54, + "learning_rate": 9.218786833649983e-06, + "loss": 1.1472, + "step": 22419 + }, + { + "epoch": 0.54, + "learning_rate": 9.218010322035648e-06, + "loss": 0.9623, + "step": 22420 + }, + { + "epoch": 0.54, + "learning_rate": 9.217233815165458e-06, + "loss": 1.3549, + "step": 22421 + }, + { + "epoch": 0.54, + "learning_rate": 9.216457313044127e-06, + "loss": 1.1591, + "step": 22422 + }, + { + "epoch": 0.54, + "learning_rate": 9.215680815676366e-06, + "loss": 0.9485, + "step": 22423 + }, + { + "epoch": 0.54, + "learning_rate": 9.214904323066884e-06, + "loss": 0.9153, + "step": 22424 + }, + { + "epoch": 0.54, + "learning_rate": 9.214127835220394e-06, + "loss": 1.0875, + "step": 22425 + }, + { + "epoch": 0.54, + "learning_rate": 9.213351352141605e-06, + "loss": 1.0264, + "step": 22426 + }, + { + "epoch": 0.54, + "learning_rate": 9.212574873835229e-06, + "loss": 1.1595, + "step": 22427 + }, + { + "epoch": 0.54, + "learning_rate": 9.211798400305975e-06, + "loss": 0.9345, + "step": 22428 + }, + { + "epoch": 0.54, + "learning_rate": 9.211021931558556e-06, + "loss": 1.1856, + "step": 22429 + }, + { + "epoch": 0.54, + "learning_rate": 9.210245467597678e-06, + "loss": 1.0452, + "step": 22430 + }, + { + "epoch": 0.54, + "learning_rate": 9.209469008428058e-06, + "loss": 1.2289, + "step": 22431 + }, + { + "epoch": 0.54, + "learning_rate": 9.208692554054404e-06, + "loss": 1.0994, + "step": 22432 + }, + { + "epoch": 0.54, + "learning_rate": 9.207916104481425e-06, + "loss": 0.7943, + "step": 22433 + }, + { + "epoch": 0.54, + "learning_rate": 9.207139659713832e-06, + "loss": 0.8907, + "step": 22434 + }, + { + "epoch": 0.54, + "learning_rate": 9.206363219756336e-06, + "loss": 1.2106, + "step": 22435 + }, + { + "epoch": 0.54, + "learning_rate": 9.20558678461365e-06, + "loss": 1.0117, + "step": 22436 + }, + { + "epoch": 0.54, + "learning_rate": 9.204810354290483e-06, + "loss": 1.0379, + "step": 22437 + }, + { + "epoch": 0.54, + "learning_rate": 9.20403392879154e-06, + "loss": 0.6888, + "step": 22438 + }, + { + "epoch": 0.54, + "learning_rate": 9.203257508121537e-06, + "loss": 1.082, + "step": 22439 + }, + { + "epoch": 0.54, + "learning_rate": 9.202481092285187e-06, + "loss": 1.1399, + "step": 22440 + }, + { + "epoch": 0.54, + "learning_rate": 9.201704681287195e-06, + "loss": 1.1988, + "step": 22441 + }, + { + "epoch": 0.54, + "learning_rate": 9.20092827513227e-06, + "loss": 1.1208, + "step": 22442 + }, + { + "epoch": 0.54, + "learning_rate": 9.200151873825128e-06, + "loss": 0.8486, + "step": 22443 + }, + { + "epoch": 0.54, + "learning_rate": 9.199375477370476e-06, + "loss": 1.2173, + "step": 22444 + }, + { + "epoch": 0.54, + "learning_rate": 9.198599085773022e-06, + "loss": 1.1459, + "step": 22445 + }, + { + "epoch": 0.54, + "learning_rate": 9.197822699037483e-06, + "loss": 1.194, + "step": 22446 + }, + { + "epoch": 0.54, + "learning_rate": 9.197046317168562e-06, + "loss": 1.1813, + "step": 22447 + }, + { + "epoch": 0.54, + "learning_rate": 9.196269940170975e-06, + "loss": 1.0295, + "step": 22448 + }, + { + "epoch": 0.54, + "learning_rate": 9.195493568049427e-06, + "loss": 1.1357, + "step": 22449 + }, + { + "epoch": 0.54, + "learning_rate": 9.19471720080863e-06, + "loss": 1.1566, + "step": 22450 + }, + { + "epoch": 0.54, + "learning_rate": 9.193940838453298e-06, + "loss": 1.0991, + "step": 22451 + }, + { + "epoch": 0.54, + "learning_rate": 9.193164480988134e-06, + "loss": 1.1472, + "step": 22452 + }, + { + "epoch": 0.54, + "learning_rate": 9.19238812841785e-06, + "loss": 1.0795, + "step": 22453 + }, + { + "epoch": 0.54, + "learning_rate": 9.19161178074716e-06, + "loss": 1.2951, + "step": 22454 + }, + { + "epoch": 0.54, + "learning_rate": 9.19083543798077e-06, + "loss": 1.2074, + "step": 22455 + }, + { + "epoch": 0.54, + "learning_rate": 9.190059100123392e-06, + "loss": 0.8327, + "step": 22456 + }, + { + "epoch": 0.54, + "learning_rate": 9.189282767179732e-06, + "loss": 1.252, + "step": 22457 + }, + { + "epoch": 0.54, + "learning_rate": 9.188506439154506e-06, + "loss": 0.9623, + "step": 22458 + }, + { + "epoch": 0.54, + "learning_rate": 9.187730116052422e-06, + "loss": 1.0882, + "step": 22459 + }, + { + "epoch": 0.54, + "learning_rate": 9.186953797878185e-06, + "loss": 1.1522, + "step": 22460 + }, + { + "epoch": 0.54, + "learning_rate": 9.186177484636508e-06, + "loss": 1.2661, + "step": 22461 + }, + { + "epoch": 0.54, + "learning_rate": 9.185401176332103e-06, + "loss": 1.2176, + "step": 22462 + }, + { + "epoch": 0.54, + "learning_rate": 9.184624872969676e-06, + "loss": 1.2519, + "step": 22463 + }, + { + "epoch": 0.54, + "learning_rate": 9.183848574553939e-06, + "loss": 1.204, + "step": 22464 + }, + { + "epoch": 0.54, + "learning_rate": 9.183072281089598e-06, + "loss": 1.1221, + "step": 22465 + }, + { + "epoch": 0.54, + "learning_rate": 9.182295992581368e-06, + "loss": 0.9534, + "step": 22466 + }, + { + "epoch": 0.54, + "learning_rate": 9.181519709033955e-06, + "loss": 0.9712, + "step": 22467 + }, + { + "epoch": 0.54, + "learning_rate": 9.180743430452072e-06, + "loss": 1.0201, + "step": 22468 + }, + { + "epoch": 0.54, + "learning_rate": 9.179967156840422e-06, + "loss": 1.092, + "step": 22469 + }, + { + "epoch": 0.54, + "learning_rate": 9.17919088820372e-06, + "loss": 1.0107, + "step": 22470 + }, + { + "epoch": 0.54, + "learning_rate": 9.178414624546674e-06, + "loss": 1.0392, + "step": 22471 + }, + { + "epoch": 0.54, + "learning_rate": 9.177638365873995e-06, + "loss": 0.9712, + "step": 22472 + }, + { + "epoch": 0.54, + "learning_rate": 9.176862112190386e-06, + "loss": 1.1777, + "step": 22473 + }, + { + "epoch": 0.54, + "learning_rate": 9.176085863500564e-06, + "loss": 0.9728, + "step": 22474 + }, + { + "epoch": 0.54, + "learning_rate": 9.175309619809235e-06, + "loss": 1.3481, + "step": 22475 + }, + { + "epoch": 0.54, + "learning_rate": 9.17453338112111e-06, + "loss": 0.9418, + "step": 22476 + }, + { + "epoch": 0.54, + "learning_rate": 9.173757147440893e-06, + "loss": 1.2801, + "step": 22477 + }, + { + "epoch": 0.54, + "learning_rate": 9.1729809187733e-06, + "loss": 1.0249, + "step": 22478 + }, + { + "epoch": 0.54, + "learning_rate": 9.172204695123037e-06, + "loss": 0.96, + "step": 22479 + }, + { + "epoch": 0.54, + "learning_rate": 9.171428476494813e-06, + "loss": 1.1951, + "step": 22480 + }, + { + "epoch": 0.54, + "learning_rate": 9.170652262893335e-06, + "loss": 1.0586, + "step": 22481 + }, + { + "epoch": 0.54, + "learning_rate": 9.169876054323319e-06, + "loss": 1.0939, + "step": 22482 + }, + { + "epoch": 0.54, + "learning_rate": 9.169099850789468e-06, + "loss": 0.6964, + "step": 22483 + }, + { + "epoch": 0.54, + "learning_rate": 9.168323652296494e-06, + "loss": 1.1457, + "step": 22484 + }, + { + "epoch": 0.54, + "learning_rate": 9.167547458849101e-06, + "loss": 0.9632, + "step": 22485 + }, + { + "epoch": 0.54, + "learning_rate": 9.166771270452005e-06, + "loss": 0.7535, + "step": 22486 + }, + { + "epoch": 0.54, + "learning_rate": 9.165995087109911e-06, + "loss": 1.1273, + "step": 22487 + }, + { + "epoch": 0.54, + "learning_rate": 9.165218908827529e-06, + "loss": 1.0784, + "step": 22488 + }, + { + "epoch": 0.54, + "learning_rate": 9.164442735609566e-06, + "loss": 1.4133, + "step": 22489 + }, + { + "epoch": 0.54, + "learning_rate": 9.163666567460735e-06, + "loss": 1.081, + "step": 22490 + }, + { + "epoch": 0.54, + "learning_rate": 9.16289040438574e-06, + "loss": 1.0233, + "step": 22491 + }, + { + "epoch": 0.54, + "learning_rate": 9.16211424638929e-06, + "loss": 1.2154, + "step": 22492 + }, + { + "epoch": 0.54, + "learning_rate": 9.1613380934761e-06, + "loss": 1.0867, + "step": 22493 + }, + { + "epoch": 0.54, + "learning_rate": 9.160561945650874e-06, + "loss": 1.1973, + "step": 22494 + }, + { + "epoch": 0.54, + "learning_rate": 9.15978580291832e-06, + "loss": 1.067, + "step": 22495 + }, + { + "epoch": 0.54, + "learning_rate": 9.159009665283146e-06, + "loss": 0.9579, + "step": 22496 + }, + { + "epoch": 0.54, + "learning_rate": 9.158233532750066e-06, + "loss": 1.0017, + "step": 22497 + }, + { + "epoch": 0.54, + "learning_rate": 9.157457405323784e-06, + "loss": 1.1394, + "step": 22498 + }, + { + "epoch": 0.54, + "learning_rate": 9.156681283009006e-06, + "loss": 0.9497, + "step": 22499 + }, + { + "epoch": 0.54, + "learning_rate": 9.155905165810449e-06, + "loss": 0.9093, + "step": 22500 + }, + { + "epoch": 0.54, + "learning_rate": 9.155129053732815e-06, + "loss": 1.0743, + "step": 22501 + }, + { + "epoch": 0.54, + "learning_rate": 9.154352946780813e-06, + "loss": 1.3173, + "step": 22502 + }, + { + "epoch": 0.54, + "learning_rate": 9.153576844959156e-06, + "loss": 1.1884, + "step": 22503 + }, + { + "epoch": 0.54, + "learning_rate": 9.152800748272548e-06, + "loss": 1.2281, + "step": 22504 + }, + { + "epoch": 0.54, + "learning_rate": 9.1520246567257e-06, + "loss": 1.0666, + "step": 22505 + }, + { + "epoch": 0.54, + "learning_rate": 9.151248570323314e-06, + "loss": 0.9952, + "step": 22506 + }, + { + "epoch": 0.54, + "learning_rate": 9.150472489070108e-06, + "loss": 1.1518, + "step": 22507 + }, + { + "epoch": 0.54, + "learning_rate": 9.149696412970785e-06, + "loss": 0.8051, + "step": 22508 + }, + { + "epoch": 0.54, + "learning_rate": 9.148920342030054e-06, + "loss": 1.1633, + "step": 22509 + }, + { + "epoch": 0.54, + "learning_rate": 9.148144276252621e-06, + "loss": 1.1438, + "step": 22510 + }, + { + "epoch": 0.54, + "learning_rate": 9.147368215643198e-06, + "loss": 1.0033, + "step": 22511 + }, + { + "epoch": 0.54, + "learning_rate": 9.146592160206492e-06, + "loss": 1.0447, + "step": 22512 + }, + { + "epoch": 0.54, + "learning_rate": 9.14581610994721e-06, + "loss": 1.1032, + "step": 22513 + }, + { + "epoch": 0.54, + "learning_rate": 9.145040064870058e-06, + "loss": 1.4021, + "step": 22514 + }, + { + "epoch": 0.54, + "learning_rate": 9.14426402497975e-06, + "loss": 0.9064, + "step": 22515 + }, + { + "epoch": 0.54, + "learning_rate": 9.14348799028099e-06, + "loss": 1.2995, + "step": 22516 + }, + { + "epoch": 0.54, + "learning_rate": 9.14271196077849e-06, + "loss": 1.0037, + "step": 22517 + }, + { + "epoch": 0.54, + "learning_rate": 9.141935936476949e-06, + "loss": 1.3046, + "step": 22518 + }, + { + "epoch": 0.54, + "learning_rate": 9.141159917381084e-06, + "loss": 1.2526, + "step": 22519 + }, + { + "epoch": 0.54, + "learning_rate": 9.1403839034956e-06, + "loss": 1.0392, + "step": 22520 + }, + { + "epoch": 0.54, + "learning_rate": 9.139607894825205e-06, + "loss": 1.1948, + "step": 22521 + }, + { + "epoch": 0.54, + "learning_rate": 9.138831891374605e-06, + "loss": 0.9452, + "step": 22522 + }, + { + "epoch": 0.54, + "learning_rate": 9.13805589314851e-06, + "loss": 1.3826, + "step": 22523 + }, + { + "epoch": 0.54, + "learning_rate": 9.137279900151627e-06, + "loss": 1.2011, + "step": 22524 + }, + { + "epoch": 0.54, + "learning_rate": 9.136503912388664e-06, + "loss": 0.7504, + "step": 22525 + }, + { + "epoch": 0.54, + "learning_rate": 9.135727929864328e-06, + "loss": 1.163, + "step": 22526 + }, + { + "epoch": 0.54, + "learning_rate": 9.134951952583328e-06, + "loss": 1.2799, + "step": 22527 + }, + { + "epoch": 0.54, + "learning_rate": 9.134175980550373e-06, + "loss": 0.8799, + "step": 22528 + }, + { + "epoch": 0.54, + "learning_rate": 9.133400013770165e-06, + "loss": 1.084, + "step": 22529 + }, + { + "epoch": 0.54, + "learning_rate": 9.132624052247417e-06, + "loss": 1.1061, + "step": 22530 + }, + { + "epoch": 0.54, + "learning_rate": 9.131848095986837e-06, + "loss": 1.0215, + "step": 22531 + }, + { + "epoch": 0.54, + "learning_rate": 9.131072144993128e-06, + "loss": 0.9996, + "step": 22532 + }, + { + "epoch": 0.54, + "learning_rate": 9.130296199271e-06, + "loss": 1.0847, + "step": 22533 + }, + { + "epoch": 0.54, + "learning_rate": 9.12952025882516e-06, + "loss": 0.918, + "step": 22534 + }, + { + "epoch": 0.54, + "learning_rate": 9.128744323660317e-06, + "loss": 1.0091, + "step": 22535 + }, + { + "epoch": 0.54, + "learning_rate": 9.127968393781178e-06, + "loss": 0.997, + "step": 22536 + }, + { + "epoch": 0.54, + "learning_rate": 9.127192469192447e-06, + "loss": 0.9497, + "step": 22537 + }, + { + "epoch": 0.54, + "learning_rate": 9.126416549898836e-06, + "loss": 0.9695, + "step": 22538 + }, + { + "epoch": 0.54, + "learning_rate": 9.12564063590505e-06, + "loss": 1.1516, + "step": 22539 + }, + { + "epoch": 0.54, + "learning_rate": 9.124864727215796e-06, + "loss": 1.0446, + "step": 22540 + }, + { + "epoch": 0.54, + "learning_rate": 9.124088823835781e-06, + "loss": 1.0085, + "step": 22541 + }, + { + "epoch": 0.54, + "learning_rate": 9.123312925769716e-06, + "loss": 1.0926, + "step": 22542 + }, + { + "epoch": 0.54, + "learning_rate": 9.122537033022304e-06, + "loss": 1.2569, + "step": 22543 + }, + { + "epoch": 0.54, + "learning_rate": 9.121761145598253e-06, + "loss": 1.246, + "step": 22544 + }, + { + "epoch": 0.54, + "learning_rate": 9.120985263502269e-06, + "loss": 0.9198, + "step": 22545 + }, + { + "epoch": 0.54, + "learning_rate": 9.120209386739063e-06, + "loss": 1.1872, + "step": 22546 + }, + { + "epoch": 0.54, + "learning_rate": 9.11943351531334e-06, + "loss": 1.1767, + "step": 22547 + }, + { + "epoch": 0.54, + "learning_rate": 9.118657649229806e-06, + "loss": 0.733, + "step": 22548 + }, + { + "epoch": 0.54, + "learning_rate": 9.117881788493166e-06, + "loss": 1.0392, + "step": 22549 + }, + { + "epoch": 0.54, + "learning_rate": 9.117105933108132e-06, + "loss": 1.0872, + "step": 22550 + }, + { + "epoch": 0.54, + "learning_rate": 9.11633008307941e-06, + "loss": 0.7583, + "step": 22551 + }, + { + "epoch": 0.54, + "learning_rate": 9.115554238411704e-06, + "loss": 0.6711, + "step": 22552 + }, + { + "epoch": 0.54, + "learning_rate": 9.114778399109722e-06, + "loss": 1.246, + "step": 22553 + }, + { + "epoch": 0.54, + "learning_rate": 9.114002565178172e-06, + "loss": 1.0821, + "step": 22554 + }, + { + "epoch": 0.54, + "learning_rate": 9.113226736621761e-06, + "loss": 1.1864, + "step": 22555 + }, + { + "epoch": 0.54, + "learning_rate": 9.112450913445195e-06, + "loss": 1.2608, + "step": 22556 + }, + { + "epoch": 0.54, + "learning_rate": 9.111675095653177e-06, + "loss": 1.1042, + "step": 22557 + }, + { + "epoch": 0.54, + "learning_rate": 9.11089928325042e-06, + "loss": 1.0257, + "step": 22558 + }, + { + "epoch": 0.54, + "learning_rate": 9.110123476241627e-06, + "loss": 1.3265, + "step": 22559 + }, + { + "epoch": 0.54, + "learning_rate": 9.109347674631508e-06, + "loss": 0.9408, + "step": 22560 + }, + { + "epoch": 0.54, + "learning_rate": 9.108571878424763e-06, + "loss": 1.148, + "step": 22561 + }, + { + "epoch": 0.54, + "learning_rate": 9.107796087626106e-06, + "loss": 1.1414, + "step": 22562 + }, + { + "epoch": 0.54, + "learning_rate": 9.107020302240238e-06, + "loss": 0.9842, + "step": 22563 + }, + { + "epoch": 0.54, + "learning_rate": 9.10624452227187e-06, + "loss": 0.9153, + "step": 22564 + }, + { + "epoch": 0.54, + "learning_rate": 9.105468747725704e-06, + "loss": 0.8402, + "step": 22565 + }, + { + "epoch": 0.54, + "learning_rate": 9.10469297860645e-06, + "loss": 0.899, + "step": 22566 + }, + { + "epoch": 0.54, + "learning_rate": 9.103917214918814e-06, + "loss": 1.2504, + "step": 22567 + }, + { + "epoch": 0.54, + "learning_rate": 9.103141456667502e-06, + "loss": 1.0333, + "step": 22568 + }, + { + "epoch": 0.54, + "learning_rate": 9.102365703857217e-06, + "loss": 0.9028, + "step": 22569 + }, + { + "epoch": 0.54, + "learning_rate": 9.10158995649267e-06, + "loss": 1.1277, + "step": 22570 + }, + { + "epoch": 0.54, + "learning_rate": 9.100814214578564e-06, + "loss": 1.0974, + "step": 22571 + }, + { + "epoch": 0.54, + "learning_rate": 9.100038478119605e-06, + "loss": 1.167, + "step": 22572 + }, + { + "epoch": 0.54, + "learning_rate": 9.099262747120506e-06, + "loss": 1.0323, + "step": 22573 + }, + { + "epoch": 0.54, + "learning_rate": 9.098487021585966e-06, + "loss": 0.8605, + "step": 22574 + }, + { + "epoch": 0.54, + "learning_rate": 9.097711301520693e-06, + "loss": 1.1715, + "step": 22575 + }, + { + "epoch": 0.54, + "learning_rate": 9.096935586929392e-06, + "loss": 0.8278, + "step": 22576 + }, + { + "epoch": 0.54, + "learning_rate": 9.096159877816771e-06, + "loss": 1.0835, + "step": 22577 + }, + { + "epoch": 0.54, + "learning_rate": 9.095384174187535e-06, + "loss": 1.1334, + "step": 22578 + }, + { + "epoch": 0.54, + "learning_rate": 9.094608476046394e-06, + "loss": 1.1062, + "step": 22579 + }, + { + "epoch": 0.54, + "learning_rate": 9.093832783398045e-06, + "loss": 1.0648, + "step": 22580 + }, + { + "epoch": 0.54, + "learning_rate": 9.093057096247204e-06, + "loss": 1.0694, + "step": 22581 + }, + { + "epoch": 0.54, + "learning_rate": 9.09228141459857e-06, + "loss": 1.0862, + "step": 22582 + }, + { + "epoch": 0.54, + "learning_rate": 9.091505738456852e-06, + "loss": 0.7536, + "step": 22583 + }, + { + "epoch": 0.54, + "learning_rate": 9.090730067826752e-06, + "loss": 1.0034, + "step": 22584 + }, + { + "epoch": 0.54, + "learning_rate": 9.089954402712982e-06, + "loss": 1.1697, + "step": 22585 + }, + { + "epoch": 0.54, + "learning_rate": 9.089178743120245e-06, + "loss": 1.1723, + "step": 22586 + }, + { + "epoch": 0.54, + "learning_rate": 9.088403089053245e-06, + "loss": 0.9824, + "step": 22587 + }, + { + "epoch": 0.54, + "learning_rate": 9.087627440516687e-06, + "loss": 1.0902, + "step": 22588 + }, + { + "epoch": 0.54, + "learning_rate": 9.086851797515283e-06, + "loss": 1.2187, + "step": 22589 + }, + { + "epoch": 0.54, + "learning_rate": 9.086076160053732e-06, + "loss": 0.9612, + "step": 22590 + }, + { + "epoch": 0.54, + "learning_rate": 9.085300528136743e-06, + "loss": 1.1534, + "step": 22591 + }, + { + "epoch": 0.54, + "learning_rate": 9.084524901769017e-06, + "loss": 1.128, + "step": 22592 + }, + { + "epoch": 0.54, + "learning_rate": 9.083749280955267e-06, + "loss": 1.2141, + "step": 22593 + }, + { + "epoch": 0.54, + "learning_rate": 9.082973665700194e-06, + "loss": 0.9971, + "step": 22594 + }, + { + "epoch": 0.54, + "learning_rate": 9.082198056008504e-06, + "loss": 1.1662, + "step": 22595 + }, + { + "epoch": 0.54, + "learning_rate": 9.0814224518849e-06, + "loss": 1.1551, + "step": 22596 + }, + { + "epoch": 0.54, + "learning_rate": 9.080646853334093e-06, + "loss": 1.1779, + "step": 22597 + }, + { + "epoch": 0.54, + "learning_rate": 9.079871260360786e-06, + "loss": 1.1466, + "step": 22598 + }, + { + "epoch": 0.54, + "learning_rate": 9.079095672969681e-06, + "loss": 0.9278, + "step": 22599 + }, + { + "epoch": 0.54, + "learning_rate": 9.078320091165486e-06, + "loss": 0.8699, + "step": 22600 + }, + { + "epoch": 0.54, + "learning_rate": 9.077544514952907e-06, + "loss": 1.4682, + "step": 22601 + }, + { + "epoch": 0.54, + "learning_rate": 9.07676894433665e-06, + "loss": 1.2172, + "step": 22602 + }, + { + "epoch": 0.54, + "learning_rate": 9.075993379321417e-06, + "loss": 1.072, + "step": 22603 + }, + { + "epoch": 0.54, + "learning_rate": 9.075217819911913e-06, + "loss": 1.0423, + "step": 22604 + }, + { + "epoch": 0.54, + "learning_rate": 9.074442266112846e-06, + "loss": 1.1079, + "step": 22605 + }, + { + "epoch": 0.54, + "learning_rate": 9.073666717928924e-06, + "loss": 1.2251, + "step": 22606 + }, + { + "epoch": 0.54, + "learning_rate": 9.072891175364846e-06, + "loss": 1.1905, + "step": 22607 + }, + { + "epoch": 0.54, + "learning_rate": 9.072115638425316e-06, + "loss": 1.2281, + "step": 22608 + }, + { + "epoch": 0.54, + "learning_rate": 9.071340107115045e-06, + "loss": 1.0486, + "step": 22609 + }, + { + "epoch": 0.54, + "learning_rate": 9.070564581438734e-06, + "loss": 1.1049, + "step": 22610 + }, + { + "epoch": 0.54, + "learning_rate": 9.06978906140109e-06, + "loss": 1.0862, + "step": 22611 + }, + { + "epoch": 0.54, + "learning_rate": 9.069013547006816e-06, + "loss": 1.2755, + "step": 22612 + }, + { + "epoch": 0.54, + "learning_rate": 9.06823803826062e-06, + "loss": 1.0556, + "step": 22613 + }, + { + "epoch": 0.54, + "learning_rate": 9.067462535167204e-06, + "loss": 0.9937, + "step": 22614 + }, + { + "epoch": 0.54, + "learning_rate": 9.066687037731273e-06, + "loss": 1.3293, + "step": 22615 + }, + { + "epoch": 0.54, + "learning_rate": 9.065911545957528e-06, + "loss": 0.8999, + "step": 22616 + }, + { + "epoch": 0.54, + "learning_rate": 9.065136059850683e-06, + "loss": 0.808, + "step": 22617 + }, + { + "epoch": 0.54, + "learning_rate": 9.064360579415437e-06, + "loss": 1.1325, + "step": 22618 + }, + { + "epoch": 0.54, + "learning_rate": 9.063585104656494e-06, + "loss": 1.1746, + "step": 22619 + }, + { + "epoch": 0.54, + "learning_rate": 9.06280963557856e-06, + "loss": 1.104, + "step": 22620 + }, + { + "epoch": 0.54, + "learning_rate": 9.062034172186342e-06, + "loss": 1.1028, + "step": 22621 + }, + { + "epoch": 0.54, + "learning_rate": 9.06125871448454e-06, + "loss": 1.0581, + "step": 22622 + }, + { + "epoch": 0.54, + "learning_rate": 9.060483262477859e-06, + "loss": 0.7172, + "step": 22623 + }, + { + "epoch": 0.54, + "learning_rate": 9.059707816171007e-06, + "loss": 0.9939, + "step": 22624 + }, + { + "epoch": 0.54, + "learning_rate": 9.058932375568688e-06, + "loss": 0.8627, + "step": 22625 + }, + { + "epoch": 0.54, + "learning_rate": 9.058156940675603e-06, + "loss": 0.9649, + "step": 22626 + }, + { + "epoch": 0.54, + "learning_rate": 9.057381511496456e-06, + "loss": 1.162, + "step": 22627 + }, + { + "epoch": 0.54, + "learning_rate": 9.056606088035958e-06, + "loss": 1.046, + "step": 22628 + }, + { + "epoch": 0.54, + "learning_rate": 9.055830670298807e-06, + "loss": 0.8722, + "step": 22629 + }, + { + "epoch": 0.54, + "learning_rate": 9.05505525828971e-06, + "loss": 1.3074, + "step": 22630 + }, + { + "epoch": 0.54, + "learning_rate": 9.054279852013367e-06, + "loss": 0.9862, + "step": 22631 + }, + { + "epoch": 0.54, + "learning_rate": 9.05350445147449e-06, + "loss": 1.064, + "step": 22632 + }, + { + "epoch": 0.54, + "learning_rate": 9.052729056677778e-06, + "loss": 0.9423, + "step": 22633 + }, + { + "epoch": 0.54, + "learning_rate": 9.051953667627936e-06, + "loss": 1.1498, + "step": 22634 + }, + { + "epoch": 0.54, + "learning_rate": 9.051178284329665e-06, + "loss": 1.1938, + "step": 22635 + }, + { + "epoch": 0.54, + "learning_rate": 9.050402906787677e-06, + "loss": 1.1161, + "step": 22636 + }, + { + "epoch": 0.54, + "learning_rate": 9.049627535006668e-06, + "loss": 1.1155, + "step": 22637 + }, + { + "epoch": 0.54, + "learning_rate": 9.048852168991348e-06, + "loss": 1.2053, + "step": 22638 + }, + { + "epoch": 0.54, + "learning_rate": 9.048076808746414e-06, + "loss": 1.1593, + "step": 22639 + }, + { + "epoch": 0.54, + "learning_rate": 9.047301454276577e-06, + "loss": 1.2754, + "step": 22640 + }, + { + "epoch": 0.54, + "learning_rate": 9.046526105586538e-06, + "loss": 1.1633, + "step": 22641 + }, + { + "epoch": 0.54, + "learning_rate": 9.045750762681e-06, + "loss": 1.3379, + "step": 22642 + }, + { + "epoch": 0.54, + "learning_rate": 9.044975425564668e-06, + "loss": 0.8467, + "step": 22643 + }, + { + "epoch": 0.54, + "learning_rate": 9.044200094242245e-06, + "loss": 1.1324, + "step": 22644 + }, + { + "epoch": 0.54, + "learning_rate": 9.043424768718437e-06, + "loss": 1.2833, + "step": 22645 + }, + { + "epoch": 0.54, + "learning_rate": 9.042649448997943e-06, + "loss": 1.049, + "step": 22646 + }, + { + "epoch": 0.54, + "learning_rate": 9.04187413508547e-06, + "loss": 0.922, + "step": 22647 + }, + { + "epoch": 0.54, + "learning_rate": 9.041098826985722e-06, + "loss": 1.2084, + "step": 22648 + }, + { + "epoch": 0.54, + "learning_rate": 9.040323524703404e-06, + "loss": 1.1035, + "step": 22649 + }, + { + "epoch": 0.54, + "learning_rate": 9.039548228243216e-06, + "loss": 1.0077, + "step": 22650 + }, + { + "epoch": 0.54, + "learning_rate": 9.03877293760986e-06, + "loss": 0.8856, + "step": 22651 + }, + { + "epoch": 0.54, + "learning_rate": 9.037997652808046e-06, + "loss": 0.9077, + "step": 22652 + }, + { + "epoch": 0.54, + "learning_rate": 9.037222373842473e-06, + "loss": 1.1072, + "step": 22653 + }, + { + "epoch": 0.54, + "learning_rate": 9.036447100717847e-06, + "loss": 0.8689, + "step": 22654 + }, + { + "epoch": 0.54, + "learning_rate": 9.035671833438868e-06, + "loss": 1.0826, + "step": 22655 + }, + { + "epoch": 0.54, + "learning_rate": 9.034896572010241e-06, + "loss": 0.9689, + "step": 22656 + }, + { + "epoch": 0.54, + "learning_rate": 9.03412131643667e-06, + "loss": 1.1352, + "step": 22657 + }, + { + "epoch": 0.54, + "learning_rate": 9.033346066722858e-06, + "loss": 1.1154, + "step": 22658 + }, + { + "epoch": 0.54, + "learning_rate": 9.032570822873506e-06, + "loss": 1.2213, + "step": 22659 + }, + { + "epoch": 0.54, + "learning_rate": 9.031795584893323e-06, + "loss": 0.7911, + "step": 22660 + }, + { + "epoch": 0.54, + "learning_rate": 9.031020352787007e-06, + "loss": 0.9744, + "step": 22661 + }, + { + "epoch": 0.54, + "learning_rate": 9.030245126559262e-06, + "loss": 1.1928, + "step": 22662 + }, + { + "epoch": 0.55, + "learning_rate": 9.02946990621479e-06, + "loss": 1.0017, + "step": 22663 + }, + { + "epoch": 0.55, + "learning_rate": 9.028694691758299e-06, + "loss": 1.1571, + "step": 22664 + }, + { + "epoch": 0.55, + "learning_rate": 9.027919483194487e-06, + "loss": 0.9911, + "step": 22665 + }, + { + "epoch": 0.55, + "learning_rate": 9.027144280528055e-06, + "loss": 1.4363, + "step": 22666 + }, + { + "epoch": 0.55, + "learning_rate": 9.026369083763716e-06, + "loss": 0.983, + "step": 22667 + }, + { + "epoch": 0.55, + "learning_rate": 9.025593892906163e-06, + "loss": 0.9174, + "step": 22668 + }, + { + "epoch": 0.55, + "learning_rate": 9.024818707960105e-06, + "loss": 1.1124, + "step": 22669 + }, + { + "epoch": 0.55, + "learning_rate": 9.024043528930242e-06, + "loss": 1.2504, + "step": 22670 + }, + { + "epoch": 0.55, + "learning_rate": 9.023268355821278e-06, + "loss": 1.1076, + "step": 22671 + }, + { + "epoch": 0.55, + "learning_rate": 9.022493188637913e-06, + "loss": 1.0688, + "step": 22672 + }, + { + "epoch": 0.55, + "learning_rate": 9.021718027384854e-06, + "loss": 0.9395, + "step": 22673 + }, + { + "epoch": 0.55, + "learning_rate": 9.020942872066801e-06, + "loss": 1.2616, + "step": 22674 + }, + { + "epoch": 0.55, + "learning_rate": 9.020167722688458e-06, + "loss": 0.9622, + "step": 22675 + }, + { + "epoch": 0.55, + "learning_rate": 9.019392579254525e-06, + "loss": 1.0758, + "step": 22676 + }, + { + "epoch": 0.55, + "learning_rate": 9.018617441769709e-06, + "loss": 1.2408, + "step": 22677 + }, + { + "epoch": 0.55, + "learning_rate": 9.017842310238709e-06, + "loss": 0.8842, + "step": 22678 + }, + { + "epoch": 0.55, + "learning_rate": 9.017067184666229e-06, + "loss": 0.849, + "step": 22679 + }, + { + "epoch": 0.55, + "learning_rate": 9.01629206505697e-06, + "loss": 1.2597, + "step": 22680 + }, + { + "epoch": 0.55, + "learning_rate": 9.015516951415637e-06, + "loss": 1.0836, + "step": 22681 + }, + { + "epoch": 0.55, + "learning_rate": 9.014741843746931e-06, + "loss": 1.0922, + "step": 22682 + }, + { + "epoch": 0.55, + "learning_rate": 9.013966742055554e-06, + "loss": 1.1086, + "step": 22683 + }, + { + "epoch": 0.55, + "learning_rate": 9.013191646346208e-06, + "loss": 1.2966, + "step": 22684 + }, + { + "epoch": 0.55, + "learning_rate": 9.012416556623597e-06, + "loss": 1.1595, + "step": 22685 + }, + { + "epoch": 0.55, + "learning_rate": 9.011641472892424e-06, + "loss": 1.1333, + "step": 22686 + }, + { + "epoch": 0.55, + "learning_rate": 9.010866395157388e-06, + "loss": 1.1241, + "step": 22687 + }, + { + "epoch": 0.55, + "learning_rate": 9.010091323423192e-06, + "loss": 1.0106, + "step": 22688 + }, + { + "epoch": 0.55, + "learning_rate": 9.009316257694542e-06, + "loss": 1.0554, + "step": 22689 + }, + { + "epoch": 0.55, + "learning_rate": 9.008541197976136e-06, + "loss": 1.2905, + "step": 22690 + }, + { + "epoch": 0.55, + "learning_rate": 9.007766144272676e-06, + "loss": 1.2076, + "step": 22691 + }, + { + "epoch": 0.55, + "learning_rate": 9.006991096588865e-06, + "loss": 1.1594, + "step": 22692 + }, + { + "epoch": 0.55, + "learning_rate": 9.006216054929408e-06, + "loss": 1.0816, + "step": 22693 + }, + { + "epoch": 0.55, + "learning_rate": 9.005441019299004e-06, + "loss": 0.9915, + "step": 22694 + }, + { + "epoch": 0.55, + "learning_rate": 9.004665989702355e-06, + "loss": 0.8656, + "step": 22695 + }, + { + "epoch": 0.55, + "learning_rate": 9.003890966144162e-06, + "loss": 1.2425, + "step": 22696 + }, + { + "epoch": 0.55, + "learning_rate": 9.003115948629128e-06, + "loss": 1.0121, + "step": 22697 + }, + { + "epoch": 0.55, + "learning_rate": 9.002340937161957e-06, + "loss": 0.9522, + "step": 22698 + }, + { + "epoch": 0.55, + "learning_rate": 9.001565931747346e-06, + "loss": 0.9062, + "step": 22699 + }, + { + "epoch": 0.55, + "learning_rate": 9.000790932390004e-06, + "loss": 1.0952, + "step": 22700 + }, + { + "epoch": 0.55, + "learning_rate": 9.000015939094627e-06, + "loss": 1.1062, + "step": 22701 + }, + { + "epoch": 0.55, + "learning_rate": 8.999240951865917e-06, + "loss": 1.2216, + "step": 22702 + }, + { + "epoch": 0.55, + "learning_rate": 8.998465970708575e-06, + "loss": 1.2337, + "step": 22703 + }, + { + "epoch": 0.55, + "learning_rate": 8.997690995627307e-06, + "loss": 1.0884, + "step": 22704 + }, + { + "epoch": 0.55, + "learning_rate": 8.996916026626812e-06, + "loss": 0.9226, + "step": 22705 + }, + { + "epoch": 0.55, + "learning_rate": 8.996141063711791e-06, + "loss": 0.9718, + "step": 22706 + }, + { + "epoch": 0.55, + "learning_rate": 8.995366106886944e-06, + "loss": 0.994, + "step": 22707 + }, + { + "epoch": 0.55, + "learning_rate": 8.994591156156977e-06, + "loss": 0.8378, + "step": 22708 + }, + { + "epoch": 0.55, + "learning_rate": 8.99381621152659e-06, + "loss": 1.1888, + "step": 22709 + }, + { + "epoch": 0.55, + "learning_rate": 8.993041273000483e-06, + "loss": 1.1557, + "step": 22710 + }, + { + "epoch": 0.55, + "learning_rate": 8.992266340583353e-06, + "loss": 0.9981, + "step": 22711 + }, + { + "epoch": 0.55, + "learning_rate": 8.991491414279912e-06, + "loss": 0.9665, + "step": 22712 + }, + { + "epoch": 0.55, + "learning_rate": 8.990716494094853e-06, + "loss": 1.262, + "step": 22713 + }, + { + "epoch": 0.55, + "learning_rate": 8.989941580032879e-06, + "loss": 1.3789, + "step": 22714 + }, + { + "epoch": 0.55, + "learning_rate": 8.98916667209869e-06, + "loss": 0.9294, + "step": 22715 + }, + { + "epoch": 0.55, + "learning_rate": 8.988391770296994e-06, + "loss": 0.9342, + "step": 22716 + }, + { + "epoch": 0.55, + "learning_rate": 8.987616874632485e-06, + "loss": 1.1547, + "step": 22717 + }, + { + "epoch": 0.55, + "learning_rate": 8.986841985109867e-06, + "loss": 1.2772, + "step": 22718 + }, + { + "epoch": 0.55, + "learning_rate": 8.986067101733837e-06, + "loss": 1.1715, + "step": 22719 + }, + { + "epoch": 0.55, + "learning_rate": 8.985292224509103e-06, + "loss": 0.8978, + "step": 22720 + }, + { + "epoch": 0.55, + "learning_rate": 8.98451735344036e-06, + "loss": 1.1152, + "step": 22721 + }, + { + "epoch": 0.55, + "learning_rate": 8.983742488532314e-06, + "loss": 1.016, + "step": 22722 + }, + { + "epoch": 0.55, + "learning_rate": 8.982967629789659e-06, + "loss": 1.3065, + "step": 22723 + }, + { + "epoch": 0.55, + "learning_rate": 8.982192777217105e-06, + "loss": 1.1282, + "step": 22724 + }, + { + "epoch": 0.55, + "learning_rate": 8.981417930819346e-06, + "loss": 0.8894, + "step": 22725 + }, + { + "epoch": 0.55, + "learning_rate": 8.980643090601086e-06, + "loss": 1.076, + "step": 22726 + }, + { + "epoch": 0.55, + "learning_rate": 8.979868256567021e-06, + "loss": 1.2335, + "step": 22727 + }, + { + "epoch": 0.55, + "learning_rate": 8.97909342872186e-06, + "loss": 1.0489, + "step": 22728 + }, + { + "epoch": 0.55, + "learning_rate": 8.978318607070295e-06, + "loss": 1.3953, + "step": 22729 + }, + { + "epoch": 0.55, + "learning_rate": 8.977543791617034e-06, + "loss": 1.0711, + "step": 22730 + }, + { + "epoch": 0.55, + "learning_rate": 8.97676898236677e-06, + "loss": 0.8588, + "step": 22731 + }, + { + "epoch": 0.55, + "learning_rate": 8.975994179324212e-06, + "loss": 1.1241, + "step": 22732 + }, + { + "epoch": 0.55, + "learning_rate": 8.975219382494055e-06, + "loss": 1.0493, + "step": 22733 + }, + { + "epoch": 0.55, + "learning_rate": 8.974444591881001e-06, + "loss": 0.7501, + "step": 22734 + }, + { + "epoch": 0.55, + "learning_rate": 8.97366980748975e-06, + "loss": 1.1742, + "step": 22735 + }, + { + "epoch": 0.55, + "learning_rate": 8.972895029325004e-06, + "loss": 1.0012, + "step": 22736 + }, + { + "epoch": 0.55, + "learning_rate": 8.972120257391462e-06, + "loss": 1.0639, + "step": 22737 + }, + { + "epoch": 0.55, + "learning_rate": 8.971345491693827e-06, + "loss": 1.2921, + "step": 22738 + }, + { + "epoch": 0.55, + "learning_rate": 8.970570732236791e-06, + "loss": 0.9827, + "step": 22739 + }, + { + "epoch": 0.55, + "learning_rate": 8.969795979025065e-06, + "loss": 1.1647, + "step": 22740 + }, + { + "epoch": 0.55, + "learning_rate": 8.969021232063344e-06, + "loss": 1.0511, + "step": 22741 + }, + { + "epoch": 0.55, + "learning_rate": 8.968246491356326e-06, + "loss": 1.0828, + "step": 22742 + }, + { + "epoch": 0.55, + "learning_rate": 8.967471756908714e-06, + "loss": 1.1226, + "step": 22743 + }, + { + "epoch": 0.55, + "learning_rate": 8.966697028725212e-06, + "loss": 1.4891, + "step": 22744 + }, + { + "epoch": 0.55, + "learning_rate": 8.965922306810513e-06, + "loss": 1.1645, + "step": 22745 + }, + { + "epoch": 0.55, + "learning_rate": 8.965147591169319e-06, + "loss": 0.9718, + "step": 22746 + }, + { + "epoch": 0.55, + "learning_rate": 8.964372881806333e-06, + "loss": 1.2275, + "step": 22747 + }, + { + "epoch": 0.55, + "learning_rate": 8.963598178726254e-06, + "loss": 1.2552, + "step": 22748 + }, + { + "epoch": 0.55, + "learning_rate": 8.96282348193378e-06, + "loss": 1.0082, + "step": 22749 + }, + { + "epoch": 0.55, + "learning_rate": 8.962048791433611e-06, + "loss": 1.3435, + "step": 22750 + }, + { + "epoch": 0.55, + "learning_rate": 8.961274107230448e-06, + "loss": 1.038, + "step": 22751 + }, + { + "epoch": 0.55, + "learning_rate": 8.960499429328994e-06, + "loss": 1.0167, + "step": 22752 + }, + { + "epoch": 0.55, + "learning_rate": 8.959724757733944e-06, + "loss": 0.8782, + "step": 22753 + }, + { + "epoch": 0.55, + "learning_rate": 8.958950092449995e-06, + "loss": 1.2877, + "step": 22754 + }, + { + "epoch": 0.55, + "learning_rate": 8.958175433481855e-06, + "loss": 1.1336, + "step": 22755 + }, + { + "epoch": 0.55, + "learning_rate": 8.957400780834219e-06, + "loss": 1.2124, + "step": 22756 + }, + { + "epoch": 0.55, + "learning_rate": 8.956626134511788e-06, + "loss": 1.0303, + "step": 22757 + }, + { + "epoch": 0.55, + "learning_rate": 8.955851494519257e-06, + "loss": 1.0402, + "step": 22758 + }, + { + "epoch": 0.55, + "learning_rate": 8.955076860861334e-06, + "loss": 1.2972, + "step": 22759 + }, + { + "epoch": 0.55, + "learning_rate": 8.954302233542712e-06, + "loss": 0.9953, + "step": 22760 + }, + { + "epoch": 0.55, + "learning_rate": 8.953527612568094e-06, + "loss": 1.1191, + "step": 22761 + }, + { + "epoch": 0.55, + "learning_rate": 8.952752997942174e-06, + "loss": 0.8298, + "step": 22762 + }, + { + "epoch": 0.55, + "learning_rate": 8.951978389669657e-06, + "loss": 1.0478, + "step": 22763 + }, + { + "epoch": 0.55, + "learning_rate": 8.951203787755242e-06, + "loss": 1.1386, + "step": 22764 + }, + { + "epoch": 0.55, + "learning_rate": 8.950429192203627e-06, + "loss": 0.9245, + "step": 22765 + }, + { + "epoch": 0.55, + "learning_rate": 8.949654603019509e-06, + "loss": 1.0166, + "step": 22766 + }, + { + "epoch": 0.55, + "learning_rate": 8.94888002020759e-06, + "loss": 1.1294, + "step": 22767 + }, + { + "epoch": 0.55, + "learning_rate": 8.94810544377257e-06, + "loss": 1.172, + "step": 22768 + }, + { + "epoch": 0.55, + "learning_rate": 8.947330873719147e-06, + "loss": 1.0853, + "step": 22769 + }, + { + "epoch": 0.55, + "learning_rate": 8.946556310052017e-06, + "loss": 1.1166, + "step": 22770 + }, + { + "epoch": 0.55, + "learning_rate": 8.945781752775886e-06, + "loss": 1.0685, + "step": 22771 + }, + { + "epoch": 0.55, + "learning_rate": 8.945007201895449e-06, + "loss": 1.1604, + "step": 22772 + }, + { + "epoch": 0.55, + "learning_rate": 8.944232657415404e-06, + "loss": 0.9027, + "step": 22773 + }, + { + "epoch": 0.55, + "learning_rate": 8.943458119340449e-06, + "loss": 1.1406, + "step": 22774 + }, + { + "epoch": 0.55, + "learning_rate": 8.942683587675288e-06, + "loss": 1.1044, + "step": 22775 + }, + { + "epoch": 0.55, + "learning_rate": 8.941909062424615e-06, + "loss": 1.0965, + "step": 22776 + }, + { + "epoch": 0.55, + "learning_rate": 8.941134543593133e-06, + "loss": 1.0844, + "step": 22777 + }, + { + "epoch": 0.55, + "learning_rate": 8.940360031185536e-06, + "loss": 1.2522, + "step": 22778 + }, + { + "epoch": 0.55, + "learning_rate": 8.939585525206527e-06, + "loss": 1.1944, + "step": 22779 + }, + { + "epoch": 0.55, + "learning_rate": 8.938811025660802e-06, + "loss": 1.1052, + "step": 22780 + }, + { + "epoch": 0.55, + "learning_rate": 8.938036532553063e-06, + "loss": 1.1604, + "step": 22781 + }, + { + "epoch": 0.55, + "learning_rate": 8.937262045888004e-06, + "loss": 1.0245, + "step": 22782 + }, + { + "epoch": 0.55, + "learning_rate": 8.936487565670328e-06, + "loss": 1.0793, + "step": 22783 + }, + { + "epoch": 0.55, + "learning_rate": 8.935713091904732e-06, + "loss": 1.1351, + "step": 22784 + }, + { + "epoch": 0.55, + "learning_rate": 8.934938624595914e-06, + "loss": 1.1642, + "step": 22785 + }, + { + "epoch": 0.55, + "learning_rate": 8.934164163748572e-06, + "loss": 1.1318, + "step": 22786 + }, + { + "epoch": 0.55, + "learning_rate": 8.933389709367406e-06, + "loss": 0.9261, + "step": 22787 + }, + { + "epoch": 0.55, + "learning_rate": 8.932615261457113e-06, + "loss": 1.1295, + "step": 22788 + }, + { + "epoch": 0.55, + "learning_rate": 8.931840820022392e-06, + "loss": 1.1416, + "step": 22789 + }, + { + "epoch": 0.55, + "learning_rate": 8.931066385067944e-06, + "loss": 1.1481, + "step": 22790 + }, + { + "epoch": 0.55, + "learning_rate": 8.930291956598462e-06, + "loss": 1.5036, + "step": 22791 + }, + { + "epoch": 0.55, + "learning_rate": 8.929517534618651e-06, + "loss": 1.3356, + "step": 22792 + }, + { + "epoch": 0.55, + "learning_rate": 8.9287431191332e-06, + "loss": 1.1452, + "step": 22793 + }, + { + "epoch": 0.55, + "learning_rate": 8.927968710146815e-06, + "loss": 1.0912, + "step": 22794 + }, + { + "epoch": 0.55, + "learning_rate": 8.927194307664191e-06, + "loss": 1.0903, + "step": 22795 + }, + { + "epoch": 0.55, + "learning_rate": 8.92641991169003e-06, + "loss": 0.9721, + "step": 22796 + }, + { + "epoch": 0.55, + "learning_rate": 8.925645522229024e-06, + "loss": 1.0438, + "step": 22797 + }, + { + "epoch": 0.55, + "learning_rate": 8.924871139285873e-06, + "loss": 1.1719, + "step": 22798 + }, + { + "epoch": 0.55, + "learning_rate": 8.924096762865278e-06, + "loss": 1.0946, + "step": 22799 + }, + { + "epoch": 0.55, + "learning_rate": 8.923322392971935e-06, + "loss": 1.0205, + "step": 22800 + }, + { + "epoch": 0.55, + "learning_rate": 8.922548029610538e-06, + "loss": 1.0688, + "step": 22801 + }, + { + "epoch": 0.55, + "learning_rate": 8.921773672785793e-06, + "loss": 1.1256, + "step": 22802 + }, + { + "epoch": 0.55, + "learning_rate": 8.920999322502393e-06, + "loss": 1.2029, + "step": 22803 + }, + { + "epoch": 0.55, + "learning_rate": 8.920224978765035e-06, + "loss": 1.0322, + "step": 22804 + }, + { + "epoch": 0.55, + "learning_rate": 8.919450641578418e-06, + "loss": 1.1065, + "step": 22805 + }, + { + "epoch": 0.55, + "learning_rate": 8.918676310947239e-06, + "loss": 1.2132, + "step": 22806 + }, + { + "epoch": 0.55, + "learning_rate": 8.917901986876199e-06, + "loss": 1.2189, + "step": 22807 + }, + { + "epoch": 0.55, + "learning_rate": 8.917127669369992e-06, + "loss": 0.9765, + "step": 22808 + }, + { + "epoch": 0.55, + "learning_rate": 8.916353358433316e-06, + "loss": 0.9341, + "step": 22809 + }, + { + "epoch": 0.55, + "learning_rate": 8.915579054070869e-06, + "loss": 1.1151, + "step": 22810 + }, + { + "epoch": 0.55, + "learning_rate": 8.914804756287352e-06, + "loss": 0.8791, + "step": 22811 + }, + { + "epoch": 0.55, + "learning_rate": 8.914030465087457e-06, + "loss": 1.1639, + "step": 22812 + }, + { + "epoch": 0.55, + "learning_rate": 8.913256180475881e-06, + "loss": 1.149, + "step": 22813 + }, + { + "epoch": 0.55, + "learning_rate": 8.912481902457328e-06, + "loss": 1.0446, + "step": 22814 + }, + { + "epoch": 0.55, + "learning_rate": 8.911707631036492e-06, + "loss": 0.9875, + "step": 22815 + }, + { + "epoch": 0.55, + "learning_rate": 8.910933366218068e-06, + "loss": 1.1428, + "step": 22816 + }, + { + "epoch": 0.55, + "learning_rate": 8.910159108006753e-06, + "loss": 1.2221, + "step": 22817 + }, + { + "epoch": 0.55, + "learning_rate": 8.90938485640725e-06, + "loss": 1.1008, + "step": 22818 + }, + { + "epoch": 0.55, + "learning_rate": 8.90861061142425e-06, + "loss": 1.238, + "step": 22819 + }, + { + "epoch": 0.55, + "learning_rate": 8.907836373062456e-06, + "loss": 1.1537, + "step": 22820 + }, + { + "epoch": 0.55, + "learning_rate": 8.907062141326557e-06, + "loss": 1.1527, + "step": 22821 + }, + { + "epoch": 0.55, + "learning_rate": 8.906287916221259e-06, + "loss": 0.9346, + "step": 22822 + }, + { + "epoch": 0.55, + "learning_rate": 8.905513697751253e-06, + "loss": 0.9384, + "step": 22823 + }, + { + "epoch": 0.55, + "learning_rate": 8.90473948592124e-06, + "loss": 0.8995, + "step": 22824 + }, + { + "epoch": 0.55, + "learning_rate": 8.903965280735912e-06, + "loss": 1.1175, + "step": 22825 + }, + { + "epoch": 0.55, + "learning_rate": 8.903191082199972e-06, + "loss": 1.1815, + "step": 22826 + }, + { + "epoch": 0.55, + "learning_rate": 8.902416890318111e-06, + "loss": 0.9897, + "step": 22827 + }, + { + "epoch": 0.55, + "learning_rate": 8.901642705095031e-06, + "loss": 1.2289, + "step": 22828 + }, + { + "epoch": 0.55, + "learning_rate": 8.900868526535422e-06, + "loss": 1.1653, + "step": 22829 + }, + { + "epoch": 0.55, + "learning_rate": 8.90009435464399e-06, + "loss": 0.8498, + "step": 22830 + }, + { + "epoch": 0.55, + "learning_rate": 8.899320189425425e-06, + "loss": 1.2278, + "step": 22831 + }, + { + "epoch": 0.55, + "learning_rate": 8.898546030884421e-06, + "loss": 1.0118, + "step": 22832 + }, + { + "epoch": 0.55, + "learning_rate": 8.897771879025684e-06, + "loss": 1.0137, + "step": 22833 + }, + { + "epoch": 0.55, + "learning_rate": 8.896997733853904e-06, + "loss": 0.9555, + "step": 22834 + }, + { + "epoch": 0.55, + "learning_rate": 8.896223595373781e-06, + "loss": 0.8072, + "step": 22835 + }, + { + "epoch": 0.55, + "learning_rate": 8.895449463590011e-06, + "loss": 0.7612, + "step": 22836 + }, + { + "epoch": 0.55, + "learning_rate": 8.894675338507288e-06, + "loss": 1.1563, + "step": 22837 + }, + { + "epoch": 0.55, + "learning_rate": 8.893901220130306e-06, + "loss": 0.9976, + "step": 22838 + }, + { + "epoch": 0.55, + "learning_rate": 8.89312710846377e-06, + "loss": 1.123, + "step": 22839 + }, + { + "epoch": 0.55, + "learning_rate": 8.89235300351237e-06, + "loss": 1.1574, + "step": 22840 + }, + { + "epoch": 0.55, + "learning_rate": 8.891578905280806e-06, + "loss": 0.9826, + "step": 22841 + }, + { + "epoch": 0.55, + "learning_rate": 8.890804813773767e-06, + "loss": 1.0844, + "step": 22842 + }, + { + "epoch": 0.55, + "learning_rate": 8.890030728995958e-06, + "loss": 0.8861, + "step": 22843 + }, + { + "epoch": 0.55, + "learning_rate": 8.889256650952072e-06, + "loss": 1.0369, + "step": 22844 + }, + { + "epoch": 0.55, + "learning_rate": 8.888482579646803e-06, + "loss": 0.6727, + "step": 22845 + }, + { + "epoch": 0.55, + "learning_rate": 8.887708515084848e-06, + "loss": 1.2658, + "step": 22846 + }, + { + "epoch": 0.55, + "learning_rate": 8.886934457270905e-06, + "loss": 1.2486, + "step": 22847 + }, + { + "epoch": 0.55, + "learning_rate": 8.88616040620967e-06, + "loss": 1.2636, + "step": 22848 + }, + { + "epoch": 0.55, + "learning_rate": 8.885386361905836e-06, + "loss": 0.9409, + "step": 22849 + }, + { + "epoch": 0.55, + "learning_rate": 8.8846123243641e-06, + "loss": 1.2065, + "step": 22850 + }, + { + "epoch": 0.55, + "learning_rate": 8.883838293589162e-06, + "loss": 1.0564, + "step": 22851 + }, + { + "epoch": 0.55, + "learning_rate": 8.883064269585711e-06, + "loss": 1.06, + "step": 22852 + }, + { + "epoch": 0.55, + "learning_rate": 8.88229025235845e-06, + "loss": 1.1187, + "step": 22853 + }, + { + "epoch": 0.55, + "learning_rate": 8.881516241912066e-06, + "loss": 1.1013, + "step": 22854 + }, + { + "epoch": 0.55, + "learning_rate": 8.880742238251263e-06, + "loss": 1.2046, + "step": 22855 + }, + { + "epoch": 0.55, + "learning_rate": 8.879968241380735e-06, + "loss": 0.9932, + "step": 22856 + }, + { + "epoch": 0.55, + "learning_rate": 8.879194251305175e-06, + "loss": 1.1629, + "step": 22857 + }, + { + "epoch": 0.55, + "learning_rate": 8.878420268029278e-06, + "loss": 0.8698, + "step": 22858 + }, + { + "epoch": 0.55, + "learning_rate": 8.877646291557743e-06, + "loss": 1.3224, + "step": 22859 + }, + { + "epoch": 0.55, + "learning_rate": 8.876872321895264e-06, + "loss": 0.7474, + "step": 22860 + }, + { + "epoch": 0.55, + "learning_rate": 8.876098359046537e-06, + "loss": 0.9311, + "step": 22861 + }, + { + "epoch": 0.55, + "learning_rate": 8.875324403016255e-06, + "loss": 1.3195, + "step": 22862 + }, + { + "epoch": 0.55, + "learning_rate": 8.874550453809117e-06, + "loss": 1.2046, + "step": 22863 + }, + { + "epoch": 0.55, + "learning_rate": 8.873776511429817e-06, + "loss": 0.9577, + "step": 22864 + }, + { + "epoch": 0.55, + "learning_rate": 8.87300257588305e-06, + "loss": 0.976, + "step": 22865 + }, + { + "epoch": 0.55, + "learning_rate": 8.872228647173509e-06, + "loss": 1.2283, + "step": 22866 + }, + { + "epoch": 0.55, + "learning_rate": 8.871454725305894e-06, + "loss": 1.1404, + "step": 22867 + }, + { + "epoch": 0.55, + "learning_rate": 8.870680810284897e-06, + "loss": 1.0032, + "step": 22868 + }, + { + "epoch": 0.55, + "learning_rate": 8.869906902115211e-06, + "loss": 1.0208, + "step": 22869 + }, + { + "epoch": 0.55, + "learning_rate": 8.869133000801539e-06, + "loss": 1.234, + "step": 22870 + }, + { + "epoch": 0.55, + "learning_rate": 8.868359106348569e-06, + "loss": 0.8633, + "step": 22871 + }, + { + "epoch": 0.55, + "learning_rate": 8.867585218760997e-06, + "loss": 1.1212, + "step": 22872 + }, + { + "epoch": 0.55, + "learning_rate": 8.866811338043519e-06, + "loss": 1.1702, + "step": 22873 + }, + { + "epoch": 0.55, + "learning_rate": 8.866037464200832e-06, + "loss": 1.1641, + "step": 22874 + }, + { + "epoch": 0.55, + "learning_rate": 8.865263597237629e-06, + "loss": 1.122, + "step": 22875 + }, + { + "epoch": 0.55, + "learning_rate": 8.864489737158604e-06, + "loss": 0.9851, + "step": 22876 + }, + { + "epoch": 0.55, + "learning_rate": 8.863715883968451e-06, + "loss": 1.0723, + "step": 22877 + }, + { + "epoch": 0.55, + "learning_rate": 8.86294203767187e-06, + "loss": 1.0968, + "step": 22878 + }, + { + "epoch": 0.55, + "learning_rate": 8.862168198273549e-06, + "loss": 0.7297, + "step": 22879 + }, + { + "epoch": 0.55, + "learning_rate": 8.861394365778188e-06, + "loss": 0.9808, + "step": 22880 + }, + { + "epoch": 0.55, + "learning_rate": 8.860620540190477e-06, + "loss": 1.1331, + "step": 22881 + }, + { + "epoch": 0.55, + "learning_rate": 8.859846721515116e-06, + "loss": 1.0579, + "step": 22882 + }, + { + "epoch": 0.55, + "learning_rate": 8.859072909756795e-06, + "loss": 1.1915, + "step": 22883 + }, + { + "epoch": 0.55, + "learning_rate": 8.858299104920213e-06, + "loss": 1.0049, + "step": 22884 + }, + { + "epoch": 0.55, + "learning_rate": 8.857525307010058e-06, + "loss": 1.233, + "step": 22885 + }, + { + "epoch": 0.55, + "learning_rate": 8.85675151603103e-06, + "loss": 1.1719, + "step": 22886 + }, + { + "epoch": 0.55, + "learning_rate": 8.855977731987823e-06, + "loss": 0.7339, + "step": 22887 + }, + { + "epoch": 0.55, + "learning_rate": 8.855203954885128e-06, + "loss": 1.1531, + "step": 22888 + }, + { + "epoch": 0.55, + "learning_rate": 8.854430184727642e-06, + "loss": 1.1173, + "step": 22889 + }, + { + "epoch": 0.55, + "learning_rate": 8.853656421520059e-06, + "loss": 0.9932, + "step": 22890 + }, + { + "epoch": 0.55, + "learning_rate": 8.852882665267073e-06, + "loss": 1.1969, + "step": 22891 + }, + { + "epoch": 0.55, + "learning_rate": 8.852108915973377e-06, + "loss": 1.2347, + "step": 22892 + }, + { + "epoch": 0.55, + "learning_rate": 8.851335173643665e-06, + "loss": 1.0353, + "step": 22893 + }, + { + "epoch": 0.55, + "learning_rate": 8.850561438282636e-06, + "loss": 1.0756, + "step": 22894 + }, + { + "epoch": 0.55, + "learning_rate": 8.849787709894978e-06, + "loss": 0.9958, + "step": 22895 + }, + { + "epoch": 0.55, + "learning_rate": 8.84901398848539e-06, + "loss": 1.0956, + "step": 22896 + }, + { + "epoch": 0.55, + "learning_rate": 8.848240274058558e-06, + "loss": 1.0053, + "step": 22897 + }, + { + "epoch": 0.55, + "learning_rate": 8.847466566619185e-06, + "loss": 1.0763, + "step": 22898 + }, + { + "epoch": 0.55, + "learning_rate": 8.846692866171963e-06, + "loss": 1.3423, + "step": 22899 + }, + { + "epoch": 0.55, + "learning_rate": 8.845919172721581e-06, + "loss": 0.9002, + "step": 22900 + }, + { + "epoch": 0.55, + "learning_rate": 8.845145486272736e-06, + "loss": 1.0474, + "step": 22901 + }, + { + "epoch": 0.55, + "learning_rate": 8.844371806830123e-06, + "loss": 1.2128, + "step": 22902 + }, + { + "epoch": 0.55, + "learning_rate": 8.843598134398434e-06, + "loss": 0.8027, + "step": 22903 + }, + { + "epoch": 0.55, + "learning_rate": 8.842824468982363e-06, + "loss": 1.0662, + "step": 22904 + }, + { + "epoch": 0.55, + "learning_rate": 8.842050810586603e-06, + "loss": 1.0945, + "step": 22905 + }, + { + "epoch": 0.55, + "learning_rate": 8.84127715921585e-06, + "loss": 0.8005, + "step": 22906 + }, + { + "epoch": 0.55, + "learning_rate": 8.840503514874795e-06, + "loss": 0.9746, + "step": 22907 + }, + { + "epoch": 0.55, + "learning_rate": 8.839729877568135e-06, + "loss": 1.2159, + "step": 22908 + }, + { + "epoch": 0.55, + "learning_rate": 8.838956247300555e-06, + "loss": 1.0213, + "step": 22909 + }, + { + "epoch": 0.55, + "learning_rate": 8.838182624076759e-06, + "loss": 0.8436, + "step": 22910 + }, + { + "epoch": 0.55, + "learning_rate": 8.837409007901436e-06, + "loss": 0.9586, + "step": 22911 + }, + { + "epoch": 0.55, + "learning_rate": 8.836635398779278e-06, + "loss": 1.1553, + "step": 22912 + }, + { + "epoch": 0.55, + "learning_rate": 8.835861796714977e-06, + "loss": 1.0139, + "step": 22913 + }, + { + "epoch": 0.55, + "learning_rate": 8.83508820171323e-06, + "loss": 1.036, + "step": 22914 + }, + { + "epoch": 0.55, + "learning_rate": 8.834314613778731e-06, + "loss": 0.9003, + "step": 22915 + }, + { + "epoch": 0.55, + "learning_rate": 8.833541032916168e-06, + "loss": 1.0372, + "step": 22916 + }, + { + "epoch": 0.55, + "learning_rate": 8.83276745913024e-06, + "loss": 0.9609, + "step": 22917 + }, + { + "epoch": 0.55, + "learning_rate": 8.831993892425636e-06, + "loss": 0.9729, + "step": 22918 + }, + { + "epoch": 0.55, + "learning_rate": 8.831220332807052e-06, + "loss": 1.0293, + "step": 22919 + }, + { + "epoch": 0.55, + "learning_rate": 8.830446780279175e-06, + "loss": 1.106, + "step": 22920 + }, + { + "epoch": 0.55, + "learning_rate": 8.829673234846706e-06, + "loss": 0.9793, + "step": 22921 + }, + { + "epoch": 0.55, + "learning_rate": 8.828899696514335e-06, + "loss": 0.9726, + "step": 22922 + }, + { + "epoch": 0.55, + "learning_rate": 8.828126165286752e-06, + "loss": 1.1201, + "step": 22923 + }, + { + "epoch": 0.55, + "learning_rate": 8.827352641168652e-06, + "loss": 1.0772, + "step": 22924 + }, + { + "epoch": 0.55, + "learning_rate": 8.826579124164729e-06, + "loss": 0.8877, + "step": 22925 + }, + { + "epoch": 0.55, + "learning_rate": 8.825805614279672e-06, + "loss": 1.4331, + "step": 22926 + }, + { + "epoch": 0.55, + "learning_rate": 8.82503211151818e-06, + "loss": 1.0517, + "step": 22927 + }, + { + "epoch": 0.55, + "learning_rate": 8.824258615884937e-06, + "loss": 0.8492, + "step": 22928 + }, + { + "epoch": 0.55, + "learning_rate": 8.823485127384643e-06, + "loss": 0.8744, + "step": 22929 + }, + { + "epoch": 0.55, + "learning_rate": 8.82271164602199e-06, + "loss": 1.1511, + "step": 22930 + }, + { + "epoch": 0.55, + "learning_rate": 8.821938171801665e-06, + "loss": 1.0162, + "step": 22931 + }, + { + "epoch": 0.55, + "learning_rate": 8.821164704728364e-06, + "loss": 1.0554, + "step": 22932 + }, + { + "epoch": 0.55, + "learning_rate": 8.820391244806782e-06, + "loss": 1.1275, + "step": 22933 + }, + { + "epoch": 0.55, + "learning_rate": 8.819617792041607e-06, + "loss": 0.9613, + "step": 22934 + }, + { + "epoch": 0.55, + "learning_rate": 8.818844346437536e-06, + "loss": 0.8556, + "step": 22935 + }, + { + "epoch": 0.55, + "learning_rate": 8.818070907999254e-06, + "loss": 0.7352, + "step": 22936 + }, + { + "epoch": 0.55, + "learning_rate": 8.81729747673146e-06, + "loss": 0.885, + "step": 22937 + }, + { + "epoch": 0.55, + "learning_rate": 8.816524052638846e-06, + "loss": 1.1106, + "step": 22938 + }, + { + "epoch": 0.55, + "learning_rate": 8.8157506357261e-06, + "loss": 1.0982, + "step": 22939 + }, + { + "epoch": 0.55, + "learning_rate": 8.814977225997916e-06, + "loss": 1.1061, + "step": 22940 + }, + { + "epoch": 0.55, + "learning_rate": 8.814203823458988e-06, + "loss": 0.8786, + "step": 22941 + }, + { + "epoch": 0.55, + "learning_rate": 8.813430428114006e-06, + "loss": 0.9853, + "step": 22942 + }, + { + "epoch": 0.55, + "learning_rate": 8.812657039967663e-06, + "loss": 1.1288, + "step": 22943 + }, + { + "epoch": 0.55, + "learning_rate": 8.811883659024647e-06, + "loss": 1.2661, + "step": 22944 + }, + { + "epoch": 0.55, + "learning_rate": 8.811110285289657e-06, + "loss": 1.0068, + "step": 22945 + }, + { + "epoch": 0.55, + "learning_rate": 8.810336918767381e-06, + "loss": 1.1875, + "step": 22946 + }, + { + "epoch": 0.55, + "learning_rate": 8.80956355946251e-06, + "loss": 1.2224, + "step": 22947 + }, + { + "epoch": 0.55, + "learning_rate": 8.808790207379736e-06, + "loss": 1.0712, + "step": 22948 + }, + { + "epoch": 0.55, + "learning_rate": 8.808016862523753e-06, + "loss": 1.1738, + "step": 22949 + }, + { + "epoch": 0.55, + "learning_rate": 8.807243524899253e-06, + "loss": 1.0619, + "step": 22950 + }, + { + "epoch": 0.55, + "learning_rate": 8.806470194510924e-06, + "loss": 0.8256, + "step": 22951 + }, + { + "epoch": 0.55, + "learning_rate": 8.805696871363459e-06, + "loss": 1.1536, + "step": 22952 + }, + { + "epoch": 0.55, + "learning_rate": 8.804923555461551e-06, + "loss": 1.2007, + "step": 22953 + }, + { + "epoch": 0.55, + "learning_rate": 8.804150246809891e-06, + "loss": 1.1897, + "step": 22954 + }, + { + "epoch": 0.55, + "learning_rate": 8.803376945413172e-06, + "loss": 0.8294, + "step": 22955 + }, + { + "epoch": 0.55, + "learning_rate": 8.802603651276079e-06, + "loss": 1.1204, + "step": 22956 + }, + { + "epoch": 0.55, + "learning_rate": 8.80183036440331e-06, + "loss": 1.0256, + "step": 22957 + }, + { + "epoch": 0.55, + "learning_rate": 8.801057084799555e-06, + "loss": 1.2424, + "step": 22958 + }, + { + "epoch": 0.55, + "learning_rate": 8.800283812469505e-06, + "loss": 1.1077, + "step": 22959 + }, + { + "epoch": 0.55, + "learning_rate": 8.79951054741785e-06, + "loss": 1.1763, + "step": 22960 + }, + { + "epoch": 0.55, + "learning_rate": 8.798737289649283e-06, + "loss": 1.0469, + "step": 22961 + }, + { + "epoch": 0.55, + "learning_rate": 8.797964039168493e-06, + "loss": 1.2583, + "step": 22962 + }, + { + "epoch": 0.55, + "learning_rate": 8.797190795980173e-06, + "loss": 1.0949, + "step": 22963 + }, + { + "epoch": 0.55, + "learning_rate": 8.796417560089014e-06, + "loss": 0.9749, + "step": 22964 + }, + { + "epoch": 0.55, + "learning_rate": 8.795644331499706e-06, + "loss": 1.1353, + "step": 22965 + }, + { + "epoch": 0.55, + "learning_rate": 8.794871110216942e-06, + "loss": 0.6534, + "step": 22966 + }, + { + "epoch": 0.55, + "learning_rate": 8.794097896245409e-06, + "loss": 1.0439, + "step": 22967 + }, + { + "epoch": 0.55, + "learning_rate": 8.793324689589802e-06, + "loss": 0.9052, + "step": 22968 + }, + { + "epoch": 0.55, + "learning_rate": 8.79255149025481e-06, + "loss": 1.09, + "step": 22969 + }, + { + "epoch": 0.55, + "learning_rate": 8.791778298245126e-06, + "loss": 1.1253, + "step": 22970 + }, + { + "epoch": 0.55, + "learning_rate": 8.791005113565436e-06, + "loss": 1.2674, + "step": 22971 + }, + { + "epoch": 0.55, + "learning_rate": 8.790231936220435e-06, + "loss": 1.0545, + "step": 22972 + }, + { + "epoch": 0.55, + "learning_rate": 8.789458766214811e-06, + "loss": 1.007, + "step": 22973 + }, + { + "epoch": 0.55, + "learning_rate": 8.788685603553259e-06, + "loss": 1.027, + "step": 22974 + }, + { + "epoch": 0.55, + "learning_rate": 8.787912448240461e-06, + "loss": 0.9726, + "step": 22975 + }, + { + "epoch": 0.55, + "learning_rate": 8.787139300281117e-06, + "loss": 1.2301, + "step": 22976 + }, + { + "epoch": 0.55, + "learning_rate": 8.786366159679915e-06, + "loss": 1.1477, + "step": 22977 + }, + { + "epoch": 0.55, + "learning_rate": 8.785593026441543e-06, + "loss": 1.2162, + "step": 22978 + }, + { + "epoch": 0.55, + "learning_rate": 8.78481990057069e-06, + "loss": 0.8563, + "step": 22979 + }, + { + "epoch": 0.55, + "learning_rate": 8.78404678207205e-06, + "loss": 1.1949, + "step": 22980 + }, + { + "epoch": 0.55, + "learning_rate": 8.783273670950312e-06, + "loss": 1.0389, + "step": 22981 + }, + { + "epoch": 0.55, + "learning_rate": 8.78250056721017e-06, + "loss": 0.9918, + "step": 22982 + }, + { + "epoch": 0.55, + "learning_rate": 8.781727470856305e-06, + "loss": 1.0139, + "step": 22983 + }, + { + "epoch": 0.55, + "learning_rate": 8.780954381893415e-06, + "loss": 1.1244, + "step": 22984 + }, + { + "epoch": 0.55, + "learning_rate": 8.78018130032619e-06, + "loss": 1.2566, + "step": 22985 + }, + { + "epoch": 0.55, + "learning_rate": 8.779408226159318e-06, + "loss": 0.8689, + "step": 22986 + }, + { + "epoch": 0.55, + "learning_rate": 8.778635159397486e-06, + "loss": 1.1519, + "step": 22987 + }, + { + "epoch": 0.55, + "learning_rate": 8.77786210004539e-06, + "loss": 0.8866, + "step": 22988 + }, + { + "epoch": 0.55, + "learning_rate": 8.777089048107717e-06, + "loss": 1.1437, + "step": 22989 + }, + { + "epoch": 0.55, + "learning_rate": 8.776316003589156e-06, + "loss": 1.1137, + "step": 22990 + }, + { + "epoch": 0.55, + "learning_rate": 8.775542966494397e-06, + "loss": 1.0948, + "step": 22991 + }, + { + "epoch": 0.55, + "learning_rate": 8.774769936828133e-06, + "loss": 0.861, + "step": 22992 + }, + { + "epoch": 0.55, + "learning_rate": 8.773996914595051e-06, + "loss": 1.0576, + "step": 22993 + }, + { + "epoch": 0.55, + "learning_rate": 8.773223899799843e-06, + "loss": 1.113, + "step": 22994 + }, + { + "epoch": 0.55, + "learning_rate": 8.772450892447192e-06, + "loss": 1.0008, + "step": 22995 + }, + { + "epoch": 0.55, + "learning_rate": 8.771677892541796e-06, + "loss": 1.187, + "step": 22996 + }, + { + "epoch": 0.55, + "learning_rate": 8.770904900088344e-06, + "loss": 0.906, + "step": 22997 + }, + { + "epoch": 0.55, + "learning_rate": 8.77013191509152e-06, + "loss": 1.0908, + "step": 22998 + }, + { + "epoch": 0.55, + "learning_rate": 8.769358937556012e-06, + "loss": 0.8138, + "step": 22999 + }, + { + "epoch": 0.55, + "learning_rate": 8.768585967486518e-06, + "loss": 0.8238, + "step": 23000 + }, + { + "epoch": 0.55, + "learning_rate": 8.767813004887722e-06, + "loss": 0.8318, + "step": 23001 + }, + { + "epoch": 0.55, + "learning_rate": 8.767040049764318e-06, + "loss": 1.2459, + "step": 23002 + }, + { + "epoch": 0.55, + "learning_rate": 8.766267102120992e-06, + "loss": 1.0059, + "step": 23003 + }, + { + "epoch": 0.55, + "learning_rate": 8.765494161962428e-06, + "loss": 1.0383, + "step": 23004 + }, + { + "epoch": 0.55, + "learning_rate": 8.764721229293325e-06, + "loss": 1.0032, + "step": 23005 + }, + { + "epoch": 0.55, + "learning_rate": 8.763948304118368e-06, + "loss": 0.9713, + "step": 23006 + }, + { + "epoch": 0.55, + "learning_rate": 8.763175386442246e-06, + "loss": 0.9667, + "step": 23007 + }, + { + "epoch": 0.55, + "learning_rate": 8.762402476269645e-06, + "loss": 1.0363, + "step": 23008 + }, + { + "epoch": 0.55, + "learning_rate": 8.761629573605259e-06, + "loss": 1.1931, + "step": 23009 + }, + { + "epoch": 0.55, + "learning_rate": 8.760856678453776e-06, + "loss": 1.0542, + "step": 23010 + }, + { + "epoch": 0.55, + "learning_rate": 8.760083790819881e-06, + "loss": 0.773, + "step": 23011 + }, + { + "epoch": 0.55, + "learning_rate": 8.759310910708268e-06, + "loss": 0.9275, + "step": 23012 + }, + { + "epoch": 0.55, + "learning_rate": 8.758538038123625e-06, + "loss": 1.0975, + "step": 23013 + }, + { + "epoch": 0.55, + "learning_rate": 8.757765173070637e-06, + "loss": 1.192, + "step": 23014 + }, + { + "epoch": 0.55, + "learning_rate": 8.756992315553999e-06, + "loss": 1.0816, + "step": 23015 + }, + { + "epoch": 0.55, + "learning_rate": 8.756219465578393e-06, + "loss": 0.8163, + "step": 23016 + }, + { + "epoch": 0.55, + "learning_rate": 8.755446623148514e-06, + "loss": 1.1312, + "step": 23017 + }, + { + "epoch": 0.55, + "learning_rate": 8.754673788269046e-06, + "loss": 1.1944, + "step": 23018 + }, + { + "epoch": 0.55, + "learning_rate": 8.753900960944678e-06, + "loss": 0.9482, + "step": 23019 + }, + { + "epoch": 0.55, + "learning_rate": 8.7531281411801e-06, + "loss": 0.9421, + "step": 23020 + }, + { + "epoch": 0.55, + "learning_rate": 8.752355328980002e-06, + "loss": 1.2474, + "step": 23021 + }, + { + "epoch": 0.55, + "learning_rate": 8.75158252434907e-06, + "loss": 0.9917, + "step": 23022 + }, + { + "epoch": 0.55, + "learning_rate": 8.750809727291995e-06, + "loss": 1.1153, + "step": 23023 + }, + { + "epoch": 0.55, + "learning_rate": 8.75003693781346e-06, + "loss": 1.1747, + "step": 23024 + }, + { + "epoch": 0.55, + "learning_rate": 8.749264155918158e-06, + "loss": 1.2323, + "step": 23025 + }, + { + "epoch": 0.55, + "learning_rate": 8.748491381610777e-06, + "loss": 0.914, + "step": 23026 + }, + { + "epoch": 0.55, + "learning_rate": 8.747718614896006e-06, + "loss": 1.007, + "step": 23027 + }, + { + "epoch": 0.55, + "learning_rate": 8.746945855778528e-06, + "loss": 1.1227, + "step": 23028 + }, + { + "epoch": 0.55, + "learning_rate": 8.746173104263036e-06, + "loss": 1.0986, + "step": 23029 + }, + { + "epoch": 0.55, + "learning_rate": 8.745400360354219e-06, + "loss": 1.2013, + "step": 23030 + }, + { + "epoch": 0.55, + "learning_rate": 8.744627624056762e-06, + "loss": 0.8867, + "step": 23031 + }, + { + "epoch": 0.55, + "learning_rate": 8.74385489537535e-06, + "loss": 1.2054, + "step": 23032 + }, + { + "epoch": 0.55, + "learning_rate": 8.74308217431468e-06, + "loss": 1.024, + "step": 23033 + }, + { + "epoch": 0.55, + "learning_rate": 8.742309460879433e-06, + "loss": 1.1629, + "step": 23034 + }, + { + "epoch": 0.55, + "learning_rate": 8.7415367550743e-06, + "loss": 1.0561, + "step": 23035 + }, + { + "epoch": 0.55, + "learning_rate": 8.740764056903962e-06, + "loss": 0.8704, + "step": 23036 + }, + { + "epoch": 0.55, + "learning_rate": 8.739991366373118e-06, + "loss": 1.075, + "step": 23037 + }, + { + "epoch": 0.55, + "learning_rate": 8.73921868348645e-06, + "loss": 0.9572, + "step": 23038 + }, + { + "epoch": 0.55, + "learning_rate": 8.738446008248642e-06, + "loss": 0.8928, + "step": 23039 + }, + { + "epoch": 0.55, + "learning_rate": 8.737673340664387e-06, + "loss": 1.1503, + "step": 23040 + }, + { + "epoch": 0.55, + "learning_rate": 8.736900680738373e-06, + "loss": 1.3151, + "step": 23041 + }, + { + "epoch": 0.55, + "learning_rate": 8.736128028475283e-06, + "loss": 0.9156, + "step": 23042 + }, + { + "epoch": 0.55, + "learning_rate": 8.735355383879807e-06, + "loss": 1.1645, + "step": 23043 + }, + { + "epoch": 0.55, + "learning_rate": 8.734582746956633e-06, + "loss": 0.8444, + "step": 23044 + }, + { + "epoch": 0.55, + "learning_rate": 8.73381011771045e-06, + "loss": 1.0007, + "step": 23045 + }, + { + "epoch": 0.55, + "learning_rate": 8.733037496145941e-06, + "loss": 1.0845, + "step": 23046 + }, + { + "epoch": 0.55, + "learning_rate": 8.732264882267794e-06, + "loss": 1.1383, + "step": 23047 + }, + { + "epoch": 0.55, + "learning_rate": 8.7314922760807e-06, + "loss": 1.0581, + "step": 23048 + }, + { + "epoch": 0.55, + "learning_rate": 8.730719677589344e-06, + "loss": 0.7196, + "step": 23049 + }, + { + "epoch": 0.55, + "learning_rate": 8.729947086798413e-06, + "loss": 1.1939, + "step": 23050 + }, + { + "epoch": 0.55, + "learning_rate": 8.729174503712592e-06, + "loss": 0.9013, + "step": 23051 + }, + { + "epoch": 0.55, + "learning_rate": 8.728401928336574e-06, + "loss": 0.8814, + "step": 23052 + }, + { + "epoch": 0.55, + "learning_rate": 8.72762936067504e-06, + "loss": 1.1227, + "step": 23053 + }, + { + "epoch": 0.55, + "learning_rate": 8.726856800732682e-06, + "loss": 0.8933, + "step": 23054 + }, + { + "epoch": 0.55, + "learning_rate": 8.72608424851418e-06, + "loss": 0.9135, + "step": 23055 + }, + { + "epoch": 0.55, + "learning_rate": 8.725311704024229e-06, + "loss": 1.1357, + "step": 23056 + }, + { + "epoch": 0.55, + "learning_rate": 8.72453916726751e-06, + "loss": 0.9807, + "step": 23057 + }, + { + "epoch": 0.55, + "learning_rate": 8.723766638248714e-06, + "loss": 1.1541, + "step": 23058 + }, + { + "epoch": 0.55, + "learning_rate": 8.722994116972522e-06, + "loss": 1.1516, + "step": 23059 + }, + { + "epoch": 0.55, + "learning_rate": 8.722221603443627e-06, + "loss": 1.1536, + "step": 23060 + }, + { + "epoch": 0.55, + "learning_rate": 8.721449097666714e-06, + "loss": 0.9179, + "step": 23061 + }, + { + "epoch": 0.55, + "learning_rate": 8.720676599646467e-06, + "loss": 0.9985, + "step": 23062 + }, + { + "epoch": 0.55, + "learning_rate": 8.719904109387573e-06, + "loss": 1.0278, + "step": 23063 + }, + { + "epoch": 0.55, + "learning_rate": 8.719131626894721e-06, + "loss": 0.7627, + "step": 23064 + }, + { + "epoch": 0.55, + "learning_rate": 8.718359152172597e-06, + "loss": 1.1299, + "step": 23065 + }, + { + "epoch": 0.55, + "learning_rate": 8.717586685225884e-06, + "loss": 0.9261, + "step": 23066 + }, + { + "epoch": 0.55, + "learning_rate": 8.716814226059273e-06, + "loss": 1.1656, + "step": 23067 + }, + { + "epoch": 0.55, + "learning_rate": 8.716041774677448e-06, + "loss": 0.9769, + "step": 23068 + }, + { + "epoch": 0.55, + "learning_rate": 8.715269331085096e-06, + "loss": 1.2547, + "step": 23069 + }, + { + "epoch": 0.55, + "learning_rate": 8.714496895286903e-06, + "loss": 0.8826, + "step": 23070 + }, + { + "epoch": 0.55, + "learning_rate": 8.713724467287552e-06, + "loss": 1.1607, + "step": 23071 + }, + { + "epoch": 0.55, + "learning_rate": 8.712952047091735e-06, + "loss": 1.0676, + "step": 23072 + }, + { + "epoch": 0.55, + "learning_rate": 8.712179634704134e-06, + "loss": 1.1121, + "step": 23073 + }, + { + "epoch": 0.55, + "learning_rate": 8.711407230129438e-06, + "loss": 1.24, + "step": 23074 + }, + { + "epoch": 0.55, + "learning_rate": 8.710634833372327e-06, + "loss": 1.2494, + "step": 23075 + }, + { + "epoch": 0.55, + "learning_rate": 8.709862444437495e-06, + "loss": 1.3416, + "step": 23076 + }, + { + "epoch": 0.55, + "learning_rate": 8.709090063329623e-06, + "loss": 0.9511, + "step": 23077 + }, + { + "epoch": 0.55, + "learning_rate": 8.708317690053401e-06, + "loss": 0.9268, + "step": 23078 + }, + { + "epoch": 0.56, + "learning_rate": 8.707545324613507e-06, + "loss": 1.0531, + "step": 23079 + }, + { + "epoch": 0.56, + "learning_rate": 8.706772967014633e-06, + "loss": 1.0539, + "step": 23080 + }, + { + "epoch": 0.56, + "learning_rate": 8.706000617261465e-06, + "loss": 0.8699, + "step": 23081 + }, + { + "epoch": 0.56, + "learning_rate": 8.705228275358686e-06, + "loss": 0.8885, + "step": 23082 + }, + { + "epoch": 0.56, + "learning_rate": 8.70445594131098e-06, + "loss": 1.0734, + "step": 23083 + }, + { + "epoch": 0.56, + "learning_rate": 8.703683615123037e-06, + "loss": 1.1624, + "step": 23084 + }, + { + "epoch": 0.56, + "learning_rate": 8.702911296799541e-06, + "loss": 0.6737, + "step": 23085 + }, + { + "epoch": 0.56, + "learning_rate": 8.702138986345176e-06, + "loss": 1.215, + "step": 23086 + }, + { + "epoch": 0.56, + "learning_rate": 8.701366683764629e-06, + "loss": 0.9479, + "step": 23087 + }, + { + "epoch": 0.56, + "learning_rate": 8.700594389062586e-06, + "loss": 0.8023, + "step": 23088 + }, + { + "epoch": 0.56, + "learning_rate": 8.699822102243732e-06, + "loss": 1.1951, + "step": 23089 + }, + { + "epoch": 0.56, + "learning_rate": 8.699049823312747e-06, + "loss": 1.0389, + "step": 23090 + }, + { + "epoch": 0.56, + "learning_rate": 8.698277552274323e-06, + "loss": 0.9961, + "step": 23091 + }, + { + "epoch": 0.56, + "learning_rate": 8.697505289133146e-06, + "loss": 1.1107, + "step": 23092 + }, + { + "epoch": 0.56, + "learning_rate": 8.696733033893894e-06, + "loss": 1.1976, + "step": 23093 + }, + { + "epoch": 0.56, + "learning_rate": 8.695960786561258e-06, + "loss": 0.8882, + "step": 23094 + }, + { + "epoch": 0.56, + "learning_rate": 8.69518854713992e-06, + "loss": 0.9354, + "step": 23095 + }, + { + "epoch": 0.56, + "learning_rate": 8.69441631563457e-06, + "loss": 1.2869, + "step": 23096 + }, + { + "epoch": 0.56, + "learning_rate": 8.693644092049884e-06, + "loss": 1.0351, + "step": 23097 + }, + { + "epoch": 0.56, + "learning_rate": 8.692871876390554e-06, + "loss": 1.078, + "step": 23098 + }, + { + "epoch": 0.56, + "learning_rate": 8.692099668661263e-06, + "loss": 1.125, + "step": 23099 + }, + { + "epoch": 0.56, + "learning_rate": 8.691327468866695e-06, + "loss": 1.0396, + "step": 23100 + }, + { + "epoch": 0.56, + "learning_rate": 8.690555277011537e-06, + "loss": 1.1285, + "step": 23101 + }, + { + "epoch": 0.56, + "learning_rate": 8.68978309310047e-06, + "loss": 1.1123, + "step": 23102 + }, + { + "epoch": 0.56, + "learning_rate": 8.689010917138181e-06, + "loss": 1.1101, + "step": 23103 + }, + { + "epoch": 0.56, + "learning_rate": 8.688238749129357e-06, + "loss": 0.8348, + "step": 23104 + }, + { + "epoch": 0.56, + "learning_rate": 8.687466589078678e-06, + "loss": 0.9921, + "step": 23105 + }, + { + "epoch": 0.56, + "learning_rate": 8.68669443699083e-06, + "loss": 1.1168, + "step": 23106 + }, + { + "epoch": 0.56, + "learning_rate": 8.6859222928705e-06, + "loss": 0.9664, + "step": 23107 + }, + { + "epoch": 0.56, + "learning_rate": 8.685150156722369e-06, + "loss": 0.9372, + "step": 23108 + }, + { + "epoch": 0.56, + "learning_rate": 8.684378028551123e-06, + "loss": 0.9218, + "step": 23109 + }, + { + "epoch": 0.56, + "learning_rate": 8.683605908361446e-06, + "loss": 1.1328, + "step": 23110 + }, + { + "epoch": 0.56, + "learning_rate": 8.682833796158022e-06, + "loss": 1.1571, + "step": 23111 + }, + { + "epoch": 0.56, + "learning_rate": 8.682061691945536e-06, + "loss": 1.0333, + "step": 23112 + }, + { + "epoch": 0.56, + "learning_rate": 8.681289595728673e-06, + "loss": 0.9007, + "step": 23113 + }, + { + "epoch": 0.56, + "learning_rate": 8.680517507512112e-06, + "loss": 1.2216, + "step": 23114 + }, + { + "epoch": 0.56, + "learning_rate": 8.679745427300543e-06, + "loss": 0.9829, + "step": 23115 + }, + { + "epoch": 0.56, + "learning_rate": 8.67897335509865e-06, + "loss": 1.1392, + "step": 23116 + }, + { + "epoch": 0.56, + "learning_rate": 8.678201290911113e-06, + "loss": 1.0221, + "step": 23117 + }, + { + "epoch": 0.56, + "learning_rate": 8.677429234742615e-06, + "loss": 1.2391, + "step": 23118 + }, + { + "epoch": 0.56, + "learning_rate": 8.676657186597848e-06, + "loss": 0.882, + "step": 23119 + }, + { + "epoch": 0.56, + "learning_rate": 8.675885146481489e-06, + "loss": 0.9647, + "step": 23120 + }, + { + "epoch": 0.56, + "learning_rate": 8.675113114398223e-06, + "loss": 0.8984, + "step": 23121 + }, + { + "epoch": 0.56, + "learning_rate": 8.674341090352731e-06, + "loss": 1.1645, + "step": 23122 + }, + { + "epoch": 0.56, + "learning_rate": 8.673569074349704e-06, + "loss": 1.2266, + "step": 23123 + }, + { + "epoch": 0.56, + "learning_rate": 8.672797066393822e-06, + "loss": 0.9535, + "step": 23124 + }, + { + "epoch": 0.56, + "learning_rate": 8.672025066489765e-06, + "loss": 1.1989, + "step": 23125 + }, + { + "epoch": 0.56, + "learning_rate": 8.671253074642218e-06, + "loss": 1.0641, + "step": 23126 + }, + { + "epoch": 0.56, + "learning_rate": 8.67048109085587e-06, + "loss": 1.1776, + "step": 23127 + }, + { + "epoch": 0.56, + "learning_rate": 8.6697091151354e-06, + "loss": 1.119, + "step": 23128 + }, + { + "epoch": 0.56, + "learning_rate": 8.66893714748549e-06, + "loss": 1.0117, + "step": 23129 + }, + { + "epoch": 0.56, + "learning_rate": 8.668165187910826e-06, + "loss": 0.8785, + "step": 23130 + }, + { + "epoch": 0.56, + "learning_rate": 8.667393236416091e-06, + "loss": 1.2673, + "step": 23131 + }, + { + "epoch": 0.56, + "learning_rate": 8.66662129300597e-06, + "loss": 1.1056, + "step": 23132 + }, + { + "epoch": 0.56, + "learning_rate": 8.665849357685139e-06, + "loss": 0.9936, + "step": 23133 + }, + { + "epoch": 0.56, + "learning_rate": 8.665077430458288e-06, + "loss": 1.2057, + "step": 23134 + }, + { + "epoch": 0.56, + "learning_rate": 8.6643055113301e-06, + "loss": 0.8301, + "step": 23135 + }, + { + "epoch": 0.56, + "learning_rate": 8.663533600305256e-06, + "loss": 1.1457, + "step": 23136 + }, + { + "epoch": 0.56, + "learning_rate": 8.662761697388437e-06, + "loss": 1.1304, + "step": 23137 + }, + { + "epoch": 0.56, + "learning_rate": 8.66198980258433e-06, + "loss": 1.034, + "step": 23138 + }, + { + "epoch": 0.56, + "learning_rate": 8.661217915897618e-06, + "loss": 1.0536, + "step": 23139 + }, + { + "epoch": 0.56, + "learning_rate": 8.660446037332979e-06, + "loss": 1.0649, + "step": 23140 + }, + { + "epoch": 0.56, + "learning_rate": 8.659674166895098e-06, + "loss": 1.0607, + "step": 23141 + }, + { + "epoch": 0.56, + "learning_rate": 8.658902304588662e-06, + "loss": 1.0831, + "step": 23142 + }, + { + "epoch": 0.56, + "learning_rate": 8.658130450418348e-06, + "loss": 0.9677, + "step": 23143 + }, + { + "epoch": 0.56, + "learning_rate": 8.657358604388843e-06, + "loss": 0.9473, + "step": 23144 + }, + { + "epoch": 0.56, + "learning_rate": 8.656586766504825e-06, + "loss": 1.0899, + "step": 23145 + }, + { + "epoch": 0.56, + "learning_rate": 8.65581493677098e-06, + "loss": 1.1119, + "step": 23146 + }, + { + "epoch": 0.56, + "learning_rate": 8.65504311519199e-06, + "loss": 1.1132, + "step": 23147 + }, + { + "epoch": 0.56, + "learning_rate": 8.654271301772537e-06, + "loss": 1.0243, + "step": 23148 + }, + { + "epoch": 0.56, + "learning_rate": 8.653499496517302e-06, + "loss": 1.1696, + "step": 23149 + }, + { + "epoch": 0.56, + "learning_rate": 8.652727699430971e-06, + "loss": 1.0175, + "step": 23150 + }, + { + "epoch": 0.56, + "learning_rate": 8.651955910518224e-06, + "loss": 1.179, + "step": 23151 + }, + { + "epoch": 0.56, + "learning_rate": 8.651184129783742e-06, + "loss": 0.9551, + "step": 23152 + }, + { + "epoch": 0.56, + "learning_rate": 8.650412357232209e-06, + "loss": 1.1958, + "step": 23153 + }, + { + "epoch": 0.56, + "learning_rate": 8.649640592868306e-06, + "loss": 1.0154, + "step": 23154 + }, + { + "epoch": 0.56, + "learning_rate": 8.64886883669672e-06, + "loss": 1.2346, + "step": 23155 + }, + { + "epoch": 0.56, + "learning_rate": 8.648097088722125e-06, + "loss": 1.0126, + "step": 23156 + }, + { + "epoch": 0.56, + "learning_rate": 8.647325348949207e-06, + "loss": 1.1015, + "step": 23157 + }, + { + "epoch": 0.56, + "learning_rate": 8.64655361738265e-06, + "loss": 1.0344, + "step": 23158 + }, + { + "epoch": 0.56, + "learning_rate": 8.645781894027132e-06, + "loss": 1.2441, + "step": 23159 + }, + { + "epoch": 0.56, + "learning_rate": 8.645010178887338e-06, + "loss": 0.9883, + "step": 23160 + }, + { + "epoch": 0.56, + "learning_rate": 8.644238471967946e-06, + "loss": 0.7897, + "step": 23161 + }, + { + "epoch": 0.56, + "learning_rate": 8.643466773273643e-06, + "loss": 1.1589, + "step": 23162 + }, + { + "epoch": 0.56, + "learning_rate": 8.642695082809108e-06, + "loss": 1.125, + "step": 23163 + }, + { + "epoch": 0.56, + "learning_rate": 8.641923400579022e-06, + "loss": 1.0738, + "step": 23164 + }, + { + "epoch": 0.56, + "learning_rate": 8.641151726588066e-06, + "loss": 0.8481, + "step": 23165 + }, + { + "epoch": 0.56, + "learning_rate": 8.640380060840921e-06, + "loss": 1.054, + "step": 23166 + }, + { + "epoch": 0.56, + "learning_rate": 8.639608403342273e-06, + "loss": 1.0135, + "step": 23167 + }, + { + "epoch": 0.56, + "learning_rate": 8.638836754096802e-06, + "loss": 1.0749, + "step": 23168 + }, + { + "epoch": 0.56, + "learning_rate": 8.638065113109189e-06, + "loss": 1.1888, + "step": 23169 + }, + { + "epoch": 0.56, + "learning_rate": 8.637293480384111e-06, + "loss": 1.0078, + "step": 23170 + }, + { + "epoch": 0.56, + "learning_rate": 8.636521855926256e-06, + "loss": 1.1596, + "step": 23171 + }, + { + "epoch": 0.56, + "learning_rate": 8.635750239740302e-06, + "loss": 1.0268, + "step": 23172 + }, + { + "epoch": 0.56, + "learning_rate": 8.63497863183093e-06, + "loss": 1.2236, + "step": 23173 + }, + { + "epoch": 0.56, + "learning_rate": 8.634207032202819e-06, + "loss": 1.3131, + "step": 23174 + }, + { + "epoch": 0.56, + "learning_rate": 8.633435440860656e-06, + "loss": 1.2422, + "step": 23175 + }, + { + "epoch": 0.56, + "learning_rate": 8.632663857809119e-06, + "loss": 1.3015, + "step": 23176 + }, + { + "epoch": 0.56, + "learning_rate": 8.631892283052887e-06, + "loss": 1.1969, + "step": 23177 + }, + { + "epoch": 0.56, + "learning_rate": 8.631120716596642e-06, + "loss": 1.0648, + "step": 23178 + }, + { + "epoch": 0.56, + "learning_rate": 8.630349158445067e-06, + "loss": 1.2003, + "step": 23179 + }, + { + "epoch": 0.56, + "learning_rate": 8.629577608602842e-06, + "loss": 1.1166, + "step": 23180 + }, + { + "epoch": 0.56, + "learning_rate": 8.628806067074646e-06, + "loss": 1.0494, + "step": 23181 + }, + { + "epoch": 0.56, + "learning_rate": 8.628034533865162e-06, + "loss": 0.9757, + "step": 23182 + }, + { + "epoch": 0.56, + "learning_rate": 8.62726300897907e-06, + "loss": 1.2625, + "step": 23183 + }, + { + "epoch": 0.56, + "learning_rate": 8.62649149242105e-06, + "loss": 1.1794, + "step": 23184 + }, + { + "epoch": 0.56, + "learning_rate": 8.625719984195782e-06, + "loss": 1.1572, + "step": 23185 + }, + { + "epoch": 0.56, + "learning_rate": 8.624948484307948e-06, + "loss": 0.5584, + "step": 23186 + }, + { + "epoch": 0.56, + "learning_rate": 8.624176992762229e-06, + "loss": 0.87, + "step": 23187 + }, + { + "epoch": 0.56, + "learning_rate": 8.623405509563305e-06, + "loss": 0.912, + "step": 23188 + }, + { + "epoch": 0.56, + "learning_rate": 8.622634034715856e-06, + "loss": 1.1213, + "step": 23189 + }, + { + "epoch": 0.56, + "learning_rate": 8.621862568224559e-06, + "loss": 1.1235, + "step": 23190 + }, + { + "epoch": 0.56, + "learning_rate": 8.621091110094102e-06, + "loss": 1.2046, + "step": 23191 + }, + { + "epoch": 0.56, + "learning_rate": 8.620319660329159e-06, + "loss": 1.1629, + "step": 23192 + }, + { + "epoch": 0.56, + "learning_rate": 8.619548218934412e-06, + "loss": 1.2333, + "step": 23193 + }, + { + "epoch": 0.56, + "learning_rate": 8.61877678591454e-06, + "loss": 1.2539, + "step": 23194 + }, + { + "epoch": 0.56, + "learning_rate": 8.618005361274226e-06, + "loss": 0.9981, + "step": 23195 + }, + { + "epoch": 0.56, + "learning_rate": 8.61723394501815e-06, + "loss": 1.1788, + "step": 23196 + }, + { + "epoch": 0.56, + "learning_rate": 8.616462537150989e-06, + "loss": 1.1666, + "step": 23197 + }, + { + "epoch": 0.56, + "learning_rate": 8.615691137677423e-06, + "loss": 1.1221, + "step": 23198 + }, + { + "epoch": 0.56, + "learning_rate": 8.614919746602134e-06, + "loss": 1.3603, + "step": 23199 + }, + { + "epoch": 0.56, + "learning_rate": 8.614148363929802e-06, + "loss": 1.0806, + "step": 23200 + }, + { + "epoch": 0.56, + "learning_rate": 8.613376989665105e-06, + "loss": 1.1416, + "step": 23201 + }, + { + "epoch": 0.56, + "learning_rate": 8.612605623812721e-06, + "loss": 1.235, + "step": 23202 + }, + { + "epoch": 0.56, + "learning_rate": 8.611834266377336e-06, + "loss": 0.8647, + "step": 23203 + }, + { + "epoch": 0.56, + "learning_rate": 8.611062917363627e-06, + "loss": 1.027, + "step": 23204 + }, + { + "epoch": 0.56, + "learning_rate": 8.61029157677627e-06, + "loss": 1.0361, + "step": 23205 + }, + { + "epoch": 0.56, + "learning_rate": 8.609520244619948e-06, + "loss": 1.2886, + "step": 23206 + }, + { + "epoch": 0.56, + "learning_rate": 8.608748920899339e-06, + "loss": 1.1329, + "step": 23207 + }, + { + "epoch": 0.56, + "learning_rate": 8.607977605619125e-06, + "loss": 1.041, + "step": 23208 + }, + { + "epoch": 0.56, + "learning_rate": 8.607206298783982e-06, + "loss": 1.1033, + "step": 23209 + }, + { + "epoch": 0.56, + "learning_rate": 8.60643500039859e-06, + "loss": 1.1933, + "step": 23210 + }, + { + "epoch": 0.56, + "learning_rate": 8.60566371046763e-06, + "loss": 0.9853, + "step": 23211 + }, + { + "epoch": 0.56, + "learning_rate": 8.604892428995782e-06, + "loss": 1.0235, + "step": 23212 + }, + { + "epoch": 0.56, + "learning_rate": 8.60412115598772e-06, + "loss": 1.0025, + "step": 23213 + }, + { + "epoch": 0.56, + "learning_rate": 8.603349891448128e-06, + "loss": 0.9996, + "step": 23214 + }, + { + "epoch": 0.56, + "learning_rate": 8.602578635381685e-06, + "loss": 1.2413, + "step": 23215 + }, + { + "epoch": 0.56, + "learning_rate": 8.601807387793068e-06, + "loss": 0.9517, + "step": 23216 + }, + { + "epoch": 0.56, + "learning_rate": 8.601036148686954e-06, + "loss": 1.1079, + "step": 23217 + }, + { + "epoch": 0.56, + "learning_rate": 8.600264918068028e-06, + "loss": 0.9862, + "step": 23218 + }, + { + "epoch": 0.56, + "learning_rate": 8.599493695940964e-06, + "loss": 1.1935, + "step": 23219 + }, + { + "epoch": 0.56, + "learning_rate": 8.598722482310445e-06, + "loss": 1.132, + "step": 23220 + }, + { + "epoch": 0.56, + "learning_rate": 8.597951277181143e-06, + "loss": 0.8953, + "step": 23221 + }, + { + "epoch": 0.56, + "learning_rate": 8.597180080557743e-06, + "loss": 1.1374, + "step": 23222 + }, + { + "epoch": 0.56, + "learning_rate": 8.596408892444921e-06, + "loss": 1.1485, + "step": 23223 + }, + { + "epoch": 0.56, + "learning_rate": 8.595637712847357e-06, + "loss": 0.9371, + "step": 23224 + }, + { + "epoch": 0.56, + "learning_rate": 8.594866541769727e-06, + "loss": 0.8874, + "step": 23225 + }, + { + "epoch": 0.56, + "learning_rate": 8.594095379216712e-06, + "loss": 1.2061, + "step": 23226 + }, + { + "epoch": 0.56, + "learning_rate": 8.593324225192991e-06, + "loss": 1.0479, + "step": 23227 + }, + { + "epoch": 0.56, + "learning_rate": 8.59255307970324e-06, + "loss": 0.8238, + "step": 23228 + }, + { + "epoch": 0.56, + "learning_rate": 8.591781942752137e-06, + "loss": 0.7339, + "step": 23229 + }, + { + "epoch": 0.56, + "learning_rate": 8.591010814344364e-06, + "loss": 1.1719, + "step": 23230 + }, + { + "epoch": 0.56, + "learning_rate": 8.590239694484597e-06, + "loss": 1.0074, + "step": 23231 + }, + { + "epoch": 0.56, + "learning_rate": 8.589468583177514e-06, + "loss": 1.34, + "step": 23232 + }, + { + "epoch": 0.56, + "learning_rate": 8.58869748042779e-06, + "loss": 1.066, + "step": 23233 + }, + { + "epoch": 0.56, + "learning_rate": 8.58792638624011e-06, + "loss": 0.7184, + "step": 23234 + }, + { + "epoch": 0.56, + "learning_rate": 8.587155300619147e-06, + "loss": 0.8717, + "step": 23235 + }, + { + "epoch": 0.56, + "learning_rate": 8.586384223569582e-06, + "loss": 0.866, + "step": 23236 + }, + { + "epoch": 0.56, + "learning_rate": 8.585613155096088e-06, + "loss": 1.0497, + "step": 23237 + }, + { + "epoch": 0.56, + "learning_rate": 8.58484209520335e-06, + "loss": 1.2864, + "step": 23238 + }, + { + "epoch": 0.56, + "learning_rate": 8.584071043896041e-06, + "loss": 1.1029, + "step": 23239 + }, + { + "epoch": 0.56, + "learning_rate": 8.58330000117884e-06, + "loss": 1.34, + "step": 23240 + }, + { + "epoch": 0.56, + "learning_rate": 8.582528967056423e-06, + "loss": 1.0387, + "step": 23241 + }, + { + "epoch": 0.56, + "learning_rate": 8.581757941533471e-06, + "loss": 1.0691, + "step": 23242 + }, + { + "epoch": 0.56, + "learning_rate": 8.58098692461466e-06, + "loss": 1.1701, + "step": 23243 + }, + { + "epoch": 0.56, + "learning_rate": 8.580215916304668e-06, + "loss": 1.2272, + "step": 23244 + }, + { + "epoch": 0.56, + "learning_rate": 8.57944491660817e-06, + "loss": 1.0525, + "step": 23245 + }, + { + "epoch": 0.56, + "learning_rate": 8.578673925529848e-06, + "loss": 1.0689, + "step": 23246 + }, + { + "epoch": 0.56, + "learning_rate": 8.577902943074376e-06, + "loss": 1.0491, + "step": 23247 + }, + { + "epoch": 0.56, + "learning_rate": 8.577131969246433e-06, + "loss": 0.9705, + "step": 23248 + }, + { + "epoch": 0.56, + "learning_rate": 8.576361004050693e-06, + "loss": 0.9564, + "step": 23249 + }, + { + "epoch": 0.56, + "learning_rate": 8.575590047491839e-06, + "loss": 1.1143, + "step": 23250 + }, + { + "epoch": 0.56, + "learning_rate": 8.574819099574544e-06, + "loss": 1.0384, + "step": 23251 + }, + { + "epoch": 0.56, + "learning_rate": 8.574048160303487e-06, + "loss": 1.2815, + "step": 23252 + }, + { + "epoch": 0.56, + "learning_rate": 8.573277229683343e-06, + "loss": 0.7892, + "step": 23253 + }, + { + "epoch": 0.56, + "learning_rate": 8.57250630771879e-06, + "loss": 1.2444, + "step": 23254 + }, + { + "epoch": 0.56, + "learning_rate": 8.571735394414507e-06, + "loss": 1.1976, + "step": 23255 + }, + { + "epoch": 0.56, + "learning_rate": 8.57096448977517e-06, + "loss": 1.0313, + "step": 23256 + }, + { + "epoch": 0.56, + "learning_rate": 8.570193593805452e-06, + "loss": 1.1709, + "step": 23257 + }, + { + "epoch": 0.56, + "learning_rate": 8.569422706510037e-06, + "loss": 1.0653, + "step": 23258 + }, + { + "epoch": 0.56, + "learning_rate": 8.568651827893597e-06, + "loss": 1.1131, + "step": 23259 + }, + { + "epoch": 0.56, + "learning_rate": 8.567880957960807e-06, + "loss": 1.1473, + "step": 23260 + }, + { + "epoch": 0.56, + "learning_rate": 8.567110096716348e-06, + "loss": 1.1111, + "step": 23261 + }, + { + "epoch": 0.56, + "learning_rate": 8.566339244164897e-06, + "loss": 1.248, + "step": 23262 + }, + { + "epoch": 0.56, + "learning_rate": 8.565568400311129e-06, + "loss": 1.2221, + "step": 23263 + }, + { + "epoch": 0.56, + "learning_rate": 8.564797565159715e-06, + "loss": 1.1115, + "step": 23264 + }, + { + "epoch": 0.56, + "learning_rate": 8.56402673871534e-06, + "loss": 1.1647, + "step": 23265 + }, + { + "epoch": 0.56, + "learning_rate": 8.563255920982678e-06, + "loss": 0.975, + "step": 23266 + }, + { + "epoch": 0.56, + "learning_rate": 8.562485111966402e-06, + "loss": 1.0316, + "step": 23267 + }, + { + "epoch": 0.56, + "learning_rate": 8.561714311671192e-06, + "loss": 0.9654, + "step": 23268 + }, + { + "epoch": 0.56, + "learning_rate": 8.560943520101724e-06, + "loss": 1.039, + "step": 23269 + }, + { + "epoch": 0.56, + "learning_rate": 8.560172737262671e-06, + "loss": 0.9113, + "step": 23270 + }, + { + "epoch": 0.56, + "learning_rate": 8.559401963158714e-06, + "loss": 0.9468, + "step": 23271 + }, + { + "epoch": 0.56, + "learning_rate": 8.558631197794522e-06, + "loss": 1.1716, + "step": 23272 + }, + { + "epoch": 0.56, + "learning_rate": 8.55786044117478e-06, + "loss": 1.1786, + "step": 23273 + }, + { + "epoch": 0.56, + "learning_rate": 8.557089693304158e-06, + "loss": 1.0638, + "step": 23274 + }, + { + "epoch": 0.56, + "learning_rate": 8.556318954187334e-06, + "loss": 0.9855, + "step": 23275 + }, + { + "epoch": 0.56, + "learning_rate": 8.55554822382898e-06, + "loss": 1.2062, + "step": 23276 + }, + { + "epoch": 0.56, + "learning_rate": 8.554777502233778e-06, + "loss": 1.0909, + "step": 23277 + }, + { + "epoch": 0.56, + "learning_rate": 8.554006789406402e-06, + "loss": 1.254, + "step": 23278 + }, + { + "epoch": 0.56, + "learning_rate": 8.553236085351525e-06, + "loss": 0.9098, + "step": 23279 + }, + { + "epoch": 0.56, + "learning_rate": 8.552465390073825e-06, + "loss": 0.992, + "step": 23280 + }, + { + "epoch": 0.56, + "learning_rate": 8.551694703577977e-06, + "loss": 1.3067, + "step": 23281 + }, + { + "epoch": 0.56, + "learning_rate": 8.550924025868656e-06, + "loss": 1.138, + "step": 23282 + }, + { + "epoch": 0.56, + "learning_rate": 8.55015335695054e-06, + "loss": 1.3299, + "step": 23283 + }, + { + "epoch": 0.56, + "learning_rate": 8.549382696828299e-06, + "loss": 0.882, + "step": 23284 + }, + { + "epoch": 0.56, + "learning_rate": 8.548612045506614e-06, + "loss": 1.0959, + "step": 23285 + }, + { + "epoch": 0.56, + "learning_rate": 8.547841402990158e-06, + "loss": 1.1394, + "step": 23286 + }, + { + "epoch": 0.56, + "learning_rate": 8.547070769283608e-06, + "loss": 1.1392, + "step": 23287 + }, + { + "epoch": 0.56, + "learning_rate": 8.546300144391634e-06, + "loss": 0.9021, + "step": 23288 + }, + { + "epoch": 0.56, + "learning_rate": 8.54552952831892e-06, + "loss": 1.0981, + "step": 23289 + }, + { + "epoch": 0.56, + "learning_rate": 8.544758921070134e-06, + "loss": 1.0582, + "step": 23290 + }, + { + "epoch": 0.56, + "learning_rate": 8.543988322649954e-06, + "loss": 1.0839, + "step": 23291 + }, + { + "epoch": 0.56, + "learning_rate": 8.543217733063053e-06, + "loss": 1.0076, + "step": 23292 + }, + { + "epoch": 0.56, + "learning_rate": 8.542447152314108e-06, + "loss": 1.0212, + "step": 23293 + }, + { + "epoch": 0.56, + "learning_rate": 8.541676580407793e-06, + "loss": 1.05, + "step": 23294 + }, + { + "epoch": 0.56, + "learning_rate": 8.540906017348786e-06, + "loss": 0.8156, + "step": 23295 + }, + { + "epoch": 0.56, + "learning_rate": 8.540135463141755e-06, + "loss": 1.0423, + "step": 23296 + }, + { + "epoch": 0.56, + "learning_rate": 8.53936491779138e-06, + "loss": 1.2777, + "step": 23297 + }, + { + "epoch": 0.56, + "learning_rate": 8.538594381302337e-06, + "loss": 1.1038, + "step": 23298 + }, + { + "epoch": 0.56, + "learning_rate": 8.537823853679296e-06, + "loss": 1.0339, + "step": 23299 + }, + { + "epoch": 0.56, + "learning_rate": 8.537053334926932e-06, + "loss": 0.9757, + "step": 23300 + }, + { + "epoch": 0.56, + "learning_rate": 8.536282825049922e-06, + "loss": 1.146, + "step": 23301 + }, + { + "epoch": 0.56, + "learning_rate": 8.535512324052943e-06, + "loss": 0.9796, + "step": 23302 + }, + { + "epoch": 0.56, + "learning_rate": 8.534741831940664e-06, + "loss": 1.2694, + "step": 23303 + }, + { + "epoch": 0.56, + "learning_rate": 8.53397134871776e-06, + "loss": 0.8145, + "step": 23304 + }, + { + "epoch": 0.56, + "learning_rate": 8.53320087438891e-06, + "loss": 1.0129, + "step": 23305 + }, + { + "epoch": 0.56, + "learning_rate": 8.532430408958785e-06, + "loss": 1.2597, + "step": 23306 + }, + { + "epoch": 0.56, + "learning_rate": 8.531659952432057e-06, + "loss": 1.1674, + "step": 23307 + }, + { + "epoch": 0.56, + "learning_rate": 8.530889504813405e-06, + "loss": 1.0068, + "step": 23308 + }, + { + "epoch": 0.56, + "learning_rate": 8.530119066107501e-06, + "loss": 1.0563, + "step": 23309 + }, + { + "epoch": 0.56, + "learning_rate": 8.529348636319017e-06, + "loss": 0.9315, + "step": 23310 + }, + { + "epoch": 0.56, + "learning_rate": 8.528578215452628e-06, + "loss": 1.2689, + "step": 23311 + }, + { + "epoch": 0.56, + "learning_rate": 8.527807803513012e-06, + "loss": 1.2009, + "step": 23312 + }, + { + "epoch": 0.56, + "learning_rate": 8.52703740050484e-06, + "loss": 1.1324, + "step": 23313 + }, + { + "epoch": 0.56, + "learning_rate": 8.526267006432785e-06, + "loss": 1.0944, + "step": 23314 + }, + { + "epoch": 0.56, + "learning_rate": 8.525496621301519e-06, + "loss": 1.2754, + "step": 23315 + }, + { + "epoch": 0.56, + "learning_rate": 8.52472624511572e-06, + "loss": 1.3337, + "step": 23316 + }, + { + "epoch": 0.56, + "learning_rate": 8.52395587788006e-06, + "loss": 0.9821, + "step": 23317 + }, + { + "epoch": 0.56, + "learning_rate": 8.523185519599211e-06, + "loss": 1.1276, + "step": 23318 + }, + { + "epoch": 0.56, + "learning_rate": 8.522415170277849e-06, + "loss": 1.1255, + "step": 23319 + }, + { + "epoch": 0.56, + "learning_rate": 8.521644829920647e-06, + "loss": 1.1152, + "step": 23320 + }, + { + "epoch": 0.56, + "learning_rate": 8.520874498532279e-06, + "loss": 1.0609, + "step": 23321 + }, + { + "epoch": 0.56, + "learning_rate": 8.520104176117416e-06, + "loss": 0.9497, + "step": 23322 + }, + { + "epoch": 0.56, + "learning_rate": 8.519333862680733e-06, + "loss": 0.9165, + "step": 23323 + }, + { + "epoch": 0.56, + "learning_rate": 8.518563558226903e-06, + "loss": 1.1217, + "step": 23324 + }, + { + "epoch": 0.56, + "learning_rate": 8.5177932627606e-06, + "loss": 0.9095, + "step": 23325 + }, + { + "epoch": 0.56, + "learning_rate": 8.517022976286497e-06, + "loss": 0.8143, + "step": 23326 + }, + { + "epoch": 0.56, + "learning_rate": 8.516252698809267e-06, + "loss": 1.1117, + "step": 23327 + }, + { + "epoch": 0.56, + "learning_rate": 8.515482430333582e-06, + "loss": 0.9721, + "step": 23328 + }, + { + "epoch": 0.56, + "learning_rate": 8.514712170864117e-06, + "loss": 1.0744, + "step": 23329 + }, + { + "epoch": 0.56, + "learning_rate": 8.513941920405544e-06, + "loss": 1.2504, + "step": 23330 + }, + { + "epoch": 0.56, + "learning_rate": 8.513171678962531e-06, + "loss": 1.0187, + "step": 23331 + }, + { + "epoch": 0.56, + "learning_rate": 8.512401446539759e-06, + "loss": 1.0564, + "step": 23332 + }, + { + "epoch": 0.56, + "learning_rate": 8.511631223141897e-06, + "loss": 0.9602, + "step": 23333 + }, + { + "epoch": 0.56, + "learning_rate": 8.510861008773618e-06, + "loss": 0.9235, + "step": 23334 + }, + { + "epoch": 0.56, + "learning_rate": 8.510090803439598e-06, + "loss": 1.1927, + "step": 23335 + }, + { + "epoch": 0.56, + "learning_rate": 8.509320607144505e-06, + "loss": 1.0533, + "step": 23336 + }, + { + "epoch": 0.56, + "learning_rate": 8.508550419893012e-06, + "loss": 1.1085, + "step": 23337 + }, + { + "epoch": 0.56, + "learning_rate": 8.507780241689795e-06, + "loss": 1.0868, + "step": 23338 + }, + { + "epoch": 0.56, + "learning_rate": 8.507010072539525e-06, + "loss": 1.0991, + "step": 23339 + }, + { + "epoch": 0.56, + "learning_rate": 8.506239912446869e-06, + "loss": 1.1487, + "step": 23340 + }, + { + "epoch": 0.56, + "learning_rate": 8.505469761416509e-06, + "loss": 1.0056, + "step": 23341 + }, + { + "epoch": 0.56, + "learning_rate": 8.504699619453111e-06, + "loss": 0.9922, + "step": 23342 + }, + { + "epoch": 0.56, + "learning_rate": 8.503929486561349e-06, + "loss": 1.2196, + "step": 23343 + }, + { + "epoch": 0.56, + "learning_rate": 8.503159362745891e-06, + "loss": 0.8823, + "step": 23344 + }, + { + "epoch": 0.56, + "learning_rate": 8.502389248011418e-06, + "loss": 1.2367, + "step": 23345 + }, + { + "epoch": 0.56, + "learning_rate": 8.501619142362596e-06, + "loss": 1.0064, + "step": 23346 + }, + { + "epoch": 0.56, + "learning_rate": 8.500849045804098e-06, + "loss": 1.0628, + "step": 23347 + }, + { + "epoch": 0.56, + "learning_rate": 8.500078958340595e-06, + "loss": 0.9755, + "step": 23348 + }, + { + "epoch": 0.56, + "learning_rate": 8.49930887997676e-06, + "loss": 1.3477, + "step": 23349 + }, + { + "epoch": 0.56, + "learning_rate": 8.498538810717267e-06, + "loss": 1.063, + "step": 23350 + }, + { + "epoch": 0.56, + "learning_rate": 8.497768750566787e-06, + "loss": 1.1779, + "step": 23351 + }, + { + "epoch": 0.56, + "learning_rate": 8.496998699529986e-06, + "loss": 1.0105, + "step": 23352 + }, + { + "epoch": 0.56, + "learning_rate": 8.496228657611542e-06, + "loss": 1.0684, + "step": 23353 + }, + { + "epoch": 0.56, + "learning_rate": 8.495458624816127e-06, + "loss": 0.8534, + "step": 23354 + }, + { + "epoch": 0.56, + "learning_rate": 8.494688601148411e-06, + "loss": 1.0445, + "step": 23355 + }, + { + "epoch": 0.56, + "learning_rate": 8.493918586613062e-06, + "loss": 0.7319, + "step": 23356 + }, + { + "epoch": 0.56, + "learning_rate": 8.493148581214757e-06, + "loss": 0.9577, + "step": 23357 + }, + { + "epoch": 0.56, + "learning_rate": 8.492378584958165e-06, + "loss": 1.0912, + "step": 23358 + }, + { + "epoch": 0.56, + "learning_rate": 8.491608597847959e-06, + "loss": 1.3043, + "step": 23359 + }, + { + "epoch": 0.56, + "learning_rate": 8.490838619888803e-06, + "loss": 1.0979, + "step": 23360 + }, + { + "epoch": 0.56, + "learning_rate": 8.49006865108538e-06, + "loss": 0.8867, + "step": 23361 + }, + { + "epoch": 0.56, + "learning_rate": 8.489298691442354e-06, + "loss": 1.1771, + "step": 23362 + }, + { + "epoch": 0.56, + "learning_rate": 8.488528740964397e-06, + "loss": 1.0353, + "step": 23363 + }, + { + "epoch": 0.56, + "learning_rate": 8.487758799656179e-06, + "loss": 1.2486, + "step": 23364 + }, + { + "epoch": 0.56, + "learning_rate": 8.486988867522375e-06, + "loss": 0.9097, + "step": 23365 + }, + { + "epoch": 0.56, + "learning_rate": 8.486218944567652e-06, + "loss": 1.1757, + "step": 23366 + }, + { + "epoch": 0.56, + "learning_rate": 8.485449030796684e-06, + "loss": 0.9548, + "step": 23367 + }, + { + "epoch": 0.56, + "learning_rate": 8.48467912621414e-06, + "loss": 1.2563, + "step": 23368 + }, + { + "epoch": 0.56, + "learning_rate": 8.48390923082469e-06, + "loss": 0.9994, + "step": 23369 + }, + { + "epoch": 0.56, + "learning_rate": 8.483139344633009e-06, + "loss": 1.0581, + "step": 23370 + }, + { + "epoch": 0.56, + "learning_rate": 8.482369467643763e-06, + "loss": 0.9583, + "step": 23371 + }, + { + "epoch": 0.56, + "learning_rate": 8.481599599861622e-06, + "loss": 0.9612, + "step": 23372 + }, + { + "epoch": 0.56, + "learning_rate": 8.480829741291263e-06, + "loss": 1.1352, + "step": 23373 + }, + { + "epoch": 0.56, + "learning_rate": 8.480059891937352e-06, + "loss": 1.0318, + "step": 23374 + }, + { + "epoch": 0.56, + "learning_rate": 8.47929005180456e-06, + "loss": 1.3323, + "step": 23375 + }, + { + "epoch": 0.56, + "learning_rate": 8.478520220897555e-06, + "loss": 1.074, + "step": 23376 + }, + { + "epoch": 0.56, + "learning_rate": 8.477750399221013e-06, + "loss": 1.3073, + "step": 23377 + }, + { + "epoch": 0.56, + "learning_rate": 8.476980586779601e-06, + "loss": 1.2002, + "step": 23378 + }, + { + "epoch": 0.56, + "learning_rate": 8.47621078357799e-06, + "loss": 1.2186, + "step": 23379 + }, + { + "epoch": 0.56, + "learning_rate": 8.475440989620845e-06, + "loss": 0.929, + "step": 23380 + }, + { + "epoch": 0.56, + "learning_rate": 8.474671204912846e-06, + "loss": 1.2132, + "step": 23381 + }, + { + "epoch": 0.56, + "learning_rate": 8.473901429458657e-06, + "loss": 1.1561, + "step": 23382 + }, + { + "epoch": 0.56, + "learning_rate": 8.473131663262947e-06, + "loss": 1.1647, + "step": 23383 + }, + { + "epoch": 0.56, + "learning_rate": 8.47236190633039e-06, + "loss": 0.9531, + "step": 23384 + }, + { + "epoch": 0.56, + "learning_rate": 8.471592158665653e-06, + "loss": 1.4808, + "step": 23385 + }, + { + "epoch": 0.56, + "learning_rate": 8.470822420273409e-06, + "loss": 1.0943, + "step": 23386 + }, + { + "epoch": 0.56, + "learning_rate": 8.470052691158323e-06, + "loss": 0.7705, + "step": 23387 + }, + { + "epoch": 0.56, + "learning_rate": 8.469282971325067e-06, + "loss": 1.0185, + "step": 23388 + }, + { + "epoch": 0.56, + "learning_rate": 8.468513260778315e-06, + "loss": 0.9955, + "step": 23389 + }, + { + "epoch": 0.56, + "learning_rate": 8.46774355952273e-06, + "loss": 1.0651, + "step": 23390 + }, + { + "epoch": 0.56, + "learning_rate": 8.466973867562984e-06, + "loss": 1.0452, + "step": 23391 + }, + { + "epoch": 0.56, + "learning_rate": 8.46620418490375e-06, + "loss": 1.2813, + "step": 23392 + }, + { + "epoch": 0.56, + "learning_rate": 8.465434511549692e-06, + "loss": 1.0081, + "step": 23393 + }, + { + "epoch": 0.56, + "learning_rate": 8.464664847505482e-06, + "loss": 0.9636, + "step": 23394 + }, + { + "epoch": 0.56, + "learning_rate": 8.463895192775788e-06, + "loss": 1.222, + "step": 23395 + }, + { + "epoch": 0.56, + "learning_rate": 8.463125547365283e-06, + "loss": 1.2513, + "step": 23396 + }, + { + "epoch": 0.56, + "learning_rate": 8.462355911278634e-06, + "loss": 1.0338, + "step": 23397 + }, + { + "epoch": 0.56, + "learning_rate": 8.461586284520509e-06, + "loss": 1.149, + "step": 23398 + }, + { + "epoch": 0.56, + "learning_rate": 8.460816667095576e-06, + "loss": 1.1102, + "step": 23399 + }, + { + "epoch": 0.56, + "learning_rate": 8.460047059008508e-06, + "loss": 1.3313, + "step": 23400 + }, + { + "epoch": 0.56, + "learning_rate": 8.459277460263974e-06, + "loss": 1.2758, + "step": 23401 + }, + { + "epoch": 0.56, + "learning_rate": 8.45850787086664e-06, + "loss": 1.1336, + "step": 23402 + }, + { + "epoch": 0.56, + "learning_rate": 8.457738290821174e-06, + "loss": 1.0688, + "step": 23403 + }, + { + "epoch": 0.56, + "learning_rate": 8.45696872013225e-06, + "loss": 1.1259, + "step": 23404 + }, + { + "epoch": 0.56, + "learning_rate": 8.456199158804533e-06, + "loss": 1.0585, + "step": 23405 + }, + { + "epoch": 0.56, + "learning_rate": 8.455429606842694e-06, + "loss": 0.953, + "step": 23406 + }, + { + "epoch": 0.56, + "learning_rate": 8.454660064251396e-06, + "loss": 1.2878, + "step": 23407 + }, + { + "epoch": 0.56, + "learning_rate": 8.453890531035315e-06, + "loss": 1.2132, + "step": 23408 + }, + { + "epoch": 0.56, + "learning_rate": 8.453121007199117e-06, + "loss": 1.0816, + "step": 23409 + }, + { + "epoch": 0.56, + "learning_rate": 8.45235149274747e-06, + "loss": 1.0286, + "step": 23410 + }, + { + "epoch": 0.56, + "learning_rate": 8.451581987685038e-06, + "loss": 1.28, + "step": 23411 + }, + { + "epoch": 0.56, + "learning_rate": 8.450812492016499e-06, + "loss": 0.8262, + "step": 23412 + }, + { + "epoch": 0.56, + "learning_rate": 8.450043005746515e-06, + "loss": 1.0268, + "step": 23413 + }, + { + "epoch": 0.56, + "learning_rate": 8.449273528879756e-06, + "loss": 1.0684, + "step": 23414 + }, + { + "epoch": 0.56, + "learning_rate": 8.448504061420888e-06, + "loss": 1.2125, + "step": 23415 + }, + { + "epoch": 0.56, + "learning_rate": 8.447734603374582e-06, + "loss": 0.8598, + "step": 23416 + }, + { + "epoch": 0.56, + "learning_rate": 8.446965154745506e-06, + "loss": 0.9321, + "step": 23417 + }, + { + "epoch": 0.56, + "learning_rate": 8.446195715538326e-06, + "loss": 1.1581, + "step": 23418 + }, + { + "epoch": 0.56, + "learning_rate": 8.44542628575771e-06, + "loss": 1.4377, + "step": 23419 + }, + { + "epoch": 0.56, + "learning_rate": 8.444656865408328e-06, + "loss": 0.7187, + "step": 23420 + }, + { + "epoch": 0.56, + "learning_rate": 8.443887454494849e-06, + "loss": 1.0636, + "step": 23421 + }, + { + "epoch": 0.56, + "learning_rate": 8.443118053021936e-06, + "loss": 1.3148, + "step": 23422 + }, + { + "epoch": 0.56, + "learning_rate": 8.442348660994258e-06, + "loss": 1.2375, + "step": 23423 + }, + { + "epoch": 0.56, + "learning_rate": 8.441579278416487e-06, + "loss": 1.0638, + "step": 23424 + }, + { + "epoch": 0.56, + "learning_rate": 8.44080990529329e-06, + "loss": 1.3136, + "step": 23425 + }, + { + "epoch": 0.56, + "learning_rate": 8.44004054162933e-06, + "loss": 1.1222, + "step": 23426 + }, + { + "epoch": 0.56, + "learning_rate": 8.439271187429275e-06, + "loss": 0.9307, + "step": 23427 + }, + { + "epoch": 0.56, + "learning_rate": 8.438501842697799e-06, + "loss": 1.0897, + "step": 23428 + }, + { + "epoch": 0.56, + "learning_rate": 8.437732507439563e-06, + "loss": 0.9605, + "step": 23429 + }, + { + "epoch": 0.56, + "learning_rate": 8.436963181659234e-06, + "loss": 1.1301, + "step": 23430 + }, + { + "epoch": 0.56, + "learning_rate": 8.436193865361485e-06, + "loss": 1.1919, + "step": 23431 + }, + { + "epoch": 0.56, + "learning_rate": 8.435424558550979e-06, + "loss": 0.9379, + "step": 23432 + }, + { + "epoch": 0.56, + "learning_rate": 8.434655261232385e-06, + "loss": 1.1816, + "step": 23433 + }, + { + "epoch": 0.56, + "learning_rate": 8.433885973410368e-06, + "loss": 1.1436, + "step": 23434 + }, + { + "epoch": 0.56, + "learning_rate": 8.433116695089596e-06, + "loss": 1.0602, + "step": 23435 + }, + { + "epoch": 0.56, + "learning_rate": 8.432347426274739e-06, + "loss": 0.9853, + "step": 23436 + }, + { + "epoch": 0.56, + "learning_rate": 8.43157816697046e-06, + "loss": 1.0926, + "step": 23437 + }, + { + "epoch": 0.56, + "learning_rate": 8.430808917181427e-06, + "loss": 0.9499, + "step": 23438 + }, + { + "epoch": 0.56, + "learning_rate": 8.430039676912308e-06, + "loss": 1.1924, + "step": 23439 + }, + { + "epoch": 0.56, + "learning_rate": 8.429270446167768e-06, + "loss": 1.2442, + "step": 23440 + }, + { + "epoch": 0.56, + "learning_rate": 8.428501224952476e-06, + "loss": 1.2345, + "step": 23441 + }, + { + "epoch": 0.56, + "learning_rate": 8.427732013271095e-06, + "loss": 0.8567, + "step": 23442 + }, + { + "epoch": 0.56, + "learning_rate": 8.426962811128297e-06, + "loss": 1.0713, + "step": 23443 + }, + { + "epoch": 0.56, + "learning_rate": 8.426193618528744e-06, + "loss": 0.9404, + "step": 23444 + }, + { + "epoch": 0.56, + "learning_rate": 8.425424435477105e-06, + "loss": 1.019, + "step": 23445 + }, + { + "epoch": 0.56, + "learning_rate": 8.424655261978044e-06, + "loss": 0.9899, + "step": 23446 + }, + { + "epoch": 0.56, + "learning_rate": 8.42388609803623e-06, + "loss": 1.2196, + "step": 23447 + }, + { + "epoch": 0.56, + "learning_rate": 8.42311694365633e-06, + "loss": 0.8695, + "step": 23448 + }, + { + "epoch": 0.56, + "learning_rate": 8.422347798843008e-06, + "loss": 0.7986, + "step": 23449 + }, + { + "epoch": 0.56, + "learning_rate": 8.421578663600926e-06, + "loss": 0.9767, + "step": 23450 + }, + { + "epoch": 0.56, + "learning_rate": 8.42080953793476e-06, + "loss": 1.0329, + "step": 23451 + }, + { + "epoch": 0.56, + "learning_rate": 8.420040421849172e-06, + "loss": 1.1993, + "step": 23452 + }, + { + "epoch": 0.56, + "learning_rate": 8.419271315348825e-06, + "loss": 1.0838, + "step": 23453 + }, + { + "epoch": 0.56, + "learning_rate": 8.418502218438385e-06, + "loss": 1.1158, + "step": 23454 + }, + { + "epoch": 0.56, + "learning_rate": 8.417733131122522e-06, + "loss": 0.9856, + "step": 23455 + }, + { + "epoch": 0.56, + "learning_rate": 8.416964053405902e-06, + "loss": 0.9904, + "step": 23456 + }, + { + "epoch": 0.56, + "learning_rate": 8.416194985293186e-06, + "loss": 1.0743, + "step": 23457 + }, + { + "epoch": 0.56, + "learning_rate": 8.415425926789042e-06, + "loss": 0.9839, + "step": 23458 + }, + { + "epoch": 0.56, + "learning_rate": 8.414656877898139e-06, + "loss": 1.0796, + "step": 23459 + }, + { + "epoch": 0.56, + "learning_rate": 8.41388783862514e-06, + "loss": 0.9486, + "step": 23460 + }, + { + "epoch": 0.56, + "learning_rate": 8.413118808974707e-06, + "loss": 1.0544, + "step": 23461 + }, + { + "epoch": 0.56, + "learning_rate": 8.41234978895151e-06, + "loss": 1.0886, + "step": 23462 + }, + { + "epoch": 0.56, + "learning_rate": 8.411580778560214e-06, + "loss": 1.1664, + "step": 23463 + }, + { + "epoch": 0.56, + "learning_rate": 8.410811777805484e-06, + "loss": 0.976, + "step": 23464 + }, + { + "epoch": 0.56, + "learning_rate": 8.410042786691986e-06, + "loss": 1.0902, + "step": 23465 + }, + { + "epoch": 0.56, + "learning_rate": 8.409273805224382e-06, + "loss": 1.0604, + "step": 23466 + }, + { + "epoch": 0.56, + "learning_rate": 8.408504833407342e-06, + "loss": 0.8208, + "step": 23467 + }, + { + "epoch": 0.56, + "learning_rate": 8.407735871245528e-06, + "loss": 1.0012, + "step": 23468 + }, + { + "epoch": 0.56, + "learning_rate": 8.406966918743607e-06, + "loss": 1.1217, + "step": 23469 + }, + { + "epoch": 0.56, + "learning_rate": 8.40619797590624e-06, + "loss": 0.9373, + "step": 23470 + }, + { + "epoch": 0.56, + "learning_rate": 8.405429042738096e-06, + "loss": 1.2582, + "step": 23471 + }, + { + "epoch": 0.56, + "learning_rate": 8.404660119243841e-06, + "loss": 0.8665, + "step": 23472 + }, + { + "epoch": 0.56, + "learning_rate": 8.403891205428137e-06, + "loss": 0.9536, + "step": 23473 + }, + { + "epoch": 0.56, + "learning_rate": 8.403122301295648e-06, + "loss": 1.1262, + "step": 23474 + }, + { + "epoch": 0.56, + "learning_rate": 8.402353406851042e-06, + "loss": 1.2242, + "step": 23475 + }, + { + "epoch": 0.56, + "learning_rate": 8.401584522098981e-06, + "loss": 1.0985, + "step": 23476 + }, + { + "epoch": 0.56, + "learning_rate": 8.400815647044132e-06, + "loss": 1.297, + "step": 23477 + }, + { + "epoch": 0.56, + "learning_rate": 8.400046781691159e-06, + "loss": 1.2233, + "step": 23478 + }, + { + "epoch": 0.56, + "learning_rate": 8.399277926044724e-06, + "loss": 1.0735, + "step": 23479 + }, + { + "epoch": 0.56, + "learning_rate": 8.398509080109495e-06, + "loss": 1.035, + "step": 23480 + }, + { + "epoch": 0.56, + "learning_rate": 8.39774024389013e-06, + "loss": 1.0993, + "step": 23481 + }, + { + "epoch": 0.56, + "learning_rate": 8.396971417391304e-06, + "loss": 1.1398, + "step": 23482 + }, + { + "epoch": 0.56, + "learning_rate": 8.396202600617673e-06, + "loss": 0.8277, + "step": 23483 + }, + { + "epoch": 0.56, + "learning_rate": 8.395433793573904e-06, + "loss": 0.8964, + "step": 23484 + }, + { + "epoch": 0.56, + "learning_rate": 8.394664996264658e-06, + "loss": 1.0667, + "step": 23485 + }, + { + "epoch": 0.56, + "learning_rate": 8.393896208694607e-06, + "loss": 0.9921, + "step": 23486 + }, + { + "epoch": 0.56, + "learning_rate": 8.393127430868406e-06, + "loss": 0.9768, + "step": 23487 + }, + { + "epoch": 0.56, + "learning_rate": 8.392358662790724e-06, + "loss": 1.2624, + "step": 23488 + }, + { + "epoch": 0.56, + "learning_rate": 8.391589904466224e-06, + "loss": 1.2681, + "step": 23489 + }, + { + "epoch": 0.56, + "learning_rate": 8.39082115589957e-06, + "loss": 0.9691, + "step": 23490 + }, + { + "epoch": 0.56, + "learning_rate": 8.390052417095425e-06, + "loss": 1.0394, + "step": 23491 + }, + { + "epoch": 0.56, + "learning_rate": 8.389283688058454e-06, + "loss": 1.0891, + "step": 23492 + }, + { + "epoch": 0.56, + "learning_rate": 8.38851496879332e-06, + "loss": 0.9684, + "step": 23493 + }, + { + "epoch": 0.57, + "learning_rate": 8.387746259304685e-06, + "loss": 0.8927, + "step": 23494 + }, + { + "epoch": 0.57, + "learning_rate": 8.386977559597217e-06, + "loss": 1.0503, + "step": 23495 + }, + { + "epoch": 0.57, + "learning_rate": 8.386208869675576e-06, + "loss": 0.8855, + "step": 23496 + }, + { + "epoch": 0.57, + "learning_rate": 8.385440189544422e-06, + "loss": 0.9376, + "step": 23497 + }, + { + "epoch": 0.57, + "learning_rate": 8.384671519208427e-06, + "loss": 1.23, + "step": 23498 + }, + { + "epoch": 0.57, + "learning_rate": 8.383902858672245e-06, + "loss": 1.0732, + "step": 23499 + }, + { + "epoch": 0.57, + "learning_rate": 8.383134207940548e-06, + "loss": 1.2071, + "step": 23500 + }, + { + "epoch": 0.57, + "learning_rate": 8.382365567017997e-06, + "loss": 0.9458, + "step": 23501 + }, + { + "epoch": 0.57, + "learning_rate": 8.381596935909253e-06, + "loss": 1.0742, + "step": 23502 + }, + { + "epoch": 0.57, + "learning_rate": 8.380828314618978e-06, + "loss": 1.0723, + "step": 23503 + }, + { + "epoch": 0.57, + "learning_rate": 8.380059703151837e-06, + "loss": 1.0461, + "step": 23504 + }, + { + "epoch": 0.57, + "learning_rate": 8.379291101512494e-06, + "loss": 1.1997, + "step": 23505 + }, + { + "epoch": 0.57, + "learning_rate": 8.378522509705611e-06, + "loss": 0.8914, + "step": 23506 + }, + { + "epoch": 0.57, + "learning_rate": 8.377753927735847e-06, + "loss": 1.085, + "step": 23507 + }, + { + "epoch": 0.57, + "learning_rate": 8.376985355607872e-06, + "loss": 1.0841, + "step": 23508 + }, + { + "epoch": 0.57, + "learning_rate": 8.376216793326345e-06, + "loss": 1.377, + "step": 23509 + }, + { + "epoch": 0.57, + "learning_rate": 8.375448240895927e-06, + "loss": 1.0394, + "step": 23510 + }, + { + "epoch": 0.57, + "learning_rate": 8.374679698321282e-06, + "loss": 0.9391, + "step": 23511 + }, + { + "epoch": 0.57, + "learning_rate": 8.373911165607075e-06, + "loss": 1.0741, + "step": 23512 + }, + { + "epoch": 0.57, + "learning_rate": 8.373142642757968e-06, + "loss": 0.9806, + "step": 23513 + }, + { + "epoch": 0.57, + "learning_rate": 8.372374129778617e-06, + "loss": 0.8907, + "step": 23514 + }, + { + "epoch": 0.57, + "learning_rate": 8.371605626673694e-06, + "loss": 1.0638, + "step": 23515 + }, + { + "epoch": 0.57, + "learning_rate": 8.370837133447853e-06, + "loss": 0.8113, + "step": 23516 + }, + { + "epoch": 0.57, + "learning_rate": 8.370068650105763e-06, + "loss": 1.1637, + "step": 23517 + }, + { + "epoch": 0.57, + "learning_rate": 8.36930017665208e-06, + "loss": 0.9536, + "step": 23518 + }, + { + "epoch": 0.57, + "learning_rate": 8.36853171309147e-06, + "loss": 1.0496, + "step": 23519 + }, + { + "epoch": 0.57, + "learning_rate": 8.367763259428597e-06, + "loss": 0.8862, + "step": 23520 + }, + { + "epoch": 0.57, + "learning_rate": 8.366994815668118e-06, + "loss": 1.2281, + "step": 23521 + }, + { + "epoch": 0.57, + "learning_rate": 8.366226381814698e-06, + "loss": 1.2932, + "step": 23522 + }, + { + "epoch": 0.57, + "learning_rate": 8.365457957872998e-06, + "loss": 1.0874, + "step": 23523 + }, + { + "epoch": 0.57, + "learning_rate": 8.364689543847681e-06, + "loss": 1.3303, + "step": 23524 + }, + { + "epoch": 0.57, + "learning_rate": 8.363921139743408e-06, + "loss": 0.771, + "step": 23525 + }, + { + "epoch": 0.57, + "learning_rate": 8.363152745564838e-06, + "loss": 1.0989, + "step": 23526 + }, + { + "epoch": 0.57, + "learning_rate": 8.362384361316638e-06, + "loss": 1.0686, + "step": 23527 + }, + { + "epoch": 0.57, + "learning_rate": 8.361615987003467e-06, + "loss": 1.1114, + "step": 23528 + }, + { + "epoch": 0.57, + "learning_rate": 8.360847622629986e-06, + "loss": 0.9317, + "step": 23529 + }, + { + "epoch": 0.57, + "learning_rate": 8.360079268200854e-06, + "loss": 0.9142, + "step": 23530 + }, + { + "epoch": 0.57, + "learning_rate": 8.359310923720737e-06, + "loss": 1.066, + "step": 23531 + }, + { + "epoch": 0.57, + "learning_rate": 8.358542589194297e-06, + "loss": 0.9567, + "step": 23532 + }, + { + "epoch": 0.57, + "learning_rate": 8.357774264626192e-06, + "loss": 1.1299, + "step": 23533 + }, + { + "epoch": 0.57, + "learning_rate": 8.357005950021081e-06, + "loss": 1.0551, + "step": 23534 + }, + { + "epoch": 0.57, + "learning_rate": 8.356237645383633e-06, + "loss": 0.86, + "step": 23535 + }, + { + "epoch": 0.57, + "learning_rate": 8.355469350718503e-06, + "loss": 1.0479, + "step": 23536 + }, + { + "epoch": 0.57, + "learning_rate": 8.354701066030354e-06, + "loss": 1.1189, + "step": 23537 + }, + { + "epoch": 0.57, + "learning_rate": 8.353932791323845e-06, + "loss": 0.9694, + "step": 23538 + }, + { + "epoch": 0.57, + "learning_rate": 8.353164526603639e-06, + "loss": 1.1359, + "step": 23539 + }, + { + "epoch": 0.57, + "learning_rate": 8.352396271874398e-06, + "loss": 1.1603, + "step": 23540 + }, + { + "epoch": 0.57, + "learning_rate": 8.35162802714078e-06, + "loss": 1.2961, + "step": 23541 + }, + { + "epoch": 0.57, + "learning_rate": 8.350859792407446e-06, + "loss": 0.7571, + "step": 23542 + }, + { + "epoch": 0.57, + "learning_rate": 8.350091567679059e-06, + "loss": 0.9252, + "step": 23543 + }, + { + "epoch": 0.57, + "learning_rate": 8.349323352960279e-06, + "loss": 1.1542, + "step": 23544 + }, + { + "epoch": 0.57, + "learning_rate": 8.348555148255765e-06, + "loss": 1.362, + "step": 23545 + }, + { + "epoch": 0.57, + "learning_rate": 8.347786953570178e-06, + "loss": 1.2517, + "step": 23546 + }, + { + "epoch": 0.57, + "learning_rate": 8.347018768908182e-06, + "loss": 1.1424, + "step": 23547 + }, + { + "epoch": 0.57, + "learning_rate": 8.346250594274433e-06, + "loss": 1.3264, + "step": 23548 + }, + { + "epoch": 0.57, + "learning_rate": 8.34548242967359e-06, + "loss": 0.8877, + "step": 23549 + }, + { + "epoch": 0.57, + "learning_rate": 8.344714275110318e-06, + "loss": 1.0297, + "step": 23550 + }, + { + "epoch": 0.57, + "learning_rate": 8.343946130589274e-06, + "loss": 1.0699, + "step": 23551 + }, + { + "epoch": 0.57, + "learning_rate": 8.343177996115122e-06, + "loss": 0.9273, + "step": 23552 + }, + { + "epoch": 0.57, + "learning_rate": 8.342409871692517e-06, + "loss": 1.1919, + "step": 23553 + }, + { + "epoch": 0.57, + "learning_rate": 8.34164175732612e-06, + "loss": 1.3533, + "step": 23554 + }, + { + "epoch": 0.57, + "learning_rate": 8.340873653020596e-06, + "loss": 0.8712, + "step": 23555 + }, + { + "epoch": 0.57, + "learning_rate": 8.3401055587806e-06, + "loss": 1.1774, + "step": 23556 + }, + { + "epoch": 0.57, + "learning_rate": 8.33933747461079e-06, + "loss": 0.8707, + "step": 23557 + }, + { + "epoch": 0.57, + "learning_rate": 8.338569400515833e-06, + "loss": 0.9402, + "step": 23558 + }, + { + "epoch": 0.57, + "learning_rate": 8.337801336500384e-06, + "loss": 1.149, + "step": 23559 + }, + { + "epoch": 0.57, + "learning_rate": 8.337033282569102e-06, + "loss": 0.784, + "step": 23560 + }, + { + "epoch": 0.57, + "learning_rate": 8.336265238726649e-06, + "loss": 1.1617, + "step": 23561 + }, + { + "epoch": 0.57, + "learning_rate": 8.335497204977682e-06, + "loss": 1.2628, + "step": 23562 + }, + { + "epoch": 0.57, + "learning_rate": 8.334729181326863e-06, + "loss": 0.977, + "step": 23563 + }, + { + "epoch": 0.57, + "learning_rate": 8.333961167778853e-06, + "loss": 1.2915, + "step": 23564 + }, + { + "epoch": 0.57, + "learning_rate": 8.333193164338303e-06, + "loss": 1.2748, + "step": 23565 + }, + { + "epoch": 0.57, + "learning_rate": 8.332425171009882e-06, + "loss": 0.9802, + "step": 23566 + }, + { + "epoch": 0.57, + "learning_rate": 8.331657187798246e-06, + "loss": 0.9588, + "step": 23567 + }, + { + "epoch": 0.57, + "learning_rate": 8.330889214708051e-06, + "loss": 1.1397, + "step": 23568 + }, + { + "epoch": 0.57, + "learning_rate": 8.330121251743958e-06, + "loss": 1.1982, + "step": 23569 + }, + { + "epoch": 0.57, + "learning_rate": 8.329353298910629e-06, + "loss": 1.1449, + "step": 23570 + }, + { + "epoch": 0.57, + "learning_rate": 8.328585356212719e-06, + "loss": 1.0907, + "step": 23571 + }, + { + "epoch": 0.57, + "learning_rate": 8.32781742365489e-06, + "loss": 1.2357, + "step": 23572 + }, + { + "epoch": 0.57, + "learning_rate": 8.327049501241797e-06, + "loss": 1.0153, + "step": 23573 + }, + { + "epoch": 0.57, + "learning_rate": 8.326281588978102e-06, + "loss": 1.1517, + "step": 23574 + }, + { + "epoch": 0.57, + "learning_rate": 8.325513686868465e-06, + "loss": 0.9494, + "step": 23575 + }, + { + "epoch": 0.57, + "learning_rate": 8.32474579491754e-06, + "loss": 0.7985, + "step": 23576 + }, + { + "epoch": 0.57, + "learning_rate": 8.32397791312999e-06, + "loss": 1.0981, + "step": 23577 + }, + { + "epoch": 0.57, + "learning_rate": 8.32321004151047e-06, + "loss": 1.087, + "step": 23578 + }, + { + "epoch": 0.57, + "learning_rate": 8.322442180063643e-06, + "loss": 0.8617, + "step": 23579 + }, + { + "epoch": 0.57, + "learning_rate": 8.321674328794162e-06, + "loss": 0.8723, + "step": 23580 + }, + { + "epoch": 0.57, + "learning_rate": 8.320906487706689e-06, + "loss": 1.0841, + "step": 23581 + }, + { + "epoch": 0.57, + "learning_rate": 8.32013865680588e-06, + "loss": 0.9536, + "step": 23582 + }, + { + "epoch": 0.57, + "learning_rate": 8.319370836096398e-06, + "loss": 1.1234, + "step": 23583 + }, + { + "epoch": 0.57, + "learning_rate": 8.318603025582897e-06, + "loss": 1.2469, + "step": 23584 + }, + { + "epoch": 0.57, + "learning_rate": 8.317835225270033e-06, + "loss": 1.3025, + "step": 23585 + }, + { + "epoch": 0.57, + "learning_rate": 8.317067435162469e-06, + "loss": 0.8376, + "step": 23586 + }, + { + "epoch": 0.57, + "learning_rate": 8.316299655264862e-06, + "loss": 1.0114, + "step": 23587 + }, + { + "epoch": 0.57, + "learning_rate": 8.315531885581867e-06, + "loss": 1.0813, + "step": 23588 + }, + { + "epoch": 0.57, + "learning_rate": 8.314764126118144e-06, + "loss": 1.065, + "step": 23589 + }, + { + "epoch": 0.57, + "learning_rate": 8.313996376878353e-06, + "loss": 1.2134, + "step": 23590 + }, + { + "epoch": 0.57, + "learning_rate": 8.313228637867148e-06, + "loss": 0.9416, + "step": 23591 + }, + { + "epoch": 0.57, + "learning_rate": 8.31246090908919e-06, + "loss": 0.9665, + "step": 23592 + }, + { + "epoch": 0.57, + "learning_rate": 8.311693190549128e-06, + "loss": 1.0456, + "step": 23593 + }, + { + "epoch": 0.57, + "learning_rate": 8.310925482251633e-06, + "loss": 0.9187, + "step": 23594 + }, + { + "epoch": 0.57, + "learning_rate": 8.310157784201354e-06, + "loss": 1.0541, + "step": 23595 + }, + { + "epoch": 0.57, + "learning_rate": 8.30939009640295e-06, + "loss": 0.9407, + "step": 23596 + }, + { + "epoch": 0.57, + "learning_rate": 8.308622418861078e-06, + "loss": 0.9343, + "step": 23597 + }, + { + "epoch": 0.57, + "learning_rate": 8.307854751580396e-06, + "loss": 1.1906, + "step": 23598 + }, + { + "epoch": 0.57, + "learning_rate": 8.307087094565562e-06, + "loss": 1.1535, + "step": 23599 + }, + { + "epoch": 0.57, + "learning_rate": 8.306319447821234e-06, + "loss": 1.2025, + "step": 23600 + }, + { + "epoch": 0.57, + "learning_rate": 8.305551811352064e-06, + "loss": 1.2697, + "step": 23601 + }, + { + "epoch": 0.57, + "learning_rate": 8.304784185162714e-06, + "loss": 1.1516, + "step": 23602 + }, + { + "epoch": 0.57, + "learning_rate": 8.304016569257841e-06, + "loss": 0.9889, + "step": 23603 + }, + { + "epoch": 0.57, + "learning_rate": 8.303248963642098e-06, + "loss": 0.9322, + "step": 23604 + }, + { + "epoch": 0.57, + "learning_rate": 8.302481368320146e-06, + "loss": 1.0582, + "step": 23605 + }, + { + "epoch": 0.57, + "learning_rate": 8.301713783296641e-06, + "loss": 0.8663, + "step": 23606 + }, + { + "epoch": 0.57, + "learning_rate": 8.300946208576239e-06, + "loss": 1.2269, + "step": 23607 + }, + { + "epoch": 0.57, + "learning_rate": 8.300178644163593e-06, + "loss": 1.0584, + "step": 23608 + }, + { + "epoch": 0.57, + "learning_rate": 8.299411090063367e-06, + "loss": 1.0567, + "step": 23609 + }, + { + "epoch": 0.57, + "learning_rate": 8.298643546280214e-06, + "loss": 1.0415, + "step": 23610 + }, + { + "epoch": 0.57, + "learning_rate": 8.29787601281879e-06, + "loss": 1.2075, + "step": 23611 + }, + { + "epoch": 0.57, + "learning_rate": 8.297108489683749e-06, + "loss": 0.9773, + "step": 23612 + }, + { + "epoch": 0.57, + "learning_rate": 8.296340976879754e-06, + "loss": 1.0529, + "step": 23613 + }, + { + "epoch": 0.57, + "learning_rate": 8.295573474411456e-06, + "loss": 1.3458, + "step": 23614 + }, + { + "epoch": 0.57, + "learning_rate": 8.294805982283515e-06, + "loss": 0.8047, + "step": 23615 + }, + { + "epoch": 0.57, + "learning_rate": 8.29403850050058e-06, + "loss": 1.3097, + "step": 23616 + }, + { + "epoch": 0.57, + "learning_rate": 8.293271029067318e-06, + "loss": 1.248, + "step": 23617 + }, + { + "epoch": 0.57, + "learning_rate": 8.292503567988375e-06, + "loss": 1.1722, + "step": 23618 + }, + { + "epoch": 0.57, + "learning_rate": 8.291736117268414e-06, + "loss": 1.2375, + "step": 23619 + }, + { + "epoch": 0.57, + "learning_rate": 8.290968676912085e-06, + "loss": 1.0299, + "step": 23620 + }, + { + "epoch": 0.57, + "learning_rate": 8.29020124692405e-06, + "loss": 0.9104, + "step": 23621 + }, + { + "epoch": 0.57, + "learning_rate": 8.289433827308961e-06, + "loss": 1.2405, + "step": 23622 + }, + { + "epoch": 0.57, + "learning_rate": 8.288666418071474e-06, + "loss": 1.269, + "step": 23623 + }, + { + "epoch": 0.57, + "learning_rate": 8.287899019216245e-06, + "loss": 0.9375, + "step": 23624 + }, + { + "epoch": 0.57, + "learning_rate": 8.287131630747931e-06, + "loss": 1.2104, + "step": 23625 + }, + { + "epoch": 0.57, + "learning_rate": 8.286364252671188e-06, + "loss": 1.0017, + "step": 23626 + }, + { + "epoch": 0.57, + "learning_rate": 8.285596884990667e-06, + "loss": 1.0774, + "step": 23627 + }, + { + "epoch": 0.57, + "learning_rate": 8.284829527711026e-06, + "loss": 0.9857, + "step": 23628 + }, + { + "epoch": 0.57, + "learning_rate": 8.284062180836923e-06, + "loss": 0.9509, + "step": 23629 + }, + { + "epoch": 0.57, + "learning_rate": 8.283294844373012e-06, + "loss": 1.188, + "step": 23630 + }, + { + "epoch": 0.57, + "learning_rate": 8.282527518323946e-06, + "loss": 1.1131, + "step": 23631 + }, + { + "epoch": 0.57, + "learning_rate": 8.28176020269438e-06, + "loss": 1.0757, + "step": 23632 + }, + { + "epoch": 0.57, + "learning_rate": 8.280992897488973e-06, + "loss": 0.9383, + "step": 23633 + }, + { + "epoch": 0.57, + "learning_rate": 8.280225602712378e-06, + "loss": 1.0902, + "step": 23634 + }, + { + "epoch": 0.57, + "learning_rate": 8.279458318369249e-06, + "loss": 0.9609, + "step": 23635 + }, + { + "epoch": 0.57, + "learning_rate": 8.27869104446424e-06, + "loss": 1.1499, + "step": 23636 + }, + { + "epoch": 0.57, + "learning_rate": 8.277923781002008e-06, + "loss": 1.1032, + "step": 23637 + }, + { + "epoch": 0.57, + "learning_rate": 8.27715652798721e-06, + "loss": 1.0123, + "step": 23638 + }, + { + "epoch": 0.57, + "learning_rate": 8.276389285424498e-06, + "loss": 1.359, + "step": 23639 + }, + { + "epoch": 0.57, + "learning_rate": 8.275622053318523e-06, + "loss": 1.0259, + "step": 23640 + }, + { + "epoch": 0.57, + "learning_rate": 8.274854831673946e-06, + "loss": 0.9022, + "step": 23641 + }, + { + "epoch": 0.57, + "learning_rate": 8.274087620495419e-06, + "loss": 1.0939, + "step": 23642 + }, + { + "epoch": 0.57, + "learning_rate": 8.273320419787596e-06, + "loss": 1.1954, + "step": 23643 + }, + { + "epoch": 0.57, + "learning_rate": 8.272553229555132e-06, + "loss": 1.0128, + "step": 23644 + }, + { + "epoch": 0.57, + "learning_rate": 8.27178604980268e-06, + "loss": 0.8581, + "step": 23645 + }, + { + "epoch": 0.57, + "learning_rate": 8.271018880534898e-06, + "loss": 1.0639, + "step": 23646 + }, + { + "epoch": 0.57, + "learning_rate": 8.270251721756434e-06, + "loss": 0.9912, + "step": 23647 + }, + { + "epoch": 0.57, + "learning_rate": 8.26948457347195e-06, + "loss": 1.0517, + "step": 23648 + }, + { + "epoch": 0.57, + "learning_rate": 8.268717435686095e-06, + "loss": 1.1766, + "step": 23649 + }, + { + "epoch": 0.57, + "learning_rate": 8.267950308403523e-06, + "loss": 0.8958, + "step": 23650 + }, + { + "epoch": 0.57, + "learning_rate": 8.26718319162889e-06, + "loss": 1.1168, + "step": 23651 + }, + { + "epoch": 0.57, + "learning_rate": 8.266416085366847e-06, + "loss": 1.1416, + "step": 23652 + }, + { + "epoch": 0.57, + "learning_rate": 8.265648989622052e-06, + "loss": 0.8426, + "step": 23653 + }, + { + "epoch": 0.57, + "learning_rate": 8.264881904399157e-06, + "loss": 0.927, + "step": 23654 + }, + { + "epoch": 0.57, + "learning_rate": 8.264114829702813e-06, + "loss": 0.9825, + "step": 23655 + }, + { + "epoch": 0.57, + "learning_rate": 8.263347765537678e-06, + "loss": 1.2639, + "step": 23656 + }, + { + "epoch": 0.57, + "learning_rate": 8.262580711908403e-06, + "loss": 1.2228, + "step": 23657 + }, + { + "epoch": 0.57, + "learning_rate": 8.261813668819644e-06, + "loss": 1.1405, + "step": 23658 + }, + { + "epoch": 0.57, + "learning_rate": 8.26104663627605e-06, + "loss": 1.006, + "step": 23659 + }, + { + "epoch": 0.57, + "learning_rate": 8.260279614282277e-06, + "loss": 1.1947, + "step": 23660 + }, + { + "epoch": 0.57, + "learning_rate": 8.25951260284298e-06, + "loss": 1.0758, + "step": 23661 + }, + { + "epoch": 0.57, + "learning_rate": 8.258745601962811e-06, + "loss": 1.021, + "step": 23662 + }, + { + "epoch": 0.57, + "learning_rate": 8.25797861164642e-06, + "loss": 1.0643, + "step": 23663 + }, + { + "epoch": 0.57, + "learning_rate": 8.257211631898465e-06, + "loss": 0.7948, + "step": 23664 + }, + { + "epoch": 0.57, + "learning_rate": 8.256444662723597e-06, + "loss": 1.0225, + "step": 23665 + }, + { + "epoch": 0.57, + "learning_rate": 8.255677704126466e-06, + "loss": 1.227, + "step": 23666 + }, + { + "epoch": 0.57, + "learning_rate": 8.254910756111731e-06, + "loss": 1.2653, + "step": 23667 + }, + { + "epoch": 0.57, + "learning_rate": 8.254143818684042e-06, + "loss": 1.2705, + "step": 23668 + }, + { + "epoch": 0.57, + "learning_rate": 8.25337689184805e-06, + "loss": 1.2106, + "step": 23669 + }, + { + "epoch": 0.57, + "learning_rate": 8.252609975608413e-06, + "loss": 0.9061, + "step": 23670 + }, + { + "epoch": 0.57, + "learning_rate": 8.251843069969779e-06, + "loss": 1.1702, + "step": 23671 + }, + { + "epoch": 0.57, + "learning_rate": 8.251076174936802e-06, + "loss": 1.173, + "step": 23672 + }, + { + "epoch": 0.57, + "learning_rate": 8.250309290514133e-06, + "loss": 1.1365, + "step": 23673 + }, + { + "epoch": 0.57, + "learning_rate": 8.249542416706428e-06, + "loss": 1.2263, + "step": 23674 + }, + { + "epoch": 0.57, + "learning_rate": 8.248775553518336e-06, + "loss": 1.1023, + "step": 23675 + }, + { + "epoch": 0.57, + "learning_rate": 8.248008700954512e-06, + "loss": 0.6758, + "step": 23676 + }, + { + "epoch": 0.57, + "learning_rate": 8.247241859019605e-06, + "loss": 1.0054, + "step": 23677 + }, + { + "epoch": 0.57, + "learning_rate": 8.246475027718272e-06, + "loss": 0.759, + "step": 23678 + }, + { + "epoch": 0.57, + "learning_rate": 8.245708207055163e-06, + "loss": 0.9828, + "step": 23679 + }, + { + "epoch": 0.57, + "learning_rate": 8.244941397034924e-06, + "loss": 1.0102, + "step": 23680 + }, + { + "epoch": 0.57, + "learning_rate": 8.244174597662218e-06, + "loss": 1.2039, + "step": 23681 + }, + { + "epoch": 0.57, + "learning_rate": 8.243407808941692e-06, + "loss": 1.0805, + "step": 23682 + }, + { + "epoch": 0.57, + "learning_rate": 8.242641030877995e-06, + "loss": 0.8658, + "step": 23683 + }, + { + "epoch": 0.57, + "learning_rate": 8.241874263475782e-06, + "loss": 1.0979, + "step": 23684 + }, + { + "epoch": 0.57, + "learning_rate": 8.241107506739704e-06, + "loss": 1.1787, + "step": 23685 + }, + { + "epoch": 0.57, + "learning_rate": 8.240340760674415e-06, + "loss": 1.0777, + "step": 23686 + }, + { + "epoch": 0.57, + "learning_rate": 8.239574025284564e-06, + "loss": 0.9578, + "step": 23687 + }, + { + "epoch": 0.57, + "learning_rate": 8.238807300574802e-06, + "loss": 1.0867, + "step": 23688 + }, + { + "epoch": 0.57, + "learning_rate": 8.238040586549783e-06, + "loss": 1.1251, + "step": 23689 + }, + { + "epoch": 0.57, + "learning_rate": 8.237273883214159e-06, + "loss": 0.987, + "step": 23690 + }, + { + "epoch": 0.57, + "learning_rate": 8.236507190572578e-06, + "loss": 0.9577, + "step": 23691 + }, + { + "epoch": 0.57, + "learning_rate": 8.23574050862969e-06, + "loss": 1.1929, + "step": 23692 + }, + { + "epoch": 0.57, + "learning_rate": 8.234973837390154e-06, + "loss": 1.0435, + "step": 23693 + }, + { + "epoch": 0.57, + "learning_rate": 8.234207176858615e-06, + "loss": 1.1363, + "step": 23694 + }, + { + "epoch": 0.57, + "learning_rate": 8.233440527039727e-06, + "loss": 1.082, + "step": 23695 + }, + { + "epoch": 0.57, + "learning_rate": 8.232673887938136e-06, + "loss": 1.2134, + "step": 23696 + }, + { + "epoch": 0.57, + "learning_rate": 8.2319072595585e-06, + "loss": 1.12, + "step": 23697 + }, + { + "epoch": 0.57, + "learning_rate": 8.231140641905467e-06, + "loss": 0.9515, + "step": 23698 + }, + { + "epoch": 0.57, + "learning_rate": 8.230374034983686e-06, + "loss": 1.0124, + "step": 23699 + }, + { + "epoch": 0.57, + "learning_rate": 8.22960743879781e-06, + "loss": 1.1238, + "step": 23700 + }, + { + "epoch": 0.57, + "learning_rate": 8.22884085335249e-06, + "loss": 1.2338, + "step": 23701 + }, + { + "epoch": 0.57, + "learning_rate": 8.228074278652376e-06, + "loss": 0.7544, + "step": 23702 + }, + { + "epoch": 0.57, + "learning_rate": 8.227307714702117e-06, + "loss": 0.6975, + "step": 23703 + }, + { + "epoch": 0.57, + "learning_rate": 8.226541161506364e-06, + "loss": 1.325, + "step": 23704 + }, + { + "epoch": 0.57, + "learning_rate": 8.225774619069773e-06, + "loss": 1.3535, + "step": 23705 + }, + { + "epoch": 0.57, + "learning_rate": 8.225008087396988e-06, + "loss": 0.9443, + "step": 23706 + }, + { + "epoch": 0.57, + "learning_rate": 8.22424156649266e-06, + "loss": 1.0896, + "step": 23707 + }, + { + "epoch": 0.57, + "learning_rate": 8.223475056361442e-06, + "loss": 0.8399, + "step": 23708 + }, + { + "epoch": 0.57, + "learning_rate": 8.222708557007984e-06, + "loss": 0.7069, + "step": 23709 + }, + { + "epoch": 0.57, + "learning_rate": 8.221942068436935e-06, + "loss": 1.1652, + "step": 23710 + }, + { + "epoch": 0.57, + "learning_rate": 8.221175590652944e-06, + "loss": 1.1349, + "step": 23711 + }, + { + "epoch": 0.57, + "learning_rate": 8.22040912366066e-06, + "loss": 0.9738, + "step": 23712 + }, + { + "epoch": 0.57, + "learning_rate": 8.219642667464739e-06, + "loss": 1.0106, + "step": 23713 + }, + { + "epoch": 0.57, + "learning_rate": 8.218876222069827e-06, + "loss": 1.1344, + "step": 23714 + }, + { + "epoch": 0.57, + "learning_rate": 8.218109787480573e-06, + "loss": 0.9983, + "step": 23715 + }, + { + "epoch": 0.57, + "learning_rate": 8.217343363701627e-06, + "loss": 1.2271, + "step": 23716 + }, + { + "epoch": 0.57, + "learning_rate": 8.21657695073764e-06, + "loss": 0.9479, + "step": 23717 + }, + { + "epoch": 0.57, + "learning_rate": 8.215810548593261e-06, + "loss": 1.1652, + "step": 23718 + }, + { + "epoch": 0.57, + "learning_rate": 8.215044157273143e-06, + "loss": 0.8314, + "step": 23719 + }, + { + "epoch": 0.57, + "learning_rate": 8.214277776781928e-06, + "loss": 1.152, + "step": 23720 + }, + { + "epoch": 0.57, + "learning_rate": 8.21351140712427e-06, + "loss": 1.1989, + "step": 23721 + }, + { + "epoch": 0.57, + "learning_rate": 8.21274504830482e-06, + "loss": 1.2328, + "step": 23722 + }, + { + "epoch": 0.57, + "learning_rate": 8.211978700328225e-06, + "loss": 0.9669, + "step": 23723 + }, + { + "epoch": 0.57, + "learning_rate": 8.211212363199132e-06, + "loss": 1.0582, + "step": 23724 + }, + { + "epoch": 0.57, + "learning_rate": 8.210446036922198e-06, + "loss": 1.1299, + "step": 23725 + }, + { + "epoch": 0.57, + "learning_rate": 8.209679721502063e-06, + "loss": 1.1213, + "step": 23726 + }, + { + "epoch": 0.57, + "learning_rate": 8.20891341694338e-06, + "loss": 1.0825, + "step": 23727 + }, + { + "epoch": 0.57, + "learning_rate": 8.208147123250799e-06, + "loss": 1.0501, + "step": 23728 + }, + { + "epoch": 0.57, + "learning_rate": 8.20738084042897e-06, + "loss": 1.1005, + "step": 23729 + }, + { + "epoch": 0.57, + "learning_rate": 8.206614568482538e-06, + "loss": 1.2709, + "step": 23730 + }, + { + "epoch": 0.57, + "learning_rate": 8.205848307416151e-06, + "loss": 1.0582, + "step": 23731 + }, + { + "epoch": 0.57, + "learning_rate": 8.205082057234463e-06, + "loss": 1.1097, + "step": 23732 + }, + { + "epoch": 0.57, + "learning_rate": 8.20431581794212e-06, + "loss": 0.999, + "step": 23733 + }, + { + "epoch": 0.57, + "learning_rate": 8.203549589543772e-06, + "loss": 1.1714, + "step": 23734 + }, + { + "epoch": 0.57, + "learning_rate": 8.202783372044063e-06, + "loss": 1.2195, + "step": 23735 + }, + { + "epoch": 0.57, + "learning_rate": 8.202017165447646e-06, + "loss": 1.2027, + "step": 23736 + }, + { + "epoch": 0.57, + "learning_rate": 8.201250969759169e-06, + "loss": 0.879, + "step": 23737 + }, + { + "epoch": 0.57, + "learning_rate": 8.200484784983278e-06, + "loss": 0.9712, + "step": 23738 + }, + { + "epoch": 0.57, + "learning_rate": 8.19971861112462e-06, + "loss": 1.0539, + "step": 23739 + }, + { + "epoch": 0.57, + "learning_rate": 8.19895244818785e-06, + "loss": 1.1175, + "step": 23740 + }, + { + "epoch": 0.57, + "learning_rate": 8.19818629617761e-06, + "loss": 1.0127, + "step": 23741 + }, + { + "epoch": 0.57, + "learning_rate": 8.197420155098551e-06, + "loss": 1.0491, + "step": 23742 + }, + { + "epoch": 0.57, + "learning_rate": 8.196654024955317e-06, + "loss": 1.1653, + "step": 23743 + }, + { + "epoch": 0.57, + "learning_rate": 8.195887905752563e-06, + "loss": 1.0358, + "step": 23744 + }, + { + "epoch": 0.57, + "learning_rate": 8.19512179749493e-06, + "loss": 0.9823, + "step": 23745 + }, + { + "epoch": 0.57, + "learning_rate": 8.19435570018707e-06, + "loss": 1.0151, + "step": 23746 + }, + { + "epoch": 0.57, + "learning_rate": 8.193589613833627e-06, + "loss": 1.1342, + "step": 23747 + }, + { + "epoch": 0.57, + "learning_rate": 8.192823538439253e-06, + "loss": 1.0236, + "step": 23748 + }, + { + "epoch": 0.57, + "learning_rate": 8.192057474008594e-06, + "loss": 1.1818, + "step": 23749 + }, + { + "epoch": 0.57, + "learning_rate": 8.191291420546296e-06, + "loss": 1.085, + "step": 23750 + }, + { + "epoch": 0.57, + "learning_rate": 8.190525378057005e-06, + "loss": 1.1285, + "step": 23751 + }, + { + "epoch": 0.57, + "learning_rate": 8.189759346545375e-06, + "loss": 0.9885, + "step": 23752 + }, + { + "epoch": 0.57, + "learning_rate": 8.188993326016047e-06, + "loss": 1.0546, + "step": 23753 + }, + { + "epoch": 0.57, + "learning_rate": 8.188227316473672e-06, + "loss": 1.1723, + "step": 23754 + }, + { + "epoch": 0.57, + "learning_rate": 8.187461317922894e-06, + "loss": 1.1258, + "step": 23755 + }, + { + "epoch": 0.57, + "learning_rate": 8.186695330368364e-06, + "loss": 1.0221, + "step": 23756 + }, + { + "epoch": 0.57, + "learning_rate": 8.185929353814725e-06, + "loss": 0.9738, + "step": 23757 + }, + { + "epoch": 0.57, + "learning_rate": 8.185163388266628e-06, + "loss": 1.2963, + "step": 23758 + }, + { + "epoch": 0.57, + "learning_rate": 8.184397433728715e-06, + "loss": 1.1726, + "step": 23759 + }, + { + "epoch": 0.57, + "learning_rate": 8.183631490205636e-06, + "loss": 1.24, + "step": 23760 + }, + { + "epoch": 0.57, + "learning_rate": 8.18286555770204e-06, + "loss": 0.8879, + "step": 23761 + }, + { + "epoch": 0.57, + "learning_rate": 8.18209963622257e-06, + "loss": 1.0201, + "step": 23762 + }, + { + "epoch": 0.57, + "learning_rate": 8.181333725771873e-06, + "loss": 0.9752, + "step": 23763 + }, + { + "epoch": 0.57, + "learning_rate": 8.180567826354597e-06, + "loss": 0.9797, + "step": 23764 + }, + { + "epoch": 0.57, + "learning_rate": 8.179801937975389e-06, + "loss": 1.1694, + "step": 23765 + }, + { + "epoch": 0.57, + "learning_rate": 8.179036060638895e-06, + "loss": 0.8711, + "step": 23766 + }, + { + "epoch": 0.57, + "learning_rate": 8.178270194349757e-06, + "loss": 1.0913, + "step": 23767 + }, + { + "epoch": 0.57, + "learning_rate": 8.177504339112629e-06, + "loss": 1.0652, + "step": 23768 + }, + { + "epoch": 0.57, + "learning_rate": 8.176738494932154e-06, + "loss": 1.3248, + "step": 23769 + }, + { + "epoch": 0.57, + "learning_rate": 8.175972661812977e-06, + "loss": 1.1971, + "step": 23770 + }, + { + "epoch": 0.57, + "learning_rate": 8.175206839759742e-06, + "loss": 1.3428, + "step": 23771 + }, + { + "epoch": 0.57, + "learning_rate": 8.1744410287771e-06, + "loss": 1.047, + "step": 23772 + }, + { + "epoch": 0.57, + "learning_rate": 8.173675228869696e-06, + "loss": 1.0353, + "step": 23773 + }, + { + "epoch": 0.57, + "learning_rate": 8.172909440042172e-06, + "loss": 1.0494, + "step": 23774 + }, + { + "epoch": 0.57, + "learning_rate": 8.172143662299178e-06, + "loss": 1.0891, + "step": 23775 + }, + { + "epoch": 0.57, + "learning_rate": 8.171377895645359e-06, + "loss": 0.9963, + "step": 23776 + }, + { + "epoch": 0.57, + "learning_rate": 8.17061214008536e-06, + "loss": 1.0371, + "step": 23777 + }, + { + "epoch": 0.57, + "learning_rate": 8.169846395623826e-06, + "loss": 1.1007, + "step": 23778 + }, + { + "epoch": 0.57, + "learning_rate": 8.169080662265405e-06, + "loss": 0.8832, + "step": 23779 + }, + { + "epoch": 0.57, + "learning_rate": 8.168314940014742e-06, + "loss": 1.21, + "step": 23780 + }, + { + "epoch": 0.57, + "learning_rate": 8.16754922887648e-06, + "loss": 1.2262, + "step": 23781 + }, + { + "epoch": 0.57, + "learning_rate": 8.166783528855263e-06, + "loss": 0.8955, + "step": 23782 + }, + { + "epoch": 0.57, + "learning_rate": 8.16601783995574e-06, + "loss": 0.6929, + "step": 23783 + }, + { + "epoch": 0.57, + "learning_rate": 8.165252162182559e-06, + "loss": 1.1237, + "step": 23784 + }, + { + "epoch": 0.57, + "learning_rate": 8.16448649554036e-06, + "loss": 1.018, + "step": 23785 + }, + { + "epoch": 0.57, + "learning_rate": 8.163720840033788e-06, + "loss": 1.0916, + "step": 23786 + }, + { + "epoch": 0.57, + "learning_rate": 8.16295519566749e-06, + "loss": 1.1052, + "step": 23787 + }, + { + "epoch": 0.57, + "learning_rate": 8.162189562446112e-06, + "loss": 1.0349, + "step": 23788 + }, + { + "epoch": 0.57, + "learning_rate": 8.161423940374298e-06, + "loss": 1.2235, + "step": 23789 + }, + { + "epoch": 0.57, + "learning_rate": 8.160658329456692e-06, + "loss": 1.1806, + "step": 23790 + }, + { + "epoch": 0.57, + "learning_rate": 8.159892729697938e-06, + "loss": 1.2301, + "step": 23791 + }, + { + "epoch": 0.57, + "learning_rate": 8.159127141102684e-06, + "loss": 1.2742, + "step": 23792 + }, + { + "epoch": 0.57, + "learning_rate": 8.158361563675572e-06, + "loss": 1.0874, + "step": 23793 + }, + { + "epoch": 0.57, + "learning_rate": 8.157595997421246e-06, + "loss": 0.9995, + "step": 23794 + }, + { + "epoch": 0.57, + "learning_rate": 8.156830442344352e-06, + "loss": 0.8931, + "step": 23795 + }, + { + "epoch": 0.57, + "learning_rate": 8.156064898449536e-06, + "loss": 0.9778, + "step": 23796 + }, + { + "epoch": 0.57, + "learning_rate": 8.15529936574144e-06, + "loss": 1.0508, + "step": 23797 + }, + { + "epoch": 0.57, + "learning_rate": 8.154533844224706e-06, + "loss": 0.6094, + "step": 23798 + }, + { + "epoch": 0.57, + "learning_rate": 8.153768333903985e-06, + "loss": 1.1122, + "step": 23799 + }, + { + "epoch": 0.57, + "learning_rate": 8.153002834783916e-06, + "loss": 1.114, + "step": 23800 + }, + { + "epoch": 0.57, + "learning_rate": 8.152237346869142e-06, + "loss": 1.2697, + "step": 23801 + }, + { + "epoch": 0.57, + "learning_rate": 8.151471870164311e-06, + "loss": 1.2226, + "step": 23802 + }, + { + "epoch": 0.57, + "learning_rate": 8.150706404674068e-06, + "loss": 1.2942, + "step": 23803 + }, + { + "epoch": 0.57, + "learning_rate": 8.14994095040305e-06, + "loss": 1.0803, + "step": 23804 + }, + { + "epoch": 0.57, + "learning_rate": 8.149175507355908e-06, + "loss": 0.8995, + "step": 23805 + }, + { + "epoch": 0.57, + "learning_rate": 8.14841007553728e-06, + "loss": 0.7431, + "step": 23806 + }, + { + "epoch": 0.57, + "learning_rate": 8.147644654951814e-06, + "loss": 1.1221, + "step": 23807 + }, + { + "epoch": 0.57, + "learning_rate": 8.146879245604155e-06, + "loss": 1.1293, + "step": 23808 + }, + { + "epoch": 0.57, + "learning_rate": 8.146113847498943e-06, + "loss": 1.1113, + "step": 23809 + }, + { + "epoch": 0.57, + "learning_rate": 8.145348460640817e-06, + "loss": 1.0362, + "step": 23810 + }, + { + "epoch": 0.57, + "learning_rate": 8.14458308503443e-06, + "loss": 1.0076, + "step": 23811 + }, + { + "epoch": 0.57, + "learning_rate": 8.14381772068442e-06, + "loss": 1.1571, + "step": 23812 + }, + { + "epoch": 0.57, + "learning_rate": 8.143052367595433e-06, + "loss": 0.996, + "step": 23813 + }, + { + "epoch": 0.57, + "learning_rate": 8.142287025772107e-06, + "loss": 1.1435, + "step": 23814 + }, + { + "epoch": 0.57, + "learning_rate": 8.14152169521909e-06, + "loss": 1.1649, + "step": 23815 + }, + { + "epoch": 0.57, + "learning_rate": 8.140756375941026e-06, + "loss": 1.0392, + "step": 23816 + }, + { + "epoch": 0.57, + "learning_rate": 8.139991067942554e-06, + "loss": 1.0808, + "step": 23817 + }, + { + "epoch": 0.57, + "learning_rate": 8.139225771228318e-06, + "loss": 1.0551, + "step": 23818 + }, + { + "epoch": 0.57, + "learning_rate": 8.138460485802962e-06, + "loss": 1.0325, + "step": 23819 + }, + { + "epoch": 0.57, + "learning_rate": 8.137695211671131e-06, + "loss": 1.1223, + "step": 23820 + }, + { + "epoch": 0.57, + "learning_rate": 8.13692994883746e-06, + "loss": 1.1506, + "step": 23821 + }, + { + "epoch": 0.57, + "learning_rate": 8.1361646973066e-06, + "loss": 1.1462, + "step": 23822 + }, + { + "epoch": 0.57, + "learning_rate": 8.13539945708319e-06, + "loss": 1.121, + "step": 23823 + }, + { + "epoch": 0.57, + "learning_rate": 8.134634228171875e-06, + "loss": 1.0075, + "step": 23824 + }, + { + "epoch": 0.57, + "learning_rate": 8.13386901057729e-06, + "loss": 0.9233, + "step": 23825 + }, + { + "epoch": 0.57, + "learning_rate": 8.133103804304087e-06, + "loss": 0.9125, + "step": 23826 + }, + { + "epoch": 0.57, + "learning_rate": 8.132338609356904e-06, + "loss": 1.255, + "step": 23827 + }, + { + "epoch": 0.57, + "learning_rate": 8.131573425740384e-06, + "loss": 1.1004, + "step": 23828 + }, + { + "epoch": 0.57, + "learning_rate": 8.130808253459165e-06, + "loss": 1.0037, + "step": 23829 + }, + { + "epoch": 0.57, + "learning_rate": 8.130043092517895e-06, + "loss": 1.1753, + "step": 23830 + }, + { + "epoch": 0.57, + "learning_rate": 8.129277942921214e-06, + "loss": 1.1216, + "step": 23831 + }, + { + "epoch": 0.57, + "learning_rate": 8.12851280467376e-06, + "loss": 1.1474, + "step": 23832 + }, + { + "epoch": 0.57, + "learning_rate": 8.127747677780183e-06, + "loss": 0.7036, + "step": 23833 + }, + { + "epoch": 0.57, + "learning_rate": 8.126982562245118e-06, + "loss": 1.2614, + "step": 23834 + }, + { + "epoch": 0.57, + "learning_rate": 8.12621745807321e-06, + "loss": 1.2657, + "step": 23835 + }, + { + "epoch": 0.57, + "learning_rate": 8.1254523652691e-06, + "loss": 1.1275, + "step": 23836 + }, + { + "epoch": 0.57, + "learning_rate": 8.124687283837432e-06, + "loss": 1.2511, + "step": 23837 + }, + { + "epoch": 0.57, + "learning_rate": 8.123922213782843e-06, + "loss": 1.0227, + "step": 23838 + }, + { + "epoch": 0.57, + "learning_rate": 8.123157155109975e-06, + "loss": 1.2005, + "step": 23839 + }, + { + "epoch": 0.57, + "learning_rate": 8.122392107823474e-06, + "loss": 1.0323, + "step": 23840 + }, + { + "epoch": 0.57, + "learning_rate": 8.121627071927977e-06, + "loss": 1.2144, + "step": 23841 + }, + { + "epoch": 0.57, + "learning_rate": 8.120862047428127e-06, + "loss": 1.1608, + "step": 23842 + }, + { + "epoch": 0.57, + "learning_rate": 8.120097034328565e-06, + "loss": 0.6944, + "step": 23843 + }, + { + "epoch": 0.57, + "learning_rate": 8.119332032633933e-06, + "loss": 1.2172, + "step": 23844 + }, + { + "epoch": 0.57, + "learning_rate": 8.11856704234887e-06, + "loss": 0.9136, + "step": 23845 + }, + { + "epoch": 0.57, + "learning_rate": 8.117802063478019e-06, + "loss": 1.1839, + "step": 23846 + }, + { + "epoch": 0.57, + "learning_rate": 8.117037096026018e-06, + "loss": 1.1971, + "step": 23847 + }, + { + "epoch": 0.57, + "learning_rate": 8.116272139997511e-06, + "loss": 1.0073, + "step": 23848 + }, + { + "epoch": 0.57, + "learning_rate": 8.11550719539714e-06, + "loss": 1.0563, + "step": 23849 + }, + { + "epoch": 0.57, + "learning_rate": 8.114742262229542e-06, + "loss": 1.0232, + "step": 23850 + }, + { + "epoch": 0.57, + "learning_rate": 8.113977340499358e-06, + "loss": 1.049, + "step": 23851 + }, + { + "epoch": 0.57, + "learning_rate": 8.113212430211232e-06, + "loss": 1.12, + "step": 23852 + }, + { + "epoch": 0.57, + "learning_rate": 8.112447531369803e-06, + "loss": 1.0436, + "step": 23853 + }, + { + "epoch": 0.57, + "learning_rate": 8.111682643979707e-06, + "loss": 1.0476, + "step": 23854 + }, + { + "epoch": 0.57, + "learning_rate": 8.110917768045592e-06, + "loss": 0.897, + "step": 23855 + }, + { + "epoch": 0.57, + "learning_rate": 8.110152903572094e-06, + "loss": 1.0849, + "step": 23856 + }, + { + "epoch": 0.57, + "learning_rate": 8.109388050563853e-06, + "loss": 1.126, + "step": 23857 + }, + { + "epoch": 0.57, + "learning_rate": 8.108623209025507e-06, + "loss": 1.274, + "step": 23858 + }, + { + "epoch": 0.57, + "learning_rate": 8.107858378961705e-06, + "loss": 1.1146, + "step": 23859 + }, + { + "epoch": 0.57, + "learning_rate": 8.107093560377079e-06, + "loss": 1.2562, + "step": 23860 + }, + { + "epoch": 0.57, + "learning_rate": 8.10632875327627e-06, + "loss": 0.8763, + "step": 23861 + }, + { + "epoch": 0.57, + "learning_rate": 8.105563957663918e-06, + "loss": 1.0864, + "step": 23862 + }, + { + "epoch": 0.57, + "learning_rate": 8.104799173544667e-06, + "loss": 1.1273, + "step": 23863 + }, + { + "epoch": 0.57, + "learning_rate": 8.104034400923153e-06, + "loss": 1.0088, + "step": 23864 + }, + { + "epoch": 0.57, + "learning_rate": 8.103269639804015e-06, + "loss": 1.0843, + "step": 23865 + }, + { + "epoch": 0.57, + "learning_rate": 8.102504890191892e-06, + "loss": 1.0223, + "step": 23866 + }, + { + "epoch": 0.57, + "learning_rate": 8.101740152091429e-06, + "loss": 1.0072, + "step": 23867 + }, + { + "epoch": 0.57, + "learning_rate": 8.100975425507261e-06, + "loss": 1.1925, + "step": 23868 + }, + { + "epoch": 0.57, + "learning_rate": 8.100210710444029e-06, + "loss": 0.9599, + "step": 23869 + }, + { + "epoch": 0.57, + "learning_rate": 8.09944600690637e-06, + "loss": 1.0312, + "step": 23870 + }, + { + "epoch": 0.57, + "learning_rate": 8.098681314898927e-06, + "loss": 1.3502, + "step": 23871 + }, + { + "epoch": 0.57, + "learning_rate": 8.097916634426335e-06, + "loss": 1.205, + "step": 23872 + }, + { + "epoch": 0.57, + "learning_rate": 8.097151965493238e-06, + "loss": 1.3399, + "step": 23873 + }, + { + "epoch": 0.57, + "learning_rate": 8.096387308104269e-06, + "loss": 1.0985, + "step": 23874 + }, + { + "epoch": 0.57, + "learning_rate": 8.095622662264073e-06, + "loss": 1.2137, + "step": 23875 + }, + { + "epoch": 0.57, + "learning_rate": 8.094858027977288e-06, + "loss": 1.2422, + "step": 23876 + }, + { + "epoch": 0.57, + "learning_rate": 8.094093405248548e-06, + "loss": 1.1164, + "step": 23877 + }, + { + "epoch": 0.57, + "learning_rate": 8.093328794082496e-06, + "loss": 1.1149, + "step": 23878 + }, + { + "epoch": 0.57, + "learning_rate": 8.09256419448377e-06, + "loss": 0.9228, + "step": 23879 + }, + { + "epoch": 0.57, + "learning_rate": 8.09179960645701e-06, + "loss": 1.1554, + "step": 23880 + }, + { + "epoch": 0.57, + "learning_rate": 8.09103503000685e-06, + "loss": 0.7951, + "step": 23881 + }, + { + "epoch": 0.57, + "learning_rate": 8.090270465137932e-06, + "loss": 0.9079, + "step": 23882 + }, + { + "epoch": 0.57, + "learning_rate": 8.089505911854894e-06, + "loss": 1.2292, + "step": 23883 + }, + { + "epoch": 0.57, + "learning_rate": 8.088741370162376e-06, + "loss": 0.7754, + "step": 23884 + }, + { + "epoch": 0.57, + "learning_rate": 8.087976840065014e-06, + "loss": 1.2268, + "step": 23885 + }, + { + "epoch": 0.57, + "learning_rate": 8.087212321567445e-06, + "loss": 1.1389, + "step": 23886 + }, + { + "epoch": 0.57, + "learning_rate": 8.08644781467431e-06, + "loss": 0.9485, + "step": 23887 + }, + { + "epoch": 0.57, + "learning_rate": 8.085683319390247e-06, + "loss": 1.3797, + "step": 23888 + }, + { + "epoch": 0.57, + "learning_rate": 8.084918835719893e-06, + "loss": 0.8816, + "step": 23889 + }, + { + "epoch": 0.57, + "learning_rate": 8.084154363667884e-06, + "loss": 1.3453, + "step": 23890 + }, + { + "epoch": 0.57, + "learning_rate": 8.083389903238861e-06, + "loss": 1.1491, + "step": 23891 + }, + { + "epoch": 0.57, + "learning_rate": 8.08262545443746e-06, + "loss": 1.0264, + "step": 23892 + }, + { + "epoch": 0.57, + "learning_rate": 8.081861017268321e-06, + "loss": 1.1372, + "step": 23893 + }, + { + "epoch": 0.57, + "learning_rate": 8.081096591736077e-06, + "loss": 0.9217, + "step": 23894 + }, + { + "epoch": 0.57, + "learning_rate": 8.08033217784537e-06, + "loss": 0.86, + "step": 23895 + }, + { + "epoch": 0.57, + "learning_rate": 8.079567775600839e-06, + "loss": 0.9552, + "step": 23896 + }, + { + "epoch": 0.57, + "learning_rate": 8.078803385007116e-06, + "loss": 1.1647, + "step": 23897 + }, + { + "epoch": 0.57, + "learning_rate": 8.07803900606884e-06, + "loss": 1.1942, + "step": 23898 + }, + { + "epoch": 0.57, + "learning_rate": 8.077274638790651e-06, + "loss": 0.8845, + "step": 23899 + }, + { + "epoch": 0.57, + "learning_rate": 8.076510283177186e-06, + "loss": 1.0, + "step": 23900 + }, + { + "epoch": 0.57, + "learning_rate": 8.075745939233078e-06, + "loss": 0.9364, + "step": 23901 + }, + { + "epoch": 0.57, + "learning_rate": 8.074981606962967e-06, + "loss": 1.2475, + "step": 23902 + }, + { + "epoch": 0.57, + "learning_rate": 8.07421728637149e-06, + "loss": 1.3185, + "step": 23903 + }, + { + "epoch": 0.57, + "learning_rate": 8.073452977463285e-06, + "loss": 1.1632, + "step": 23904 + }, + { + "epoch": 0.57, + "learning_rate": 8.072688680242985e-06, + "loss": 0.8648, + "step": 23905 + }, + { + "epoch": 0.57, + "learning_rate": 8.071924394715233e-06, + "loss": 0.9986, + "step": 23906 + }, + { + "epoch": 0.57, + "learning_rate": 8.07116012088466e-06, + "loss": 0.8593, + "step": 23907 + }, + { + "epoch": 0.57, + "learning_rate": 8.070395858755906e-06, + "loss": 1.1151, + "step": 23908 + }, + { + "epoch": 0.57, + "learning_rate": 8.069631608333604e-06, + "loss": 1.2832, + "step": 23909 + }, + { + "epoch": 0.58, + "learning_rate": 8.068867369622394e-06, + "loss": 0.9408, + "step": 23910 + }, + { + "epoch": 0.58, + "learning_rate": 8.068103142626914e-06, + "loss": 0.9334, + "step": 23911 + }, + { + "epoch": 0.58, + "learning_rate": 8.067338927351797e-06, + "loss": 1.0369, + "step": 23912 + }, + { + "epoch": 0.58, + "learning_rate": 8.066574723801677e-06, + "loss": 0.9996, + "step": 23913 + }, + { + "epoch": 0.58, + "learning_rate": 8.065810531981196e-06, + "loss": 1.0537, + "step": 23914 + }, + { + "epoch": 0.58, + "learning_rate": 8.065046351894987e-06, + "loss": 0.8699, + "step": 23915 + }, + { + "epoch": 0.58, + "learning_rate": 8.064282183547688e-06, + "loss": 1.1064, + "step": 23916 + }, + { + "epoch": 0.58, + "learning_rate": 8.063518026943931e-06, + "loss": 1.1815, + "step": 23917 + }, + { + "epoch": 0.58, + "learning_rate": 8.062753882088356e-06, + "loss": 1.076, + "step": 23918 + }, + { + "epoch": 0.58, + "learning_rate": 8.061989748985599e-06, + "loss": 0.7803, + "step": 23919 + }, + { + "epoch": 0.58, + "learning_rate": 8.061225627640294e-06, + "loss": 1.1152, + "step": 23920 + }, + { + "epoch": 0.58, + "learning_rate": 8.060461518057074e-06, + "loss": 0.8567, + "step": 23921 + }, + { + "epoch": 0.58, + "learning_rate": 8.059697420240579e-06, + "loss": 1.2068, + "step": 23922 + }, + { + "epoch": 0.58, + "learning_rate": 8.058933334195445e-06, + "loss": 0.9769, + "step": 23923 + }, + { + "epoch": 0.58, + "learning_rate": 8.058169259926306e-06, + "loss": 1.2143, + "step": 23924 + }, + { + "epoch": 0.58, + "learning_rate": 8.057405197437793e-06, + "loss": 1.1538, + "step": 23925 + }, + { + "epoch": 0.58, + "learning_rate": 8.056641146734552e-06, + "loss": 0.9723, + "step": 23926 + }, + { + "epoch": 0.58, + "learning_rate": 8.055877107821208e-06, + "loss": 1.22, + "step": 23927 + }, + { + "epoch": 0.58, + "learning_rate": 8.055113080702403e-06, + "loss": 1.0964, + "step": 23928 + }, + { + "epoch": 0.58, + "learning_rate": 8.054349065382765e-06, + "loss": 1.1478, + "step": 23929 + }, + { + "epoch": 0.58, + "learning_rate": 8.053585061866937e-06, + "loss": 0.9186, + "step": 23930 + }, + { + "epoch": 0.58, + "learning_rate": 8.052821070159552e-06, + "loss": 0.8488, + "step": 23931 + }, + { + "epoch": 0.58, + "learning_rate": 8.052057090265241e-06, + "loss": 1.2079, + "step": 23932 + }, + { + "epoch": 0.58, + "learning_rate": 8.05129312218864e-06, + "loss": 0.9197, + "step": 23933 + }, + { + "epoch": 0.58, + "learning_rate": 8.05052916593439e-06, + "loss": 1.1554, + "step": 23934 + }, + { + "epoch": 0.58, + "learning_rate": 8.049765221507118e-06, + "loss": 0.7756, + "step": 23935 + }, + { + "epoch": 0.58, + "learning_rate": 8.049001288911462e-06, + "loss": 1.2553, + "step": 23936 + }, + { + "epoch": 0.58, + "learning_rate": 8.048237368152055e-06, + "loss": 1.0779, + "step": 23937 + }, + { + "epoch": 0.58, + "learning_rate": 8.047473459233535e-06, + "loss": 1.0403, + "step": 23938 + }, + { + "epoch": 0.58, + "learning_rate": 8.046709562160534e-06, + "loss": 1.0141, + "step": 23939 + }, + { + "epoch": 0.58, + "learning_rate": 8.045945676937688e-06, + "loss": 1.1038, + "step": 23940 + }, + { + "epoch": 0.58, + "learning_rate": 8.045181803569625e-06, + "loss": 0.9205, + "step": 23941 + }, + { + "epoch": 0.58, + "learning_rate": 8.044417942060988e-06, + "loss": 1.2206, + "step": 23942 + }, + { + "epoch": 0.58, + "learning_rate": 8.043654092416407e-06, + "loss": 1.0461, + "step": 23943 + }, + { + "epoch": 0.58, + "learning_rate": 8.042890254640515e-06, + "loss": 1.0173, + "step": 23944 + }, + { + "epoch": 0.58, + "learning_rate": 8.042126428737949e-06, + "loss": 1.0368, + "step": 23945 + }, + { + "epoch": 0.58, + "learning_rate": 8.041362614713341e-06, + "loss": 1.0827, + "step": 23946 + }, + { + "epoch": 0.58, + "learning_rate": 8.040598812571327e-06, + "loss": 1.0915, + "step": 23947 + }, + { + "epoch": 0.58, + "learning_rate": 8.039835022316534e-06, + "loss": 1.1016, + "step": 23948 + }, + { + "epoch": 0.58, + "learning_rate": 8.039071243953606e-06, + "loss": 0.9801, + "step": 23949 + }, + { + "epoch": 0.58, + "learning_rate": 8.03830747748717e-06, + "loss": 1.0602, + "step": 23950 + }, + { + "epoch": 0.58, + "learning_rate": 8.037543722921861e-06, + "loss": 1.186, + "step": 23951 + }, + { + "epoch": 0.58, + "learning_rate": 8.036779980262312e-06, + "loss": 0.9019, + "step": 23952 + }, + { + "epoch": 0.58, + "learning_rate": 8.036016249513158e-06, + "loss": 1.1814, + "step": 23953 + }, + { + "epoch": 0.58, + "learning_rate": 8.035252530679031e-06, + "loss": 1.1311, + "step": 23954 + }, + { + "epoch": 0.58, + "learning_rate": 8.034488823764566e-06, + "loss": 0.9131, + "step": 23955 + }, + { + "epoch": 0.58, + "learning_rate": 8.033725128774391e-06, + "loss": 1.0791, + "step": 23956 + }, + { + "epoch": 0.58, + "learning_rate": 8.032961445713147e-06, + "loss": 1.0347, + "step": 23957 + }, + { + "epoch": 0.58, + "learning_rate": 8.032197774585463e-06, + "loss": 1.0425, + "step": 23958 + }, + { + "epoch": 0.58, + "learning_rate": 8.031434115395971e-06, + "loss": 1.2166, + "step": 23959 + }, + { + "epoch": 0.58, + "learning_rate": 8.030670468149305e-06, + "loss": 1.1107, + "step": 23960 + }, + { + "epoch": 0.58, + "learning_rate": 8.029906832850098e-06, + "loss": 1.2641, + "step": 23961 + }, + { + "epoch": 0.58, + "learning_rate": 8.029143209502985e-06, + "loss": 1.0868, + "step": 23962 + }, + { + "epoch": 0.58, + "learning_rate": 8.028379598112595e-06, + "loss": 1.1525, + "step": 23963 + }, + { + "epoch": 0.58, + "learning_rate": 8.027615998683561e-06, + "loss": 0.7306, + "step": 23964 + }, + { + "epoch": 0.58, + "learning_rate": 8.02685241122052e-06, + "loss": 1.1296, + "step": 23965 + }, + { + "epoch": 0.58, + "learning_rate": 8.026088835728098e-06, + "loss": 1.1488, + "step": 23966 + }, + { + "epoch": 0.58, + "learning_rate": 8.025325272210933e-06, + "loss": 1.1437, + "step": 23967 + }, + { + "epoch": 0.58, + "learning_rate": 8.024561720673654e-06, + "loss": 1.1119, + "step": 23968 + }, + { + "epoch": 0.58, + "learning_rate": 8.023798181120895e-06, + "loss": 1.1308, + "step": 23969 + }, + { + "epoch": 0.58, + "learning_rate": 8.023034653557287e-06, + "loss": 1.0358, + "step": 23970 + }, + { + "epoch": 0.58, + "learning_rate": 8.022271137987464e-06, + "loss": 1.139, + "step": 23971 + }, + { + "epoch": 0.58, + "learning_rate": 8.021507634416055e-06, + "loss": 1.1282, + "step": 23972 + }, + { + "epoch": 0.58, + "learning_rate": 8.020744142847693e-06, + "loss": 0.9615, + "step": 23973 + }, + { + "epoch": 0.58, + "learning_rate": 8.019980663287015e-06, + "loss": 1.3414, + "step": 23974 + }, + { + "epoch": 0.58, + "learning_rate": 8.019217195738647e-06, + "loss": 1.0699, + "step": 23975 + }, + { + "epoch": 0.58, + "learning_rate": 8.01845374020722e-06, + "loss": 1.3418, + "step": 23976 + }, + { + "epoch": 0.58, + "learning_rate": 8.01769029669737e-06, + "loss": 1.157, + "step": 23977 + }, + { + "epoch": 0.58, + "learning_rate": 8.016926865213726e-06, + "loss": 1.1834, + "step": 23978 + }, + { + "epoch": 0.58, + "learning_rate": 8.016163445760921e-06, + "loss": 1.1307, + "step": 23979 + }, + { + "epoch": 0.58, + "learning_rate": 8.015400038343582e-06, + "loss": 0.9739, + "step": 23980 + }, + { + "epoch": 0.58, + "learning_rate": 8.014636642966349e-06, + "loss": 1.2162, + "step": 23981 + }, + { + "epoch": 0.58, + "learning_rate": 8.013873259633846e-06, + "loss": 0.9655, + "step": 23982 + }, + { + "epoch": 0.58, + "learning_rate": 8.01310988835071e-06, + "loss": 0.7623, + "step": 23983 + }, + { + "epoch": 0.58, + "learning_rate": 8.012346529121564e-06, + "loss": 1.2846, + "step": 23984 + }, + { + "epoch": 0.58, + "learning_rate": 8.011583181951047e-06, + "loss": 0.7791, + "step": 23985 + }, + { + "epoch": 0.58, + "learning_rate": 8.010819846843788e-06, + "loss": 0.6965, + "step": 23986 + }, + { + "epoch": 0.58, + "learning_rate": 8.010056523804417e-06, + "loss": 1.0796, + "step": 23987 + }, + { + "epoch": 0.58, + "learning_rate": 8.00929321283756e-06, + "loss": 1.2753, + "step": 23988 + }, + { + "epoch": 0.58, + "learning_rate": 8.008529913947858e-06, + "loss": 1.1561, + "step": 23989 + }, + { + "epoch": 0.58, + "learning_rate": 8.007766627139935e-06, + "loss": 1.0256, + "step": 23990 + }, + { + "epoch": 0.58, + "learning_rate": 8.007003352418423e-06, + "loss": 0.9801, + "step": 23991 + }, + { + "epoch": 0.58, + "learning_rate": 8.006240089787953e-06, + "loss": 1.1735, + "step": 23992 + }, + { + "epoch": 0.58, + "learning_rate": 8.005476839253157e-06, + "loss": 0.8344, + "step": 23993 + }, + { + "epoch": 0.58, + "learning_rate": 8.004713600818662e-06, + "loss": 1.1418, + "step": 23994 + }, + { + "epoch": 0.58, + "learning_rate": 8.003950374489102e-06, + "loss": 1.0174, + "step": 23995 + }, + { + "epoch": 0.58, + "learning_rate": 8.003187160269104e-06, + "loss": 1.0783, + "step": 23996 + }, + { + "epoch": 0.58, + "learning_rate": 8.002423958163302e-06, + "loss": 1.1037, + "step": 23997 + }, + { + "epoch": 0.58, + "learning_rate": 8.001660768176322e-06, + "loss": 0.8375, + "step": 23998 + }, + { + "epoch": 0.58, + "learning_rate": 8.000897590312792e-06, + "loss": 1.0226, + "step": 23999 + }, + { + "epoch": 0.58, + "learning_rate": 8.000134424577351e-06, + "loss": 1.0133, + "step": 24000 + }, + { + "epoch": 0.58, + "learning_rate": 7.999371270974623e-06, + "loss": 1.2198, + "step": 24001 + }, + { + "epoch": 0.58, + "learning_rate": 7.99860812950924e-06, + "loss": 1.0759, + "step": 24002 + }, + { + "epoch": 0.58, + "learning_rate": 7.99784500018583e-06, + "loss": 0.8641, + "step": 24003 + }, + { + "epoch": 0.58, + "learning_rate": 7.997081883009025e-06, + "loss": 1.1637, + "step": 24004 + }, + { + "epoch": 0.58, + "learning_rate": 7.996318777983449e-06, + "loss": 1.1026, + "step": 24005 + }, + { + "epoch": 0.58, + "learning_rate": 7.99555568511374e-06, + "loss": 1.095, + "step": 24006 + }, + { + "epoch": 0.58, + "learning_rate": 7.994792604404522e-06, + "loss": 1.1109, + "step": 24007 + }, + { + "epoch": 0.58, + "learning_rate": 7.994029535860426e-06, + "loss": 1.1075, + "step": 24008 + }, + { + "epoch": 0.58, + "learning_rate": 7.993266479486079e-06, + "loss": 0.8877, + "step": 24009 + }, + { + "epoch": 0.58, + "learning_rate": 7.992503435286113e-06, + "loss": 1.1198, + "step": 24010 + }, + { + "epoch": 0.58, + "learning_rate": 7.991740403265159e-06, + "loss": 1.1697, + "step": 24011 + }, + { + "epoch": 0.58, + "learning_rate": 7.99097738342784e-06, + "loss": 1.1351, + "step": 24012 + }, + { + "epoch": 0.58, + "learning_rate": 7.99021437577879e-06, + "loss": 1.0589, + "step": 24013 + }, + { + "epoch": 0.58, + "learning_rate": 7.989451380322636e-06, + "loss": 0.8817, + "step": 24014 + }, + { + "epoch": 0.58, + "learning_rate": 7.98868839706401e-06, + "loss": 1.0695, + "step": 24015 + }, + { + "epoch": 0.58, + "learning_rate": 7.987925426007537e-06, + "loss": 1.1202, + "step": 24016 + }, + { + "epoch": 0.58, + "learning_rate": 7.987162467157845e-06, + "loss": 1.1832, + "step": 24017 + }, + { + "epoch": 0.58, + "learning_rate": 7.986399520519567e-06, + "loss": 0.9987, + "step": 24018 + }, + { + "epoch": 0.58, + "learning_rate": 7.985636586097331e-06, + "loss": 1.0884, + "step": 24019 + }, + { + "epoch": 0.58, + "learning_rate": 7.984873663895762e-06, + "loss": 1.0927, + "step": 24020 + }, + { + "epoch": 0.58, + "learning_rate": 7.98411075391949e-06, + "loss": 0.742, + "step": 24021 + }, + { + "epoch": 0.58, + "learning_rate": 7.983347856173146e-06, + "loss": 0.8057, + "step": 24022 + }, + { + "epoch": 0.58, + "learning_rate": 7.982584970661355e-06, + "loss": 1.1538, + "step": 24023 + }, + { + "epoch": 0.58, + "learning_rate": 7.981822097388745e-06, + "loss": 1.155, + "step": 24024 + }, + { + "epoch": 0.58, + "learning_rate": 7.981059236359946e-06, + "loss": 1.1319, + "step": 24025 + }, + { + "epoch": 0.58, + "learning_rate": 7.98029638757959e-06, + "loss": 0.8494, + "step": 24026 + }, + { + "epoch": 0.58, + "learning_rate": 7.979533551052296e-06, + "loss": 1.1425, + "step": 24027 + }, + { + "epoch": 0.58, + "learning_rate": 7.978770726782697e-06, + "loss": 1.1343, + "step": 24028 + }, + { + "epoch": 0.58, + "learning_rate": 7.978007914775423e-06, + "loss": 1.0258, + "step": 24029 + }, + { + "epoch": 0.58, + "learning_rate": 7.9772451150351e-06, + "loss": 1.0833, + "step": 24030 + }, + { + "epoch": 0.58, + "learning_rate": 7.976482327566353e-06, + "loss": 1.0318, + "step": 24031 + }, + { + "epoch": 0.58, + "learning_rate": 7.975719552373811e-06, + "loss": 0.8618, + "step": 24032 + }, + { + "epoch": 0.58, + "learning_rate": 7.974956789462102e-06, + "loss": 0.9781, + "step": 24033 + }, + { + "epoch": 0.58, + "learning_rate": 7.974194038835857e-06, + "loss": 1.2272, + "step": 24034 + }, + { + "epoch": 0.58, + "learning_rate": 7.973431300499699e-06, + "loss": 1.1044, + "step": 24035 + }, + { + "epoch": 0.58, + "learning_rate": 7.972668574458255e-06, + "loss": 1.1705, + "step": 24036 + }, + { + "epoch": 0.58, + "learning_rate": 7.971905860716154e-06, + "loss": 1.1902, + "step": 24037 + }, + { + "epoch": 0.58, + "learning_rate": 7.971143159278025e-06, + "loss": 1.0959, + "step": 24038 + }, + { + "epoch": 0.58, + "learning_rate": 7.970380470148493e-06, + "loss": 1.1054, + "step": 24039 + }, + { + "epoch": 0.58, + "learning_rate": 7.969617793332184e-06, + "loss": 0.9378, + "step": 24040 + }, + { + "epoch": 0.58, + "learning_rate": 7.968855128833728e-06, + "loss": 1.2325, + "step": 24041 + }, + { + "epoch": 0.58, + "learning_rate": 7.96809247665775e-06, + "loss": 1.1535, + "step": 24042 + }, + { + "epoch": 0.58, + "learning_rate": 7.967329836808875e-06, + "loss": 1.0581, + "step": 24043 + }, + { + "epoch": 0.58, + "learning_rate": 7.966567209291732e-06, + "loss": 1.1538, + "step": 24044 + }, + { + "epoch": 0.58, + "learning_rate": 7.965804594110949e-06, + "loss": 1.065, + "step": 24045 + }, + { + "epoch": 0.58, + "learning_rate": 7.96504199127115e-06, + "loss": 1.0204, + "step": 24046 + }, + { + "epoch": 0.58, + "learning_rate": 7.964279400776965e-06, + "loss": 1.1406, + "step": 24047 + }, + { + "epoch": 0.58, + "learning_rate": 7.963516822633014e-06, + "loss": 1.0632, + "step": 24048 + }, + { + "epoch": 0.58, + "learning_rate": 7.96275425684393e-06, + "loss": 1.1355, + "step": 24049 + }, + { + "epoch": 0.58, + "learning_rate": 7.961991703414337e-06, + "loss": 0.9907, + "step": 24050 + }, + { + "epoch": 0.58, + "learning_rate": 7.96122916234886e-06, + "loss": 0.8872, + "step": 24051 + }, + { + "epoch": 0.58, + "learning_rate": 7.960466633652125e-06, + "loss": 1.1473, + "step": 24052 + }, + { + "epoch": 0.58, + "learning_rate": 7.959704117328761e-06, + "loss": 0.9571, + "step": 24053 + }, + { + "epoch": 0.58, + "learning_rate": 7.958941613383391e-06, + "loss": 0.9335, + "step": 24054 + }, + { + "epoch": 0.58, + "learning_rate": 7.958179121820645e-06, + "loss": 1.0564, + "step": 24055 + }, + { + "epoch": 0.58, + "learning_rate": 7.957416642645142e-06, + "loss": 1.0294, + "step": 24056 + }, + { + "epoch": 0.58, + "learning_rate": 7.956654175861513e-06, + "loss": 0.8094, + "step": 24057 + }, + { + "epoch": 0.58, + "learning_rate": 7.955891721474385e-06, + "loss": 0.9309, + "step": 24058 + }, + { + "epoch": 0.58, + "learning_rate": 7.95512927948838e-06, + "loss": 1.0024, + "step": 24059 + }, + { + "epoch": 0.58, + "learning_rate": 7.954366849908121e-06, + "loss": 1.065, + "step": 24060 + }, + { + "epoch": 0.58, + "learning_rate": 7.953604432738242e-06, + "loss": 1.1469, + "step": 24061 + }, + { + "epoch": 0.58, + "learning_rate": 7.952842027983362e-06, + "loss": 0.9924, + "step": 24062 + }, + { + "epoch": 0.58, + "learning_rate": 7.952079635648109e-06, + "loss": 1.0414, + "step": 24063 + }, + { + "epoch": 0.58, + "learning_rate": 7.951317255737103e-06, + "loss": 1.2143, + "step": 24064 + }, + { + "epoch": 0.58, + "learning_rate": 7.950554888254977e-06, + "loss": 1.1244, + "step": 24065 + }, + { + "epoch": 0.58, + "learning_rate": 7.949792533206352e-06, + "loss": 0.923, + "step": 24066 + }, + { + "epoch": 0.58, + "learning_rate": 7.949030190595854e-06, + "loss": 1.1549, + "step": 24067 + }, + { + "epoch": 0.58, + "learning_rate": 7.948267860428104e-06, + "loss": 1.1402, + "step": 24068 + }, + { + "epoch": 0.58, + "learning_rate": 7.947505542707733e-06, + "loss": 1.0249, + "step": 24069 + }, + { + "epoch": 0.58, + "learning_rate": 7.946743237439363e-06, + "loss": 1.3561, + "step": 24070 + }, + { + "epoch": 0.58, + "learning_rate": 7.945980944627618e-06, + "loss": 1.2068, + "step": 24071 + }, + { + "epoch": 0.58, + "learning_rate": 7.945218664277123e-06, + "loss": 1.2575, + "step": 24072 + }, + { + "epoch": 0.58, + "learning_rate": 7.944456396392506e-06, + "loss": 1.1732, + "step": 24073 + }, + { + "epoch": 0.58, + "learning_rate": 7.943694140978387e-06, + "loss": 1.3718, + "step": 24074 + }, + { + "epoch": 0.58, + "learning_rate": 7.94293189803939e-06, + "loss": 1.001, + "step": 24075 + }, + { + "epoch": 0.58, + "learning_rate": 7.942169667580142e-06, + "loss": 0.952, + "step": 24076 + }, + { + "epoch": 0.58, + "learning_rate": 7.94140744960527e-06, + "loss": 0.9136, + "step": 24077 + }, + { + "epoch": 0.58, + "learning_rate": 7.940645244119391e-06, + "loss": 1.3832, + "step": 24078 + }, + { + "epoch": 0.58, + "learning_rate": 7.939883051127132e-06, + "loss": 0.7452, + "step": 24079 + }, + { + "epoch": 0.58, + "learning_rate": 7.939120870633123e-06, + "loss": 1.1237, + "step": 24080 + }, + { + "epoch": 0.58, + "learning_rate": 7.93835870264198e-06, + "loss": 1.1506, + "step": 24081 + }, + { + "epoch": 0.58, + "learning_rate": 7.937596547158331e-06, + "loss": 0.9769, + "step": 24082 + }, + { + "epoch": 0.58, + "learning_rate": 7.936834404186795e-06, + "loss": 0.9615, + "step": 24083 + }, + { + "epoch": 0.58, + "learning_rate": 7.936072273732002e-06, + "loss": 1.2532, + "step": 24084 + }, + { + "epoch": 0.58, + "learning_rate": 7.935310155798574e-06, + "loss": 0.8288, + "step": 24085 + }, + { + "epoch": 0.58, + "learning_rate": 7.934548050391133e-06, + "loss": 1.1023, + "step": 24086 + }, + { + "epoch": 0.58, + "learning_rate": 7.933785957514301e-06, + "loss": 1.006, + "step": 24087 + }, + { + "epoch": 0.58, + "learning_rate": 7.933023877172707e-06, + "loss": 1.2292, + "step": 24088 + }, + { + "epoch": 0.58, + "learning_rate": 7.932261809370971e-06, + "loss": 1.0072, + "step": 24089 + }, + { + "epoch": 0.58, + "learning_rate": 7.931499754113713e-06, + "loss": 1.0749, + "step": 24090 + }, + { + "epoch": 0.58, + "learning_rate": 7.93073771140556e-06, + "loss": 0.8699, + "step": 24091 + }, + { + "epoch": 0.58, + "learning_rate": 7.929975681251136e-06, + "loss": 1.1057, + "step": 24092 + }, + { + "epoch": 0.58, + "learning_rate": 7.929213663655063e-06, + "loss": 1.1235, + "step": 24093 + }, + { + "epoch": 0.58, + "learning_rate": 7.928451658621964e-06, + "loss": 1.289, + "step": 24094 + }, + { + "epoch": 0.58, + "learning_rate": 7.927689666156458e-06, + "loss": 1.1001, + "step": 24095 + }, + { + "epoch": 0.58, + "learning_rate": 7.926927686263176e-06, + "loss": 0.9976, + "step": 24096 + }, + { + "epoch": 0.58, + "learning_rate": 7.926165718946734e-06, + "loss": 1.0005, + "step": 24097 + }, + { + "epoch": 0.58, + "learning_rate": 7.925403764211757e-06, + "loss": 1.0523, + "step": 24098 + }, + { + "epoch": 0.58, + "learning_rate": 7.924641822062866e-06, + "loss": 1.0568, + "step": 24099 + }, + { + "epoch": 0.58, + "learning_rate": 7.923879892504686e-06, + "loss": 0.9073, + "step": 24100 + }, + { + "epoch": 0.58, + "learning_rate": 7.923117975541839e-06, + "loss": 0.956, + "step": 24101 + }, + { + "epoch": 0.58, + "learning_rate": 7.922356071178945e-06, + "loss": 0.6569, + "step": 24102 + }, + { + "epoch": 0.58, + "learning_rate": 7.921594179420629e-06, + "loss": 1.1784, + "step": 24103 + }, + { + "epoch": 0.58, + "learning_rate": 7.920832300271512e-06, + "loss": 1.1158, + "step": 24104 + }, + { + "epoch": 0.58, + "learning_rate": 7.920070433736218e-06, + "loss": 1.259, + "step": 24105 + }, + { + "epoch": 0.58, + "learning_rate": 7.919308579819368e-06, + "loss": 1.0469, + "step": 24106 + }, + { + "epoch": 0.58, + "learning_rate": 7.918546738525578e-06, + "loss": 1.3836, + "step": 24107 + }, + { + "epoch": 0.58, + "learning_rate": 7.91778490985948e-06, + "loss": 1.296, + "step": 24108 + }, + { + "epoch": 0.58, + "learning_rate": 7.91702309382569e-06, + "loss": 1.0061, + "step": 24109 + }, + { + "epoch": 0.58, + "learning_rate": 7.916261290428833e-06, + "loss": 0.8344, + "step": 24110 + }, + { + "epoch": 0.58, + "learning_rate": 7.915499499673524e-06, + "loss": 1.2418, + "step": 24111 + }, + { + "epoch": 0.58, + "learning_rate": 7.914737721564393e-06, + "loss": 1.0673, + "step": 24112 + }, + { + "epoch": 0.58, + "learning_rate": 7.913975956106057e-06, + "loss": 0.9872, + "step": 24113 + }, + { + "epoch": 0.58, + "learning_rate": 7.91321420330314e-06, + "loss": 1.2028, + "step": 24114 + }, + { + "epoch": 0.58, + "learning_rate": 7.912452463160256e-06, + "loss": 1.0508, + "step": 24115 + }, + { + "epoch": 0.58, + "learning_rate": 7.911690735682037e-06, + "loss": 0.7652, + "step": 24116 + }, + { + "epoch": 0.58, + "learning_rate": 7.910929020873098e-06, + "loss": 1.08, + "step": 24117 + }, + { + "epoch": 0.58, + "learning_rate": 7.910167318738058e-06, + "loss": 0.8775, + "step": 24118 + }, + { + "epoch": 0.58, + "learning_rate": 7.909405629281545e-06, + "loss": 0.7913, + "step": 24119 + }, + { + "epoch": 0.58, + "learning_rate": 7.908643952508177e-06, + "loss": 1.002, + "step": 24120 + }, + { + "epoch": 0.58, + "learning_rate": 7.907882288422572e-06, + "loss": 1.2827, + "step": 24121 + }, + { + "epoch": 0.58, + "learning_rate": 7.907120637029353e-06, + "loss": 1.1492, + "step": 24122 + }, + { + "epoch": 0.58, + "learning_rate": 7.906358998333141e-06, + "loss": 0.7349, + "step": 24123 + }, + { + "epoch": 0.58, + "learning_rate": 7.905597372338558e-06, + "loss": 1.1975, + "step": 24124 + }, + { + "epoch": 0.58, + "learning_rate": 7.904835759050222e-06, + "loss": 0.9557, + "step": 24125 + }, + { + "epoch": 0.58, + "learning_rate": 7.904074158472754e-06, + "loss": 1.0737, + "step": 24126 + }, + { + "epoch": 0.58, + "learning_rate": 7.903312570610776e-06, + "loss": 1.0297, + "step": 24127 + }, + { + "epoch": 0.58, + "learning_rate": 7.902550995468908e-06, + "loss": 1.1411, + "step": 24128 + }, + { + "epoch": 0.58, + "learning_rate": 7.90178943305177e-06, + "loss": 1.122, + "step": 24129 + }, + { + "epoch": 0.58, + "learning_rate": 7.90102788336398e-06, + "loss": 0.8919, + "step": 24130 + }, + { + "epoch": 0.58, + "learning_rate": 7.900266346410161e-06, + "loss": 1.085, + "step": 24131 + }, + { + "epoch": 0.58, + "learning_rate": 7.899504822194933e-06, + "loss": 1.1444, + "step": 24132 + }, + { + "epoch": 0.58, + "learning_rate": 7.898743310722916e-06, + "loss": 1.0913, + "step": 24133 + }, + { + "epoch": 0.58, + "learning_rate": 7.897981811998725e-06, + "loss": 0.9651, + "step": 24134 + }, + { + "epoch": 0.58, + "learning_rate": 7.897220326026987e-06, + "loss": 0.7353, + "step": 24135 + }, + { + "epoch": 0.58, + "learning_rate": 7.896458852812319e-06, + "loss": 1.201, + "step": 24136 + }, + { + "epoch": 0.58, + "learning_rate": 7.895697392359339e-06, + "loss": 1.1814, + "step": 24137 + }, + { + "epoch": 0.58, + "learning_rate": 7.894935944672668e-06, + "loss": 0.9956, + "step": 24138 + }, + { + "epoch": 0.58, + "learning_rate": 7.894174509756927e-06, + "loss": 1.1996, + "step": 24139 + }, + { + "epoch": 0.58, + "learning_rate": 7.893413087616733e-06, + "loss": 1.211, + "step": 24140 + }, + { + "epoch": 0.58, + "learning_rate": 7.892651678256707e-06, + "loss": 1.1116, + "step": 24141 + }, + { + "epoch": 0.58, + "learning_rate": 7.891890281681465e-06, + "loss": 1.1322, + "step": 24142 + }, + { + "epoch": 0.58, + "learning_rate": 7.89112889789563e-06, + "loss": 1.2414, + "step": 24143 + }, + { + "epoch": 0.58, + "learning_rate": 7.890367526903822e-06, + "loss": 1.1437, + "step": 24144 + }, + { + "epoch": 0.58, + "learning_rate": 7.889606168710657e-06, + "loss": 1.0894, + "step": 24145 + }, + { + "epoch": 0.58, + "learning_rate": 7.888844823320753e-06, + "loss": 1.0223, + "step": 24146 + }, + { + "epoch": 0.58, + "learning_rate": 7.888083490738733e-06, + "loss": 0.879, + "step": 24147 + }, + { + "epoch": 0.58, + "learning_rate": 7.887322170969215e-06, + "loss": 1.1384, + "step": 24148 + }, + { + "epoch": 0.58, + "learning_rate": 7.886560864016814e-06, + "loss": 1.045, + "step": 24149 + }, + { + "epoch": 0.58, + "learning_rate": 7.88579956988615e-06, + "loss": 0.8997, + "step": 24150 + }, + { + "epoch": 0.58, + "learning_rate": 7.885038288581845e-06, + "loss": 0.9194, + "step": 24151 + }, + { + "epoch": 0.58, + "learning_rate": 7.884277020108515e-06, + "loss": 1.1948, + "step": 24152 + }, + { + "epoch": 0.58, + "learning_rate": 7.88351576447078e-06, + "loss": 1.2205, + "step": 24153 + }, + { + "epoch": 0.58, + "learning_rate": 7.882754521673253e-06, + "loss": 0.9649, + "step": 24154 + }, + { + "epoch": 0.58, + "learning_rate": 7.881993291720559e-06, + "loss": 1.3045, + "step": 24155 + }, + { + "epoch": 0.58, + "learning_rate": 7.881232074617313e-06, + "loss": 1.1652, + "step": 24156 + }, + { + "epoch": 0.58, + "learning_rate": 7.880470870368136e-06, + "loss": 1.0482, + "step": 24157 + }, + { + "epoch": 0.58, + "learning_rate": 7.879709678977638e-06, + "loss": 0.9693, + "step": 24158 + }, + { + "epoch": 0.58, + "learning_rate": 7.878948500450448e-06, + "loss": 1.1397, + "step": 24159 + }, + { + "epoch": 0.58, + "learning_rate": 7.878187334791176e-06, + "loss": 1.3725, + "step": 24160 + }, + { + "epoch": 0.58, + "learning_rate": 7.877426182004444e-06, + "loss": 1.1648, + "step": 24161 + }, + { + "epoch": 0.58, + "learning_rate": 7.876665042094867e-06, + "loss": 1.2113, + "step": 24162 + }, + { + "epoch": 0.58, + "learning_rate": 7.875903915067064e-06, + "loss": 1.1222, + "step": 24163 + }, + { + "epoch": 0.58, + "learning_rate": 7.875142800925652e-06, + "loss": 1.0027, + "step": 24164 + }, + { + "epoch": 0.58, + "learning_rate": 7.874381699675247e-06, + "loss": 0.9495, + "step": 24165 + }, + { + "epoch": 0.58, + "learning_rate": 7.87362061132047e-06, + "loss": 1.1787, + "step": 24166 + }, + { + "epoch": 0.58, + "learning_rate": 7.872859535865936e-06, + "loss": 1.0985, + "step": 24167 + }, + { + "epoch": 0.58, + "learning_rate": 7.872098473316263e-06, + "loss": 0.9988, + "step": 24168 + }, + { + "epoch": 0.58, + "learning_rate": 7.87133742367607e-06, + "loss": 0.9008, + "step": 24169 + }, + { + "epoch": 0.58, + "learning_rate": 7.870576386949971e-06, + "loss": 0.9452, + "step": 24170 + }, + { + "epoch": 0.58, + "learning_rate": 7.869815363142582e-06, + "loss": 1.1202, + "step": 24171 + }, + { + "epoch": 0.58, + "learning_rate": 7.869054352258526e-06, + "loss": 1.1971, + "step": 24172 + }, + { + "epoch": 0.58, + "learning_rate": 7.868293354302414e-06, + "loss": 0.8453, + "step": 24173 + }, + { + "epoch": 0.58, + "learning_rate": 7.867532369278867e-06, + "loss": 0.9662, + "step": 24174 + }, + { + "epoch": 0.58, + "learning_rate": 7.866771397192495e-06, + "loss": 1.223, + "step": 24175 + }, + { + "epoch": 0.58, + "learning_rate": 7.866010438047921e-06, + "loss": 1.1216, + "step": 24176 + }, + { + "epoch": 0.58, + "learning_rate": 7.865249491849762e-06, + "loss": 0.9519, + "step": 24177 + }, + { + "epoch": 0.58, + "learning_rate": 7.864488558602633e-06, + "loss": 0.9062, + "step": 24178 + }, + { + "epoch": 0.58, + "learning_rate": 7.863727638311145e-06, + "loss": 1.2068, + "step": 24179 + }, + { + "epoch": 0.58, + "learning_rate": 7.862966730979922e-06, + "loss": 1.0105, + "step": 24180 + }, + { + "epoch": 0.58, + "learning_rate": 7.862205836613577e-06, + "loss": 0.7715, + "step": 24181 + }, + { + "epoch": 0.58, + "learning_rate": 7.861444955216726e-06, + "loss": 1.0328, + "step": 24182 + }, + { + "epoch": 0.58, + "learning_rate": 7.860684086793984e-06, + "loss": 1.1896, + "step": 24183 + }, + { + "epoch": 0.58, + "learning_rate": 7.85992323134997e-06, + "loss": 1.1767, + "step": 24184 + }, + { + "epoch": 0.58, + "learning_rate": 7.8591623888893e-06, + "loss": 1.1609, + "step": 24185 + }, + { + "epoch": 0.58, + "learning_rate": 7.858401559416586e-06, + "loss": 1.0476, + "step": 24186 + }, + { + "epoch": 0.58, + "learning_rate": 7.857640742936444e-06, + "loss": 1.1714, + "step": 24187 + }, + { + "epoch": 0.58, + "learning_rate": 7.856879939453495e-06, + "loss": 1.2214, + "step": 24188 + }, + { + "epoch": 0.58, + "learning_rate": 7.856119148972351e-06, + "loss": 0.9067, + "step": 24189 + }, + { + "epoch": 0.58, + "learning_rate": 7.855358371497629e-06, + "loss": 1.1252, + "step": 24190 + }, + { + "epoch": 0.58, + "learning_rate": 7.854597607033939e-06, + "loss": 1.2112, + "step": 24191 + }, + { + "epoch": 0.58, + "learning_rate": 7.853836855585904e-06, + "loss": 1.1789, + "step": 24192 + }, + { + "epoch": 0.58, + "learning_rate": 7.853076117158135e-06, + "loss": 1.3544, + "step": 24193 + }, + { + "epoch": 0.58, + "learning_rate": 7.852315391755249e-06, + "loss": 1.0047, + "step": 24194 + }, + { + "epoch": 0.58, + "learning_rate": 7.85155467938186e-06, + "loss": 1.2189, + "step": 24195 + }, + { + "epoch": 0.58, + "learning_rate": 7.850793980042582e-06, + "loss": 0.8663, + "step": 24196 + }, + { + "epoch": 0.58, + "learning_rate": 7.850033293742034e-06, + "loss": 0.9473, + "step": 24197 + }, + { + "epoch": 0.58, + "learning_rate": 7.849272620484825e-06, + "loss": 0.7656, + "step": 24198 + }, + { + "epoch": 0.58, + "learning_rate": 7.848511960275575e-06, + "loss": 1.2033, + "step": 24199 + }, + { + "epoch": 0.58, + "learning_rate": 7.847751313118898e-06, + "loss": 1.0517, + "step": 24200 + }, + { + "epoch": 0.58, + "learning_rate": 7.846990679019406e-06, + "loss": 1.1713, + "step": 24201 + }, + { + "epoch": 0.58, + "learning_rate": 7.846230057981714e-06, + "loss": 1.3063, + "step": 24202 + }, + { + "epoch": 0.58, + "learning_rate": 7.84546945001044e-06, + "loss": 0.9716, + "step": 24203 + }, + { + "epoch": 0.58, + "learning_rate": 7.844708855110196e-06, + "loss": 1.0099, + "step": 24204 + }, + { + "epoch": 0.58, + "learning_rate": 7.843948273285596e-06, + "loss": 1.1707, + "step": 24205 + }, + { + "epoch": 0.58, + "learning_rate": 7.843187704541252e-06, + "loss": 0.9737, + "step": 24206 + }, + { + "epoch": 0.58, + "learning_rate": 7.842427148881785e-06, + "loss": 0.8995, + "step": 24207 + }, + { + "epoch": 0.58, + "learning_rate": 7.841666606311805e-06, + "loss": 0.7415, + "step": 24208 + }, + { + "epoch": 0.58, + "learning_rate": 7.840906076835923e-06, + "loss": 1.1844, + "step": 24209 + }, + { + "epoch": 0.58, + "learning_rate": 7.840145560458756e-06, + "loss": 1.1874, + "step": 24210 + }, + { + "epoch": 0.58, + "learning_rate": 7.839385057184921e-06, + "loss": 1.2302, + "step": 24211 + }, + { + "epoch": 0.58, + "learning_rate": 7.838624567019026e-06, + "loss": 0.997, + "step": 24212 + }, + { + "epoch": 0.58, + "learning_rate": 7.837864089965688e-06, + "loss": 1.029, + "step": 24213 + }, + { + "epoch": 0.58, + "learning_rate": 7.83710362602952e-06, + "loss": 1.1796, + "step": 24214 + }, + { + "epoch": 0.58, + "learning_rate": 7.836343175215136e-06, + "loss": 0.7869, + "step": 24215 + }, + { + "epoch": 0.58, + "learning_rate": 7.83558273752715e-06, + "loss": 1.0142, + "step": 24216 + }, + { + "epoch": 0.58, + "learning_rate": 7.834822312970173e-06, + "loss": 1.037, + "step": 24217 + }, + { + "epoch": 0.58, + "learning_rate": 7.834061901548818e-06, + "loss": 0.9842, + "step": 24218 + }, + { + "epoch": 0.58, + "learning_rate": 7.833301503267703e-06, + "loss": 1.0693, + "step": 24219 + }, + { + "epoch": 0.58, + "learning_rate": 7.832541118131437e-06, + "loss": 1.0364, + "step": 24220 + }, + { + "epoch": 0.58, + "learning_rate": 7.831780746144635e-06, + "loss": 1.3507, + "step": 24221 + }, + { + "epoch": 0.58, + "learning_rate": 7.831020387311907e-06, + "loss": 1.041, + "step": 24222 + }, + { + "epoch": 0.58, + "learning_rate": 7.83026004163787e-06, + "loss": 1.234, + "step": 24223 + }, + { + "epoch": 0.58, + "learning_rate": 7.829499709127134e-06, + "loss": 0.7147, + "step": 24224 + }, + { + "epoch": 0.58, + "learning_rate": 7.828739389784315e-06, + "loss": 1.1858, + "step": 24225 + }, + { + "epoch": 0.58, + "learning_rate": 7.82797908361402e-06, + "loss": 1.1369, + "step": 24226 + }, + { + "epoch": 0.58, + "learning_rate": 7.827218790620867e-06, + "loss": 1.0904, + "step": 24227 + }, + { + "epoch": 0.58, + "learning_rate": 7.826458510809467e-06, + "loss": 0.9533, + "step": 24228 + }, + { + "epoch": 0.58, + "learning_rate": 7.825698244184432e-06, + "loss": 0.9978, + "step": 24229 + }, + { + "epoch": 0.58, + "learning_rate": 7.824937990750372e-06, + "loss": 1.0977, + "step": 24230 + }, + { + "epoch": 0.58, + "learning_rate": 7.824177750511904e-06, + "loss": 0.6763, + "step": 24231 + }, + { + "epoch": 0.58, + "learning_rate": 7.823417523473638e-06, + "loss": 1.1211, + "step": 24232 + }, + { + "epoch": 0.58, + "learning_rate": 7.822657309640185e-06, + "loss": 1.0133, + "step": 24233 + }, + { + "epoch": 0.58, + "learning_rate": 7.821897109016156e-06, + "loss": 1.2101, + "step": 24234 + }, + { + "epoch": 0.58, + "learning_rate": 7.821136921606167e-06, + "loss": 1.2061, + "step": 24235 + }, + { + "epoch": 0.58, + "learning_rate": 7.820376747414829e-06, + "loss": 1.201, + "step": 24236 + }, + { + "epoch": 0.58, + "learning_rate": 7.819616586446752e-06, + "loss": 1.0653, + "step": 24237 + }, + { + "epoch": 0.58, + "learning_rate": 7.818856438706546e-06, + "loss": 0.9825, + "step": 24238 + }, + { + "epoch": 0.58, + "learning_rate": 7.818096304198825e-06, + "loss": 0.9852, + "step": 24239 + }, + { + "epoch": 0.58, + "learning_rate": 7.817336182928204e-06, + "loss": 1.0934, + "step": 24240 + }, + { + "epoch": 0.58, + "learning_rate": 7.816576074899286e-06, + "loss": 1.0004, + "step": 24241 + }, + { + "epoch": 0.58, + "learning_rate": 7.815815980116692e-06, + "loss": 1.0753, + "step": 24242 + }, + { + "epoch": 0.58, + "learning_rate": 7.815055898585027e-06, + "loss": 0.8046, + "step": 24243 + }, + { + "epoch": 0.58, + "learning_rate": 7.814295830308904e-06, + "loss": 1.2635, + "step": 24244 + }, + { + "epoch": 0.58, + "learning_rate": 7.813535775292932e-06, + "loss": 1.1411, + "step": 24245 + }, + { + "epoch": 0.58, + "learning_rate": 7.812775733541727e-06, + "loss": 1.0783, + "step": 24246 + }, + { + "epoch": 0.58, + "learning_rate": 7.812015705059898e-06, + "loss": 0.9122, + "step": 24247 + }, + { + "epoch": 0.58, + "learning_rate": 7.811255689852053e-06, + "loss": 1.2945, + "step": 24248 + }, + { + "epoch": 0.58, + "learning_rate": 7.810495687922804e-06, + "loss": 1.4329, + "step": 24249 + }, + { + "epoch": 0.58, + "learning_rate": 7.809735699276765e-06, + "loss": 1.2961, + "step": 24250 + }, + { + "epoch": 0.58, + "learning_rate": 7.808975723918544e-06, + "loss": 1.2777, + "step": 24251 + }, + { + "epoch": 0.58, + "learning_rate": 7.808215761852751e-06, + "loss": 1.0371, + "step": 24252 + }, + { + "epoch": 0.58, + "learning_rate": 7.807455813083997e-06, + "loss": 0.813, + "step": 24253 + }, + { + "epoch": 0.58, + "learning_rate": 7.806695877616894e-06, + "loss": 0.947, + "step": 24254 + }, + { + "epoch": 0.58, + "learning_rate": 7.805935955456051e-06, + "loss": 1.0762, + "step": 24255 + }, + { + "epoch": 0.58, + "learning_rate": 7.805176046606079e-06, + "loss": 0.9716, + "step": 24256 + }, + { + "epoch": 0.58, + "learning_rate": 7.804416151071585e-06, + "loss": 1.0789, + "step": 24257 + }, + { + "epoch": 0.58, + "learning_rate": 7.803656268857184e-06, + "loss": 0.8294, + "step": 24258 + }, + { + "epoch": 0.58, + "learning_rate": 7.802896399967484e-06, + "loss": 1.1659, + "step": 24259 + }, + { + "epoch": 0.58, + "learning_rate": 7.802136544407096e-06, + "loss": 1.0624, + "step": 24260 + }, + { + "epoch": 0.58, + "learning_rate": 7.801376702180624e-06, + "loss": 0.99, + "step": 24261 + }, + { + "epoch": 0.58, + "learning_rate": 7.800616873292686e-06, + "loss": 0.988, + "step": 24262 + }, + { + "epoch": 0.58, + "learning_rate": 7.799857057747888e-06, + "loss": 1.16, + "step": 24263 + }, + { + "epoch": 0.58, + "learning_rate": 7.79909725555084e-06, + "loss": 1.072, + "step": 24264 + }, + { + "epoch": 0.58, + "learning_rate": 7.798337466706149e-06, + "loss": 1.07, + "step": 24265 + }, + { + "epoch": 0.58, + "learning_rate": 7.797577691218428e-06, + "loss": 0.8847, + "step": 24266 + }, + { + "epoch": 0.58, + "learning_rate": 7.796817929092286e-06, + "loss": 1.1432, + "step": 24267 + }, + { + "epoch": 0.58, + "learning_rate": 7.796058180332332e-06, + "loss": 1.3266, + "step": 24268 + }, + { + "epoch": 0.58, + "learning_rate": 7.79529844494317e-06, + "loss": 1.1215, + "step": 24269 + }, + { + "epoch": 0.58, + "learning_rate": 7.794538722929417e-06, + "loss": 1.1231, + "step": 24270 + }, + { + "epoch": 0.58, + "learning_rate": 7.79377901429568e-06, + "loss": 0.9379, + "step": 24271 + }, + { + "epoch": 0.58, + "learning_rate": 7.793019319046566e-06, + "loss": 1.0892, + "step": 24272 + }, + { + "epoch": 0.58, + "learning_rate": 7.792259637186683e-06, + "loss": 1.0601, + "step": 24273 + }, + { + "epoch": 0.58, + "learning_rate": 7.791499968720644e-06, + "loss": 0.9909, + "step": 24274 + }, + { + "epoch": 0.58, + "learning_rate": 7.790740313653054e-06, + "loss": 1.1301, + "step": 24275 + }, + { + "epoch": 0.58, + "learning_rate": 7.789980671988524e-06, + "loss": 0.9295, + "step": 24276 + }, + { + "epoch": 0.58, + "learning_rate": 7.789221043731658e-06, + "loss": 1.1487, + "step": 24277 + }, + { + "epoch": 0.58, + "learning_rate": 7.78846142888707e-06, + "loss": 1.2113, + "step": 24278 + }, + { + "epoch": 0.58, + "learning_rate": 7.787701827459367e-06, + "loss": 1.185, + "step": 24279 + }, + { + "epoch": 0.58, + "learning_rate": 7.786942239453156e-06, + "loss": 0.9043, + "step": 24280 + }, + { + "epoch": 0.58, + "learning_rate": 7.786182664873046e-06, + "loss": 0.8985, + "step": 24281 + }, + { + "epoch": 0.58, + "learning_rate": 7.785423103723644e-06, + "loss": 0.9835, + "step": 24282 + }, + { + "epoch": 0.58, + "learning_rate": 7.784663556009561e-06, + "loss": 1.1417, + "step": 24283 + }, + { + "epoch": 0.58, + "learning_rate": 7.783904021735405e-06, + "loss": 0.8423, + "step": 24284 + }, + { + "epoch": 0.58, + "learning_rate": 7.783144500905777e-06, + "loss": 1.2272, + "step": 24285 + }, + { + "epoch": 0.58, + "learning_rate": 7.782384993525293e-06, + "loss": 0.8834, + "step": 24286 + }, + { + "epoch": 0.58, + "learning_rate": 7.781625499598557e-06, + "loss": 0.9851, + "step": 24287 + }, + { + "epoch": 0.58, + "learning_rate": 7.780866019130175e-06, + "loss": 0.8691, + "step": 24288 + }, + { + "epoch": 0.58, + "learning_rate": 7.78010655212476e-06, + "loss": 1.1763, + "step": 24289 + }, + { + "epoch": 0.58, + "learning_rate": 7.779347098586916e-06, + "loss": 1.2091, + "step": 24290 + }, + { + "epoch": 0.58, + "learning_rate": 7.77858765852125e-06, + "loss": 0.9639, + "step": 24291 + }, + { + "epoch": 0.58, + "learning_rate": 7.77782823193237e-06, + "loss": 1.2025, + "step": 24292 + }, + { + "epoch": 0.58, + "learning_rate": 7.777068818824885e-06, + "loss": 1.1759, + "step": 24293 + }, + { + "epoch": 0.58, + "learning_rate": 7.776309419203399e-06, + "loss": 1.2095, + "step": 24294 + }, + { + "epoch": 0.58, + "learning_rate": 7.775550033072522e-06, + "loss": 0.9912, + "step": 24295 + }, + { + "epoch": 0.58, + "learning_rate": 7.774790660436857e-06, + "loss": 1.008, + "step": 24296 + }, + { + "epoch": 0.58, + "learning_rate": 7.774031301301016e-06, + "loss": 0.9408, + "step": 24297 + }, + { + "epoch": 0.58, + "learning_rate": 7.773271955669603e-06, + "loss": 1.0605, + "step": 24298 + }, + { + "epoch": 0.58, + "learning_rate": 7.772512623547228e-06, + "loss": 0.9596, + "step": 24299 + }, + { + "epoch": 0.58, + "learning_rate": 7.77175330493849e-06, + "loss": 1.1619, + "step": 24300 + }, + { + "epoch": 0.58, + "learning_rate": 7.770993999848003e-06, + "loss": 1.1763, + "step": 24301 + }, + { + "epoch": 0.58, + "learning_rate": 7.770234708280372e-06, + "loss": 0.9745, + "step": 24302 + }, + { + "epoch": 0.58, + "learning_rate": 7.769475430240202e-06, + "loss": 0.9559, + "step": 24303 + }, + { + "epoch": 0.58, + "learning_rate": 7.768716165732097e-06, + "loss": 1.0622, + "step": 24304 + }, + { + "epoch": 0.58, + "learning_rate": 7.76795691476067e-06, + "loss": 1.1834, + "step": 24305 + }, + { + "epoch": 0.58, + "learning_rate": 7.767197677330522e-06, + "loss": 0.9859, + "step": 24306 + }, + { + "epoch": 0.58, + "learning_rate": 7.766438453446262e-06, + "loss": 1.1195, + "step": 24307 + }, + { + "epoch": 0.58, + "learning_rate": 7.765679243112491e-06, + "loss": 1.1274, + "step": 24308 + }, + { + "epoch": 0.58, + "learning_rate": 7.76492004633382e-06, + "loss": 1.175, + "step": 24309 + }, + { + "epoch": 0.58, + "learning_rate": 7.764160863114857e-06, + "loss": 1.1334, + "step": 24310 + }, + { + "epoch": 0.58, + "learning_rate": 7.7634016934602e-06, + "loss": 1.0502, + "step": 24311 + }, + { + "epoch": 0.58, + "learning_rate": 7.76264253737446e-06, + "loss": 1.1637, + "step": 24312 + }, + { + "epoch": 0.58, + "learning_rate": 7.761883394862241e-06, + "loss": 1.047, + "step": 24313 + }, + { + "epoch": 0.58, + "learning_rate": 7.76112426592815e-06, + "loss": 1.3472, + "step": 24314 + }, + { + "epoch": 0.58, + "learning_rate": 7.760365150576792e-06, + "loss": 1.2193, + "step": 24315 + }, + { + "epoch": 0.58, + "learning_rate": 7.75960604881277e-06, + "loss": 1.0722, + "step": 24316 + }, + { + "epoch": 0.58, + "learning_rate": 7.758846960640691e-06, + "loss": 1.1899, + "step": 24317 + }, + { + "epoch": 0.58, + "learning_rate": 7.758087886065163e-06, + "loss": 0.9916, + "step": 24318 + }, + { + "epoch": 0.58, + "learning_rate": 7.757328825090787e-06, + "loss": 0.9514, + "step": 24319 + }, + { + "epoch": 0.58, + "learning_rate": 7.756569777722167e-06, + "loss": 0.9503, + "step": 24320 + }, + { + "epoch": 0.58, + "learning_rate": 7.755810743963913e-06, + "loss": 1.0539, + "step": 24321 + }, + { + "epoch": 0.58, + "learning_rate": 7.755051723820628e-06, + "loss": 1.2198, + "step": 24322 + }, + { + "epoch": 0.58, + "learning_rate": 7.754292717296917e-06, + "loss": 1.2362, + "step": 24323 + }, + { + "epoch": 0.58, + "learning_rate": 7.753533724397379e-06, + "loss": 1.1467, + "step": 24324 + }, + { + "epoch": 0.58, + "learning_rate": 7.752774745126627e-06, + "loss": 1.1684, + "step": 24325 + }, + { + "epoch": 0.59, + "learning_rate": 7.752015779489262e-06, + "loss": 1.0233, + "step": 24326 + }, + { + "epoch": 0.59, + "learning_rate": 7.751256827489889e-06, + "loss": 1.2198, + "step": 24327 + }, + { + "epoch": 0.59, + "learning_rate": 7.750497889133109e-06, + "loss": 1.1659, + "step": 24328 + }, + { + "epoch": 0.59, + "learning_rate": 7.749738964423532e-06, + "loss": 1.0377, + "step": 24329 + }, + { + "epoch": 0.59, + "learning_rate": 7.74898005336576e-06, + "loss": 1.0035, + "step": 24330 + }, + { + "epoch": 0.59, + "learning_rate": 7.748221155964396e-06, + "loss": 0.9854, + "step": 24331 + }, + { + "epoch": 0.59, + "learning_rate": 7.747462272224039e-06, + "loss": 0.8996, + "step": 24332 + }, + { + "epoch": 0.59, + "learning_rate": 7.746703402149302e-06, + "loss": 1.1999, + "step": 24333 + }, + { + "epoch": 0.59, + "learning_rate": 7.745944545744787e-06, + "loss": 0.892, + "step": 24334 + }, + { + "epoch": 0.59, + "learning_rate": 7.745185703015097e-06, + "loss": 1.0409, + "step": 24335 + }, + { + "epoch": 0.59, + "learning_rate": 7.744426873964834e-06, + "loss": 0.9577, + "step": 24336 + }, + { + "epoch": 0.59, + "learning_rate": 7.7436680585986e-06, + "loss": 1.1668, + "step": 24337 + }, + { + "epoch": 0.59, + "learning_rate": 7.742909256921005e-06, + "loss": 1.1256, + "step": 24338 + }, + { + "epoch": 0.59, + "learning_rate": 7.742150468936647e-06, + "loss": 1.1559, + "step": 24339 + }, + { + "epoch": 0.59, + "learning_rate": 7.741391694650133e-06, + "loss": 1.1846, + "step": 24340 + }, + { + "epoch": 0.59, + "learning_rate": 7.740632934066061e-06, + "loss": 0.6204, + "step": 24341 + }, + { + "epoch": 0.59, + "learning_rate": 7.73987418718904e-06, + "loss": 0.9336, + "step": 24342 + }, + { + "epoch": 0.59, + "learning_rate": 7.739115454023671e-06, + "loss": 1.0148, + "step": 24343 + }, + { + "epoch": 0.59, + "learning_rate": 7.738356734574556e-06, + "loss": 1.01, + "step": 24344 + }, + { + "epoch": 0.59, + "learning_rate": 7.737598028846296e-06, + "loss": 0.9271, + "step": 24345 + }, + { + "epoch": 0.59, + "learning_rate": 7.7368393368435e-06, + "loss": 1.2051, + "step": 24346 + }, + { + "epoch": 0.59, + "learning_rate": 7.736080658570767e-06, + "loss": 1.1844, + "step": 24347 + }, + { + "epoch": 0.59, + "learning_rate": 7.7353219940327e-06, + "loss": 1.0784, + "step": 24348 + }, + { + "epoch": 0.59, + "learning_rate": 7.7345633432339e-06, + "loss": 0.979, + "step": 24349 + }, + { + "epoch": 0.59, + "learning_rate": 7.733804706178974e-06, + "loss": 0.9322, + "step": 24350 + }, + { + "epoch": 0.59, + "learning_rate": 7.73304608287252e-06, + "loss": 1.2438, + "step": 24351 + }, + { + "epoch": 0.59, + "learning_rate": 7.732287473319143e-06, + "loss": 0.8202, + "step": 24352 + }, + { + "epoch": 0.59, + "learning_rate": 7.731528877523444e-06, + "loss": 1.0236, + "step": 24353 + }, + { + "epoch": 0.59, + "learning_rate": 7.730770295490026e-06, + "loss": 1.0305, + "step": 24354 + }, + { + "epoch": 0.59, + "learning_rate": 7.730011727223492e-06, + "loss": 0.9442, + "step": 24355 + }, + { + "epoch": 0.59, + "learning_rate": 7.729253172728441e-06, + "loss": 1.1522, + "step": 24356 + }, + { + "epoch": 0.59, + "learning_rate": 7.728494632009477e-06, + "loss": 1.213, + "step": 24357 + }, + { + "epoch": 0.59, + "learning_rate": 7.727736105071202e-06, + "loss": 1.0845, + "step": 24358 + }, + { + "epoch": 0.59, + "learning_rate": 7.726977591918219e-06, + "loss": 1.0976, + "step": 24359 + }, + { + "epoch": 0.59, + "learning_rate": 7.726219092555127e-06, + "loss": 1.1683, + "step": 24360 + }, + { + "epoch": 0.59, + "learning_rate": 7.725460606986529e-06, + "loss": 1.1708, + "step": 24361 + }, + { + "epoch": 0.59, + "learning_rate": 7.724702135217026e-06, + "loss": 0.9708, + "step": 24362 + }, + { + "epoch": 0.59, + "learning_rate": 7.723943677251222e-06, + "loss": 1.1806, + "step": 24363 + }, + { + "epoch": 0.59, + "learning_rate": 7.723185233093716e-06, + "loss": 1.116, + "step": 24364 + }, + { + "epoch": 0.59, + "learning_rate": 7.722426802749106e-06, + "loss": 0.9687, + "step": 24365 + }, + { + "epoch": 0.59, + "learning_rate": 7.721668386222001e-06, + "loss": 1.1606, + "step": 24366 + }, + { + "epoch": 0.59, + "learning_rate": 7.720909983516996e-06, + "loss": 1.1692, + "step": 24367 + }, + { + "epoch": 0.59, + "learning_rate": 7.720151594638695e-06, + "loss": 1.216, + "step": 24368 + }, + { + "epoch": 0.59, + "learning_rate": 7.719393219591696e-06, + "loss": 1.0215, + "step": 24369 + }, + { + "epoch": 0.59, + "learning_rate": 7.718634858380604e-06, + "loss": 0.9714, + "step": 24370 + }, + { + "epoch": 0.59, + "learning_rate": 7.717876511010019e-06, + "loss": 0.8936, + "step": 24371 + }, + { + "epoch": 0.59, + "learning_rate": 7.717118177484538e-06, + "loss": 1.0202, + "step": 24372 + }, + { + "epoch": 0.59, + "learning_rate": 7.716359857808764e-06, + "loss": 1.0257, + "step": 24373 + }, + { + "epoch": 0.59, + "learning_rate": 7.715601551987298e-06, + "loss": 0.939, + "step": 24374 + }, + { + "epoch": 0.59, + "learning_rate": 7.714843260024741e-06, + "loss": 1.1811, + "step": 24375 + }, + { + "epoch": 0.59, + "learning_rate": 7.714084981925691e-06, + "loss": 1.0389, + "step": 24376 + }, + { + "epoch": 0.59, + "learning_rate": 7.71332671769475e-06, + "loss": 0.7921, + "step": 24377 + }, + { + "epoch": 0.59, + "learning_rate": 7.712568467336519e-06, + "loss": 1.0741, + "step": 24378 + }, + { + "epoch": 0.59, + "learning_rate": 7.711810230855598e-06, + "loss": 1.2198, + "step": 24379 + }, + { + "epoch": 0.59, + "learning_rate": 7.711052008256583e-06, + "loss": 0.8956, + "step": 24380 + }, + { + "epoch": 0.59, + "learning_rate": 7.71029379954408e-06, + "loss": 1.1355, + "step": 24381 + }, + { + "epoch": 0.59, + "learning_rate": 7.709535604722685e-06, + "loss": 0.8864, + "step": 24382 + }, + { + "epoch": 0.59, + "learning_rate": 7.708777423796998e-06, + "loss": 1.1042, + "step": 24383 + }, + { + "epoch": 0.59, + "learning_rate": 7.708019256771617e-06, + "loss": 0.9728, + "step": 24384 + }, + { + "epoch": 0.59, + "learning_rate": 7.707261103651149e-06, + "loss": 0.8906, + "step": 24385 + }, + { + "epoch": 0.59, + "learning_rate": 7.706502964440188e-06, + "loss": 0.9951, + "step": 24386 + }, + { + "epoch": 0.59, + "learning_rate": 7.705744839143332e-06, + "loss": 1.0909, + "step": 24387 + }, + { + "epoch": 0.59, + "learning_rate": 7.704986727765182e-06, + "loss": 0.8379, + "step": 24388 + }, + { + "epoch": 0.59, + "learning_rate": 7.704228630310338e-06, + "loss": 0.786, + "step": 24389 + }, + { + "epoch": 0.59, + "learning_rate": 7.703470546783401e-06, + "loss": 1.0692, + "step": 24390 + }, + { + "epoch": 0.59, + "learning_rate": 7.702712477188966e-06, + "loss": 0.9648, + "step": 24391 + }, + { + "epoch": 0.59, + "learning_rate": 7.701954421531634e-06, + "loss": 1.0256, + "step": 24392 + }, + { + "epoch": 0.59, + "learning_rate": 7.701196379816005e-06, + "loss": 1.0747, + "step": 24393 + }, + { + "epoch": 0.59, + "learning_rate": 7.700438352046677e-06, + "loss": 1.2897, + "step": 24394 + }, + { + "epoch": 0.59, + "learning_rate": 7.699680338228249e-06, + "loss": 1.388, + "step": 24395 + }, + { + "epoch": 0.59, + "learning_rate": 7.698922338365317e-06, + "loss": 1.0526, + "step": 24396 + }, + { + "epoch": 0.59, + "learning_rate": 7.698164352462483e-06, + "loss": 1.1679, + "step": 24397 + }, + { + "epoch": 0.59, + "learning_rate": 7.697406380524346e-06, + "loss": 1.0636, + "step": 24398 + }, + { + "epoch": 0.59, + "learning_rate": 7.696648422555503e-06, + "loss": 0.6876, + "step": 24399 + }, + { + "epoch": 0.59, + "learning_rate": 7.695890478560548e-06, + "loss": 1.1211, + "step": 24400 + }, + { + "epoch": 0.59, + "learning_rate": 7.69513254854409e-06, + "loss": 1.1436, + "step": 24401 + }, + { + "epoch": 0.59, + "learning_rate": 7.694374632510716e-06, + "loss": 1.0659, + "step": 24402 + }, + { + "epoch": 0.59, + "learning_rate": 7.69361673046503e-06, + "loss": 1.2572, + "step": 24403 + }, + { + "epoch": 0.59, + "learning_rate": 7.692858842411628e-06, + "loss": 1.0476, + "step": 24404 + }, + { + "epoch": 0.59, + "learning_rate": 7.69210096835511e-06, + "loss": 1.0323, + "step": 24405 + }, + { + "epoch": 0.59, + "learning_rate": 7.691343108300072e-06, + "loss": 0.9868, + "step": 24406 + }, + { + "epoch": 0.59, + "learning_rate": 7.690585262251114e-06, + "loss": 1.2135, + "step": 24407 + }, + { + "epoch": 0.59, + "learning_rate": 7.689827430212827e-06, + "loss": 1.0669, + "step": 24408 + }, + { + "epoch": 0.59, + "learning_rate": 7.689069612189818e-06, + "loss": 1.2514, + "step": 24409 + }, + { + "epoch": 0.59, + "learning_rate": 7.688311808186679e-06, + "loss": 1.0527, + "step": 24410 + }, + { + "epoch": 0.59, + "learning_rate": 7.68755401820801e-06, + "loss": 1.3741, + "step": 24411 + }, + { + "epoch": 0.59, + "learning_rate": 7.686796242258404e-06, + "loss": 1.0482, + "step": 24412 + }, + { + "epoch": 0.59, + "learning_rate": 7.686038480342462e-06, + "loss": 1.2432, + "step": 24413 + }, + { + "epoch": 0.59, + "learning_rate": 7.685280732464781e-06, + "loss": 0.9884, + "step": 24414 + }, + { + "epoch": 0.59, + "learning_rate": 7.684522998629957e-06, + "loss": 1.1426, + "step": 24415 + }, + { + "epoch": 0.59, + "learning_rate": 7.683765278842587e-06, + "loss": 1.1575, + "step": 24416 + }, + { + "epoch": 0.59, + "learning_rate": 7.683007573107271e-06, + "loss": 1.18, + "step": 24417 + }, + { + "epoch": 0.59, + "learning_rate": 7.6822498814286e-06, + "loss": 1.0145, + "step": 24418 + }, + { + "epoch": 0.59, + "learning_rate": 7.681492203811174e-06, + "loss": 1.1046, + "step": 24419 + }, + { + "epoch": 0.59, + "learning_rate": 7.68073454025959e-06, + "loss": 0.9526, + "step": 24420 + }, + { + "epoch": 0.59, + "learning_rate": 7.679976890778445e-06, + "loss": 0.9395, + "step": 24421 + }, + { + "epoch": 0.59, + "learning_rate": 7.679219255372334e-06, + "loss": 1.068, + "step": 24422 + }, + { + "epoch": 0.59, + "learning_rate": 7.678461634045851e-06, + "loss": 1.2698, + "step": 24423 + }, + { + "epoch": 0.59, + "learning_rate": 7.677704026803599e-06, + "loss": 1.2333, + "step": 24424 + }, + { + "epoch": 0.59, + "learning_rate": 7.67694643365017e-06, + "loss": 0.8738, + "step": 24425 + }, + { + "epoch": 0.59, + "learning_rate": 7.67618885459016e-06, + "loss": 1.0957, + "step": 24426 + }, + { + "epoch": 0.59, + "learning_rate": 7.675431289628164e-06, + "loss": 1.0792, + "step": 24427 + }, + { + "epoch": 0.59, + "learning_rate": 7.674673738768781e-06, + "loss": 1.0578, + "step": 24428 + }, + { + "epoch": 0.59, + "learning_rate": 7.673916202016606e-06, + "loss": 1.2187, + "step": 24429 + }, + { + "epoch": 0.59, + "learning_rate": 7.673158679376233e-06, + "loss": 1.2818, + "step": 24430 + }, + { + "epoch": 0.59, + "learning_rate": 7.67240117085226e-06, + "loss": 1.2836, + "step": 24431 + }, + { + "epoch": 0.59, + "learning_rate": 7.67164367644928e-06, + "loss": 1.1066, + "step": 24432 + }, + { + "epoch": 0.59, + "learning_rate": 7.670886196171893e-06, + "loss": 0.9988, + "step": 24433 + }, + { + "epoch": 0.59, + "learning_rate": 7.670128730024691e-06, + "loss": 0.9918, + "step": 24434 + }, + { + "epoch": 0.59, + "learning_rate": 7.669371278012267e-06, + "loss": 1.0759, + "step": 24435 + }, + { + "epoch": 0.59, + "learning_rate": 7.66861384013922e-06, + "loss": 1.1879, + "step": 24436 + }, + { + "epoch": 0.59, + "learning_rate": 7.667856416410148e-06, + "loss": 1.1786, + "step": 24437 + }, + { + "epoch": 0.59, + "learning_rate": 7.66709900682964e-06, + "loss": 1.2102, + "step": 24438 + }, + { + "epoch": 0.59, + "learning_rate": 7.666341611402293e-06, + "loss": 1.1127, + "step": 24439 + }, + { + "epoch": 0.59, + "learning_rate": 7.665584230132703e-06, + "loss": 0.9897, + "step": 24440 + }, + { + "epoch": 0.59, + "learning_rate": 7.664826863025466e-06, + "loss": 1.0782, + "step": 24441 + }, + { + "epoch": 0.59, + "learning_rate": 7.664069510085174e-06, + "loss": 0.8969, + "step": 24442 + }, + { + "epoch": 0.59, + "learning_rate": 7.663312171316421e-06, + "loss": 1.2292, + "step": 24443 + }, + { + "epoch": 0.59, + "learning_rate": 7.662554846723807e-06, + "loss": 1.0229, + "step": 24444 + }, + { + "epoch": 0.59, + "learning_rate": 7.661797536311921e-06, + "loss": 0.7089, + "step": 24445 + }, + { + "epoch": 0.59, + "learning_rate": 7.661040240085361e-06, + "loss": 1.3251, + "step": 24446 + }, + { + "epoch": 0.59, + "learning_rate": 7.660282958048715e-06, + "loss": 1.0997, + "step": 24447 + }, + { + "epoch": 0.59, + "learning_rate": 7.659525690206586e-06, + "loss": 1.3137, + "step": 24448 + }, + { + "epoch": 0.59, + "learning_rate": 7.658768436563564e-06, + "loss": 0.69, + "step": 24449 + }, + { + "epoch": 0.59, + "learning_rate": 7.658011197124244e-06, + "loss": 1.0546, + "step": 24450 + }, + { + "epoch": 0.59, + "learning_rate": 7.657253971893215e-06, + "loss": 1.1015, + "step": 24451 + }, + { + "epoch": 0.59, + "learning_rate": 7.656496760875078e-06, + "loss": 1.1212, + "step": 24452 + }, + { + "epoch": 0.59, + "learning_rate": 7.655739564074426e-06, + "loss": 0.9662, + "step": 24453 + }, + { + "epoch": 0.59, + "learning_rate": 7.654982381495848e-06, + "loss": 1.2626, + "step": 24454 + }, + { + "epoch": 0.59, + "learning_rate": 7.654225213143939e-06, + "loss": 1.2401, + "step": 24455 + }, + { + "epoch": 0.59, + "learning_rate": 7.653468059023295e-06, + "loss": 0.9738, + "step": 24456 + }, + { + "epoch": 0.59, + "learning_rate": 7.65271091913851e-06, + "loss": 1.2022, + "step": 24457 + }, + { + "epoch": 0.59, + "learning_rate": 7.651953793494175e-06, + "loss": 1.0929, + "step": 24458 + }, + { + "epoch": 0.59, + "learning_rate": 7.651196682094883e-06, + "loss": 1.3462, + "step": 24459 + }, + { + "epoch": 0.59, + "learning_rate": 7.65043958494523e-06, + "loss": 1.2043, + "step": 24460 + }, + { + "epoch": 0.59, + "learning_rate": 7.649682502049807e-06, + "loss": 1.1723, + "step": 24461 + }, + { + "epoch": 0.59, + "learning_rate": 7.648925433413204e-06, + "loss": 0.8592, + "step": 24462 + }, + { + "epoch": 0.59, + "learning_rate": 7.64816837904002e-06, + "loss": 1.3337, + "step": 24463 + }, + { + "epoch": 0.59, + "learning_rate": 7.647411338934847e-06, + "loss": 1.038, + "step": 24464 + }, + { + "epoch": 0.59, + "learning_rate": 7.646654313102275e-06, + "loss": 0.96, + "step": 24465 + }, + { + "epoch": 0.59, + "learning_rate": 7.645897301546896e-06, + "loss": 1.1078, + "step": 24466 + }, + { + "epoch": 0.59, + "learning_rate": 7.645140304273307e-06, + "loss": 0.967, + "step": 24467 + }, + { + "epoch": 0.59, + "learning_rate": 7.644383321286095e-06, + "loss": 1.1131, + "step": 24468 + }, + { + "epoch": 0.59, + "learning_rate": 7.643626352589858e-06, + "loss": 0.9934, + "step": 24469 + }, + { + "epoch": 0.59, + "learning_rate": 7.642869398189183e-06, + "loss": 1.172, + "step": 24470 + }, + { + "epoch": 0.59, + "learning_rate": 7.642112458088666e-06, + "loss": 1.0698, + "step": 24471 + }, + { + "epoch": 0.59, + "learning_rate": 7.6413555322929e-06, + "loss": 1.0629, + "step": 24472 + }, + { + "epoch": 0.59, + "learning_rate": 7.640598620806473e-06, + "loss": 1.0579, + "step": 24473 + }, + { + "epoch": 0.59, + "learning_rate": 7.639841723633979e-06, + "loss": 1.162, + "step": 24474 + }, + { + "epoch": 0.59, + "learning_rate": 7.639084840780012e-06, + "loss": 1.013, + "step": 24475 + }, + { + "epoch": 0.59, + "learning_rate": 7.63832797224916e-06, + "loss": 1.2298, + "step": 24476 + }, + { + "epoch": 0.59, + "learning_rate": 7.637571118046018e-06, + "loss": 1.2639, + "step": 24477 + }, + { + "epoch": 0.59, + "learning_rate": 7.636814278175176e-06, + "loss": 0.8877, + "step": 24478 + }, + { + "epoch": 0.59, + "learning_rate": 7.636057452641226e-06, + "loss": 1.1174, + "step": 24479 + }, + { + "epoch": 0.59, + "learning_rate": 7.63530064144876e-06, + "loss": 1.1869, + "step": 24480 + }, + { + "epoch": 0.59, + "learning_rate": 7.63454384460237e-06, + "loss": 0.8262, + "step": 24481 + }, + { + "epoch": 0.59, + "learning_rate": 7.633787062106642e-06, + "loss": 1.1093, + "step": 24482 + }, + { + "epoch": 0.59, + "learning_rate": 7.633030293966172e-06, + "loss": 1.0337, + "step": 24483 + }, + { + "epoch": 0.59, + "learning_rate": 7.632273540185554e-06, + "loss": 0.9822, + "step": 24484 + }, + { + "epoch": 0.59, + "learning_rate": 7.631516800769373e-06, + "loss": 1.2164, + "step": 24485 + }, + { + "epoch": 0.59, + "learning_rate": 7.630760075722222e-06, + "loss": 1.0954, + "step": 24486 + }, + { + "epoch": 0.59, + "learning_rate": 7.630003365048694e-06, + "loss": 1.2837, + "step": 24487 + }, + { + "epoch": 0.59, + "learning_rate": 7.629246668753378e-06, + "loss": 1.2393, + "step": 24488 + }, + { + "epoch": 0.59, + "learning_rate": 7.628489986840865e-06, + "loss": 1.0942, + "step": 24489 + }, + { + "epoch": 0.59, + "learning_rate": 7.6277333193157435e-06, + "loss": 0.8095, + "step": 24490 + }, + { + "epoch": 0.59, + "learning_rate": 7.6269766661826074e-06, + "loss": 1.1866, + "step": 24491 + }, + { + "epoch": 0.59, + "learning_rate": 7.626220027446045e-06, + "loss": 0.9015, + "step": 24492 + }, + { + "epoch": 0.59, + "learning_rate": 7.625463403110649e-06, + "loss": 1.2335, + "step": 24493 + }, + { + "epoch": 0.59, + "learning_rate": 7.624706793181005e-06, + "loss": 0.6737, + "step": 24494 + }, + { + "epoch": 0.59, + "learning_rate": 7.62395019766171e-06, + "loss": 1.0102, + "step": 24495 + }, + { + "epoch": 0.59, + "learning_rate": 7.6231936165573475e-06, + "loss": 0.8858, + "step": 24496 + }, + { + "epoch": 0.59, + "learning_rate": 7.622437049872512e-06, + "loss": 1.2231, + "step": 24497 + }, + { + "epoch": 0.59, + "learning_rate": 7.621680497611788e-06, + "loss": 0.9508, + "step": 24498 + }, + { + "epoch": 0.59, + "learning_rate": 7.620923959779769e-06, + "loss": 1.1237, + "step": 24499 + }, + { + "epoch": 0.59, + "learning_rate": 7.6201674363810475e-06, + "loss": 0.9835, + "step": 24500 + }, + { + "epoch": 0.59, + "learning_rate": 7.619410927420211e-06, + "loss": 1.1318, + "step": 24501 + }, + { + "epoch": 0.59, + "learning_rate": 7.618654432901849e-06, + "loss": 1.1889, + "step": 24502 + }, + { + "epoch": 0.59, + "learning_rate": 7.617897952830548e-06, + "loss": 1.1066, + "step": 24503 + }, + { + "epoch": 0.59, + "learning_rate": 7.617141487210901e-06, + "loss": 1.244, + "step": 24504 + }, + { + "epoch": 0.59, + "learning_rate": 7.6163850360474965e-06, + "loss": 1.116, + "step": 24505 + }, + { + "epoch": 0.59, + "learning_rate": 7.615628599344924e-06, + "loss": 1.079, + "step": 24506 + }, + { + "epoch": 0.59, + "learning_rate": 7.614872177107768e-06, + "loss": 1.0629, + "step": 24507 + }, + { + "epoch": 0.59, + "learning_rate": 7.614115769340626e-06, + "loss": 1.1235, + "step": 24508 + }, + { + "epoch": 0.59, + "learning_rate": 7.613359376048082e-06, + "loss": 1.1237, + "step": 24509 + }, + { + "epoch": 0.59, + "learning_rate": 7.612602997234723e-06, + "loss": 1.235, + "step": 24510 + }, + { + "epoch": 0.59, + "learning_rate": 7.61184663290514e-06, + "loss": 0.8547, + "step": 24511 + }, + { + "epoch": 0.59, + "learning_rate": 7.611090283063924e-06, + "loss": 1.2025, + "step": 24512 + }, + { + "epoch": 0.59, + "learning_rate": 7.610333947715661e-06, + "loss": 0.8291, + "step": 24513 + }, + { + "epoch": 0.59, + "learning_rate": 7.609577626864939e-06, + "loss": 0.8355, + "step": 24514 + }, + { + "epoch": 0.59, + "learning_rate": 7.608821320516346e-06, + "loss": 1.0238, + "step": 24515 + }, + { + "epoch": 0.59, + "learning_rate": 7.6080650286744735e-06, + "loss": 1.1028, + "step": 24516 + }, + { + "epoch": 0.59, + "learning_rate": 7.607308751343908e-06, + "loss": 1.1153, + "step": 24517 + }, + { + "epoch": 0.59, + "learning_rate": 7.606552488529238e-06, + "loss": 1.3685, + "step": 24518 + }, + { + "epoch": 0.59, + "learning_rate": 7.605796240235048e-06, + "loss": 1.0476, + "step": 24519 + }, + { + "epoch": 0.59, + "learning_rate": 7.605040006465931e-06, + "loss": 1.1652, + "step": 24520 + }, + { + "epoch": 0.59, + "learning_rate": 7.604283787226474e-06, + "loss": 1.0864, + "step": 24521 + }, + { + "epoch": 0.59, + "learning_rate": 7.603527582521264e-06, + "loss": 1.023, + "step": 24522 + }, + { + "epoch": 0.59, + "learning_rate": 7.6027713923548855e-06, + "loss": 0.965, + "step": 24523 + }, + { + "epoch": 0.59, + "learning_rate": 7.602015216731932e-06, + "loss": 0.8354, + "step": 24524 + }, + { + "epoch": 0.59, + "learning_rate": 7.6012590556569865e-06, + "loss": 0.6485, + "step": 24525 + }, + { + "epoch": 0.59, + "learning_rate": 7.600502909134639e-06, + "loss": 1.0127, + "step": 24526 + }, + { + "epoch": 0.59, + "learning_rate": 7.599746777169474e-06, + "loss": 1.2513, + "step": 24527 + }, + { + "epoch": 0.59, + "learning_rate": 7.598990659766083e-06, + "loss": 0.9197, + "step": 24528 + }, + { + "epoch": 0.59, + "learning_rate": 7.59823455692905e-06, + "loss": 1.2117, + "step": 24529 + }, + { + "epoch": 0.59, + "learning_rate": 7.597478468662964e-06, + "loss": 1.263, + "step": 24530 + }, + { + "epoch": 0.59, + "learning_rate": 7.596722394972409e-06, + "loss": 0.7766, + "step": 24531 + }, + { + "epoch": 0.59, + "learning_rate": 7.5959663358619765e-06, + "loss": 0.8477, + "step": 24532 + }, + { + "epoch": 0.59, + "learning_rate": 7.595210291336249e-06, + "loss": 0.6832, + "step": 24533 + }, + { + "epoch": 0.59, + "learning_rate": 7.594454261399816e-06, + "loss": 0.9761, + "step": 24534 + }, + { + "epoch": 0.59, + "learning_rate": 7.5936982460572616e-06, + "loss": 1.015, + "step": 24535 + }, + { + "epoch": 0.59, + "learning_rate": 7.592942245313175e-06, + "loss": 1.1481, + "step": 24536 + }, + { + "epoch": 0.59, + "learning_rate": 7.592186259172143e-06, + "loss": 1.095, + "step": 24537 + }, + { + "epoch": 0.59, + "learning_rate": 7.591430287638747e-06, + "loss": 0.9926, + "step": 24538 + }, + { + "epoch": 0.59, + "learning_rate": 7.59067433071758e-06, + "loss": 1.1308, + "step": 24539 + }, + { + "epoch": 0.59, + "learning_rate": 7.589918388413224e-06, + "loss": 0.9626, + "step": 24540 + }, + { + "epoch": 0.59, + "learning_rate": 7.589162460730267e-06, + "loss": 1.129, + "step": 24541 + }, + { + "epoch": 0.59, + "learning_rate": 7.588406547673292e-06, + "loss": 1.0972, + "step": 24542 + }, + { + "epoch": 0.59, + "learning_rate": 7.58765064924689e-06, + "loss": 1.1562, + "step": 24543 + }, + { + "epoch": 0.59, + "learning_rate": 7.586894765455643e-06, + "loss": 0.9663, + "step": 24544 + }, + { + "epoch": 0.59, + "learning_rate": 7.58613889630414e-06, + "loss": 1.2121, + "step": 24545 + }, + { + "epoch": 0.59, + "learning_rate": 7.58538304179696e-06, + "loss": 0.8436, + "step": 24546 + }, + { + "epoch": 0.59, + "learning_rate": 7.584627201938697e-06, + "loss": 1.0127, + "step": 24547 + }, + { + "epoch": 0.59, + "learning_rate": 7.583871376733931e-06, + "loss": 1.3108, + "step": 24548 + }, + { + "epoch": 0.59, + "learning_rate": 7.58311556618725e-06, + "loss": 1.0547, + "step": 24549 + }, + { + "epoch": 0.59, + "learning_rate": 7.582359770303236e-06, + "loss": 1.0218, + "step": 24550 + }, + { + "epoch": 0.59, + "learning_rate": 7.581603989086479e-06, + "loss": 1.084, + "step": 24551 + }, + { + "epoch": 0.59, + "learning_rate": 7.580848222541562e-06, + "loss": 1.1718, + "step": 24552 + }, + { + "epoch": 0.59, + "learning_rate": 7.58009247067307e-06, + "loss": 1.1013, + "step": 24553 + }, + { + "epoch": 0.59, + "learning_rate": 7.579336733485584e-06, + "loss": 0.9897, + "step": 24554 + }, + { + "epoch": 0.59, + "learning_rate": 7.578581010983697e-06, + "loss": 1.1044, + "step": 24555 + }, + { + "epoch": 0.59, + "learning_rate": 7.577825303171987e-06, + "loss": 0.944, + "step": 24556 + }, + { + "epoch": 0.59, + "learning_rate": 7.577069610055043e-06, + "loss": 1.0783, + "step": 24557 + }, + { + "epoch": 0.59, + "learning_rate": 7.576313931637444e-06, + "loss": 0.993, + "step": 24558 + }, + { + "epoch": 0.59, + "learning_rate": 7.575558267923783e-06, + "loss": 0.7811, + "step": 24559 + }, + { + "epoch": 0.59, + "learning_rate": 7.5748026189186376e-06, + "loss": 0.9044, + "step": 24560 + }, + { + "epoch": 0.59, + "learning_rate": 7.574046984626595e-06, + "loss": 0.8118, + "step": 24561 + }, + { + "epoch": 0.59, + "learning_rate": 7.573291365052236e-06, + "loss": 1.0677, + "step": 24562 + }, + { + "epoch": 0.59, + "learning_rate": 7.5725357602001505e-06, + "loss": 1.0153, + "step": 24563 + }, + { + "epoch": 0.59, + "learning_rate": 7.571780170074919e-06, + "loss": 1.124, + "step": 24564 + }, + { + "epoch": 0.59, + "learning_rate": 7.571024594681126e-06, + "loss": 0.8521, + "step": 24565 + }, + { + "epoch": 0.59, + "learning_rate": 7.570269034023354e-06, + "loss": 1.036, + "step": 24566 + }, + { + "epoch": 0.59, + "learning_rate": 7.569513488106191e-06, + "loss": 0.8337, + "step": 24567 + }, + { + "epoch": 0.59, + "learning_rate": 7.568757956934216e-06, + "loss": 1.0231, + "step": 24568 + }, + { + "epoch": 0.59, + "learning_rate": 7.568002440512016e-06, + "loss": 0.9326, + "step": 24569 + }, + { + "epoch": 0.59, + "learning_rate": 7.567246938844171e-06, + "loss": 1.0811, + "step": 24570 + }, + { + "epoch": 0.59, + "learning_rate": 7.5664914519352674e-06, + "loss": 1.1332, + "step": 24571 + }, + { + "epoch": 0.59, + "learning_rate": 7.565735979789889e-06, + "loss": 0.9251, + "step": 24572 + }, + { + "epoch": 0.59, + "learning_rate": 7.5649805224126176e-06, + "loss": 1.1602, + "step": 24573 + }, + { + "epoch": 0.59, + "learning_rate": 7.564225079808034e-06, + "loss": 0.9052, + "step": 24574 + }, + { + "epoch": 0.59, + "learning_rate": 7.563469651980727e-06, + "loss": 0.9843, + "step": 24575 + }, + { + "epoch": 0.59, + "learning_rate": 7.562714238935276e-06, + "loss": 0.9448, + "step": 24576 + }, + { + "epoch": 0.59, + "learning_rate": 7.561958840676265e-06, + "loss": 1.3531, + "step": 24577 + }, + { + "epoch": 0.59, + "learning_rate": 7.5612034572082726e-06, + "loss": 1.2479, + "step": 24578 + }, + { + "epoch": 0.59, + "learning_rate": 7.5604480885358885e-06, + "loss": 1.1283, + "step": 24579 + }, + { + "epoch": 0.59, + "learning_rate": 7.559692734663692e-06, + "loss": 0.9492, + "step": 24580 + }, + { + "epoch": 0.59, + "learning_rate": 7.5589373955962654e-06, + "loss": 1.0698, + "step": 24581 + }, + { + "epoch": 0.59, + "learning_rate": 7.5581820713381895e-06, + "loss": 1.1752, + "step": 24582 + }, + { + "epoch": 0.59, + "learning_rate": 7.557426761894051e-06, + "loss": 0.9634, + "step": 24583 + }, + { + "epoch": 0.59, + "learning_rate": 7.556671467268428e-06, + "loss": 0.8812, + "step": 24584 + }, + { + "epoch": 0.59, + "learning_rate": 7.555916187465905e-06, + "loss": 1.2429, + "step": 24585 + }, + { + "epoch": 0.59, + "learning_rate": 7.555160922491065e-06, + "loss": 1.0089, + "step": 24586 + }, + { + "epoch": 0.59, + "learning_rate": 7.5544056723484874e-06, + "loss": 0.9862, + "step": 24587 + }, + { + "epoch": 0.59, + "learning_rate": 7.553650437042756e-06, + "loss": 1.0275, + "step": 24588 + }, + { + "epoch": 0.59, + "learning_rate": 7.552895216578451e-06, + "loss": 1.1744, + "step": 24589 + }, + { + "epoch": 0.59, + "learning_rate": 7.552140010960155e-06, + "loss": 0.8887, + "step": 24590 + }, + { + "epoch": 0.59, + "learning_rate": 7.551384820192451e-06, + "loss": 1.2106, + "step": 24591 + }, + { + "epoch": 0.59, + "learning_rate": 7.5506296442799195e-06, + "loss": 1.0163, + "step": 24592 + }, + { + "epoch": 0.59, + "learning_rate": 7.54987448322714e-06, + "loss": 0.9405, + "step": 24593 + }, + { + "epoch": 0.59, + "learning_rate": 7.549119337038697e-06, + "loss": 1.0933, + "step": 24594 + }, + { + "epoch": 0.59, + "learning_rate": 7.548364205719171e-06, + "loss": 1.0263, + "step": 24595 + }, + { + "epoch": 0.59, + "learning_rate": 7.5476090892731425e-06, + "loss": 1.1248, + "step": 24596 + }, + { + "epoch": 0.59, + "learning_rate": 7.546853987705192e-06, + "loss": 0.6931, + "step": 24597 + }, + { + "epoch": 0.59, + "learning_rate": 7.546098901019902e-06, + "loss": 1.2548, + "step": 24598 + }, + { + "epoch": 0.59, + "learning_rate": 7.5453438292218515e-06, + "loss": 1.1295, + "step": 24599 + }, + { + "epoch": 0.59, + "learning_rate": 7.544588772315625e-06, + "loss": 1.1357, + "step": 24600 + }, + { + "epoch": 0.59, + "learning_rate": 7.543833730305797e-06, + "loss": 1.1245, + "step": 24601 + }, + { + "epoch": 0.59, + "learning_rate": 7.543078703196956e-06, + "loss": 1.1347, + "step": 24602 + }, + { + "epoch": 0.59, + "learning_rate": 7.542323690993678e-06, + "loss": 1.2445, + "step": 24603 + }, + { + "epoch": 0.59, + "learning_rate": 7.541568693700543e-06, + "loss": 1.169, + "step": 24604 + }, + { + "epoch": 0.59, + "learning_rate": 7.540813711322131e-06, + "loss": 1.0142, + "step": 24605 + }, + { + "epoch": 0.59, + "learning_rate": 7.540058743863026e-06, + "loss": 1.1331, + "step": 24606 + }, + { + "epoch": 0.59, + "learning_rate": 7.539303791327806e-06, + "loss": 0.7956, + "step": 24607 + }, + { + "epoch": 0.59, + "learning_rate": 7.538548853721052e-06, + "loss": 0.9302, + "step": 24608 + }, + { + "epoch": 0.59, + "learning_rate": 7.537793931047339e-06, + "loss": 1.0896, + "step": 24609 + }, + { + "epoch": 0.59, + "learning_rate": 7.537039023311255e-06, + "loss": 1.0215, + "step": 24610 + }, + { + "epoch": 0.59, + "learning_rate": 7.536284130517375e-06, + "loss": 1.0119, + "step": 24611 + }, + { + "epoch": 0.59, + "learning_rate": 7.535529252670279e-06, + "loss": 1.3084, + "step": 24612 + }, + { + "epoch": 0.59, + "learning_rate": 7.534774389774546e-06, + "loss": 0.8789, + "step": 24613 + }, + { + "epoch": 0.59, + "learning_rate": 7.534019541834761e-06, + "loss": 0.8996, + "step": 24614 + }, + { + "epoch": 0.59, + "learning_rate": 7.533264708855496e-06, + "loss": 1.0414, + "step": 24615 + }, + { + "epoch": 0.59, + "learning_rate": 7.532509890841336e-06, + "loss": 1.2156, + "step": 24616 + }, + { + "epoch": 0.59, + "learning_rate": 7.5317550877968546e-06, + "loss": 0.8494, + "step": 24617 + }, + { + "epoch": 0.59, + "learning_rate": 7.531000299726638e-06, + "loss": 1.0032, + "step": 24618 + }, + { + "epoch": 0.59, + "learning_rate": 7.530245526635262e-06, + "loss": 1.006, + "step": 24619 + }, + { + "epoch": 0.59, + "learning_rate": 7.5294907685273035e-06, + "loss": 1.0596, + "step": 24620 + }, + { + "epoch": 0.59, + "learning_rate": 7.5287360254073425e-06, + "loss": 1.1755, + "step": 24621 + }, + { + "epoch": 0.59, + "learning_rate": 7.527981297279961e-06, + "loss": 1.3475, + "step": 24622 + }, + { + "epoch": 0.59, + "learning_rate": 7.527226584149736e-06, + "loss": 1.1199, + "step": 24623 + }, + { + "epoch": 0.59, + "learning_rate": 7.526471886021245e-06, + "loss": 1.1289, + "step": 24624 + }, + { + "epoch": 0.59, + "learning_rate": 7.5257172028990655e-06, + "loss": 0.9269, + "step": 24625 + }, + { + "epoch": 0.59, + "learning_rate": 7.5249625347877785e-06, + "loss": 1.0903, + "step": 24626 + }, + { + "epoch": 0.59, + "learning_rate": 7.524207881691963e-06, + "loss": 0.7361, + "step": 24627 + }, + { + "epoch": 0.59, + "learning_rate": 7.523453243616195e-06, + "loss": 1.1748, + "step": 24628 + }, + { + "epoch": 0.59, + "learning_rate": 7.522698620565053e-06, + "loss": 1.1898, + "step": 24629 + }, + { + "epoch": 0.59, + "learning_rate": 7.521944012543117e-06, + "loss": 1.1082, + "step": 24630 + }, + { + "epoch": 0.59, + "learning_rate": 7.521189419554964e-06, + "loss": 1.1019, + "step": 24631 + }, + { + "epoch": 0.59, + "learning_rate": 7.520434841605168e-06, + "loss": 1.1073, + "step": 24632 + }, + { + "epoch": 0.59, + "learning_rate": 7.519680278698316e-06, + "loss": 1.0034, + "step": 24633 + }, + { + "epoch": 0.59, + "learning_rate": 7.518925730838978e-06, + "loss": 1.0303, + "step": 24634 + }, + { + "epoch": 0.59, + "learning_rate": 7.518171198031734e-06, + "loss": 0.8528, + "step": 24635 + }, + { + "epoch": 0.59, + "learning_rate": 7.51741668028116e-06, + "loss": 0.9691, + "step": 24636 + }, + { + "epoch": 0.59, + "learning_rate": 7.516662177591838e-06, + "loss": 0.9392, + "step": 24637 + }, + { + "epoch": 0.59, + "learning_rate": 7.515907689968341e-06, + "loss": 1.1782, + "step": 24638 + }, + { + "epoch": 0.59, + "learning_rate": 7.515153217415248e-06, + "loss": 0.8825, + "step": 24639 + }, + { + "epoch": 0.59, + "learning_rate": 7.514398759937135e-06, + "loss": 1.0888, + "step": 24640 + }, + { + "epoch": 0.59, + "learning_rate": 7.513644317538581e-06, + "loss": 1.1374, + "step": 24641 + }, + { + "epoch": 0.59, + "learning_rate": 7.512889890224163e-06, + "loss": 1.0364, + "step": 24642 + }, + { + "epoch": 0.59, + "learning_rate": 7.5121354779984565e-06, + "loss": 1.208, + "step": 24643 + }, + { + "epoch": 0.59, + "learning_rate": 7.511381080866036e-06, + "loss": 0.9354, + "step": 24644 + }, + { + "epoch": 0.59, + "learning_rate": 7.510626698831485e-06, + "loss": 1.0663, + "step": 24645 + }, + { + "epoch": 0.59, + "learning_rate": 7.509872331899375e-06, + "loss": 1.1858, + "step": 24646 + }, + { + "epoch": 0.59, + "learning_rate": 7.509117980074284e-06, + "loss": 1.1012, + "step": 24647 + }, + { + "epoch": 0.59, + "learning_rate": 7.508363643360786e-06, + "loss": 1.1003, + "step": 24648 + }, + { + "epoch": 0.59, + "learning_rate": 7.507609321763462e-06, + "loss": 0.9723, + "step": 24649 + }, + { + "epoch": 0.59, + "learning_rate": 7.506855015286887e-06, + "loss": 1.2919, + "step": 24650 + }, + { + "epoch": 0.59, + "learning_rate": 7.506100723935635e-06, + "loss": 1.1215, + "step": 24651 + }, + { + "epoch": 0.59, + "learning_rate": 7.505346447714281e-06, + "loss": 1.3455, + "step": 24652 + }, + { + "epoch": 0.59, + "learning_rate": 7.504592186627406e-06, + "loss": 1.0997, + "step": 24653 + }, + { + "epoch": 0.59, + "learning_rate": 7.503837940679584e-06, + "loss": 1.1907, + "step": 24654 + }, + { + "epoch": 0.59, + "learning_rate": 7.503083709875389e-06, + "loss": 1.2545, + "step": 24655 + }, + { + "epoch": 0.59, + "learning_rate": 7.502329494219396e-06, + "loss": 0.8943, + "step": 24656 + }, + { + "epoch": 0.59, + "learning_rate": 7.501575293716185e-06, + "loss": 1.3062, + "step": 24657 + }, + { + "epoch": 0.59, + "learning_rate": 7.500821108370329e-06, + "loss": 1.0961, + "step": 24658 + }, + { + "epoch": 0.59, + "learning_rate": 7.500066938186404e-06, + "loss": 1.0335, + "step": 24659 + }, + { + "epoch": 0.59, + "learning_rate": 7.499312783168981e-06, + "loss": 1.0591, + "step": 24660 + }, + { + "epoch": 0.59, + "learning_rate": 7.498558643322642e-06, + "loss": 1.2704, + "step": 24661 + }, + { + "epoch": 0.59, + "learning_rate": 7.497804518651959e-06, + "loss": 1.1632, + "step": 24662 + }, + { + "epoch": 0.59, + "learning_rate": 7.497050409161508e-06, + "loss": 1.141, + "step": 24663 + }, + { + "epoch": 0.59, + "learning_rate": 7.496296314855862e-06, + "loss": 0.993, + "step": 24664 + }, + { + "epoch": 0.59, + "learning_rate": 7.495542235739599e-06, + "loss": 1.1384, + "step": 24665 + }, + { + "epoch": 0.59, + "learning_rate": 7.494788171817286e-06, + "loss": 0.9751, + "step": 24666 + }, + { + "epoch": 0.59, + "learning_rate": 7.494034123093511e-06, + "loss": 1.3351, + "step": 24667 + }, + { + "epoch": 0.59, + "learning_rate": 7.493280089572841e-06, + "loss": 1.0544, + "step": 24668 + }, + { + "epoch": 0.59, + "learning_rate": 7.492526071259847e-06, + "loss": 1.2043, + "step": 24669 + }, + { + "epoch": 0.59, + "learning_rate": 7.491772068159111e-06, + "loss": 1.1315, + "step": 24670 + }, + { + "epoch": 0.59, + "learning_rate": 7.491018080275204e-06, + "loss": 1.2547, + "step": 24671 + }, + { + "epoch": 0.59, + "learning_rate": 7.4902641076127e-06, + "loss": 1.1125, + "step": 24672 + }, + { + "epoch": 0.59, + "learning_rate": 7.489510150176172e-06, + "loss": 0.9052, + "step": 24673 + }, + { + "epoch": 0.59, + "learning_rate": 7.488756207970196e-06, + "loss": 1.1019, + "step": 24674 + }, + { + "epoch": 0.59, + "learning_rate": 7.488002280999347e-06, + "loss": 0.9902, + "step": 24675 + }, + { + "epoch": 0.59, + "learning_rate": 7.4872483692681966e-06, + "loss": 1.1665, + "step": 24676 + }, + { + "epoch": 0.59, + "learning_rate": 7.486494472781316e-06, + "loss": 1.0015, + "step": 24677 + }, + { + "epoch": 0.59, + "learning_rate": 7.485740591543286e-06, + "loss": 1.1329, + "step": 24678 + }, + { + "epoch": 0.59, + "learning_rate": 7.484986725558676e-06, + "loss": 0.963, + "step": 24679 + }, + { + "epoch": 0.59, + "learning_rate": 7.484232874832059e-06, + "loss": 1.1199, + "step": 24680 + }, + { + "epoch": 0.59, + "learning_rate": 7.483479039368009e-06, + "loss": 1.1425, + "step": 24681 + }, + { + "epoch": 0.59, + "learning_rate": 7.482725219171101e-06, + "loss": 0.8663, + "step": 24682 + }, + { + "epoch": 0.59, + "learning_rate": 7.481971414245907e-06, + "loss": 0.9536, + "step": 24683 + }, + { + "epoch": 0.59, + "learning_rate": 7.481217624596999e-06, + "loss": 1.1389, + "step": 24684 + }, + { + "epoch": 0.59, + "learning_rate": 7.48046385022895e-06, + "loss": 1.0972, + "step": 24685 + }, + { + "epoch": 0.59, + "learning_rate": 7.479710091146337e-06, + "loss": 0.8938, + "step": 24686 + }, + { + "epoch": 0.59, + "learning_rate": 7.4789563473537295e-06, + "loss": 1.0337, + "step": 24687 + }, + { + "epoch": 0.59, + "learning_rate": 7.478202618855701e-06, + "loss": 1.1581, + "step": 24688 + }, + { + "epoch": 0.59, + "learning_rate": 7.477448905656822e-06, + "loss": 1.227, + "step": 24689 + }, + { + "epoch": 0.59, + "learning_rate": 7.476695207761668e-06, + "loss": 1.1889, + "step": 24690 + }, + { + "epoch": 0.59, + "learning_rate": 7.475941525174813e-06, + "loss": 0.9631, + "step": 24691 + }, + { + "epoch": 0.59, + "learning_rate": 7.475187857900825e-06, + "loss": 1.1238, + "step": 24692 + }, + { + "epoch": 0.59, + "learning_rate": 7.4744342059442785e-06, + "loss": 0.9133, + "step": 24693 + }, + { + "epoch": 0.59, + "learning_rate": 7.473680569309746e-06, + "loss": 0.9604, + "step": 24694 + }, + { + "epoch": 0.59, + "learning_rate": 7.4729269480018e-06, + "loss": 1.0904, + "step": 24695 + }, + { + "epoch": 0.59, + "learning_rate": 7.472173342025011e-06, + "loss": 1.2006, + "step": 24696 + }, + { + "epoch": 0.59, + "learning_rate": 7.47141975138395e-06, + "loss": 0.9879, + "step": 24697 + }, + { + "epoch": 0.59, + "learning_rate": 7.470666176083193e-06, + "loss": 1.2385, + "step": 24698 + }, + { + "epoch": 0.59, + "learning_rate": 7.4699126161273085e-06, + "loss": 0.8153, + "step": 24699 + }, + { + "epoch": 0.59, + "learning_rate": 7.469159071520871e-06, + "loss": 1.0431, + "step": 24700 + }, + { + "epoch": 0.59, + "learning_rate": 7.468405542268446e-06, + "loss": 0.8247, + "step": 24701 + }, + { + "epoch": 0.59, + "learning_rate": 7.467652028374613e-06, + "loss": 1.0259, + "step": 24702 + }, + { + "epoch": 0.59, + "learning_rate": 7.4668985298439376e-06, + "loss": 0.964, + "step": 24703 + }, + { + "epoch": 0.59, + "learning_rate": 7.466145046680994e-06, + "loss": 1.2762, + "step": 24704 + }, + { + "epoch": 0.59, + "learning_rate": 7.46539157889035e-06, + "loss": 0.9015, + "step": 24705 + }, + { + "epoch": 0.59, + "learning_rate": 7.464638126476581e-06, + "loss": 1.1258, + "step": 24706 + }, + { + "epoch": 0.59, + "learning_rate": 7.4638846894442566e-06, + "loss": 1.0609, + "step": 24707 + }, + { + "epoch": 0.59, + "learning_rate": 7.463131267797947e-06, + "loss": 0.9799, + "step": 24708 + }, + { + "epoch": 0.59, + "learning_rate": 7.4623778615422205e-06, + "loss": 1.0316, + "step": 24709 + }, + { + "epoch": 0.59, + "learning_rate": 7.461624470681654e-06, + "loss": 0.9783, + "step": 24710 + }, + { + "epoch": 0.59, + "learning_rate": 7.460871095220814e-06, + "loss": 0.9241, + "step": 24711 + }, + { + "epoch": 0.59, + "learning_rate": 7.460117735164271e-06, + "loss": 1.2223, + "step": 24712 + }, + { + "epoch": 0.59, + "learning_rate": 7.459364390516597e-06, + "loss": 1.1307, + "step": 24713 + }, + { + "epoch": 0.59, + "learning_rate": 7.458611061282363e-06, + "loss": 1.0611, + "step": 24714 + }, + { + "epoch": 0.59, + "learning_rate": 7.457857747466136e-06, + "loss": 1.1938, + "step": 24715 + }, + { + "epoch": 0.59, + "learning_rate": 7.457104449072488e-06, + "loss": 1.075, + "step": 24716 + }, + { + "epoch": 0.59, + "learning_rate": 7.45635116610599e-06, + "loss": 0.9616, + "step": 24717 + }, + { + "epoch": 0.59, + "learning_rate": 7.455597898571212e-06, + "loss": 1.1625, + "step": 24718 + }, + { + "epoch": 0.59, + "learning_rate": 7.454844646472721e-06, + "loss": 0.8304, + "step": 24719 + }, + { + "epoch": 0.59, + "learning_rate": 7.45409140981509e-06, + "loss": 1.0747, + "step": 24720 + }, + { + "epoch": 0.59, + "learning_rate": 7.453338188602888e-06, + "loss": 1.2595, + "step": 24721 + }, + { + "epoch": 0.59, + "learning_rate": 7.452584982840682e-06, + "loss": 1.1883, + "step": 24722 + }, + { + "epoch": 0.59, + "learning_rate": 7.451831792533046e-06, + "loss": 1.2228, + "step": 24723 + }, + { + "epoch": 0.59, + "learning_rate": 7.451078617684546e-06, + "loss": 0.8841, + "step": 24724 + }, + { + "epoch": 0.59, + "learning_rate": 7.450325458299754e-06, + "loss": 1.0722, + "step": 24725 + }, + { + "epoch": 0.59, + "learning_rate": 7.449572314383237e-06, + "loss": 1.0227, + "step": 24726 + }, + { + "epoch": 0.59, + "learning_rate": 7.448819185939564e-06, + "loss": 1.2205, + "step": 24727 + }, + { + "epoch": 0.59, + "learning_rate": 7.448066072973306e-06, + "loss": 0.9224, + "step": 24728 + }, + { + "epoch": 0.59, + "learning_rate": 7.447312975489029e-06, + "loss": 1.1041, + "step": 24729 + }, + { + "epoch": 0.59, + "learning_rate": 7.446559893491306e-06, + "loss": 0.9164, + "step": 24730 + }, + { + "epoch": 0.59, + "learning_rate": 7.445806826984704e-06, + "loss": 1.0028, + "step": 24731 + }, + { + "epoch": 0.59, + "learning_rate": 7.445053775973787e-06, + "loss": 1.0528, + "step": 24732 + }, + { + "epoch": 0.59, + "learning_rate": 7.444300740463133e-06, + "loss": 1.303, + "step": 24733 + }, + { + "epoch": 0.59, + "learning_rate": 7.443547720457302e-06, + "loss": 0.9912, + "step": 24734 + }, + { + "epoch": 0.59, + "learning_rate": 7.442794715960868e-06, + "loss": 1.0412, + "step": 24735 + }, + { + "epoch": 0.59, + "learning_rate": 7.442041726978394e-06, + "loss": 0.8692, + "step": 24736 + }, + { + "epoch": 0.59, + "learning_rate": 7.441288753514453e-06, + "loss": 1.1251, + "step": 24737 + }, + { + "epoch": 0.59, + "learning_rate": 7.440535795573611e-06, + "loss": 0.9494, + "step": 24738 + }, + { + "epoch": 0.59, + "learning_rate": 7.439782853160438e-06, + "loss": 1.0505, + "step": 24739 + }, + { + "epoch": 0.59, + "learning_rate": 7.439029926279498e-06, + "loss": 1.1252, + "step": 24740 + }, + { + "epoch": 0.59, + "learning_rate": 7.43827701493536e-06, + "loss": 1.0059, + "step": 24741 + }, + { + "epoch": 0.6, + "learning_rate": 7.437524119132596e-06, + "loss": 1.0431, + "step": 24742 + }, + { + "epoch": 0.6, + "learning_rate": 7.436771238875768e-06, + "loss": 0.6813, + "step": 24743 + }, + { + "epoch": 0.6, + "learning_rate": 7.436018374169447e-06, + "loss": 1.3672, + "step": 24744 + }, + { + "epoch": 0.6, + "learning_rate": 7.435265525018199e-06, + "loss": 1.0237, + "step": 24745 + }, + { + "epoch": 0.6, + "learning_rate": 7.434512691426593e-06, + "loss": 1.1146, + "step": 24746 + }, + { + "epoch": 0.6, + "learning_rate": 7.433759873399194e-06, + "loss": 1.321, + "step": 24747 + }, + { + "epoch": 0.6, + "learning_rate": 7.43300707094057e-06, + "loss": 1.0598, + "step": 24748 + }, + { + "epoch": 0.6, + "learning_rate": 7.432254284055289e-06, + "loss": 1.2876, + "step": 24749 + }, + { + "epoch": 0.6, + "learning_rate": 7.431501512747916e-06, + "loss": 1.0567, + "step": 24750 + }, + { + "epoch": 0.6, + "learning_rate": 7.430748757023022e-06, + "loss": 1.1868, + "step": 24751 + }, + { + "epoch": 0.6, + "learning_rate": 7.429996016885167e-06, + "loss": 1.076, + "step": 24752 + }, + { + "epoch": 0.6, + "learning_rate": 7.429243292338925e-06, + "loss": 0.9034, + "step": 24753 + }, + { + "epoch": 0.6, + "learning_rate": 7.428490583388859e-06, + "loss": 0.9626, + "step": 24754 + }, + { + "epoch": 0.6, + "learning_rate": 7.4277378900395345e-06, + "loss": 1.2253, + "step": 24755 + }, + { + "epoch": 0.6, + "learning_rate": 7.4269852122955174e-06, + "loss": 0.8855, + "step": 24756 + }, + { + "epoch": 0.6, + "learning_rate": 7.426232550161378e-06, + "loss": 1.1175, + "step": 24757 + }, + { + "epoch": 0.6, + "learning_rate": 7.425479903641682e-06, + "loss": 1.2571, + "step": 24758 + }, + { + "epoch": 0.6, + "learning_rate": 7.4247272727409904e-06, + "loss": 1.0251, + "step": 24759 + }, + { + "epoch": 0.6, + "learning_rate": 7.423974657463875e-06, + "loss": 1.0124, + "step": 24760 + }, + { + "epoch": 0.6, + "learning_rate": 7.423222057814898e-06, + "loss": 1.2175, + "step": 24761 + }, + { + "epoch": 0.6, + "learning_rate": 7.4224694737986295e-06, + "loss": 0.9888, + "step": 24762 + }, + { + "epoch": 0.6, + "learning_rate": 7.421716905419629e-06, + "loss": 0.9402, + "step": 24763 + }, + { + "epoch": 0.6, + "learning_rate": 7.420964352682467e-06, + "loss": 1.119, + "step": 24764 + }, + { + "epoch": 0.6, + "learning_rate": 7.420211815591709e-06, + "loss": 1.023, + "step": 24765 + }, + { + "epoch": 0.6, + "learning_rate": 7.4194592941519185e-06, + "loss": 0.8877, + "step": 24766 + }, + { + "epoch": 0.6, + "learning_rate": 7.4187067883676575e-06, + "loss": 1.2251, + "step": 24767 + }, + { + "epoch": 0.6, + "learning_rate": 7.417954298243499e-06, + "loss": 1.2733, + "step": 24768 + }, + { + "epoch": 0.6, + "learning_rate": 7.417201823784006e-06, + "loss": 0.9917, + "step": 24769 + }, + { + "epoch": 0.6, + "learning_rate": 7.41644936499374e-06, + "loss": 1.0228, + "step": 24770 + }, + { + "epoch": 0.6, + "learning_rate": 7.415696921877267e-06, + "loss": 1.1091, + "step": 24771 + }, + { + "epoch": 0.6, + "learning_rate": 7.414944494439153e-06, + "loss": 1.1204, + "step": 24772 + }, + { + "epoch": 0.6, + "learning_rate": 7.414192082683964e-06, + "loss": 0.9971, + "step": 24773 + }, + { + "epoch": 0.6, + "learning_rate": 7.413439686616264e-06, + "loss": 1.0308, + "step": 24774 + }, + { + "epoch": 0.6, + "learning_rate": 7.4126873062406155e-06, + "loss": 1.188, + "step": 24775 + }, + { + "epoch": 0.6, + "learning_rate": 7.411934941561585e-06, + "loss": 1.0967, + "step": 24776 + }, + { + "epoch": 0.6, + "learning_rate": 7.411182592583737e-06, + "loss": 1.0179, + "step": 24777 + }, + { + "epoch": 0.6, + "learning_rate": 7.410430259311636e-06, + "loss": 1.0704, + "step": 24778 + }, + { + "epoch": 0.6, + "learning_rate": 7.409677941749842e-06, + "loss": 1.0153, + "step": 24779 + }, + { + "epoch": 0.6, + "learning_rate": 7.408925639902926e-06, + "loss": 0.967, + "step": 24780 + }, + { + "epoch": 0.6, + "learning_rate": 7.4081733537754475e-06, + "loss": 1.1346, + "step": 24781 + }, + { + "epoch": 0.6, + "learning_rate": 7.407421083371973e-06, + "loss": 1.3811, + "step": 24782 + }, + { + "epoch": 0.6, + "learning_rate": 7.406668828697062e-06, + "loss": 1.1887, + "step": 24783 + }, + { + "epoch": 0.6, + "learning_rate": 7.4059165897552845e-06, + "loss": 1.1386, + "step": 24784 + }, + { + "epoch": 0.6, + "learning_rate": 7.405164366551199e-06, + "loss": 1.1292, + "step": 24785 + }, + { + "epoch": 0.6, + "learning_rate": 7.404412159089372e-06, + "loss": 1.0862, + "step": 24786 + }, + { + "epoch": 0.6, + "learning_rate": 7.403659967374363e-06, + "loss": 1.0644, + "step": 24787 + }, + { + "epoch": 0.6, + "learning_rate": 7.402907791410741e-06, + "loss": 0.8399, + "step": 24788 + }, + { + "epoch": 0.6, + "learning_rate": 7.402155631203067e-06, + "loss": 1.0206, + "step": 24789 + }, + { + "epoch": 0.6, + "learning_rate": 7.401403486755903e-06, + "loss": 1.0491, + "step": 24790 + }, + { + "epoch": 0.6, + "learning_rate": 7.400651358073811e-06, + "loss": 0.9248, + "step": 24791 + }, + { + "epoch": 0.6, + "learning_rate": 7.3998992451613596e-06, + "loss": 0.9016, + "step": 24792 + }, + { + "epoch": 0.6, + "learning_rate": 7.399147148023106e-06, + "loss": 0.9866, + "step": 24793 + }, + { + "epoch": 0.6, + "learning_rate": 7.3983950666636154e-06, + "loss": 1.0361, + "step": 24794 + }, + { + "epoch": 0.6, + "learning_rate": 7.3976430010874466e-06, + "loss": 1.4237, + "step": 24795 + }, + { + "epoch": 0.6, + "learning_rate": 7.39689095129917e-06, + "loss": 0.9828, + "step": 24796 + }, + { + "epoch": 0.6, + "learning_rate": 7.396138917303342e-06, + "loss": 1.1309, + "step": 24797 + }, + { + "epoch": 0.6, + "learning_rate": 7.395386899104526e-06, + "loss": 1.0245, + "step": 24798 + }, + { + "epoch": 0.6, + "learning_rate": 7.394634896707284e-06, + "loss": 1.0126, + "step": 24799 + }, + { + "epoch": 0.6, + "learning_rate": 7.393882910116182e-06, + "loss": 0.8866, + "step": 24800 + }, + { + "epoch": 0.6, + "learning_rate": 7.39313093933578e-06, + "loss": 1.0247, + "step": 24801 + }, + { + "epoch": 0.6, + "learning_rate": 7.392378984370636e-06, + "loss": 1.0395, + "step": 24802 + }, + { + "epoch": 0.6, + "learning_rate": 7.391627045225318e-06, + "loss": 1.1628, + "step": 24803 + }, + { + "epoch": 0.6, + "learning_rate": 7.3908751219043835e-06, + "loss": 0.7694, + "step": 24804 + }, + { + "epoch": 0.6, + "learning_rate": 7.390123214412398e-06, + "loss": 1.0505, + "step": 24805 + }, + { + "epoch": 0.6, + "learning_rate": 7.389371322753918e-06, + "loss": 1.0577, + "step": 24806 + }, + { + "epoch": 0.6, + "learning_rate": 7.388619446933512e-06, + "loss": 1.2159, + "step": 24807 + }, + { + "epoch": 0.6, + "learning_rate": 7.387867586955735e-06, + "loss": 1.0785, + "step": 24808 + }, + { + "epoch": 0.6, + "learning_rate": 7.387115742825153e-06, + "loss": 1.1968, + "step": 24809 + }, + { + "epoch": 0.6, + "learning_rate": 7.386363914546322e-06, + "loss": 1.0495, + "step": 24810 + }, + { + "epoch": 0.6, + "learning_rate": 7.385612102123807e-06, + "loss": 0.9425, + "step": 24811 + }, + { + "epoch": 0.6, + "learning_rate": 7.384860305562171e-06, + "loss": 1.1551, + "step": 24812 + }, + { + "epoch": 0.6, + "learning_rate": 7.384108524865972e-06, + "loss": 1.1674, + "step": 24813 + }, + { + "epoch": 0.6, + "learning_rate": 7.383356760039768e-06, + "loss": 1.1302, + "step": 24814 + }, + { + "epoch": 0.6, + "learning_rate": 7.382605011088125e-06, + "loss": 0.9677, + "step": 24815 + }, + { + "epoch": 0.6, + "learning_rate": 7.381853278015603e-06, + "loss": 1.1189, + "step": 24816 + }, + { + "epoch": 0.6, + "learning_rate": 7.381101560826761e-06, + "loss": 1.1976, + "step": 24817 + }, + { + "epoch": 0.6, + "learning_rate": 7.3803498595261565e-06, + "loss": 1.2171, + "step": 24818 + }, + { + "epoch": 0.6, + "learning_rate": 7.379598174118358e-06, + "loss": 1.106, + "step": 24819 + }, + { + "epoch": 0.6, + "learning_rate": 7.3788465046079195e-06, + "loss": 0.9119, + "step": 24820 + }, + { + "epoch": 0.6, + "learning_rate": 7.378094850999402e-06, + "loss": 1.1515, + "step": 24821 + }, + { + "epoch": 0.6, + "learning_rate": 7.377343213297366e-06, + "loss": 1.1069, + "step": 24822 + }, + { + "epoch": 0.6, + "learning_rate": 7.376591591506373e-06, + "loss": 0.9965, + "step": 24823 + }, + { + "epoch": 0.6, + "learning_rate": 7.3758399856309805e-06, + "loss": 1.0087, + "step": 24824 + }, + { + "epoch": 0.6, + "learning_rate": 7.375088395675751e-06, + "loss": 0.9879, + "step": 24825 + }, + { + "epoch": 0.6, + "learning_rate": 7.37433682164524e-06, + "loss": 1.0413, + "step": 24826 + }, + { + "epoch": 0.6, + "learning_rate": 7.373585263544012e-06, + "loss": 1.1212, + "step": 24827 + }, + { + "epoch": 0.6, + "learning_rate": 7.3728337213766245e-06, + "loss": 0.9555, + "step": 24828 + }, + { + "epoch": 0.6, + "learning_rate": 7.372082195147637e-06, + "loss": 1.1264, + "step": 24829 + }, + { + "epoch": 0.6, + "learning_rate": 7.371330684861607e-06, + "loss": 1.1111, + "step": 24830 + }, + { + "epoch": 0.6, + "learning_rate": 7.370579190523097e-06, + "loss": 1.0189, + "step": 24831 + }, + { + "epoch": 0.6, + "learning_rate": 7.369827712136661e-06, + "loss": 0.9589, + "step": 24832 + }, + { + "epoch": 0.6, + "learning_rate": 7.369076249706866e-06, + "loss": 1.0799, + "step": 24833 + }, + { + "epoch": 0.6, + "learning_rate": 7.368324803238266e-06, + "loss": 1.0433, + "step": 24834 + }, + { + "epoch": 0.6, + "learning_rate": 7.3675733727354185e-06, + "loss": 1.1101, + "step": 24835 + }, + { + "epoch": 0.6, + "learning_rate": 7.366821958202885e-06, + "loss": 1.134, + "step": 24836 + }, + { + "epoch": 0.6, + "learning_rate": 7.366070559645224e-06, + "loss": 1.1691, + "step": 24837 + }, + { + "epoch": 0.6, + "learning_rate": 7.365319177066994e-06, + "loss": 0.8168, + "step": 24838 + }, + { + "epoch": 0.6, + "learning_rate": 7.364567810472752e-06, + "loss": 1.2504, + "step": 24839 + }, + { + "epoch": 0.6, + "learning_rate": 7.363816459867057e-06, + "loss": 1.1671, + "step": 24840 + }, + { + "epoch": 0.6, + "learning_rate": 7.3630651252544685e-06, + "loss": 1.4092, + "step": 24841 + }, + { + "epoch": 0.6, + "learning_rate": 7.362313806639543e-06, + "loss": 0.9892, + "step": 24842 + }, + { + "epoch": 0.6, + "learning_rate": 7.361562504026838e-06, + "loss": 1.1998, + "step": 24843 + }, + { + "epoch": 0.6, + "learning_rate": 7.360811217420915e-06, + "loss": 1.0623, + "step": 24844 + }, + { + "epoch": 0.6, + "learning_rate": 7.3600599468263275e-06, + "loss": 1.3567, + "step": 24845 + }, + { + "epoch": 0.6, + "learning_rate": 7.3593086922476375e-06, + "loss": 1.0843, + "step": 24846 + }, + { + "epoch": 0.6, + "learning_rate": 7.3585574536893974e-06, + "loss": 1.0447, + "step": 24847 + }, + { + "epoch": 0.6, + "learning_rate": 7.35780623115617e-06, + "loss": 0.9804, + "step": 24848 + }, + { + "epoch": 0.6, + "learning_rate": 7.3570550246525116e-06, + "loss": 1.0615, + "step": 24849 + }, + { + "epoch": 0.6, + "learning_rate": 7.356303834182978e-06, + "loss": 1.1598, + "step": 24850 + }, + { + "epoch": 0.6, + "learning_rate": 7.355552659752125e-06, + "loss": 1.0975, + "step": 24851 + }, + { + "epoch": 0.6, + "learning_rate": 7.354801501364514e-06, + "loss": 0.9258, + "step": 24852 + }, + { + "epoch": 0.6, + "learning_rate": 7.354050359024702e-06, + "loss": 1.1431, + "step": 24853 + }, + { + "epoch": 0.6, + "learning_rate": 7.353299232737242e-06, + "loss": 1.3567, + "step": 24854 + }, + { + "epoch": 0.6, + "learning_rate": 7.352548122506691e-06, + "loss": 1.0345, + "step": 24855 + }, + { + "epoch": 0.6, + "learning_rate": 7.351797028337611e-06, + "loss": 1.0635, + "step": 24856 + }, + { + "epoch": 0.6, + "learning_rate": 7.351045950234554e-06, + "loss": 1.2313, + "step": 24857 + }, + { + "epoch": 0.6, + "learning_rate": 7.35029488820208e-06, + "loss": 1.0524, + "step": 24858 + }, + { + "epoch": 0.6, + "learning_rate": 7.349543842244739e-06, + "loss": 1.0279, + "step": 24859 + }, + { + "epoch": 0.6, + "learning_rate": 7.348792812367097e-06, + "loss": 1.2347, + "step": 24860 + }, + { + "epoch": 0.6, + "learning_rate": 7.348041798573704e-06, + "loss": 1.0861, + "step": 24861 + }, + { + "epoch": 0.6, + "learning_rate": 7.347290800869118e-06, + "loss": 0.9079, + "step": 24862 + }, + { + "epoch": 0.6, + "learning_rate": 7.346539819257893e-06, + "loss": 0.8136, + "step": 24863 + }, + { + "epoch": 0.6, + "learning_rate": 7.345788853744589e-06, + "loss": 1.0309, + "step": 24864 + }, + { + "epoch": 0.6, + "learning_rate": 7.345037904333758e-06, + "loss": 1.1587, + "step": 24865 + }, + { + "epoch": 0.6, + "learning_rate": 7.344286971029959e-06, + "loss": 1.3024, + "step": 24866 + }, + { + "epoch": 0.6, + "learning_rate": 7.343536053837743e-06, + "loss": 1.1484, + "step": 24867 + }, + { + "epoch": 0.6, + "learning_rate": 7.342785152761673e-06, + "loss": 1.3654, + "step": 24868 + }, + { + "epoch": 0.6, + "learning_rate": 7.342034267806299e-06, + "loss": 0.9528, + "step": 24869 + }, + { + "epoch": 0.6, + "learning_rate": 7.341283398976178e-06, + "loss": 0.933, + "step": 24870 + }, + { + "epoch": 0.6, + "learning_rate": 7.340532546275864e-06, + "loss": 1.2523, + "step": 24871 + }, + { + "epoch": 0.6, + "learning_rate": 7.3397817097099145e-06, + "loss": 1.1197, + "step": 24872 + }, + { + "epoch": 0.6, + "learning_rate": 7.3390308892828845e-06, + "loss": 0.9816, + "step": 24873 + }, + { + "epoch": 0.6, + "learning_rate": 7.338280084999328e-06, + "loss": 0.9929, + "step": 24874 + }, + { + "epoch": 0.6, + "learning_rate": 7.337529296863797e-06, + "loss": 1.1423, + "step": 24875 + }, + { + "epoch": 0.6, + "learning_rate": 7.3367785248808535e-06, + "loss": 1.197, + "step": 24876 + }, + { + "epoch": 0.6, + "learning_rate": 7.336027769055048e-06, + "loss": 1.051, + "step": 24877 + }, + { + "epoch": 0.6, + "learning_rate": 7.335277029390934e-06, + "loss": 1.1975, + "step": 24878 + }, + { + "epoch": 0.6, + "learning_rate": 7.334526305893066e-06, + "loss": 0.9063, + "step": 24879 + }, + { + "epoch": 0.6, + "learning_rate": 7.333775598566002e-06, + "loss": 0.9252, + "step": 24880 + }, + { + "epoch": 0.6, + "learning_rate": 7.3330249074142946e-06, + "loss": 0.8072, + "step": 24881 + }, + { + "epoch": 0.6, + "learning_rate": 7.332274232442496e-06, + "loss": 1.1559, + "step": 24882 + }, + { + "epoch": 0.6, + "learning_rate": 7.331523573655163e-06, + "loss": 1.1516, + "step": 24883 + }, + { + "epoch": 0.6, + "learning_rate": 7.330772931056849e-06, + "loss": 1.2425, + "step": 24884 + }, + { + "epoch": 0.6, + "learning_rate": 7.330022304652109e-06, + "loss": 1.3456, + "step": 24885 + }, + { + "epoch": 0.6, + "learning_rate": 7.329271694445492e-06, + "loss": 0.9145, + "step": 24886 + }, + { + "epoch": 0.6, + "learning_rate": 7.328521100441559e-06, + "loss": 1.0548, + "step": 24887 + }, + { + "epoch": 0.6, + "learning_rate": 7.327770522644859e-06, + "loss": 0.8698, + "step": 24888 + }, + { + "epoch": 0.6, + "learning_rate": 7.327019961059946e-06, + "loss": 1.3502, + "step": 24889 + }, + { + "epoch": 0.6, + "learning_rate": 7.326269415691375e-06, + "loss": 0.8777, + "step": 24890 + }, + { + "epoch": 0.6, + "learning_rate": 7.325518886543699e-06, + "loss": 1.0229, + "step": 24891 + }, + { + "epoch": 0.6, + "learning_rate": 7.324768373621471e-06, + "loss": 1.0793, + "step": 24892 + }, + { + "epoch": 0.6, + "learning_rate": 7.324017876929244e-06, + "loss": 1.0668, + "step": 24893 + }, + { + "epoch": 0.6, + "learning_rate": 7.323267396471568e-06, + "loss": 1.1831, + "step": 24894 + }, + { + "epoch": 0.6, + "learning_rate": 7.322516932253002e-06, + "loss": 1.278, + "step": 24895 + }, + { + "epoch": 0.6, + "learning_rate": 7.321766484278097e-06, + "loss": 1.2253, + "step": 24896 + }, + { + "epoch": 0.6, + "learning_rate": 7.321016052551403e-06, + "loss": 1.1135, + "step": 24897 + }, + { + "epoch": 0.6, + "learning_rate": 7.320265637077473e-06, + "loss": 0.8578, + "step": 24898 + }, + { + "epoch": 0.6, + "learning_rate": 7.319515237860863e-06, + "loss": 1.1346, + "step": 24899 + }, + { + "epoch": 0.6, + "learning_rate": 7.318764854906124e-06, + "loss": 1.1578, + "step": 24900 + }, + { + "epoch": 0.6, + "learning_rate": 7.318014488217809e-06, + "loss": 0.697, + "step": 24901 + }, + { + "epoch": 0.6, + "learning_rate": 7.317264137800466e-06, + "loss": 1.0052, + "step": 24902 + }, + { + "epoch": 0.6, + "learning_rate": 7.3165138036586515e-06, + "loss": 1.0591, + "step": 24903 + }, + { + "epoch": 0.6, + "learning_rate": 7.315763485796918e-06, + "loss": 1.1315, + "step": 24904 + }, + { + "epoch": 0.6, + "learning_rate": 7.315013184219816e-06, + "loss": 1.1063, + "step": 24905 + }, + { + "epoch": 0.6, + "learning_rate": 7.314262898931894e-06, + "loss": 1.18, + "step": 24906 + }, + { + "epoch": 0.6, + "learning_rate": 7.3135126299377115e-06, + "loss": 0.8734, + "step": 24907 + }, + { + "epoch": 0.6, + "learning_rate": 7.312762377241815e-06, + "loss": 0.8694, + "step": 24908 + }, + { + "epoch": 0.6, + "learning_rate": 7.312012140848757e-06, + "loss": 0.9986, + "step": 24909 + }, + { + "epoch": 0.6, + "learning_rate": 7.311261920763087e-06, + "loss": 0.8472, + "step": 24910 + }, + { + "epoch": 0.6, + "learning_rate": 7.31051171698936e-06, + "loss": 1.2224, + "step": 24911 + }, + { + "epoch": 0.6, + "learning_rate": 7.309761529532127e-06, + "loss": 1.1763, + "step": 24912 + }, + { + "epoch": 0.6, + "learning_rate": 7.309011358395937e-06, + "loss": 0.9496, + "step": 24913 + }, + { + "epoch": 0.6, + "learning_rate": 7.30826120358534e-06, + "loss": 1.2116, + "step": 24914 + }, + { + "epoch": 0.6, + "learning_rate": 7.307511065104892e-06, + "loss": 1.1474, + "step": 24915 + }, + { + "epoch": 0.6, + "learning_rate": 7.306760942959141e-06, + "loss": 1.0649, + "step": 24916 + }, + { + "epoch": 0.6, + "learning_rate": 7.306010837152637e-06, + "loss": 0.8837, + "step": 24917 + }, + { + "epoch": 0.6, + "learning_rate": 7.3052607476899305e-06, + "loss": 1.06, + "step": 24918 + }, + { + "epoch": 0.6, + "learning_rate": 7.3045106745755755e-06, + "loss": 1.1267, + "step": 24919 + }, + { + "epoch": 0.6, + "learning_rate": 7.30376061781412e-06, + "loss": 1.0155, + "step": 24920 + }, + { + "epoch": 0.6, + "learning_rate": 7.303010577410114e-06, + "loss": 1.08, + "step": 24921 + }, + { + "epoch": 0.6, + "learning_rate": 7.302260553368106e-06, + "loss": 1.0714, + "step": 24922 + }, + { + "epoch": 0.6, + "learning_rate": 7.301510545692652e-06, + "loss": 1.0918, + "step": 24923 + }, + { + "epoch": 0.6, + "learning_rate": 7.300760554388297e-06, + "loss": 1.0286, + "step": 24924 + }, + { + "epoch": 0.6, + "learning_rate": 7.300010579459594e-06, + "loss": 1.0063, + "step": 24925 + }, + { + "epoch": 0.6, + "learning_rate": 7.29926062091109e-06, + "loss": 1.1249, + "step": 24926 + }, + { + "epoch": 0.6, + "learning_rate": 7.2985106787473385e-06, + "loss": 1.1646, + "step": 24927 + }, + { + "epoch": 0.6, + "learning_rate": 7.2977607529728865e-06, + "loss": 1.1807, + "step": 24928 + }, + { + "epoch": 0.6, + "learning_rate": 7.297010843592283e-06, + "loss": 0.9534, + "step": 24929 + }, + { + "epoch": 0.6, + "learning_rate": 7.296260950610081e-06, + "loss": 1.2752, + "step": 24930 + }, + { + "epoch": 0.6, + "learning_rate": 7.295511074030828e-06, + "loss": 1.221, + "step": 24931 + }, + { + "epoch": 0.6, + "learning_rate": 7.2947612138590725e-06, + "loss": 1.1073, + "step": 24932 + }, + { + "epoch": 0.6, + "learning_rate": 7.294011370099363e-06, + "loss": 1.1217, + "step": 24933 + }, + { + "epoch": 0.6, + "learning_rate": 7.293261542756251e-06, + "loss": 1.1076, + "step": 24934 + }, + { + "epoch": 0.6, + "learning_rate": 7.292511731834286e-06, + "loss": 1.0395, + "step": 24935 + }, + { + "epoch": 0.6, + "learning_rate": 7.291761937338016e-06, + "loss": 1.0971, + "step": 24936 + }, + { + "epoch": 0.6, + "learning_rate": 7.291012159271985e-06, + "loss": 0.9998, + "step": 24937 + }, + { + "epoch": 0.6, + "learning_rate": 7.290262397640749e-06, + "loss": 1.0975, + "step": 24938 + }, + { + "epoch": 0.6, + "learning_rate": 7.289512652448855e-06, + "loss": 1.3236, + "step": 24939 + }, + { + "epoch": 0.6, + "learning_rate": 7.2887629237008485e-06, + "loss": 1.3187, + "step": 24940 + }, + { + "epoch": 0.6, + "learning_rate": 7.288013211401278e-06, + "loss": 0.881, + "step": 24941 + }, + { + "epoch": 0.6, + "learning_rate": 7.287263515554697e-06, + "loss": 1.0038, + "step": 24942 + }, + { + "epoch": 0.6, + "learning_rate": 7.286513836165648e-06, + "loss": 1.0811, + "step": 24943 + }, + { + "epoch": 0.6, + "learning_rate": 7.285764173238682e-06, + "loss": 0.8273, + "step": 24944 + }, + { + "epoch": 0.6, + "learning_rate": 7.285014526778345e-06, + "loss": 1.174, + "step": 24945 + }, + { + "epoch": 0.6, + "learning_rate": 7.284264896789187e-06, + "loss": 1.1823, + "step": 24946 + }, + { + "epoch": 0.6, + "learning_rate": 7.283515283275756e-06, + "loss": 0.8073, + "step": 24947 + }, + { + "epoch": 0.6, + "learning_rate": 7.282765686242599e-06, + "loss": 0.9912, + "step": 24948 + }, + { + "epoch": 0.6, + "learning_rate": 7.282016105694261e-06, + "loss": 1.1359, + "step": 24949 + }, + { + "epoch": 0.6, + "learning_rate": 7.281266541635293e-06, + "loss": 0.8785, + "step": 24950 + }, + { + "epoch": 0.6, + "learning_rate": 7.280516994070243e-06, + "loss": 1.3782, + "step": 24951 + }, + { + "epoch": 0.6, + "learning_rate": 7.279767463003656e-06, + "loss": 0.9531, + "step": 24952 + }, + { + "epoch": 0.6, + "learning_rate": 7.279017948440079e-06, + "loss": 0.9558, + "step": 24953 + }, + { + "epoch": 0.6, + "learning_rate": 7.278268450384061e-06, + "loss": 1.0269, + "step": 24954 + }, + { + "epoch": 0.6, + "learning_rate": 7.277518968840148e-06, + "loss": 1.0849, + "step": 24955 + }, + { + "epoch": 0.6, + "learning_rate": 7.276769503812889e-06, + "loss": 1.119, + "step": 24956 + }, + { + "epoch": 0.6, + "learning_rate": 7.276020055306824e-06, + "loss": 1.2277, + "step": 24957 + }, + { + "epoch": 0.6, + "learning_rate": 7.275270623326508e-06, + "loss": 1.0014, + "step": 24958 + }, + { + "epoch": 0.6, + "learning_rate": 7.274521207876483e-06, + "loss": 1.0521, + "step": 24959 + }, + { + "epoch": 0.6, + "learning_rate": 7.2737718089612985e-06, + "loss": 1.0282, + "step": 24960 + }, + { + "epoch": 0.6, + "learning_rate": 7.273022426585497e-06, + "loss": 1.0947, + "step": 24961 + }, + { + "epoch": 0.6, + "learning_rate": 7.272273060753629e-06, + "loss": 1.3569, + "step": 24962 + }, + { + "epoch": 0.6, + "learning_rate": 7.271523711470239e-06, + "loss": 0.8735, + "step": 24963 + }, + { + "epoch": 0.6, + "learning_rate": 7.270774378739871e-06, + "loss": 1.0749, + "step": 24964 + }, + { + "epoch": 0.6, + "learning_rate": 7.270025062567072e-06, + "loss": 0.6384, + "step": 24965 + }, + { + "epoch": 0.6, + "learning_rate": 7.269275762956392e-06, + "loss": 1.1007, + "step": 24966 + }, + { + "epoch": 0.6, + "learning_rate": 7.268526479912372e-06, + "loss": 1.0956, + "step": 24967 + }, + { + "epoch": 0.6, + "learning_rate": 7.26777721343956e-06, + "loss": 0.982, + "step": 24968 + }, + { + "epoch": 0.6, + "learning_rate": 7.267027963542499e-06, + "loss": 1.0315, + "step": 24969 + }, + { + "epoch": 0.6, + "learning_rate": 7.266278730225738e-06, + "loss": 1.1137, + "step": 24970 + }, + { + "epoch": 0.6, + "learning_rate": 7.265529513493821e-06, + "loss": 1.1004, + "step": 24971 + }, + { + "epoch": 0.6, + "learning_rate": 7.264780313351294e-06, + "loss": 0.751, + "step": 24972 + }, + { + "epoch": 0.6, + "learning_rate": 7.2640311298027e-06, + "loss": 0.8275, + "step": 24973 + }, + { + "epoch": 0.6, + "learning_rate": 7.263281962852586e-06, + "loss": 0.9223, + "step": 24974 + }, + { + "epoch": 0.6, + "learning_rate": 7.262532812505497e-06, + "loss": 1.0932, + "step": 24975 + }, + { + "epoch": 0.6, + "learning_rate": 7.261783678765977e-06, + "loss": 0.9579, + "step": 24976 + }, + { + "epoch": 0.6, + "learning_rate": 7.261034561638571e-06, + "loss": 0.9969, + "step": 24977 + }, + { + "epoch": 0.6, + "learning_rate": 7.260285461127825e-06, + "loss": 0.9627, + "step": 24978 + }, + { + "epoch": 0.6, + "learning_rate": 7.259536377238283e-06, + "loss": 1.1308, + "step": 24979 + }, + { + "epoch": 0.6, + "learning_rate": 7.258787309974488e-06, + "loss": 0.9784, + "step": 24980 + }, + { + "epoch": 0.6, + "learning_rate": 7.258038259340987e-06, + "loss": 1.0236, + "step": 24981 + }, + { + "epoch": 0.6, + "learning_rate": 7.257289225342323e-06, + "loss": 1.1703, + "step": 24982 + }, + { + "epoch": 0.6, + "learning_rate": 7.256540207983038e-06, + "loss": 0.9646, + "step": 24983 + }, + { + "epoch": 0.6, + "learning_rate": 7.255791207267679e-06, + "loss": 0.9339, + "step": 24984 + }, + { + "epoch": 0.6, + "learning_rate": 7.25504222320079e-06, + "loss": 0.7569, + "step": 24985 + }, + { + "epoch": 0.6, + "learning_rate": 7.254293255786915e-06, + "loss": 0.9544, + "step": 24986 + }, + { + "epoch": 0.6, + "learning_rate": 7.253544305030597e-06, + "loss": 1.0841, + "step": 24987 + }, + { + "epoch": 0.6, + "learning_rate": 7.2527953709363765e-06, + "loss": 1.214, + "step": 24988 + }, + { + "epoch": 0.6, + "learning_rate": 7.252046453508803e-06, + "loss": 0.9054, + "step": 24989 + }, + { + "epoch": 0.6, + "learning_rate": 7.251297552752418e-06, + "loss": 1.086, + "step": 24990 + }, + { + "epoch": 0.6, + "learning_rate": 7.250548668671764e-06, + "loss": 1.1893, + "step": 24991 + }, + { + "epoch": 0.6, + "learning_rate": 7.249799801271381e-06, + "loss": 1.2075, + "step": 24992 + }, + { + "epoch": 0.6, + "learning_rate": 7.2490509505558196e-06, + "loss": 1.123, + "step": 24993 + }, + { + "epoch": 0.6, + "learning_rate": 7.248302116529618e-06, + "loss": 1.035, + "step": 24994 + }, + { + "epoch": 0.6, + "learning_rate": 7.2475532991973205e-06, + "loss": 1.1089, + "step": 24995 + }, + { + "epoch": 0.6, + "learning_rate": 7.2468044985634665e-06, + "loss": 1.0209, + "step": 24996 + }, + { + "epoch": 0.6, + "learning_rate": 7.246055714632606e-06, + "loss": 1.0266, + "step": 24997 + }, + { + "epoch": 0.6, + "learning_rate": 7.245306947409276e-06, + "loss": 1.0437, + "step": 24998 + }, + { + "epoch": 0.6, + "learning_rate": 7.244558196898018e-06, + "loss": 1.2002, + "step": 24999 + }, + { + "epoch": 0.6, + "learning_rate": 7.243809463103381e-06, + "loss": 0.8851, + "step": 25000 + }, + { + "epoch": 0.6, + "learning_rate": 7.243060746029904e-06, + "loss": 1.2093, + "step": 25001 + }, + { + "epoch": 0.6, + "learning_rate": 7.242312045682127e-06, + "loss": 1.2627, + "step": 25002 + }, + { + "epoch": 0.6, + "learning_rate": 7.2415633620645966e-06, + "loss": 1.001, + "step": 25003 + }, + { + "epoch": 0.6, + "learning_rate": 7.24081469518185e-06, + "loss": 1.1723, + "step": 25004 + }, + { + "epoch": 0.6, + "learning_rate": 7.240066045038434e-06, + "loss": 1.0761, + "step": 25005 + }, + { + "epoch": 0.6, + "learning_rate": 7.239317411638886e-06, + "loss": 0.6923, + "step": 25006 + }, + { + "epoch": 0.6, + "learning_rate": 7.23856879498775e-06, + "loss": 1.1421, + "step": 25007 + }, + { + "epoch": 0.6, + "learning_rate": 7.2378201950895696e-06, + "loss": 1.2389, + "step": 25008 + }, + { + "epoch": 0.6, + "learning_rate": 7.237071611948883e-06, + "loss": 1.0457, + "step": 25009 + }, + { + "epoch": 0.6, + "learning_rate": 7.236323045570233e-06, + "loss": 1.0645, + "step": 25010 + }, + { + "epoch": 0.6, + "learning_rate": 7.235574495958162e-06, + "loss": 0.9495, + "step": 25011 + }, + { + "epoch": 0.6, + "learning_rate": 7.234825963117209e-06, + "loss": 1.0597, + "step": 25012 + }, + { + "epoch": 0.6, + "learning_rate": 7.234077447051917e-06, + "loss": 0.956, + "step": 25013 + }, + { + "epoch": 0.6, + "learning_rate": 7.2333289477668265e-06, + "loss": 0.7787, + "step": 25014 + }, + { + "epoch": 0.6, + "learning_rate": 7.232580465266479e-06, + "loss": 1.4178, + "step": 25015 + }, + { + "epoch": 0.6, + "learning_rate": 7.231831999555415e-06, + "loss": 0.9964, + "step": 25016 + }, + { + "epoch": 0.6, + "learning_rate": 7.231083550638174e-06, + "loss": 0.8525, + "step": 25017 + }, + { + "epoch": 0.6, + "learning_rate": 7.230335118519298e-06, + "loss": 1.216, + "step": 25018 + }, + { + "epoch": 0.6, + "learning_rate": 7.2295867032033285e-06, + "loss": 1.326, + "step": 25019 + }, + { + "epoch": 0.6, + "learning_rate": 7.228838304694805e-06, + "loss": 0.9241, + "step": 25020 + }, + { + "epoch": 0.6, + "learning_rate": 7.228089922998265e-06, + "loss": 0.9728, + "step": 25021 + }, + { + "epoch": 0.6, + "learning_rate": 7.227341558118254e-06, + "loss": 0.6231, + "step": 25022 + }, + { + "epoch": 0.6, + "learning_rate": 7.226593210059309e-06, + "loss": 1.2004, + "step": 25023 + }, + { + "epoch": 0.6, + "learning_rate": 7.225844878825971e-06, + "loss": 1.1646, + "step": 25024 + }, + { + "epoch": 0.6, + "learning_rate": 7.225096564422777e-06, + "loss": 1.1545, + "step": 25025 + }, + { + "epoch": 0.6, + "learning_rate": 7.224348266854273e-06, + "loss": 1.0675, + "step": 25026 + }, + { + "epoch": 0.6, + "learning_rate": 7.223599986124994e-06, + "loss": 1.0908, + "step": 25027 + }, + { + "epoch": 0.6, + "learning_rate": 7.2228517222394815e-06, + "loss": 1.1821, + "step": 25028 + }, + { + "epoch": 0.6, + "learning_rate": 7.222103475202272e-06, + "loss": 1.3173, + "step": 25029 + }, + { + "epoch": 0.6, + "learning_rate": 7.221355245017909e-06, + "loss": 1.252, + "step": 25030 + }, + { + "epoch": 0.6, + "learning_rate": 7.220607031690931e-06, + "loss": 0.9801, + "step": 25031 + }, + { + "epoch": 0.6, + "learning_rate": 7.219858835225876e-06, + "loss": 1.0475, + "step": 25032 + }, + { + "epoch": 0.6, + "learning_rate": 7.2191106556272815e-06, + "loss": 1.0549, + "step": 25033 + }, + { + "epoch": 0.6, + "learning_rate": 7.2183624928996905e-06, + "loss": 0.9681, + "step": 25034 + }, + { + "epoch": 0.6, + "learning_rate": 7.217614347047639e-06, + "loss": 1.1462, + "step": 25035 + }, + { + "epoch": 0.6, + "learning_rate": 7.216866218075668e-06, + "loss": 1.2386, + "step": 25036 + }, + { + "epoch": 0.6, + "learning_rate": 7.216118105988312e-06, + "loss": 1.2595, + "step": 25037 + }, + { + "epoch": 0.6, + "learning_rate": 7.215370010790117e-06, + "loss": 1.212, + "step": 25038 + }, + { + "epoch": 0.6, + "learning_rate": 7.214621932485615e-06, + "loss": 1.1733, + "step": 25039 + }, + { + "epoch": 0.6, + "learning_rate": 7.213873871079347e-06, + "loss": 0.9972, + "step": 25040 + }, + { + "epoch": 0.6, + "learning_rate": 7.2131258265758485e-06, + "loss": 1.1972, + "step": 25041 + }, + { + "epoch": 0.6, + "learning_rate": 7.212377798979664e-06, + "loss": 1.0963, + "step": 25042 + }, + { + "epoch": 0.6, + "learning_rate": 7.211629788295325e-06, + "loss": 1.0762, + "step": 25043 + }, + { + "epoch": 0.6, + "learning_rate": 7.210881794527373e-06, + "loss": 0.9215, + "step": 25044 + }, + { + "epoch": 0.6, + "learning_rate": 7.2101338176803445e-06, + "loss": 1.1771, + "step": 25045 + }, + { + "epoch": 0.6, + "learning_rate": 7.2093858577587795e-06, + "loss": 0.8483, + "step": 25046 + }, + { + "epoch": 0.6, + "learning_rate": 7.208637914767214e-06, + "loss": 1.067, + "step": 25047 + }, + { + "epoch": 0.6, + "learning_rate": 7.207889988710185e-06, + "loss": 1.1188, + "step": 25048 + }, + { + "epoch": 0.6, + "learning_rate": 7.207142079592228e-06, + "loss": 1.0224, + "step": 25049 + }, + { + "epoch": 0.6, + "learning_rate": 7.206394187417887e-06, + "loss": 0.9514, + "step": 25050 + }, + { + "epoch": 0.6, + "learning_rate": 7.205646312191694e-06, + "loss": 1.0519, + "step": 25051 + }, + { + "epoch": 0.6, + "learning_rate": 7.2048984539181876e-06, + "loss": 1.067, + "step": 25052 + }, + { + "epoch": 0.6, + "learning_rate": 7.204150612601904e-06, + "loss": 1.166, + "step": 25053 + }, + { + "epoch": 0.6, + "learning_rate": 7.203402788247383e-06, + "loss": 0.9152, + "step": 25054 + }, + { + "epoch": 0.6, + "learning_rate": 7.2026549808591585e-06, + "loss": 1.1492, + "step": 25055 + }, + { + "epoch": 0.6, + "learning_rate": 7.201907190441766e-06, + "loss": 1.1381, + "step": 25056 + }, + { + "epoch": 0.6, + "learning_rate": 7.201159416999747e-06, + "loss": 1.1081, + "step": 25057 + }, + { + "epoch": 0.6, + "learning_rate": 7.2004116605376364e-06, + "loss": 1.2727, + "step": 25058 + }, + { + "epoch": 0.6, + "learning_rate": 7.199663921059967e-06, + "loss": 1.0484, + "step": 25059 + }, + { + "epoch": 0.6, + "learning_rate": 7.198916198571278e-06, + "loss": 0.92, + "step": 25060 + }, + { + "epoch": 0.6, + "learning_rate": 7.1981684930761075e-06, + "loss": 0.7729, + "step": 25061 + }, + { + "epoch": 0.6, + "learning_rate": 7.19742080457899e-06, + "loss": 0.9768, + "step": 25062 + }, + { + "epoch": 0.6, + "learning_rate": 7.196673133084461e-06, + "loss": 1.1106, + "step": 25063 + }, + { + "epoch": 0.6, + "learning_rate": 7.195925478597054e-06, + "loss": 1.2514, + "step": 25064 + }, + { + "epoch": 0.6, + "learning_rate": 7.19517784112131e-06, + "loss": 1.0368, + "step": 25065 + }, + { + "epoch": 0.6, + "learning_rate": 7.194430220661762e-06, + "loss": 0.9228, + "step": 25066 + }, + { + "epoch": 0.6, + "learning_rate": 7.193682617222947e-06, + "loss": 1.0591, + "step": 25067 + }, + { + "epoch": 0.6, + "learning_rate": 7.192935030809396e-06, + "loss": 0.8799, + "step": 25068 + }, + { + "epoch": 0.6, + "learning_rate": 7.192187461425651e-06, + "loss": 1.2342, + "step": 25069 + }, + { + "epoch": 0.6, + "learning_rate": 7.191439909076243e-06, + "loss": 1.1038, + "step": 25070 + }, + { + "epoch": 0.6, + "learning_rate": 7.190692373765711e-06, + "loss": 0.8946, + "step": 25071 + }, + { + "epoch": 0.6, + "learning_rate": 7.189944855498583e-06, + "loss": 0.9448, + "step": 25072 + }, + { + "epoch": 0.6, + "learning_rate": 7.189197354279402e-06, + "loss": 0.7889, + "step": 25073 + }, + { + "epoch": 0.6, + "learning_rate": 7.188449870112699e-06, + "loss": 1.0577, + "step": 25074 + }, + { + "epoch": 0.6, + "learning_rate": 7.18770240300301e-06, + "loss": 1.1017, + "step": 25075 + }, + { + "epoch": 0.6, + "learning_rate": 7.186954952954865e-06, + "loss": 1.1849, + "step": 25076 + }, + { + "epoch": 0.6, + "learning_rate": 7.186207519972807e-06, + "loss": 0.9898, + "step": 25077 + }, + { + "epoch": 0.6, + "learning_rate": 7.185460104061364e-06, + "loss": 1.0872, + "step": 25078 + }, + { + "epoch": 0.6, + "learning_rate": 7.184712705225074e-06, + "loss": 1.1406, + "step": 25079 + }, + { + "epoch": 0.6, + "learning_rate": 7.183965323468468e-06, + "loss": 1.0585, + "step": 25080 + }, + { + "epoch": 0.6, + "learning_rate": 7.1832179587960825e-06, + "loss": 0.915, + "step": 25081 + }, + { + "epoch": 0.6, + "learning_rate": 7.182470611212453e-06, + "loss": 1.2639, + "step": 25082 + }, + { + "epoch": 0.6, + "learning_rate": 7.181723280722111e-06, + "loss": 0.9924, + "step": 25083 + }, + { + "epoch": 0.6, + "learning_rate": 7.180975967329589e-06, + "loss": 0.7386, + "step": 25084 + }, + { + "epoch": 0.6, + "learning_rate": 7.180228671039424e-06, + "loss": 1.1169, + "step": 25085 + }, + { + "epoch": 0.6, + "learning_rate": 7.179481391856149e-06, + "loss": 1.2129, + "step": 25086 + }, + { + "epoch": 0.6, + "learning_rate": 7.178734129784296e-06, + "loss": 1.2433, + "step": 25087 + }, + { + "epoch": 0.6, + "learning_rate": 7.177986884828398e-06, + "loss": 1.1774, + "step": 25088 + }, + { + "epoch": 0.6, + "learning_rate": 7.177239656992992e-06, + "loss": 0.908, + "step": 25089 + }, + { + "epoch": 0.6, + "learning_rate": 7.176492446282611e-06, + "loss": 0.9199, + "step": 25090 + }, + { + "epoch": 0.6, + "learning_rate": 7.175745252701782e-06, + "loss": 1.279, + "step": 25091 + }, + { + "epoch": 0.6, + "learning_rate": 7.1749980762550434e-06, + "loss": 1.1759, + "step": 25092 + }, + { + "epoch": 0.6, + "learning_rate": 7.174250916946927e-06, + "loss": 1.0966, + "step": 25093 + }, + { + "epoch": 0.6, + "learning_rate": 7.173503774781967e-06, + "loss": 0.7907, + "step": 25094 + }, + { + "epoch": 0.6, + "learning_rate": 7.172756649764695e-06, + "loss": 1.081, + "step": 25095 + }, + { + "epoch": 0.6, + "learning_rate": 7.17200954189964e-06, + "loss": 1.2747, + "step": 25096 + }, + { + "epoch": 0.6, + "learning_rate": 7.171262451191341e-06, + "loss": 1.0524, + "step": 25097 + }, + { + "epoch": 0.6, + "learning_rate": 7.170515377644327e-06, + "loss": 1.2934, + "step": 25098 + }, + { + "epoch": 0.6, + "learning_rate": 7.169768321263128e-06, + "loss": 0.9042, + "step": 25099 + }, + { + "epoch": 0.6, + "learning_rate": 7.169021282052283e-06, + "loss": 1.1427, + "step": 25100 + }, + { + "epoch": 0.6, + "learning_rate": 7.168274260016319e-06, + "loss": 1.0455, + "step": 25101 + }, + { + "epoch": 0.6, + "learning_rate": 7.167527255159769e-06, + "loss": 1.0196, + "step": 25102 + }, + { + "epoch": 0.6, + "learning_rate": 7.166780267487161e-06, + "loss": 1.1468, + "step": 25103 + }, + { + "epoch": 0.6, + "learning_rate": 7.166033297003035e-06, + "loss": 1.1595, + "step": 25104 + }, + { + "epoch": 0.6, + "learning_rate": 7.165286343711918e-06, + "loss": 0.8071, + "step": 25105 + }, + { + "epoch": 0.6, + "learning_rate": 7.164539407618342e-06, + "loss": 1.5268, + "step": 25106 + }, + { + "epoch": 0.6, + "learning_rate": 7.163792488726835e-06, + "loss": 0.7036, + "step": 25107 + }, + { + "epoch": 0.6, + "learning_rate": 7.163045587041936e-06, + "loss": 0.9236, + "step": 25108 + }, + { + "epoch": 0.6, + "learning_rate": 7.162298702568171e-06, + "loss": 1.1525, + "step": 25109 + }, + { + "epoch": 0.6, + "learning_rate": 7.1615518353100726e-06, + "loss": 1.0105, + "step": 25110 + }, + { + "epoch": 0.6, + "learning_rate": 7.160804985272169e-06, + "loss": 1.0398, + "step": 25111 + }, + { + "epoch": 0.6, + "learning_rate": 7.160058152458996e-06, + "loss": 1.2915, + "step": 25112 + }, + { + "epoch": 0.6, + "learning_rate": 7.159311336875083e-06, + "loss": 0.762, + "step": 25113 + }, + { + "epoch": 0.6, + "learning_rate": 7.158564538524959e-06, + "loss": 1.2592, + "step": 25114 + }, + { + "epoch": 0.6, + "learning_rate": 7.157817757413153e-06, + "loss": 0.9795, + "step": 25115 + }, + { + "epoch": 0.6, + "learning_rate": 7.1570709935442015e-06, + "loss": 0.9801, + "step": 25116 + }, + { + "epoch": 0.6, + "learning_rate": 7.156324246922631e-06, + "loss": 1.0259, + "step": 25117 + }, + { + "epoch": 0.6, + "learning_rate": 7.155577517552973e-06, + "loss": 1.1817, + "step": 25118 + }, + { + "epoch": 0.6, + "learning_rate": 7.1548308054397544e-06, + "loss": 1.0666, + "step": 25119 + }, + { + "epoch": 0.6, + "learning_rate": 7.1540841105875105e-06, + "loss": 1.2698, + "step": 25120 + }, + { + "epoch": 0.6, + "learning_rate": 7.153337433000769e-06, + "loss": 1.2266, + "step": 25121 + }, + { + "epoch": 0.6, + "learning_rate": 7.1525907726840595e-06, + "loss": 1.048, + "step": 25122 + }, + { + "epoch": 0.6, + "learning_rate": 7.15184412964191e-06, + "loss": 1.0057, + "step": 25123 + }, + { + "epoch": 0.6, + "learning_rate": 7.151097503878854e-06, + "loss": 1.2801, + "step": 25124 + }, + { + "epoch": 0.6, + "learning_rate": 7.15035089539942e-06, + "loss": 0.9818, + "step": 25125 + }, + { + "epoch": 0.6, + "learning_rate": 7.149604304208135e-06, + "loss": 0.8642, + "step": 25126 + }, + { + "epoch": 0.6, + "learning_rate": 7.1488577303095305e-06, + "loss": 0.8368, + "step": 25127 + }, + { + "epoch": 0.6, + "learning_rate": 7.1481111737081365e-06, + "loss": 1.0242, + "step": 25128 + }, + { + "epoch": 0.6, + "learning_rate": 7.147364634408482e-06, + "loss": 1.0142, + "step": 25129 + }, + { + "epoch": 0.6, + "learning_rate": 7.1466181124150935e-06, + "loss": 1.0632, + "step": 25130 + }, + { + "epoch": 0.6, + "learning_rate": 7.145871607732502e-06, + "loss": 1.1924, + "step": 25131 + }, + { + "epoch": 0.6, + "learning_rate": 7.145125120365237e-06, + "loss": 0.9015, + "step": 25132 + }, + { + "epoch": 0.6, + "learning_rate": 7.144378650317827e-06, + "loss": 0.9212, + "step": 25133 + }, + { + "epoch": 0.6, + "learning_rate": 7.143632197594799e-06, + "loss": 0.951, + "step": 25134 + }, + { + "epoch": 0.6, + "learning_rate": 7.142885762200681e-06, + "loss": 1.0401, + "step": 25135 + }, + { + "epoch": 0.6, + "learning_rate": 7.1421393441400046e-06, + "loss": 0.874, + "step": 25136 + }, + { + "epoch": 0.6, + "learning_rate": 7.141392943417298e-06, + "loss": 1.1563, + "step": 25137 + }, + { + "epoch": 0.6, + "learning_rate": 7.140646560037087e-06, + "loss": 1.1713, + "step": 25138 + }, + { + "epoch": 0.6, + "learning_rate": 7.139900194003898e-06, + "loss": 1.0892, + "step": 25139 + }, + { + "epoch": 0.6, + "learning_rate": 7.139153845322264e-06, + "loss": 1.1287, + "step": 25140 + }, + { + "epoch": 0.6, + "learning_rate": 7.138407513996713e-06, + "loss": 1.3613, + "step": 25141 + }, + { + "epoch": 0.6, + "learning_rate": 7.137661200031769e-06, + "loss": 0.8315, + "step": 25142 + }, + { + "epoch": 0.6, + "learning_rate": 7.136914903431959e-06, + "loss": 0.9948, + "step": 25143 + }, + { + "epoch": 0.6, + "learning_rate": 7.136168624201815e-06, + "loss": 1.3243, + "step": 25144 + }, + { + "epoch": 0.6, + "learning_rate": 7.135422362345862e-06, + "loss": 1.0726, + "step": 25145 + }, + { + "epoch": 0.6, + "learning_rate": 7.1346761178686265e-06, + "loss": 1.0892, + "step": 25146 + }, + { + "epoch": 0.6, + "learning_rate": 7.133929890774638e-06, + "loss": 1.2208, + "step": 25147 + }, + { + "epoch": 0.6, + "learning_rate": 7.133183681068423e-06, + "loss": 0.9603, + "step": 25148 + }, + { + "epoch": 0.6, + "learning_rate": 7.132437488754509e-06, + "loss": 0.9837, + "step": 25149 + }, + { + "epoch": 0.6, + "learning_rate": 7.131691313837419e-06, + "loss": 1.1835, + "step": 25150 + }, + { + "epoch": 0.6, + "learning_rate": 7.130945156321687e-06, + "loss": 1.2038, + "step": 25151 + }, + { + "epoch": 0.6, + "learning_rate": 7.130199016211834e-06, + "loss": 0.9669, + "step": 25152 + }, + { + "epoch": 0.6, + "learning_rate": 7.12945289351239e-06, + "loss": 0.9414, + "step": 25153 + }, + { + "epoch": 0.6, + "learning_rate": 7.128706788227877e-06, + "loss": 1.0811, + "step": 25154 + }, + { + "epoch": 0.6, + "learning_rate": 7.127960700362828e-06, + "loss": 0.8899, + "step": 25155 + }, + { + "epoch": 0.6, + "learning_rate": 7.127214629921765e-06, + "loss": 1.0259, + "step": 25156 + }, + { + "epoch": 0.6, + "learning_rate": 7.126468576909214e-06, + "loss": 0.954, + "step": 25157 + }, + { + "epoch": 0.61, + "learning_rate": 7.125722541329702e-06, + "loss": 1.1218, + "step": 25158 + }, + { + "epoch": 0.61, + "learning_rate": 7.124976523187757e-06, + "loss": 1.18, + "step": 25159 + }, + { + "epoch": 0.61, + "learning_rate": 7.124230522487903e-06, + "loss": 0.9105, + "step": 25160 + }, + { + "epoch": 0.61, + "learning_rate": 7.123484539234665e-06, + "loss": 1.027, + "step": 25161 + }, + { + "epoch": 0.61, + "learning_rate": 7.122738573432568e-06, + "loss": 0.8892, + "step": 25162 + }, + { + "epoch": 0.61, + "learning_rate": 7.121992625086143e-06, + "loss": 1.2171, + "step": 25163 + }, + { + "epoch": 0.61, + "learning_rate": 7.121246694199911e-06, + "loss": 0.9686, + "step": 25164 + }, + { + "epoch": 0.61, + "learning_rate": 7.120500780778397e-06, + "loss": 1.3247, + "step": 25165 + }, + { + "epoch": 0.61, + "learning_rate": 7.119754884826124e-06, + "loss": 0.8039, + "step": 25166 + }, + { + "epoch": 0.61, + "learning_rate": 7.119009006347625e-06, + "loss": 1.1471, + "step": 25167 + }, + { + "epoch": 0.61, + "learning_rate": 7.118263145347418e-06, + "loss": 1.0966, + "step": 25168 + }, + { + "epoch": 0.61, + "learning_rate": 7.117517301830034e-06, + "loss": 1.0292, + "step": 25169 + }, + { + "epoch": 0.61, + "learning_rate": 7.116771475799993e-06, + "loss": 1.1536, + "step": 25170 + }, + { + "epoch": 0.61, + "learning_rate": 7.116025667261823e-06, + "loss": 1.1012, + "step": 25171 + }, + { + "epoch": 0.61, + "learning_rate": 7.115279876220044e-06, + "loss": 1.204, + "step": 25172 + }, + { + "epoch": 0.61, + "learning_rate": 7.1145341026791845e-06, + "loss": 1.154, + "step": 25173 + }, + { + "epoch": 0.61, + "learning_rate": 7.11378834664377e-06, + "loss": 1.0626, + "step": 25174 + }, + { + "epoch": 0.61, + "learning_rate": 7.113042608118321e-06, + "loss": 1.0167, + "step": 25175 + }, + { + "epoch": 0.61, + "learning_rate": 7.112296887107361e-06, + "loss": 1.0841, + "step": 25176 + }, + { + "epoch": 0.61, + "learning_rate": 7.1115511836154205e-06, + "loss": 0.9302, + "step": 25177 + }, + { + "epoch": 0.61, + "learning_rate": 7.1108054976470185e-06, + "loss": 1.2399, + "step": 25178 + }, + { + "epoch": 0.61, + "learning_rate": 7.110059829206678e-06, + "loss": 0.8016, + "step": 25179 + }, + { + "epoch": 0.61, + "learning_rate": 7.1093141782989275e-06, + "loss": 1.0544, + "step": 25180 + }, + { + "epoch": 0.61, + "learning_rate": 7.108568544928287e-06, + "loss": 1.0562, + "step": 25181 + }, + { + "epoch": 0.61, + "learning_rate": 7.107822929099282e-06, + "loss": 1.124, + "step": 25182 + }, + { + "epoch": 0.61, + "learning_rate": 7.107077330816432e-06, + "loss": 0.9496, + "step": 25183 + }, + { + "epoch": 0.61, + "learning_rate": 7.106331750084266e-06, + "loss": 0.981, + "step": 25184 + }, + { + "epoch": 0.61, + "learning_rate": 7.105586186907304e-06, + "loss": 0.9164, + "step": 25185 + }, + { + "epoch": 0.61, + "learning_rate": 7.1048406412900695e-06, + "loss": 1.0635, + "step": 25186 + }, + { + "epoch": 0.61, + "learning_rate": 7.104095113237085e-06, + "loss": 0.8932, + "step": 25187 + }, + { + "epoch": 0.61, + "learning_rate": 7.1033496027528755e-06, + "loss": 1.2569, + "step": 25188 + }, + { + "epoch": 0.61, + "learning_rate": 7.1026041098419626e-06, + "loss": 0.9118, + "step": 25189 + }, + { + "epoch": 0.61, + "learning_rate": 7.101858634508869e-06, + "loss": 1.1168, + "step": 25190 + }, + { + "epoch": 0.61, + "learning_rate": 7.1011131767581165e-06, + "loss": 0.7662, + "step": 25191 + }, + { + "epoch": 0.61, + "learning_rate": 7.100367736594229e-06, + "loss": 1.0091, + "step": 25192 + }, + { + "epoch": 0.61, + "learning_rate": 7.09962231402173e-06, + "loss": 1.1202, + "step": 25193 + }, + { + "epoch": 0.61, + "learning_rate": 7.098876909045139e-06, + "loss": 1.0723, + "step": 25194 + }, + { + "epoch": 0.61, + "learning_rate": 7.098131521668978e-06, + "loss": 1.1265, + "step": 25195 + }, + { + "epoch": 0.61, + "learning_rate": 7.097386151897773e-06, + "loss": 0.8714, + "step": 25196 + }, + { + "epoch": 0.61, + "learning_rate": 7.096640799736042e-06, + "loss": 1.0717, + "step": 25197 + }, + { + "epoch": 0.61, + "learning_rate": 7.095895465188309e-06, + "loss": 1.0892, + "step": 25198 + }, + { + "epoch": 0.61, + "learning_rate": 7.095150148259093e-06, + "loss": 0.8183, + "step": 25199 + }, + { + "epoch": 0.61, + "learning_rate": 7.0944048489529205e-06, + "loss": 1.0802, + "step": 25200 + }, + { + "epoch": 0.61, + "learning_rate": 7.09365956727431e-06, + "loss": 1.1531, + "step": 25201 + }, + { + "epoch": 0.61, + "learning_rate": 7.092914303227783e-06, + "loss": 1.0942, + "step": 25202 + }, + { + "epoch": 0.61, + "learning_rate": 7.0921690568178595e-06, + "loss": 1.0119, + "step": 25203 + }, + { + "epoch": 0.61, + "learning_rate": 7.091423828049064e-06, + "loss": 1.1518, + "step": 25204 + }, + { + "epoch": 0.61, + "learning_rate": 7.090678616925916e-06, + "loss": 1.1253, + "step": 25205 + }, + { + "epoch": 0.61, + "learning_rate": 7.089933423452937e-06, + "loss": 1.2271, + "step": 25206 + }, + { + "epoch": 0.61, + "learning_rate": 7.0891882476346454e-06, + "loss": 0.9635, + "step": 25207 + }, + { + "epoch": 0.61, + "learning_rate": 7.0884430894755655e-06, + "loss": 0.846, + "step": 25208 + }, + { + "epoch": 0.61, + "learning_rate": 7.087697948980217e-06, + "loss": 1.1266, + "step": 25209 + }, + { + "epoch": 0.61, + "learning_rate": 7.08695282615312e-06, + "loss": 1.1519, + "step": 25210 + }, + { + "epoch": 0.61, + "learning_rate": 7.0862077209987925e-06, + "loss": 1.0415, + "step": 25211 + }, + { + "epoch": 0.61, + "learning_rate": 7.08546263352176e-06, + "loss": 1.1254, + "step": 25212 + }, + { + "epoch": 0.61, + "learning_rate": 7.0847175637265395e-06, + "loss": 0.9143, + "step": 25213 + }, + { + "epoch": 0.61, + "learning_rate": 7.0839725116176515e-06, + "loss": 1.2585, + "step": 25214 + }, + { + "epoch": 0.61, + "learning_rate": 7.083227477199616e-06, + "loss": 1.1116, + "step": 25215 + }, + { + "epoch": 0.61, + "learning_rate": 7.082482460476954e-06, + "loss": 1.0736, + "step": 25216 + }, + { + "epoch": 0.61, + "learning_rate": 7.0817374614541856e-06, + "loss": 1.0673, + "step": 25217 + }, + { + "epoch": 0.61, + "learning_rate": 7.080992480135829e-06, + "loss": 1.0873, + "step": 25218 + }, + { + "epoch": 0.61, + "learning_rate": 7.080247516526402e-06, + "loss": 1.3315, + "step": 25219 + }, + { + "epoch": 0.61, + "learning_rate": 7.07950257063043e-06, + "loss": 1.1595, + "step": 25220 + }, + { + "epoch": 0.61, + "learning_rate": 7.078757642452428e-06, + "loss": 1.1167, + "step": 25221 + }, + { + "epoch": 0.61, + "learning_rate": 7.078012731996915e-06, + "loss": 0.7854, + "step": 25222 + }, + { + "epoch": 0.61, + "learning_rate": 7.077267839268411e-06, + "loss": 1.2095, + "step": 25223 + }, + { + "epoch": 0.61, + "learning_rate": 7.0765229642714374e-06, + "loss": 1.0899, + "step": 25224 + }, + { + "epoch": 0.61, + "learning_rate": 7.075778107010511e-06, + "loss": 1.1693, + "step": 25225 + }, + { + "epoch": 0.61, + "learning_rate": 7.07503326749015e-06, + "loss": 1.2519, + "step": 25226 + }, + { + "epoch": 0.61, + "learning_rate": 7.074288445714875e-06, + "loss": 1.0722, + "step": 25227 + }, + { + "epoch": 0.61, + "learning_rate": 7.073543641689205e-06, + "loss": 1.0202, + "step": 25228 + }, + { + "epoch": 0.61, + "learning_rate": 7.072798855417656e-06, + "loss": 1.124, + "step": 25229 + }, + { + "epoch": 0.61, + "learning_rate": 7.072054086904747e-06, + "loss": 0.9773, + "step": 25230 + }, + { + "epoch": 0.61, + "learning_rate": 7.071309336154999e-06, + "loss": 1.0821, + "step": 25231 + }, + { + "epoch": 0.61, + "learning_rate": 7.070564603172927e-06, + "loss": 1.1634, + "step": 25232 + }, + { + "epoch": 0.61, + "learning_rate": 7.069819887963052e-06, + "loss": 1.2704, + "step": 25233 + }, + { + "epoch": 0.61, + "learning_rate": 7.069075190529888e-06, + "loss": 1.2309, + "step": 25234 + }, + { + "epoch": 0.61, + "learning_rate": 7.068330510877957e-06, + "loss": 1.1062, + "step": 25235 + }, + { + "epoch": 0.61, + "learning_rate": 7.067585849011775e-06, + "loss": 1.0388, + "step": 25236 + }, + { + "epoch": 0.61, + "learning_rate": 7.066841204935861e-06, + "loss": 1.1509, + "step": 25237 + }, + { + "epoch": 0.61, + "learning_rate": 7.066096578654729e-06, + "loss": 0.9953, + "step": 25238 + }, + { + "epoch": 0.61, + "learning_rate": 7.0653519701729025e-06, + "loss": 1.1855, + "step": 25239 + }, + { + "epoch": 0.61, + "learning_rate": 7.064607379494892e-06, + "loss": 0.8804, + "step": 25240 + }, + { + "epoch": 0.61, + "learning_rate": 7.0638628066252215e-06, + "loss": 1.074, + "step": 25241 + }, + { + "epoch": 0.61, + "learning_rate": 7.0631182515684e-06, + "loss": 1.1832, + "step": 25242 + }, + { + "epoch": 0.61, + "learning_rate": 7.0623737143289525e-06, + "loss": 0.9605, + "step": 25243 + }, + { + "epoch": 0.61, + "learning_rate": 7.061629194911392e-06, + "loss": 1.0785, + "step": 25244 + }, + { + "epoch": 0.61, + "learning_rate": 7.060884693320238e-06, + "loss": 1.1788, + "step": 25245 + }, + { + "epoch": 0.61, + "learning_rate": 7.0601402095600015e-06, + "loss": 1.1878, + "step": 25246 + }, + { + "epoch": 0.61, + "learning_rate": 7.0593957436352044e-06, + "loss": 0.8768, + "step": 25247 + }, + { + "epoch": 0.61, + "learning_rate": 7.058651295550363e-06, + "loss": 1.0603, + "step": 25248 + }, + { + "epoch": 0.61, + "learning_rate": 7.0579068653099925e-06, + "loss": 1.2644, + "step": 25249 + }, + { + "epoch": 0.61, + "learning_rate": 7.057162452918607e-06, + "loss": 1.0688, + "step": 25250 + }, + { + "epoch": 0.61, + "learning_rate": 7.056418058380725e-06, + "loss": 1.0419, + "step": 25251 + }, + { + "epoch": 0.61, + "learning_rate": 7.055673681700863e-06, + "loss": 1.0256, + "step": 25252 + }, + { + "epoch": 0.61, + "learning_rate": 7.054929322883537e-06, + "loss": 1.1115, + "step": 25253 + }, + { + "epoch": 0.61, + "learning_rate": 7.05418498193326e-06, + "loss": 1.2484, + "step": 25254 + }, + { + "epoch": 0.61, + "learning_rate": 7.053440658854551e-06, + "loss": 1.0711, + "step": 25255 + }, + { + "epoch": 0.61, + "learning_rate": 7.052696353651926e-06, + "loss": 0.9406, + "step": 25256 + }, + { + "epoch": 0.61, + "learning_rate": 7.0519520663298964e-06, + "loss": 0.6558, + "step": 25257 + }, + { + "epoch": 0.61, + "learning_rate": 7.0512077968929805e-06, + "loss": 1.0583, + "step": 25258 + }, + { + "epoch": 0.61, + "learning_rate": 7.0504635453456935e-06, + "loss": 0.9486, + "step": 25259 + }, + { + "epoch": 0.61, + "learning_rate": 7.049719311692552e-06, + "loss": 1.0514, + "step": 25260 + }, + { + "epoch": 0.61, + "learning_rate": 7.048975095938068e-06, + "loss": 1.0523, + "step": 25261 + }, + { + "epoch": 0.61, + "learning_rate": 7.048230898086759e-06, + "loss": 1.1981, + "step": 25262 + }, + { + "epoch": 0.61, + "learning_rate": 7.047486718143137e-06, + "loss": 1.1662, + "step": 25263 + }, + { + "epoch": 0.61, + "learning_rate": 7.046742556111721e-06, + "loss": 1.2883, + "step": 25264 + }, + { + "epoch": 0.61, + "learning_rate": 7.045998411997022e-06, + "loss": 0.8132, + "step": 25265 + }, + { + "epoch": 0.61, + "learning_rate": 7.045254285803556e-06, + "loss": 0.99, + "step": 25266 + }, + { + "epoch": 0.61, + "learning_rate": 7.044510177535837e-06, + "loss": 0.9893, + "step": 25267 + }, + { + "epoch": 0.61, + "learning_rate": 7.043766087198381e-06, + "loss": 1.1529, + "step": 25268 + }, + { + "epoch": 0.61, + "learning_rate": 7.043022014795698e-06, + "loss": 0.9118, + "step": 25269 + }, + { + "epoch": 0.61, + "learning_rate": 7.042277960332305e-06, + "loss": 0.8388, + "step": 25270 + }, + { + "epoch": 0.61, + "learning_rate": 7.041533923812718e-06, + "loss": 0.9205, + "step": 25271 + }, + { + "epoch": 0.61, + "learning_rate": 7.04078990524145e-06, + "loss": 0.8627, + "step": 25272 + }, + { + "epoch": 0.61, + "learning_rate": 7.04004590462301e-06, + "loss": 1.0489, + "step": 25273 + }, + { + "epoch": 0.61, + "learning_rate": 7.039301921961916e-06, + "loss": 1.0341, + "step": 25274 + }, + { + "epoch": 0.61, + "learning_rate": 7.038557957262682e-06, + "loss": 1.0204, + "step": 25275 + }, + { + "epoch": 0.61, + "learning_rate": 7.037814010529822e-06, + "loss": 0.8409, + "step": 25276 + }, + { + "epoch": 0.61, + "learning_rate": 7.037070081767844e-06, + "loss": 1.2137, + "step": 25277 + }, + { + "epoch": 0.61, + "learning_rate": 7.036326170981267e-06, + "loss": 1.1109, + "step": 25278 + }, + { + "epoch": 0.61, + "learning_rate": 7.035582278174602e-06, + "loss": 1.0394, + "step": 25279 + }, + { + "epoch": 0.61, + "learning_rate": 7.034838403352362e-06, + "loss": 1.2197, + "step": 25280 + }, + { + "epoch": 0.61, + "learning_rate": 7.034094546519059e-06, + "loss": 1.2176, + "step": 25281 + }, + { + "epoch": 0.61, + "learning_rate": 7.033350707679207e-06, + "loss": 1.1046, + "step": 25282 + }, + { + "epoch": 0.61, + "learning_rate": 7.03260688683732e-06, + "loss": 1.018, + "step": 25283 + }, + { + "epoch": 0.61, + "learning_rate": 7.0318630839979086e-06, + "loss": 0.8384, + "step": 25284 + }, + { + "epoch": 0.61, + "learning_rate": 7.031119299165483e-06, + "loss": 0.9365, + "step": 25285 + }, + { + "epoch": 0.61, + "learning_rate": 7.030375532344562e-06, + "loss": 1.1633, + "step": 25286 + }, + { + "epoch": 0.61, + "learning_rate": 7.0296317835396525e-06, + "loss": 1.1595, + "step": 25287 + }, + { + "epoch": 0.61, + "learning_rate": 7.02888805275527e-06, + "loss": 1.1352, + "step": 25288 + }, + { + "epoch": 0.61, + "learning_rate": 7.028144339995921e-06, + "loss": 1.0534, + "step": 25289 + }, + { + "epoch": 0.61, + "learning_rate": 7.027400645266123e-06, + "loss": 1.2084, + "step": 25290 + }, + { + "epoch": 0.61, + "learning_rate": 7.026656968570387e-06, + "loss": 1.0672, + "step": 25291 + }, + { + "epoch": 0.61, + "learning_rate": 7.025913309913223e-06, + "loss": 1.139, + "step": 25292 + }, + { + "epoch": 0.61, + "learning_rate": 7.025169669299143e-06, + "loss": 1.1007, + "step": 25293 + }, + { + "epoch": 0.61, + "learning_rate": 7.02442604673266e-06, + "loss": 1.112, + "step": 25294 + }, + { + "epoch": 0.61, + "learning_rate": 7.0236824422182844e-06, + "loss": 1.0251, + "step": 25295 + }, + { + "epoch": 0.61, + "learning_rate": 7.022938855760528e-06, + "loss": 1.1008, + "step": 25296 + }, + { + "epoch": 0.61, + "learning_rate": 7.022195287363898e-06, + "loss": 1.1906, + "step": 25297 + }, + { + "epoch": 0.61, + "learning_rate": 7.021451737032912e-06, + "loss": 1.0713, + "step": 25298 + }, + { + "epoch": 0.61, + "learning_rate": 7.020708204772077e-06, + "loss": 0.9951, + "step": 25299 + }, + { + "epoch": 0.61, + "learning_rate": 7.0199646905859055e-06, + "loss": 0.9501, + "step": 25300 + }, + { + "epoch": 0.61, + "learning_rate": 7.019221194478903e-06, + "loss": 1.0346, + "step": 25301 + }, + { + "epoch": 0.61, + "learning_rate": 7.018477716455588e-06, + "loss": 1.1838, + "step": 25302 + }, + { + "epoch": 0.61, + "learning_rate": 7.017734256520467e-06, + "loss": 1.1247, + "step": 25303 + }, + { + "epoch": 0.61, + "learning_rate": 7.016990814678052e-06, + "loss": 1.2716, + "step": 25304 + }, + { + "epoch": 0.61, + "learning_rate": 7.016247390932848e-06, + "loss": 1.2505, + "step": 25305 + }, + { + "epoch": 0.61, + "learning_rate": 7.015503985289374e-06, + "loss": 1.1375, + "step": 25306 + }, + { + "epoch": 0.61, + "learning_rate": 7.0147605977521325e-06, + "loss": 1.0383, + "step": 25307 + }, + { + "epoch": 0.61, + "learning_rate": 7.014017228325638e-06, + "loss": 1.0289, + "step": 25308 + }, + { + "epoch": 0.61, + "learning_rate": 7.013273877014396e-06, + "loss": 0.9711, + "step": 25309 + }, + { + "epoch": 0.61, + "learning_rate": 7.012530543822921e-06, + "loss": 1.1866, + "step": 25310 + }, + { + "epoch": 0.61, + "learning_rate": 7.01178722875572e-06, + "loss": 0.9948, + "step": 25311 + }, + { + "epoch": 0.61, + "learning_rate": 7.011043931817304e-06, + "loss": 1.024, + "step": 25312 + }, + { + "epoch": 0.61, + "learning_rate": 7.010300653012179e-06, + "loss": 1.0033, + "step": 25313 + }, + { + "epoch": 0.61, + "learning_rate": 7.009557392344859e-06, + "loss": 0.7471, + "step": 25314 + }, + { + "epoch": 0.61, + "learning_rate": 7.008814149819849e-06, + "loss": 0.9861, + "step": 25315 + }, + { + "epoch": 0.61, + "learning_rate": 7.008070925441663e-06, + "loss": 1.0147, + "step": 25316 + }, + { + "epoch": 0.61, + "learning_rate": 7.0073277192148026e-06, + "loss": 0.954, + "step": 25317 + }, + { + "epoch": 0.61, + "learning_rate": 7.006584531143784e-06, + "loss": 0.837, + "step": 25318 + }, + { + "epoch": 0.61, + "learning_rate": 7.005841361233112e-06, + "loss": 0.9842, + "step": 25319 + }, + { + "epoch": 0.61, + "learning_rate": 7.005098209487295e-06, + "loss": 0.8834, + "step": 25320 + }, + { + "epoch": 0.61, + "learning_rate": 7.004355075910845e-06, + "loss": 0.92, + "step": 25321 + }, + { + "epoch": 0.61, + "learning_rate": 7.003611960508268e-06, + "loss": 1.0189, + "step": 25322 + }, + { + "epoch": 0.61, + "learning_rate": 7.002868863284073e-06, + "loss": 0.9647, + "step": 25323 + }, + { + "epoch": 0.61, + "learning_rate": 7.0021257842427645e-06, + "loss": 1.0124, + "step": 25324 + }, + { + "epoch": 0.61, + "learning_rate": 7.001382723388856e-06, + "loss": 0.9556, + "step": 25325 + }, + { + "epoch": 0.61, + "learning_rate": 7.000639680726854e-06, + "loss": 1.015, + "step": 25326 + }, + { + "epoch": 0.61, + "learning_rate": 6.999896656261265e-06, + "loss": 1.0432, + "step": 25327 + }, + { + "epoch": 0.61, + "learning_rate": 6.999153649996595e-06, + "loss": 1.0463, + "step": 25328 + }, + { + "epoch": 0.61, + "learning_rate": 6.998410661937356e-06, + "loss": 1.2072, + "step": 25329 + }, + { + "epoch": 0.61, + "learning_rate": 6.997667692088053e-06, + "loss": 1.1122, + "step": 25330 + }, + { + "epoch": 0.61, + "learning_rate": 6.996924740453195e-06, + "loss": 0.93, + "step": 25331 + }, + { + "epoch": 0.61, + "learning_rate": 6.9961818070372845e-06, + "loss": 1.1102, + "step": 25332 + }, + { + "epoch": 0.61, + "learning_rate": 6.9954388918448354e-06, + "loss": 1.1785, + "step": 25333 + }, + { + "epoch": 0.61, + "learning_rate": 6.994695994880349e-06, + "loss": 1.1783, + "step": 25334 + }, + { + "epoch": 0.61, + "learning_rate": 6.99395311614834e-06, + "loss": 1.1657, + "step": 25335 + }, + { + "epoch": 0.61, + "learning_rate": 6.993210255653309e-06, + "loss": 0.958, + "step": 25336 + }, + { + "epoch": 0.61, + "learning_rate": 6.992467413399764e-06, + "loss": 1.2624, + "step": 25337 + }, + { + "epoch": 0.61, + "learning_rate": 6.991724589392208e-06, + "loss": 1.2344, + "step": 25338 + }, + { + "epoch": 0.61, + "learning_rate": 6.9909817836351554e-06, + "loss": 1.3748, + "step": 25339 + }, + { + "epoch": 0.61, + "learning_rate": 6.990238996133109e-06, + "loss": 0.9736, + "step": 25340 + }, + { + "epoch": 0.61, + "learning_rate": 6.989496226890573e-06, + "loss": 0.9614, + "step": 25341 + }, + { + "epoch": 0.61, + "learning_rate": 6.988753475912054e-06, + "loss": 0.9124, + "step": 25342 + }, + { + "epoch": 0.61, + "learning_rate": 6.988010743202062e-06, + "loss": 1.0634, + "step": 25343 + }, + { + "epoch": 0.61, + "learning_rate": 6.9872680287651e-06, + "loss": 1.0573, + "step": 25344 + }, + { + "epoch": 0.61, + "learning_rate": 6.986525332605673e-06, + "loss": 1.2051, + "step": 25345 + }, + { + "epoch": 0.61, + "learning_rate": 6.985782654728288e-06, + "loss": 1.0333, + "step": 25346 + }, + { + "epoch": 0.61, + "learning_rate": 6.985039995137452e-06, + "loss": 1.3275, + "step": 25347 + }, + { + "epoch": 0.61, + "learning_rate": 6.9842973538376675e-06, + "loss": 1.1339, + "step": 25348 + }, + { + "epoch": 0.61, + "learning_rate": 6.983554730833444e-06, + "loss": 0.9714, + "step": 25349 + }, + { + "epoch": 0.61, + "learning_rate": 6.982812126129281e-06, + "loss": 1.1032, + "step": 25350 + }, + { + "epoch": 0.61, + "learning_rate": 6.98206953972969e-06, + "loss": 1.0092, + "step": 25351 + }, + { + "epoch": 0.61, + "learning_rate": 6.981326971639173e-06, + "loss": 0.8651, + "step": 25352 + }, + { + "epoch": 0.61, + "learning_rate": 6.980584421862233e-06, + "loss": 0.8713, + "step": 25353 + }, + { + "epoch": 0.61, + "learning_rate": 6.979841890403378e-06, + "loss": 1.0458, + "step": 25354 + }, + { + "epoch": 0.61, + "learning_rate": 6.979099377267113e-06, + "loss": 0.9402, + "step": 25355 + }, + { + "epoch": 0.61, + "learning_rate": 6.978356882457942e-06, + "loss": 0.9399, + "step": 25356 + }, + { + "epoch": 0.61, + "learning_rate": 6.9776144059803665e-06, + "loss": 0.7551, + "step": 25357 + }, + { + "epoch": 0.61, + "learning_rate": 6.9768719478388956e-06, + "loss": 1.1259, + "step": 25358 + }, + { + "epoch": 0.61, + "learning_rate": 6.976129508038032e-06, + "loss": 1.0763, + "step": 25359 + }, + { + "epoch": 0.61, + "learning_rate": 6.975387086582279e-06, + "loss": 0.9775, + "step": 25360 + }, + { + "epoch": 0.61, + "learning_rate": 6.97464468347614e-06, + "loss": 1.005, + "step": 25361 + }, + { + "epoch": 0.61, + "learning_rate": 6.97390229872412e-06, + "loss": 1.0589, + "step": 25362 + }, + { + "epoch": 0.61, + "learning_rate": 6.973159932330724e-06, + "loss": 1.1385, + "step": 25363 + }, + { + "epoch": 0.61, + "learning_rate": 6.972417584300455e-06, + "loss": 1.0091, + "step": 25364 + }, + { + "epoch": 0.61, + "learning_rate": 6.971675254637814e-06, + "loss": 1.0894, + "step": 25365 + }, + { + "epoch": 0.61, + "learning_rate": 6.970932943347308e-06, + "loss": 0.8316, + "step": 25366 + }, + { + "epoch": 0.61, + "learning_rate": 6.9701906504334415e-06, + "loss": 1.1997, + "step": 25367 + }, + { + "epoch": 0.61, + "learning_rate": 6.969448375900715e-06, + "loss": 0.9112, + "step": 25368 + }, + { + "epoch": 0.61, + "learning_rate": 6.9687061197536295e-06, + "loss": 1.1149, + "step": 25369 + }, + { + "epoch": 0.61, + "learning_rate": 6.967963881996693e-06, + "loss": 0.8642, + "step": 25370 + }, + { + "epoch": 0.61, + "learning_rate": 6.9672216626344065e-06, + "loss": 0.5733, + "step": 25371 + }, + { + "epoch": 0.61, + "learning_rate": 6.966479461671274e-06, + "loss": 1.1093, + "step": 25372 + }, + { + "epoch": 0.61, + "learning_rate": 6.965737279111795e-06, + "loss": 1.0291, + "step": 25373 + }, + { + "epoch": 0.61, + "learning_rate": 6.964995114960475e-06, + "loss": 1.2837, + "step": 25374 + }, + { + "epoch": 0.61, + "learning_rate": 6.964252969221816e-06, + "loss": 0.9541, + "step": 25375 + }, + { + "epoch": 0.61, + "learning_rate": 6.963510841900321e-06, + "loss": 1.372, + "step": 25376 + }, + { + "epoch": 0.61, + "learning_rate": 6.962768733000489e-06, + "loss": 1.1361, + "step": 25377 + }, + { + "epoch": 0.61, + "learning_rate": 6.962026642526827e-06, + "loss": 1.1556, + "step": 25378 + }, + { + "epoch": 0.61, + "learning_rate": 6.961284570483835e-06, + "loss": 1.2823, + "step": 25379 + }, + { + "epoch": 0.61, + "learning_rate": 6.960542516876014e-06, + "loss": 1.092, + "step": 25380 + }, + { + "epoch": 0.61, + "learning_rate": 6.959800481707865e-06, + "loss": 1.1783, + "step": 25381 + }, + { + "epoch": 0.61, + "learning_rate": 6.959058464983892e-06, + "loss": 1.1353, + "step": 25382 + }, + { + "epoch": 0.61, + "learning_rate": 6.9583164667085974e-06, + "loss": 1.227, + "step": 25383 + }, + { + "epoch": 0.61, + "learning_rate": 6.957574486886483e-06, + "loss": 0.8915, + "step": 25384 + }, + { + "epoch": 0.61, + "learning_rate": 6.9568325255220435e-06, + "loss": 1.1114, + "step": 25385 + }, + { + "epoch": 0.61, + "learning_rate": 6.9560905826197885e-06, + "loss": 1.2274, + "step": 25386 + }, + { + "epoch": 0.61, + "learning_rate": 6.9553486581842155e-06, + "loss": 0.9476, + "step": 25387 + }, + { + "epoch": 0.61, + "learning_rate": 6.954606752219828e-06, + "loss": 1.1912, + "step": 25388 + }, + { + "epoch": 0.61, + "learning_rate": 6.953864864731122e-06, + "loss": 1.1963, + "step": 25389 + }, + { + "epoch": 0.61, + "learning_rate": 6.953122995722601e-06, + "loss": 1.1831, + "step": 25390 + }, + { + "epoch": 0.61, + "learning_rate": 6.952381145198769e-06, + "loss": 1.0759, + "step": 25391 + }, + { + "epoch": 0.61, + "learning_rate": 6.951639313164122e-06, + "loss": 0.9887, + "step": 25392 + }, + { + "epoch": 0.61, + "learning_rate": 6.9508974996231614e-06, + "loss": 0.8741, + "step": 25393 + }, + { + "epoch": 0.61, + "learning_rate": 6.950155704580391e-06, + "loss": 1.114, + "step": 25394 + }, + { + "epoch": 0.61, + "learning_rate": 6.949413928040307e-06, + "loss": 1.167, + "step": 25395 + }, + { + "epoch": 0.61, + "learning_rate": 6.94867217000741e-06, + "loss": 1.1376, + "step": 25396 + }, + { + "epoch": 0.61, + "learning_rate": 6.9479304304862025e-06, + "loss": 0.7085, + "step": 25397 + }, + { + "epoch": 0.61, + "learning_rate": 6.947188709481185e-06, + "loss": 1.1758, + "step": 25398 + }, + { + "epoch": 0.61, + "learning_rate": 6.946447006996853e-06, + "loss": 1.0867, + "step": 25399 + }, + { + "epoch": 0.61, + "learning_rate": 6.945705323037709e-06, + "loss": 0.7329, + "step": 25400 + }, + { + "epoch": 0.61, + "learning_rate": 6.944963657608254e-06, + "loss": 1.0312, + "step": 25401 + }, + { + "epoch": 0.61, + "learning_rate": 6.9442220107129844e-06, + "loss": 0.9917, + "step": 25402 + }, + { + "epoch": 0.61, + "learning_rate": 6.9434803823564025e-06, + "loss": 1.08, + "step": 25403 + }, + { + "epoch": 0.61, + "learning_rate": 6.942738772543004e-06, + "loss": 1.1969, + "step": 25404 + }, + { + "epoch": 0.61, + "learning_rate": 6.941997181277291e-06, + "loss": 1.1293, + "step": 25405 + }, + { + "epoch": 0.61, + "learning_rate": 6.941255608563764e-06, + "loss": 0.8688, + "step": 25406 + }, + { + "epoch": 0.61, + "learning_rate": 6.940514054406918e-06, + "loss": 0.8889, + "step": 25407 + }, + { + "epoch": 0.61, + "learning_rate": 6.939772518811252e-06, + "loss": 1.1207, + "step": 25408 + }, + { + "epoch": 0.61, + "learning_rate": 6.939031001781269e-06, + "loss": 0.8955, + "step": 25409 + }, + { + "epoch": 0.61, + "learning_rate": 6.938289503321465e-06, + "loss": 1.0787, + "step": 25410 + }, + { + "epoch": 0.61, + "learning_rate": 6.937548023436338e-06, + "loss": 1.0829, + "step": 25411 + }, + { + "epoch": 0.61, + "learning_rate": 6.936806562130385e-06, + "loss": 1.3113, + "step": 25412 + }, + { + "epoch": 0.61, + "learning_rate": 6.936065119408106e-06, + "loss": 1.1119, + "step": 25413 + }, + { + "epoch": 0.61, + "learning_rate": 6.935323695274002e-06, + "loss": 1.25, + "step": 25414 + }, + { + "epoch": 0.61, + "learning_rate": 6.9345822897325675e-06, + "loss": 1.0216, + "step": 25415 + }, + { + "epoch": 0.61, + "learning_rate": 6.933840902788298e-06, + "loss": 1.0698, + "step": 25416 + }, + { + "epoch": 0.61, + "learning_rate": 6.9330995344456985e-06, + "loss": 1.0063, + "step": 25417 + }, + { + "epoch": 0.61, + "learning_rate": 6.932358184709263e-06, + "loss": 1.1561, + "step": 25418 + }, + { + "epoch": 0.61, + "learning_rate": 6.931616853583487e-06, + "loss": 1.1995, + "step": 25419 + }, + { + "epoch": 0.61, + "learning_rate": 6.930875541072868e-06, + "loss": 1.2557, + "step": 25420 + }, + { + "epoch": 0.61, + "learning_rate": 6.930134247181908e-06, + "loss": 0.7564, + "step": 25421 + }, + { + "epoch": 0.61, + "learning_rate": 6.929392971915101e-06, + "loss": 0.9518, + "step": 25422 + }, + { + "epoch": 0.61, + "learning_rate": 6.928651715276944e-06, + "loss": 0.9098, + "step": 25423 + }, + { + "epoch": 0.61, + "learning_rate": 6.927910477271933e-06, + "loss": 1.0176, + "step": 25424 + }, + { + "epoch": 0.61, + "learning_rate": 6.9271692579045695e-06, + "loss": 1.1156, + "step": 25425 + }, + { + "epoch": 0.61, + "learning_rate": 6.926428057179346e-06, + "loss": 1.1835, + "step": 25426 + }, + { + "epoch": 0.61, + "learning_rate": 6.925686875100761e-06, + "loss": 1.1637, + "step": 25427 + }, + { + "epoch": 0.61, + "learning_rate": 6.92494571167331e-06, + "loss": 1.0525, + "step": 25428 + }, + { + "epoch": 0.61, + "learning_rate": 6.924204566901492e-06, + "loss": 1.0154, + "step": 25429 + }, + { + "epoch": 0.61, + "learning_rate": 6.923463440789799e-06, + "loss": 0.9178, + "step": 25430 + }, + { + "epoch": 0.61, + "learning_rate": 6.922722333342731e-06, + "loss": 1.0809, + "step": 25431 + }, + { + "epoch": 0.61, + "learning_rate": 6.9219812445647805e-06, + "loss": 0.6009, + "step": 25432 + }, + { + "epoch": 0.61, + "learning_rate": 6.921240174460448e-06, + "loss": 0.895, + "step": 25433 + }, + { + "epoch": 0.61, + "learning_rate": 6.920499123034228e-06, + "loss": 0.9533, + "step": 25434 + }, + { + "epoch": 0.61, + "learning_rate": 6.919758090290613e-06, + "loss": 1.2788, + "step": 25435 + }, + { + "epoch": 0.61, + "learning_rate": 6.919017076234103e-06, + "loss": 1.0439, + "step": 25436 + }, + { + "epoch": 0.61, + "learning_rate": 6.9182760808691904e-06, + "loss": 1.1018, + "step": 25437 + }, + { + "epoch": 0.61, + "learning_rate": 6.917535104200372e-06, + "loss": 0.9052, + "step": 25438 + }, + { + "epoch": 0.61, + "learning_rate": 6.916794146232146e-06, + "loss": 1.1717, + "step": 25439 + }, + { + "epoch": 0.61, + "learning_rate": 6.916053206969e-06, + "loss": 1.1264, + "step": 25440 + }, + { + "epoch": 0.61, + "learning_rate": 6.915312286415436e-06, + "loss": 1.1139, + "step": 25441 + }, + { + "epoch": 0.61, + "learning_rate": 6.914571384575949e-06, + "loss": 1.2887, + "step": 25442 + }, + { + "epoch": 0.61, + "learning_rate": 6.913830501455028e-06, + "loss": 1.0959, + "step": 25443 + }, + { + "epoch": 0.61, + "learning_rate": 6.913089637057173e-06, + "loss": 1.071, + "step": 25444 + }, + { + "epoch": 0.61, + "learning_rate": 6.912348791386878e-06, + "loss": 1.1368, + "step": 25445 + }, + { + "epoch": 0.61, + "learning_rate": 6.911607964448637e-06, + "loss": 1.0379, + "step": 25446 + }, + { + "epoch": 0.61, + "learning_rate": 6.910867156246941e-06, + "loss": 0.9995, + "step": 25447 + }, + { + "epoch": 0.61, + "learning_rate": 6.910126366786289e-06, + "loss": 1.0466, + "step": 25448 + }, + { + "epoch": 0.61, + "learning_rate": 6.909385596071175e-06, + "loss": 1.1793, + "step": 25449 + }, + { + "epoch": 0.61, + "learning_rate": 6.908644844106091e-06, + "loss": 1.1135, + "step": 25450 + }, + { + "epoch": 0.61, + "learning_rate": 6.907904110895529e-06, + "loss": 1.0022, + "step": 25451 + }, + { + "epoch": 0.61, + "learning_rate": 6.907163396443988e-06, + "loss": 1.0863, + "step": 25452 + }, + { + "epoch": 0.61, + "learning_rate": 6.906422700755959e-06, + "loss": 1.2631, + "step": 25453 + }, + { + "epoch": 0.61, + "learning_rate": 6.9056820238359355e-06, + "loss": 1.1269, + "step": 25454 + }, + { + "epoch": 0.61, + "learning_rate": 6.904941365688411e-06, + "loss": 1.2635, + "step": 25455 + }, + { + "epoch": 0.61, + "learning_rate": 6.90420072631788e-06, + "loss": 1.1958, + "step": 25456 + }, + { + "epoch": 0.61, + "learning_rate": 6.903460105728834e-06, + "loss": 1.1718, + "step": 25457 + }, + { + "epoch": 0.61, + "learning_rate": 6.902719503925769e-06, + "loss": 0.8824, + "step": 25458 + }, + { + "epoch": 0.61, + "learning_rate": 6.901978920913173e-06, + "loss": 0.9455, + "step": 25459 + }, + { + "epoch": 0.61, + "learning_rate": 6.901238356695545e-06, + "loss": 0.9765, + "step": 25460 + }, + { + "epoch": 0.61, + "learning_rate": 6.9004978112773745e-06, + "loss": 0.9456, + "step": 25461 + }, + { + "epoch": 0.61, + "learning_rate": 6.899757284663156e-06, + "loss": 1.0028, + "step": 25462 + }, + { + "epoch": 0.61, + "learning_rate": 6.899016776857377e-06, + "loss": 1.1635, + "step": 25463 + }, + { + "epoch": 0.61, + "learning_rate": 6.898276287864536e-06, + "loss": 0.7959, + "step": 25464 + }, + { + "epoch": 0.61, + "learning_rate": 6.897535817689125e-06, + "loss": 1.1945, + "step": 25465 + }, + { + "epoch": 0.61, + "learning_rate": 6.896795366335633e-06, + "loss": 1.2149, + "step": 25466 + }, + { + "epoch": 0.61, + "learning_rate": 6.896054933808551e-06, + "loss": 0.9429, + "step": 25467 + }, + { + "epoch": 0.61, + "learning_rate": 6.8953145201123776e-06, + "loss": 0.9355, + "step": 25468 + }, + { + "epoch": 0.61, + "learning_rate": 6.8945741252515985e-06, + "loss": 1.1604, + "step": 25469 + }, + { + "epoch": 0.61, + "learning_rate": 6.893833749230709e-06, + "loss": 1.1198, + "step": 25470 + }, + { + "epoch": 0.61, + "learning_rate": 6.893093392054197e-06, + "loss": 1.1565, + "step": 25471 + }, + { + "epoch": 0.61, + "learning_rate": 6.89235305372656e-06, + "loss": 1.1194, + "step": 25472 + }, + { + "epoch": 0.61, + "learning_rate": 6.8916127342522846e-06, + "loss": 1.1185, + "step": 25473 + }, + { + "epoch": 0.61, + "learning_rate": 6.890872433635863e-06, + "loss": 0.8877, + "step": 25474 + }, + { + "epoch": 0.61, + "learning_rate": 6.8901321518817864e-06, + "loss": 1.2101, + "step": 25475 + }, + { + "epoch": 0.61, + "learning_rate": 6.889391888994548e-06, + "loss": 1.2038, + "step": 25476 + }, + { + "epoch": 0.61, + "learning_rate": 6.8886516449786374e-06, + "loss": 1.1365, + "step": 25477 + }, + { + "epoch": 0.61, + "learning_rate": 6.887911419838546e-06, + "loss": 1.1076, + "step": 25478 + }, + { + "epoch": 0.61, + "learning_rate": 6.887171213578762e-06, + "loss": 1.1649, + "step": 25479 + }, + { + "epoch": 0.61, + "learning_rate": 6.8864310262037795e-06, + "loss": 1.0649, + "step": 25480 + }, + { + "epoch": 0.61, + "learning_rate": 6.8856908577180885e-06, + "loss": 1.1141, + "step": 25481 + }, + { + "epoch": 0.61, + "learning_rate": 6.884950708126177e-06, + "loss": 0.8017, + "step": 25482 + }, + { + "epoch": 0.61, + "learning_rate": 6.8842105774325355e-06, + "loss": 1.1395, + "step": 25483 + }, + { + "epoch": 0.61, + "learning_rate": 6.883470465641657e-06, + "loss": 1.2111, + "step": 25484 + }, + { + "epoch": 0.61, + "learning_rate": 6.882730372758032e-06, + "loss": 1.0576, + "step": 25485 + }, + { + "epoch": 0.61, + "learning_rate": 6.881990298786148e-06, + "loss": 1.0042, + "step": 25486 + }, + { + "epoch": 0.61, + "learning_rate": 6.881250243730492e-06, + "loss": 1.1108, + "step": 25487 + }, + { + "epoch": 0.61, + "learning_rate": 6.88051020759556e-06, + "loss": 1.2864, + "step": 25488 + }, + { + "epoch": 0.61, + "learning_rate": 6.879770190385839e-06, + "loss": 1.0377, + "step": 25489 + }, + { + "epoch": 0.61, + "learning_rate": 6.8790301921058155e-06, + "loss": 0.8343, + "step": 25490 + }, + { + "epoch": 0.61, + "learning_rate": 6.878290212759986e-06, + "loss": 1.1165, + "step": 25491 + }, + { + "epoch": 0.61, + "learning_rate": 6.877550252352834e-06, + "loss": 1.209, + "step": 25492 + }, + { + "epoch": 0.61, + "learning_rate": 6.8768103108888496e-06, + "loss": 0.9138, + "step": 25493 + }, + { + "epoch": 0.61, + "learning_rate": 6.87607038837252e-06, + "loss": 0.9898, + "step": 25494 + }, + { + "epoch": 0.61, + "learning_rate": 6.875330484808341e-06, + "loss": 1.0532, + "step": 25495 + }, + { + "epoch": 0.61, + "learning_rate": 6.874590600200794e-06, + "loss": 1.0894, + "step": 25496 + }, + { + "epoch": 0.61, + "learning_rate": 6.873850734554373e-06, + "loss": 1.1679, + "step": 25497 + }, + { + "epoch": 0.61, + "learning_rate": 6.873110887873562e-06, + "loss": 1.1852, + "step": 25498 + }, + { + "epoch": 0.61, + "learning_rate": 6.87237106016285e-06, + "loss": 1.067, + "step": 25499 + }, + { + "epoch": 0.61, + "learning_rate": 6.871631251426729e-06, + "loss": 0.8391, + "step": 25500 + }, + { + "epoch": 0.61, + "learning_rate": 6.870891461669686e-06, + "loss": 1.0719, + "step": 25501 + }, + { + "epoch": 0.61, + "learning_rate": 6.87015169089621e-06, + "loss": 1.0911, + "step": 25502 + }, + { + "epoch": 0.61, + "learning_rate": 6.869411939110786e-06, + "loss": 1.1293, + "step": 25503 + }, + { + "epoch": 0.61, + "learning_rate": 6.868672206317903e-06, + "loss": 0.8185, + "step": 25504 + }, + { + "epoch": 0.61, + "learning_rate": 6.867932492522049e-06, + "loss": 1.1232, + "step": 25505 + }, + { + "epoch": 0.61, + "learning_rate": 6.867192797727714e-06, + "loss": 1.1086, + "step": 25506 + }, + { + "epoch": 0.61, + "learning_rate": 6.866453121939382e-06, + "loss": 0.9759, + "step": 25507 + }, + { + "epoch": 0.61, + "learning_rate": 6.865713465161541e-06, + "loss": 1.1015, + "step": 25508 + }, + { + "epoch": 0.61, + "learning_rate": 6.86497382739868e-06, + "loss": 1.1216, + "step": 25509 + }, + { + "epoch": 0.61, + "learning_rate": 6.864234208655287e-06, + "loss": 1.0932, + "step": 25510 + }, + { + "epoch": 0.61, + "learning_rate": 6.8634946089358456e-06, + "loss": 1.2615, + "step": 25511 + }, + { + "epoch": 0.61, + "learning_rate": 6.862755028244844e-06, + "loss": 1.1635, + "step": 25512 + }, + { + "epoch": 0.61, + "learning_rate": 6.862015466586771e-06, + "loss": 0.9757, + "step": 25513 + }, + { + "epoch": 0.61, + "learning_rate": 6.861275923966113e-06, + "loss": 0.9973, + "step": 25514 + }, + { + "epoch": 0.61, + "learning_rate": 6.860536400387355e-06, + "loss": 0.9142, + "step": 25515 + }, + { + "epoch": 0.61, + "learning_rate": 6.859796895854982e-06, + "loss": 0.9542, + "step": 25516 + }, + { + "epoch": 0.61, + "learning_rate": 6.859057410373484e-06, + "loss": 1.082, + "step": 25517 + }, + { + "epoch": 0.61, + "learning_rate": 6.858317943947348e-06, + "loss": 1.0833, + "step": 25518 + }, + { + "epoch": 0.61, + "learning_rate": 6.857578496581056e-06, + "loss": 1.1926, + "step": 25519 + }, + { + "epoch": 0.61, + "learning_rate": 6.856839068279096e-06, + "loss": 1.0132, + "step": 25520 + }, + { + "epoch": 0.61, + "learning_rate": 6.856099659045954e-06, + "loss": 1.1481, + "step": 25521 + }, + { + "epoch": 0.61, + "learning_rate": 6.8553602688861175e-06, + "loss": 1.1867, + "step": 25522 + }, + { + "epoch": 0.61, + "learning_rate": 6.8546208978040675e-06, + "loss": 0.8242, + "step": 25523 + }, + { + "epoch": 0.61, + "learning_rate": 6.853881545804295e-06, + "loss": 1.1843, + "step": 25524 + }, + { + "epoch": 0.61, + "learning_rate": 6.8531422128912835e-06, + "loss": 1.3408, + "step": 25525 + }, + { + "epoch": 0.61, + "learning_rate": 6.852402899069517e-06, + "loss": 0.9485, + "step": 25526 + }, + { + "epoch": 0.61, + "learning_rate": 6.851663604343482e-06, + "loss": 0.9733, + "step": 25527 + }, + { + "epoch": 0.61, + "learning_rate": 6.850924328717663e-06, + "loss": 0.9866, + "step": 25528 + }, + { + "epoch": 0.61, + "learning_rate": 6.8501850721965465e-06, + "loss": 1.0836, + "step": 25529 + }, + { + "epoch": 0.61, + "learning_rate": 6.849445834784616e-06, + "loss": 0.8949, + "step": 25530 + }, + { + "epoch": 0.61, + "learning_rate": 6.848706616486353e-06, + "loss": 1.034, + "step": 25531 + }, + { + "epoch": 0.61, + "learning_rate": 6.847967417306251e-06, + "loss": 1.1902, + "step": 25532 + }, + { + "epoch": 0.61, + "learning_rate": 6.847228237248787e-06, + "loss": 1.1222, + "step": 25533 + }, + { + "epoch": 0.61, + "learning_rate": 6.846489076318449e-06, + "loss": 0.9761, + "step": 25534 + }, + { + "epoch": 0.61, + "learning_rate": 6.845749934519716e-06, + "loss": 0.9524, + "step": 25535 + }, + { + "epoch": 0.61, + "learning_rate": 6.8450108118570805e-06, + "loss": 0.9876, + "step": 25536 + }, + { + "epoch": 0.61, + "learning_rate": 6.844271708335022e-06, + "loss": 0.9887, + "step": 25537 + }, + { + "epoch": 0.61, + "learning_rate": 6.843532623958023e-06, + "loss": 1.3223, + "step": 25538 + }, + { + "epoch": 0.61, + "learning_rate": 6.84279355873057e-06, + "loss": 0.7993, + "step": 25539 + }, + { + "epoch": 0.61, + "learning_rate": 6.842054512657146e-06, + "loss": 1.0597, + "step": 25540 + }, + { + "epoch": 0.61, + "learning_rate": 6.841315485742234e-06, + "loss": 0.938, + "step": 25541 + }, + { + "epoch": 0.61, + "learning_rate": 6.840576477990318e-06, + "loss": 0.9512, + "step": 25542 + }, + { + "epoch": 0.61, + "learning_rate": 6.8398374894058805e-06, + "loss": 1.2135, + "step": 25543 + }, + { + "epoch": 0.61, + "learning_rate": 6.839098519993407e-06, + "loss": 1.0402, + "step": 25544 + }, + { + "epoch": 0.61, + "learning_rate": 6.83835956975738e-06, + "loss": 1.1053, + "step": 25545 + }, + { + "epoch": 0.61, + "learning_rate": 6.83762063870228e-06, + "loss": 1.0337, + "step": 25546 + }, + { + "epoch": 0.61, + "learning_rate": 6.836881726832592e-06, + "loss": 0.9804, + "step": 25547 + }, + { + "epoch": 0.61, + "learning_rate": 6.8361428341527995e-06, + "loss": 1.003, + "step": 25548 + }, + { + "epoch": 0.61, + "learning_rate": 6.835403960667384e-06, + "loss": 1.1411, + "step": 25549 + }, + { + "epoch": 0.61, + "learning_rate": 6.834665106380829e-06, + "loss": 1.0201, + "step": 25550 + }, + { + "epoch": 0.61, + "learning_rate": 6.833926271297613e-06, + "loss": 1.2678, + "step": 25551 + }, + { + "epoch": 0.61, + "learning_rate": 6.833187455422226e-06, + "loss": 0.9178, + "step": 25552 + }, + { + "epoch": 0.61, + "learning_rate": 6.832448658759143e-06, + "loss": 1.1027, + "step": 25553 + }, + { + "epoch": 0.61, + "learning_rate": 6.83170988131285e-06, + "loss": 1.388, + "step": 25554 + }, + { + "epoch": 0.61, + "learning_rate": 6.830971123087826e-06, + "loss": 1.1992, + "step": 25555 + }, + { + "epoch": 0.61, + "learning_rate": 6.830232384088557e-06, + "loss": 1.1911, + "step": 25556 + }, + { + "epoch": 0.61, + "learning_rate": 6.829493664319523e-06, + "loss": 0.9915, + "step": 25557 + }, + { + "epoch": 0.61, + "learning_rate": 6.8287549637852046e-06, + "loss": 1.3353, + "step": 25558 + }, + { + "epoch": 0.61, + "learning_rate": 6.828016282490082e-06, + "loss": 1.1643, + "step": 25559 + }, + { + "epoch": 0.61, + "learning_rate": 6.827277620438641e-06, + "loss": 1.0415, + "step": 25560 + }, + { + "epoch": 0.61, + "learning_rate": 6.8265389776353594e-06, + "loss": 0.9064, + "step": 25561 + }, + { + "epoch": 0.61, + "learning_rate": 6.8258003540847196e-06, + "loss": 0.9842, + "step": 25562 + }, + { + "epoch": 0.61, + "learning_rate": 6.825061749791201e-06, + "loss": 1.2461, + "step": 25563 + }, + { + "epoch": 0.61, + "learning_rate": 6.824323164759288e-06, + "loss": 1.0968, + "step": 25564 + }, + { + "epoch": 0.61, + "learning_rate": 6.823584598993458e-06, + "loss": 0.9864, + "step": 25565 + }, + { + "epoch": 0.61, + "learning_rate": 6.822846052498195e-06, + "loss": 0.9682, + "step": 25566 + }, + { + "epoch": 0.61, + "learning_rate": 6.8221075252779746e-06, + "loss": 1.2088, + "step": 25567 + }, + { + "epoch": 0.61, + "learning_rate": 6.821369017337283e-06, + "loss": 1.055, + "step": 25568 + }, + { + "epoch": 0.61, + "learning_rate": 6.820630528680597e-06, + "loss": 0.8224, + "step": 25569 + }, + { + "epoch": 0.61, + "learning_rate": 6.8198920593123966e-06, + "loss": 1.0531, + "step": 25570 + }, + { + "epoch": 0.61, + "learning_rate": 6.819153609237165e-06, + "loss": 0.8843, + "step": 25571 + }, + { + "epoch": 0.61, + "learning_rate": 6.81841517845938e-06, + "loss": 0.8444, + "step": 25572 + }, + { + "epoch": 0.62, + "learning_rate": 6.817676766983521e-06, + "loss": 0.9568, + "step": 25573 + }, + { + "epoch": 0.62, + "learning_rate": 6.8169383748140685e-06, + "loss": 1.0083, + "step": 25574 + }, + { + "epoch": 0.62, + "learning_rate": 6.816200001955502e-06, + "loss": 1.0216, + "step": 25575 + }, + { + "epoch": 0.62, + "learning_rate": 6.815461648412303e-06, + "loss": 0.8729, + "step": 25576 + }, + { + "epoch": 0.62, + "learning_rate": 6.814723314188948e-06, + "loss": 0.8262, + "step": 25577 + }, + { + "epoch": 0.62, + "learning_rate": 6.8139849992899155e-06, + "loss": 0.9139, + "step": 25578 + }, + { + "epoch": 0.62, + "learning_rate": 6.813246703719689e-06, + "loss": 1.0401, + "step": 25579 + }, + { + "epoch": 0.62, + "learning_rate": 6.812508427482745e-06, + "loss": 0.7937, + "step": 25580 + }, + { + "epoch": 0.62, + "learning_rate": 6.811770170583562e-06, + "loss": 0.6171, + "step": 25581 + }, + { + "epoch": 0.62, + "learning_rate": 6.811031933026619e-06, + "loss": 1.1258, + "step": 25582 + }, + { + "epoch": 0.62, + "learning_rate": 6.810293714816396e-06, + "loss": 1.173, + "step": 25583 + }, + { + "epoch": 0.62, + "learning_rate": 6.809555515957371e-06, + "loss": 1.1246, + "step": 25584 + }, + { + "epoch": 0.62, + "learning_rate": 6.808817336454023e-06, + "loss": 1.0443, + "step": 25585 + }, + { + "epoch": 0.62, + "learning_rate": 6.808079176310826e-06, + "loss": 1.1937, + "step": 25586 + }, + { + "epoch": 0.62, + "learning_rate": 6.807341035532265e-06, + "loss": 1.1927, + "step": 25587 + }, + { + "epoch": 0.62, + "learning_rate": 6.806602914122815e-06, + "loss": 1.029, + "step": 25588 + }, + { + "epoch": 0.62, + "learning_rate": 6.805864812086954e-06, + "loss": 1.2751, + "step": 25589 + }, + { + "epoch": 0.62, + "learning_rate": 6.805126729429158e-06, + "loss": 0.7991, + "step": 25590 + }, + { + "epoch": 0.62, + "learning_rate": 6.804388666153908e-06, + "loss": 1.1629, + "step": 25591 + }, + { + "epoch": 0.62, + "learning_rate": 6.803650622265681e-06, + "loss": 1.086, + "step": 25592 + }, + { + "epoch": 0.62, + "learning_rate": 6.802912597768954e-06, + "loss": 0.9471, + "step": 25593 + }, + { + "epoch": 0.62, + "learning_rate": 6.802174592668201e-06, + "loss": 1.1637, + "step": 25594 + }, + { + "epoch": 0.62, + "learning_rate": 6.801436606967905e-06, + "loss": 1.0145, + "step": 25595 + }, + { + "epoch": 0.62, + "learning_rate": 6.800698640672541e-06, + "loss": 0.9262, + "step": 25596 + }, + { + "epoch": 0.62, + "learning_rate": 6.799960693786586e-06, + "loss": 1.2165, + "step": 25597 + }, + { + "epoch": 0.62, + "learning_rate": 6.799222766314515e-06, + "loss": 1.416, + "step": 25598 + }, + { + "epoch": 0.62, + "learning_rate": 6.7984848582608075e-06, + "loss": 0.8917, + "step": 25599 + }, + { + "epoch": 0.62, + "learning_rate": 6.797746969629941e-06, + "loss": 0.9766, + "step": 25600 + }, + { + "epoch": 0.62, + "learning_rate": 6.797009100426389e-06, + "loss": 1.1844, + "step": 25601 + }, + { + "epoch": 0.62, + "learning_rate": 6.796271250654628e-06, + "loss": 1.1594, + "step": 25602 + }, + { + "epoch": 0.62, + "learning_rate": 6.795533420319138e-06, + "loss": 0.8676, + "step": 25603 + }, + { + "epoch": 0.62, + "learning_rate": 6.794795609424392e-06, + "loss": 0.9753, + "step": 25604 + }, + { + "epoch": 0.62, + "learning_rate": 6.7940578179748685e-06, + "loss": 1.1913, + "step": 25605 + }, + { + "epoch": 0.62, + "learning_rate": 6.793320045975038e-06, + "loss": 1.4334, + "step": 25606 + }, + { + "epoch": 0.62, + "learning_rate": 6.792582293429385e-06, + "loss": 1.0306, + "step": 25607 + }, + { + "epoch": 0.62, + "learning_rate": 6.79184456034238e-06, + "loss": 1.2711, + "step": 25608 + }, + { + "epoch": 0.62, + "learning_rate": 6.791106846718498e-06, + "loss": 0.9385, + "step": 25609 + }, + { + "epoch": 0.62, + "learning_rate": 6.7903691525622166e-06, + "loss": 1.3116, + "step": 25610 + }, + { + "epoch": 0.62, + "learning_rate": 6.7896314778780105e-06, + "loss": 1.1504, + "step": 25611 + }, + { + "epoch": 0.62, + "learning_rate": 6.7888938226703564e-06, + "loss": 0.9154, + "step": 25612 + }, + { + "epoch": 0.62, + "learning_rate": 6.788156186943727e-06, + "loss": 1.1811, + "step": 25613 + }, + { + "epoch": 0.62, + "learning_rate": 6.7874185707025976e-06, + "loss": 1.1528, + "step": 25614 + }, + { + "epoch": 0.62, + "learning_rate": 6.786680973951445e-06, + "loss": 0.8572, + "step": 25615 + }, + { + "epoch": 0.62, + "learning_rate": 6.785943396694743e-06, + "loss": 1.1065, + "step": 25616 + }, + { + "epoch": 0.62, + "learning_rate": 6.785205838936966e-06, + "loss": 0.9666, + "step": 25617 + }, + { + "epoch": 0.62, + "learning_rate": 6.78446830068259e-06, + "loss": 1.0804, + "step": 25618 + }, + { + "epoch": 0.62, + "learning_rate": 6.783730781936089e-06, + "loss": 1.0874, + "step": 25619 + }, + { + "epoch": 0.62, + "learning_rate": 6.782993282701936e-06, + "loss": 1.1512, + "step": 25620 + }, + { + "epoch": 0.62, + "learning_rate": 6.7822558029846045e-06, + "loss": 1.0388, + "step": 25621 + }, + { + "epoch": 0.62, + "learning_rate": 6.7815183427885725e-06, + "loss": 1.2825, + "step": 25622 + }, + { + "epoch": 0.62, + "learning_rate": 6.780780902118311e-06, + "loss": 1.1042, + "step": 25623 + }, + { + "epoch": 0.62, + "learning_rate": 6.780043480978296e-06, + "loss": 0.9909, + "step": 25624 + }, + { + "epoch": 0.62, + "learning_rate": 6.779306079372996e-06, + "loss": 1.0503, + "step": 25625 + }, + { + "epoch": 0.62, + "learning_rate": 6.778568697306891e-06, + "loss": 0.9942, + "step": 25626 + }, + { + "epoch": 0.62, + "learning_rate": 6.7778313347844526e-06, + "loss": 0.961, + "step": 25627 + }, + { + "epoch": 0.62, + "learning_rate": 6.777093991810153e-06, + "loss": 0.8878, + "step": 25628 + }, + { + "epoch": 0.62, + "learning_rate": 6.776356668388464e-06, + "loss": 1.0141, + "step": 25629 + }, + { + "epoch": 0.62, + "learning_rate": 6.775619364523863e-06, + "loss": 1.2112, + "step": 25630 + }, + { + "epoch": 0.62, + "learning_rate": 6.774882080220822e-06, + "loss": 1.0073, + "step": 25631 + }, + { + "epoch": 0.62, + "learning_rate": 6.7741448154838105e-06, + "loss": 1.228, + "step": 25632 + }, + { + "epoch": 0.62, + "learning_rate": 6.773407570317303e-06, + "loss": 1.152, + "step": 25633 + }, + { + "epoch": 0.62, + "learning_rate": 6.7726703447257755e-06, + "loss": 1.0973, + "step": 25634 + }, + { + "epoch": 0.62, + "learning_rate": 6.7719331387136965e-06, + "loss": 1.217, + "step": 25635 + }, + { + "epoch": 0.62, + "learning_rate": 6.771195952285541e-06, + "loss": 1.0604, + "step": 25636 + }, + { + "epoch": 0.62, + "learning_rate": 6.770458785445777e-06, + "loss": 1.0561, + "step": 25637 + }, + { + "epoch": 0.62, + "learning_rate": 6.769721638198883e-06, + "loss": 0.9708, + "step": 25638 + }, + { + "epoch": 0.62, + "learning_rate": 6.768984510549328e-06, + "loss": 1.1187, + "step": 25639 + }, + { + "epoch": 0.62, + "learning_rate": 6.768247402501585e-06, + "loss": 0.9409, + "step": 25640 + }, + { + "epoch": 0.62, + "learning_rate": 6.76751031406012e-06, + "loss": 1.0472, + "step": 25641 + }, + { + "epoch": 0.62, + "learning_rate": 6.766773245229413e-06, + "loss": 1.0864, + "step": 25642 + }, + { + "epoch": 0.62, + "learning_rate": 6.766036196013932e-06, + "loss": 1.0516, + "step": 25643 + }, + { + "epoch": 0.62, + "learning_rate": 6.76529916641815e-06, + "loss": 1.016, + "step": 25644 + }, + { + "epoch": 0.62, + "learning_rate": 6.7645621564465345e-06, + "loss": 1.0948, + "step": 25645 + }, + { + "epoch": 0.62, + "learning_rate": 6.76382516610356e-06, + "loss": 1.2002, + "step": 25646 + }, + { + "epoch": 0.62, + "learning_rate": 6.7630881953936985e-06, + "loss": 1.0984, + "step": 25647 + }, + { + "epoch": 0.62, + "learning_rate": 6.762351244321419e-06, + "loss": 1.2315, + "step": 25648 + }, + { + "epoch": 0.62, + "learning_rate": 6.761614312891191e-06, + "loss": 1.0707, + "step": 25649 + }, + { + "epoch": 0.62, + "learning_rate": 6.76087740110749e-06, + "loss": 1.1969, + "step": 25650 + }, + { + "epoch": 0.62, + "learning_rate": 6.760140508974783e-06, + "loss": 1.2465, + "step": 25651 + }, + { + "epoch": 0.62, + "learning_rate": 6.759403636497541e-06, + "loss": 1.0785, + "step": 25652 + }, + { + "epoch": 0.62, + "learning_rate": 6.758666783680234e-06, + "loss": 0.9272, + "step": 25653 + }, + { + "epoch": 0.62, + "learning_rate": 6.757929950527335e-06, + "loss": 1.0063, + "step": 25654 + }, + { + "epoch": 0.62, + "learning_rate": 6.757193137043312e-06, + "loss": 1.0637, + "step": 25655 + }, + { + "epoch": 0.62, + "learning_rate": 6.756456343232636e-06, + "loss": 1.2338, + "step": 25656 + }, + { + "epoch": 0.62, + "learning_rate": 6.755719569099774e-06, + "loss": 0.9114, + "step": 25657 + }, + { + "epoch": 0.62, + "learning_rate": 6.7549828146492e-06, + "loss": 1.0058, + "step": 25658 + }, + { + "epoch": 0.62, + "learning_rate": 6.754246079885383e-06, + "loss": 1.1273, + "step": 25659 + }, + { + "epoch": 0.62, + "learning_rate": 6.753509364812789e-06, + "loss": 1.1506, + "step": 25660 + }, + { + "epoch": 0.62, + "learning_rate": 6.752772669435892e-06, + "loss": 1.0885, + "step": 25661 + }, + { + "epoch": 0.62, + "learning_rate": 6.7520359937591595e-06, + "loss": 1.1004, + "step": 25662 + }, + { + "epoch": 0.62, + "learning_rate": 6.751299337787061e-06, + "loss": 1.0161, + "step": 25663 + }, + { + "epoch": 0.62, + "learning_rate": 6.750562701524063e-06, + "loss": 1.1947, + "step": 25664 + }, + { + "epoch": 0.62, + "learning_rate": 6.749826084974638e-06, + "loss": 1.0645, + "step": 25665 + }, + { + "epoch": 0.62, + "learning_rate": 6.74908948814325e-06, + "loss": 1.0896, + "step": 25666 + }, + { + "epoch": 0.62, + "learning_rate": 6.748352911034377e-06, + "loss": 0.9132, + "step": 25667 + }, + { + "epoch": 0.62, + "learning_rate": 6.7476163536524796e-06, + "loss": 1.0548, + "step": 25668 + }, + { + "epoch": 0.62, + "learning_rate": 6.746879816002031e-06, + "loss": 1.0556, + "step": 25669 + }, + { + "epoch": 0.62, + "learning_rate": 6.746143298087494e-06, + "loss": 1.1496, + "step": 25670 + }, + { + "epoch": 0.62, + "learning_rate": 6.745406799913343e-06, + "loss": 1.0104, + "step": 25671 + }, + { + "epoch": 0.62, + "learning_rate": 6.744670321484044e-06, + "loss": 1.0618, + "step": 25672 + }, + { + "epoch": 0.62, + "learning_rate": 6.743933862804061e-06, + "loss": 0.896, + "step": 25673 + }, + { + "epoch": 0.62, + "learning_rate": 6.743197423877867e-06, + "loss": 0.9149, + "step": 25674 + }, + { + "epoch": 0.62, + "learning_rate": 6.7424610047099295e-06, + "loss": 0.7456, + "step": 25675 + }, + { + "epoch": 0.62, + "learning_rate": 6.741724605304714e-06, + "loss": 0.9389, + "step": 25676 + }, + { + "epoch": 0.62, + "learning_rate": 6.740988225666689e-06, + "loss": 1.261, + "step": 25677 + }, + { + "epoch": 0.62, + "learning_rate": 6.740251865800321e-06, + "loss": 1.2612, + "step": 25678 + }, + { + "epoch": 0.62, + "learning_rate": 6.7395155257100795e-06, + "loss": 1.1498, + "step": 25679 + }, + { + "epoch": 0.62, + "learning_rate": 6.738779205400429e-06, + "loss": 1.2102, + "step": 25680 + }, + { + "epoch": 0.62, + "learning_rate": 6.738042904875841e-06, + "loss": 1.1246, + "step": 25681 + }, + { + "epoch": 0.62, + "learning_rate": 6.737306624140776e-06, + "loss": 1.1328, + "step": 25682 + }, + { + "epoch": 0.62, + "learning_rate": 6.736570363199706e-06, + "loss": 1.2331, + "step": 25683 + }, + { + "epoch": 0.62, + "learning_rate": 6.735834122057096e-06, + "loss": 0.7388, + "step": 25684 + }, + { + "epoch": 0.62, + "learning_rate": 6.735097900717412e-06, + "loss": 0.7604, + "step": 25685 + }, + { + "epoch": 0.62, + "learning_rate": 6.734361699185119e-06, + "loss": 1.012, + "step": 25686 + }, + { + "epoch": 0.62, + "learning_rate": 6.733625517464688e-06, + "loss": 0.9196, + "step": 25687 + }, + { + "epoch": 0.62, + "learning_rate": 6.732889355560583e-06, + "loss": 1.0767, + "step": 25688 + }, + { + "epoch": 0.62, + "learning_rate": 6.73215321347727e-06, + "loss": 1.0588, + "step": 25689 + }, + { + "epoch": 0.62, + "learning_rate": 6.731417091219212e-06, + "loss": 0.9499, + "step": 25690 + }, + { + "epoch": 0.62, + "learning_rate": 6.7306809887908804e-06, + "loss": 1.0786, + "step": 25691 + }, + { + "epoch": 0.62, + "learning_rate": 6.729944906196739e-06, + "loss": 1.0918, + "step": 25692 + }, + { + "epoch": 0.62, + "learning_rate": 6.72920884344125e-06, + "loss": 1.1339, + "step": 25693 + }, + { + "epoch": 0.62, + "learning_rate": 6.728472800528883e-06, + "loss": 1.1955, + "step": 25694 + }, + { + "epoch": 0.62, + "learning_rate": 6.727736777464102e-06, + "loss": 0.9161, + "step": 25695 + }, + { + "epoch": 0.62, + "learning_rate": 6.727000774251373e-06, + "loss": 1.2536, + "step": 25696 + }, + { + "epoch": 0.62, + "learning_rate": 6.726264790895157e-06, + "loss": 1.1328, + "step": 25697 + }, + { + "epoch": 0.62, + "learning_rate": 6.725528827399926e-06, + "loss": 1.2273, + "step": 25698 + }, + { + "epoch": 0.62, + "learning_rate": 6.724792883770141e-06, + "loss": 1.1149, + "step": 25699 + }, + { + "epoch": 0.62, + "learning_rate": 6.724056960010268e-06, + "loss": 0.8767, + "step": 25700 + }, + { + "epoch": 0.62, + "learning_rate": 6.723321056124768e-06, + "loss": 1.124, + "step": 25701 + }, + { + "epoch": 0.62, + "learning_rate": 6.72258517211811e-06, + "loss": 1.1309, + "step": 25702 + }, + { + "epoch": 0.62, + "learning_rate": 6.721849307994756e-06, + "loss": 1.285, + "step": 25703 + }, + { + "epoch": 0.62, + "learning_rate": 6.721113463759174e-06, + "loss": 1.0186, + "step": 25704 + }, + { + "epoch": 0.62, + "learning_rate": 6.72037763941582e-06, + "loss": 1.2114, + "step": 25705 + }, + { + "epoch": 0.62, + "learning_rate": 6.719641834969168e-06, + "loss": 1.02, + "step": 25706 + }, + { + "epoch": 0.62, + "learning_rate": 6.718906050423677e-06, + "loss": 1.1702, + "step": 25707 + }, + { + "epoch": 0.62, + "learning_rate": 6.718170285783811e-06, + "loss": 0.8828, + "step": 25708 + }, + { + "epoch": 0.62, + "learning_rate": 6.717434541054032e-06, + "loss": 1.0375, + "step": 25709 + }, + { + "epoch": 0.62, + "learning_rate": 6.716698816238807e-06, + "loss": 1.1293, + "step": 25710 + }, + { + "epoch": 0.62, + "learning_rate": 6.715963111342598e-06, + "loss": 1.1378, + "step": 25711 + }, + { + "epoch": 0.62, + "learning_rate": 6.71522742636987e-06, + "loss": 0.8635, + "step": 25712 + }, + { + "epoch": 0.62, + "learning_rate": 6.714491761325082e-06, + "loss": 1.0984, + "step": 25713 + }, + { + "epoch": 0.62, + "learning_rate": 6.7137561162127e-06, + "loss": 1.239, + "step": 25714 + }, + { + "epoch": 0.62, + "learning_rate": 6.713020491037188e-06, + "loss": 1.1121, + "step": 25715 + }, + { + "epoch": 0.62, + "learning_rate": 6.712284885803007e-06, + "loss": 1.0899, + "step": 25716 + }, + { + "epoch": 0.62, + "learning_rate": 6.711549300514618e-06, + "loss": 0.9963, + "step": 25717 + }, + { + "epoch": 0.62, + "learning_rate": 6.710813735176489e-06, + "loss": 0.9514, + "step": 25718 + }, + { + "epoch": 0.62, + "learning_rate": 6.7100781897930785e-06, + "loss": 1.0688, + "step": 25719 + }, + { + "epoch": 0.62, + "learning_rate": 6.70934266436885e-06, + "loss": 1.0637, + "step": 25720 + }, + { + "epoch": 0.62, + "learning_rate": 6.708607158908264e-06, + "loss": 0.7649, + "step": 25721 + }, + { + "epoch": 0.62, + "learning_rate": 6.707871673415785e-06, + "loss": 0.9507, + "step": 25722 + }, + { + "epoch": 0.62, + "learning_rate": 6.707136207895876e-06, + "loss": 1.2772, + "step": 25723 + }, + { + "epoch": 0.62, + "learning_rate": 6.706400762352996e-06, + "loss": 1.046, + "step": 25724 + }, + { + "epoch": 0.62, + "learning_rate": 6.705665336791606e-06, + "loss": 1.1099, + "step": 25725 + }, + { + "epoch": 0.62, + "learning_rate": 6.704929931216171e-06, + "loss": 1.2143, + "step": 25726 + }, + { + "epoch": 0.62, + "learning_rate": 6.7041945456311515e-06, + "loss": 0.9368, + "step": 25727 + }, + { + "epoch": 0.62, + "learning_rate": 6.70345918004101e-06, + "loss": 1.1762, + "step": 25728 + }, + { + "epoch": 0.62, + "learning_rate": 6.702723834450202e-06, + "loss": 1.0761, + "step": 25729 + }, + { + "epoch": 0.62, + "learning_rate": 6.701988508863195e-06, + "loss": 1.2885, + "step": 25730 + }, + { + "epoch": 0.62, + "learning_rate": 6.701253203284449e-06, + "loss": 1.0917, + "step": 25731 + }, + { + "epoch": 0.62, + "learning_rate": 6.700517917718423e-06, + "loss": 1.3113, + "step": 25732 + }, + { + "epoch": 0.62, + "learning_rate": 6.699782652169578e-06, + "loss": 1.1976, + "step": 25733 + }, + { + "epoch": 0.62, + "learning_rate": 6.699047406642377e-06, + "loss": 1.0359, + "step": 25734 + }, + { + "epoch": 0.62, + "learning_rate": 6.698312181141279e-06, + "loss": 1.0761, + "step": 25735 + }, + { + "epoch": 0.62, + "learning_rate": 6.6975769756707434e-06, + "loss": 1.2892, + "step": 25736 + }, + { + "epoch": 0.62, + "learning_rate": 6.696841790235231e-06, + "loss": 1.2799, + "step": 25737 + }, + { + "epoch": 0.62, + "learning_rate": 6.696106624839203e-06, + "loss": 1.2118, + "step": 25738 + }, + { + "epoch": 0.62, + "learning_rate": 6.695371479487121e-06, + "loss": 1.3114, + "step": 25739 + }, + { + "epoch": 0.62, + "learning_rate": 6.694636354183439e-06, + "loss": 1.2069, + "step": 25740 + }, + { + "epoch": 0.62, + "learning_rate": 6.693901248932623e-06, + "loss": 0.9071, + "step": 25741 + }, + { + "epoch": 0.62, + "learning_rate": 6.693166163739131e-06, + "loss": 1.0585, + "step": 25742 + }, + { + "epoch": 0.62, + "learning_rate": 6.692431098607421e-06, + "loss": 0.9232, + "step": 25743 + }, + { + "epoch": 0.62, + "learning_rate": 6.691696053541954e-06, + "loss": 1.112, + "step": 25744 + }, + { + "epoch": 0.62, + "learning_rate": 6.69096102854719e-06, + "loss": 1.0964, + "step": 25745 + }, + { + "epoch": 0.62, + "learning_rate": 6.690226023627585e-06, + "loss": 1.0428, + "step": 25746 + }, + { + "epoch": 0.62, + "learning_rate": 6.689491038787601e-06, + "loss": 1.0895, + "step": 25747 + }, + { + "epoch": 0.62, + "learning_rate": 6.688756074031695e-06, + "loss": 0.9965, + "step": 25748 + }, + { + "epoch": 0.62, + "learning_rate": 6.688021129364329e-06, + "loss": 1.2123, + "step": 25749 + }, + { + "epoch": 0.62, + "learning_rate": 6.68728620478996e-06, + "loss": 1.1897, + "step": 25750 + }, + { + "epoch": 0.62, + "learning_rate": 6.686551300313047e-06, + "loss": 1.0265, + "step": 25751 + }, + { + "epoch": 0.62, + "learning_rate": 6.6858164159380445e-06, + "loss": 0.8066, + "step": 25752 + }, + { + "epoch": 0.62, + "learning_rate": 6.685081551669416e-06, + "loss": 1.1772, + "step": 25753 + }, + { + "epoch": 0.62, + "learning_rate": 6.6843467075116196e-06, + "loss": 1.3595, + "step": 25754 + }, + { + "epoch": 0.62, + "learning_rate": 6.6836118834691096e-06, + "loss": 0.8754, + "step": 25755 + }, + { + "epoch": 0.62, + "learning_rate": 6.682877079546346e-06, + "loss": 0.9987, + "step": 25756 + }, + { + "epoch": 0.62, + "learning_rate": 6.682142295747789e-06, + "loss": 0.9646, + "step": 25757 + }, + { + "epoch": 0.62, + "learning_rate": 6.681407532077895e-06, + "loss": 1.1159, + "step": 25758 + }, + { + "epoch": 0.62, + "learning_rate": 6.680672788541118e-06, + "loss": 1.3013, + "step": 25759 + }, + { + "epoch": 0.62, + "learning_rate": 6.679938065141918e-06, + "loss": 0.9133, + "step": 25760 + }, + { + "epoch": 0.62, + "learning_rate": 6.6792033618847566e-06, + "loss": 0.9391, + "step": 25761 + }, + { + "epoch": 0.62, + "learning_rate": 6.6784686787740845e-06, + "loss": 1.1445, + "step": 25762 + }, + { + "epoch": 0.62, + "learning_rate": 6.677734015814364e-06, + "loss": 1.031, + "step": 25763 + }, + { + "epoch": 0.62, + "learning_rate": 6.676999373010047e-06, + "loss": 0.9401, + "step": 25764 + }, + { + "epoch": 0.62, + "learning_rate": 6.676264750365594e-06, + "loss": 1.2353, + "step": 25765 + }, + { + "epoch": 0.62, + "learning_rate": 6.675530147885462e-06, + "loss": 1.1216, + "step": 25766 + }, + { + "epoch": 0.62, + "learning_rate": 6.674795565574106e-06, + "loss": 1.2966, + "step": 25767 + }, + { + "epoch": 0.62, + "learning_rate": 6.674061003435982e-06, + "loss": 1.0394, + "step": 25768 + }, + { + "epoch": 0.62, + "learning_rate": 6.673326461475549e-06, + "loss": 0.9705, + "step": 25769 + }, + { + "epoch": 0.62, + "learning_rate": 6.672591939697261e-06, + "loss": 1.0767, + "step": 25770 + }, + { + "epoch": 0.62, + "learning_rate": 6.671857438105578e-06, + "loss": 0.9235, + "step": 25771 + }, + { + "epoch": 0.62, + "learning_rate": 6.6711229567049485e-06, + "loss": 1.23, + "step": 25772 + }, + { + "epoch": 0.62, + "learning_rate": 6.670388495499836e-06, + "loss": 1.036, + "step": 25773 + }, + { + "epoch": 0.62, + "learning_rate": 6.669654054494693e-06, + "loss": 1.185, + "step": 25774 + }, + { + "epoch": 0.62, + "learning_rate": 6.668919633693975e-06, + "loss": 0.9949, + "step": 25775 + }, + { + "epoch": 0.62, + "learning_rate": 6.668185233102137e-06, + "loss": 1.022, + "step": 25776 + }, + { + "epoch": 0.62, + "learning_rate": 6.667450852723637e-06, + "loss": 0.9961, + "step": 25777 + }, + { + "epoch": 0.62, + "learning_rate": 6.666716492562929e-06, + "loss": 0.9977, + "step": 25778 + }, + { + "epoch": 0.62, + "learning_rate": 6.665982152624468e-06, + "loss": 0.944, + "step": 25779 + }, + { + "epoch": 0.62, + "learning_rate": 6.665247832912707e-06, + "loss": 1.2803, + "step": 25780 + }, + { + "epoch": 0.62, + "learning_rate": 6.6645135334321056e-06, + "loss": 0.9654, + "step": 25781 + }, + { + "epoch": 0.62, + "learning_rate": 6.663779254187115e-06, + "loss": 0.8702, + "step": 25782 + }, + { + "epoch": 0.62, + "learning_rate": 6.663044995182192e-06, + "loss": 1.2462, + "step": 25783 + }, + { + "epoch": 0.62, + "learning_rate": 6.662310756421787e-06, + "loss": 0.8176, + "step": 25784 + }, + { + "epoch": 0.62, + "learning_rate": 6.66157653791036e-06, + "loss": 1.1147, + "step": 25785 + }, + { + "epoch": 0.62, + "learning_rate": 6.660842339652362e-06, + "loss": 1.2118, + "step": 25786 + }, + { + "epoch": 0.62, + "learning_rate": 6.660108161652248e-06, + "loss": 1.1979, + "step": 25787 + }, + { + "epoch": 0.62, + "learning_rate": 6.659374003914474e-06, + "loss": 0.8649, + "step": 25788 + }, + { + "epoch": 0.62, + "learning_rate": 6.65863986644349e-06, + "loss": 1.0699, + "step": 25789 + }, + { + "epoch": 0.62, + "learning_rate": 6.657905749243754e-06, + "loss": 1.2827, + "step": 25790 + }, + { + "epoch": 0.62, + "learning_rate": 6.6571716523197165e-06, + "loss": 1.1339, + "step": 25791 + }, + { + "epoch": 0.62, + "learning_rate": 6.6564375756758314e-06, + "loss": 0.9842, + "step": 25792 + }, + { + "epoch": 0.62, + "learning_rate": 6.655703519316555e-06, + "loss": 1.1616, + "step": 25793 + }, + { + "epoch": 0.62, + "learning_rate": 6.6549694832463385e-06, + "loss": 1.229, + "step": 25794 + }, + { + "epoch": 0.62, + "learning_rate": 6.654235467469634e-06, + "loss": 1.1554, + "step": 25795 + }, + { + "epoch": 0.62, + "learning_rate": 6.653501471990896e-06, + "loss": 1.0334, + "step": 25796 + }, + { + "epoch": 0.62, + "learning_rate": 6.65276749681458e-06, + "loss": 1.0538, + "step": 25797 + }, + { + "epoch": 0.62, + "learning_rate": 6.652033541945134e-06, + "loss": 1.2219, + "step": 25798 + }, + { + "epoch": 0.62, + "learning_rate": 6.651299607387012e-06, + "loss": 1.354, + "step": 25799 + }, + { + "epoch": 0.62, + "learning_rate": 6.650565693144669e-06, + "loss": 1.184, + "step": 25800 + }, + { + "epoch": 0.62, + "learning_rate": 6.649831799222557e-06, + "loss": 1.3054, + "step": 25801 + }, + { + "epoch": 0.62, + "learning_rate": 6.649097925625128e-06, + "loss": 0.9745, + "step": 25802 + }, + { + "epoch": 0.62, + "learning_rate": 6.64836407235683e-06, + "loss": 1.0384, + "step": 25803 + }, + { + "epoch": 0.62, + "learning_rate": 6.647630239422121e-06, + "loss": 0.9804, + "step": 25804 + }, + { + "epoch": 0.62, + "learning_rate": 6.6468964268254525e-06, + "loss": 0.9582, + "step": 25805 + }, + { + "epoch": 0.62, + "learning_rate": 6.6461626345712735e-06, + "loss": 0.9687, + "step": 25806 + }, + { + "epoch": 0.62, + "learning_rate": 6.645428862664034e-06, + "loss": 1.1216, + "step": 25807 + }, + { + "epoch": 0.62, + "learning_rate": 6.644695111108192e-06, + "loss": 0.8512, + "step": 25808 + }, + { + "epoch": 0.62, + "learning_rate": 6.643961379908195e-06, + "loss": 1.0824, + "step": 25809 + }, + { + "epoch": 0.62, + "learning_rate": 6.643227669068495e-06, + "loss": 1.1962, + "step": 25810 + }, + { + "epoch": 0.62, + "learning_rate": 6.642493978593541e-06, + "loss": 0.9163, + "step": 25811 + }, + { + "epoch": 0.62, + "learning_rate": 6.6417603084877894e-06, + "loss": 0.8801, + "step": 25812 + }, + { + "epoch": 0.62, + "learning_rate": 6.641026658755687e-06, + "loss": 1.1296, + "step": 25813 + }, + { + "epoch": 0.62, + "learning_rate": 6.640293029401686e-06, + "loss": 1.1586, + "step": 25814 + }, + { + "epoch": 0.62, + "learning_rate": 6.639559420430235e-06, + "loss": 1.0635, + "step": 25815 + }, + { + "epoch": 0.62, + "learning_rate": 6.638825831845788e-06, + "loss": 0.8729, + "step": 25816 + }, + { + "epoch": 0.62, + "learning_rate": 6.638092263652794e-06, + "loss": 0.8427, + "step": 25817 + }, + { + "epoch": 0.62, + "learning_rate": 6.637358715855704e-06, + "loss": 1.0753, + "step": 25818 + }, + { + "epoch": 0.62, + "learning_rate": 6.636625188458966e-06, + "loss": 1.0355, + "step": 25819 + }, + { + "epoch": 0.62, + "learning_rate": 6.635891681467034e-06, + "loss": 1.2042, + "step": 25820 + }, + { + "epoch": 0.62, + "learning_rate": 6.635158194884356e-06, + "loss": 1.205, + "step": 25821 + }, + { + "epoch": 0.62, + "learning_rate": 6.634424728715381e-06, + "loss": 1.3363, + "step": 25822 + }, + { + "epoch": 0.62, + "learning_rate": 6.633691282964558e-06, + "loss": 1.1363, + "step": 25823 + }, + { + "epoch": 0.62, + "learning_rate": 6.632957857636339e-06, + "loss": 0.9399, + "step": 25824 + }, + { + "epoch": 0.62, + "learning_rate": 6.632224452735175e-06, + "loss": 1.0907, + "step": 25825 + }, + { + "epoch": 0.62, + "learning_rate": 6.631491068265511e-06, + "loss": 0.9036, + "step": 25826 + }, + { + "epoch": 0.62, + "learning_rate": 6.630757704231797e-06, + "loss": 1.1999, + "step": 25827 + }, + { + "epoch": 0.62, + "learning_rate": 6.630024360638486e-06, + "loss": 1.0504, + "step": 25828 + }, + { + "epoch": 0.62, + "learning_rate": 6.6292910374900244e-06, + "loss": 1.3434, + "step": 25829 + }, + { + "epoch": 0.62, + "learning_rate": 6.628557734790862e-06, + "loss": 1.0947, + "step": 25830 + }, + { + "epoch": 0.62, + "learning_rate": 6.627824452545443e-06, + "loss": 0.7997, + "step": 25831 + }, + { + "epoch": 0.62, + "learning_rate": 6.627091190758218e-06, + "loss": 1.1356, + "step": 25832 + }, + { + "epoch": 0.62, + "learning_rate": 6.6263579494336415e-06, + "loss": 1.0037, + "step": 25833 + }, + { + "epoch": 0.62, + "learning_rate": 6.625624728576159e-06, + "loss": 1.2081, + "step": 25834 + }, + { + "epoch": 0.62, + "learning_rate": 6.624891528190217e-06, + "loss": 1.1789, + "step": 25835 + }, + { + "epoch": 0.62, + "learning_rate": 6.624158348280263e-06, + "loss": 0.8734, + "step": 25836 + }, + { + "epoch": 0.62, + "learning_rate": 6.6234251888507475e-06, + "loss": 1.0487, + "step": 25837 + }, + { + "epoch": 0.62, + "learning_rate": 6.6226920499061164e-06, + "loss": 1.0657, + "step": 25838 + }, + { + "epoch": 0.62, + "learning_rate": 6.621958931450819e-06, + "loss": 0.9052, + "step": 25839 + }, + { + "epoch": 0.62, + "learning_rate": 6.6212258334893e-06, + "loss": 1.3511, + "step": 25840 + }, + { + "epoch": 0.62, + "learning_rate": 6.6204927560260115e-06, + "loss": 1.1418, + "step": 25841 + }, + { + "epoch": 0.62, + "learning_rate": 6.619759699065398e-06, + "loss": 0.9936, + "step": 25842 + }, + { + "epoch": 0.62, + "learning_rate": 6.6190266626119085e-06, + "loss": 0.9536, + "step": 25843 + }, + { + "epoch": 0.62, + "learning_rate": 6.618293646669987e-06, + "loss": 0.9211, + "step": 25844 + }, + { + "epoch": 0.62, + "learning_rate": 6.617560651244082e-06, + "loss": 1.3547, + "step": 25845 + }, + { + "epoch": 0.62, + "learning_rate": 6.616827676338643e-06, + "loss": 1.0728, + "step": 25846 + }, + { + "epoch": 0.62, + "learning_rate": 6.616094721958115e-06, + "loss": 1.1004, + "step": 25847 + }, + { + "epoch": 0.62, + "learning_rate": 6.6153617881069405e-06, + "loss": 1.179, + "step": 25848 + }, + { + "epoch": 0.62, + "learning_rate": 6.614628874789574e-06, + "loss": 1.163, + "step": 25849 + }, + { + "epoch": 0.62, + "learning_rate": 6.613895982010456e-06, + "loss": 1.181, + "step": 25850 + }, + { + "epoch": 0.62, + "learning_rate": 6.613163109774037e-06, + "loss": 0.6892, + "step": 25851 + }, + { + "epoch": 0.62, + "learning_rate": 6.6124302580847565e-06, + "loss": 0.9512, + "step": 25852 + }, + { + "epoch": 0.62, + "learning_rate": 6.611697426947068e-06, + "loss": 1.1396, + "step": 25853 + }, + { + "epoch": 0.62, + "learning_rate": 6.610964616365414e-06, + "loss": 1.1742, + "step": 25854 + }, + { + "epoch": 0.62, + "learning_rate": 6.61023182634424e-06, + "loss": 1.2202, + "step": 25855 + }, + { + "epoch": 0.62, + "learning_rate": 6.60949905688799e-06, + "loss": 1.0356, + "step": 25856 + }, + { + "epoch": 0.62, + "learning_rate": 6.608766308001116e-06, + "loss": 1.0303, + "step": 25857 + }, + { + "epoch": 0.62, + "learning_rate": 6.608033579688056e-06, + "loss": 0.9366, + "step": 25858 + }, + { + "epoch": 0.62, + "learning_rate": 6.607300871953259e-06, + "loss": 1.3015, + "step": 25859 + }, + { + "epoch": 0.62, + "learning_rate": 6.60656818480117e-06, + "loss": 0.9356, + "step": 25860 + }, + { + "epoch": 0.62, + "learning_rate": 6.605835518236235e-06, + "loss": 1.0907, + "step": 25861 + }, + { + "epoch": 0.62, + "learning_rate": 6.605102872262896e-06, + "loss": 1.2454, + "step": 25862 + }, + { + "epoch": 0.62, + "learning_rate": 6.6043702468856e-06, + "loss": 0.9736, + "step": 25863 + }, + { + "epoch": 0.62, + "learning_rate": 6.603637642108789e-06, + "loss": 1.0734, + "step": 25864 + }, + { + "epoch": 0.62, + "learning_rate": 6.6029050579369116e-06, + "loss": 1.1192, + "step": 25865 + }, + { + "epoch": 0.62, + "learning_rate": 6.60217249437441e-06, + "loss": 1.081, + "step": 25866 + }, + { + "epoch": 0.62, + "learning_rate": 6.601439951425726e-06, + "loss": 0.9914, + "step": 25867 + }, + { + "epoch": 0.62, + "learning_rate": 6.60070742909531e-06, + "loss": 1.0396, + "step": 25868 + }, + { + "epoch": 0.62, + "learning_rate": 6.599974927387601e-06, + "loss": 1.0626, + "step": 25869 + }, + { + "epoch": 0.62, + "learning_rate": 6.5992424463070455e-06, + "loss": 1.1281, + "step": 25870 + }, + { + "epoch": 0.62, + "learning_rate": 6.598509985858084e-06, + "loss": 1.1458, + "step": 25871 + }, + { + "epoch": 0.62, + "learning_rate": 6.597777546045163e-06, + "loss": 0.7397, + "step": 25872 + }, + { + "epoch": 0.62, + "learning_rate": 6.597045126872728e-06, + "loss": 1.3275, + "step": 25873 + }, + { + "epoch": 0.62, + "learning_rate": 6.596312728345218e-06, + "loss": 1.1063, + "step": 25874 + }, + { + "epoch": 0.62, + "learning_rate": 6.595580350467077e-06, + "loss": 1.0141, + "step": 25875 + }, + { + "epoch": 0.62, + "learning_rate": 6.59484799324275e-06, + "loss": 0.8855, + "step": 25876 + }, + { + "epoch": 0.62, + "learning_rate": 6.5941156566766795e-06, + "loss": 1.218, + "step": 25877 + }, + { + "epoch": 0.62, + "learning_rate": 6.593383340773309e-06, + "loss": 1.1834, + "step": 25878 + }, + { + "epoch": 0.62, + "learning_rate": 6.592651045537077e-06, + "loss": 1.2076, + "step": 25879 + }, + { + "epoch": 0.62, + "learning_rate": 6.591918770972434e-06, + "loss": 1.1517, + "step": 25880 + }, + { + "epoch": 0.62, + "learning_rate": 6.591186517083816e-06, + "loss": 1.2568, + "step": 25881 + }, + { + "epoch": 0.62, + "learning_rate": 6.590454283875669e-06, + "loss": 0.9097, + "step": 25882 + }, + { + "epoch": 0.62, + "learning_rate": 6.589722071352431e-06, + "loss": 1.0626, + "step": 25883 + }, + { + "epoch": 0.62, + "learning_rate": 6.58898987951855e-06, + "loss": 1.0095, + "step": 25884 + }, + { + "epoch": 0.62, + "learning_rate": 6.588257708378465e-06, + "loss": 1.1918, + "step": 25885 + }, + { + "epoch": 0.62, + "learning_rate": 6.587525557936616e-06, + "loss": 1.1053, + "step": 25886 + }, + { + "epoch": 0.62, + "learning_rate": 6.586793428197446e-06, + "loss": 1.1198, + "step": 25887 + }, + { + "epoch": 0.62, + "learning_rate": 6.5860613191654e-06, + "loss": 1.0639, + "step": 25888 + }, + { + "epoch": 0.62, + "learning_rate": 6.585329230844917e-06, + "loss": 1.1305, + "step": 25889 + }, + { + "epoch": 0.62, + "learning_rate": 6.584597163240438e-06, + "loss": 1.2225, + "step": 25890 + }, + { + "epoch": 0.62, + "learning_rate": 6.5838651163564025e-06, + "loss": 1.1628, + "step": 25891 + }, + { + "epoch": 0.62, + "learning_rate": 6.583133090197257e-06, + "loss": 1.019, + "step": 25892 + }, + { + "epoch": 0.62, + "learning_rate": 6.582401084767438e-06, + "loss": 1.0187, + "step": 25893 + }, + { + "epoch": 0.62, + "learning_rate": 6.581669100071388e-06, + "loss": 1.0642, + "step": 25894 + }, + { + "epoch": 0.62, + "learning_rate": 6.580937136113546e-06, + "loss": 0.8293, + "step": 25895 + }, + { + "epoch": 0.62, + "learning_rate": 6.580205192898356e-06, + "loss": 1.1038, + "step": 25896 + }, + { + "epoch": 0.62, + "learning_rate": 6.579473270430256e-06, + "loss": 1.1218, + "step": 25897 + }, + { + "epoch": 0.62, + "learning_rate": 6.578741368713688e-06, + "loss": 0.9628, + "step": 25898 + }, + { + "epoch": 0.62, + "learning_rate": 6.57800948775309e-06, + "loss": 1.1194, + "step": 25899 + }, + { + "epoch": 0.62, + "learning_rate": 6.577277627552904e-06, + "loss": 1.0645, + "step": 25900 + }, + { + "epoch": 0.62, + "learning_rate": 6.57654578811757e-06, + "loss": 0.9727, + "step": 25901 + }, + { + "epoch": 0.62, + "learning_rate": 6.575813969451528e-06, + "loss": 1.0623, + "step": 25902 + }, + { + "epoch": 0.62, + "learning_rate": 6.575082171559214e-06, + "loss": 1.0123, + "step": 25903 + }, + { + "epoch": 0.62, + "learning_rate": 6.574350394445074e-06, + "loss": 1.1439, + "step": 25904 + }, + { + "epoch": 0.62, + "learning_rate": 6.573618638113544e-06, + "loss": 1.0762, + "step": 25905 + }, + { + "epoch": 0.62, + "learning_rate": 6.572886902569064e-06, + "loss": 0.9492, + "step": 25906 + }, + { + "epoch": 0.62, + "learning_rate": 6.572155187816072e-06, + "loss": 1.0889, + "step": 25907 + }, + { + "epoch": 0.62, + "learning_rate": 6.571423493859008e-06, + "loss": 1.0358, + "step": 25908 + }, + { + "epoch": 0.62, + "learning_rate": 6.570691820702313e-06, + "loss": 1.1827, + "step": 25909 + }, + { + "epoch": 0.62, + "learning_rate": 6.569960168350422e-06, + "loss": 0.8531, + "step": 25910 + }, + { + "epoch": 0.62, + "learning_rate": 6.569228536807777e-06, + "loss": 1.2775, + "step": 25911 + }, + { + "epoch": 0.62, + "learning_rate": 6.5684969260788155e-06, + "loss": 1.204, + "step": 25912 + }, + { + "epoch": 0.62, + "learning_rate": 6.5677653361679765e-06, + "loss": 0.8639, + "step": 25913 + }, + { + "epoch": 0.62, + "learning_rate": 6.567033767079696e-06, + "loss": 1.0282, + "step": 25914 + }, + { + "epoch": 0.62, + "learning_rate": 6.566302218818415e-06, + "loss": 0.8399, + "step": 25915 + }, + { + "epoch": 0.62, + "learning_rate": 6.565570691388572e-06, + "loss": 0.9748, + "step": 25916 + }, + { + "epoch": 0.62, + "learning_rate": 6.5648391847946025e-06, + "loss": 1.1345, + "step": 25917 + }, + { + "epoch": 0.62, + "learning_rate": 6.564107699040944e-06, + "loss": 1.025, + "step": 25918 + }, + { + "epoch": 0.62, + "learning_rate": 6.563376234132039e-06, + "loss": 1.0547, + "step": 25919 + }, + { + "epoch": 0.62, + "learning_rate": 6.562644790072322e-06, + "loss": 0.9469, + "step": 25920 + }, + { + "epoch": 0.62, + "learning_rate": 6.56191336686623e-06, + "loss": 0.9749, + "step": 25921 + }, + { + "epoch": 0.62, + "learning_rate": 6.5611819645182e-06, + "loss": 1.0808, + "step": 25922 + }, + { + "epoch": 0.62, + "learning_rate": 6.560450583032672e-06, + "loss": 1.1563, + "step": 25923 + }, + { + "epoch": 0.62, + "learning_rate": 6.559719222414079e-06, + "loss": 0.925, + "step": 25924 + }, + { + "epoch": 0.62, + "learning_rate": 6.5589878826668615e-06, + "loss": 0.9215, + "step": 25925 + }, + { + "epoch": 0.62, + "learning_rate": 6.558256563795453e-06, + "loss": 0.9666, + "step": 25926 + }, + { + "epoch": 0.62, + "learning_rate": 6.557525265804297e-06, + "loss": 0.8995, + "step": 25927 + }, + { + "epoch": 0.62, + "learning_rate": 6.5567939886978224e-06, + "loss": 1.0748, + "step": 25928 + }, + { + "epoch": 0.62, + "learning_rate": 6.556062732480469e-06, + "loss": 0.9916, + "step": 25929 + }, + { + "epoch": 0.62, + "learning_rate": 6.555331497156671e-06, + "loss": 0.8999, + "step": 25930 + }, + { + "epoch": 0.62, + "learning_rate": 6.55460028273087e-06, + "loss": 1.0148, + "step": 25931 + }, + { + "epoch": 0.62, + "learning_rate": 6.5538690892074986e-06, + "loss": 1.2383, + "step": 25932 + }, + { + "epoch": 0.62, + "learning_rate": 6.553137916590992e-06, + "loss": 1.134, + "step": 25933 + }, + { + "epoch": 0.62, + "learning_rate": 6.5524067648857836e-06, + "loss": 1.1469, + "step": 25934 + }, + { + "epoch": 0.62, + "learning_rate": 6.551675634096318e-06, + "loss": 1.0822, + "step": 25935 + }, + { + "epoch": 0.62, + "learning_rate": 6.550944524227022e-06, + "loss": 1.0557, + "step": 25936 + }, + { + "epoch": 0.62, + "learning_rate": 6.550213435282336e-06, + "loss": 0.967, + "step": 25937 + }, + { + "epoch": 0.62, + "learning_rate": 6.549482367266691e-06, + "loss": 1.3487, + "step": 25938 + }, + { + "epoch": 0.62, + "learning_rate": 6.548751320184526e-06, + "loss": 1.0016, + "step": 25939 + }, + { + "epoch": 0.62, + "learning_rate": 6.548020294040276e-06, + "loss": 1.2274, + "step": 25940 + }, + { + "epoch": 0.62, + "learning_rate": 6.547289288838375e-06, + "loss": 1.0936, + "step": 25941 + }, + { + "epoch": 0.62, + "learning_rate": 6.546558304583255e-06, + "loss": 1.1778, + "step": 25942 + }, + { + "epoch": 0.62, + "learning_rate": 6.545827341279356e-06, + "loss": 1.0035, + "step": 25943 + }, + { + "epoch": 0.62, + "learning_rate": 6.54509639893111e-06, + "loss": 1.1499, + "step": 25944 + }, + { + "epoch": 0.62, + "learning_rate": 6.544365477542951e-06, + "loss": 1.1223, + "step": 25945 + }, + { + "epoch": 0.62, + "learning_rate": 6.543634577119313e-06, + "loss": 0.9661, + "step": 25946 + }, + { + "epoch": 0.62, + "learning_rate": 6.5429036976646335e-06, + "loss": 1.0362, + "step": 25947 + }, + { + "epoch": 0.62, + "learning_rate": 6.542172839183343e-06, + "loss": 0.9144, + "step": 25948 + }, + { + "epoch": 0.62, + "learning_rate": 6.541442001679876e-06, + "loss": 0.8448, + "step": 25949 + }, + { + "epoch": 0.62, + "learning_rate": 6.5407111851586655e-06, + "loss": 1.114, + "step": 25950 + }, + { + "epoch": 0.62, + "learning_rate": 6.539980389624147e-06, + "loss": 0.987, + "step": 25951 + }, + { + "epoch": 0.62, + "learning_rate": 6.539249615080756e-06, + "loss": 1.3542, + "step": 25952 + }, + { + "epoch": 0.62, + "learning_rate": 6.538518861532923e-06, + "loss": 1.058, + "step": 25953 + }, + { + "epoch": 0.62, + "learning_rate": 6.5377881289850794e-06, + "loss": 0.9356, + "step": 25954 + }, + { + "epoch": 0.62, + "learning_rate": 6.537057417441664e-06, + "loss": 1.154, + "step": 25955 + }, + { + "epoch": 0.62, + "learning_rate": 6.5363267269071055e-06, + "loss": 1.1953, + "step": 25956 + }, + { + "epoch": 0.62, + "learning_rate": 6.5355960573858355e-06, + "loss": 1.1357, + "step": 25957 + }, + { + "epoch": 0.62, + "learning_rate": 6.534865408882293e-06, + "loss": 1.0541, + "step": 25958 + }, + { + "epoch": 0.62, + "learning_rate": 6.534134781400905e-06, + "loss": 1.1493, + "step": 25959 + }, + { + "epoch": 0.62, + "learning_rate": 6.533404174946107e-06, + "loss": 1.1119, + "step": 25960 + }, + { + "epoch": 0.62, + "learning_rate": 6.53267358952233e-06, + "loss": 1.0932, + "step": 25961 + }, + { + "epoch": 0.62, + "learning_rate": 6.531943025134006e-06, + "loss": 1.0421, + "step": 25962 + }, + { + "epoch": 0.62, + "learning_rate": 6.531212481785571e-06, + "loss": 0.9112, + "step": 25963 + }, + { + "epoch": 0.62, + "learning_rate": 6.530481959481452e-06, + "loss": 1.2231, + "step": 25964 + }, + { + "epoch": 0.62, + "learning_rate": 6.529751458226081e-06, + "loss": 1.0085, + "step": 25965 + }, + { + "epoch": 0.62, + "learning_rate": 6.529020978023894e-06, + "loss": 1.2396, + "step": 25966 + }, + { + "epoch": 0.62, + "learning_rate": 6.52829051887932e-06, + "loss": 1.058, + "step": 25967 + }, + { + "epoch": 0.62, + "learning_rate": 6.527560080796791e-06, + "loss": 1.0491, + "step": 25968 + }, + { + "epoch": 0.62, + "learning_rate": 6.526829663780736e-06, + "loss": 0.9177, + "step": 25969 + }, + { + "epoch": 0.62, + "learning_rate": 6.5260992678355904e-06, + "loss": 0.9209, + "step": 25970 + }, + { + "epoch": 0.62, + "learning_rate": 6.525368892965784e-06, + "loss": 1.0149, + "step": 25971 + }, + { + "epoch": 0.62, + "learning_rate": 6.524638539175747e-06, + "loss": 1.0608, + "step": 25972 + }, + { + "epoch": 0.62, + "learning_rate": 6.523908206469909e-06, + "loss": 0.8799, + "step": 25973 + }, + { + "epoch": 0.62, + "learning_rate": 6.523177894852704e-06, + "loss": 1.1034, + "step": 25974 + }, + { + "epoch": 0.62, + "learning_rate": 6.5224476043285614e-06, + "loss": 0.9937, + "step": 25975 + }, + { + "epoch": 0.62, + "learning_rate": 6.521717334901909e-06, + "loss": 1.3264, + "step": 25976 + }, + { + "epoch": 0.62, + "learning_rate": 6.52098708657718e-06, + "loss": 1.0918, + "step": 25977 + }, + { + "epoch": 0.62, + "learning_rate": 6.520256859358804e-06, + "loss": 1.279, + "step": 25978 + }, + { + "epoch": 0.62, + "learning_rate": 6.519526653251212e-06, + "loss": 1.1224, + "step": 25979 + }, + { + "epoch": 0.62, + "learning_rate": 6.518796468258834e-06, + "loss": 1.0733, + "step": 25980 + }, + { + "epoch": 0.62, + "learning_rate": 6.5180663043860955e-06, + "loss": 1.0822, + "step": 25981 + }, + { + "epoch": 0.62, + "learning_rate": 6.517336161637433e-06, + "loss": 0.6708, + "step": 25982 + }, + { + "epoch": 0.62, + "learning_rate": 6.516606040017273e-06, + "loss": 1.0774, + "step": 25983 + }, + { + "epoch": 0.62, + "learning_rate": 6.515875939530043e-06, + "loss": 0.9449, + "step": 25984 + }, + { + "epoch": 0.62, + "learning_rate": 6.5151458601801745e-06, + "loss": 1.1641, + "step": 25985 + }, + { + "epoch": 0.62, + "learning_rate": 6.514415801972096e-06, + "loss": 1.1006, + "step": 25986 + }, + { + "epoch": 0.62, + "learning_rate": 6.513685764910238e-06, + "loss": 0.999, + "step": 25987 + }, + { + "epoch": 0.62, + "learning_rate": 6.512955748999029e-06, + "loss": 1.1487, + "step": 25988 + }, + { + "epoch": 0.63, + "learning_rate": 6.512225754242895e-06, + "loss": 1.0449, + "step": 25989 + }, + { + "epoch": 0.63, + "learning_rate": 6.511495780646269e-06, + "loss": 1.0705, + "step": 25990 + }, + { + "epoch": 0.63, + "learning_rate": 6.5107658282135765e-06, + "loss": 1.0282, + "step": 25991 + }, + { + "epoch": 0.63, + "learning_rate": 6.510035896949248e-06, + "loss": 1.2301, + "step": 25992 + }, + { + "epoch": 0.63, + "learning_rate": 6.509305986857709e-06, + "loss": 1.3851, + "step": 25993 + }, + { + "epoch": 0.63, + "learning_rate": 6.508576097943391e-06, + "loss": 0.6233, + "step": 25994 + }, + { + "epoch": 0.63, + "learning_rate": 6.5078462302107214e-06, + "loss": 1.0963, + "step": 25995 + }, + { + "epoch": 0.63, + "learning_rate": 6.507116383664127e-06, + "loss": 1.1059, + "step": 25996 + }, + { + "epoch": 0.63, + "learning_rate": 6.5063865583080335e-06, + "loss": 1.2188, + "step": 25997 + }, + { + "epoch": 0.63, + "learning_rate": 6.505656754146873e-06, + "loss": 0.993, + "step": 25998 + }, + { + "epoch": 0.63, + "learning_rate": 6.504926971185066e-06, + "loss": 1.0846, + "step": 25999 + }, + { + "epoch": 0.63, + "learning_rate": 6.504197209427051e-06, + "loss": 0.9454, + "step": 26000 + }, + { + "epoch": 0.63, + "learning_rate": 6.503467468877248e-06, + "loss": 0.9738, + "step": 26001 + }, + { + "epoch": 0.63, + "learning_rate": 6.502737749540084e-06, + "loss": 0.8928, + "step": 26002 + }, + { + "epoch": 0.63, + "learning_rate": 6.5020080514199905e-06, + "loss": 1.0463, + "step": 26003 + }, + { + "epoch": 0.63, + "learning_rate": 6.501278374521389e-06, + "loss": 1.1911, + "step": 26004 + }, + { + "epoch": 0.63, + "learning_rate": 6.500548718848709e-06, + "loss": 0.9255, + "step": 26005 + }, + { + "epoch": 0.63, + "learning_rate": 6.499819084406377e-06, + "loss": 1.2029, + "step": 26006 + }, + { + "epoch": 0.63, + "learning_rate": 6.49908947119882e-06, + "loss": 1.0311, + "step": 26007 + }, + { + "epoch": 0.63, + "learning_rate": 6.498359879230463e-06, + "loss": 1.0134, + "step": 26008 + }, + { + "epoch": 0.63, + "learning_rate": 6.497630308505733e-06, + "loss": 1.014, + "step": 26009 + }, + { + "epoch": 0.63, + "learning_rate": 6.496900759029055e-06, + "loss": 1.0806, + "step": 26010 + }, + { + "epoch": 0.63, + "learning_rate": 6.496171230804859e-06, + "loss": 1.0702, + "step": 26011 + }, + { + "epoch": 0.63, + "learning_rate": 6.495441723837567e-06, + "loss": 1.0489, + "step": 26012 + }, + { + "epoch": 0.63, + "learning_rate": 6.494712238131606e-06, + "loss": 1.2402, + "step": 26013 + }, + { + "epoch": 0.63, + "learning_rate": 6.493982773691398e-06, + "loss": 1.0845, + "step": 26014 + }, + { + "epoch": 0.63, + "learning_rate": 6.493253330521375e-06, + "loss": 0.7876, + "step": 26015 + }, + { + "epoch": 0.63, + "learning_rate": 6.492523908625958e-06, + "loss": 1.1277, + "step": 26016 + }, + { + "epoch": 0.63, + "learning_rate": 6.491794508009576e-06, + "loss": 1.0045, + "step": 26017 + }, + { + "epoch": 0.63, + "learning_rate": 6.491065128676648e-06, + "loss": 1.1933, + "step": 26018 + }, + { + "epoch": 0.63, + "learning_rate": 6.490335770631604e-06, + "loss": 1.0188, + "step": 26019 + }, + { + "epoch": 0.63, + "learning_rate": 6.4896064338788674e-06, + "loss": 1.0183, + "step": 26020 + }, + { + "epoch": 0.63, + "learning_rate": 6.4888771184228635e-06, + "loss": 0.9933, + "step": 26021 + }, + { + "epoch": 0.63, + "learning_rate": 6.488147824268014e-06, + "loss": 1.1032, + "step": 26022 + }, + { + "epoch": 0.63, + "learning_rate": 6.4874185514187485e-06, + "loss": 0.9471, + "step": 26023 + }, + { + "epoch": 0.63, + "learning_rate": 6.486689299879487e-06, + "loss": 0.7345, + "step": 26024 + }, + { + "epoch": 0.63, + "learning_rate": 6.485960069654656e-06, + "loss": 1.0731, + "step": 26025 + }, + { + "epoch": 0.63, + "learning_rate": 6.485230860748677e-06, + "loss": 1.224, + "step": 26026 + }, + { + "epoch": 0.63, + "learning_rate": 6.484501673165977e-06, + "loss": 0.9347, + "step": 26027 + }, + { + "epoch": 0.63, + "learning_rate": 6.483772506910978e-06, + "loss": 1.1061, + "step": 26028 + }, + { + "epoch": 0.63, + "learning_rate": 6.483043361988105e-06, + "loss": 0.8181, + "step": 26029 + }, + { + "epoch": 0.63, + "learning_rate": 6.482314238401779e-06, + "loss": 1.217, + "step": 26030 + }, + { + "epoch": 0.63, + "learning_rate": 6.481585136156425e-06, + "loss": 0.8717, + "step": 26031 + }, + { + "epoch": 0.63, + "learning_rate": 6.4808560552564685e-06, + "loss": 1.1537, + "step": 26032 + }, + { + "epoch": 0.63, + "learning_rate": 6.48012699570633e-06, + "loss": 0.9498, + "step": 26033 + }, + { + "epoch": 0.63, + "learning_rate": 6.479397957510432e-06, + "loss": 0.872, + "step": 26034 + }, + { + "epoch": 0.63, + "learning_rate": 6.478668940673199e-06, + "loss": 1.1264, + "step": 26035 + }, + { + "epoch": 0.63, + "learning_rate": 6.477939945199054e-06, + "loss": 1.0928, + "step": 26036 + }, + { + "epoch": 0.63, + "learning_rate": 6.477210971092416e-06, + "loss": 1.1164, + "step": 26037 + }, + { + "epoch": 0.63, + "learning_rate": 6.476482018357713e-06, + "loss": 1.2194, + "step": 26038 + }, + { + "epoch": 0.63, + "learning_rate": 6.475753086999366e-06, + "loss": 1.15, + "step": 26039 + }, + { + "epoch": 0.63, + "learning_rate": 6.475024177021795e-06, + "loss": 0.9993, + "step": 26040 + }, + { + "epoch": 0.63, + "learning_rate": 6.474295288429421e-06, + "loss": 1.051, + "step": 26041 + }, + { + "epoch": 0.63, + "learning_rate": 6.4735664212266705e-06, + "loss": 1.0118, + "step": 26042 + }, + { + "epoch": 0.63, + "learning_rate": 6.472837575417963e-06, + "loss": 0.9679, + "step": 26043 + }, + { + "epoch": 0.63, + "learning_rate": 6.472108751007719e-06, + "loss": 1.3297, + "step": 26044 + }, + { + "epoch": 0.63, + "learning_rate": 6.471379948000362e-06, + "loss": 1.0407, + "step": 26045 + }, + { + "epoch": 0.63, + "learning_rate": 6.470651166400313e-06, + "loss": 1.2444, + "step": 26046 + }, + { + "epoch": 0.63, + "learning_rate": 6.4699224062119936e-06, + "loss": 1.1418, + "step": 26047 + }, + { + "epoch": 0.63, + "learning_rate": 6.469193667439826e-06, + "loss": 1.1399, + "step": 26048 + }, + { + "epoch": 0.63, + "learning_rate": 6.468464950088225e-06, + "loss": 1.0217, + "step": 26049 + }, + { + "epoch": 0.63, + "learning_rate": 6.467736254161621e-06, + "loss": 1.1635, + "step": 26050 + }, + { + "epoch": 0.63, + "learning_rate": 6.467007579664429e-06, + "loss": 0.9637, + "step": 26051 + }, + { + "epoch": 0.63, + "learning_rate": 6.466278926601073e-06, + "loss": 1.1644, + "step": 26052 + }, + { + "epoch": 0.63, + "learning_rate": 6.4655502949759665e-06, + "loss": 1.4858, + "step": 26053 + }, + { + "epoch": 0.63, + "learning_rate": 6.464821684793538e-06, + "loss": 1.0014, + "step": 26054 + }, + { + "epoch": 0.63, + "learning_rate": 6.464093096058206e-06, + "loss": 0.6985, + "step": 26055 + }, + { + "epoch": 0.63, + "learning_rate": 6.46336452877439e-06, + "loss": 1.0545, + "step": 26056 + }, + { + "epoch": 0.63, + "learning_rate": 6.462635982946506e-06, + "loss": 1.0418, + "step": 26057 + }, + { + "epoch": 0.63, + "learning_rate": 6.46190745857898e-06, + "loss": 1.1651, + "step": 26058 + }, + { + "epoch": 0.63, + "learning_rate": 6.461178955676228e-06, + "loss": 1.1152, + "step": 26059 + }, + { + "epoch": 0.63, + "learning_rate": 6.460450474242672e-06, + "loss": 1.0715, + "step": 26060 + }, + { + "epoch": 0.63, + "learning_rate": 6.45972201428273e-06, + "loss": 1.2115, + "step": 26061 + }, + { + "epoch": 0.63, + "learning_rate": 6.458993575800822e-06, + "loss": 1.1058, + "step": 26062 + }, + { + "epoch": 0.63, + "learning_rate": 6.4582651588013665e-06, + "loss": 0.9197, + "step": 26063 + }, + { + "epoch": 0.63, + "learning_rate": 6.457536763288785e-06, + "loss": 1.0549, + "step": 26064 + }, + { + "epoch": 0.63, + "learning_rate": 6.45680838926749e-06, + "loss": 1.1231, + "step": 26065 + }, + { + "epoch": 0.63, + "learning_rate": 6.456080036741909e-06, + "loss": 1.2284, + "step": 26066 + }, + { + "epoch": 0.63, + "learning_rate": 6.455351705716457e-06, + "loss": 0.8382, + "step": 26067 + }, + { + "epoch": 0.63, + "learning_rate": 6.4546233961955525e-06, + "loss": 1.0845, + "step": 26068 + }, + { + "epoch": 0.63, + "learning_rate": 6.453895108183612e-06, + "loss": 1.1681, + "step": 26069 + }, + { + "epoch": 0.63, + "learning_rate": 6.453166841685057e-06, + "loss": 1.3139, + "step": 26070 + }, + { + "epoch": 0.63, + "learning_rate": 6.452438596704305e-06, + "loss": 1.1615, + "step": 26071 + }, + { + "epoch": 0.63, + "learning_rate": 6.451710373245774e-06, + "loss": 1.1884, + "step": 26072 + }, + { + "epoch": 0.63, + "learning_rate": 6.450982171313879e-06, + "loss": 1.0478, + "step": 26073 + }, + { + "epoch": 0.63, + "learning_rate": 6.450253990913042e-06, + "loss": 0.9931, + "step": 26074 + }, + { + "epoch": 0.63, + "learning_rate": 6.44952583204768e-06, + "loss": 1.0087, + "step": 26075 + }, + { + "epoch": 0.63, + "learning_rate": 6.4487976947222085e-06, + "loss": 1.144, + "step": 26076 + }, + { + "epoch": 0.63, + "learning_rate": 6.448069578941045e-06, + "loss": 1.0498, + "step": 26077 + }, + { + "epoch": 0.63, + "learning_rate": 6.447341484708609e-06, + "loss": 1.1248, + "step": 26078 + }, + { + "epoch": 0.63, + "learning_rate": 6.446613412029318e-06, + "loss": 1.0389, + "step": 26079 + }, + { + "epoch": 0.63, + "learning_rate": 6.445885360907586e-06, + "loss": 0.9784, + "step": 26080 + }, + { + "epoch": 0.63, + "learning_rate": 6.445157331347831e-06, + "loss": 0.8155, + "step": 26081 + }, + { + "epoch": 0.63, + "learning_rate": 6.444429323354472e-06, + "loss": 0.8696, + "step": 26082 + }, + { + "epoch": 0.63, + "learning_rate": 6.443701336931924e-06, + "loss": 1.1044, + "step": 26083 + }, + { + "epoch": 0.63, + "learning_rate": 6.4429733720846e-06, + "loss": 1.1304, + "step": 26084 + }, + { + "epoch": 0.63, + "learning_rate": 6.4422454288169234e-06, + "loss": 1.1301, + "step": 26085 + }, + { + "epoch": 0.63, + "learning_rate": 6.441517507133307e-06, + "loss": 0.8422, + "step": 26086 + }, + { + "epoch": 0.63, + "learning_rate": 6.440789607038167e-06, + "loss": 1.1573, + "step": 26087 + }, + { + "epoch": 0.63, + "learning_rate": 6.440061728535916e-06, + "loss": 1.088, + "step": 26088 + }, + { + "epoch": 0.63, + "learning_rate": 6.439333871630975e-06, + "loss": 0.855, + "step": 26089 + }, + { + "epoch": 0.63, + "learning_rate": 6.438606036327759e-06, + "loss": 1.0954, + "step": 26090 + }, + { + "epoch": 0.63, + "learning_rate": 6.437878222630683e-06, + "loss": 0.9517, + "step": 26091 + }, + { + "epoch": 0.63, + "learning_rate": 6.437150430544159e-06, + "loss": 0.839, + "step": 26092 + }, + { + "epoch": 0.63, + "learning_rate": 6.436422660072608e-06, + "loss": 0.9096, + "step": 26093 + }, + { + "epoch": 0.63, + "learning_rate": 6.435694911220441e-06, + "loss": 1.0996, + "step": 26094 + }, + { + "epoch": 0.63, + "learning_rate": 6.434967183992076e-06, + "loss": 1.1094, + "step": 26095 + }, + { + "epoch": 0.63, + "learning_rate": 6.434239478391923e-06, + "loss": 0.8941, + "step": 26096 + }, + { + "epoch": 0.63, + "learning_rate": 6.4335117944244056e-06, + "loss": 1.0579, + "step": 26097 + }, + { + "epoch": 0.63, + "learning_rate": 6.43278413209393e-06, + "loss": 0.9621, + "step": 26098 + }, + { + "epoch": 0.63, + "learning_rate": 6.432056491404917e-06, + "loss": 1.1712, + "step": 26099 + }, + { + "epoch": 0.63, + "learning_rate": 6.431328872361773e-06, + "loss": 0.9469, + "step": 26100 + }, + { + "epoch": 0.63, + "learning_rate": 6.43060127496892e-06, + "loss": 1.144, + "step": 26101 + }, + { + "epoch": 0.63, + "learning_rate": 6.429873699230771e-06, + "loss": 1.1742, + "step": 26102 + }, + { + "epoch": 0.63, + "learning_rate": 6.429146145151737e-06, + "loss": 1.1497, + "step": 26103 + }, + { + "epoch": 0.63, + "learning_rate": 6.428418612736232e-06, + "loss": 0.9817, + "step": 26104 + }, + { + "epoch": 0.63, + "learning_rate": 6.427691101988673e-06, + "loss": 1.1819, + "step": 26105 + }, + { + "epoch": 0.63, + "learning_rate": 6.426963612913473e-06, + "loss": 0.8956, + "step": 26106 + }, + { + "epoch": 0.63, + "learning_rate": 6.426236145515043e-06, + "loss": 1.2166, + "step": 26107 + }, + { + "epoch": 0.63, + "learning_rate": 6.425508699797796e-06, + "loss": 1.0324, + "step": 26108 + }, + { + "epoch": 0.63, + "learning_rate": 6.42478127576615e-06, + "loss": 1.1877, + "step": 26109 + }, + { + "epoch": 0.63, + "learning_rate": 6.424053873424514e-06, + "loss": 1.2322, + "step": 26110 + }, + { + "epoch": 0.63, + "learning_rate": 6.423326492777302e-06, + "loss": 1.0381, + "step": 26111 + }, + { + "epoch": 0.63, + "learning_rate": 6.422599133828924e-06, + "loss": 1.0281, + "step": 26112 + }, + { + "epoch": 0.63, + "learning_rate": 6.421871796583798e-06, + "loss": 0.962, + "step": 26113 + }, + { + "epoch": 0.63, + "learning_rate": 6.421144481046335e-06, + "loss": 0.9781, + "step": 26114 + }, + { + "epoch": 0.63, + "learning_rate": 6.420417187220946e-06, + "loss": 1.0086, + "step": 26115 + }, + { + "epoch": 0.63, + "learning_rate": 6.419689915112044e-06, + "loss": 1.1858, + "step": 26116 + }, + { + "epoch": 0.63, + "learning_rate": 6.418962664724041e-06, + "loss": 0.9457, + "step": 26117 + }, + { + "epoch": 0.63, + "learning_rate": 6.41823543606135e-06, + "loss": 1.0069, + "step": 26118 + }, + { + "epoch": 0.63, + "learning_rate": 6.417508229128382e-06, + "loss": 0.8329, + "step": 26119 + }, + { + "epoch": 0.63, + "learning_rate": 6.416781043929547e-06, + "loss": 0.923, + "step": 26120 + }, + { + "epoch": 0.63, + "learning_rate": 6.4160538804692595e-06, + "loss": 1.1301, + "step": 26121 + }, + { + "epoch": 0.63, + "learning_rate": 6.41532673875193e-06, + "loss": 1.1203, + "step": 26122 + }, + { + "epoch": 0.63, + "learning_rate": 6.414599618781971e-06, + "loss": 1.2106, + "step": 26123 + }, + { + "epoch": 0.63, + "learning_rate": 6.413872520563792e-06, + "loss": 1.0552, + "step": 26124 + }, + { + "epoch": 0.63, + "learning_rate": 6.413145444101805e-06, + "loss": 0.9262, + "step": 26125 + }, + { + "epoch": 0.63, + "learning_rate": 6.4124183894004224e-06, + "loss": 1.2085, + "step": 26126 + }, + { + "epoch": 0.63, + "learning_rate": 6.411691356464052e-06, + "loss": 0.9431, + "step": 26127 + }, + { + "epoch": 0.63, + "learning_rate": 6.410964345297105e-06, + "loss": 0.9382, + "step": 26128 + }, + { + "epoch": 0.63, + "learning_rate": 6.410237355903994e-06, + "loss": 0.9513, + "step": 26129 + }, + { + "epoch": 0.63, + "learning_rate": 6.409510388289131e-06, + "loss": 1.0856, + "step": 26130 + }, + { + "epoch": 0.63, + "learning_rate": 6.40878344245692e-06, + "loss": 1.1838, + "step": 26131 + }, + { + "epoch": 0.63, + "learning_rate": 6.408056518411778e-06, + "loss": 1.1847, + "step": 26132 + }, + { + "epoch": 0.63, + "learning_rate": 6.407329616158111e-06, + "loss": 1.1255, + "step": 26133 + }, + { + "epoch": 0.63, + "learning_rate": 6.406602735700331e-06, + "loss": 0.9511, + "step": 26134 + }, + { + "epoch": 0.63, + "learning_rate": 6.405875877042847e-06, + "loss": 1.0095, + "step": 26135 + }, + { + "epoch": 0.63, + "learning_rate": 6.405149040190067e-06, + "loss": 1.4309, + "step": 26136 + }, + { + "epoch": 0.63, + "learning_rate": 6.404422225146405e-06, + "loss": 1.0533, + "step": 26137 + }, + { + "epoch": 0.63, + "learning_rate": 6.403695431916266e-06, + "loss": 0.8708, + "step": 26138 + }, + { + "epoch": 0.63, + "learning_rate": 6.402968660504059e-06, + "loss": 0.8662, + "step": 26139 + }, + { + "epoch": 0.63, + "learning_rate": 6.402241910914198e-06, + "loss": 1.2425, + "step": 26140 + }, + { + "epoch": 0.63, + "learning_rate": 6.401515183151088e-06, + "loss": 0.6504, + "step": 26141 + }, + { + "epoch": 0.63, + "learning_rate": 6.4007884772191395e-06, + "loss": 0.9942, + "step": 26142 + }, + { + "epoch": 0.63, + "learning_rate": 6.400061793122759e-06, + "loss": 1.0288, + "step": 26143 + }, + { + "epoch": 0.63, + "learning_rate": 6.39933513086636e-06, + "loss": 1.1098, + "step": 26144 + }, + { + "epoch": 0.63, + "learning_rate": 6.398608490454345e-06, + "loss": 0.9136, + "step": 26145 + }, + { + "epoch": 0.63, + "learning_rate": 6.397881871891127e-06, + "loss": 1.0033, + "step": 26146 + }, + { + "epoch": 0.63, + "learning_rate": 6.397155275181111e-06, + "loss": 1.1063, + "step": 26147 + }, + { + "epoch": 0.63, + "learning_rate": 6.396428700328707e-06, + "loss": 1.2039, + "step": 26148 + }, + { + "epoch": 0.63, + "learning_rate": 6.395702147338323e-06, + "loss": 1.1086, + "step": 26149 + }, + { + "epoch": 0.63, + "learning_rate": 6.3949756162143665e-06, + "loss": 0.9191, + "step": 26150 + }, + { + "epoch": 0.63, + "learning_rate": 6.394249106961243e-06, + "loss": 1.0488, + "step": 26151 + }, + { + "epoch": 0.63, + "learning_rate": 6.393522619583363e-06, + "loss": 1.2438, + "step": 26152 + }, + { + "epoch": 0.63, + "learning_rate": 6.392796154085135e-06, + "loss": 1.1332, + "step": 26153 + }, + { + "epoch": 0.63, + "learning_rate": 6.392069710470963e-06, + "loss": 0.8352, + "step": 26154 + }, + { + "epoch": 0.63, + "learning_rate": 6.391343288745253e-06, + "loss": 1.1023, + "step": 26155 + }, + { + "epoch": 0.63, + "learning_rate": 6.390616888912418e-06, + "loss": 0.9599, + "step": 26156 + }, + { + "epoch": 0.63, + "learning_rate": 6.3898905109768595e-06, + "loss": 1.2698, + "step": 26157 + }, + { + "epoch": 0.63, + "learning_rate": 6.3891641549429865e-06, + "loss": 1.1194, + "step": 26158 + }, + { + "epoch": 0.63, + "learning_rate": 6.388437820815204e-06, + "loss": 0.955, + "step": 26159 + }, + { + "epoch": 0.63, + "learning_rate": 6.387711508597921e-06, + "loss": 0.9675, + "step": 26160 + }, + { + "epoch": 0.63, + "learning_rate": 6.386985218295544e-06, + "loss": 0.8647, + "step": 26161 + }, + { + "epoch": 0.63, + "learning_rate": 6.386258949912476e-06, + "loss": 1.2154, + "step": 26162 + }, + { + "epoch": 0.63, + "learning_rate": 6.385532703453124e-06, + "loss": 1.1726, + "step": 26163 + }, + { + "epoch": 0.63, + "learning_rate": 6.3848064789218965e-06, + "loss": 1.0137, + "step": 26164 + }, + { + "epoch": 0.63, + "learning_rate": 6.384080276323199e-06, + "loss": 1.0549, + "step": 26165 + }, + { + "epoch": 0.63, + "learning_rate": 6.38335409566143e-06, + "loss": 0.9711, + "step": 26166 + }, + { + "epoch": 0.63, + "learning_rate": 6.382627936941006e-06, + "loss": 1.1807, + "step": 26167 + }, + { + "epoch": 0.63, + "learning_rate": 6.3819018001663255e-06, + "loss": 0.9636, + "step": 26168 + }, + { + "epoch": 0.63, + "learning_rate": 6.381175685341798e-06, + "loss": 1.1751, + "step": 26169 + }, + { + "epoch": 0.63, + "learning_rate": 6.380449592471826e-06, + "loss": 1.1398, + "step": 26170 + }, + { + "epoch": 0.63, + "learning_rate": 6.379723521560816e-06, + "loss": 1.0486, + "step": 26171 + }, + { + "epoch": 0.63, + "learning_rate": 6.378997472613168e-06, + "loss": 1.1886, + "step": 26172 + }, + { + "epoch": 0.63, + "learning_rate": 6.3782714456332936e-06, + "loss": 0.9376, + "step": 26173 + }, + { + "epoch": 0.63, + "learning_rate": 6.377545440625594e-06, + "loss": 1.0714, + "step": 26174 + }, + { + "epoch": 0.63, + "learning_rate": 6.376819457594475e-06, + "loss": 1.2123, + "step": 26175 + }, + { + "epoch": 0.63, + "learning_rate": 6.376093496544338e-06, + "loss": 1.2915, + "step": 26176 + }, + { + "epoch": 0.63, + "learning_rate": 6.37536755747959e-06, + "loss": 0.9147, + "step": 26177 + }, + { + "epoch": 0.63, + "learning_rate": 6.374641640404636e-06, + "loss": 1.1755, + "step": 26178 + }, + { + "epoch": 0.63, + "learning_rate": 6.373915745323879e-06, + "loss": 1.0387, + "step": 26179 + }, + { + "epoch": 0.63, + "learning_rate": 6.3731898722417186e-06, + "loss": 0.9568, + "step": 26180 + }, + { + "epoch": 0.63, + "learning_rate": 6.372464021162565e-06, + "loss": 1.0233, + "step": 26181 + }, + { + "epoch": 0.63, + "learning_rate": 6.371738192090818e-06, + "loss": 1.2063, + "step": 26182 + }, + { + "epoch": 0.63, + "learning_rate": 6.371012385030884e-06, + "loss": 1.1931, + "step": 26183 + }, + { + "epoch": 0.63, + "learning_rate": 6.37028659998716e-06, + "loss": 0.9601, + "step": 26184 + }, + { + "epoch": 0.63, + "learning_rate": 6.369560836964057e-06, + "loss": 0.969, + "step": 26185 + }, + { + "epoch": 0.63, + "learning_rate": 6.368835095965973e-06, + "loss": 1.0436, + "step": 26186 + }, + { + "epoch": 0.63, + "learning_rate": 6.368109376997314e-06, + "loss": 1.1184, + "step": 26187 + }, + { + "epoch": 0.63, + "learning_rate": 6.3673836800624775e-06, + "loss": 0.9232, + "step": 26188 + }, + { + "epoch": 0.63, + "learning_rate": 6.366658005165873e-06, + "loss": 0.9698, + "step": 26189 + }, + { + "epoch": 0.63, + "learning_rate": 6.365932352311899e-06, + "loss": 0.9843, + "step": 26190 + }, + { + "epoch": 0.63, + "learning_rate": 6.365206721504959e-06, + "loss": 0.9779, + "step": 26191 + }, + { + "epoch": 0.63, + "learning_rate": 6.364481112749453e-06, + "loss": 0.8925, + "step": 26192 + }, + { + "epoch": 0.63, + "learning_rate": 6.3637555260497875e-06, + "loss": 0.9197, + "step": 26193 + }, + { + "epoch": 0.63, + "learning_rate": 6.363029961410361e-06, + "loss": 0.8557, + "step": 26194 + }, + { + "epoch": 0.63, + "learning_rate": 6.362304418835577e-06, + "loss": 1.0806, + "step": 26195 + }, + { + "epoch": 0.63, + "learning_rate": 6.361578898329834e-06, + "loss": 1.0148, + "step": 26196 + }, + { + "epoch": 0.63, + "learning_rate": 6.360853399897538e-06, + "loss": 0.8978, + "step": 26197 + }, + { + "epoch": 0.63, + "learning_rate": 6.360127923543088e-06, + "loss": 1.0917, + "step": 26198 + }, + { + "epoch": 0.63, + "learning_rate": 6.359402469270887e-06, + "loss": 0.9878, + "step": 26199 + }, + { + "epoch": 0.63, + "learning_rate": 6.358677037085331e-06, + "loss": 1.0612, + "step": 26200 + }, + { + "epoch": 0.63, + "learning_rate": 6.357951626990828e-06, + "loss": 0.9579, + "step": 26201 + }, + { + "epoch": 0.63, + "learning_rate": 6.357226238991776e-06, + "loss": 1.1706, + "step": 26202 + }, + { + "epoch": 0.63, + "learning_rate": 6.356500873092576e-06, + "loss": 1.1217, + "step": 26203 + }, + { + "epoch": 0.63, + "learning_rate": 6.355775529297625e-06, + "loss": 0.8692, + "step": 26204 + }, + { + "epoch": 0.63, + "learning_rate": 6.35505020761133e-06, + "loss": 1.1709, + "step": 26205 + }, + { + "epoch": 0.63, + "learning_rate": 6.354324908038087e-06, + "loss": 1.1439, + "step": 26206 + }, + { + "epoch": 0.63, + "learning_rate": 6.353599630582298e-06, + "loss": 0.7076, + "step": 26207 + }, + { + "epoch": 0.63, + "learning_rate": 6.352874375248359e-06, + "loss": 1.209, + "step": 26208 + }, + { + "epoch": 0.63, + "learning_rate": 6.3521491420406765e-06, + "loss": 1.0785, + "step": 26209 + }, + { + "epoch": 0.63, + "learning_rate": 6.3514239309636474e-06, + "loss": 1.1001, + "step": 26210 + }, + { + "epoch": 0.63, + "learning_rate": 6.350698742021668e-06, + "loss": 1.1006, + "step": 26211 + }, + { + "epoch": 0.63, + "learning_rate": 6.349973575219144e-06, + "loss": 0.9843, + "step": 26212 + }, + { + "epoch": 0.63, + "learning_rate": 6.3492484305604705e-06, + "loss": 0.9776, + "step": 26213 + }, + { + "epoch": 0.63, + "learning_rate": 6.34852330805005e-06, + "loss": 1.1483, + "step": 26214 + }, + { + "epoch": 0.63, + "learning_rate": 6.347798207692277e-06, + "loss": 1.1309, + "step": 26215 + }, + { + "epoch": 0.63, + "learning_rate": 6.347073129491554e-06, + "loss": 1.3093, + "step": 26216 + }, + { + "epoch": 0.63, + "learning_rate": 6.34634807345228e-06, + "loss": 1.1444, + "step": 26217 + }, + { + "epoch": 0.63, + "learning_rate": 6.345623039578853e-06, + "loss": 0.9352, + "step": 26218 + }, + { + "epoch": 0.63, + "learning_rate": 6.34489802787567e-06, + "loss": 1.0011, + "step": 26219 + }, + { + "epoch": 0.63, + "learning_rate": 6.344173038347132e-06, + "loss": 0.8886, + "step": 26220 + }, + { + "epoch": 0.63, + "learning_rate": 6.343448070997636e-06, + "loss": 1.1274, + "step": 26221 + }, + { + "epoch": 0.63, + "learning_rate": 6.3427231258315816e-06, + "loss": 1.077, + "step": 26222 + }, + { + "epoch": 0.63, + "learning_rate": 6.341998202853363e-06, + "loss": 1.1438, + "step": 26223 + }, + { + "epoch": 0.63, + "learning_rate": 6.341273302067384e-06, + "loss": 1.0274, + "step": 26224 + }, + { + "epoch": 0.63, + "learning_rate": 6.340548423478037e-06, + "loss": 1.0939, + "step": 26225 + }, + { + "epoch": 0.63, + "learning_rate": 6.339823567089722e-06, + "loss": 0.9473, + "step": 26226 + }, + { + "epoch": 0.63, + "learning_rate": 6.339098732906837e-06, + "loss": 0.8437, + "step": 26227 + }, + { + "epoch": 0.63, + "learning_rate": 6.3383739209337806e-06, + "loss": 0.9509, + "step": 26228 + }, + { + "epoch": 0.63, + "learning_rate": 6.337649131174946e-06, + "loss": 1.2822, + "step": 26229 + }, + { + "epoch": 0.63, + "learning_rate": 6.336924363634734e-06, + "loss": 1.0912, + "step": 26230 + }, + { + "epoch": 0.63, + "learning_rate": 6.336199618317538e-06, + "loss": 1.1153, + "step": 26231 + }, + { + "epoch": 0.63, + "learning_rate": 6.33547489522776e-06, + "loss": 1.0202, + "step": 26232 + }, + { + "epoch": 0.63, + "learning_rate": 6.3347501943697945e-06, + "loss": 0.9912, + "step": 26233 + }, + { + "epoch": 0.63, + "learning_rate": 6.3340255157480344e-06, + "loss": 1.1474, + "step": 26234 + }, + { + "epoch": 0.63, + "learning_rate": 6.33330085936688e-06, + "loss": 1.1894, + "step": 26235 + }, + { + "epoch": 0.63, + "learning_rate": 6.332576225230727e-06, + "loss": 0.8482, + "step": 26236 + }, + { + "epoch": 0.63, + "learning_rate": 6.331851613343973e-06, + "loss": 1.0108, + "step": 26237 + }, + { + "epoch": 0.63, + "learning_rate": 6.331127023711012e-06, + "loss": 0.9537, + "step": 26238 + }, + { + "epoch": 0.63, + "learning_rate": 6.330402456336237e-06, + "loss": 0.9559, + "step": 26239 + }, + { + "epoch": 0.63, + "learning_rate": 6.329677911224052e-06, + "loss": 1.1414, + "step": 26240 + }, + { + "epoch": 0.63, + "learning_rate": 6.3289533883788454e-06, + "loss": 1.165, + "step": 26241 + }, + { + "epoch": 0.63, + "learning_rate": 6.328228887805017e-06, + "loss": 1.0838, + "step": 26242 + }, + { + "epoch": 0.63, + "learning_rate": 6.327504409506956e-06, + "loss": 1.2914, + "step": 26243 + }, + { + "epoch": 0.63, + "learning_rate": 6.326779953489065e-06, + "loss": 0.9457, + "step": 26244 + }, + { + "epoch": 0.63, + "learning_rate": 6.326055519755736e-06, + "loss": 0.9873, + "step": 26245 + }, + { + "epoch": 0.63, + "learning_rate": 6.325331108311364e-06, + "loss": 0.782, + "step": 26246 + }, + { + "epoch": 0.63, + "learning_rate": 6.324606719160342e-06, + "loss": 0.9884, + "step": 26247 + }, + { + "epoch": 0.63, + "learning_rate": 6.323882352307069e-06, + "loss": 1.1874, + "step": 26248 + }, + { + "epoch": 0.63, + "learning_rate": 6.323158007755936e-06, + "loss": 1.2122, + "step": 26249 + }, + { + "epoch": 0.63, + "learning_rate": 6.32243368551134e-06, + "loss": 1.0344, + "step": 26250 + }, + { + "epoch": 0.63, + "learning_rate": 6.32170938557767e-06, + "loss": 1.1796, + "step": 26251 + }, + { + "epoch": 0.63, + "learning_rate": 6.320985107959328e-06, + "loss": 1.2045, + "step": 26252 + }, + { + "epoch": 0.63, + "learning_rate": 6.320260852660702e-06, + "loss": 0.8347, + "step": 26253 + }, + { + "epoch": 0.63, + "learning_rate": 6.319536619686187e-06, + "loss": 1.0252, + "step": 26254 + }, + { + "epoch": 0.63, + "learning_rate": 6.318812409040178e-06, + "loss": 0.8075, + "step": 26255 + }, + { + "epoch": 0.63, + "learning_rate": 6.31808822072707e-06, + "loss": 1.0177, + "step": 26256 + }, + { + "epoch": 0.63, + "learning_rate": 6.317364054751255e-06, + "loss": 1.2705, + "step": 26257 + }, + { + "epoch": 0.63, + "learning_rate": 6.316639911117124e-06, + "loss": 1.0322, + "step": 26258 + }, + { + "epoch": 0.63, + "learning_rate": 6.315915789829072e-06, + "loss": 1.0687, + "step": 26259 + }, + { + "epoch": 0.63, + "learning_rate": 6.315191690891496e-06, + "loss": 0.8954, + "step": 26260 + }, + { + "epoch": 0.63, + "learning_rate": 6.314467614308782e-06, + "loss": 0.9659, + "step": 26261 + }, + { + "epoch": 0.63, + "learning_rate": 6.313743560085326e-06, + "loss": 1.0278, + "step": 26262 + }, + { + "epoch": 0.63, + "learning_rate": 6.313019528225522e-06, + "loss": 1.1724, + "step": 26263 + }, + { + "epoch": 0.63, + "learning_rate": 6.312295518733761e-06, + "loss": 1.3027, + "step": 26264 + }, + { + "epoch": 0.63, + "learning_rate": 6.311571531614435e-06, + "loss": 0.9316, + "step": 26265 + }, + { + "epoch": 0.63, + "learning_rate": 6.310847566871937e-06, + "loss": 0.9322, + "step": 26266 + }, + { + "epoch": 0.63, + "learning_rate": 6.31012362451066e-06, + "loss": 1.0752, + "step": 26267 + }, + { + "epoch": 0.63, + "learning_rate": 6.309399704534995e-06, + "loss": 1.3166, + "step": 26268 + }, + { + "epoch": 0.63, + "learning_rate": 6.308675806949335e-06, + "loss": 1.0644, + "step": 26269 + }, + { + "epoch": 0.63, + "learning_rate": 6.307951931758067e-06, + "loss": 1.1506, + "step": 26270 + }, + { + "epoch": 0.63, + "learning_rate": 6.3072280789655895e-06, + "loss": 1.1422, + "step": 26271 + }, + { + "epoch": 0.63, + "learning_rate": 6.3065042485762906e-06, + "loss": 1.1666, + "step": 26272 + }, + { + "epoch": 0.63, + "learning_rate": 6.3057804405945596e-06, + "loss": 1.189, + "step": 26273 + }, + { + "epoch": 0.63, + "learning_rate": 6.30505665502479e-06, + "loss": 0.8994, + "step": 26274 + }, + { + "epoch": 0.63, + "learning_rate": 6.304332891871374e-06, + "loss": 1.1393, + "step": 26275 + }, + { + "epoch": 0.63, + "learning_rate": 6.303609151138701e-06, + "loss": 0.99, + "step": 26276 + }, + { + "epoch": 0.63, + "learning_rate": 6.302885432831162e-06, + "loss": 1.4285, + "step": 26277 + }, + { + "epoch": 0.63, + "learning_rate": 6.3021617369531455e-06, + "loss": 0.8691, + "step": 26278 + }, + { + "epoch": 0.63, + "learning_rate": 6.301438063509046e-06, + "loss": 1.0529, + "step": 26279 + }, + { + "epoch": 0.63, + "learning_rate": 6.30071441250325e-06, + "loss": 0.9614, + "step": 26280 + }, + { + "epoch": 0.63, + "learning_rate": 6.299990783940152e-06, + "loss": 0.8812, + "step": 26281 + }, + { + "epoch": 0.63, + "learning_rate": 6.299267177824137e-06, + "loss": 0.8395, + "step": 26282 + }, + { + "epoch": 0.63, + "learning_rate": 6.298543594159598e-06, + "loss": 1.0332, + "step": 26283 + }, + { + "epoch": 0.63, + "learning_rate": 6.2978200329509265e-06, + "loss": 0.9755, + "step": 26284 + }, + { + "epoch": 0.63, + "learning_rate": 6.297096494202509e-06, + "loss": 1.1225, + "step": 26285 + }, + { + "epoch": 0.63, + "learning_rate": 6.296372977918734e-06, + "loss": 1.0878, + "step": 26286 + }, + { + "epoch": 0.63, + "learning_rate": 6.295649484103996e-06, + "loss": 1.2137, + "step": 26287 + }, + { + "epoch": 0.63, + "learning_rate": 6.294926012762681e-06, + "loss": 0.9875, + "step": 26288 + }, + { + "epoch": 0.63, + "learning_rate": 6.294202563899176e-06, + "loss": 1.0271, + "step": 26289 + }, + { + "epoch": 0.63, + "learning_rate": 6.2934791375178725e-06, + "loss": 1.1522, + "step": 26290 + }, + { + "epoch": 0.63, + "learning_rate": 6.292755733623162e-06, + "loss": 0.6861, + "step": 26291 + }, + { + "epoch": 0.63, + "learning_rate": 6.2920323522194295e-06, + "loss": 0.9668, + "step": 26292 + }, + { + "epoch": 0.63, + "learning_rate": 6.2913089933110645e-06, + "loss": 1.2281, + "step": 26293 + }, + { + "epoch": 0.63, + "learning_rate": 6.2905856569024535e-06, + "loss": 1.1526, + "step": 26294 + }, + { + "epoch": 0.63, + "learning_rate": 6.289862342997989e-06, + "loss": 1.1956, + "step": 26295 + }, + { + "epoch": 0.63, + "learning_rate": 6.2891390516020566e-06, + "loss": 1.0973, + "step": 26296 + }, + { + "epoch": 0.63, + "learning_rate": 6.288415782719045e-06, + "loss": 0.8478, + "step": 26297 + }, + { + "epoch": 0.63, + "learning_rate": 6.287692536353341e-06, + "loss": 1.114, + "step": 26298 + }, + { + "epoch": 0.63, + "learning_rate": 6.2869693125093325e-06, + "loss": 0.8541, + "step": 26299 + }, + { + "epoch": 0.63, + "learning_rate": 6.28624611119141e-06, + "loss": 0.7014, + "step": 26300 + }, + { + "epoch": 0.63, + "learning_rate": 6.285522932403956e-06, + "loss": 1.0601, + "step": 26301 + }, + { + "epoch": 0.63, + "learning_rate": 6.284799776151363e-06, + "loss": 1.0876, + "step": 26302 + }, + { + "epoch": 0.63, + "learning_rate": 6.2840766424380154e-06, + "loss": 1.0464, + "step": 26303 + }, + { + "epoch": 0.63, + "learning_rate": 6.283353531268301e-06, + "loss": 1.0486, + "step": 26304 + }, + { + "epoch": 0.63, + "learning_rate": 6.2826304426466045e-06, + "loss": 1.038, + "step": 26305 + }, + { + "epoch": 0.63, + "learning_rate": 6.281907376577316e-06, + "loss": 0.8557, + "step": 26306 + }, + { + "epoch": 0.63, + "learning_rate": 6.2811843330648225e-06, + "loss": 0.8464, + "step": 26307 + }, + { + "epoch": 0.63, + "learning_rate": 6.280461312113508e-06, + "loss": 1.0148, + "step": 26308 + }, + { + "epoch": 0.63, + "learning_rate": 6.279738313727759e-06, + "loss": 0.8455, + "step": 26309 + }, + { + "epoch": 0.63, + "learning_rate": 6.279015337911963e-06, + "loss": 1.179, + "step": 26310 + }, + { + "epoch": 0.63, + "learning_rate": 6.278292384670505e-06, + "loss": 1.0456, + "step": 26311 + }, + { + "epoch": 0.63, + "learning_rate": 6.277569454007775e-06, + "loss": 1.1512, + "step": 26312 + }, + { + "epoch": 0.63, + "learning_rate": 6.27684654592815e-06, + "loss": 0.8951, + "step": 26313 + }, + { + "epoch": 0.63, + "learning_rate": 6.276123660436026e-06, + "loss": 1.3182, + "step": 26314 + }, + { + "epoch": 0.63, + "learning_rate": 6.2754007975357826e-06, + "loss": 0.8347, + "step": 26315 + }, + { + "epoch": 0.63, + "learning_rate": 6.2746779572318065e-06, + "loss": 1.2742, + "step": 26316 + }, + { + "epoch": 0.63, + "learning_rate": 6.27395513952848e-06, + "loss": 1.0496, + "step": 26317 + }, + { + "epoch": 0.63, + "learning_rate": 6.273232344430195e-06, + "loss": 0.9666, + "step": 26318 + }, + { + "epoch": 0.63, + "learning_rate": 6.272509571941332e-06, + "loss": 0.9066, + "step": 26319 + }, + { + "epoch": 0.63, + "learning_rate": 6.271786822066277e-06, + "loss": 1.0753, + "step": 26320 + }, + { + "epoch": 0.63, + "learning_rate": 6.2710640948094115e-06, + "loss": 0.9625, + "step": 26321 + }, + { + "epoch": 0.63, + "learning_rate": 6.270341390175126e-06, + "loss": 0.8156, + "step": 26322 + }, + { + "epoch": 0.63, + "learning_rate": 6.269618708167802e-06, + "loss": 1.1791, + "step": 26323 + }, + { + "epoch": 0.63, + "learning_rate": 6.268896048791824e-06, + "loss": 1.2561, + "step": 26324 + }, + { + "epoch": 0.63, + "learning_rate": 6.2681734120515725e-06, + "loss": 0.9759, + "step": 26325 + }, + { + "epoch": 0.63, + "learning_rate": 6.267450797951439e-06, + "loss": 1.1526, + "step": 26326 + }, + { + "epoch": 0.63, + "learning_rate": 6.266728206495803e-06, + "loss": 1.1014, + "step": 26327 + }, + { + "epoch": 0.63, + "learning_rate": 6.2660056376890476e-06, + "loss": 0.793, + "step": 26328 + }, + { + "epoch": 0.63, + "learning_rate": 6.265283091535557e-06, + "loss": 1.1198, + "step": 26329 + }, + { + "epoch": 0.63, + "learning_rate": 6.264560568039717e-06, + "loss": 1.2038, + "step": 26330 + }, + { + "epoch": 0.63, + "learning_rate": 6.26383806720591e-06, + "loss": 1.0918, + "step": 26331 + }, + { + "epoch": 0.63, + "learning_rate": 6.2631155890385145e-06, + "loss": 1.0574, + "step": 26332 + }, + { + "epoch": 0.63, + "learning_rate": 6.262393133541922e-06, + "loss": 1.0524, + "step": 26333 + }, + { + "epoch": 0.63, + "learning_rate": 6.261670700720508e-06, + "loss": 1.0418, + "step": 26334 + }, + { + "epoch": 0.63, + "learning_rate": 6.260948290578661e-06, + "loss": 1.1511, + "step": 26335 + }, + { + "epoch": 0.63, + "learning_rate": 6.260225903120763e-06, + "loss": 0.8758, + "step": 26336 + }, + { + "epoch": 0.63, + "learning_rate": 6.2595035383511924e-06, + "loss": 0.9403, + "step": 26337 + }, + { + "epoch": 0.63, + "learning_rate": 6.258781196274333e-06, + "loss": 0.8864, + "step": 26338 + }, + { + "epoch": 0.63, + "learning_rate": 6.258058876894571e-06, + "loss": 1.2227, + "step": 26339 + }, + { + "epoch": 0.63, + "learning_rate": 6.2573365802162834e-06, + "loss": 1.1198, + "step": 26340 + }, + { + "epoch": 0.63, + "learning_rate": 6.256614306243855e-06, + "loss": 1.2053, + "step": 26341 + }, + { + "epoch": 0.63, + "learning_rate": 6.255892054981667e-06, + "loss": 1.1054, + "step": 26342 + }, + { + "epoch": 0.63, + "learning_rate": 6.255169826434102e-06, + "loss": 1.2023, + "step": 26343 + }, + { + "epoch": 0.63, + "learning_rate": 6.254447620605541e-06, + "loss": 1.0184, + "step": 26344 + }, + { + "epoch": 0.63, + "learning_rate": 6.253725437500364e-06, + "loss": 1.0342, + "step": 26345 + }, + { + "epoch": 0.63, + "learning_rate": 6.253003277122953e-06, + "loss": 1.0434, + "step": 26346 + }, + { + "epoch": 0.63, + "learning_rate": 6.2522811394776916e-06, + "loss": 1.0759, + "step": 26347 + }, + { + "epoch": 0.63, + "learning_rate": 6.251559024568958e-06, + "loss": 1.0641, + "step": 26348 + }, + { + "epoch": 0.63, + "learning_rate": 6.250836932401136e-06, + "loss": 1.0142, + "step": 26349 + }, + { + "epoch": 0.63, + "learning_rate": 6.250114862978601e-06, + "loss": 1.2124, + "step": 26350 + }, + { + "epoch": 0.63, + "learning_rate": 6.249392816305739e-06, + "loss": 1.2287, + "step": 26351 + }, + { + "epoch": 0.63, + "learning_rate": 6.2486707923869305e-06, + "loss": 1.0902, + "step": 26352 + }, + { + "epoch": 0.63, + "learning_rate": 6.247948791226552e-06, + "loss": 1.0494, + "step": 26353 + }, + { + "epoch": 0.63, + "learning_rate": 6.247226812828985e-06, + "loss": 1.1082, + "step": 26354 + }, + { + "epoch": 0.63, + "learning_rate": 6.246504857198611e-06, + "loss": 1.0932, + "step": 26355 + }, + { + "epoch": 0.63, + "learning_rate": 6.245782924339809e-06, + "loss": 1.1924, + "step": 26356 + }, + { + "epoch": 0.63, + "learning_rate": 6.24506101425696e-06, + "loss": 1.2979, + "step": 26357 + }, + { + "epoch": 0.63, + "learning_rate": 6.2443391269544405e-06, + "loss": 1.1117, + "step": 26358 + }, + { + "epoch": 0.63, + "learning_rate": 6.2436172624366344e-06, + "loss": 1.1809, + "step": 26359 + }, + { + "epoch": 0.63, + "learning_rate": 6.242895420707917e-06, + "loss": 1.2405, + "step": 26360 + }, + { + "epoch": 0.63, + "learning_rate": 6.242173601772672e-06, + "loss": 1.1218, + "step": 26361 + }, + { + "epoch": 0.63, + "learning_rate": 6.241451805635272e-06, + "loss": 1.1446, + "step": 26362 + }, + { + "epoch": 0.63, + "learning_rate": 6.240730032300102e-06, + "loss": 0.9815, + "step": 26363 + }, + { + "epoch": 0.63, + "learning_rate": 6.240008281771539e-06, + "loss": 1.2645, + "step": 26364 + }, + { + "epoch": 0.63, + "learning_rate": 6.239286554053962e-06, + "loss": 0.9609, + "step": 26365 + }, + { + "epoch": 0.63, + "learning_rate": 6.238564849151745e-06, + "loss": 1.0178, + "step": 26366 + }, + { + "epoch": 0.63, + "learning_rate": 6.237843167069274e-06, + "loss": 1.268, + "step": 26367 + }, + { + "epoch": 0.63, + "learning_rate": 6.237121507810924e-06, + "loss": 1.1658, + "step": 26368 + }, + { + "epoch": 0.63, + "learning_rate": 6.236399871381071e-06, + "loss": 0.9285, + "step": 26369 + }, + { + "epoch": 0.63, + "learning_rate": 6.235678257784094e-06, + "loss": 0.9215, + "step": 26370 + }, + { + "epoch": 0.63, + "learning_rate": 6.234956667024373e-06, + "loss": 0.7851, + "step": 26371 + }, + { + "epoch": 0.63, + "learning_rate": 6.234235099106284e-06, + "loss": 1.2412, + "step": 26372 + }, + { + "epoch": 0.63, + "learning_rate": 6.233513554034205e-06, + "loss": 1.1801, + "step": 26373 + }, + { + "epoch": 0.63, + "learning_rate": 6.232792031812512e-06, + "loss": 0.9613, + "step": 26374 + }, + { + "epoch": 0.63, + "learning_rate": 6.232070532445585e-06, + "loss": 1.1433, + "step": 26375 + }, + { + "epoch": 0.63, + "learning_rate": 6.231349055937801e-06, + "loss": 1.0284, + "step": 26376 + }, + { + "epoch": 0.63, + "learning_rate": 6.230627602293535e-06, + "loss": 1.1303, + "step": 26377 + }, + { + "epoch": 0.63, + "learning_rate": 6.229906171517162e-06, + "loss": 0.8989, + "step": 26378 + }, + { + "epoch": 0.63, + "learning_rate": 6.229184763613065e-06, + "loss": 0.9411, + "step": 26379 + }, + { + "epoch": 0.63, + "learning_rate": 6.228463378585616e-06, + "loss": 0.9977, + "step": 26380 + }, + { + "epoch": 0.63, + "learning_rate": 6.227742016439191e-06, + "loss": 1.0674, + "step": 26381 + }, + { + "epoch": 0.63, + "learning_rate": 6.227020677178169e-06, + "loss": 1.0735, + "step": 26382 + }, + { + "epoch": 0.63, + "learning_rate": 6.226299360806927e-06, + "loss": 1.2589, + "step": 26383 + }, + { + "epoch": 0.63, + "learning_rate": 6.2255780673298375e-06, + "loss": 1.1694, + "step": 26384 + }, + { + "epoch": 0.63, + "learning_rate": 6.224856796751277e-06, + "loss": 1.1153, + "step": 26385 + }, + { + "epoch": 0.63, + "learning_rate": 6.224135549075623e-06, + "loss": 1.2908, + "step": 26386 + }, + { + "epoch": 0.63, + "learning_rate": 6.2234143243072505e-06, + "loss": 1.2908, + "step": 26387 + }, + { + "epoch": 0.63, + "learning_rate": 6.222693122450537e-06, + "loss": 0.9386, + "step": 26388 + }, + { + "epoch": 0.63, + "learning_rate": 6.221971943509852e-06, + "loss": 1.1748, + "step": 26389 + }, + { + "epoch": 0.63, + "learning_rate": 6.221250787489578e-06, + "loss": 0.9569, + "step": 26390 + }, + { + "epoch": 0.63, + "learning_rate": 6.220529654394085e-06, + "loss": 1.0048, + "step": 26391 + }, + { + "epoch": 0.63, + "learning_rate": 6.219808544227751e-06, + "loss": 1.3179, + "step": 26392 + }, + { + "epoch": 0.63, + "learning_rate": 6.2190874569949475e-06, + "loss": 0.9252, + "step": 26393 + }, + { + "epoch": 0.63, + "learning_rate": 6.218366392700053e-06, + "loss": 0.9505, + "step": 26394 + }, + { + "epoch": 0.63, + "learning_rate": 6.217645351347439e-06, + "loss": 1.0597, + "step": 26395 + }, + { + "epoch": 0.63, + "learning_rate": 6.216924332941481e-06, + "loss": 0.9069, + "step": 26396 + }, + { + "epoch": 0.63, + "learning_rate": 6.216203337486553e-06, + "loss": 1.2225, + "step": 26397 + }, + { + "epoch": 0.63, + "learning_rate": 6.215482364987029e-06, + "loss": 1.0107, + "step": 26398 + }, + { + "epoch": 0.63, + "learning_rate": 6.214761415447286e-06, + "loss": 0.9759, + "step": 26399 + }, + { + "epoch": 0.63, + "learning_rate": 6.214040488871693e-06, + "loss": 1.0769, + "step": 26400 + }, + { + "epoch": 0.63, + "learning_rate": 6.213319585264625e-06, + "loss": 0.7694, + "step": 26401 + }, + { + "epoch": 0.63, + "learning_rate": 6.212598704630457e-06, + "loss": 0.9015, + "step": 26402 + }, + { + "epoch": 0.63, + "learning_rate": 6.211877846973563e-06, + "loss": 1.0486, + "step": 26403 + }, + { + "epoch": 0.63, + "learning_rate": 6.211157012298314e-06, + "loss": 0.8763, + "step": 26404 + }, + { + "epoch": 0.64, + "learning_rate": 6.2104362006090835e-06, + "loss": 1.2316, + "step": 26405 + }, + { + "epoch": 0.64, + "learning_rate": 6.209715411910246e-06, + "loss": 1.0436, + "step": 26406 + }, + { + "epoch": 0.64, + "learning_rate": 6.208994646206174e-06, + "loss": 0.794, + "step": 26407 + }, + { + "epoch": 0.64, + "learning_rate": 6.2082739035012404e-06, + "loss": 0.9014, + "step": 26408 + }, + { + "epoch": 0.64, + "learning_rate": 6.207553183799816e-06, + "loss": 0.9012, + "step": 26409 + }, + { + "epoch": 0.64, + "learning_rate": 6.206832487106276e-06, + "loss": 1.1218, + "step": 26410 + }, + { + "epoch": 0.64, + "learning_rate": 6.206111813424989e-06, + "loss": 1.2369, + "step": 26411 + }, + { + "epoch": 0.64, + "learning_rate": 6.2053911627603305e-06, + "loss": 0.964, + "step": 26412 + }, + { + "epoch": 0.64, + "learning_rate": 6.20467053511667e-06, + "loss": 0.9153, + "step": 26413 + }, + { + "epoch": 0.64, + "learning_rate": 6.203949930498382e-06, + "loss": 1.1139, + "step": 26414 + }, + { + "epoch": 0.64, + "learning_rate": 6.203229348909836e-06, + "loss": 1.0985, + "step": 26415 + }, + { + "epoch": 0.64, + "learning_rate": 6.202508790355406e-06, + "loss": 1.0681, + "step": 26416 + }, + { + "epoch": 0.64, + "learning_rate": 6.201788254839459e-06, + "loss": 1.0287, + "step": 26417 + }, + { + "epoch": 0.64, + "learning_rate": 6.201067742366371e-06, + "loss": 1.0854, + "step": 26418 + }, + { + "epoch": 0.64, + "learning_rate": 6.200347252940513e-06, + "loss": 1.2873, + "step": 26419 + }, + { + "epoch": 0.64, + "learning_rate": 6.199626786566253e-06, + "loss": 1.1206, + "step": 26420 + }, + { + "epoch": 0.64, + "learning_rate": 6.198906343247962e-06, + "loss": 0.8001, + "step": 26421 + }, + { + "epoch": 0.64, + "learning_rate": 6.198185922990014e-06, + "loss": 1.0795, + "step": 26422 + }, + { + "epoch": 0.64, + "learning_rate": 6.197465525796777e-06, + "loss": 0.9882, + "step": 26423 + }, + { + "epoch": 0.64, + "learning_rate": 6.196745151672624e-06, + "loss": 0.994, + "step": 26424 + }, + { + "epoch": 0.64, + "learning_rate": 6.19602480062192e-06, + "loss": 1.2656, + "step": 26425 + }, + { + "epoch": 0.64, + "learning_rate": 6.195304472649042e-06, + "loss": 0.889, + "step": 26426 + }, + { + "epoch": 0.64, + "learning_rate": 6.194584167758356e-06, + "loss": 1.1997, + "step": 26427 + }, + { + "epoch": 0.64, + "learning_rate": 6.193863885954231e-06, + "loss": 1.2045, + "step": 26428 + }, + { + "epoch": 0.64, + "learning_rate": 6.19314362724104e-06, + "loss": 1.1887, + "step": 26429 + }, + { + "epoch": 0.64, + "learning_rate": 6.192423391623152e-06, + "loss": 1.0024, + "step": 26430 + }, + { + "epoch": 0.64, + "learning_rate": 6.191703179104936e-06, + "loss": 1.2882, + "step": 26431 + }, + { + "epoch": 0.64, + "learning_rate": 6.190982989690759e-06, + "loss": 1.1523, + "step": 26432 + }, + { + "epoch": 0.64, + "learning_rate": 6.190262823384994e-06, + "loss": 1.0193, + "step": 26433 + }, + { + "epoch": 0.64, + "learning_rate": 6.189542680192007e-06, + "loss": 0.9983, + "step": 26434 + }, + { + "epoch": 0.64, + "learning_rate": 6.18882256011617e-06, + "loss": 1.1202, + "step": 26435 + }, + { + "epoch": 0.64, + "learning_rate": 6.188102463161847e-06, + "loss": 1.146, + "step": 26436 + }, + { + "epoch": 0.64, + "learning_rate": 6.1873823893334114e-06, + "loss": 1.0605, + "step": 26437 + }, + { + "epoch": 0.64, + "learning_rate": 6.186662338635232e-06, + "loss": 1.2343, + "step": 26438 + }, + { + "epoch": 0.64, + "learning_rate": 6.185942311071674e-06, + "loss": 1.2704, + "step": 26439 + }, + { + "epoch": 0.64, + "learning_rate": 6.185222306647105e-06, + "loss": 1.035, + "step": 26440 + }, + { + "epoch": 0.64, + "learning_rate": 6.184502325365897e-06, + "loss": 0.9913, + "step": 26441 + }, + { + "epoch": 0.64, + "learning_rate": 6.183782367232416e-06, + "loss": 0.9543, + "step": 26442 + }, + { + "epoch": 0.64, + "learning_rate": 6.1830624322510295e-06, + "loss": 1.1515, + "step": 26443 + }, + { + "epoch": 0.64, + "learning_rate": 6.1823425204261035e-06, + "loss": 1.1439, + "step": 26444 + }, + { + "epoch": 0.64, + "learning_rate": 6.18162263176201e-06, + "loss": 1.197, + "step": 26445 + }, + { + "epoch": 0.64, + "learning_rate": 6.180902766263113e-06, + "loss": 1.1909, + "step": 26446 + }, + { + "epoch": 0.64, + "learning_rate": 6.180182923933781e-06, + "loss": 0.9882, + "step": 26447 + }, + { + "epoch": 0.64, + "learning_rate": 6.179463104778378e-06, + "loss": 1.2102, + "step": 26448 + }, + { + "epoch": 0.64, + "learning_rate": 6.178743308801276e-06, + "loss": 1.0394, + "step": 26449 + }, + { + "epoch": 0.64, + "learning_rate": 6.1780235360068404e-06, + "loss": 0.9495, + "step": 26450 + }, + { + "epoch": 0.64, + "learning_rate": 6.177303786399437e-06, + "loss": 1.0872, + "step": 26451 + }, + { + "epoch": 0.64, + "learning_rate": 6.17658405998343e-06, + "loss": 1.0516, + "step": 26452 + }, + { + "epoch": 0.64, + "learning_rate": 6.1758643567631906e-06, + "loss": 1.1148, + "step": 26453 + }, + { + "epoch": 0.64, + "learning_rate": 6.175144676743081e-06, + "loss": 1.1121, + "step": 26454 + }, + { + "epoch": 0.64, + "learning_rate": 6.1744250199274705e-06, + "loss": 1.2288, + "step": 26455 + }, + { + "epoch": 0.64, + "learning_rate": 6.173705386320721e-06, + "loss": 1.1532, + "step": 26456 + }, + { + "epoch": 0.64, + "learning_rate": 6.1729857759272035e-06, + "loss": 1.0595, + "step": 26457 + }, + { + "epoch": 0.64, + "learning_rate": 6.172266188751282e-06, + "loss": 1.0392, + "step": 26458 + }, + { + "epoch": 0.64, + "learning_rate": 6.171546624797319e-06, + "loss": 1.3786, + "step": 26459 + }, + { + "epoch": 0.64, + "learning_rate": 6.170827084069682e-06, + "loss": 1.1956, + "step": 26460 + }, + { + "epoch": 0.64, + "learning_rate": 6.170107566572736e-06, + "loss": 0.9742, + "step": 26461 + }, + { + "epoch": 0.64, + "learning_rate": 6.169388072310849e-06, + "loss": 0.9412, + "step": 26462 + }, + { + "epoch": 0.64, + "learning_rate": 6.168668601288382e-06, + "loss": 1.18, + "step": 26463 + }, + { + "epoch": 0.64, + "learning_rate": 6.1679491535097e-06, + "loss": 0.9362, + "step": 26464 + }, + { + "epoch": 0.64, + "learning_rate": 6.167229728979171e-06, + "loss": 0.9081, + "step": 26465 + }, + { + "epoch": 0.64, + "learning_rate": 6.166510327701159e-06, + "loss": 1.0663, + "step": 26466 + }, + { + "epoch": 0.64, + "learning_rate": 6.165790949680025e-06, + "loss": 1.1378, + "step": 26467 + }, + { + "epoch": 0.64, + "learning_rate": 6.165071594920134e-06, + "loss": 1.0341, + "step": 26468 + }, + { + "epoch": 0.64, + "learning_rate": 6.164352263425854e-06, + "loss": 1.2145, + "step": 26469 + }, + { + "epoch": 0.64, + "learning_rate": 6.163632955201547e-06, + "loss": 0.9918, + "step": 26470 + }, + { + "epoch": 0.64, + "learning_rate": 6.162913670251576e-06, + "loss": 1.3233, + "step": 26471 + }, + { + "epoch": 0.64, + "learning_rate": 6.162194408580304e-06, + "loss": 0.8685, + "step": 26472 + }, + { + "epoch": 0.64, + "learning_rate": 6.1614751701920975e-06, + "loss": 1.1816, + "step": 26473 + }, + { + "epoch": 0.64, + "learning_rate": 6.160755955091318e-06, + "loss": 1.1475, + "step": 26474 + }, + { + "epoch": 0.64, + "learning_rate": 6.160036763282326e-06, + "loss": 0.9784, + "step": 26475 + }, + { + "epoch": 0.64, + "learning_rate": 6.159317594769492e-06, + "loss": 1.0508, + "step": 26476 + }, + { + "epoch": 0.64, + "learning_rate": 6.158598449557174e-06, + "loss": 1.0079, + "step": 26477 + }, + { + "epoch": 0.64, + "learning_rate": 6.157879327649734e-06, + "loss": 0.8261, + "step": 26478 + }, + { + "epoch": 0.64, + "learning_rate": 6.157160229051535e-06, + "loss": 0.9981, + "step": 26479 + }, + { + "epoch": 0.64, + "learning_rate": 6.156441153766944e-06, + "loss": 0.7147, + "step": 26480 + }, + { + "epoch": 0.64, + "learning_rate": 6.155722101800321e-06, + "loss": 0.9695, + "step": 26481 + }, + { + "epoch": 0.64, + "learning_rate": 6.155003073156026e-06, + "loss": 1.1182, + "step": 26482 + }, + { + "epoch": 0.64, + "learning_rate": 6.154284067838422e-06, + "loss": 1.0103, + "step": 26483 + }, + { + "epoch": 0.64, + "learning_rate": 6.153565085851874e-06, + "loss": 1.2639, + "step": 26484 + }, + { + "epoch": 0.64, + "learning_rate": 6.1528461272007425e-06, + "loss": 0.8938, + "step": 26485 + }, + { + "epoch": 0.64, + "learning_rate": 6.152127191889388e-06, + "loss": 1.0265, + "step": 26486 + }, + { + "epoch": 0.64, + "learning_rate": 6.15140827992217e-06, + "loss": 0.7612, + "step": 26487 + }, + { + "epoch": 0.64, + "learning_rate": 6.150689391303455e-06, + "loss": 1.1808, + "step": 26488 + }, + { + "epoch": 0.64, + "learning_rate": 6.149970526037603e-06, + "loss": 0.9429, + "step": 26489 + }, + { + "epoch": 0.64, + "learning_rate": 6.149251684128973e-06, + "loss": 1.2322, + "step": 26490 + }, + { + "epoch": 0.64, + "learning_rate": 6.148532865581926e-06, + "loss": 1.1558, + "step": 26491 + }, + { + "epoch": 0.64, + "learning_rate": 6.147814070400826e-06, + "loss": 0.8535, + "step": 26492 + }, + { + "epoch": 0.64, + "learning_rate": 6.147095298590031e-06, + "loss": 0.851, + "step": 26493 + }, + { + "epoch": 0.64, + "learning_rate": 6.146376550153904e-06, + "loss": 0.7663, + "step": 26494 + }, + { + "epoch": 0.64, + "learning_rate": 6.145657825096802e-06, + "loss": 0.972, + "step": 26495 + }, + { + "epoch": 0.64, + "learning_rate": 6.144939123423088e-06, + "loss": 1.0198, + "step": 26496 + }, + { + "epoch": 0.64, + "learning_rate": 6.1442204451371235e-06, + "loss": 0.901, + "step": 26497 + }, + { + "epoch": 0.64, + "learning_rate": 6.143501790243266e-06, + "loss": 1.0853, + "step": 26498 + }, + { + "epoch": 0.64, + "learning_rate": 6.142783158745871e-06, + "loss": 1.1877, + "step": 26499 + }, + { + "epoch": 0.64, + "learning_rate": 6.142064550649308e-06, + "loss": 1.1027, + "step": 26500 + }, + { + "epoch": 0.64, + "learning_rate": 6.14134596595793e-06, + "loss": 1.1349, + "step": 26501 + }, + { + "epoch": 0.64, + "learning_rate": 6.140627404676101e-06, + "loss": 0.9689, + "step": 26502 + }, + { + "epoch": 0.64, + "learning_rate": 6.139908866808177e-06, + "loss": 1.2872, + "step": 26503 + }, + { + "epoch": 0.64, + "learning_rate": 6.139190352358517e-06, + "loss": 0.7955, + "step": 26504 + }, + { + "epoch": 0.64, + "learning_rate": 6.1384718613314805e-06, + "loss": 1.0487, + "step": 26505 + }, + { + "epoch": 0.64, + "learning_rate": 6.13775339373143e-06, + "loss": 0.9624, + "step": 26506 + }, + { + "epoch": 0.64, + "learning_rate": 6.137034949562719e-06, + "loss": 0.7795, + "step": 26507 + }, + { + "epoch": 0.64, + "learning_rate": 6.136316528829707e-06, + "loss": 1.0743, + "step": 26508 + }, + { + "epoch": 0.64, + "learning_rate": 6.135598131536755e-06, + "loss": 1.0576, + "step": 26509 + }, + { + "epoch": 0.64, + "learning_rate": 6.1348797576882215e-06, + "loss": 1.1186, + "step": 26510 + }, + { + "epoch": 0.64, + "learning_rate": 6.134161407288464e-06, + "loss": 1.1482, + "step": 26511 + }, + { + "epoch": 0.64, + "learning_rate": 6.133443080341836e-06, + "loss": 0.8582, + "step": 26512 + }, + { + "epoch": 0.64, + "learning_rate": 6.132724776852702e-06, + "loss": 1.0392, + "step": 26513 + }, + { + "epoch": 0.64, + "learning_rate": 6.132006496825418e-06, + "loss": 0.9992, + "step": 26514 + }, + { + "epoch": 0.64, + "learning_rate": 6.131288240264339e-06, + "loss": 1.0201, + "step": 26515 + }, + { + "epoch": 0.64, + "learning_rate": 6.130570007173824e-06, + "loss": 1.2962, + "step": 26516 + }, + { + "epoch": 0.64, + "learning_rate": 6.129851797558233e-06, + "loss": 1.1256, + "step": 26517 + }, + { + "epoch": 0.64, + "learning_rate": 6.129133611421919e-06, + "loss": 0.9499, + "step": 26518 + }, + { + "epoch": 0.64, + "learning_rate": 6.128415448769243e-06, + "loss": 1.1944, + "step": 26519 + }, + { + "epoch": 0.64, + "learning_rate": 6.1276973096045555e-06, + "loss": 0.9379, + "step": 26520 + }, + { + "epoch": 0.64, + "learning_rate": 6.126979193932221e-06, + "loss": 0.9383, + "step": 26521 + }, + { + "epoch": 0.64, + "learning_rate": 6.126261101756592e-06, + "loss": 0.8659, + "step": 26522 + }, + { + "epoch": 0.64, + "learning_rate": 6.1255430330820264e-06, + "loss": 0.9186, + "step": 26523 + }, + { + "epoch": 0.64, + "learning_rate": 6.124824987912877e-06, + "loss": 1.0463, + "step": 26524 + }, + { + "epoch": 0.64, + "learning_rate": 6.124106966253507e-06, + "loss": 1.0819, + "step": 26525 + }, + { + "epoch": 0.64, + "learning_rate": 6.1233889681082655e-06, + "loss": 1.269, + "step": 26526 + }, + { + "epoch": 0.64, + "learning_rate": 6.122670993481513e-06, + "loss": 1.212, + "step": 26527 + }, + { + "epoch": 0.64, + "learning_rate": 6.121953042377602e-06, + "loss": 1.2331, + "step": 26528 + }, + { + "epoch": 0.64, + "learning_rate": 6.121235114800891e-06, + "loss": 1.121, + "step": 26529 + }, + { + "epoch": 0.64, + "learning_rate": 6.120517210755734e-06, + "loss": 0.997, + "step": 26530 + }, + { + "epoch": 0.64, + "learning_rate": 6.119799330246488e-06, + "loss": 1.1436, + "step": 26531 + }, + { + "epoch": 0.64, + "learning_rate": 6.119081473277502e-06, + "loss": 1.1686, + "step": 26532 + }, + { + "epoch": 0.64, + "learning_rate": 6.11836363985314e-06, + "loss": 1.1364, + "step": 26533 + }, + { + "epoch": 0.64, + "learning_rate": 6.117645829977752e-06, + "loss": 0.9872, + "step": 26534 + }, + { + "epoch": 0.64, + "learning_rate": 6.116928043655694e-06, + "loss": 1.076, + "step": 26535 + }, + { + "epoch": 0.64, + "learning_rate": 6.116210280891317e-06, + "loss": 1.2549, + "step": 26536 + }, + { + "epoch": 0.64, + "learning_rate": 6.1154925416889835e-06, + "loss": 0.9467, + "step": 26537 + }, + { + "epoch": 0.64, + "learning_rate": 6.114774826053041e-06, + "loss": 0.7345, + "step": 26538 + }, + { + "epoch": 0.64, + "learning_rate": 6.114057133987845e-06, + "loss": 1.0369, + "step": 26539 + }, + { + "epoch": 0.64, + "learning_rate": 6.1133394654977495e-06, + "loss": 1.0004, + "step": 26540 + }, + { + "epoch": 0.64, + "learning_rate": 6.112621820587111e-06, + "loss": 1.2281, + "step": 26541 + }, + { + "epoch": 0.64, + "learning_rate": 6.111904199260283e-06, + "loss": 1.0124, + "step": 26542 + }, + { + "epoch": 0.64, + "learning_rate": 6.111186601521615e-06, + "loss": 1.1868, + "step": 26543 + }, + { + "epoch": 0.64, + "learning_rate": 6.110469027375463e-06, + "loss": 0.8682, + "step": 26544 + }, + { + "epoch": 0.64, + "learning_rate": 6.1097514768261825e-06, + "loss": 1.1024, + "step": 26545 + }, + { + "epoch": 0.64, + "learning_rate": 6.109033949878124e-06, + "loss": 1.2544, + "step": 26546 + }, + { + "epoch": 0.64, + "learning_rate": 6.108316446535642e-06, + "loss": 0.551, + "step": 26547 + }, + { + "epoch": 0.64, + "learning_rate": 6.107598966803085e-06, + "loss": 0.7671, + "step": 26548 + }, + { + "epoch": 0.64, + "learning_rate": 6.106881510684814e-06, + "loss": 1.1175, + "step": 26549 + }, + { + "epoch": 0.64, + "learning_rate": 6.106164078185176e-06, + "loss": 0.9408, + "step": 26550 + }, + { + "epoch": 0.64, + "learning_rate": 6.105446669308523e-06, + "loss": 1.0545, + "step": 26551 + }, + { + "epoch": 0.64, + "learning_rate": 6.10472928405921e-06, + "loss": 1.0503, + "step": 26552 + }, + { + "epoch": 0.64, + "learning_rate": 6.10401192244159e-06, + "loss": 0.9553, + "step": 26553 + }, + { + "epoch": 0.64, + "learning_rate": 6.103294584460011e-06, + "loss": 0.9977, + "step": 26554 + }, + { + "epoch": 0.64, + "learning_rate": 6.102577270118827e-06, + "loss": 1.0748, + "step": 26555 + }, + { + "epoch": 0.64, + "learning_rate": 6.101859979422391e-06, + "loss": 1.0867, + "step": 26556 + }, + { + "epoch": 0.64, + "learning_rate": 6.101142712375054e-06, + "loss": 1.1163, + "step": 26557 + }, + { + "epoch": 0.64, + "learning_rate": 6.100425468981168e-06, + "loss": 0.8637, + "step": 26558 + }, + { + "epoch": 0.64, + "learning_rate": 6.09970824924508e-06, + "loss": 0.9068, + "step": 26559 + }, + { + "epoch": 0.64, + "learning_rate": 6.098991053171147e-06, + "loss": 1.2922, + "step": 26560 + }, + { + "epoch": 0.64, + "learning_rate": 6.098273880763719e-06, + "loss": 1.0809, + "step": 26561 + }, + { + "epoch": 0.64, + "learning_rate": 6.097556732027143e-06, + "loss": 0.7889, + "step": 26562 + }, + { + "epoch": 0.64, + "learning_rate": 6.096839606965772e-06, + "loss": 0.9815, + "step": 26563 + }, + { + "epoch": 0.64, + "learning_rate": 6.096122505583959e-06, + "loss": 1.1218, + "step": 26564 + }, + { + "epoch": 0.64, + "learning_rate": 6.095405427886053e-06, + "loss": 1.0995, + "step": 26565 + }, + { + "epoch": 0.64, + "learning_rate": 6.0946883738764025e-06, + "loss": 1.0232, + "step": 26566 + }, + { + "epoch": 0.64, + "learning_rate": 6.093971343559358e-06, + "loss": 1.1693, + "step": 26567 + }, + { + "epoch": 0.64, + "learning_rate": 6.093254336939272e-06, + "loss": 1.0708, + "step": 26568 + }, + { + "epoch": 0.64, + "learning_rate": 6.092537354020493e-06, + "loss": 1.1618, + "step": 26569 + }, + { + "epoch": 0.64, + "learning_rate": 6.091820394807373e-06, + "loss": 1.0219, + "step": 26570 + }, + { + "epoch": 0.64, + "learning_rate": 6.091103459304255e-06, + "loss": 0.9834, + "step": 26571 + }, + { + "epoch": 0.64, + "learning_rate": 6.0903865475154955e-06, + "loss": 1.1179, + "step": 26572 + }, + { + "epoch": 0.64, + "learning_rate": 6.089669659445441e-06, + "loss": 0.9117, + "step": 26573 + }, + { + "epoch": 0.64, + "learning_rate": 6.088952795098442e-06, + "loss": 0.9445, + "step": 26574 + }, + { + "epoch": 0.64, + "learning_rate": 6.088235954478843e-06, + "loss": 0.9801, + "step": 26575 + }, + { + "epoch": 0.64, + "learning_rate": 6.0875191375909995e-06, + "loss": 1.0668, + "step": 26576 + }, + { + "epoch": 0.64, + "learning_rate": 6.086802344439257e-06, + "loss": 1.1169, + "step": 26577 + }, + { + "epoch": 0.64, + "learning_rate": 6.086085575027962e-06, + "loss": 1.0869, + "step": 26578 + }, + { + "epoch": 0.64, + "learning_rate": 6.085368829361467e-06, + "loss": 1.0387, + "step": 26579 + }, + { + "epoch": 0.64, + "learning_rate": 6.0846521074441176e-06, + "loss": 1.0407, + "step": 26580 + }, + { + "epoch": 0.64, + "learning_rate": 6.083935409280264e-06, + "loss": 1.0898, + "step": 26581 + }, + { + "epoch": 0.64, + "learning_rate": 6.083218734874253e-06, + "loss": 0.9326, + "step": 26582 + }, + { + "epoch": 0.64, + "learning_rate": 6.08250208423043e-06, + "loss": 1.2977, + "step": 26583 + }, + { + "epoch": 0.64, + "learning_rate": 6.081785457353148e-06, + "loss": 0.9831, + "step": 26584 + }, + { + "epoch": 0.64, + "learning_rate": 6.081068854246753e-06, + "loss": 0.8723, + "step": 26585 + }, + { + "epoch": 0.64, + "learning_rate": 6.080352274915591e-06, + "loss": 1.0873, + "step": 26586 + }, + { + "epoch": 0.64, + "learning_rate": 6.079635719364008e-06, + "loss": 1.0128, + "step": 26587 + }, + { + "epoch": 0.64, + "learning_rate": 6.0789191875963525e-06, + "loss": 0.9647, + "step": 26588 + }, + { + "epoch": 0.64, + "learning_rate": 6.078202679616974e-06, + "loss": 1.1179, + "step": 26589 + }, + { + "epoch": 0.64, + "learning_rate": 6.077486195430218e-06, + "loss": 1.0373, + "step": 26590 + }, + { + "epoch": 0.64, + "learning_rate": 6.076769735040427e-06, + "loss": 1.1431, + "step": 26591 + }, + { + "epoch": 0.64, + "learning_rate": 6.076053298451955e-06, + "loss": 1.1388, + "step": 26592 + }, + { + "epoch": 0.64, + "learning_rate": 6.075336885669143e-06, + "loss": 1.1754, + "step": 26593 + }, + { + "epoch": 0.64, + "learning_rate": 6.07462049669634e-06, + "loss": 1.0056, + "step": 26594 + }, + { + "epoch": 0.64, + "learning_rate": 6.073904131537889e-06, + "loss": 0.9614, + "step": 26595 + }, + { + "epoch": 0.64, + "learning_rate": 6.073187790198141e-06, + "loss": 0.9924, + "step": 26596 + }, + { + "epoch": 0.64, + "learning_rate": 6.072471472681437e-06, + "loss": 1.0742, + "step": 26597 + }, + { + "epoch": 0.64, + "learning_rate": 6.071755178992124e-06, + "loss": 1.1412, + "step": 26598 + }, + { + "epoch": 0.64, + "learning_rate": 6.07103890913455e-06, + "loss": 1.0196, + "step": 26599 + }, + { + "epoch": 0.64, + "learning_rate": 6.070322663113059e-06, + "loss": 1.18, + "step": 26600 + }, + { + "epoch": 0.64, + "learning_rate": 6.069606440931996e-06, + "loss": 1.0157, + "step": 26601 + }, + { + "epoch": 0.64, + "learning_rate": 6.068890242595704e-06, + "loss": 1.2203, + "step": 26602 + }, + { + "epoch": 0.64, + "learning_rate": 6.068174068108532e-06, + "loss": 0.8527, + "step": 26603 + }, + { + "epoch": 0.64, + "learning_rate": 6.0674579174748234e-06, + "loss": 1.2644, + "step": 26604 + }, + { + "epoch": 0.64, + "learning_rate": 6.066741790698924e-06, + "loss": 0.9937, + "step": 26605 + }, + { + "epoch": 0.64, + "learning_rate": 6.066025687785173e-06, + "loss": 1.2214, + "step": 26606 + }, + { + "epoch": 0.64, + "learning_rate": 6.065309608737923e-06, + "loss": 1.3904, + "step": 26607 + }, + { + "epoch": 0.64, + "learning_rate": 6.064593553561512e-06, + "loss": 0.9232, + "step": 26608 + }, + { + "epoch": 0.64, + "learning_rate": 6.0638775222602865e-06, + "loss": 1.0857, + "step": 26609 + }, + { + "epoch": 0.64, + "learning_rate": 6.0631615148385895e-06, + "loss": 1.1535, + "step": 26610 + }, + { + "epoch": 0.64, + "learning_rate": 6.062445531300767e-06, + "loss": 1.1454, + "step": 26611 + }, + { + "epoch": 0.64, + "learning_rate": 6.0617295716511605e-06, + "loss": 0.8287, + "step": 26612 + }, + { + "epoch": 0.64, + "learning_rate": 6.061013635894115e-06, + "loss": 1.0849, + "step": 26613 + }, + { + "epoch": 0.64, + "learning_rate": 6.060297724033972e-06, + "loss": 0.9342, + "step": 26614 + }, + { + "epoch": 0.64, + "learning_rate": 6.059581836075077e-06, + "loss": 1.2778, + "step": 26615 + }, + { + "epoch": 0.64, + "learning_rate": 6.0588659720217725e-06, + "loss": 0.6686, + "step": 26616 + }, + { + "epoch": 0.64, + "learning_rate": 6.058150131878401e-06, + "loss": 1.1115, + "step": 26617 + }, + { + "epoch": 0.64, + "learning_rate": 6.057434315649304e-06, + "loss": 1.0951, + "step": 26618 + }, + { + "epoch": 0.64, + "learning_rate": 6.056718523338827e-06, + "loss": 1.2649, + "step": 26619 + }, + { + "epoch": 0.64, + "learning_rate": 6.056002754951311e-06, + "loss": 0.9148, + "step": 26620 + }, + { + "epoch": 0.64, + "learning_rate": 6.0552870104911e-06, + "loss": 0.9816, + "step": 26621 + }, + { + "epoch": 0.64, + "learning_rate": 6.054571289962531e-06, + "loss": 1.04, + "step": 26622 + }, + { + "epoch": 0.64, + "learning_rate": 6.0538555933699526e-06, + "loss": 1.1511, + "step": 26623 + }, + { + "epoch": 0.64, + "learning_rate": 6.053139920717704e-06, + "loss": 0.8847, + "step": 26624 + }, + { + "epoch": 0.64, + "learning_rate": 6.052424272010129e-06, + "loss": 1.1347, + "step": 26625 + }, + { + "epoch": 0.64, + "learning_rate": 6.051708647251563e-06, + "loss": 1.2009, + "step": 26626 + }, + { + "epoch": 0.64, + "learning_rate": 6.050993046446355e-06, + "loss": 1.2293, + "step": 26627 + }, + { + "epoch": 0.64, + "learning_rate": 6.050277469598843e-06, + "loss": 1.1454, + "step": 26628 + }, + { + "epoch": 0.64, + "learning_rate": 6.049561916713369e-06, + "loss": 1.0005, + "step": 26629 + }, + { + "epoch": 0.64, + "learning_rate": 6.0488463877942715e-06, + "loss": 0.709, + "step": 26630 + }, + { + "epoch": 0.64, + "learning_rate": 6.048130882845895e-06, + "loss": 1.2123, + "step": 26631 + }, + { + "epoch": 0.64, + "learning_rate": 6.04741540187258e-06, + "loss": 0.9367, + "step": 26632 + }, + { + "epoch": 0.64, + "learning_rate": 6.046699944878664e-06, + "loss": 1.198, + "step": 26633 + }, + { + "epoch": 0.64, + "learning_rate": 6.0459845118684905e-06, + "loss": 1.0689, + "step": 26634 + }, + { + "epoch": 0.64, + "learning_rate": 6.045269102846398e-06, + "loss": 0.9925, + "step": 26635 + }, + { + "epoch": 0.64, + "learning_rate": 6.04455371781673e-06, + "loss": 1.0939, + "step": 26636 + }, + { + "epoch": 0.64, + "learning_rate": 6.043838356783823e-06, + "loss": 1.1544, + "step": 26637 + }, + { + "epoch": 0.64, + "learning_rate": 6.043123019752017e-06, + "loss": 1.0542, + "step": 26638 + }, + { + "epoch": 0.64, + "learning_rate": 6.0424077067256546e-06, + "loss": 0.992, + "step": 26639 + }, + { + "epoch": 0.64, + "learning_rate": 6.041692417709073e-06, + "loss": 0.6823, + "step": 26640 + }, + { + "epoch": 0.64, + "learning_rate": 6.040977152706613e-06, + "loss": 0.9843, + "step": 26641 + }, + { + "epoch": 0.64, + "learning_rate": 6.0402619117226114e-06, + "loss": 1.1282, + "step": 26642 + }, + { + "epoch": 0.64, + "learning_rate": 6.039546694761411e-06, + "loss": 0.8431, + "step": 26643 + }, + { + "epoch": 0.64, + "learning_rate": 6.038831501827349e-06, + "loss": 1.1479, + "step": 26644 + }, + { + "epoch": 0.64, + "learning_rate": 6.038116332924764e-06, + "loss": 0.915, + "step": 26645 + }, + { + "epoch": 0.64, + "learning_rate": 6.037401188057996e-06, + "loss": 1.0316, + "step": 26646 + }, + { + "epoch": 0.64, + "learning_rate": 6.036686067231383e-06, + "loss": 1.0813, + "step": 26647 + }, + { + "epoch": 0.64, + "learning_rate": 6.035970970449265e-06, + "loss": 1.036, + "step": 26648 + }, + { + "epoch": 0.64, + "learning_rate": 6.035255897715975e-06, + "loss": 1.2635, + "step": 26649 + }, + { + "epoch": 0.64, + "learning_rate": 6.034540849035857e-06, + "loss": 1.0038, + "step": 26650 + }, + { + "epoch": 0.64, + "learning_rate": 6.033825824413249e-06, + "loss": 1.166, + "step": 26651 + }, + { + "epoch": 0.64, + "learning_rate": 6.0331108238524835e-06, + "loss": 1.1748, + "step": 26652 + }, + { + "epoch": 0.64, + "learning_rate": 6.032395847357902e-06, + "loss": 1.1954, + "step": 26653 + }, + { + "epoch": 0.64, + "learning_rate": 6.0316808949338425e-06, + "loss": 1.0481, + "step": 26654 + }, + { + "epoch": 0.64, + "learning_rate": 6.030965966584643e-06, + "loss": 1.0786, + "step": 26655 + }, + { + "epoch": 0.64, + "learning_rate": 6.030251062314638e-06, + "loss": 1.0569, + "step": 26656 + }, + { + "epoch": 0.64, + "learning_rate": 6.029536182128164e-06, + "loss": 1.1291, + "step": 26657 + }, + { + "epoch": 0.64, + "learning_rate": 6.028821326029563e-06, + "loss": 1.1258, + "step": 26658 + }, + { + "epoch": 0.64, + "learning_rate": 6.028106494023168e-06, + "loss": 1.0363, + "step": 26659 + }, + { + "epoch": 0.64, + "learning_rate": 6.0273916861133176e-06, + "loss": 1.344, + "step": 26660 + }, + { + "epoch": 0.64, + "learning_rate": 6.0266769023043446e-06, + "loss": 1.0526, + "step": 26661 + }, + { + "epoch": 0.64, + "learning_rate": 6.025962142600591e-06, + "loss": 1.2015, + "step": 26662 + }, + { + "epoch": 0.64, + "learning_rate": 6.025247407006389e-06, + "loss": 1.0316, + "step": 26663 + }, + { + "epoch": 0.64, + "learning_rate": 6.024532695526077e-06, + "loss": 0.9249, + "step": 26664 + }, + { + "epoch": 0.64, + "learning_rate": 6.023818008163989e-06, + "loss": 0.8759, + "step": 26665 + }, + { + "epoch": 0.64, + "learning_rate": 6.023103344924461e-06, + "loss": 1.0864, + "step": 26666 + }, + { + "epoch": 0.64, + "learning_rate": 6.02238870581183e-06, + "loss": 1.0694, + "step": 26667 + }, + { + "epoch": 0.64, + "learning_rate": 6.021674090830431e-06, + "loss": 1.3284, + "step": 26668 + }, + { + "epoch": 0.64, + "learning_rate": 6.020959499984602e-06, + "loss": 1.1889, + "step": 26669 + }, + { + "epoch": 0.64, + "learning_rate": 6.020244933278676e-06, + "loss": 1.0307, + "step": 26670 + }, + { + "epoch": 0.64, + "learning_rate": 6.019530390716985e-06, + "loss": 1.1596, + "step": 26671 + }, + { + "epoch": 0.64, + "learning_rate": 6.018815872303868e-06, + "loss": 1.2227, + "step": 26672 + }, + { + "epoch": 0.64, + "learning_rate": 6.018101378043659e-06, + "loss": 1.0812, + "step": 26673 + }, + { + "epoch": 0.64, + "learning_rate": 6.017386907940693e-06, + "loss": 1.1246, + "step": 26674 + }, + { + "epoch": 0.64, + "learning_rate": 6.016672461999302e-06, + "loss": 0.9587, + "step": 26675 + }, + { + "epoch": 0.64, + "learning_rate": 6.015958040223824e-06, + "loss": 1.2056, + "step": 26676 + }, + { + "epoch": 0.64, + "learning_rate": 6.015243642618591e-06, + "loss": 0.868, + "step": 26677 + }, + { + "epoch": 0.64, + "learning_rate": 6.014529269187936e-06, + "loss": 1.078, + "step": 26678 + }, + { + "epoch": 0.64, + "learning_rate": 6.013814919936196e-06, + "loss": 0.8348, + "step": 26679 + }, + { + "epoch": 0.64, + "learning_rate": 6.013100594867704e-06, + "loss": 1.0768, + "step": 26680 + }, + { + "epoch": 0.64, + "learning_rate": 6.012386293986791e-06, + "loss": 1.1092, + "step": 26681 + }, + { + "epoch": 0.64, + "learning_rate": 6.011672017297794e-06, + "loss": 1.1946, + "step": 26682 + }, + { + "epoch": 0.64, + "learning_rate": 6.010957764805043e-06, + "loss": 1.1155, + "step": 26683 + }, + { + "epoch": 0.64, + "learning_rate": 6.010243536512877e-06, + "loss": 1.0966, + "step": 26684 + }, + { + "epoch": 0.64, + "learning_rate": 6.009529332425621e-06, + "loss": 0.9177, + "step": 26685 + }, + { + "epoch": 0.64, + "learning_rate": 6.008815152547611e-06, + "loss": 1.057, + "step": 26686 + }, + { + "epoch": 0.64, + "learning_rate": 6.008100996883184e-06, + "loss": 1.065, + "step": 26687 + }, + { + "epoch": 0.64, + "learning_rate": 6.007386865436669e-06, + "loss": 0.9926, + "step": 26688 + }, + { + "epoch": 0.64, + "learning_rate": 6.006672758212397e-06, + "loss": 0.9682, + "step": 26689 + }, + { + "epoch": 0.64, + "learning_rate": 6.005958675214702e-06, + "loss": 1.0297, + "step": 26690 + }, + { + "epoch": 0.64, + "learning_rate": 6.005244616447918e-06, + "loss": 1.1187, + "step": 26691 + }, + { + "epoch": 0.64, + "learning_rate": 6.004530581916375e-06, + "loss": 1.0982, + "step": 26692 + }, + { + "epoch": 0.64, + "learning_rate": 6.003816571624406e-06, + "loss": 1.1909, + "step": 26693 + }, + { + "epoch": 0.64, + "learning_rate": 6.0031025855763396e-06, + "loss": 1.1491, + "step": 26694 + }, + { + "epoch": 0.64, + "learning_rate": 6.002388623776509e-06, + "loss": 1.1602, + "step": 26695 + }, + { + "epoch": 0.64, + "learning_rate": 6.00167468622925e-06, + "loss": 1.2434, + "step": 26696 + }, + { + "epoch": 0.64, + "learning_rate": 6.000960772938889e-06, + "loss": 1.1187, + "step": 26697 + }, + { + "epoch": 0.64, + "learning_rate": 6.0002468839097546e-06, + "loss": 1.1197, + "step": 26698 + }, + { + "epoch": 0.64, + "learning_rate": 5.999533019146186e-06, + "loss": 0.8843, + "step": 26699 + }, + { + "epoch": 0.64, + "learning_rate": 5.998819178652508e-06, + "loss": 1.1065, + "step": 26700 + }, + { + "epoch": 0.64, + "learning_rate": 5.998105362433053e-06, + "loss": 1.0082, + "step": 26701 + }, + { + "epoch": 0.64, + "learning_rate": 5.997391570492151e-06, + "loss": 0.8499, + "step": 26702 + }, + { + "epoch": 0.64, + "learning_rate": 5.996677802834135e-06, + "loss": 0.8103, + "step": 26703 + }, + { + "epoch": 0.64, + "learning_rate": 5.9959640594633304e-06, + "loss": 1.1859, + "step": 26704 + }, + { + "epoch": 0.64, + "learning_rate": 5.995250340384071e-06, + "loss": 1.1074, + "step": 26705 + }, + { + "epoch": 0.64, + "learning_rate": 5.994536645600686e-06, + "loss": 0.8736, + "step": 26706 + }, + { + "epoch": 0.64, + "learning_rate": 5.993822975117505e-06, + "loss": 0.7797, + "step": 26707 + }, + { + "epoch": 0.64, + "learning_rate": 5.9931093289388575e-06, + "loss": 1.099, + "step": 26708 + }, + { + "epoch": 0.64, + "learning_rate": 5.992395707069074e-06, + "loss": 1.1342, + "step": 26709 + }, + { + "epoch": 0.64, + "learning_rate": 5.991682109512481e-06, + "loss": 0.9687, + "step": 26710 + }, + { + "epoch": 0.64, + "learning_rate": 5.990968536273411e-06, + "loss": 1.0034, + "step": 26711 + }, + { + "epoch": 0.64, + "learning_rate": 5.990254987356195e-06, + "loss": 1.2534, + "step": 26712 + }, + { + "epoch": 0.64, + "learning_rate": 5.989541462765156e-06, + "loss": 1.0283, + "step": 26713 + }, + { + "epoch": 0.64, + "learning_rate": 5.988827962504624e-06, + "loss": 0.8706, + "step": 26714 + }, + { + "epoch": 0.64, + "learning_rate": 5.988114486578931e-06, + "loss": 0.9908, + "step": 26715 + }, + { + "epoch": 0.64, + "learning_rate": 5.987401034992404e-06, + "loss": 1.0167, + "step": 26716 + }, + { + "epoch": 0.64, + "learning_rate": 5.986687607749372e-06, + "loss": 1.0328, + "step": 26717 + }, + { + "epoch": 0.64, + "learning_rate": 5.985974204854159e-06, + "loss": 1.1154, + "step": 26718 + }, + { + "epoch": 0.64, + "learning_rate": 5.985260826311099e-06, + "loss": 1.0419, + "step": 26719 + }, + { + "epoch": 0.64, + "learning_rate": 5.9845474721245175e-06, + "loss": 1.0186, + "step": 26720 + }, + { + "epoch": 0.64, + "learning_rate": 5.9838341422987415e-06, + "loss": 0.8031, + "step": 26721 + }, + { + "epoch": 0.64, + "learning_rate": 5.983120836838097e-06, + "loss": 1.2422, + "step": 26722 + }, + { + "epoch": 0.64, + "learning_rate": 5.982407555746916e-06, + "loss": 1.0076, + "step": 26723 + }, + { + "epoch": 0.64, + "learning_rate": 5.981694299029524e-06, + "loss": 1.2533, + "step": 26724 + }, + { + "epoch": 0.64, + "learning_rate": 5.9809810666902435e-06, + "loss": 1.3583, + "step": 26725 + }, + { + "epoch": 0.64, + "learning_rate": 5.98026785873341e-06, + "loss": 0.938, + "step": 26726 + }, + { + "epoch": 0.64, + "learning_rate": 5.979554675163344e-06, + "loss": 1.0648, + "step": 26727 + }, + { + "epoch": 0.64, + "learning_rate": 5.978841515984375e-06, + "loss": 0.7486, + "step": 26728 + }, + { + "epoch": 0.64, + "learning_rate": 5.978128381200826e-06, + "loss": 0.9972, + "step": 26729 + }, + { + "epoch": 0.64, + "learning_rate": 5.977415270817029e-06, + "loss": 1.0911, + "step": 26730 + }, + { + "epoch": 0.64, + "learning_rate": 5.976702184837307e-06, + "loss": 0.8756, + "step": 26731 + }, + { + "epoch": 0.64, + "learning_rate": 5.975989123265986e-06, + "loss": 1.3159, + "step": 26732 + }, + { + "epoch": 0.64, + "learning_rate": 5.975276086107391e-06, + "loss": 0.9543, + "step": 26733 + }, + { + "epoch": 0.64, + "learning_rate": 5.974563073365851e-06, + "loss": 1.083, + "step": 26734 + }, + { + "epoch": 0.64, + "learning_rate": 5.973850085045691e-06, + "loss": 1.0917, + "step": 26735 + }, + { + "epoch": 0.64, + "learning_rate": 5.973137121151234e-06, + "loss": 1.0206, + "step": 26736 + }, + { + "epoch": 0.64, + "learning_rate": 5.972424181686807e-06, + "loss": 1.1225, + "step": 26737 + }, + { + "epoch": 0.64, + "learning_rate": 5.971711266656736e-06, + "loss": 1.1366, + "step": 26738 + }, + { + "epoch": 0.64, + "learning_rate": 5.9709983760653445e-06, + "loss": 1.2186, + "step": 26739 + }, + { + "epoch": 0.64, + "learning_rate": 5.970285509916959e-06, + "loss": 1.0271, + "step": 26740 + }, + { + "epoch": 0.64, + "learning_rate": 5.969572668215902e-06, + "loss": 0.9439, + "step": 26741 + }, + { + "epoch": 0.64, + "learning_rate": 5.9688598509665e-06, + "loss": 1.1795, + "step": 26742 + }, + { + "epoch": 0.64, + "learning_rate": 5.9681470581730775e-06, + "loss": 0.7712, + "step": 26743 + }, + { + "epoch": 0.64, + "learning_rate": 5.967434289839959e-06, + "loss": 1.1364, + "step": 26744 + }, + { + "epoch": 0.64, + "learning_rate": 5.966721545971466e-06, + "loss": 0.916, + "step": 26745 + }, + { + "epoch": 0.64, + "learning_rate": 5.966008826571927e-06, + "loss": 1.0832, + "step": 26746 + }, + { + "epoch": 0.64, + "learning_rate": 5.9652961316456635e-06, + "loss": 0.9411, + "step": 26747 + }, + { + "epoch": 0.64, + "learning_rate": 5.964583461196999e-06, + "loss": 0.8982, + "step": 26748 + }, + { + "epoch": 0.64, + "learning_rate": 5.963870815230255e-06, + "loss": 1.0065, + "step": 26749 + }, + { + "epoch": 0.64, + "learning_rate": 5.963158193749761e-06, + "loss": 0.8094, + "step": 26750 + }, + { + "epoch": 0.64, + "learning_rate": 5.962445596759836e-06, + "loss": 1.0509, + "step": 26751 + }, + { + "epoch": 0.64, + "learning_rate": 5.961733024264803e-06, + "loss": 1.2231, + "step": 26752 + }, + { + "epoch": 0.64, + "learning_rate": 5.961020476268984e-06, + "loss": 1.2817, + "step": 26753 + }, + { + "epoch": 0.64, + "learning_rate": 5.960307952776707e-06, + "loss": 0.8944, + "step": 26754 + }, + { + "epoch": 0.64, + "learning_rate": 5.959595453792291e-06, + "loss": 1.0794, + "step": 26755 + }, + { + "epoch": 0.64, + "learning_rate": 5.958882979320057e-06, + "loss": 0.9797, + "step": 26756 + }, + { + "epoch": 0.64, + "learning_rate": 5.95817052936433e-06, + "loss": 0.993, + "step": 26757 + }, + { + "epoch": 0.64, + "learning_rate": 5.9574581039294335e-06, + "loss": 1.1141, + "step": 26758 + }, + { + "epoch": 0.64, + "learning_rate": 5.956745703019686e-06, + "loss": 0.9912, + "step": 26759 + }, + { + "epoch": 0.64, + "learning_rate": 5.9560333266394135e-06, + "loss": 1.1624, + "step": 26760 + }, + { + "epoch": 0.64, + "learning_rate": 5.955320974792933e-06, + "loss": 1.0691, + "step": 26761 + }, + { + "epoch": 0.64, + "learning_rate": 5.95460864748457e-06, + "loss": 0.9137, + "step": 26762 + }, + { + "epoch": 0.64, + "learning_rate": 5.953896344718645e-06, + "loss": 0.8729, + "step": 26763 + }, + { + "epoch": 0.64, + "learning_rate": 5.95318406649948e-06, + "loss": 1.0762, + "step": 26764 + }, + { + "epoch": 0.64, + "learning_rate": 5.952471812831392e-06, + "loss": 1.0864, + "step": 26765 + }, + { + "epoch": 0.64, + "learning_rate": 5.951759583718708e-06, + "loss": 0.9943, + "step": 26766 + }, + { + "epoch": 0.64, + "learning_rate": 5.951047379165746e-06, + "loss": 1.2262, + "step": 26767 + }, + { + "epoch": 0.64, + "learning_rate": 5.950335199176829e-06, + "loss": 1.0315, + "step": 26768 + }, + { + "epoch": 0.64, + "learning_rate": 5.949623043756272e-06, + "loss": 1.0238, + "step": 26769 + }, + { + "epoch": 0.64, + "learning_rate": 5.948910912908402e-06, + "loss": 1.1605, + "step": 26770 + }, + { + "epoch": 0.64, + "learning_rate": 5.9481988066375365e-06, + "loss": 0.8546, + "step": 26771 + }, + { + "epoch": 0.64, + "learning_rate": 5.947486724947994e-06, + "loss": 0.7228, + "step": 26772 + }, + { + "epoch": 0.64, + "learning_rate": 5.946774667844097e-06, + "loss": 1.0796, + "step": 26773 + }, + { + "epoch": 0.64, + "learning_rate": 5.9460626353301665e-06, + "loss": 0.9935, + "step": 26774 + }, + { + "epoch": 0.64, + "learning_rate": 5.94535062741052e-06, + "loss": 1.0381, + "step": 26775 + }, + { + "epoch": 0.64, + "learning_rate": 5.944638644089475e-06, + "loss": 0.8144, + "step": 26776 + }, + { + "epoch": 0.64, + "learning_rate": 5.943926685371356e-06, + "loss": 0.9695, + "step": 26777 + }, + { + "epoch": 0.64, + "learning_rate": 5.9432147512604795e-06, + "loss": 1.0288, + "step": 26778 + }, + { + "epoch": 0.64, + "learning_rate": 5.942502841761165e-06, + "loss": 1.1792, + "step": 26779 + }, + { + "epoch": 0.64, + "learning_rate": 5.941790956877729e-06, + "loss": 1.0345, + "step": 26780 + }, + { + "epoch": 0.64, + "learning_rate": 5.941079096614495e-06, + "loss": 1.0093, + "step": 26781 + }, + { + "epoch": 0.64, + "learning_rate": 5.94036726097578e-06, + "loss": 1.1979, + "step": 26782 + }, + { + "epoch": 0.64, + "learning_rate": 5.939655449965902e-06, + "loss": 0.9954, + "step": 26783 + }, + { + "epoch": 0.64, + "learning_rate": 5.938943663589176e-06, + "loss": 1.1116, + "step": 26784 + }, + { + "epoch": 0.64, + "learning_rate": 5.938231901849927e-06, + "loss": 0.9555, + "step": 26785 + }, + { + "epoch": 0.64, + "learning_rate": 5.937520164752469e-06, + "loss": 1.0313, + "step": 26786 + }, + { + "epoch": 0.64, + "learning_rate": 5.93680845230112e-06, + "loss": 1.1689, + "step": 26787 + }, + { + "epoch": 0.64, + "learning_rate": 5.936096764500196e-06, + "loss": 1.0838, + "step": 26788 + }, + { + "epoch": 0.64, + "learning_rate": 5.935385101354021e-06, + "loss": 0.9468, + "step": 26789 + }, + { + "epoch": 0.64, + "learning_rate": 5.934673462866907e-06, + "loss": 1.1361, + "step": 26790 + }, + { + "epoch": 0.64, + "learning_rate": 5.9339618490431735e-06, + "loss": 0.9881, + "step": 26791 + }, + { + "epoch": 0.64, + "learning_rate": 5.933250259887135e-06, + "loss": 1.0931, + "step": 26792 + }, + { + "epoch": 0.64, + "learning_rate": 5.932538695403113e-06, + "loss": 0.971, + "step": 26793 + }, + { + "epoch": 0.64, + "learning_rate": 5.931827155595422e-06, + "loss": 1.1291, + "step": 26794 + }, + { + "epoch": 0.64, + "learning_rate": 5.931115640468378e-06, + "loss": 1.0331, + "step": 26795 + }, + { + "epoch": 0.64, + "learning_rate": 5.930404150026296e-06, + "loss": 1.2675, + "step": 26796 + }, + { + "epoch": 0.64, + "learning_rate": 5.929692684273498e-06, + "loss": 0.9176, + "step": 26797 + }, + { + "epoch": 0.64, + "learning_rate": 5.928981243214296e-06, + "loss": 0.8397, + "step": 26798 + }, + { + "epoch": 0.64, + "learning_rate": 5.928269826853008e-06, + "loss": 0.9704, + "step": 26799 + }, + { + "epoch": 0.64, + "learning_rate": 5.9275584351939476e-06, + "loss": 0.9612, + "step": 26800 + }, + { + "epoch": 0.64, + "learning_rate": 5.926847068241434e-06, + "loss": 0.9704, + "step": 26801 + }, + { + "epoch": 0.64, + "learning_rate": 5.92613572599978e-06, + "loss": 0.9958, + "step": 26802 + }, + { + "epoch": 0.64, + "learning_rate": 5.925424408473304e-06, + "loss": 0.9289, + "step": 26803 + }, + { + "epoch": 0.64, + "learning_rate": 5.924713115666316e-06, + "loss": 1.0211, + "step": 26804 + }, + { + "epoch": 0.64, + "learning_rate": 5.924001847583138e-06, + "loss": 1.0913, + "step": 26805 + }, + { + "epoch": 0.64, + "learning_rate": 5.923290604228082e-06, + "loss": 1.0897, + "step": 26806 + }, + { + "epoch": 0.64, + "learning_rate": 5.9225793856054625e-06, + "loss": 1.1824, + "step": 26807 + }, + { + "epoch": 0.64, + "learning_rate": 5.921868191719594e-06, + "loss": 1.1722, + "step": 26808 + }, + { + "epoch": 0.64, + "learning_rate": 5.921157022574793e-06, + "loss": 1.1491, + "step": 26809 + }, + { + "epoch": 0.64, + "learning_rate": 5.920445878175374e-06, + "loss": 0.8592, + "step": 26810 + }, + { + "epoch": 0.64, + "learning_rate": 5.919734758525648e-06, + "loss": 1.2455, + "step": 26811 + }, + { + "epoch": 0.64, + "learning_rate": 5.919023663629933e-06, + "loss": 1.1157, + "step": 26812 + }, + { + "epoch": 0.64, + "learning_rate": 5.91831259349254e-06, + "loss": 0.8261, + "step": 26813 + }, + { + "epoch": 0.64, + "learning_rate": 5.917601548117786e-06, + "loss": 1.0326, + "step": 26814 + }, + { + "epoch": 0.64, + "learning_rate": 5.916890527509982e-06, + "loss": 0.7868, + "step": 26815 + }, + { + "epoch": 0.64, + "learning_rate": 5.916179531673443e-06, + "loss": 1.152, + "step": 26816 + }, + { + "epoch": 0.64, + "learning_rate": 5.9154685606124826e-06, + "loss": 0.9616, + "step": 26817 + }, + { + "epoch": 0.64, + "learning_rate": 5.914757614331415e-06, + "loss": 1.108, + "step": 26818 + }, + { + "epoch": 0.64, + "learning_rate": 5.9140466928345485e-06, + "loss": 0.8852, + "step": 26819 + }, + { + "epoch": 0.64, + "learning_rate": 5.9133357961262024e-06, + "loss": 1.205, + "step": 26820 + }, + { + "epoch": 0.65, + "learning_rate": 5.912624924210686e-06, + "loss": 1.0477, + "step": 26821 + }, + { + "epoch": 0.65, + "learning_rate": 5.911914077092314e-06, + "loss": 1.0507, + "step": 26822 + }, + { + "epoch": 0.65, + "learning_rate": 5.911203254775395e-06, + "loss": 1.0249, + "step": 26823 + }, + { + "epoch": 0.65, + "learning_rate": 5.910492457264246e-06, + "loss": 0.9218, + "step": 26824 + }, + { + "epoch": 0.65, + "learning_rate": 5.9097816845631784e-06, + "loss": 0.8732, + "step": 26825 + }, + { + "epoch": 0.65, + "learning_rate": 5.909070936676502e-06, + "loss": 0.9091, + "step": 26826 + }, + { + "epoch": 0.65, + "learning_rate": 5.908360213608528e-06, + "loss": 1.3421, + "step": 26827 + }, + { + "epoch": 0.65, + "learning_rate": 5.907649515363574e-06, + "loss": 1.1225, + "step": 26828 + }, + { + "epoch": 0.65, + "learning_rate": 5.906938841945946e-06, + "loss": 1.2534, + "step": 26829 + }, + { + "epoch": 0.65, + "learning_rate": 5.906228193359957e-06, + "loss": 0.9298, + "step": 26830 + }, + { + "epoch": 0.65, + "learning_rate": 5.905517569609917e-06, + "loss": 1.0659, + "step": 26831 + }, + { + "epoch": 0.65, + "learning_rate": 5.904806970700141e-06, + "loss": 1.0576, + "step": 26832 + }, + { + "epoch": 0.65, + "learning_rate": 5.904096396634935e-06, + "loss": 0.9251, + "step": 26833 + }, + { + "epoch": 0.65, + "learning_rate": 5.9033858474186165e-06, + "loss": 1.0178, + "step": 26834 + }, + { + "epoch": 0.65, + "learning_rate": 5.9026753230554915e-06, + "loss": 0.9663, + "step": 26835 + }, + { + "epoch": 0.65, + "learning_rate": 5.901964823549871e-06, + "loss": 1.2401, + "step": 26836 + }, + { + "epoch": 0.65, + "learning_rate": 5.901254348906065e-06, + "loss": 0.9943, + "step": 26837 + }, + { + "epoch": 0.65, + "learning_rate": 5.900543899128387e-06, + "loss": 1.071, + "step": 26838 + }, + { + "epoch": 0.65, + "learning_rate": 5.899833474221145e-06, + "loss": 1.1273, + "step": 26839 + }, + { + "epoch": 0.65, + "learning_rate": 5.899123074188648e-06, + "loss": 1.2153, + "step": 26840 + }, + { + "epoch": 0.65, + "learning_rate": 5.898412699035205e-06, + "loss": 1.0318, + "step": 26841 + }, + { + "epoch": 0.65, + "learning_rate": 5.89770234876513e-06, + "loss": 1.2031, + "step": 26842 + }, + { + "epoch": 0.65, + "learning_rate": 5.896992023382729e-06, + "loss": 0.9485, + "step": 26843 + }, + { + "epoch": 0.65, + "learning_rate": 5.896281722892313e-06, + "loss": 1.2306, + "step": 26844 + }, + { + "epoch": 0.65, + "learning_rate": 5.895571447298188e-06, + "loss": 1.1414, + "step": 26845 + }, + { + "epoch": 0.65, + "learning_rate": 5.894861196604668e-06, + "loss": 0.9502, + "step": 26846 + }, + { + "epoch": 0.65, + "learning_rate": 5.8941509708160596e-06, + "loss": 1.0252, + "step": 26847 + }, + { + "epoch": 0.65, + "learning_rate": 5.8934407699366695e-06, + "loss": 1.2536, + "step": 26848 + }, + { + "epoch": 0.65, + "learning_rate": 5.8927305939708106e-06, + "loss": 0.9384, + "step": 26849 + }, + { + "epoch": 0.65, + "learning_rate": 5.892020442922788e-06, + "loss": 1.099, + "step": 26850 + }, + { + "epoch": 0.65, + "learning_rate": 5.891310316796912e-06, + "loss": 0.8718, + "step": 26851 + }, + { + "epoch": 0.65, + "learning_rate": 5.890600215597488e-06, + "loss": 1.0402, + "step": 26852 + }, + { + "epoch": 0.65, + "learning_rate": 5.8898901393288276e-06, + "loss": 1.0368, + "step": 26853 + }, + { + "epoch": 0.65, + "learning_rate": 5.889180087995236e-06, + "loss": 1.2619, + "step": 26854 + }, + { + "epoch": 0.65, + "learning_rate": 5.888470061601023e-06, + "loss": 1.1802, + "step": 26855 + }, + { + "epoch": 0.65, + "learning_rate": 5.887760060150493e-06, + "loss": 1.2729, + "step": 26856 + }, + { + "epoch": 0.65, + "learning_rate": 5.887050083647958e-06, + "loss": 0.9256, + "step": 26857 + }, + { + "epoch": 0.65, + "learning_rate": 5.886340132097723e-06, + "loss": 1.0724, + "step": 26858 + }, + { + "epoch": 0.65, + "learning_rate": 5.885630205504094e-06, + "loss": 1.2201, + "step": 26859 + }, + { + "epoch": 0.65, + "learning_rate": 5.884920303871376e-06, + "loss": 1.1964, + "step": 26860 + }, + { + "epoch": 0.65, + "learning_rate": 5.884210427203881e-06, + "loss": 1.2168, + "step": 26861 + }, + { + "epoch": 0.65, + "learning_rate": 5.883500575505914e-06, + "loss": 0.9319, + "step": 26862 + }, + { + "epoch": 0.65, + "learning_rate": 5.882790748781779e-06, + "loss": 1.0727, + "step": 26863 + }, + { + "epoch": 0.65, + "learning_rate": 5.882080947035784e-06, + "loss": 1.2723, + "step": 26864 + }, + { + "epoch": 0.65, + "learning_rate": 5.881371170272237e-06, + "loss": 1.161, + "step": 26865 + }, + { + "epoch": 0.65, + "learning_rate": 5.880661418495441e-06, + "loss": 1.0049, + "step": 26866 + }, + { + "epoch": 0.65, + "learning_rate": 5.879951691709706e-06, + "loss": 1.0473, + "step": 26867 + }, + { + "epoch": 0.65, + "learning_rate": 5.879241989919331e-06, + "loss": 1.0991, + "step": 26868 + }, + { + "epoch": 0.65, + "learning_rate": 5.878532313128627e-06, + "loss": 1.2158, + "step": 26869 + }, + { + "epoch": 0.65, + "learning_rate": 5.877822661341899e-06, + "loss": 1.215, + "step": 26870 + }, + { + "epoch": 0.65, + "learning_rate": 5.877113034563451e-06, + "loss": 1.1488, + "step": 26871 + }, + { + "epoch": 0.65, + "learning_rate": 5.876403432797587e-06, + "loss": 1.3008, + "step": 26872 + }, + { + "epoch": 0.65, + "learning_rate": 5.8756938560486155e-06, + "loss": 0.9401, + "step": 26873 + }, + { + "epoch": 0.65, + "learning_rate": 5.874984304320839e-06, + "loss": 1.0022, + "step": 26874 + }, + { + "epoch": 0.65, + "learning_rate": 5.874274777618563e-06, + "loss": 1.0453, + "step": 26875 + }, + { + "epoch": 0.65, + "learning_rate": 5.873565275946088e-06, + "loss": 1.1597, + "step": 26876 + }, + { + "epoch": 0.65, + "learning_rate": 5.872855799307724e-06, + "loss": 1.118, + "step": 26877 + }, + { + "epoch": 0.65, + "learning_rate": 5.8721463477077745e-06, + "loss": 1.0907, + "step": 26878 + }, + { + "epoch": 0.65, + "learning_rate": 5.871436921150542e-06, + "loss": 1.1167, + "step": 26879 + }, + { + "epoch": 0.65, + "learning_rate": 5.870727519640327e-06, + "loss": 0.9766, + "step": 26880 + }, + { + "epoch": 0.65, + "learning_rate": 5.87001814318144e-06, + "loss": 0.9636, + "step": 26881 + }, + { + "epoch": 0.65, + "learning_rate": 5.869308791778182e-06, + "loss": 0.8254, + "step": 26882 + }, + { + "epoch": 0.65, + "learning_rate": 5.868599465434856e-06, + "loss": 1.1317, + "step": 26883 + }, + { + "epoch": 0.65, + "learning_rate": 5.867890164155763e-06, + "loss": 1.2486, + "step": 26884 + }, + { + "epoch": 0.65, + "learning_rate": 5.867180887945209e-06, + "loss": 1.0404, + "step": 26885 + }, + { + "epoch": 0.65, + "learning_rate": 5.866471636807498e-06, + "loss": 1.0175, + "step": 26886 + }, + { + "epoch": 0.65, + "learning_rate": 5.865762410746931e-06, + "loss": 1.0378, + "step": 26887 + }, + { + "epoch": 0.65, + "learning_rate": 5.865053209767808e-06, + "loss": 1.0116, + "step": 26888 + }, + { + "epoch": 0.65, + "learning_rate": 5.8643440338744386e-06, + "loss": 1.0496, + "step": 26889 + }, + { + "epoch": 0.65, + "learning_rate": 5.86363488307112e-06, + "loss": 0.8464, + "step": 26890 + }, + { + "epoch": 0.65, + "learning_rate": 5.862925757362155e-06, + "loss": 0.9581, + "step": 26891 + }, + { + "epoch": 0.65, + "learning_rate": 5.862216656751846e-06, + "loss": 1.1528, + "step": 26892 + }, + { + "epoch": 0.65, + "learning_rate": 5.861507581244497e-06, + "loss": 1.0297, + "step": 26893 + }, + { + "epoch": 0.65, + "learning_rate": 5.860798530844407e-06, + "loss": 1.2843, + "step": 26894 + }, + { + "epoch": 0.65, + "learning_rate": 5.860089505555877e-06, + "loss": 0.9338, + "step": 26895 + }, + { + "epoch": 0.65, + "learning_rate": 5.859380505383212e-06, + "loss": 0.8361, + "step": 26896 + }, + { + "epoch": 0.65, + "learning_rate": 5.8586715303307115e-06, + "loss": 1.131, + "step": 26897 + }, + { + "epoch": 0.65, + "learning_rate": 5.857962580402678e-06, + "loss": 0.9866, + "step": 26898 + }, + { + "epoch": 0.65, + "learning_rate": 5.8572536556034076e-06, + "loss": 1.2941, + "step": 26899 + }, + { + "epoch": 0.65, + "learning_rate": 5.856544755937208e-06, + "loss": 1.0847, + "step": 26900 + }, + { + "epoch": 0.65, + "learning_rate": 5.855835881408376e-06, + "loss": 0.9697, + "step": 26901 + }, + { + "epoch": 0.65, + "learning_rate": 5.855127032021213e-06, + "loss": 1.3582, + "step": 26902 + }, + { + "epoch": 0.65, + "learning_rate": 5.854418207780017e-06, + "loss": 1.168, + "step": 26903 + }, + { + "epoch": 0.65, + "learning_rate": 5.853709408689094e-06, + "loss": 1.0582, + "step": 26904 + }, + { + "epoch": 0.65, + "learning_rate": 5.8530006347527405e-06, + "loss": 1.2279, + "step": 26905 + }, + { + "epoch": 0.65, + "learning_rate": 5.852291885975257e-06, + "loss": 0.8237, + "step": 26906 + }, + { + "epoch": 0.65, + "learning_rate": 5.851583162360939e-06, + "loss": 1.1643, + "step": 26907 + }, + { + "epoch": 0.65, + "learning_rate": 5.850874463914094e-06, + "loss": 1.3205, + "step": 26908 + }, + { + "epoch": 0.65, + "learning_rate": 5.850165790639018e-06, + "loss": 1.0864, + "step": 26909 + }, + { + "epoch": 0.65, + "learning_rate": 5.849457142540008e-06, + "loss": 1.1376, + "step": 26910 + }, + { + "epoch": 0.65, + "learning_rate": 5.848748519621365e-06, + "loss": 1.1319, + "step": 26911 + }, + { + "epoch": 0.65, + "learning_rate": 5.84803992188739e-06, + "loss": 1.0691, + "step": 26912 + }, + { + "epoch": 0.65, + "learning_rate": 5.84733134934238e-06, + "loss": 0.9361, + "step": 26913 + }, + { + "epoch": 0.65, + "learning_rate": 5.846622801990633e-06, + "loss": 0.9911, + "step": 26914 + }, + { + "epoch": 0.65, + "learning_rate": 5.845914279836448e-06, + "loss": 1.1523, + "step": 26915 + }, + { + "epoch": 0.65, + "learning_rate": 5.845205782884125e-06, + "loss": 1.0559, + "step": 26916 + }, + { + "epoch": 0.65, + "learning_rate": 5.844497311137962e-06, + "loss": 0.9321, + "step": 26917 + }, + { + "epoch": 0.65, + "learning_rate": 5.843788864602256e-06, + "loss": 0.8423, + "step": 26918 + }, + { + "epoch": 0.65, + "learning_rate": 5.8430804432813025e-06, + "loss": 1.3526, + "step": 26919 + }, + { + "epoch": 0.65, + "learning_rate": 5.842372047179405e-06, + "loss": 1.3125, + "step": 26920 + }, + { + "epoch": 0.65, + "learning_rate": 5.84166367630086e-06, + "loss": 1.2285, + "step": 26921 + }, + { + "epoch": 0.65, + "learning_rate": 5.84095533064996e-06, + "loss": 1.074, + "step": 26922 + }, + { + "epoch": 0.65, + "learning_rate": 5.840247010231006e-06, + "loss": 1.1344, + "step": 26923 + }, + { + "epoch": 0.65, + "learning_rate": 5.839538715048297e-06, + "loss": 1.022, + "step": 26924 + }, + { + "epoch": 0.65, + "learning_rate": 5.838830445106127e-06, + "loss": 1.093, + "step": 26925 + }, + { + "epoch": 0.65, + "learning_rate": 5.838122200408794e-06, + "loss": 0.8778, + "step": 26926 + }, + { + "epoch": 0.65, + "learning_rate": 5.8374139809605935e-06, + "loss": 1.2237, + "step": 26927 + }, + { + "epoch": 0.65, + "learning_rate": 5.836705786765827e-06, + "loss": 0.9245, + "step": 26928 + }, + { + "epoch": 0.65, + "learning_rate": 5.8359976178287835e-06, + "loss": 1.1869, + "step": 26929 + }, + { + "epoch": 0.65, + "learning_rate": 5.835289474153762e-06, + "loss": 1.0384, + "step": 26930 + }, + { + "epoch": 0.65, + "learning_rate": 5.834581355745064e-06, + "loss": 1.1636, + "step": 26931 + }, + { + "epoch": 0.65, + "learning_rate": 5.833873262606977e-06, + "loss": 1.1847, + "step": 26932 + }, + { + "epoch": 0.65, + "learning_rate": 5.833165194743802e-06, + "loss": 1.0977, + "step": 26933 + }, + { + "epoch": 0.65, + "learning_rate": 5.832457152159837e-06, + "loss": 1.0284, + "step": 26934 + }, + { + "epoch": 0.65, + "learning_rate": 5.83174913485937e-06, + "loss": 1.1979, + "step": 26935 + }, + { + "epoch": 0.65, + "learning_rate": 5.831041142846705e-06, + "loss": 0.9098, + "step": 26936 + }, + { + "epoch": 0.65, + "learning_rate": 5.830333176126128e-06, + "loss": 0.7427, + "step": 26937 + }, + { + "epoch": 0.65, + "learning_rate": 5.829625234701941e-06, + "loss": 1.0755, + "step": 26938 + }, + { + "epoch": 0.65, + "learning_rate": 5.8289173185784385e-06, + "loss": 1.0971, + "step": 26939 + }, + { + "epoch": 0.65, + "learning_rate": 5.82820942775991e-06, + "loss": 1.0463, + "step": 26940 + }, + { + "epoch": 0.65, + "learning_rate": 5.827501562250655e-06, + "loss": 1.1851, + "step": 26941 + }, + { + "epoch": 0.65, + "learning_rate": 5.826793722054969e-06, + "loss": 1.2848, + "step": 26942 + }, + { + "epoch": 0.65, + "learning_rate": 5.826085907177141e-06, + "loss": 1.2218, + "step": 26943 + }, + { + "epoch": 0.65, + "learning_rate": 5.8253781176214675e-06, + "loss": 1.058, + "step": 26944 + }, + { + "epoch": 0.65, + "learning_rate": 5.8246703533922455e-06, + "loss": 1.3469, + "step": 26945 + }, + { + "epoch": 0.65, + "learning_rate": 5.823962614493764e-06, + "loss": 1.0269, + "step": 26946 + }, + { + "epoch": 0.65, + "learning_rate": 5.823254900930324e-06, + "loss": 1.2848, + "step": 26947 + }, + { + "epoch": 0.65, + "learning_rate": 5.822547212706209e-06, + "loss": 1.0379, + "step": 26948 + }, + { + "epoch": 0.65, + "learning_rate": 5.821839549825717e-06, + "loss": 0.9904, + "step": 26949 + }, + { + "epoch": 0.65, + "learning_rate": 5.821131912293145e-06, + "loss": 0.9136, + "step": 26950 + }, + { + "epoch": 0.65, + "learning_rate": 5.8204243001127814e-06, + "loss": 1.1672, + "step": 26951 + }, + { + "epoch": 0.65, + "learning_rate": 5.819716713288919e-06, + "loss": 0.9551, + "step": 26952 + }, + { + "epoch": 0.65, + "learning_rate": 5.819009151825855e-06, + "loss": 0.9606, + "step": 26953 + }, + { + "epoch": 0.65, + "learning_rate": 5.818301615727876e-06, + "loss": 0.9754, + "step": 26954 + }, + { + "epoch": 0.65, + "learning_rate": 5.817594104999283e-06, + "loss": 1.138, + "step": 26955 + }, + { + "epoch": 0.65, + "learning_rate": 5.816886619644357e-06, + "loss": 1.1216, + "step": 26956 + }, + { + "epoch": 0.65, + "learning_rate": 5.816179159667396e-06, + "loss": 0.7526, + "step": 26957 + }, + { + "epoch": 0.65, + "learning_rate": 5.815471725072695e-06, + "loss": 1.0799, + "step": 26958 + }, + { + "epoch": 0.65, + "learning_rate": 5.81476431586454e-06, + "loss": 1.1058, + "step": 26959 + }, + { + "epoch": 0.65, + "learning_rate": 5.814056932047226e-06, + "loss": 0.8887, + "step": 26960 + }, + { + "epoch": 0.65, + "learning_rate": 5.813349573625047e-06, + "loss": 1.138, + "step": 26961 + }, + { + "epoch": 0.65, + "learning_rate": 5.812642240602289e-06, + "loss": 1.1664, + "step": 26962 + }, + { + "epoch": 0.65, + "learning_rate": 5.811934932983247e-06, + "loss": 1.2157, + "step": 26963 + }, + { + "epoch": 0.65, + "learning_rate": 5.811227650772209e-06, + "loss": 1.0546, + "step": 26964 + }, + { + "epoch": 0.65, + "learning_rate": 5.8105203939734664e-06, + "loss": 1.1992, + "step": 26965 + }, + { + "epoch": 0.65, + "learning_rate": 5.809813162591315e-06, + "loss": 0.8776, + "step": 26966 + }, + { + "epoch": 0.65, + "learning_rate": 5.809105956630039e-06, + "loss": 1.0767, + "step": 26967 + }, + { + "epoch": 0.65, + "learning_rate": 5.80839877609393e-06, + "loss": 1.1555, + "step": 26968 + }, + { + "epoch": 0.65, + "learning_rate": 5.807691620987284e-06, + "loss": 0.9103, + "step": 26969 + }, + { + "epoch": 0.65, + "learning_rate": 5.8069844913143824e-06, + "loss": 1.1657, + "step": 26970 + }, + { + "epoch": 0.65, + "learning_rate": 5.8062773870795246e-06, + "loss": 0.9714, + "step": 26971 + }, + { + "epoch": 0.65, + "learning_rate": 5.805570308286992e-06, + "loss": 1.1647, + "step": 26972 + }, + { + "epoch": 0.65, + "learning_rate": 5.804863254941076e-06, + "loss": 1.0868, + "step": 26973 + }, + { + "epoch": 0.65, + "learning_rate": 5.804156227046073e-06, + "loss": 1.2113, + "step": 26974 + }, + { + "epoch": 0.65, + "learning_rate": 5.803449224606264e-06, + "loss": 1.4067, + "step": 26975 + }, + { + "epoch": 0.65, + "learning_rate": 5.802742247625939e-06, + "loss": 1.2022, + "step": 26976 + }, + { + "epoch": 0.65, + "learning_rate": 5.802035296109395e-06, + "loss": 0.8593, + "step": 26977 + }, + { + "epoch": 0.65, + "learning_rate": 5.801328370060911e-06, + "loss": 1.0965, + "step": 26978 + }, + { + "epoch": 0.65, + "learning_rate": 5.800621469484785e-06, + "loss": 0.973, + "step": 26979 + }, + { + "epoch": 0.65, + "learning_rate": 5.799914594385296e-06, + "loss": 1.0348, + "step": 26980 + }, + { + "epoch": 0.65, + "learning_rate": 5.799207744766737e-06, + "loss": 1.0921, + "step": 26981 + }, + { + "epoch": 0.65, + "learning_rate": 5.798500920633402e-06, + "loss": 1.0735, + "step": 26982 + }, + { + "epoch": 0.65, + "learning_rate": 5.797794121989568e-06, + "loss": 0.8336, + "step": 26983 + }, + { + "epoch": 0.65, + "learning_rate": 5.797087348839529e-06, + "loss": 1.0323, + "step": 26984 + }, + { + "epoch": 0.65, + "learning_rate": 5.7963806011875775e-06, + "loss": 1.0783, + "step": 26985 + }, + { + "epoch": 0.65, + "learning_rate": 5.795673879037992e-06, + "loss": 0.9739, + "step": 26986 + }, + { + "epoch": 0.65, + "learning_rate": 5.794967182395064e-06, + "loss": 1.1321, + "step": 26987 + }, + { + "epoch": 0.65, + "learning_rate": 5.794260511263084e-06, + "loss": 1.2345, + "step": 26988 + }, + { + "epoch": 0.65, + "learning_rate": 5.793553865646333e-06, + "loss": 1.2374, + "step": 26989 + }, + { + "epoch": 0.65, + "learning_rate": 5.792847245549104e-06, + "loss": 0.7176, + "step": 26990 + }, + { + "epoch": 0.65, + "learning_rate": 5.792140650975679e-06, + "loss": 1.0029, + "step": 26991 + }, + { + "epoch": 0.65, + "learning_rate": 5.7914340819303475e-06, + "loss": 1.1786, + "step": 26992 + }, + { + "epoch": 0.65, + "learning_rate": 5.7907275384173975e-06, + "loss": 1.1296, + "step": 26993 + }, + { + "epoch": 0.65, + "learning_rate": 5.79002102044111e-06, + "loss": 1.1787, + "step": 26994 + }, + { + "epoch": 0.65, + "learning_rate": 5.789314528005776e-06, + "loss": 0.7325, + "step": 26995 + }, + { + "epoch": 0.65, + "learning_rate": 5.788608061115684e-06, + "loss": 0.93, + "step": 26996 + }, + { + "epoch": 0.65, + "learning_rate": 5.787901619775111e-06, + "loss": 1.0689, + "step": 26997 + }, + { + "epoch": 0.65, + "learning_rate": 5.787195203988352e-06, + "loss": 1.2942, + "step": 26998 + }, + { + "epoch": 0.65, + "learning_rate": 5.786488813759682e-06, + "loss": 1.0705, + "step": 26999 + }, + { + "epoch": 0.65, + "learning_rate": 5.7857824490934e-06, + "loss": 1.0845, + "step": 27000 + }, + { + "epoch": 0.65, + "learning_rate": 5.785076109993781e-06, + "loss": 0.9595, + "step": 27001 + }, + { + "epoch": 0.65, + "learning_rate": 5.7843697964651144e-06, + "loss": 1.1062, + "step": 27002 + }, + { + "epoch": 0.65, + "learning_rate": 5.78366350851169e-06, + "loss": 1.161, + "step": 27003 + }, + { + "epoch": 0.65, + "learning_rate": 5.782957246137781e-06, + "loss": 1.1228, + "step": 27004 + }, + { + "epoch": 0.65, + "learning_rate": 5.782251009347679e-06, + "loss": 0.9684, + "step": 27005 + }, + { + "epoch": 0.65, + "learning_rate": 5.781544798145673e-06, + "loss": 1.0947, + "step": 27006 + }, + { + "epoch": 0.65, + "learning_rate": 5.7808386125360385e-06, + "loss": 1.3319, + "step": 27007 + }, + { + "epoch": 0.65, + "learning_rate": 5.780132452523066e-06, + "loss": 1.2157, + "step": 27008 + }, + { + "epoch": 0.65, + "learning_rate": 5.779426318111036e-06, + "loss": 1.0101, + "step": 27009 + }, + { + "epoch": 0.65, + "learning_rate": 5.7787202093042316e-06, + "loss": 1.1209, + "step": 27010 + }, + { + "epoch": 0.65, + "learning_rate": 5.778014126106943e-06, + "loss": 1.141, + "step": 27011 + }, + { + "epoch": 0.65, + "learning_rate": 5.777308068523446e-06, + "loss": 1.0707, + "step": 27012 + }, + { + "epoch": 0.65, + "learning_rate": 5.776602036558028e-06, + "loss": 1.0617, + "step": 27013 + }, + { + "epoch": 0.65, + "learning_rate": 5.7758960302149755e-06, + "loss": 0.9922, + "step": 27014 + }, + { + "epoch": 0.65, + "learning_rate": 5.7751900494985644e-06, + "loss": 1.2359, + "step": 27015 + }, + { + "epoch": 0.65, + "learning_rate": 5.774484094413087e-06, + "loss": 1.005, + "step": 27016 + }, + { + "epoch": 0.65, + "learning_rate": 5.773778164962815e-06, + "loss": 1.1202, + "step": 27017 + }, + { + "epoch": 0.65, + "learning_rate": 5.773072261152037e-06, + "loss": 0.8635, + "step": 27018 + }, + { + "epoch": 0.65, + "learning_rate": 5.772366382985039e-06, + "loss": 1.1078, + "step": 27019 + }, + { + "epoch": 0.65, + "learning_rate": 5.771660530466097e-06, + "loss": 0.8998, + "step": 27020 + }, + { + "epoch": 0.65, + "learning_rate": 5.770954703599495e-06, + "loss": 1.099, + "step": 27021 + }, + { + "epoch": 0.65, + "learning_rate": 5.77024890238952e-06, + "loss": 1.1861, + "step": 27022 + }, + { + "epoch": 0.65, + "learning_rate": 5.769543126840447e-06, + "loss": 1.0457, + "step": 27023 + }, + { + "epoch": 0.65, + "learning_rate": 5.76883737695656e-06, + "loss": 1.1283, + "step": 27024 + }, + { + "epoch": 0.65, + "learning_rate": 5.768131652742146e-06, + "loss": 1.2561, + "step": 27025 + }, + { + "epoch": 0.65, + "learning_rate": 5.767425954201478e-06, + "loss": 1.1085, + "step": 27026 + }, + { + "epoch": 0.65, + "learning_rate": 5.766720281338843e-06, + "loss": 0.9565, + "step": 27027 + }, + { + "epoch": 0.65, + "learning_rate": 5.766014634158519e-06, + "loss": 1.2485, + "step": 27028 + }, + { + "epoch": 0.65, + "learning_rate": 5.765309012664786e-06, + "loss": 1.1278, + "step": 27029 + }, + { + "epoch": 0.65, + "learning_rate": 5.7646034168619335e-06, + "loss": 0.9962, + "step": 27030 + }, + { + "epoch": 0.65, + "learning_rate": 5.76389784675423e-06, + "loss": 0.8933, + "step": 27031 + }, + { + "epoch": 0.65, + "learning_rate": 5.763192302345963e-06, + "loss": 1.0467, + "step": 27032 + }, + { + "epoch": 0.65, + "learning_rate": 5.762486783641415e-06, + "loss": 1.066, + "step": 27033 + }, + { + "epoch": 0.65, + "learning_rate": 5.76178129064486e-06, + "loss": 1.1707, + "step": 27034 + }, + { + "epoch": 0.65, + "learning_rate": 5.761075823360584e-06, + "loss": 0.9345, + "step": 27035 + }, + { + "epoch": 0.65, + "learning_rate": 5.76037038179286e-06, + "loss": 0.9974, + "step": 27036 + }, + { + "epoch": 0.65, + "learning_rate": 5.759664965945972e-06, + "loss": 0.9639, + "step": 27037 + }, + { + "epoch": 0.65, + "learning_rate": 5.7589595758242036e-06, + "loss": 1.1167, + "step": 27038 + }, + { + "epoch": 0.65, + "learning_rate": 5.758254211431826e-06, + "loss": 1.1069, + "step": 27039 + }, + { + "epoch": 0.65, + "learning_rate": 5.757548872773122e-06, + "loss": 1.0712, + "step": 27040 + }, + { + "epoch": 0.65, + "learning_rate": 5.756843559852374e-06, + "loss": 1.136, + "step": 27041 + }, + { + "epoch": 0.65, + "learning_rate": 5.756138272673854e-06, + "loss": 0.8644, + "step": 27042 + }, + { + "epoch": 0.65, + "learning_rate": 5.755433011241851e-06, + "loss": 1.1845, + "step": 27043 + }, + { + "epoch": 0.65, + "learning_rate": 5.754727775560633e-06, + "loss": 1.2752, + "step": 27044 + }, + { + "epoch": 0.65, + "learning_rate": 5.754022565634483e-06, + "loss": 1.1067, + "step": 27045 + }, + { + "epoch": 0.65, + "learning_rate": 5.753317381467682e-06, + "loss": 1.0677, + "step": 27046 + }, + { + "epoch": 0.65, + "learning_rate": 5.752612223064502e-06, + "loss": 0.9213, + "step": 27047 + }, + { + "epoch": 0.65, + "learning_rate": 5.7519070904292255e-06, + "loss": 1.0682, + "step": 27048 + }, + { + "epoch": 0.65, + "learning_rate": 5.7512019835661325e-06, + "loss": 0.8458, + "step": 27049 + }, + { + "epoch": 0.65, + "learning_rate": 5.750496902479495e-06, + "loss": 1.3431, + "step": 27050 + }, + { + "epoch": 0.65, + "learning_rate": 5.749791847173596e-06, + "loss": 1.0484, + "step": 27051 + }, + { + "epoch": 0.65, + "learning_rate": 5.749086817652707e-06, + "loss": 0.8748, + "step": 27052 + }, + { + "epoch": 0.65, + "learning_rate": 5.748381813921109e-06, + "loss": 1.1712, + "step": 27053 + }, + { + "epoch": 0.65, + "learning_rate": 5.74767683598308e-06, + "loss": 0.9027, + "step": 27054 + }, + { + "epoch": 0.65, + "learning_rate": 5.7469718838428935e-06, + "loss": 0.9435, + "step": 27055 + }, + { + "epoch": 0.65, + "learning_rate": 5.746266957504828e-06, + "loss": 1.0262, + "step": 27056 + }, + { + "epoch": 0.65, + "learning_rate": 5.745562056973164e-06, + "loss": 1.0912, + "step": 27057 + }, + { + "epoch": 0.65, + "learning_rate": 5.744857182252171e-06, + "loss": 1.1055, + "step": 27058 + }, + { + "epoch": 0.65, + "learning_rate": 5.744152333346131e-06, + "loss": 1.229, + "step": 27059 + }, + { + "epoch": 0.65, + "learning_rate": 5.743447510259315e-06, + "loss": 1.109, + "step": 27060 + }, + { + "epoch": 0.65, + "learning_rate": 5.7427427129960015e-06, + "loss": 1.0827, + "step": 27061 + }, + { + "epoch": 0.65, + "learning_rate": 5.7420379415604696e-06, + "loss": 1.1462, + "step": 27062 + }, + { + "epoch": 0.65, + "learning_rate": 5.741333195956989e-06, + "loss": 1.044, + "step": 27063 + }, + { + "epoch": 0.65, + "learning_rate": 5.740628476189838e-06, + "loss": 1.2636, + "step": 27064 + }, + { + "epoch": 0.65, + "learning_rate": 5.739923782263297e-06, + "loss": 1.0318, + "step": 27065 + }, + { + "epoch": 0.65, + "learning_rate": 5.7392191141816315e-06, + "loss": 1.0844, + "step": 27066 + }, + { + "epoch": 0.65, + "learning_rate": 5.7385144719491215e-06, + "loss": 1.1137, + "step": 27067 + }, + { + "epoch": 0.65, + "learning_rate": 5.737809855570046e-06, + "loss": 1.019, + "step": 27068 + }, + { + "epoch": 0.65, + "learning_rate": 5.73710526504867e-06, + "loss": 1.0431, + "step": 27069 + }, + { + "epoch": 0.65, + "learning_rate": 5.73640070038928e-06, + "loss": 1.1138, + "step": 27070 + }, + { + "epoch": 0.65, + "learning_rate": 5.735696161596139e-06, + "loss": 1.0704, + "step": 27071 + }, + { + "epoch": 0.65, + "learning_rate": 5.734991648673527e-06, + "loss": 1.0718, + "step": 27072 + }, + { + "epoch": 0.65, + "learning_rate": 5.734287161625721e-06, + "loss": 1.1733, + "step": 27073 + }, + { + "epoch": 0.65, + "learning_rate": 5.733582700456988e-06, + "loss": 0.9181, + "step": 27074 + }, + { + "epoch": 0.65, + "learning_rate": 5.732878265171604e-06, + "loss": 1.3309, + "step": 27075 + }, + { + "epoch": 0.65, + "learning_rate": 5.732173855773848e-06, + "loss": 1.0955, + "step": 27076 + }, + { + "epoch": 0.65, + "learning_rate": 5.731469472267986e-06, + "loss": 1.1274, + "step": 27077 + }, + { + "epoch": 0.65, + "learning_rate": 5.7307651146582985e-06, + "loss": 1.1998, + "step": 27078 + }, + { + "epoch": 0.65, + "learning_rate": 5.730060782949052e-06, + "loss": 0.8953, + "step": 27079 + }, + { + "epoch": 0.65, + "learning_rate": 5.729356477144521e-06, + "loss": 1.0658, + "step": 27080 + }, + { + "epoch": 0.65, + "learning_rate": 5.728652197248985e-06, + "loss": 1.1171, + "step": 27081 + }, + { + "epoch": 0.65, + "learning_rate": 5.727947943266706e-06, + "loss": 1.0927, + "step": 27082 + }, + { + "epoch": 0.65, + "learning_rate": 5.7272437152019626e-06, + "loss": 1.0002, + "step": 27083 + }, + { + "epoch": 0.65, + "learning_rate": 5.72653951305903e-06, + "loss": 1.0808, + "step": 27084 + }, + { + "epoch": 0.65, + "learning_rate": 5.725835336842174e-06, + "loss": 1.1864, + "step": 27085 + }, + { + "epoch": 0.65, + "learning_rate": 5.725131186555672e-06, + "loss": 1.1469, + "step": 27086 + }, + { + "epoch": 0.65, + "learning_rate": 5.724427062203793e-06, + "loss": 1.0332, + "step": 27087 + }, + { + "epoch": 0.65, + "learning_rate": 5.7237229637908055e-06, + "loss": 1.0506, + "step": 27088 + }, + { + "epoch": 0.65, + "learning_rate": 5.723018891320991e-06, + "loss": 0.9029, + "step": 27089 + }, + { + "epoch": 0.65, + "learning_rate": 5.722314844798611e-06, + "loss": 1.3226, + "step": 27090 + }, + { + "epoch": 0.65, + "learning_rate": 5.721610824227939e-06, + "loss": 1.0263, + "step": 27091 + }, + { + "epoch": 0.65, + "learning_rate": 5.720906829613253e-06, + "loss": 0.9905, + "step": 27092 + }, + { + "epoch": 0.65, + "learning_rate": 5.720202860958815e-06, + "loss": 0.9055, + "step": 27093 + }, + { + "epoch": 0.65, + "learning_rate": 5.719498918268903e-06, + "loss": 0.938, + "step": 27094 + }, + { + "epoch": 0.65, + "learning_rate": 5.7187950015477815e-06, + "loss": 1.0066, + "step": 27095 + }, + { + "epoch": 0.65, + "learning_rate": 5.718091110799722e-06, + "loss": 1.1891, + "step": 27096 + }, + { + "epoch": 0.65, + "learning_rate": 5.717387246029e-06, + "loss": 1.1296, + "step": 27097 + }, + { + "epoch": 0.65, + "learning_rate": 5.71668340723988e-06, + "loss": 1.1795, + "step": 27098 + }, + { + "epoch": 0.65, + "learning_rate": 5.715979594436634e-06, + "loss": 1.2022, + "step": 27099 + }, + { + "epoch": 0.65, + "learning_rate": 5.715275807623536e-06, + "loss": 1.1495, + "step": 27100 + }, + { + "epoch": 0.65, + "learning_rate": 5.714572046804847e-06, + "loss": 0.8765, + "step": 27101 + }, + { + "epoch": 0.65, + "learning_rate": 5.713868311984847e-06, + "loss": 1.0717, + "step": 27102 + }, + { + "epoch": 0.65, + "learning_rate": 5.713164603167795e-06, + "loss": 1.2586, + "step": 27103 + }, + { + "epoch": 0.65, + "learning_rate": 5.712460920357965e-06, + "loss": 1.0613, + "step": 27104 + }, + { + "epoch": 0.65, + "learning_rate": 5.711757263559629e-06, + "loss": 1.0243, + "step": 27105 + }, + { + "epoch": 0.65, + "learning_rate": 5.711053632777051e-06, + "loss": 1.1741, + "step": 27106 + }, + { + "epoch": 0.65, + "learning_rate": 5.710350028014501e-06, + "loss": 1.0283, + "step": 27107 + }, + { + "epoch": 0.65, + "learning_rate": 5.709646449276252e-06, + "loss": 1.0431, + "step": 27108 + }, + { + "epoch": 0.65, + "learning_rate": 5.708942896566565e-06, + "loss": 0.8306, + "step": 27109 + }, + { + "epoch": 0.65, + "learning_rate": 5.7082393698897164e-06, + "loss": 0.8887, + "step": 27110 + }, + { + "epoch": 0.65, + "learning_rate": 5.707535869249965e-06, + "loss": 0.9342, + "step": 27111 + }, + { + "epoch": 0.65, + "learning_rate": 5.706832394651585e-06, + "loss": 0.9816, + "step": 27112 + }, + { + "epoch": 0.65, + "learning_rate": 5.706128946098848e-06, + "loss": 0.8888, + "step": 27113 + }, + { + "epoch": 0.65, + "learning_rate": 5.705425523596011e-06, + "loss": 1.1027, + "step": 27114 + }, + { + "epoch": 0.65, + "learning_rate": 5.704722127147347e-06, + "loss": 0.8605, + "step": 27115 + }, + { + "epoch": 0.65, + "learning_rate": 5.704018756757128e-06, + "loss": 1.3099, + "step": 27116 + }, + { + "epoch": 0.65, + "learning_rate": 5.703315412429613e-06, + "loss": 1.0318, + "step": 27117 + }, + { + "epoch": 0.65, + "learning_rate": 5.702612094169073e-06, + "loss": 0.8637, + "step": 27118 + }, + { + "epoch": 0.65, + "learning_rate": 5.701908801979779e-06, + "loss": 1.1324, + "step": 27119 + }, + { + "epoch": 0.65, + "learning_rate": 5.701205535865988e-06, + "loss": 0.9092, + "step": 27120 + }, + { + "epoch": 0.65, + "learning_rate": 5.700502295831976e-06, + "loss": 0.9847, + "step": 27121 + }, + { + "epoch": 0.65, + "learning_rate": 5.699799081882002e-06, + "loss": 1.0938, + "step": 27122 + }, + { + "epoch": 0.65, + "learning_rate": 5.6990958940203355e-06, + "loss": 1.162, + "step": 27123 + }, + { + "epoch": 0.65, + "learning_rate": 5.698392732251246e-06, + "loss": 1.0889, + "step": 27124 + }, + { + "epoch": 0.65, + "learning_rate": 5.697689596578992e-06, + "loss": 1.1067, + "step": 27125 + }, + { + "epoch": 0.65, + "learning_rate": 5.696986487007844e-06, + "loss": 1.0803, + "step": 27126 + }, + { + "epoch": 0.65, + "learning_rate": 5.6962834035420715e-06, + "loss": 1.1579, + "step": 27127 + }, + { + "epoch": 0.65, + "learning_rate": 5.6955803461859295e-06, + "loss": 1.0231, + "step": 27128 + }, + { + "epoch": 0.65, + "learning_rate": 5.694877314943695e-06, + "loss": 0.9014, + "step": 27129 + }, + { + "epoch": 0.65, + "learning_rate": 5.69417430981962e-06, + "loss": 1.1244, + "step": 27130 + }, + { + "epoch": 0.65, + "learning_rate": 5.693471330817979e-06, + "loss": 1.1991, + "step": 27131 + }, + { + "epoch": 0.65, + "learning_rate": 5.692768377943039e-06, + "loss": 0.9388, + "step": 27132 + }, + { + "epoch": 0.65, + "learning_rate": 5.692065451199055e-06, + "loss": 0.929, + "step": 27133 + }, + { + "epoch": 0.65, + "learning_rate": 5.6913625505902966e-06, + "loss": 1.0117, + "step": 27134 + }, + { + "epoch": 0.65, + "learning_rate": 5.690659676121033e-06, + "loss": 1.1833, + "step": 27135 + }, + { + "epoch": 0.65, + "learning_rate": 5.689956827795519e-06, + "loss": 0.9869, + "step": 27136 + }, + { + "epoch": 0.65, + "learning_rate": 5.689254005618026e-06, + "loss": 0.8534, + "step": 27137 + }, + { + "epoch": 0.65, + "learning_rate": 5.688551209592813e-06, + "loss": 1.0483, + "step": 27138 + }, + { + "epoch": 0.65, + "learning_rate": 5.687848439724144e-06, + "loss": 1.0186, + "step": 27139 + }, + { + "epoch": 0.65, + "learning_rate": 5.68714569601629e-06, + "loss": 1.1871, + "step": 27140 + }, + { + "epoch": 0.65, + "learning_rate": 5.686442978473503e-06, + "loss": 0.9222, + "step": 27141 + }, + { + "epoch": 0.65, + "learning_rate": 5.685740287100052e-06, + "loss": 1.0109, + "step": 27142 + }, + { + "epoch": 0.65, + "learning_rate": 5.685037621900204e-06, + "loss": 1.0494, + "step": 27143 + }, + { + "epoch": 0.65, + "learning_rate": 5.684334982878216e-06, + "loss": 1.0413, + "step": 27144 + }, + { + "epoch": 0.65, + "learning_rate": 5.683632370038354e-06, + "loss": 1.0463, + "step": 27145 + }, + { + "epoch": 0.65, + "learning_rate": 5.682929783384876e-06, + "loss": 1.2356, + "step": 27146 + }, + { + "epoch": 0.65, + "learning_rate": 5.682227222922047e-06, + "loss": 0.9534, + "step": 27147 + }, + { + "epoch": 0.65, + "learning_rate": 5.681524688654135e-06, + "loss": 1.1794, + "step": 27148 + }, + { + "epoch": 0.65, + "learning_rate": 5.680822180585392e-06, + "loss": 0.8128, + "step": 27149 + }, + { + "epoch": 0.65, + "learning_rate": 5.680119698720086e-06, + "loss": 0.8943, + "step": 27150 + }, + { + "epoch": 0.65, + "learning_rate": 5.679417243062481e-06, + "loss": 0.9856, + "step": 27151 + }, + { + "epoch": 0.65, + "learning_rate": 5.67871481361683e-06, + "loss": 0.9849, + "step": 27152 + }, + { + "epoch": 0.65, + "learning_rate": 5.678012410387405e-06, + "loss": 1.1899, + "step": 27153 + }, + { + "epoch": 0.65, + "learning_rate": 5.677310033378458e-06, + "loss": 1.1534, + "step": 27154 + }, + { + "epoch": 0.65, + "learning_rate": 5.676607682594255e-06, + "loss": 1.1758, + "step": 27155 + }, + { + "epoch": 0.65, + "learning_rate": 5.6759053580390596e-06, + "loss": 1.1636, + "step": 27156 + }, + { + "epoch": 0.65, + "learning_rate": 5.675203059717125e-06, + "loss": 0.9447, + "step": 27157 + }, + { + "epoch": 0.65, + "learning_rate": 5.6745007876327176e-06, + "loss": 1.1853, + "step": 27158 + }, + { + "epoch": 0.65, + "learning_rate": 5.673798541790099e-06, + "loss": 1.1867, + "step": 27159 + }, + { + "epoch": 0.65, + "learning_rate": 5.673096322193523e-06, + "loss": 1.1014, + "step": 27160 + }, + { + "epoch": 0.65, + "learning_rate": 5.672394128847254e-06, + "loss": 0.9286, + "step": 27161 + }, + { + "epoch": 0.65, + "learning_rate": 5.671691961755557e-06, + "loss": 1.192, + "step": 27162 + }, + { + "epoch": 0.65, + "learning_rate": 5.670989820922681e-06, + "loss": 0.9495, + "step": 27163 + }, + { + "epoch": 0.65, + "learning_rate": 5.6702877063528975e-06, + "loss": 1.1243, + "step": 27164 + }, + { + "epoch": 0.65, + "learning_rate": 5.669585618050455e-06, + "loss": 1.0723, + "step": 27165 + }, + { + "epoch": 0.65, + "learning_rate": 5.668883556019617e-06, + "loss": 1.0951, + "step": 27166 + }, + { + "epoch": 0.65, + "learning_rate": 5.668181520264646e-06, + "loss": 1.1496, + "step": 27167 + }, + { + "epoch": 0.65, + "learning_rate": 5.667479510789797e-06, + "loss": 0.9881, + "step": 27168 + }, + { + "epoch": 0.65, + "learning_rate": 5.666777527599337e-06, + "loss": 1.1419, + "step": 27169 + }, + { + "epoch": 0.65, + "learning_rate": 5.666075570697511e-06, + "loss": 0.9196, + "step": 27170 + }, + { + "epoch": 0.65, + "learning_rate": 5.665373640088587e-06, + "loss": 1.2379, + "step": 27171 + }, + { + "epoch": 0.65, + "learning_rate": 5.6646717357768264e-06, + "loss": 1.005, + "step": 27172 + }, + { + "epoch": 0.65, + "learning_rate": 5.663969857766478e-06, + "loss": 0.9865, + "step": 27173 + }, + { + "epoch": 0.65, + "learning_rate": 5.663268006061809e-06, + "loss": 0.6658, + "step": 27174 + }, + { + "epoch": 0.65, + "learning_rate": 5.662566180667068e-06, + "loss": 0.9972, + "step": 27175 + }, + { + "epoch": 0.65, + "learning_rate": 5.661864381586517e-06, + "loss": 0.766, + "step": 27176 + }, + { + "epoch": 0.65, + "learning_rate": 5.66116260882442e-06, + "loss": 1.081, + "step": 27177 + }, + { + "epoch": 0.65, + "learning_rate": 5.660460862385025e-06, + "loss": 0.9511, + "step": 27178 + }, + { + "epoch": 0.65, + "learning_rate": 5.659759142272592e-06, + "loss": 1.101, + "step": 27179 + }, + { + "epoch": 0.65, + "learning_rate": 5.659057448491384e-06, + "loss": 1.0638, + "step": 27180 + }, + { + "epoch": 0.65, + "learning_rate": 5.6583557810456495e-06, + "loss": 1.0525, + "step": 27181 + }, + { + "epoch": 0.65, + "learning_rate": 5.6576541399396524e-06, + "loss": 1.191, + "step": 27182 + }, + { + "epoch": 0.65, + "learning_rate": 5.656952525177644e-06, + "loss": 0.9126, + "step": 27183 + }, + { + "epoch": 0.65, + "learning_rate": 5.656250936763882e-06, + "loss": 1.2479, + "step": 27184 + }, + { + "epoch": 0.65, + "learning_rate": 5.655549374702627e-06, + "loss": 1.1414, + "step": 27185 + }, + { + "epoch": 0.65, + "learning_rate": 5.65484783899813e-06, + "loss": 1.2089, + "step": 27186 + }, + { + "epoch": 0.65, + "learning_rate": 5.6541463296546485e-06, + "loss": 0.7149, + "step": 27187 + }, + { + "epoch": 0.65, + "learning_rate": 5.6534448466764416e-06, + "loss": 0.9081, + "step": 27188 + }, + { + "epoch": 0.65, + "learning_rate": 5.652743390067759e-06, + "loss": 0.7078, + "step": 27189 + }, + { + "epoch": 0.65, + "learning_rate": 5.652041959832866e-06, + "loss": 1.1662, + "step": 27190 + }, + { + "epoch": 0.65, + "learning_rate": 5.651340555976005e-06, + "loss": 1.261, + "step": 27191 + }, + { + "epoch": 0.65, + "learning_rate": 5.650639178501438e-06, + "loss": 0.6605, + "step": 27192 + }, + { + "epoch": 0.65, + "learning_rate": 5.649937827413425e-06, + "loss": 1.0897, + "step": 27193 + }, + { + "epoch": 0.65, + "learning_rate": 5.6492365027162135e-06, + "loss": 0.899, + "step": 27194 + }, + { + "epoch": 0.65, + "learning_rate": 5.648535204414058e-06, + "loss": 0.8466, + "step": 27195 + }, + { + "epoch": 0.65, + "learning_rate": 5.647833932511222e-06, + "loss": 1.0826, + "step": 27196 + }, + { + "epoch": 0.65, + "learning_rate": 5.647132687011949e-06, + "loss": 0.9624, + "step": 27197 + }, + { + "epoch": 0.65, + "learning_rate": 5.6464314679204976e-06, + "loss": 1.188, + "step": 27198 + }, + { + "epoch": 0.65, + "learning_rate": 5.645730275241128e-06, + "loss": 1.0802, + "step": 27199 + }, + { + "epoch": 0.65, + "learning_rate": 5.645029108978083e-06, + "loss": 0.9408, + "step": 27200 + }, + { + "epoch": 0.65, + "learning_rate": 5.6443279691356265e-06, + "loss": 1.0902, + "step": 27201 + }, + { + "epoch": 0.65, + "learning_rate": 5.643626855718004e-06, + "loss": 0.9474, + "step": 27202 + }, + { + "epoch": 0.65, + "learning_rate": 5.6429257687294725e-06, + "loss": 1.2742, + "step": 27203 + }, + { + "epoch": 0.65, + "learning_rate": 5.642224708174289e-06, + "loss": 1.024, + "step": 27204 + }, + { + "epoch": 0.65, + "learning_rate": 5.641523674056701e-06, + "loss": 1.1573, + "step": 27205 + }, + { + "epoch": 0.65, + "learning_rate": 5.640822666380962e-06, + "loss": 1.0838, + "step": 27206 + }, + { + "epoch": 0.65, + "learning_rate": 5.640121685151331e-06, + "loss": 0.7523, + "step": 27207 + }, + { + "epoch": 0.65, + "learning_rate": 5.639420730372052e-06, + "loss": 0.8467, + "step": 27208 + }, + { + "epoch": 0.65, + "learning_rate": 5.638719802047385e-06, + "loss": 1.2144, + "step": 27209 + }, + { + "epoch": 0.65, + "learning_rate": 5.638018900181576e-06, + "loss": 0.9366, + "step": 27210 + }, + { + "epoch": 0.65, + "learning_rate": 5.637318024778882e-06, + "loss": 0.9673, + "step": 27211 + }, + { + "epoch": 0.65, + "learning_rate": 5.636617175843555e-06, + "loss": 0.8402, + "step": 27212 + }, + { + "epoch": 0.65, + "learning_rate": 5.635916353379842e-06, + "loss": 1.2246, + "step": 27213 + }, + { + "epoch": 0.65, + "learning_rate": 5.635215557391998e-06, + "loss": 1.0067, + "step": 27214 + }, + { + "epoch": 0.65, + "learning_rate": 5.634514787884279e-06, + "loss": 1.1158, + "step": 27215 + }, + { + "epoch": 0.65, + "learning_rate": 5.633814044860928e-06, + "loss": 1.0348, + "step": 27216 + }, + { + "epoch": 0.65, + "learning_rate": 5.633113328326203e-06, + "loss": 0.8449, + "step": 27217 + }, + { + "epoch": 0.65, + "learning_rate": 5.632412638284349e-06, + "loss": 1.0121, + "step": 27218 + }, + { + "epoch": 0.65, + "learning_rate": 5.63171197473962e-06, + "loss": 1.003, + "step": 27219 + }, + { + "epoch": 0.65, + "learning_rate": 5.631011337696272e-06, + "loss": 1.1945, + "step": 27220 + }, + { + "epoch": 0.65, + "learning_rate": 5.630310727158546e-06, + "loss": 1.1048, + "step": 27221 + }, + { + "epoch": 0.65, + "learning_rate": 5.629610143130696e-06, + "loss": 0.939, + "step": 27222 + }, + { + "epoch": 0.65, + "learning_rate": 5.628909585616979e-06, + "loss": 1.2018, + "step": 27223 + }, + { + "epoch": 0.65, + "learning_rate": 5.628209054621634e-06, + "loss": 0.9122, + "step": 27224 + }, + { + "epoch": 0.65, + "learning_rate": 5.627508550148921e-06, + "loss": 0.7577, + "step": 27225 + }, + { + "epoch": 0.65, + "learning_rate": 5.62680807220308e-06, + "loss": 0.9675, + "step": 27226 + }, + { + "epoch": 0.65, + "learning_rate": 5.626107620788368e-06, + "loss": 1.1196, + "step": 27227 + }, + { + "epoch": 0.65, + "learning_rate": 5.6254071959090335e-06, + "loss": 1.1135, + "step": 27228 + }, + { + "epoch": 0.65, + "learning_rate": 5.6247067975693236e-06, + "loss": 1.0034, + "step": 27229 + }, + { + "epoch": 0.65, + "learning_rate": 5.624006425773486e-06, + "loss": 0.9757, + "step": 27230 + }, + { + "epoch": 0.65, + "learning_rate": 5.623306080525776e-06, + "loss": 1.0092, + "step": 27231 + }, + { + "epoch": 0.65, + "learning_rate": 5.622605761830437e-06, + "loss": 1.0444, + "step": 27232 + }, + { + "epoch": 0.65, + "learning_rate": 5.6219054696917216e-06, + "loss": 0.8721, + "step": 27233 + }, + { + "epoch": 0.65, + "learning_rate": 5.621205204113872e-06, + "loss": 1.0066, + "step": 27234 + }, + { + "epoch": 0.65, + "learning_rate": 5.620504965101142e-06, + "loss": 1.073, + "step": 27235 + }, + { + "epoch": 0.65, + "learning_rate": 5.619804752657781e-06, + "loss": 1.0837, + "step": 27236 + }, + { + "epoch": 0.66, + "learning_rate": 5.6191045667880294e-06, + "loss": 0.9993, + "step": 27237 + }, + { + "epoch": 0.66, + "learning_rate": 5.618404407496141e-06, + "loss": 0.9101, + "step": 27238 + }, + { + "epoch": 0.66, + "learning_rate": 5.6177042747863655e-06, + "loss": 1.016, + "step": 27239 + }, + { + "epoch": 0.66, + "learning_rate": 5.617004168662944e-06, + "loss": 0.7707, + "step": 27240 + }, + { + "epoch": 0.66, + "learning_rate": 5.616304089130127e-06, + "loss": 0.9685, + "step": 27241 + }, + { + "epoch": 0.66, + "learning_rate": 5.615604036192166e-06, + "loss": 0.9168, + "step": 27242 + }, + { + "epoch": 0.66, + "learning_rate": 5.614904009853302e-06, + "loss": 1.2121, + "step": 27243 + }, + { + "epoch": 0.66, + "learning_rate": 5.614204010117785e-06, + "loss": 1.0334, + "step": 27244 + }, + { + "epoch": 0.66, + "learning_rate": 5.613504036989858e-06, + "loss": 1.0826, + "step": 27245 + }, + { + "epoch": 0.66, + "learning_rate": 5.61280409047377e-06, + "loss": 0.9443, + "step": 27246 + }, + { + "epoch": 0.66, + "learning_rate": 5.612104170573771e-06, + "loss": 1.0462, + "step": 27247 + }, + { + "epoch": 0.66, + "learning_rate": 5.611404277294101e-06, + "loss": 1.2044, + "step": 27248 + }, + { + "epoch": 0.66, + "learning_rate": 5.610704410639009e-06, + "loss": 1.1607, + "step": 27249 + }, + { + "epoch": 0.66, + "learning_rate": 5.610004570612745e-06, + "loss": 1.2022, + "step": 27250 + }, + { + "epoch": 0.66, + "learning_rate": 5.609304757219546e-06, + "loss": 0.8444, + "step": 27251 + }, + { + "epoch": 0.66, + "learning_rate": 5.608604970463667e-06, + "loss": 1.0413, + "step": 27252 + }, + { + "epoch": 0.66, + "learning_rate": 5.6079052103493445e-06, + "loss": 1.2588, + "step": 27253 + }, + { + "epoch": 0.66, + "learning_rate": 5.607205476880827e-06, + "loss": 1.1175, + "step": 27254 + }, + { + "epoch": 0.66, + "learning_rate": 5.6065057700623645e-06, + "loss": 1.0111, + "step": 27255 + }, + { + "epoch": 0.66, + "learning_rate": 5.6058060898981956e-06, + "loss": 1.0517, + "step": 27256 + }, + { + "epoch": 0.66, + "learning_rate": 5.6051064363925676e-06, + "loss": 0.7223, + "step": 27257 + }, + { + "epoch": 0.66, + "learning_rate": 5.604406809549728e-06, + "loss": 1.073, + "step": 27258 + }, + { + "epoch": 0.66, + "learning_rate": 5.6037072093739155e-06, + "loss": 1.0942, + "step": 27259 + }, + { + "epoch": 0.66, + "learning_rate": 5.60300763586938e-06, + "loss": 1.1457, + "step": 27260 + }, + { + "epoch": 0.66, + "learning_rate": 5.60230808904036e-06, + "loss": 1.0363, + "step": 27261 + }, + { + "epoch": 0.66, + "learning_rate": 5.601608568891101e-06, + "loss": 0.9308, + "step": 27262 + }, + { + "epoch": 0.66, + "learning_rate": 5.600909075425854e-06, + "loss": 0.9607, + "step": 27263 + }, + { + "epoch": 0.66, + "learning_rate": 5.600209608648852e-06, + "loss": 0.9159, + "step": 27264 + }, + { + "epoch": 0.66, + "learning_rate": 5.599510168564344e-06, + "loss": 0.9922, + "step": 27265 + }, + { + "epoch": 0.66, + "learning_rate": 5.598810755176576e-06, + "loss": 0.8566, + "step": 27266 + }, + { + "epoch": 0.66, + "learning_rate": 5.5981113684897845e-06, + "loss": 0.7767, + "step": 27267 + }, + { + "epoch": 0.66, + "learning_rate": 5.597412008508221e-06, + "loss": 1.1161, + "step": 27268 + }, + { + "epoch": 0.66, + "learning_rate": 5.596712675236117e-06, + "loss": 1.2703, + "step": 27269 + }, + { + "epoch": 0.66, + "learning_rate": 5.596013368677722e-06, + "loss": 1.1297, + "step": 27270 + }, + { + "epoch": 0.66, + "learning_rate": 5.595314088837283e-06, + "loss": 1.0867, + "step": 27271 + }, + { + "epoch": 0.66, + "learning_rate": 5.594614835719032e-06, + "loss": 1.0887, + "step": 27272 + }, + { + "epoch": 0.66, + "learning_rate": 5.593915609327218e-06, + "loss": 0.9015, + "step": 27273 + }, + { + "epoch": 0.66, + "learning_rate": 5.5932164096660845e-06, + "loss": 0.6111, + "step": 27274 + }, + { + "epoch": 0.66, + "learning_rate": 5.592517236739867e-06, + "loss": 0.9833, + "step": 27275 + }, + { + "epoch": 0.66, + "learning_rate": 5.591818090552817e-06, + "loss": 1.2826, + "step": 27276 + }, + { + "epoch": 0.66, + "learning_rate": 5.591118971109161e-06, + "loss": 1.0253, + "step": 27277 + }, + { + "epoch": 0.66, + "learning_rate": 5.590419878413153e-06, + "loss": 1.293, + "step": 27278 + }, + { + "epoch": 0.66, + "learning_rate": 5.589720812469032e-06, + "loss": 1.1433, + "step": 27279 + }, + { + "epoch": 0.66, + "learning_rate": 5.589021773281036e-06, + "loss": 0.981, + "step": 27280 + }, + { + "epoch": 0.66, + "learning_rate": 5.588322760853406e-06, + "loss": 0.9399, + "step": 27281 + }, + { + "epoch": 0.66, + "learning_rate": 5.587623775190388e-06, + "loss": 1.1873, + "step": 27282 + }, + { + "epoch": 0.66, + "learning_rate": 5.586924816296215e-06, + "loss": 1.0562, + "step": 27283 + }, + { + "epoch": 0.66, + "learning_rate": 5.58622588417513e-06, + "loss": 0.9779, + "step": 27284 + }, + { + "epoch": 0.66, + "learning_rate": 5.585526978831379e-06, + "loss": 1.0268, + "step": 27285 + }, + { + "epoch": 0.66, + "learning_rate": 5.584828100269195e-06, + "loss": 0.7499, + "step": 27286 + }, + { + "epoch": 0.66, + "learning_rate": 5.584129248492822e-06, + "loss": 1.0187, + "step": 27287 + }, + { + "epoch": 0.66, + "learning_rate": 5.583430423506497e-06, + "loss": 0.8424, + "step": 27288 + }, + { + "epoch": 0.66, + "learning_rate": 5.582731625314459e-06, + "loss": 0.9642, + "step": 27289 + }, + { + "epoch": 0.66, + "learning_rate": 5.582032853920953e-06, + "loss": 0.9974, + "step": 27290 + }, + { + "epoch": 0.66, + "learning_rate": 5.581334109330211e-06, + "loss": 1.0154, + "step": 27291 + }, + { + "epoch": 0.66, + "learning_rate": 5.580635391546475e-06, + "loss": 0.8668, + "step": 27292 + }, + { + "epoch": 0.66, + "learning_rate": 5.579936700573989e-06, + "loss": 1.0021, + "step": 27293 + }, + { + "epoch": 0.66, + "learning_rate": 5.579238036416983e-06, + "loss": 1.0416, + "step": 27294 + }, + { + "epoch": 0.66, + "learning_rate": 5.578539399079705e-06, + "loss": 1.2488, + "step": 27295 + }, + { + "epoch": 0.66, + "learning_rate": 5.577840788566384e-06, + "loss": 0.8901, + "step": 27296 + }, + { + "epoch": 0.66, + "learning_rate": 5.577142204881263e-06, + "loss": 0.8534, + "step": 27297 + }, + { + "epoch": 0.66, + "learning_rate": 5.576443648028584e-06, + "loss": 0.9186, + "step": 27298 + }, + { + "epoch": 0.66, + "learning_rate": 5.575745118012577e-06, + "loss": 1.089, + "step": 27299 + }, + { + "epoch": 0.66, + "learning_rate": 5.5750466148374825e-06, + "loss": 0.9904, + "step": 27300 + }, + { + "epoch": 0.66, + "learning_rate": 5.574348138507544e-06, + "loss": 0.9351, + "step": 27301 + }, + { + "epoch": 0.66, + "learning_rate": 5.5736496890269895e-06, + "loss": 1.0111, + "step": 27302 + }, + { + "epoch": 0.66, + "learning_rate": 5.572951266400066e-06, + "loss": 0.978, + "step": 27303 + }, + { + "epoch": 0.66, + "learning_rate": 5.5722528706310035e-06, + "loss": 1.2817, + "step": 27304 + }, + { + "epoch": 0.66, + "learning_rate": 5.571554501724039e-06, + "loss": 0.9501, + "step": 27305 + }, + { + "epoch": 0.66, + "learning_rate": 5.570856159683418e-06, + "loss": 1.1597, + "step": 27306 + }, + { + "epoch": 0.66, + "learning_rate": 5.570157844513364e-06, + "loss": 0.9413, + "step": 27307 + }, + { + "epoch": 0.66, + "learning_rate": 5.569459556218122e-06, + "loss": 1.056, + "step": 27308 + }, + { + "epoch": 0.66, + "learning_rate": 5.568761294801931e-06, + "loss": 0.9553, + "step": 27309 + }, + { + "epoch": 0.66, + "learning_rate": 5.568063060269019e-06, + "loss": 1.0291, + "step": 27310 + }, + { + "epoch": 0.66, + "learning_rate": 5.567364852623629e-06, + "loss": 0.8447, + "step": 27311 + }, + { + "epoch": 0.66, + "learning_rate": 5.566666671869991e-06, + "loss": 0.701, + "step": 27312 + }, + { + "epoch": 0.66, + "learning_rate": 5.565968518012342e-06, + "loss": 1.3459, + "step": 27313 + }, + { + "epoch": 0.66, + "learning_rate": 5.565270391054925e-06, + "loss": 1.1697, + "step": 27314 + }, + { + "epoch": 0.66, + "learning_rate": 5.564572291001964e-06, + "loss": 1.0214, + "step": 27315 + }, + { + "epoch": 0.66, + "learning_rate": 5.5638742178577e-06, + "loss": 0.9244, + "step": 27316 + }, + { + "epoch": 0.66, + "learning_rate": 5.563176171626372e-06, + "loss": 0.9779, + "step": 27317 + }, + { + "epoch": 0.66, + "learning_rate": 5.562478152312205e-06, + "loss": 0.6678, + "step": 27318 + }, + { + "epoch": 0.66, + "learning_rate": 5.561780159919445e-06, + "loss": 1.2385, + "step": 27319 + }, + { + "epoch": 0.66, + "learning_rate": 5.561082194452316e-06, + "loss": 0.917, + "step": 27320 + }, + { + "epoch": 0.66, + "learning_rate": 5.560384255915059e-06, + "loss": 0.9493, + "step": 27321 + }, + { + "epoch": 0.66, + "learning_rate": 5.559686344311908e-06, + "loss": 1.0682, + "step": 27322 + }, + { + "epoch": 0.66, + "learning_rate": 5.558988459647092e-06, + "loss": 1.2178, + "step": 27323 + }, + { + "epoch": 0.66, + "learning_rate": 5.55829060192485e-06, + "loss": 1.0878, + "step": 27324 + }, + { + "epoch": 0.66, + "learning_rate": 5.557592771149417e-06, + "loss": 0.7674, + "step": 27325 + }, + { + "epoch": 0.66, + "learning_rate": 5.556894967325019e-06, + "loss": 0.9653, + "step": 27326 + }, + { + "epoch": 0.66, + "learning_rate": 5.5561971904559e-06, + "loss": 1.1431, + "step": 27327 + }, + { + "epoch": 0.66, + "learning_rate": 5.555499440546282e-06, + "loss": 1.1613, + "step": 27328 + }, + { + "epoch": 0.66, + "learning_rate": 5.554801717600404e-06, + "loss": 1.0745, + "step": 27329 + }, + { + "epoch": 0.66, + "learning_rate": 5.554104021622503e-06, + "loss": 0.9283, + "step": 27330 + }, + { + "epoch": 0.66, + "learning_rate": 5.553406352616801e-06, + "loss": 1.0952, + "step": 27331 + }, + { + "epoch": 0.66, + "learning_rate": 5.55270871058754e-06, + "loss": 1.0551, + "step": 27332 + }, + { + "epoch": 0.66, + "learning_rate": 5.552011095538947e-06, + "loss": 1.1057, + "step": 27333 + }, + { + "epoch": 0.66, + "learning_rate": 5.551313507475256e-06, + "loss": 1.0676, + "step": 27334 + }, + { + "epoch": 0.66, + "learning_rate": 5.550615946400706e-06, + "loss": 1.0591, + "step": 27335 + }, + { + "epoch": 0.66, + "learning_rate": 5.549918412319516e-06, + "loss": 1.0293, + "step": 27336 + }, + { + "epoch": 0.66, + "learning_rate": 5.549220905235926e-06, + "loss": 1.3012, + "step": 27337 + }, + { + "epoch": 0.66, + "learning_rate": 5.5485234251541685e-06, + "loss": 1.0944, + "step": 27338 + }, + { + "epoch": 0.66, + "learning_rate": 5.5478259720784685e-06, + "loss": 0.9153, + "step": 27339 + }, + { + "epoch": 0.66, + "learning_rate": 5.547128546013065e-06, + "loss": 1.1155, + "step": 27340 + }, + { + "epoch": 0.66, + "learning_rate": 5.546431146962182e-06, + "loss": 0.9605, + "step": 27341 + }, + { + "epoch": 0.66, + "learning_rate": 5.545733774930053e-06, + "loss": 0.6512, + "step": 27342 + }, + { + "epoch": 0.66, + "learning_rate": 5.545036429920915e-06, + "loss": 1.2825, + "step": 27343 + }, + { + "epoch": 0.66, + "learning_rate": 5.544339111938987e-06, + "loss": 0.8042, + "step": 27344 + }, + { + "epoch": 0.66, + "learning_rate": 5.543641820988508e-06, + "loss": 1.0538, + "step": 27345 + }, + { + "epoch": 0.66, + "learning_rate": 5.5429445570737085e-06, + "loss": 0.9611, + "step": 27346 + }, + { + "epoch": 0.66, + "learning_rate": 5.542247320198811e-06, + "loss": 0.9544, + "step": 27347 + }, + { + "epoch": 0.66, + "learning_rate": 5.541550110368055e-06, + "loss": 0.9757, + "step": 27348 + }, + { + "epoch": 0.66, + "learning_rate": 5.540852927585663e-06, + "loss": 1.0983, + "step": 27349 + }, + { + "epoch": 0.66, + "learning_rate": 5.540155771855866e-06, + "loss": 0.9911, + "step": 27350 + }, + { + "epoch": 0.66, + "learning_rate": 5.5394586431829e-06, + "loss": 1.0245, + "step": 27351 + }, + { + "epoch": 0.66, + "learning_rate": 5.538761541570985e-06, + "loss": 1.021, + "step": 27352 + }, + { + "epoch": 0.66, + "learning_rate": 5.538064467024354e-06, + "loss": 0.8166, + "step": 27353 + }, + { + "epoch": 0.66, + "learning_rate": 5.537367419547242e-06, + "loss": 0.9157, + "step": 27354 + }, + { + "epoch": 0.66, + "learning_rate": 5.536670399143867e-06, + "loss": 1.2591, + "step": 27355 + }, + { + "epoch": 0.66, + "learning_rate": 5.535973405818466e-06, + "loss": 1.0464, + "step": 27356 + }, + { + "epoch": 0.66, + "learning_rate": 5.5352764395752615e-06, + "loss": 1.1282, + "step": 27357 + }, + { + "epoch": 0.66, + "learning_rate": 5.5345795004184835e-06, + "loss": 1.1903, + "step": 27358 + }, + { + "epoch": 0.66, + "learning_rate": 5.533882588352366e-06, + "loss": 0.9372, + "step": 27359 + }, + { + "epoch": 0.66, + "learning_rate": 5.533185703381128e-06, + "loss": 0.9586, + "step": 27360 + }, + { + "epoch": 0.66, + "learning_rate": 5.532488845509002e-06, + "loss": 0.777, + "step": 27361 + }, + { + "epoch": 0.66, + "learning_rate": 5.53179201474022e-06, + "loss": 1.2024, + "step": 27362 + }, + { + "epoch": 0.66, + "learning_rate": 5.531095211078999e-06, + "loss": 1.1145, + "step": 27363 + }, + { + "epoch": 0.66, + "learning_rate": 5.530398434529574e-06, + "loss": 1.0372, + "step": 27364 + }, + { + "epoch": 0.66, + "learning_rate": 5.529701685096174e-06, + "loss": 1.0873, + "step": 27365 + }, + { + "epoch": 0.66, + "learning_rate": 5.5290049627830175e-06, + "loss": 1.0641, + "step": 27366 + }, + { + "epoch": 0.66, + "learning_rate": 5.5283082675943415e-06, + "loss": 0.844, + "step": 27367 + }, + { + "epoch": 0.66, + "learning_rate": 5.527611599534363e-06, + "loss": 0.6799, + "step": 27368 + }, + { + "epoch": 0.66, + "learning_rate": 5.526914958607312e-06, + "loss": 0.9082, + "step": 27369 + }, + { + "epoch": 0.66, + "learning_rate": 5.526218344817421e-06, + "loss": 1.0581, + "step": 27370 + }, + { + "epoch": 0.66, + "learning_rate": 5.525521758168907e-06, + "loss": 1.0703, + "step": 27371 + }, + { + "epoch": 0.66, + "learning_rate": 5.524825198665999e-06, + "loss": 1.1633, + "step": 27372 + }, + { + "epoch": 0.66, + "learning_rate": 5.524128666312929e-06, + "loss": 1.029, + "step": 27373 + }, + { + "epoch": 0.66, + "learning_rate": 5.523432161113913e-06, + "loss": 1.0608, + "step": 27374 + }, + { + "epoch": 0.66, + "learning_rate": 5.522735683073186e-06, + "loss": 1.1752, + "step": 27375 + }, + { + "epoch": 0.66, + "learning_rate": 5.522039232194963e-06, + "loss": 1.0204, + "step": 27376 + }, + { + "epoch": 0.66, + "learning_rate": 5.5213428084834765e-06, + "loss": 1.0437, + "step": 27377 + }, + { + "epoch": 0.66, + "learning_rate": 5.520646411942952e-06, + "loss": 1.095, + "step": 27378 + }, + { + "epoch": 0.66, + "learning_rate": 5.519950042577609e-06, + "loss": 1.0366, + "step": 27379 + }, + { + "epoch": 0.66, + "learning_rate": 5.519253700391676e-06, + "loss": 0.8871, + "step": 27380 + }, + { + "epoch": 0.66, + "learning_rate": 5.518557385389379e-06, + "loss": 0.7762, + "step": 27381 + }, + { + "epoch": 0.66, + "learning_rate": 5.517861097574937e-06, + "loss": 0.7995, + "step": 27382 + }, + { + "epoch": 0.66, + "learning_rate": 5.517164836952582e-06, + "loss": 1.0365, + "step": 27383 + }, + { + "epoch": 0.66, + "learning_rate": 5.516468603526529e-06, + "loss": 1.1493, + "step": 27384 + }, + { + "epoch": 0.66, + "learning_rate": 5.515772397301007e-06, + "loss": 0.9163, + "step": 27385 + }, + { + "epoch": 0.66, + "learning_rate": 5.515076218280242e-06, + "loss": 1.2261, + "step": 27386 + }, + { + "epoch": 0.66, + "learning_rate": 5.514380066468452e-06, + "loss": 1.0786, + "step": 27387 + }, + { + "epoch": 0.66, + "learning_rate": 5.5136839418698625e-06, + "loss": 0.8064, + "step": 27388 + }, + { + "epoch": 0.66, + "learning_rate": 5.5129878444887015e-06, + "loss": 1.1578, + "step": 27389 + }, + { + "epoch": 0.66, + "learning_rate": 5.512291774329184e-06, + "loss": 1.3678, + "step": 27390 + }, + { + "epoch": 0.66, + "learning_rate": 5.5115957313955396e-06, + "loss": 1.2, + "step": 27391 + }, + { + "epoch": 0.66, + "learning_rate": 5.510899715691984e-06, + "loss": 0.9487, + "step": 27392 + }, + { + "epoch": 0.66, + "learning_rate": 5.510203727222745e-06, + "loss": 1.1382, + "step": 27393 + }, + { + "epoch": 0.66, + "learning_rate": 5.509507765992048e-06, + "loss": 0.8174, + "step": 27394 + }, + { + "epoch": 0.66, + "learning_rate": 5.5088118320041075e-06, + "loss": 1.147, + "step": 27395 + }, + { + "epoch": 0.66, + "learning_rate": 5.5081159252631485e-06, + "loss": 1.0288, + "step": 27396 + }, + { + "epoch": 0.66, + "learning_rate": 5.507420045773398e-06, + "loss": 0.9, + "step": 27397 + }, + { + "epoch": 0.66, + "learning_rate": 5.506724193539068e-06, + "loss": 1.0786, + "step": 27398 + }, + { + "epoch": 0.66, + "learning_rate": 5.5060283685643915e-06, + "loss": 0.922, + "step": 27399 + }, + { + "epoch": 0.66, + "learning_rate": 5.505332570853579e-06, + "loss": 1.0178, + "step": 27400 + }, + { + "epoch": 0.66, + "learning_rate": 5.504636800410856e-06, + "loss": 1.0306, + "step": 27401 + }, + { + "epoch": 0.66, + "learning_rate": 5.503941057240448e-06, + "loss": 1.1145, + "step": 27402 + }, + { + "epoch": 0.66, + "learning_rate": 5.50324534134657e-06, + "loss": 1.0281, + "step": 27403 + }, + { + "epoch": 0.66, + "learning_rate": 5.502549652733442e-06, + "loss": 1.1545, + "step": 27404 + }, + { + "epoch": 0.66, + "learning_rate": 5.501853991405293e-06, + "loss": 0.7231, + "step": 27405 + }, + { + "epoch": 0.66, + "learning_rate": 5.501158357366334e-06, + "loss": 1.1621, + "step": 27406 + }, + { + "epoch": 0.66, + "learning_rate": 5.500462750620792e-06, + "loss": 1.0185, + "step": 27407 + }, + { + "epoch": 0.66, + "learning_rate": 5.499767171172882e-06, + "loss": 0.9156, + "step": 27408 + }, + { + "epoch": 0.66, + "learning_rate": 5.499071619026825e-06, + "loss": 1.1851, + "step": 27409 + }, + { + "epoch": 0.66, + "learning_rate": 5.498376094186846e-06, + "loss": 1.1687, + "step": 27410 + }, + { + "epoch": 0.66, + "learning_rate": 5.4976805966571555e-06, + "loss": 0.8706, + "step": 27411 + }, + { + "epoch": 0.66, + "learning_rate": 5.496985126441978e-06, + "loss": 1.0162, + "step": 27412 + }, + { + "epoch": 0.66, + "learning_rate": 5.496289683545537e-06, + "loss": 1.0833, + "step": 27413 + }, + { + "epoch": 0.66, + "learning_rate": 5.495594267972043e-06, + "loss": 1.1018, + "step": 27414 + }, + { + "epoch": 0.66, + "learning_rate": 5.49489887972572e-06, + "loss": 1.1265, + "step": 27415 + }, + { + "epoch": 0.66, + "learning_rate": 5.494203518810788e-06, + "loss": 0.8479, + "step": 27416 + }, + { + "epoch": 0.66, + "learning_rate": 5.493508185231461e-06, + "loss": 1.352, + "step": 27417 + }, + { + "epoch": 0.66, + "learning_rate": 5.492812878991963e-06, + "loss": 0.9226, + "step": 27418 + }, + { + "epoch": 0.66, + "learning_rate": 5.492117600096505e-06, + "loss": 0.8759, + "step": 27419 + }, + { + "epoch": 0.66, + "learning_rate": 5.49142234854931e-06, + "loss": 0.9364, + "step": 27420 + }, + { + "epoch": 0.66, + "learning_rate": 5.490727124354599e-06, + "loss": 1.1531, + "step": 27421 + }, + { + "epoch": 0.66, + "learning_rate": 5.490031927516581e-06, + "loss": 1.0447, + "step": 27422 + }, + { + "epoch": 0.66, + "learning_rate": 5.48933675803948e-06, + "loss": 0.9616, + "step": 27423 + }, + { + "epoch": 0.66, + "learning_rate": 5.488641615927515e-06, + "loss": 1.0853, + "step": 27424 + }, + { + "epoch": 0.66, + "learning_rate": 5.487946501184895e-06, + "loss": 1.0802, + "step": 27425 + }, + { + "epoch": 0.66, + "learning_rate": 5.487251413815848e-06, + "loss": 1.12, + "step": 27426 + }, + { + "epoch": 0.66, + "learning_rate": 5.486556353824581e-06, + "loss": 1.2123, + "step": 27427 + }, + { + "epoch": 0.66, + "learning_rate": 5.485861321215316e-06, + "loss": 1.1753, + "step": 27428 + }, + { + "epoch": 0.66, + "learning_rate": 5.48516631599227e-06, + "loss": 1.0122, + "step": 27429 + }, + { + "epoch": 0.66, + "learning_rate": 5.484471338159656e-06, + "loss": 0.7836, + "step": 27430 + }, + { + "epoch": 0.66, + "learning_rate": 5.483776387721691e-06, + "loss": 1.1492, + "step": 27431 + }, + { + "epoch": 0.66, + "learning_rate": 5.483081464682597e-06, + "loss": 0.8442, + "step": 27432 + }, + { + "epoch": 0.66, + "learning_rate": 5.48238656904658e-06, + "loss": 1.2684, + "step": 27433 + }, + { + "epoch": 0.66, + "learning_rate": 5.481691700817867e-06, + "loss": 1.0745, + "step": 27434 + }, + { + "epoch": 0.66, + "learning_rate": 5.480996860000664e-06, + "loss": 1.0643, + "step": 27435 + }, + { + "epoch": 0.66, + "learning_rate": 5.4803020465991884e-06, + "loss": 0.7765, + "step": 27436 + }, + { + "epoch": 0.66, + "learning_rate": 5.479607260617662e-06, + "loss": 1.0868, + "step": 27437 + }, + { + "epoch": 0.66, + "learning_rate": 5.478912502060291e-06, + "loss": 0.8869, + "step": 27438 + }, + { + "epoch": 0.66, + "learning_rate": 5.478217770931294e-06, + "loss": 0.7713, + "step": 27439 + }, + { + "epoch": 0.66, + "learning_rate": 5.477523067234891e-06, + "loss": 1.0891, + "step": 27440 + }, + { + "epoch": 0.66, + "learning_rate": 5.476828390975287e-06, + "loss": 1.019, + "step": 27441 + }, + { + "epoch": 0.66, + "learning_rate": 5.476133742156706e-06, + "loss": 0.8934, + "step": 27442 + }, + { + "epoch": 0.66, + "learning_rate": 5.475439120783353e-06, + "loss": 1.0649, + "step": 27443 + }, + { + "epoch": 0.66, + "learning_rate": 5.4747445268594464e-06, + "loss": 1.2139, + "step": 27444 + }, + { + "epoch": 0.66, + "learning_rate": 5.474049960389205e-06, + "loss": 1.0508, + "step": 27445 + }, + { + "epoch": 0.66, + "learning_rate": 5.473355421376834e-06, + "loss": 1.0974, + "step": 27446 + }, + { + "epoch": 0.66, + "learning_rate": 5.4726609098265505e-06, + "loss": 1.0284, + "step": 27447 + }, + { + "epoch": 0.66, + "learning_rate": 5.471966425742572e-06, + "loss": 1.1266, + "step": 27448 + }, + { + "epoch": 0.66, + "learning_rate": 5.471271969129104e-06, + "loss": 1.2037, + "step": 27449 + }, + { + "epoch": 0.66, + "learning_rate": 5.470577539990369e-06, + "loss": 1.2457, + "step": 27450 + }, + { + "epoch": 0.66, + "learning_rate": 5.469883138330569e-06, + "loss": 0.7847, + "step": 27451 + }, + { + "epoch": 0.66, + "learning_rate": 5.469188764153924e-06, + "loss": 0.9431, + "step": 27452 + }, + { + "epoch": 0.66, + "learning_rate": 5.468494417464649e-06, + "loss": 0.9099, + "step": 27453 + }, + { + "epoch": 0.66, + "learning_rate": 5.467800098266948e-06, + "loss": 1.1564, + "step": 27454 + }, + { + "epoch": 0.66, + "learning_rate": 5.467105806565038e-06, + "loss": 1.3895, + "step": 27455 + }, + { + "epoch": 0.66, + "learning_rate": 5.4664115423631345e-06, + "loss": 1.1853, + "step": 27456 + }, + { + "epoch": 0.66, + "learning_rate": 5.465717305665441e-06, + "loss": 1.0945, + "step": 27457 + }, + { + "epoch": 0.66, + "learning_rate": 5.465023096476175e-06, + "loss": 1.0115, + "step": 27458 + }, + { + "epoch": 0.66, + "learning_rate": 5.464328914799552e-06, + "loss": 1.0191, + "step": 27459 + }, + { + "epoch": 0.66, + "learning_rate": 5.463634760639773e-06, + "loss": 1.0265, + "step": 27460 + }, + { + "epoch": 0.66, + "learning_rate": 5.462940634001059e-06, + "loss": 0.9278, + "step": 27461 + }, + { + "epoch": 0.66, + "learning_rate": 5.462246534887614e-06, + "loss": 1.1752, + "step": 27462 + }, + { + "epoch": 0.66, + "learning_rate": 5.461552463303652e-06, + "loss": 1.0357, + "step": 27463 + }, + { + "epoch": 0.66, + "learning_rate": 5.460858419253388e-06, + "loss": 1.0902, + "step": 27464 + }, + { + "epoch": 0.66, + "learning_rate": 5.4601644027410235e-06, + "loss": 1.0965, + "step": 27465 + }, + { + "epoch": 0.66, + "learning_rate": 5.459470413770773e-06, + "loss": 0.9764, + "step": 27466 + }, + { + "epoch": 0.66, + "learning_rate": 5.458776452346852e-06, + "loss": 1.1103, + "step": 27467 + }, + { + "epoch": 0.66, + "learning_rate": 5.458082518473463e-06, + "loss": 1.0619, + "step": 27468 + }, + { + "epoch": 0.66, + "learning_rate": 5.457388612154823e-06, + "loss": 1.0942, + "step": 27469 + }, + { + "epoch": 0.66, + "learning_rate": 5.456694733395133e-06, + "loss": 1.0559, + "step": 27470 + }, + { + "epoch": 0.66, + "learning_rate": 5.456000882198608e-06, + "loss": 1.0633, + "step": 27471 + }, + { + "epoch": 0.66, + "learning_rate": 5.45530705856946e-06, + "loss": 1.1947, + "step": 27472 + }, + { + "epoch": 0.66, + "learning_rate": 5.454613262511892e-06, + "loss": 1.0357, + "step": 27473 + }, + { + "epoch": 0.66, + "learning_rate": 5.453919494030116e-06, + "loss": 1.2328, + "step": 27474 + }, + { + "epoch": 0.66, + "learning_rate": 5.4532257531283436e-06, + "loss": 1.1937, + "step": 27475 + }, + { + "epoch": 0.66, + "learning_rate": 5.45253203981078e-06, + "loss": 1.2018, + "step": 27476 + }, + { + "epoch": 0.66, + "learning_rate": 5.451838354081636e-06, + "loss": 1.0898, + "step": 27477 + }, + { + "epoch": 0.66, + "learning_rate": 5.451144695945116e-06, + "loss": 1.1899, + "step": 27478 + }, + { + "epoch": 0.66, + "learning_rate": 5.450451065405432e-06, + "loss": 1.0509, + "step": 27479 + }, + { + "epoch": 0.66, + "learning_rate": 5.449757462466795e-06, + "loss": 1.0339, + "step": 27480 + }, + { + "epoch": 0.66, + "learning_rate": 5.449063887133404e-06, + "loss": 1.1213, + "step": 27481 + }, + { + "epoch": 0.66, + "learning_rate": 5.448370339409473e-06, + "loss": 1.0303, + "step": 27482 + }, + { + "epoch": 0.66, + "learning_rate": 5.447676819299212e-06, + "loss": 1.0856, + "step": 27483 + }, + { + "epoch": 0.66, + "learning_rate": 5.446983326806823e-06, + "loss": 1.1734, + "step": 27484 + }, + { + "epoch": 0.66, + "learning_rate": 5.446289861936517e-06, + "loss": 1.0385, + "step": 27485 + }, + { + "epoch": 0.66, + "learning_rate": 5.445596424692497e-06, + "loss": 0.8888, + "step": 27486 + }, + { + "epoch": 0.66, + "learning_rate": 5.4449030150789695e-06, + "loss": 0.8859, + "step": 27487 + }, + { + "epoch": 0.66, + "learning_rate": 5.444209633100151e-06, + "loss": 1.162, + "step": 27488 + }, + { + "epoch": 0.66, + "learning_rate": 5.4435162787602345e-06, + "loss": 1.08, + "step": 27489 + }, + { + "epoch": 0.66, + "learning_rate": 5.4428229520634335e-06, + "loss": 0.9294, + "step": 27490 + }, + { + "epoch": 0.66, + "learning_rate": 5.442129653013959e-06, + "loss": 1.0454, + "step": 27491 + }, + { + "epoch": 0.66, + "learning_rate": 5.441436381616008e-06, + "loss": 1.1817, + "step": 27492 + }, + { + "epoch": 0.66, + "learning_rate": 5.440743137873792e-06, + "loss": 1.2693, + "step": 27493 + }, + { + "epoch": 0.66, + "learning_rate": 5.440049921791512e-06, + "loss": 1.0464, + "step": 27494 + }, + { + "epoch": 0.66, + "learning_rate": 5.439356733373378e-06, + "loss": 1.1175, + "step": 27495 + }, + { + "epoch": 0.66, + "learning_rate": 5.438663572623597e-06, + "loss": 0.9829, + "step": 27496 + }, + { + "epoch": 0.66, + "learning_rate": 5.437970439546367e-06, + "loss": 1.2768, + "step": 27497 + }, + { + "epoch": 0.66, + "learning_rate": 5.437277334145897e-06, + "loss": 1.3721, + "step": 27498 + }, + { + "epoch": 0.66, + "learning_rate": 5.436584256426392e-06, + "loss": 0.987, + "step": 27499 + }, + { + "epoch": 0.66, + "learning_rate": 5.435891206392059e-06, + "loss": 0.8104, + "step": 27500 + }, + { + "epoch": 0.66, + "learning_rate": 5.435198184047102e-06, + "loss": 1.0937, + "step": 27501 + }, + { + "epoch": 0.66, + "learning_rate": 5.434505189395721e-06, + "loss": 0.8901, + "step": 27502 + }, + { + "epoch": 0.66, + "learning_rate": 5.433812222442123e-06, + "loss": 1.1407, + "step": 27503 + }, + { + "epoch": 0.66, + "learning_rate": 5.433119283190517e-06, + "loss": 1.0609, + "step": 27504 + }, + { + "epoch": 0.66, + "learning_rate": 5.432426371645097e-06, + "loss": 0.9574, + "step": 27505 + }, + { + "epoch": 0.66, + "learning_rate": 5.431733487810076e-06, + "loss": 0.8984, + "step": 27506 + }, + { + "epoch": 0.66, + "learning_rate": 5.43104063168965e-06, + "loss": 1.1502, + "step": 27507 + }, + { + "epoch": 0.66, + "learning_rate": 5.430347803288025e-06, + "loss": 1.0689, + "step": 27508 + }, + { + "epoch": 0.66, + "learning_rate": 5.42965500260941e-06, + "loss": 1.1382, + "step": 27509 + }, + { + "epoch": 0.66, + "learning_rate": 5.428962229657999e-06, + "loss": 1.1143, + "step": 27510 + }, + { + "epoch": 0.66, + "learning_rate": 5.428269484437998e-06, + "loss": 1.1267, + "step": 27511 + }, + { + "epoch": 0.66, + "learning_rate": 5.427576766953615e-06, + "loss": 1.0368, + "step": 27512 + }, + { + "epoch": 0.66, + "learning_rate": 5.426884077209045e-06, + "loss": 1.0629, + "step": 27513 + }, + { + "epoch": 0.66, + "learning_rate": 5.426191415208497e-06, + "loss": 0.8078, + "step": 27514 + }, + { + "epoch": 0.66, + "learning_rate": 5.425498780956167e-06, + "loss": 0.8333, + "step": 27515 + }, + { + "epoch": 0.66, + "learning_rate": 5.424806174456258e-06, + "loss": 0.8584, + "step": 27516 + }, + { + "epoch": 0.66, + "learning_rate": 5.424113595712979e-06, + "loss": 0.9213, + "step": 27517 + }, + { + "epoch": 0.66, + "learning_rate": 5.423421044730521e-06, + "loss": 1.1144, + "step": 27518 + }, + { + "epoch": 0.66, + "learning_rate": 5.422728521513093e-06, + "loss": 1.2885, + "step": 27519 + }, + { + "epoch": 0.66, + "learning_rate": 5.422036026064899e-06, + "loss": 1.2457, + "step": 27520 + }, + { + "epoch": 0.66, + "learning_rate": 5.421343558390129e-06, + "loss": 0.9953, + "step": 27521 + }, + { + "epoch": 0.66, + "learning_rate": 5.4206511184929964e-06, + "loss": 1.0337, + "step": 27522 + }, + { + "epoch": 0.66, + "learning_rate": 5.4199587063776925e-06, + "loss": 0.9694, + "step": 27523 + }, + { + "epoch": 0.66, + "learning_rate": 5.41926632204842e-06, + "loss": 1.1417, + "step": 27524 + }, + { + "epoch": 0.66, + "learning_rate": 5.418573965509387e-06, + "loss": 1.2476, + "step": 27525 + }, + { + "epoch": 0.66, + "learning_rate": 5.417881636764784e-06, + "loss": 1.1933, + "step": 27526 + }, + { + "epoch": 0.66, + "learning_rate": 5.417189335818815e-06, + "loss": 0.9957, + "step": 27527 + }, + { + "epoch": 0.66, + "learning_rate": 5.416497062675684e-06, + "loss": 1.1823, + "step": 27528 + }, + { + "epoch": 0.66, + "learning_rate": 5.415804817339584e-06, + "loss": 1.0158, + "step": 27529 + }, + { + "epoch": 0.66, + "learning_rate": 5.41511259981472e-06, + "loss": 1.251, + "step": 27530 + }, + { + "epoch": 0.66, + "learning_rate": 5.414420410105287e-06, + "loss": 1.1681, + "step": 27531 + }, + { + "epoch": 0.66, + "learning_rate": 5.413728248215488e-06, + "loss": 1.0263, + "step": 27532 + }, + { + "epoch": 0.66, + "learning_rate": 5.413036114149523e-06, + "loss": 1.3015, + "step": 27533 + }, + { + "epoch": 0.66, + "learning_rate": 5.412344007911585e-06, + "loss": 1.2932, + "step": 27534 + }, + { + "epoch": 0.66, + "learning_rate": 5.411651929505879e-06, + "loss": 1.1517, + "step": 27535 + }, + { + "epoch": 0.66, + "learning_rate": 5.410959878936603e-06, + "loss": 1.0514, + "step": 27536 + }, + { + "epoch": 0.66, + "learning_rate": 5.410267856207952e-06, + "loss": 1.0494, + "step": 27537 + }, + { + "epoch": 0.66, + "learning_rate": 5.409575861324124e-06, + "loss": 1.2833, + "step": 27538 + }, + { + "epoch": 0.66, + "learning_rate": 5.4088838942893255e-06, + "loss": 1.0255, + "step": 27539 + }, + { + "epoch": 0.66, + "learning_rate": 5.408191955107743e-06, + "loss": 1.1244, + "step": 27540 + }, + { + "epoch": 0.66, + "learning_rate": 5.407500043783585e-06, + "loss": 0.9381, + "step": 27541 + }, + { + "epoch": 0.66, + "learning_rate": 5.4068081603210395e-06, + "loss": 1.1273, + "step": 27542 + }, + { + "epoch": 0.66, + "learning_rate": 5.406116304724309e-06, + "loss": 1.1356, + "step": 27543 + }, + { + "epoch": 0.66, + "learning_rate": 5.405424476997594e-06, + "loss": 1.0074, + "step": 27544 + }, + { + "epoch": 0.66, + "learning_rate": 5.404732677145084e-06, + "loss": 1.0573, + "step": 27545 + }, + { + "epoch": 0.66, + "learning_rate": 5.404040905170979e-06, + "loss": 1.0825, + "step": 27546 + }, + { + "epoch": 0.66, + "learning_rate": 5.403349161079483e-06, + "loss": 1.0147, + "step": 27547 + }, + { + "epoch": 0.66, + "learning_rate": 5.402657444874782e-06, + "loss": 1.0778, + "step": 27548 + }, + { + "epoch": 0.66, + "learning_rate": 5.401965756561078e-06, + "loss": 1.132, + "step": 27549 + }, + { + "epoch": 0.66, + "learning_rate": 5.401274096142566e-06, + "loss": 0.8872, + "step": 27550 + }, + { + "epoch": 0.66, + "learning_rate": 5.400582463623441e-06, + "loss": 1.0619, + "step": 27551 + }, + { + "epoch": 0.66, + "learning_rate": 5.399890859007903e-06, + "loss": 1.2012, + "step": 27552 + }, + { + "epoch": 0.66, + "learning_rate": 5.399199282300142e-06, + "loss": 1.2534, + "step": 27553 + }, + { + "epoch": 0.66, + "learning_rate": 5.3985077335043565e-06, + "loss": 0.9039, + "step": 27554 + }, + { + "epoch": 0.66, + "learning_rate": 5.397816212624748e-06, + "loss": 1.0538, + "step": 27555 + }, + { + "epoch": 0.66, + "learning_rate": 5.397124719665501e-06, + "loss": 0.9876, + "step": 27556 + }, + { + "epoch": 0.66, + "learning_rate": 5.3964332546308196e-06, + "loss": 1.1078, + "step": 27557 + }, + { + "epoch": 0.66, + "learning_rate": 5.39574181752489e-06, + "loss": 1.0087, + "step": 27558 + }, + { + "epoch": 0.66, + "learning_rate": 5.395050408351912e-06, + "loss": 0.9606, + "step": 27559 + }, + { + "epoch": 0.66, + "learning_rate": 5.394359027116085e-06, + "loss": 0.7182, + "step": 27560 + }, + { + "epoch": 0.66, + "learning_rate": 5.3936676738215945e-06, + "loss": 1.142, + "step": 27561 + }, + { + "epoch": 0.66, + "learning_rate": 5.3929763484726385e-06, + "loss": 1.001, + "step": 27562 + }, + { + "epoch": 0.66, + "learning_rate": 5.392285051073413e-06, + "loss": 0.8607, + "step": 27563 + }, + { + "epoch": 0.66, + "learning_rate": 5.391593781628109e-06, + "loss": 1.1391, + "step": 27564 + }, + { + "epoch": 0.66, + "learning_rate": 5.390902540140926e-06, + "loss": 1.3039, + "step": 27565 + }, + { + "epoch": 0.66, + "learning_rate": 5.390211326616048e-06, + "loss": 0.9628, + "step": 27566 + }, + { + "epoch": 0.66, + "learning_rate": 5.3895201410576736e-06, + "loss": 1.0607, + "step": 27567 + }, + { + "epoch": 0.66, + "learning_rate": 5.3888289834700006e-06, + "loss": 0.994, + "step": 27568 + }, + { + "epoch": 0.66, + "learning_rate": 5.388137853857212e-06, + "loss": 1.1273, + "step": 27569 + }, + { + "epoch": 0.66, + "learning_rate": 5.387446752223507e-06, + "loss": 1.2652, + "step": 27570 + }, + { + "epoch": 0.66, + "learning_rate": 5.386755678573082e-06, + "loss": 1.1797, + "step": 27571 + }, + { + "epoch": 0.66, + "learning_rate": 5.386064632910122e-06, + "loss": 1.0831, + "step": 27572 + }, + { + "epoch": 0.66, + "learning_rate": 5.385373615238825e-06, + "loss": 1.2178, + "step": 27573 + }, + { + "epoch": 0.66, + "learning_rate": 5.384682625563378e-06, + "loss": 1.0016, + "step": 27574 + }, + { + "epoch": 0.66, + "learning_rate": 5.3839916638879755e-06, + "loss": 1.1682, + "step": 27575 + }, + { + "epoch": 0.66, + "learning_rate": 5.383300730216814e-06, + "loss": 1.0965, + "step": 27576 + }, + { + "epoch": 0.66, + "learning_rate": 5.382609824554077e-06, + "loss": 1.1014, + "step": 27577 + }, + { + "epoch": 0.66, + "learning_rate": 5.38191894690396e-06, + "loss": 0.8239, + "step": 27578 + }, + { + "epoch": 0.66, + "learning_rate": 5.381228097270658e-06, + "loss": 0.8798, + "step": 27579 + }, + { + "epoch": 0.66, + "learning_rate": 5.380537275658355e-06, + "loss": 0.8544, + "step": 27580 + }, + { + "epoch": 0.66, + "learning_rate": 5.379846482071247e-06, + "loss": 1.0877, + "step": 27581 + }, + { + "epoch": 0.66, + "learning_rate": 5.379155716513526e-06, + "loss": 0.9583, + "step": 27582 + }, + { + "epoch": 0.66, + "learning_rate": 5.378464978989377e-06, + "loss": 0.9674, + "step": 27583 + }, + { + "epoch": 0.66, + "learning_rate": 5.377774269502999e-06, + "loss": 1.062, + "step": 27584 + }, + { + "epoch": 0.66, + "learning_rate": 5.377083588058571e-06, + "loss": 1.0922, + "step": 27585 + }, + { + "epoch": 0.66, + "learning_rate": 5.376392934660292e-06, + "loss": 0.8856, + "step": 27586 + }, + { + "epoch": 0.66, + "learning_rate": 5.375702309312352e-06, + "loss": 0.9287, + "step": 27587 + }, + { + "epoch": 0.66, + "learning_rate": 5.375011712018935e-06, + "loss": 1.1916, + "step": 27588 + }, + { + "epoch": 0.66, + "learning_rate": 5.374321142784233e-06, + "loss": 1.0779, + "step": 27589 + }, + { + "epoch": 0.66, + "learning_rate": 5.373630601612443e-06, + "loss": 0.965, + "step": 27590 + }, + { + "epoch": 0.66, + "learning_rate": 5.372940088507741e-06, + "loss": 1.106, + "step": 27591 + }, + { + "epoch": 0.66, + "learning_rate": 5.372249603474329e-06, + "loss": 1.1416, + "step": 27592 + }, + { + "epoch": 0.66, + "learning_rate": 5.371559146516385e-06, + "loss": 1.044, + "step": 27593 + }, + { + "epoch": 0.66, + "learning_rate": 5.370868717638105e-06, + "loss": 0.914, + "step": 27594 + }, + { + "epoch": 0.66, + "learning_rate": 5.370178316843677e-06, + "loss": 1.0187, + "step": 27595 + }, + { + "epoch": 0.66, + "learning_rate": 5.369487944137286e-06, + "loss": 0.9234, + "step": 27596 + }, + { + "epoch": 0.66, + "learning_rate": 5.368797599523121e-06, + "loss": 1.0139, + "step": 27597 + }, + { + "epoch": 0.66, + "learning_rate": 5.368107283005377e-06, + "loss": 0.9528, + "step": 27598 + }, + { + "epoch": 0.66, + "learning_rate": 5.367416994588232e-06, + "loss": 1.1106, + "step": 27599 + }, + { + "epoch": 0.66, + "learning_rate": 5.366726734275883e-06, + "loss": 1.0312, + "step": 27600 + }, + { + "epoch": 0.66, + "learning_rate": 5.366036502072509e-06, + "loss": 0.8004, + "step": 27601 + }, + { + "epoch": 0.66, + "learning_rate": 5.365346297982301e-06, + "loss": 0.8097, + "step": 27602 + }, + { + "epoch": 0.66, + "learning_rate": 5.364656122009451e-06, + "loss": 1.2279, + "step": 27603 + }, + { + "epoch": 0.66, + "learning_rate": 5.363965974158139e-06, + "loss": 1.3399, + "step": 27604 + }, + { + "epoch": 0.66, + "learning_rate": 5.363275854432553e-06, + "loss": 1.2704, + "step": 27605 + }, + { + "epoch": 0.66, + "learning_rate": 5.362585762836887e-06, + "loss": 1.0901, + "step": 27606 + }, + { + "epoch": 0.66, + "learning_rate": 5.3618956993753195e-06, + "loss": 1.2022, + "step": 27607 + }, + { + "epoch": 0.66, + "learning_rate": 5.3612056640520415e-06, + "loss": 0.9261, + "step": 27608 + }, + { + "epoch": 0.66, + "learning_rate": 5.360515656871235e-06, + "loss": 1.0993, + "step": 27609 + }, + { + "epoch": 0.66, + "learning_rate": 5.3598256778370886e-06, + "loss": 1.1718, + "step": 27610 + }, + { + "epoch": 0.66, + "learning_rate": 5.359135726953792e-06, + "loss": 0.9772, + "step": 27611 + }, + { + "epoch": 0.66, + "learning_rate": 5.358445804225524e-06, + "loss": 1.086, + "step": 27612 + }, + { + "epoch": 0.66, + "learning_rate": 5.357755909656472e-06, + "loss": 1.0425, + "step": 27613 + }, + { + "epoch": 0.66, + "learning_rate": 5.357066043250829e-06, + "loss": 0.9576, + "step": 27614 + }, + { + "epoch": 0.66, + "learning_rate": 5.35637620501277e-06, + "loss": 1.2055, + "step": 27615 + }, + { + "epoch": 0.66, + "learning_rate": 5.355686394946486e-06, + "loss": 0.9064, + "step": 27616 + }, + { + "epoch": 0.66, + "learning_rate": 5.354996613056158e-06, + "loss": 1.0756, + "step": 27617 + }, + { + "epoch": 0.66, + "learning_rate": 5.354306859345975e-06, + "loss": 1.0286, + "step": 27618 + }, + { + "epoch": 0.66, + "learning_rate": 5.35361713382012e-06, + "loss": 1.0104, + "step": 27619 + }, + { + "epoch": 0.66, + "learning_rate": 5.352927436482776e-06, + "loss": 0.9821, + "step": 27620 + }, + { + "epoch": 0.66, + "learning_rate": 5.352237767338125e-06, + "loss": 0.9649, + "step": 27621 + }, + { + "epoch": 0.66, + "learning_rate": 5.35154812639036e-06, + "loss": 1.2476, + "step": 27622 + }, + { + "epoch": 0.66, + "learning_rate": 5.350858513643655e-06, + "loss": 1.1415, + "step": 27623 + }, + { + "epoch": 0.66, + "learning_rate": 5.350168929102202e-06, + "loss": 1.1694, + "step": 27624 + }, + { + "epoch": 0.66, + "learning_rate": 5.349479372770177e-06, + "loss": 1.0464, + "step": 27625 + }, + { + "epoch": 0.66, + "learning_rate": 5.348789844651765e-06, + "loss": 0.8407, + "step": 27626 + }, + { + "epoch": 0.66, + "learning_rate": 5.348100344751156e-06, + "loss": 0.8026, + "step": 27627 + }, + { + "epoch": 0.66, + "learning_rate": 5.347410873072524e-06, + "loss": 0.9848, + "step": 27628 + }, + { + "epoch": 0.66, + "learning_rate": 5.346721429620057e-06, + "loss": 1.1721, + "step": 27629 + }, + { + "epoch": 0.66, + "learning_rate": 5.346032014397938e-06, + "loss": 1.1649, + "step": 27630 + }, + { + "epoch": 0.66, + "learning_rate": 5.345342627410346e-06, + "loss": 1.1913, + "step": 27631 + }, + { + "epoch": 0.66, + "learning_rate": 5.344653268661465e-06, + "loss": 0.9313, + "step": 27632 + }, + { + "epoch": 0.66, + "learning_rate": 5.343963938155481e-06, + "loss": 1.0899, + "step": 27633 + }, + { + "epoch": 0.66, + "learning_rate": 5.343274635896569e-06, + "loss": 0.927, + "step": 27634 + }, + { + "epoch": 0.66, + "learning_rate": 5.342585361888919e-06, + "loss": 1.0582, + "step": 27635 + }, + { + "epoch": 0.66, + "learning_rate": 5.341896116136705e-06, + "loss": 1.105, + "step": 27636 + }, + { + "epoch": 0.66, + "learning_rate": 5.341206898644109e-06, + "loss": 0.8315, + "step": 27637 + }, + { + "epoch": 0.66, + "learning_rate": 5.34051770941532e-06, + "loss": 0.8289, + "step": 27638 + }, + { + "epoch": 0.66, + "learning_rate": 5.339828548454511e-06, + "loss": 1.0111, + "step": 27639 + }, + { + "epoch": 0.66, + "learning_rate": 5.339139415765866e-06, + "loss": 0.524, + "step": 27640 + }, + { + "epoch": 0.66, + "learning_rate": 5.338450311353569e-06, + "loss": 1.1384, + "step": 27641 + }, + { + "epoch": 0.66, + "learning_rate": 5.337761235221793e-06, + "loss": 1.1364, + "step": 27642 + }, + { + "epoch": 0.66, + "learning_rate": 5.3370721873747275e-06, + "loss": 1.2329, + "step": 27643 + }, + { + "epoch": 0.66, + "learning_rate": 5.3363831678165456e-06, + "loss": 0.8094, + "step": 27644 + }, + { + "epoch": 0.66, + "learning_rate": 5.335694176551429e-06, + "loss": 1.173, + "step": 27645 + }, + { + "epoch": 0.66, + "learning_rate": 5.3350052135835616e-06, + "loss": 1.2413, + "step": 27646 + }, + { + "epoch": 0.66, + "learning_rate": 5.334316278917118e-06, + "loss": 1.2973, + "step": 27647 + }, + { + "epoch": 0.66, + "learning_rate": 5.333627372556279e-06, + "loss": 1.2846, + "step": 27648 + }, + { + "epoch": 0.66, + "learning_rate": 5.33293849450523e-06, + "loss": 0.9453, + "step": 27649 + }, + { + "epoch": 0.66, + "learning_rate": 5.332249644768142e-06, + "loss": 1.0911, + "step": 27650 + }, + { + "epoch": 0.66, + "learning_rate": 5.3315608233492e-06, + "loss": 0.8956, + "step": 27651 + }, + { + "epoch": 0.66, + "learning_rate": 5.330872030252576e-06, + "loss": 0.9956, + "step": 27652 + }, + { + "epoch": 0.67, + "learning_rate": 5.330183265482455e-06, + "loss": 0.8906, + "step": 27653 + }, + { + "epoch": 0.67, + "learning_rate": 5.329494529043015e-06, + "loss": 1.3384, + "step": 27654 + }, + { + "epoch": 0.67, + "learning_rate": 5.328805820938432e-06, + "loss": 1.0815, + "step": 27655 + }, + { + "epoch": 0.67, + "learning_rate": 5.328117141172884e-06, + "loss": 1.0184, + "step": 27656 + }, + { + "epoch": 0.67, + "learning_rate": 5.3274284897505545e-06, + "loss": 1.3104, + "step": 27657 + }, + { + "epoch": 0.67, + "learning_rate": 5.326739866675614e-06, + "loss": 1.1629, + "step": 27658 + }, + { + "epoch": 0.67, + "learning_rate": 5.326051271952248e-06, + "loss": 1.2421, + "step": 27659 + }, + { + "epoch": 0.67, + "learning_rate": 5.325362705584623e-06, + "loss": 1.083, + "step": 27660 + }, + { + "epoch": 0.67, + "learning_rate": 5.324674167576924e-06, + "loss": 0.7747, + "step": 27661 + }, + { + "epoch": 0.67, + "learning_rate": 5.323985657933333e-06, + "loss": 1.1563, + "step": 27662 + }, + { + "epoch": 0.67, + "learning_rate": 5.323297176658015e-06, + "loss": 0.9722, + "step": 27663 + }, + { + "epoch": 0.67, + "learning_rate": 5.322608723755154e-06, + "loss": 1.1739, + "step": 27664 + }, + { + "epoch": 0.67, + "learning_rate": 5.321920299228928e-06, + "loss": 0.8574, + "step": 27665 + }, + { + "epoch": 0.67, + "learning_rate": 5.321231903083506e-06, + "loss": 1.245, + "step": 27666 + }, + { + "epoch": 0.67, + "learning_rate": 5.320543535323076e-06, + "loss": 1.2444, + "step": 27667 + }, + { + "epoch": 0.67, + "learning_rate": 5.319855195951804e-06, + "loss": 1.0322, + "step": 27668 + }, + { + "epoch": 0.67, + "learning_rate": 5.31916688497387e-06, + "loss": 1.0922, + "step": 27669 + }, + { + "epoch": 0.67, + "learning_rate": 5.318478602393453e-06, + "loss": 1.1594, + "step": 27670 + }, + { + "epoch": 0.67, + "learning_rate": 5.317790348214721e-06, + "loss": 1.149, + "step": 27671 + }, + { + "epoch": 0.67, + "learning_rate": 5.3171021224418574e-06, + "loss": 1.2528, + "step": 27672 + }, + { + "epoch": 0.67, + "learning_rate": 5.31641392507903e-06, + "loss": 1.0163, + "step": 27673 + }, + { + "epoch": 0.67, + "learning_rate": 5.315725756130419e-06, + "loss": 1.086, + "step": 27674 + }, + { + "epoch": 0.67, + "learning_rate": 5.3150376156002005e-06, + "loss": 1.0831, + "step": 27675 + }, + { + "epoch": 0.67, + "learning_rate": 5.314349503492542e-06, + "loss": 0.9755, + "step": 27676 + }, + { + "epoch": 0.67, + "learning_rate": 5.313661419811624e-06, + "loss": 0.9442, + "step": 27677 + }, + { + "epoch": 0.67, + "learning_rate": 5.3129733645616245e-06, + "loss": 1.1056, + "step": 27678 + }, + { + "epoch": 0.67, + "learning_rate": 5.312285337746709e-06, + "loss": 1.0207, + "step": 27679 + }, + { + "epoch": 0.67, + "learning_rate": 5.31159733937106e-06, + "loss": 1.1368, + "step": 27680 + }, + { + "epoch": 0.67, + "learning_rate": 5.310909369438842e-06, + "loss": 1.2288, + "step": 27681 + }, + { + "epoch": 0.67, + "learning_rate": 5.310221427954234e-06, + "loss": 0.9844, + "step": 27682 + }, + { + "epoch": 0.67, + "learning_rate": 5.309533514921414e-06, + "loss": 1.1415, + "step": 27683 + }, + { + "epoch": 0.67, + "learning_rate": 5.308845630344547e-06, + "loss": 1.0143, + "step": 27684 + }, + { + "epoch": 0.67, + "learning_rate": 5.308157774227809e-06, + "loss": 0.9794, + "step": 27685 + }, + { + "epoch": 0.67, + "learning_rate": 5.30746994657538e-06, + "loss": 1.1888, + "step": 27686 + }, + { + "epoch": 0.67, + "learning_rate": 5.306782147391423e-06, + "loss": 1.0921, + "step": 27687 + }, + { + "epoch": 0.67, + "learning_rate": 5.306094376680117e-06, + "loss": 1.2693, + "step": 27688 + }, + { + "epoch": 0.67, + "learning_rate": 5.305406634445629e-06, + "loss": 1.0529, + "step": 27689 + }, + { + "epoch": 0.67, + "learning_rate": 5.3047189206921354e-06, + "loss": 1.0933, + "step": 27690 + }, + { + "epoch": 0.67, + "learning_rate": 5.304031235423811e-06, + "loss": 1.1874, + "step": 27691 + }, + { + "epoch": 0.67, + "learning_rate": 5.303343578644822e-06, + "loss": 1.0773, + "step": 27692 + }, + { + "epoch": 0.67, + "learning_rate": 5.302655950359341e-06, + "loss": 1.2367, + "step": 27693 + }, + { + "epoch": 0.67, + "learning_rate": 5.301968350571547e-06, + "loss": 1.3166, + "step": 27694 + }, + { + "epoch": 0.67, + "learning_rate": 5.3012807792856e-06, + "loss": 1.3274, + "step": 27695 + }, + { + "epoch": 0.67, + "learning_rate": 5.300593236505682e-06, + "loss": 0.9344, + "step": 27696 + }, + { + "epoch": 0.67, + "learning_rate": 5.299905722235956e-06, + "loss": 1.3546, + "step": 27697 + }, + { + "epoch": 0.67, + "learning_rate": 5.299218236480597e-06, + "loss": 0.7746, + "step": 27698 + }, + { + "epoch": 0.67, + "learning_rate": 5.298530779243778e-06, + "loss": 1.1697, + "step": 27699 + }, + { + "epoch": 0.67, + "learning_rate": 5.297843350529664e-06, + "loss": 1.2354, + "step": 27700 + }, + { + "epoch": 0.67, + "learning_rate": 5.297155950342428e-06, + "loss": 1.1373, + "step": 27701 + }, + { + "epoch": 0.67, + "learning_rate": 5.296468578686244e-06, + "loss": 0.8228, + "step": 27702 + }, + { + "epoch": 0.67, + "learning_rate": 5.295781235565276e-06, + "loss": 0.8204, + "step": 27703 + }, + { + "epoch": 0.67, + "learning_rate": 5.2950939209837e-06, + "loss": 1.2012, + "step": 27704 + }, + { + "epoch": 0.67, + "learning_rate": 5.294406634945679e-06, + "loss": 1.0707, + "step": 27705 + }, + { + "epoch": 0.67, + "learning_rate": 5.293719377455386e-06, + "loss": 1.2052, + "step": 27706 + }, + { + "epoch": 0.67, + "learning_rate": 5.293032148516994e-06, + "loss": 0.842, + "step": 27707 + }, + { + "epoch": 0.67, + "learning_rate": 5.292344948134665e-06, + "loss": 1.1022, + "step": 27708 + }, + { + "epoch": 0.67, + "learning_rate": 5.291657776312572e-06, + "loss": 1.0786, + "step": 27709 + }, + { + "epoch": 0.67, + "learning_rate": 5.290970633054888e-06, + "loss": 1.3025, + "step": 27710 + }, + { + "epoch": 0.67, + "learning_rate": 5.290283518365773e-06, + "loss": 1.3615, + "step": 27711 + }, + { + "epoch": 0.67, + "learning_rate": 5.289596432249401e-06, + "loss": 1.0559, + "step": 27712 + }, + { + "epoch": 0.67, + "learning_rate": 5.288909374709943e-06, + "loss": 0.8892, + "step": 27713 + }, + { + "epoch": 0.67, + "learning_rate": 5.288222345751559e-06, + "loss": 0.8285, + "step": 27714 + }, + { + "epoch": 0.67, + "learning_rate": 5.287535345378428e-06, + "loss": 1.1693, + "step": 27715 + }, + { + "epoch": 0.67, + "learning_rate": 5.286848373594705e-06, + "loss": 1.0503, + "step": 27716 + }, + { + "epoch": 0.67, + "learning_rate": 5.286161430404566e-06, + "loss": 1.3351, + "step": 27717 + }, + { + "epoch": 0.67, + "learning_rate": 5.285474515812179e-06, + "loss": 1.0382, + "step": 27718 + }, + { + "epoch": 0.67, + "learning_rate": 5.284787629821707e-06, + "loss": 1.2083, + "step": 27719 + }, + { + "epoch": 0.67, + "learning_rate": 5.2841007724373195e-06, + "loss": 1.3393, + "step": 27720 + }, + { + "epoch": 0.67, + "learning_rate": 5.283413943663186e-06, + "loss": 1.154, + "step": 27721 + }, + { + "epoch": 0.67, + "learning_rate": 5.282727143503468e-06, + "loss": 1.1556, + "step": 27722 + }, + { + "epoch": 0.67, + "learning_rate": 5.282040371962338e-06, + "loss": 0.7804, + "step": 27723 + }, + { + "epoch": 0.67, + "learning_rate": 5.281353629043955e-06, + "loss": 0.7078, + "step": 27724 + }, + { + "epoch": 0.67, + "learning_rate": 5.280666914752491e-06, + "loss": 1.1579, + "step": 27725 + }, + { + "epoch": 0.67, + "learning_rate": 5.279980229092112e-06, + "loss": 0.994, + "step": 27726 + }, + { + "epoch": 0.67, + "learning_rate": 5.279293572066981e-06, + "loss": 1.1067, + "step": 27727 + }, + { + "epoch": 0.67, + "learning_rate": 5.278606943681266e-06, + "loss": 1.1435, + "step": 27728 + }, + { + "epoch": 0.67, + "learning_rate": 5.277920343939135e-06, + "loss": 1.3252, + "step": 27729 + }, + { + "epoch": 0.67, + "learning_rate": 5.277233772844748e-06, + "loss": 1.0111, + "step": 27730 + }, + { + "epoch": 0.67, + "learning_rate": 5.276547230402274e-06, + "loss": 0.9301, + "step": 27731 + }, + { + "epoch": 0.67, + "learning_rate": 5.275860716615875e-06, + "loss": 1.0714, + "step": 27732 + }, + { + "epoch": 0.67, + "learning_rate": 5.275174231489717e-06, + "loss": 1.1257, + "step": 27733 + }, + { + "epoch": 0.67, + "learning_rate": 5.27448777502797e-06, + "loss": 1.2526, + "step": 27734 + }, + { + "epoch": 0.67, + "learning_rate": 5.27380134723479e-06, + "loss": 0.9074, + "step": 27735 + }, + { + "epoch": 0.67, + "learning_rate": 5.273114948114346e-06, + "loss": 0.6996, + "step": 27736 + }, + { + "epoch": 0.67, + "learning_rate": 5.272428577670806e-06, + "loss": 0.6834, + "step": 27737 + }, + { + "epoch": 0.67, + "learning_rate": 5.271742235908326e-06, + "loss": 0.8243, + "step": 27738 + }, + { + "epoch": 0.67, + "learning_rate": 5.271055922831077e-06, + "loss": 1.185, + "step": 27739 + }, + { + "epoch": 0.67, + "learning_rate": 5.270369638443216e-06, + "loss": 1.4433, + "step": 27740 + }, + { + "epoch": 0.67, + "learning_rate": 5.269683382748909e-06, + "loss": 1.006, + "step": 27741 + }, + { + "epoch": 0.67, + "learning_rate": 5.268997155752324e-06, + "loss": 1.2227, + "step": 27742 + }, + { + "epoch": 0.67, + "learning_rate": 5.268310957457617e-06, + "loss": 1.0509, + "step": 27743 + }, + { + "epoch": 0.67, + "learning_rate": 5.267624787868955e-06, + "loss": 0.8845, + "step": 27744 + }, + { + "epoch": 0.67, + "learning_rate": 5.266938646990505e-06, + "loss": 1.0724, + "step": 27745 + }, + { + "epoch": 0.67, + "learning_rate": 5.266252534826419e-06, + "loss": 0.7289, + "step": 27746 + }, + { + "epoch": 0.67, + "learning_rate": 5.2655664513808705e-06, + "loss": 0.9375, + "step": 27747 + }, + { + "epoch": 0.67, + "learning_rate": 5.264880396658012e-06, + "loss": 1.2898, + "step": 27748 + }, + { + "epoch": 0.67, + "learning_rate": 5.264194370662011e-06, + "loss": 0.9802, + "step": 27749 + }, + { + "epoch": 0.67, + "learning_rate": 5.263508373397033e-06, + "loss": 1.1473, + "step": 27750 + }, + { + "epoch": 0.67, + "learning_rate": 5.2628224048672315e-06, + "loss": 0.9982, + "step": 27751 + }, + { + "epoch": 0.67, + "learning_rate": 5.262136465076771e-06, + "loss": 0.9767, + "step": 27752 + }, + { + "epoch": 0.67, + "learning_rate": 5.26145055402982e-06, + "loss": 1.0097, + "step": 27753 + }, + { + "epoch": 0.67, + "learning_rate": 5.2607646717305294e-06, + "loss": 0.9152, + "step": 27754 + }, + { + "epoch": 0.67, + "learning_rate": 5.260078818183064e-06, + "loss": 1.0205, + "step": 27755 + }, + { + "epoch": 0.67, + "learning_rate": 5.2593929933915896e-06, + "loss": 1.0106, + "step": 27756 + }, + { + "epoch": 0.67, + "learning_rate": 5.25870719736026e-06, + "loss": 1.0512, + "step": 27757 + }, + { + "epoch": 0.67, + "learning_rate": 5.258021430093241e-06, + "loss": 1.1453, + "step": 27758 + }, + { + "epoch": 0.67, + "learning_rate": 5.257335691594687e-06, + "loss": 0.7069, + "step": 27759 + }, + { + "epoch": 0.67, + "learning_rate": 5.256649981868763e-06, + "loss": 1.155, + "step": 27760 + }, + { + "epoch": 0.67, + "learning_rate": 5.255964300919631e-06, + "loss": 0.881, + "step": 27761 + }, + { + "epoch": 0.67, + "learning_rate": 5.255278648751444e-06, + "loss": 0.8384, + "step": 27762 + }, + { + "epoch": 0.67, + "learning_rate": 5.254593025368366e-06, + "loss": 0.932, + "step": 27763 + }, + { + "epoch": 0.67, + "learning_rate": 5.253907430774559e-06, + "loss": 1.3416, + "step": 27764 + }, + { + "epoch": 0.67, + "learning_rate": 5.253221864974175e-06, + "loss": 1.1668, + "step": 27765 + }, + { + "epoch": 0.67, + "learning_rate": 5.252536327971383e-06, + "loss": 1.0, + "step": 27766 + }, + { + "epoch": 0.67, + "learning_rate": 5.25185081977033e-06, + "loss": 1.1011, + "step": 27767 + }, + { + "epoch": 0.67, + "learning_rate": 5.251165340375183e-06, + "loss": 1.1264, + "step": 27768 + }, + { + "epoch": 0.67, + "learning_rate": 5.250479889790102e-06, + "loss": 1.1682, + "step": 27769 + }, + { + "epoch": 0.67, + "learning_rate": 5.2497944680192395e-06, + "loss": 1.0939, + "step": 27770 + }, + { + "epoch": 0.67, + "learning_rate": 5.249109075066756e-06, + "loss": 0.9864, + "step": 27771 + }, + { + "epoch": 0.67, + "learning_rate": 5.2484237109368145e-06, + "loss": 0.8601, + "step": 27772 + }, + { + "epoch": 0.67, + "learning_rate": 5.247738375633564e-06, + "loss": 0.9341, + "step": 27773 + }, + { + "epoch": 0.67, + "learning_rate": 5.247053069161171e-06, + "loss": 0.7981, + "step": 27774 + }, + { + "epoch": 0.67, + "learning_rate": 5.246367791523786e-06, + "loss": 1.1481, + "step": 27775 + }, + { + "epoch": 0.67, + "learning_rate": 5.2456825427255684e-06, + "loss": 0.9251, + "step": 27776 + }, + { + "epoch": 0.67, + "learning_rate": 5.244997322770682e-06, + "loss": 0.9902, + "step": 27777 + }, + { + "epoch": 0.67, + "learning_rate": 5.2443121316632725e-06, + "loss": 1.1008, + "step": 27778 + }, + { + "epoch": 0.67, + "learning_rate": 5.243626969407505e-06, + "loss": 0.8983, + "step": 27779 + }, + { + "epoch": 0.67, + "learning_rate": 5.242941836007536e-06, + "loss": 1.1615, + "step": 27780 + }, + { + "epoch": 0.67, + "learning_rate": 5.242256731467517e-06, + "loss": 0.9442, + "step": 27781 + }, + { + "epoch": 0.67, + "learning_rate": 5.241571655791611e-06, + "loss": 1.0283, + "step": 27782 + }, + { + "epoch": 0.67, + "learning_rate": 5.240886608983966e-06, + "loss": 0.8848, + "step": 27783 + }, + { + "epoch": 0.67, + "learning_rate": 5.240201591048742e-06, + "loss": 1.1654, + "step": 27784 + }, + { + "epoch": 0.67, + "learning_rate": 5.2395166019901e-06, + "loss": 0.7772, + "step": 27785 + }, + { + "epoch": 0.67, + "learning_rate": 5.238831641812187e-06, + "loss": 0.9837, + "step": 27786 + }, + { + "epoch": 0.67, + "learning_rate": 5.238146710519163e-06, + "loss": 1.0017, + "step": 27787 + }, + { + "epoch": 0.67, + "learning_rate": 5.237461808115187e-06, + "loss": 1.171, + "step": 27788 + }, + { + "epoch": 0.67, + "learning_rate": 5.236776934604406e-06, + "loss": 1.0376, + "step": 27789 + }, + { + "epoch": 0.67, + "learning_rate": 5.236092089990983e-06, + "loss": 1.0737, + "step": 27790 + }, + { + "epoch": 0.67, + "learning_rate": 5.235407274279064e-06, + "loss": 1.0472, + "step": 27791 + }, + { + "epoch": 0.67, + "learning_rate": 5.2347224874728095e-06, + "loss": 1.1195, + "step": 27792 + }, + { + "epoch": 0.67, + "learning_rate": 5.234037729576376e-06, + "loss": 0.8041, + "step": 27793 + }, + { + "epoch": 0.67, + "learning_rate": 5.23335300059391e-06, + "loss": 1.1758, + "step": 27794 + }, + { + "epoch": 0.67, + "learning_rate": 5.232668300529571e-06, + "loss": 1.2843, + "step": 27795 + }, + { + "epoch": 0.67, + "learning_rate": 5.231983629387516e-06, + "loss": 0.9502, + "step": 27796 + }, + { + "epoch": 0.67, + "learning_rate": 5.231298987171891e-06, + "loss": 1.1913, + "step": 27797 + }, + { + "epoch": 0.67, + "learning_rate": 5.230614373886856e-06, + "loss": 1.0628, + "step": 27798 + }, + { + "epoch": 0.67, + "learning_rate": 5.22992978953656e-06, + "loss": 1.0146, + "step": 27799 + }, + { + "epoch": 0.67, + "learning_rate": 5.229245234125156e-06, + "loss": 1.0862, + "step": 27800 + }, + { + "epoch": 0.67, + "learning_rate": 5.228560707656805e-06, + "loss": 1.1667, + "step": 27801 + }, + { + "epoch": 0.67, + "learning_rate": 5.2278762101356475e-06, + "loss": 1.0482, + "step": 27802 + }, + { + "epoch": 0.67, + "learning_rate": 5.227191741565844e-06, + "loss": 0.8988, + "step": 27803 + }, + { + "epoch": 0.67, + "learning_rate": 5.226507301951549e-06, + "loss": 0.9899, + "step": 27804 + }, + { + "epoch": 0.67, + "learning_rate": 5.225822891296908e-06, + "loss": 1.0225, + "step": 27805 + }, + { + "epoch": 0.67, + "learning_rate": 5.225138509606076e-06, + "loss": 1.0217, + "step": 27806 + }, + { + "epoch": 0.67, + "learning_rate": 5.2244541568832096e-06, + "loss": 0.9636, + "step": 27807 + }, + { + "epoch": 0.67, + "learning_rate": 5.223769833132454e-06, + "loss": 1.3011, + "step": 27808 + }, + { + "epoch": 0.67, + "learning_rate": 5.223085538357966e-06, + "loss": 1.1029, + "step": 27809 + }, + { + "epoch": 0.67, + "learning_rate": 5.222401272563891e-06, + "loss": 0.958, + "step": 27810 + }, + { + "epoch": 0.67, + "learning_rate": 5.221717035754383e-06, + "loss": 1.073, + "step": 27811 + }, + { + "epoch": 0.67, + "learning_rate": 5.221032827933598e-06, + "loss": 0.8238, + "step": 27812 + }, + { + "epoch": 0.67, + "learning_rate": 5.220348649105679e-06, + "loss": 1.1967, + "step": 27813 + }, + { + "epoch": 0.67, + "learning_rate": 5.21966449927478e-06, + "loss": 1.2562, + "step": 27814 + }, + { + "epoch": 0.67, + "learning_rate": 5.2189803784450575e-06, + "loss": 0.9741, + "step": 27815 + }, + { + "epoch": 0.67, + "learning_rate": 5.218296286620651e-06, + "loss": 1.0213, + "step": 27816 + }, + { + "epoch": 0.67, + "learning_rate": 5.21761222380572e-06, + "loss": 0.925, + "step": 27817 + }, + { + "epoch": 0.67, + "learning_rate": 5.216928190004408e-06, + "loss": 0.9618, + "step": 27818 + }, + { + "epoch": 0.67, + "learning_rate": 5.216244185220867e-06, + "loss": 1.0979, + "step": 27819 + }, + { + "epoch": 0.67, + "learning_rate": 5.215560209459252e-06, + "loss": 1.0574, + "step": 27820 + }, + { + "epoch": 0.67, + "learning_rate": 5.214876262723703e-06, + "loss": 1.0953, + "step": 27821 + }, + { + "epoch": 0.67, + "learning_rate": 5.214192345018374e-06, + "loss": 1.0596, + "step": 27822 + }, + { + "epoch": 0.67, + "learning_rate": 5.2135084563474185e-06, + "loss": 1.0012, + "step": 27823 + }, + { + "epoch": 0.67, + "learning_rate": 5.212824596714978e-06, + "loss": 1.0185, + "step": 27824 + }, + { + "epoch": 0.67, + "learning_rate": 5.212140766125209e-06, + "loss": 1.0992, + "step": 27825 + }, + { + "epoch": 0.67, + "learning_rate": 5.2114569645822515e-06, + "loss": 0.9824, + "step": 27826 + }, + { + "epoch": 0.67, + "learning_rate": 5.210773192090257e-06, + "loss": 1.1332, + "step": 27827 + }, + { + "epoch": 0.67, + "learning_rate": 5.210089448653379e-06, + "loss": 1.1699, + "step": 27828 + }, + { + "epoch": 0.67, + "learning_rate": 5.209405734275759e-06, + "loss": 0.9044, + "step": 27829 + }, + { + "epoch": 0.67, + "learning_rate": 5.208722048961548e-06, + "loss": 0.9322, + "step": 27830 + }, + { + "epoch": 0.67, + "learning_rate": 5.208038392714896e-06, + "loss": 1.0561, + "step": 27831 + }, + { + "epoch": 0.67, + "learning_rate": 5.207354765539941e-06, + "loss": 1.0529, + "step": 27832 + }, + { + "epoch": 0.67, + "learning_rate": 5.206671167440844e-06, + "loss": 1.0957, + "step": 27833 + }, + { + "epoch": 0.67, + "learning_rate": 5.205987598421743e-06, + "loss": 1.2524, + "step": 27834 + }, + { + "epoch": 0.67, + "learning_rate": 5.2053040584867865e-06, + "loss": 1.0348, + "step": 27835 + }, + { + "epoch": 0.67, + "learning_rate": 5.204620547640127e-06, + "loss": 1.3197, + "step": 27836 + }, + { + "epoch": 0.67, + "learning_rate": 5.203937065885902e-06, + "loss": 0.9556, + "step": 27837 + }, + { + "epoch": 0.67, + "learning_rate": 5.203253613228268e-06, + "loss": 1.0025, + "step": 27838 + }, + { + "epoch": 0.67, + "learning_rate": 5.202570189671361e-06, + "loss": 1.3006, + "step": 27839 + }, + { + "epoch": 0.67, + "learning_rate": 5.201886795219333e-06, + "loss": 1.1394, + "step": 27840 + }, + { + "epoch": 0.67, + "learning_rate": 5.2012034298763326e-06, + "loss": 1.1029, + "step": 27841 + }, + { + "epoch": 0.67, + "learning_rate": 5.200520093646498e-06, + "loss": 1.1331, + "step": 27842 + }, + { + "epoch": 0.67, + "learning_rate": 5.1998367865339806e-06, + "loss": 1.0875, + "step": 27843 + }, + { + "epoch": 0.67, + "learning_rate": 5.1991535085429266e-06, + "loss": 0.9892, + "step": 27844 + }, + { + "epoch": 0.67, + "learning_rate": 5.1984702596774765e-06, + "loss": 1.1009, + "step": 27845 + }, + { + "epoch": 0.67, + "learning_rate": 5.197787039941781e-06, + "loss": 0.7558, + "step": 27846 + }, + { + "epoch": 0.67, + "learning_rate": 5.197103849339978e-06, + "loss": 0.9811, + "step": 27847 + }, + { + "epoch": 0.67, + "learning_rate": 5.196420687876217e-06, + "loss": 0.9409, + "step": 27848 + }, + { + "epoch": 0.67, + "learning_rate": 5.195737555554645e-06, + "loss": 1.0918, + "step": 27849 + }, + { + "epoch": 0.67, + "learning_rate": 5.1950544523794e-06, + "loss": 1.0988, + "step": 27850 + }, + { + "epoch": 0.67, + "learning_rate": 5.194371378354628e-06, + "loss": 0.7502, + "step": 27851 + }, + { + "epoch": 0.67, + "learning_rate": 5.193688333484479e-06, + "loss": 1.0688, + "step": 27852 + }, + { + "epoch": 0.67, + "learning_rate": 5.19300531777309e-06, + "loss": 1.1031, + "step": 27853 + }, + { + "epoch": 0.67, + "learning_rate": 5.1923223312246095e-06, + "loss": 0.9977, + "step": 27854 + }, + { + "epoch": 0.67, + "learning_rate": 5.191639373843175e-06, + "loss": 0.7112, + "step": 27855 + }, + { + "epoch": 0.67, + "learning_rate": 5.190956445632933e-06, + "loss": 1.1499, + "step": 27856 + }, + { + "epoch": 0.67, + "learning_rate": 5.190273546598031e-06, + "loss": 0.8806, + "step": 27857 + }, + { + "epoch": 0.67, + "learning_rate": 5.1895906767426055e-06, + "loss": 1.1493, + "step": 27858 + }, + { + "epoch": 0.67, + "learning_rate": 5.1889078360708e-06, + "loss": 1.151, + "step": 27859 + }, + { + "epoch": 0.67, + "learning_rate": 5.188225024586764e-06, + "loss": 0.926, + "step": 27860 + }, + { + "epoch": 0.67, + "learning_rate": 5.187542242294632e-06, + "loss": 1.128, + "step": 27861 + }, + { + "epoch": 0.67, + "learning_rate": 5.186859489198552e-06, + "loss": 1.1455, + "step": 27862 + }, + { + "epoch": 0.67, + "learning_rate": 5.186176765302659e-06, + "loss": 1.063, + "step": 27863 + }, + { + "epoch": 0.67, + "learning_rate": 5.1854940706111e-06, + "loss": 1.0492, + "step": 27864 + }, + { + "epoch": 0.67, + "learning_rate": 5.184811405128019e-06, + "loss": 0.9546, + "step": 27865 + }, + { + "epoch": 0.67, + "learning_rate": 5.1841287688575525e-06, + "loss": 0.8343, + "step": 27866 + }, + { + "epoch": 0.67, + "learning_rate": 5.183446161803842e-06, + "loss": 1.2074, + "step": 27867 + }, + { + "epoch": 0.67, + "learning_rate": 5.182763583971036e-06, + "loss": 1.1196, + "step": 27868 + }, + { + "epoch": 0.67, + "learning_rate": 5.182081035363267e-06, + "loss": 0.8717, + "step": 27869 + }, + { + "epoch": 0.67, + "learning_rate": 5.181398515984682e-06, + "loss": 1.1282, + "step": 27870 + }, + { + "epoch": 0.67, + "learning_rate": 5.180716025839414e-06, + "loss": 0.9012, + "step": 27871 + }, + { + "epoch": 0.67, + "learning_rate": 5.180033564931609e-06, + "loss": 1.1616, + "step": 27872 + }, + { + "epoch": 0.67, + "learning_rate": 5.179351133265411e-06, + "loss": 1.275, + "step": 27873 + }, + { + "epoch": 0.67, + "learning_rate": 5.178668730844951e-06, + "loss": 0.8676, + "step": 27874 + }, + { + "epoch": 0.67, + "learning_rate": 5.177986357674375e-06, + "loss": 1.0202, + "step": 27875 + }, + { + "epoch": 0.67, + "learning_rate": 5.177304013757825e-06, + "loss": 1.1352, + "step": 27876 + }, + { + "epoch": 0.67, + "learning_rate": 5.176621699099433e-06, + "loss": 1.2333, + "step": 27877 + }, + { + "epoch": 0.67, + "learning_rate": 5.175939413703343e-06, + "loss": 1.1301, + "step": 27878 + }, + { + "epoch": 0.67, + "learning_rate": 5.175257157573697e-06, + "loss": 0.8901, + "step": 27879 + }, + { + "epoch": 0.67, + "learning_rate": 5.174574930714627e-06, + "loss": 1.1089, + "step": 27880 + }, + { + "epoch": 0.67, + "learning_rate": 5.173892733130282e-06, + "loss": 1.0367, + "step": 27881 + }, + { + "epoch": 0.67, + "learning_rate": 5.173210564824789e-06, + "loss": 1.0733, + "step": 27882 + }, + { + "epoch": 0.67, + "learning_rate": 5.172528425802292e-06, + "loss": 1.3238, + "step": 27883 + }, + { + "epoch": 0.67, + "learning_rate": 5.171846316066933e-06, + "loss": 0.9881, + "step": 27884 + }, + { + "epoch": 0.67, + "learning_rate": 5.171164235622843e-06, + "loss": 0.8607, + "step": 27885 + }, + { + "epoch": 0.67, + "learning_rate": 5.170482184474166e-06, + "loss": 1.1789, + "step": 27886 + }, + { + "epoch": 0.67, + "learning_rate": 5.169800162625038e-06, + "loss": 1.2643, + "step": 27887 + }, + { + "epoch": 0.67, + "learning_rate": 5.1691181700795955e-06, + "loss": 1.1701, + "step": 27888 + }, + { + "epoch": 0.67, + "learning_rate": 5.168436206841979e-06, + "loss": 0.9233, + "step": 27889 + }, + { + "epoch": 0.67, + "learning_rate": 5.167754272916321e-06, + "loss": 0.9892, + "step": 27890 + }, + { + "epoch": 0.67, + "learning_rate": 5.16707236830676e-06, + "loss": 1.0091, + "step": 27891 + }, + { + "epoch": 0.67, + "learning_rate": 5.166390493017438e-06, + "loss": 1.0641, + "step": 27892 + }, + { + "epoch": 0.67, + "learning_rate": 5.165708647052484e-06, + "loss": 1.1806, + "step": 27893 + }, + { + "epoch": 0.67, + "learning_rate": 5.16502683041604e-06, + "loss": 0.932, + "step": 27894 + }, + { + "epoch": 0.67, + "learning_rate": 5.164345043112244e-06, + "loss": 1.0169, + "step": 27895 + }, + { + "epoch": 0.67, + "learning_rate": 5.163663285145224e-06, + "loss": 0.8672, + "step": 27896 + }, + { + "epoch": 0.67, + "learning_rate": 5.162981556519127e-06, + "loss": 1.1333, + "step": 27897 + }, + { + "epoch": 0.67, + "learning_rate": 5.1622998572380776e-06, + "loss": 1.2816, + "step": 27898 + }, + { + "epoch": 0.67, + "learning_rate": 5.161618187306218e-06, + "loss": 1.2529, + "step": 27899 + }, + { + "epoch": 0.67, + "learning_rate": 5.160936546727687e-06, + "loss": 1.0216, + "step": 27900 + }, + { + "epoch": 0.67, + "learning_rate": 5.160254935506612e-06, + "loss": 0.8212, + "step": 27901 + }, + { + "epoch": 0.67, + "learning_rate": 5.159573353647131e-06, + "loss": 0.9346, + "step": 27902 + }, + { + "epoch": 0.67, + "learning_rate": 5.158891801153385e-06, + "loss": 0.9947, + "step": 27903 + }, + { + "epoch": 0.67, + "learning_rate": 5.158210278029498e-06, + "loss": 1.0331, + "step": 27904 + }, + { + "epoch": 0.67, + "learning_rate": 5.157528784279616e-06, + "loss": 0.9782, + "step": 27905 + }, + { + "epoch": 0.67, + "learning_rate": 5.156847319907864e-06, + "loss": 1.1515, + "step": 27906 + }, + { + "epoch": 0.67, + "learning_rate": 5.1561658849183796e-06, + "loss": 1.0453, + "step": 27907 + }, + { + "epoch": 0.67, + "learning_rate": 5.1554844793153005e-06, + "loss": 0.9179, + "step": 27908 + }, + { + "epoch": 0.67, + "learning_rate": 5.1548031031027545e-06, + "loss": 0.9116, + "step": 27909 + }, + { + "epoch": 0.67, + "learning_rate": 5.154121756284879e-06, + "loss": 1.0416, + "step": 27910 + }, + { + "epoch": 0.67, + "learning_rate": 5.15344043886581e-06, + "loss": 1.3186, + "step": 27911 + }, + { + "epoch": 0.67, + "learning_rate": 5.152759150849674e-06, + "loss": 1.1476, + "step": 27912 + }, + { + "epoch": 0.67, + "learning_rate": 5.1520778922406125e-06, + "loss": 0.9748, + "step": 27913 + }, + { + "epoch": 0.67, + "learning_rate": 5.1513966630427495e-06, + "loss": 0.9736, + "step": 27914 + }, + { + "epoch": 0.67, + "learning_rate": 5.150715463260223e-06, + "loss": 1.0629, + "step": 27915 + }, + { + "epoch": 0.67, + "learning_rate": 5.150034292897169e-06, + "loss": 1.0033, + "step": 27916 + }, + { + "epoch": 0.67, + "learning_rate": 5.149353151957712e-06, + "loss": 1.2089, + "step": 27917 + }, + { + "epoch": 0.67, + "learning_rate": 5.14867204044599e-06, + "loss": 0.9911, + "step": 27918 + }, + { + "epoch": 0.67, + "learning_rate": 5.147990958366136e-06, + "loss": 1.0331, + "step": 27919 + }, + { + "epoch": 0.67, + "learning_rate": 5.147309905722276e-06, + "loss": 1.083, + "step": 27920 + }, + { + "epoch": 0.67, + "learning_rate": 5.14662888251855e-06, + "loss": 1.098, + "step": 27921 + }, + { + "epoch": 0.67, + "learning_rate": 5.14594788875908e-06, + "loss": 1.0408, + "step": 27922 + }, + { + "epoch": 0.67, + "learning_rate": 5.145266924448002e-06, + "loss": 0.991, + "step": 27923 + }, + { + "epoch": 0.67, + "learning_rate": 5.1445859895894525e-06, + "loss": 0.9952, + "step": 27924 + }, + { + "epoch": 0.67, + "learning_rate": 5.143905084187553e-06, + "loss": 1.0126, + "step": 27925 + }, + { + "epoch": 0.67, + "learning_rate": 5.14322420824644e-06, + "loss": 1.065, + "step": 27926 + }, + { + "epoch": 0.67, + "learning_rate": 5.142543361770247e-06, + "loss": 1.2149, + "step": 27927 + }, + { + "epoch": 0.67, + "learning_rate": 5.1418625447630966e-06, + "loss": 1.1619, + "step": 27928 + }, + { + "epoch": 0.67, + "learning_rate": 5.141181757229125e-06, + "loss": 1.1728, + "step": 27929 + }, + { + "epoch": 0.67, + "learning_rate": 5.140500999172463e-06, + "loss": 1.0476, + "step": 27930 + }, + { + "epoch": 0.67, + "learning_rate": 5.139820270597236e-06, + "loss": 1.2954, + "step": 27931 + }, + { + "epoch": 0.67, + "learning_rate": 5.139139571507579e-06, + "loss": 1.3182, + "step": 27932 + }, + { + "epoch": 0.67, + "learning_rate": 5.138458901907614e-06, + "loss": 1.118, + "step": 27933 + }, + { + "epoch": 0.67, + "learning_rate": 5.137778261801477e-06, + "loss": 0.8541, + "step": 27934 + }, + { + "epoch": 0.67, + "learning_rate": 5.137097651193299e-06, + "loss": 0.8564, + "step": 27935 + }, + { + "epoch": 0.67, + "learning_rate": 5.136417070087202e-06, + "loss": 1.2553, + "step": 27936 + }, + { + "epoch": 0.67, + "learning_rate": 5.135736518487318e-06, + "loss": 0.8363, + "step": 27937 + }, + { + "epoch": 0.67, + "learning_rate": 5.135055996397782e-06, + "loss": 1.0514, + "step": 27938 + }, + { + "epoch": 0.67, + "learning_rate": 5.1343755038227105e-06, + "loss": 0.6453, + "step": 27939 + }, + { + "epoch": 0.67, + "learning_rate": 5.133695040766244e-06, + "loss": 1.0524, + "step": 27940 + }, + { + "epoch": 0.67, + "learning_rate": 5.1330146072325e-06, + "loss": 1.0611, + "step": 27941 + }, + { + "epoch": 0.67, + "learning_rate": 5.132334203225613e-06, + "loss": 1.1983, + "step": 27942 + }, + { + "epoch": 0.67, + "learning_rate": 5.131653828749713e-06, + "loss": 1.0042, + "step": 27943 + }, + { + "epoch": 0.67, + "learning_rate": 5.13097348380892e-06, + "loss": 0.8088, + "step": 27944 + }, + { + "epoch": 0.67, + "learning_rate": 5.130293168407366e-06, + "loss": 1.1709, + "step": 27945 + }, + { + "epoch": 0.67, + "learning_rate": 5.1296128825491805e-06, + "loss": 1.1472, + "step": 27946 + }, + { + "epoch": 0.67, + "learning_rate": 5.128932626238485e-06, + "loss": 1.0614, + "step": 27947 + }, + { + "epoch": 0.67, + "learning_rate": 5.128252399479414e-06, + "loss": 1.0583, + "step": 27948 + }, + { + "epoch": 0.67, + "learning_rate": 5.127572202276086e-06, + "loss": 0.7471, + "step": 27949 + }, + { + "epoch": 0.67, + "learning_rate": 5.12689203463263e-06, + "loss": 1.1821, + "step": 27950 + }, + { + "epoch": 0.67, + "learning_rate": 5.126211896553178e-06, + "loss": 1.0172, + "step": 27951 + }, + { + "epoch": 0.67, + "learning_rate": 5.1255317880418485e-06, + "loss": 0.888, + "step": 27952 + }, + { + "epoch": 0.67, + "learning_rate": 5.124851709102771e-06, + "loss": 0.8671, + "step": 27953 + }, + { + "epoch": 0.67, + "learning_rate": 5.124171659740076e-06, + "loss": 1.1597, + "step": 27954 + }, + { + "epoch": 0.67, + "learning_rate": 5.123491639957881e-06, + "loss": 0.8352, + "step": 27955 + }, + { + "epoch": 0.67, + "learning_rate": 5.122811649760318e-06, + "loss": 1.1417, + "step": 27956 + }, + { + "epoch": 0.67, + "learning_rate": 5.122131689151506e-06, + "loss": 0.8759, + "step": 27957 + }, + { + "epoch": 0.67, + "learning_rate": 5.121451758135574e-06, + "loss": 1.0276, + "step": 27958 + }, + { + "epoch": 0.67, + "learning_rate": 5.12077185671665e-06, + "loss": 1.3258, + "step": 27959 + }, + { + "epoch": 0.67, + "learning_rate": 5.120091984898852e-06, + "loss": 1.1883, + "step": 27960 + }, + { + "epoch": 0.67, + "learning_rate": 5.119412142686308e-06, + "loss": 1.2416, + "step": 27961 + }, + { + "epoch": 0.67, + "learning_rate": 5.118732330083147e-06, + "loss": 1.023, + "step": 27962 + }, + { + "epoch": 0.67, + "learning_rate": 5.118052547093484e-06, + "loss": 0.9581, + "step": 27963 + }, + { + "epoch": 0.67, + "learning_rate": 5.117372793721451e-06, + "loss": 0.9082, + "step": 27964 + }, + { + "epoch": 0.67, + "learning_rate": 5.116693069971167e-06, + "loss": 1.1563, + "step": 27965 + }, + { + "epoch": 0.67, + "learning_rate": 5.116013375846756e-06, + "loss": 0.9723, + "step": 27966 + }, + { + "epoch": 0.67, + "learning_rate": 5.115333711352348e-06, + "loss": 0.9321, + "step": 27967 + }, + { + "epoch": 0.67, + "learning_rate": 5.114654076492057e-06, + "loss": 1.1421, + "step": 27968 + }, + { + "epoch": 0.67, + "learning_rate": 5.113974471270011e-06, + "loss": 1.2019, + "step": 27969 + }, + { + "epoch": 0.67, + "learning_rate": 5.113294895690335e-06, + "loss": 1.1833, + "step": 27970 + }, + { + "epoch": 0.67, + "learning_rate": 5.112615349757147e-06, + "loss": 1.002, + "step": 27971 + }, + { + "epoch": 0.67, + "learning_rate": 5.111935833474573e-06, + "loss": 1.1611, + "step": 27972 + }, + { + "epoch": 0.67, + "learning_rate": 5.111256346846737e-06, + "loss": 1.1707, + "step": 27973 + }, + { + "epoch": 0.67, + "learning_rate": 5.110576889877754e-06, + "loss": 1.0422, + "step": 27974 + }, + { + "epoch": 0.67, + "learning_rate": 5.109897462571757e-06, + "loss": 1.1495, + "step": 27975 + }, + { + "epoch": 0.67, + "learning_rate": 5.109218064932857e-06, + "loss": 0.9012, + "step": 27976 + }, + { + "epoch": 0.67, + "learning_rate": 5.1085386969651806e-06, + "loss": 1.1883, + "step": 27977 + }, + { + "epoch": 0.67, + "learning_rate": 5.107859358672853e-06, + "loss": 1.0492, + "step": 27978 + }, + { + "epoch": 0.67, + "learning_rate": 5.107180050059989e-06, + "loss": 1.1137, + "step": 27979 + }, + { + "epoch": 0.67, + "learning_rate": 5.106500771130711e-06, + "loss": 0.8037, + "step": 27980 + }, + { + "epoch": 0.67, + "learning_rate": 5.105821521889147e-06, + "loss": 1.1005, + "step": 27981 + }, + { + "epoch": 0.67, + "learning_rate": 5.1051423023394095e-06, + "loss": 1.0045, + "step": 27982 + }, + { + "epoch": 0.67, + "learning_rate": 5.104463112485624e-06, + "loss": 1.2144, + "step": 27983 + }, + { + "epoch": 0.67, + "learning_rate": 5.1037839523319086e-06, + "loss": 0.7481, + "step": 27984 + }, + { + "epoch": 0.67, + "learning_rate": 5.103104821882382e-06, + "loss": 1.0725, + "step": 27985 + }, + { + "epoch": 0.67, + "learning_rate": 5.102425721141171e-06, + "loss": 1.0291, + "step": 27986 + }, + { + "epoch": 0.67, + "learning_rate": 5.101746650112387e-06, + "loss": 1.2161, + "step": 27987 + }, + { + "epoch": 0.67, + "learning_rate": 5.101067608800154e-06, + "loss": 1.2522, + "step": 27988 + }, + { + "epoch": 0.67, + "learning_rate": 5.100388597208597e-06, + "loss": 0.8482, + "step": 27989 + }, + { + "epoch": 0.67, + "learning_rate": 5.099709615341824e-06, + "loss": 1.0934, + "step": 27990 + }, + { + "epoch": 0.67, + "learning_rate": 5.0990306632039644e-06, + "loss": 1.0967, + "step": 27991 + }, + { + "epoch": 0.67, + "learning_rate": 5.098351740799129e-06, + "loss": 0.9756, + "step": 27992 + }, + { + "epoch": 0.67, + "learning_rate": 5.097672848131441e-06, + "loss": 0.8455, + "step": 27993 + }, + { + "epoch": 0.67, + "learning_rate": 5.096993985205023e-06, + "loss": 1.1329, + "step": 27994 + }, + { + "epoch": 0.67, + "learning_rate": 5.096315152023985e-06, + "loss": 1.1048, + "step": 27995 + }, + { + "epoch": 0.67, + "learning_rate": 5.09563634859245e-06, + "loss": 0.7266, + "step": 27996 + }, + { + "epoch": 0.67, + "learning_rate": 5.094957574914539e-06, + "loss": 1.1594, + "step": 27997 + }, + { + "epoch": 0.67, + "learning_rate": 5.094278830994362e-06, + "loss": 0.8438, + "step": 27998 + }, + { + "epoch": 0.67, + "learning_rate": 5.093600116836042e-06, + "loss": 1.299, + "step": 27999 + }, + { + "epoch": 0.67, + "learning_rate": 5.092921432443696e-06, + "loss": 0.8694, + "step": 28000 + }, + { + "epoch": 0.67, + "learning_rate": 5.092242777821446e-06, + "loss": 0.8053, + "step": 28001 + }, + { + "epoch": 0.67, + "learning_rate": 5.0915641529734e-06, + "loss": 1.1863, + "step": 28002 + }, + { + "epoch": 0.67, + "learning_rate": 5.090885557903682e-06, + "loss": 1.1813, + "step": 28003 + }, + { + "epoch": 0.67, + "learning_rate": 5.090206992616408e-06, + "loss": 1.1122, + "step": 28004 + }, + { + "epoch": 0.67, + "learning_rate": 5.08952845711569e-06, + "loss": 0.7816, + "step": 28005 + }, + { + "epoch": 0.67, + "learning_rate": 5.088849951405647e-06, + "loss": 1.126, + "step": 28006 + }, + { + "epoch": 0.67, + "learning_rate": 5.088171475490402e-06, + "loss": 1.037, + "step": 28007 + }, + { + "epoch": 0.67, + "learning_rate": 5.087493029374061e-06, + "loss": 0.8198, + "step": 28008 + }, + { + "epoch": 0.67, + "learning_rate": 5.086814613060745e-06, + "loss": 0.9302, + "step": 28009 + }, + { + "epoch": 0.67, + "learning_rate": 5.08613622655457e-06, + "loss": 1.0884, + "step": 28010 + }, + { + "epoch": 0.67, + "learning_rate": 5.085457869859649e-06, + "loss": 1.145, + "step": 28011 + }, + { + "epoch": 0.67, + "learning_rate": 5.084779542980102e-06, + "loss": 1.0104, + "step": 28012 + }, + { + "epoch": 0.67, + "learning_rate": 5.084101245920039e-06, + "loss": 1.1065, + "step": 28013 + }, + { + "epoch": 0.67, + "learning_rate": 5.083422978683576e-06, + "loss": 1.2983, + "step": 28014 + }, + { + "epoch": 0.67, + "learning_rate": 5.082744741274834e-06, + "loss": 0.9167, + "step": 28015 + }, + { + "epoch": 0.67, + "learning_rate": 5.082066533697919e-06, + "loss": 1.0271, + "step": 28016 + }, + { + "epoch": 0.67, + "learning_rate": 5.0813883559569486e-06, + "loss": 0.7305, + "step": 28017 + }, + { + "epoch": 0.67, + "learning_rate": 5.080710208056043e-06, + "loss": 1.0678, + "step": 28018 + }, + { + "epoch": 0.67, + "learning_rate": 5.080032089999309e-06, + "loss": 0.8992, + "step": 28019 + }, + { + "epoch": 0.67, + "learning_rate": 5.079354001790865e-06, + "loss": 1.0239, + "step": 28020 + }, + { + "epoch": 0.67, + "learning_rate": 5.078675943434819e-06, + "loss": 1.0677, + "step": 28021 + }, + { + "epoch": 0.67, + "learning_rate": 5.077997914935289e-06, + "loss": 1.1468, + "step": 28022 + }, + { + "epoch": 0.67, + "learning_rate": 5.077319916296392e-06, + "loss": 1.0195, + "step": 28023 + }, + { + "epoch": 0.67, + "learning_rate": 5.0766419475222335e-06, + "loss": 1.1713, + "step": 28024 + }, + { + "epoch": 0.67, + "learning_rate": 5.075964008616929e-06, + "loss": 1.0993, + "step": 28025 + }, + { + "epoch": 0.67, + "learning_rate": 5.075286099584598e-06, + "loss": 1.0331, + "step": 28026 + }, + { + "epoch": 0.67, + "learning_rate": 5.0746082204293435e-06, + "loss": 1.0545, + "step": 28027 + }, + { + "epoch": 0.67, + "learning_rate": 5.073930371155286e-06, + "loss": 1.0725, + "step": 28028 + }, + { + "epoch": 0.67, + "learning_rate": 5.07325255176653e-06, + "loss": 0.7894, + "step": 28029 + }, + { + "epoch": 0.67, + "learning_rate": 5.072574762267193e-06, + "loss": 1.1809, + "step": 28030 + }, + { + "epoch": 0.67, + "learning_rate": 5.071897002661389e-06, + "loss": 1.2386, + "step": 28031 + }, + { + "epoch": 0.67, + "learning_rate": 5.071219272953223e-06, + "loss": 1.0626, + "step": 28032 + }, + { + "epoch": 0.67, + "learning_rate": 5.07054157314681e-06, + "loss": 1.0714, + "step": 28033 + }, + { + "epoch": 0.67, + "learning_rate": 5.069863903246266e-06, + "loss": 1.0399, + "step": 28034 + }, + { + "epoch": 0.67, + "learning_rate": 5.069186263255694e-06, + "loss": 1.0382, + "step": 28035 + }, + { + "epoch": 0.67, + "learning_rate": 5.068508653179214e-06, + "loss": 1.1668, + "step": 28036 + }, + { + "epoch": 0.67, + "learning_rate": 5.067831073020928e-06, + "loss": 0.8797, + "step": 28037 + }, + { + "epoch": 0.67, + "learning_rate": 5.06715352278495e-06, + "loss": 1.2504, + "step": 28038 + }, + { + "epoch": 0.67, + "learning_rate": 5.066476002475395e-06, + "loss": 0.8842, + "step": 28039 + }, + { + "epoch": 0.67, + "learning_rate": 5.065798512096367e-06, + "loss": 0.9945, + "step": 28040 + }, + { + "epoch": 0.67, + "learning_rate": 5.065121051651978e-06, + "loss": 1.0382, + "step": 28041 + }, + { + "epoch": 0.67, + "learning_rate": 5.064443621146342e-06, + "loss": 0.8569, + "step": 28042 + }, + { + "epoch": 0.67, + "learning_rate": 5.063766220583563e-06, + "loss": 1.0118, + "step": 28043 + }, + { + "epoch": 0.67, + "learning_rate": 5.063088849967757e-06, + "loss": 0.9639, + "step": 28044 + }, + { + "epoch": 0.67, + "learning_rate": 5.0624115093030245e-06, + "loss": 1.1119, + "step": 28045 + }, + { + "epoch": 0.67, + "learning_rate": 5.061734198593481e-06, + "loss": 1.1774, + "step": 28046 + }, + { + "epoch": 0.67, + "learning_rate": 5.061056917843239e-06, + "loss": 0.8489, + "step": 28047 + }, + { + "epoch": 0.67, + "learning_rate": 5.060379667056399e-06, + "loss": 1.098, + "step": 28048 + }, + { + "epoch": 0.67, + "learning_rate": 5.059702446237072e-06, + "loss": 0.9506, + "step": 28049 + }, + { + "epoch": 0.67, + "learning_rate": 5.059025255389373e-06, + "loss": 1.0123, + "step": 28050 + }, + { + "epoch": 0.67, + "learning_rate": 5.058348094517402e-06, + "loss": 1.0601, + "step": 28051 + }, + { + "epoch": 0.67, + "learning_rate": 5.057670963625272e-06, + "loss": 0.8312, + "step": 28052 + }, + { + "epoch": 0.67, + "learning_rate": 5.056993862717092e-06, + "loss": 0.8392, + "step": 28053 + }, + { + "epoch": 0.67, + "learning_rate": 5.056316791796964e-06, + "loss": 1.1077, + "step": 28054 + }, + { + "epoch": 0.67, + "learning_rate": 5.055639750869002e-06, + "loss": 0.9577, + "step": 28055 + }, + { + "epoch": 0.67, + "learning_rate": 5.054962739937308e-06, + "loss": 1.0408, + "step": 28056 + }, + { + "epoch": 0.67, + "learning_rate": 5.0542857590059925e-06, + "loss": 1.1089, + "step": 28057 + }, + { + "epoch": 0.67, + "learning_rate": 5.0536088080791645e-06, + "loss": 1.1614, + "step": 28058 + }, + { + "epoch": 0.67, + "learning_rate": 5.052931887160926e-06, + "loss": 1.0611, + "step": 28059 + }, + { + "epoch": 0.67, + "learning_rate": 5.052254996255386e-06, + "loss": 1.1403, + "step": 28060 + }, + { + "epoch": 0.67, + "learning_rate": 5.0515781353666535e-06, + "loss": 1.0705, + "step": 28061 + }, + { + "epoch": 0.67, + "learning_rate": 5.050901304498829e-06, + "loss": 1.0992, + "step": 28062 + }, + { + "epoch": 0.67, + "learning_rate": 5.050224503656027e-06, + "loss": 1.1325, + "step": 28063 + }, + { + "epoch": 0.67, + "learning_rate": 5.049547732842343e-06, + "loss": 1.0069, + "step": 28064 + }, + { + "epoch": 0.67, + "learning_rate": 5.048870992061892e-06, + "loss": 1.0665, + "step": 28065 + }, + { + "epoch": 0.67, + "learning_rate": 5.048194281318777e-06, + "loss": 1.1811, + "step": 28066 + }, + { + "epoch": 0.67, + "learning_rate": 5.0475176006171e-06, + "loss": 1.029, + "step": 28067 + }, + { + "epoch": 0.68, + "learning_rate": 5.046840949960968e-06, + "loss": 1.0216, + "step": 28068 + }, + { + "epoch": 0.68, + "learning_rate": 5.0461643293544915e-06, + "loss": 1.0951, + "step": 28069 + }, + { + "epoch": 0.68, + "learning_rate": 5.045487738801766e-06, + "loss": 0.9538, + "step": 28070 + }, + { + "epoch": 0.68, + "learning_rate": 5.0448111783069055e-06, + "loss": 0.9559, + "step": 28071 + }, + { + "epoch": 0.68, + "learning_rate": 5.044134647874006e-06, + "loss": 0.921, + "step": 28072 + }, + { + "epoch": 0.68, + "learning_rate": 5.043458147507176e-06, + "loss": 1.0588, + "step": 28073 + }, + { + "epoch": 0.68, + "learning_rate": 5.042781677210524e-06, + "loss": 1.0563, + "step": 28074 + }, + { + "epoch": 0.68, + "learning_rate": 5.042105236988145e-06, + "loss": 1.0976, + "step": 28075 + }, + { + "epoch": 0.68, + "learning_rate": 5.041428826844148e-06, + "loss": 1.2012, + "step": 28076 + }, + { + "epoch": 0.68, + "learning_rate": 5.0407524467826396e-06, + "loss": 0.9868, + "step": 28077 + }, + { + "epoch": 0.68, + "learning_rate": 5.040076096807715e-06, + "loss": 1.1057, + "step": 28078 + }, + { + "epoch": 0.68, + "learning_rate": 5.039399776923486e-06, + "loss": 0.7493, + "step": 28079 + }, + { + "epoch": 0.68, + "learning_rate": 5.038723487134049e-06, + "loss": 0.9772, + "step": 28080 + }, + { + "epoch": 0.68, + "learning_rate": 5.038047227443509e-06, + "loss": 1.1575, + "step": 28081 + }, + { + "epoch": 0.68, + "learning_rate": 5.0373709978559725e-06, + "loss": 1.3789, + "step": 28082 + }, + { + "epoch": 0.68, + "learning_rate": 5.036694798375537e-06, + "loss": 1.2157, + "step": 28083 + }, + { + "epoch": 0.68, + "learning_rate": 5.036018629006305e-06, + "loss": 1.1084, + "step": 28084 + }, + { + "epoch": 0.68, + "learning_rate": 5.035342489752384e-06, + "loss": 0.9519, + "step": 28085 + }, + { + "epoch": 0.68, + "learning_rate": 5.03466638061787e-06, + "loss": 0.9354, + "step": 28086 + }, + { + "epoch": 0.68, + "learning_rate": 5.033990301606869e-06, + "loss": 1.0661, + "step": 28087 + }, + { + "epoch": 0.68, + "learning_rate": 5.033314252723478e-06, + "loss": 1.1301, + "step": 28088 + }, + { + "epoch": 0.68, + "learning_rate": 5.032638233971802e-06, + "loss": 1.203, + "step": 28089 + }, + { + "epoch": 0.68, + "learning_rate": 5.031962245355943e-06, + "loss": 1.002, + "step": 28090 + }, + { + "epoch": 0.68, + "learning_rate": 5.0312862868799995e-06, + "loss": 1.1586, + "step": 28091 + }, + { + "epoch": 0.68, + "learning_rate": 5.030610358548071e-06, + "loss": 0.9259, + "step": 28092 + }, + { + "epoch": 0.68, + "learning_rate": 5.029934460364267e-06, + "loss": 1.0928, + "step": 28093 + }, + { + "epoch": 0.68, + "learning_rate": 5.029258592332676e-06, + "loss": 0.886, + "step": 28094 + }, + { + "epoch": 0.68, + "learning_rate": 5.0285827544574075e-06, + "loss": 1.1335, + "step": 28095 + }, + { + "epoch": 0.68, + "learning_rate": 5.027906946742556e-06, + "loss": 0.8588, + "step": 28096 + }, + { + "epoch": 0.68, + "learning_rate": 5.027231169192221e-06, + "loss": 1.1122, + "step": 28097 + }, + { + "epoch": 0.68, + "learning_rate": 5.026555421810511e-06, + "loss": 1.0027, + "step": 28098 + }, + { + "epoch": 0.68, + "learning_rate": 5.025879704601515e-06, + "loss": 1.3393, + "step": 28099 + }, + { + "epoch": 0.68, + "learning_rate": 5.025204017569336e-06, + "loss": 1.2825, + "step": 28100 + }, + { + "epoch": 0.68, + "learning_rate": 5.0245283607180776e-06, + "loss": 1.2241, + "step": 28101 + }, + { + "epoch": 0.68, + "learning_rate": 5.023852734051833e-06, + "loss": 1.1063, + "step": 28102 + }, + { + "epoch": 0.68, + "learning_rate": 5.023177137574702e-06, + "loss": 0.9384, + "step": 28103 + }, + { + "epoch": 0.68, + "learning_rate": 5.0225015712907894e-06, + "loss": 1.0901, + "step": 28104 + }, + { + "epoch": 0.68, + "learning_rate": 5.0218260352041845e-06, + "loss": 1.3849, + "step": 28105 + }, + { + "epoch": 0.68, + "learning_rate": 5.021150529318994e-06, + "loss": 0.8338, + "step": 28106 + }, + { + "epoch": 0.68, + "learning_rate": 5.02047505363931e-06, + "loss": 1.0068, + "step": 28107 + }, + { + "epoch": 0.68, + "learning_rate": 5.019799608169231e-06, + "loss": 1.3487, + "step": 28108 + }, + { + "epoch": 0.68, + "learning_rate": 5.0191241929128606e-06, + "loss": 0.956, + "step": 28109 + }, + { + "epoch": 0.68, + "learning_rate": 5.018448807874288e-06, + "loss": 1.132, + "step": 28110 + }, + { + "epoch": 0.68, + "learning_rate": 5.017773453057615e-06, + "loss": 1.019, + "step": 28111 + }, + { + "epoch": 0.68, + "learning_rate": 5.017098128466942e-06, + "loss": 1.0696, + "step": 28112 + }, + { + "epoch": 0.68, + "learning_rate": 5.016422834106359e-06, + "loss": 0.8672, + "step": 28113 + }, + { + "epoch": 0.68, + "learning_rate": 5.015747569979971e-06, + "loss": 1.2478, + "step": 28114 + }, + { + "epoch": 0.68, + "learning_rate": 5.015072336091866e-06, + "loss": 0.8443, + "step": 28115 + }, + { + "epoch": 0.68, + "learning_rate": 5.014397132446144e-06, + "loss": 1.191, + "step": 28116 + }, + { + "epoch": 0.68, + "learning_rate": 5.013721959046907e-06, + "loss": 0.9339, + "step": 28117 + }, + { + "epoch": 0.68, + "learning_rate": 5.0130468158982405e-06, + "loss": 1.2089, + "step": 28118 + }, + { + "epoch": 0.68, + "learning_rate": 5.012371703004246e-06, + "loss": 0.9914, + "step": 28119 + }, + { + "epoch": 0.68, + "learning_rate": 5.011696620369024e-06, + "loss": 1.0773, + "step": 28120 + }, + { + "epoch": 0.68, + "learning_rate": 5.011021567996661e-06, + "loss": 1.0889, + "step": 28121 + }, + { + "epoch": 0.68, + "learning_rate": 5.010346545891261e-06, + "loss": 0.8973, + "step": 28122 + }, + { + "epoch": 0.68, + "learning_rate": 5.009671554056911e-06, + "loss": 1.2919, + "step": 28123 + }, + { + "epoch": 0.68, + "learning_rate": 5.00899659249771e-06, + "loss": 1.2656, + "step": 28124 + }, + { + "epoch": 0.68, + "learning_rate": 5.008321661217756e-06, + "loss": 0.9846, + "step": 28125 + }, + { + "epoch": 0.68, + "learning_rate": 5.007646760221136e-06, + "loss": 1.0203, + "step": 28126 + }, + { + "epoch": 0.68, + "learning_rate": 5.00697188951195e-06, + "loss": 1.0819, + "step": 28127 + }, + { + "epoch": 0.68, + "learning_rate": 5.006297049094294e-06, + "loss": 1.238, + "step": 28128 + }, + { + "epoch": 0.68, + "learning_rate": 5.005622238972256e-06, + "loss": 1.0182, + "step": 28129 + }, + { + "epoch": 0.68, + "learning_rate": 5.004947459149937e-06, + "loss": 1.067, + "step": 28130 + }, + { + "epoch": 0.68, + "learning_rate": 5.004272709631424e-06, + "loss": 0.9823, + "step": 28131 + }, + { + "epoch": 0.68, + "learning_rate": 5.003597990420812e-06, + "loss": 1.1126, + "step": 28132 + }, + { + "epoch": 0.68, + "learning_rate": 5.002923301522199e-06, + "loss": 1.0185, + "step": 28133 + }, + { + "epoch": 0.68, + "learning_rate": 5.002248642939672e-06, + "loss": 0.9643, + "step": 28134 + }, + { + "epoch": 0.68, + "learning_rate": 5.001574014677327e-06, + "loss": 1.1106, + "step": 28135 + }, + { + "epoch": 0.68, + "learning_rate": 5.00089941673926e-06, + "loss": 1.2616, + "step": 28136 + }, + { + "epoch": 0.68, + "learning_rate": 5.000224849129558e-06, + "loss": 1.0916, + "step": 28137 + }, + { + "epoch": 0.68, + "learning_rate": 4.999550311852318e-06, + "loss": 0.8605, + "step": 28138 + }, + { + "epoch": 0.68, + "learning_rate": 4.998875804911627e-06, + "loss": 1.1153, + "step": 28139 + }, + { + "epoch": 0.68, + "learning_rate": 4.998201328311581e-06, + "loss": 1.1464, + "step": 28140 + }, + { + "epoch": 0.68, + "learning_rate": 4.9975268820562725e-06, + "loss": 0.9826, + "step": 28141 + }, + { + "epoch": 0.68, + "learning_rate": 4.99685246614979e-06, + "loss": 1.1092, + "step": 28142 + }, + { + "epoch": 0.68, + "learning_rate": 4.996178080596227e-06, + "loss": 0.8742, + "step": 28143 + }, + { + "epoch": 0.68, + "learning_rate": 4.995503725399677e-06, + "loss": 1.0242, + "step": 28144 + }, + { + "epoch": 0.68, + "learning_rate": 4.994829400564225e-06, + "loss": 1.0102, + "step": 28145 + }, + { + "epoch": 0.68, + "learning_rate": 4.994155106093966e-06, + "loss": 1.0162, + "step": 28146 + }, + { + "epoch": 0.68, + "learning_rate": 4.993480841992996e-06, + "loss": 1.0446, + "step": 28147 + }, + { + "epoch": 0.68, + "learning_rate": 4.992806608265395e-06, + "loss": 0.7748, + "step": 28148 + }, + { + "epoch": 0.68, + "learning_rate": 4.992132404915262e-06, + "loss": 1.1139, + "step": 28149 + }, + { + "epoch": 0.68, + "learning_rate": 4.99145823194668e-06, + "loss": 0.7665, + "step": 28150 + }, + { + "epoch": 0.68, + "learning_rate": 4.990784089363743e-06, + "loss": 1.2454, + "step": 28151 + }, + { + "epoch": 0.68, + "learning_rate": 4.990109977170545e-06, + "loss": 0.9724, + "step": 28152 + }, + { + "epoch": 0.68, + "learning_rate": 4.989435895371168e-06, + "loss": 0.8241, + "step": 28153 + }, + { + "epoch": 0.68, + "learning_rate": 4.988761843969703e-06, + "loss": 1.1278, + "step": 28154 + }, + { + "epoch": 0.68, + "learning_rate": 4.9880878229702456e-06, + "loss": 0.9923, + "step": 28155 + }, + { + "epoch": 0.68, + "learning_rate": 4.987413832376877e-06, + "loss": 1.0886, + "step": 28156 + }, + { + "epoch": 0.68, + "learning_rate": 4.986739872193694e-06, + "loss": 0.9355, + "step": 28157 + }, + { + "epoch": 0.68, + "learning_rate": 4.986065942424777e-06, + "loss": 1.0095, + "step": 28158 + }, + { + "epoch": 0.68, + "learning_rate": 4.9853920430742185e-06, + "loss": 1.1828, + "step": 28159 + }, + { + "epoch": 0.68, + "learning_rate": 4.984718174146111e-06, + "loss": 0.7925, + "step": 28160 + }, + { + "epoch": 0.68, + "learning_rate": 4.984044335644533e-06, + "loss": 1.022, + "step": 28161 + }, + { + "epoch": 0.68, + "learning_rate": 4.98337052757358e-06, + "loss": 0.9559, + "step": 28162 + }, + { + "epoch": 0.68, + "learning_rate": 4.9826967499373425e-06, + "loss": 0.9994, + "step": 28163 + }, + { + "epoch": 0.68, + "learning_rate": 4.982023002739899e-06, + "loss": 1.1617, + "step": 28164 + }, + { + "epoch": 0.68, + "learning_rate": 4.981349285985345e-06, + "loss": 0.729, + "step": 28165 + }, + { + "epoch": 0.68, + "learning_rate": 4.9806755996777565e-06, + "loss": 1.1402, + "step": 28166 + }, + { + "epoch": 0.68, + "learning_rate": 4.980001943821238e-06, + "loss": 1.105, + "step": 28167 + }, + { + "epoch": 0.68, + "learning_rate": 4.979328318419862e-06, + "loss": 1.1287, + "step": 28168 + }, + { + "epoch": 0.68, + "learning_rate": 4.9786547234777194e-06, + "loss": 0.9623, + "step": 28169 + }, + { + "epoch": 0.68, + "learning_rate": 4.977981158998901e-06, + "loss": 1.3324, + "step": 28170 + }, + { + "epoch": 0.68, + "learning_rate": 4.977307624987487e-06, + "loss": 0.9749, + "step": 28171 + }, + { + "epoch": 0.68, + "learning_rate": 4.976634121447567e-06, + "loss": 1.1287, + "step": 28172 + }, + { + "epoch": 0.68, + "learning_rate": 4.975960648383229e-06, + "loss": 1.0241, + "step": 28173 + }, + { + "epoch": 0.68, + "learning_rate": 4.975287205798551e-06, + "loss": 1.1226, + "step": 28174 + }, + { + "epoch": 0.68, + "learning_rate": 4.974613793697626e-06, + "loss": 0.9088, + "step": 28175 + }, + { + "epoch": 0.68, + "learning_rate": 4.973940412084539e-06, + "loss": 1.2875, + "step": 28176 + }, + { + "epoch": 0.68, + "learning_rate": 4.97326706096337e-06, + "loss": 1.0579, + "step": 28177 + }, + { + "epoch": 0.68, + "learning_rate": 4.972593740338212e-06, + "loss": 1.2009, + "step": 28178 + }, + { + "epoch": 0.68, + "learning_rate": 4.9719204502131415e-06, + "loss": 0.8673, + "step": 28179 + }, + { + "epoch": 0.68, + "learning_rate": 4.9712471905922465e-06, + "loss": 1.1386, + "step": 28180 + }, + { + "epoch": 0.68, + "learning_rate": 4.970573961479616e-06, + "loss": 0.9451, + "step": 28181 + }, + { + "epoch": 0.68, + "learning_rate": 4.969900762879325e-06, + "loss": 1.202, + "step": 28182 + }, + { + "epoch": 0.68, + "learning_rate": 4.969227594795465e-06, + "loss": 1.1147, + "step": 28183 + }, + { + "epoch": 0.68, + "learning_rate": 4.96855445723212e-06, + "loss": 1.0231, + "step": 28184 + }, + { + "epoch": 0.68, + "learning_rate": 4.967881350193369e-06, + "loss": 1.2074, + "step": 28185 + }, + { + "epoch": 0.68, + "learning_rate": 4.967208273683302e-06, + "loss": 1.0825, + "step": 28186 + }, + { + "epoch": 0.68, + "learning_rate": 4.966535227705994e-06, + "loss": 0.951, + "step": 28187 + }, + { + "epoch": 0.68, + "learning_rate": 4.965862212265533e-06, + "loss": 1.0528, + "step": 28188 + }, + { + "epoch": 0.68, + "learning_rate": 4.965189227366007e-06, + "loss": 1.0798, + "step": 28189 + }, + { + "epoch": 0.68, + "learning_rate": 4.964516273011488e-06, + "loss": 1.0981, + "step": 28190 + }, + { + "epoch": 0.68, + "learning_rate": 4.963843349206065e-06, + "loss": 1.1796, + "step": 28191 + }, + { + "epoch": 0.68, + "learning_rate": 4.963170455953824e-06, + "loss": 0.9953, + "step": 28192 + }, + { + "epoch": 0.68, + "learning_rate": 4.962497593258839e-06, + "loss": 1.0823, + "step": 28193 + }, + { + "epoch": 0.68, + "learning_rate": 4.9618247611252e-06, + "loss": 1.1063, + "step": 28194 + }, + { + "epoch": 0.68, + "learning_rate": 4.961151959556981e-06, + "loss": 1.2217, + "step": 28195 + }, + { + "epoch": 0.68, + "learning_rate": 4.960479188558268e-06, + "loss": 1.0311, + "step": 28196 + }, + { + "epoch": 0.68, + "learning_rate": 4.959806448133148e-06, + "loss": 1.3171, + "step": 28197 + }, + { + "epoch": 0.68, + "learning_rate": 4.9591337382856905e-06, + "loss": 1.0108, + "step": 28198 + }, + { + "epoch": 0.68, + "learning_rate": 4.9584610590199835e-06, + "loss": 1.085, + "step": 28199 + }, + { + "epoch": 0.68, + "learning_rate": 4.957788410340112e-06, + "loss": 1.0243, + "step": 28200 + }, + { + "epoch": 0.68, + "learning_rate": 4.9571157922501475e-06, + "loss": 0.9447, + "step": 28201 + }, + { + "epoch": 0.68, + "learning_rate": 4.95644320475418e-06, + "loss": 0.8291, + "step": 28202 + }, + { + "epoch": 0.68, + "learning_rate": 4.955770647856281e-06, + "loss": 0.8035, + "step": 28203 + }, + { + "epoch": 0.68, + "learning_rate": 4.9550981215605356e-06, + "loss": 0.8612, + "step": 28204 + }, + { + "epoch": 0.68, + "learning_rate": 4.954425625871026e-06, + "loss": 1.0729, + "step": 28205 + }, + { + "epoch": 0.68, + "learning_rate": 4.953753160791827e-06, + "loss": 0.8014, + "step": 28206 + }, + { + "epoch": 0.68, + "learning_rate": 4.953080726327019e-06, + "loss": 1.2482, + "step": 28207 + }, + { + "epoch": 0.68, + "learning_rate": 4.9524083224806875e-06, + "loss": 1.1359, + "step": 28208 + }, + { + "epoch": 0.68, + "learning_rate": 4.951735949256904e-06, + "loss": 1.0943, + "step": 28209 + }, + { + "epoch": 0.68, + "learning_rate": 4.951063606659754e-06, + "loss": 1.212, + "step": 28210 + }, + { + "epoch": 0.68, + "learning_rate": 4.950391294693311e-06, + "loss": 1.0565, + "step": 28211 + }, + { + "epoch": 0.68, + "learning_rate": 4.949719013361655e-06, + "loss": 1.1164, + "step": 28212 + }, + { + "epoch": 0.68, + "learning_rate": 4.949046762668871e-06, + "loss": 1.0221, + "step": 28213 + }, + { + "epoch": 0.68, + "learning_rate": 4.948374542619026e-06, + "loss": 1.0754, + "step": 28214 + }, + { + "epoch": 0.68, + "learning_rate": 4.947702353216207e-06, + "loss": 1.1283, + "step": 28215 + }, + { + "epoch": 0.68, + "learning_rate": 4.947030194464493e-06, + "loss": 1.0551, + "step": 28216 + }, + { + "epoch": 0.68, + "learning_rate": 4.9463580663679536e-06, + "loss": 1.1745, + "step": 28217 + }, + { + "epoch": 0.68, + "learning_rate": 4.945685968930675e-06, + "loss": 1.2008, + "step": 28218 + }, + { + "epoch": 0.68, + "learning_rate": 4.945013902156726e-06, + "loss": 1.0715, + "step": 28219 + }, + { + "epoch": 0.68, + "learning_rate": 4.94434186605019e-06, + "loss": 0.8959, + "step": 28220 + }, + { + "epoch": 0.68, + "learning_rate": 4.943669860615148e-06, + "loss": 1.1644, + "step": 28221 + }, + { + "epoch": 0.68, + "learning_rate": 4.942997885855666e-06, + "loss": 1.0816, + "step": 28222 + }, + { + "epoch": 0.68, + "learning_rate": 4.942325941775826e-06, + "loss": 1.1044, + "step": 28223 + }, + { + "epoch": 0.68, + "learning_rate": 4.94165402837971e-06, + "loss": 1.1721, + "step": 28224 + }, + { + "epoch": 0.68, + "learning_rate": 4.9409821456713834e-06, + "loss": 0.9955, + "step": 28225 + }, + { + "epoch": 0.68, + "learning_rate": 4.94031029365493e-06, + "loss": 1.0145, + "step": 28226 + }, + { + "epoch": 0.68, + "learning_rate": 4.939638472334427e-06, + "loss": 1.1267, + "step": 28227 + }, + { + "epoch": 0.68, + "learning_rate": 4.938966681713944e-06, + "loss": 0.937, + "step": 28228 + }, + { + "epoch": 0.68, + "learning_rate": 4.938294921797563e-06, + "loss": 1.5004, + "step": 28229 + }, + { + "epoch": 0.68, + "learning_rate": 4.937623192589351e-06, + "loss": 0.9995, + "step": 28230 + }, + { + "epoch": 0.68, + "learning_rate": 4.93695149409339e-06, + "loss": 1.1558, + "step": 28231 + }, + { + "epoch": 0.68, + "learning_rate": 4.936279826313757e-06, + "loss": 0.9925, + "step": 28232 + }, + { + "epoch": 0.68, + "learning_rate": 4.93560818925452e-06, + "loss": 1.1515, + "step": 28233 + }, + { + "epoch": 0.68, + "learning_rate": 4.934936582919756e-06, + "loss": 0.9843, + "step": 28234 + }, + { + "epoch": 0.68, + "learning_rate": 4.934265007313544e-06, + "loss": 1.1371, + "step": 28235 + }, + { + "epoch": 0.68, + "learning_rate": 4.9335934624399505e-06, + "loss": 1.004, + "step": 28236 + }, + { + "epoch": 0.68, + "learning_rate": 4.9329219483030585e-06, + "loss": 1.1333, + "step": 28237 + }, + { + "epoch": 0.68, + "learning_rate": 4.9322504649069335e-06, + "loss": 1.0291, + "step": 28238 + }, + { + "epoch": 0.68, + "learning_rate": 4.931579012255653e-06, + "loss": 0.772, + "step": 28239 + }, + { + "epoch": 0.68, + "learning_rate": 4.9309075903532935e-06, + "loss": 1.0151, + "step": 28240 + }, + { + "epoch": 0.68, + "learning_rate": 4.930236199203922e-06, + "loss": 1.1834, + "step": 28241 + }, + { + "epoch": 0.68, + "learning_rate": 4.929564838811615e-06, + "loss": 0.9315, + "step": 28242 + }, + { + "epoch": 0.68, + "learning_rate": 4.928893509180449e-06, + "loss": 0.4878, + "step": 28243 + }, + { + "epoch": 0.68, + "learning_rate": 4.928222210314489e-06, + "loss": 1.1722, + "step": 28244 + }, + { + "epoch": 0.68, + "learning_rate": 4.927550942217817e-06, + "loss": 0.999, + "step": 28245 + }, + { + "epoch": 0.68, + "learning_rate": 4.926879704894495e-06, + "loss": 1.0087, + "step": 28246 + }, + { + "epoch": 0.68, + "learning_rate": 4.926208498348601e-06, + "loss": 1.309, + "step": 28247 + }, + { + "epoch": 0.68, + "learning_rate": 4.9255373225842115e-06, + "loss": 0.904, + "step": 28248 + }, + { + "epoch": 0.68, + "learning_rate": 4.924866177605389e-06, + "loss": 0.8747, + "step": 28249 + }, + { + "epoch": 0.68, + "learning_rate": 4.92419506341621e-06, + "loss": 1.1598, + "step": 28250 + }, + { + "epoch": 0.68, + "learning_rate": 4.9235239800207495e-06, + "loss": 1.2082, + "step": 28251 + }, + { + "epoch": 0.68, + "learning_rate": 4.92285292742307e-06, + "loss": 0.7577, + "step": 28252 + }, + { + "epoch": 0.68, + "learning_rate": 4.9221819056272524e-06, + "loss": 0.7894, + "step": 28253 + }, + { + "epoch": 0.68, + "learning_rate": 4.921510914637358e-06, + "loss": 0.8997, + "step": 28254 + }, + { + "epoch": 0.68, + "learning_rate": 4.920839954457463e-06, + "loss": 1.0208, + "step": 28255 + }, + { + "epoch": 0.68, + "learning_rate": 4.920169025091642e-06, + "loss": 0.8317, + "step": 28256 + }, + { + "epoch": 0.68, + "learning_rate": 4.9194981265439555e-06, + "loss": 1.0172, + "step": 28257 + }, + { + "epoch": 0.68, + "learning_rate": 4.918827258818479e-06, + "loss": 1.0801, + "step": 28258 + }, + { + "epoch": 0.68, + "learning_rate": 4.918156421919288e-06, + "loss": 1.1028, + "step": 28259 + }, + { + "epoch": 0.68, + "learning_rate": 4.9174856158504405e-06, + "loss": 0.8831, + "step": 28260 + }, + { + "epoch": 0.68, + "learning_rate": 4.916814840616018e-06, + "loss": 0.8833, + "step": 28261 + }, + { + "epoch": 0.68, + "learning_rate": 4.916144096220079e-06, + "loss": 0.8185, + "step": 28262 + }, + { + "epoch": 0.68, + "learning_rate": 4.915473382666698e-06, + "loss": 0.9322, + "step": 28263 + }, + { + "epoch": 0.68, + "learning_rate": 4.914802699959949e-06, + "loss": 0.838, + "step": 28264 + }, + { + "epoch": 0.68, + "learning_rate": 4.9141320481038914e-06, + "loss": 1.1689, + "step": 28265 + }, + { + "epoch": 0.68, + "learning_rate": 4.913461427102598e-06, + "loss": 1.1436, + "step": 28266 + }, + { + "epoch": 0.68, + "learning_rate": 4.9127908369601415e-06, + "loss": 0.9797, + "step": 28267 + }, + { + "epoch": 0.68, + "learning_rate": 4.912120277680583e-06, + "loss": 1.2979, + "step": 28268 + }, + { + "epoch": 0.68, + "learning_rate": 4.911449749267994e-06, + "loss": 0.9911, + "step": 28269 + }, + { + "epoch": 0.68, + "learning_rate": 4.910779251726447e-06, + "loss": 0.8966, + "step": 28270 + }, + { + "epoch": 0.68, + "learning_rate": 4.91010878506e-06, + "loss": 1.0709, + "step": 28271 + }, + { + "epoch": 0.68, + "learning_rate": 4.90943834927273e-06, + "loss": 0.993, + "step": 28272 + }, + { + "epoch": 0.68, + "learning_rate": 4.908767944368697e-06, + "loss": 1.215, + "step": 28273 + }, + { + "epoch": 0.68, + "learning_rate": 4.908097570351971e-06, + "loss": 1.2722, + "step": 28274 + }, + { + "epoch": 0.68, + "learning_rate": 4.907427227226623e-06, + "loss": 1.1781, + "step": 28275 + }, + { + "epoch": 0.68, + "learning_rate": 4.906756914996712e-06, + "loss": 1.0281, + "step": 28276 + }, + { + "epoch": 0.68, + "learning_rate": 4.90608663366631e-06, + "loss": 1.0183, + "step": 28277 + }, + { + "epoch": 0.68, + "learning_rate": 4.905416383239485e-06, + "loss": 0.9661, + "step": 28278 + }, + { + "epoch": 0.68, + "learning_rate": 4.904746163720298e-06, + "loss": 1.2696, + "step": 28279 + }, + { + "epoch": 0.68, + "learning_rate": 4.90407597511282e-06, + "loss": 1.3785, + "step": 28280 + }, + { + "epoch": 0.68, + "learning_rate": 4.903405817421111e-06, + "loss": 1.2072, + "step": 28281 + }, + { + "epoch": 0.68, + "learning_rate": 4.90273569064924e-06, + "loss": 0.8785, + "step": 28282 + }, + { + "epoch": 0.68, + "learning_rate": 4.902065594801277e-06, + "loss": 0.9844, + "step": 28283 + }, + { + "epoch": 0.68, + "learning_rate": 4.901395529881278e-06, + "loss": 1.1435, + "step": 28284 + }, + { + "epoch": 0.68, + "learning_rate": 4.900725495893313e-06, + "loss": 1.1849, + "step": 28285 + }, + { + "epoch": 0.68, + "learning_rate": 4.9000554928414525e-06, + "loss": 1.1849, + "step": 28286 + }, + { + "epoch": 0.68, + "learning_rate": 4.89938552072975e-06, + "loss": 1.1234, + "step": 28287 + }, + { + "epoch": 0.68, + "learning_rate": 4.898715579562281e-06, + "loss": 1.2336, + "step": 28288 + }, + { + "epoch": 0.68, + "learning_rate": 4.8980456693431e-06, + "loss": 1.2775, + "step": 28289 + }, + { + "epoch": 0.68, + "learning_rate": 4.8973757900762765e-06, + "loss": 0.9604, + "step": 28290 + }, + { + "epoch": 0.68, + "learning_rate": 4.8967059417658765e-06, + "loss": 0.976, + "step": 28291 + }, + { + "epoch": 0.68, + "learning_rate": 4.89603612441596e-06, + "loss": 0.8999, + "step": 28292 + }, + { + "epoch": 0.68, + "learning_rate": 4.895366338030589e-06, + "loss": 1.2694, + "step": 28293 + }, + { + "epoch": 0.68, + "learning_rate": 4.894696582613835e-06, + "loss": 1.1438, + "step": 28294 + }, + { + "epoch": 0.68, + "learning_rate": 4.894026858169753e-06, + "loss": 1.0291, + "step": 28295 + }, + { + "epoch": 0.68, + "learning_rate": 4.89335716470241e-06, + "loss": 0.9959, + "step": 28296 + }, + { + "epoch": 0.68, + "learning_rate": 4.892687502215867e-06, + "loss": 0.9317, + "step": 28297 + }, + { + "epoch": 0.68, + "learning_rate": 4.8920178707141866e-06, + "loss": 1.1696, + "step": 28298 + }, + { + "epoch": 0.68, + "learning_rate": 4.8913482702014355e-06, + "loss": 1.1315, + "step": 28299 + }, + { + "epoch": 0.68, + "learning_rate": 4.890678700681669e-06, + "loss": 0.9209, + "step": 28300 + }, + { + "epoch": 0.68, + "learning_rate": 4.890009162158954e-06, + "loss": 0.8794, + "step": 28301 + }, + { + "epoch": 0.68, + "learning_rate": 4.889339654637355e-06, + "loss": 0.9802, + "step": 28302 + }, + { + "epoch": 0.68, + "learning_rate": 4.888670178120925e-06, + "loss": 0.8631, + "step": 28303 + }, + { + "epoch": 0.68, + "learning_rate": 4.8880007326137365e-06, + "loss": 0.9808, + "step": 28304 + }, + { + "epoch": 0.68, + "learning_rate": 4.887331318119841e-06, + "loss": 1.0553, + "step": 28305 + }, + { + "epoch": 0.68, + "learning_rate": 4.886661934643303e-06, + "loss": 1.1593, + "step": 28306 + }, + { + "epoch": 0.68, + "learning_rate": 4.885992582188189e-06, + "loss": 0.6724, + "step": 28307 + }, + { + "epoch": 0.68, + "learning_rate": 4.885323260758551e-06, + "loss": 0.7934, + "step": 28308 + }, + { + "epoch": 0.68, + "learning_rate": 4.884653970358454e-06, + "loss": 1.031, + "step": 28309 + }, + { + "epoch": 0.68, + "learning_rate": 4.883984710991961e-06, + "loss": 1.1431, + "step": 28310 + }, + { + "epoch": 0.68, + "learning_rate": 4.8833154826631255e-06, + "loss": 1.1554, + "step": 28311 + }, + { + "epoch": 0.68, + "learning_rate": 4.882646285376017e-06, + "loss": 0.9335, + "step": 28312 + }, + { + "epoch": 0.68, + "learning_rate": 4.881977119134685e-06, + "loss": 1.0524, + "step": 28313 + }, + { + "epoch": 0.68, + "learning_rate": 4.881307983943193e-06, + "loss": 1.022, + "step": 28314 + }, + { + "epoch": 0.68, + "learning_rate": 4.880638879805607e-06, + "loss": 0.9567, + "step": 28315 + }, + { + "epoch": 0.68, + "learning_rate": 4.879969806725976e-06, + "loss": 1.2272, + "step": 28316 + }, + { + "epoch": 0.68, + "learning_rate": 4.879300764708363e-06, + "loss": 0.9347, + "step": 28317 + }, + { + "epoch": 0.68, + "learning_rate": 4.878631753756834e-06, + "loss": 0.9709, + "step": 28318 + }, + { + "epoch": 0.68, + "learning_rate": 4.877962773875436e-06, + "loss": 0.9675, + "step": 28319 + }, + { + "epoch": 0.68, + "learning_rate": 4.877293825068233e-06, + "loss": 0.788, + "step": 28320 + }, + { + "epoch": 0.68, + "learning_rate": 4.876624907339288e-06, + "loss": 1.1746, + "step": 28321 + }, + { + "epoch": 0.68, + "learning_rate": 4.875956020692651e-06, + "loss": 0.9757, + "step": 28322 + }, + { + "epoch": 0.68, + "learning_rate": 4.875287165132387e-06, + "loss": 1.0928, + "step": 28323 + }, + { + "epoch": 0.68, + "learning_rate": 4.874618340662546e-06, + "loss": 1.1319, + "step": 28324 + }, + { + "epoch": 0.68, + "learning_rate": 4.87394954728719e-06, + "loss": 0.6758, + "step": 28325 + }, + { + "epoch": 0.68, + "learning_rate": 4.87328078501038e-06, + "loss": 1.1472, + "step": 28326 + }, + { + "epoch": 0.68, + "learning_rate": 4.8726120538361665e-06, + "loss": 1.0102, + "step": 28327 + }, + { + "epoch": 0.68, + "learning_rate": 4.87194335376861e-06, + "loss": 0.8959, + "step": 28328 + }, + { + "epoch": 0.68, + "learning_rate": 4.87127468481177e-06, + "loss": 1.1545, + "step": 28329 + }, + { + "epoch": 0.68, + "learning_rate": 4.8706060469696966e-06, + "loss": 1.2465, + "step": 28330 + }, + { + "epoch": 0.68, + "learning_rate": 4.869937440246454e-06, + "loss": 1.0148, + "step": 28331 + }, + { + "epoch": 0.68, + "learning_rate": 4.869268864646086e-06, + "loss": 1.115, + "step": 28332 + }, + { + "epoch": 0.68, + "learning_rate": 4.868600320172665e-06, + "loss": 1.0956, + "step": 28333 + }, + { + "epoch": 0.68, + "learning_rate": 4.867931806830235e-06, + "loss": 1.1516, + "step": 28334 + }, + { + "epoch": 0.68, + "learning_rate": 4.867263324622855e-06, + "loss": 0.8383, + "step": 28335 + }, + { + "epoch": 0.68, + "learning_rate": 4.866594873554585e-06, + "loss": 1.2161, + "step": 28336 + }, + { + "epoch": 0.68, + "learning_rate": 4.865926453629474e-06, + "loss": 1.1589, + "step": 28337 + }, + { + "epoch": 0.68, + "learning_rate": 4.865258064851579e-06, + "loss": 1.0877, + "step": 28338 + }, + { + "epoch": 0.68, + "learning_rate": 4.864589707224958e-06, + "loss": 0.8797, + "step": 28339 + }, + { + "epoch": 0.68, + "learning_rate": 4.863921380753661e-06, + "loss": 1.0409, + "step": 28340 + }, + { + "epoch": 0.68, + "learning_rate": 4.8632530854417485e-06, + "loss": 1.1636, + "step": 28341 + }, + { + "epoch": 0.68, + "learning_rate": 4.862584821293267e-06, + "loss": 0.9935, + "step": 28342 + }, + { + "epoch": 0.68, + "learning_rate": 4.861916588312275e-06, + "loss": 1.2362, + "step": 28343 + }, + { + "epoch": 0.68, + "learning_rate": 4.861248386502831e-06, + "loss": 0.785, + "step": 28344 + }, + { + "epoch": 0.68, + "learning_rate": 4.86058021586898e-06, + "loss": 1.0806, + "step": 28345 + }, + { + "epoch": 0.68, + "learning_rate": 4.85991207641478e-06, + "loss": 1.1555, + "step": 28346 + }, + { + "epoch": 0.68, + "learning_rate": 4.859243968144288e-06, + "loss": 0.8879, + "step": 28347 + }, + { + "epoch": 0.68, + "learning_rate": 4.858575891061551e-06, + "loss": 1.1045, + "step": 28348 + }, + { + "epoch": 0.68, + "learning_rate": 4.857907845170624e-06, + "loss": 0.9412, + "step": 28349 + }, + { + "epoch": 0.68, + "learning_rate": 4.857239830475564e-06, + "loss": 1.1134, + "step": 28350 + }, + { + "epoch": 0.68, + "learning_rate": 4.856571846980417e-06, + "loss": 0.8216, + "step": 28351 + }, + { + "epoch": 0.68, + "learning_rate": 4.855903894689242e-06, + "loss": 1.2848, + "step": 28352 + }, + { + "epoch": 0.68, + "learning_rate": 4.855235973606085e-06, + "loss": 1.2364, + "step": 28353 + }, + { + "epoch": 0.68, + "learning_rate": 4.854568083735001e-06, + "loss": 1.236, + "step": 28354 + }, + { + "epoch": 0.68, + "learning_rate": 4.853900225080046e-06, + "loss": 1.2579, + "step": 28355 + }, + { + "epoch": 0.68, + "learning_rate": 4.853232397645263e-06, + "loss": 0.835, + "step": 28356 + }, + { + "epoch": 0.68, + "learning_rate": 4.85256460143471e-06, + "loss": 1.252, + "step": 28357 + }, + { + "epoch": 0.68, + "learning_rate": 4.85189683645244e-06, + "loss": 0.9005, + "step": 28358 + }, + { + "epoch": 0.68, + "learning_rate": 4.851229102702497e-06, + "loss": 1.1733, + "step": 28359 + }, + { + "epoch": 0.68, + "learning_rate": 4.85056140018894e-06, + "loss": 1.1366, + "step": 28360 + }, + { + "epoch": 0.68, + "learning_rate": 4.849893728915812e-06, + "loss": 0.9952, + "step": 28361 + }, + { + "epoch": 0.68, + "learning_rate": 4.849226088887168e-06, + "loss": 1.2232, + "step": 28362 + }, + { + "epoch": 0.68, + "learning_rate": 4.8485584801070605e-06, + "loss": 1.159, + "step": 28363 + }, + { + "epoch": 0.68, + "learning_rate": 4.847890902579533e-06, + "loss": 1.1529, + "step": 28364 + }, + { + "epoch": 0.68, + "learning_rate": 4.8472233563086406e-06, + "loss": 1.1229, + "step": 28365 + }, + { + "epoch": 0.68, + "learning_rate": 4.846555841298435e-06, + "loss": 0.8484, + "step": 28366 + }, + { + "epoch": 0.68, + "learning_rate": 4.8458883575529605e-06, + "loss": 1.022, + "step": 28367 + }, + { + "epoch": 0.68, + "learning_rate": 4.845220905076271e-06, + "loss": 0.9814, + "step": 28368 + }, + { + "epoch": 0.68, + "learning_rate": 4.844553483872412e-06, + "loss": 1.4275, + "step": 28369 + }, + { + "epoch": 0.68, + "learning_rate": 4.8438860939454325e-06, + "loss": 0.8811, + "step": 28370 + }, + { + "epoch": 0.68, + "learning_rate": 4.8432187352993876e-06, + "loss": 1.0828, + "step": 28371 + }, + { + "epoch": 0.68, + "learning_rate": 4.842551407938318e-06, + "loss": 1.0723, + "step": 28372 + }, + { + "epoch": 0.68, + "learning_rate": 4.8418841118662765e-06, + "loss": 1.1769, + "step": 28373 + }, + { + "epoch": 0.68, + "learning_rate": 4.841216847087315e-06, + "loss": 1.1458, + "step": 28374 + }, + { + "epoch": 0.68, + "learning_rate": 4.840549613605472e-06, + "loss": 1.3048, + "step": 28375 + }, + { + "epoch": 0.68, + "learning_rate": 4.8398824114248055e-06, + "loss": 1.1217, + "step": 28376 + }, + { + "epoch": 0.68, + "learning_rate": 4.8392152405493555e-06, + "loss": 0.993, + "step": 28377 + }, + { + "epoch": 0.68, + "learning_rate": 4.838548100983171e-06, + "loss": 1.2665, + "step": 28378 + }, + { + "epoch": 0.68, + "learning_rate": 4.837880992730306e-06, + "loss": 1.0118, + "step": 28379 + }, + { + "epoch": 0.68, + "learning_rate": 4.8372139157948e-06, + "loss": 1.1926, + "step": 28380 + }, + { + "epoch": 0.68, + "learning_rate": 4.836546870180702e-06, + "loss": 0.982, + "step": 28381 + }, + { + "epoch": 0.68, + "learning_rate": 4.835879855892063e-06, + "loss": 1.1101, + "step": 28382 + }, + { + "epoch": 0.68, + "learning_rate": 4.835212872932923e-06, + "loss": 1.0432, + "step": 28383 + }, + { + "epoch": 0.68, + "learning_rate": 4.834545921307335e-06, + "loss": 1.1511, + "step": 28384 + }, + { + "epoch": 0.68, + "learning_rate": 4.8338790010193386e-06, + "loss": 1.1796, + "step": 28385 + }, + { + "epoch": 0.68, + "learning_rate": 4.8332121120729825e-06, + "loss": 1.2115, + "step": 28386 + }, + { + "epoch": 0.68, + "learning_rate": 4.832545254472317e-06, + "loss": 0.9471, + "step": 28387 + }, + { + "epoch": 0.68, + "learning_rate": 4.831878428221382e-06, + "loss": 0.9305, + "step": 28388 + }, + { + "epoch": 0.68, + "learning_rate": 4.831211633324223e-06, + "loss": 1.1539, + "step": 28389 + }, + { + "epoch": 0.68, + "learning_rate": 4.830544869784891e-06, + "loss": 1.1753, + "step": 28390 + }, + { + "epoch": 0.68, + "learning_rate": 4.829878137607424e-06, + "loss": 0.9231, + "step": 28391 + }, + { + "epoch": 0.68, + "learning_rate": 4.8292114367958745e-06, + "loss": 1.0246, + "step": 28392 + }, + { + "epoch": 0.68, + "learning_rate": 4.82854476735428e-06, + "loss": 1.2694, + "step": 28393 + }, + { + "epoch": 0.68, + "learning_rate": 4.8278781292866865e-06, + "loss": 1.1479, + "step": 28394 + }, + { + "epoch": 0.68, + "learning_rate": 4.827211522597144e-06, + "loss": 1.1812, + "step": 28395 + }, + { + "epoch": 0.68, + "learning_rate": 4.826544947289689e-06, + "loss": 1.0069, + "step": 28396 + }, + { + "epoch": 0.68, + "learning_rate": 4.825878403368368e-06, + "loss": 1.0799, + "step": 28397 + }, + { + "epoch": 0.68, + "learning_rate": 4.82521189083723e-06, + "loss": 0.842, + "step": 28398 + }, + { + "epoch": 0.68, + "learning_rate": 4.824545409700311e-06, + "loss": 1.1482, + "step": 28399 + }, + { + "epoch": 0.68, + "learning_rate": 4.823878959961656e-06, + "loss": 1.1797, + "step": 28400 + }, + { + "epoch": 0.68, + "learning_rate": 4.823212541625315e-06, + "loss": 1.1621, + "step": 28401 + }, + { + "epoch": 0.68, + "learning_rate": 4.822546154695322e-06, + "loss": 0.9364, + "step": 28402 + }, + { + "epoch": 0.68, + "learning_rate": 4.8218797991757285e-06, + "loss": 1.4415, + "step": 28403 + }, + { + "epoch": 0.68, + "learning_rate": 4.821213475070567e-06, + "loss": 0.8208, + "step": 28404 + }, + { + "epoch": 0.68, + "learning_rate": 4.820547182383886e-06, + "loss": 1.1922, + "step": 28405 + }, + { + "epoch": 0.68, + "learning_rate": 4.81988092111973e-06, + "loss": 0.9695, + "step": 28406 + }, + { + "epoch": 0.68, + "learning_rate": 4.819214691282134e-06, + "loss": 1.1209, + "step": 28407 + }, + { + "epoch": 0.68, + "learning_rate": 4.818548492875145e-06, + "loss": 0.7291, + "step": 28408 + }, + { + "epoch": 0.68, + "learning_rate": 4.817882325902806e-06, + "loss": 1.1137, + "step": 28409 + }, + { + "epoch": 0.68, + "learning_rate": 4.817216190369154e-06, + "loss": 1.0454, + "step": 28410 + }, + { + "epoch": 0.68, + "learning_rate": 4.816550086278234e-06, + "loss": 0.9879, + "step": 28411 + }, + { + "epoch": 0.68, + "learning_rate": 4.815884013634083e-06, + "loss": 0.7672, + "step": 28412 + }, + { + "epoch": 0.68, + "learning_rate": 4.815217972440745e-06, + "loss": 0.7596, + "step": 28413 + }, + { + "epoch": 0.68, + "learning_rate": 4.814551962702263e-06, + "loss": 1.0568, + "step": 28414 + }, + { + "epoch": 0.68, + "learning_rate": 4.81388598442267e-06, + "loss": 0.9364, + "step": 28415 + }, + { + "epoch": 0.68, + "learning_rate": 4.813220037606012e-06, + "loss": 0.7806, + "step": 28416 + }, + { + "epoch": 0.68, + "learning_rate": 4.812554122256331e-06, + "loss": 1.1575, + "step": 28417 + }, + { + "epoch": 0.68, + "learning_rate": 4.811888238377661e-06, + "loss": 1.075, + "step": 28418 + }, + { + "epoch": 0.68, + "learning_rate": 4.811222385974048e-06, + "loss": 0.963, + "step": 28419 + }, + { + "epoch": 0.68, + "learning_rate": 4.810556565049524e-06, + "loss": 1.1022, + "step": 28420 + }, + { + "epoch": 0.68, + "learning_rate": 4.809890775608134e-06, + "loss": 1.1339, + "step": 28421 + }, + { + "epoch": 0.68, + "learning_rate": 4.809225017653919e-06, + "loss": 1.0143, + "step": 28422 + }, + { + "epoch": 0.68, + "learning_rate": 4.808559291190911e-06, + "loss": 1.0163, + "step": 28423 + }, + { + "epoch": 0.68, + "learning_rate": 4.807893596223152e-06, + "loss": 1.2596, + "step": 28424 + }, + { + "epoch": 0.68, + "learning_rate": 4.807227932754686e-06, + "loss": 1.0928, + "step": 28425 + }, + { + "epoch": 0.68, + "learning_rate": 4.806562300789542e-06, + "loss": 0.9991, + "step": 28426 + }, + { + "epoch": 0.68, + "learning_rate": 4.805896700331768e-06, + "loss": 1.1049, + "step": 28427 + }, + { + "epoch": 0.68, + "learning_rate": 4.805231131385393e-06, + "loss": 1.0308, + "step": 28428 + }, + { + "epoch": 0.68, + "learning_rate": 4.8045655939544585e-06, + "loss": 1.2417, + "step": 28429 + }, + { + "epoch": 0.68, + "learning_rate": 4.803900088043007e-06, + "loss": 1.0248, + "step": 28430 + }, + { + "epoch": 0.68, + "learning_rate": 4.803234613655066e-06, + "loss": 1.162, + "step": 28431 + }, + { + "epoch": 0.68, + "learning_rate": 4.80256917079468e-06, + "loss": 0.987, + "step": 28432 + }, + { + "epoch": 0.68, + "learning_rate": 4.801903759465887e-06, + "loss": 1.0985, + "step": 28433 + }, + { + "epoch": 0.68, + "learning_rate": 4.801238379672717e-06, + "loss": 1.2014, + "step": 28434 + }, + { + "epoch": 0.68, + "learning_rate": 4.800573031419216e-06, + "loss": 1.1545, + "step": 28435 + }, + { + "epoch": 0.68, + "learning_rate": 4.799907714709411e-06, + "loss": 1.1795, + "step": 28436 + }, + { + "epoch": 0.68, + "learning_rate": 4.799242429547343e-06, + "loss": 0.9646, + "step": 28437 + }, + { + "epoch": 0.68, + "learning_rate": 4.798577175937051e-06, + "loss": 1.0136, + "step": 28438 + }, + { + "epoch": 0.68, + "learning_rate": 4.797911953882565e-06, + "loss": 1.0376, + "step": 28439 + }, + { + "epoch": 0.68, + "learning_rate": 4.797246763387922e-06, + "loss": 0.9849, + "step": 28440 + }, + { + "epoch": 0.68, + "learning_rate": 4.796581604457165e-06, + "loss": 0.9807, + "step": 28441 + }, + { + "epoch": 0.68, + "learning_rate": 4.795916477094319e-06, + "loss": 0.8571, + "step": 28442 + }, + { + "epoch": 0.68, + "learning_rate": 4.795251381303423e-06, + "loss": 0.868, + "step": 28443 + }, + { + "epoch": 0.68, + "learning_rate": 4.7945863170885175e-06, + "loss": 1.1254, + "step": 28444 + }, + { + "epoch": 0.68, + "learning_rate": 4.793921284453627e-06, + "loss": 1.0684, + "step": 28445 + }, + { + "epoch": 0.68, + "learning_rate": 4.7932562834027975e-06, + "loss": 1.036, + "step": 28446 + }, + { + "epoch": 0.68, + "learning_rate": 4.792591313940052e-06, + "loss": 1.0632, + "step": 28447 + }, + { + "epoch": 0.68, + "learning_rate": 4.791926376069431e-06, + "loss": 1.2084, + "step": 28448 + }, + { + "epoch": 0.68, + "learning_rate": 4.791261469794971e-06, + "loss": 1.0097, + "step": 28449 + }, + { + "epoch": 0.68, + "learning_rate": 4.790596595120699e-06, + "loss": 1.1443, + "step": 28450 + }, + { + "epoch": 0.68, + "learning_rate": 4.789931752050651e-06, + "loss": 1.0642, + "step": 28451 + }, + { + "epoch": 0.68, + "learning_rate": 4.789266940588867e-06, + "loss": 1.0119, + "step": 28452 + }, + { + "epoch": 0.68, + "learning_rate": 4.7886021607393695e-06, + "loss": 1.1508, + "step": 28453 + }, + { + "epoch": 0.68, + "learning_rate": 4.787937412506201e-06, + "loss": 0.877, + "step": 28454 + }, + { + "epoch": 0.68, + "learning_rate": 4.787272695893388e-06, + "loss": 1.0942, + "step": 28455 + }, + { + "epoch": 0.68, + "learning_rate": 4.786608010904964e-06, + "loss": 0.7019, + "step": 28456 + }, + { + "epoch": 0.68, + "learning_rate": 4.785943357544967e-06, + "loss": 0.7975, + "step": 28457 + }, + { + "epoch": 0.68, + "learning_rate": 4.785278735817423e-06, + "loss": 1.1657, + "step": 28458 + }, + { + "epoch": 0.68, + "learning_rate": 4.784614145726364e-06, + "loss": 1.1004, + "step": 28459 + }, + { + "epoch": 0.68, + "learning_rate": 4.783949587275829e-06, + "loss": 1.2167, + "step": 28460 + }, + { + "epoch": 0.68, + "learning_rate": 4.783285060469841e-06, + "loss": 1.1229, + "step": 28461 + }, + { + "epoch": 0.68, + "learning_rate": 4.782620565312439e-06, + "loss": 1.0163, + "step": 28462 + }, + { + "epoch": 0.68, + "learning_rate": 4.781956101807648e-06, + "loss": 1.0418, + "step": 28463 + }, + { + "epoch": 0.68, + "learning_rate": 4.781291669959502e-06, + "loss": 1.0202, + "step": 28464 + }, + { + "epoch": 0.68, + "learning_rate": 4.780627269772034e-06, + "loss": 1.0751, + "step": 28465 + }, + { + "epoch": 0.68, + "learning_rate": 4.77996290124927e-06, + "loss": 1.2594, + "step": 28466 + }, + { + "epoch": 0.68, + "learning_rate": 4.779298564395244e-06, + "loss": 0.8682, + "step": 28467 + }, + { + "epoch": 0.68, + "learning_rate": 4.778634259213988e-06, + "loss": 1.2588, + "step": 28468 + }, + { + "epoch": 0.68, + "learning_rate": 4.777969985709526e-06, + "loss": 1.1247, + "step": 28469 + }, + { + "epoch": 0.68, + "learning_rate": 4.7773057438858975e-06, + "loss": 1.0425, + "step": 28470 + }, + { + "epoch": 0.68, + "learning_rate": 4.77664153374712e-06, + "loss": 0.9146, + "step": 28471 + }, + { + "epoch": 0.68, + "learning_rate": 4.775977355297232e-06, + "loss": 0.9462, + "step": 28472 + }, + { + "epoch": 0.68, + "learning_rate": 4.775313208540264e-06, + "loss": 0.9857, + "step": 28473 + }, + { + "epoch": 0.68, + "learning_rate": 4.774649093480238e-06, + "loss": 1.1664, + "step": 28474 + }, + { + "epoch": 0.68, + "learning_rate": 4.773985010121186e-06, + "loss": 1.1191, + "step": 28475 + }, + { + "epoch": 0.68, + "learning_rate": 4.773320958467143e-06, + "loss": 0.8845, + "step": 28476 + }, + { + "epoch": 0.68, + "learning_rate": 4.772656938522127e-06, + "loss": 0.7354, + "step": 28477 + }, + { + "epoch": 0.68, + "learning_rate": 4.771992950290176e-06, + "loss": 0.9902, + "step": 28478 + }, + { + "epoch": 0.68, + "learning_rate": 4.771328993775311e-06, + "loss": 1.1788, + "step": 28479 + }, + { + "epoch": 0.68, + "learning_rate": 4.770665068981563e-06, + "loss": 1.0188, + "step": 28480 + }, + { + "epoch": 0.68, + "learning_rate": 4.770001175912962e-06, + "loss": 1.0778, + "step": 28481 + }, + { + "epoch": 0.68, + "learning_rate": 4.769337314573533e-06, + "loss": 0.986, + "step": 28482 + }, + { + "epoch": 0.68, + "learning_rate": 4.768673484967302e-06, + "loss": 1.0381, + "step": 28483 + }, + { + "epoch": 0.69, + "learning_rate": 4.768009687098302e-06, + "loss": 1.194, + "step": 28484 + }, + { + "epoch": 0.69, + "learning_rate": 4.7673459209705555e-06, + "loss": 1.0542, + "step": 28485 + }, + { + "epoch": 0.69, + "learning_rate": 4.766682186588087e-06, + "loss": 1.2024, + "step": 28486 + }, + { + "epoch": 0.69, + "learning_rate": 4.766018483954933e-06, + "loss": 0.8881, + "step": 28487 + }, + { + "epoch": 0.69, + "learning_rate": 4.765354813075109e-06, + "loss": 0.9803, + "step": 28488 + }, + { + "epoch": 0.69, + "learning_rate": 4.76469117395265e-06, + "loss": 1.1223, + "step": 28489 + }, + { + "epoch": 0.69, + "learning_rate": 4.7640275665915745e-06, + "loss": 1.0353, + "step": 28490 + }, + { + "epoch": 0.69, + "learning_rate": 4.7633639909959126e-06, + "loss": 0.9899, + "step": 28491 + }, + { + "epoch": 0.69, + "learning_rate": 4.762700447169693e-06, + "loss": 1.0726, + "step": 28492 + }, + { + "epoch": 0.69, + "learning_rate": 4.762036935116934e-06, + "loss": 0.8398, + "step": 28493 + }, + { + "epoch": 0.69, + "learning_rate": 4.761373454841665e-06, + "loss": 1.0494, + "step": 28494 + }, + { + "epoch": 0.69, + "learning_rate": 4.760710006347914e-06, + "loss": 1.2248, + "step": 28495 + }, + { + "epoch": 0.69, + "learning_rate": 4.760046589639701e-06, + "loss": 1.1746, + "step": 28496 + }, + { + "epoch": 0.69, + "learning_rate": 4.759383204721056e-06, + "loss": 1.2485, + "step": 28497 + }, + { + "epoch": 0.69, + "learning_rate": 4.758719851595995e-06, + "loss": 1.113, + "step": 28498 + }, + { + "epoch": 0.69, + "learning_rate": 4.75805653026855e-06, + "loss": 0.8556, + "step": 28499 + }, + { + "epoch": 0.69, + "learning_rate": 4.757393240742742e-06, + "loss": 1.1047, + "step": 28500 + }, + { + "epoch": 0.69, + "learning_rate": 4.756729983022595e-06, + "loss": 1.1414, + "step": 28501 + }, + { + "epoch": 0.69, + "learning_rate": 4.75606675711214e-06, + "loss": 1.2863, + "step": 28502 + }, + { + "epoch": 0.69, + "learning_rate": 4.7554035630153895e-06, + "loss": 1.2221, + "step": 28503 + }, + { + "epoch": 0.69, + "learning_rate": 4.754740400736372e-06, + "loss": 1.024, + "step": 28504 + }, + { + "epoch": 0.69, + "learning_rate": 4.754077270279115e-06, + "loss": 1.1533, + "step": 28505 + }, + { + "epoch": 0.69, + "learning_rate": 4.753414171647634e-06, + "loss": 1.1877, + "step": 28506 + }, + { + "epoch": 0.69, + "learning_rate": 4.752751104845957e-06, + "loss": 1.0139, + "step": 28507 + }, + { + "epoch": 0.69, + "learning_rate": 4.7520880698781026e-06, + "loss": 1.0677, + "step": 28508 + }, + { + "epoch": 0.69, + "learning_rate": 4.7514250667480944e-06, + "loss": 1.3164, + "step": 28509 + }, + { + "epoch": 0.69, + "learning_rate": 4.75076209545996e-06, + "loss": 1.2395, + "step": 28510 + }, + { + "epoch": 0.69, + "learning_rate": 4.750099156017714e-06, + "loss": 0.94, + "step": 28511 + }, + { + "epoch": 0.69, + "learning_rate": 4.749436248425381e-06, + "loss": 1.1187, + "step": 28512 + }, + { + "epoch": 0.69, + "learning_rate": 4.7487733726869875e-06, + "loss": 1.1518, + "step": 28513 + }, + { + "epoch": 0.69, + "learning_rate": 4.748110528806547e-06, + "loss": 1.0546, + "step": 28514 + }, + { + "epoch": 0.69, + "learning_rate": 4.747447716788088e-06, + "loss": 0.7852, + "step": 28515 + }, + { + "epoch": 0.69, + "learning_rate": 4.746784936635624e-06, + "loss": 1.0766, + "step": 28516 + }, + { + "epoch": 0.69, + "learning_rate": 4.746122188353182e-06, + "loss": 1.0287, + "step": 28517 + }, + { + "epoch": 0.69, + "learning_rate": 4.745459471944784e-06, + "loss": 1.077, + "step": 28518 + }, + { + "epoch": 0.69, + "learning_rate": 4.7447967874144415e-06, + "loss": 0.9799, + "step": 28519 + }, + { + "epoch": 0.69, + "learning_rate": 4.744134134766183e-06, + "loss": 1.0416, + "step": 28520 + }, + { + "epoch": 0.69, + "learning_rate": 4.7434715140040295e-06, + "loss": 0.8898, + "step": 28521 + }, + { + "epoch": 0.69, + "learning_rate": 4.742808925131994e-06, + "loss": 0.9398, + "step": 28522 + }, + { + "epoch": 0.69, + "learning_rate": 4.742146368154101e-06, + "loss": 0.8847, + "step": 28523 + }, + { + "epoch": 0.69, + "learning_rate": 4.741483843074372e-06, + "loss": 0.8858, + "step": 28524 + }, + { + "epoch": 0.69, + "learning_rate": 4.7408213498968205e-06, + "loss": 1.0312, + "step": 28525 + }, + { + "epoch": 0.69, + "learning_rate": 4.740158888625473e-06, + "loss": 0.9701, + "step": 28526 + }, + { + "epoch": 0.69, + "learning_rate": 4.7394964592643396e-06, + "loss": 1.0096, + "step": 28527 + }, + { + "epoch": 0.69, + "learning_rate": 4.7388340618174444e-06, + "loss": 1.04, + "step": 28528 + }, + { + "epoch": 0.69, + "learning_rate": 4.738171696288809e-06, + "loss": 1.0533, + "step": 28529 + }, + { + "epoch": 0.69, + "learning_rate": 4.737509362682444e-06, + "loss": 1.0774, + "step": 28530 + }, + { + "epoch": 0.69, + "learning_rate": 4.736847061002373e-06, + "loss": 1.1502, + "step": 28531 + }, + { + "epoch": 0.69, + "learning_rate": 4.736184791252616e-06, + "loss": 1.0606, + "step": 28532 + }, + { + "epoch": 0.69, + "learning_rate": 4.735522553437184e-06, + "loss": 1.1092, + "step": 28533 + }, + { + "epoch": 0.69, + "learning_rate": 4.734860347560103e-06, + "loss": 1.1124, + "step": 28534 + }, + { + "epoch": 0.69, + "learning_rate": 4.734198173625381e-06, + "loss": 1.0646, + "step": 28535 + }, + { + "epoch": 0.69, + "learning_rate": 4.73353603163704e-06, + "loss": 1.0683, + "step": 28536 + }, + { + "epoch": 0.69, + "learning_rate": 4.732873921599102e-06, + "loss": 0.9489, + "step": 28537 + }, + { + "epoch": 0.69, + "learning_rate": 4.732211843515574e-06, + "loss": 1.0772, + "step": 28538 + }, + { + "epoch": 0.69, + "learning_rate": 4.731549797390479e-06, + "loss": 0.9293, + "step": 28539 + }, + { + "epoch": 0.69, + "learning_rate": 4.730887783227834e-06, + "loss": 1.0022, + "step": 28540 + }, + { + "epoch": 0.69, + "learning_rate": 4.730225801031652e-06, + "loss": 1.0684, + "step": 28541 + }, + { + "epoch": 0.69, + "learning_rate": 4.729563850805953e-06, + "loss": 1.156, + "step": 28542 + }, + { + "epoch": 0.69, + "learning_rate": 4.728901932554747e-06, + "loss": 1.1393, + "step": 28543 + }, + { + "epoch": 0.69, + "learning_rate": 4.728240046282053e-06, + "loss": 0.9442, + "step": 28544 + }, + { + "epoch": 0.69, + "learning_rate": 4.7275781919918904e-06, + "loss": 0.8723, + "step": 28545 + }, + { + "epoch": 0.69, + "learning_rate": 4.726916369688267e-06, + "loss": 1.1461, + "step": 28546 + }, + { + "epoch": 0.69, + "learning_rate": 4.726254579375201e-06, + "loss": 1.0876, + "step": 28547 + }, + { + "epoch": 0.69, + "learning_rate": 4.7255928210567125e-06, + "loss": 1.0271, + "step": 28548 + }, + { + "epoch": 0.69, + "learning_rate": 4.724931094736809e-06, + "loss": 0.8161, + "step": 28549 + }, + { + "epoch": 0.69, + "learning_rate": 4.72426940041951e-06, + "loss": 0.9827, + "step": 28550 + }, + { + "epoch": 0.69, + "learning_rate": 4.723607738108824e-06, + "loss": 1.0938, + "step": 28551 + }, + { + "epoch": 0.69, + "learning_rate": 4.722946107808769e-06, + "loss": 1.0901, + "step": 28552 + }, + { + "epoch": 0.69, + "learning_rate": 4.722284509523362e-06, + "loss": 1.1167, + "step": 28553 + }, + { + "epoch": 0.69, + "learning_rate": 4.721622943256612e-06, + "loss": 0.7642, + "step": 28554 + }, + { + "epoch": 0.69, + "learning_rate": 4.720961409012532e-06, + "loss": 0.8825, + "step": 28555 + }, + { + "epoch": 0.69, + "learning_rate": 4.720299906795142e-06, + "loss": 0.85, + "step": 28556 + }, + { + "epoch": 0.69, + "learning_rate": 4.7196384366084455e-06, + "loss": 1.0814, + "step": 28557 + }, + { + "epoch": 0.69, + "learning_rate": 4.718976998456466e-06, + "loss": 1.3192, + "step": 28558 + }, + { + "epoch": 0.69, + "learning_rate": 4.718315592343206e-06, + "loss": 1.2107, + "step": 28559 + }, + { + "epoch": 0.69, + "learning_rate": 4.717654218272684e-06, + "loss": 0.8933, + "step": 28560 + }, + { + "epoch": 0.69, + "learning_rate": 4.716992876248914e-06, + "loss": 1.011, + "step": 28561 + }, + { + "epoch": 0.69, + "learning_rate": 4.716331566275903e-06, + "loss": 0.9672, + "step": 28562 + }, + { + "epoch": 0.69, + "learning_rate": 4.715670288357663e-06, + "loss": 1.0689, + "step": 28563 + }, + { + "epoch": 0.69, + "learning_rate": 4.715009042498216e-06, + "loss": 0.9534, + "step": 28564 + }, + { + "epoch": 0.69, + "learning_rate": 4.714347828701559e-06, + "loss": 1.0478, + "step": 28565 + }, + { + "epoch": 0.69, + "learning_rate": 4.713686646971712e-06, + "loss": 0.9122, + "step": 28566 + }, + { + "epoch": 0.69, + "learning_rate": 4.713025497312689e-06, + "loss": 0.9785, + "step": 28567 + }, + { + "epoch": 0.69, + "learning_rate": 4.712364379728491e-06, + "loss": 1.2798, + "step": 28568 + }, + { + "epoch": 0.69, + "learning_rate": 4.71170329422314e-06, + "loss": 1.2402, + "step": 28569 + }, + { + "epoch": 0.69, + "learning_rate": 4.711042240800637e-06, + "loss": 0.8368, + "step": 28570 + }, + { + "epoch": 0.69, + "learning_rate": 4.710381219464997e-06, + "loss": 1.1362, + "step": 28571 + }, + { + "epoch": 0.69, + "learning_rate": 4.709720230220233e-06, + "loss": 1.0606, + "step": 28572 + }, + { + "epoch": 0.69, + "learning_rate": 4.709059273070349e-06, + "loss": 1.0917, + "step": 28573 + }, + { + "epoch": 0.69, + "learning_rate": 4.708398348019357e-06, + "loss": 0.964, + "step": 28574 + }, + { + "epoch": 0.69, + "learning_rate": 4.707737455071273e-06, + "loss": 0.8008, + "step": 28575 + }, + { + "epoch": 0.69, + "learning_rate": 4.7070765942300955e-06, + "loss": 1.0791, + "step": 28576 + }, + { + "epoch": 0.69, + "learning_rate": 4.706415765499843e-06, + "loss": 1.1926, + "step": 28577 + }, + { + "epoch": 0.69, + "learning_rate": 4.705754968884518e-06, + "loss": 1.1571, + "step": 28578 + }, + { + "epoch": 0.69, + "learning_rate": 4.705094204388133e-06, + "loss": 1.1232, + "step": 28579 + }, + { + "epoch": 0.69, + "learning_rate": 4.704433472014699e-06, + "loss": 0.9424, + "step": 28580 + }, + { + "epoch": 0.69, + "learning_rate": 4.703772771768218e-06, + "loss": 1.0678, + "step": 28581 + }, + { + "epoch": 0.69, + "learning_rate": 4.703112103652702e-06, + "loss": 1.1496, + "step": 28582 + }, + { + "epoch": 0.69, + "learning_rate": 4.702451467672162e-06, + "loss": 1.1407, + "step": 28583 + }, + { + "epoch": 0.69, + "learning_rate": 4.7017908638306e-06, + "loss": 1.1438, + "step": 28584 + }, + { + "epoch": 0.69, + "learning_rate": 4.701130292132029e-06, + "loss": 1.0416, + "step": 28585 + }, + { + "epoch": 0.69, + "learning_rate": 4.700469752580451e-06, + "loss": 1.0661, + "step": 28586 + }, + { + "epoch": 0.69, + "learning_rate": 4.6998092451798775e-06, + "loss": 1.1857, + "step": 28587 + }, + { + "epoch": 0.69, + "learning_rate": 4.699148769934318e-06, + "loss": 0.9524, + "step": 28588 + }, + { + "epoch": 0.69, + "learning_rate": 4.698488326847771e-06, + "loss": 0.9592, + "step": 28589 + }, + { + "epoch": 0.69, + "learning_rate": 4.697827915924249e-06, + "loss": 1.0457, + "step": 28590 + }, + { + "epoch": 0.69, + "learning_rate": 4.6971675371677625e-06, + "loss": 1.087, + "step": 28591 + }, + { + "epoch": 0.69, + "learning_rate": 4.696507190582309e-06, + "loss": 1.0945, + "step": 28592 + }, + { + "epoch": 0.69, + "learning_rate": 4.695846876171902e-06, + "loss": 1.0446, + "step": 28593 + }, + { + "epoch": 0.69, + "learning_rate": 4.6951865939405415e-06, + "loss": 0.9203, + "step": 28594 + }, + { + "epoch": 0.69, + "learning_rate": 4.694526343892236e-06, + "loss": 1.0743, + "step": 28595 + }, + { + "epoch": 0.69, + "learning_rate": 4.693866126030995e-06, + "loss": 0.9824, + "step": 28596 + }, + { + "epoch": 0.69, + "learning_rate": 4.693205940360817e-06, + "loss": 0.8877, + "step": 28597 + }, + { + "epoch": 0.69, + "learning_rate": 4.6925457868857096e-06, + "loss": 0.9487, + "step": 28598 + }, + { + "epoch": 0.69, + "learning_rate": 4.6918856656096815e-06, + "loss": 1.2354, + "step": 28599 + }, + { + "epoch": 0.69, + "learning_rate": 4.691225576536732e-06, + "loss": 1.0633, + "step": 28600 + }, + { + "epoch": 0.69, + "learning_rate": 4.690565519670871e-06, + "loss": 0.9337, + "step": 28601 + }, + { + "epoch": 0.69, + "learning_rate": 4.689905495016097e-06, + "loss": 1.2605, + "step": 28602 + }, + { + "epoch": 0.69, + "learning_rate": 4.689245502576417e-06, + "loss": 0.9717, + "step": 28603 + }, + { + "epoch": 0.69, + "learning_rate": 4.68858554235584e-06, + "loss": 1.2216, + "step": 28604 + }, + { + "epoch": 0.69, + "learning_rate": 4.687925614358361e-06, + "loss": 0.9454, + "step": 28605 + }, + { + "epoch": 0.69, + "learning_rate": 4.687265718587987e-06, + "loss": 0.997, + "step": 28606 + }, + { + "epoch": 0.69, + "learning_rate": 4.686605855048727e-06, + "loss": 0.8843, + "step": 28607 + }, + { + "epoch": 0.69, + "learning_rate": 4.685946023744575e-06, + "loss": 1.1623, + "step": 28608 + }, + { + "epoch": 0.69, + "learning_rate": 4.685286224679543e-06, + "loss": 1.0855, + "step": 28609 + }, + { + "epoch": 0.69, + "learning_rate": 4.684626457857625e-06, + "loss": 0.9981, + "step": 28610 + }, + { + "epoch": 0.69, + "learning_rate": 4.683966723282829e-06, + "loss": 1.0814, + "step": 28611 + }, + { + "epoch": 0.69, + "learning_rate": 4.683307020959159e-06, + "loss": 0.9891, + "step": 28612 + }, + { + "epoch": 0.69, + "learning_rate": 4.682647350890611e-06, + "loss": 1.1114, + "step": 28613 + }, + { + "epoch": 0.69, + "learning_rate": 4.6819877130811915e-06, + "loss": 1.2341, + "step": 28614 + }, + { + "epoch": 0.69, + "learning_rate": 4.681328107534906e-06, + "loss": 1.0793, + "step": 28615 + }, + { + "epoch": 0.69, + "learning_rate": 4.680668534255747e-06, + "loss": 0.8974, + "step": 28616 + }, + { + "epoch": 0.69, + "learning_rate": 4.680008993247721e-06, + "loss": 1.2071, + "step": 28617 + }, + { + "epoch": 0.69, + "learning_rate": 4.679349484514833e-06, + "loss": 0.8567, + "step": 28618 + }, + { + "epoch": 0.69, + "learning_rate": 4.678690008061076e-06, + "loss": 1.2569, + "step": 28619 + }, + { + "epoch": 0.69, + "learning_rate": 4.67803056389046e-06, + "loss": 1.0054, + "step": 28620 + }, + { + "epoch": 0.69, + "learning_rate": 4.677371152006977e-06, + "loss": 0.9927, + "step": 28621 + }, + { + "epoch": 0.69, + "learning_rate": 4.676711772414631e-06, + "loss": 1.0198, + "step": 28622 + }, + { + "epoch": 0.69, + "learning_rate": 4.676052425117425e-06, + "loss": 1.2456, + "step": 28623 + }, + { + "epoch": 0.69, + "learning_rate": 4.675393110119354e-06, + "loss": 0.9836, + "step": 28624 + }, + { + "epoch": 0.69, + "learning_rate": 4.6747338274244206e-06, + "loss": 0.9953, + "step": 28625 + }, + { + "epoch": 0.69, + "learning_rate": 4.674074577036627e-06, + "loss": 1.1923, + "step": 28626 + }, + { + "epoch": 0.69, + "learning_rate": 4.673415358959969e-06, + "loss": 0.8821, + "step": 28627 + }, + { + "epoch": 0.69, + "learning_rate": 4.6727561731984485e-06, + "loss": 1.2679, + "step": 28628 + }, + { + "epoch": 0.69, + "learning_rate": 4.67209701975606e-06, + "loss": 1.2102, + "step": 28629 + }, + { + "epoch": 0.69, + "learning_rate": 4.671437898636806e-06, + "loss": 1.0414, + "step": 28630 + }, + { + "epoch": 0.69, + "learning_rate": 4.670778809844689e-06, + "loss": 1.1737, + "step": 28631 + }, + { + "epoch": 0.69, + "learning_rate": 4.6701197533836995e-06, + "loss": 1.0282, + "step": 28632 + }, + { + "epoch": 0.69, + "learning_rate": 4.6694607292578395e-06, + "loss": 1.1112, + "step": 28633 + }, + { + "epoch": 0.69, + "learning_rate": 4.668801737471111e-06, + "loss": 1.0215, + "step": 28634 + }, + { + "epoch": 0.69, + "learning_rate": 4.668142778027506e-06, + "loss": 0.816, + "step": 28635 + }, + { + "epoch": 0.69, + "learning_rate": 4.6674838509310276e-06, + "loss": 0.7755, + "step": 28636 + }, + { + "epoch": 0.69, + "learning_rate": 4.666824956185666e-06, + "loss": 0.8436, + "step": 28637 + }, + { + "epoch": 0.69, + "learning_rate": 4.666166093795423e-06, + "loss": 0.8727, + "step": 28638 + }, + { + "epoch": 0.69, + "learning_rate": 4.665507263764299e-06, + "loss": 0.9482, + "step": 28639 + }, + { + "epoch": 0.69, + "learning_rate": 4.664848466096286e-06, + "loss": 0.9703, + "step": 28640 + }, + { + "epoch": 0.69, + "learning_rate": 4.6641897007953805e-06, + "loss": 1.1023, + "step": 28641 + }, + { + "epoch": 0.69, + "learning_rate": 4.663530967865586e-06, + "loss": 0.9948, + "step": 28642 + }, + { + "epoch": 0.69, + "learning_rate": 4.662872267310889e-06, + "loss": 1.3054, + "step": 28643 + }, + { + "epoch": 0.69, + "learning_rate": 4.662213599135296e-06, + "loss": 1.1313, + "step": 28644 + }, + { + "epoch": 0.69, + "learning_rate": 4.661554963342792e-06, + "loss": 1.0924, + "step": 28645 + }, + { + "epoch": 0.69, + "learning_rate": 4.660896359937381e-06, + "loss": 1.1984, + "step": 28646 + }, + { + "epoch": 0.69, + "learning_rate": 4.660237788923057e-06, + "loss": 0.8654, + "step": 28647 + }, + { + "epoch": 0.69, + "learning_rate": 4.6595792503038105e-06, + "loss": 1.0168, + "step": 28648 + }, + { + "epoch": 0.69, + "learning_rate": 4.658920744083642e-06, + "loss": 1.0536, + "step": 28649 + }, + { + "epoch": 0.69, + "learning_rate": 4.658262270266548e-06, + "loss": 0.8869, + "step": 28650 + }, + { + "epoch": 0.69, + "learning_rate": 4.657603828856517e-06, + "loss": 1.2304, + "step": 28651 + }, + { + "epoch": 0.69, + "learning_rate": 4.656945419857551e-06, + "loss": 0.9201, + "step": 28652 + }, + { + "epoch": 0.69, + "learning_rate": 4.656287043273636e-06, + "loss": 1.2335, + "step": 28653 + }, + { + "epoch": 0.69, + "learning_rate": 4.655628699108771e-06, + "loss": 1.0425, + "step": 28654 + }, + { + "epoch": 0.69, + "learning_rate": 4.654970387366954e-06, + "loss": 0.9763, + "step": 28655 + }, + { + "epoch": 0.69, + "learning_rate": 4.6543121080521695e-06, + "loss": 0.9303, + "step": 28656 + }, + { + "epoch": 0.69, + "learning_rate": 4.653653861168416e-06, + "loss": 1.1478, + "step": 28657 + }, + { + "epoch": 0.69, + "learning_rate": 4.6529956467196915e-06, + "loss": 1.3149, + "step": 28658 + }, + { + "epoch": 0.69, + "learning_rate": 4.65233746470998e-06, + "loss": 1.1766, + "step": 28659 + }, + { + "epoch": 0.69, + "learning_rate": 4.65167931514328e-06, + "loss": 0.9903, + "step": 28660 + }, + { + "epoch": 0.69, + "learning_rate": 4.651021198023587e-06, + "loss": 0.9893, + "step": 28661 + }, + { + "epoch": 0.69, + "learning_rate": 4.650363113354888e-06, + "loss": 0.954, + "step": 28662 + }, + { + "epoch": 0.69, + "learning_rate": 4.649705061141179e-06, + "loss": 1.2874, + "step": 28663 + }, + { + "epoch": 0.69, + "learning_rate": 4.649047041386448e-06, + "loss": 0.886, + "step": 28664 + }, + { + "epoch": 0.69, + "learning_rate": 4.648389054094692e-06, + "loss": 0.745, + "step": 28665 + }, + { + "epoch": 0.69, + "learning_rate": 4.647731099269898e-06, + "loss": 1.052, + "step": 28666 + }, + { + "epoch": 0.69, + "learning_rate": 4.647073176916063e-06, + "loss": 1.367, + "step": 28667 + }, + { + "epoch": 0.69, + "learning_rate": 4.646415287037177e-06, + "loss": 1.0165, + "step": 28668 + }, + { + "epoch": 0.69, + "learning_rate": 4.645757429637229e-06, + "loss": 1.1299, + "step": 28669 + }, + { + "epoch": 0.69, + "learning_rate": 4.645099604720209e-06, + "loss": 0.9814, + "step": 28670 + }, + { + "epoch": 0.69, + "learning_rate": 4.644441812290116e-06, + "loss": 0.9117, + "step": 28671 + }, + { + "epoch": 0.69, + "learning_rate": 4.64378405235093e-06, + "loss": 1.0605, + "step": 28672 + }, + { + "epoch": 0.69, + "learning_rate": 4.6431263249066495e-06, + "loss": 1.1531, + "step": 28673 + }, + { + "epoch": 0.69, + "learning_rate": 4.642468629961258e-06, + "loss": 0.7765, + "step": 28674 + }, + { + "epoch": 0.69, + "learning_rate": 4.641810967518751e-06, + "loss": 1.113, + "step": 28675 + }, + { + "epoch": 0.69, + "learning_rate": 4.641153337583119e-06, + "loss": 1.2337, + "step": 28676 + }, + { + "epoch": 0.69, + "learning_rate": 4.640495740158344e-06, + "loss": 0.9312, + "step": 28677 + }, + { + "epoch": 0.69, + "learning_rate": 4.639838175248422e-06, + "loss": 1.0468, + "step": 28678 + }, + { + "epoch": 0.69, + "learning_rate": 4.639180642857345e-06, + "loss": 0.9008, + "step": 28679 + }, + { + "epoch": 0.69, + "learning_rate": 4.638523142989094e-06, + "loss": 0.9607, + "step": 28680 + }, + { + "epoch": 0.69, + "learning_rate": 4.637865675647666e-06, + "loss": 1.2707, + "step": 28681 + }, + { + "epoch": 0.69, + "learning_rate": 4.637208240837042e-06, + "loss": 1.191, + "step": 28682 + }, + { + "epoch": 0.69, + "learning_rate": 4.636550838561215e-06, + "loss": 0.9054, + "step": 28683 + }, + { + "epoch": 0.69, + "learning_rate": 4.6358934688241745e-06, + "loss": 1.1203, + "step": 28684 + }, + { + "epoch": 0.69, + "learning_rate": 4.635236131629904e-06, + "loss": 0.9911, + "step": 28685 + }, + { + "epoch": 0.69, + "learning_rate": 4.634578826982394e-06, + "loss": 1.1418, + "step": 28686 + }, + { + "epoch": 0.69, + "learning_rate": 4.6339215548856355e-06, + "loss": 0.9531, + "step": 28687 + }, + { + "epoch": 0.69, + "learning_rate": 4.633264315343609e-06, + "loss": 0.93, + "step": 28688 + }, + { + "epoch": 0.69, + "learning_rate": 4.6326071083603095e-06, + "loss": 1.2515, + "step": 28689 + }, + { + "epoch": 0.69, + "learning_rate": 4.6319499339397155e-06, + "loss": 0.9172, + "step": 28690 + }, + { + "epoch": 0.69, + "learning_rate": 4.63129279208582e-06, + "loss": 1.1276, + "step": 28691 + }, + { + "epoch": 0.69, + "learning_rate": 4.630635682802612e-06, + "loss": 1.1732, + "step": 28692 + }, + { + "epoch": 0.69, + "learning_rate": 4.62997860609407e-06, + "loss": 1.1952, + "step": 28693 + }, + { + "epoch": 0.69, + "learning_rate": 4.629321561964185e-06, + "loss": 1.0141, + "step": 28694 + }, + { + "epoch": 0.69, + "learning_rate": 4.628664550416946e-06, + "loss": 1.0842, + "step": 28695 + }, + { + "epoch": 0.69, + "learning_rate": 4.628007571456333e-06, + "loss": 1.2081, + "step": 28696 + }, + { + "epoch": 0.69, + "learning_rate": 4.6273506250863346e-06, + "loss": 1.0961, + "step": 28697 + }, + { + "epoch": 0.69, + "learning_rate": 4.626693711310939e-06, + "loss": 1.0829, + "step": 28698 + }, + { + "epoch": 0.69, + "learning_rate": 4.626036830134125e-06, + "loss": 0.9075, + "step": 28699 + }, + { + "epoch": 0.69, + "learning_rate": 4.625379981559885e-06, + "loss": 1.1381, + "step": 28700 + }, + { + "epoch": 0.69, + "learning_rate": 4.624723165592197e-06, + "loss": 1.0369, + "step": 28701 + }, + { + "epoch": 0.69, + "learning_rate": 4.6240663822350505e-06, + "loss": 1.1778, + "step": 28702 + }, + { + "epoch": 0.69, + "learning_rate": 4.6234096314924315e-06, + "loss": 0.9169, + "step": 28703 + }, + { + "epoch": 0.69, + "learning_rate": 4.622752913368318e-06, + "loss": 0.8771, + "step": 28704 + }, + { + "epoch": 0.69, + "learning_rate": 4.622096227866698e-06, + "loss": 0.9904, + "step": 28705 + }, + { + "epoch": 0.69, + "learning_rate": 4.621439574991559e-06, + "loss": 1.2176, + "step": 28706 + }, + { + "epoch": 0.69, + "learning_rate": 4.620782954746876e-06, + "loss": 0.7484, + "step": 28707 + }, + { + "epoch": 0.69, + "learning_rate": 4.620126367136642e-06, + "loss": 1.1406, + "step": 28708 + }, + { + "epoch": 0.69, + "learning_rate": 4.619469812164833e-06, + "loss": 1.0223, + "step": 28709 + }, + { + "epoch": 0.69, + "learning_rate": 4.6188132898354345e-06, + "loss": 0.7468, + "step": 28710 + }, + { + "epoch": 0.69, + "learning_rate": 4.618156800152434e-06, + "loss": 1.0422, + "step": 28711 + }, + { + "epoch": 0.69, + "learning_rate": 4.617500343119806e-06, + "loss": 1.146, + "step": 28712 + }, + { + "epoch": 0.69, + "learning_rate": 4.616843918741538e-06, + "loss": 1.2529, + "step": 28713 + }, + { + "epoch": 0.69, + "learning_rate": 4.616187527021615e-06, + "loss": 1.0309, + "step": 28714 + }, + { + "epoch": 0.69, + "learning_rate": 4.615531167964013e-06, + "loss": 1.0844, + "step": 28715 + }, + { + "epoch": 0.69, + "learning_rate": 4.61487484157272e-06, + "loss": 0.9021, + "step": 28716 + }, + { + "epoch": 0.69, + "learning_rate": 4.61421854785171e-06, + "loss": 0.9388, + "step": 28717 + }, + { + "epoch": 0.69, + "learning_rate": 4.61356228680497e-06, + "loss": 0.989, + "step": 28718 + }, + { + "epoch": 0.69, + "learning_rate": 4.612906058436485e-06, + "loss": 1.1339, + "step": 28719 + }, + { + "epoch": 0.69, + "learning_rate": 4.6122498627502276e-06, + "loss": 1.0813, + "step": 28720 + }, + { + "epoch": 0.69, + "learning_rate": 4.611593699750183e-06, + "loss": 0.9073, + "step": 28721 + }, + { + "epoch": 0.69, + "learning_rate": 4.6109375694403356e-06, + "loss": 1.0731, + "step": 28722 + }, + { + "epoch": 0.69, + "learning_rate": 4.610281471824659e-06, + "loss": 0.9997, + "step": 28723 + }, + { + "epoch": 0.69, + "learning_rate": 4.609625406907141e-06, + "loss": 0.6025, + "step": 28724 + }, + { + "epoch": 0.69, + "learning_rate": 4.608969374691753e-06, + "loss": 1.1324, + "step": 28725 + }, + { + "epoch": 0.69, + "learning_rate": 4.60831337518248e-06, + "loss": 0.7837, + "step": 28726 + }, + { + "epoch": 0.69, + "learning_rate": 4.607657408383306e-06, + "loss": 1.1236, + "step": 28727 + }, + { + "epoch": 0.69, + "learning_rate": 4.607001474298203e-06, + "loss": 1.3165, + "step": 28728 + }, + { + "epoch": 0.69, + "learning_rate": 4.606345572931153e-06, + "loss": 0.8669, + "step": 28729 + }, + { + "epoch": 0.69, + "learning_rate": 4.6056897042861385e-06, + "loss": 1.1005, + "step": 28730 + }, + { + "epoch": 0.69, + "learning_rate": 4.6050338683671324e-06, + "loss": 1.0767, + "step": 28731 + }, + { + "epoch": 0.69, + "learning_rate": 4.60437806517812e-06, + "loss": 0.9927, + "step": 28732 + }, + { + "epoch": 0.69, + "learning_rate": 4.603722294723074e-06, + "loss": 1.0205, + "step": 28733 + }, + { + "epoch": 0.69, + "learning_rate": 4.6030665570059755e-06, + "loss": 1.209, + "step": 28734 + }, + { + "epoch": 0.69, + "learning_rate": 4.602410852030807e-06, + "loss": 1.0451, + "step": 28735 + }, + { + "epoch": 0.69, + "learning_rate": 4.601755179801537e-06, + "loss": 1.1738, + "step": 28736 + }, + { + "epoch": 0.69, + "learning_rate": 4.601099540322149e-06, + "loss": 1.1318, + "step": 28737 + }, + { + "epoch": 0.69, + "learning_rate": 4.600443933596625e-06, + "loss": 1.0942, + "step": 28738 + }, + { + "epoch": 0.69, + "learning_rate": 4.599788359628933e-06, + "loss": 0.8948, + "step": 28739 + }, + { + "epoch": 0.69, + "learning_rate": 4.599132818423056e-06, + "loss": 0.7908, + "step": 28740 + }, + { + "epoch": 0.69, + "learning_rate": 4.598477309982971e-06, + "loss": 0.9895, + "step": 28741 + }, + { + "epoch": 0.69, + "learning_rate": 4.597821834312652e-06, + "loss": 0.851, + "step": 28742 + }, + { + "epoch": 0.69, + "learning_rate": 4.59716639141608e-06, + "loss": 0.8876, + "step": 28743 + }, + { + "epoch": 0.69, + "learning_rate": 4.596510981297226e-06, + "loss": 0.5482, + "step": 28744 + }, + { + "epoch": 0.69, + "learning_rate": 4.595855603960069e-06, + "loss": 1.0493, + "step": 28745 + }, + { + "epoch": 0.69, + "learning_rate": 4.595200259408587e-06, + "loss": 1.1693, + "step": 28746 + }, + { + "epoch": 0.69, + "learning_rate": 4.5945449476467515e-06, + "loss": 1.0475, + "step": 28747 + }, + { + "epoch": 0.69, + "learning_rate": 4.59388966867854e-06, + "loss": 0.7614, + "step": 28748 + }, + { + "epoch": 0.69, + "learning_rate": 4.593234422507932e-06, + "loss": 0.9323, + "step": 28749 + }, + { + "epoch": 0.69, + "learning_rate": 4.592579209138896e-06, + "loss": 1.2124, + "step": 28750 + }, + { + "epoch": 0.69, + "learning_rate": 4.591924028575412e-06, + "loss": 1.0773, + "step": 28751 + }, + { + "epoch": 0.69, + "learning_rate": 4.59126888082145e-06, + "loss": 1.1666, + "step": 28752 + }, + { + "epoch": 0.69, + "learning_rate": 4.5906137658809875e-06, + "loss": 1.0738, + "step": 28753 + }, + { + "epoch": 0.69, + "learning_rate": 4.589958683758002e-06, + "loss": 1.0867, + "step": 28754 + }, + { + "epoch": 0.69, + "learning_rate": 4.589303634456462e-06, + "loss": 0.9439, + "step": 28755 + }, + { + "epoch": 0.69, + "learning_rate": 4.588648617980343e-06, + "loss": 0.76, + "step": 28756 + }, + { + "epoch": 0.69, + "learning_rate": 4.5879936343336255e-06, + "loss": 1.2418, + "step": 28757 + }, + { + "epoch": 0.69, + "learning_rate": 4.587338683520273e-06, + "loss": 1.0429, + "step": 28758 + }, + { + "epoch": 0.69, + "learning_rate": 4.586683765544268e-06, + "loss": 0.9976, + "step": 28759 + }, + { + "epoch": 0.69, + "learning_rate": 4.586028880409575e-06, + "loss": 1.1087, + "step": 28760 + }, + { + "epoch": 0.69, + "learning_rate": 4.58537402812017e-06, + "loss": 1.2348, + "step": 28761 + }, + { + "epoch": 0.69, + "learning_rate": 4.584719208680032e-06, + "loss": 1.1975, + "step": 28762 + }, + { + "epoch": 0.69, + "learning_rate": 4.584064422093125e-06, + "loss": 1.0983, + "step": 28763 + }, + { + "epoch": 0.69, + "learning_rate": 4.583409668363425e-06, + "loss": 1.3015, + "step": 28764 + }, + { + "epoch": 0.69, + "learning_rate": 4.582754947494909e-06, + "loss": 1.0912, + "step": 28765 + }, + { + "epoch": 0.69, + "learning_rate": 4.582100259491541e-06, + "loss": 0.9349, + "step": 28766 + }, + { + "epoch": 0.69, + "learning_rate": 4.581445604357299e-06, + "loss": 1.0086, + "step": 28767 + }, + { + "epoch": 0.69, + "learning_rate": 4.580790982096149e-06, + "loss": 0.9697, + "step": 28768 + }, + { + "epoch": 0.69, + "learning_rate": 4.5801363927120645e-06, + "loss": 0.9808, + "step": 28769 + }, + { + "epoch": 0.69, + "learning_rate": 4.579481836209023e-06, + "loss": 0.9449, + "step": 28770 + }, + { + "epoch": 0.69, + "learning_rate": 4.578827312590985e-06, + "loss": 1.0321, + "step": 28771 + }, + { + "epoch": 0.69, + "learning_rate": 4.578172821861927e-06, + "loss": 1.1668, + "step": 28772 + }, + { + "epoch": 0.69, + "learning_rate": 4.577518364025823e-06, + "loss": 0.9569, + "step": 28773 + }, + { + "epoch": 0.69, + "learning_rate": 4.576863939086636e-06, + "loss": 0.9904, + "step": 28774 + }, + { + "epoch": 0.69, + "learning_rate": 4.576209547048344e-06, + "loss": 0.987, + "step": 28775 + }, + { + "epoch": 0.69, + "learning_rate": 4.575555187914908e-06, + "loss": 1.1063, + "step": 28776 + }, + { + "epoch": 0.69, + "learning_rate": 4.574900861690304e-06, + "loss": 1.0783, + "step": 28777 + }, + { + "epoch": 0.69, + "learning_rate": 4.574246568378504e-06, + "loss": 0.9739, + "step": 28778 + }, + { + "epoch": 0.69, + "learning_rate": 4.5735923079834695e-06, + "loss": 1.0285, + "step": 28779 + }, + { + "epoch": 0.69, + "learning_rate": 4.572938080509175e-06, + "loss": 1.035, + "step": 28780 + }, + { + "epoch": 0.69, + "learning_rate": 4.572283885959591e-06, + "loss": 1.1331, + "step": 28781 + }, + { + "epoch": 0.69, + "learning_rate": 4.571629724338681e-06, + "loss": 0.8057, + "step": 28782 + }, + { + "epoch": 0.69, + "learning_rate": 4.570975595650417e-06, + "loss": 1.0976, + "step": 28783 + }, + { + "epoch": 0.69, + "learning_rate": 4.57032149989877e-06, + "loss": 1.0737, + "step": 28784 + }, + { + "epoch": 0.69, + "learning_rate": 4.569667437087703e-06, + "loss": 1.3611, + "step": 28785 + }, + { + "epoch": 0.69, + "learning_rate": 4.569013407221189e-06, + "loss": 1.3036, + "step": 28786 + }, + { + "epoch": 0.69, + "learning_rate": 4.568359410303189e-06, + "loss": 1.0759, + "step": 28787 + }, + { + "epoch": 0.69, + "learning_rate": 4.567705446337676e-06, + "loss": 0.9509, + "step": 28788 + }, + { + "epoch": 0.69, + "learning_rate": 4.5670515153286186e-06, + "loss": 1.1251, + "step": 28789 + }, + { + "epoch": 0.69, + "learning_rate": 4.566397617279979e-06, + "loss": 0.9078, + "step": 28790 + }, + { + "epoch": 0.69, + "learning_rate": 4.565743752195727e-06, + "loss": 1.1865, + "step": 28791 + }, + { + "epoch": 0.69, + "learning_rate": 4.565089920079834e-06, + "loss": 0.9022, + "step": 28792 + }, + { + "epoch": 0.69, + "learning_rate": 4.564436120936257e-06, + "loss": 1.012, + "step": 28793 + }, + { + "epoch": 0.69, + "learning_rate": 4.563782354768972e-06, + "loss": 0.8055, + "step": 28794 + }, + { + "epoch": 0.69, + "learning_rate": 4.563128621581937e-06, + "loss": 0.9919, + "step": 28795 + }, + { + "epoch": 0.69, + "learning_rate": 4.562474921379121e-06, + "loss": 1.2119, + "step": 28796 + }, + { + "epoch": 0.69, + "learning_rate": 4.561821254164495e-06, + "loss": 1.0056, + "step": 28797 + }, + { + "epoch": 0.69, + "learning_rate": 4.561167619942017e-06, + "loss": 1.1286, + "step": 28798 + }, + { + "epoch": 0.69, + "learning_rate": 4.560514018715655e-06, + "loss": 0.9847, + "step": 28799 + }, + { + "epoch": 0.69, + "learning_rate": 4.55986045048938e-06, + "loss": 1.263, + "step": 28800 + }, + { + "epoch": 0.69, + "learning_rate": 4.559206915267147e-06, + "loss": 1.1696, + "step": 28801 + }, + { + "epoch": 0.69, + "learning_rate": 4.5585534130529296e-06, + "loss": 1.1209, + "step": 28802 + }, + { + "epoch": 0.69, + "learning_rate": 4.557899943850686e-06, + "loss": 1.4623, + "step": 28803 + }, + { + "epoch": 0.69, + "learning_rate": 4.557246507664383e-06, + "loss": 1.2936, + "step": 28804 + }, + { + "epoch": 0.69, + "learning_rate": 4.5565931044979886e-06, + "loss": 0.8911, + "step": 28805 + }, + { + "epoch": 0.69, + "learning_rate": 4.555939734355459e-06, + "loss": 1.0237, + "step": 28806 + }, + { + "epoch": 0.69, + "learning_rate": 4.555286397240763e-06, + "loss": 1.1927, + "step": 28807 + }, + { + "epoch": 0.69, + "learning_rate": 4.554633093157868e-06, + "loss": 1.1064, + "step": 28808 + }, + { + "epoch": 0.69, + "learning_rate": 4.55397982211073e-06, + "loss": 1.1117, + "step": 28809 + }, + { + "epoch": 0.69, + "learning_rate": 4.553326584103318e-06, + "loss": 1.2105, + "step": 28810 + }, + { + "epoch": 0.69, + "learning_rate": 4.552673379139589e-06, + "loss": 1.0143, + "step": 28811 + }, + { + "epoch": 0.69, + "learning_rate": 4.55202020722351e-06, + "loss": 1.0279, + "step": 28812 + }, + { + "epoch": 0.69, + "learning_rate": 4.551367068359046e-06, + "loss": 1.1021, + "step": 28813 + }, + { + "epoch": 0.69, + "learning_rate": 4.550713962550153e-06, + "loss": 0.8883, + "step": 28814 + }, + { + "epoch": 0.69, + "learning_rate": 4.5500608898007966e-06, + "loss": 1.07, + "step": 28815 + }, + { + "epoch": 0.69, + "learning_rate": 4.549407850114943e-06, + "loss": 1.0275, + "step": 28816 + }, + { + "epoch": 0.69, + "learning_rate": 4.548754843496545e-06, + "loss": 1.2707, + "step": 28817 + }, + { + "epoch": 0.69, + "learning_rate": 4.5481018699495735e-06, + "loss": 0.7594, + "step": 28818 + }, + { + "epoch": 0.69, + "learning_rate": 4.547448929477983e-06, + "loss": 0.9227, + "step": 28819 + }, + { + "epoch": 0.69, + "learning_rate": 4.546796022085736e-06, + "loss": 0.8793, + "step": 28820 + }, + { + "epoch": 0.69, + "learning_rate": 4.546143147776799e-06, + "loss": 0.7827, + "step": 28821 + }, + { + "epoch": 0.69, + "learning_rate": 4.545490306555125e-06, + "loss": 1.0597, + "step": 28822 + }, + { + "epoch": 0.69, + "learning_rate": 4.544837498424678e-06, + "loss": 1.3164, + "step": 28823 + }, + { + "epoch": 0.69, + "learning_rate": 4.544184723389423e-06, + "loss": 1.0297, + "step": 28824 + }, + { + "epoch": 0.69, + "learning_rate": 4.543531981453312e-06, + "loss": 1.245, + "step": 28825 + }, + { + "epoch": 0.69, + "learning_rate": 4.542879272620313e-06, + "loss": 0.9969, + "step": 28826 + }, + { + "epoch": 0.69, + "learning_rate": 4.542226596894378e-06, + "loss": 1.1007, + "step": 28827 + }, + { + "epoch": 0.69, + "learning_rate": 4.54157395427947e-06, + "loss": 1.2597, + "step": 28828 + }, + { + "epoch": 0.69, + "learning_rate": 4.540921344779553e-06, + "loss": 1.1254, + "step": 28829 + }, + { + "epoch": 0.69, + "learning_rate": 4.540268768398578e-06, + "loss": 1.0581, + "step": 28830 + }, + { + "epoch": 0.69, + "learning_rate": 4.539616225140509e-06, + "loss": 1.2452, + "step": 28831 + }, + { + "epoch": 0.69, + "learning_rate": 4.538963715009304e-06, + "loss": 1.0785, + "step": 28832 + }, + { + "epoch": 0.69, + "learning_rate": 4.5383112380089215e-06, + "loss": 0.8994, + "step": 28833 + }, + { + "epoch": 0.69, + "learning_rate": 4.537658794143322e-06, + "loss": 1.0539, + "step": 28834 + }, + { + "epoch": 0.69, + "learning_rate": 4.53700638341646e-06, + "loss": 1.2182, + "step": 28835 + }, + { + "epoch": 0.69, + "learning_rate": 4.536354005832296e-06, + "loss": 1.2434, + "step": 28836 + }, + { + "epoch": 0.69, + "learning_rate": 4.535701661394789e-06, + "loss": 0.9392, + "step": 28837 + }, + { + "epoch": 0.69, + "learning_rate": 4.535049350107892e-06, + "loss": 1.1097, + "step": 28838 + }, + { + "epoch": 0.69, + "learning_rate": 4.534397071975568e-06, + "loss": 1.0616, + "step": 28839 + }, + { + "epoch": 0.69, + "learning_rate": 4.533744827001768e-06, + "loss": 1.1756, + "step": 28840 + }, + { + "epoch": 0.69, + "learning_rate": 4.533092615190453e-06, + "loss": 0.9247, + "step": 28841 + }, + { + "epoch": 0.69, + "learning_rate": 4.532440436545583e-06, + "loss": 1.3089, + "step": 28842 + }, + { + "epoch": 0.69, + "learning_rate": 4.531788291071106e-06, + "loss": 0.8685, + "step": 28843 + }, + { + "epoch": 0.69, + "learning_rate": 4.531136178770984e-06, + "loss": 1.0201, + "step": 28844 + }, + { + "epoch": 0.69, + "learning_rate": 4.530484099649176e-06, + "loss": 0.8654, + "step": 28845 + }, + { + "epoch": 0.69, + "learning_rate": 4.529832053709632e-06, + "loss": 1.1078, + "step": 28846 + }, + { + "epoch": 0.69, + "learning_rate": 4.529180040956312e-06, + "loss": 0.7826, + "step": 28847 + }, + { + "epoch": 0.69, + "learning_rate": 4.528528061393167e-06, + "loss": 0.8951, + "step": 28848 + }, + { + "epoch": 0.69, + "learning_rate": 4.527876115024156e-06, + "loss": 0.9713, + "step": 28849 + }, + { + "epoch": 0.69, + "learning_rate": 4.527224201853235e-06, + "loss": 1.3135, + "step": 28850 + }, + { + "epoch": 0.69, + "learning_rate": 4.526572321884356e-06, + "loss": 1.0676, + "step": 28851 + }, + { + "epoch": 0.69, + "learning_rate": 4.525920475121475e-06, + "loss": 0.962, + "step": 28852 + }, + { + "epoch": 0.69, + "learning_rate": 4.525268661568549e-06, + "loss": 0.758, + "step": 28853 + }, + { + "epoch": 0.69, + "learning_rate": 4.5246168812295286e-06, + "loss": 1.1866, + "step": 28854 + }, + { + "epoch": 0.69, + "learning_rate": 4.523965134108372e-06, + "loss": 0.9716, + "step": 28855 + }, + { + "epoch": 0.69, + "learning_rate": 4.523313420209028e-06, + "loss": 1.01, + "step": 28856 + }, + { + "epoch": 0.69, + "learning_rate": 4.522661739535453e-06, + "loss": 1.1404, + "step": 28857 + }, + { + "epoch": 0.69, + "learning_rate": 4.522010092091605e-06, + "loss": 1.2717, + "step": 28858 + }, + { + "epoch": 0.69, + "learning_rate": 4.521358477881429e-06, + "loss": 1.0398, + "step": 28859 + }, + { + "epoch": 0.69, + "learning_rate": 4.520706896908882e-06, + "loss": 0.8857, + "step": 28860 + }, + { + "epoch": 0.69, + "learning_rate": 4.520055349177922e-06, + "loss": 1.1341, + "step": 28861 + }, + { + "epoch": 0.69, + "learning_rate": 4.519403834692493e-06, + "loss": 0.9244, + "step": 28862 + }, + { + "epoch": 0.69, + "learning_rate": 4.5187523534565526e-06, + "loss": 1.3674, + "step": 28863 + }, + { + "epoch": 0.69, + "learning_rate": 4.518100905474056e-06, + "loss": 1.1587, + "step": 28864 + }, + { + "epoch": 0.69, + "learning_rate": 4.517449490748947e-06, + "loss": 1.3424, + "step": 28865 + }, + { + "epoch": 0.69, + "learning_rate": 4.516798109285186e-06, + "loss": 1.0167, + "step": 28866 + }, + { + "epoch": 0.69, + "learning_rate": 4.51614676108672e-06, + "loss": 0.9594, + "step": 28867 + }, + { + "epoch": 0.69, + "learning_rate": 4.515495446157498e-06, + "loss": 1.1463, + "step": 28868 + }, + { + "epoch": 0.69, + "learning_rate": 4.514844164501482e-06, + "loss": 1.1102, + "step": 28869 + }, + { + "epoch": 0.69, + "learning_rate": 4.514192916122611e-06, + "loss": 1.1601, + "step": 28870 + }, + { + "epoch": 0.69, + "learning_rate": 4.513541701024841e-06, + "loss": 0.8108, + "step": 28871 + }, + { + "epoch": 0.69, + "learning_rate": 4.512890519212127e-06, + "loss": 0.9897, + "step": 28872 + }, + { + "epoch": 0.69, + "learning_rate": 4.5122393706884125e-06, + "loss": 1.1714, + "step": 28873 + }, + { + "epoch": 0.69, + "learning_rate": 4.511588255457655e-06, + "loss": 0.8954, + "step": 28874 + }, + { + "epoch": 0.69, + "learning_rate": 4.510937173523796e-06, + "loss": 0.9831, + "step": 28875 + }, + { + "epoch": 0.69, + "learning_rate": 4.5102861248907906e-06, + "loss": 1.31, + "step": 28876 + }, + { + "epoch": 0.69, + "learning_rate": 4.50963510956259e-06, + "loss": 1.0543, + "step": 28877 + }, + { + "epoch": 0.69, + "learning_rate": 4.5089841275431405e-06, + "loss": 1.0189, + "step": 28878 + }, + { + "epoch": 0.69, + "learning_rate": 4.508333178836391e-06, + "loss": 0.9602, + "step": 28879 + }, + { + "epoch": 0.69, + "learning_rate": 4.507682263446297e-06, + "loss": 1.026, + "step": 28880 + }, + { + "epoch": 0.69, + "learning_rate": 4.507031381376799e-06, + "loss": 0.7579, + "step": 28881 + }, + { + "epoch": 0.69, + "learning_rate": 4.506380532631853e-06, + "loss": 1.074, + "step": 28882 + }, + { + "epoch": 0.69, + "learning_rate": 4.5057297172154005e-06, + "loss": 1.1584, + "step": 28883 + }, + { + "epoch": 0.69, + "learning_rate": 4.505078935131393e-06, + "loss": 1.1721, + "step": 28884 + }, + { + "epoch": 0.69, + "learning_rate": 4.504428186383784e-06, + "loss": 0.7371, + "step": 28885 + }, + { + "epoch": 0.69, + "learning_rate": 4.503777470976511e-06, + "loss": 0.789, + "step": 28886 + }, + { + "epoch": 0.69, + "learning_rate": 4.50312678891353e-06, + "loss": 1.1642, + "step": 28887 + }, + { + "epoch": 0.69, + "learning_rate": 4.502476140198787e-06, + "loss": 0.9381, + "step": 28888 + }, + { + "epoch": 0.69, + "learning_rate": 4.501825524836226e-06, + "loss": 1.0782, + "step": 28889 + }, + { + "epoch": 0.69, + "learning_rate": 4.501174942829799e-06, + "loss": 1.1698, + "step": 28890 + }, + { + "epoch": 0.69, + "learning_rate": 4.5005243941834476e-06, + "loss": 1.0681, + "step": 28891 + }, + { + "epoch": 0.69, + "learning_rate": 4.49987387890112e-06, + "loss": 1.07, + "step": 28892 + }, + { + "epoch": 0.69, + "learning_rate": 4.499223396986769e-06, + "loss": 1.0086, + "step": 28893 + }, + { + "epoch": 0.69, + "learning_rate": 4.498572948444331e-06, + "loss": 1.1172, + "step": 28894 + }, + { + "epoch": 0.69, + "learning_rate": 4.4979225332777566e-06, + "loss": 0.8802, + "step": 28895 + }, + { + "epoch": 0.69, + "learning_rate": 4.497272151490997e-06, + "loss": 0.9808, + "step": 28896 + }, + { + "epoch": 0.69, + "learning_rate": 4.496621803087988e-06, + "loss": 1.1528, + "step": 28897 + }, + { + "epoch": 0.69, + "learning_rate": 4.4959714880726845e-06, + "loss": 0.9735, + "step": 28898 + }, + { + "epoch": 0.69, + "learning_rate": 4.495321206449022e-06, + "loss": 1.2762, + "step": 28899 + }, + { + "epoch": 0.7, + "learning_rate": 4.494670958220954e-06, + "loss": 0.9441, + "step": 28900 + }, + { + "epoch": 0.7, + "learning_rate": 4.494020743392424e-06, + "loss": 1.0983, + "step": 28901 + }, + { + "epoch": 0.7, + "learning_rate": 4.493370561967372e-06, + "loss": 1.0056, + "step": 28902 + }, + { + "epoch": 0.7, + "learning_rate": 4.492720413949745e-06, + "loss": 1.1584, + "step": 28903 + }, + { + "epoch": 0.7, + "learning_rate": 4.492070299343492e-06, + "loss": 1.1586, + "step": 28904 + }, + { + "epoch": 0.7, + "learning_rate": 4.491420218152549e-06, + "loss": 0.937, + "step": 28905 + }, + { + "epoch": 0.7, + "learning_rate": 4.490770170380867e-06, + "loss": 1.0933, + "step": 28906 + }, + { + "epoch": 0.7, + "learning_rate": 4.490120156032384e-06, + "loss": 1.2226, + "step": 28907 + }, + { + "epoch": 0.7, + "learning_rate": 4.489470175111046e-06, + "loss": 1.2792, + "step": 28908 + }, + { + "epoch": 0.7, + "learning_rate": 4.4888202276208e-06, + "loss": 0.9681, + "step": 28909 + }, + { + "epoch": 0.7, + "learning_rate": 4.48817031356558e-06, + "loss": 0.9586, + "step": 28910 + }, + { + "epoch": 0.7, + "learning_rate": 4.487520432949335e-06, + "loss": 0.9718, + "step": 28911 + }, + { + "epoch": 0.7, + "learning_rate": 4.486870585776012e-06, + "loss": 1.078, + "step": 28912 + }, + { + "epoch": 0.7, + "learning_rate": 4.486220772049543e-06, + "loss": 1.168, + "step": 28913 + }, + { + "epoch": 0.7, + "learning_rate": 4.485570991773877e-06, + "loss": 0.7166, + "step": 28914 + }, + { + "epoch": 0.7, + "learning_rate": 4.484921244952957e-06, + "loss": 1.0456, + "step": 28915 + }, + { + "epoch": 0.7, + "learning_rate": 4.4842715315907205e-06, + "loss": 1.09, + "step": 28916 + }, + { + "epoch": 0.7, + "learning_rate": 4.483621851691113e-06, + "loss": 0.987, + "step": 28917 + }, + { + "epoch": 0.7, + "learning_rate": 4.482972205258072e-06, + "loss": 0.9624, + "step": 28918 + }, + { + "epoch": 0.7, + "learning_rate": 4.482322592295541e-06, + "loss": 1.112, + "step": 28919 + }, + { + "epoch": 0.7, + "learning_rate": 4.4816730128074635e-06, + "loss": 1.0305, + "step": 28920 + }, + { + "epoch": 0.7, + "learning_rate": 4.481023466797776e-06, + "loss": 1.185, + "step": 28921 + }, + { + "epoch": 0.7, + "learning_rate": 4.480373954270419e-06, + "loss": 1.0856, + "step": 28922 + }, + { + "epoch": 0.7, + "learning_rate": 4.47972447522934e-06, + "loss": 1.2298, + "step": 28923 + }, + { + "epoch": 0.7, + "learning_rate": 4.47907502967847e-06, + "loss": 0.9568, + "step": 28924 + }, + { + "epoch": 0.7, + "learning_rate": 4.478425617621756e-06, + "loss": 1.1166, + "step": 28925 + }, + { + "epoch": 0.7, + "learning_rate": 4.477776239063133e-06, + "loss": 1.1182, + "step": 28926 + }, + { + "epoch": 0.7, + "learning_rate": 4.477126894006542e-06, + "loss": 1.0908, + "step": 28927 + }, + { + "epoch": 0.7, + "learning_rate": 4.476477582455926e-06, + "loss": 1.1095, + "step": 28928 + }, + { + "epoch": 0.7, + "learning_rate": 4.475828304415219e-06, + "loss": 1.0808, + "step": 28929 + }, + { + "epoch": 0.7, + "learning_rate": 4.475179059888361e-06, + "loss": 1.1029, + "step": 28930 + }, + { + "epoch": 0.7, + "learning_rate": 4.474529848879296e-06, + "loss": 1.2702, + "step": 28931 + }, + { + "epoch": 0.7, + "learning_rate": 4.4738806713919555e-06, + "loss": 1.1753, + "step": 28932 + }, + { + "epoch": 0.7, + "learning_rate": 4.4732315274302855e-06, + "loss": 0.9061, + "step": 28933 + }, + { + "epoch": 0.7, + "learning_rate": 4.472582416998215e-06, + "loss": 1.1654, + "step": 28934 + }, + { + "epoch": 0.7, + "learning_rate": 4.471933340099687e-06, + "loss": 1.1279, + "step": 28935 + }, + { + "epoch": 0.7, + "learning_rate": 4.471284296738643e-06, + "loss": 1.027, + "step": 28936 + }, + { + "epoch": 0.7, + "learning_rate": 4.4706352869190124e-06, + "loss": 0.8768, + "step": 28937 + }, + { + "epoch": 0.7, + "learning_rate": 4.469986310644739e-06, + "loss": 0.9636, + "step": 28938 + }, + { + "epoch": 0.7, + "learning_rate": 4.4693373679197595e-06, + "loss": 1.0599, + "step": 28939 + }, + { + "epoch": 0.7, + "learning_rate": 4.468688458748006e-06, + "loss": 1.0327, + "step": 28940 + }, + { + "epoch": 0.7, + "learning_rate": 4.468039583133423e-06, + "loss": 1.129, + "step": 28941 + }, + { + "epoch": 0.7, + "learning_rate": 4.467390741079939e-06, + "loss": 1.1204, + "step": 28942 + }, + { + "epoch": 0.7, + "learning_rate": 4.466741932591493e-06, + "loss": 1.1143, + "step": 28943 + }, + { + "epoch": 0.7, + "learning_rate": 4.466093157672027e-06, + "loss": 1.0564, + "step": 28944 + }, + { + "epoch": 0.7, + "learning_rate": 4.465444416325467e-06, + "loss": 1.0231, + "step": 28945 + }, + { + "epoch": 0.7, + "learning_rate": 4.464795708555756e-06, + "loss": 1.0696, + "step": 28946 + }, + { + "epoch": 0.7, + "learning_rate": 4.464147034366829e-06, + "loss": 0.909, + "step": 28947 + }, + { + "epoch": 0.7, + "learning_rate": 4.463498393762618e-06, + "loss": 1.0955, + "step": 28948 + }, + { + "epoch": 0.7, + "learning_rate": 4.4628497867470626e-06, + "loss": 0.9431, + "step": 28949 + }, + { + "epoch": 0.7, + "learning_rate": 4.462201213324092e-06, + "loss": 0.9477, + "step": 28950 + }, + { + "epoch": 0.7, + "learning_rate": 4.461552673497643e-06, + "loss": 1.1841, + "step": 28951 + }, + { + "epoch": 0.7, + "learning_rate": 4.460904167271656e-06, + "loss": 0.9103, + "step": 28952 + }, + { + "epoch": 0.7, + "learning_rate": 4.460255694650056e-06, + "loss": 1.2757, + "step": 28953 + }, + { + "epoch": 0.7, + "learning_rate": 4.4596072556367835e-06, + "loss": 1.0759, + "step": 28954 + }, + { + "epoch": 0.7, + "learning_rate": 4.458958850235772e-06, + "loss": 1.0265, + "step": 28955 + }, + { + "epoch": 0.7, + "learning_rate": 4.458310478450951e-06, + "loss": 1.17, + "step": 28956 + }, + { + "epoch": 0.7, + "learning_rate": 4.457662140286257e-06, + "loss": 0.925, + "step": 28957 + }, + { + "epoch": 0.7, + "learning_rate": 4.457013835745627e-06, + "loss": 1.1504, + "step": 28958 + }, + { + "epoch": 0.7, + "learning_rate": 4.456365564832986e-06, + "loss": 1.1709, + "step": 28959 + }, + { + "epoch": 0.7, + "learning_rate": 4.455717327552276e-06, + "loss": 1.0943, + "step": 28960 + }, + { + "epoch": 0.7, + "learning_rate": 4.4550691239074194e-06, + "loss": 1.0902, + "step": 28961 + }, + { + "epoch": 0.7, + "learning_rate": 4.454420953902356e-06, + "loss": 0.8663, + "step": 28962 + }, + { + "epoch": 0.7, + "learning_rate": 4.453772817541019e-06, + "loss": 1.098, + "step": 28963 + }, + { + "epoch": 0.7, + "learning_rate": 4.453124714827334e-06, + "loss": 1.0159, + "step": 28964 + }, + { + "epoch": 0.7, + "learning_rate": 4.452476645765237e-06, + "loss": 0.9093, + "step": 28965 + }, + { + "epoch": 0.7, + "learning_rate": 4.451828610358663e-06, + "loss": 1.0789, + "step": 28966 + }, + { + "epoch": 0.7, + "learning_rate": 4.451180608611536e-06, + "loss": 1.2593, + "step": 28967 + }, + { + "epoch": 0.7, + "learning_rate": 4.4505326405277945e-06, + "loss": 0.9299, + "step": 28968 + }, + { + "epoch": 0.7, + "learning_rate": 4.449884706111363e-06, + "loss": 1.2697, + "step": 28969 + }, + { + "epoch": 0.7, + "learning_rate": 4.449236805366175e-06, + "loss": 1.0057, + "step": 28970 + }, + { + "epoch": 0.7, + "learning_rate": 4.448588938296166e-06, + "loss": 1.1179, + "step": 28971 + }, + { + "epoch": 0.7, + "learning_rate": 4.447941104905258e-06, + "loss": 0.7485, + "step": 28972 + }, + { + "epoch": 0.7, + "learning_rate": 4.447293305197385e-06, + "loss": 0.9271, + "step": 28973 + }, + { + "epoch": 0.7, + "learning_rate": 4.446645539176483e-06, + "loss": 1.3136, + "step": 28974 + }, + { + "epoch": 0.7, + "learning_rate": 4.445997806846471e-06, + "loss": 1.2005, + "step": 28975 + }, + { + "epoch": 0.7, + "learning_rate": 4.445350108211287e-06, + "loss": 1.0106, + "step": 28976 + }, + { + "epoch": 0.7, + "learning_rate": 4.444702443274855e-06, + "loss": 1.0778, + "step": 28977 + }, + { + "epoch": 0.7, + "learning_rate": 4.444054812041107e-06, + "loss": 0.8466, + "step": 28978 + }, + { + "epoch": 0.7, + "learning_rate": 4.443407214513974e-06, + "loss": 1.1744, + "step": 28979 + }, + { + "epoch": 0.7, + "learning_rate": 4.442759650697379e-06, + "loss": 0.9981, + "step": 28980 + }, + { + "epoch": 0.7, + "learning_rate": 4.442112120595254e-06, + "loss": 0.9019, + "step": 28981 + }, + { + "epoch": 0.7, + "learning_rate": 4.441464624211531e-06, + "loss": 1.0282, + "step": 28982 + }, + { + "epoch": 0.7, + "learning_rate": 4.440817161550132e-06, + "loss": 0.9073, + "step": 28983 + }, + { + "epoch": 0.7, + "learning_rate": 4.440169732614991e-06, + "loss": 1.1818, + "step": 28984 + }, + { + "epoch": 0.7, + "learning_rate": 4.439522337410028e-06, + "loss": 1.1466, + "step": 28985 + }, + { + "epoch": 0.7, + "learning_rate": 4.438874975939176e-06, + "loss": 0.8876, + "step": 28986 + }, + { + "epoch": 0.7, + "learning_rate": 4.438227648206365e-06, + "loss": 0.9215, + "step": 28987 + }, + { + "epoch": 0.7, + "learning_rate": 4.437580354215514e-06, + "loss": 1.0566, + "step": 28988 + }, + { + "epoch": 0.7, + "learning_rate": 4.436933093970556e-06, + "loss": 0.9753, + "step": 28989 + }, + { + "epoch": 0.7, + "learning_rate": 4.436285867475418e-06, + "loss": 1.0163, + "step": 28990 + }, + { + "epoch": 0.7, + "learning_rate": 4.435638674734023e-06, + "loss": 1.1387, + "step": 28991 + }, + { + "epoch": 0.7, + "learning_rate": 4.434991515750303e-06, + "loss": 1.1435, + "step": 28992 + }, + { + "epoch": 0.7, + "learning_rate": 4.434344390528177e-06, + "loss": 1.2058, + "step": 28993 + }, + { + "epoch": 0.7, + "learning_rate": 4.433697299071574e-06, + "loss": 1.194, + "step": 28994 + }, + { + "epoch": 0.7, + "learning_rate": 4.4330502413844236e-06, + "loss": 0.6054, + "step": 28995 + }, + { + "epoch": 0.7, + "learning_rate": 4.432403217470645e-06, + "loss": 1.2238, + "step": 28996 + }, + { + "epoch": 0.7, + "learning_rate": 4.4317562273341666e-06, + "loss": 1.0836, + "step": 28997 + }, + { + "epoch": 0.7, + "learning_rate": 4.431109270978916e-06, + "loss": 1.0214, + "step": 28998 + }, + { + "epoch": 0.7, + "learning_rate": 4.430462348408808e-06, + "loss": 1.1593, + "step": 28999 + }, + { + "epoch": 0.7, + "learning_rate": 4.429815459627784e-06, + "loss": 1.1487, + "step": 29000 + }, + { + "epoch": 0.7, + "learning_rate": 4.4291686046397554e-06, + "loss": 1.1061, + "step": 29001 + }, + { + "epoch": 0.7, + "learning_rate": 4.42852178344865e-06, + "loss": 0.9601, + "step": 29002 + }, + { + "epoch": 0.7, + "learning_rate": 4.427874996058395e-06, + "loss": 1.2011, + "step": 29003 + }, + { + "epoch": 0.7, + "learning_rate": 4.42722824247291e-06, + "loss": 1.1097, + "step": 29004 + }, + { + "epoch": 0.7, + "learning_rate": 4.426581522696123e-06, + "loss": 0.7987, + "step": 29005 + }, + { + "epoch": 0.7, + "learning_rate": 4.425934836731951e-06, + "loss": 0.992, + "step": 29006 + }, + { + "epoch": 0.7, + "learning_rate": 4.425288184584322e-06, + "loss": 0.8807, + "step": 29007 + }, + { + "epoch": 0.7, + "learning_rate": 4.424641566257162e-06, + "loss": 1.3029, + "step": 29008 + }, + { + "epoch": 0.7, + "learning_rate": 4.423994981754387e-06, + "loss": 1.1925, + "step": 29009 + }, + { + "epoch": 0.7, + "learning_rate": 4.4233484310799215e-06, + "loss": 1.34, + "step": 29010 + }, + { + "epoch": 0.7, + "learning_rate": 4.4227019142376935e-06, + "loss": 0.8458, + "step": 29011 + }, + { + "epoch": 0.7, + "learning_rate": 4.422055431231619e-06, + "loss": 1.0338, + "step": 29012 + }, + { + "epoch": 0.7, + "learning_rate": 4.421408982065623e-06, + "loss": 1.0891, + "step": 29013 + }, + { + "epoch": 0.7, + "learning_rate": 4.420762566743625e-06, + "loss": 1.0472, + "step": 29014 + }, + { + "epoch": 0.7, + "learning_rate": 4.420116185269549e-06, + "loss": 1.0109, + "step": 29015 + }, + { + "epoch": 0.7, + "learning_rate": 4.4194698376473165e-06, + "loss": 1.1075, + "step": 29016 + }, + { + "epoch": 0.7, + "learning_rate": 4.418823523880846e-06, + "loss": 0.9167, + "step": 29017 + }, + { + "epoch": 0.7, + "learning_rate": 4.41817724397406e-06, + "loss": 0.9615, + "step": 29018 + }, + { + "epoch": 0.7, + "learning_rate": 4.4175309979308834e-06, + "loss": 1.1602, + "step": 29019 + }, + { + "epoch": 0.7, + "learning_rate": 4.41688478575523e-06, + "loss": 1.0297, + "step": 29020 + }, + { + "epoch": 0.7, + "learning_rate": 4.4162386074510265e-06, + "loss": 0.8085, + "step": 29021 + }, + { + "epoch": 0.7, + "learning_rate": 4.415592463022187e-06, + "loss": 1.033, + "step": 29022 + }, + { + "epoch": 0.7, + "learning_rate": 4.414946352472634e-06, + "loss": 1.1862, + "step": 29023 + }, + { + "epoch": 0.7, + "learning_rate": 4.414300275806291e-06, + "loss": 1.2883, + "step": 29024 + }, + { + "epoch": 0.7, + "learning_rate": 4.413654233027071e-06, + "loss": 0.9899, + "step": 29025 + }, + { + "epoch": 0.7, + "learning_rate": 4.4130082241388974e-06, + "loss": 1.1636, + "step": 29026 + }, + { + "epoch": 0.7, + "learning_rate": 4.412362249145692e-06, + "loss": 0.9464, + "step": 29027 + }, + { + "epoch": 0.7, + "learning_rate": 4.411716308051366e-06, + "loss": 1.2631, + "step": 29028 + }, + { + "epoch": 0.7, + "learning_rate": 4.411070400859847e-06, + "loss": 1.1023, + "step": 29029 + }, + { + "epoch": 0.7, + "learning_rate": 4.410424527575047e-06, + "loss": 1.0694, + "step": 29030 + }, + { + "epoch": 0.7, + "learning_rate": 4.409778688200885e-06, + "loss": 1.3058, + "step": 29031 + }, + { + "epoch": 0.7, + "learning_rate": 4.409132882741285e-06, + "loss": 0.7484, + "step": 29032 + }, + { + "epoch": 0.7, + "learning_rate": 4.408487111200157e-06, + "loss": 1.0216, + "step": 29033 + }, + { + "epoch": 0.7, + "learning_rate": 4.407841373581422e-06, + "loss": 0.9956, + "step": 29034 + }, + { + "epoch": 0.7, + "learning_rate": 4.407195669889003e-06, + "loss": 1.0348, + "step": 29035 + }, + { + "epoch": 0.7, + "learning_rate": 4.406550000126807e-06, + "loss": 1.0704, + "step": 29036 + }, + { + "epoch": 0.7, + "learning_rate": 4.405904364298758e-06, + "loss": 0.9414, + "step": 29037 + }, + { + "epoch": 0.7, + "learning_rate": 4.405258762408774e-06, + "loss": 0.8904, + "step": 29038 + }, + { + "epoch": 0.7, + "learning_rate": 4.404613194460766e-06, + "loss": 0.6181, + "step": 29039 + }, + { + "epoch": 0.7, + "learning_rate": 4.403967660458657e-06, + "loss": 1.0825, + "step": 29040 + }, + { + "epoch": 0.7, + "learning_rate": 4.403322160406357e-06, + "loss": 1.0459, + "step": 29041 + }, + { + "epoch": 0.7, + "learning_rate": 4.402676694307784e-06, + "loss": 0.89, + "step": 29042 + }, + { + "epoch": 0.7, + "learning_rate": 4.402031262166859e-06, + "loss": 1.0163, + "step": 29043 + }, + { + "epoch": 0.7, + "learning_rate": 4.4013858639874905e-06, + "loss": 0.8707, + "step": 29044 + }, + { + "epoch": 0.7, + "learning_rate": 4.400740499773597e-06, + "loss": 1.1738, + "step": 29045 + }, + { + "epoch": 0.7, + "learning_rate": 4.400095169529098e-06, + "loss": 0.9364, + "step": 29046 + }, + { + "epoch": 0.7, + "learning_rate": 4.399449873257898e-06, + "loss": 0.9517, + "step": 29047 + }, + { + "epoch": 0.7, + "learning_rate": 4.398804610963925e-06, + "loss": 1.0368, + "step": 29048 + }, + { + "epoch": 0.7, + "learning_rate": 4.398159382651082e-06, + "loss": 1.2366, + "step": 29049 + }, + { + "epoch": 0.7, + "learning_rate": 4.397514188323289e-06, + "loss": 1.0105, + "step": 29050 + }, + { + "epoch": 0.7, + "learning_rate": 4.396869027984462e-06, + "loss": 1.1742, + "step": 29051 + }, + { + "epoch": 0.7, + "learning_rate": 4.39622390163851e-06, + "loss": 1.0397, + "step": 29052 + }, + { + "epoch": 0.7, + "learning_rate": 4.395578809289349e-06, + "loss": 1.1481, + "step": 29053 + }, + { + "epoch": 0.7, + "learning_rate": 4.394933750940898e-06, + "loss": 1.0531, + "step": 29054 + }, + { + "epoch": 0.7, + "learning_rate": 4.39428872659706e-06, + "loss": 0.767, + "step": 29055 + }, + { + "epoch": 0.7, + "learning_rate": 4.393643736261759e-06, + "loss": 1.2682, + "step": 29056 + }, + { + "epoch": 0.7, + "learning_rate": 4.392998779938899e-06, + "loss": 0.8623, + "step": 29057 + }, + { + "epoch": 0.7, + "learning_rate": 4.392353857632395e-06, + "loss": 1.0059, + "step": 29058 + }, + { + "epoch": 0.7, + "learning_rate": 4.391708969346167e-06, + "loss": 1.0589, + "step": 29059 + }, + { + "epoch": 0.7, + "learning_rate": 4.391064115084116e-06, + "loss": 1.1041, + "step": 29060 + }, + { + "epoch": 0.7, + "learning_rate": 4.3904192948501604e-06, + "loss": 1.1484, + "step": 29061 + }, + { + "epoch": 0.7, + "learning_rate": 4.389774508648216e-06, + "loss": 0.8328, + "step": 29062 + }, + { + "epoch": 0.7, + "learning_rate": 4.389129756482185e-06, + "loss": 1.0075, + "step": 29063 + }, + { + "epoch": 0.7, + "learning_rate": 4.388485038355988e-06, + "loss": 1.0737, + "step": 29064 + }, + { + "epoch": 0.7, + "learning_rate": 4.387840354273529e-06, + "loss": 1.1292, + "step": 29065 + }, + { + "epoch": 0.7, + "learning_rate": 4.3871957042387225e-06, + "loss": 1.0614, + "step": 29066 + }, + { + "epoch": 0.7, + "learning_rate": 4.386551088255483e-06, + "loss": 1.1812, + "step": 29067 + }, + { + "epoch": 0.7, + "learning_rate": 4.385906506327714e-06, + "loss": 1.2953, + "step": 29068 + }, + { + "epoch": 0.7, + "learning_rate": 4.38526195845933e-06, + "loss": 1.2768, + "step": 29069 + }, + { + "epoch": 0.7, + "learning_rate": 4.3846174446542445e-06, + "loss": 1.2842, + "step": 29070 + }, + { + "epoch": 0.7, + "learning_rate": 4.3839729649163605e-06, + "loss": 1.1499, + "step": 29071 + }, + { + "epoch": 0.7, + "learning_rate": 4.383328519249596e-06, + "loss": 1.0366, + "step": 29072 + }, + { + "epoch": 0.7, + "learning_rate": 4.382684107657852e-06, + "loss": 1.2177, + "step": 29073 + }, + { + "epoch": 0.7, + "learning_rate": 4.382039730145041e-06, + "loss": 1.3757, + "step": 29074 + }, + { + "epoch": 0.7, + "learning_rate": 4.381395386715078e-06, + "loss": 1.0268, + "step": 29075 + }, + { + "epoch": 0.7, + "learning_rate": 4.380751077371865e-06, + "loss": 0.7048, + "step": 29076 + }, + { + "epoch": 0.7, + "learning_rate": 4.380106802119313e-06, + "loss": 0.6905, + "step": 29077 + }, + { + "epoch": 0.7, + "learning_rate": 4.379462560961334e-06, + "loss": 1.0129, + "step": 29078 + }, + { + "epoch": 0.7, + "learning_rate": 4.37881835390183e-06, + "loss": 1.0515, + "step": 29079 + }, + { + "epoch": 0.7, + "learning_rate": 4.378174180944714e-06, + "loss": 1.1428, + "step": 29080 + }, + { + "epoch": 0.7, + "learning_rate": 4.377530042093896e-06, + "loss": 1.0901, + "step": 29081 + }, + { + "epoch": 0.7, + "learning_rate": 4.376885937353277e-06, + "loss": 1.0852, + "step": 29082 + }, + { + "epoch": 0.7, + "learning_rate": 4.376241866726772e-06, + "loss": 0.8538, + "step": 29083 + }, + { + "epoch": 0.7, + "learning_rate": 4.375597830218281e-06, + "loss": 1.0093, + "step": 29084 + }, + { + "epoch": 0.7, + "learning_rate": 4.374953827831715e-06, + "loss": 1.0053, + "step": 29085 + }, + { + "epoch": 0.7, + "learning_rate": 4.374309859570986e-06, + "loss": 0.9998, + "step": 29086 + }, + { + "epoch": 0.7, + "learning_rate": 4.3736659254399905e-06, + "loss": 1.0699, + "step": 29087 + }, + { + "epoch": 0.7, + "learning_rate": 4.37302202544264e-06, + "loss": 1.0371, + "step": 29088 + }, + { + "epoch": 0.7, + "learning_rate": 4.372378159582847e-06, + "loss": 1.2572, + "step": 29089 + }, + { + "epoch": 0.7, + "learning_rate": 4.371734327864508e-06, + "loss": 0.9098, + "step": 29090 + }, + { + "epoch": 0.7, + "learning_rate": 4.371090530291535e-06, + "loss": 1.0651, + "step": 29091 + }, + { + "epoch": 0.7, + "learning_rate": 4.370446766867829e-06, + "loss": 1.1083, + "step": 29092 + }, + { + "epoch": 0.7, + "learning_rate": 4.369803037597299e-06, + "loss": 1.328, + "step": 29093 + }, + { + "epoch": 0.7, + "learning_rate": 4.369159342483852e-06, + "loss": 0.9986, + "step": 29094 + }, + { + "epoch": 0.7, + "learning_rate": 4.368515681531388e-06, + "loss": 1.2027, + "step": 29095 + }, + { + "epoch": 0.7, + "learning_rate": 4.367872054743815e-06, + "loss": 0.9128, + "step": 29096 + }, + { + "epoch": 0.7, + "learning_rate": 4.36722846212504e-06, + "loss": 1.0496, + "step": 29097 + }, + { + "epoch": 0.7, + "learning_rate": 4.366584903678962e-06, + "loss": 0.9334, + "step": 29098 + }, + { + "epoch": 0.7, + "learning_rate": 4.365941379409493e-06, + "loss": 0.9464, + "step": 29099 + }, + { + "epoch": 0.7, + "learning_rate": 4.365297889320528e-06, + "loss": 1.1547, + "step": 29100 + }, + { + "epoch": 0.7, + "learning_rate": 4.364654433415975e-06, + "loss": 0.8253, + "step": 29101 + }, + { + "epoch": 0.7, + "learning_rate": 4.364011011699741e-06, + "loss": 1.1542, + "step": 29102 + }, + { + "epoch": 0.7, + "learning_rate": 4.363367624175724e-06, + "loss": 1.0124, + "step": 29103 + }, + { + "epoch": 0.7, + "learning_rate": 4.36272427084783e-06, + "loss": 1.1374, + "step": 29104 + }, + { + "epoch": 0.7, + "learning_rate": 4.362080951719965e-06, + "loss": 1.0503, + "step": 29105 + }, + { + "epoch": 0.7, + "learning_rate": 4.361437666796026e-06, + "loss": 1.0713, + "step": 29106 + }, + { + "epoch": 0.7, + "learning_rate": 4.360794416079921e-06, + "loss": 1.1513, + "step": 29107 + }, + { + "epoch": 0.7, + "learning_rate": 4.360151199575546e-06, + "loss": 1.186, + "step": 29108 + }, + { + "epoch": 0.7, + "learning_rate": 4.359508017286808e-06, + "loss": 1.0242, + "step": 29109 + }, + { + "epoch": 0.7, + "learning_rate": 4.3588648692176115e-06, + "loss": 0.8716, + "step": 29110 + }, + { + "epoch": 0.7, + "learning_rate": 4.358221755371851e-06, + "loss": 1.1691, + "step": 29111 + }, + { + "epoch": 0.7, + "learning_rate": 4.357578675753432e-06, + "loss": 1.1766, + "step": 29112 + }, + { + "epoch": 0.7, + "learning_rate": 4.35693563036626e-06, + "loss": 1.1273, + "step": 29113 + }, + { + "epoch": 0.7, + "learning_rate": 4.356292619214229e-06, + "loss": 1.1449, + "step": 29114 + }, + { + "epoch": 0.7, + "learning_rate": 4.355649642301246e-06, + "loss": 0.8016, + "step": 29115 + }, + { + "epoch": 0.7, + "learning_rate": 4.355006699631206e-06, + "loss": 0.9482, + "step": 29116 + }, + { + "epoch": 0.7, + "learning_rate": 4.354363791208012e-06, + "loss": 0.9393, + "step": 29117 + }, + { + "epoch": 0.7, + "learning_rate": 4.3537209170355685e-06, + "loss": 1.1563, + "step": 29118 + }, + { + "epoch": 0.7, + "learning_rate": 4.353078077117768e-06, + "loss": 1.1838, + "step": 29119 + }, + { + "epoch": 0.7, + "learning_rate": 4.3524352714585165e-06, + "loss": 0.9992, + "step": 29120 + }, + { + "epoch": 0.7, + "learning_rate": 4.351792500061715e-06, + "loss": 1.071, + "step": 29121 + }, + { + "epoch": 0.7, + "learning_rate": 4.351149762931255e-06, + "loss": 1.0267, + "step": 29122 + }, + { + "epoch": 0.7, + "learning_rate": 4.350507060071044e-06, + "loss": 1.032, + "step": 29123 + }, + { + "epoch": 0.7, + "learning_rate": 4.349864391484976e-06, + "loss": 0.8955, + "step": 29124 + }, + { + "epoch": 0.7, + "learning_rate": 4.34922175717695e-06, + "loss": 0.9336, + "step": 29125 + }, + { + "epoch": 0.7, + "learning_rate": 4.348579157150871e-06, + "loss": 1.4293, + "step": 29126 + }, + { + "epoch": 0.7, + "learning_rate": 4.3479365914106295e-06, + "loss": 1.0519, + "step": 29127 + }, + { + "epoch": 0.7, + "learning_rate": 4.347294059960127e-06, + "loss": 1.039, + "step": 29128 + }, + { + "epoch": 0.7, + "learning_rate": 4.3466515628032656e-06, + "loss": 1.0696, + "step": 29129 + }, + { + "epoch": 0.7, + "learning_rate": 4.346009099943936e-06, + "loss": 1.1437, + "step": 29130 + }, + { + "epoch": 0.7, + "learning_rate": 4.345366671386038e-06, + "loss": 1.0641, + "step": 29131 + }, + { + "epoch": 0.7, + "learning_rate": 4.344724277133475e-06, + "loss": 1.0215, + "step": 29132 + }, + { + "epoch": 0.7, + "learning_rate": 4.344081917190136e-06, + "loss": 1.0569, + "step": 29133 + }, + { + "epoch": 0.7, + "learning_rate": 4.343439591559925e-06, + "loss": 1.0402, + "step": 29134 + }, + { + "epoch": 0.7, + "learning_rate": 4.3427973002467315e-06, + "loss": 0.9662, + "step": 29135 + }, + { + "epoch": 0.7, + "learning_rate": 4.3421550432544565e-06, + "loss": 1.0733, + "step": 29136 + }, + { + "epoch": 0.7, + "learning_rate": 4.341512820587001e-06, + "loss": 1.1125, + "step": 29137 + }, + { + "epoch": 0.7, + "learning_rate": 4.3408706322482505e-06, + "loss": 1.053, + "step": 29138 + }, + { + "epoch": 0.7, + "learning_rate": 4.340228478242108e-06, + "loss": 1.1556, + "step": 29139 + }, + { + "epoch": 0.7, + "learning_rate": 4.339586358572472e-06, + "loss": 0.8324, + "step": 29140 + }, + { + "epoch": 0.7, + "learning_rate": 4.33894427324323e-06, + "loss": 0.9191, + "step": 29141 + }, + { + "epoch": 0.7, + "learning_rate": 4.338302222258285e-06, + "loss": 0.9474, + "step": 29142 + }, + { + "epoch": 0.7, + "learning_rate": 4.337660205621524e-06, + "loss": 1.0558, + "step": 29143 + }, + { + "epoch": 0.7, + "learning_rate": 4.337018223336849e-06, + "loss": 1.3662, + "step": 29144 + }, + { + "epoch": 0.7, + "learning_rate": 4.336376275408155e-06, + "loss": 1.0439, + "step": 29145 + }, + { + "epoch": 0.7, + "learning_rate": 4.33573436183933e-06, + "loss": 1.1075, + "step": 29146 + }, + { + "epoch": 0.7, + "learning_rate": 4.3350924826342725e-06, + "loss": 1.181, + "step": 29147 + }, + { + "epoch": 0.7, + "learning_rate": 4.3344506377968794e-06, + "loss": 1.0455, + "step": 29148 + }, + { + "epoch": 0.7, + "learning_rate": 4.33380882733104e-06, + "loss": 0.882, + "step": 29149 + }, + { + "epoch": 0.7, + "learning_rate": 4.333167051240652e-06, + "loss": 1.1347, + "step": 29150 + }, + { + "epoch": 0.7, + "learning_rate": 4.332525309529603e-06, + "loss": 1.1108, + "step": 29151 + }, + { + "epoch": 0.7, + "learning_rate": 4.33188360220179e-06, + "loss": 0.9174, + "step": 29152 + }, + { + "epoch": 0.7, + "learning_rate": 4.331241929261108e-06, + "loss": 1.0594, + "step": 29153 + }, + { + "epoch": 0.7, + "learning_rate": 4.330600290711446e-06, + "loss": 0.9458, + "step": 29154 + }, + { + "epoch": 0.7, + "learning_rate": 4.329958686556699e-06, + "loss": 0.887, + "step": 29155 + }, + { + "epoch": 0.7, + "learning_rate": 4.329317116800762e-06, + "loss": 1.1468, + "step": 29156 + }, + { + "epoch": 0.7, + "learning_rate": 4.32867558144752e-06, + "loss": 1.1486, + "step": 29157 + }, + { + "epoch": 0.7, + "learning_rate": 4.328034080500873e-06, + "loss": 0.9992, + "step": 29158 + }, + { + "epoch": 0.7, + "learning_rate": 4.327392613964706e-06, + "loss": 1.1113, + "step": 29159 + }, + { + "epoch": 0.7, + "learning_rate": 4.3267511818429135e-06, + "loss": 1.0585, + "step": 29160 + }, + { + "epoch": 0.7, + "learning_rate": 4.326109784139391e-06, + "loss": 1.1341, + "step": 29161 + }, + { + "epoch": 0.7, + "learning_rate": 4.325468420858022e-06, + "loss": 1.0129, + "step": 29162 + }, + { + "epoch": 0.7, + "learning_rate": 4.324827092002701e-06, + "loss": 1.0191, + "step": 29163 + }, + { + "epoch": 0.7, + "learning_rate": 4.324185797577324e-06, + "loss": 1.061, + "step": 29164 + }, + { + "epoch": 0.7, + "learning_rate": 4.323544537585772e-06, + "loss": 0.8916, + "step": 29165 + }, + { + "epoch": 0.7, + "learning_rate": 4.32290331203194e-06, + "loss": 1.0709, + "step": 29166 + }, + { + "epoch": 0.7, + "learning_rate": 4.322262120919719e-06, + "loss": 0.9902, + "step": 29167 + }, + { + "epoch": 0.7, + "learning_rate": 4.321620964252997e-06, + "loss": 1.0638, + "step": 29168 + }, + { + "epoch": 0.7, + "learning_rate": 4.3209798420356696e-06, + "loss": 0.9805, + "step": 29169 + }, + { + "epoch": 0.7, + "learning_rate": 4.320338754271618e-06, + "loss": 1.1318, + "step": 29170 + }, + { + "epoch": 0.7, + "learning_rate": 4.319697700964739e-06, + "loss": 0.9154, + "step": 29171 + }, + { + "epoch": 0.7, + "learning_rate": 4.319056682118914e-06, + "loss": 1.0416, + "step": 29172 + }, + { + "epoch": 0.7, + "learning_rate": 4.3184156977380364e-06, + "loss": 1.0758, + "step": 29173 + }, + { + "epoch": 0.7, + "learning_rate": 4.317774747825998e-06, + "loss": 1.0729, + "step": 29174 + }, + { + "epoch": 0.7, + "learning_rate": 4.3171338323866795e-06, + "loss": 0.9819, + "step": 29175 + }, + { + "epoch": 0.7, + "learning_rate": 4.316492951423973e-06, + "loss": 1.2048, + "step": 29176 + }, + { + "epoch": 0.7, + "learning_rate": 4.315852104941771e-06, + "loss": 1.0539, + "step": 29177 + }, + { + "epoch": 0.7, + "learning_rate": 4.3152112929439535e-06, + "loss": 1.0191, + "step": 29178 + }, + { + "epoch": 0.7, + "learning_rate": 4.314570515434416e-06, + "loss": 0.9713, + "step": 29179 + }, + { + "epoch": 0.7, + "learning_rate": 4.313929772417038e-06, + "loss": 1.1036, + "step": 29180 + }, + { + "epoch": 0.7, + "learning_rate": 4.313289063895712e-06, + "loss": 1.0127, + "step": 29181 + }, + { + "epoch": 0.7, + "learning_rate": 4.312648389874325e-06, + "loss": 1.0907, + "step": 29182 + }, + { + "epoch": 0.7, + "learning_rate": 4.31200775035676e-06, + "loss": 1.1161, + "step": 29183 + }, + { + "epoch": 0.7, + "learning_rate": 4.311367145346905e-06, + "loss": 1.093, + "step": 29184 + }, + { + "epoch": 0.7, + "learning_rate": 4.310726574848653e-06, + "loss": 1.1714, + "step": 29185 + }, + { + "epoch": 0.7, + "learning_rate": 4.3100860388658794e-06, + "loss": 0.9136, + "step": 29186 + }, + { + "epoch": 0.7, + "learning_rate": 4.3094455374024805e-06, + "loss": 1.1392, + "step": 29187 + }, + { + "epoch": 0.7, + "learning_rate": 4.308805070462332e-06, + "loss": 1.0201, + "step": 29188 + }, + { + "epoch": 0.7, + "learning_rate": 4.308164638049325e-06, + "loss": 0.8043, + "step": 29189 + }, + { + "epoch": 0.7, + "learning_rate": 4.307524240167348e-06, + "loss": 0.8548, + "step": 29190 + }, + { + "epoch": 0.7, + "learning_rate": 4.306883876820279e-06, + "loss": 0.9742, + "step": 29191 + }, + { + "epoch": 0.7, + "learning_rate": 4.306243548012007e-06, + "loss": 0.9658, + "step": 29192 + }, + { + "epoch": 0.7, + "learning_rate": 4.305603253746419e-06, + "loss": 1.1773, + "step": 29193 + }, + { + "epoch": 0.7, + "learning_rate": 4.304962994027393e-06, + "loss": 1.0218, + "step": 29194 + }, + { + "epoch": 0.7, + "learning_rate": 4.304322768858821e-06, + "loss": 1.1415, + "step": 29195 + }, + { + "epoch": 0.7, + "learning_rate": 4.303682578244579e-06, + "loss": 1.0678, + "step": 29196 + }, + { + "epoch": 0.7, + "learning_rate": 4.303042422188555e-06, + "loss": 0.9728, + "step": 29197 + }, + { + "epoch": 0.7, + "learning_rate": 4.302402300694636e-06, + "loss": 0.9318, + "step": 29198 + }, + { + "epoch": 0.7, + "learning_rate": 4.3017622137667e-06, + "loss": 1.1998, + "step": 29199 + }, + { + "epoch": 0.7, + "learning_rate": 4.3011221614086295e-06, + "loss": 0.9694, + "step": 29200 + }, + { + "epoch": 0.7, + "learning_rate": 4.300482143624316e-06, + "loss": 1.1294, + "step": 29201 + }, + { + "epoch": 0.7, + "learning_rate": 4.299842160417633e-06, + "loss": 1.0727, + "step": 29202 + }, + { + "epoch": 0.7, + "learning_rate": 4.29920221179247e-06, + "loss": 1.1116, + "step": 29203 + }, + { + "epoch": 0.7, + "learning_rate": 4.298562297752703e-06, + "loss": 1.1863, + "step": 29204 + }, + { + "epoch": 0.7, + "learning_rate": 4.297922418302217e-06, + "loss": 0.9779, + "step": 29205 + }, + { + "epoch": 0.7, + "learning_rate": 4.297282573444899e-06, + "loss": 0.9703, + "step": 29206 + }, + { + "epoch": 0.7, + "learning_rate": 4.296642763184622e-06, + "loss": 1.2168, + "step": 29207 + }, + { + "epoch": 0.7, + "learning_rate": 4.296002987525272e-06, + "loss": 1.0206, + "step": 29208 + }, + { + "epoch": 0.7, + "learning_rate": 4.295363246470734e-06, + "loss": 0.9091, + "step": 29209 + }, + { + "epoch": 0.7, + "learning_rate": 4.294723540024881e-06, + "loss": 1.0226, + "step": 29210 + }, + { + "epoch": 0.7, + "learning_rate": 4.2940838681915985e-06, + "loss": 1.0948, + "step": 29211 + }, + { + "epoch": 0.7, + "learning_rate": 4.2934442309747714e-06, + "loss": 1.0785, + "step": 29212 + }, + { + "epoch": 0.7, + "learning_rate": 4.292804628378272e-06, + "loss": 0.8887, + "step": 29213 + }, + { + "epoch": 0.7, + "learning_rate": 4.292165060405988e-06, + "loss": 1.0392, + "step": 29214 + }, + { + "epoch": 0.7, + "learning_rate": 4.291525527061793e-06, + "loss": 1.2081, + "step": 29215 + }, + { + "epoch": 0.7, + "learning_rate": 4.290886028349569e-06, + "loss": 1.0505, + "step": 29216 + }, + { + "epoch": 0.7, + "learning_rate": 4.2902465642732e-06, + "loss": 0.8726, + "step": 29217 + }, + { + "epoch": 0.7, + "learning_rate": 4.28960713483656e-06, + "loss": 1.0188, + "step": 29218 + }, + { + "epoch": 0.7, + "learning_rate": 4.288967740043528e-06, + "loss": 1.455, + "step": 29219 + }, + { + "epoch": 0.7, + "learning_rate": 4.288328379897991e-06, + "loss": 0.9765, + "step": 29220 + }, + { + "epoch": 0.7, + "learning_rate": 4.287689054403818e-06, + "loss": 1.0691, + "step": 29221 + }, + { + "epoch": 0.7, + "learning_rate": 4.2870497635648956e-06, + "loss": 1.1017, + "step": 29222 + }, + { + "epoch": 0.7, + "learning_rate": 4.2864105073850945e-06, + "loss": 0.7247, + "step": 29223 + }, + { + "epoch": 0.7, + "learning_rate": 4.285771285868297e-06, + "loss": 1.2133, + "step": 29224 + }, + { + "epoch": 0.7, + "learning_rate": 4.285132099018385e-06, + "loss": 0.9913, + "step": 29225 + }, + { + "epoch": 0.7, + "learning_rate": 4.284492946839229e-06, + "loss": 1.1426, + "step": 29226 + }, + { + "epoch": 0.7, + "learning_rate": 4.28385382933471e-06, + "loss": 1.3195, + "step": 29227 + }, + { + "epoch": 0.7, + "learning_rate": 4.283214746508708e-06, + "loss": 0.9178, + "step": 29228 + }, + { + "epoch": 0.7, + "learning_rate": 4.282575698365095e-06, + "loss": 1.0898, + "step": 29229 + }, + { + "epoch": 0.7, + "learning_rate": 4.281936684907753e-06, + "loss": 0.8986, + "step": 29230 + }, + { + "epoch": 0.7, + "learning_rate": 4.281297706140552e-06, + "loss": 1.1881, + "step": 29231 + }, + { + "epoch": 0.7, + "learning_rate": 4.280658762067374e-06, + "loss": 1.2617, + "step": 29232 + }, + { + "epoch": 0.7, + "learning_rate": 4.280019852692098e-06, + "loss": 0.9726, + "step": 29233 + }, + { + "epoch": 0.7, + "learning_rate": 4.279380978018591e-06, + "loss": 0.9041, + "step": 29234 + }, + { + "epoch": 0.7, + "learning_rate": 4.278742138050734e-06, + "loss": 0.9619, + "step": 29235 + }, + { + "epoch": 0.7, + "learning_rate": 4.278103332792408e-06, + "loss": 1.0168, + "step": 29236 + }, + { + "epoch": 0.7, + "learning_rate": 4.2774645622474786e-06, + "loss": 1.2703, + "step": 29237 + }, + { + "epoch": 0.7, + "learning_rate": 4.276825826419829e-06, + "loss": 1.0961, + "step": 29238 + }, + { + "epoch": 0.7, + "learning_rate": 4.276187125313328e-06, + "loss": 1.2538, + "step": 29239 + }, + { + "epoch": 0.7, + "learning_rate": 4.275548458931853e-06, + "loss": 0.9039, + "step": 29240 + }, + { + "epoch": 0.7, + "learning_rate": 4.274909827279283e-06, + "loss": 1.0617, + "step": 29241 + }, + { + "epoch": 0.7, + "learning_rate": 4.274271230359485e-06, + "loss": 1.0719, + "step": 29242 + }, + { + "epoch": 0.7, + "learning_rate": 4.273632668176335e-06, + "loss": 0.7619, + "step": 29243 + }, + { + "epoch": 0.7, + "learning_rate": 4.272994140733714e-06, + "loss": 1.0669, + "step": 29244 + }, + { + "epoch": 0.7, + "learning_rate": 4.2723556480354865e-06, + "loss": 1.0759, + "step": 29245 + }, + { + "epoch": 0.7, + "learning_rate": 4.271717190085533e-06, + "loss": 0.8834, + "step": 29246 + }, + { + "epoch": 0.7, + "learning_rate": 4.271078766887722e-06, + "loss": 0.9692, + "step": 29247 + }, + { + "epoch": 0.7, + "learning_rate": 4.270440378445927e-06, + "loss": 0.9148, + "step": 29248 + }, + { + "epoch": 0.7, + "learning_rate": 4.269802024764027e-06, + "loss": 0.8943, + "step": 29249 + }, + { + "epoch": 0.7, + "learning_rate": 4.269163705845886e-06, + "loss": 1.1739, + "step": 29250 + }, + { + "epoch": 0.7, + "learning_rate": 4.268525421695381e-06, + "loss": 1.1621, + "step": 29251 + }, + { + "epoch": 0.7, + "learning_rate": 4.267887172316389e-06, + "loss": 0.7027, + "step": 29252 + }, + { + "epoch": 0.7, + "learning_rate": 4.2672489577127715e-06, + "loss": 1.1106, + "step": 29253 + }, + { + "epoch": 0.7, + "learning_rate": 4.266610777888407e-06, + "loss": 1.0497, + "step": 29254 + }, + { + "epoch": 0.7, + "learning_rate": 4.26597263284717e-06, + "loss": 0.8144, + "step": 29255 + }, + { + "epoch": 0.7, + "learning_rate": 4.265334522592926e-06, + "loss": 1.0473, + "step": 29256 + }, + { + "epoch": 0.7, + "learning_rate": 4.264696447129551e-06, + "loss": 0.8075, + "step": 29257 + }, + { + "epoch": 0.7, + "learning_rate": 4.264058406460909e-06, + "loss": 1.1336, + "step": 29258 + }, + { + "epoch": 0.7, + "learning_rate": 4.2634204005908775e-06, + "loss": 1.1072, + "step": 29259 + }, + { + "epoch": 0.7, + "learning_rate": 4.262782429523328e-06, + "loss": 1.2304, + "step": 29260 + }, + { + "epoch": 0.7, + "learning_rate": 4.262144493262123e-06, + "loss": 0.9196, + "step": 29261 + }, + { + "epoch": 0.7, + "learning_rate": 4.2615065918111396e-06, + "loss": 1.091, + "step": 29262 + }, + { + "epoch": 0.7, + "learning_rate": 4.260868725174248e-06, + "loss": 0.9645, + "step": 29263 + }, + { + "epoch": 0.7, + "learning_rate": 4.260230893355314e-06, + "loss": 1.1785, + "step": 29264 + }, + { + "epoch": 0.7, + "learning_rate": 4.259593096358211e-06, + "loss": 0.843, + "step": 29265 + }, + { + "epoch": 0.7, + "learning_rate": 4.258955334186804e-06, + "loss": 1.0692, + "step": 29266 + }, + { + "epoch": 0.7, + "learning_rate": 4.2583176068449646e-06, + "loss": 1.2907, + "step": 29267 + }, + { + "epoch": 0.7, + "learning_rate": 4.257679914336566e-06, + "loss": 1.1639, + "step": 29268 + }, + { + "epoch": 0.7, + "learning_rate": 4.257042256665466e-06, + "loss": 1.2143, + "step": 29269 + }, + { + "epoch": 0.7, + "learning_rate": 4.2564046338355436e-06, + "loss": 0.9468, + "step": 29270 + }, + { + "epoch": 0.7, + "learning_rate": 4.255767045850665e-06, + "loss": 0.8124, + "step": 29271 + }, + { + "epoch": 0.7, + "learning_rate": 4.2551294927146925e-06, + "loss": 0.7447, + "step": 29272 + }, + { + "epoch": 0.7, + "learning_rate": 4.254491974431503e-06, + "loss": 1.0363, + "step": 29273 + }, + { + "epoch": 0.7, + "learning_rate": 4.253854491004955e-06, + "loss": 0.8905, + "step": 29274 + }, + { + "epoch": 0.7, + "learning_rate": 4.253217042438921e-06, + "loss": 1.0369, + "step": 29275 + }, + { + "epoch": 0.7, + "learning_rate": 4.25257962873727e-06, + "loss": 0.8866, + "step": 29276 + }, + { + "epoch": 0.7, + "learning_rate": 4.251942249903863e-06, + "loss": 0.836, + "step": 29277 + }, + { + "epoch": 0.7, + "learning_rate": 4.25130490594257e-06, + "loss": 1.1144, + "step": 29278 + }, + { + "epoch": 0.7, + "learning_rate": 4.250667596857263e-06, + "loss": 0.9898, + "step": 29279 + }, + { + "epoch": 0.7, + "learning_rate": 4.250030322651799e-06, + "loss": 0.9253, + "step": 29280 + }, + { + "epoch": 0.7, + "learning_rate": 4.249393083330052e-06, + "loss": 1.0157, + "step": 29281 + }, + { + "epoch": 0.7, + "learning_rate": 4.2487558788958815e-06, + "loss": 0.9274, + "step": 29282 + }, + { + "epoch": 0.7, + "learning_rate": 4.248118709353156e-06, + "loss": 1.0468, + "step": 29283 + }, + { + "epoch": 0.7, + "learning_rate": 4.247481574705744e-06, + "loss": 1.2216, + "step": 29284 + }, + { + "epoch": 0.7, + "learning_rate": 4.2468444749575054e-06, + "loss": 1.1041, + "step": 29285 + }, + { + "epoch": 0.7, + "learning_rate": 4.246207410112309e-06, + "loss": 0.9553, + "step": 29286 + }, + { + "epoch": 0.7, + "learning_rate": 4.245570380174021e-06, + "loss": 0.903, + "step": 29287 + }, + { + "epoch": 0.7, + "learning_rate": 4.244933385146501e-06, + "loss": 1.2085, + "step": 29288 + }, + { + "epoch": 0.7, + "learning_rate": 4.244296425033621e-06, + "loss": 1.1222, + "step": 29289 + }, + { + "epoch": 0.7, + "learning_rate": 4.243659499839236e-06, + "loss": 1.0357, + "step": 29290 + }, + { + "epoch": 0.7, + "learning_rate": 4.243022609567216e-06, + "loss": 1.1722, + "step": 29291 + }, + { + "epoch": 0.7, + "learning_rate": 4.2423857542214254e-06, + "loss": 0.9717, + "step": 29292 + }, + { + "epoch": 0.7, + "learning_rate": 4.241748933805724e-06, + "loss": 1.2287, + "step": 29293 + }, + { + "epoch": 0.7, + "learning_rate": 4.241112148323977e-06, + "loss": 1.0889, + "step": 29294 + }, + { + "epoch": 0.7, + "learning_rate": 4.240475397780052e-06, + "loss": 1.1042, + "step": 29295 + }, + { + "epoch": 0.7, + "learning_rate": 4.239838682177804e-06, + "loss": 1.0375, + "step": 29296 + }, + { + "epoch": 0.7, + "learning_rate": 4.2392020015211e-06, + "loss": 1.1008, + "step": 29297 + }, + { + "epoch": 0.7, + "learning_rate": 4.238565355813805e-06, + "loss": 1.0435, + "step": 29298 + }, + { + "epoch": 0.7, + "learning_rate": 4.237928745059777e-06, + "loss": 1.1473, + "step": 29299 + }, + { + "epoch": 0.7, + "learning_rate": 4.2372921692628825e-06, + "loss": 0.939, + "step": 29300 + }, + { + "epoch": 0.7, + "learning_rate": 4.2366556284269775e-06, + "loss": 1.1512, + "step": 29301 + }, + { + "epoch": 0.7, + "learning_rate": 4.2360191225559275e-06, + "loss": 1.1402, + "step": 29302 + }, + { + "epoch": 0.7, + "learning_rate": 4.235382651653597e-06, + "loss": 1.0104, + "step": 29303 + }, + { + "epoch": 0.7, + "learning_rate": 4.23474621572384e-06, + "loss": 1.0736, + "step": 29304 + }, + { + "epoch": 0.7, + "learning_rate": 4.234109814770523e-06, + "loss": 1.0598, + "step": 29305 + }, + { + "epoch": 0.7, + "learning_rate": 4.233473448797508e-06, + "loss": 1.1899, + "step": 29306 + }, + { + "epoch": 0.7, + "learning_rate": 4.23283711780865e-06, + "loss": 1.0141, + "step": 29307 + }, + { + "epoch": 0.7, + "learning_rate": 4.2322008218078156e-06, + "loss": 1.0229, + "step": 29308 + }, + { + "epoch": 0.7, + "learning_rate": 4.23156456079886e-06, + "loss": 1.1626, + "step": 29309 + }, + { + "epoch": 0.7, + "learning_rate": 4.230928334785644e-06, + "loss": 1.0971, + "step": 29310 + }, + { + "epoch": 0.7, + "learning_rate": 4.230292143772033e-06, + "loss": 0.9963, + "step": 29311 + }, + { + "epoch": 0.7, + "learning_rate": 4.229655987761879e-06, + "loss": 0.9463, + "step": 29312 + }, + { + "epoch": 0.7, + "learning_rate": 4.229019866759044e-06, + "loss": 1.1027, + "step": 29313 + }, + { + "epoch": 0.7, + "learning_rate": 4.2283837807673924e-06, + "loss": 1.0051, + "step": 29314 + }, + { + "epoch": 0.7, + "learning_rate": 4.2277477297907756e-06, + "loss": 1.1729, + "step": 29315 + }, + { + "epoch": 0.71, + "learning_rate": 4.227111713833059e-06, + "loss": 1.0877, + "step": 29316 + }, + { + "epoch": 0.71, + "learning_rate": 4.2264757328980935e-06, + "loss": 1.2338, + "step": 29317 + }, + { + "epoch": 0.71, + "learning_rate": 4.2258397869897414e-06, + "loss": 0.836, + "step": 29318 + }, + { + "epoch": 0.71, + "learning_rate": 4.225203876111867e-06, + "loss": 1.1358, + "step": 29319 + }, + { + "epoch": 0.71, + "learning_rate": 4.2245680002683165e-06, + "loss": 1.0123, + "step": 29320 + }, + { + "epoch": 0.71, + "learning_rate": 4.223932159462955e-06, + "loss": 0.9332, + "step": 29321 + }, + { + "epoch": 0.71, + "learning_rate": 4.22329635369964e-06, + "loss": 1.1567, + "step": 29322 + }, + { + "epoch": 0.71, + "learning_rate": 4.222660582982225e-06, + "loss": 1.0653, + "step": 29323 + }, + { + "epoch": 0.71, + "learning_rate": 4.222024847314573e-06, + "loss": 1.0192, + "step": 29324 + }, + { + "epoch": 0.71, + "learning_rate": 4.2213891467005345e-06, + "loss": 0.9359, + "step": 29325 + }, + { + "epoch": 0.71, + "learning_rate": 4.220753481143966e-06, + "loss": 0.9076, + "step": 29326 + }, + { + "epoch": 0.71, + "learning_rate": 4.220117850648733e-06, + "loss": 1.1716, + "step": 29327 + }, + { + "epoch": 0.71, + "learning_rate": 4.219482255218681e-06, + "loss": 1.1181, + "step": 29328 + }, + { + "epoch": 0.71, + "learning_rate": 4.218846694857671e-06, + "loss": 1.0567, + "step": 29329 + }, + { + "epoch": 0.71, + "learning_rate": 4.218211169569562e-06, + "loss": 0.9599, + "step": 29330 + }, + { + "epoch": 0.71, + "learning_rate": 4.217575679358201e-06, + "loss": 1.2049, + "step": 29331 + }, + { + "epoch": 0.71, + "learning_rate": 4.216940224227449e-06, + "loss": 1.2187, + "step": 29332 + }, + { + "epoch": 0.71, + "learning_rate": 4.216304804181161e-06, + "loss": 1.1631, + "step": 29333 + }, + { + "epoch": 0.71, + "learning_rate": 4.215669419223189e-06, + "loss": 0.8959, + "step": 29334 + }, + { + "epoch": 0.71, + "learning_rate": 4.215034069357396e-06, + "loss": 1.0689, + "step": 29335 + }, + { + "epoch": 0.71, + "learning_rate": 4.214398754587627e-06, + "loss": 0.9176, + "step": 29336 + }, + { + "epoch": 0.71, + "learning_rate": 4.213763474917743e-06, + "loss": 1.1009, + "step": 29337 + }, + { + "epoch": 0.71, + "learning_rate": 4.213128230351591e-06, + "loss": 0.9327, + "step": 29338 + }, + { + "epoch": 0.71, + "learning_rate": 4.212493020893029e-06, + "loss": 1.2945, + "step": 29339 + }, + { + "epoch": 0.71, + "learning_rate": 4.211857846545914e-06, + "loss": 1.023, + "step": 29340 + }, + { + "epoch": 0.71, + "learning_rate": 4.211222707314093e-06, + "loss": 0.9886, + "step": 29341 + }, + { + "epoch": 0.71, + "learning_rate": 4.210587603201421e-06, + "loss": 0.999, + "step": 29342 + }, + { + "epoch": 0.71, + "learning_rate": 4.209952534211758e-06, + "loss": 0.8712, + "step": 29343 + }, + { + "epoch": 0.71, + "learning_rate": 4.209317500348946e-06, + "loss": 0.9049, + "step": 29344 + }, + { + "epoch": 0.71, + "learning_rate": 4.208682501616847e-06, + "loss": 1.0744, + "step": 29345 + }, + { + "epoch": 0.71, + "learning_rate": 4.208047538019304e-06, + "loss": 0.8985, + "step": 29346 + }, + { + "epoch": 0.71, + "learning_rate": 4.207412609560175e-06, + "loss": 1.1695, + "step": 29347 + }, + { + "epoch": 0.71, + "learning_rate": 4.2067777162433155e-06, + "loss": 1.2433, + "step": 29348 + }, + { + "epoch": 0.71, + "learning_rate": 4.206142858072568e-06, + "loss": 0.9655, + "step": 29349 + }, + { + "epoch": 0.71, + "learning_rate": 4.20550803505179e-06, + "loss": 1.0493, + "step": 29350 + }, + { + "epoch": 0.71, + "learning_rate": 4.204873247184836e-06, + "loss": 1.0683, + "step": 29351 + }, + { + "epoch": 0.71, + "learning_rate": 4.2042384944755484e-06, + "loss": 1.0006, + "step": 29352 + }, + { + "epoch": 0.71, + "learning_rate": 4.203603776927786e-06, + "loss": 1.0827, + "step": 29353 + }, + { + "epoch": 0.71, + "learning_rate": 4.202969094545393e-06, + "loss": 0.9562, + "step": 29354 + }, + { + "epoch": 0.71, + "learning_rate": 4.202334447332222e-06, + "loss": 1.0358, + "step": 29355 + }, + { + "epoch": 0.71, + "learning_rate": 4.20169983529213e-06, + "loss": 0.9837, + "step": 29356 + }, + { + "epoch": 0.71, + "learning_rate": 4.2010652584289556e-06, + "loss": 0.9269, + "step": 29357 + }, + { + "epoch": 0.71, + "learning_rate": 4.200430716746555e-06, + "loss": 0.9481, + "step": 29358 + }, + { + "epoch": 0.71, + "learning_rate": 4.199796210248781e-06, + "loss": 1.2027, + "step": 29359 + }, + { + "epoch": 0.71, + "learning_rate": 4.199161738939474e-06, + "loss": 1.0801, + "step": 29360 + }, + { + "epoch": 0.71, + "learning_rate": 4.198527302822493e-06, + "loss": 1.1965, + "step": 29361 + }, + { + "epoch": 0.71, + "learning_rate": 4.19789290190168e-06, + "loss": 1.0973, + "step": 29362 + }, + { + "epoch": 0.71, + "learning_rate": 4.197258536180885e-06, + "loss": 1.0695, + "step": 29363 + }, + { + "epoch": 0.71, + "learning_rate": 4.1966242056639605e-06, + "loss": 0.9722, + "step": 29364 + }, + { + "epoch": 0.71, + "learning_rate": 4.195989910354749e-06, + "loss": 1.1789, + "step": 29365 + }, + { + "epoch": 0.71, + "learning_rate": 4.195355650257101e-06, + "loss": 0.9802, + "step": 29366 + }, + { + "epoch": 0.71, + "learning_rate": 4.194721425374869e-06, + "loss": 1.01, + "step": 29367 + }, + { + "epoch": 0.71, + "learning_rate": 4.194087235711892e-06, + "loss": 0.9672, + "step": 29368 + }, + { + "epoch": 0.71, + "learning_rate": 4.1934530812720264e-06, + "loss": 0.9985, + "step": 29369 + }, + { + "epoch": 0.71, + "learning_rate": 4.192818962059112e-06, + "loss": 1.2823, + "step": 29370 + }, + { + "epoch": 0.71, + "learning_rate": 4.192184878076999e-06, + "loss": 1.1351, + "step": 29371 + }, + { + "epoch": 0.71, + "learning_rate": 4.191550829329537e-06, + "loss": 1.0871, + "step": 29372 + }, + { + "epoch": 0.71, + "learning_rate": 4.190916815820567e-06, + "loss": 1.0612, + "step": 29373 + }, + { + "epoch": 0.71, + "learning_rate": 4.190282837553938e-06, + "loss": 1.0371, + "step": 29374 + }, + { + "epoch": 0.71, + "learning_rate": 4.189648894533501e-06, + "loss": 0.67, + "step": 29375 + }, + { + "epoch": 0.71, + "learning_rate": 4.189014986763093e-06, + "loss": 1.1642, + "step": 29376 + }, + { + "epoch": 0.71, + "learning_rate": 4.188381114246565e-06, + "loss": 0.9185, + "step": 29377 + }, + { + "epoch": 0.71, + "learning_rate": 4.187747276987764e-06, + "loss": 1.0421, + "step": 29378 + }, + { + "epoch": 0.71, + "learning_rate": 4.187113474990531e-06, + "loss": 1.2851, + "step": 29379 + }, + { + "epoch": 0.71, + "learning_rate": 4.186479708258716e-06, + "loss": 1.2223, + "step": 29380 + }, + { + "epoch": 0.71, + "learning_rate": 4.1858459767961565e-06, + "loss": 0.9419, + "step": 29381 + }, + { + "epoch": 0.71, + "learning_rate": 4.185212280606704e-06, + "loss": 1.0362, + "step": 29382 + }, + { + "epoch": 0.71, + "learning_rate": 4.184578619694204e-06, + "loss": 0.6743, + "step": 29383 + }, + { + "epoch": 0.71, + "learning_rate": 4.1839449940624935e-06, + "loss": 1.0423, + "step": 29384 + }, + { + "epoch": 0.71, + "learning_rate": 4.183311403715421e-06, + "loss": 1.2326, + "step": 29385 + }, + { + "epoch": 0.71, + "learning_rate": 4.182677848656833e-06, + "loss": 0.7135, + "step": 29386 + }, + { + "epoch": 0.71, + "learning_rate": 4.182044328890568e-06, + "loss": 0.9015, + "step": 29387 + }, + { + "epoch": 0.71, + "learning_rate": 4.181410844420473e-06, + "loss": 1.1172, + "step": 29388 + }, + { + "epoch": 0.71, + "learning_rate": 4.180777395250388e-06, + "loss": 1.2442, + "step": 29389 + }, + { + "epoch": 0.71, + "learning_rate": 4.180143981384157e-06, + "loss": 1.2738, + "step": 29390 + }, + { + "epoch": 0.71, + "learning_rate": 4.179510602825628e-06, + "loss": 0.9197, + "step": 29391 + }, + { + "epoch": 0.71, + "learning_rate": 4.178877259578634e-06, + "loss": 0.9882, + "step": 29392 + }, + { + "epoch": 0.71, + "learning_rate": 4.178243951647024e-06, + "loss": 1.0784, + "step": 29393 + }, + { + "epoch": 0.71, + "learning_rate": 4.177610679034642e-06, + "loss": 1.1921, + "step": 29394 + }, + { + "epoch": 0.71, + "learning_rate": 4.176977441745323e-06, + "loss": 1.1804, + "step": 29395 + }, + { + "epoch": 0.71, + "learning_rate": 4.1763442397829145e-06, + "loss": 1.084, + "step": 29396 + }, + { + "epoch": 0.71, + "learning_rate": 4.1757110731512525e-06, + "loss": 1.0486, + "step": 29397 + }, + { + "epoch": 0.71, + "learning_rate": 4.175077941854182e-06, + "loss": 1.1761, + "step": 29398 + }, + { + "epoch": 0.71, + "learning_rate": 4.174444845895547e-06, + "loss": 1.0467, + "step": 29399 + }, + { + "epoch": 0.71, + "learning_rate": 4.173811785279182e-06, + "loss": 1.1419, + "step": 29400 + }, + { + "epoch": 0.71, + "learning_rate": 4.173178760008929e-06, + "loss": 0.7437, + "step": 29401 + }, + { + "epoch": 0.71, + "learning_rate": 4.1725457700886345e-06, + "loss": 1.159, + "step": 29402 + }, + { + "epoch": 0.71, + "learning_rate": 4.17191281552213e-06, + "loss": 1.0246, + "step": 29403 + }, + { + "epoch": 0.71, + "learning_rate": 4.171279896313264e-06, + "loss": 1.0245, + "step": 29404 + }, + { + "epoch": 0.71, + "learning_rate": 4.1706470124658675e-06, + "loss": 0.9977, + "step": 29405 + }, + { + "epoch": 0.71, + "learning_rate": 4.170014163983784e-06, + "loss": 1.2688, + "step": 29406 + }, + { + "epoch": 0.71, + "learning_rate": 4.169381350870858e-06, + "loss": 1.0098, + "step": 29407 + }, + { + "epoch": 0.71, + "learning_rate": 4.1687485731309194e-06, + "loss": 1.0906, + "step": 29408 + }, + { + "epoch": 0.71, + "learning_rate": 4.168115830767812e-06, + "loss": 0.888, + "step": 29409 + }, + { + "epoch": 0.71, + "learning_rate": 4.167483123785378e-06, + "loss": 0.6483, + "step": 29410 + }, + { + "epoch": 0.71, + "learning_rate": 4.166850452187448e-06, + "loss": 1.3262, + "step": 29411 + }, + { + "epoch": 0.71, + "learning_rate": 4.166217815977867e-06, + "loss": 0.8629, + "step": 29412 + }, + { + "epoch": 0.71, + "learning_rate": 4.1655852151604685e-06, + "loss": 1.0245, + "step": 29413 + }, + { + "epoch": 0.71, + "learning_rate": 4.164952649739091e-06, + "loss": 1.2328, + "step": 29414 + }, + { + "epoch": 0.71, + "learning_rate": 4.1643201197175774e-06, + "loss": 1.1045, + "step": 29415 + }, + { + "epoch": 0.71, + "learning_rate": 4.163687625099757e-06, + "loss": 1.1036, + "step": 29416 + }, + { + "epoch": 0.71, + "learning_rate": 4.163055165889472e-06, + "loss": 0.9937, + "step": 29417 + }, + { + "epoch": 0.71, + "learning_rate": 4.16242274209056e-06, + "loss": 1.1159, + "step": 29418 + }, + { + "epoch": 0.71, + "learning_rate": 4.161790353706855e-06, + "loss": 0.9255, + "step": 29419 + }, + { + "epoch": 0.71, + "learning_rate": 4.161158000742197e-06, + "loss": 1.0919, + "step": 29420 + }, + { + "epoch": 0.71, + "learning_rate": 4.160525683200416e-06, + "loss": 1.1661, + "step": 29421 + }, + { + "epoch": 0.71, + "learning_rate": 4.159893401085353e-06, + "loss": 1.0597, + "step": 29422 + }, + { + "epoch": 0.71, + "learning_rate": 4.159261154400847e-06, + "loss": 0.8153, + "step": 29423 + }, + { + "epoch": 0.71, + "learning_rate": 4.158628943150724e-06, + "loss": 1.1463, + "step": 29424 + }, + { + "epoch": 0.71, + "learning_rate": 4.157996767338828e-06, + "loss": 1.0074, + "step": 29425 + }, + { + "epoch": 0.71, + "learning_rate": 4.1573646269689926e-06, + "loss": 1.0554, + "step": 29426 + }, + { + "epoch": 0.71, + "learning_rate": 4.156732522045049e-06, + "loss": 1.0195, + "step": 29427 + }, + { + "epoch": 0.71, + "learning_rate": 4.156100452570836e-06, + "loss": 0.9554, + "step": 29428 + }, + { + "epoch": 0.71, + "learning_rate": 4.155468418550189e-06, + "loss": 1.0628, + "step": 29429 + }, + { + "epoch": 0.71, + "learning_rate": 4.154836419986939e-06, + "loss": 1.2154, + "step": 29430 + }, + { + "epoch": 0.71, + "learning_rate": 4.1542044568849225e-06, + "loss": 1.2203, + "step": 29431 + }, + { + "epoch": 0.71, + "learning_rate": 4.15357252924797e-06, + "loss": 1.1366, + "step": 29432 + }, + { + "epoch": 0.71, + "learning_rate": 4.152940637079917e-06, + "loss": 1.0463, + "step": 29433 + }, + { + "epoch": 0.71, + "learning_rate": 4.152308780384603e-06, + "loss": 1.1876, + "step": 29434 + }, + { + "epoch": 0.71, + "learning_rate": 4.151676959165852e-06, + "loss": 1.0185, + "step": 29435 + }, + { + "epoch": 0.71, + "learning_rate": 4.151045173427502e-06, + "loss": 1.0356, + "step": 29436 + }, + { + "epoch": 0.71, + "learning_rate": 4.150413423173387e-06, + "loss": 1.0827, + "step": 29437 + }, + { + "epoch": 0.71, + "learning_rate": 4.149781708407335e-06, + "loss": 1.2732, + "step": 29438 + }, + { + "epoch": 0.71, + "learning_rate": 4.149150029133186e-06, + "loss": 1.2005, + "step": 29439 + }, + { + "epoch": 0.71, + "learning_rate": 4.148518385354763e-06, + "loss": 1.168, + "step": 29440 + }, + { + "epoch": 0.71, + "learning_rate": 4.147886777075902e-06, + "loss": 1.1394, + "step": 29441 + }, + { + "epoch": 0.71, + "learning_rate": 4.14725520430044e-06, + "loss": 0.9364, + "step": 29442 + }, + { + "epoch": 0.71, + "learning_rate": 4.146623667032201e-06, + "loss": 1.2447, + "step": 29443 + }, + { + "epoch": 0.71, + "learning_rate": 4.145992165275018e-06, + "loss": 1.0556, + "step": 29444 + }, + { + "epoch": 0.71, + "learning_rate": 4.145360699032728e-06, + "loss": 1.2921, + "step": 29445 + }, + { + "epoch": 0.71, + "learning_rate": 4.144729268309154e-06, + "loss": 1.1304, + "step": 29446 + }, + { + "epoch": 0.71, + "learning_rate": 4.144097873108132e-06, + "loss": 1.0225, + "step": 29447 + }, + { + "epoch": 0.71, + "learning_rate": 4.14346651343349e-06, + "loss": 1.0209, + "step": 29448 + }, + { + "epoch": 0.71, + "learning_rate": 4.142835189289058e-06, + "loss": 1.2429, + "step": 29449 + }, + { + "epoch": 0.71, + "learning_rate": 4.14220390067867e-06, + "loss": 1.3209, + "step": 29450 + }, + { + "epoch": 0.71, + "learning_rate": 4.14157264760615e-06, + "loss": 1.0132, + "step": 29451 + }, + { + "epoch": 0.71, + "learning_rate": 4.140941430075331e-06, + "loss": 0.9081, + "step": 29452 + }, + { + "epoch": 0.71, + "learning_rate": 4.140310248090046e-06, + "loss": 0.98, + "step": 29453 + }, + { + "epoch": 0.71, + "learning_rate": 4.139679101654116e-06, + "loss": 1.0513, + "step": 29454 + }, + { + "epoch": 0.71, + "learning_rate": 4.139047990771379e-06, + "loss": 0.849, + "step": 29455 + }, + { + "epoch": 0.71, + "learning_rate": 4.138416915445656e-06, + "loss": 0.7956, + "step": 29456 + }, + { + "epoch": 0.71, + "learning_rate": 4.1377858756807774e-06, + "loss": 1.1603, + "step": 29457 + }, + { + "epoch": 0.71, + "learning_rate": 4.137154871480578e-06, + "loss": 0.7574, + "step": 29458 + }, + { + "epoch": 0.71, + "learning_rate": 4.136523902848876e-06, + "loss": 0.9989, + "step": 29459 + }, + { + "epoch": 0.71, + "learning_rate": 4.135892969789506e-06, + "loss": 1.1207, + "step": 29460 + }, + { + "epoch": 0.71, + "learning_rate": 4.135262072306296e-06, + "loss": 1.2462, + "step": 29461 + }, + { + "epoch": 0.71, + "learning_rate": 4.134631210403068e-06, + "loss": 1.1991, + "step": 29462 + }, + { + "epoch": 0.71, + "learning_rate": 4.1340003840836575e-06, + "loss": 1.0994, + "step": 29463 + }, + { + "epoch": 0.71, + "learning_rate": 4.133369593351884e-06, + "loss": 1.0256, + "step": 29464 + }, + { + "epoch": 0.71, + "learning_rate": 4.132738838211575e-06, + "loss": 1.2021, + "step": 29465 + }, + { + "epoch": 0.71, + "learning_rate": 4.132108118666565e-06, + "loss": 0.9982, + "step": 29466 + }, + { + "epoch": 0.71, + "learning_rate": 4.1314774347206694e-06, + "loss": 1.271, + "step": 29467 + }, + { + "epoch": 0.71, + "learning_rate": 4.130846786377721e-06, + "loss": 0.9651, + "step": 29468 + }, + { + "epoch": 0.71, + "learning_rate": 4.130216173641548e-06, + "loss": 1.1411, + "step": 29469 + }, + { + "epoch": 0.71, + "learning_rate": 4.129585596515969e-06, + "loss": 1.0381, + "step": 29470 + }, + { + "epoch": 0.71, + "learning_rate": 4.1289550550048144e-06, + "loss": 0.7892, + "step": 29471 + }, + { + "epoch": 0.71, + "learning_rate": 4.128324549111911e-06, + "loss": 1.1144, + "step": 29472 + }, + { + "epoch": 0.71, + "learning_rate": 4.127694078841078e-06, + "loss": 1.0737, + "step": 29473 + }, + { + "epoch": 0.71, + "learning_rate": 4.127063644196148e-06, + "loss": 1.0163, + "step": 29474 + }, + { + "epoch": 0.71, + "learning_rate": 4.1264332451809375e-06, + "loss": 1.1215, + "step": 29475 + }, + { + "epoch": 0.71, + "learning_rate": 4.125802881799275e-06, + "loss": 1.0237, + "step": 29476 + }, + { + "epoch": 0.71, + "learning_rate": 4.125172554054988e-06, + "loss": 0.9292, + "step": 29477 + }, + { + "epoch": 0.71, + "learning_rate": 4.124542261951894e-06, + "loss": 1.0972, + "step": 29478 + }, + { + "epoch": 0.71, + "learning_rate": 4.12391200549382e-06, + "loss": 1.016, + "step": 29479 + }, + { + "epoch": 0.71, + "learning_rate": 4.123281784684593e-06, + "loss": 0.7664, + "step": 29480 + }, + { + "epoch": 0.71, + "learning_rate": 4.12265159952803e-06, + "loss": 1.0233, + "step": 29481 + }, + { + "epoch": 0.71, + "learning_rate": 4.122021450027961e-06, + "loss": 0.8812, + "step": 29482 + }, + { + "epoch": 0.71, + "learning_rate": 4.1213913361882005e-06, + "loss": 0.9861, + "step": 29483 + }, + { + "epoch": 0.71, + "learning_rate": 4.120761258012578e-06, + "loss": 0.7104, + "step": 29484 + }, + { + "epoch": 0.71, + "learning_rate": 4.120131215504916e-06, + "loss": 0.8737, + "step": 29485 + }, + { + "epoch": 0.71, + "learning_rate": 4.119501208669033e-06, + "loss": 1.0278, + "step": 29486 + }, + { + "epoch": 0.71, + "learning_rate": 4.118871237508751e-06, + "loss": 1.2929, + "step": 29487 + }, + { + "epoch": 0.71, + "learning_rate": 4.118241302027899e-06, + "loss": 1.0529, + "step": 29488 + }, + { + "epoch": 0.71, + "learning_rate": 4.1176114022302895e-06, + "loss": 1.1447, + "step": 29489 + }, + { + "epoch": 0.71, + "learning_rate": 4.116981538119751e-06, + "loss": 0.8437, + "step": 29490 + }, + { + "epoch": 0.71, + "learning_rate": 4.116351709700098e-06, + "loss": 0.9098, + "step": 29491 + }, + { + "epoch": 0.71, + "learning_rate": 4.115721916975156e-06, + "loss": 1.0495, + "step": 29492 + }, + { + "epoch": 0.71, + "learning_rate": 4.1150921599487496e-06, + "loss": 0.8578, + "step": 29493 + }, + { + "epoch": 0.71, + "learning_rate": 4.114462438624689e-06, + "loss": 1.1119, + "step": 29494 + }, + { + "epoch": 0.71, + "learning_rate": 4.1138327530068025e-06, + "loss": 0.9949, + "step": 29495 + }, + { + "epoch": 0.71, + "learning_rate": 4.11320310309891e-06, + "loss": 0.8054, + "step": 29496 + }, + { + "epoch": 0.71, + "learning_rate": 4.1125734889048275e-06, + "loss": 1.0301, + "step": 29497 + }, + { + "epoch": 0.71, + "learning_rate": 4.11194391042838e-06, + "loss": 1.2294, + "step": 29498 + }, + { + "epoch": 0.71, + "learning_rate": 4.111314367673376e-06, + "loss": 1.196, + "step": 29499 + }, + { + "epoch": 0.71, + "learning_rate": 4.11068486064365e-06, + "loss": 1.0367, + "step": 29500 + }, + { + "epoch": 0.71, + "learning_rate": 4.110055389343011e-06, + "loss": 1.153, + "step": 29501 + }, + { + "epoch": 0.71, + "learning_rate": 4.109425953775279e-06, + "loss": 1.1888, + "step": 29502 + }, + { + "epoch": 0.71, + "learning_rate": 4.108796553944278e-06, + "loss": 1.0366, + "step": 29503 + }, + { + "epoch": 0.71, + "learning_rate": 4.108167189853821e-06, + "loss": 1.0386, + "step": 29504 + }, + { + "epoch": 0.71, + "learning_rate": 4.107537861507727e-06, + "loss": 1.3347, + "step": 29505 + }, + { + "epoch": 0.71, + "learning_rate": 4.106908568909818e-06, + "loss": 1.0647, + "step": 29506 + }, + { + "epoch": 0.71, + "learning_rate": 4.106279312063905e-06, + "loss": 1.2844, + "step": 29507 + }, + { + "epoch": 0.71, + "learning_rate": 4.105650090973809e-06, + "loss": 1.1052, + "step": 29508 + }, + { + "epoch": 0.71, + "learning_rate": 4.10502090564335e-06, + "loss": 0.8688, + "step": 29509 + }, + { + "epoch": 0.71, + "learning_rate": 4.10439175607634e-06, + "loss": 1.0525, + "step": 29510 + }, + { + "epoch": 0.71, + "learning_rate": 4.103762642276603e-06, + "loss": 1.1205, + "step": 29511 + }, + { + "epoch": 0.71, + "learning_rate": 4.103133564247947e-06, + "loss": 1.1728, + "step": 29512 + }, + { + "epoch": 0.71, + "learning_rate": 4.102504521994192e-06, + "loss": 1.0423, + "step": 29513 + }, + { + "epoch": 0.71, + "learning_rate": 4.101875515519158e-06, + "loss": 0.8554, + "step": 29514 + }, + { + "epoch": 0.71, + "learning_rate": 4.101246544826656e-06, + "loss": 1.0775, + "step": 29515 + }, + { + "epoch": 0.71, + "learning_rate": 4.100617609920503e-06, + "loss": 1.0213, + "step": 29516 + }, + { + "epoch": 0.71, + "learning_rate": 4.099988710804519e-06, + "loss": 0.7749, + "step": 29517 + }, + { + "epoch": 0.71, + "learning_rate": 4.099359847482511e-06, + "loss": 0.873, + "step": 29518 + }, + { + "epoch": 0.71, + "learning_rate": 4.0987310199583045e-06, + "loss": 1.0519, + "step": 29519 + }, + { + "epoch": 0.71, + "learning_rate": 4.0981022282357045e-06, + "loss": 1.0754, + "step": 29520 + }, + { + "epoch": 0.71, + "learning_rate": 4.09747347231853e-06, + "loss": 1.2162, + "step": 29521 + }, + { + "epoch": 0.71, + "learning_rate": 4.096844752210598e-06, + "loss": 0.8488, + "step": 29522 + }, + { + "epoch": 0.71, + "learning_rate": 4.096216067915718e-06, + "loss": 1.0463, + "step": 29523 + }, + { + "epoch": 0.71, + "learning_rate": 4.095587419437706e-06, + "loss": 0.8244, + "step": 29524 + }, + { + "epoch": 0.71, + "learning_rate": 4.09495880678038e-06, + "loss": 1.2713, + "step": 29525 + }, + { + "epoch": 0.71, + "learning_rate": 4.094330229947546e-06, + "loss": 1.0931, + "step": 29526 + }, + { + "epoch": 0.71, + "learning_rate": 4.093701688943026e-06, + "loss": 1.0323, + "step": 29527 + }, + { + "epoch": 0.71, + "learning_rate": 4.093073183770625e-06, + "loss": 1.1341, + "step": 29528 + }, + { + "epoch": 0.71, + "learning_rate": 4.092444714434158e-06, + "loss": 1.0901, + "step": 29529 + }, + { + "epoch": 0.71, + "learning_rate": 4.091816280937445e-06, + "loss": 0.8769, + "step": 29530 + }, + { + "epoch": 0.71, + "learning_rate": 4.091187883284289e-06, + "loss": 1.3763, + "step": 29531 + }, + { + "epoch": 0.71, + "learning_rate": 4.090559521478505e-06, + "loss": 0.9135, + "step": 29532 + }, + { + "epoch": 0.71, + "learning_rate": 4.08993119552391e-06, + "loss": 1.0487, + "step": 29533 + }, + { + "epoch": 0.71, + "learning_rate": 4.089302905424311e-06, + "loss": 0.9664, + "step": 29534 + }, + { + "epoch": 0.71, + "learning_rate": 4.0886746511835215e-06, + "loss": 1.031, + "step": 29535 + }, + { + "epoch": 0.71, + "learning_rate": 4.088046432805352e-06, + "loss": 1.2949, + "step": 29536 + }, + { + "epoch": 0.71, + "learning_rate": 4.087418250293611e-06, + "loss": 1.128, + "step": 29537 + }, + { + "epoch": 0.71, + "learning_rate": 4.0867901036521186e-06, + "loss": 0.9094, + "step": 29538 + }, + { + "epoch": 0.71, + "learning_rate": 4.086161992884675e-06, + "loss": 0.9999, + "step": 29539 + }, + { + "epoch": 0.71, + "learning_rate": 4.085533917995097e-06, + "loss": 0.966, + "step": 29540 + }, + { + "epoch": 0.71, + "learning_rate": 4.084905878987196e-06, + "loss": 1.1235, + "step": 29541 + }, + { + "epoch": 0.71, + "learning_rate": 4.084277875864776e-06, + "loss": 0.7651, + "step": 29542 + }, + { + "epoch": 0.71, + "learning_rate": 4.083649908631655e-06, + "loss": 0.8353, + "step": 29543 + }, + { + "epoch": 0.71, + "learning_rate": 4.0830219772916336e-06, + "loss": 0.8612, + "step": 29544 + }, + { + "epoch": 0.71, + "learning_rate": 4.082394081848527e-06, + "loss": 1.0176, + "step": 29545 + }, + { + "epoch": 0.71, + "learning_rate": 4.0817662223061475e-06, + "loss": 1.1149, + "step": 29546 + }, + { + "epoch": 0.71, + "learning_rate": 4.081138398668296e-06, + "loss": 1.1601, + "step": 29547 + }, + { + "epoch": 0.71, + "learning_rate": 4.0805106109387856e-06, + "loss": 1.3039, + "step": 29548 + }, + { + "epoch": 0.71, + "learning_rate": 4.079882859121429e-06, + "loss": 0.9982, + "step": 29549 + }, + { + "epoch": 0.71, + "learning_rate": 4.079255143220027e-06, + "loss": 1.1836, + "step": 29550 + }, + { + "epoch": 0.71, + "learning_rate": 4.078627463238392e-06, + "loss": 1.1002, + "step": 29551 + }, + { + "epoch": 0.71, + "learning_rate": 4.077999819180334e-06, + "loss": 0.7807, + "step": 29552 + }, + { + "epoch": 0.71, + "learning_rate": 4.077372211049655e-06, + "loss": 1.1062, + "step": 29553 + }, + { + "epoch": 0.71, + "learning_rate": 4.0767446388501685e-06, + "loss": 0.8372, + "step": 29554 + }, + { + "epoch": 0.71, + "learning_rate": 4.076117102585676e-06, + "loss": 1.1176, + "step": 29555 + }, + { + "epoch": 0.71, + "learning_rate": 4.075489602259988e-06, + "loss": 1.093, + "step": 29556 + }, + { + "epoch": 0.71, + "learning_rate": 4.0748621378769145e-06, + "loss": 1.0109, + "step": 29557 + }, + { + "epoch": 0.71, + "learning_rate": 4.074234709440254e-06, + "loss": 0.9858, + "step": 29558 + }, + { + "epoch": 0.71, + "learning_rate": 4.073607316953819e-06, + "loss": 0.9436, + "step": 29559 + }, + { + "epoch": 0.71, + "learning_rate": 4.072979960421418e-06, + "loss": 0.9391, + "step": 29560 + }, + { + "epoch": 0.71, + "learning_rate": 4.072352639846851e-06, + "loss": 1.0357, + "step": 29561 + }, + { + "epoch": 0.71, + "learning_rate": 4.071725355233928e-06, + "loss": 0.7399, + "step": 29562 + }, + { + "epoch": 0.71, + "learning_rate": 4.07109810658645e-06, + "loss": 1.0242, + "step": 29563 + }, + { + "epoch": 0.71, + "learning_rate": 4.070470893908224e-06, + "loss": 0.9412, + "step": 29564 + }, + { + "epoch": 0.71, + "learning_rate": 4.069843717203061e-06, + "loss": 1.2591, + "step": 29565 + }, + { + "epoch": 0.71, + "learning_rate": 4.0692165764747585e-06, + "loss": 0.962, + "step": 29566 + }, + { + "epoch": 0.71, + "learning_rate": 4.068589471727123e-06, + "loss": 1.0352, + "step": 29567 + }, + { + "epoch": 0.71, + "learning_rate": 4.067962402963963e-06, + "loss": 0.9498, + "step": 29568 + }, + { + "epoch": 0.71, + "learning_rate": 4.0673353701890774e-06, + "loss": 0.9233, + "step": 29569 + }, + { + "epoch": 0.71, + "learning_rate": 4.066708373406275e-06, + "loss": 1.1401, + "step": 29570 + }, + { + "epoch": 0.71, + "learning_rate": 4.0660814126193546e-06, + "loss": 0.7765, + "step": 29571 + }, + { + "epoch": 0.71, + "learning_rate": 4.0654544878321235e-06, + "loss": 1.1183, + "step": 29572 + }, + { + "epoch": 0.71, + "learning_rate": 4.064827599048386e-06, + "loss": 1.0807, + "step": 29573 + }, + { + "epoch": 0.71, + "learning_rate": 4.06420074627194e-06, + "loss": 1.1422, + "step": 29574 + }, + { + "epoch": 0.71, + "learning_rate": 4.063573929506594e-06, + "loss": 1.0865, + "step": 29575 + }, + { + "epoch": 0.71, + "learning_rate": 4.06294714875615e-06, + "loss": 1.131, + "step": 29576 + }, + { + "epoch": 0.71, + "learning_rate": 4.0623204040244055e-06, + "loss": 0.9556, + "step": 29577 + }, + { + "epoch": 0.71, + "learning_rate": 4.061693695315171e-06, + "loss": 1.2464, + "step": 29578 + }, + { + "epoch": 0.71, + "learning_rate": 4.06106702263224e-06, + "loss": 1.0122, + "step": 29579 + }, + { + "epoch": 0.71, + "learning_rate": 4.0604403859794195e-06, + "loss": 1.0028, + "step": 29580 + }, + { + "epoch": 0.71, + "learning_rate": 4.0598137853605125e-06, + "loss": 1.0408, + "step": 29581 + }, + { + "epoch": 0.71, + "learning_rate": 4.059187220779315e-06, + "loss": 0.8347, + "step": 29582 + }, + { + "epoch": 0.71, + "learning_rate": 4.058560692239631e-06, + "loss": 0.9242, + "step": 29583 + }, + { + "epoch": 0.71, + "learning_rate": 4.057934199745266e-06, + "loss": 1.1886, + "step": 29584 + }, + { + "epoch": 0.71, + "learning_rate": 4.057307743300012e-06, + "loss": 0.7023, + "step": 29585 + }, + { + "epoch": 0.71, + "learning_rate": 4.056681322907678e-06, + "loss": 0.9247, + "step": 29586 + }, + { + "epoch": 0.71, + "learning_rate": 4.056054938572057e-06, + "loss": 1.1897, + "step": 29587 + }, + { + "epoch": 0.71, + "learning_rate": 4.055428590296953e-06, + "loss": 1.1647, + "step": 29588 + }, + { + "epoch": 0.71, + "learning_rate": 4.054802278086168e-06, + "loss": 1.0449, + "step": 29589 + }, + { + "epoch": 0.71, + "learning_rate": 4.054176001943497e-06, + "loss": 1.0671, + "step": 29590 + }, + { + "epoch": 0.71, + "learning_rate": 4.053549761872742e-06, + "loss": 1.227, + "step": 29591 + }, + { + "epoch": 0.71, + "learning_rate": 4.052923557877704e-06, + "loss": 1.2501, + "step": 29592 + }, + { + "epoch": 0.71, + "learning_rate": 4.052297389962177e-06, + "loss": 1.0003, + "step": 29593 + }, + { + "epoch": 0.71, + "learning_rate": 4.051671258129966e-06, + "loss": 0.8794, + "step": 29594 + }, + { + "epoch": 0.71, + "learning_rate": 4.051045162384862e-06, + "loss": 1.2216, + "step": 29595 + }, + { + "epoch": 0.71, + "learning_rate": 4.0504191027306695e-06, + "loss": 0.9421, + "step": 29596 + }, + { + "epoch": 0.71, + "learning_rate": 4.049793079171187e-06, + "loss": 0.9111, + "step": 29597 + }, + { + "epoch": 0.71, + "learning_rate": 4.049167091710208e-06, + "loss": 1.0947, + "step": 29598 + }, + { + "epoch": 0.71, + "learning_rate": 4.0485411403515306e-06, + "loss": 1.3108, + "step": 29599 + }, + { + "epoch": 0.71, + "learning_rate": 4.047915225098959e-06, + "loss": 1.233, + "step": 29600 + }, + { + "epoch": 0.71, + "learning_rate": 4.047289345956282e-06, + "loss": 0.8164, + "step": 29601 + }, + { + "epoch": 0.71, + "learning_rate": 4.046663502927302e-06, + "loss": 1.0395, + "step": 29602 + }, + { + "epoch": 0.71, + "learning_rate": 4.046037696015815e-06, + "loss": 1.1146, + "step": 29603 + }, + { + "epoch": 0.71, + "learning_rate": 4.045411925225615e-06, + "loss": 0.9589, + "step": 29604 + }, + { + "epoch": 0.71, + "learning_rate": 4.044786190560504e-06, + "loss": 0.974, + "step": 29605 + }, + { + "epoch": 0.71, + "learning_rate": 4.0441604920242705e-06, + "loss": 1.0477, + "step": 29606 + }, + { + "epoch": 0.71, + "learning_rate": 4.043534829620714e-06, + "loss": 0.9668, + "step": 29607 + }, + { + "epoch": 0.71, + "learning_rate": 4.042909203353634e-06, + "loss": 0.8944, + "step": 29608 + }, + { + "epoch": 0.71, + "learning_rate": 4.042283613226819e-06, + "loss": 0.876, + "step": 29609 + }, + { + "epoch": 0.71, + "learning_rate": 4.041658059244069e-06, + "loss": 0.9858, + "step": 29610 + }, + { + "epoch": 0.71, + "learning_rate": 4.041032541409181e-06, + "loss": 1.149, + "step": 29611 + }, + { + "epoch": 0.71, + "learning_rate": 4.0404070597259434e-06, + "loss": 0.7741, + "step": 29612 + }, + { + "epoch": 0.71, + "learning_rate": 4.0397816141981586e-06, + "loss": 1.2233, + "step": 29613 + }, + { + "epoch": 0.71, + "learning_rate": 4.039156204829613e-06, + "loss": 1.0963, + "step": 29614 + }, + { + "epoch": 0.71, + "learning_rate": 4.038530831624103e-06, + "loss": 0.965, + "step": 29615 + }, + { + "epoch": 0.71, + "learning_rate": 4.03790549458543e-06, + "loss": 1.1105, + "step": 29616 + }, + { + "epoch": 0.71, + "learning_rate": 4.0372801937173775e-06, + "loss": 0.9879, + "step": 29617 + }, + { + "epoch": 0.71, + "learning_rate": 4.036654929023743e-06, + "loss": 0.9658, + "step": 29618 + }, + { + "epoch": 0.71, + "learning_rate": 4.0360297005083245e-06, + "loss": 1.0638, + "step": 29619 + }, + { + "epoch": 0.71, + "learning_rate": 4.035404508174908e-06, + "loss": 1.0826, + "step": 29620 + }, + { + "epoch": 0.71, + "learning_rate": 4.034779352027293e-06, + "loss": 0.8551, + "step": 29621 + }, + { + "epoch": 0.71, + "learning_rate": 4.034154232069264e-06, + "loss": 1.0792, + "step": 29622 + }, + { + "epoch": 0.71, + "learning_rate": 4.033529148304618e-06, + "loss": 1.1824, + "step": 29623 + }, + { + "epoch": 0.71, + "learning_rate": 4.032904100737152e-06, + "loss": 0.8666, + "step": 29624 + }, + { + "epoch": 0.71, + "learning_rate": 4.032279089370649e-06, + "loss": 1.2528, + "step": 29625 + }, + { + "epoch": 0.71, + "learning_rate": 4.031654114208905e-06, + "loss": 0.9696, + "step": 29626 + }, + { + "epoch": 0.71, + "learning_rate": 4.0310291752557155e-06, + "loss": 1.0536, + "step": 29627 + }, + { + "epoch": 0.71, + "learning_rate": 4.030404272514864e-06, + "loss": 1.2099, + "step": 29628 + }, + { + "epoch": 0.71, + "learning_rate": 4.02977940599015e-06, + "loss": 1.1256, + "step": 29629 + }, + { + "epoch": 0.71, + "learning_rate": 4.029154575685354e-06, + "loss": 1.1553, + "step": 29630 + }, + { + "epoch": 0.71, + "learning_rate": 4.028529781604276e-06, + "loss": 1.1737, + "step": 29631 + }, + { + "epoch": 0.71, + "learning_rate": 4.027905023750706e-06, + "loss": 1.1031, + "step": 29632 + }, + { + "epoch": 0.71, + "learning_rate": 4.027280302128427e-06, + "loss": 0.8374, + "step": 29633 + }, + { + "epoch": 0.71, + "learning_rate": 4.026655616741233e-06, + "loss": 1.0082, + "step": 29634 + }, + { + "epoch": 0.71, + "learning_rate": 4.02603096759292e-06, + "loss": 1.1362, + "step": 29635 + }, + { + "epoch": 0.71, + "learning_rate": 4.025406354687267e-06, + "loss": 1.0584, + "step": 29636 + }, + { + "epoch": 0.71, + "learning_rate": 4.024781778028072e-06, + "loss": 1.0501, + "step": 29637 + }, + { + "epoch": 0.71, + "learning_rate": 4.024157237619118e-06, + "loss": 1.1589, + "step": 29638 + }, + { + "epoch": 0.71, + "learning_rate": 4.023532733464195e-06, + "loss": 0.8625, + "step": 29639 + }, + { + "epoch": 0.71, + "learning_rate": 4.022908265567097e-06, + "loss": 1.0852, + "step": 29640 + }, + { + "epoch": 0.71, + "learning_rate": 4.022283833931605e-06, + "loss": 1.1083, + "step": 29641 + }, + { + "epoch": 0.71, + "learning_rate": 4.021659438561511e-06, + "loss": 0.9856, + "step": 29642 + }, + { + "epoch": 0.71, + "learning_rate": 4.021035079460607e-06, + "loss": 0.9179, + "step": 29643 + }, + { + "epoch": 0.71, + "learning_rate": 4.020410756632674e-06, + "loss": 0.712, + "step": 29644 + }, + { + "epoch": 0.71, + "learning_rate": 4.0197864700815e-06, + "loss": 1.2224, + "step": 29645 + }, + { + "epoch": 0.71, + "learning_rate": 4.0191622198108806e-06, + "loss": 1.1458, + "step": 29646 + }, + { + "epoch": 0.71, + "learning_rate": 4.018538005824593e-06, + "loss": 1.0453, + "step": 29647 + }, + { + "epoch": 0.71, + "learning_rate": 4.017913828126432e-06, + "loss": 0.9046, + "step": 29648 + }, + { + "epoch": 0.71, + "learning_rate": 4.017289686720177e-06, + "loss": 1.0551, + "step": 29649 + }, + { + "epoch": 0.71, + "learning_rate": 4.016665581609619e-06, + "loss": 1.0855, + "step": 29650 + }, + { + "epoch": 0.71, + "learning_rate": 4.016041512798546e-06, + "loss": 1.1743, + "step": 29651 + }, + { + "epoch": 0.71, + "learning_rate": 4.0154174802907386e-06, + "loss": 0.9195, + "step": 29652 + }, + { + "epoch": 0.71, + "learning_rate": 4.014793484089986e-06, + "loss": 1.1073, + "step": 29653 + }, + { + "epoch": 0.71, + "learning_rate": 4.014169524200076e-06, + "loss": 1.0068, + "step": 29654 + }, + { + "epoch": 0.71, + "learning_rate": 4.013545600624789e-06, + "loss": 0.9823, + "step": 29655 + }, + { + "epoch": 0.71, + "learning_rate": 4.012921713367917e-06, + "loss": 1.0205, + "step": 29656 + }, + { + "epoch": 0.71, + "learning_rate": 4.012297862433235e-06, + "loss": 1.0582, + "step": 29657 + }, + { + "epoch": 0.71, + "learning_rate": 4.011674047824535e-06, + "loss": 1.059, + "step": 29658 + }, + { + "epoch": 0.71, + "learning_rate": 4.011050269545603e-06, + "loss": 1.1386, + "step": 29659 + }, + { + "epoch": 0.71, + "learning_rate": 4.010426527600218e-06, + "loss": 1.1484, + "step": 29660 + }, + { + "epoch": 0.71, + "learning_rate": 4.0098028219921645e-06, + "loss": 0.6711, + "step": 29661 + }, + { + "epoch": 0.71, + "learning_rate": 4.009179152725233e-06, + "loss": 1.0886, + "step": 29662 + }, + { + "epoch": 0.71, + "learning_rate": 4.008555519803199e-06, + "loss": 0.794, + "step": 29663 + }, + { + "epoch": 0.71, + "learning_rate": 4.007931923229853e-06, + "loss": 1.1837, + "step": 29664 + }, + { + "epoch": 0.71, + "learning_rate": 4.007308363008971e-06, + "loss": 1.1819, + "step": 29665 + }, + { + "epoch": 0.71, + "learning_rate": 4.006684839144339e-06, + "loss": 1.0024, + "step": 29666 + }, + { + "epoch": 0.71, + "learning_rate": 4.006061351639741e-06, + "loss": 1.039, + "step": 29667 + }, + { + "epoch": 0.71, + "learning_rate": 4.005437900498959e-06, + "loss": 0.8733, + "step": 29668 + }, + { + "epoch": 0.71, + "learning_rate": 4.004814485725778e-06, + "loss": 1.0377, + "step": 29669 + }, + { + "epoch": 0.71, + "learning_rate": 4.004191107323974e-06, + "loss": 1.0746, + "step": 29670 + }, + { + "epoch": 0.71, + "learning_rate": 4.003567765297331e-06, + "loss": 1.1754, + "step": 29671 + }, + { + "epoch": 0.71, + "learning_rate": 4.002944459649638e-06, + "loss": 1.225, + "step": 29672 + }, + { + "epoch": 0.71, + "learning_rate": 4.002321190384666e-06, + "loss": 1.2067, + "step": 29673 + }, + { + "epoch": 0.71, + "learning_rate": 4.0016979575062e-06, + "loss": 0.9075, + "step": 29674 + }, + { + "epoch": 0.71, + "learning_rate": 4.001074761018026e-06, + "loss": 1.056, + "step": 29675 + }, + { + "epoch": 0.71, + "learning_rate": 4.000451600923915e-06, + "loss": 1.1741, + "step": 29676 + }, + { + "epoch": 0.71, + "learning_rate": 3.999828477227659e-06, + "loss": 1.1536, + "step": 29677 + }, + { + "epoch": 0.71, + "learning_rate": 3.999205389933028e-06, + "loss": 1.0615, + "step": 29678 + }, + { + "epoch": 0.71, + "learning_rate": 3.998582339043805e-06, + "loss": 1.1688, + "step": 29679 + }, + { + "epoch": 0.71, + "learning_rate": 3.997959324563776e-06, + "loss": 0.8544, + "step": 29680 + }, + { + "epoch": 0.71, + "learning_rate": 3.997336346496712e-06, + "loss": 0.7034, + "step": 29681 + }, + { + "epoch": 0.71, + "learning_rate": 3.996713404846397e-06, + "loss": 1.1423, + "step": 29682 + }, + { + "epoch": 0.71, + "learning_rate": 3.996090499616613e-06, + "loss": 0.9221, + "step": 29683 + }, + { + "epoch": 0.71, + "learning_rate": 3.995467630811133e-06, + "loss": 0.9889, + "step": 29684 + }, + { + "epoch": 0.71, + "learning_rate": 3.994844798433741e-06, + "loss": 1.1404, + "step": 29685 + }, + { + "epoch": 0.71, + "learning_rate": 3.994222002488209e-06, + "loss": 1.0553, + "step": 29686 + }, + { + "epoch": 0.71, + "learning_rate": 3.993599242978321e-06, + "loss": 1.1964, + "step": 29687 + }, + { + "epoch": 0.71, + "learning_rate": 3.9929765199078565e-06, + "loss": 1.1139, + "step": 29688 + }, + { + "epoch": 0.71, + "learning_rate": 3.992353833280588e-06, + "loss": 0.7555, + "step": 29689 + }, + { + "epoch": 0.71, + "learning_rate": 3.991731183100293e-06, + "loss": 0.843, + "step": 29690 + }, + { + "epoch": 0.71, + "learning_rate": 3.991108569370757e-06, + "loss": 1.157, + "step": 29691 + }, + { + "epoch": 0.71, + "learning_rate": 3.99048599209575e-06, + "loss": 1.2128, + "step": 29692 + }, + { + "epoch": 0.71, + "learning_rate": 3.989863451279053e-06, + "loss": 1.0145, + "step": 29693 + }, + { + "epoch": 0.71, + "learning_rate": 3.9892409469244365e-06, + "loss": 0.8644, + "step": 29694 + }, + { + "epoch": 0.71, + "learning_rate": 3.988618479035682e-06, + "loss": 1.1308, + "step": 29695 + }, + { + "epoch": 0.71, + "learning_rate": 3.987996047616568e-06, + "loss": 1.0902, + "step": 29696 + }, + { + "epoch": 0.71, + "learning_rate": 3.987373652670865e-06, + "loss": 1.0928, + "step": 29697 + }, + { + "epoch": 0.71, + "learning_rate": 3.986751294202352e-06, + "loss": 1.2207, + "step": 29698 + }, + { + "epoch": 0.71, + "learning_rate": 3.986128972214808e-06, + "loss": 1.3436, + "step": 29699 + }, + { + "epoch": 0.71, + "learning_rate": 3.985506686712001e-06, + "loss": 0.8299, + "step": 29700 + }, + { + "epoch": 0.71, + "learning_rate": 3.984884437697714e-06, + "loss": 0.9021, + "step": 29701 + }, + { + "epoch": 0.71, + "learning_rate": 3.984262225175714e-06, + "loss": 1.0574, + "step": 29702 + }, + { + "epoch": 0.71, + "learning_rate": 3.983640049149782e-06, + "loss": 0.9367, + "step": 29703 + }, + { + "epoch": 0.71, + "learning_rate": 3.9830179096236935e-06, + "loss": 0.8919, + "step": 29704 + }, + { + "epoch": 0.71, + "learning_rate": 3.982395806601216e-06, + "loss": 1.0381, + "step": 29705 + }, + { + "epoch": 0.71, + "learning_rate": 3.981773740086128e-06, + "loss": 0.9708, + "step": 29706 + }, + { + "epoch": 0.71, + "learning_rate": 3.981151710082206e-06, + "loss": 1.1601, + "step": 29707 + }, + { + "epoch": 0.71, + "learning_rate": 3.980529716593218e-06, + "loss": 1.108, + "step": 29708 + }, + { + "epoch": 0.71, + "learning_rate": 3.979907759622943e-06, + "loss": 1.0168, + "step": 29709 + }, + { + "epoch": 0.71, + "learning_rate": 3.979285839175149e-06, + "loss": 1.1248, + "step": 29710 + }, + { + "epoch": 0.71, + "learning_rate": 3.978663955253611e-06, + "loss": 1.1706, + "step": 29711 + }, + { + "epoch": 0.71, + "learning_rate": 3.978042107862107e-06, + "loss": 0.8405, + "step": 29712 + }, + { + "epoch": 0.71, + "learning_rate": 3.977420297004401e-06, + "loss": 0.9364, + "step": 29713 + }, + { + "epoch": 0.71, + "learning_rate": 3.97679852268427e-06, + "loss": 1.111, + "step": 29714 + }, + { + "epoch": 0.71, + "learning_rate": 3.976176784905488e-06, + "loss": 0.9328, + "step": 29715 + }, + { + "epoch": 0.71, + "learning_rate": 3.975555083671821e-06, + "loss": 1.0923, + "step": 29716 + }, + { + "epoch": 0.71, + "learning_rate": 3.974933418987049e-06, + "loss": 1.143, + "step": 29717 + }, + { + "epoch": 0.71, + "learning_rate": 3.974311790854934e-06, + "loss": 0.9244, + "step": 29718 + }, + { + "epoch": 0.71, + "learning_rate": 3.9736901992792535e-06, + "loss": 1.0517, + "step": 29719 + }, + { + "epoch": 0.71, + "learning_rate": 3.97306864426378e-06, + "loss": 1.1062, + "step": 29720 + }, + { + "epoch": 0.71, + "learning_rate": 3.9724471258122775e-06, + "loss": 1.2427, + "step": 29721 + }, + { + "epoch": 0.71, + "learning_rate": 3.971825643928521e-06, + "loss": 1.0788, + "step": 29722 + }, + { + "epoch": 0.71, + "learning_rate": 3.971204198616284e-06, + "loss": 1.1105, + "step": 29723 + }, + { + "epoch": 0.71, + "learning_rate": 3.970582789879329e-06, + "loss": 1.1666, + "step": 29724 + }, + { + "epoch": 0.71, + "learning_rate": 3.96996141772143e-06, + "loss": 0.9791, + "step": 29725 + }, + { + "epoch": 0.71, + "learning_rate": 3.96934008214636e-06, + "loss": 1.0883, + "step": 29726 + }, + { + "epoch": 0.71, + "learning_rate": 3.9687187831578826e-06, + "loss": 0.9671, + "step": 29727 + }, + { + "epoch": 0.71, + "learning_rate": 3.968097520759773e-06, + "loss": 1.1438, + "step": 29728 + }, + { + "epoch": 0.71, + "learning_rate": 3.967476294955793e-06, + "loss": 1.0348, + "step": 29729 + }, + { + "epoch": 0.71, + "learning_rate": 3.966855105749716e-06, + "loss": 0.8023, + "step": 29730 + }, + { + "epoch": 0.71, + "learning_rate": 3.966233953145313e-06, + "loss": 1.1454, + "step": 29731 + }, + { + "epoch": 0.72, + "learning_rate": 3.965612837146346e-06, + "loss": 0.9583, + "step": 29732 + }, + { + "epoch": 0.72, + "learning_rate": 3.9649917577565866e-06, + "loss": 1.0677, + "step": 29733 + }, + { + "epoch": 0.72, + "learning_rate": 3.964370714979806e-06, + "loss": 0.9164, + "step": 29734 + }, + { + "epoch": 0.72, + "learning_rate": 3.9637497088197654e-06, + "loss": 0.9802, + "step": 29735 + }, + { + "epoch": 0.72, + "learning_rate": 3.963128739280239e-06, + "loss": 0.8806, + "step": 29736 + }, + { + "epoch": 0.72, + "learning_rate": 3.962507806364987e-06, + "loss": 0.8919, + "step": 29737 + }, + { + "epoch": 0.72, + "learning_rate": 3.961886910077779e-06, + "loss": 1.0097, + "step": 29738 + }, + { + "epoch": 0.72, + "learning_rate": 3.961266050422387e-06, + "loss": 0.9851, + "step": 29739 + }, + { + "epoch": 0.72, + "learning_rate": 3.960645227402571e-06, + "loss": 1.2514, + "step": 29740 + }, + { + "epoch": 0.72, + "learning_rate": 3.960024441022099e-06, + "loss": 0.939, + "step": 29741 + }, + { + "epoch": 0.72, + "learning_rate": 3.959403691284741e-06, + "loss": 1.215, + "step": 29742 + }, + { + "epoch": 0.72, + "learning_rate": 3.958782978194258e-06, + "loss": 1.0294, + "step": 29743 + }, + { + "epoch": 0.72, + "learning_rate": 3.95816230175442e-06, + "loss": 1.0672, + "step": 29744 + }, + { + "epoch": 0.72, + "learning_rate": 3.957541661968986e-06, + "loss": 1.1374, + "step": 29745 + }, + { + "epoch": 0.72, + "learning_rate": 3.956921058841727e-06, + "loss": 1.0892, + "step": 29746 + }, + { + "epoch": 0.72, + "learning_rate": 3.956300492376411e-06, + "loss": 1.1357, + "step": 29747 + }, + { + "epoch": 0.72, + "learning_rate": 3.955679962576793e-06, + "loss": 1.0859, + "step": 29748 + }, + { + "epoch": 0.72, + "learning_rate": 3.955059469446643e-06, + "loss": 1.1665, + "step": 29749 + }, + { + "epoch": 0.72, + "learning_rate": 3.9544390129897295e-06, + "loss": 0.8471, + "step": 29750 + }, + { + "epoch": 0.72, + "learning_rate": 3.9538185932098085e-06, + "loss": 1.2275, + "step": 29751 + }, + { + "epoch": 0.72, + "learning_rate": 3.953198210110651e-06, + "loss": 1.0458, + "step": 29752 + }, + { + "epoch": 0.72, + "learning_rate": 3.9525778636960145e-06, + "loss": 1.0264, + "step": 29753 + }, + { + "epoch": 0.72, + "learning_rate": 3.9519575539696665e-06, + "loss": 1.2096, + "step": 29754 + }, + { + "epoch": 0.72, + "learning_rate": 3.951337280935372e-06, + "loss": 0.996, + "step": 29755 + }, + { + "epoch": 0.72, + "learning_rate": 3.950717044596889e-06, + "loss": 0.8168, + "step": 29756 + }, + { + "epoch": 0.72, + "learning_rate": 3.950096844957981e-06, + "loss": 1.1535, + "step": 29757 + }, + { + "epoch": 0.72, + "learning_rate": 3.9494766820224175e-06, + "loss": 1.2366, + "step": 29758 + }, + { + "epoch": 0.72, + "learning_rate": 3.9488565557939516e-06, + "loss": 0.8858, + "step": 29759 + }, + { + "epoch": 0.72, + "learning_rate": 3.948236466276354e-06, + "loss": 0.9635, + "step": 29760 + }, + { + "epoch": 0.72, + "learning_rate": 3.947616413473378e-06, + "loss": 0.7151, + "step": 29761 + }, + { + "epoch": 0.72, + "learning_rate": 3.946996397388791e-06, + "loss": 0.8626, + "step": 29762 + }, + { + "epoch": 0.72, + "learning_rate": 3.946376418026354e-06, + "loss": 1.1225, + "step": 29763 + }, + { + "epoch": 0.72, + "learning_rate": 3.945756475389826e-06, + "loss": 1.0716, + "step": 29764 + }, + { + "epoch": 0.72, + "learning_rate": 3.945136569482968e-06, + "loss": 0.9481, + "step": 29765 + }, + { + "epoch": 0.72, + "learning_rate": 3.944516700309547e-06, + "loss": 0.9465, + "step": 29766 + }, + { + "epoch": 0.72, + "learning_rate": 3.943896867873315e-06, + "loss": 0.9566, + "step": 29767 + }, + { + "epoch": 0.72, + "learning_rate": 3.943277072178035e-06, + "loss": 0.9016, + "step": 29768 + }, + { + "epoch": 0.72, + "learning_rate": 3.942657313227474e-06, + "loss": 0.9408, + "step": 29769 + }, + { + "epoch": 0.72, + "learning_rate": 3.9420375910253816e-06, + "loss": 1.1143, + "step": 29770 + }, + { + "epoch": 0.72, + "learning_rate": 3.941417905575525e-06, + "loss": 0.9722, + "step": 29771 + }, + { + "epoch": 0.72, + "learning_rate": 3.940798256881658e-06, + "loss": 0.9945, + "step": 29772 + }, + { + "epoch": 0.72, + "learning_rate": 3.940178644947544e-06, + "loss": 0.9359, + "step": 29773 + }, + { + "epoch": 0.72, + "learning_rate": 3.939559069776943e-06, + "loss": 1.0505, + "step": 29774 + }, + { + "epoch": 0.72, + "learning_rate": 3.938939531373609e-06, + "loss": 0.8561, + "step": 29775 + }, + { + "epoch": 0.72, + "learning_rate": 3.938320029741302e-06, + "loss": 1.0748, + "step": 29776 + }, + { + "epoch": 0.72, + "learning_rate": 3.937700564883786e-06, + "loss": 1.0486, + "step": 29777 + }, + { + "epoch": 0.72, + "learning_rate": 3.937081136804811e-06, + "loss": 1.0037, + "step": 29778 + }, + { + "epoch": 0.72, + "learning_rate": 3.936461745508142e-06, + "loss": 1.0325, + "step": 29779 + }, + { + "epoch": 0.72, + "learning_rate": 3.93584239099753e-06, + "loss": 0.9507, + "step": 29780 + }, + { + "epoch": 0.72, + "learning_rate": 3.935223073276736e-06, + "loss": 1.0358, + "step": 29781 + }, + { + "epoch": 0.72, + "learning_rate": 3.93460379234952e-06, + "loss": 1.0425, + "step": 29782 + }, + { + "epoch": 0.72, + "learning_rate": 3.9339845482196325e-06, + "loss": 1.0626, + "step": 29783 + }, + { + "epoch": 0.72, + "learning_rate": 3.933365340890834e-06, + "loss": 1.1376, + "step": 29784 + }, + { + "epoch": 0.72, + "learning_rate": 3.932746170366886e-06, + "loss": 0.9241, + "step": 29785 + }, + { + "epoch": 0.72, + "learning_rate": 3.9321270366515355e-06, + "loss": 1.1095, + "step": 29786 + }, + { + "epoch": 0.72, + "learning_rate": 3.931507939748546e-06, + "loss": 0.999, + "step": 29787 + }, + { + "epoch": 0.72, + "learning_rate": 3.930888879661667e-06, + "loss": 0.862, + "step": 29788 + }, + { + "epoch": 0.72, + "learning_rate": 3.930269856394658e-06, + "loss": 1.0923, + "step": 29789 + }, + { + "epoch": 0.72, + "learning_rate": 3.929650869951278e-06, + "loss": 1.2176, + "step": 29790 + }, + { + "epoch": 0.72, + "learning_rate": 3.929031920335275e-06, + "loss": 0.8193, + "step": 29791 + }, + { + "epoch": 0.72, + "learning_rate": 3.928413007550406e-06, + "loss": 1.1664, + "step": 29792 + }, + { + "epoch": 0.72, + "learning_rate": 3.927794131600431e-06, + "loss": 0.9454, + "step": 29793 + }, + { + "epoch": 0.72, + "learning_rate": 3.927175292489098e-06, + "loss": 0.8942, + "step": 29794 + }, + { + "epoch": 0.72, + "learning_rate": 3.926556490220167e-06, + "loss": 1.1286, + "step": 29795 + }, + { + "epoch": 0.72, + "learning_rate": 3.925937724797386e-06, + "loss": 1.1435, + "step": 29796 + }, + { + "epoch": 0.72, + "learning_rate": 3.925318996224512e-06, + "loss": 0.9, + "step": 29797 + }, + { + "epoch": 0.72, + "learning_rate": 3.924700304505302e-06, + "loss": 0.8123, + "step": 29798 + }, + { + "epoch": 0.72, + "learning_rate": 3.924081649643503e-06, + "loss": 1.1824, + "step": 29799 + }, + { + "epoch": 0.72, + "learning_rate": 3.923463031642873e-06, + "loss": 1.0636, + "step": 29800 + }, + { + "epoch": 0.72, + "learning_rate": 3.922844450507165e-06, + "loss": 0.9946, + "step": 29801 + }, + { + "epoch": 0.72, + "learning_rate": 3.922225906240127e-06, + "loss": 1.1454, + "step": 29802 + }, + { + "epoch": 0.72, + "learning_rate": 3.921607398845518e-06, + "loss": 1.0907, + "step": 29803 + }, + { + "epoch": 0.72, + "learning_rate": 3.920988928327085e-06, + "loss": 0.918, + "step": 29804 + }, + { + "epoch": 0.72, + "learning_rate": 3.9203704946885816e-06, + "loss": 1.2843, + "step": 29805 + }, + { + "epoch": 0.72, + "learning_rate": 3.919752097933764e-06, + "loss": 0.9168, + "step": 29806 + }, + { + "epoch": 0.72, + "learning_rate": 3.919133738066377e-06, + "loss": 1.0432, + "step": 29807 + }, + { + "epoch": 0.72, + "learning_rate": 3.918515415090175e-06, + "loss": 1.2552, + "step": 29808 + }, + { + "epoch": 0.72, + "learning_rate": 3.917897129008913e-06, + "loss": 0.957, + "step": 29809 + }, + { + "epoch": 0.72, + "learning_rate": 3.917278879826335e-06, + "loss": 1.0976, + "step": 29810 + }, + { + "epoch": 0.72, + "learning_rate": 3.9166606675462e-06, + "loss": 0.962, + "step": 29811 + }, + { + "epoch": 0.72, + "learning_rate": 3.916042492172248e-06, + "loss": 0.8826, + "step": 29812 + }, + { + "epoch": 0.72, + "learning_rate": 3.915424353708237e-06, + "loss": 1.2507, + "step": 29813 + }, + { + "epoch": 0.72, + "learning_rate": 3.914806252157918e-06, + "loss": 1.2678, + "step": 29814 + }, + { + "epoch": 0.72, + "learning_rate": 3.914188187525034e-06, + "loss": 0.7606, + "step": 29815 + }, + { + "epoch": 0.72, + "learning_rate": 3.9135701598133405e-06, + "loss": 0.8042, + "step": 29816 + }, + { + "epoch": 0.72, + "learning_rate": 3.912952169026587e-06, + "loss": 1.1393, + "step": 29817 + }, + { + "epoch": 0.72, + "learning_rate": 3.912334215168519e-06, + "loss": 0.8324, + "step": 29818 + }, + { + "epoch": 0.72, + "learning_rate": 3.911716298242886e-06, + "loss": 1.1669, + "step": 29819 + }, + { + "epoch": 0.72, + "learning_rate": 3.911098418253443e-06, + "loss": 0.9478, + "step": 29820 + }, + { + "epoch": 0.72, + "learning_rate": 3.910480575203929e-06, + "loss": 0.9285, + "step": 29821 + }, + { + "epoch": 0.72, + "learning_rate": 3.909862769098102e-06, + "loss": 0.9706, + "step": 29822 + }, + { + "epoch": 0.72, + "learning_rate": 3.9092449999397e-06, + "loss": 1.1015, + "step": 29823 + }, + { + "epoch": 0.72, + "learning_rate": 3.908627267732477e-06, + "loss": 0.8555, + "step": 29824 + }, + { + "epoch": 0.72, + "learning_rate": 3.908009572480182e-06, + "loss": 0.9279, + "step": 29825 + }, + { + "epoch": 0.72, + "learning_rate": 3.907391914186557e-06, + "loss": 1.1152, + "step": 29826 + }, + { + "epoch": 0.72, + "learning_rate": 3.9067742928553524e-06, + "loss": 0.8465, + "step": 29827 + }, + { + "epoch": 0.72, + "learning_rate": 3.906156708490319e-06, + "loss": 0.9992, + "step": 29828 + }, + { + "epoch": 0.72, + "learning_rate": 3.905539161095196e-06, + "loss": 1.0686, + "step": 29829 + }, + { + "epoch": 0.72, + "learning_rate": 3.904921650673735e-06, + "loss": 1.1593, + "step": 29830 + }, + { + "epoch": 0.72, + "learning_rate": 3.904304177229678e-06, + "loss": 1.1409, + "step": 29831 + }, + { + "epoch": 0.72, + "learning_rate": 3.903686740766774e-06, + "loss": 1.1786, + "step": 29832 + }, + { + "epoch": 0.72, + "learning_rate": 3.90306934128877e-06, + "loss": 1.0748, + "step": 29833 + }, + { + "epoch": 0.72, + "learning_rate": 3.902451978799409e-06, + "loss": 1.0725, + "step": 29834 + }, + { + "epoch": 0.72, + "learning_rate": 3.901834653302441e-06, + "loss": 0.9407, + "step": 29835 + }, + { + "epoch": 0.72, + "learning_rate": 3.901217364801605e-06, + "loss": 1.2352, + "step": 29836 + }, + { + "epoch": 0.72, + "learning_rate": 3.900600113300648e-06, + "loss": 1.0528, + "step": 29837 + }, + { + "epoch": 0.72, + "learning_rate": 3.899982898803319e-06, + "loss": 0.9635, + "step": 29838 + }, + { + "epoch": 0.72, + "learning_rate": 3.899365721313355e-06, + "loss": 0.9503, + "step": 29839 + }, + { + "epoch": 0.72, + "learning_rate": 3.898748580834508e-06, + "loss": 0.9356, + "step": 29840 + }, + { + "epoch": 0.72, + "learning_rate": 3.898131477370516e-06, + "loss": 0.7738, + "step": 29841 + }, + { + "epoch": 0.72, + "learning_rate": 3.897514410925124e-06, + "loss": 1.2307, + "step": 29842 + }, + { + "epoch": 0.72, + "learning_rate": 3.896897381502081e-06, + "loss": 0.839, + "step": 29843 + }, + { + "epoch": 0.72, + "learning_rate": 3.896280389105121e-06, + "loss": 0.7846, + "step": 29844 + }, + { + "epoch": 0.72, + "learning_rate": 3.895663433737993e-06, + "loss": 1.0042, + "step": 29845 + }, + { + "epoch": 0.72, + "learning_rate": 3.895046515404442e-06, + "loss": 1.0432, + "step": 29846 + }, + { + "epoch": 0.72, + "learning_rate": 3.894429634108205e-06, + "loss": 0.8164, + "step": 29847 + }, + { + "epoch": 0.72, + "learning_rate": 3.893812789853028e-06, + "loss": 0.9388, + "step": 29848 + }, + { + "epoch": 0.72, + "learning_rate": 3.893195982642656e-06, + "loss": 0.9997, + "step": 29849 + }, + { + "epoch": 0.72, + "learning_rate": 3.892579212480823e-06, + "loss": 1.2593, + "step": 29850 + }, + { + "epoch": 0.72, + "learning_rate": 3.891962479371279e-06, + "loss": 1.0919, + "step": 29851 + }, + { + "epoch": 0.72, + "learning_rate": 3.891345783317759e-06, + "loss": 1.0555, + "step": 29852 + }, + { + "epoch": 0.72, + "learning_rate": 3.890729124324006e-06, + "loss": 0.9963, + "step": 29853 + }, + { + "epoch": 0.72, + "learning_rate": 3.890112502393768e-06, + "loss": 1.006, + "step": 29854 + }, + { + "epoch": 0.72, + "learning_rate": 3.8894959175307755e-06, + "loss": 1.0246, + "step": 29855 + }, + { + "epoch": 0.72, + "learning_rate": 3.888879369738774e-06, + "loss": 1.0651, + "step": 29856 + }, + { + "epoch": 0.72, + "learning_rate": 3.888262859021508e-06, + "loss": 0.6524, + "step": 29857 + }, + { + "epoch": 0.72, + "learning_rate": 3.88764638538271e-06, + "loss": 0.7965, + "step": 29858 + }, + { + "epoch": 0.72, + "learning_rate": 3.8870299488261265e-06, + "loss": 1.1396, + "step": 29859 + }, + { + "epoch": 0.72, + "learning_rate": 3.886413549355491e-06, + "loss": 1.1199, + "step": 29860 + }, + { + "epoch": 0.72, + "learning_rate": 3.885797186974547e-06, + "loss": 1.0937, + "step": 29861 + }, + { + "epoch": 0.72, + "learning_rate": 3.8851808616870365e-06, + "loss": 0.813, + "step": 29862 + }, + { + "epoch": 0.72, + "learning_rate": 3.884564573496693e-06, + "loss": 0.8123, + "step": 29863 + }, + { + "epoch": 0.72, + "learning_rate": 3.883948322407256e-06, + "loss": 1.3832, + "step": 29864 + }, + { + "epoch": 0.72, + "learning_rate": 3.883332108422471e-06, + "loss": 1.021, + "step": 29865 + }, + { + "epoch": 0.72, + "learning_rate": 3.882715931546067e-06, + "loss": 0.9231, + "step": 29866 + }, + { + "epoch": 0.72, + "learning_rate": 3.8820997917817905e-06, + "loss": 1.3115, + "step": 29867 + }, + { + "epoch": 0.72, + "learning_rate": 3.881483689133372e-06, + "loss": 1.1735, + "step": 29868 + }, + { + "epoch": 0.72, + "learning_rate": 3.880867623604553e-06, + "loss": 1.0503, + "step": 29869 + }, + { + "epoch": 0.72, + "learning_rate": 3.880251595199074e-06, + "loss": 1.0573, + "step": 29870 + }, + { + "epoch": 0.72, + "learning_rate": 3.879635603920665e-06, + "loss": 1.0841, + "step": 29871 + }, + { + "epoch": 0.72, + "learning_rate": 3.879019649773068e-06, + "loss": 0.8919, + "step": 29872 + }, + { + "epoch": 0.72, + "learning_rate": 3.878403732760022e-06, + "loss": 1.0563, + "step": 29873 + }, + { + "epoch": 0.72, + "learning_rate": 3.877787852885258e-06, + "loss": 1.0849, + "step": 29874 + }, + { + "epoch": 0.72, + "learning_rate": 3.877172010152518e-06, + "loss": 1.2154, + "step": 29875 + }, + { + "epoch": 0.72, + "learning_rate": 3.876556204565531e-06, + "loss": 1.0398, + "step": 29876 + }, + { + "epoch": 0.72, + "learning_rate": 3.8759404361280385e-06, + "loss": 0.9521, + "step": 29877 + }, + { + "epoch": 0.72, + "learning_rate": 3.8753247048437785e-06, + "loss": 1.1287, + "step": 29878 + }, + { + "epoch": 0.72, + "learning_rate": 3.874709010716479e-06, + "loss": 1.0732, + "step": 29879 + }, + { + "epoch": 0.72, + "learning_rate": 3.874093353749878e-06, + "loss": 1.2621, + "step": 29880 + }, + { + "epoch": 0.72, + "learning_rate": 3.873477733947717e-06, + "loss": 1.089, + "step": 29881 + }, + { + "epoch": 0.72, + "learning_rate": 3.872862151313721e-06, + "loss": 0.9391, + "step": 29882 + }, + { + "epoch": 0.72, + "learning_rate": 3.872246605851633e-06, + "loss": 1.0533, + "step": 29883 + }, + { + "epoch": 0.72, + "learning_rate": 3.871631097565179e-06, + "loss": 1.1527, + "step": 29884 + }, + { + "epoch": 0.72, + "learning_rate": 3.871015626458099e-06, + "loss": 0.7681, + "step": 29885 + }, + { + "epoch": 0.72, + "learning_rate": 3.870400192534128e-06, + "loss": 1.1513, + "step": 29886 + }, + { + "epoch": 0.72, + "learning_rate": 3.869784795796995e-06, + "loss": 1.2704, + "step": 29887 + }, + { + "epoch": 0.72, + "learning_rate": 3.869169436250435e-06, + "loss": 1.0564, + "step": 29888 + }, + { + "epoch": 0.72, + "learning_rate": 3.8685541138981865e-06, + "loss": 1.068, + "step": 29889 + }, + { + "epoch": 0.72, + "learning_rate": 3.867938828743973e-06, + "loss": 0.9606, + "step": 29890 + }, + { + "epoch": 0.72, + "learning_rate": 3.867323580791537e-06, + "loss": 0.9556, + "step": 29891 + }, + { + "epoch": 0.72, + "learning_rate": 3.866708370044602e-06, + "loss": 1.2544, + "step": 29892 + }, + { + "epoch": 0.72, + "learning_rate": 3.866093196506905e-06, + "loss": 0.9897, + "step": 29893 + }, + { + "epoch": 0.72, + "learning_rate": 3.865478060182181e-06, + "loss": 0.925, + "step": 29894 + }, + { + "epoch": 0.72, + "learning_rate": 3.864862961074155e-06, + "loss": 0.9941, + "step": 29895 + }, + { + "epoch": 0.72, + "learning_rate": 3.864247899186563e-06, + "loss": 0.969, + "step": 29896 + }, + { + "epoch": 0.72, + "learning_rate": 3.863632874523139e-06, + "loss": 1.0117, + "step": 29897 + }, + { + "epoch": 0.72, + "learning_rate": 3.863017887087606e-06, + "loss": 0.9351, + "step": 29898 + }, + { + "epoch": 0.72, + "learning_rate": 3.862402936883702e-06, + "loss": 1.2196, + "step": 29899 + }, + { + "epoch": 0.72, + "learning_rate": 3.861788023915159e-06, + "loss": 1.0982, + "step": 29900 + }, + { + "epoch": 0.72, + "learning_rate": 3.861173148185701e-06, + "loss": 0.8224, + "step": 29901 + }, + { + "epoch": 0.72, + "learning_rate": 3.860558309699064e-06, + "loss": 1.0149, + "step": 29902 + }, + { + "epoch": 0.72, + "learning_rate": 3.859943508458972e-06, + "loss": 1.0104, + "step": 29903 + }, + { + "epoch": 0.72, + "learning_rate": 3.859328744469159e-06, + "loss": 1.0345, + "step": 29904 + }, + { + "epoch": 0.72, + "learning_rate": 3.858714017733358e-06, + "loss": 1.2501, + "step": 29905 + }, + { + "epoch": 0.72, + "learning_rate": 3.858099328255291e-06, + "loss": 1.0931, + "step": 29906 + }, + { + "epoch": 0.72, + "learning_rate": 3.857484676038691e-06, + "loss": 1.1945, + "step": 29907 + }, + { + "epoch": 0.72, + "learning_rate": 3.856870061087289e-06, + "loss": 1.1352, + "step": 29908 + }, + { + "epoch": 0.72, + "learning_rate": 3.856255483404809e-06, + "loss": 0.9881, + "step": 29909 + }, + { + "epoch": 0.72, + "learning_rate": 3.855640942994985e-06, + "loss": 0.9156, + "step": 29910 + }, + { + "epoch": 0.72, + "learning_rate": 3.855026439861539e-06, + "loss": 1.0981, + "step": 29911 + }, + { + "epoch": 0.72, + "learning_rate": 3.8544119740082005e-06, + "loss": 1.2, + "step": 29912 + }, + { + "epoch": 0.72, + "learning_rate": 3.853797545438705e-06, + "loss": 1.1458, + "step": 29913 + }, + { + "epoch": 0.72, + "learning_rate": 3.853183154156769e-06, + "loss": 1.2424, + "step": 29914 + }, + { + "epoch": 0.72, + "learning_rate": 3.852568800166126e-06, + "loss": 0.5503, + "step": 29915 + }, + { + "epoch": 0.72, + "learning_rate": 3.851954483470505e-06, + "loss": 0.9027, + "step": 29916 + }, + { + "epoch": 0.72, + "learning_rate": 3.851340204073626e-06, + "loss": 1.1768, + "step": 29917 + }, + { + "epoch": 0.72, + "learning_rate": 3.850725961979223e-06, + "loss": 1.1117, + "step": 29918 + }, + { + "epoch": 0.72, + "learning_rate": 3.850111757191018e-06, + "loss": 0.6971, + "step": 29919 + }, + { + "epoch": 0.72, + "learning_rate": 3.8494975897127365e-06, + "loss": 1.3185, + "step": 29920 + }, + { + "epoch": 0.72, + "learning_rate": 3.848883459548111e-06, + "loss": 1.0296, + "step": 29921 + }, + { + "epoch": 0.72, + "learning_rate": 3.848269366700857e-06, + "loss": 0.8753, + "step": 29922 + }, + { + "epoch": 0.72, + "learning_rate": 3.847655311174708e-06, + "loss": 0.8292, + "step": 29923 + }, + { + "epoch": 0.72, + "learning_rate": 3.84704129297339e-06, + "loss": 1.206, + "step": 29924 + }, + { + "epoch": 0.72, + "learning_rate": 3.8464273121006225e-06, + "loss": 1.1188, + "step": 29925 + }, + { + "epoch": 0.72, + "learning_rate": 3.845813368560135e-06, + "loss": 1.2373, + "step": 29926 + }, + { + "epoch": 0.72, + "learning_rate": 3.845199462355648e-06, + "loss": 0.8994, + "step": 29927 + }, + { + "epoch": 0.72, + "learning_rate": 3.844585593490887e-06, + "loss": 0.9945, + "step": 29928 + }, + { + "epoch": 0.72, + "learning_rate": 3.843971761969583e-06, + "loss": 1.0341, + "step": 29929 + }, + { + "epoch": 0.72, + "learning_rate": 3.843357967795448e-06, + "loss": 1.1074, + "step": 29930 + }, + { + "epoch": 0.72, + "learning_rate": 3.842744210972213e-06, + "loss": 1.1942, + "step": 29931 + }, + { + "epoch": 0.72, + "learning_rate": 3.842130491503605e-06, + "loss": 1.3916, + "step": 29932 + }, + { + "epoch": 0.72, + "learning_rate": 3.84151680939334e-06, + "loss": 1.1164, + "step": 29933 + }, + { + "epoch": 0.72, + "learning_rate": 3.840903164645145e-06, + "loss": 0.8473, + "step": 29934 + }, + { + "epoch": 0.72, + "learning_rate": 3.840289557262741e-06, + "loss": 1.0656, + "step": 29935 + }, + { + "epoch": 0.72, + "learning_rate": 3.83967598724985e-06, + "loss": 0.9176, + "step": 29936 + }, + { + "epoch": 0.72, + "learning_rate": 3.8390624546102e-06, + "loss": 0.9222, + "step": 29937 + }, + { + "epoch": 0.72, + "learning_rate": 3.8384489593475035e-06, + "loss": 0.9344, + "step": 29938 + }, + { + "epoch": 0.72, + "learning_rate": 3.83783550146549e-06, + "loss": 1.0126, + "step": 29939 + }, + { + "epoch": 0.72, + "learning_rate": 3.837222080967882e-06, + "loss": 1.0035, + "step": 29940 + }, + { + "epoch": 0.72, + "learning_rate": 3.836608697858394e-06, + "loss": 0.8917, + "step": 29941 + }, + { + "epoch": 0.72, + "learning_rate": 3.835995352140751e-06, + "loss": 0.9469, + "step": 29942 + }, + { + "epoch": 0.72, + "learning_rate": 3.835382043818679e-06, + "loss": 1.0023, + "step": 29943 + }, + { + "epoch": 0.72, + "learning_rate": 3.8347687728958906e-06, + "loss": 0.986, + "step": 29944 + }, + { + "epoch": 0.72, + "learning_rate": 3.834155539376112e-06, + "loss": 1.2414, + "step": 29945 + }, + { + "epoch": 0.72, + "learning_rate": 3.833542343263059e-06, + "loss": 1.0537, + "step": 29946 + }, + { + "epoch": 0.72, + "learning_rate": 3.832929184560455e-06, + "loss": 0.8161, + "step": 29947 + }, + { + "epoch": 0.72, + "learning_rate": 3.832316063272021e-06, + "loss": 0.8773, + "step": 29948 + }, + { + "epoch": 0.72, + "learning_rate": 3.831702979401472e-06, + "loss": 1.2915, + "step": 29949 + }, + { + "epoch": 0.72, + "learning_rate": 3.83108993295253e-06, + "loss": 1.1423, + "step": 29950 + }, + { + "epoch": 0.72, + "learning_rate": 3.8304769239289176e-06, + "loss": 0.8594, + "step": 29951 + }, + { + "epoch": 0.72, + "learning_rate": 3.829863952334348e-06, + "loss": 1.1843, + "step": 29952 + }, + { + "epoch": 0.72, + "learning_rate": 3.829251018172545e-06, + "loss": 1.0408, + "step": 29953 + }, + { + "epoch": 0.72, + "learning_rate": 3.828638121447221e-06, + "loss": 0.9353, + "step": 29954 + }, + { + "epoch": 0.72, + "learning_rate": 3.828025262162098e-06, + "loss": 1.1407, + "step": 29955 + }, + { + "epoch": 0.72, + "learning_rate": 3.827412440320898e-06, + "loss": 1.1582, + "step": 29956 + }, + { + "epoch": 0.72, + "learning_rate": 3.826799655927331e-06, + "loss": 1.2377, + "step": 29957 + }, + { + "epoch": 0.72, + "learning_rate": 3.826186908985119e-06, + "loss": 1.1279, + "step": 29958 + }, + { + "epoch": 0.72, + "learning_rate": 3.825574199497982e-06, + "loss": 1.0324, + "step": 29959 + }, + { + "epoch": 0.72, + "learning_rate": 3.82496152746963e-06, + "loss": 1.2023, + "step": 29960 + }, + { + "epoch": 0.72, + "learning_rate": 3.824348892903788e-06, + "loss": 1.1277, + "step": 29961 + }, + { + "epoch": 0.72, + "learning_rate": 3.823736295804166e-06, + "loss": 1.0041, + "step": 29962 + }, + { + "epoch": 0.72, + "learning_rate": 3.8231237361744824e-06, + "loss": 1.0145, + "step": 29963 + }, + { + "epoch": 0.72, + "learning_rate": 3.822511214018457e-06, + "loss": 1.049, + "step": 29964 + }, + { + "epoch": 0.72, + "learning_rate": 3.8218987293398005e-06, + "loss": 0.9495, + "step": 29965 + }, + { + "epoch": 0.72, + "learning_rate": 3.8212862821422305e-06, + "loss": 1.1456, + "step": 29966 + }, + { + "epoch": 0.72, + "learning_rate": 3.820673872429467e-06, + "loss": 0.9806, + "step": 29967 + }, + { + "epoch": 0.72, + "learning_rate": 3.820061500205219e-06, + "loss": 0.9157, + "step": 29968 + }, + { + "epoch": 0.72, + "learning_rate": 3.8194491654732066e-06, + "loss": 1.0815, + "step": 29969 + }, + { + "epoch": 0.72, + "learning_rate": 3.818836868237139e-06, + "loss": 0.9796, + "step": 29970 + }, + { + "epoch": 0.72, + "learning_rate": 3.818224608500735e-06, + "loss": 0.7235, + "step": 29971 + }, + { + "epoch": 0.72, + "learning_rate": 3.81761238626771e-06, + "loss": 1.1366, + "step": 29972 + }, + { + "epoch": 0.72, + "learning_rate": 3.817000201541775e-06, + "loss": 1.004, + "step": 29973 + }, + { + "epoch": 0.72, + "learning_rate": 3.816388054326645e-06, + "loss": 0.8098, + "step": 29974 + }, + { + "epoch": 0.72, + "learning_rate": 3.815775944626038e-06, + "loss": 1.0201, + "step": 29975 + }, + { + "epoch": 0.72, + "learning_rate": 3.815163872443659e-06, + "loss": 1.2902, + "step": 29976 + }, + { + "epoch": 0.72, + "learning_rate": 3.81455183778323e-06, + "loss": 0.9393, + "step": 29977 + }, + { + "epoch": 0.72, + "learning_rate": 3.813939840648457e-06, + "loss": 1.0819, + "step": 29978 + }, + { + "epoch": 0.72, + "learning_rate": 3.8133278810430553e-06, + "loss": 0.8225, + "step": 29979 + }, + { + "epoch": 0.72, + "learning_rate": 3.8127159589707417e-06, + "loss": 1.2102, + "step": 29980 + }, + { + "epoch": 0.72, + "learning_rate": 3.812104074435222e-06, + "loss": 1.0954, + "step": 29981 + }, + { + "epoch": 0.72, + "learning_rate": 3.811492227440212e-06, + "loss": 1.3489, + "step": 29982 + }, + { + "epoch": 0.72, + "learning_rate": 3.8108804179894253e-06, + "loss": 1.0231, + "step": 29983 + }, + { + "epoch": 0.72, + "learning_rate": 3.810268646086569e-06, + "loss": 0.9806, + "step": 29984 + }, + { + "epoch": 0.72, + "learning_rate": 3.8096569117353553e-06, + "loss": 0.96, + "step": 29985 + }, + { + "epoch": 0.72, + "learning_rate": 3.809045214939502e-06, + "loss": 1.2372, + "step": 29986 + }, + { + "epoch": 0.72, + "learning_rate": 3.808433555702711e-06, + "loss": 1.3096, + "step": 29987 + }, + { + "epoch": 0.72, + "learning_rate": 3.8078219340287016e-06, + "loss": 1.0107, + "step": 29988 + }, + { + "epoch": 0.72, + "learning_rate": 3.8072103499211753e-06, + "loss": 1.0921, + "step": 29989 + }, + { + "epoch": 0.72, + "learning_rate": 3.8065988033838475e-06, + "loss": 1.0217, + "step": 29990 + }, + { + "epoch": 0.72, + "learning_rate": 3.8059872944204324e-06, + "loss": 1.2133, + "step": 29991 + }, + { + "epoch": 0.72, + "learning_rate": 3.8053758230346315e-06, + "loss": 1.2292, + "step": 29992 + }, + { + "epoch": 0.72, + "learning_rate": 3.8047643892301578e-06, + "loss": 0.9519, + "step": 29993 + }, + { + "epoch": 0.72, + "learning_rate": 3.8041529930107255e-06, + "loss": 0.9496, + "step": 29994 + }, + { + "epoch": 0.72, + "learning_rate": 3.803541634380036e-06, + "loss": 0.9389, + "step": 29995 + }, + { + "epoch": 0.72, + "learning_rate": 3.802930313341805e-06, + "loss": 0.7818, + "step": 29996 + }, + { + "epoch": 0.72, + "learning_rate": 3.802319029899735e-06, + "loss": 1.0637, + "step": 29997 + }, + { + "epoch": 0.72, + "learning_rate": 3.801707784057538e-06, + "loss": 1.1591, + "step": 29998 + }, + { + "epoch": 0.72, + "learning_rate": 3.801096575818921e-06, + "loss": 1.0455, + "step": 29999 + }, + { + "epoch": 0.72, + "learning_rate": 3.800485405187594e-06, + "loss": 1.0383, + "step": 30000 + }, + { + "epoch": 0.72, + "learning_rate": 3.7998742721672667e-06, + "loss": 1.0878, + "step": 30001 + }, + { + "epoch": 0.72, + "learning_rate": 3.7992631767616406e-06, + "loss": 1.2316, + "step": 30002 + }, + { + "epoch": 0.72, + "learning_rate": 3.798652118974426e-06, + "loss": 1.1881, + "step": 30003 + }, + { + "epoch": 0.72, + "learning_rate": 3.798041098809334e-06, + "loss": 1.1138, + "step": 30004 + }, + { + "epoch": 0.72, + "learning_rate": 3.7974301162700644e-06, + "loss": 1.0371, + "step": 30005 + }, + { + "epoch": 0.72, + "learning_rate": 3.7968191713603308e-06, + "loss": 1.1024, + "step": 30006 + }, + { + "epoch": 0.72, + "learning_rate": 3.796208264083834e-06, + "loss": 1.0973, + "step": 30007 + }, + { + "epoch": 0.72, + "learning_rate": 3.7955973944442804e-06, + "loss": 1.3427, + "step": 30008 + }, + { + "epoch": 0.72, + "learning_rate": 3.794986562445383e-06, + "loss": 1.1001, + "step": 30009 + }, + { + "epoch": 0.72, + "learning_rate": 3.7943757680908377e-06, + "loss": 0.7347, + "step": 30010 + }, + { + "epoch": 0.72, + "learning_rate": 3.793765011384356e-06, + "loss": 1.0953, + "step": 30011 + }, + { + "epoch": 0.72, + "learning_rate": 3.793154292329646e-06, + "loss": 1.2808, + "step": 30012 + }, + { + "epoch": 0.72, + "learning_rate": 3.792543610930405e-06, + "loss": 0.898, + "step": 30013 + }, + { + "epoch": 0.72, + "learning_rate": 3.7919329671903447e-06, + "loss": 0.8068, + "step": 30014 + }, + { + "epoch": 0.72, + "learning_rate": 3.7913223611131646e-06, + "loss": 1.0695, + "step": 30015 + }, + { + "epoch": 0.72, + "learning_rate": 3.7907117927025706e-06, + "loss": 1.0783, + "step": 30016 + }, + { + "epoch": 0.72, + "learning_rate": 3.790101261962271e-06, + "loss": 1.0684, + "step": 30017 + }, + { + "epoch": 0.72, + "learning_rate": 3.7894907688959625e-06, + "loss": 1.0004, + "step": 30018 + }, + { + "epoch": 0.72, + "learning_rate": 3.7888803135073536e-06, + "loss": 1.2052, + "step": 30019 + }, + { + "epoch": 0.72, + "learning_rate": 3.788269895800151e-06, + "loss": 1.0795, + "step": 30020 + }, + { + "epoch": 0.72, + "learning_rate": 3.7876595157780493e-06, + "loss": 0.948, + "step": 30021 + }, + { + "epoch": 0.72, + "learning_rate": 3.7870491734447557e-06, + "loss": 0.9764, + "step": 30022 + }, + { + "epoch": 0.72, + "learning_rate": 3.7864388688039777e-06, + "loss": 1.1985, + "step": 30023 + }, + { + "epoch": 0.72, + "learning_rate": 3.785828601859409e-06, + "loss": 0.9556, + "step": 30024 + }, + { + "epoch": 0.72, + "learning_rate": 3.7852183726147616e-06, + "loss": 1.1459, + "step": 30025 + }, + { + "epoch": 0.72, + "learning_rate": 3.784608181073728e-06, + "loss": 1.002, + "step": 30026 + }, + { + "epoch": 0.72, + "learning_rate": 3.7839980272400147e-06, + "loss": 0.8477, + "step": 30027 + }, + { + "epoch": 0.72, + "learning_rate": 3.7833879111173276e-06, + "loss": 1.1119, + "step": 30028 + }, + { + "epoch": 0.72, + "learning_rate": 3.782777832709359e-06, + "loss": 0.9893, + "step": 30029 + }, + { + "epoch": 0.72, + "learning_rate": 3.7821677920198164e-06, + "loss": 0.8182, + "step": 30030 + }, + { + "epoch": 0.72, + "learning_rate": 3.7815577890524013e-06, + "loss": 0.9753, + "step": 30031 + }, + { + "epoch": 0.72, + "learning_rate": 3.78094782381081e-06, + "loss": 1.0337, + "step": 30032 + }, + { + "epoch": 0.72, + "learning_rate": 3.780337896298748e-06, + "loss": 0.9718, + "step": 30033 + }, + { + "epoch": 0.72, + "learning_rate": 3.7797280065199102e-06, + "loss": 1.2598, + "step": 30034 + }, + { + "epoch": 0.72, + "learning_rate": 3.7791181544779997e-06, + "loss": 0.7677, + "step": 30035 + }, + { + "epoch": 0.72, + "learning_rate": 3.7785083401767187e-06, + "loss": 1.1963, + "step": 30036 + }, + { + "epoch": 0.72, + "learning_rate": 3.777898563619762e-06, + "loss": 0.9348, + "step": 30037 + }, + { + "epoch": 0.72, + "learning_rate": 3.77728882481083e-06, + "loss": 0.9981, + "step": 30038 + }, + { + "epoch": 0.72, + "learning_rate": 3.776679123753627e-06, + "loss": 0.948, + "step": 30039 + }, + { + "epoch": 0.72, + "learning_rate": 3.7760694604518445e-06, + "loss": 1.0747, + "step": 30040 + }, + { + "epoch": 0.72, + "learning_rate": 3.7754598349091877e-06, + "loss": 1.0735, + "step": 30041 + }, + { + "epoch": 0.72, + "learning_rate": 3.7748502471293492e-06, + "loss": 0.9986, + "step": 30042 + }, + { + "epoch": 0.72, + "learning_rate": 3.7742406971160295e-06, + "loss": 1.0918, + "step": 30043 + }, + { + "epoch": 0.72, + "learning_rate": 3.77363118487293e-06, + "loss": 0.9495, + "step": 30044 + }, + { + "epoch": 0.72, + "learning_rate": 3.7730217104037437e-06, + "loss": 0.8924, + "step": 30045 + }, + { + "epoch": 0.72, + "learning_rate": 3.7724122737121692e-06, + "loss": 0.9889, + "step": 30046 + }, + { + "epoch": 0.72, + "learning_rate": 3.7718028748019074e-06, + "loss": 1.0856, + "step": 30047 + }, + { + "epoch": 0.72, + "learning_rate": 3.77119351367665e-06, + "loss": 1.1744, + "step": 30048 + }, + { + "epoch": 0.72, + "learning_rate": 3.7705841903401e-06, + "loss": 1.2528, + "step": 30049 + }, + { + "epoch": 0.72, + "learning_rate": 3.7699749047959466e-06, + "loss": 0.6726, + "step": 30050 + }, + { + "epoch": 0.72, + "learning_rate": 3.7693656570478897e-06, + "loss": 0.7854, + "step": 30051 + }, + { + "epoch": 0.72, + "learning_rate": 3.7687564470996296e-06, + "loss": 1.2747, + "step": 30052 + }, + { + "epoch": 0.72, + "learning_rate": 3.7681472749548553e-06, + "loss": 1.2293, + "step": 30053 + }, + { + "epoch": 0.72, + "learning_rate": 3.767538140617265e-06, + "loss": 0.8313, + "step": 30054 + }, + { + "epoch": 0.72, + "learning_rate": 3.7669290440905595e-06, + "loss": 0.8037, + "step": 30055 + }, + { + "epoch": 0.72, + "learning_rate": 3.766319985378425e-06, + "loss": 1.0006, + "step": 30056 + }, + { + "epoch": 0.72, + "learning_rate": 3.7657109644845644e-06, + "loss": 1.0712, + "step": 30057 + }, + { + "epoch": 0.72, + "learning_rate": 3.7651019814126656e-06, + "loss": 1.0735, + "step": 30058 + }, + { + "epoch": 0.72, + "learning_rate": 3.7644930361664266e-06, + "loss": 1.1433, + "step": 30059 + }, + { + "epoch": 0.72, + "learning_rate": 3.7638841287495455e-06, + "loss": 0.9692, + "step": 30060 + }, + { + "epoch": 0.72, + "learning_rate": 3.7632752591657086e-06, + "loss": 1.0183, + "step": 30061 + }, + { + "epoch": 0.72, + "learning_rate": 3.762666427418614e-06, + "loss": 0.9882, + "step": 30062 + }, + { + "epoch": 0.72, + "learning_rate": 3.7620576335119585e-06, + "loss": 0.9767, + "step": 30063 + }, + { + "epoch": 0.72, + "learning_rate": 3.7614488774494283e-06, + "loss": 0.9763, + "step": 30064 + }, + { + "epoch": 0.72, + "learning_rate": 3.7608401592347198e-06, + "loss": 1.0517, + "step": 30065 + }, + { + "epoch": 0.72, + "learning_rate": 3.760231478871531e-06, + "loss": 0.9556, + "step": 30066 + }, + { + "epoch": 0.72, + "learning_rate": 3.7596228363635468e-06, + "loss": 1.1207, + "step": 30067 + }, + { + "epoch": 0.72, + "learning_rate": 3.7590142317144663e-06, + "loss": 0.9523, + "step": 30068 + }, + { + "epoch": 0.72, + "learning_rate": 3.7584056649279742e-06, + "loss": 1.0599, + "step": 30069 + }, + { + "epoch": 0.72, + "learning_rate": 3.757797136007768e-06, + "loss": 0.9676, + "step": 30070 + }, + { + "epoch": 0.72, + "learning_rate": 3.757188644957541e-06, + "loss": 1.1699, + "step": 30071 + }, + { + "epoch": 0.72, + "learning_rate": 3.7565801917809796e-06, + "loss": 0.9764, + "step": 30072 + }, + { + "epoch": 0.72, + "learning_rate": 3.755971776481776e-06, + "loss": 1.4323, + "step": 30073 + }, + { + "epoch": 0.72, + "learning_rate": 3.7553633990636274e-06, + "loss": 1.0889, + "step": 30074 + }, + { + "epoch": 0.72, + "learning_rate": 3.7547550595302174e-06, + "loss": 0.9714, + "step": 30075 + }, + { + "epoch": 0.72, + "learning_rate": 3.754146757885242e-06, + "loss": 0.9039, + "step": 30076 + }, + { + "epoch": 0.72, + "learning_rate": 3.753538494132386e-06, + "loss": 1.0686, + "step": 30077 + }, + { + "epoch": 0.72, + "learning_rate": 3.7529302682753422e-06, + "loss": 1.0068, + "step": 30078 + }, + { + "epoch": 0.72, + "learning_rate": 3.752322080317805e-06, + "loss": 1.1817, + "step": 30079 + }, + { + "epoch": 0.72, + "learning_rate": 3.7517139302634564e-06, + "loss": 0.9443, + "step": 30080 + }, + { + "epoch": 0.72, + "learning_rate": 3.75110581811599e-06, + "loss": 0.8514, + "step": 30081 + }, + { + "epoch": 0.72, + "learning_rate": 3.750497743879098e-06, + "loss": 0.7879, + "step": 30082 + }, + { + "epoch": 0.72, + "learning_rate": 3.7498897075564633e-06, + "loss": 1.168, + "step": 30083 + }, + { + "epoch": 0.72, + "learning_rate": 3.749281709151781e-06, + "loss": 1.1786, + "step": 30084 + }, + { + "epoch": 0.72, + "learning_rate": 3.748673748668732e-06, + "loss": 0.8681, + "step": 30085 + }, + { + "epoch": 0.72, + "learning_rate": 3.7480658261110105e-06, + "loss": 1.2495, + "step": 30086 + }, + { + "epoch": 0.72, + "learning_rate": 3.7474579414823056e-06, + "loss": 1.2476, + "step": 30087 + }, + { + "epoch": 0.72, + "learning_rate": 3.7468500947862987e-06, + "loss": 0.7627, + "step": 30088 + }, + { + "epoch": 0.72, + "learning_rate": 3.746242286026682e-06, + "loss": 1.0316, + "step": 30089 + }, + { + "epoch": 0.72, + "learning_rate": 3.745634515207146e-06, + "loss": 1.0805, + "step": 30090 + }, + { + "epoch": 0.72, + "learning_rate": 3.74502678233137e-06, + "loss": 1.119, + "step": 30091 + }, + { + "epoch": 0.72, + "learning_rate": 3.7444190874030496e-06, + "loss": 1.1466, + "step": 30092 + }, + { + "epoch": 0.72, + "learning_rate": 3.7438114304258634e-06, + "loss": 1.1086, + "step": 30093 + }, + { + "epoch": 0.72, + "learning_rate": 3.743203811403502e-06, + "loss": 1.0817, + "step": 30094 + }, + { + "epoch": 0.72, + "learning_rate": 3.7425962303396548e-06, + "loss": 1.0581, + "step": 30095 + }, + { + "epoch": 0.72, + "learning_rate": 3.7419886872380007e-06, + "loss": 0.7671, + "step": 30096 + }, + { + "epoch": 0.72, + "learning_rate": 3.7413811821022284e-06, + "loss": 1.0695, + "step": 30097 + }, + { + "epoch": 0.72, + "learning_rate": 3.740773714936029e-06, + "loss": 1.1194, + "step": 30098 + }, + { + "epoch": 0.72, + "learning_rate": 3.7401662857430798e-06, + "loss": 1.1755, + "step": 30099 + }, + { + "epoch": 0.72, + "learning_rate": 3.739558894527072e-06, + "loss": 1.0715, + "step": 30100 + }, + { + "epoch": 0.72, + "learning_rate": 3.7389515412916855e-06, + "loss": 1.0456, + "step": 30101 + }, + { + "epoch": 0.72, + "learning_rate": 3.7383442260406067e-06, + "loss": 0.8621, + "step": 30102 + }, + { + "epoch": 0.72, + "learning_rate": 3.7377369487775237e-06, + "loss": 1.1164, + "step": 30103 + }, + { + "epoch": 0.72, + "learning_rate": 3.7371297095061143e-06, + "loss": 0.9152, + "step": 30104 + }, + { + "epoch": 0.72, + "learning_rate": 3.7365225082300673e-06, + "loss": 1.0205, + "step": 30105 + }, + { + "epoch": 0.72, + "learning_rate": 3.7359153449530673e-06, + "loss": 1.2383, + "step": 30106 + }, + { + "epoch": 0.72, + "learning_rate": 3.7353082196787914e-06, + "loss": 0.9385, + "step": 30107 + }, + { + "epoch": 0.72, + "learning_rate": 3.7347011324109307e-06, + "loss": 1.1309, + "step": 30108 + }, + { + "epoch": 0.72, + "learning_rate": 3.7340940831531615e-06, + "loss": 1.2102, + "step": 30109 + }, + { + "epoch": 0.72, + "learning_rate": 3.7334870719091697e-06, + "loss": 0.8993, + "step": 30110 + }, + { + "epoch": 0.72, + "learning_rate": 3.732880098682642e-06, + "loss": 0.9111, + "step": 30111 + }, + { + "epoch": 0.72, + "learning_rate": 3.7322731634772514e-06, + "loss": 1.1941, + "step": 30112 + }, + { + "epoch": 0.72, + "learning_rate": 3.7316662662966864e-06, + "loss": 1.2321, + "step": 30113 + }, + { + "epoch": 0.72, + "learning_rate": 3.7310594071446317e-06, + "loss": 1.1153, + "step": 30114 + }, + { + "epoch": 0.72, + "learning_rate": 3.7304525860247607e-06, + "loss": 1.1531, + "step": 30115 + }, + { + "epoch": 0.72, + "learning_rate": 3.729845802940759e-06, + "loss": 0.9575, + "step": 30116 + }, + { + "epoch": 0.72, + "learning_rate": 3.7292390578963124e-06, + "loss": 0.7218, + "step": 30117 + }, + { + "epoch": 0.72, + "learning_rate": 3.7286323508950928e-06, + "loss": 1.2609, + "step": 30118 + }, + { + "epoch": 0.72, + "learning_rate": 3.72802568194079e-06, + "loss": 1.12, + "step": 30119 + }, + { + "epoch": 0.72, + "learning_rate": 3.7274190510370766e-06, + "loss": 1.1068, + "step": 30120 + }, + { + "epoch": 0.72, + "learning_rate": 3.7268124581876366e-06, + "loss": 0.9928, + "step": 30121 + }, + { + "epoch": 0.72, + "learning_rate": 3.726205903396153e-06, + "loss": 1.0393, + "step": 30122 + }, + { + "epoch": 0.72, + "learning_rate": 3.7255993866662986e-06, + "loss": 1.0759, + "step": 30123 + }, + { + "epoch": 0.72, + "learning_rate": 3.7249929080017576e-06, + "loss": 0.9731, + "step": 30124 + }, + { + "epoch": 0.72, + "learning_rate": 3.724386467406211e-06, + "loss": 0.8622, + "step": 30125 + }, + { + "epoch": 0.72, + "learning_rate": 3.7237800648833333e-06, + "loss": 1.1395, + "step": 30126 + }, + { + "epoch": 0.72, + "learning_rate": 3.7231737004368086e-06, + "loss": 1.1619, + "step": 30127 + }, + { + "epoch": 0.72, + "learning_rate": 3.7225673740703094e-06, + "loss": 1.1402, + "step": 30128 + }, + { + "epoch": 0.72, + "learning_rate": 3.7219610857875177e-06, + "loss": 0.846, + "step": 30129 + }, + { + "epoch": 0.72, + "learning_rate": 3.721354835592115e-06, + "loss": 0.9709, + "step": 30130 + }, + { + "epoch": 0.72, + "learning_rate": 3.720748623487771e-06, + "loss": 1.1862, + "step": 30131 + }, + { + "epoch": 0.72, + "learning_rate": 3.720142449478169e-06, + "loss": 1.0058, + "step": 30132 + }, + { + "epoch": 0.72, + "learning_rate": 3.7195363135669895e-06, + "loss": 1.1501, + "step": 30133 + }, + { + "epoch": 0.72, + "learning_rate": 3.7189302157579023e-06, + "loss": 0.9397, + "step": 30134 + }, + { + "epoch": 0.72, + "learning_rate": 3.71832415605459e-06, + "loss": 1.0313, + "step": 30135 + }, + { + "epoch": 0.72, + "learning_rate": 3.717718134460726e-06, + "loss": 1.0992, + "step": 30136 + }, + { + "epoch": 0.72, + "learning_rate": 3.717112150979988e-06, + "loss": 1.2235, + "step": 30137 + }, + { + "epoch": 0.72, + "learning_rate": 3.716506205616055e-06, + "loss": 1.1152, + "step": 30138 + }, + { + "epoch": 0.72, + "learning_rate": 3.715900298372599e-06, + "loss": 0.9277, + "step": 30139 + }, + { + "epoch": 0.72, + "learning_rate": 3.7152944292532966e-06, + "loss": 1.1274, + "step": 30140 + }, + { + "epoch": 0.72, + "learning_rate": 3.7146885982618286e-06, + "loss": 0.9946, + "step": 30141 + }, + { + "epoch": 0.72, + "learning_rate": 3.714082805401863e-06, + "loss": 0.8975, + "step": 30142 + }, + { + "epoch": 0.72, + "learning_rate": 3.7134770506770822e-06, + "loss": 0.7852, + "step": 30143 + }, + { + "epoch": 0.72, + "learning_rate": 3.712871334091154e-06, + "loss": 0.977, + "step": 30144 + }, + { + "epoch": 0.72, + "learning_rate": 3.7122656556477554e-06, + "loss": 1.1434, + "step": 30145 + }, + { + "epoch": 0.72, + "learning_rate": 3.7116600153505665e-06, + "loss": 0.9046, + "step": 30146 + }, + { + "epoch": 0.73, + "learning_rate": 3.7110544132032545e-06, + "loss": 1.0172, + "step": 30147 + }, + { + "epoch": 0.73, + "learning_rate": 3.7104488492094947e-06, + "loss": 1.1032, + "step": 30148 + }, + { + "epoch": 0.73, + "learning_rate": 3.709843323372967e-06, + "loss": 1.1008, + "step": 30149 + }, + { + "epoch": 0.73, + "learning_rate": 3.7092378356973357e-06, + "loss": 1.1896, + "step": 30150 + }, + { + "epoch": 0.73, + "learning_rate": 3.7086323861862828e-06, + "loss": 1.1646, + "step": 30151 + }, + { + "epoch": 0.73, + "learning_rate": 3.708026974843474e-06, + "loss": 0.8417, + "step": 30152 + }, + { + "epoch": 0.73, + "learning_rate": 3.7074216016725863e-06, + "loss": 1.0163, + "step": 30153 + }, + { + "epoch": 0.73, + "learning_rate": 3.7068162666772945e-06, + "loss": 0.9819, + "step": 30154 + }, + { + "epoch": 0.73, + "learning_rate": 3.7062109698612636e-06, + "loss": 1.0325, + "step": 30155 + }, + { + "epoch": 0.73, + "learning_rate": 3.705605711228172e-06, + "loss": 1.225, + "step": 30156 + }, + { + "epoch": 0.73, + "learning_rate": 3.705000490781693e-06, + "loss": 0.8948, + "step": 30157 + }, + { + "epoch": 0.73, + "learning_rate": 3.704395308525491e-06, + "loss": 1.0657, + "step": 30158 + }, + { + "epoch": 0.73, + "learning_rate": 3.7037901644632422e-06, + "loss": 1.0635, + "step": 30159 + }, + { + "epoch": 0.73, + "learning_rate": 3.7031850585986217e-06, + "loss": 1.1109, + "step": 30160 + }, + { + "epoch": 0.73, + "learning_rate": 3.702579990935292e-06, + "loss": 1.0095, + "step": 30161 + }, + { + "epoch": 0.73, + "learning_rate": 3.701974961476933e-06, + "loss": 1.3092, + "step": 30162 + }, + { + "epoch": 0.73, + "learning_rate": 3.7013699702272055e-06, + "loss": 0.9072, + "step": 30163 + }, + { + "epoch": 0.73, + "learning_rate": 3.7007650171897856e-06, + "loss": 0.9791, + "step": 30164 + }, + { + "epoch": 0.73, + "learning_rate": 3.7001601023683463e-06, + "loss": 1.0019, + "step": 30165 + }, + { + "epoch": 0.73, + "learning_rate": 3.699555225766547e-06, + "loss": 0.8584, + "step": 30166 + }, + { + "epoch": 0.73, + "learning_rate": 3.6989503873880705e-06, + "loss": 0.9754, + "step": 30167 + }, + { + "epoch": 0.73, + "learning_rate": 3.698345587236577e-06, + "loss": 0.9377, + "step": 30168 + }, + { + "epoch": 0.73, + "learning_rate": 3.6977408253157387e-06, + "loss": 1.0999, + "step": 30169 + }, + { + "epoch": 0.73, + "learning_rate": 3.6971361016292274e-06, + "loss": 0.9701, + "step": 30170 + }, + { + "epoch": 0.73, + "learning_rate": 3.6965314161807053e-06, + "loss": 1.237, + "step": 30171 + }, + { + "epoch": 0.73, + "learning_rate": 3.6959267689738488e-06, + "loss": 1.3315, + "step": 30172 + }, + { + "epoch": 0.73, + "learning_rate": 3.695322160012318e-06, + "loss": 0.9127, + "step": 30173 + }, + { + "epoch": 0.73, + "learning_rate": 3.6947175892997846e-06, + "loss": 0.9213, + "step": 30174 + }, + { + "epoch": 0.73, + "learning_rate": 3.6941130568399186e-06, + "loss": 1.1729, + "step": 30175 + }, + { + "epoch": 0.73, + "learning_rate": 3.6935085626363833e-06, + "loss": 1.2088, + "step": 30176 + }, + { + "epoch": 0.73, + "learning_rate": 3.6929041066928485e-06, + "loss": 1.0079, + "step": 30177 + }, + { + "epoch": 0.73, + "learning_rate": 3.6922996890129836e-06, + "loss": 1.1011, + "step": 30178 + }, + { + "epoch": 0.73, + "learning_rate": 3.6916953096004506e-06, + "loss": 0.8292, + "step": 30179 + }, + { + "epoch": 0.73, + "learning_rate": 3.6910909684589203e-06, + "loss": 0.907, + "step": 30180 + }, + { + "epoch": 0.73, + "learning_rate": 3.6904866655920547e-06, + "loss": 1.1773, + "step": 30181 + }, + { + "epoch": 0.73, + "learning_rate": 3.6898824010035216e-06, + "loss": 1.1745, + "step": 30182 + }, + { + "epoch": 0.73, + "learning_rate": 3.689278174696992e-06, + "loss": 0.8612, + "step": 30183 + }, + { + "epoch": 0.73, + "learning_rate": 3.6886739866761235e-06, + "loss": 0.912, + "step": 30184 + }, + { + "epoch": 0.73, + "learning_rate": 3.6880698369445846e-06, + "loss": 1.1246, + "step": 30185 + }, + { + "epoch": 0.73, + "learning_rate": 3.6874657255060466e-06, + "loss": 0.9059, + "step": 30186 + }, + { + "epoch": 0.73, + "learning_rate": 3.686861652364164e-06, + "loss": 1.1715, + "step": 30187 + }, + { + "epoch": 0.73, + "learning_rate": 3.6862576175226118e-06, + "loss": 1.2139, + "step": 30188 + }, + { + "epoch": 0.73, + "learning_rate": 3.6856536209850447e-06, + "loss": 1.2092, + "step": 30189 + }, + { + "epoch": 0.73, + "learning_rate": 3.6850496627551322e-06, + "loss": 1.039, + "step": 30190 + }, + { + "epoch": 0.73, + "learning_rate": 3.684445742836542e-06, + "loss": 0.9443, + "step": 30191 + }, + { + "epoch": 0.73, + "learning_rate": 3.6838418612329306e-06, + "loss": 1.0095, + "step": 30192 + }, + { + "epoch": 0.73, + "learning_rate": 3.6832380179479644e-06, + "loss": 1.2764, + "step": 30193 + }, + { + "epoch": 0.73, + "learning_rate": 3.6826342129853108e-06, + "loss": 1.0078, + "step": 30194 + }, + { + "epoch": 0.73, + "learning_rate": 3.6820304463486267e-06, + "loss": 1.0147, + "step": 30195 + }, + { + "epoch": 0.73, + "learning_rate": 3.6814267180415776e-06, + "loss": 1.1463, + "step": 30196 + }, + { + "epoch": 0.73, + "learning_rate": 3.680823028067829e-06, + "loss": 1.0823, + "step": 30197 + }, + { + "epoch": 0.73, + "learning_rate": 3.680219376431039e-06, + "loss": 1.096, + "step": 30198 + }, + { + "epoch": 0.73, + "learning_rate": 3.679615763134874e-06, + "loss": 1.1323, + "step": 30199 + }, + { + "epoch": 0.73, + "learning_rate": 3.67901218818299e-06, + "loss": 1.145, + "step": 30200 + }, + { + "epoch": 0.73, + "learning_rate": 3.678408651579052e-06, + "loss": 1.0653, + "step": 30201 + }, + { + "epoch": 0.73, + "learning_rate": 3.6778051533267257e-06, + "loss": 1.0684, + "step": 30202 + }, + { + "epoch": 0.73, + "learning_rate": 3.677201693429665e-06, + "loss": 1.1035, + "step": 30203 + }, + { + "epoch": 0.73, + "learning_rate": 3.676598271891534e-06, + "loss": 0.9422, + "step": 30204 + }, + { + "epoch": 0.73, + "learning_rate": 3.675994888715998e-06, + "loss": 1.0891, + "step": 30205 + }, + { + "epoch": 0.73, + "learning_rate": 3.675391543906709e-06, + "loss": 0.9145, + "step": 30206 + }, + { + "epoch": 0.73, + "learning_rate": 3.674788237467336e-06, + "loss": 1.0391, + "step": 30207 + }, + { + "epoch": 0.73, + "learning_rate": 3.674184969401532e-06, + "loss": 1.0203, + "step": 30208 + }, + { + "epoch": 0.73, + "learning_rate": 3.6735817397129593e-06, + "loss": 1.0028, + "step": 30209 + }, + { + "epoch": 0.73, + "learning_rate": 3.6729785484052815e-06, + "loss": 0.926, + "step": 30210 + }, + { + "epoch": 0.73, + "learning_rate": 3.672375395482152e-06, + "loss": 1.1031, + "step": 30211 + }, + { + "epoch": 0.73, + "learning_rate": 3.671772280947231e-06, + "loss": 0.9453, + "step": 30212 + }, + { + "epoch": 0.73, + "learning_rate": 3.6711692048041834e-06, + "loss": 1.2718, + "step": 30213 + }, + { + "epoch": 0.73, + "learning_rate": 3.6705661670566604e-06, + "loss": 0.9459, + "step": 30214 + }, + { + "epoch": 0.73, + "learning_rate": 3.669963167708327e-06, + "loss": 0.8291, + "step": 30215 + }, + { + "epoch": 0.73, + "learning_rate": 3.669360206762835e-06, + "loss": 1.0053, + "step": 30216 + }, + { + "epoch": 0.73, + "learning_rate": 3.668757284223845e-06, + "loss": 1.0439, + "step": 30217 + }, + { + "epoch": 0.73, + "learning_rate": 3.66815440009502e-06, + "loss": 1.0697, + "step": 30218 + }, + { + "epoch": 0.73, + "learning_rate": 3.667551554380008e-06, + "loss": 1.0297, + "step": 30219 + }, + { + "epoch": 0.73, + "learning_rate": 3.6669487470824717e-06, + "loss": 1.0343, + "step": 30220 + }, + { + "epoch": 0.73, + "learning_rate": 3.6663459782060718e-06, + "loss": 0.9066, + "step": 30221 + }, + { + "epoch": 0.73, + "learning_rate": 3.6657432477544562e-06, + "loss": 1.2703, + "step": 30222 + }, + { + "epoch": 0.73, + "learning_rate": 3.6651405557312915e-06, + "loss": 0.99, + "step": 30223 + }, + { + "epoch": 0.73, + "learning_rate": 3.664537902140225e-06, + "loss": 0.9732, + "step": 30224 + }, + { + "epoch": 0.73, + "learning_rate": 3.6639352869849164e-06, + "loss": 0.8637, + "step": 30225 + }, + { + "epoch": 0.73, + "learning_rate": 3.6633327102690253e-06, + "loss": 1.1367, + "step": 30226 + }, + { + "epoch": 0.73, + "learning_rate": 3.6627301719962018e-06, + "loss": 0.8087, + "step": 30227 + }, + { + "epoch": 0.73, + "learning_rate": 3.6621276721701025e-06, + "loss": 0.9448, + "step": 30228 + }, + { + "epoch": 0.73, + "learning_rate": 3.661525210794388e-06, + "loss": 1.105, + "step": 30229 + }, + { + "epoch": 0.73, + "learning_rate": 3.6609227878727062e-06, + "loss": 1.1482, + "step": 30230 + }, + { + "epoch": 0.73, + "learning_rate": 3.660320403408717e-06, + "loss": 1.0676, + "step": 30231 + }, + { + "epoch": 0.73, + "learning_rate": 3.6597180574060698e-06, + "loss": 1.1404, + "step": 30232 + }, + { + "epoch": 0.73, + "learning_rate": 3.6591157498684227e-06, + "loss": 0.9695, + "step": 30233 + }, + { + "epoch": 0.73, + "learning_rate": 3.658513480799432e-06, + "loss": 1.0649, + "step": 30234 + }, + { + "epoch": 0.73, + "learning_rate": 3.657911250202744e-06, + "loss": 0.997, + "step": 30235 + }, + { + "epoch": 0.73, + "learning_rate": 3.657309058082017e-06, + "loss": 0.9448, + "step": 30236 + }, + { + "epoch": 0.73, + "learning_rate": 3.6567069044409075e-06, + "loss": 1.2596, + "step": 30237 + }, + { + "epoch": 0.73, + "learning_rate": 3.6561047892830627e-06, + "loss": 1.0563, + "step": 30238 + }, + { + "epoch": 0.73, + "learning_rate": 3.655502712612137e-06, + "loss": 0.9971, + "step": 30239 + }, + { + "epoch": 0.73, + "learning_rate": 3.6549006744317882e-06, + "loss": 1.1493, + "step": 30240 + }, + { + "epoch": 0.73, + "learning_rate": 3.65429867474566e-06, + "loss": 0.9909, + "step": 30241 + }, + { + "epoch": 0.73, + "learning_rate": 3.653696713557413e-06, + "loss": 0.8154, + "step": 30242 + }, + { + "epoch": 0.73, + "learning_rate": 3.6530947908706924e-06, + "loss": 0.8373, + "step": 30243 + }, + { + "epoch": 0.73, + "learning_rate": 3.6524929066891524e-06, + "loss": 0.7682, + "step": 30244 + }, + { + "epoch": 0.73, + "learning_rate": 3.651891061016449e-06, + "loss": 1.072, + "step": 30245 + }, + { + "epoch": 0.73, + "learning_rate": 3.651289253856225e-06, + "loss": 1.2376, + "step": 30246 + }, + { + "epoch": 0.73, + "learning_rate": 3.6506874852121367e-06, + "loss": 1.1544, + "step": 30247 + }, + { + "epoch": 0.73, + "learning_rate": 3.6500857550878366e-06, + "loss": 1.0817, + "step": 30248 + }, + { + "epoch": 0.73, + "learning_rate": 3.64948406348697e-06, + "loss": 1.1158, + "step": 30249 + }, + { + "epoch": 0.73, + "learning_rate": 3.6488824104131927e-06, + "loss": 0.9926, + "step": 30250 + }, + { + "epoch": 0.73, + "learning_rate": 3.6482807958701482e-06, + "loss": 1.0211, + "step": 30251 + }, + { + "epoch": 0.73, + "learning_rate": 3.647679219861491e-06, + "loss": 0.8793, + "step": 30252 + }, + { + "epoch": 0.73, + "learning_rate": 3.6470776823908717e-06, + "loss": 0.8969, + "step": 30253 + }, + { + "epoch": 0.73, + "learning_rate": 3.6464761834619356e-06, + "loss": 1.2259, + "step": 30254 + }, + { + "epoch": 0.73, + "learning_rate": 3.645874723078334e-06, + "loss": 1.0567, + "step": 30255 + }, + { + "epoch": 0.73, + "learning_rate": 3.645273301243719e-06, + "loss": 1.1786, + "step": 30256 + }, + { + "epoch": 0.73, + "learning_rate": 3.6446719179617328e-06, + "loss": 1.098, + "step": 30257 + }, + { + "epoch": 0.73, + "learning_rate": 3.6440705732360315e-06, + "loss": 1.0215, + "step": 30258 + }, + { + "epoch": 0.73, + "learning_rate": 3.643469267070254e-06, + "loss": 1.2075, + "step": 30259 + }, + { + "epoch": 0.73, + "learning_rate": 3.642867999468055e-06, + "loss": 1.0057, + "step": 30260 + }, + { + "epoch": 0.73, + "learning_rate": 3.6422667704330827e-06, + "loss": 1.1308, + "step": 30261 + }, + { + "epoch": 0.73, + "learning_rate": 3.6416655799689794e-06, + "loss": 0.951, + "step": 30262 + }, + { + "epoch": 0.73, + "learning_rate": 3.641064428079395e-06, + "loss": 0.8961, + "step": 30263 + }, + { + "epoch": 0.73, + "learning_rate": 3.6404633147679813e-06, + "loss": 0.8132, + "step": 30264 + }, + { + "epoch": 0.73, + "learning_rate": 3.6398622400383777e-06, + "loss": 1.0949, + "step": 30265 + }, + { + "epoch": 0.73, + "learning_rate": 3.6392612038942364e-06, + "loss": 1.233, + "step": 30266 + }, + { + "epoch": 0.73, + "learning_rate": 3.638660206339197e-06, + "loss": 0.9377, + "step": 30267 + }, + { + "epoch": 0.73, + "learning_rate": 3.6380592473769115e-06, + "loss": 0.9305, + "step": 30268 + }, + { + "epoch": 0.73, + "learning_rate": 3.6374583270110276e-06, + "loss": 0.7896, + "step": 30269 + }, + { + "epoch": 0.73, + "learning_rate": 3.6368574452451834e-06, + "loss": 1.0695, + "step": 30270 + }, + { + "epoch": 0.73, + "learning_rate": 3.636256602083028e-06, + "loss": 1.0158, + "step": 30271 + }, + { + "epoch": 0.73, + "learning_rate": 3.6356557975282104e-06, + "loss": 1.0528, + "step": 30272 + }, + { + "epoch": 0.73, + "learning_rate": 3.635055031584369e-06, + "loss": 1.3044, + "step": 30273 + }, + { + "epoch": 0.73, + "learning_rate": 3.634454304255155e-06, + "loss": 0.8898, + "step": 30274 + }, + { + "epoch": 0.73, + "learning_rate": 3.633853615544206e-06, + "loss": 0.9757, + "step": 30275 + }, + { + "epoch": 0.73, + "learning_rate": 3.633252965455171e-06, + "loss": 1.2936, + "step": 30276 + }, + { + "epoch": 0.73, + "learning_rate": 3.6326523539916946e-06, + "loss": 1.095, + "step": 30277 + }, + { + "epoch": 0.73, + "learning_rate": 3.6320517811574154e-06, + "loss": 0.9036, + "step": 30278 + }, + { + "epoch": 0.73, + "learning_rate": 3.6314512469559803e-06, + "loss": 0.9402, + "step": 30279 + }, + { + "epoch": 0.73, + "learning_rate": 3.630850751391036e-06, + "loss": 1.448, + "step": 30280 + }, + { + "epoch": 0.73, + "learning_rate": 3.630250294466219e-06, + "loss": 1.0023, + "step": 30281 + }, + { + "epoch": 0.73, + "learning_rate": 3.629649876185175e-06, + "loss": 1.0835, + "step": 30282 + }, + { + "epoch": 0.73, + "learning_rate": 3.62904949655155e-06, + "loss": 1.009, + "step": 30283 + }, + { + "epoch": 0.73, + "learning_rate": 3.6284491555689795e-06, + "loss": 1.0877, + "step": 30284 + }, + { + "epoch": 0.73, + "learning_rate": 3.627848853241113e-06, + "loss": 0.8733, + "step": 30285 + }, + { + "epoch": 0.73, + "learning_rate": 3.627248589571586e-06, + "loss": 1.0245, + "step": 30286 + }, + { + "epoch": 0.73, + "learning_rate": 3.6266483645640416e-06, + "loss": 1.0128, + "step": 30287 + }, + { + "epoch": 0.73, + "learning_rate": 3.626048178222127e-06, + "loss": 0.934, + "step": 30288 + }, + { + "epoch": 0.73, + "learning_rate": 3.6254480305494745e-06, + "loss": 1.047, + "step": 30289 + }, + { + "epoch": 0.73, + "learning_rate": 3.62484792154973e-06, + "loss": 1.0525, + "step": 30290 + }, + { + "epoch": 0.73, + "learning_rate": 3.624247851226537e-06, + "loss": 1.3252, + "step": 30291 + }, + { + "epoch": 0.73, + "learning_rate": 3.623647819583529e-06, + "loss": 1.02, + "step": 30292 + }, + { + "epoch": 0.73, + "learning_rate": 3.623047826624353e-06, + "loss": 1.0641, + "step": 30293 + }, + { + "epoch": 0.73, + "learning_rate": 3.6224478723526425e-06, + "loss": 1.024, + "step": 30294 + }, + { + "epoch": 0.73, + "learning_rate": 3.621847956772041e-06, + "loss": 0.6304, + "step": 30295 + }, + { + "epoch": 0.73, + "learning_rate": 3.6212480798861915e-06, + "loss": 1.0847, + "step": 30296 + }, + { + "epoch": 0.73, + "learning_rate": 3.6206482416987256e-06, + "loss": 0.9999, + "step": 30297 + }, + { + "epoch": 0.73, + "learning_rate": 3.6200484422132865e-06, + "loss": 1.0563, + "step": 30298 + }, + { + "epoch": 0.73, + "learning_rate": 3.619448681433516e-06, + "loss": 0.9655, + "step": 30299 + }, + { + "epoch": 0.73, + "learning_rate": 3.618848959363046e-06, + "loss": 1.037, + "step": 30300 + }, + { + "epoch": 0.73, + "learning_rate": 3.618249276005523e-06, + "loss": 1.2441, + "step": 30301 + }, + { + "epoch": 0.73, + "learning_rate": 3.6176496313645757e-06, + "loss": 1.3803, + "step": 30302 + }, + { + "epoch": 0.73, + "learning_rate": 3.6170500254438477e-06, + "loss": 1.0256, + "step": 30303 + }, + { + "epoch": 0.73, + "learning_rate": 3.6164504582469796e-06, + "loss": 1.2437, + "step": 30304 + }, + { + "epoch": 0.73, + "learning_rate": 3.6158509297776014e-06, + "loss": 0.8748, + "step": 30305 + }, + { + "epoch": 0.73, + "learning_rate": 3.6152514400393535e-06, + "loss": 1.3148, + "step": 30306 + }, + { + "epoch": 0.73, + "learning_rate": 3.6146519890358766e-06, + "loss": 1.1723, + "step": 30307 + }, + { + "epoch": 0.73, + "learning_rate": 3.6140525767708014e-06, + "loss": 1.1617, + "step": 30308 + }, + { + "epoch": 0.73, + "learning_rate": 3.61345320324777e-06, + "loss": 1.1833, + "step": 30309 + }, + { + "epoch": 0.73, + "learning_rate": 3.6128538684704117e-06, + "loss": 1.0278, + "step": 30310 + }, + { + "epoch": 0.73, + "learning_rate": 3.6122545724423685e-06, + "loss": 0.9849, + "step": 30311 + }, + { + "epoch": 0.73, + "learning_rate": 3.611655315167276e-06, + "loss": 0.8953, + "step": 30312 + }, + { + "epoch": 0.73, + "learning_rate": 3.6110560966487663e-06, + "loss": 1.1568, + "step": 30313 + }, + { + "epoch": 0.73, + "learning_rate": 3.6104569168904747e-06, + "loss": 0.912, + "step": 30314 + }, + { + "epoch": 0.73, + "learning_rate": 3.609857775896043e-06, + "loss": 1.0607, + "step": 30315 + }, + { + "epoch": 0.73, + "learning_rate": 3.609258673669097e-06, + "loss": 1.0682, + "step": 30316 + }, + { + "epoch": 0.73, + "learning_rate": 3.6086596102132788e-06, + "loss": 0.8956, + "step": 30317 + }, + { + "epoch": 0.73, + "learning_rate": 3.608060585532216e-06, + "loss": 1.1438, + "step": 30318 + }, + { + "epoch": 0.73, + "learning_rate": 3.6074615996295457e-06, + "loss": 0.9316, + "step": 30319 + }, + { + "epoch": 0.73, + "learning_rate": 3.606862652508907e-06, + "loss": 1.2257, + "step": 30320 + }, + { + "epoch": 0.73, + "learning_rate": 3.6062637441739245e-06, + "loss": 1.0715, + "step": 30321 + }, + { + "epoch": 0.73, + "learning_rate": 3.605664874628236e-06, + "loss": 0.8745, + "step": 30322 + }, + { + "epoch": 0.73, + "learning_rate": 3.605066043875478e-06, + "loss": 0.7553, + "step": 30323 + }, + { + "epoch": 0.73, + "learning_rate": 3.604467251919278e-06, + "loss": 1.0245, + "step": 30324 + }, + { + "epoch": 0.73, + "learning_rate": 3.6038684987632735e-06, + "loss": 1.0771, + "step": 30325 + }, + { + "epoch": 0.73, + "learning_rate": 3.6032697844110896e-06, + "loss": 0.9317, + "step": 30326 + }, + { + "epoch": 0.73, + "learning_rate": 3.6026711088663644e-06, + "loss": 1.1238, + "step": 30327 + }, + { + "epoch": 0.73, + "learning_rate": 3.602072472132733e-06, + "loss": 1.0953, + "step": 30328 + }, + { + "epoch": 0.73, + "learning_rate": 3.6014738742138187e-06, + "loss": 1.0736, + "step": 30329 + }, + { + "epoch": 0.73, + "learning_rate": 3.600875315113257e-06, + "loss": 0.9696, + "step": 30330 + }, + { + "epoch": 0.73, + "learning_rate": 3.600276794834684e-06, + "loss": 1.0809, + "step": 30331 + }, + { + "epoch": 0.73, + "learning_rate": 3.5996783133817192e-06, + "loss": 1.0966, + "step": 30332 + }, + { + "epoch": 0.73, + "learning_rate": 3.5990798707580064e-06, + "loss": 0.9108, + "step": 30333 + }, + { + "epoch": 0.73, + "learning_rate": 3.5984814669671687e-06, + "loss": 0.9368, + "step": 30334 + }, + { + "epoch": 0.73, + "learning_rate": 3.597883102012837e-06, + "loss": 1.2234, + "step": 30335 + }, + { + "epoch": 0.73, + "learning_rate": 3.5972847758986475e-06, + "loss": 1.0135, + "step": 30336 + }, + { + "epoch": 0.73, + "learning_rate": 3.5966864886282205e-06, + "loss": 1.226, + "step": 30337 + }, + { + "epoch": 0.73, + "learning_rate": 3.5960882402051945e-06, + "loss": 1.0485, + "step": 30338 + }, + { + "epoch": 0.73, + "learning_rate": 3.5954900306331908e-06, + "loss": 1.3379, + "step": 30339 + }, + { + "epoch": 0.73, + "learning_rate": 3.5948918599158432e-06, + "loss": 0.885, + "step": 30340 + }, + { + "epoch": 0.73, + "learning_rate": 3.5942937280567823e-06, + "loss": 1.0546, + "step": 30341 + }, + { + "epoch": 0.73, + "learning_rate": 3.593695635059633e-06, + "loss": 1.1685, + "step": 30342 + }, + { + "epoch": 0.73, + "learning_rate": 3.5930975809280246e-06, + "loss": 0.9057, + "step": 30343 + }, + { + "epoch": 0.73, + "learning_rate": 3.592499565665589e-06, + "loss": 1.0259, + "step": 30344 + }, + { + "epoch": 0.73, + "learning_rate": 3.591901589275949e-06, + "loss": 1.1253, + "step": 30345 + }, + { + "epoch": 0.73, + "learning_rate": 3.591303651762739e-06, + "loss": 1.2721, + "step": 30346 + }, + { + "epoch": 0.73, + "learning_rate": 3.5907057531295776e-06, + "loss": 1.2974, + "step": 30347 + }, + { + "epoch": 0.73, + "learning_rate": 3.5901078933800957e-06, + "loss": 0.975, + "step": 30348 + }, + { + "epoch": 0.73, + "learning_rate": 3.5895100725179266e-06, + "loss": 1.009, + "step": 30349 + }, + { + "epoch": 0.73, + "learning_rate": 3.588912290546689e-06, + "loss": 0.879, + "step": 30350 + }, + { + "epoch": 0.73, + "learning_rate": 3.588314547470011e-06, + "loss": 0.9403, + "step": 30351 + }, + { + "epoch": 0.73, + "learning_rate": 3.5877168432915256e-06, + "loss": 1.3465, + "step": 30352 + }, + { + "epoch": 0.73, + "learning_rate": 3.5871191780148496e-06, + "loss": 1.2912, + "step": 30353 + }, + { + "epoch": 0.73, + "learning_rate": 3.5865215516436157e-06, + "loss": 0.9509, + "step": 30354 + }, + { + "epoch": 0.73, + "learning_rate": 3.5859239641814446e-06, + "loss": 0.973, + "step": 30355 + }, + { + "epoch": 0.73, + "learning_rate": 3.5853264156319645e-06, + "loss": 1.0131, + "step": 30356 + }, + { + "epoch": 0.73, + "learning_rate": 3.5847289059988033e-06, + "loss": 1.2948, + "step": 30357 + }, + { + "epoch": 0.73, + "learning_rate": 3.584131435285579e-06, + "loss": 0.8917, + "step": 30358 + }, + { + "epoch": 0.73, + "learning_rate": 3.58353400349592e-06, + "loss": 0.9166, + "step": 30359 + }, + { + "epoch": 0.73, + "learning_rate": 3.582936610633455e-06, + "loss": 1.0913, + "step": 30360 + }, + { + "epoch": 0.73, + "learning_rate": 3.582339256701801e-06, + "loss": 1.219, + "step": 30361 + }, + { + "epoch": 0.73, + "learning_rate": 3.581741941704584e-06, + "loss": 1.1492, + "step": 30362 + }, + { + "epoch": 0.73, + "learning_rate": 3.581144665645434e-06, + "loss": 1.2096, + "step": 30363 + }, + { + "epoch": 0.73, + "learning_rate": 3.580547428527965e-06, + "loss": 1.0303, + "step": 30364 + }, + { + "epoch": 0.73, + "learning_rate": 3.5799502303558086e-06, + "loss": 1.1577, + "step": 30365 + }, + { + "epoch": 0.73, + "learning_rate": 3.5793530711325807e-06, + "loss": 0.9095, + "step": 30366 + }, + { + "epoch": 0.73, + "learning_rate": 3.5787559508619073e-06, + "loss": 1.0726, + "step": 30367 + }, + { + "epoch": 0.73, + "learning_rate": 3.5781588695474146e-06, + "loss": 1.1332, + "step": 30368 + }, + { + "epoch": 0.73, + "learning_rate": 3.577561827192718e-06, + "loss": 0.9591, + "step": 30369 + }, + { + "epoch": 0.73, + "learning_rate": 3.5769648238014433e-06, + "loss": 1.1372, + "step": 30370 + }, + { + "epoch": 0.73, + "learning_rate": 3.5763678593772157e-06, + "loss": 0.9485, + "step": 30371 + }, + { + "epoch": 0.73, + "learning_rate": 3.5757709339236492e-06, + "loss": 1.0226, + "step": 30372 + }, + { + "epoch": 0.73, + "learning_rate": 3.5751740474443743e-06, + "loss": 1.0086, + "step": 30373 + }, + { + "epoch": 0.73, + "learning_rate": 3.5745771999430022e-06, + "loss": 0.9466, + "step": 30374 + }, + { + "epoch": 0.73, + "learning_rate": 3.5739803914231596e-06, + "loss": 0.9913, + "step": 30375 + }, + { + "epoch": 0.73, + "learning_rate": 3.57338362188847e-06, + "loss": 1.1356, + "step": 30376 + }, + { + "epoch": 0.73, + "learning_rate": 3.572786891342547e-06, + "loss": 1.0894, + "step": 30377 + }, + { + "epoch": 0.73, + "learning_rate": 3.572190199789014e-06, + "loss": 1.1261, + "step": 30378 + }, + { + "epoch": 0.73, + "learning_rate": 3.5715935472314943e-06, + "loss": 0.9913, + "step": 30379 + }, + { + "epoch": 0.73, + "learning_rate": 3.5709969336736017e-06, + "loss": 1.1239, + "step": 30380 + }, + { + "epoch": 0.73, + "learning_rate": 3.570400359118962e-06, + "loss": 1.2233, + "step": 30381 + }, + { + "epoch": 0.73, + "learning_rate": 3.5698038235711875e-06, + "loss": 0.9734, + "step": 30382 + }, + { + "epoch": 0.73, + "learning_rate": 3.5692073270339e-06, + "loss": 1.0316, + "step": 30383 + }, + { + "epoch": 0.73, + "learning_rate": 3.5686108695107223e-06, + "loss": 0.9119, + "step": 30384 + }, + { + "epoch": 0.73, + "learning_rate": 3.5680144510052673e-06, + "loss": 0.8794, + "step": 30385 + }, + { + "epoch": 0.73, + "learning_rate": 3.5674180715211548e-06, + "loss": 1.0771, + "step": 30386 + }, + { + "epoch": 0.73, + "learning_rate": 3.566821731062008e-06, + "loss": 0.8896, + "step": 30387 + }, + { + "epoch": 0.73, + "learning_rate": 3.566225429631437e-06, + "loss": 1.1936, + "step": 30388 + }, + { + "epoch": 0.73, + "learning_rate": 3.5656291672330656e-06, + "loss": 0.91, + "step": 30389 + }, + { + "epoch": 0.73, + "learning_rate": 3.5650329438705067e-06, + "loss": 1.0146, + "step": 30390 + }, + { + "epoch": 0.73, + "learning_rate": 3.5644367595473783e-06, + "loss": 1.0846, + "step": 30391 + }, + { + "epoch": 0.73, + "learning_rate": 3.5638406142673022e-06, + "loss": 0.9173, + "step": 30392 + }, + { + "epoch": 0.73, + "learning_rate": 3.5632445080338874e-06, + "loss": 1.1116, + "step": 30393 + }, + { + "epoch": 0.73, + "learning_rate": 3.5626484408507543e-06, + "loss": 1.0627, + "step": 30394 + }, + { + "epoch": 0.73, + "learning_rate": 3.5620524127215216e-06, + "loss": 1.1494, + "step": 30395 + }, + { + "epoch": 0.73, + "learning_rate": 3.5614564236498006e-06, + "loss": 1.2913, + "step": 30396 + }, + { + "epoch": 0.73, + "learning_rate": 3.5608604736392115e-06, + "loss": 1.0413, + "step": 30397 + }, + { + "epoch": 0.73, + "learning_rate": 3.5602645626933642e-06, + "loss": 1.1611, + "step": 30398 + }, + { + "epoch": 0.73, + "learning_rate": 3.559668690815877e-06, + "loss": 1.1066, + "step": 30399 + }, + { + "epoch": 0.73, + "learning_rate": 3.5590728580103685e-06, + "loss": 0.9812, + "step": 30400 + }, + { + "epoch": 0.73, + "learning_rate": 3.558477064280447e-06, + "loss": 1.1357, + "step": 30401 + }, + { + "epoch": 0.73, + "learning_rate": 3.5578813096297293e-06, + "loss": 1.0344, + "step": 30402 + }, + { + "epoch": 0.73, + "learning_rate": 3.5572855940618345e-06, + "loss": 0.9744, + "step": 30403 + }, + { + "epoch": 0.73, + "learning_rate": 3.5566899175803683e-06, + "loss": 0.9786, + "step": 30404 + }, + { + "epoch": 0.73, + "learning_rate": 3.5560942801889533e-06, + "loss": 0.9853, + "step": 30405 + }, + { + "epoch": 0.73, + "learning_rate": 3.555498681891194e-06, + "loss": 1.0213, + "step": 30406 + }, + { + "epoch": 0.73, + "learning_rate": 3.5549031226907094e-06, + "loss": 0.8643, + "step": 30407 + }, + { + "epoch": 0.73, + "learning_rate": 3.5543076025911136e-06, + "loss": 0.8685, + "step": 30408 + }, + { + "epoch": 0.73, + "learning_rate": 3.5537121215960158e-06, + "loss": 1.0775, + "step": 30409 + }, + { + "epoch": 0.73, + "learning_rate": 3.553116679709029e-06, + "loss": 1.1813, + "step": 30410 + }, + { + "epoch": 0.73, + "learning_rate": 3.552521276933771e-06, + "loss": 0.9568, + "step": 30411 + }, + { + "epoch": 0.73, + "learning_rate": 3.551925913273846e-06, + "loss": 0.9714, + "step": 30412 + }, + { + "epoch": 0.73, + "learning_rate": 3.551330588732871e-06, + "loss": 1.0023, + "step": 30413 + }, + { + "epoch": 0.73, + "learning_rate": 3.550735303314459e-06, + "loss": 1.0257, + "step": 30414 + }, + { + "epoch": 0.73, + "learning_rate": 3.5501400570222143e-06, + "loss": 1.2515, + "step": 30415 + }, + { + "epoch": 0.73, + "learning_rate": 3.5495448498597584e-06, + "loss": 1.1853, + "step": 30416 + }, + { + "epoch": 0.73, + "learning_rate": 3.548949681830692e-06, + "loss": 1.1089, + "step": 30417 + }, + { + "epoch": 0.73, + "learning_rate": 3.5483545529386307e-06, + "loss": 1.1339, + "step": 30418 + }, + { + "epoch": 0.73, + "learning_rate": 3.547759463187188e-06, + "loss": 1.0393, + "step": 30419 + }, + { + "epoch": 0.73, + "learning_rate": 3.5471644125799675e-06, + "loss": 1.1415, + "step": 30420 + }, + { + "epoch": 0.73, + "learning_rate": 3.546569401120583e-06, + "loss": 1.2413, + "step": 30421 + }, + { + "epoch": 0.73, + "learning_rate": 3.5459744288126472e-06, + "loss": 0.93, + "step": 30422 + }, + { + "epoch": 0.73, + "learning_rate": 3.5453794956597644e-06, + "loss": 1.1127, + "step": 30423 + }, + { + "epoch": 0.73, + "learning_rate": 3.5447846016655475e-06, + "loss": 0.8046, + "step": 30424 + }, + { + "epoch": 0.73, + "learning_rate": 3.5441897468336017e-06, + "loss": 1.1527, + "step": 30425 + }, + { + "epoch": 0.73, + "learning_rate": 3.543594931167538e-06, + "loss": 1.0542, + "step": 30426 + }, + { + "epoch": 0.73, + "learning_rate": 3.5430001546709692e-06, + "loss": 1.0348, + "step": 30427 + }, + { + "epoch": 0.73, + "learning_rate": 3.5424054173474955e-06, + "loss": 1.09, + "step": 30428 + }, + { + "epoch": 0.73, + "learning_rate": 3.5418107192007287e-06, + "loss": 1.132, + "step": 30429 + }, + { + "epoch": 0.73, + "learning_rate": 3.5412160602342816e-06, + "loss": 0.9142, + "step": 30430 + }, + { + "epoch": 0.73, + "learning_rate": 3.5406214404517535e-06, + "loss": 1.0712, + "step": 30431 + }, + { + "epoch": 0.73, + "learning_rate": 3.54002685985676e-06, + "loss": 0.7945, + "step": 30432 + }, + { + "epoch": 0.73, + "learning_rate": 3.5394323184529e-06, + "loss": 1.0193, + "step": 30433 + }, + { + "epoch": 0.73, + "learning_rate": 3.538837816243784e-06, + "loss": 1.2504, + "step": 30434 + }, + { + "epoch": 0.73, + "learning_rate": 3.538243353233024e-06, + "loss": 1.0067, + "step": 30435 + }, + { + "epoch": 0.73, + "learning_rate": 3.5376489294242167e-06, + "loss": 1.137, + "step": 30436 + }, + { + "epoch": 0.73, + "learning_rate": 3.5370545448209747e-06, + "loss": 1.1809, + "step": 30437 + }, + { + "epoch": 0.73, + "learning_rate": 3.5364601994269055e-06, + "loss": 0.8177, + "step": 30438 + }, + { + "epoch": 0.73, + "learning_rate": 3.535865893245608e-06, + "loss": 0.9778, + "step": 30439 + }, + { + "epoch": 0.73, + "learning_rate": 3.5352716262806953e-06, + "loss": 0.8668, + "step": 30440 + }, + { + "epoch": 0.73, + "learning_rate": 3.5346773985357663e-06, + "loss": 0.9503, + "step": 30441 + }, + { + "epoch": 0.73, + "learning_rate": 3.5340832100144284e-06, + "loss": 1.183, + "step": 30442 + }, + { + "epoch": 0.73, + "learning_rate": 3.53348906072029e-06, + "loss": 1.1621, + "step": 30443 + }, + { + "epoch": 0.73, + "learning_rate": 3.5328949506569488e-06, + "loss": 1.1287, + "step": 30444 + }, + { + "epoch": 0.73, + "learning_rate": 3.532300879828013e-06, + "loss": 1.2729, + "step": 30445 + }, + { + "epoch": 0.73, + "learning_rate": 3.5317068482370907e-06, + "loss": 1.341, + "step": 30446 + }, + { + "epoch": 0.73, + "learning_rate": 3.531112855887777e-06, + "loss": 1.1585, + "step": 30447 + }, + { + "epoch": 0.73, + "learning_rate": 3.530518902783684e-06, + "loss": 1.1475, + "step": 30448 + }, + { + "epoch": 0.73, + "learning_rate": 3.5299249889284073e-06, + "loss": 1.0342, + "step": 30449 + }, + { + "epoch": 0.73, + "learning_rate": 3.529331114325555e-06, + "loss": 1.0448, + "step": 30450 + }, + { + "epoch": 0.73, + "learning_rate": 3.528737278978731e-06, + "loss": 0.9559, + "step": 30451 + }, + { + "epoch": 0.73, + "learning_rate": 3.528143482891533e-06, + "loss": 1.0586, + "step": 30452 + }, + { + "epoch": 0.73, + "learning_rate": 3.527549726067565e-06, + "loss": 1.2219, + "step": 30453 + }, + { + "epoch": 0.73, + "learning_rate": 3.5269560085104347e-06, + "loss": 0.8382, + "step": 30454 + }, + { + "epoch": 0.73, + "learning_rate": 3.526362330223737e-06, + "loss": 1.0624, + "step": 30455 + }, + { + "epoch": 0.73, + "learning_rate": 3.5257686912110758e-06, + "loss": 0.7705, + "step": 30456 + }, + { + "epoch": 0.73, + "learning_rate": 3.5251750914760563e-06, + "loss": 1.1902, + "step": 30457 + }, + { + "epoch": 0.73, + "learning_rate": 3.5245815310222733e-06, + "loss": 1.0573, + "step": 30458 + }, + { + "epoch": 0.73, + "learning_rate": 3.523988009853334e-06, + "loss": 1.0662, + "step": 30459 + }, + { + "epoch": 0.73, + "learning_rate": 3.523394527972833e-06, + "loss": 0.9237, + "step": 30460 + }, + { + "epoch": 0.73, + "learning_rate": 3.522801085384375e-06, + "loss": 1.0889, + "step": 30461 + }, + { + "epoch": 0.73, + "learning_rate": 3.5222076820915618e-06, + "loss": 0.9576, + "step": 30462 + }, + { + "epoch": 0.73, + "learning_rate": 3.5216143180979877e-06, + "loss": 1.0447, + "step": 30463 + }, + { + "epoch": 0.73, + "learning_rate": 3.521020993407256e-06, + "loss": 0.9971, + "step": 30464 + }, + { + "epoch": 0.73, + "learning_rate": 3.520427708022969e-06, + "loss": 1.091, + "step": 30465 + }, + { + "epoch": 0.73, + "learning_rate": 3.519834461948719e-06, + "loss": 0.9672, + "step": 30466 + }, + { + "epoch": 0.73, + "learning_rate": 3.5192412551881137e-06, + "loss": 0.97, + "step": 30467 + }, + { + "epoch": 0.73, + "learning_rate": 3.5186480877447437e-06, + "loss": 0.9683, + "step": 30468 + }, + { + "epoch": 0.73, + "learning_rate": 3.51805495962221e-06, + "loss": 1.1829, + "step": 30469 + }, + { + "epoch": 0.73, + "learning_rate": 3.517461870824117e-06, + "loss": 1.0499, + "step": 30470 + }, + { + "epoch": 0.73, + "learning_rate": 3.5168688213540547e-06, + "loss": 1.2077, + "step": 30471 + }, + { + "epoch": 0.73, + "learning_rate": 3.516275811215624e-06, + "loss": 1.0855, + "step": 30472 + }, + { + "epoch": 0.73, + "learning_rate": 3.5156828404124255e-06, + "loss": 0.8258, + "step": 30473 + }, + { + "epoch": 0.73, + "learning_rate": 3.5150899089480516e-06, + "loss": 1.0183, + "step": 30474 + }, + { + "epoch": 0.73, + "learning_rate": 3.514497016826105e-06, + "loss": 1.3756, + "step": 30475 + }, + { + "epoch": 0.73, + "learning_rate": 3.513904164050176e-06, + "loss": 1.1902, + "step": 30476 + }, + { + "epoch": 0.73, + "learning_rate": 3.513311350623865e-06, + "loss": 1.0287, + "step": 30477 + }, + { + "epoch": 0.73, + "learning_rate": 3.5127185765507708e-06, + "loss": 0.7817, + "step": 30478 + }, + { + "epoch": 0.73, + "learning_rate": 3.512125841834485e-06, + "loss": 1.1147, + "step": 30479 + }, + { + "epoch": 0.73, + "learning_rate": 3.5115331464786053e-06, + "loss": 0.852, + "step": 30480 + }, + { + "epoch": 0.73, + "learning_rate": 3.5109404904867304e-06, + "loss": 0.7804, + "step": 30481 + }, + { + "epoch": 0.73, + "learning_rate": 3.5103478738624506e-06, + "loss": 1.002, + "step": 30482 + }, + { + "epoch": 0.73, + "learning_rate": 3.5097552966093672e-06, + "loss": 0.9832, + "step": 30483 + }, + { + "epoch": 0.73, + "learning_rate": 3.5091627587310682e-06, + "loss": 1.2774, + "step": 30484 + }, + { + "epoch": 0.73, + "learning_rate": 3.508570260231152e-06, + "loss": 1.208, + "step": 30485 + }, + { + "epoch": 0.73, + "learning_rate": 3.507977801113216e-06, + "loss": 1.1049, + "step": 30486 + }, + { + "epoch": 0.73, + "learning_rate": 3.5073853813808488e-06, + "loss": 0.8923, + "step": 30487 + }, + { + "epoch": 0.73, + "learning_rate": 3.5067930010376484e-06, + "loss": 1.1473, + "step": 30488 + }, + { + "epoch": 0.73, + "learning_rate": 3.506200660087209e-06, + "loss": 0.8545, + "step": 30489 + }, + { + "epoch": 0.73, + "learning_rate": 3.5056083585331204e-06, + "loss": 0.6168, + "step": 30490 + }, + { + "epoch": 0.73, + "learning_rate": 3.5050160963789814e-06, + "loss": 0.9242, + "step": 30491 + }, + { + "epoch": 0.73, + "learning_rate": 3.50442387362838e-06, + "loss": 1.044, + "step": 30492 + }, + { + "epoch": 0.73, + "learning_rate": 3.503831690284909e-06, + "loss": 0.9533, + "step": 30493 + }, + { + "epoch": 0.73, + "learning_rate": 3.5032395463521674e-06, + "loss": 0.8182, + "step": 30494 + }, + { + "epoch": 0.73, + "learning_rate": 3.5026474418337406e-06, + "loss": 1.1309, + "step": 30495 + }, + { + "epoch": 0.73, + "learning_rate": 3.5020553767332233e-06, + "loss": 1.1187, + "step": 30496 + }, + { + "epoch": 0.73, + "learning_rate": 3.5014633510542105e-06, + "loss": 1.2788, + "step": 30497 + }, + { + "epoch": 0.73, + "learning_rate": 3.500871364800288e-06, + "loss": 1.1671, + "step": 30498 + }, + { + "epoch": 0.73, + "learning_rate": 3.50027941797505e-06, + "loss": 0.9987, + "step": 30499 + }, + { + "epoch": 0.73, + "learning_rate": 3.499687510582087e-06, + "loss": 1.1111, + "step": 30500 + }, + { + "epoch": 0.73, + "learning_rate": 3.4990956426249923e-06, + "loss": 1.1544, + "step": 30501 + }, + { + "epoch": 0.73, + "learning_rate": 3.498503814107358e-06, + "loss": 1.1495, + "step": 30502 + }, + { + "epoch": 0.73, + "learning_rate": 3.497912025032769e-06, + "loss": 1.1009, + "step": 30503 + }, + { + "epoch": 0.73, + "learning_rate": 3.4973202754048207e-06, + "loss": 0.9951, + "step": 30504 + }, + { + "epoch": 0.73, + "learning_rate": 3.496728565227099e-06, + "loss": 1.109, + "step": 30505 + }, + { + "epoch": 0.73, + "learning_rate": 3.4961368945031936e-06, + "loss": 1.0144, + "step": 30506 + }, + { + "epoch": 0.73, + "learning_rate": 3.4955452632367006e-06, + "loss": 1.1059, + "step": 30507 + }, + { + "epoch": 0.73, + "learning_rate": 3.4949536714312005e-06, + "loss": 0.8708, + "step": 30508 + }, + { + "epoch": 0.73, + "learning_rate": 3.4943621190902864e-06, + "loss": 0.8973, + "step": 30509 + }, + { + "epoch": 0.73, + "learning_rate": 3.4937706062175513e-06, + "loss": 1.0363, + "step": 30510 + }, + { + "epoch": 0.73, + "learning_rate": 3.4931791328165753e-06, + "loss": 1.1038, + "step": 30511 + }, + { + "epoch": 0.73, + "learning_rate": 3.4925876988909544e-06, + "loss": 0.7267, + "step": 30512 + }, + { + "epoch": 0.73, + "learning_rate": 3.491996304444271e-06, + "loss": 1.1366, + "step": 30513 + }, + { + "epoch": 0.73, + "learning_rate": 3.4914049494801137e-06, + "loss": 0.8591, + "step": 30514 + }, + { + "epoch": 0.73, + "learning_rate": 3.490813634002076e-06, + "loss": 1.1049, + "step": 30515 + }, + { + "epoch": 0.73, + "learning_rate": 3.4902223580137372e-06, + "loss": 1.1574, + "step": 30516 + }, + { + "epoch": 0.73, + "learning_rate": 3.4896311215186884e-06, + "loss": 0.9643, + "step": 30517 + }, + { + "epoch": 0.73, + "learning_rate": 3.4890399245205186e-06, + "loss": 1.0781, + "step": 30518 + }, + { + "epoch": 0.73, + "learning_rate": 3.4884487670228097e-06, + "loss": 0.9854, + "step": 30519 + }, + { + "epoch": 0.73, + "learning_rate": 3.4878576490291537e-06, + "loss": 1.1434, + "step": 30520 + }, + { + "epoch": 0.73, + "learning_rate": 3.4872665705431308e-06, + "loss": 1.2648, + "step": 30521 + }, + { + "epoch": 0.73, + "learning_rate": 3.486675531568329e-06, + "loss": 1.0663, + "step": 30522 + }, + { + "epoch": 0.73, + "learning_rate": 3.4860845321083382e-06, + "loss": 1.0421, + "step": 30523 + }, + { + "epoch": 0.73, + "learning_rate": 3.485493572166737e-06, + "loss": 1.1471, + "step": 30524 + }, + { + "epoch": 0.73, + "learning_rate": 3.4849026517471143e-06, + "loss": 1.0029, + "step": 30525 + }, + { + "epoch": 0.73, + "learning_rate": 3.4843117708530583e-06, + "loss": 1.0856, + "step": 30526 + }, + { + "epoch": 0.73, + "learning_rate": 3.483720929488147e-06, + "loss": 1.0815, + "step": 30527 + }, + { + "epoch": 0.73, + "learning_rate": 3.4831301276559706e-06, + "loss": 0.9792, + "step": 30528 + }, + { + "epoch": 0.73, + "learning_rate": 3.482539365360108e-06, + "loss": 0.8267, + "step": 30529 + }, + { + "epoch": 0.73, + "learning_rate": 3.481948642604146e-06, + "loss": 0.8015, + "step": 30530 + }, + { + "epoch": 0.73, + "learning_rate": 3.4813579593916713e-06, + "loss": 1.2931, + "step": 30531 + }, + { + "epoch": 0.73, + "learning_rate": 3.480767315726261e-06, + "loss": 0.6988, + "step": 30532 + }, + { + "epoch": 0.73, + "learning_rate": 3.4801767116115036e-06, + "loss": 1.2248, + "step": 30533 + }, + { + "epoch": 0.73, + "learning_rate": 3.479586147050983e-06, + "loss": 1.2062, + "step": 30534 + }, + { + "epoch": 0.73, + "learning_rate": 3.478995622048276e-06, + "loss": 1.0976, + "step": 30535 + }, + { + "epoch": 0.73, + "learning_rate": 3.478405136606968e-06, + "loss": 1.0473, + "step": 30536 + }, + { + "epoch": 0.73, + "learning_rate": 3.4778146907306465e-06, + "loss": 0.8002, + "step": 30537 + }, + { + "epoch": 0.73, + "learning_rate": 3.4772242844228866e-06, + "loss": 1.1632, + "step": 30538 + }, + { + "epoch": 0.73, + "learning_rate": 3.4766339176872745e-06, + "loss": 0.9118, + "step": 30539 + }, + { + "epoch": 0.73, + "learning_rate": 3.476043590527387e-06, + "loss": 0.9235, + "step": 30540 + }, + { + "epoch": 0.73, + "learning_rate": 3.4754533029468086e-06, + "loss": 1.1558, + "step": 30541 + }, + { + "epoch": 0.73, + "learning_rate": 3.4748630549491237e-06, + "loss": 0.7996, + "step": 30542 + }, + { + "epoch": 0.73, + "learning_rate": 3.4742728465379073e-06, + "loss": 1.1485, + "step": 30543 + }, + { + "epoch": 0.73, + "learning_rate": 3.473682677716742e-06, + "loss": 1.0176, + "step": 30544 + }, + { + "epoch": 0.73, + "learning_rate": 3.473092548489212e-06, + "loss": 1.1082, + "step": 30545 + }, + { + "epoch": 0.73, + "learning_rate": 3.4725024588588906e-06, + "loss": 1.0387, + "step": 30546 + }, + { + "epoch": 0.73, + "learning_rate": 3.4719124088293654e-06, + "loss": 1.0596, + "step": 30547 + }, + { + "epoch": 0.73, + "learning_rate": 3.4713223984042086e-06, + "loss": 0.9548, + "step": 30548 + }, + { + "epoch": 0.73, + "learning_rate": 3.4707324275870024e-06, + "loss": 1.1548, + "step": 30549 + }, + { + "epoch": 0.73, + "learning_rate": 3.4701424963813313e-06, + "loss": 0.8565, + "step": 30550 + }, + { + "epoch": 0.73, + "learning_rate": 3.4695526047907656e-06, + "loss": 1.1672, + "step": 30551 + }, + { + "epoch": 0.73, + "learning_rate": 3.4689627528188883e-06, + "loss": 1.0569, + "step": 30552 + }, + { + "epoch": 0.73, + "learning_rate": 3.4683729404692823e-06, + "loss": 0.9943, + "step": 30553 + }, + { + "epoch": 0.73, + "learning_rate": 3.4677831677455166e-06, + "loss": 1.4287, + "step": 30554 + }, + { + "epoch": 0.73, + "learning_rate": 3.4671934346511783e-06, + "loss": 0.6766, + "step": 30555 + }, + { + "epoch": 0.73, + "learning_rate": 3.466603741189837e-06, + "loss": 1.143, + "step": 30556 + }, + { + "epoch": 0.73, + "learning_rate": 3.4660140873650736e-06, + "loss": 1.1557, + "step": 30557 + }, + { + "epoch": 0.73, + "learning_rate": 3.4654244731804697e-06, + "loss": 1.0813, + "step": 30558 + }, + { + "epoch": 0.73, + "learning_rate": 3.464834898639595e-06, + "loss": 0.8519, + "step": 30559 + }, + { + "epoch": 0.73, + "learning_rate": 3.4642453637460295e-06, + "loss": 0.916, + "step": 30560 + }, + { + "epoch": 0.73, + "learning_rate": 3.463655868503354e-06, + "loss": 0.9928, + "step": 30561 + }, + { + "epoch": 0.73, + "learning_rate": 3.463066412915137e-06, + "loss": 0.8989, + "step": 30562 + }, + { + "epoch": 0.74, + "learning_rate": 3.4624769969849624e-06, + "loss": 0.9923, + "step": 30563 + }, + { + "epoch": 0.74, + "learning_rate": 3.4618876207163987e-06, + "loss": 1.1416, + "step": 30564 + }, + { + "epoch": 0.74, + "learning_rate": 3.461298284113025e-06, + "loss": 1.0012, + "step": 30565 + }, + { + "epoch": 0.74, + "learning_rate": 3.4607089871784193e-06, + "loss": 1.2493, + "step": 30566 + }, + { + "epoch": 0.74, + "learning_rate": 3.4601197299161517e-06, + "loss": 1.1117, + "step": 30567 + }, + { + "epoch": 0.74, + "learning_rate": 3.4595305123297984e-06, + "loss": 1.024, + "step": 30568 + }, + { + "epoch": 0.74, + "learning_rate": 3.4589413344229395e-06, + "loss": 0.9812, + "step": 30569 + }, + { + "epoch": 0.74, + "learning_rate": 3.458352196199141e-06, + "loss": 1.1086, + "step": 30570 + }, + { + "epoch": 0.74, + "learning_rate": 3.457763097661984e-06, + "loss": 1.2691, + "step": 30571 + }, + { + "epoch": 0.74, + "learning_rate": 3.4571740388150365e-06, + "loss": 1.0413, + "step": 30572 + }, + { + "epoch": 0.74, + "learning_rate": 3.4565850196618754e-06, + "loss": 1.0088, + "step": 30573 + }, + { + "epoch": 0.74, + "learning_rate": 3.4559960402060764e-06, + "loss": 1.2197, + "step": 30574 + }, + { + "epoch": 0.74, + "learning_rate": 3.455407100451208e-06, + "loss": 1.0334, + "step": 30575 + }, + { + "epoch": 0.74, + "learning_rate": 3.4548182004008447e-06, + "loss": 1.1744, + "step": 30576 + }, + { + "epoch": 0.74, + "learning_rate": 3.4542293400585635e-06, + "loss": 0.8615, + "step": 30577 + }, + { + "epoch": 0.74, + "learning_rate": 3.4536405194279286e-06, + "loss": 1.1389, + "step": 30578 + }, + { + "epoch": 0.74, + "learning_rate": 3.4530517385125185e-06, + "loss": 1.1705, + "step": 30579 + }, + { + "epoch": 0.74, + "learning_rate": 3.4524629973159063e-06, + "loss": 0.9045, + "step": 30580 + }, + { + "epoch": 0.74, + "learning_rate": 3.4518742958416572e-06, + "loss": 1.21, + "step": 30581 + }, + { + "epoch": 0.74, + "learning_rate": 3.4512856340933507e-06, + "loss": 1.0799, + "step": 30582 + }, + { + "epoch": 0.74, + "learning_rate": 3.4506970120745508e-06, + "loss": 0.9096, + "step": 30583 + }, + { + "epoch": 0.74, + "learning_rate": 3.450108429788831e-06, + "loss": 1.0269, + "step": 30584 + }, + { + "epoch": 0.74, + "learning_rate": 3.4495198872397662e-06, + "loss": 1.1949, + "step": 30585 + }, + { + "epoch": 0.74, + "learning_rate": 3.4489313844309204e-06, + "loss": 1.0362, + "step": 30586 + }, + { + "epoch": 0.74, + "learning_rate": 3.448342921365866e-06, + "loss": 1.2732, + "step": 30587 + }, + { + "epoch": 0.74, + "learning_rate": 3.447754498048178e-06, + "loss": 1.0539, + "step": 30588 + }, + { + "epoch": 0.74, + "learning_rate": 3.4471661144814195e-06, + "loss": 1.3954, + "step": 30589 + }, + { + "epoch": 0.74, + "learning_rate": 3.446577770669165e-06, + "loss": 0.9195, + "step": 30590 + }, + { + "epoch": 0.74, + "learning_rate": 3.4459894666149796e-06, + "loss": 1.1057, + "step": 30591 + }, + { + "epoch": 0.74, + "learning_rate": 3.4454012023224337e-06, + "loss": 1.0903, + "step": 30592 + }, + { + "epoch": 0.74, + "learning_rate": 3.4448129777951014e-06, + "loss": 1.0907, + "step": 30593 + }, + { + "epoch": 0.74, + "learning_rate": 3.4442247930365423e-06, + "loss": 1.0253, + "step": 30594 + }, + { + "epoch": 0.74, + "learning_rate": 3.4436366480503293e-06, + "loss": 1.0053, + "step": 30595 + }, + { + "epoch": 0.74, + "learning_rate": 3.443048542840035e-06, + "loss": 1.0499, + "step": 30596 + }, + { + "epoch": 0.74, + "learning_rate": 3.4424604774092183e-06, + "loss": 1.0115, + "step": 30597 + }, + { + "epoch": 0.74, + "learning_rate": 3.4418724517614565e-06, + "loss": 1.0486, + "step": 30598 + }, + { + "epoch": 0.74, + "learning_rate": 3.4412844659003074e-06, + "loss": 0.9619, + "step": 30599 + }, + { + "epoch": 0.74, + "learning_rate": 3.4406965198293428e-06, + "loss": 1.0832, + "step": 30600 + }, + { + "epoch": 0.74, + "learning_rate": 3.4401086135521335e-06, + "loss": 1.2647, + "step": 30601 + }, + { + "epoch": 0.74, + "learning_rate": 3.439520747072238e-06, + "loss": 0.6555, + "step": 30602 + }, + { + "epoch": 0.74, + "learning_rate": 3.4389329203932276e-06, + "loss": 0.9167, + "step": 30603 + }, + { + "epoch": 0.74, + "learning_rate": 3.438345133518671e-06, + "loss": 1.1883, + "step": 30604 + }, + { + "epoch": 0.74, + "learning_rate": 3.4377573864521285e-06, + "loss": 1.1587, + "step": 30605 + }, + { + "epoch": 0.74, + "learning_rate": 3.4371696791971708e-06, + "loss": 1.0207, + "step": 30606 + }, + { + "epoch": 0.74, + "learning_rate": 3.436582011757358e-06, + "loss": 1.0772, + "step": 30607 + }, + { + "epoch": 0.74, + "learning_rate": 3.4359943841362586e-06, + "loss": 1.1208, + "step": 30608 + }, + { + "epoch": 0.74, + "learning_rate": 3.4354067963374404e-06, + "loss": 1.1543, + "step": 30609 + }, + { + "epoch": 0.74, + "learning_rate": 3.4348192483644614e-06, + "loss": 1.0624, + "step": 30610 + }, + { + "epoch": 0.74, + "learning_rate": 3.43423174022089e-06, + "loss": 1.1065, + "step": 30611 + }, + { + "epoch": 0.74, + "learning_rate": 3.4336442719102934e-06, + "loss": 0.8975, + "step": 30612 + }, + { + "epoch": 0.74, + "learning_rate": 3.43305684343623e-06, + "loss": 0.9441, + "step": 30613 + }, + { + "epoch": 0.74, + "learning_rate": 3.4324694548022686e-06, + "loss": 1.0691, + "step": 30614 + }, + { + "epoch": 0.74, + "learning_rate": 3.431882106011968e-06, + "loss": 1.0156, + "step": 30615 + }, + { + "epoch": 0.74, + "learning_rate": 3.4312947970688927e-06, + "loss": 1.0254, + "step": 30616 + }, + { + "epoch": 0.74, + "learning_rate": 3.43070752797661e-06, + "loss": 1.0972, + "step": 30617 + }, + { + "epoch": 0.74, + "learning_rate": 3.4301202987386772e-06, + "loss": 1.0868, + "step": 30618 + }, + { + "epoch": 0.74, + "learning_rate": 3.429533109358658e-06, + "loss": 0.9539, + "step": 30619 + }, + { + "epoch": 0.74, + "learning_rate": 3.4289459598401197e-06, + "loss": 1.1056, + "step": 30620 + }, + { + "epoch": 0.74, + "learning_rate": 3.4283588501866184e-06, + "loss": 0.8511, + "step": 30621 + }, + { + "epoch": 0.74, + "learning_rate": 3.4277717804017208e-06, + "loss": 1.167, + "step": 30622 + }, + { + "epoch": 0.74, + "learning_rate": 3.4271847504889823e-06, + "loss": 1.0445, + "step": 30623 + }, + { + "epoch": 0.74, + "learning_rate": 3.4265977604519686e-06, + "loss": 0.995, + "step": 30624 + }, + { + "epoch": 0.74, + "learning_rate": 3.4260108102942422e-06, + "loss": 0.8919, + "step": 30625 + }, + { + "epoch": 0.74, + "learning_rate": 3.425423900019359e-06, + "loss": 0.8468, + "step": 30626 + }, + { + "epoch": 0.74, + "learning_rate": 3.424837029630883e-06, + "loss": 1.2546, + "step": 30627 + }, + { + "epoch": 0.74, + "learning_rate": 3.424250199132377e-06, + "loss": 0.856, + "step": 30628 + }, + { + "epoch": 0.74, + "learning_rate": 3.4236634085273955e-06, + "loss": 0.9933, + "step": 30629 + }, + { + "epoch": 0.74, + "learning_rate": 3.4230766578195016e-06, + "loss": 0.8181, + "step": 30630 + }, + { + "epoch": 0.74, + "learning_rate": 3.4224899470122573e-06, + "loss": 1.1284, + "step": 30631 + }, + { + "epoch": 0.74, + "learning_rate": 3.4219032761092163e-06, + "loss": 1.1485, + "step": 30632 + }, + { + "epoch": 0.74, + "learning_rate": 3.4213166451139445e-06, + "loss": 0.744, + "step": 30633 + }, + { + "epoch": 0.74, + "learning_rate": 3.4207300540299935e-06, + "loss": 0.7617, + "step": 30634 + }, + { + "epoch": 0.74, + "learning_rate": 3.4201435028609253e-06, + "loss": 0.9573, + "step": 30635 + }, + { + "epoch": 0.74, + "learning_rate": 3.419556991610303e-06, + "loss": 1.0793, + "step": 30636 + }, + { + "epoch": 0.74, + "learning_rate": 3.418970520281677e-06, + "loss": 1.2899, + "step": 30637 + }, + { + "epoch": 0.74, + "learning_rate": 3.418384088878609e-06, + "loss": 0.9088, + "step": 30638 + }, + { + "epoch": 0.74, + "learning_rate": 3.417797697404659e-06, + "loss": 1.1097, + "step": 30639 + }, + { + "epoch": 0.74, + "learning_rate": 3.4172113458633803e-06, + "loss": 0.9905, + "step": 30640 + }, + { + "epoch": 0.74, + "learning_rate": 3.416625034258334e-06, + "loss": 0.9618, + "step": 30641 + }, + { + "epoch": 0.74, + "learning_rate": 3.4160387625930723e-06, + "loss": 0.9031, + "step": 30642 + }, + { + "epoch": 0.74, + "learning_rate": 3.415452530871155e-06, + "loss": 1.1767, + "step": 30643 + }, + { + "epoch": 0.74, + "learning_rate": 3.414866339096142e-06, + "loss": 1.1112, + "step": 30644 + }, + { + "epoch": 0.74, + "learning_rate": 3.4142801872715824e-06, + "loss": 1.076, + "step": 30645 + }, + { + "epoch": 0.74, + "learning_rate": 3.413694075401035e-06, + "loss": 1.1534, + "step": 30646 + }, + { + "epoch": 0.74, + "learning_rate": 3.4131080034880603e-06, + "loss": 1.049, + "step": 30647 + }, + { + "epoch": 0.74, + "learning_rate": 3.412521971536208e-06, + "loss": 1.1022, + "step": 30648 + }, + { + "epoch": 0.74, + "learning_rate": 3.411935979549037e-06, + "loss": 0.9728, + "step": 30649 + }, + { + "epoch": 0.74, + "learning_rate": 3.411350027530098e-06, + "loss": 0.8801, + "step": 30650 + }, + { + "epoch": 0.74, + "learning_rate": 3.4107641154829485e-06, + "loss": 0.9606, + "step": 30651 + }, + { + "epoch": 0.74, + "learning_rate": 3.4101782434111476e-06, + "loss": 1.1019, + "step": 30652 + }, + { + "epoch": 0.74, + "learning_rate": 3.409592411318241e-06, + "loss": 1.0147, + "step": 30653 + }, + { + "epoch": 0.74, + "learning_rate": 3.4090066192077864e-06, + "loss": 1.1211, + "step": 30654 + }, + { + "epoch": 0.74, + "learning_rate": 3.408420867083343e-06, + "loss": 0.8991, + "step": 30655 + }, + { + "epoch": 0.74, + "learning_rate": 3.4078351549484555e-06, + "loss": 1.1311, + "step": 30656 + }, + { + "epoch": 0.74, + "learning_rate": 3.407249482806686e-06, + "loss": 1.0827, + "step": 30657 + }, + { + "epoch": 0.74, + "learning_rate": 3.406663850661579e-06, + "loss": 0.7467, + "step": 30658 + }, + { + "epoch": 0.74, + "learning_rate": 3.406078258516692e-06, + "loss": 1.0028, + "step": 30659 + }, + { + "epoch": 0.74, + "learning_rate": 3.4054927063755793e-06, + "loss": 0.9537, + "step": 30660 + }, + { + "epoch": 0.74, + "learning_rate": 3.40490719424179e-06, + "loss": 1.0518, + "step": 30661 + }, + { + "epoch": 0.74, + "learning_rate": 3.4043217221188764e-06, + "loss": 1.0312, + "step": 30662 + }, + { + "epoch": 0.74, + "learning_rate": 3.403736290010394e-06, + "loss": 1.3372, + "step": 30663 + }, + { + "epoch": 0.74, + "learning_rate": 3.40315089791989e-06, + "loss": 0.9917, + "step": 30664 + }, + { + "epoch": 0.74, + "learning_rate": 3.40256554585092e-06, + "loss": 0.9955, + "step": 30665 + }, + { + "epoch": 0.74, + "learning_rate": 3.4019802338070276e-06, + "loss": 0.8779, + "step": 30666 + }, + { + "epoch": 0.74, + "learning_rate": 3.4013949617917717e-06, + "loss": 1.1462, + "step": 30667 + }, + { + "epoch": 0.74, + "learning_rate": 3.400809729808704e-06, + "loss": 1.0019, + "step": 30668 + }, + { + "epoch": 0.74, + "learning_rate": 3.4002245378613686e-06, + "loss": 0.8029, + "step": 30669 + }, + { + "epoch": 0.74, + "learning_rate": 3.3996393859533204e-06, + "loss": 1.25, + "step": 30670 + }, + { + "epoch": 0.74, + "learning_rate": 3.399054274088106e-06, + "loss": 1.047, + "step": 30671 + }, + { + "epoch": 0.74, + "learning_rate": 3.3984692022692756e-06, + "loss": 1.0308, + "step": 30672 + }, + { + "epoch": 0.74, + "learning_rate": 3.397884170500384e-06, + "loss": 1.2554, + "step": 30673 + }, + { + "epoch": 0.74, + "learning_rate": 3.397299178784972e-06, + "loss": 1.2482, + "step": 30674 + }, + { + "epoch": 0.74, + "learning_rate": 3.396714227126593e-06, + "loss": 1.0779, + "step": 30675 + }, + { + "epoch": 0.74, + "learning_rate": 3.396129315528799e-06, + "loss": 1.1573, + "step": 30676 + }, + { + "epoch": 0.74, + "learning_rate": 3.3955444439951323e-06, + "loss": 0.9362, + "step": 30677 + }, + { + "epoch": 0.74, + "learning_rate": 3.3949596125291463e-06, + "loss": 0.8329, + "step": 30678 + }, + { + "epoch": 0.74, + "learning_rate": 3.394374821134384e-06, + "loss": 1.0224, + "step": 30679 + }, + { + "epoch": 0.74, + "learning_rate": 3.3937900698143967e-06, + "loss": 1.3142, + "step": 30680 + }, + { + "epoch": 0.74, + "learning_rate": 3.3932053585727333e-06, + "loss": 1.158, + "step": 30681 + }, + { + "epoch": 0.74, + "learning_rate": 3.3926206874129363e-06, + "loss": 0.9275, + "step": 30682 + }, + { + "epoch": 0.74, + "learning_rate": 3.3920360563385546e-06, + "loss": 1.1428, + "step": 30683 + }, + { + "epoch": 0.74, + "learning_rate": 3.3914514653531404e-06, + "loss": 1.1065, + "step": 30684 + }, + { + "epoch": 0.74, + "learning_rate": 3.390866914460231e-06, + "loss": 1.0525, + "step": 30685 + }, + { + "epoch": 0.74, + "learning_rate": 3.390282403663382e-06, + "loss": 0.9241, + "step": 30686 + }, + { + "epoch": 0.74, + "learning_rate": 3.3896979329661317e-06, + "loss": 1.0825, + "step": 30687 + }, + { + "epoch": 0.74, + "learning_rate": 3.389113502372028e-06, + "loss": 1.2222, + "step": 30688 + }, + { + "epoch": 0.74, + "learning_rate": 3.388529111884622e-06, + "loss": 1.131, + "step": 30689 + }, + { + "epoch": 0.74, + "learning_rate": 3.3879447615074514e-06, + "loss": 1.0291, + "step": 30690 + }, + { + "epoch": 0.74, + "learning_rate": 3.387360451244064e-06, + "loss": 0.9701, + "step": 30691 + }, + { + "epoch": 0.74, + "learning_rate": 3.386776181098009e-06, + "loss": 0.9081, + "step": 30692 + }, + { + "epoch": 0.74, + "learning_rate": 3.386191951072825e-06, + "loss": 1.0146, + "step": 30693 + }, + { + "epoch": 0.74, + "learning_rate": 3.3856077611720607e-06, + "loss": 0.9577, + "step": 30694 + }, + { + "epoch": 0.74, + "learning_rate": 3.385023611399256e-06, + "loss": 1.2293, + "step": 30695 + }, + { + "epoch": 0.74, + "learning_rate": 3.3844395017579566e-06, + "loss": 0.885, + "step": 30696 + }, + { + "epoch": 0.74, + "learning_rate": 3.3838554322517103e-06, + "loss": 1.0891, + "step": 30697 + }, + { + "epoch": 0.74, + "learning_rate": 3.383271402884053e-06, + "loss": 1.0572, + "step": 30698 + }, + { + "epoch": 0.74, + "learning_rate": 3.382687413658532e-06, + "loss": 0.968, + "step": 30699 + }, + { + "epoch": 0.74, + "learning_rate": 3.3821034645786934e-06, + "loss": 1.1945, + "step": 30700 + }, + { + "epoch": 0.74, + "learning_rate": 3.381519555648073e-06, + "loss": 1.0286, + "step": 30701 + }, + { + "epoch": 0.74, + "learning_rate": 3.3809356868702202e-06, + "loss": 0.9781, + "step": 30702 + }, + { + "epoch": 0.74, + "learning_rate": 3.3803518582486706e-06, + "loss": 1.0888, + "step": 30703 + }, + { + "epoch": 0.74, + "learning_rate": 3.3797680697869685e-06, + "loss": 0.9976, + "step": 30704 + }, + { + "epoch": 0.74, + "learning_rate": 3.37918432148866e-06, + "loss": 0.8905, + "step": 30705 + }, + { + "epoch": 0.74, + "learning_rate": 3.3786006133572803e-06, + "loss": 1.0926, + "step": 30706 + }, + { + "epoch": 0.74, + "learning_rate": 3.378016945396373e-06, + "loss": 1.2384, + "step": 30707 + }, + { + "epoch": 0.74, + "learning_rate": 3.377433317609482e-06, + "loss": 1.0024, + "step": 30708 + }, + { + "epoch": 0.74, + "learning_rate": 3.3768497300001425e-06, + "loss": 1.0828, + "step": 30709 + }, + { + "epoch": 0.74, + "learning_rate": 3.376266182571897e-06, + "loss": 1.0726, + "step": 30710 + }, + { + "epoch": 0.74, + "learning_rate": 3.375682675328291e-06, + "loss": 0.8438, + "step": 30711 + }, + { + "epoch": 0.74, + "learning_rate": 3.375099208272856e-06, + "loss": 1.0639, + "step": 30712 + }, + { + "epoch": 0.74, + "learning_rate": 3.374515781409139e-06, + "loss": 1.1615, + "step": 30713 + }, + { + "epoch": 0.74, + "learning_rate": 3.3739323947406744e-06, + "loss": 0.8672, + "step": 30714 + }, + { + "epoch": 0.74, + "learning_rate": 3.373349048271002e-06, + "loss": 1.0756, + "step": 30715 + }, + { + "epoch": 0.74, + "learning_rate": 3.3727657420036665e-06, + "loss": 0.9433, + "step": 30716 + }, + { + "epoch": 0.74, + "learning_rate": 3.3721824759421985e-06, + "loss": 1.1889, + "step": 30717 + }, + { + "epoch": 0.74, + "learning_rate": 3.3715992500901405e-06, + "loss": 1.0593, + "step": 30718 + }, + { + "epoch": 0.74, + "learning_rate": 3.371016064451034e-06, + "loss": 1.1992, + "step": 30719 + }, + { + "epoch": 0.74, + "learning_rate": 3.3704329190284104e-06, + "loss": 0.7909, + "step": 30720 + }, + { + "epoch": 0.74, + "learning_rate": 3.369849813825814e-06, + "loss": 0.975, + "step": 30721 + }, + { + "epoch": 0.74, + "learning_rate": 3.369266748846777e-06, + "loss": 0.8842, + "step": 30722 + }, + { + "epoch": 0.74, + "learning_rate": 3.3686837240948388e-06, + "loss": 1.137, + "step": 30723 + }, + { + "epoch": 0.74, + "learning_rate": 3.368100739573539e-06, + "loss": 1.0786, + "step": 30724 + }, + { + "epoch": 0.74, + "learning_rate": 3.367517795286409e-06, + "loss": 0.97, + "step": 30725 + }, + { + "epoch": 0.74, + "learning_rate": 3.366934891236989e-06, + "loss": 1.0687, + "step": 30726 + }, + { + "epoch": 0.74, + "learning_rate": 3.3663520274288187e-06, + "loss": 1.1294, + "step": 30727 + }, + { + "epoch": 0.74, + "learning_rate": 3.3657692038654256e-06, + "loss": 1.163, + "step": 30728 + }, + { + "epoch": 0.74, + "learning_rate": 3.365186420550355e-06, + "loss": 1.194, + "step": 30729 + }, + { + "epoch": 0.74, + "learning_rate": 3.3646036774871338e-06, + "loss": 1.127, + "step": 30730 + }, + { + "epoch": 0.74, + "learning_rate": 3.3640209746793005e-06, + "loss": 1.0295, + "step": 30731 + }, + { + "epoch": 0.74, + "learning_rate": 3.363438312130396e-06, + "loss": 1.1245, + "step": 30732 + }, + { + "epoch": 0.74, + "learning_rate": 3.3628556898439456e-06, + "loss": 0.8824, + "step": 30733 + }, + { + "epoch": 0.74, + "learning_rate": 3.362273107823489e-06, + "loss": 0.9106, + "step": 30734 + }, + { + "epoch": 0.74, + "learning_rate": 3.361690566072564e-06, + "loss": 0.9283, + "step": 30735 + }, + { + "epoch": 0.74, + "learning_rate": 3.361108064594697e-06, + "loss": 0.995, + "step": 30736 + }, + { + "epoch": 0.74, + "learning_rate": 3.3605256033934295e-06, + "loss": 1.0291, + "step": 30737 + }, + { + "epoch": 0.74, + "learning_rate": 3.3599431824722873e-06, + "loss": 0.8747, + "step": 30738 + }, + { + "epoch": 0.74, + "learning_rate": 3.359360801834809e-06, + "loss": 1.2047, + "step": 30739 + }, + { + "epoch": 0.74, + "learning_rate": 3.35877846148453e-06, + "loss": 1.0621, + "step": 30740 + }, + { + "epoch": 0.74, + "learning_rate": 3.3581961614249757e-06, + "loss": 1.3042, + "step": 30741 + }, + { + "epoch": 0.74, + "learning_rate": 3.3576139016596843e-06, + "loss": 0.7102, + "step": 30742 + }, + { + "epoch": 0.74, + "learning_rate": 3.357031682192189e-06, + "loss": 0.8815, + "step": 30743 + }, + { + "epoch": 0.74, + "learning_rate": 3.356449503026018e-06, + "loss": 0.9517, + "step": 30744 + }, + { + "epoch": 0.74, + "learning_rate": 3.355867364164708e-06, + "loss": 1.1402, + "step": 30745 + }, + { + "epoch": 0.74, + "learning_rate": 3.355285265611784e-06, + "loss": 0.8773, + "step": 30746 + }, + { + "epoch": 0.74, + "learning_rate": 3.354703207370782e-06, + "loss": 0.9154, + "step": 30747 + }, + { + "epoch": 0.74, + "learning_rate": 3.354121189445235e-06, + "loss": 0.9352, + "step": 30748 + }, + { + "epoch": 0.74, + "learning_rate": 3.353539211838669e-06, + "loss": 1.1928, + "step": 30749 + }, + { + "epoch": 0.74, + "learning_rate": 3.352957274554618e-06, + "loss": 0.7822, + "step": 30750 + }, + { + "epoch": 0.74, + "learning_rate": 3.352375377596614e-06, + "loss": 1.0742, + "step": 30751 + }, + { + "epoch": 0.74, + "learning_rate": 3.351793520968182e-06, + "loss": 1.2861, + "step": 30752 + }, + { + "epoch": 0.74, + "learning_rate": 3.3512117046728545e-06, + "loss": 1.423, + "step": 30753 + }, + { + "epoch": 0.74, + "learning_rate": 3.350629928714165e-06, + "loss": 1.0803, + "step": 30754 + }, + { + "epoch": 0.74, + "learning_rate": 3.3500481930956363e-06, + "loss": 1.1579, + "step": 30755 + }, + { + "epoch": 0.74, + "learning_rate": 3.3494664978208056e-06, + "loss": 1.1432, + "step": 30756 + }, + { + "epoch": 0.74, + "learning_rate": 3.348884842893193e-06, + "loss": 0.8673, + "step": 30757 + }, + { + "epoch": 0.74, + "learning_rate": 3.348303228316332e-06, + "loss": 1.0266, + "step": 30758 + }, + { + "epoch": 0.74, + "learning_rate": 3.347721654093754e-06, + "loss": 1.1081, + "step": 30759 + }, + { + "epoch": 0.74, + "learning_rate": 3.34714012022898e-06, + "loss": 1.3315, + "step": 30760 + }, + { + "epoch": 0.74, + "learning_rate": 3.3465586267255425e-06, + "loss": 1.0069, + "step": 30761 + }, + { + "epoch": 0.74, + "learning_rate": 3.345977173586973e-06, + "loss": 1.0755, + "step": 30762 + }, + { + "epoch": 0.74, + "learning_rate": 3.345395760816791e-06, + "loss": 1.0403, + "step": 30763 + }, + { + "epoch": 0.74, + "learning_rate": 3.3448143884185302e-06, + "loss": 1.3211, + "step": 30764 + }, + { + "epoch": 0.74, + "learning_rate": 3.3442330563957128e-06, + "loss": 0.9704, + "step": 30765 + }, + { + "epoch": 0.74, + "learning_rate": 3.3436517647518673e-06, + "loss": 0.7672, + "step": 30766 + }, + { + "epoch": 0.74, + "learning_rate": 3.343070513490524e-06, + "loss": 0.9916, + "step": 30767 + }, + { + "epoch": 0.74, + "learning_rate": 3.3424893026152027e-06, + "loss": 1.0852, + "step": 30768 + }, + { + "epoch": 0.74, + "learning_rate": 3.3419081321294343e-06, + "loss": 0.9638, + "step": 30769 + }, + { + "epoch": 0.74, + "learning_rate": 3.341327002036745e-06, + "loss": 1.1124, + "step": 30770 + }, + { + "epoch": 0.74, + "learning_rate": 3.340745912340656e-06, + "loss": 0.9385, + "step": 30771 + }, + { + "epoch": 0.74, + "learning_rate": 3.3401648630446983e-06, + "loss": 0.8373, + "step": 30772 + }, + { + "epoch": 0.74, + "learning_rate": 3.339583854152391e-06, + "loss": 0.8512, + "step": 30773 + }, + { + "epoch": 0.74, + "learning_rate": 3.3390028856672607e-06, + "loss": 0.9174, + "step": 30774 + }, + { + "epoch": 0.74, + "learning_rate": 3.3384219575928367e-06, + "loss": 0.9364, + "step": 30775 + }, + { + "epoch": 0.74, + "learning_rate": 3.337841069932637e-06, + "loss": 0.835, + "step": 30776 + }, + { + "epoch": 0.74, + "learning_rate": 3.3372602226901884e-06, + "loss": 1.0717, + "step": 30777 + }, + { + "epoch": 0.74, + "learning_rate": 3.3366794158690187e-06, + "loss": 1.1186, + "step": 30778 + }, + { + "epoch": 0.74, + "learning_rate": 3.3360986494726433e-06, + "loss": 1.4279, + "step": 30779 + }, + { + "epoch": 0.74, + "learning_rate": 3.3355179235045944e-06, + "loss": 0.9642, + "step": 30780 + }, + { + "epoch": 0.74, + "learning_rate": 3.3349372379683864e-06, + "loss": 1.0188, + "step": 30781 + }, + { + "epoch": 0.74, + "learning_rate": 3.334356592867547e-06, + "loss": 1.2477, + "step": 30782 + }, + { + "epoch": 0.74, + "learning_rate": 3.333775988205601e-06, + "loss": 1.2174, + "step": 30783 + }, + { + "epoch": 0.74, + "learning_rate": 3.3331954239860655e-06, + "loss": 0.8296, + "step": 30784 + }, + { + "epoch": 0.74, + "learning_rate": 3.3326149002124654e-06, + "loss": 0.7193, + "step": 30785 + }, + { + "epoch": 0.74, + "learning_rate": 3.3320344168883245e-06, + "loss": 1.04, + "step": 30786 + }, + { + "epoch": 0.74, + "learning_rate": 3.3314539740171604e-06, + "loss": 1.2115, + "step": 30787 + }, + { + "epoch": 0.74, + "learning_rate": 3.330873571602499e-06, + "loss": 1.1796, + "step": 30788 + }, + { + "epoch": 0.74, + "learning_rate": 3.3302932096478556e-06, + "loss": 1.0276, + "step": 30789 + }, + { + "epoch": 0.74, + "learning_rate": 3.329712888156755e-06, + "loss": 1.0506, + "step": 30790 + }, + { + "epoch": 0.74, + "learning_rate": 3.32913260713272e-06, + "loss": 0.9786, + "step": 30791 + }, + { + "epoch": 0.74, + "learning_rate": 3.3285523665792663e-06, + "loss": 0.8641, + "step": 30792 + }, + { + "epoch": 0.74, + "learning_rate": 3.3279721664999152e-06, + "loss": 0.9321, + "step": 30793 + }, + { + "epoch": 0.74, + "learning_rate": 3.3273920068981913e-06, + "loss": 1.0129, + "step": 30794 + }, + { + "epoch": 0.74, + "learning_rate": 3.326811887777607e-06, + "loss": 0.9344, + "step": 30795 + }, + { + "epoch": 0.74, + "learning_rate": 3.326231809141686e-06, + "loss": 1.0783, + "step": 30796 + }, + { + "epoch": 0.74, + "learning_rate": 3.3256517709939495e-06, + "loss": 0.8545, + "step": 30797 + }, + { + "epoch": 0.74, + "learning_rate": 3.32507177333791e-06, + "loss": 1.2273, + "step": 30798 + }, + { + "epoch": 0.74, + "learning_rate": 3.3244918161770943e-06, + "loss": 1.216, + "step": 30799 + }, + { + "epoch": 0.74, + "learning_rate": 3.323911899515013e-06, + "loss": 0.9292, + "step": 30800 + }, + { + "epoch": 0.74, + "learning_rate": 3.323332023355188e-06, + "loss": 1.2433, + "step": 30801 + }, + { + "epoch": 0.74, + "learning_rate": 3.3227521877011405e-06, + "loss": 1.1484, + "step": 30802 + }, + { + "epoch": 0.74, + "learning_rate": 3.3221723925563807e-06, + "loss": 1.0483, + "step": 30803 + }, + { + "epoch": 0.74, + "learning_rate": 3.3215926379244314e-06, + "loss": 0.7742, + "step": 30804 + }, + { + "epoch": 0.74, + "learning_rate": 3.321012923808812e-06, + "loss": 1.0803, + "step": 30805 + }, + { + "epoch": 0.74, + "learning_rate": 3.320433250213032e-06, + "loss": 1.0904, + "step": 30806 + }, + { + "epoch": 0.74, + "learning_rate": 3.319853617140616e-06, + "loss": 0.8656, + "step": 30807 + }, + { + "epoch": 0.74, + "learning_rate": 3.319274024595074e-06, + "loss": 1.1592, + "step": 30808 + }, + { + "epoch": 0.74, + "learning_rate": 3.318694472579924e-06, + "loss": 1.1213, + "step": 30809 + }, + { + "epoch": 0.74, + "learning_rate": 3.3181149610986874e-06, + "loss": 0.9779, + "step": 30810 + }, + { + "epoch": 0.74, + "learning_rate": 3.3175354901548717e-06, + "loss": 1.3434, + "step": 30811 + }, + { + "epoch": 0.74, + "learning_rate": 3.316956059751997e-06, + "loss": 1.0314, + "step": 30812 + }, + { + "epoch": 0.74, + "learning_rate": 3.3163766698935805e-06, + "loss": 1.1963, + "step": 30813 + }, + { + "epoch": 0.74, + "learning_rate": 3.3157973205831327e-06, + "loss": 1.1812, + "step": 30814 + }, + { + "epoch": 0.74, + "learning_rate": 3.3152180118241726e-06, + "loss": 0.8014, + "step": 30815 + }, + { + "epoch": 0.74, + "learning_rate": 3.3146387436202086e-06, + "loss": 1.0135, + "step": 30816 + }, + { + "epoch": 0.74, + "learning_rate": 3.31405951597476e-06, + "loss": 1.0885, + "step": 30817 + }, + { + "epoch": 0.74, + "learning_rate": 3.3134803288913433e-06, + "loss": 0.7541, + "step": 30818 + }, + { + "epoch": 0.74, + "learning_rate": 3.312901182373465e-06, + "loss": 1.063, + "step": 30819 + }, + { + "epoch": 0.74, + "learning_rate": 3.312322076424641e-06, + "loss": 0.8821, + "step": 30820 + }, + { + "epoch": 0.74, + "learning_rate": 3.3117430110483904e-06, + "loss": 1.0948, + "step": 30821 + }, + { + "epoch": 0.74, + "learning_rate": 3.311163986248218e-06, + "loss": 0.9108, + "step": 30822 + }, + { + "epoch": 0.74, + "learning_rate": 3.3105850020276443e-06, + "loss": 0.9025, + "step": 30823 + }, + { + "epoch": 0.74, + "learning_rate": 3.3100060583901737e-06, + "loss": 1.2996, + "step": 30824 + }, + { + "epoch": 0.74, + "learning_rate": 3.3094271553393243e-06, + "loss": 1.1814, + "step": 30825 + }, + { + "epoch": 0.74, + "learning_rate": 3.3088482928786093e-06, + "loss": 0.9671, + "step": 30826 + }, + { + "epoch": 0.74, + "learning_rate": 3.3082694710115337e-06, + "loss": 0.9894, + "step": 30827 + }, + { + "epoch": 0.74, + "learning_rate": 3.307690689741614e-06, + "loss": 1.0822, + "step": 30828 + }, + { + "epoch": 0.74, + "learning_rate": 3.3071119490723634e-06, + "loss": 0.9817, + "step": 30829 + }, + { + "epoch": 0.74, + "learning_rate": 3.3065332490072877e-06, + "loss": 0.9818, + "step": 30830 + }, + { + "epoch": 0.74, + "learning_rate": 3.3059545895499036e-06, + "loss": 0.9669, + "step": 30831 + }, + { + "epoch": 0.74, + "learning_rate": 3.3053759707037116e-06, + "loss": 0.9963, + "step": 30832 + }, + { + "epoch": 0.74, + "learning_rate": 3.3047973924722323e-06, + "loss": 0.8697, + "step": 30833 + }, + { + "epoch": 0.74, + "learning_rate": 3.304218854858976e-06, + "loss": 1.1473, + "step": 30834 + }, + { + "epoch": 0.74, + "learning_rate": 3.303640357867445e-06, + "loss": 1.1238, + "step": 30835 + }, + { + "epoch": 0.74, + "learning_rate": 3.3030619015011567e-06, + "loss": 0.8517, + "step": 30836 + }, + { + "epoch": 0.74, + "learning_rate": 3.3024834857636136e-06, + "loss": 1.1598, + "step": 30837 + }, + { + "epoch": 0.74, + "learning_rate": 3.3019051106583278e-06, + "loss": 1.0134, + "step": 30838 + }, + { + "epoch": 0.74, + "learning_rate": 3.30132677618881e-06, + "loss": 0.9753, + "step": 30839 + }, + { + "epoch": 0.74, + "learning_rate": 3.3007484823585655e-06, + "loss": 1.0681, + "step": 30840 + }, + { + "epoch": 0.74, + "learning_rate": 3.3001702291711037e-06, + "loss": 1.2643, + "step": 30841 + }, + { + "epoch": 0.74, + "learning_rate": 3.2995920166299367e-06, + "loss": 1.1218, + "step": 30842 + }, + { + "epoch": 0.74, + "learning_rate": 3.2990138447385644e-06, + "loss": 0.8996, + "step": 30843 + }, + { + "epoch": 0.74, + "learning_rate": 3.298435713500504e-06, + "loss": 1.08, + "step": 30844 + }, + { + "epoch": 0.74, + "learning_rate": 3.297857622919254e-06, + "loss": 1.1519, + "step": 30845 + }, + { + "epoch": 0.74, + "learning_rate": 3.2972795729983242e-06, + "loss": 0.9268, + "step": 30846 + }, + { + "epoch": 0.74, + "learning_rate": 3.296701563741227e-06, + "loss": 1.1123, + "step": 30847 + }, + { + "epoch": 0.74, + "learning_rate": 3.2961235951514616e-06, + "loss": 1.187, + "step": 30848 + }, + { + "epoch": 0.74, + "learning_rate": 3.295545667232537e-06, + "loss": 0.8469, + "step": 30849 + }, + { + "epoch": 0.74, + "learning_rate": 3.294967779987963e-06, + "loss": 1.1652, + "step": 30850 + }, + { + "epoch": 0.74, + "learning_rate": 3.294389933421239e-06, + "loss": 1.1234, + "step": 30851 + }, + { + "epoch": 0.74, + "learning_rate": 3.293812127535878e-06, + "loss": 1.1505, + "step": 30852 + }, + { + "epoch": 0.74, + "learning_rate": 3.293234362335378e-06, + "loss": 0.8839, + "step": 30853 + }, + { + "epoch": 0.74, + "learning_rate": 3.2926566378232474e-06, + "loss": 1.1123, + "step": 30854 + }, + { + "epoch": 0.74, + "learning_rate": 3.292078954002994e-06, + "loss": 1.1751, + "step": 30855 + }, + { + "epoch": 0.74, + "learning_rate": 3.291501310878117e-06, + "loss": 1.1644, + "step": 30856 + }, + { + "epoch": 0.74, + "learning_rate": 3.290923708452124e-06, + "loss": 1.2055, + "step": 30857 + }, + { + "epoch": 0.74, + "learning_rate": 3.2903461467285216e-06, + "loss": 1.0104, + "step": 30858 + }, + { + "epoch": 0.74, + "learning_rate": 3.2897686257108074e-06, + "loss": 1.1501, + "step": 30859 + }, + { + "epoch": 0.74, + "learning_rate": 3.289191145402493e-06, + "loss": 1.2344, + "step": 30860 + }, + { + "epoch": 0.74, + "learning_rate": 3.288613705807073e-06, + "loss": 1.1505, + "step": 30861 + }, + { + "epoch": 0.74, + "learning_rate": 3.2880363069280553e-06, + "loss": 1.1994, + "step": 30862 + }, + { + "epoch": 0.74, + "learning_rate": 3.2874589487689467e-06, + "loss": 1.1085, + "step": 30863 + }, + { + "epoch": 0.74, + "learning_rate": 3.286881631333242e-06, + "loss": 1.121, + "step": 30864 + }, + { + "epoch": 0.74, + "learning_rate": 3.2863043546244468e-06, + "loss": 1.1299, + "step": 30865 + }, + { + "epoch": 0.74, + "learning_rate": 3.2857271186460672e-06, + "loss": 0.976, + "step": 30866 + }, + { + "epoch": 0.74, + "learning_rate": 3.285149923401599e-06, + "loss": 0.8063, + "step": 30867 + }, + { + "epoch": 0.74, + "learning_rate": 3.2845727688945504e-06, + "loss": 1.2629, + "step": 30868 + }, + { + "epoch": 0.74, + "learning_rate": 3.283995655128416e-06, + "loss": 1.1331, + "step": 30869 + }, + { + "epoch": 0.74, + "learning_rate": 3.2834185821067e-06, + "loss": 0.8579, + "step": 30870 + }, + { + "epoch": 0.74, + "learning_rate": 3.282841549832907e-06, + "loss": 1.2303, + "step": 30871 + }, + { + "epoch": 0.74, + "learning_rate": 3.2822645583105306e-06, + "loss": 1.1456, + "step": 30872 + }, + { + "epoch": 0.74, + "learning_rate": 3.2816876075430746e-06, + "loss": 0.9608, + "step": 30873 + }, + { + "epoch": 0.74, + "learning_rate": 3.2811106975340445e-06, + "loss": 0.8793, + "step": 30874 + }, + { + "epoch": 0.74, + "learning_rate": 3.2805338282869304e-06, + "loss": 1.0714, + "step": 30875 + }, + { + "epoch": 0.74, + "learning_rate": 3.279956999805238e-06, + "loss": 1.1363, + "step": 30876 + }, + { + "epoch": 0.74, + "learning_rate": 3.279380212092469e-06, + "loss": 1.1787, + "step": 30877 + }, + { + "epoch": 0.74, + "learning_rate": 3.278803465152116e-06, + "loss": 1.1322, + "step": 30878 + }, + { + "epoch": 0.74, + "learning_rate": 3.278226758987686e-06, + "loss": 0.9379, + "step": 30879 + }, + { + "epoch": 0.74, + "learning_rate": 3.277650093602668e-06, + "loss": 0.987, + "step": 30880 + }, + { + "epoch": 0.74, + "learning_rate": 3.277073469000567e-06, + "loss": 0.8487, + "step": 30881 + }, + { + "epoch": 0.74, + "learning_rate": 3.2764968851848822e-06, + "loss": 1.2369, + "step": 30882 + }, + { + "epoch": 0.74, + "learning_rate": 3.275920342159107e-06, + "loss": 1.061, + "step": 30883 + }, + { + "epoch": 0.74, + "learning_rate": 3.2753438399267414e-06, + "loss": 0.9305, + "step": 30884 + }, + { + "epoch": 0.74, + "learning_rate": 3.274767378491286e-06, + "loss": 0.9824, + "step": 30885 + }, + { + "epoch": 0.74, + "learning_rate": 3.2741909578562314e-06, + "loss": 0.9887, + "step": 30886 + }, + { + "epoch": 0.74, + "learning_rate": 3.2736145780250815e-06, + "loss": 1.0096, + "step": 30887 + }, + { + "epoch": 0.74, + "learning_rate": 3.273038239001327e-06, + "loss": 0.8647, + "step": 30888 + }, + { + "epoch": 0.74, + "learning_rate": 3.2724619407884674e-06, + "loss": 0.9762, + "step": 30889 + }, + { + "epoch": 0.74, + "learning_rate": 3.2718856833900013e-06, + "loss": 1.1662, + "step": 30890 + }, + { + "epoch": 0.74, + "learning_rate": 3.27130946680942e-06, + "loss": 1.0484, + "step": 30891 + }, + { + "epoch": 0.74, + "learning_rate": 3.270733291050221e-06, + "loss": 0.9861, + "step": 30892 + }, + { + "epoch": 0.74, + "learning_rate": 3.270157156115904e-06, + "loss": 1.053, + "step": 30893 + }, + { + "epoch": 0.74, + "learning_rate": 3.2695810620099566e-06, + "loss": 1.0233, + "step": 30894 + }, + { + "epoch": 0.74, + "learning_rate": 3.2690050087358805e-06, + "loss": 0.8437, + "step": 30895 + }, + { + "epoch": 0.74, + "learning_rate": 3.268428996297166e-06, + "loss": 0.793, + "step": 30896 + }, + { + "epoch": 0.74, + "learning_rate": 3.2678530246973084e-06, + "loss": 1.187, + "step": 30897 + }, + { + "epoch": 0.74, + "learning_rate": 3.267277093939807e-06, + "loss": 0.9629, + "step": 30898 + }, + { + "epoch": 0.74, + "learning_rate": 3.2667012040281475e-06, + "loss": 0.8172, + "step": 30899 + }, + { + "epoch": 0.74, + "learning_rate": 3.266125354965829e-06, + "loss": 1.0219, + "step": 30900 + }, + { + "epoch": 0.74, + "learning_rate": 3.265549546756347e-06, + "loss": 0.9333, + "step": 30901 + }, + { + "epoch": 0.74, + "learning_rate": 3.264973779403189e-06, + "loss": 0.8971, + "step": 30902 + }, + { + "epoch": 0.74, + "learning_rate": 3.264398052909853e-06, + "loss": 1.2128, + "step": 30903 + }, + { + "epoch": 0.74, + "learning_rate": 3.2638223672798274e-06, + "loss": 1.0488, + "step": 30904 + }, + { + "epoch": 0.74, + "learning_rate": 3.2632467225166065e-06, + "loss": 1.066, + "step": 30905 + }, + { + "epoch": 0.74, + "learning_rate": 3.262671118623687e-06, + "loss": 0.9431, + "step": 30906 + }, + { + "epoch": 0.74, + "learning_rate": 3.2620955556045542e-06, + "loss": 1.1753, + "step": 30907 + }, + { + "epoch": 0.74, + "learning_rate": 3.2615200334627017e-06, + "loss": 0.9392, + "step": 30908 + }, + { + "epoch": 0.74, + "learning_rate": 3.260944552201626e-06, + "loss": 0.9636, + "step": 30909 + }, + { + "epoch": 0.74, + "learning_rate": 3.2603691118248115e-06, + "loss": 0.9456, + "step": 30910 + }, + { + "epoch": 0.74, + "learning_rate": 3.2597937123357547e-06, + "loss": 1.1025, + "step": 30911 + }, + { + "epoch": 0.74, + "learning_rate": 3.259218353737942e-06, + "loss": 1.156, + "step": 30912 + }, + { + "epoch": 0.74, + "learning_rate": 3.2586430360348654e-06, + "loss": 0.9161, + "step": 30913 + }, + { + "epoch": 0.74, + "learning_rate": 3.2580677592300192e-06, + "loss": 1.0854, + "step": 30914 + }, + { + "epoch": 0.74, + "learning_rate": 3.257492523326887e-06, + "loss": 0.5935, + "step": 30915 + }, + { + "epoch": 0.74, + "learning_rate": 3.256917328328961e-06, + "loss": 0.9367, + "step": 30916 + }, + { + "epoch": 0.74, + "learning_rate": 3.256342174239736e-06, + "loss": 0.8769, + "step": 30917 + }, + { + "epoch": 0.74, + "learning_rate": 3.2557670610626924e-06, + "loss": 1.1895, + "step": 30918 + }, + { + "epoch": 0.74, + "learning_rate": 3.2551919888013282e-06, + "loss": 1.0646, + "step": 30919 + }, + { + "epoch": 0.74, + "learning_rate": 3.2546169574591233e-06, + "loss": 0.9583, + "step": 30920 + }, + { + "epoch": 0.74, + "learning_rate": 3.2540419670395717e-06, + "loss": 1.1089, + "step": 30921 + }, + { + "epoch": 0.74, + "learning_rate": 3.2534670175461626e-06, + "loss": 1.1122, + "step": 30922 + }, + { + "epoch": 0.74, + "learning_rate": 3.2528921089823807e-06, + "loss": 0.9973, + "step": 30923 + }, + { + "epoch": 0.74, + "learning_rate": 3.2523172413517134e-06, + "loss": 1.1546, + "step": 30924 + }, + { + "epoch": 0.74, + "learning_rate": 3.2517424146576546e-06, + "loss": 1.0673, + "step": 30925 + }, + { + "epoch": 0.74, + "learning_rate": 3.2511676289036844e-06, + "loss": 0.7875, + "step": 30926 + }, + { + "epoch": 0.74, + "learning_rate": 3.250592884093292e-06, + "loss": 1.1078, + "step": 30927 + }, + { + "epoch": 0.74, + "learning_rate": 3.250018180229969e-06, + "loss": 0.8719, + "step": 30928 + }, + { + "epoch": 0.74, + "learning_rate": 3.2494435173171947e-06, + "loss": 0.8805, + "step": 30929 + }, + { + "epoch": 0.74, + "learning_rate": 3.2488688953584612e-06, + "loss": 0.9682, + "step": 30930 + }, + { + "epoch": 0.74, + "learning_rate": 3.24829431435725e-06, + "loss": 1.048, + "step": 30931 + }, + { + "epoch": 0.74, + "learning_rate": 3.2477197743170476e-06, + "loss": 1.0157, + "step": 30932 + }, + { + "epoch": 0.74, + "learning_rate": 3.2471452752413458e-06, + "loss": 1.2187, + "step": 30933 + }, + { + "epoch": 0.74, + "learning_rate": 3.2465708171336206e-06, + "loss": 1.0955, + "step": 30934 + }, + { + "epoch": 0.74, + "learning_rate": 3.245996399997363e-06, + "loss": 1.2729, + "step": 30935 + }, + { + "epoch": 0.74, + "learning_rate": 3.24542202383606e-06, + "loss": 0.9774, + "step": 30936 + }, + { + "epoch": 0.74, + "learning_rate": 3.2448476886531886e-06, + "loss": 1.0717, + "step": 30937 + }, + { + "epoch": 0.74, + "learning_rate": 3.244273394452241e-06, + "loss": 1.2503, + "step": 30938 + }, + { + "epoch": 0.74, + "learning_rate": 3.2436991412366946e-06, + "loss": 1.1731, + "step": 30939 + }, + { + "epoch": 0.74, + "learning_rate": 3.2431249290100365e-06, + "loss": 0.92, + "step": 30940 + }, + { + "epoch": 0.74, + "learning_rate": 3.242550757775752e-06, + "loss": 0.9906, + "step": 30941 + }, + { + "epoch": 0.74, + "learning_rate": 3.241976627537321e-06, + "loss": 0.9676, + "step": 30942 + }, + { + "epoch": 0.74, + "learning_rate": 3.2414025382982273e-06, + "loss": 1.1175, + "step": 30943 + }, + { + "epoch": 0.74, + "learning_rate": 3.240828490061958e-06, + "loss": 0.7319, + "step": 30944 + }, + { + "epoch": 0.74, + "learning_rate": 3.240254482831989e-06, + "loss": 1.2104, + "step": 30945 + }, + { + "epoch": 0.74, + "learning_rate": 3.2396805166118108e-06, + "loss": 1.2128, + "step": 30946 + }, + { + "epoch": 0.74, + "learning_rate": 3.2391065914048956e-06, + "loss": 0.8157, + "step": 30947 + }, + { + "epoch": 0.74, + "learning_rate": 3.2385327072147323e-06, + "loss": 1.114, + "step": 30948 + }, + { + "epoch": 0.74, + "learning_rate": 3.237958864044802e-06, + "loss": 1.2603, + "step": 30949 + }, + { + "epoch": 0.74, + "learning_rate": 3.2373850618985826e-06, + "loss": 1.1765, + "step": 30950 + }, + { + "epoch": 0.74, + "learning_rate": 3.2368113007795574e-06, + "loss": 1.1654, + "step": 30951 + }, + { + "epoch": 0.74, + "learning_rate": 3.2362375806912106e-06, + "loss": 1.1484, + "step": 30952 + }, + { + "epoch": 0.74, + "learning_rate": 3.235663901637015e-06, + "loss": 1.009, + "step": 30953 + }, + { + "epoch": 0.74, + "learning_rate": 3.2350902636204594e-06, + "loss": 0.9823, + "step": 30954 + }, + { + "epoch": 0.74, + "learning_rate": 3.234516666645018e-06, + "loss": 1.2431, + "step": 30955 + }, + { + "epoch": 0.74, + "learning_rate": 3.2339431107141707e-06, + "loss": 1.0448, + "step": 30956 + }, + { + "epoch": 0.74, + "learning_rate": 3.2333695958314037e-06, + "loss": 0.6662, + "step": 30957 + }, + { + "epoch": 0.74, + "learning_rate": 3.232796122000188e-06, + "loss": 1.0287, + "step": 30958 + }, + { + "epoch": 0.74, + "learning_rate": 3.232222689224006e-06, + "loss": 1.1792, + "step": 30959 + }, + { + "epoch": 0.74, + "learning_rate": 3.2316492975063407e-06, + "loss": 1.0114, + "step": 30960 + }, + { + "epoch": 0.74, + "learning_rate": 3.231075946850664e-06, + "loss": 1.2569, + "step": 30961 + }, + { + "epoch": 0.74, + "learning_rate": 3.2305026372604596e-06, + "loss": 1.2204, + "step": 30962 + }, + { + "epoch": 0.74, + "learning_rate": 3.229929368739202e-06, + "loss": 0.9803, + "step": 30963 + }, + { + "epoch": 0.74, + "learning_rate": 3.2293561412903697e-06, + "loss": 1.265, + "step": 30964 + }, + { + "epoch": 0.74, + "learning_rate": 3.2287829549174444e-06, + "loss": 0.9641, + "step": 30965 + }, + { + "epoch": 0.74, + "learning_rate": 3.2282098096238967e-06, + "loss": 0.8968, + "step": 30966 + }, + { + "epoch": 0.74, + "learning_rate": 3.2276367054132075e-06, + "loss": 1.2002, + "step": 30967 + }, + { + "epoch": 0.74, + "learning_rate": 3.2270636422888556e-06, + "loss": 1.1461, + "step": 30968 + }, + { + "epoch": 0.74, + "learning_rate": 3.2264906202543134e-06, + "loss": 0.9363, + "step": 30969 + }, + { + "epoch": 0.74, + "learning_rate": 3.2259176393130596e-06, + "loss": 0.9013, + "step": 30970 + }, + { + "epoch": 0.74, + "learning_rate": 3.2253446994685734e-06, + "loss": 1.1105, + "step": 30971 + }, + { + "epoch": 0.74, + "learning_rate": 3.2247718007243235e-06, + "loss": 1.3031, + "step": 30972 + }, + { + "epoch": 0.74, + "learning_rate": 3.224198943083793e-06, + "loss": 0.9633, + "step": 30973 + }, + { + "epoch": 0.74, + "learning_rate": 3.2236261265504508e-06, + "loss": 1.0647, + "step": 30974 + }, + { + "epoch": 0.74, + "learning_rate": 3.223053351127774e-06, + "loss": 1.184, + "step": 30975 + }, + { + "epoch": 0.74, + "learning_rate": 3.2224806168192423e-06, + "loss": 1.2707, + "step": 30976 + }, + { + "epoch": 0.74, + "learning_rate": 3.2219079236283236e-06, + "loss": 1.0899, + "step": 30977 + }, + { + "epoch": 0.74, + "learning_rate": 3.2213352715584955e-06, + "loss": 0.9771, + "step": 30978 + }, + { + "epoch": 0.75, + "learning_rate": 3.2207626606132347e-06, + "loss": 1.142, + "step": 30979 + }, + { + "epoch": 0.75, + "learning_rate": 3.220190090796008e-06, + "loss": 0.9617, + "step": 30980 + }, + { + "epoch": 0.75, + "learning_rate": 3.2196175621102977e-06, + "loss": 1.0973, + "step": 30981 + }, + { + "epoch": 0.75, + "learning_rate": 3.2190450745595693e-06, + "loss": 1.0042, + "step": 30982 + }, + { + "epoch": 0.75, + "learning_rate": 3.2184726281473e-06, + "loss": 1.2283, + "step": 30983 + }, + { + "epoch": 0.75, + "learning_rate": 3.2179002228769653e-06, + "loss": 1.0556, + "step": 30984 + }, + { + "epoch": 0.75, + "learning_rate": 3.2173278587520307e-06, + "loss": 1.0948, + "step": 30985 + }, + { + "epoch": 0.75, + "learning_rate": 3.216755535775973e-06, + "loss": 1.2589, + "step": 30986 + }, + { + "epoch": 0.75, + "learning_rate": 3.2161832539522666e-06, + "loss": 1.0347, + "step": 30987 + }, + { + "epoch": 0.75, + "learning_rate": 3.2156110132843775e-06, + "loss": 1.2651, + "step": 30988 + }, + { + "epoch": 0.75, + "learning_rate": 3.215038813775785e-06, + "loss": 1.1219, + "step": 30989 + }, + { + "epoch": 0.75, + "learning_rate": 3.214466655429952e-06, + "loss": 1.1145, + "step": 30990 + }, + { + "epoch": 0.75, + "learning_rate": 3.213894538250354e-06, + "loss": 1.1226, + "step": 30991 + }, + { + "epoch": 0.75, + "learning_rate": 3.2133224622404648e-06, + "loss": 1.0758, + "step": 30992 + }, + { + "epoch": 0.75, + "learning_rate": 3.212750427403749e-06, + "loss": 0.9256, + "step": 30993 + }, + { + "epoch": 0.75, + "learning_rate": 3.21217843374368e-06, + "loss": 0.9606, + "step": 30994 + }, + { + "epoch": 0.75, + "learning_rate": 3.2116064812637305e-06, + "loss": 1.1876, + "step": 30995 + }, + { + "epoch": 0.75, + "learning_rate": 3.211034569967365e-06, + "loss": 1.0936, + "step": 30996 + }, + { + "epoch": 0.75, + "learning_rate": 3.210462699858059e-06, + "loss": 1.0935, + "step": 30997 + }, + { + "epoch": 0.75, + "learning_rate": 3.209890870939275e-06, + "loss": 0.8905, + "step": 30998 + }, + { + "epoch": 0.75, + "learning_rate": 3.209319083214486e-06, + "loss": 0.8606, + "step": 30999 + }, + { + "epoch": 0.75, + "learning_rate": 3.2087473366871615e-06, + "loss": 1.042, + "step": 31000 + }, + { + "epoch": 0.75, + "learning_rate": 3.2081756313607694e-06, + "loss": 1.0433, + "step": 31001 + }, + { + "epoch": 0.75, + "learning_rate": 3.2076039672387817e-06, + "loss": 0.9807, + "step": 31002 + }, + { + "epoch": 0.75, + "learning_rate": 3.207032344324659e-06, + "loss": 1.0608, + "step": 31003 + }, + { + "epoch": 0.75, + "learning_rate": 3.2064607626218737e-06, + "loss": 0.883, + "step": 31004 + }, + { + "epoch": 0.75, + "learning_rate": 3.2058892221338966e-06, + "loss": 0.9252, + "step": 31005 + }, + { + "epoch": 0.75, + "learning_rate": 3.2053177228641872e-06, + "loss": 0.9815, + "step": 31006 + }, + { + "epoch": 0.75, + "learning_rate": 3.2047462648162176e-06, + "loss": 1.164, + "step": 31007 + }, + { + "epoch": 0.75, + "learning_rate": 3.2041748479934575e-06, + "loss": 1.1102, + "step": 31008 + }, + { + "epoch": 0.75, + "learning_rate": 3.203603472399367e-06, + "loss": 0.9021, + "step": 31009 + }, + { + "epoch": 0.75, + "learning_rate": 3.2030321380374185e-06, + "loss": 0.8511, + "step": 31010 + }, + { + "epoch": 0.75, + "learning_rate": 3.202460844911073e-06, + "loss": 0.9599, + "step": 31011 + }, + { + "epoch": 0.75, + "learning_rate": 3.201889593023797e-06, + "loss": 1.0058, + "step": 31012 + }, + { + "epoch": 0.75, + "learning_rate": 3.2013183823790618e-06, + "loss": 0.9647, + "step": 31013 + }, + { + "epoch": 0.75, + "learning_rate": 3.200747212980326e-06, + "loss": 0.995, + "step": 31014 + }, + { + "epoch": 0.75, + "learning_rate": 3.200176084831057e-06, + "loss": 1.0798, + "step": 31015 + }, + { + "epoch": 0.75, + "learning_rate": 3.199604997934724e-06, + "loss": 0.9876, + "step": 31016 + }, + { + "epoch": 0.75, + "learning_rate": 3.1990339522947845e-06, + "loss": 1.0111, + "step": 31017 + }, + { + "epoch": 0.75, + "learning_rate": 3.19846294791471e-06, + "loss": 0.6918, + "step": 31018 + }, + { + "epoch": 0.75, + "learning_rate": 3.197891984797957e-06, + "loss": 1.1585, + "step": 31019 + }, + { + "epoch": 0.75, + "learning_rate": 3.1973210629479936e-06, + "loss": 1.346, + "step": 31020 + }, + { + "epoch": 0.75, + "learning_rate": 3.196750182368287e-06, + "loss": 0.9796, + "step": 31021 + }, + { + "epoch": 0.75, + "learning_rate": 3.196179343062292e-06, + "loss": 1.0765, + "step": 31022 + }, + { + "epoch": 0.75, + "learning_rate": 3.1956085450334785e-06, + "loss": 1.2064, + "step": 31023 + }, + { + "epoch": 0.75, + "learning_rate": 3.19503778828531e-06, + "loss": 0.9409, + "step": 31024 + }, + { + "epoch": 0.75, + "learning_rate": 3.1944670728212426e-06, + "loss": 0.9211, + "step": 31025 + }, + { + "epoch": 0.75, + "learning_rate": 3.193896398644747e-06, + "loss": 0.9899, + "step": 31026 + }, + { + "epoch": 0.75, + "learning_rate": 3.1933257657592775e-06, + "loss": 1.0482, + "step": 31027 + }, + { + "epoch": 0.75, + "learning_rate": 3.1927551741682995e-06, + "loss": 0.9889, + "step": 31028 + }, + { + "epoch": 0.75, + "learning_rate": 3.192184623875277e-06, + "loss": 1.1657, + "step": 31029 + }, + { + "epoch": 0.75, + "learning_rate": 3.1916141148836678e-06, + "loss": 0.8865, + "step": 31030 + }, + { + "epoch": 0.75, + "learning_rate": 3.191043647196933e-06, + "loss": 1.0793, + "step": 31031 + }, + { + "epoch": 0.75, + "learning_rate": 3.1904732208185384e-06, + "loss": 1.2679, + "step": 31032 + }, + { + "epoch": 0.75, + "learning_rate": 3.1899028357519378e-06, + "loss": 1.3122, + "step": 31033 + }, + { + "epoch": 0.75, + "learning_rate": 3.1893324920005974e-06, + "loss": 1.0747, + "step": 31034 + }, + { + "epoch": 0.75, + "learning_rate": 3.188762189567973e-06, + "loss": 1.2033, + "step": 31035 + }, + { + "epoch": 0.75, + "learning_rate": 3.188191928457526e-06, + "loss": 1.0402, + "step": 31036 + }, + { + "epoch": 0.75, + "learning_rate": 3.1876217086727192e-06, + "loss": 1.1104, + "step": 31037 + }, + { + "epoch": 0.75, + "learning_rate": 3.187051530217005e-06, + "loss": 1.067, + "step": 31038 + }, + { + "epoch": 0.75, + "learning_rate": 3.1864813930938475e-06, + "loss": 1.0734, + "step": 31039 + }, + { + "epoch": 0.75, + "learning_rate": 3.1859112973067086e-06, + "loss": 1.2673, + "step": 31040 + }, + { + "epoch": 0.75, + "learning_rate": 3.185341242859039e-06, + "loss": 0.9143, + "step": 31041 + }, + { + "epoch": 0.75, + "learning_rate": 3.184771229754304e-06, + "loss": 1.0447, + "step": 31042 + }, + { + "epoch": 0.75, + "learning_rate": 3.1842012579959567e-06, + "loss": 0.938, + "step": 31043 + }, + { + "epoch": 0.75, + "learning_rate": 3.1836313275874564e-06, + "loss": 1.2695, + "step": 31044 + }, + { + "epoch": 0.75, + "learning_rate": 3.1830614385322655e-06, + "loss": 1.0001, + "step": 31045 + }, + { + "epoch": 0.75, + "learning_rate": 3.182491590833834e-06, + "loss": 1.1674, + "step": 31046 + }, + { + "epoch": 0.75, + "learning_rate": 3.1819217844956216e-06, + "loss": 0.9603, + "step": 31047 + }, + { + "epoch": 0.75, + "learning_rate": 3.1813520195210903e-06, + "loss": 1.1038, + "step": 31048 + }, + { + "epoch": 0.75, + "learning_rate": 3.1807822959136893e-06, + "loss": 1.2301, + "step": 31049 + }, + { + "epoch": 0.75, + "learning_rate": 3.1802126136768776e-06, + "loss": 0.8565, + "step": 31050 + }, + { + "epoch": 0.75, + "learning_rate": 3.179642972814115e-06, + "loss": 0.9845, + "step": 31051 + }, + { + "epoch": 0.75, + "learning_rate": 3.1790733733288513e-06, + "loss": 1.1122, + "step": 31052 + }, + { + "epoch": 0.75, + "learning_rate": 3.1785038152245485e-06, + "loss": 1.0769, + "step": 31053 + }, + { + "epoch": 0.75, + "learning_rate": 3.177934298504655e-06, + "loss": 1.0189, + "step": 31054 + }, + { + "epoch": 0.75, + "learning_rate": 3.1773648231726285e-06, + "loss": 1.0268, + "step": 31055 + }, + { + "epoch": 0.75, + "learning_rate": 3.1767953892319294e-06, + "loss": 0.9395, + "step": 31056 + }, + { + "epoch": 0.75, + "learning_rate": 3.1762259966860044e-06, + "loss": 1.048, + "step": 31057 + }, + { + "epoch": 0.75, + "learning_rate": 3.1756566455383107e-06, + "loss": 1.1638, + "step": 31058 + }, + { + "epoch": 0.75, + "learning_rate": 3.1750873357923064e-06, + "loss": 1.0358, + "step": 31059 + }, + { + "epoch": 0.75, + "learning_rate": 3.1745180674514388e-06, + "loss": 0.8361, + "step": 31060 + }, + { + "epoch": 0.75, + "learning_rate": 3.1739488405191676e-06, + "loss": 0.8759, + "step": 31061 + }, + { + "epoch": 0.75, + "learning_rate": 3.1733796549989403e-06, + "loss": 0.8567, + "step": 31062 + }, + { + "epoch": 0.75, + "learning_rate": 3.1728105108942133e-06, + "loss": 1.2009, + "step": 31063 + }, + { + "epoch": 0.75, + "learning_rate": 3.172241408208442e-06, + "loss": 0.9401, + "step": 31064 + }, + { + "epoch": 0.75, + "learning_rate": 3.171672346945072e-06, + "loss": 1.0249, + "step": 31065 + }, + { + "epoch": 0.75, + "learning_rate": 3.171103327107561e-06, + "loss": 1.1407, + "step": 31066 + }, + { + "epoch": 0.75, + "learning_rate": 3.170534348699362e-06, + "loss": 1.1696, + "step": 31067 + }, + { + "epoch": 0.75, + "learning_rate": 3.169965411723922e-06, + "loss": 1.2312, + "step": 31068 + }, + { + "epoch": 0.75, + "learning_rate": 3.1693965161846987e-06, + "loss": 1.0671, + "step": 31069 + }, + { + "epoch": 0.75, + "learning_rate": 3.1688276620851368e-06, + "loss": 1.1785, + "step": 31070 + }, + { + "epoch": 0.75, + "learning_rate": 3.1682588494286905e-06, + "loss": 1.0926, + "step": 31071 + }, + { + "epoch": 0.75, + "learning_rate": 3.1676900782188136e-06, + "loss": 0.9602, + "step": 31072 + }, + { + "epoch": 0.75, + "learning_rate": 3.1671213484589513e-06, + "loss": 1.1383, + "step": 31073 + }, + { + "epoch": 0.75, + "learning_rate": 3.1665526601525567e-06, + "loss": 1.0326, + "step": 31074 + }, + { + "epoch": 0.75, + "learning_rate": 3.165984013303083e-06, + "loss": 0.9409, + "step": 31075 + }, + { + "epoch": 0.75, + "learning_rate": 3.165415407913973e-06, + "loss": 1.1647, + "step": 31076 + }, + { + "epoch": 0.75, + "learning_rate": 3.1648468439886835e-06, + "loss": 1.0047, + "step": 31077 + }, + { + "epoch": 0.75, + "learning_rate": 3.1642783215306585e-06, + "loss": 0.6492, + "step": 31078 + }, + { + "epoch": 0.75, + "learning_rate": 3.1637098405433475e-06, + "loss": 1.2779, + "step": 31079 + }, + { + "epoch": 0.75, + "learning_rate": 3.163141401030205e-06, + "loss": 0.9794, + "step": 31080 + }, + { + "epoch": 0.75, + "learning_rate": 3.162573002994672e-06, + "loss": 1.0741, + "step": 31081 + }, + { + "epoch": 0.75, + "learning_rate": 3.1620046464401997e-06, + "loss": 1.1299, + "step": 31082 + }, + { + "epoch": 0.75, + "learning_rate": 3.161436331370241e-06, + "loss": 1.2715, + "step": 31083 + }, + { + "epoch": 0.75, + "learning_rate": 3.1608680577882367e-06, + "loss": 1.0105, + "step": 31084 + }, + { + "epoch": 0.75, + "learning_rate": 3.1602998256976394e-06, + "loss": 1.3011, + "step": 31085 + }, + { + "epoch": 0.75, + "learning_rate": 3.1597316351018914e-06, + "loss": 1.1633, + "step": 31086 + }, + { + "epoch": 0.75, + "learning_rate": 3.1591634860044428e-06, + "loss": 0.8578, + "step": 31087 + }, + { + "epoch": 0.75, + "learning_rate": 3.158595378408743e-06, + "loss": 1.1983, + "step": 31088 + }, + { + "epoch": 0.75, + "learning_rate": 3.1580273123182337e-06, + "loss": 1.1525, + "step": 31089 + }, + { + "epoch": 0.75, + "learning_rate": 3.157459287736362e-06, + "loss": 1.1284, + "step": 31090 + }, + { + "epoch": 0.75, + "learning_rate": 3.1568913046665794e-06, + "loss": 1.0442, + "step": 31091 + }, + { + "epoch": 0.75, + "learning_rate": 3.1563233631123237e-06, + "loss": 1.0567, + "step": 31092 + }, + { + "epoch": 0.75, + "learning_rate": 3.1557554630770447e-06, + "loss": 1.0134, + "step": 31093 + }, + { + "epoch": 0.75, + "learning_rate": 3.1551876045641915e-06, + "loss": 1.1249, + "step": 31094 + }, + { + "epoch": 0.75, + "learning_rate": 3.154619787577201e-06, + "loss": 0.8207, + "step": 31095 + }, + { + "epoch": 0.75, + "learning_rate": 3.154052012119526e-06, + "loss": 0.8044, + "step": 31096 + }, + { + "epoch": 0.75, + "learning_rate": 3.1534842781946027e-06, + "loss": 1.0803, + "step": 31097 + }, + { + "epoch": 0.75, + "learning_rate": 3.152916585805882e-06, + "loss": 1.0018, + "step": 31098 + }, + { + "epoch": 0.75, + "learning_rate": 3.152348934956807e-06, + "loss": 1.0362, + "step": 31099 + }, + { + "epoch": 0.75, + "learning_rate": 3.1517813256508177e-06, + "loss": 0.7949, + "step": 31100 + }, + { + "epoch": 0.75, + "learning_rate": 3.1512137578913614e-06, + "loss": 0.9065, + "step": 31101 + }, + { + "epoch": 0.75, + "learning_rate": 3.150646231681882e-06, + "loss": 1.1688, + "step": 31102 + }, + { + "epoch": 0.75, + "learning_rate": 3.1500787470258186e-06, + "loss": 0.8705, + "step": 31103 + }, + { + "epoch": 0.75, + "learning_rate": 3.149511303926619e-06, + "loss": 0.987, + "step": 31104 + }, + { + "epoch": 0.75, + "learning_rate": 3.14894390238772e-06, + "loss": 0.8315, + "step": 31105 + }, + { + "epoch": 0.75, + "learning_rate": 3.148376542412567e-06, + "loss": 1.1024, + "step": 31106 + }, + { + "epoch": 0.75, + "learning_rate": 3.1478092240046042e-06, + "loss": 1.1641, + "step": 31107 + }, + { + "epoch": 0.75, + "learning_rate": 3.147241947167269e-06, + "loss": 1.2096, + "step": 31108 + }, + { + "epoch": 0.75, + "learning_rate": 3.1466747119040054e-06, + "loss": 1.009, + "step": 31109 + }, + { + "epoch": 0.75, + "learning_rate": 3.1461075182182565e-06, + "loss": 0.9638, + "step": 31110 + }, + { + "epoch": 0.75, + "learning_rate": 3.1455403661134574e-06, + "loss": 1.0059, + "step": 31111 + }, + { + "epoch": 0.75, + "learning_rate": 3.1449732555930578e-06, + "loss": 1.0567, + "step": 31112 + }, + { + "epoch": 0.75, + "learning_rate": 3.144406186660488e-06, + "loss": 1.045, + "step": 31113 + }, + { + "epoch": 0.75, + "learning_rate": 3.1438391593191943e-06, + "loss": 1.0408, + "step": 31114 + }, + { + "epoch": 0.75, + "learning_rate": 3.143272173572619e-06, + "loss": 1.1226, + "step": 31115 + }, + { + "epoch": 0.75, + "learning_rate": 3.1427052294241943e-06, + "loss": 1.0304, + "step": 31116 + }, + { + "epoch": 0.75, + "learning_rate": 3.1421383268773655e-06, + "loss": 0.8918, + "step": 31117 + }, + { + "epoch": 0.75, + "learning_rate": 3.1415714659355736e-06, + "loss": 0.986, + "step": 31118 + }, + { + "epoch": 0.75, + "learning_rate": 3.1410046466022504e-06, + "loss": 1.1377, + "step": 31119 + }, + { + "epoch": 0.75, + "learning_rate": 3.140437868880841e-06, + "loss": 1.0295, + "step": 31120 + }, + { + "epoch": 0.75, + "learning_rate": 3.13987113277478e-06, + "loss": 0.9377, + "step": 31121 + }, + { + "epoch": 0.75, + "learning_rate": 3.139304438287507e-06, + "loss": 1.2306, + "step": 31122 + }, + { + "epoch": 0.75, + "learning_rate": 3.138737785422463e-06, + "loss": 1.0907, + "step": 31123 + }, + { + "epoch": 0.75, + "learning_rate": 3.13817117418308e-06, + "loss": 0.9292, + "step": 31124 + }, + { + "epoch": 0.75, + "learning_rate": 3.137604604572798e-06, + "loss": 1.0033, + "step": 31125 + }, + { + "epoch": 0.75, + "learning_rate": 3.1370380765950583e-06, + "loss": 1.0665, + "step": 31126 + }, + { + "epoch": 0.75, + "learning_rate": 3.1364715902532917e-06, + "loss": 0.9627, + "step": 31127 + }, + { + "epoch": 0.75, + "learning_rate": 3.1359051455509393e-06, + "loss": 0.8814, + "step": 31128 + }, + { + "epoch": 0.75, + "learning_rate": 3.1353387424914337e-06, + "loss": 1.0489, + "step": 31129 + }, + { + "epoch": 0.75, + "learning_rate": 3.134772381078214e-06, + "loss": 1.2804, + "step": 31130 + }, + { + "epoch": 0.75, + "learning_rate": 3.1342060613147175e-06, + "loss": 1.0518, + "step": 31131 + }, + { + "epoch": 0.75, + "learning_rate": 3.133639783204375e-06, + "loss": 1.1083, + "step": 31132 + }, + { + "epoch": 0.75, + "learning_rate": 3.1330735467506246e-06, + "loss": 1.1996, + "step": 31133 + }, + { + "epoch": 0.75, + "learning_rate": 3.1325073519569047e-06, + "loss": 1.2646, + "step": 31134 + }, + { + "epoch": 0.75, + "learning_rate": 3.131941198826646e-06, + "loss": 1.0967, + "step": 31135 + }, + { + "epoch": 0.75, + "learning_rate": 3.1313750873632854e-06, + "loss": 1.2457, + "step": 31136 + }, + { + "epoch": 0.75, + "learning_rate": 3.1308090175702545e-06, + "loss": 0.9299, + "step": 31137 + }, + { + "epoch": 0.75, + "learning_rate": 3.13024298945099e-06, + "loss": 1.0949, + "step": 31138 + }, + { + "epoch": 0.75, + "learning_rate": 3.1296770030089286e-06, + "loss": 1.0973, + "step": 31139 + }, + { + "epoch": 0.75, + "learning_rate": 3.129111058247497e-06, + "loss": 1.056, + "step": 31140 + }, + { + "epoch": 0.75, + "learning_rate": 3.128545155170132e-06, + "loss": 1.1755, + "step": 31141 + }, + { + "epoch": 0.75, + "learning_rate": 3.127979293780271e-06, + "loss": 0.8732, + "step": 31142 + }, + { + "epoch": 0.75, + "learning_rate": 3.12741347408134e-06, + "loss": 0.9837, + "step": 31143 + }, + { + "epoch": 0.75, + "learning_rate": 3.1268476960767756e-06, + "loss": 1.0582, + "step": 31144 + }, + { + "epoch": 0.75, + "learning_rate": 3.126281959770012e-06, + "loss": 1.252, + "step": 31145 + }, + { + "epoch": 0.75, + "learning_rate": 3.1257162651644756e-06, + "loss": 0.992, + "step": 31146 + }, + { + "epoch": 0.75, + "learning_rate": 3.1251506122636057e-06, + "loss": 0.788, + "step": 31147 + }, + { + "epoch": 0.75, + "learning_rate": 3.1245850010708266e-06, + "loss": 1.0046, + "step": 31148 + }, + { + "epoch": 0.75, + "learning_rate": 3.124019431589572e-06, + "loss": 1.04, + "step": 31149 + }, + { + "epoch": 0.75, + "learning_rate": 3.123453903823279e-06, + "loss": 1.0531, + "step": 31150 + }, + { + "epoch": 0.75, + "learning_rate": 3.1228884177753693e-06, + "loss": 1.0381, + "step": 31151 + }, + { + "epoch": 0.75, + "learning_rate": 3.122322973449279e-06, + "loss": 1.219, + "step": 31152 + }, + { + "epoch": 0.75, + "learning_rate": 3.1217575708484394e-06, + "loss": 1.0516, + "step": 31153 + }, + { + "epoch": 0.75, + "learning_rate": 3.121192209976277e-06, + "loss": 1.146, + "step": 31154 + }, + { + "epoch": 0.75, + "learning_rate": 3.120626890836226e-06, + "loss": 1.1141, + "step": 31155 + }, + { + "epoch": 0.75, + "learning_rate": 3.12006161343171e-06, + "loss": 0.9677, + "step": 31156 + }, + { + "epoch": 0.75, + "learning_rate": 3.1194963777661615e-06, + "loss": 1.1844, + "step": 31157 + }, + { + "epoch": 0.75, + "learning_rate": 3.1189311838430147e-06, + "loss": 1.0324, + "step": 31158 + }, + { + "epoch": 0.75, + "learning_rate": 3.1183660316656893e-06, + "loss": 0.9706, + "step": 31159 + }, + { + "epoch": 0.75, + "learning_rate": 3.1178009212376193e-06, + "loss": 1.113, + "step": 31160 + }, + { + "epoch": 0.75, + "learning_rate": 3.1172358525622347e-06, + "loss": 1.0887, + "step": 31161 + }, + { + "epoch": 0.75, + "learning_rate": 3.116670825642959e-06, + "loss": 0.9612, + "step": 31162 + }, + { + "epoch": 0.75, + "learning_rate": 3.1161058404832256e-06, + "loss": 1.0034, + "step": 31163 + }, + { + "epoch": 0.75, + "learning_rate": 3.1155408970864555e-06, + "loss": 0.8653, + "step": 31164 + }, + { + "epoch": 0.75, + "learning_rate": 3.1149759954560786e-06, + "loss": 0.8492, + "step": 31165 + }, + { + "epoch": 0.75, + "learning_rate": 3.1144111355955243e-06, + "loss": 0.8974, + "step": 31166 + }, + { + "epoch": 0.75, + "learning_rate": 3.1138463175082167e-06, + "loss": 1.095, + "step": 31167 + }, + { + "epoch": 0.75, + "learning_rate": 3.113281541197588e-06, + "loss": 0.8938, + "step": 31168 + }, + { + "epoch": 0.75, + "learning_rate": 3.1127168066670567e-06, + "loss": 1.036, + "step": 31169 + }, + { + "epoch": 0.75, + "learning_rate": 3.112152113920053e-06, + "loss": 0.963, + "step": 31170 + }, + { + "epoch": 0.75, + "learning_rate": 3.111587462960005e-06, + "loss": 1.1273, + "step": 31171 + }, + { + "epoch": 0.75, + "learning_rate": 3.111022853790332e-06, + "loss": 1.021, + "step": 31172 + }, + { + "epoch": 0.75, + "learning_rate": 3.110458286414464e-06, + "loss": 1.1622, + "step": 31173 + }, + { + "epoch": 0.75, + "learning_rate": 3.109893760835827e-06, + "loss": 0.8142, + "step": 31174 + }, + { + "epoch": 0.75, + "learning_rate": 3.109329277057842e-06, + "loss": 1.2463, + "step": 31175 + }, + { + "epoch": 0.75, + "learning_rate": 3.1087648350839382e-06, + "loss": 1.0849, + "step": 31176 + }, + { + "epoch": 0.75, + "learning_rate": 3.108200434917533e-06, + "loss": 1.1723, + "step": 31177 + }, + { + "epoch": 0.75, + "learning_rate": 3.107636076562056e-06, + "loss": 1.0988, + "step": 31178 + }, + { + "epoch": 0.75, + "learning_rate": 3.107071760020933e-06, + "loss": 1.0329, + "step": 31179 + }, + { + "epoch": 0.75, + "learning_rate": 3.1065074852975806e-06, + "loss": 0.9705, + "step": 31180 + }, + { + "epoch": 0.75, + "learning_rate": 3.105943252395426e-06, + "loss": 0.8723, + "step": 31181 + }, + { + "epoch": 0.75, + "learning_rate": 3.1053790613178945e-06, + "loss": 1.2671, + "step": 31182 + }, + { + "epoch": 0.75, + "learning_rate": 3.1048149120684045e-06, + "loss": 1.0391, + "step": 31183 + }, + { + "epoch": 0.75, + "learning_rate": 3.104250804650384e-06, + "loss": 1.1084, + "step": 31184 + }, + { + "epoch": 0.75, + "learning_rate": 3.1036867390672475e-06, + "loss": 1.0528, + "step": 31185 + }, + { + "epoch": 0.75, + "learning_rate": 3.1031227153224217e-06, + "loss": 1.0736, + "step": 31186 + }, + { + "epoch": 0.75, + "learning_rate": 3.102558733419332e-06, + "loss": 0.973, + "step": 31187 + }, + { + "epoch": 0.75, + "learning_rate": 3.1019947933613927e-06, + "loss": 0.9748, + "step": 31188 + }, + { + "epoch": 0.75, + "learning_rate": 3.1014308951520276e-06, + "loss": 1.1351, + "step": 31189 + }, + { + "epoch": 0.75, + "learning_rate": 3.1008670387946625e-06, + "loss": 1.1377, + "step": 31190 + }, + { + "epoch": 0.75, + "learning_rate": 3.1003032242927113e-06, + "loss": 1.0232, + "step": 31191 + }, + { + "epoch": 0.75, + "learning_rate": 3.0997394516496004e-06, + "loss": 0.8316, + "step": 31192 + }, + { + "epoch": 0.75, + "learning_rate": 3.0991757208687444e-06, + "loss": 0.7979, + "step": 31193 + }, + { + "epoch": 0.75, + "learning_rate": 3.0986120319535664e-06, + "loss": 0.9875, + "step": 31194 + }, + { + "epoch": 0.75, + "learning_rate": 3.0980483849074882e-06, + "loss": 0.8875, + "step": 31195 + }, + { + "epoch": 0.75, + "learning_rate": 3.0974847797339235e-06, + "loss": 1.3187, + "step": 31196 + }, + { + "epoch": 0.75, + "learning_rate": 3.096921216436296e-06, + "loss": 1.1486, + "step": 31197 + }, + { + "epoch": 0.75, + "learning_rate": 3.0963576950180275e-06, + "loss": 0.919, + "step": 31198 + }, + { + "epoch": 0.75, + "learning_rate": 3.0957942154825284e-06, + "loss": 0.9427, + "step": 31199 + }, + { + "epoch": 0.75, + "learning_rate": 3.0952307778332256e-06, + "loss": 0.8299, + "step": 31200 + }, + { + "epoch": 0.75, + "learning_rate": 3.0946673820735308e-06, + "loss": 0.7712, + "step": 31201 + }, + { + "epoch": 0.75, + "learning_rate": 3.0941040282068645e-06, + "loss": 1.1207, + "step": 31202 + }, + { + "epoch": 0.75, + "learning_rate": 3.0935407162366483e-06, + "loss": 1.2796, + "step": 31203 + }, + { + "epoch": 0.75, + "learning_rate": 3.092977446166292e-06, + "loss": 0.948, + "step": 31204 + }, + { + "epoch": 0.75, + "learning_rate": 3.092414217999217e-06, + "loss": 1.0566, + "step": 31205 + }, + { + "epoch": 0.75, + "learning_rate": 3.091851031738844e-06, + "loss": 0.9109, + "step": 31206 + }, + { + "epoch": 0.75, + "learning_rate": 3.091287887388582e-06, + "loss": 1.0041, + "step": 31207 + }, + { + "epoch": 0.75, + "learning_rate": 3.090724784951855e-06, + "loss": 0.9875, + "step": 31208 + }, + { + "epoch": 0.75, + "learning_rate": 3.0901617244320725e-06, + "loss": 0.958, + "step": 31209 + }, + { + "epoch": 0.75, + "learning_rate": 3.0895987058326528e-06, + "loss": 1.0872, + "step": 31210 + }, + { + "epoch": 0.75, + "learning_rate": 3.089035729157016e-06, + "loss": 1.0159, + "step": 31211 + }, + { + "epoch": 0.75, + "learning_rate": 3.0884727944085714e-06, + "loss": 0.8749, + "step": 31212 + }, + { + "epoch": 0.75, + "learning_rate": 3.0879099015907355e-06, + "loss": 0.9971, + "step": 31213 + }, + { + "epoch": 0.75, + "learning_rate": 3.087347050706928e-06, + "loss": 1.1603, + "step": 31214 + }, + { + "epoch": 0.75, + "learning_rate": 3.0867842417605565e-06, + "loss": 0.9564, + "step": 31215 + }, + { + "epoch": 0.75, + "learning_rate": 3.0862214747550434e-06, + "loss": 1.1335, + "step": 31216 + }, + { + "epoch": 0.75, + "learning_rate": 3.085658749693794e-06, + "loss": 0.7148, + "step": 31217 + }, + { + "epoch": 0.75, + "learning_rate": 3.0850960665802267e-06, + "loss": 0.8236, + "step": 31218 + }, + { + "epoch": 0.75, + "learning_rate": 3.0845334254177582e-06, + "loss": 0.9571, + "step": 31219 + }, + { + "epoch": 0.75, + "learning_rate": 3.0839708262097955e-06, + "loss": 0.9604, + "step": 31220 + }, + { + "epoch": 0.75, + "learning_rate": 3.0834082689597556e-06, + "loss": 1.0236, + "step": 31221 + }, + { + "epoch": 0.75, + "learning_rate": 3.0828457536710544e-06, + "loss": 1.0792, + "step": 31222 + }, + { + "epoch": 0.75, + "learning_rate": 3.082283280347097e-06, + "loss": 0.7226, + "step": 31223 + }, + { + "epoch": 0.75, + "learning_rate": 3.0817208489913e-06, + "loss": 1.0049, + "step": 31224 + }, + { + "epoch": 0.75, + "learning_rate": 3.0811584596070787e-06, + "loss": 1.1179, + "step": 31225 + }, + { + "epoch": 0.75, + "learning_rate": 3.0805961121978377e-06, + "loss": 1.0282, + "step": 31226 + }, + { + "epoch": 0.75, + "learning_rate": 3.0800338067669966e-06, + "loss": 1.0215, + "step": 31227 + }, + { + "epoch": 0.75, + "learning_rate": 3.07947154331796e-06, + "loss": 1.1836, + "step": 31228 + }, + { + "epoch": 0.75, + "learning_rate": 3.078909321854141e-06, + "loss": 1.1848, + "step": 31229 + }, + { + "epoch": 0.75, + "learning_rate": 3.0783471423789536e-06, + "loss": 0.8473, + "step": 31230 + }, + { + "epoch": 0.75, + "learning_rate": 3.0777850048958036e-06, + "loss": 1.0806, + "step": 31231 + }, + { + "epoch": 0.75, + "learning_rate": 3.0772229094081042e-06, + "loss": 1.0727, + "step": 31232 + }, + { + "epoch": 0.75, + "learning_rate": 3.076660855919268e-06, + "loss": 1.2627, + "step": 31233 + }, + { + "epoch": 0.75, + "learning_rate": 3.076098844432698e-06, + "loss": 0.8668, + "step": 31234 + }, + { + "epoch": 0.75, + "learning_rate": 3.0755368749518124e-06, + "loss": 1.1114, + "step": 31235 + }, + { + "epoch": 0.75, + "learning_rate": 3.0749749474800107e-06, + "loss": 0.9219, + "step": 31236 + }, + { + "epoch": 0.75, + "learning_rate": 3.074413062020708e-06, + "loss": 0.8409, + "step": 31237 + }, + { + "epoch": 0.75, + "learning_rate": 3.0738512185773163e-06, + "loss": 0.8538, + "step": 31238 + }, + { + "epoch": 0.75, + "learning_rate": 3.073289417153236e-06, + "loss": 1.2401, + "step": 31239 + }, + { + "epoch": 0.75, + "learning_rate": 3.0727276577518795e-06, + "loss": 0.9851, + "step": 31240 + }, + { + "epoch": 0.75, + "learning_rate": 3.072165940376659e-06, + "loss": 0.8543, + "step": 31241 + }, + { + "epoch": 0.75, + "learning_rate": 3.071604265030974e-06, + "loss": 1.2691, + "step": 31242 + }, + { + "epoch": 0.75, + "learning_rate": 3.0710426317182395e-06, + "loss": 1.051, + "step": 31243 + }, + { + "epoch": 0.75, + "learning_rate": 3.070481040441857e-06, + "loss": 0.9708, + "step": 31244 + }, + { + "epoch": 0.75, + "learning_rate": 3.0699194912052366e-06, + "loss": 1.0967, + "step": 31245 + }, + { + "epoch": 0.75, + "learning_rate": 3.069357984011786e-06, + "loss": 0.9623, + "step": 31246 + }, + { + "epoch": 0.75, + "learning_rate": 3.0687965188649083e-06, + "loss": 1.1604, + "step": 31247 + }, + { + "epoch": 0.75, + "learning_rate": 3.068235095768012e-06, + "loss": 1.1428, + "step": 31248 + }, + { + "epoch": 0.75, + "learning_rate": 3.0676737147245063e-06, + "loss": 1.1192, + "step": 31249 + }, + { + "epoch": 0.75, + "learning_rate": 3.0671123757377898e-06, + "loss": 0.9659, + "step": 31250 + }, + { + "epoch": 0.75, + "learning_rate": 3.0665510788112752e-06, + "loss": 1.0902, + "step": 31251 + }, + { + "epoch": 0.75, + "learning_rate": 3.065989823948361e-06, + "loss": 1.2327, + "step": 31252 + }, + { + "epoch": 0.75, + "learning_rate": 3.0654286111524545e-06, + "loss": 0.8409, + "step": 31253 + }, + { + "epoch": 0.75, + "learning_rate": 3.064867440426966e-06, + "loss": 0.9671, + "step": 31254 + }, + { + "epoch": 0.75, + "learning_rate": 3.0643063117752924e-06, + "loss": 1.1026, + "step": 31255 + }, + { + "epoch": 0.75, + "learning_rate": 3.0637452252008404e-06, + "loss": 0.9469, + "step": 31256 + }, + { + "epoch": 0.75, + "learning_rate": 3.0631841807070184e-06, + "loss": 0.9138, + "step": 31257 + }, + { + "epoch": 0.75, + "learning_rate": 3.0626231782972227e-06, + "loss": 1.0344, + "step": 31258 + }, + { + "epoch": 0.75, + "learning_rate": 3.0620622179748637e-06, + "loss": 1.0977, + "step": 31259 + }, + { + "epoch": 0.75, + "learning_rate": 3.061501299743338e-06, + "loss": 0.9732, + "step": 31260 + }, + { + "epoch": 0.75, + "learning_rate": 3.0609404236060525e-06, + "loss": 0.9687, + "step": 31261 + }, + { + "epoch": 0.75, + "learning_rate": 3.0603795895664125e-06, + "loss": 0.9822, + "step": 31262 + }, + { + "epoch": 0.75, + "learning_rate": 3.0598187976278126e-06, + "loss": 0.8755, + "step": 31263 + }, + { + "epoch": 0.75, + "learning_rate": 3.0592580477936606e-06, + "loss": 1.1698, + "step": 31264 + }, + { + "epoch": 0.75, + "learning_rate": 3.0586973400673613e-06, + "loss": 1.0626, + "step": 31265 + }, + { + "epoch": 0.75, + "learning_rate": 3.058136674452309e-06, + "loss": 1.0877, + "step": 31266 + }, + { + "epoch": 0.75, + "learning_rate": 3.0575760509519082e-06, + "loss": 0.9156, + "step": 31267 + }, + { + "epoch": 0.75, + "learning_rate": 3.0570154695695642e-06, + "loss": 0.892, + "step": 31268 + }, + { + "epoch": 0.75, + "learning_rate": 3.05645493030867e-06, + "loss": 1.1386, + "step": 31269 + }, + { + "epoch": 0.75, + "learning_rate": 3.055894433172635e-06, + "loss": 1.0315, + "step": 31270 + }, + { + "epoch": 0.75, + "learning_rate": 3.0553339781648517e-06, + "loss": 1.1613, + "step": 31271 + }, + { + "epoch": 0.75, + "learning_rate": 3.0547735652887236e-06, + "loss": 1.009, + "step": 31272 + }, + { + "epoch": 0.75, + "learning_rate": 3.0542131945476527e-06, + "loss": 0.8529, + "step": 31273 + }, + { + "epoch": 0.75, + "learning_rate": 3.0536528659450348e-06, + "loss": 0.9411, + "step": 31274 + }, + { + "epoch": 0.75, + "learning_rate": 3.0530925794842702e-06, + "loss": 1.074, + "step": 31275 + }, + { + "epoch": 0.75, + "learning_rate": 3.052532335168763e-06, + "loss": 1.0916, + "step": 31276 + }, + { + "epoch": 0.75, + "learning_rate": 3.0519721330019048e-06, + "loss": 0.9645, + "step": 31277 + }, + { + "epoch": 0.75, + "learning_rate": 3.0514119729870995e-06, + "loss": 1.0252, + "step": 31278 + }, + { + "epoch": 0.75, + "learning_rate": 3.050851855127741e-06, + "loss": 1.0071, + "step": 31279 + }, + { + "epoch": 0.75, + "learning_rate": 3.050291779427229e-06, + "loss": 1.1806, + "step": 31280 + }, + { + "epoch": 0.75, + "learning_rate": 3.0497317458889662e-06, + "loss": 1.151, + "step": 31281 + }, + { + "epoch": 0.75, + "learning_rate": 3.0491717545163426e-06, + "loss": 0.9788, + "step": 31282 + }, + { + "epoch": 0.75, + "learning_rate": 3.0486118053127587e-06, + "loss": 1.0957, + "step": 31283 + }, + { + "epoch": 0.75, + "learning_rate": 3.0480518982816155e-06, + "loss": 0.8956, + "step": 31284 + }, + { + "epoch": 0.75, + "learning_rate": 3.0474920334263026e-06, + "loss": 1.0154, + "step": 31285 + }, + { + "epoch": 0.75, + "learning_rate": 3.046932210750224e-06, + "loss": 1.1919, + "step": 31286 + }, + { + "epoch": 0.75, + "learning_rate": 3.0463724302567686e-06, + "loss": 0.9788, + "step": 31287 + }, + { + "epoch": 0.75, + "learning_rate": 3.045812691949336e-06, + "loss": 0.9729, + "step": 31288 + }, + { + "epoch": 0.75, + "learning_rate": 3.045252995831326e-06, + "loss": 1.1643, + "step": 31289 + }, + { + "epoch": 0.75, + "learning_rate": 3.0446933419061263e-06, + "loss": 1.1674, + "step": 31290 + }, + { + "epoch": 0.75, + "learning_rate": 3.0441337301771366e-06, + "loss": 1.0706, + "step": 31291 + }, + { + "epoch": 0.75, + "learning_rate": 3.0435741606477543e-06, + "loss": 0.9208, + "step": 31292 + }, + { + "epoch": 0.75, + "learning_rate": 3.0430146333213672e-06, + "loss": 1.2407, + "step": 31293 + }, + { + "epoch": 0.75, + "learning_rate": 3.0424551482013787e-06, + "loss": 0.8205, + "step": 31294 + }, + { + "epoch": 0.75, + "learning_rate": 3.041895705291175e-06, + "loss": 1.0504, + "step": 31295 + }, + { + "epoch": 0.75, + "learning_rate": 3.0413363045941525e-06, + "loss": 1.1986, + "step": 31296 + }, + { + "epoch": 0.75, + "learning_rate": 3.0407769461137093e-06, + "loss": 0.9777, + "step": 31297 + }, + { + "epoch": 0.75, + "learning_rate": 3.040217629853233e-06, + "loss": 1.3293, + "step": 31298 + }, + { + "epoch": 0.75, + "learning_rate": 3.0396583558161185e-06, + "loss": 1.0377, + "step": 31299 + }, + { + "epoch": 0.75, + "learning_rate": 3.039099124005762e-06, + "loss": 0.9368, + "step": 31300 + }, + { + "epoch": 0.75, + "learning_rate": 3.038539934425552e-06, + "loss": 1.2194, + "step": 31301 + }, + { + "epoch": 0.75, + "learning_rate": 3.0379807870788857e-06, + "loss": 0.8377, + "step": 31302 + }, + { + "epoch": 0.75, + "learning_rate": 3.0374216819691494e-06, + "loss": 1.1994, + "step": 31303 + }, + { + "epoch": 0.75, + "learning_rate": 3.0368626190997374e-06, + "loss": 0.9964, + "step": 31304 + }, + { + "epoch": 0.75, + "learning_rate": 3.036303598474045e-06, + "loss": 1.058, + "step": 31305 + }, + { + "epoch": 0.75, + "learning_rate": 3.0357446200954586e-06, + "loss": 1.1622, + "step": 31306 + }, + { + "epoch": 0.75, + "learning_rate": 3.035185683967371e-06, + "loss": 1.0647, + "step": 31307 + }, + { + "epoch": 0.75, + "learning_rate": 3.034626790093177e-06, + "loss": 1.0276, + "step": 31308 + }, + { + "epoch": 0.75, + "learning_rate": 3.0340679384762606e-06, + "loss": 0.8793, + "step": 31309 + }, + { + "epoch": 0.75, + "learning_rate": 3.0335091291200193e-06, + "loss": 1.1376, + "step": 31310 + }, + { + "epoch": 0.75, + "learning_rate": 3.0329503620278354e-06, + "loss": 0.8111, + "step": 31311 + }, + { + "epoch": 0.75, + "learning_rate": 3.0323916372031024e-06, + "loss": 1.0343, + "step": 31312 + }, + { + "epoch": 0.75, + "learning_rate": 3.0318329546492154e-06, + "loss": 1.1022, + "step": 31313 + }, + { + "epoch": 0.75, + "learning_rate": 3.0312743143695544e-06, + "loss": 0.8571, + "step": 31314 + }, + { + "epoch": 0.75, + "learning_rate": 3.0307157163675127e-06, + "loss": 1.1044, + "step": 31315 + }, + { + "epoch": 0.75, + "learning_rate": 3.030157160646483e-06, + "loss": 1.0522, + "step": 31316 + }, + { + "epoch": 0.75, + "learning_rate": 3.0295986472098482e-06, + "loss": 1.1426, + "step": 31317 + }, + { + "epoch": 0.75, + "learning_rate": 3.0290401760609976e-06, + "loss": 0.7492, + "step": 31318 + }, + { + "epoch": 0.75, + "learning_rate": 3.0284817472033246e-06, + "loss": 1.0346, + "step": 31319 + }, + { + "epoch": 0.75, + "learning_rate": 3.0279233606402094e-06, + "loss": 0.703, + "step": 31320 + }, + { + "epoch": 0.75, + "learning_rate": 3.027365016375047e-06, + "loss": 0.9868, + "step": 31321 + }, + { + "epoch": 0.75, + "learning_rate": 3.0268067144112167e-06, + "loss": 0.7694, + "step": 31322 + }, + { + "epoch": 0.75, + "learning_rate": 3.0262484547521108e-06, + "loss": 0.7989, + "step": 31323 + }, + { + "epoch": 0.75, + "learning_rate": 3.0256902374011177e-06, + "loss": 1.1435, + "step": 31324 + }, + { + "epoch": 0.75, + "learning_rate": 3.025132062361619e-06, + "loss": 1.1212, + "step": 31325 + }, + { + "epoch": 0.75, + "learning_rate": 3.0245739296370035e-06, + "loss": 0.9713, + "step": 31326 + }, + { + "epoch": 0.75, + "learning_rate": 3.0240158392306595e-06, + "loss": 1.0429, + "step": 31327 + }, + { + "epoch": 0.75, + "learning_rate": 3.0234577911459686e-06, + "loss": 1.148, + "step": 31328 + }, + { + "epoch": 0.75, + "learning_rate": 3.0228997853863206e-06, + "loss": 1.028, + "step": 31329 + }, + { + "epoch": 0.75, + "learning_rate": 3.022341821955096e-06, + "loss": 1.123, + "step": 31330 + }, + { + "epoch": 0.75, + "learning_rate": 3.0217839008556816e-06, + "loss": 0.9636, + "step": 31331 + }, + { + "epoch": 0.75, + "learning_rate": 3.0212260220914636e-06, + "loss": 1.315, + "step": 31332 + }, + { + "epoch": 0.75, + "learning_rate": 3.020668185665825e-06, + "loss": 1.1274, + "step": 31333 + }, + { + "epoch": 0.75, + "learning_rate": 3.020110391582154e-06, + "loss": 0.8714, + "step": 31334 + }, + { + "epoch": 0.75, + "learning_rate": 3.019552639843829e-06, + "loss": 1.2366, + "step": 31335 + }, + { + "epoch": 0.75, + "learning_rate": 3.0189949304542365e-06, + "loss": 0.791, + "step": 31336 + }, + { + "epoch": 0.75, + "learning_rate": 3.0184372634167626e-06, + "loss": 0.9805, + "step": 31337 + }, + { + "epoch": 0.75, + "learning_rate": 3.017879638734784e-06, + "loss": 1.0583, + "step": 31338 + }, + { + "epoch": 0.75, + "learning_rate": 3.01732205641169e-06, + "loss": 1.0937, + "step": 31339 + }, + { + "epoch": 0.75, + "learning_rate": 3.0167645164508584e-06, + "loss": 1.0515, + "step": 31340 + }, + { + "epoch": 0.75, + "learning_rate": 3.0162070188556747e-06, + "loss": 0.9686, + "step": 31341 + }, + { + "epoch": 0.75, + "learning_rate": 3.0156495636295224e-06, + "loss": 0.8499, + "step": 31342 + }, + { + "epoch": 0.75, + "learning_rate": 3.015092150775778e-06, + "loss": 1.1207, + "step": 31343 + }, + { + "epoch": 0.75, + "learning_rate": 3.0145347802978274e-06, + "loss": 0.9546, + "step": 31344 + }, + { + "epoch": 0.75, + "learning_rate": 3.013977452199054e-06, + "loss": 0.969, + "step": 31345 + }, + { + "epoch": 0.75, + "learning_rate": 3.0134201664828324e-06, + "loss": 0.7935, + "step": 31346 + }, + { + "epoch": 0.75, + "learning_rate": 3.012862923152549e-06, + "loss": 1.3406, + "step": 31347 + }, + { + "epoch": 0.75, + "learning_rate": 3.0123057222115835e-06, + "loss": 1.1951, + "step": 31348 + }, + { + "epoch": 0.75, + "learning_rate": 3.0117485636633136e-06, + "loss": 1.0548, + "step": 31349 + }, + { + "epoch": 0.75, + "learning_rate": 3.0111914475111247e-06, + "loss": 1.1025, + "step": 31350 + }, + { + "epoch": 0.75, + "learning_rate": 3.010634373758391e-06, + "loss": 1.1341, + "step": 31351 + }, + { + "epoch": 0.75, + "learning_rate": 3.010077342408494e-06, + "loss": 1.1563, + "step": 31352 + }, + { + "epoch": 0.75, + "learning_rate": 3.0095203534648167e-06, + "loss": 1.0717, + "step": 31353 + }, + { + "epoch": 0.75, + "learning_rate": 3.0089634069307316e-06, + "loss": 1.0496, + "step": 31354 + }, + { + "epoch": 0.75, + "learning_rate": 3.008406502809621e-06, + "loss": 0.8677, + "step": 31355 + }, + { + "epoch": 0.75, + "learning_rate": 3.0078496411048675e-06, + "loss": 1.1713, + "step": 31356 + }, + { + "epoch": 0.75, + "learning_rate": 3.007292821819843e-06, + "loss": 1.2206, + "step": 31357 + }, + { + "epoch": 0.75, + "learning_rate": 3.0067360449579297e-06, + "loss": 1.2377, + "step": 31358 + }, + { + "epoch": 0.75, + "learning_rate": 3.0061793105225014e-06, + "loss": 0.9339, + "step": 31359 + }, + { + "epoch": 0.75, + "learning_rate": 3.005622618516938e-06, + "loss": 1.3019, + "step": 31360 + }, + { + "epoch": 0.75, + "learning_rate": 3.0050659689446205e-06, + "loss": 1.0451, + "step": 31361 + }, + { + "epoch": 0.75, + "learning_rate": 3.004509361808919e-06, + "loss": 1.0559, + "step": 31362 + }, + { + "epoch": 0.75, + "learning_rate": 3.003952797113214e-06, + "loss": 1.0917, + "step": 31363 + }, + { + "epoch": 0.75, + "learning_rate": 3.003396274860886e-06, + "loss": 1.2203, + "step": 31364 + }, + { + "epoch": 0.75, + "learning_rate": 3.0028397950553023e-06, + "loss": 0.8215, + "step": 31365 + }, + { + "epoch": 0.75, + "learning_rate": 3.002283357699848e-06, + "loss": 1.1315, + "step": 31366 + }, + { + "epoch": 0.75, + "learning_rate": 3.001726962797891e-06, + "loss": 1.3999, + "step": 31367 + }, + { + "epoch": 0.75, + "learning_rate": 3.001170610352809e-06, + "loss": 1.0996, + "step": 31368 + }, + { + "epoch": 0.75, + "learning_rate": 3.0006143003679845e-06, + "loss": 0.9404, + "step": 31369 + }, + { + "epoch": 0.75, + "learning_rate": 3.000058032846782e-06, + "loss": 1.1054, + "step": 31370 + }, + { + "epoch": 0.75, + "learning_rate": 2.9995018077925807e-06, + "loss": 0.8405, + "step": 31371 + }, + { + "epoch": 0.75, + "learning_rate": 2.998945625208759e-06, + "loss": 1.0499, + "step": 31372 + }, + { + "epoch": 0.75, + "learning_rate": 2.998389485098684e-06, + "loss": 0.6357, + "step": 31373 + }, + { + "epoch": 0.75, + "learning_rate": 2.9978333874657373e-06, + "loss": 1.0035, + "step": 31374 + }, + { + "epoch": 0.75, + "learning_rate": 2.997277332313284e-06, + "loss": 0.9196, + "step": 31375 + }, + { + "epoch": 0.75, + "learning_rate": 2.996721319644702e-06, + "loss": 1.0823, + "step": 31376 + }, + { + "epoch": 0.75, + "learning_rate": 2.996165349463368e-06, + "loss": 0.9961, + "step": 31377 + }, + { + "epoch": 0.75, + "learning_rate": 2.995609421772648e-06, + "loss": 1.0158, + "step": 31378 + }, + { + "epoch": 0.75, + "learning_rate": 2.9950535365759192e-06, + "loss": 1.1936, + "step": 31379 + }, + { + "epoch": 0.75, + "learning_rate": 2.9944976938765557e-06, + "loss": 1.0987, + "step": 31380 + }, + { + "epoch": 0.75, + "learning_rate": 2.993941893677923e-06, + "loss": 1.1352, + "step": 31381 + }, + { + "epoch": 0.75, + "learning_rate": 2.9933861359834004e-06, + "loss": 1.2647, + "step": 31382 + }, + { + "epoch": 0.75, + "learning_rate": 2.992830420796353e-06, + "loss": 0.996, + "step": 31383 + }, + { + "epoch": 0.75, + "learning_rate": 2.9922747481201554e-06, + "loss": 1.0942, + "step": 31384 + }, + { + "epoch": 0.75, + "learning_rate": 2.9917191179581818e-06, + "loss": 1.0229, + "step": 31385 + }, + { + "epoch": 0.75, + "learning_rate": 2.9911635303137964e-06, + "loss": 0.9667, + "step": 31386 + }, + { + "epoch": 0.75, + "learning_rate": 2.9906079851903726e-06, + "loss": 1.0013, + "step": 31387 + }, + { + "epoch": 0.75, + "learning_rate": 2.990052482591286e-06, + "loss": 1.1151, + "step": 31388 + }, + { + "epoch": 0.75, + "learning_rate": 2.989497022519898e-06, + "loss": 0.9738, + "step": 31389 + }, + { + "epoch": 0.75, + "learning_rate": 2.9889416049795826e-06, + "loss": 1.0702, + "step": 31390 + }, + { + "epoch": 0.75, + "learning_rate": 2.988386229973713e-06, + "loss": 0.8926, + "step": 31391 + }, + { + "epoch": 0.75, + "learning_rate": 2.9878308975056513e-06, + "loss": 0.9821, + "step": 31392 + }, + { + "epoch": 0.75, + "learning_rate": 2.9872756075787733e-06, + "loss": 1.0293, + "step": 31393 + }, + { + "epoch": 0.75, + "learning_rate": 2.9867203601964425e-06, + "loss": 1.2086, + "step": 31394 + }, + { + "epoch": 0.76, + "learning_rate": 2.9861651553620285e-06, + "loss": 1.2127, + "step": 31395 + }, + { + "epoch": 0.76, + "learning_rate": 2.985609993078905e-06, + "loss": 1.3033, + "step": 31396 + }, + { + "epoch": 0.76, + "learning_rate": 2.9850548733504316e-06, + "loss": 1.1342, + "step": 31397 + }, + { + "epoch": 0.76, + "learning_rate": 2.9844997961799814e-06, + "loss": 0.906, + "step": 31398 + }, + { + "epoch": 0.76, + "learning_rate": 2.9839447615709226e-06, + "loss": 0.9396, + "step": 31399 + }, + { + "epoch": 0.76, + "learning_rate": 2.983389769526619e-06, + "loss": 0.9615, + "step": 31400 + }, + { + "epoch": 0.76, + "learning_rate": 2.982834820050441e-06, + "loss": 0.9325, + "step": 31401 + }, + { + "epoch": 0.76, + "learning_rate": 2.982279913145751e-06, + "loss": 1.0028, + "step": 31402 + }, + { + "epoch": 0.76, + "learning_rate": 2.9817250488159176e-06, + "loss": 1.1053, + "step": 31403 + }, + { + "epoch": 0.76, + "learning_rate": 2.981170227064312e-06, + "loss": 1.0266, + "step": 31404 + }, + { + "epoch": 0.76, + "learning_rate": 2.9806154478942916e-06, + "loss": 0.9001, + "step": 31405 + }, + { + "epoch": 0.76, + "learning_rate": 2.9800607113092263e-06, + "loss": 0.8305, + "step": 31406 + }, + { + "epoch": 0.76, + "learning_rate": 2.9795060173124846e-06, + "loss": 0.9872, + "step": 31407 + }, + { + "epoch": 0.76, + "learning_rate": 2.9789513659074254e-06, + "loss": 0.9446, + "step": 31408 + }, + { + "epoch": 0.76, + "learning_rate": 2.9783967570974204e-06, + "loss": 1.0697, + "step": 31409 + }, + { + "epoch": 0.76, + "learning_rate": 2.9778421908858267e-06, + "loss": 1.1354, + "step": 31410 + }, + { + "epoch": 0.76, + "learning_rate": 2.9772876672760143e-06, + "loss": 1.0519, + "step": 31411 + }, + { + "epoch": 0.76, + "learning_rate": 2.976733186271348e-06, + "loss": 1.0909, + "step": 31412 + }, + { + "epoch": 0.76, + "learning_rate": 2.976178747875186e-06, + "loss": 0.8897, + "step": 31413 + }, + { + "epoch": 0.76, + "learning_rate": 2.9756243520908957e-06, + "loss": 0.9226, + "step": 31414 + }, + { + "epoch": 0.76, + "learning_rate": 2.9750699989218424e-06, + "loss": 1.1546, + "step": 31415 + }, + { + "epoch": 0.76, + "learning_rate": 2.9745156883713856e-06, + "loss": 0.9673, + "step": 31416 + }, + { + "epoch": 0.76, + "learning_rate": 2.973961420442891e-06, + "loss": 0.8962, + "step": 31417 + }, + { + "epoch": 0.76, + "learning_rate": 2.9734071951397182e-06, + "loss": 0.6831, + "step": 31418 + }, + { + "epoch": 0.76, + "learning_rate": 2.972853012465231e-06, + "loss": 1.0714, + "step": 31419 + }, + { + "epoch": 0.76, + "learning_rate": 2.972298872422794e-06, + "loss": 1.0747, + "step": 31420 + }, + { + "epoch": 0.76, + "learning_rate": 2.9717447750157647e-06, + "loss": 1.1681, + "step": 31421 + }, + { + "epoch": 0.76, + "learning_rate": 2.9711907202475054e-06, + "loss": 1.0363, + "step": 31422 + }, + { + "epoch": 0.76, + "learning_rate": 2.9706367081213825e-06, + "loss": 1.0684, + "step": 31423 + }, + { + "epoch": 0.76, + "learning_rate": 2.9700827386407504e-06, + "loss": 0.9649, + "step": 31424 + }, + { + "epoch": 0.76, + "learning_rate": 2.969528811808976e-06, + "loss": 1.0201, + "step": 31425 + }, + { + "epoch": 0.76, + "learning_rate": 2.9689749276294133e-06, + "loss": 0.9232, + "step": 31426 + }, + { + "epoch": 0.76, + "learning_rate": 2.9684210861054254e-06, + "loss": 1.164, + "step": 31427 + }, + { + "epoch": 0.76, + "learning_rate": 2.967867287240377e-06, + "loss": 1.1764, + "step": 31428 + }, + { + "epoch": 0.76, + "learning_rate": 2.96731353103762e-06, + "loss": 1.1152, + "step": 31429 + }, + { + "epoch": 0.76, + "learning_rate": 2.9667598175005176e-06, + "loss": 1.0219, + "step": 31430 + }, + { + "epoch": 0.76, + "learning_rate": 2.966206146632432e-06, + "loss": 0.978, + "step": 31431 + }, + { + "epoch": 0.76, + "learning_rate": 2.965652518436716e-06, + "loss": 1.2855, + "step": 31432 + }, + { + "epoch": 0.76, + "learning_rate": 2.965098932916736e-06, + "loss": 1.2119, + "step": 31433 + }, + { + "epoch": 0.76, + "learning_rate": 2.9645453900758415e-06, + "loss": 0.7412, + "step": 31434 + }, + { + "epoch": 0.76, + "learning_rate": 2.963991889917396e-06, + "loss": 0.8891, + "step": 31435 + }, + { + "epoch": 0.76, + "learning_rate": 2.9634384324447596e-06, + "loss": 0.9295, + "step": 31436 + }, + { + "epoch": 0.76, + "learning_rate": 2.9628850176612835e-06, + "loss": 0.9929, + "step": 31437 + }, + { + "epoch": 0.76, + "learning_rate": 2.96233164557033e-06, + "loss": 1.0235, + "step": 31438 + }, + { + "epoch": 0.76, + "learning_rate": 2.9617783161752565e-06, + "loss": 0.9131, + "step": 31439 + }, + { + "epoch": 0.76, + "learning_rate": 2.9612250294794166e-06, + "loss": 1.1605, + "step": 31440 + }, + { + "epoch": 0.76, + "learning_rate": 2.960671785486169e-06, + "loss": 1.1634, + "step": 31441 + }, + { + "epoch": 0.76, + "learning_rate": 2.9601185841988722e-06, + "loss": 0.9083, + "step": 31442 + }, + { + "epoch": 0.76, + "learning_rate": 2.959565425620877e-06, + "loss": 1.4556, + "step": 31443 + }, + { + "epoch": 0.76, + "learning_rate": 2.9590123097555466e-06, + "loss": 1.1422, + "step": 31444 + }, + { + "epoch": 0.76, + "learning_rate": 2.9584592366062283e-06, + "loss": 1.0637, + "step": 31445 + }, + { + "epoch": 0.76, + "learning_rate": 2.957906206176282e-06, + "loss": 1.1442, + "step": 31446 + }, + { + "epoch": 0.76, + "learning_rate": 2.9573532184690646e-06, + "loss": 0.8761, + "step": 31447 + }, + { + "epoch": 0.76, + "learning_rate": 2.9568002734879264e-06, + "loss": 0.9052, + "step": 31448 + }, + { + "epoch": 0.76, + "learning_rate": 2.9562473712362237e-06, + "loss": 1.1792, + "step": 31449 + }, + { + "epoch": 0.76, + "learning_rate": 2.9556945117173154e-06, + "loss": 1.2431, + "step": 31450 + }, + { + "epoch": 0.76, + "learning_rate": 2.955141694934548e-06, + "loss": 1.0876, + "step": 31451 + }, + { + "epoch": 0.76, + "learning_rate": 2.9545889208912816e-06, + "loss": 1.0495, + "step": 31452 + }, + { + "epoch": 0.76, + "learning_rate": 2.9540361895908642e-06, + "loss": 1.1646, + "step": 31453 + }, + { + "epoch": 0.76, + "learning_rate": 2.9534835010366515e-06, + "loss": 1.0728, + "step": 31454 + }, + { + "epoch": 0.76, + "learning_rate": 2.9529308552320014e-06, + "loss": 0.99, + "step": 31455 + }, + { + "epoch": 0.76, + "learning_rate": 2.952378252180258e-06, + "loss": 1.0838, + "step": 31456 + }, + { + "epoch": 0.76, + "learning_rate": 2.9518256918847776e-06, + "loss": 1.0851, + "step": 31457 + }, + { + "epoch": 0.76, + "learning_rate": 2.9512731743489165e-06, + "loss": 1.0411, + "step": 31458 + }, + { + "epoch": 0.76, + "learning_rate": 2.9507206995760207e-06, + "loss": 1.1733, + "step": 31459 + }, + { + "epoch": 0.76, + "learning_rate": 2.9501682675694466e-06, + "loss": 1.0099, + "step": 31460 + }, + { + "epoch": 0.76, + "learning_rate": 2.9496158783325412e-06, + "loss": 1.0873, + "step": 31461 + }, + { + "epoch": 0.76, + "learning_rate": 2.9490635318686567e-06, + "loss": 1.1111, + "step": 31462 + }, + { + "epoch": 0.76, + "learning_rate": 2.948511228181149e-06, + "loss": 1.1216, + "step": 31463 + }, + { + "epoch": 0.76, + "learning_rate": 2.947958967273362e-06, + "loss": 1.0835, + "step": 31464 + }, + { + "epoch": 0.76, + "learning_rate": 2.947406749148649e-06, + "loss": 1.1516, + "step": 31465 + }, + { + "epoch": 0.76, + "learning_rate": 2.9468545738103636e-06, + "loss": 1.1285, + "step": 31466 + }, + { + "epoch": 0.76, + "learning_rate": 2.94630244126185e-06, + "loss": 1.1612, + "step": 31467 + }, + { + "epoch": 0.76, + "learning_rate": 2.9457503515064624e-06, + "loss": 0.984, + "step": 31468 + }, + { + "epoch": 0.76, + "learning_rate": 2.9451983045475454e-06, + "loss": 0.9241, + "step": 31469 + }, + { + "epoch": 0.76, + "learning_rate": 2.944646300388452e-06, + "loss": 1.0203, + "step": 31470 + }, + { + "epoch": 0.76, + "learning_rate": 2.944094339032532e-06, + "loss": 1.31, + "step": 31471 + }, + { + "epoch": 0.76, + "learning_rate": 2.9435424204831285e-06, + "loss": 1.1728, + "step": 31472 + }, + { + "epoch": 0.76, + "learning_rate": 2.9429905447435937e-06, + "loss": 1.3156, + "step": 31473 + }, + { + "epoch": 0.76, + "learning_rate": 2.942438711817279e-06, + "loss": 1.0414, + "step": 31474 + }, + { + "epoch": 0.76, + "learning_rate": 2.9418869217075254e-06, + "loss": 1.1342, + "step": 31475 + }, + { + "epoch": 0.76, + "learning_rate": 2.9413351744176855e-06, + "loss": 0.9754, + "step": 31476 + }, + { + "epoch": 0.76, + "learning_rate": 2.9407834699511027e-06, + "loss": 0.8847, + "step": 31477 + }, + { + "epoch": 0.76, + "learning_rate": 2.940231808311127e-06, + "loss": 0.964, + "step": 31478 + }, + { + "epoch": 0.76, + "learning_rate": 2.9396801895011064e-06, + "loss": 0.8952, + "step": 31479 + }, + { + "epoch": 0.76, + "learning_rate": 2.9391286135243823e-06, + "loss": 1.0984, + "step": 31480 + }, + { + "epoch": 0.76, + "learning_rate": 2.938577080384304e-06, + "loss": 1.0939, + "step": 31481 + }, + { + "epoch": 0.76, + "learning_rate": 2.9380255900842213e-06, + "loss": 1.0516, + "step": 31482 + }, + { + "epoch": 0.76, + "learning_rate": 2.9374741426274734e-06, + "loss": 1.1606, + "step": 31483 + }, + { + "epoch": 0.76, + "learning_rate": 2.936922738017407e-06, + "loss": 1.1166, + "step": 31484 + }, + { + "epoch": 0.76, + "learning_rate": 2.9363713762573742e-06, + "loss": 0.9321, + "step": 31485 + }, + { + "epoch": 0.76, + "learning_rate": 2.93582005735071e-06, + "loss": 1.0366, + "step": 31486 + }, + { + "epoch": 0.76, + "learning_rate": 2.9352687813007685e-06, + "loss": 0.9013, + "step": 31487 + }, + { + "epoch": 0.76, + "learning_rate": 2.9347175481108857e-06, + "loss": 0.9929, + "step": 31488 + }, + { + "epoch": 0.76, + "learning_rate": 2.9341663577844093e-06, + "loss": 1.0785, + "step": 31489 + }, + { + "epoch": 0.76, + "learning_rate": 2.9336152103246873e-06, + "loss": 1.0639, + "step": 31490 + }, + { + "epoch": 0.76, + "learning_rate": 2.933064105735056e-06, + "loss": 1.2118, + "step": 31491 + }, + { + "epoch": 0.76, + "learning_rate": 2.932513044018863e-06, + "loss": 1.1566, + "step": 31492 + }, + { + "epoch": 0.76, + "learning_rate": 2.9319620251794544e-06, + "loss": 0.7533, + "step": 31493 + }, + { + "epoch": 0.76, + "learning_rate": 2.9314110492201665e-06, + "loss": 0.7847, + "step": 31494 + }, + { + "epoch": 0.76, + "learning_rate": 2.9308601161443475e-06, + "loss": 0.9796, + "step": 31495 + }, + { + "epoch": 0.76, + "learning_rate": 2.9303092259553346e-06, + "loss": 1.0282, + "step": 31496 + }, + { + "epoch": 0.76, + "learning_rate": 2.9297583786564733e-06, + "loss": 1.055, + "step": 31497 + }, + { + "epoch": 0.76, + "learning_rate": 2.929207574251107e-06, + "loss": 1.1326, + "step": 31498 + }, + { + "epoch": 0.76, + "learning_rate": 2.9286568127425697e-06, + "loss": 0.9559, + "step": 31499 + }, + { + "epoch": 0.76, + "learning_rate": 2.9281060941342143e-06, + "loss": 1.0066, + "step": 31500 + }, + { + "epoch": 0.76, + "learning_rate": 2.9275554184293733e-06, + "loss": 1.016, + "step": 31501 + }, + { + "epoch": 0.76, + "learning_rate": 2.927004785631389e-06, + "loss": 0.9503, + "step": 31502 + }, + { + "epoch": 0.76, + "learning_rate": 2.9264541957436067e-06, + "loss": 1.0714, + "step": 31503 + }, + { + "epoch": 0.76, + "learning_rate": 2.9259036487693604e-06, + "loss": 0.9316, + "step": 31504 + }, + { + "epoch": 0.76, + "learning_rate": 2.9253531447119954e-06, + "loss": 0.974, + "step": 31505 + }, + { + "epoch": 0.76, + "learning_rate": 2.9248026835748444e-06, + "loss": 1.0073, + "step": 31506 + }, + { + "epoch": 0.76, + "learning_rate": 2.9242522653612524e-06, + "loss": 0.9398, + "step": 31507 + }, + { + "epoch": 0.76, + "learning_rate": 2.9237018900745617e-06, + "loss": 1.1391, + "step": 31508 + }, + { + "epoch": 0.76, + "learning_rate": 2.9231515577181024e-06, + "loss": 1.0351, + "step": 31509 + }, + { + "epoch": 0.76, + "learning_rate": 2.9226012682952186e-06, + "loss": 1.1115, + "step": 31510 + }, + { + "epoch": 0.76, + "learning_rate": 2.9220510218092524e-06, + "loss": 1.1271, + "step": 31511 + }, + { + "epoch": 0.76, + "learning_rate": 2.921500818263533e-06, + "loss": 0.9198, + "step": 31512 + }, + { + "epoch": 0.76, + "learning_rate": 2.920950657661408e-06, + "loss": 1.0116, + "step": 31513 + }, + { + "epoch": 0.76, + "learning_rate": 2.9204005400062052e-06, + "loss": 0.9348, + "step": 31514 + }, + { + "epoch": 0.76, + "learning_rate": 2.9198504653012684e-06, + "loss": 0.9118, + "step": 31515 + }, + { + "epoch": 0.76, + "learning_rate": 2.9193004335499365e-06, + "loss": 0.7945, + "step": 31516 + }, + { + "epoch": 0.76, + "learning_rate": 2.9187504447555404e-06, + "loss": 1.0852, + "step": 31517 + }, + { + "epoch": 0.76, + "learning_rate": 2.9182004989214187e-06, + "loss": 1.1312, + "step": 31518 + }, + { + "epoch": 0.76, + "learning_rate": 2.9176505960509126e-06, + "loss": 0.8951, + "step": 31519 + }, + { + "epoch": 0.76, + "learning_rate": 2.9171007361473512e-06, + "loss": 1.2219, + "step": 31520 + }, + { + "epoch": 0.76, + "learning_rate": 2.916550919214074e-06, + "loss": 0.9738, + "step": 31521 + }, + { + "epoch": 0.76, + "learning_rate": 2.916001145254419e-06, + "loss": 0.9904, + "step": 31522 + }, + { + "epoch": 0.76, + "learning_rate": 2.915451414271716e-06, + "loss": 0.7358, + "step": 31523 + }, + { + "epoch": 0.76, + "learning_rate": 2.9149017262693056e-06, + "loss": 1.0144, + "step": 31524 + }, + { + "epoch": 0.76, + "learning_rate": 2.9143520812505177e-06, + "loss": 1.0307, + "step": 31525 + }, + { + "epoch": 0.76, + "learning_rate": 2.913802479218688e-06, + "loss": 0.9373, + "step": 31526 + }, + { + "epoch": 0.76, + "learning_rate": 2.9132529201771565e-06, + "loss": 0.9508, + "step": 31527 + }, + { + "epoch": 0.76, + "learning_rate": 2.912703404129248e-06, + "loss": 0.8133, + "step": 31528 + }, + { + "epoch": 0.76, + "learning_rate": 2.912153931078301e-06, + "loss": 1.0751, + "step": 31529 + }, + { + "epoch": 0.76, + "learning_rate": 2.911604501027653e-06, + "loss": 1.1066, + "step": 31530 + }, + { + "epoch": 0.76, + "learning_rate": 2.9110551139806288e-06, + "loss": 1.118, + "step": 31531 + }, + { + "epoch": 0.76, + "learning_rate": 2.9105057699405704e-06, + "loss": 1.159, + "step": 31532 + }, + { + "epoch": 0.76, + "learning_rate": 2.9099564689108015e-06, + "loss": 0.9226, + "step": 31533 + }, + { + "epoch": 0.76, + "learning_rate": 2.9094072108946592e-06, + "loss": 1.275, + "step": 31534 + }, + { + "epoch": 0.76, + "learning_rate": 2.90885799589548e-06, + "loss": 0.9796, + "step": 31535 + }, + { + "epoch": 0.76, + "learning_rate": 2.9083088239165867e-06, + "loss": 1.1378, + "step": 31536 + }, + { + "epoch": 0.76, + "learning_rate": 2.9077596949613164e-06, + "loss": 1.0684, + "step": 31537 + }, + { + "epoch": 0.76, + "learning_rate": 2.907210609033003e-06, + "loss": 1.089, + "step": 31538 + }, + { + "epoch": 0.76, + "learning_rate": 2.9066615661349706e-06, + "loss": 0.9429, + "step": 31539 + }, + { + "epoch": 0.76, + "learning_rate": 2.906112566270558e-06, + "loss": 1.0887, + "step": 31540 + }, + { + "epoch": 0.76, + "learning_rate": 2.9055636094430882e-06, + "loss": 1.0274, + "step": 31541 + }, + { + "epoch": 0.76, + "learning_rate": 2.9050146956558968e-06, + "loss": 0.9591, + "step": 31542 + }, + { + "epoch": 0.76, + "learning_rate": 2.9044658249123138e-06, + "loss": 1.1829, + "step": 31543 + }, + { + "epoch": 0.76, + "learning_rate": 2.9039169972156656e-06, + "loss": 1.0288, + "step": 31544 + }, + { + "epoch": 0.76, + "learning_rate": 2.903368212569283e-06, + "loss": 1.3513, + "step": 31545 + }, + { + "epoch": 0.76, + "learning_rate": 2.9028194709765e-06, + "loss": 1.0307, + "step": 31546 + }, + { + "epoch": 0.76, + "learning_rate": 2.9022707724406384e-06, + "loss": 1.0045, + "step": 31547 + }, + { + "epoch": 0.76, + "learning_rate": 2.9017221169650334e-06, + "loss": 0.8728, + "step": 31548 + }, + { + "epoch": 0.76, + "learning_rate": 2.9011735045530078e-06, + "loss": 1.0016, + "step": 31549 + }, + { + "epoch": 0.76, + "learning_rate": 2.900624935207893e-06, + "loss": 1.1707, + "step": 31550 + }, + { + "epoch": 0.76, + "learning_rate": 2.9000764089330203e-06, + "loss": 0.9058, + "step": 31551 + }, + { + "epoch": 0.76, + "learning_rate": 2.8995279257317097e-06, + "loss": 1.1129, + "step": 31552 + }, + { + "epoch": 0.76, + "learning_rate": 2.898979485607294e-06, + "loss": 0.9541, + "step": 31553 + }, + { + "epoch": 0.76, + "learning_rate": 2.898431088563103e-06, + "loss": 1.0737, + "step": 31554 + }, + { + "epoch": 0.76, + "learning_rate": 2.897882734602456e-06, + "loss": 1.0028, + "step": 31555 + }, + { + "epoch": 0.76, + "learning_rate": 2.8973344237286883e-06, + "loss": 0.9715, + "step": 31556 + }, + { + "epoch": 0.76, + "learning_rate": 2.8967861559451183e-06, + "loss": 0.9074, + "step": 31557 + }, + { + "epoch": 0.76, + "learning_rate": 2.8962379312550746e-06, + "loss": 1.0397, + "step": 31558 + }, + { + "epoch": 0.76, + "learning_rate": 2.89568974966189e-06, + "loss": 0.9172, + "step": 31559 + }, + { + "epoch": 0.76, + "learning_rate": 2.89514161116888e-06, + "loss": 0.8513, + "step": 31560 + }, + { + "epoch": 0.76, + "learning_rate": 2.8945935157793746e-06, + "loss": 1.3417, + "step": 31561 + }, + { + "epoch": 0.76, + "learning_rate": 2.894045463496703e-06, + "loss": 1.1953, + "step": 31562 + }, + { + "epoch": 0.76, + "learning_rate": 2.893497454324182e-06, + "loss": 0.983, + "step": 31563 + }, + { + "epoch": 0.76, + "learning_rate": 2.892949488265141e-06, + "loss": 1.3715, + "step": 31564 + }, + { + "epoch": 0.76, + "learning_rate": 2.8924015653229064e-06, + "loss": 0.9033, + "step": 31565 + }, + { + "epoch": 0.76, + "learning_rate": 2.8918536855007975e-06, + "loss": 0.8676, + "step": 31566 + }, + { + "epoch": 0.76, + "learning_rate": 2.891305848802143e-06, + "loss": 1.1483, + "step": 31567 + }, + { + "epoch": 0.76, + "learning_rate": 2.8907580552302607e-06, + "loss": 1.278, + "step": 31568 + }, + { + "epoch": 0.76, + "learning_rate": 2.8902103047884775e-06, + "loss": 1.0209, + "step": 31569 + }, + { + "epoch": 0.76, + "learning_rate": 2.8896625974801184e-06, + "loss": 1.0632, + "step": 31570 + }, + { + "epoch": 0.76, + "learning_rate": 2.889114933308501e-06, + "loss": 1.2082, + "step": 31571 + }, + { + "epoch": 0.76, + "learning_rate": 2.8885673122769507e-06, + "loss": 1.0446, + "step": 31572 + }, + { + "epoch": 0.76, + "learning_rate": 2.888019734388794e-06, + "loss": 1.0736, + "step": 31573 + }, + { + "epoch": 0.76, + "learning_rate": 2.8874721996473444e-06, + "loss": 0.882, + "step": 31574 + }, + { + "epoch": 0.76, + "learning_rate": 2.8869247080559314e-06, + "loss": 1.0367, + "step": 31575 + }, + { + "epoch": 0.76, + "learning_rate": 2.8863772596178697e-06, + "loss": 1.2296, + "step": 31576 + }, + { + "epoch": 0.76, + "learning_rate": 2.8858298543364847e-06, + "loss": 1.2498, + "step": 31577 + }, + { + "epoch": 0.76, + "learning_rate": 2.885282492215099e-06, + "loss": 0.8927, + "step": 31578 + }, + { + "epoch": 0.76, + "learning_rate": 2.8847351732570284e-06, + "loss": 1.2447, + "step": 31579 + }, + { + "epoch": 0.76, + "learning_rate": 2.8841878974655957e-06, + "loss": 0.9868, + "step": 31580 + }, + { + "epoch": 0.76, + "learning_rate": 2.883640664844124e-06, + "loss": 0.988, + "step": 31581 + }, + { + "epoch": 0.76, + "learning_rate": 2.883093475395927e-06, + "loss": 1.1519, + "step": 31582 + }, + { + "epoch": 0.76, + "learning_rate": 2.8825463291243315e-06, + "loss": 1.1539, + "step": 31583 + }, + { + "epoch": 0.76, + "learning_rate": 2.881999226032651e-06, + "loss": 1.159, + "step": 31584 + }, + { + "epoch": 0.76, + "learning_rate": 2.8814521661242066e-06, + "loss": 1.0905, + "step": 31585 + }, + { + "epoch": 0.76, + "learning_rate": 2.88090514940232e-06, + "loss": 1.0115, + "step": 31586 + }, + { + "epoch": 0.76, + "learning_rate": 2.8803581758703047e-06, + "loss": 1.0163, + "step": 31587 + }, + { + "epoch": 0.76, + "learning_rate": 2.879811245531482e-06, + "loss": 1.1563, + "step": 31588 + }, + { + "epoch": 0.76, + "learning_rate": 2.879264358389172e-06, + "loss": 1.0155, + "step": 31589 + }, + { + "epoch": 0.76, + "learning_rate": 2.8787175144466873e-06, + "loss": 0.8367, + "step": 31590 + }, + { + "epoch": 0.76, + "learning_rate": 2.8781707137073523e-06, + "loss": 0.9885, + "step": 31591 + }, + { + "epoch": 0.76, + "learning_rate": 2.8776239561744768e-06, + "loss": 1.17, + "step": 31592 + }, + { + "epoch": 0.76, + "learning_rate": 2.8770772418513802e-06, + "loss": 1.1527, + "step": 31593 + }, + { + "epoch": 0.76, + "learning_rate": 2.876530570741386e-06, + "loss": 1.2398, + "step": 31594 + }, + { + "epoch": 0.76, + "learning_rate": 2.875983942847801e-06, + "loss": 1.1318, + "step": 31595 + }, + { + "epoch": 0.76, + "learning_rate": 2.875437358173946e-06, + "loss": 1.0049, + "step": 31596 + }, + { + "epoch": 0.76, + "learning_rate": 2.8748908167231395e-06, + "loss": 1.2622, + "step": 31597 + }, + { + "epoch": 0.76, + "learning_rate": 2.874344318498691e-06, + "loss": 1.0598, + "step": 31598 + }, + { + "epoch": 0.76, + "learning_rate": 2.873797863503923e-06, + "loss": 1.0951, + "step": 31599 + }, + { + "epoch": 0.76, + "learning_rate": 2.873251451742145e-06, + "loss": 0.9344, + "step": 31600 + }, + { + "epoch": 0.76, + "learning_rate": 2.8727050832166727e-06, + "loss": 1.1644, + "step": 31601 + }, + { + "epoch": 0.76, + "learning_rate": 2.8721587579308265e-06, + "loss": 1.1265, + "step": 31602 + }, + { + "epoch": 0.76, + "learning_rate": 2.8716124758879127e-06, + "loss": 1.2651, + "step": 31603 + }, + { + "epoch": 0.76, + "learning_rate": 2.871066237091249e-06, + "loss": 0.9357, + "step": 31604 + }, + { + "epoch": 0.76, + "learning_rate": 2.8705200415441535e-06, + "loss": 1.0193, + "step": 31605 + }, + { + "epoch": 0.76, + "learning_rate": 2.869973889249933e-06, + "loss": 0.9915, + "step": 31606 + }, + { + "epoch": 0.76, + "learning_rate": 2.8694277802119064e-06, + "loss": 1.3263, + "step": 31607 + }, + { + "epoch": 0.76, + "learning_rate": 2.8688817144333816e-06, + "loss": 0.9078, + "step": 31608 + }, + { + "epoch": 0.76, + "learning_rate": 2.868335691917673e-06, + "loss": 1.1231, + "step": 31609 + }, + { + "epoch": 0.76, + "learning_rate": 2.8677897126680988e-06, + "loss": 1.0777, + "step": 31610 + }, + { + "epoch": 0.76, + "learning_rate": 2.8672437766879625e-06, + "loss": 0.8519, + "step": 31611 + }, + { + "epoch": 0.76, + "learning_rate": 2.8666978839805805e-06, + "loss": 0.706, + "step": 31612 + }, + { + "epoch": 0.76, + "learning_rate": 2.8661520345492687e-06, + "loss": 0.9596, + "step": 31613 + }, + { + "epoch": 0.76, + "learning_rate": 2.8656062283973307e-06, + "loss": 0.9295, + "step": 31614 + }, + { + "epoch": 0.76, + "learning_rate": 2.8650604655280822e-06, + "loss": 0.9221, + "step": 31615 + }, + { + "epoch": 0.76, + "learning_rate": 2.8645147459448363e-06, + "loss": 0.7437, + "step": 31616 + }, + { + "epoch": 0.76, + "learning_rate": 2.863969069650898e-06, + "loss": 0.9251, + "step": 31617 + }, + { + "epoch": 0.76, + "learning_rate": 2.863423436649584e-06, + "loss": 1.0895, + "step": 31618 + }, + { + "epoch": 0.76, + "learning_rate": 2.8628778469441974e-06, + "loss": 1.0362, + "step": 31619 + }, + { + "epoch": 0.76, + "learning_rate": 2.862332300538053e-06, + "loss": 1.1145, + "step": 31620 + }, + { + "epoch": 0.76, + "learning_rate": 2.8617867974344637e-06, + "loss": 0.9929, + "step": 31621 + }, + { + "epoch": 0.76, + "learning_rate": 2.861241337636731e-06, + "loss": 0.8803, + "step": 31622 + }, + { + "epoch": 0.76, + "learning_rate": 2.8606959211481676e-06, + "loss": 1.0822, + "step": 31623 + }, + { + "epoch": 0.76, + "learning_rate": 2.860150547972086e-06, + "loss": 1.1429, + "step": 31624 + }, + { + "epoch": 0.76, + "learning_rate": 2.8596052181117882e-06, + "loss": 0.9046, + "step": 31625 + }, + { + "epoch": 0.76, + "learning_rate": 2.8590599315705892e-06, + "loss": 1.0381, + "step": 31626 + }, + { + "epoch": 0.76, + "learning_rate": 2.858514688351791e-06, + "loss": 1.161, + "step": 31627 + }, + { + "epoch": 0.76, + "learning_rate": 2.857969488458705e-06, + "loss": 0.9641, + "step": 31628 + }, + { + "epoch": 0.76, + "learning_rate": 2.8574243318946415e-06, + "loss": 1.0991, + "step": 31629 + }, + { + "epoch": 0.76, + "learning_rate": 2.8568792186629e-06, + "loss": 1.2302, + "step": 31630 + }, + { + "epoch": 0.76, + "learning_rate": 2.856334148766793e-06, + "loss": 0.9925, + "step": 31631 + }, + { + "epoch": 0.76, + "learning_rate": 2.85578912220963e-06, + "loss": 1.028, + "step": 31632 + }, + { + "epoch": 0.76, + "learning_rate": 2.8552441389947106e-06, + "loss": 1.0327, + "step": 31633 + }, + { + "epoch": 0.76, + "learning_rate": 2.854699199125347e-06, + "loss": 1.0213, + "step": 31634 + }, + { + "epoch": 0.76, + "learning_rate": 2.85415430260484e-06, + "loss": 1.1457, + "step": 31635 + }, + { + "epoch": 0.76, + "learning_rate": 2.8536094494364976e-06, + "loss": 0.853, + "step": 31636 + }, + { + "epoch": 0.76, + "learning_rate": 2.8530646396236293e-06, + "loss": 0.9316, + "step": 31637 + }, + { + "epoch": 0.76, + "learning_rate": 2.8525198731695334e-06, + "loss": 0.9419, + "step": 31638 + }, + { + "epoch": 0.76, + "learning_rate": 2.851975150077518e-06, + "loss": 1.2309, + "step": 31639 + }, + { + "epoch": 0.76, + "learning_rate": 2.851430470350892e-06, + "loss": 1.0362, + "step": 31640 + }, + { + "epoch": 0.76, + "learning_rate": 2.8508858339929524e-06, + "loss": 0.9141, + "step": 31641 + }, + { + "epoch": 0.76, + "learning_rate": 2.8503412410070097e-06, + "loss": 1.0927, + "step": 31642 + }, + { + "epoch": 0.76, + "learning_rate": 2.8497966913963625e-06, + "loss": 1.1001, + "step": 31643 + }, + { + "epoch": 0.76, + "learning_rate": 2.8492521851643164e-06, + "loss": 0.7479, + "step": 31644 + }, + { + "epoch": 0.76, + "learning_rate": 2.848707722314178e-06, + "loss": 0.955, + "step": 31645 + }, + { + "epoch": 0.76, + "learning_rate": 2.848163302849245e-06, + "loss": 1.0169, + "step": 31646 + }, + { + "epoch": 0.76, + "learning_rate": 2.847618926772824e-06, + "loss": 0.9757, + "step": 31647 + }, + { + "epoch": 0.76, + "learning_rate": 2.8470745940882182e-06, + "loss": 1.0263, + "step": 31648 + }, + { + "epoch": 0.76, + "learning_rate": 2.846530304798727e-06, + "loss": 1.0134, + "step": 31649 + }, + { + "epoch": 0.76, + "learning_rate": 2.845986058907655e-06, + "loss": 1.2532, + "step": 31650 + }, + { + "epoch": 0.76, + "learning_rate": 2.8454418564183006e-06, + "loss": 1.1318, + "step": 31651 + }, + { + "epoch": 0.76, + "learning_rate": 2.8448976973339682e-06, + "loss": 1.0392, + "step": 31652 + }, + { + "epoch": 0.76, + "learning_rate": 2.8443535816579615e-06, + "loss": 1.1242, + "step": 31653 + }, + { + "epoch": 0.76, + "learning_rate": 2.8438095093935746e-06, + "loss": 0.8811, + "step": 31654 + }, + { + "epoch": 0.76, + "learning_rate": 2.843265480544113e-06, + "loss": 0.9603, + "step": 31655 + }, + { + "epoch": 0.76, + "learning_rate": 2.842721495112879e-06, + "loss": 0.9117, + "step": 31656 + }, + { + "epoch": 0.76, + "learning_rate": 2.842177553103167e-06, + "loss": 0.9775, + "step": 31657 + }, + { + "epoch": 0.76, + "learning_rate": 2.84163365451828e-06, + "loss": 1.1403, + "step": 31658 + }, + { + "epoch": 0.76, + "learning_rate": 2.8410897993615214e-06, + "loss": 1.0693, + "step": 31659 + }, + { + "epoch": 0.76, + "learning_rate": 2.8405459876361842e-06, + "loss": 0.9951, + "step": 31660 + }, + { + "epoch": 0.76, + "learning_rate": 2.8400022193455734e-06, + "loss": 1.2286, + "step": 31661 + }, + { + "epoch": 0.76, + "learning_rate": 2.8394584944929813e-06, + "loss": 0.7667, + "step": 31662 + }, + { + "epoch": 0.76, + "learning_rate": 2.838914813081711e-06, + "loss": 0.9096, + "step": 31663 + }, + { + "epoch": 0.76, + "learning_rate": 2.838371175115062e-06, + "loss": 0.9449, + "step": 31664 + }, + { + "epoch": 0.76, + "learning_rate": 2.8378275805963297e-06, + "loss": 0.9989, + "step": 31665 + }, + { + "epoch": 0.76, + "learning_rate": 2.8372840295288106e-06, + "loss": 1.2582, + "step": 31666 + }, + { + "epoch": 0.76, + "learning_rate": 2.8367405219158063e-06, + "loss": 0.8419, + "step": 31667 + }, + { + "epoch": 0.76, + "learning_rate": 2.8361970577606113e-06, + "loss": 1.1805, + "step": 31668 + }, + { + "epoch": 0.76, + "learning_rate": 2.8356536370665265e-06, + "loss": 1.1204, + "step": 31669 + }, + { + "epoch": 0.76, + "learning_rate": 2.8351102598368442e-06, + "loss": 0.8892, + "step": 31670 + }, + { + "epoch": 0.76, + "learning_rate": 2.834566926074864e-06, + "loss": 0.9996, + "step": 31671 + }, + { + "epoch": 0.76, + "learning_rate": 2.834023635783879e-06, + "loss": 0.896, + "step": 31672 + }, + { + "epoch": 0.76, + "learning_rate": 2.8334803889671868e-06, + "loss": 1.0353, + "step": 31673 + }, + { + "epoch": 0.76, + "learning_rate": 2.8329371856280865e-06, + "loss": 0.9948, + "step": 31674 + }, + { + "epoch": 0.76, + "learning_rate": 2.832394025769868e-06, + "loss": 0.7717, + "step": 31675 + }, + { + "epoch": 0.76, + "learning_rate": 2.831850909395828e-06, + "loss": 0.8356, + "step": 31676 + }, + { + "epoch": 0.76, + "learning_rate": 2.8313078365092662e-06, + "loss": 1.035, + "step": 31677 + }, + { + "epoch": 0.76, + "learning_rate": 2.83076480711347e-06, + "loss": 1.1485, + "step": 31678 + }, + { + "epoch": 0.76, + "learning_rate": 2.8302218212117403e-06, + "loss": 1.0747, + "step": 31679 + }, + { + "epoch": 0.76, + "learning_rate": 2.829678878807366e-06, + "loss": 0.8967, + "step": 31680 + }, + { + "epoch": 0.76, + "learning_rate": 2.8291359799036435e-06, + "loss": 0.9954, + "step": 31681 + }, + { + "epoch": 0.76, + "learning_rate": 2.828593124503869e-06, + "loss": 1.0819, + "step": 31682 + }, + { + "epoch": 0.76, + "learning_rate": 2.82805031261133e-06, + "loss": 1.0493, + "step": 31683 + }, + { + "epoch": 0.76, + "learning_rate": 2.827507544229322e-06, + "loss": 1.1395, + "step": 31684 + }, + { + "epoch": 0.76, + "learning_rate": 2.8269648193611433e-06, + "loss": 1.0942, + "step": 31685 + }, + { + "epoch": 0.76, + "learning_rate": 2.8264221380100778e-06, + "loss": 0.9357, + "step": 31686 + }, + { + "epoch": 0.76, + "learning_rate": 2.8258795001794224e-06, + "loss": 0.7519, + "step": 31687 + }, + { + "epoch": 0.76, + "learning_rate": 2.8253369058724713e-06, + "loss": 1.2782, + "step": 31688 + }, + { + "epoch": 0.76, + "learning_rate": 2.8247943550925104e-06, + "loss": 0.9801, + "step": 31689 + }, + { + "epoch": 0.76, + "learning_rate": 2.824251847842838e-06, + "loss": 1.2173, + "step": 31690 + }, + { + "epoch": 0.76, + "learning_rate": 2.8237093841267393e-06, + "loss": 0.9005, + "step": 31691 + }, + { + "epoch": 0.76, + "learning_rate": 2.8231669639475068e-06, + "loss": 1.0629, + "step": 31692 + }, + { + "epoch": 0.76, + "learning_rate": 2.822624587308436e-06, + "loss": 0.7364, + "step": 31693 + }, + { + "epoch": 0.76, + "learning_rate": 2.822082254212809e-06, + "loss": 1.181, + "step": 31694 + }, + { + "epoch": 0.76, + "learning_rate": 2.821539964663922e-06, + "loss": 1.1366, + "step": 31695 + }, + { + "epoch": 0.76, + "learning_rate": 2.8209977186650663e-06, + "loss": 0.9766, + "step": 31696 + }, + { + "epoch": 0.76, + "learning_rate": 2.820455516219526e-06, + "loss": 0.8642, + "step": 31697 + }, + { + "epoch": 0.76, + "learning_rate": 2.819913357330596e-06, + "loss": 1.0706, + "step": 31698 + }, + { + "epoch": 0.76, + "learning_rate": 2.819371242001561e-06, + "loss": 0.9186, + "step": 31699 + }, + { + "epoch": 0.76, + "learning_rate": 2.8188291702357095e-06, + "loss": 1.1718, + "step": 31700 + }, + { + "epoch": 0.76, + "learning_rate": 2.818287142036337e-06, + "loss": 1.1704, + "step": 31701 + }, + { + "epoch": 0.76, + "learning_rate": 2.817745157406724e-06, + "loss": 1.0581, + "step": 31702 + }, + { + "epoch": 0.76, + "learning_rate": 2.8172032163501606e-06, + "loss": 0.969, + "step": 31703 + }, + { + "epoch": 0.76, + "learning_rate": 2.81666131886994e-06, + "loss": 0.7749, + "step": 31704 + }, + { + "epoch": 0.76, + "learning_rate": 2.8161194649693423e-06, + "loss": 1.0172, + "step": 31705 + }, + { + "epoch": 0.76, + "learning_rate": 2.8155776546516612e-06, + "loss": 0.8719, + "step": 31706 + }, + { + "epoch": 0.76, + "learning_rate": 2.8150358879201768e-06, + "loss": 1.0611, + "step": 31707 + }, + { + "epoch": 0.76, + "learning_rate": 2.8144941647781797e-06, + "loss": 0.977, + "step": 31708 + }, + { + "epoch": 0.76, + "learning_rate": 2.81395248522896e-06, + "loss": 1.1742, + "step": 31709 + }, + { + "epoch": 0.76, + "learning_rate": 2.813410849275797e-06, + "loss": 0.8949, + "step": 31710 + }, + { + "epoch": 0.76, + "learning_rate": 2.8128692569219806e-06, + "loss": 1.1482, + "step": 31711 + }, + { + "epoch": 0.76, + "learning_rate": 2.8123277081707977e-06, + "loss": 1.2321, + "step": 31712 + }, + { + "epoch": 0.76, + "learning_rate": 2.8117862030255304e-06, + "loss": 1.0545, + "step": 31713 + }, + { + "epoch": 0.76, + "learning_rate": 2.811244741489467e-06, + "loss": 0.9988, + "step": 31714 + }, + { + "epoch": 0.76, + "learning_rate": 2.8107033235658887e-06, + "loss": 0.9226, + "step": 31715 + }, + { + "epoch": 0.76, + "learning_rate": 2.810161949258082e-06, + "loss": 1.1088, + "step": 31716 + }, + { + "epoch": 0.76, + "learning_rate": 2.8096206185693355e-06, + "loss": 1.099, + "step": 31717 + }, + { + "epoch": 0.76, + "learning_rate": 2.8090793315029254e-06, + "loss": 1.1305, + "step": 31718 + }, + { + "epoch": 0.76, + "learning_rate": 2.8085380880621404e-06, + "loss": 1.1269, + "step": 31719 + }, + { + "epoch": 0.76, + "learning_rate": 2.807996888250265e-06, + "loss": 0.8831, + "step": 31720 + }, + { + "epoch": 0.76, + "learning_rate": 2.8074557320705788e-06, + "loss": 1.0237, + "step": 31721 + }, + { + "epoch": 0.76, + "learning_rate": 2.8069146195263695e-06, + "loss": 1.1495, + "step": 31722 + }, + { + "epoch": 0.76, + "learning_rate": 2.8063735506209144e-06, + "loss": 0.9381, + "step": 31723 + }, + { + "epoch": 0.76, + "learning_rate": 2.805832525357499e-06, + "loss": 1.1714, + "step": 31724 + }, + { + "epoch": 0.76, + "learning_rate": 2.805291543739409e-06, + "loss": 1.1381, + "step": 31725 + }, + { + "epoch": 0.76, + "learning_rate": 2.804750605769919e-06, + "loss": 1.1213, + "step": 31726 + }, + { + "epoch": 0.76, + "learning_rate": 2.8042097114523147e-06, + "loss": 1.1235, + "step": 31727 + }, + { + "epoch": 0.76, + "learning_rate": 2.803668860789881e-06, + "loss": 1.1031, + "step": 31728 + }, + { + "epoch": 0.76, + "learning_rate": 2.8031280537858927e-06, + "loss": 1.1378, + "step": 31729 + }, + { + "epoch": 0.76, + "learning_rate": 2.802587290443637e-06, + "loss": 1.1993, + "step": 31730 + }, + { + "epoch": 0.76, + "learning_rate": 2.802046570766388e-06, + "loss": 0.9409, + "step": 31731 + }, + { + "epoch": 0.76, + "learning_rate": 2.801505894757428e-06, + "loss": 1.1812, + "step": 31732 + }, + { + "epoch": 0.76, + "learning_rate": 2.8009652624200436e-06, + "loss": 1.13, + "step": 31733 + }, + { + "epoch": 0.76, + "learning_rate": 2.8004246737575057e-06, + "loss": 0.9749, + "step": 31734 + }, + { + "epoch": 0.76, + "learning_rate": 2.7998841287730972e-06, + "loss": 1.2442, + "step": 31735 + }, + { + "epoch": 0.76, + "learning_rate": 2.7993436274701024e-06, + "loss": 1.2753, + "step": 31736 + }, + { + "epoch": 0.76, + "learning_rate": 2.7988031698517926e-06, + "loss": 1.1045, + "step": 31737 + }, + { + "epoch": 0.76, + "learning_rate": 2.7982627559214504e-06, + "loss": 0.9821, + "step": 31738 + }, + { + "epoch": 0.76, + "learning_rate": 2.7977223856823567e-06, + "loss": 1.212, + "step": 31739 + }, + { + "epoch": 0.76, + "learning_rate": 2.7971820591377842e-06, + "loss": 1.0537, + "step": 31740 + }, + { + "epoch": 0.76, + "learning_rate": 2.7966417762910168e-06, + "loss": 1.0371, + "step": 31741 + }, + { + "epoch": 0.76, + "learning_rate": 2.796101537145327e-06, + "loss": 1.1421, + "step": 31742 + }, + { + "epoch": 0.76, + "learning_rate": 2.795561341703994e-06, + "loss": 0.924, + "step": 31743 + }, + { + "epoch": 0.76, + "learning_rate": 2.7950211899702996e-06, + "loss": 1.1624, + "step": 31744 + }, + { + "epoch": 0.76, + "learning_rate": 2.794481081947513e-06, + "loss": 1.1369, + "step": 31745 + }, + { + "epoch": 0.76, + "learning_rate": 2.793941017638915e-06, + "loss": 0.9273, + "step": 31746 + }, + { + "epoch": 0.76, + "learning_rate": 2.793400997047785e-06, + "loss": 0.9902, + "step": 31747 + }, + { + "epoch": 0.76, + "learning_rate": 2.7928610201773933e-06, + "loss": 1.2599, + "step": 31748 + }, + { + "epoch": 0.76, + "learning_rate": 2.792321087031021e-06, + "loss": 1.1965, + "step": 31749 + }, + { + "epoch": 0.76, + "learning_rate": 2.791781197611939e-06, + "loss": 1.2953, + "step": 31750 + }, + { + "epoch": 0.76, + "learning_rate": 2.791241351923425e-06, + "loss": 1.1743, + "step": 31751 + }, + { + "epoch": 0.76, + "learning_rate": 2.7907015499687573e-06, + "loss": 1.2636, + "step": 31752 + }, + { + "epoch": 0.76, + "learning_rate": 2.7901617917512034e-06, + "loss": 1.0689, + "step": 31753 + }, + { + "epoch": 0.76, + "learning_rate": 2.7896220772740434e-06, + "loss": 1.1235, + "step": 31754 + }, + { + "epoch": 0.76, + "learning_rate": 2.789082406540553e-06, + "loss": 1.0533, + "step": 31755 + }, + { + "epoch": 0.76, + "learning_rate": 2.7885427795539997e-06, + "loss": 1.0037, + "step": 31756 + }, + { + "epoch": 0.76, + "learning_rate": 2.7880031963176644e-06, + "loss": 1.0701, + "step": 31757 + }, + { + "epoch": 0.76, + "learning_rate": 2.7874636568348144e-06, + "loss": 0.9013, + "step": 31758 + }, + { + "epoch": 0.76, + "learning_rate": 2.786924161108727e-06, + "loss": 0.9489, + "step": 31759 + }, + { + "epoch": 0.76, + "learning_rate": 2.7863847091426755e-06, + "loss": 0.9659, + "step": 31760 + }, + { + "epoch": 0.76, + "learning_rate": 2.785845300939929e-06, + "loss": 1.2057, + "step": 31761 + }, + { + "epoch": 0.76, + "learning_rate": 2.785305936503763e-06, + "loss": 1.0521, + "step": 31762 + }, + { + "epoch": 0.76, + "learning_rate": 2.7847666158374497e-06, + "loss": 0.9776, + "step": 31763 + }, + { + "epoch": 0.76, + "learning_rate": 2.7842273389442586e-06, + "loss": 1.119, + "step": 31764 + }, + { + "epoch": 0.76, + "learning_rate": 2.7836881058274666e-06, + "loss": 0.7678, + "step": 31765 + }, + { + "epoch": 0.76, + "learning_rate": 2.783148916490338e-06, + "loss": 1.088, + "step": 31766 + }, + { + "epoch": 0.76, + "learning_rate": 2.7826097709361467e-06, + "loss": 1.2969, + "step": 31767 + }, + { + "epoch": 0.76, + "learning_rate": 2.7820706691681676e-06, + "loss": 1.1643, + "step": 31768 + }, + { + "epoch": 0.76, + "learning_rate": 2.781531611189665e-06, + "loss": 1.067, + "step": 31769 + }, + { + "epoch": 0.76, + "learning_rate": 2.780992597003913e-06, + "loss": 0.845, + "step": 31770 + }, + { + "epoch": 0.76, + "learning_rate": 2.7804536266141823e-06, + "loss": 0.8955, + "step": 31771 + }, + { + "epoch": 0.76, + "learning_rate": 2.7799147000237404e-06, + "loss": 1.0968, + "step": 31772 + }, + { + "epoch": 0.76, + "learning_rate": 2.7793758172358587e-06, + "loss": 0.6928, + "step": 31773 + }, + { + "epoch": 0.76, + "learning_rate": 2.7788369782538038e-06, + "loss": 1.2321, + "step": 31774 + }, + { + "epoch": 0.76, + "learning_rate": 2.7782981830808463e-06, + "loss": 0.9777, + "step": 31775 + }, + { + "epoch": 0.76, + "learning_rate": 2.7777594317202573e-06, + "loss": 1.0999, + "step": 31776 + }, + { + "epoch": 0.76, + "learning_rate": 2.777220724175301e-06, + "loss": 0.9864, + "step": 31777 + }, + { + "epoch": 0.76, + "learning_rate": 2.7766820604492474e-06, + "loss": 0.984, + "step": 31778 + }, + { + "epoch": 0.76, + "learning_rate": 2.776143440545367e-06, + "loss": 1.0636, + "step": 31779 + }, + { + "epoch": 0.76, + "learning_rate": 2.775604864466923e-06, + "loss": 0.9237, + "step": 31780 + }, + { + "epoch": 0.76, + "learning_rate": 2.7750663322171845e-06, + "loss": 1.1131, + "step": 31781 + }, + { + "epoch": 0.76, + "learning_rate": 2.7745278437994227e-06, + "loss": 1.1483, + "step": 31782 + }, + { + "epoch": 0.76, + "learning_rate": 2.7739893992168974e-06, + "loss": 1.018, + "step": 31783 + }, + { + "epoch": 0.76, + "learning_rate": 2.773450998472883e-06, + "loss": 1.0803, + "step": 31784 + }, + { + "epoch": 0.76, + "learning_rate": 2.7729126415706374e-06, + "loss": 1.2368, + "step": 31785 + }, + { + "epoch": 0.76, + "learning_rate": 2.7723743285134307e-06, + "loss": 1.2524, + "step": 31786 + }, + { + "epoch": 0.76, + "learning_rate": 2.7718360593045324e-06, + "loss": 0.8351, + "step": 31787 + }, + { + "epoch": 0.76, + "learning_rate": 2.771297833947202e-06, + "loss": 0.9961, + "step": 31788 + }, + { + "epoch": 0.76, + "learning_rate": 2.7707596524447067e-06, + "loss": 1.094, + "step": 31789 + }, + { + "epoch": 0.76, + "learning_rate": 2.7702215148003153e-06, + "loss": 0.9764, + "step": 31790 + }, + { + "epoch": 0.76, + "learning_rate": 2.769683421017286e-06, + "loss": 1.0803, + "step": 31791 + }, + { + "epoch": 0.76, + "learning_rate": 2.76914537109889e-06, + "loss": 0.8677, + "step": 31792 + }, + { + "epoch": 0.76, + "learning_rate": 2.768607365048386e-06, + "loss": 0.8664, + "step": 31793 + }, + { + "epoch": 0.76, + "learning_rate": 2.7680694028690393e-06, + "loss": 1.061, + "step": 31794 + }, + { + "epoch": 0.76, + "learning_rate": 2.7675314845641176e-06, + "loss": 0.969, + "step": 31795 + }, + { + "epoch": 0.76, + "learning_rate": 2.766993610136878e-06, + "loss": 1.2055, + "step": 31796 + }, + { + "epoch": 0.76, + "learning_rate": 2.7664557795905877e-06, + "loss": 1.0349, + "step": 31797 + }, + { + "epoch": 0.76, + "learning_rate": 2.765917992928511e-06, + "loss": 1.044, + "step": 31798 + }, + { + "epoch": 0.76, + "learning_rate": 2.765380250153905e-06, + "loss": 1.0204, + "step": 31799 + }, + { + "epoch": 0.76, + "learning_rate": 2.7648425512700393e-06, + "loss": 0.8618, + "step": 31800 + }, + { + "epoch": 0.76, + "learning_rate": 2.7643048962801688e-06, + "loss": 1.1555, + "step": 31801 + }, + { + "epoch": 0.76, + "learning_rate": 2.7637672851875584e-06, + "loss": 1.1121, + "step": 31802 + }, + { + "epoch": 0.76, + "learning_rate": 2.7632297179954727e-06, + "loss": 1.207, + "step": 31803 + }, + { + "epoch": 0.76, + "learning_rate": 2.762692194707167e-06, + "loss": 1.0219, + "step": 31804 + }, + { + "epoch": 0.76, + "learning_rate": 2.7621547153259054e-06, + "loss": 0.9754, + "step": 31805 + }, + { + "epoch": 0.76, + "learning_rate": 2.7616172798549523e-06, + "loss": 1.1163, + "step": 31806 + }, + { + "epoch": 0.76, + "learning_rate": 2.761079888297561e-06, + "loss": 1.1037, + "step": 31807 + }, + { + "epoch": 0.76, + "learning_rate": 2.7605425406569984e-06, + "loss": 0.8831, + "step": 31808 + }, + { + "epoch": 0.76, + "learning_rate": 2.7600052369365184e-06, + "loss": 0.9161, + "step": 31809 + }, + { + "epoch": 0.76, + "learning_rate": 2.759467977139384e-06, + "loss": 1.1626, + "step": 31810 + }, + { + "epoch": 0.77, + "learning_rate": 2.7589307612688564e-06, + "loss": 1.0662, + "step": 31811 + }, + { + "epoch": 0.77, + "learning_rate": 2.758393589328191e-06, + "loss": 0.9953, + "step": 31812 + }, + { + "epoch": 0.77, + "learning_rate": 2.757856461320647e-06, + "loss": 1.093, + "step": 31813 + }, + { + "epoch": 0.77, + "learning_rate": 2.757319377249489e-06, + "loss": 1.157, + "step": 31814 + }, + { + "epoch": 0.77, + "learning_rate": 2.756782337117967e-06, + "loss": 1.0895, + "step": 31815 + }, + { + "epoch": 0.77, + "learning_rate": 2.7562453409293456e-06, + "loss": 0.9422, + "step": 31816 + }, + { + "epoch": 0.77, + "learning_rate": 2.7557083886868765e-06, + "loss": 1.0281, + "step": 31817 + }, + { + "epoch": 0.77, + "learning_rate": 2.7551714803938214e-06, + "loss": 0.983, + "step": 31818 + }, + { + "epoch": 0.77, + "learning_rate": 2.75463461605344e-06, + "loss": 1.192, + "step": 31819 + }, + { + "epoch": 0.77, + "learning_rate": 2.7540977956689817e-06, + "loss": 1.1053, + "step": 31820 + }, + { + "epoch": 0.77, + "learning_rate": 2.753561019243709e-06, + "loss": 0.8266, + "step": 31821 + }, + { + "epoch": 0.77, + "learning_rate": 2.75302428678088e-06, + "loss": 0.8611, + "step": 31822 + }, + { + "epoch": 0.77, + "learning_rate": 2.752487598283745e-06, + "loss": 0.8789, + "step": 31823 + }, + { + "epoch": 0.77, + "learning_rate": 2.7519509537555665e-06, + "loss": 1.0115, + "step": 31824 + }, + { + "epoch": 0.77, + "learning_rate": 2.751414353199593e-06, + "loss": 0.9538, + "step": 31825 + }, + { + "epoch": 0.77, + "learning_rate": 2.750877796619084e-06, + "loss": 1.1534, + "step": 31826 + }, + { + "epoch": 0.77, + "learning_rate": 2.7503412840172973e-06, + "loss": 1.0512, + "step": 31827 + }, + { + "epoch": 0.77, + "learning_rate": 2.749804815397481e-06, + "loss": 1.1657, + "step": 31828 + }, + { + "epoch": 0.77, + "learning_rate": 2.7492683907628946e-06, + "loss": 0.957, + "step": 31829 + }, + { + "epoch": 0.77, + "learning_rate": 2.7487320101167935e-06, + "loss": 0.9813, + "step": 31830 + }, + { + "epoch": 0.77, + "learning_rate": 2.748195673462427e-06, + "loss": 1.1655, + "step": 31831 + }, + { + "epoch": 0.77, + "learning_rate": 2.747659380803053e-06, + "loss": 1.0612, + "step": 31832 + }, + { + "epoch": 0.77, + "learning_rate": 2.7471231321419225e-06, + "loss": 1.1541, + "step": 31833 + }, + { + "epoch": 0.77, + "learning_rate": 2.746586927482289e-06, + "loss": 0.8703, + "step": 31834 + }, + { + "epoch": 0.77, + "learning_rate": 2.7460507668274108e-06, + "loss": 1.0463, + "step": 31835 + }, + { + "epoch": 0.77, + "learning_rate": 2.745514650180533e-06, + "loss": 1.1866, + "step": 31836 + }, + { + "epoch": 0.77, + "learning_rate": 2.7449785775449146e-06, + "loss": 0.9514, + "step": 31837 + }, + { + "epoch": 0.77, + "learning_rate": 2.7444425489238014e-06, + "loss": 0.9895, + "step": 31838 + }, + { + "epoch": 0.77, + "learning_rate": 2.743906564320449e-06, + "loss": 1.0223, + "step": 31839 + }, + { + "epoch": 0.77, + "learning_rate": 2.743370623738112e-06, + "loss": 1.1618, + "step": 31840 + }, + { + "epoch": 0.77, + "learning_rate": 2.742834727180035e-06, + "loss": 1.0604, + "step": 31841 + }, + { + "epoch": 0.77, + "learning_rate": 2.7422988746494726e-06, + "loss": 1.1478, + "step": 31842 + }, + { + "epoch": 0.77, + "learning_rate": 2.7417630661496786e-06, + "loss": 1.1494, + "step": 31843 + }, + { + "epoch": 0.77, + "learning_rate": 2.7412273016838986e-06, + "loss": 0.762, + "step": 31844 + }, + { + "epoch": 0.77, + "learning_rate": 2.740691581255388e-06, + "loss": 1.1501, + "step": 31845 + }, + { + "epoch": 0.77, + "learning_rate": 2.7401559048673907e-06, + "loss": 0.943, + "step": 31846 + }, + { + "epoch": 0.77, + "learning_rate": 2.73962027252316e-06, + "loss": 0.9129, + "step": 31847 + }, + { + "epoch": 0.77, + "learning_rate": 2.7390846842259487e-06, + "loss": 0.9868, + "step": 31848 + }, + { + "epoch": 0.77, + "learning_rate": 2.738549139979e-06, + "loss": 0.9859, + "step": 31849 + }, + { + "epoch": 0.77, + "learning_rate": 2.738013639785565e-06, + "loss": 0.9401, + "step": 31850 + }, + { + "epoch": 0.77, + "learning_rate": 2.737478183648896e-06, + "loss": 0.7958, + "step": 31851 + }, + { + "epoch": 0.77, + "learning_rate": 2.7369427715722362e-06, + "loss": 1.1318, + "step": 31852 + }, + { + "epoch": 0.77, + "learning_rate": 2.7364074035588393e-06, + "loss": 1.0555, + "step": 31853 + }, + { + "epoch": 0.77, + "learning_rate": 2.7358720796119465e-06, + "loss": 0.8444, + "step": 31854 + }, + { + "epoch": 0.77, + "learning_rate": 2.73533679973481e-06, + "loss": 1.2781, + "step": 31855 + }, + { + "epoch": 0.77, + "learning_rate": 2.734801563930679e-06, + "loss": 1.3186, + "step": 31856 + }, + { + "epoch": 0.77, + "learning_rate": 2.734266372202795e-06, + "loss": 1.2837, + "step": 31857 + }, + { + "epoch": 0.77, + "learning_rate": 2.733731224554408e-06, + "loss": 1.1504, + "step": 31858 + }, + { + "epoch": 0.77, + "learning_rate": 2.733196120988767e-06, + "loss": 0.889, + "step": 31859 + }, + { + "epoch": 0.77, + "learning_rate": 2.732661061509113e-06, + "loss": 1.2497, + "step": 31860 + }, + { + "epoch": 0.77, + "learning_rate": 2.732126046118696e-06, + "loss": 1.1826, + "step": 31861 + }, + { + "epoch": 0.77, + "learning_rate": 2.731591074820763e-06, + "loss": 1.1418, + "step": 31862 + }, + { + "epoch": 0.77, + "learning_rate": 2.7310561476185537e-06, + "loss": 0.8585, + "step": 31863 + }, + { + "epoch": 0.77, + "learning_rate": 2.7305212645153213e-06, + "loss": 1.0487, + "step": 31864 + }, + { + "epoch": 0.77, + "learning_rate": 2.7299864255143026e-06, + "loss": 0.6609, + "step": 31865 + }, + { + "epoch": 0.77, + "learning_rate": 2.7294516306187457e-06, + "loss": 1.1658, + "step": 31866 + }, + { + "epoch": 0.77, + "learning_rate": 2.7289168798319e-06, + "loss": 1.231, + "step": 31867 + }, + { + "epoch": 0.77, + "learning_rate": 2.7283821731570014e-06, + "loss": 0.8385, + "step": 31868 + }, + { + "epoch": 0.77, + "learning_rate": 2.7278475105972977e-06, + "loss": 0.8677, + "step": 31869 + }, + { + "epoch": 0.77, + "learning_rate": 2.7273128921560365e-06, + "loss": 1.0964, + "step": 31870 + }, + { + "epoch": 0.77, + "learning_rate": 2.726778317836454e-06, + "loss": 1.1382, + "step": 31871 + }, + { + "epoch": 0.77, + "learning_rate": 2.7262437876418e-06, + "loss": 1.0251, + "step": 31872 + }, + { + "epoch": 0.77, + "learning_rate": 2.72570930157531e-06, + "loss": 1.3249, + "step": 31873 + }, + { + "epoch": 0.77, + "learning_rate": 2.7251748596402317e-06, + "loss": 1.011, + "step": 31874 + }, + { + "epoch": 0.77, + "learning_rate": 2.7246404618398093e-06, + "loss": 1.0147, + "step": 31875 + }, + { + "epoch": 0.77, + "learning_rate": 2.724106108177278e-06, + "loss": 1.0849, + "step": 31876 + }, + { + "epoch": 0.77, + "learning_rate": 2.723571798655884e-06, + "loss": 1.0588, + "step": 31877 + }, + { + "epoch": 0.77, + "learning_rate": 2.7230375332788717e-06, + "loss": 0.8933, + "step": 31878 + }, + { + "epoch": 0.77, + "learning_rate": 2.722503312049476e-06, + "loss": 1.0054, + "step": 31879 + }, + { + "epoch": 0.77, + "learning_rate": 2.7219691349709433e-06, + "loss": 1.1472, + "step": 31880 + }, + { + "epoch": 0.77, + "learning_rate": 2.7214350020465098e-06, + "loss": 1.2629, + "step": 31881 + }, + { + "epoch": 0.77, + "learning_rate": 2.7209009132794174e-06, + "loss": 1.3139, + "step": 31882 + }, + { + "epoch": 0.77, + "learning_rate": 2.7203668686729113e-06, + "loss": 0.9744, + "step": 31883 + }, + { + "epoch": 0.77, + "learning_rate": 2.7198328682302223e-06, + "loss": 0.9245, + "step": 31884 + }, + { + "epoch": 0.77, + "learning_rate": 2.7192989119545963e-06, + "loss": 1.2136, + "step": 31885 + }, + { + "epoch": 0.77, + "learning_rate": 2.7187649998492737e-06, + "loss": 0.9316, + "step": 31886 + }, + { + "epoch": 0.77, + "learning_rate": 2.718231131917488e-06, + "loss": 1.0434, + "step": 31887 + }, + { + "epoch": 0.77, + "learning_rate": 2.717697308162485e-06, + "loss": 0.8704, + "step": 31888 + }, + { + "epoch": 0.77, + "learning_rate": 2.717163528587495e-06, + "loss": 1.2629, + "step": 31889 + }, + { + "epoch": 0.77, + "learning_rate": 2.7166297931957632e-06, + "loss": 0.9162, + "step": 31890 + }, + { + "epoch": 0.77, + "learning_rate": 2.716096101990526e-06, + "loss": 0.8795, + "step": 31891 + }, + { + "epoch": 0.77, + "learning_rate": 2.715562454975019e-06, + "loss": 0.8157, + "step": 31892 + }, + { + "epoch": 0.77, + "learning_rate": 2.7150288521524802e-06, + "loss": 1.2029, + "step": 31893 + }, + { + "epoch": 0.77, + "learning_rate": 2.7144952935261517e-06, + "loss": 1.0216, + "step": 31894 + }, + { + "epoch": 0.77, + "learning_rate": 2.7139617790992635e-06, + "loss": 1.2185, + "step": 31895 + }, + { + "epoch": 0.77, + "learning_rate": 2.713428308875058e-06, + "loss": 1.0909, + "step": 31896 + }, + { + "epoch": 0.77, + "learning_rate": 2.712894882856766e-06, + "loss": 1.0857, + "step": 31897 + }, + { + "epoch": 0.77, + "learning_rate": 2.7123615010476266e-06, + "loss": 0.9142, + "step": 31898 + }, + { + "epoch": 0.77, + "learning_rate": 2.7118281634508792e-06, + "loss": 1.0623, + "step": 31899 + }, + { + "epoch": 0.77, + "learning_rate": 2.7112948700697528e-06, + "loss": 1.1558, + "step": 31900 + }, + { + "epoch": 0.77, + "learning_rate": 2.710761620907486e-06, + "loss": 0.9184, + "step": 31901 + }, + { + "epoch": 0.77, + "learning_rate": 2.7102284159673165e-06, + "loss": 1.1908, + "step": 31902 + }, + { + "epoch": 0.77, + "learning_rate": 2.7096952552524746e-06, + "loss": 0.9541, + "step": 31903 + }, + { + "epoch": 0.77, + "learning_rate": 2.7091621387661992e-06, + "loss": 0.924, + "step": 31904 + }, + { + "epoch": 0.77, + "learning_rate": 2.708629066511719e-06, + "loss": 0.9208, + "step": 31905 + }, + { + "epoch": 0.77, + "learning_rate": 2.7080960384922716e-06, + "loss": 1.189, + "step": 31906 + }, + { + "epoch": 0.77, + "learning_rate": 2.707563054711093e-06, + "loss": 0.968, + "step": 31907 + }, + { + "epoch": 0.77, + "learning_rate": 2.7070301151714117e-06, + "loss": 1.0161, + "step": 31908 + }, + { + "epoch": 0.77, + "learning_rate": 2.7064972198764627e-06, + "loss": 0.6937, + "step": 31909 + }, + { + "epoch": 0.77, + "learning_rate": 2.7059643688294824e-06, + "loss": 1.1446, + "step": 31910 + }, + { + "epoch": 0.77, + "learning_rate": 2.705431562033698e-06, + "loss": 0.9838, + "step": 31911 + }, + { + "epoch": 0.77, + "learning_rate": 2.704898799492345e-06, + "loss": 1.1256, + "step": 31912 + }, + { + "epoch": 0.77, + "learning_rate": 2.704366081208657e-06, + "loss": 1.0229, + "step": 31913 + }, + { + "epoch": 0.77, + "learning_rate": 2.703833407185862e-06, + "loss": 1.0097, + "step": 31914 + }, + { + "epoch": 0.77, + "learning_rate": 2.7033007774271958e-06, + "loss": 1.1839, + "step": 31915 + }, + { + "epoch": 0.77, + "learning_rate": 2.702768191935885e-06, + "loss": 1.0174, + "step": 31916 + }, + { + "epoch": 0.77, + "learning_rate": 2.7022356507151627e-06, + "loss": 1.0561, + "step": 31917 + }, + { + "epoch": 0.77, + "learning_rate": 2.7017031537682638e-06, + "loss": 0.6045, + "step": 31918 + }, + { + "epoch": 0.77, + "learning_rate": 2.7011707010984112e-06, + "loss": 1.232, + "step": 31919 + }, + { + "epoch": 0.77, + "learning_rate": 2.700638292708839e-06, + "loss": 1.2928, + "step": 31920 + }, + { + "epoch": 0.77, + "learning_rate": 2.7001059286027818e-06, + "loss": 1.1241, + "step": 31921 + }, + { + "epoch": 0.77, + "learning_rate": 2.6995736087834614e-06, + "loss": 0.9256, + "step": 31922 + }, + { + "epoch": 0.77, + "learning_rate": 2.699041333254113e-06, + "loss": 1.0751, + "step": 31923 + }, + { + "epoch": 0.77, + "learning_rate": 2.698509102017961e-06, + "loss": 1.1352, + "step": 31924 + }, + { + "epoch": 0.77, + "learning_rate": 2.6979769150782375e-06, + "loss": 1.0689, + "step": 31925 + }, + { + "epoch": 0.77, + "learning_rate": 2.6974447724381727e-06, + "loss": 1.1084, + "step": 31926 + }, + { + "epoch": 0.77, + "learning_rate": 2.69691267410099e-06, + "loss": 0.986, + "step": 31927 + }, + { + "epoch": 0.77, + "learning_rate": 2.6963806200699206e-06, + "loss": 1.0562, + "step": 31928 + }, + { + "epoch": 0.77, + "learning_rate": 2.695848610348195e-06, + "loss": 1.1863, + "step": 31929 + }, + { + "epoch": 0.77, + "learning_rate": 2.6953166449390344e-06, + "loss": 0.941, + "step": 31930 + }, + { + "epoch": 0.77, + "learning_rate": 2.6947847238456724e-06, + "loss": 1.0752, + "step": 31931 + }, + { + "epoch": 0.77, + "learning_rate": 2.6942528470713315e-06, + "loss": 1.2547, + "step": 31932 + }, + { + "epoch": 0.77, + "learning_rate": 2.693721014619238e-06, + "loss": 0.9945, + "step": 31933 + }, + { + "epoch": 0.77, + "learning_rate": 2.693189226492625e-06, + "loss": 0.9835, + "step": 31934 + }, + { + "epoch": 0.77, + "learning_rate": 2.6926574826947116e-06, + "loss": 1.1143, + "step": 31935 + }, + { + "epoch": 0.77, + "learning_rate": 2.692125783228726e-06, + "loss": 0.9751, + "step": 31936 + }, + { + "epoch": 0.77, + "learning_rate": 2.6915941280978976e-06, + "loss": 0.8367, + "step": 31937 + }, + { + "epoch": 0.77, + "learning_rate": 2.691062517305445e-06, + "loss": 1.2503, + "step": 31938 + }, + { + "epoch": 0.77, + "learning_rate": 2.6905309508546005e-06, + "loss": 1.1536, + "step": 31939 + }, + { + "epoch": 0.77, + "learning_rate": 2.689999428748582e-06, + "loss": 1.2726, + "step": 31940 + }, + { + "epoch": 0.77, + "learning_rate": 2.6894679509906176e-06, + "loss": 0.9861, + "step": 31941 + }, + { + "epoch": 0.77, + "learning_rate": 2.6889365175839344e-06, + "loss": 0.9911, + "step": 31942 + }, + { + "epoch": 0.77, + "learning_rate": 2.6884051285317515e-06, + "loss": 0.9301, + "step": 31943 + }, + { + "epoch": 0.77, + "learning_rate": 2.6878737838372946e-06, + "loss": 0.9197, + "step": 31944 + }, + { + "epoch": 0.77, + "learning_rate": 2.68734248350379e-06, + "loss": 0.829, + "step": 31945 + }, + { + "epoch": 0.77, + "learning_rate": 2.686811227534456e-06, + "loss": 0.9519, + "step": 31946 + }, + { + "epoch": 0.77, + "learning_rate": 2.686280015932521e-06, + "loss": 0.9538, + "step": 31947 + }, + { + "epoch": 0.77, + "learning_rate": 2.685748848701202e-06, + "loss": 1.0374, + "step": 31948 + }, + { + "epoch": 0.77, + "learning_rate": 2.685217725843724e-06, + "loss": 1.0309, + "step": 31949 + }, + { + "epoch": 0.77, + "learning_rate": 2.6846866473633126e-06, + "loss": 1.2363, + "step": 31950 + }, + { + "epoch": 0.77, + "learning_rate": 2.684155613263184e-06, + "loss": 1.2052, + "step": 31951 + }, + { + "epoch": 0.77, + "learning_rate": 2.6836246235465625e-06, + "loss": 0.9867, + "step": 31952 + }, + { + "epoch": 0.77, + "learning_rate": 2.683093678216672e-06, + "loss": 1.0797, + "step": 31953 + }, + { + "epoch": 0.77, + "learning_rate": 2.682562777276728e-06, + "loss": 1.0101, + "step": 31954 + }, + { + "epoch": 0.77, + "learning_rate": 2.682031920729955e-06, + "loss": 1.0758, + "step": 31955 + }, + { + "epoch": 0.77, + "learning_rate": 2.681501108579575e-06, + "loss": 1.0662, + "step": 31956 + }, + { + "epoch": 0.77, + "learning_rate": 2.6809703408288046e-06, + "loss": 1.0296, + "step": 31957 + }, + { + "epoch": 0.77, + "learning_rate": 2.6804396174808676e-06, + "loss": 1.0843, + "step": 31958 + }, + { + "epoch": 0.77, + "learning_rate": 2.6799089385389797e-06, + "loss": 0.9228, + "step": 31959 + }, + { + "epoch": 0.77, + "learning_rate": 2.6793783040063627e-06, + "loss": 1.2228, + "step": 31960 + }, + { + "epoch": 0.77, + "learning_rate": 2.6788477138862377e-06, + "loss": 1.1272, + "step": 31961 + }, + { + "epoch": 0.77, + "learning_rate": 2.678317168181819e-06, + "loss": 0.8862, + "step": 31962 + }, + { + "epoch": 0.77, + "learning_rate": 2.6777866668963283e-06, + "loss": 0.8765, + "step": 31963 + }, + { + "epoch": 0.77, + "learning_rate": 2.677256210032986e-06, + "loss": 0.9383, + "step": 31964 + }, + { + "epoch": 0.77, + "learning_rate": 2.6767257975950047e-06, + "loss": 0.8485, + "step": 31965 + }, + { + "epoch": 0.77, + "learning_rate": 2.6761954295856087e-06, + "loss": 1.0009, + "step": 31966 + }, + { + "epoch": 0.77, + "learning_rate": 2.6756651060080094e-06, + "loss": 1.1184, + "step": 31967 + }, + { + "epoch": 0.77, + "learning_rate": 2.6751348268654266e-06, + "loss": 1.0379, + "step": 31968 + }, + { + "epoch": 0.77, + "learning_rate": 2.674604592161082e-06, + "loss": 0.7818, + "step": 31969 + }, + { + "epoch": 0.77, + "learning_rate": 2.6740744018981846e-06, + "loss": 0.9919, + "step": 31970 + }, + { + "epoch": 0.77, + "learning_rate": 2.673544256079954e-06, + "loss": 0.8657, + "step": 31971 + }, + { + "epoch": 0.77, + "learning_rate": 2.673014154709611e-06, + "loss": 1.0051, + "step": 31972 + }, + { + "epoch": 0.77, + "learning_rate": 2.6724840977903644e-06, + "loss": 0.9657, + "step": 31973 + }, + { + "epoch": 0.77, + "learning_rate": 2.6719540853254366e-06, + "loss": 1.353, + "step": 31974 + }, + { + "epoch": 0.77, + "learning_rate": 2.671424117318037e-06, + "loss": 0.7627, + "step": 31975 + }, + { + "epoch": 0.77, + "learning_rate": 2.670894193771383e-06, + "loss": 1.1418, + "step": 31976 + }, + { + "epoch": 0.77, + "learning_rate": 2.6703643146886927e-06, + "loss": 1.1249, + "step": 31977 + }, + { + "epoch": 0.77, + "learning_rate": 2.6698344800731756e-06, + "loss": 1.2635, + "step": 31978 + }, + { + "epoch": 0.77, + "learning_rate": 2.6693046899280482e-06, + "loss": 1.2121, + "step": 31979 + }, + { + "epoch": 0.77, + "learning_rate": 2.668774944256528e-06, + "loss": 1.0699, + "step": 31980 + }, + { + "epoch": 0.77, + "learning_rate": 2.668245243061822e-06, + "loss": 0.9387, + "step": 31981 + }, + { + "epoch": 0.77, + "learning_rate": 2.6677155863471525e-06, + "loss": 0.8259, + "step": 31982 + }, + { + "epoch": 0.77, + "learning_rate": 2.6671859741157236e-06, + "loss": 0.9547, + "step": 31983 + }, + { + "epoch": 0.77, + "learning_rate": 2.666656406370752e-06, + "loss": 1.2269, + "step": 31984 + }, + { + "epoch": 0.77, + "learning_rate": 2.6661268831154552e-06, + "loss": 1.0036, + "step": 31985 + }, + { + "epoch": 0.77, + "learning_rate": 2.6655974043530385e-06, + "loss": 1.009, + "step": 31986 + }, + { + "epoch": 0.77, + "learning_rate": 2.6650679700867177e-06, + "loss": 1.2235, + "step": 31987 + }, + { + "epoch": 0.77, + "learning_rate": 2.6645385803197065e-06, + "loss": 1.0604, + "step": 31988 + }, + { + "epoch": 0.77, + "learning_rate": 2.6640092350552104e-06, + "loss": 1.1857, + "step": 31989 + }, + { + "epoch": 0.77, + "learning_rate": 2.663479934296449e-06, + "loss": 1.2262, + "step": 31990 + }, + { + "epoch": 0.77, + "learning_rate": 2.6629506780466264e-06, + "loss": 0.8778, + "step": 31991 + }, + { + "epoch": 0.77, + "learning_rate": 2.6624214663089555e-06, + "loss": 1.1454, + "step": 31992 + }, + { + "epoch": 0.77, + "learning_rate": 2.66189229908665e-06, + "loss": 0.9379, + "step": 31993 + }, + { + "epoch": 0.77, + "learning_rate": 2.6613631763829164e-06, + "loss": 1.0092, + "step": 31994 + }, + { + "epoch": 0.77, + "learning_rate": 2.660834098200966e-06, + "loss": 1.0824, + "step": 31995 + }, + { + "epoch": 0.77, + "learning_rate": 2.660305064544011e-06, + "loss": 1.1052, + "step": 31996 + }, + { + "epoch": 0.77, + "learning_rate": 2.6597760754152567e-06, + "loss": 0.8671, + "step": 31997 + }, + { + "epoch": 0.77, + "learning_rate": 2.6592471308179136e-06, + "loss": 1.0999, + "step": 31998 + }, + { + "epoch": 0.77, + "learning_rate": 2.6587182307551916e-06, + "loss": 1.0041, + "step": 31999 + }, + { + "epoch": 0.77, + "learning_rate": 2.6581893752303e-06, + "loss": 1.1021, + "step": 32000 + }, + { + "epoch": 0.77, + "learning_rate": 2.6576605642464494e-06, + "loss": 1.073, + "step": 32001 + }, + { + "epoch": 0.77, + "learning_rate": 2.6571317978068425e-06, + "loss": 1.0298, + "step": 32002 + }, + { + "epoch": 0.77, + "learning_rate": 2.656603075914691e-06, + "loss": 1.0573, + "step": 32003 + }, + { + "epoch": 0.77, + "learning_rate": 2.6560743985731994e-06, + "loss": 1.0907, + "step": 32004 + }, + { + "epoch": 0.77, + "learning_rate": 2.6555457657855774e-06, + "loss": 1.2364, + "step": 32005 + }, + { + "epoch": 0.77, + "learning_rate": 2.655017177555035e-06, + "loss": 1.0138, + "step": 32006 + }, + { + "epoch": 0.77, + "learning_rate": 2.654488633884772e-06, + "loss": 1.0601, + "step": 32007 + }, + { + "epoch": 0.77, + "learning_rate": 2.6539601347779997e-06, + "loss": 0.9454, + "step": 32008 + }, + { + "epoch": 0.77, + "learning_rate": 2.6534316802379256e-06, + "loss": 0.8257, + "step": 32009 + }, + { + "epoch": 0.77, + "learning_rate": 2.6529032702677503e-06, + "loss": 0.9804, + "step": 32010 + }, + { + "epoch": 0.77, + "learning_rate": 2.6523749048706857e-06, + "loss": 0.8459, + "step": 32011 + }, + { + "epoch": 0.77, + "learning_rate": 2.651846584049932e-06, + "loss": 1.0528, + "step": 32012 + }, + { + "epoch": 0.77, + "learning_rate": 2.6513183078086958e-06, + "loss": 1.0094, + "step": 32013 + }, + { + "epoch": 0.77, + "learning_rate": 2.650790076150186e-06, + "loss": 0.9478, + "step": 32014 + }, + { + "epoch": 0.77, + "learning_rate": 2.650261889077602e-06, + "loss": 0.8339, + "step": 32015 + }, + { + "epoch": 0.77, + "learning_rate": 2.6497337465941496e-06, + "loss": 0.9697, + "step": 32016 + }, + { + "epoch": 0.77, + "learning_rate": 2.649205648703037e-06, + "loss": 1.0227, + "step": 32017 + }, + { + "epoch": 0.77, + "learning_rate": 2.6486775954074617e-06, + "loss": 1.1226, + "step": 32018 + }, + { + "epoch": 0.77, + "learning_rate": 2.6481495867106333e-06, + "loss": 0.8803, + "step": 32019 + }, + { + "epoch": 0.77, + "learning_rate": 2.647621622615749e-06, + "loss": 1.0836, + "step": 32020 + }, + { + "epoch": 0.77, + "learning_rate": 2.6470937031260145e-06, + "loss": 1.1123, + "step": 32021 + }, + { + "epoch": 0.77, + "learning_rate": 2.646565828244637e-06, + "loss": 0.99, + "step": 32022 + }, + { + "epoch": 0.77, + "learning_rate": 2.6460379979748107e-06, + "loss": 0.9986, + "step": 32023 + }, + { + "epoch": 0.77, + "learning_rate": 2.6455102123197428e-06, + "loss": 1.0985, + "step": 32024 + }, + { + "epoch": 0.77, + "learning_rate": 2.6449824712826377e-06, + "loss": 1.0505, + "step": 32025 + }, + { + "epoch": 0.77, + "learning_rate": 2.6444547748666916e-06, + "loss": 0.8772, + "step": 32026 + }, + { + "epoch": 0.77, + "learning_rate": 2.6439271230751107e-06, + "loss": 0.9178, + "step": 32027 + }, + { + "epoch": 0.77, + "learning_rate": 2.6433995159110902e-06, + "loss": 1.3014, + "step": 32028 + }, + { + "epoch": 0.77, + "learning_rate": 2.6428719533778347e-06, + "loss": 1.27, + "step": 32029 + }, + { + "epoch": 0.77, + "learning_rate": 2.642344435478548e-06, + "loss": 1.0168, + "step": 32030 + }, + { + "epoch": 0.77, + "learning_rate": 2.6418169622164226e-06, + "loss": 1.1481, + "step": 32031 + }, + { + "epoch": 0.77, + "learning_rate": 2.641289533594664e-06, + "loss": 1.1644, + "step": 32032 + }, + { + "epoch": 0.77, + "learning_rate": 2.6407621496164735e-06, + "loss": 0.934, + "step": 32033 + }, + { + "epoch": 0.77, + "learning_rate": 2.640234810285046e-06, + "loss": 0.9783, + "step": 32034 + }, + { + "epoch": 0.77, + "learning_rate": 2.6397075156035823e-06, + "loss": 1.1623, + "step": 32035 + }, + { + "epoch": 0.77, + "learning_rate": 2.6391802655752853e-06, + "loss": 1.1466, + "step": 32036 + }, + { + "epoch": 0.77, + "learning_rate": 2.6386530602033457e-06, + "loss": 0.9601, + "step": 32037 + }, + { + "epoch": 0.77, + "learning_rate": 2.638125899490971e-06, + "loss": 0.7966, + "step": 32038 + }, + { + "epoch": 0.77, + "learning_rate": 2.6375987834413506e-06, + "loss": 1.282, + "step": 32039 + }, + { + "epoch": 0.77, + "learning_rate": 2.6370717120576885e-06, + "loss": 1.1716, + "step": 32040 + }, + { + "epoch": 0.77, + "learning_rate": 2.636544685343182e-06, + "loss": 1.0927, + "step": 32041 + }, + { + "epoch": 0.77, + "learning_rate": 2.636017703301024e-06, + "loss": 1.0674, + "step": 32042 + }, + { + "epoch": 0.77, + "learning_rate": 2.6354907659344144e-06, + "loss": 1.079, + "step": 32043 + }, + { + "epoch": 0.77, + "learning_rate": 2.6349638732465534e-06, + "loss": 0.9965, + "step": 32044 + }, + { + "epoch": 0.77, + "learning_rate": 2.6344370252406316e-06, + "loss": 0.8766, + "step": 32045 + }, + { + "epoch": 0.77, + "learning_rate": 2.63391022191985e-06, + "loss": 0.7055, + "step": 32046 + }, + { + "epoch": 0.77, + "learning_rate": 2.6333834632874e-06, + "loss": 0.987, + "step": 32047 + }, + { + "epoch": 0.77, + "learning_rate": 2.6328567493464796e-06, + "loss": 0.9879, + "step": 32048 + }, + { + "epoch": 0.77, + "learning_rate": 2.6323300801002883e-06, + "loss": 1.0975, + "step": 32049 + }, + { + "epoch": 0.77, + "learning_rate": 2.6318034555520134e-06, + "loss": 0.7847, + "step": 32050 + }, + { + "epoch": 0.77, + "learning_rate": 2.6312768757048544e-06, + "loss": 1.0701, + "step": 32051 + }, + { + "epoch": 0.77, + "learning_rate": 2.630750340562008e-06, + "loss": 0.9433, + "step": 32052 + }, + { + "epoch": 0.77, + "learning_rate": 2.630223850126663e-06, + "loss": 0.7775, + "step": 32053 + }, + { + "epoch": 0.77, + "learning_rate": 2.6296974044020195e-06, + "loss": 1.1986, + "step": 32054 + }, + { + "epoch": 0.77, + "learning_rate": 2.6291710033912654e-06, + "loss": 0.9149, + "step": 32055 + }, + { + "epoch": 0.77, + "learning_rate": 2.628644647097597e-06, + "loss": 0.9808, + "step": 32056 + }, + { + "epoch": 0.77, + "learning_rate": 2.62811833552421e-06, + "loss": 0.9644, + "step": 32057 + }, + { + "epoch": 0.77, + "learning_rate": 2.627592068674293e-06, + "loss": 1.0867, + "step": 32058 + }, + { + "epoch": 0.77, + "learning_rate": 2.62706584655104e-06, + "loss": 0.9385, + "step": 32059 + }, + { + "epoch": 0.77, + "learning_rate": 2.6265396691576485e-06, + "loss": 1.1851, + "step": 32060 + }, + { + "epoch": 0.77, + "learning_rate": 2.6260135364973027e-06, + "loss": 1.1237, + "step": 32061 + }, + { + "epoch": 0.77, + "learning_rate": 2.6254874485732006e-06, + "loss": 1.15, + "step": 32062 + }, + { + "epoch": 0.77, + "learning_rate": 2.6249614053885297e-06, + "loss": 0.9396, + "step": 32063 + }, + { + "epoch": 0.77, + "learning_rate": 2.624435406946483e-06, + "loss": 0.8814, + "step": 32064 + }, + { + "epoch": 0.77, + "learning_rate": 2.6239094532502553e-06, + "loss": 1.0996, + "step": 32065 + }, + { + "epoch": 0.77, + "learning_rate": 2.6233835443030296e-06, + "loss": 1.0541, + "step": 32066 + }, + { + "epoch": 0.77, + "learning_rate": 2.622857680108002e-06, + "loss": 1.2031, + "step": 32067 + }, + { + "epoch": 0.77, + "learning_rate": 2.6223318606683647e-06, + "loss": 1.0006, + "step": 32068 + }, + { + "epoch": 0.77, + "learning_rate": 2.6218060859873007e-06, + "loss": 1.0397, + "step": 32069 + }, + { + "epoch": 0.77, + "learning_rate": 2.621280356068008e-06, + "loss": 0.8771, + "step": 32070 + }, + { + "epoch": 0.77, + "learning_rate": 2.620754670913668e-06, + "loss": 1.1227, + "step": 32071 + }, + { + "epoch": 0.77, + "learning_rate": 2.6202290305274737e-06, + "loss": 1.0967, + "step": 32072 + }, + { + "epoch": 0.77, + "learning_rate": 2.619703434912617e-06, + "loss": 1.1182, + "step": 32073 + }, + { + "epoch": 0.77, + "learning_rate": 2.6191778840722804e-06, + "loss": 1.1603, + "step": 32074 + }, + { + "epoch": 0.77, + "learning_rate": 2.618652378009656e-06, + "loss": 1.1119, + "step": 32075 + }, + { + "epoch": 0.77, + "learning_rate": 2.6181269167279344e-06, + "loss": 1.2324, + "step": 32076 + }, + { + "epoch": 0.77, + "learning_rate": 2.617601500230297e-06, + "loss": 0.9492, + "step": 32077 + }, + { + "epoch": 0.77, + "learning_rate": 2.617076128519935e-06, + "loss": 1.0155, + "step": 32078 + }, + { + "epoch": 0.77, + "learning_rate": 2.6165508016000384e-06, + "loss": 1.1774, + "step": 32079 + }, + { + "epoch": 0.77, + "learning_rate": 2.6160255194737895e-06, + "loss": 1.0946, + "step": 32080 + }, + { + "epoch": 0.77, + "learning_rate": 2.6155002821443786e-06, + "loss": 1.1557, + "step": 32081 + }, + { + "epoch": 0.77, + "learning_rate": 2.6149750896149884e-06, + "loss": 0.9501, + "step": 32082 + }, + { + "epoch": 0.77, + "learning_rate": 2.614449941888807e-06, + "loss": 0.8072, + "step": 32083 + }, + { + "epoch": 0.77, + "learning_rate": 2.613924838969024e-06, + "loss": 1.0369, + "step": 32084 + }, + { + "epoch": 0.77, + "learning_rate": 2.6133997808588184e-06, + "loss": 1.0612, + "step": 32085 + }, + { + "epoch": 0.77, + "learning_rate": 2.6128747675613788e-06, + "loss": 0.9528, + "step": 32086 + }, + { + "epoch": 0.77, + "learning_rate": 2.612349799079893e-06, + "loss": 0.8244, + "step": 32087 + }, + { + "epoch": 0.77, + "learning_rate": 2.6118248754175413e-06, + "loss": 0.9748, + "step": 32088 + }, + { + "epoch": 0.77, + "learning_rate": 2.611299996577512e-06, + "loss": 1.0506, + "step": 32089 + }, + { + "epoch": 0.77, + "learning_rate": 2.6107751625629863e-06, + "loss": 1.3052, + "step": 32090 + }, + { + "epoch": 0.77, + "learning_rate": 2.610250373377148e-06, + "loss": 0.9287, + "step": 32091 + }, + { + "epoch": 0.77, + "learning_rate": 2.6097256290231866e-06, + "loss": 0.8384, + "step": 32092 + }, + { + "epoch": 0.77, + "learning_rate": 2.6092009295042775e-06, + "loss": 1.0, + "step": 32093 + }, + { + "epoch": 0.77, + "learning_rate": 2.6086762748236093e-06, + "loss": 1.0387, + "step": 32094 + }, + { + "epoch": 0.77, + "learning_rate": 2.6081516649843652e-06, + "loss": 1.046, + "step": 32095 + }, + { + "epoch": 0.77, + "learning_rate": 2.6076270999897245e-06, + "loss": 1.1063, + "step": 32096 + }, + { + "epoch": 0.77, + "learning_rate": 2.607102579842874e-06, + "loss": 1.0177, + "step": 32097 + }, + { + "epoch": 0.77, + "learning_rate": 2.6065781045469897e-06, + "loss": 0.9638, + "step": 32098 + }, + { + "epoch": 0.77, + "learning_rate": 2.6060536741052576e-06, + "loss": 1.3806, + "step": 32099 + }, + { + "epoch": 0.77, + "learning_rate": 2.6055292885208615e-06, + "loss": 0.9476, + "step": 32100 + }, + { + "epoch": 0.77, + "learning_rate": 2.6050049477969774e-06, + "loss": 0.8855, + "step": 32101 + }, + { + "epoch": 0.77, + "learning_rate": 2.6044806519367884e-06, + "loss": 1.0828, + "step": 32102 + }, + { + "epoch": 0.77, + "learning_rate": 2.603956400943478e-06, + "loss": 1.0422, + "step": 32103 + }, + { + "epoch": 0.77, + "learning_rate": 2.6034321948202223e-06, + "loss": 1.0501, + "step": 32104 + }, + { + "epoch": 0.77, + "learning_rate": 2.6029080335702072e-06, + "loss": 1.2152, + "step": 32105 + }, + { + "epoch": 0.77, + "learning_rate": 2.6023839171966047e-06, + "loss": 0.9005, + "step": 32106 + }, + { + "epoch": 0.77, + "learning_rate": 2.6018598457026002e-06, + "loss": 1.3814, + "step": 32107 + }, + { + "epoch": 0.77, + "learning_rate": 2.601335819091374e-06, + "loss": 1.0487, + "step": 32108 + }, + { + "epoch": 0.77, + "learning_rate": 2.6008118373661006e-06, + "loss": 0.9924, + "step": 32109 + }, + { + "epoch": 0.77, + "learning_rate": 2.600287900529962e-06, + "loss": 1.0602, + "step": 32110 + }, + { + "epoch": 0.77, + "learning_rate": 2.5997640085861387e-06, + "loss": 1.2024, + "step": 32111 + }, + { + "epoch": 0.77, + "learning_rate": 2.5992401615378027e-06, + "loss": 0.9667, + "step": 32112 + }, + { + "epoch": 0.77, + "learning_rate": 2.5987163593881405e-06, + "loss": 1.0323, + "step": 32113 + }, + { + "epoch": 0.77, + "learning_rate": 2.5981926021403203e-06, + "loss": 1.0467, + "step": 32114 + }, + { + "epoch": 0.77, + "learning_rate": 2.5976688897975254e-06, + "loss": 1.1627, + "step": 32115 + }, + { + "epoch": 0.77, + "learning_rate": 2.597145222362936e-06, + "loss": 0.9741, + "step": 32116 + }, + { + "epoch": 0.77, + "learning_rate": 2.5966215998397213e-06, + "loss": 0.7839, + "step": 32117 + }, + { + "epoch": 0.77, + "learning_rate": 2.596098022231063e-06, + "loss": 0.8643, + "step": 32118 + }, + { + "epoch": 0.77, + "learning_rate": 2.5955744895401392e-06, + "loss": 0.9883, + "step": 32119 + }, + { + "epoch": 0.77, + "learning_rate": 2.5950510017701194e-06, + "loss": 1.0352, + "step": 32120 + }, + { + "epoch": 0.77, + "learning_rate": 2.5945275589241868e-06, + "loss": 0.803, + "step": 32121 + }, + { + "epoch": 0.77, + "learning_rate": 2.594004161005511e-06, + "loss": 0.9814, + "step": 32122 + }, + { + "epoch": 0.77, + "learning_rate": 2.5934808080172702e-06, + "loss": 1.1848, + "step": 32123 + }, + { + "epoch": 0.77, + "learning_rate": 2.5929574999626416e-06, + "loss": 1.1714, + "step": 32124 + }, + { + "epoch": 0.77, + "learning_rate": 2.5924342368447943e-06, + "loss": 0.9064, + "step": 32125 + }, + { + "epoch": 0.77, + "learning_rate": 2.5919110186669072e-06, + "loss": 1.2535, + "step": 32126 + }, + { + "epoch": 0.77, + "learning_rate": 2.5913878454321564e-06, + "loss": 0.9207, + "step": 32127 + }, + { + "epoch": 0.77, + "learning_rate": 2.5908647171437083e-06, + "loss": 1.0294, + "step": 32128 + }, + { + "epoch": 0.77, + "learning_rate": 2.590341633804743e-06, + "loss": 0.9189, + "step": 32129 + }, + { + "epoch": 0.77, + "learning_rate": 2.5898185954184342e-06, + "loss": 1.0405, + "step": 32130 + }, + { + "epoch": 0.77, + "learning_rate": 2.58929560198795e-06, + "loss": 1.1745, + "step": 32131 + }, + { + "epoch": 0.77, + "learning_rate": 2.58877265351647e-06, + "loss": 1.0355, + "step": 32132 + }, + { + "epoch": 0.77, + "learning_rate": 2.588249750007159e-06, + "loss": 0.8368, + "step": 32133 + }, + { + "epoch": 0.77, + "learning_rate": 2.587726891463195e-06, + "loss": 1.166, + "step": 32134 + }, + { + "epoch": 0.77, + "learning_rate": 2.5872040778877505e-06, + "loss": 0.764, + "step": 32135 + }, + { + "epoch": 0.77, + "learning_rate": 2.5866813092839915e-06, + "loss": 0.9283, + "step": 32136 + }, + { + "epoch": 0.77, + "learning_rate": 2.5861585856550943e-06, + "loss": 0.9665, + "step": 32137 + }, + { + "epoch": 0.77, + "learning_rate": 2.5856359070042324e-06, + "loss": 1.0961, + "step": 32138 + }, + { + "epoch": 0.77, + "learning_rate": 2.5851132733345698e-06, + "loss": 1.0117, + "step": 32139 + }, + { + "epoch": 0.77, + "learning_rate": 2.584590684649284e-06, + "loss": 1.0829, + "step": 32140 + }, + { + "epoch": 0.77, + "learning_rate": 2.58406814095154e-06, + "loss": 1.0122, + "step": 32141 + }, + { + "epoch": 0.77, + "learning_rate": 2.5835456422445093e-06, + "loss": 1.0272, + "step": 32142 + }, + { + "epoch": 0.77, + "learning_rate": 2.583023188531366e-06, + "loss": 0.8367, + "step": 32143 + }, + { + "epoch": 0.77, + "learning_rate": 2.5825007798152737e-06, + "loss": 0.9343, + "step": 32144 + }, + { + "epoch": 0.77, + "learning_rate": 2.5819784160994045e-06, + "loss": 1.0371, + "step": 32145 + }, + { + "epoch": 0.77, + "learning_rate": 2.58145609738693e-06, + "loss": 1.0095, + "step": 32146 + }, + { + "epoch": 0.77, + "learning_rate": 2.5809338236810133e-06, + "loss": 1.0007, + "step": 32147 + }, + { + "epoch": 0.77, + "learning_rate": 2.5804115949848298e-06, + "loss": 1.1622, + "step": 32148 + }, + { + "epoch": 0.77, + "learning_rate": 2.5798894113015406e-06, + "loss": 0.5746, + "step": 32149 + }, + { + "epoch": 0.77, + "learning_rate": 2.579367272634317e-06, + "loss": 0.8992, + "step": 32150 + }, + { + "epoch": 0.77, + "learning_rate": 2.5788451789863288e-06, + "loss": 1.152, + "step": 32151 + }, + { + "epoch": 0.77, + "learning_rate": 2.5783231303607393e-06, + "loss": 0.9831, + "step": 32152 + }, + { + "epoch": 0.77, + "learning_rate": 2.577801126760717e-06, + "loss": 1.0031, + "step": 32153 + }, + { + "epoch": 0.77, + "learning_rate": 2.577279168189434e-06, + "loss": 1.1036, + "step": 32154 + }, + { + "epoch": 0.77, + "learning_rate": 2.5767572546500484e-06, + "loss": 1.14, + "step": 32155 + }, + { + "epoch": 0.77, + "learning_rate": 2.5762353861457335e-06, + "loss": 0.8665, + "step": 32156 + }, + { + "epoch": 0.77, + "learning_rate": 2.575713562679649e-06, + "loss": 1.1843, + "step": 32157 + }, + { + "epoch": 0.77, + "learning_rate": 2.5751917842549647e-06, + "loss": 1.0654, + "step": 32158 + }, + { + "epoch": 0.77, + "learning_rate": 2.574670050874848e-06, + "loss": 1.0887, + "step": 32159 + }, + { + "epoch": 0.77, + "learning_rate": 2.574148362542458e-06, + "loss": 1.3998, + "step": 32160 + }, + { + "epoch": 0.77, + "learning_rate": 2.5736267192609643e-06, + "loss": 0.8942, + "step": 32161 + }, + { + "epoch": 0.77, + "learning_rate": 2.573105121033532e-06, + "loss": 0.936, + "step": 32162 + }, + { + "epoch": 0.77, + "learning_rate": 2.5725835678633226e-06, + "loss": 0.9853, + "step": 32163 + }, + { + "epoch": 0.77, + "learning_rate": 2.572062059753504e-06, + "loss": 1.089, + "step": 32164 + }, + { + "epoch": 0.77, + "learning_rate": 2.571540596707235e-06, + "loss": 0.9863, + "step": 32165 + }, + { + "epoch": 0.77, + "learning_rate": 2.5710191787276783e-06, + "loss": 1.294, + "step": 32166 + }, + { + "epoch": 0.77, + "learning_rate": 2.5704978058180074e-06, + "loss": 1.1204, + "step": 32167 + }, + { + "epoch": 0.77, + "learning_rate": 2.5699764779813763e-06, + "loss": 0.868, + "step": 32168 + }, + { + "epoch": 0.77, + "learning_rate": 2.5694551952209523e-06, + "loss": 1.169, + "step": 32169 + }, + { + "epoch": 0.77, + "learning_rate": 2.568933957539893e-06, + "loss": 1.0658, + "step": 32170 + }, + { + "epoch": 0.77, + "learning_rate": 2.5684127649413636e-06, + "loss": 0.9315, + "step": 32171 + }, + { + "epoch": 0.77, + "learning_rate": 2.5678916174285296e-06, + "loss": 0.8968, + "step": 32172 + }, + { + "epoch": 0.77, + "learning_rate": 2.567370515004545e-06, + "loss": 0.8861, + "step": 32173 + }, + { + "epoch": 0.77, + "learning_rate": 2.5668494576725754e-06, + "loss": 0.8799, + "step": 32174 + }, + { + "epoch": 0.77, + "learning_rate": 2.566328445435785e-06, + "loss": 1.1598, + "step": 32175 + }, + { + "epoch": 0.77, + "learning_rate": 2.5658074782973288e-06, + "loss": 0.8719, + "step": 32176 + }, + { + "epoch": 0.77, + "learning_rate": 2.565286556260372e-06, + "loss": 1.0442, + "step": 32177 + }, + { + "epoch": 0.77, + "learning_rate": 2.564765679328071e-06, + "loss": 0.8368, + "step": 32178 + }, + { + "epoch": 0.77, + "learning_rate": 2.5642448475035874e-06, + "loss": 1.0703, + "step": 32179 + }, + { + "epoch": 0.77, + "learning_rate": 2.5637240607900837e-06, + "loss": 0.9971, + "step": 32180 + }, + { + "epoch": 0.77, + "learning_rate": 2.5632033191907147e-06, + "loss": 1.1298, + "step": 32181 + }, + { + "epoch": 0.77, + "learning_rate": 2.562682622708642e-06, + "loss": 0.9773, + "step": 32182 + }, + { + "epoch": 0.77, + "learning_rate": 2.562161971347027e-06, + "loss": 0.9288, + "step": 32183 + }, + { + "epoch": 0.77, + "learning_rate": 2.561641365109023e-06, + "loss": 1.0354, + "step": 32184 + }, + { + "epoch": 0.77, + "learning_rate": 2.5611208039977943e-06, + "loss": 0.96, + "step": 32185 + }, + { + "epoch": 0.77, + "learning_rate": 2.5606002880164925e-06, + "loss": 1.201, + "step": 32186 + }, + { + "epoch": 0.77, + "learning_rate": 2.5600798171682794e-06, + "loss": 1.1904, + "step": 32187 + }, + { + "epoch": 0.77, + "learning_rate": 2.5595593914563155e-06, + "loss": 0.8292, + "step": 32188 + }, + { + "epoch": 0.77, + "learning_rate": 2.5590390108837504e-06, + "loss": 0.846, + "step": 32189 + }, + { + "epoch": 0.77, + "learning_rate": 2.558518675453746e-06, + "loss": 1.2131, + "step": 32190 + }, + { + "epoch": 0.77, + "learning_rate": 2.5579983851694624e-06, + "loss": 1.3077, + "step": 32191 + }, + { + "epoch": 0.77, + "learning_rate": 2.5574781400340477e-06, + "loss": 0.9121, + "step": 32192 + }, + { + "epoch": 0.77, + "learning_rate": 2.5569579400506663e-06, + "loss": 1.212, + "step": 32193 + }, + { + "epoch": 0.77, + "learning_rate": 2.5564377852224676e-06, + "loss": 1.3969, + "step": 32194 + }, + { + "epoch": 0.77, + "learning_rate": 2.555917675552609e-06, + "loss": 0.7045, + "step": 32195 + }, + { + "epoch": 0.77, + "learning_rate": 2.555397611044251e-06, + "loss": 1.1244, + "step": 32196 + }, + { + "epoch": 0.77, + "learning_rate": 2.5548775917005407e-06, + "loss": 0.7661, + "step": 32197 + }, + { + "epoch": 0.77, + "learning_rate": 2.5543576175246378e-06, + "loss": 1.1728, + "step": 32198 + }, + { + "epoch": 0.77, + "learning_rate": 2.553837688519698e-06, + "loss": 0.8329, + "step": 32199 + }, + { + "epoch": 0.77, + "learning_rate": 2.553317804688872e-06, + "loss": 0.9957, + "step": 32200 + }, + { + "epoch": 0.77, + "learning_rate": 2.5527979660353166e-06, + "loss": 1.0083, + "step": 32201 + }, + { + "epoch": 0.77, + "learning_rate": 2.5522781725621814e-06, + "loss": 0.9453, + "step": 32202 + }, + { + "epoch": 0.77, + "learning_rate": 2.5517584242726233e-06, + "loss": 1.2974, + "step": 32203 + }, + { + "epoch": 0.77, + "learning_rate": 2.5512387211697988e-06, + "loss": 0.9418, + "step": 32204 + }, + { + "epoch": 0.77, + "learning_rate": 2.5507190632568524e-06, + "loss": 1.1673, + "step": 32205 + }, + { + "epoch": 0.77, + "learning_rate": 2.550199450536942e-06, + "loss": 1.1465, + "step": 32206 + }, + { + "epoch": 0.77, + "learning_rate": 2.549679883013222e-06, + "loss": 1.0963, + "step": 32207 + }, + { + "epoch": 0.77, + "learning_rate": 2.5491603606888384e-06, + "loss": 0.9061, + "step": 32208 + }, + { + "epoch": 0.77, + "learning_rate": 2.5486408835669473e-06, + "loss": 1.0413, + "step": 32209 + }, + { + "epoch": 0.77, + "learning_rate": 2.5481214516507015e-06, + "loss": 1.1747, + "step": 32210 + }, + { + "epoch": 0.77, + "learning_rate": 2.5476020649432477e-06, + "loss": 1.1152, + "step": 32211 + }, + { + "epoch": 0.77, + "learning_rate": 2.5470827234477414e-06, + "loss": 0.9977, + "step": 32212 + }, + { + "epoch": 0.77, + "learning_rate": 2.546563427167329e-06, + "loss": 1.2052, + "step": 32213 + }, + { + "epoch": 0.77, + "learning_rate": 2.5460441761051625e-06, + "loss": 1.2324, + "step": 32214 + }, + { + "epoch": 0.77, + "learning_rate": 2.545524970264396e-06, + "loss": 1.1265, + "step": 32215 + }, + { + "epoch": 0.77, + "learning_rate": 2.5450058096481733e-06, + "loss": 0.9563, + "step": 32216 + }, + { + "epoch": 0.77, + "learning_rate": 2.544486694259646e-06, + "loss": 0.8554, + "step": 32217 + }, + { + "epoch": 0.77, + "learning_rate": 2.5439676241019684e-06, + "loss": 1.0947, + "step": 32218 + }, + { + "epoch": 0.77, + "learning_rate": 2.543448599178282e-06, + "loss": 1.1312, + "step": 32219 + }, + { + "epoch": 0.77, + "learning_rate": 2.542929619491742e-06, + "loss": 1.0649, + "step": 32220 + }, + { + "epoch": 0.77, + "learning_rate": 2.542410685045491e-06, + "loss": 1.2887, + "step": 32221 + }, + { + "epoch": 0.77, + "learning_rate": 2.54189179584268e-06, + "loss": 1.0824, + "step": 32222 + }, + { + "epoch": 0.77, + "learning_rate": 2.5413729518864604e-06, + "loss": 1.0616, + "step": 32223 + }, + { + "epoch": 0.77, + "learning_rate": 2.5408541531799735e-06, + "loss": 1.095, + "step": 32224 + }, + { + "epoch": 0.77, + "learning_rate": 2.5403353997263704e-06, + "loss": 0.991, + "step": 32225 + }, + { + "epoch": 0.77, + "learning_rate": 2.5398166915288004e-06, + "loss": 1.1343, + "step": 32226 + }, + { + "epoch": 0.78, + "learning_rate": 2.5392980285904057e-06, + "loss": 0.8752, + "step": 32227 + }, + { + "epoch": 0.78, + "learning_rate": 2.538779410914337e-06, + "loss": 1.0059, + "step": 32228 + }, + { + "epoch": 0.78, + "learning_rate": 2.5382608385037355e-06, + "loss": 0.918, + "step": 32229 + }, + { + "epoch": 0.78, + "learning_rate": 2.537742311361752e-06, + "loss": 1.2373, + "step": 32230 + }, + { + "epoch": 0.78, + "learning_rate": 2.537223829491532e-06, + "loss": 0.904, + "step": 32231 + }, + { + "epoch": 0.78, + "learning_rate": 2.536705392896217e-06, + "loss": 1.0022, + "step": 32232 + }, + { + "epoch": 0.78, + "learning_rate": 2.5361870015789548e-06, + "loss": 1.2302, + "step": 32233 + }, + { + "epoch": 0.78, + "learning_rate": 2.5356686555428935e-06, + "loss": 1.0502, + "step": 32234 + }, + { + "epoch": 0.78, + "learning_rate": 2.535150354791173e-06, + "loss": 0.9746, + "step": 32235 + }, + { + "epoch": 0.78, + "learning_rate": 2.5346320993269402e-06, + "loss": 0.9914, + "step": 32236 + }, + { + "epoch": 0.78, + "learning_rate": 2.5341138891533366e-06, + "loss": 0.9724, + "step": 32237 + }, + { + "epoch": 0.78, + "learning_rate": 2.5335957242735086e-06, + "loss": 1.0306, + "step": 32238 + }, + { + "epoch": 0.78, + "learning_rate": 2.5330776046906012e-06, + "loss": 1.1887, + "step": 32239 + }, + { + "epoch": 0.78, + "learning_rate": 2.5325595304077532e-06, + "loss": 1.0921, + "step": 32240 + }, + { + "epoch": 0.78, + "learning_rate": 2.532041501428111e-06, + "loss": 1.0537, + "step": 32241 + }, + { + "epoch": 0.78, + "learning_rate": 2.531523517754818e-06, + "loss": 0.781, + "step": 32242 + }, + { + "epoch": 0.78, + "learning_rate": 2.531005579391013e-06, + "loss": 1.0525, + "step": 32243 + }, + { + "epoch": 0.78, + "learning_rate": 2.530487686339842e-06, + "loss": 1.1137, + "step": 32244 + }, + { + "epoch": 0.78, + "learning_rate": 2.5299698386044435e-06, + "loss": 0.8915, + "step": 32245 + }, + { + "epoch": 0.78, + "learning_rate": 2.5294520361879606e-06, + "loss": 1.0785, + "step": 32246 + }, + { + "epoch": 0.78, + "learning_rate": 2.528934279093538e-06, + "loss": 1.16, + "step": 32247 + }, + { + "epoch": 0.78, + "learning_rate": 2.528416567324311e-06, + "loss": 1.2013, + "step": 32248 + }, + { + "epoch": 0.78, + "learning_rate": 2.527898900883423e-06, + "loss": 1.0379, + "step": 32249 + }, + { + "epoch": 0.78, + "learning_rate": 2.527381279774017e-06, + "loss": 1.119, + "step": 32250 + }, + { + "epoch": 0.78, + "learning_rate": 2.5268637039992296e-06, + "loss": 1.0942, + "step": 32251 + }, + { + "epoch": 0.78, + "learning_rate": 2.526346173562202e-06, + "loss": 0.8847, + "step": 32252 + }, + { + "epoch": 0.78, + "learning_rate": 2.5258286884660764e-06, + "loss": 1.077, + "step": 32253 + }, + { + "epoch": 0.78, + "learning_rate": 2.5253112487139873e-06, + "loss": 0.8016, + "step": 32254 + }, + { + "epoch": 0.78, + "learning_rate": 2.524793854309081e-06, + "loss": 1.013, + "step": 32255 + }, + { + "epoch": 0.78, + "learning_rate": 2.524276505254488e-06, + "loss": 0.9442, + "step": 32256 + }, + { + "epoch": 0.78, + "learning_rate": 2.523759201553351e-06, + "loss": 0.8407, + "step": 32257 + }, + { + "epoch": 0.78, + "learning_rate": 2.523241943208812e-06, + "loss": 1.0794, + "step": 32258 + }, + { + "epoch": 0.78, + "learning_rate": 2.5227247302240022e-06, + "loss": 1.2122, + "step": 32259 + }, + { + "epoch": 0.78, + "learning_rate": 2.5222075626020624e-06, + "loss": 1.1775, + "step": 32260 + }, + { + "epoch": 0.78, + "learning_rate": 2.521690440346134e-06, + "loss": 0.9881, + "step": 32261 + }, + { + "epoch": 0.78, + "learning_rate": 2.521173363459347e-06, + "loss": 1.0934, + "step": 32262 + }, + { + "epoch": 0.78, + "learning_rate": 2.5206563319448453e-06, + "loss": 0.88, + "step": 32263 + }, + { + "epoch": 0.78, + "learning_rate": 2.52013934580576e-06, + "loss": 0.9315, + "step": 32264 + }, + { + "epoch": 0.78, + "learning_rate": 2.519622405045229e-06, + "loss": 1.164, + "step": 32265 + }, + { + "epoch": 0.78, + "learning_rate": 2.519105509666393e-06, + "loss": 0.994, + "step": 32266 + }, + { + "epoch": 0.78, + "learning_rate": 2.5185886596723806e-06, + "loss": 1.0048, + "step": 32267 + }, + { + "epoch": 0.78, + "learning_rate": 2.5180718550663317e-06, + "loss": 1.0465, + "step": 32268 + }, + { + "epoch": 0.78, + "learning_rate": 2.517555095851384e-06, + "loss": 0.8083, + "step": 32269 + }, + { + "epoch": 0.78, + "learning_rate": 2.5170383820306667e-06, + "loss": 1.152, + "step": 32270 + }, + { + "epoch": 0.78, + "learning_rate": 2.5165217136073195e-06, + "loss": 1.2365, + "step": 32271 + }, + { + "epoch": 0.78, + "learning_rate": 2.5160050905844724e-06, + "loss": 1.1251, + "step": 32272 + }, + { + "epoch": 0.78, + "learning_rate": 2.515488512965263e-06, + "loss": 1.0367, + "step": 32273 + }, + { + "epoch": 0.78, + "learning_rate": 2.514971980752826e-06, + "loss": 1.1409, + "step": 32274 + }, + { + "epoch": 0.78, + "learning_rate": 2.514455493950291e-06, + "loss": 0.9061, + "step": 32275 + }, + { + "epoch": 0.78, + "learning_rate": 2.5139390525607934e-06, + "loss": 1.0791, + "step": 32276 + }, + { + "epoch": 0.78, + "learning_rate": 2.51342265658747e-06, + "loss": 1.0883, + "step": 32277 + }, + { + "epoch": 0.78, + "learning_rate": 2.512906306033448e-06, + "loss": 1.2209, + "step": 32278 + }, + { + "epoch": 0.78, + "learning_rate": 2.5123900009018644e-06, + "loss": 0.8596, + "step": 32279 + }, + { + "epoch": 0.78, + "learning_rate": 2.511873741195847e-06, + "loss": 1.1225, + "step": 32280 + }, + { + "epoch": 0.78, + "learning_rate": 2.51135752691853e-06, + "loss": 0.933, + "step": 32281 + }, + { + "epoch": 0.78, + "learning_rate": 2.5108413580730486e-06, + "loss": 0.9633, + "step": 32282 + }, + { + "epoch": 0.78, + "learning_rate": 2.5103252346625286e-06, + "loss": 1.0147, + "step": 32283 + }, + { + "epoch": 0.78, + "learning_rate": 2.509809156690103e-06, + "loss": 1.3032, + "step": 32284 + }, + { + "epoch": 0.78, + "learning_rate": 2.509293124158907e-06, + "loss": 0.9922, + "step": 32285 + }, + { + "epoch": 0.78, + "learning_rate": 2.508777137072064e-06, + "loss": 0.9655, + "step": 32286 + }, + { + "epoch": 0.78, + "learning_rate": 2.508261195432712e-06, + "loss": 1.0311, + "step": 32287 + }, + { + "epoch": 0.78, + "learning_rate": 2.5077452992439733e-06, + "loss": 1.0497, + "step": 32288 + }, + { + "epoch": 0.78, + "learning_rate": 2.507229448508982e-06, + "loss": 0.9101, + "step": 32289 + }, + { + "epoch": 0.78, + "learning_rate": 2.506713643230869e-06, + "loss": 0.8986, + "step": 32290 + }, + { + "epoch": 0.78, + "learning_rate": 2.506197883412761e-06, + "loss": 1.0121, + "step": 32291 + }, + { + "epoch": 0.78, + "learning_rate": 2.505682169057786e-06, + "loss": 0.8634, + "step": 32292 + }, + { + "epoch": 0.78, + "learning_rate": 2.5051665001690786e-06, + "loss": 1.0874, + "step": 32293 + }, + { + "epoch": 0.78, + "learning_rate": 2.50465087674976e-06, + "loss": 1.1044, + "step": 32294 + }, + { + "epoch": 0.78, + "learning_rate": 2.504135298802961e-06, + "loss": 1.0832, + "step": 32295 + }, + { + "epoch": 0.78, + "learning_rate": 2.503619766331813e-06, + "loss": 1.0332, + "step": 32296 + }, + { + "epoch": 0.78, + "learning_rate": 2.5031042793394376e-06, + "loss": 1.4058, + "step": 32297 + }, + { + "epoch": 0.78, + "learning_rate": 2.5025888378289687e-06, + "loss": 0.9335, + "step": 32298 + }, + { + "epoch": 0.78, + "learning_rate": 2.5020734418035275e-06, + "loss": 1.1236, + "step": 32299 + }, + { + "epoch": 0.78, + "learning_rate": 2.501558091266242e-06, + "loss": 1.0852, + "step": 32300 + }, + { + "epoch": 0.78, + "learning_rate": 2.5010427862202424e-06, + "loss": 1.0333, + "step": 32301 + }, + { + "epoch": 0.78, + "learning_rate": 2.50052752666865e-06, + "loss": 0.7979, + "step": 32302 + }, + { + "epoch": 0.78, + "learning_rate": 2.5000123126145935e-06, + "loss": 1.015, + "step": 32303 + }, + { + "epoch": 0.78, + "learning_rate": 2.4994971440612004e-06, + "loss": 1.1932, + "step": 32304 + }, + { + "epoch": 0.78, + "learning_rate": 2.4989820210115912e-06, + "loss": 0.9384, + "step": 32305 + }, + { + "epoch": 0.78, + "learning_rate": 2.498466943468897e-06, + "loss": 1.1032, + "step": 32306 + }, + { + "epoch": 0.78, + "learning_rate": 2.497951911436236e-06, + "loss": 0.8274, + "step": 32307 + }, + { + "epoch": 0.78, + "learning_rate": 2.4974369249167375e-06, + "loss": 1.0574, + "step": 32308 + }, + { + "epoch": 0.78, + "learning_rate": 2.4969219839135263e-06, + "loss": 1.2088, + "step": 32309 + }, + { + "epoch": 0.78, + "learning_rate": 2.496407088429721e-06, + "loss": 1.0445, + "step": 32310 + }, + { + "epoch": 0.78, + "learning_rate": 2.495892238468449e-06, + "loss": 0.8844, + "step": 32311 + }, + { + "epoch": 0.78, + "learning_rate": 2.4953774340328373e-06, + "loss": 1.0927, + "step": 32312 + }, + { + "epoch": 0.78, + "learning_rate": 2.494862675126003e-06, + "loss": 1.2615, + "step": 32313 + }, + { + "epoch": 0.78, + "learning_rate": 2.4943479617510745e-06, + "loss": 1.244, + "step": 32314 + }, + { + "epoch": 0.78, + "learning_rate": 2.4938332939111674e-06, + "loss": 1.0842, + "step": 32315 + }, + { + "epoch": 0.78, + "learning_rate": 2.4933186716094093e-06, + "loss": 1.2842, + "step": 32316 + }, + { + "epoch": 0.78, + "learning_rate": 2.492804094848923e-06, + "loss": 1.334, + "step": 32317 + }, + { + "epoch": 0.78, + "learning_rate": 2.4922895636328257e-06, + "loss": 1.0591, + "step": 32318 + }, + { + "epoch": 0.78, + "learning_rate": 2.4917750779642425e-06, + "loss": 1.221, + "step": 32319 + }, + { + "epoch": 0.78, + "learning_rate": 2.491260637846297e-06, + "loss": 0.5772, + "step": 32320 + }, + { + "epoch": 0.78, + "learning_rate": 2.490746243282103e-06, + "loss": 1.0908, + "step": 32321 + }, + { + "epoch": 0.78, + "learning_rate": 2.4902318942747883e-06, + "loss": 1.0821, + "step": 32322 + }, + { + "epoch": 0.78, + "learning_rate": 2.4897175908274683e-06, + "loss": 0.8656, + "step": 32323 + }, + { + "epoch": 0.78, + "learning_rate": 2.489203332943264e-06, + "loss": 0.9809, + "step": 32324 + }, + { + "epoch": 0.78, + "learning_rate": 2.4886891206253004e-06, + "loss": 1.1094, + "step": 32325 + }, + { + "epoch": 0.78, + "learning_rate": 2.4881749538766895e-06, + "loss": 0.9892, + "step": 32326 + }, + { + "epoch": 0.78, + "learning_rate": 2.487660832700555e-06, + "loss": 1.1503, + "step": 32327 + }, + { + "epoch": 0.78, + "learning_rate": 2.487146757100017e-06, + "loss": 1.0083, + "step": 32328 + }, + { + "epoch": 0.78, + "learning_rate": 2.4866327270781907e-06, + "loss": 0.9995, + "step": 32329 + }, + { + "epoch": 0.78, + "learning_rate": 2.4861187426381993e-06, + "loss": 1.1761, + "step": 32330 + }, + { + "epoch": 0.78, + "learning_rate": 2.485604803783155e-06, + "loss": 1.0476, + "step": 32331 + }, + { + "epoch": 0.78, + "learning_rate": 2.4850909105161747e-06, + "loss": 1.1641, + "step": 32332 + }, + { + "epoch": 0.78, + "learning_rate": 2.4845770628403875e-06, + "loss": 0.9499, + "step": 32333 + }, + { + "epoch": 0.78, + "learning_rate": 2.4840632607589e-06, + "loss": 1.0147, + "step": 32334 + }, + { + "epoch": 0.78, + "learning_rate": 2.483549504274836e-06, + "loss": 0.8736, + "step": 32335 + }, + { + "epoch": 0.78, + "learning_rate": 2.4830357933913063e-06, + "loss": 1.0466, + "step": 32336 + }, + { + "epoch": 0.78, + "learning_rate": 2.48252212811143e-06, + "loss": 1.1994, + "step": 32337 + }, + { + "epoch": 0.78, + "learning_rate": 2.4820085084383273e-06, + "loss": 1.0592, + "step": 32338 + }, + { + "epoch": 0.78, + "learning_rate": 2.481494934375107e-06, + "loss": 1.2379, + "step": 32339 + }, + { + "epoch": 0.78, + "learning_rate": 2.4809814059248882e-06, + "loss": 1.09, + "step": 32340 + }, + { + "epoch": 0.78, + "learning_rate": 2.48046792309079e-06, + "loss": 0.9616, + "step": 32341 + }, + { + "epoch": 0.78, + "learning_rate": 2.4799544858759218e-06, + "loss": 1.0669, + "step": 32342 + }, + { + "epoch": 0.78, + "learning_rate": 2.4794410942834034e-06, + "loss": 0.8315, + "step": 32343 + }, + { + "epoch": 0.78, + "learning_rate": 2.4789277483163443e-06, + "loss": 1.0145, + "step": 32344 + }, + { + "epoch": 0.78, + "learning_rate": 2.4784144479778617e-06, + "loss": 1.134, + "step": 32345 + }, + { + "epoch": 0.78, + "learning_rate": 2.4779011932710717e-06, + "loss": 0.8442, + "step": 32346 + }, + { + "epoch": 0.78, + "learning_rate": 2.4773879841990835e-06, + "loss": 0.9024, + "step": 32347 + }, + { + "epoch": 0.78, + "learning_rate": 2.476874820765012e-06, + "loss": 1.0371, + "step": 32348 + }, + { + "epoch": 0.78, + "learning_rate": 2.4763617029719757e-06, + "loss": 1.0617, + "step": 32349 + }, + { + "epoch": 0.78, + "learning_rate": 2.475848630823079e-06, + "loss": 0.9109, + "step": 32350 + }, + { + "epoch": 0.78, + "learning_rate": 2.475335604321443e-06, + "loss": 1.0397, + "step": 32351 + }, + { + "epoch": 0.78, + "learning_rate": 2.4748226234701723e-06, + "loss": 0.7903, + "step": 32352 + }, + { + "epoch": 0.78, + "learning_rate": 2.4743096882723827e-06, + "loss": 0.8956, + "step": 32353 + }, + { + "epoch": 0.78, + "learning_rate": 2.4737967987311893e-06, + "loss": 1.3357, + "step": 32354 + }, + { + "epoch": 0.78, + "learning_rate": 2.4732839548496966e-06, + "loss": 0.9233, + "step": 32355 + }, + { + "epoch": 0.78, + "learning_rate": 2.4727711566310207e-06, + "loss": 1.0197, + "step": 32356 + }, + { + "epoch": 0.78, + "learning_rate": 2.472258404078275e-06, + "loss": 0.9469, + "step": 32357 + }, + { + "epoch": 0.78, + "learning_rate": 2.471745697194563e-06, + "loss": 0.9674, + "step": 32358 + }, + { + "epoch": 0.78, + "learning_rate": 2.4712330359830027e-06, + "loss": 1.2156, + "step": 32359 + }, + { + "epoch": 0.78, + "learning_rate": 2.470720420446697e-06, + "loss": 1.026, + "step": 32360 + }, + { + "epoch": 0.78, + "learning_rate": 2.4702078505887607e-06, + "loss": 0.8599, + "step": 32361 + }, + { + "epoch": 0.78, + "learning_rate": 2.4696953264123048e-06, + "loss": 1.0015, + "step": 32362 + }, + { + "epoch": 0.78, + "learning_rate": 2.4691828479204327e-06, + "loss": 1.3864, + "step": 32363 + }, + { + "epoch": 0.78, + "learning_rate": 2.4686704151162575e-06, + "loss": 1.0043, + "step": 32364 + }, + { + "epoch": 0.78, + "learning_rate": 2.4681580280028894e-06, + "loss": 1.0741, + "step": 32365 + }, + { + "epoch": 0.78, + "learning_rate": 2.467645686583433e-06, + "loss": 1.3094, + "step": 32366 + }, + { + "epoch": 0.78, + "learning_rate": 2.467133390861001e-06, + "loss": 0.913, + "step": 32367 + }, + { + "epoch": 0.78, + "learning_rate": 2.4666211408386963e-06, + "loss": 1.2149, + "step": 32368 + }, + { + "epoch": 0.78, + "learning_rate": 2.466108936519629e-06, + "loss": 0.8647, + "step": 32369 + }, + { + "epoch": 0.78, + "learning_rate": 2.465596777906909e-06, + "loss": 1.0858, + "step": 32370 + }, + { + "epoch": 0.78, + "learning_rate": 2.4650846650036385e-06, + "loss": 0.9549, + "step": 32371 + }, + { + "epoch": 0.78, + "learning_rate": 2.4645725978129277e-06, + "loss": 1.1059, + "step": 32372 + }, + { + "epoch": 0.78, + "learning_rate": 2.4640605763378844e-06, + "loss": 0.8372, + "step": 32373 + }, + { + "epoch": 0.78, + "learning_rate": 2.4635486005816113e-06, + "loss": 1.1725, + "step": 32374 + }, + { + "epoch": 0.78, + "learning_rate": 2.4630366705472153e-06, + "loss": 0.9881, + "step": 32375 + }, + { + "epoch": 0.78, + "learning_rate": 2.4625247862378067e-06, + "loss": 1.0424, + "step": 32376 + }, + { + "epoch": 0.78, + "learning_rate": 2.4620129476564835e-06, + "loss": 1.108, + "step": 32377 + }, + { + "epoch": 0.78, + "learning_rate": 2.4615011548063582e-06, + "loss": 1.1522, + "step": 32378 + }, + { + "epoch": 0.78, + "learning_rate": 2.4609894076905284e-06, + "loss": 1.0813, + "step": 32379 + }, + { + "epoch": 0.78, + "learning_rate": 2.460477706312103e-06, + "loss": 1.0132, + "step": 32380 + }, + { + "epoch": 0.78, + "learning_rate": 2.4599660506741896e-06, + "loss": 1.0729, + "step": 32381 + }, + { + "epoch": 0.78, + "learning_rate": 2.459454440779886e-06, + "loss": 1.0149, + "step": 32382 + }, + { + "epoch": 0.78, + "learning_rate": 2.4589428766322978e-06, + "loss": 0.9886, + "step": 32383 + }, + { + "epoch": 0.78, + "learning_rate": 2.458431358234532e-06, + "loss": 1.2892, + "step": 32384 + }, + { + "epoch": 0.78, + "learning_rate": 2.457919885589687e-06, + "loss": 1.1064, + "step": 32385 + }, + { + "epoch": 0.78, + "learning_rate": 2.4574084587008706e-06, + "loss": 0.8145, + "step": 32386 + }, + { + "epoch": 0.78, + "learning_rate": 2.4568970775711797e-06, + "loss": 1.1558, + "step": 32387 + }, + { + "epoch": 0.78, + "learning_rate": 2.456385742203721e-06, + "loss": 0.9452, + "step": 32388 + }, + { + "epoch": 0.78, + "learning_rate": 2.4558744526015977e-06, + "loss": 1.1922, + "step": 32389 + }, + { + "epoch": 0.78, + "learning_rate": 2.455363208767908e-06, + "loss": 0.9465, + "step": 32390 + }, + { + "epoch": 0.78, + "learning_rate": 2.454852010705754e-06, + "loss": 0.9131, + "step": 32391 + }, + { + "epoch": 0.78, + "learning_rate": 2.4543408584182404e-06, + "loss": 1.0631, + "step": 32392 + }, + { + "epoch": 0.78, + "learning_rate": 2.453829751908464e-06, + "loss": 0.8286, + "step": 32393 + }, + { + "epoch": 0.78, + "learning_rate": 2.4533186911795303e-06, + "loss": 0.7968, + "step": 32394 + }, + { + "epoch": 0.78, + "learning_rate": 2.4528076762345345e-06, + "loss": 0.9963, + "step": 32395 + }, + { + "epoch": 0.78, + "learning_rate": 2.4522967070765792e-06, + "loss": 1.0241, + "step": 32396 + }, + { + "epoch": 0.78, + "learning_rate": 2.4517857837087664e-06, + "loss": 0.9667, + "step": 32397 + }, + { + "epoch": 0.78, + "learning_rate": 2.451274906134191e-06, + "loss": 0.9304, + "step": 32398 + }, + { + "epoch": 0.78, + "learning_rate": 2.450764074355957e-06, + "loss": 0.915, + "step": 32399 + }, + { + "epoch": 0.78, + "learning_rate": 2.450253288377162e-06, + "loss": 0.8565, + "step": 32400 + }, + { + "epoch": 0.78, + "learning_rate": 2.449742548200904e-06, + "loss": 1.2089, + "step": 32401 + }, + { + "epoch": 0.78, + "learning_rate": 2.4492318538302837e-06, + "loss": 1.0898, + "step": 32402 + }, + { + "epoch": 0.78, + "learning_rate": 2.448721205268395e-06, + "loss": 1.0045, + "step": 32403 + }, + { + "epoch": 0.78, + "learning_rate": 2.4482106025183383e-06, + "loss": 0.9552, + "step": 32404 + }, + { + "epoch": 0.78, + "learning_rate": 2.4477000455832143e-06, + "loss": 0.6449, + "step": 32405 + }, + { + "epoch": 0.78, + "learning_rate": 2.4471895344661154e-06, + "loss": 1.0486, + "step": 32406 + }, + { + "epoch": 0.78, + "learning_rate": 2.44667906917014e-06, + "loss": 1.1808, + "step": 32407 + }, + { + "epoch": 0.78, + "learning_rate": 2.4461686496983906e-06, + "loss": 0.9099, + "step": 32408 + }, + { + "epoch": 0.78, + "learning_rate": 2.445658276053956e-06, + "loss": 0.9177, + "step": 32409 + }, + { + "epoch": 0.78, + "learning_rate": 2.4451479482399377e-06, + "loss": 0.9467, + "step": 32410 + }, + { + "epoch": 0.78, + "learning_rate": 2.4446376662594275e-06, + "loss": 1.0685, + "step": 32411 + }, + { + "epoch": 0.78, + "learning_rate": 2.444127430115523e-06, + "loss": 0.9104, + "step": 32412 + }, + { + "epoch": 0.78, + "learning_rate": 2.4436172398113235e-06, + "loss": 1.0204, + "step": 32413 + }, + { + "epoch": 0.78, + "learning_rate": 2.443107095349918e-06, + "loss": 1.1496, + "step": 32414 + }, + { + "epoch": 0.78, + "learning_rate": 2.4425969967344045e-06, + "loss": 1.1379, + "step": 32415 + }, + { + "epoch": 0.78, + "learning_rate": 2.4420869439678796e-06, + "loss": 1.2501, + "step": 32416 + }, + { + "epoch": 0.78, + "learning_rate": 2.4415769370534327e-06, + "loss": 0.9012, + "step": 32417 + }, + { + "epoch": 0.78, + "learning_rate": 2.441066975994163e-06, + "loss": 1.0609, + "step": 32418 + }, + { + "epoch": 0.78, + "learning_rate": 2.4405570607931604e-06, + "loss": 0.8726, + "step": 32419 + }, + { + "epoch": 0.78, + "learning_rate": 2.4400471914535193e-06, + "loss": 1.1448, + "step": 32420 + }, + { + "epoch": 0.78, + "learning_rate": 2.4395373679783364e-06, + "loss": 1.0789, + "step": 32421 + }, + { + "epoch": 0.78, + "learning_rate": 2.4390275903706995e-06, + "loss": 0.9972, + "step": 32422 + }, + { + "epoch": 0.78, + "learning_rate": 2.4385178586337033e-06, + "loss": 0.6794, + "step": 32423 + }, + { + "epoch": 0.78, + "learning_rate": 2.438008172770444e-06, + "loss": 0.908, + "step": 32424 + }, + { + "epoch": 0.78, + "learning_rate": 2.4374985327840075e-06, + "loss": 1.0585, + "step": 32425 + }, + { + "epoch": 0.78, + "learning_rate": 2.436988938677489e-06, + "loss": 0.9962, + "step": 32426 + }, + { + "epoch": 0.78, + "learning_rate": 2.436479390453982e-06, + "loss": 1.2408, + "step": 32427 + }, + { + "epoch": 0.78, + "learning_rate": 2.4359698881165726e-06, + "loss": 1.0315, + "step": 32428 + }, + { + "epoch": 0.78, + "learning_rate": 2.4354604316683572e-06, + "loss": 1.1348, + "step": 32429 + }, + { + "epoch": 0.78, + "learning_rate": 2.434951021112423e-06, + "loss": 0.9957, + "step": 32430 + }, + { + "epoch": 0.78, + "learning_rate": 2.4344416564518593e-06, + "loss": 1.017, + "step": 32431 + }, + { + "epoch": 0.78, + "learning_rate": 2.4339323376897626e-06, + "loss": 0.9917, + "step": 32432 + }, + { + "epoch": 0.78, + "learning_rate": 2.4334230648292157e-06, + "loss": 0.8041, + "step": 32433 + }, + { + "epoch": 0.78, + "learning_rate": 2.432913837873311e-06, + "loss": 1.2033, + "step": 32434 + }, + { + "epoch": 0.78, + "learning_rate": 2.4324046568251415e-06, + "loss": 1.0393, + "step": 32435 + }, + { + "epoch": 0.78, + "learning_rate": 2.431895521687789e-06, + "loss": 0.8101, + "step": 32436 + }, + { + "epoch": 0.78, + "learning_rate": 2.431386432464351e-06, + "loss": 0.8877, + "step": 32437 + }, + { + "epoch": 0.78, + "learning_rate": 2.430877389157907e-06, + "loss": 0.891, + "step": 32438 + }, + { + "epoch": 0.78, + "learning_rate": 2.4303683917715503e-06, + "loss": 1.0271, + "step": 32439 + }, + { + "epoch": 0.78, + "learning_rate": 2.4298594403083696e-06, + "loss": 1.2126, + "step": 32440 + }, + { + "epoch": 0.78, + "learning_rate": 2.4293505347714496e-06, + "loss": 1.1041, + "step": 32441 + }, + { + "epoch": 0.78, + "learning_rate": 2.428841675163879e-06, + "loss": 0.9015, + "step": 32442 + }, + { + "epoch": 0.78, + "learning_rate": 2.428332861488747e-06, + "loss": 1.1576, + "step": 32443 + }, + { + "epoch": 0.78, + "learning_rate": 2.427824093749137e-06, + "loss": 1.1727, + "step": 32444 + }, + { + "epoch": 0.78, + "learning_rate": 2.427315371948139e-06, + "loss": 1.0233, + "step": 32445 + }, + { + "epoch": 0.78, + "learning_rate": 2.4268066960888346e-06, + "loss": 1.0452, + "step": 32446 + }, + { + "epoch": 0.78, + "learning_rate": 2.426298066174313e-06, + "loss": 1.1183, + "step": 32447 + }, + { + "epoch": 0.78, + "learning_rate": 2.425789482207663e-06, + "loss": 1.1594, + "step": 32448 + }, + { + "epoch": 0.78, + "learning_rate": 2.4252809441919633e-06, + "loss": 1.3435, + "step": 32449 + }, + { + "epoch": 0.78, + "learning_rate": 2.4247724521303017e-06, + "loss": 1.1837, + "step": 32450 + }, + { + "epoch": 0.78, + "learning_rate": 2.4242640060257683e-06, + "loss": 1.1064, + "step": 32451 + }, + { + "epoch": 0.78, + "learning_rate": 2.423755605881439e-06, + "loss": 1.1491, + "step": 32452 + }, + { + "epoch": 0.78, + "learning_rate": 2.423247251700406e-06, + "loss": 1.0912, + "step": 32453 + }, + { + "epoch": 0.78, + "learning_rate": 2.4227389434857463e-06, + "loss": 1.1008, + "step": 32454 + }, + { + "epoch": 0.78, + "learning_rate": 2.4222306812405473e-06, + "loss": 1.2455, + "step": 32455 + }, + { + "epoch": 0.78, + "learning_rate": 2.421722464967895e-06, + "loss": 1.2699, + "step": 32456 + }, + { + "epoch": 0.78, + "learning_rate": 2.4212142946708685e-06, + "loss": 1.1915, + "step": 32457 + }, + { + "epoch": 0.78, + "learning_rate": 2.4207061703525505e-06, + "loss": 0.713, + "step": 32458 + }, + { + "epoch": 0.78, + "learning_rate": 2.4201980920160297e-06, + "loss": 1.1004, + "step": 32459 + }, + { + "epoch": 0.78, + "learning_rate": 2.4196900596643804e-06, + "loss": 1.0542, + "step": 32460 + }, + { + "epoch": 0.78, + "learning_rate": 2.419182073300691e-06, + "loss": 1.1242, + "step": 32461 + }, + { + "epoch": 0.78, + "learning_rate": 2.4186741329280394e-06, + "loss": 0.9309, + "step": 32462 + }, + { + "epoch": 0.78, + "learning_rate": 2.4181662385495077e-06, + "loss": 0.7736, + "step": 32463 + }, + { + "epoch": 0.78, + "learning_rate": 2.4176583901681803e-06, + "loss": 1.0641, + "step": 32464 + }, + { + "epoch": 0.78, + "learning_rate": 2.4171505877871337e-06, + "loss": 0.689, + "step": 32465 + }, + { + "epoch": 0.78, + "learning_rate": 2.4166428314094514e-06, + "loss": 1.1332, + "step": 32466 + }, + { + "epoch": 0.78, + "learning_rate": 2.4161351210382154e-06, + "loss": 1.3081, + "step": 32467 + }, + { + "epoch": 0.78, + "learning_rate": 2.4156274566765014e-06, + "loss": 1.1395, + "step": 32468 + }, + { + "epoch": 0.78, + "learning_rate": 2.4151198383273924e-06, + "loss": 0.9714, + "step": 32469 + }, + { + "epoch": 0.78, + "learning_rate": 2.4146122659939687e-06, + "loss": 1.2055, + "step": 32470 + }, + { + "epoch": 0.78, + "learning_rate": 2.414104739679306e-06, + "loss": 1.0663, + "step": 32471 + }, + { + "epoch": 0.78, + "learning_rate": 2.413597259386489e-06, + "loss": 1.1511, + "step": 32472 + }, + { + "epoch": 0.78, + "learning_rate": 2.4130898251185898e-06, + "loss": 1.2907, + "step": 32473 + }, + { + "epoch": 0.78, + "learning_rate": 2.412582436878689e-06, + "loss": 1.0728, + "step": 32474 + }, + { + "epoch": 0.78, + "learning_rate": 2.4120750946698703e-06, + "loss": 0.9419, + "step": 32475 + }, + { + "epoch": 0.78, + "learning_rate": 2.4115677984952023e-06, + "loss": 0.978, + "step": 32476 + }, + { + "epoch": 0.78, + "learning_rate": 2.41106054835777e-06, + "loss": 1.3173, + "step": 32477 + }, + { + "epoch": 0.78, + "learning_rate": 2.41055334426065e-06, + "loss": 0.9541, + "step": 32478 + }, + { + "epoch": 0.78, + "learning_rate": 2.4100461862069147e-06, + "loss": 1.0559, + "step": 32479 + }, + { + "epoch": 0.78, + "learning_rate": 2.409539074199646e-06, + "loss": 0.8411, + "step": 32480 + }, + { + "epoch": 0.78, + "learning_rate": 2.409032008241917e-06, + "loss": 0.959, + "step": 32481 + }, + { + "epoch": 0.78, + "learning_rate": 2.4085249883368043e-06, + "loss": 1.0332, + "step": 32482 + }, + { + "epoch": 0.78, + "learning_rate": 2.4080180144873887e-06, + "loss": 0.986, + "step": 32483 + }, + { + "epoch": 0.78, + "learning_rate": 2.407511086696739e-06, + "loss": 0.8221, + "step": 32484 + }, + { + "epoch": 0.78, + "learning_rate": 2.4070042049679333e-06, + "loss": 0.9781, + "step": 32485 + }, + { + "epoch": 0.78, + "learning_rate": 2.4064973693040495e-06, + "loss": 1.2175, + "step": 32486 + }, + { + "epoch": 0.78, + "learning_rate": 2.4059905797081582e-06, + "loss": 0.9291, + "step": 32487 + }, + { + "epoch": 0.78, + "learning_rate": 2.4054838361833376e-06, + "loss": 1.0372, + "step": 32488 + }, + { + "epoch": 0.78, + "learning_rate": 2.4049771387326582e-06, + "loss": 1.1527, + "step": 32489 + }, + { + "epoch": 0.78, + "learning_rate": 2.4044704873591964e-06, + "loss": 0.9656, + "step": 32490 + }, + { + "epoch": 0.78, + "learning_rate": 2.4039638820660273e-06, + "loss": 0.8518, + "step": 32491 + }, + { + "epoch": 0.78, + "learning_rate": 2.4034573228562206e-06, + "loss": 0.8427, + "step": 32492 + }, + { + "epoch": 0.78, + "learning_rate": 2.4029508097328515e-06, + "loss": 1.0197, + "step": 32493 + }, + { + "epoch": 0.78, + "learning_rate": 2.402444342698995e-06, + "loss": 1.0786, + "step": 32494 + }, + { + "epoch": 0.78, + "learning_rate": 2.401937921757719e-06, + "loss": 0.9576, + "step": 32495 + }, + { + "epoch": 0.78, + "learning_rate": 2.401431546912102e-06, + "loss": 0.9577, + "step": 32496 + }, + { + "epoch": 0.78, + "learning_rate": 2.4009252181652087e-06, + "loss": 0.7475, + "step": 32497 + }, + { + "epoch": 0.78, + "learning_rate": 2.4004189355201156e-06, + "loss": 0.7646, + "step": 32498 + }, + { + "epoch": 0.78, + "learning_rate": 2.399912698979893e-06, + "loss": 1.2083, + "step": 32499 + }, + { + "epoch": 0.78, + "learning_rate": 2.3994065085476125e-06, + "loss": 1.1594, + "step": 32500 + }, + { + "epoch": 0.78, + "learning_rate": 2.3989003642263476e-06, + "loss": 1.1483, + "step": 32501 + }, + { + "epoch": 0.78, + "learning_rate": 2.3983942660191628e-06, + "loss": 1.2355, + "step": 32502 + }, + { + "epoch": 0.78, + "learning_rate": 2.3978882139291327e-06, + "loss": 1.2126, + "step": 32503 + }, + { + "epoch": 0.78, + "learning_rate": 2.39738220795933e-06, + "loss": 1.0205, + "step": 32504 + }, + { + "epoch": 0.78, + "learning_rate": 2.3968762481128173e-06, + "loss": 1.0758, + "step": 32505 + }, + { + "epoch": 0.78, + "learning_rate": 2.396370334392669e-06, + "loss": 0.9843, + "step": 32506 + }, + { + "epoch": 0.78, + "learning_rate": 2.395864466801956e-06, + "loss": 0.7882, + "step": 32507 + }, + { + "epoch": 0.78, + "learning_rate": 2.395358645343742e-06, + "loss": 1.1178, + "step": 32508 + }, + { + "epoch": 0.78, + "learning_rate": 2.3948528700211005e-06, + "loss": 0.8994, + "step": 32509 + }, + { + "epoch": 0.78, + "learning_rate": 2.3943471408370943e-06, + "loss": 0.9494, + "step": 32510 + }, + { + "epoch": 0.78, + "learning_rate": 2.3938414577947966e-06, + "loss": 1.1796, + "step": 32511 + }, + { + "epoch": 0.78, + "learning_rate": 2.3933358208972755e-06, + "loss": 1.2457, + "step": 32512 + }, + { + "epoch": 0.78, + "learning_rate": 2.3928302301475946e-06, + "loss": 1.0334, + "step": 32513 + }, + { + "epoch": 0.78, + "learning_rate": 2.3923246855488236e-06, + "loss": 1.147, + "step": 32514 + }, + { + "epoch": 0.78, + "learning_rate": 2.391819187104032e-06, + "loss": 0.9441, + "step": 32515 + }, + { + "epoch": 0.78, + "learning_rate": 2.3913137348162805e-06, + "loss": 0.788, + "step": 32516 + }, + { + "epoch": 0.78, + "learning_rate": 2.3908083286886418e-06, + "loss": 0.9253, + "step": 32517 + }, + { + "epoch": 0.78, + "learning_rate": 2.3903029687241763e-06, + "loss": 0.8789, + "step": 32518 + }, + { + "epoch": 0.78, + "learning_rate": 2.389797654925953e-06, + "loss": 0.971, + "step": 32519 + }, + { + "epoch": 0.78, + "learning_rate": 2.3892923872970406e-06, + "loss": 1.1542, + "step": 32520 + }, + { + "epoch": 0.78, + "learning_rate": 2.388787165840497e-06, + "loss": 0.8013, + "step": 32521 + }, + { + "epoch": 0.78, + "learning_rate": 2.388281990559392e-06, + "loss": 1.1586, + "step": 32522 + }, + { + "epoch": 0.78, + "learning_rate": 2.3877768614567943e-06, + "loss": 1.1771, + "step": 32523 + }, + { + "epoch": 0.78, + "learning_rate": 2.3872717785357593e-06, + "loss": 1.0491, + "step": 32524 + }, + { + "epoch": 0.78, + "learning_rate": 2.386766741799359e-06, + "loss": 1.088, + "step": 32525 + }, + { + "epoch": 0.78, + "learning_rate": 2.3862617512506504e-06, + "loss": 0.9891, + "step": 32526 + }, + { + "epoch": 0.78, + "learning_rate": 2.385756806892702e-06, + "loss": 1.1487, + "step": 32527 + }, + { + "epoch": 0.78, + "learning_rate": 2.3852519087285787e-06, + "loss": 1.0934, + "step": 32528 + }, + { + "epoch": 0.78, + "learning_rate": 2.3847470567613384e-06, + "loss": 1.067, + "step": 32529 + }, + { + "epoch": 0.78, + "learning_rate": 2.3842422509940454e-06, + "loss": 1.1862, + "step": 32530 + }, + { + "epoch": 0.78, + "learning_rate": 2.3837374914297675e-06, + "loss": 0.9655, + "step": 32531 + }, + { + "epoch": 0.78, + "learning_rate": 2.383232778071559e-06, + "loss": 0.88, + "step": 32532 + }, + { + "epoch": 0.78, + "learning_rate": 2.3827281109224885e-06, + "loss": 0.7703, + "step": 32533 + }, + { + "epoch": 0.78, + "learning_rate": 2.3822234899856133e-06, + "loss": 1.0838, + "step": 32534 + }, + { + "epoch": 0.78, + "learning_rate": 2.3817189152639953e-06, + "loss": 1.1467, + "step": 32535 + }, + { + "epoch": 0.78, + "learning_rate": 2.3812143867606996e-06, + "loss": 0.9228, + "step": 32536 + }, + { + "epoch": 0.78, + "learning_rate": 2.380709904478782e-06, + "loss": 0.8502, + "step": 32537 + }, + { + "epoch": 0.78, + "learning_rate": 2.380205468421304e-06, + "loss": 0.9361, + "step": 32538 + }, + { + "epoch": 0.78, + "learning_rate": 2.3797010785913312e-06, + "loss": 0.6388, + "step": 32539 + }, + { + "epoch": 0.78, + "learning_rate": 2.3791967349919152e-06, + "loss": 1.2538, + "step": 32540 + }, + { + "epoch": 0.78, + "learning_rate": 2.3786924376261244e-06, + "loss": 1.0897, + "step": 32541 + }, + { + "epoch": 0.78, + "learning_rate": 2.3781881864970102e-06, + "loss": 0.92, + "step": 32542 + }, + { + "epoch": 0.78, + "learning_rate": 2.377683981607636e-06, + "loss": 1.109, + "step": 32543 + }, + { + "epoch": 0.78, + "learning_rate": 2.3771798229610632e-06, + "loss": 0.8779, + "step": 32544 + }, + { + "epoch": 0.78, + "learning_rate": 2.3766757105603443e-06, + "loss": 0.9051, + "step": 32545 + }, + { + "epoch": 0.78, + "learning_rate": 2.3761716444085404e-06, + "loss": 1.1059, + "step": 32546 + }, + { + "epoch": 0.78, + "learning_rate": 2.375667624508714e-06, + "loss": 1.0118, + "step": 32547 + }, + { + "epoch": 0.78, + "learning_rate": 2.375163650863914e-06, + "loss": 0.898, + "step": 32548 + }, + { + "epoch": 0.78, + "learning_rate": 2.374659723477204e-06, + "loss": 1.09, + "step": 32549 + }, + { + "epoch": 0.78, + "learning_rate": 2.374155842351643e-06, + "loss": 1.2354, + "step": 32550 + }, + { + "epoch": 0.78, + "learning_rate": 2.373652007490281e-06, + "loss": 1.0272, + "step": 32551 + }, + { + "epoch": 0.78, + "learning_rate": 2.373148218896182e-06, + "loss": 0.9531, + "step": 32552 + }, + { + "epoch": 0.78, + "learning_rate": 2.3726444765723956e-06, + "loss": 0.9516, + "step": 32553 + }, + { + "epoch": 0.78, + "learning_rate": 2.372140780521981e-06, + "loss": 1.0684, + "step": 32554 + }, + { + "epoch": 0.78, + "learning_rate": 2.3716371307479967e-06, + "loss": 1.3624, + "step": 32555 + }, + { + "epoch": 0.78, + "learning_rate": 2.371133527253493e-06, + "loss": 0.864, + "step": 32556 + }, + { + "epoch": 0.78, + "learning_rate": 2.3706299700415282e-06, + "loss": 0.7906, + "step": 32557 + }, + { + "epoch": 0.78, + "learning_rate": 2.3701264591151596e-06, + "loss": 1.0392, + "step": 32558 + }, + { + "epoch": 0.78, + "learning_rate": 2.3696229944774364e-06, + "loss": 0.8102, + "step": 32559 + }, + { + "epoch": 0.78, + "learning_rate": 2.3691195761314178e-06, + "loss": 1.001, + "step": 32560 + }, + { + "epoch": 0.78, + "learning_rate": 2.368616204080153e-06, + "loss": 1.1986, + "step": 32561 + }, + { + "epoch": 0.78, + "learning_rate": 2.3681128783266995e-06, + "loss": 1.0898, + "step": 32562 + }, + { + "epoch": 0.78, + "learning_rate": 2.367609598874112e-06, + "loss": 1.1079, + "step": 32563 + }, + { + "epoch": 0.78, + "learning_rate": 2.36710636572544e-06, + "loss": 1.3107, + "step": 32564 + }, + { + "epoch": 0.78, + "learning_rate": 2.3666031788837372e-06, + "loss": 0.7872, + "step": 32565 + }, + { + "epoch": 0.78, + "learning_rate": 2.3661000383520616e-06, + "loss": 1.1614, + "step": 32566 + }, + { + "epoch": 0.78, + "learning_rate": 2.3655969441334583e-06, + "loss": 1.024, + "step": 32567 + }, + { + "epoch": 0.78, + "learning_rate": 2.3650938962309845e-06, + "loss": 0.7638, + "step": 32568 + }, + { + "epoch": 0.78, + "learning_rate": 2.364590894647688e-06, + "loss": 1.2966, + "step": 32569 + }, + { + "epoch": 0.78, + "learning_rate": 2.364087939386622e-06, + "loss": 1.0325, + "step": 32570 + }, + { + "epoch": 0.78, + "learning_rate": 2.3635850304508424e-06, + "loss": 0.9021, + "step": 32571 + }, + { + "epoch": 0.78, + "learning_rate": 2.3630821678433925e-06, + "loss": 1.0095, + "step": 32572 + }, + { + "epoch": 0.78, + "learning_rate": 2.3625793515673268e-06, + "loss": 0.7867, + "step": 32573 + }, + { + "epoch": 0.78, + "learning_rate": 2.3620765816256984e-06, + "loss": 0.8973, + "step": 32574 + }, + { + "epoch": 0.78, + "learning_rate": 2.361573858021553e-06, + "loss": 0.9926, + "step": 32575 + }, + { + "epoch": 0.78, + "learning_rate": 2.3610711807579443e-06, + "loss": 1.1476, + "step": 32576 + }, + { + "epoch": 0.78, + "learning_rate": 2.3605685498379172e-06, + "loss": 1.2436, + "step": 32577 + }, + { + "epoch": 0.78, + "learning_rate": 2.3600659652645243e-06, + "loss": 0.7797, + "step": 32578 + }, + { + "epoch": 0.78, + "learning_rate": 2.3595634270408162e-06, + "loss": 0.9719, + "step": 32579 + }, + { + "epoch": 0.78, + "learning_rate": 2.3590609351698367e-06, + "loss": 1.0344, + "step": 32580 + }, + { + "epoch": 0.78, + "learning_rate": 2.3585584896546386e-06, + "loss": 0.9122, + "step": 32581 + }, + { + "epoch": 0.78, + "learning_rate": 2.3580560904982706e-06, + "loss": 1.1764, + "step": 32582 + }, + { + "epoch": 0.78, + "learning_rate": 2.357553737703776e-06, + "loss": 1.1754, + "step": 32583 + }, + { + "epoch": 0.78, + "learning_rate": 2.3570514312742087e-06, + "loss": 0.9734, + "step": 32584 + }, + { + "epoch": 0.78, + "learning_rate": 2.3565491712126086e-06, + "loss": 1.0439, + "step": 32585 + }, + { + "epoch": 0.78, + "learning_rate": 2.356046957522028e-06, + "loss": 1.0904, + "step": 32586 + }, + { + "epoch": 0.78, + "learning_rate": 2.355544790205515e-06, + "loss": 0.6908, + "step": 32587 + }, + { + "epoch": 0.78, + "learning_rate": 2.355042669266111e-06, + "loss": 1.075, + "step": 32588 + }, + { + "epoch": 0.78, + "learning_rate": 2.3545405947068644e-06, + "loss": 0.8477, + "step": 32589 + }, + { + "epoch": 0.78, + "learning_rate": 2.354038566530825e-06, + "loss": 0.9405, + "step": 32590 + }, + { + "epoch": 0.78, + "learning_rate": 2.353536584741032e-06, + "loss": 1.2499, + "step": 32591 + }, + { + "epoch": 0.78, + "learning_rate": 2.3530346493405343e-06, + "loss": 1.1423, + "step": 32592 + }, + { + "epoch": 0.78, + "learning_rate": 2.35253276033238e-06, + "loss": 1.1088, + "step": 32593 + }, + { + "epoch": 0.78, + "learning_rate": 2.3520309177196065e-06, + "loss": 0.8982, + "step": 32594 + }, + { + "epoch": 0.78, + "learning_rate": 2.351529121505266e-06, + "loss": 1.0691, + "step": 32595 + }, + { + "epoch": 0.78, + "learning_rate": 2.3510273716923972e-06, + "loss": 1.1589, + "step": 32596 + }, + { + "epoch": 0.78, + "learning_rate": 2.350525668284045e-06, + "loss": 1.042, + "step": 32597 + }, + { + "epoch": 0.78, + "learning_rate": 2.3500240112832575e-06, + "loss": 0.9024, + "step": 32598 + }, + { + "epoch": 0.78, + "learning_rate": 2.349522400693073e-06, + "loss": 1.203, + "step": 32599 + }, + { + "epoch": 0.78, + "learning_rate": 2.3490208365165356e-06, + "loss": 1.1726, + "step": 32600 + }, + { + "epoch": 0.78, + "learning_rate": 2.3485193187566923e-06, + "loss": 1.1657, + "step": 32601 + }, + { + "epoch": 0.78, + "learning_rate": 2.3480178474165795e-06, + "loss": 1.1164, + "step": 32602 + }, + { + "epoch": 0.78, + "learning_rate": 2.347516422499245e-06, + "loss": 1.0464, + "step": 32603 + }, + { + "epoch": 0.78, + "learning_rate": 2.3470150440077267e-06, + "loss": 0.9855, + "step": 32604 + }, + { + "epoch": 0.78, + "learning_rate": 2.3465137119450675e-06, + "loss": 0.9944, + "step": 32605 + }, + { + "epoch": 0.78, + "learning_rate": 2.3460124263143124e-06, + "loss": 1.0431, + "step": 32606 + }, + { + "epoch": 0.78, + "learning_rate": 2.3455111871184964e-06, + "loss": 1.0235, + "step": 32607 + }, + { + "epoch": 0.78, + "learning_rate": 2.345009994360663e-06, + "loss": 0.9477, + "step": 32608 + }, + { + "epoch": 0.78, + "learning_rate": 2.3445088480438573e-06, + "loss": 1.1065, + "step": 32609 + }, + { + "epoch": 0.78, + "learning_rate": 2.344007748171113e-06, + "loss": 0.9691, + "step": 32610 + }, + { + "epoch": 0.78, + "learning_rate": 2.3435066947454754e-06, + "loss": 0.9469, + "step": 32611 + }, + { + "epoch": 0.78, + "learning_rate": 2.343005687769978e-06, + "loss": 0.8559, + "step": 32612 + }, + { + "epoch": 0.78, + "learning_rate": 2.342504727247665e-06, + "loss": 0.9587, + "step": 32613 + }, + { + "epoch": 0.78, + "learning_rate": 2.342003813181578e-06, + "loss": 1.0775, + "step": 32614 + }, + { + "epoch": 0.78, + "learning_rate": 2.3415029455747486e-06, + "loss": 0.9303, + "step": 32615 + }, + { + "epoch": 0.78, + "learning_rate": 2.3410021244302205e-06, + "loss": 1.0791, + "step": 32616 + }, + { + "epoch": 0.78, + "learning_rate": 2.340501349751033e-06, + "loss": 1.1487, + "step": 32617 + }, + { + "epoch": 0.78, + "learning_rate": 2.3400006215402195e-06, + "loss": 0.8898, + "step": 32618 + }, + { + "epoch": 0.78, + "learning_rate": 2.339499939800823e-06, + "loss": 1.0534, + "step": 32619 + }, + { + "epoch": 0.78, + "learning_rate": 2.338999304535877e-06, + "loss": 1.0002, + "step": 32620 + }, + { + "epoch": 0.78, + "learning_rate": 2.3384987157484195e-06, + "loss": 0.9883, + "step": 32621 + }, + { + "epoch": 0.78, + "learning_rate": 2.3379981734414913e-06, + "loss": 1.1908, + "step": 32622 + }, + { + "epoch": 0.78, + "learning_rate": 2.3374976776181223e-06, + "loss": 1.2199, + "step": 32623 + }, + { + "epoch": 0.78, + "learning_rate": 2.3369972282813534e-06, + "loss": 1.1734, + "step": 32624 + }, + { + "epoch": 0.78, + "learning_rate": 2.3364968254342223e-06, + "loss": 1.0527, + "step": 32625 + }, + { + "epoch": 0.78, + "learning_rate": 2.3359964690797587e-06, + "loss": 0.9083, + "step": 32626 + }, + { + "epoch": 0.78, + "learning_rate": 2.3354961592210058e-06, + "loss": 1.2128, + "step": 32627 + }, + { + "epoch": 0.78, + "learning_rate": 2.3349958958609918e-06, + "loss": 0.9583, + "step": 32628 + }, + { + "epoch": 0.78, + "learning_rate": 2.334495679002755e-06, + "loss": 1.0798, + "step": 32629 + }, + { + "epoch": 0.78, + "learning_rate": 2.3339955086493327e-06, + "loss": 1.1881, + "step": 32630 + }, + { + "epoch": 0.78, + "learning_rate": 2.3334953848037533e-06, + "loss": 1.0556, + "step": 32631 + }, + { + "epoch": 0.78, + "learning_rate": 2.3329953074690546e-06, + "loss": 1.0828, + "step": 32632 + }, + { + "epoch": 0.78, + "learning_rate": 2.3324952766482723e-06, + "loss": 1.0007, + "step": 32633 + }, + { + "epoch": 0.78, + "learning_rate": 2.3319952923444355e-06, + "loss": 1.0769, + "step": 32634 + }, + { + "epoch": 0.78, + "learning_rate": 2.3314953545605824e-06, + "loss": 0.8376, + "step": 32635 + }, + { + "epoch": 0.78, + "learning_rate": 2.330995463299739e-06, + "loss": 1.1857, + "step": 32636 + }, + { + "epoch": 0.78, + "learning_rate": 2.3304956185649442e-06, + "loss": 1.1901, + "step": 32637 + }, + { + "epoch": 0.78, + "learning_rate": 2.32999582035923e-06, + "loss": 1.0897, + "step": 32638 + }, + { + "epoch": 0.78, + "learning_rate": 2.329496068685625e-06, + "loss": 1.007, + "step": 32639 + }, + { + "epoch": 0.78, + "learning_rate": 2.3289963635471622e-06, + "loss": 0.9268, + "step": 32640 + }, + { + "epoch": 0.78, + "learning_rate": 2.3284967049468775e-06, + "loss": 1.1294, + "step": 32641 + }, + { + "epoch": 0.79, + "learning_rate": 2.327997092887796e-06, + "loss": 1.2579, + "step": 32642 + }, + { + "epoch": 0.79, + "learning_rate": 2.3274975273729518e-06, + "loss": 0.8434, + "step": 32643 + }, + { + "epoch": 0.79, + "learning_rate": 2.326998008405378e-06, + "loss": 1.1003, + "step": 32644 + }, + { + "epoch": 0.79, + "learning_rate": 2.3264985359881e-06, + "loss": 0.8162, + "step": 32645 + }, + { + "epoch": 0.79, + "learning_rate": 2.3259991101241532e-06, + "loss": 1.0076, + "step": 32646 + }, + { + "epoch": 0.79, + "learning_rate": 2.3254997308165615e-06, + "loss": 1.0794, + "step": 32647 + }, + { + "epoch": 0.79, + "learning_rate": 2.325000398068358e-06, + "loss": 1.1054, + "step": 32648 + }, + { + "epoch": 0.79, + "learning_rate": 2.3245011118825743e-06, + "loss": 1.0497, + "step": 32649 + }, + { + "epoch": 0.79, + "learning_rate": 2.3240018722622347e-06, + "loss": 1.1569, + "step": 32650 + }, + { + "epoch": 0.79, + "learning_rate": 2.323502679210371e-06, + "loss": 1.162, + "step": 32651 + }, + { + "epoch": 0.79, + "learning_rate": 2.3230035327300125e-06, + "loss": 1.1451, + "step": 32652 + }, + { + "epoch": 0.79, + "learning_rate": 2.322504432824183e-06, + "loss": 1.1008, + "step": 32653 + }, + { + "epoch": 0.79, + "learning_rate": 2.3220053794959163e-06, + "loss": 1.0778, + "step": 32654 + }, + { + "epoch": 0.79, + "learning_rate": 2.3215063727482346e-06, + "loss": 0.9992, + "step": 32655 + }, + { + "epoch": 0.79, + "learning_rate": 2.3210074125841674e-06, + "loss": 0.8376, + "step": 32656 + }, + { + "epoch": 0.79, + "learning_rate": 2.320508499006745e-06, + "loss": 0.8982, + "step": 32657 + }, + { + "epoch": 0.79, + "learning_rate": 2.320009632018988e-06, + "loss": 1.0664, + "step": 32658 + }, + { + "epoch": 0.79, + "learning_rate": 2.319510811623926e-06, + "loss": 1.0899, + "step": 32659 + }, + { + "epoch": 0.79, + "learning_rate": 2.3190120378245884e-06, + "loss": 0.9156, + "step": 32660 + }, + { + "epoch": 0.79, + "learning_rate": 2.3185133106239953e-06, + "loss": 1.343, + "step": 32661 + }, + { + "epoch": 0.79, + "learning_rate": 2.318014630025178e-06, + "loss": 0.9338, + "step": 32662 + }, + { + "epoch": 0.79, + "learning_rate": 2.3175159960311554e-06, + "loss": 0.8133, + "step": 32663 + }, + { + "epoch": 0.79, + "learning_rate": 2.317017408644957e-06, + "loss": 1.0071, + "step": 32664 + }, + { + "epoch": 0.79, + "learning_rate": 2.3165188678696103e-06, + "loss": 1.2377, + "step": 32665 + }, + { + "epoch": 0.79, + "learning_rate": 2.3160203737081297e-06, + "loss": 1.1088, + "step": 32666 + }, + { + "epoch": 0.79, + "learning_rate": 2.315521926163552e-06, + "loss": 1.113, + "step": 32667 + }, + { + "epoch": 0.79, + "learning_rate": 2.315023525238893e-06, + "loss": 0.9109, + "step": 32668 + }, + { + "epoch": 0.79, + "learning_rate": 2.3145251709371785e-06, + "loss": 0.7913, + "step": 32669 + }, + { + "epoch": 0.79, + "learning_rate": 2.3140268632614337e-06, + "loss": 1.0275, + "step": 32670 + }, + { + "epoch": 0.79, + "learning_rate": 2.3135286022146785e-06, + "loss": 1.1489, + "step": 32671 + }, + { + "epoch": 0.79, + "learning_rate": 2.313030387799937e-06, + "loss": 1.0347, + "step": 32672 + }, + { + "epoch": 0.79, + "learning_rate": 2.312532220020235e-06, + "loss": 0.9512, + "step": 32673 + }, + { + "epoch": 0.79, + "learning_rate": 2.312034098878588e-06, + "loss": 1.0538, + "step": 32674 + }, + { + "epoch": 0.79, + "learning_rate": 2.311536024378026e-06, + "loss": 0.9577, + "step": 32675 + }, + { + "epoch": 0.79, + "learning_rate": 2.3110379965215637e-06, + "loss": 1.1407, + "step": 32676 + }, + { + "epoch": 0.79, + "learning_rate": 2.310540015312225e-06, + "loss": 0.9525, + "step": 32677 + }, + { + "epoch": 0.79, + "learning_rate": 2.310042080753033e-06, + "loss": 1.2777, + "step": 32678 + }, + { + "epoch": 0.79, + "learning_rate": 2.3095441928470053e-06, + "loss": 1.1658, + "step": 32679 + }, + { + "epoch": 0.79, + "learning_rate": 2.309046351597164e-06, + "loss": 1.2307, + "step": 32680 + }, + { + "epoch": 0.79, + "learning_rate": 2.3085485570065313e-06, + "loss": 0.8931, + "step": 32681 + }, + { + "epoch": 0.79, + "learning_rate": 2.308050809078124e-06, + "loss": 1.0303, + "step": 32682 + }, + { + "epoch": 0.79, + "learning_rate": 2.3075531078149647e-06, + "loss": 0.8176, + "step": 32683 + }, + { + "epoch": 0.79, + "learning_rate": 2.3070554532200693e-06, + "loss": 1.257, + "step": 32684 + }, + { + "epoch": 0.79, + "learning_rate": 2.306557845296459e-06, + "loss": 0.8712, + "step": 32685 + }, + { + "epoch": 0.79, + "learning_rate": 2.3060602840471558e-06, + "loss": 0.8348, + "step": 32686 + }, + { + "epoch": 0.79, + "learning_rate": 2.3055627694751714e-06, + "loss": 1.1758, + "step": 32687 + }, + { + "epoch": 0.79, + "learning_rate": 2.3050653015835278e-06, + "loss": 0.8279, + "step": 32688 + }, + { + "epoch": 0.79, + "learning_rate": 2.304567880375247e-06, + "loss": 1.0818, + "step": 32689 + }, + { + "epoch": 0.79, + "learning_rate": 2.3040705058533384e-06, + "loss": 1.0651, + "step": 32690 + }, + { + "epoch": 0.79, + "learning_rate": 2.3035731780208283e-06, + "loss": 1.0316, + "step": 32691 + }, + { + "epoch": 0.79, + "learning_rate": 2.303075896880725e-06, + "loss": 1.0437, + "step": 32692 + }, + { + "epoch": 0.79, + "learning_rate": 2.3025786624360514e-06, + "loss": 0.9471, + "step": 32693 + }, + { + "epoch": 0.79, + "learning_rate": 2.3020814746898248e-06, + "loss": 1.1422, + "step": 32694 + }, + { + "epoch": 0.79, + "learning_rate": 2.301584333645056e-06, + "loss": 0.937, + "step": 32695 + }, + { + "epoch": 0.79, + "learning_rate": 2.3010872393047644e-06, + "loss": 0.9293, + "step": 32696 + }, + { + "epoch": 0.79, + "learning_rate": 2.3005901916719687e-06, + "loss": 0.9425, + "step": 32697 + }, + { + "epoch": 0.79, + "learning_rate": 2.300093190749677e-06, + "loss": 0.9166, + "step": 32698 + }, + { + "epoch": 0.79, + "learning_rate": 2.299596236540913e-06, + "loss": 0.8759, + "step": 32699 + }, + { + "epoch": 0.79, + "learning_rate": 2.299099329048685e-06, + "loss": 1.1786, + "step": 32700 + }, + { + "epoch": 0.79, + "learning_rate": 2.298602468276009e-06, + "loss": 0.9376, + "step": 32701 + }, + { + "epoch": 0.79, + "learning_rate": 2.2981056542259037e-06, + "loss": 1.1172, + "step": 32702 + }, + { + "epoch": 0.79, + "learning_rate": 2.297608886901377e-06, + "loss": 1.1074, + "step": 32703 + }, + { + "epoch": 0.79, + "learning_rate": 2.297112166305445e-06, + "loss": 1.2547, + "step": 32704 + }, + { + "epoch": 0.79, + "learning_rate": 2.2966154924411255e-06, + "loss": 0.8315, + "step": 32705 + }, + { + "epoch": 0.79, + "learning_rate": 2.296118865311424e-06, + "loss": 0.978, + "step": 32706 + }, + { + "epoch": 0.79, + "learning_rate": 2.295622284919361e-06, + "loss": 1.3324, + "step": 32707 + }, + { + "epoch": 0.79, + "learning_rate": 2.2951257512679413e-06, + "loss": 1.2452, + "step": 32708 + }, + { + "epoch": 0.79, + "learning_rate": 2.294629264360182e-06, + "loss": 1.1484, + "step": 32709 + }, + { + "epoch": 0.79, + "learning_rate": 2.294132824199098e-06, + "loss": 1.0032, + "step": 32710 + }, + { + "epoch": 0.79, + "learning_rate": 2.2936364307876946e-06, + "loss": 0.9206, + "step": 32711 + }, + { + "epoch": 0.79, + "learning_rate": 2.293140084128985e-06, + "loss": 1.1254, + "step": 32712 + }, + { + "epoch": 0.79, + "learning_rate": 2.292643784225986e-06, + "loss": 1.13, + "step": 32713 + }, + { + "epoch": 0.79, + "learning_rate": 2.2921475310817008e-06, + "loss": 1.1287, + "step": 32714 + }, + { + "epoch": 0.79, + "learning_rate": 2.2916513246991465e-06, + "loss": 0.8939, + "step": 32715 + }, + { + "epoch": 0.79, + "learning_rate": 2.291155165081328e-06, + "loss": 0.9435, + "step": 32716 + }, + { + "epoch": 0.79, + "learning_rate": 2.290659052231258e-06, + "loss": 1.1029, + "step": 32717 + }, + { + "epoch": 0.79, + "learning_rate": 2.290162986151949e-06, + "loss": 0.8964, + "step": 32718 + }, + { + "epoch": 0.79, + "learning_rate": 2.2896669668464045e-06, + "loss": 1.2106, + "step": 32719 + }, + { + "epoch": 0.79, + "learning_rate": 2.2891709943176366e-06, + "loss": 1.0333, + "step": 32720 + }, + { + "epoch": 0.79, + "learning_rate": 2.2886750685686577e-06, + "loss": 1.2313, + "step": 32721 + }, + { + "epoch": 0.79, + "learning_rate": 2.2881791896024706e-06, + "loss": 0.9881, + "step": 32722 + }, + { + "epoch": 0.79, + "learning_rate": 2.287683357422087e-06, + "loss": 0.9575, + "step": 32723 + }, + { + "epoch": 0.79, + "learning_rate": 2.2871875720305158e-06, + "loss": 1.2582, + "step": 32724 + }, + { + "epoch": 0.79, + "learning_rate": 2.286691833430762e-06, + "loss": 0.947, + "step": 32725 + }, + { + "epoch": 0.79, + "learning_rate": 2.286196141625837e-06, + "loss": 1.0075, + "step": 32726 + }, + { + "epoch": 0.79, + "learning_rate": 2.2857004966187424e-06, + "loss": 0.9984, + "step": 32727 + }, + { + "epoch": 0.79, + "learning_rate": 2.2852048984124885e-06, + "loss": 1.0336, + "step": 32728 + }, + { + "epoch": 0.79, + "learning_rate": 2.2847093470100857e-06, + "loss": 1.0624, + "step": 32729 + }, + { + "epoch": 0.79, + "learning_rate": 2.284213842414532e-06, + "loss": 0.9353, + "step": 32730 + }, + { + "epoch": 0.79, + "learning_rate": 2.283718384628839e-06, + "loss": 1.0964, + "step": 32731 + }, + { + "epoch": 0.79, + "learning_rate": 2.283222973656014e-06, + "loss": 0.7551, + "step": 32732 + }, + { + "epoch": 0.79, + "learning_rate": 2.282727609499058e-06, + "loss": 0.8874, + "step": 32733 + }, + { + "epoch": 0.79, + "learning_rate": 2.282232292160981e-06, + "loss": 1.0224, + "step": 32734 + }, + { + "epoch": 0.79, + "learning_rate": 2.281737021644782e-06, + "loss": 1.0808, + "step": 32735 + }, + { + "epoch": 0.79, + "learning_rate": 2.2812417979534696e-06, + "loss": 1.1656, + "step": 32736 + }, + { + "epoch": 0.79, + "learning_rate": 2.28074662109005e-06, + "loss": 1.012, + "step": 32737 + }, + { + "epoch": 0.79, + "learning_rate": 2.2802514910575223e-06, + "loss": 0.981, + "step": 32738 + }, + { + "epoch": 0.79, + "learning_rate": 2.279756407858893e-06, + "loss": 1.1813, + "step": 32739 + }, + { + "epoch": 0.79, + "learning_rate": 2.279261371497169e-06, + "loss": 1.0611, + "step": 32740 + }, + { + "epoch": 0.79, + "learning_rate": 2.278766381975347e-06, + "loss": 1.1229, + "step": 32741 + }, + { + "epoch": 0.79, + "learning_rate": 2.2782714392964354e-06, + "loss": 1.1204, + "step": 32742 + }, + { + "epoch": 0.79, + "learning_rate": 2.2777765434634326e-06, + "loss": 0.9341, + "step": 32743 + }, + { + "epoch": 0.79, + "learning_rate": 2.277281694479343e-06, + "loss": 1.2354, + "step": 32744 + }, + { + "epoch": 0.79, + "learning_rate": 2.2767868923471724e-06, + "loss": 1.2192, + "step": 32745 + }, + { + "epoch": 0.79, + "learning_rate": 2.2762921370699152e-06, + "loss": 1.0075, + "step": 32746 + }, + { + "epoch": 0.79, + "learning_rate": 2.2757974286505766e-06, + "loss": 0.7705, + "step": 32747 + }, + { + "epoch": 0.79, + "learning_rate": 2.2753027670921623e-06, + "loss": 1.048, + "step": 32748 + }, + { + "epoch": 0.79, + "learning_rate": 2.274808152397666e-06, + "loss": 0.905, + "step": 32749 + }, + { + "epoch": 0.79, + "learning_rate": 2.2743135845700935e-06, + "loss": 0.7316, + "step": 32750 + }, + { + "epoch": 0.79, + "learning_rate": 2.273819063612441e-06, + "loss": 1.1201, + "step": 32751 + }, + { + "epoch": 0.79, + "learning_rate": 2.2733245895277113e-06, + "loss": 1.3067, + "step": 32752 + }, + { + "epoch": 0.79, + "learning_rate": 2.2728301623189062e-06, + "loss": 0.8878, + "step": 32753 + }, + { + "epoch": 0.79, + "learning_rate": 2.2723357819890214e-06, + "loss": 1.1496, + "step": 32754 + }, + { + "epoch": 0.79, + "learning_rate": 2.2718414485410554e-06, + "loss": 0.9422, + "step": 32755 + }, + { + "epoch": 0.79, + "learning_rate": 2.2713471619780148e-06, + "loss": 1.0326, + "step": 32756 + }, + { + "epoch": 0.79, + "learning_rate": 2.2708529223028887e-06, + "loss": 1.1927, + "step": 32757 + }, + { + "epoch": 0.79, + "learning_rate": 2.2703587295186836e-06, + "loss": 0.945, + "step": 32758 + }, + { + "epoch": 0.79, + "learning_rate": 2.2698645836283905e-06, + "loss": 0.9537, + "step": 32759 + }, + { + "epoch": 0.79, + "learning_rate": 2.2693704846350108e-06, + "loss": 1.1387, + "step": 32760 + }, + { + "epoch": 0.79, + "learning_rate": 2.2688764325415446e-06, + "loss": 0.965, + "step": 32761 + }, + { + "epoch": 0.79, + "learning_rate": 2.2683824273509847e-06, + "loss": 0.7224, + "step": 32762 + }, + { + "epoch": 0.79, + "learning_rate": 2.2678884690663294e-06, + "loss": 1.0273, + "step": 32763 + }, + { + "epoch": 0.79, + "learning_rate": 2.267394557690579e-06, + "loss": 1.1391, + "step": 32764 + }, + { + "epoch": 0.79, + "learning_rate": 2.2669006932267246e-06, + "loss": 1.2801, + "step": 32765 + }, + { + "epoch": 0.79, + "learning_rate": 2.2664068756777637e-06, + "loss": 1.0459, + "step": 32766 + }, + { + "epoch": 0.79, + "learning_rate": 2.2659131050466963e-06, + "loss": 1.0543, + "step": 32767 + }, + { + "epoch": 0.79, + "learning_rate": 2.2654193813365133e-06, + "loss": 1.102, + "step": 32768 + }, + { + "epoch": 0.79, + "learning_rate": 2.2649257045502137e-06, + "loss": 1.1239, + "step": 32769 + }, + { + "epoch": 0.79, + "learning_rate": 2.2644320746907867e-06, + "loss": 0.9527, + "step": 32770 + }, + { + "epoch": 0.79, + "learning_rate": 2.263938491761232e-06, + "loss": 1.1513, + "step": 32771 + }, + { + "epoch": 0.79, + "learning_rate": 2.2634449557645454e-06, + "loss": 1.1417, + "step": 32772 + }, + { + "epoch": 0.79, + "learning_rate": 2.2629514667037155e-06, + "loss": 1.0345, + "step": 32773 + }, + { + "epoch": 0.79, + "learning_rate": 2.2624580245817394e-06, + "loss": 0.8461, + "step": 32774 + }, + { + "epoch": 0.79, + "learning_rate": 2.261964629401613e-06, + "loss": 1.2796, + "step": 32775 + }, + { + "epoch": 0.79, + "learning_rate": 2.261471281166324e-06, + "loss": 0.9617, + "step": 32776 + }, + { + "epoch": 0.79, + "learning_rate": 2.2609779798788724e-06, + "loss": 1.1162, + "step": 32777 + }, + { + "epoch": 0.79, + "learning_rate": 2.2604847255422436e-06, + "loss": 1.1911, + "step": 32778 + }, + { + "epoch": 0.79, + "learning_rate": 2.2599915181594335e-06, + "loss": 1.2077, + "step": 32779 + }, + { + "epoch": 0.79, + "learning_rate": 2.2594983577334374e-06, + "loss": 1.1575, + "step": 32780 + }, + { + "epoch": 0.79, + "learning_rate": 2.2590052442672404e-06, + "loss": 1.071, + "step": 32781 + }, + { + "epoch": 0.79, + "learning_rate": 2.2585121777638387e-06, + "loss": 0.9617, + "step": 32782 + }, + { + "epoch": 0.79, + "learning_rate": 2.2580191582262256e-06, + "loss": 1.111, + "step": 32783 + }, + { + "epoch": 0.79, + "learning_rate": 2.257526185657386e-06, + "loss": 0.8291, + "step": 32784 + }, + { + "epoch": 0.79, + "learning_rate": 2.257033260060316e-06, + "loss": 0.957, + "step": 32785 + }, + { + "epoch": 0.79, + "learning_rate": 2.2565403814380026e-06, + "loss": 1.1585, + "step": 32786 + }, + { + "epoch": 0.79, + "learning_rate": 2.2560475497934366e-06, + "loss": 1.1557, + "step": 32787 + }, + { + "epoch": 0.79, + "learning_rate": 2.255554765129612e-06, + "loss": 1.137, + "step": 32788 + }, + { + "epoch": 0.79, + "learning_rate": 2.2550620274495117e-06, + "loss": 0.7917, + "step": 32789 + }, + { + "epoch": 0.79, + "learning_rate": 2.2545693367561282e-06, + "loss": 1.2499, + "step": 32790 + }, + { + "epoch": 0.79, + "learning_rate": 2.254076693052454e-06, + "loss": 0.9458, + "step": 32791 + }, + { + "epoch": 0.79, + "learning_rate": 2.2535840963414714e-06, + "loss": 0.8069, + "step": 32792 + }, + { + "epoch": 0.79, + "learning_rate": 2.2530915466261747e-06, + "loss": 1.1794, + "step": 32793 + }, + { + "epoch": 0.79, + "learning_rate": 2.252599043909547e-06, + "loss": 1.2808, + "step": 32794 + }, + { + "epoch": 0.79, + "learning_rate": 2.252106588194578e-06, + "loss": 0.8623, + "step": 32795 + }, + { + "epoch": 0.79, + "learning_rate": 2.2516141794842594e-06, + "loss": 0.9456, + "step": 32796 + }, + { + "epoch": 0.79, + "learning_rate": 2.251121817781572e-06, + "loss": 1.1719, + "step": 32797 + }, + { + "epoch": 0.79, + "learning_rate": 2.250629503089505e-06, + "loss": 0.9209, + "step": 32798 + }, + { + "epoch": 0.79, + "learning_rate": 2.2501372354110498e-06, + "loss": 1.0599, + "step": 32799 + }, + { + "epoch": 0.79, + "learning_rate": 2.2496450147491867e-06, + "loss": 0.7725, + "step": 32800 + }, + { + "epoch": 0.79, + "learning_rate": 2.2491528411069065e-06, + "loss": 1.0993, + "step": 32801 + }, + { + "epoch": 0.79, + "learning_rate": 2.24866071448719e-06, + "loss": 0.8917, + "step": 32802 + }, + { + "epoch": 0.79, + "learning_rate": 2.2481686348930254e-06, + "loss": 0.8759, + "step": 32803 + }, + { + "epoch": 0.79, + "learning_rate": 2.247676602327402e-06, + "loss": 0.9229, + "step": 32804 + }, + { + "epoch": 0.79, + "learning_rate": 2.2471846167932975e-06, + "loss": 0.8775, + "step": 32805 + }, + { + "epoch": 0.79, + "learning_rate": 2.2466926782937014e-06, + "loss": 0.9058, + "step": 32806 + }, + { + "epoch": 0.79, + "learning_rate": 2.2462007868315995e-06, + "loss": 1.0074, + "step": 32807 + }, + { + "epoch": 0.79, + "learning_rate": 2.245708942409971e-06, + "loss": 1.0014, + "step": 32808 + }, + { + "epoch": 0.79, + "learning_rate": 2.245217145031802e-06, + "loss": 1.1248, + "step": 32809 + }, + { + "epoch": 0.79, + "learning_rate": 2.244725394700079e-06, + "loss": 1.0955, + "step": 32810 + }, + { + "epoch": 0.79, + "learning_rate": 2.2442336914177808e-06, + "loss": 1.0156, + "step": 32811 + }, + { + "epoch": 0.79, + "learning_rate": 2.2437420351878947e-06, + "loss": 1.0552, + "step": 32812 + }, + { + "epoch": 0.79, + "learning_rate": 2.2432504260133994e-06, + "loss": 1.1469, + "step": 32813 + }, + { + "epoch": 0.79, + "learning_rate": 2.2427588638972784e-06, + "loss": 1.0193, + "step": 32814 + }, + { + "epoch": 0.79, + "learning_rate": 2.2422673488425185e-06, + "loss": 0.8754, + "step": 32815 + }, + { + "epoch": 0.79, + "learning_rate": 2.241775880852094e-06, + "loss": 1.1346, + "step": 32816 + }, + { + "epoch": 0.79, + "learning_rate": 2.2412844599289907e-06, + "loss": 1.1551, + "step": 32817 + }, + { + "epoch": 0.79, + "learning_rate": 2.2407930860761926e-06, + "loss": 1.0092, + "step": 32818 + }, + { + "epoch": 0.79, + "learning_rate": 2.2403017592966747e-06, + "loss": 1.1089, + "step": 32819 + }, + { + "epoch": 0.79, + "learning_rate": 2.239810479593424e-06, + "loss": 1.2951, + "step": 32820 + }, + { + "epoch": 0.79, + "learning_rate": 2.2393192469694157e-06, + "loss": 1.1518, + "step": 32821 + }, + { + "epoch": 0.79, + "learning_rate": 2.2388280614276315e-06, + "loss": 0.9627, + "step": 32822 + }, + { + "epoch": 0.79, + "learning_rate": 2.238336922971054e-06, + "loss": 1.0554, + "step": 32823 + }, + { + "epoch": 0.79, + "learning_rate": 2.2378458316026588e-06, + "loss": 1.1406, + "step": 32824 + }, + { + "epoch": 0.79, + "learning_rate": 2.2373547873254266e-06, + "loss": 0.9122, + "step": 32825 + }, + { + "epoch": 0.79, + "learning_rate": 2.2368637901423395e-06, + "loss": 1.1337, + "step": 32826 + }, + { + "epoch": 0.79, + "learning_rate": 2.236372840056372e-06, + "loss": 0.9979, + "step": 32827 + }, + { + "epoch": 0.79, + "learning_rate": 2.2358819370705067e-06, + "loss": 1.0769, + "step": 32828 + }, + { + "epoch": 0.79, + "learning_rate": 2.2353910811877164e-06, + "loss": 1.0468, + "step": 32829 + }, + { + "epoch": 0.79, + "learning_rate": 2.2349002724109826e-06, + "loss": 1.0821, + "step": 32830 + }, + { + "epoch": 0.79, + "learning_rate": 2.234409510743285e-06, + "loss": 1.0214, + "step": 32831 + }, + { + "epoch": 0.79, + "learning_rate": 2.233918796187592e-06, + "loss": 0.8717, + "step": 32832 + }, + { + "epoch": 0.79, + "learning_rate": 2.233428128746894e-06, + "loss": 0.9704, + "step": 32833 + }, + { + "epoch": 0.79, + "learning_rate": 2.2329375084241568e-06, + "loss": 0.9056, + "step": 32834 + }, + { + "epoch": 0.79, + "learning_rate": 2.232446935222361e-06, + "loss": 1.0637, + "step": 32835 + }, + { + "epoch": 0.79, + "learning_rate": 2.2319564091444857e-06, + "loss": 1.1551, + "step": 32836 + }, + { + "epoch": 0.79, + "learning_rate": 2.2314659301935005e-06, + "loss": 1.0842, + "step": 32837 + }, + { + "epoch": 0.79, + "learning_rate": 2.2309754983723876e-06, + "loss": 0.9819, + "step": 32838 + }, + { + "epoch": 0.79, + "learning_rate": 2.2304851136841157e-06, + "loss": 1.0179, + "step": 32839 + }, + { + "epoch": 0.79, + "learning_rate": 2.229994776131663e-06, + "loss": 1.0122, + "step": 32840 + }, + { + "epoch": 0.79, + "learning_rate": 2.2295044857180083e-06, + "loss": 0.8531, + "step": 32841 + }, + { + "epoch": 0.79, + "learning_rate": 2.2290142424461182e-06, + "loss": 0.8994, + "step": 32842 + }, + { + "epoch": 0.79, + "learning_rate": 2.2285240463189707e-06, + "loss": 0.8384, + "step": 32843 + }, + { + "epoch": 0.79, + "learning_rate": 2.228033897339543e-06, + "loss": 1.0045, + "step": 32844 + }, + { + "epoch": 0.79, + "learning_rate": 2.2275437955108024e-06, + "loss": 1.0781, + "step": 32845 + }, + { + "epoch": 0.79, + "learning_rate": 2.227053740835726e-06, + "loss": 0.9659, + "step": 32846 + }, + { + "epoch": 0.79, + "learning_rate": 2.2265637333172897e-06, + "loss": 1.0664, + "step": 32847 + }, + { + "epoch": 0.79, + "learning_rate": 2.226073772958459e-06, + "loss": 0.9899, + "step": 32848 + }, + { + "epoch": 0.79, + "learning_rate": 2.2255838597622127e-06, + "loss": 1.1894, + "step": 32849 + }, + { + "epoch": 0.79, + "learning_rate": 2.2250939937315186e-06, + "loss": 0.8846, + "step": 32850 + }, + { + "epoch": 0.79, + "learning_rate": 2.2246041748693505e-06, + "loss": 1.0795, + "step": 32851 + }, + { + "epoch": 0.79, + "learning_rate": 2.2241144031786823e-06, + "loss": 0.9394, + "step": 32852 + }, + { + "epoch": 0.79, + "learning_rate": 2.2236246786624794e-06, + "loss": 0.9915, + "step": 32853 + }, + { + "epoch": 0.79, + "learning_rate": 2.2231350013237175e-06, + "loss": 1.2053, + "step": 32854 + }, + { + "epoch": 0.79, + "learning_rate": 2.2226453711653684e-06, + "loss": 0.9434, + "step": 32855 + }, + { + "epoch": 0.79, + "learning_rate": 2.222155788190399e-06, + "loss": 1.2076, + "step": 32856 + }, + { + "epoch": 0.79, + "learning_rate": 2.221666252401783e-06, + "loss": 1.1176, + "step": 32857 + }, + { + "epoch": 0.79, + "learning_rate": 2.221176763802485e-06, + "loss": 1.0697, + "step": 32858 + }, + { + "epoch": 0.79, + "learning_rate": 2.2206873223954786e-06, + "loss": 1.0416, + "step": 32859 + }, + { + "epoch": 0.79, + "learning_rate": 2.220197928183735e-06, + "loss": 0.9624, + "step": 32860 + }, + { + "epoch": 0.79, + "learning_rate": 2.219708581170217e-06, + "loss": 0.8888, + "step": 32861 + }, + { + "epoch": 0.79, + "learning_rate": 2.2192192813578983e-06, + "loss": 1.2023, + "step": 32862 + }, + { + "epoch": 0.79, + "learning_rate": 2.2187300287497484e-06, + "loss": 1.1216, + "step": 32863 + }, + { + "epoch": 0.79, + "learning_rate": 2.2182408233487296e-06, + "loss": 1.0994, + "step": 32864 + }, + { + "epoch": 0.79, + "learning_rate": 2.2177516651578167e-06, + "loss": 0.8226, + "step": 32865 + }, + { + "epoch": 0.79, + "learning_rate": 2.217262554179972e-06, + "loss": 0.969, + "step": 32866 + }, + { + "epoch": 0.79, + "learning_rate": 2.2167734904181635e-06, + "loss": 0.9736, + "step": 32867 + }, + { + "epoch": 0.79, + "learning_rate": 2.216284473875363e-06, + "loss": 1.1987, + "step": 32868 + }, + { + "epoch": 0.79, + "learning_rate": 2.215795504554531e-06, + "loss": 1.1533, + "step": 32869 + }, + { + "epoch": 0.79, + "learning_rate": 2.2153065824586363e-06, + "loss": 0.8102, + "step": 32870 + }, + { + "epoch": 0.79, + "learning_rate": 2.2148177075906487e-06, + "loss": 1.2238, + "step": 32871 + }, + { + "epoch": 0.79, + "learning_rate": 2.214328879953528e-06, + "loss": 1.0138, + "step": 32872 + }, + { + "epoch": 0.79, + "learning_rate": 2.213840099550245e-06, + "loss": 1.1723, + "step": 32873 + }, + { + "epoch": 0.79, + "learning_rate": 2.2133513663837604e-06, + "loss": 1.1762, + "step": 32874 + }, + { + "epoch": 0.79, + "learning_rate": 2.2128626804570418e-06, + "loss": 0.9121, + "step": 32875 + }, + { + "epoch": 0.79, + "learning_rate": 2.2123740417730556e-06, + "loss": 0.8482, + "step": 32876 + }, + { + "epoch": 0.79, + "learning_rate": 2.211885450334762e-06, + "loss": 0.9186, + "step": 32877 + }, + { + "epoch": 0.79, + "learning_rate": 2.2113969061451278e-06, + "loss": 0.9795, + "step": 32878 + }, + { + "epoch": 0.79, + "learning_rate": 2.210908409207119e-06, + "loss": 0.9572, + "step": 32879 + }, + { + "epoch": 0.79, + "learning_rate": 2.2104199595236943e-06, + "loss": 1.0, + "step": 32880 + }, + { + "epoch": 0.79, + "learning_rate": 2.209931557097822e-06, + "loss": 1.0001, + "step": 32881 + }, + { + "epoch": 0.79, + "learning_rate": 2.2094432019324595e-06, + "loss": 0.8215, + "step": 32882 + }, + { + "epoch": 0.79, + "learning_rate": 2.2089548940305737e-06, + "loss": 1.0344, + "step": 32883 + }, + { + "epoch": 0.79, + "learning_rate": 2.208466633395128e-06, + "loss": 1.0369, + "step": 32884 + }, + { + "epoch": 0.79, + "learning_rate": 2.20797842002908e-06, + "loss": 1.4137, + "step": 32885 + }, + { + "epoch": 0.79, + "learning_rate": 2.207490253935394e-06, + "loss": 0.863, + "step": 32886 + }, + { + "epoch": 0.79, + "learning_rate": 2.2070021351170346e-06, + "loss": 1.1021, + "step": 32887 + }, + { + "epoch": 0.79, + "learning_rate": 2.2065140635769568e-06, + "loss": 0.8984, + "step": 32888 + }, + { + "epoch": 0.79, + "learning_rate": 2.2060260393181255e-06, + "loss": 1.1374, + "step": 32889 + }, + { + "epoch": 0.79, + "learning_rate": 2.205538062343504e-06, + "loss": 1.1864, + "step": 32890 + }, + { + "epoch": 0.79, + "learning_rate": 2.205050132656047e-06, + "loss": 1.0704, + "step": 32891 + }, + { + "epoch": 0.79, + "learning_rate": 2.2045622502587195e-06, + "loss": 0.8806, + "step": 32892 + }, + { + "epoch": 0.79, + "learning_rate": 2.204074415154477e-06, + "loss": 1.0054, + "step": 32893 + }, + { + "epoch": 0.79, + "learning_rate": 2.2035866273462813e-06, + "loss": 1.0938, + "step": 32894 + }, + { + "epoch": 0.79, + "learning_rate": 2.203098886837094e-06, + "loss": 1.0471, + "step": 32895 + }, + { + "epoch": 0.79, + "learning_rate": 2.2026111936298687e-06, + "loss": 0.8128, + "step": 32896 + }, + { + "epoch": 0.79, + "learning_rate": 2.2021235477275683e-06, + "loss": 1.125, + "step": 32897 + }, + { + "epoch": 0.79, + "learning_rate": 2.201635949133153e-06, + "loss": 1.0911, + "step": 32898 + }, + { + "epoch": 0.79, + "learning_rate": 2.201148397849574e-06, + "loss": 0.9684, + "step": 32899 + }, + { + "epoch": 0.79, + "learning_rate": 2.200660893879797e-06, + "loss": 1.1892, + "step": 32900 + }, + { + "epoch": 0.79, + "learning_rate": 2.2001734372267734e-06, + "loss": 1.317, + "step": 32901 + }, + { + "epoch": 0.79, + "learning_rate": 2.1996860278934628e-06, + "loss": 1.0024, + "step": 32902 + }, + { + "epoch": 0.79, + "learning_rate": 2.1991986658828245e-06, + "loss": 1.0822, + "step": 32903 + }, + { + "epoch": 0.79, + "learning_rate": 2.198711351197811e-06, + "loss": 1.0924, + "step": 32904 + }, + { + "epoch": 0.79, + "learning_rate": 2.198224083841379e-06, + "loss": 0.9819, + "step": 32905 + }, + { + "epoch": 0.79, + "learning_rate": 2.1977368638164908e-06, + "loss": 1.2305, + "step": 32906 + }, + { + "epoch": 0.79, + "learning_rate": 2.1972496911260955e-06, + "loss": 0.9142, + "step": 32907 + }, + { + "epoch": 0.79, + "learning_rate": 2.196762565773153e-06, + "loss": 1.143, + "step": 32908 + }, + { + "epoch": 0.79, + "learning_rate": 2.196275487760613e-06, + "loss": 1.0389, + "step": 32909 + }, + { + "epoch": 0.79, + "learning_rate": 2.1957884570914357e-06, + "loss": 0.8454, + "step": 32910 + }, + { + "epoch": 0.79, + "learning_rate": 2.1953014737685762e-06, + "loss": 1.0229, + "step": 32911 + }, + { + "epoch": 0.79, + "learning_rate": 2.1948145377949835e-06, + "loss": 1.051, + "step": 32912 + }, + { + "epoch": 0.79, + "learning_rate": 2.194327649173615e-06, + "loss": 1.1411, + "step": 32913 + }, + { + "epoch": 0.79, + "learning_rate": 2.1938408079074293e-06, + "loss": 1.1905, + "step": 32914 + }, + { + "epoch": 0.79, + "learning_rate": 2.19335401399937e-06, + "loss": 0.9615, + "step": 32915 + }, + { + "epoch": 0.79, + "learning_rate": 2.192867267452401e-06, + "loss": 1.1361, + "step": 32916 + }, + { + "epoch": 0.79, + "learning_rate": 2.1923805682694655e-06, + "loss": 1.2762, + "step": 32917 + }, + { + "epoch": 0.79, + "learning_rate": 2.1918939164535214e-06, + "loss": 0.793, + "step": 32918 + }, + { + "epoch": 0.79, + "learning_rate": 2.191407312007522e-06, + "loss": 0.8407, + "step": 32919 + }, + { + "epoch": 0.79, + "learning_rate": 2.190920754934416e-06, + "loss": 1.007, + "step": 32920 + }, + { + "epoch": 0.79, + "learning_rate": 2.1904342452371564e-06, + "loss": 1.075, + "step": 32921 + }, + { + "epoch": 0.79, + "learning_rate": 2.1899477829186986e-06, + "loss": 0.8847, + "step": 32922 + }, + { + "epoch": 0.79, + "learning_rate": 2.1894613679819874e-06, + "loss": 0.9428, + "step": 32923 + }, + { + "epoch": 0.79, + "learning_rate": 2.1889750004299803e-06, + "loss": 1.1512, + "step": 32924 + }, + { + "epoch": 0.79, + "learning_rate": 2.1884886802656214e-06, + "loss": 1.0192, + "step": 32925 + }, + { + "epoch": 0.79, + "learning_rate": 2.1880024074918636e-06, + "loss": 1.0202, + "step": 32926 + }, + { + "epoch": 0.79, + "learning_rate": 2.187516182111661e-06, + "loss": 0.8749, + "step": 32927 + }, + { + "epoch": 0.79, + "learning_rate": 2.187030004127958e-06, + "loss": 1.0963, + "step": 32928 + }, + { + "epoch": 0.79, + "learning_rate": 2.186543873543706e-06, + "loss": 1.1809, + "step": 32929 + }, + { + "epoch": 0.79, + "learning_rate": 2.1860577903618564e-06, + "loss": 1.1225, + "step": 32930 + }, + { + "epoch": 0.79, + "learning_rate": 2.1855717545853537e-06, + "loss": 1.0181, + "step": 32931 + }, + { + "epoch": 0.79, + "learning_rate": 2.1850857662171533e-06, + "loss": 1.1575, + "step": 32932 + }, + { + "epoch": 0.79, + "learning_rate": 2.1845998252601953e-06, + "loss": 1.2127, + "step": 32933 + }, + { + "epoch": 0.79, + "learning_rate": 2.184113931717432e-06, + "loss": 0.8575, + "step": 32934 + }, + { + "epoch": 0.79, + "learning_rate": 2.1836280855918133e-06, + "loss": 1.07, + "step": 32935 + }, + { + "epoch": 0.79, + "learning_rate": 2.183142286886283e-06, + "loss": 0.9215, + "step": 32936 + }, + { + "epoch": 0.79, + "learning_rate": 2.1826565356037887e-06, + "loss": 1.0922, + "step": 32937 + }, + { + "epoch": 0.79, + "learning_rate": 2.1821708317472823e-06, + "loss": 1.1125, + "step": 32938 + }, + { + "epoch": 0.79, + "learning_rate": 2.1816851753197023e-06, + "loss": 0.9108, + "step": 32939 + }, + { + "epoch": 0.79, + "learning_rate": 2.1811995663240006e-06, + "loss": 1.1549, + "step": 32940 + }, + { + "epoch": 0.79, + "learning_rate": 2.1807140047631247e-06, + "loss": 1.1536, + "step": 32941 + }, + { + "epoch": 0.79, + "learning_rate": 2.180228490640014e-06, + "loss": 1.2645, + "step": 32942 + }, + { + "epoch": 0.79, + "learning_rate": 2.1797430239576224e-06, + "loss": 0.9925, + "step": 32943 + }, + { + "epoch": 0.79, + "learning_rate": 2.179257604718886e-06, + "loss": 1.1271, + "step": 32944 + }, + { + "epoch": 0.79, + "learning_rate": 2.1787722329267547e-06, + "loss": 1.1194, + "step": 32945 + }, + { + "epoch": 0.79, + "learning_rate": 2.1782869085841752e-06, + "loss": 0.8537, + "step": 32946 + }, + { + "epoch": 0.79, + "learning_rate": 2.1778016316940866e-06, + "loss": 1.0417, + "step": 32947 + }, + { + "epoch": 0.79, + "learning_rate": 2.177316402259435e-06, + "loss": 0.8837, + "step": 32948 + }, + { + "epoch": 0.79, + "learning_rate": 2.176831220283169e-06, + "loss": 1.0349, + "step": 32949 + }, + { + "epoch": 0.79, + "learning_rate": 2.1763460857682237e-06, + "loss": 1.104, + "step": 32950 + }, + { + "epoch": 0.79, + "learning_rate": 2.1758609987175495e-06, + "loss": 1.1848, + "step": 32951 + }, + { + "epoch": 0.79, + "learning_rate": 2.175375959134084e-06, + "loss": 1.2141, + "step": 32952 + }, + { + "epoch": 0.79, + "learning_rate": 2.174890967020772e-06, + "loss": 1.0856, + "step": 32953 + }, + { + "epoch": 0.79, + "learning_rate": 2.1744060223805587e-06, + "loss": 1.132, + "step": 32954 + }, + { + "epoch": 0.79, + "learning_rate": 2.1739211252163804e-06, + "loss": 1.0522, + "step": 32955 + }, + { + "epoch": 0.79, + "learning_rate": 2.173436275531182e-06, + "loss": 1.2214, + "step": 32956 + }, + { + "epoch": 0.79, + "learning_rate": 2.1729514733279066e-06, + "loss": 1.1159, + "step": 32957 + }, + { + "epoch": 0.79, + "learning_rate": 2.172466718609493e-06, + "loss": 1.125, + "step": 32958 + }, + { + "epoch": 0.79, + "learning_rate": 2.1719820113788835e-06, + "loss": 1.1623, + "step": 32959 + }, + { + "epoch": 0.79, + "learning_rate": 2.1714973516390157e-06, + "loss": 1.0322, + "step": 32960 + }, + { + "epoch": 0.79, + "learning_rate": 2.1710127393928316e-06, + "loss": 1.129, + "step": 32961 + }, + { + "epoch": 0.79, + "learning_rate": 2.170528174643275e-06, + "loss": 1.2472, + "step": 32962 + }, + { + "epoch": 0.79, + "learning_rate": 2.17004365739328e-06, + "loss": 0.9187, + "step": 32963 + }, + { + "epoch": 0.79, + "learning_rate": 2.169559187645788e-06, + "loss": 1.0402, + "step": 32964 + }, + { + "epoch": 0.79, + "learning_rate": 2.1690747654037416e-06, + "loss": 0.9904, + "step": 32965 + }, + { + "epoch": 0.79, + "learning_rate": 2.1685903906700733e-06, + "loss": 0.9597, + "step": 32966 + }, + { + "epoch": 0.79, + "learning_rate": 2.168106063447728e-06, + "loss": 1.0872, + "step": 32967 + }, + { + "epoch": 0.79, + "learning_rate": 2.1676217837396386e-06, + "loss": 1.1182, + "step": 32968 + }, + { + "epoch": 0.79, + "learning_rate": 2.1671375515487447e-06, + "loss": 0.843, + "step": 32969 + }, + { + "epoch": 0.79, + "learning_rate": 2.1666533668779886e-06, + "loss": 1.1603, + "step": 32970 + }, + { + "epoch": 0.79, + "learning_rate": 2.1661692297303006e-06, + "loss": 1.0331, + "step": 32971 + }, + { + "epoch": 0.79, + "learning_rate": 2.165685140108622e-06, + "loss": 1.0937, + "step": 32972 + }, + { + "epoch": 0.79, + "learning_rate": 2.165201098015891e-06, + "loss": 1.0828, + "step": 32973 + }, + { + "epoch": 0.79, + "learning_rate": 2.1647171034550398e-06, + "loss": 1.2312, + "step": 32974 + }, + { + "epoch": 0.79, + "learning_rate": 2.16423315642901e-06, + "loss": 0.9764, + "step": 32975 + }, + { + "epoch": 0.79, + "learning_rate": 2.1637492569407306e-06, + "loss": 1.0123, + "step": 32976 + }, + { + "epoch": 0.79, + "learning_rate": 2.1632654049931423e-06, + "loss": 0.7018, + "step": 32977 + }, + { + "epoch": 0.79, + "learning_rate": 2.162781600589182e-06, + "loss": 1.1348, + "step": 32978 + }, + { + "epoch": 0.79, + "learning_rate": 2.162297843731779e-06, + "loss": 0.8286, + "step": 32979 + }, + { + "epoch": 0.79, + "learning_rate": 2.161814134423872e-06, + "loss": 1.0287, + "step": 32980 + }, + { + "epoch": 0.79, + "learning_rate": 2.1613304726683983e-06, + "loss": 1.1792, + "step": 32981 + }, + { + "epoch": 0.79, + "learning_rate": 2.160846858468285e-06, + "loss": 0.8497, + "step": 32982 + }, + { + "epoch": 0.79, + "learning_rate": 2.1603632918264704e-06, + "loss": 0.8735, + "step": 32983 + }, + { + "epoch": 0.79, + "learning_rate": 2.1598797727458898e-06, + "loss": 1.1968, + "step": 32984 + }, + { + "epoch": 0.79, + "learning_rate": 2.1593963012294727e-06, + "loss": 0.8466, + "step": 32985 + }, + { + "epoch": 0.79, + "learning_rate": 2.158912877280156e-06, + "loss": 1.1655, + "step": 32986 + }, + { + "epoch": 0.79, + "learning_rate": 2.158429500900868e-06, + "loss": 0.7899, + "step": 32987 + }, + { + "epoch": 0.79, + "learning_rate": 2.157946172094545e-06, + "loss": 1.3831, + "step": 32988 + }, + { + "epoch": 0.79, + "learning_rate": 2.1574628908641192e-06, + "loss": 0.9903, + "step": 32989 + }, + { + "epoch": 0.79, + "learning_rate": 2.156979657212519e-06, + "loss": 1.232, + "step": 32990 + }, + { + "epoch": 0.79, + "learning_rate": 2.1564964711426785e-06, + "loss": 0.81, + "step": 32991 + }, + { + "epoch": 0.79, + "learning_rate": 2.156013332657533e-06, + "loss": 1.0685, + "step": 32992 + }, + { + "epoch": 0.79, + "learning_rate": 2.1555302417600057e-06, + "loss": 1.0573, + "step": 32993 + }, + { + "epoch": 0.79, + "learning_rate": 2.155047198453033e-06, + "loss": 0.9473, + "step": 32994 + }, + { + "epoch": 0.79, + "learning_rate": 2.154564202739542e-06, + "loss": 0.8257, + "step": 32995 + }, + { + "epoch": 0.79, + "learning_rate": 2.1540812546224644e-06, + "loss": 1.1629, + "step": 32996 + }, + { + "epoch": 0.79, + "learning_rate": 2.1535983541047333e-06, + "loss": 0.8681, + "step": 32997 + }, + { + "epoch": 0.79, + "learning_rate": 2.1531155011892733e-06, + "loss": 1.0483, + "step": 32998 + }, + { + "epoch": 0.79, + "learning_rate": 2.1526326958790157e-06, + "loss": 1.035, + "step": 32999 + }, + { + "epoch": 0.79, + "learning_rate": 2.1521499381768886e-06, + "loss": 0.955, + "step": 33000 + }, + { + "epoch": 0.79, + "learning_rate": 2.151667228085822e-06, + "loss": 1.4824, + "step": 33001 + }, + { + "epoch": 0.79, + "learning_rate": 2.1511845656087482e-06, + "loss": 1.3312, + "step": 33002 + }, + { + "epoch": 0.79, + "learning_rate": 2.1507019507485883e-06, + "loss": 1.0031, + "step": 33003 + }, + { + "epoch": 0.79, + "learning_rate": 2.150219383508275e-06, + "loss": 1.1215, + "step": 33004 + }, + { + "epoch": 0.79, + "learning_rate": 2.1497368638907334e-06, + "loss": 0.9141, + "step": 33005 + }, + { + "epoch": 0.79, + "learning_rate": 2.1492543918988906e-06, + "loss": 1.1529, + "step": 33006 + }, + { + "epoch": 0.79, + "learning_rate": 2.1487719675356777e-06, + "loss": 0.6955, + "step": 33007 + }, + { + "epoch": 0.79, + "learning_rate": 2.148289590804016e-06, + "loss": 1.0778, + "step": 33008 + }, + { + "epoch": 0.79, + "learning_rate": 2.1478072617068346e-06, + "loss": 1.0703, + "step": 33009 + }, + { + "epoch": 0.79, + "learning_rate": 2.147324980247062e-06, + "loss": 1.0649, + "step": 33010 + }, + { + "epoch": 0.79, + "learning_rate": 2.146842746427619e-06, + "loss": 0.9372, + "step": 33011 + }, + { + "epoch": 0.79, + "learning_rate": 2.146360560251437e-06, + "loss": 0.9154, + "step": 33012 + }, + { + "epoch": 0.79, + "learning_rate": 2.1458784217214358e-06, + "loss": 1.0923, + "step": 33013 + }, + { + "epoch": 0.79, + "learning_rate": 2.145396330840541e-06, + "loss": 0.8549, + "step": 33014 + }, + { + "epoch": 0.79, + "learning_rate": 2.1449142876116834e-06, + "loss": 1.1323, + "step": 33015 + }, + { + "epoch": 0.79, + "learning_rate": 2.14443229203778e-06, + "loss": 1.2562, + "step": 33016 + }, + { + "epoch": 0.79, + "learning_rate": 2.143950344121759e-06, + "loss": 1.0589, + "step": 33017 + }, + { + "epoch": 0.79, + "learning_rate": 2.1434684438665455e-06, + "loss": 0.9221, + "step": 33018 + }, + { + "epoch": 0.79, + "learning_rate": 2.142986591275058e-06, + "loss": 0.9373, + "step": 33019 + }, + { + "epoch": 0.79, + "learning_rate": 2.1425047863502223e-06, + "loss": 0.9646, + "step": 33020 + }, + { + "epoch": 0.79, + "learning_rate": 2.1420230290949664e-06, + "loss": 0.9527, + "step": 33021 + }, + { + "epoch": 0.79, + "learning_rate": 2.1415413195122036e-06, + "loss": 1.2638, + "step": 33022 + }, + { + "epoch": 0.79, + "learning_rate": 2.1410596576048648e-06, + "loss": 1.1991, + "step": 33023 + }, + { + "epoch": 0.79, + "learning_rate": 2.140578043375866e-06, + "loss": 1.1105, + "step": 33024 + }, + { + "epoch": 0.79, + "learning_rate": 2.140096476828132e-06, + "loss": 0.9503, + "step": 33025 + }, + { + "epoch": 0.79, + "learning_rate": 2.139614957964585e-06, + "loss": 1.1121, + "step": 33026 + }, + { + "epoch": 0.79, + "learning_rate": 2.139133486788144e-06, + "loss": 1.1124, + "step": 33027 + }, + { + "epoch": 0.79, + "learning_rate": 2.1386520633017303e-06, + "loss": 0.8254, + "step": 33028 + }, + { + "epoch": 0.79, + "learning_rate": 2.1381706875082695e-06, + "loss": 1.0701, + "step": 33029 + }, + { + "epoch": 0.79, + "learning_rate": 2.137689359410674e-06, + "loss": 0.7393, + "step": 33030 + }, + { + "epoch": 0.79, + "learning_rate": 2.137208079011871e-06, + "loss": 0.7832, + "step": 33031 + }, + { + "epoch": 0.79, + "learning_rate": 2.136726846314774e-06, + "loss": 1.0511, + "step": 33032 + }, + { + "epoch": 0.79, + "learning_rate": 2.136245661322306e-06, + "loss": 1.2581, + "step": 33033 + }, + { + "epoch": 0.79, + "learning_rate": 2.135764524037388e-06, + "loss": 0.925, + "step": 33034 + }, + { + "epoch": 0.79, + "learning_rate": 2.135283434462935e-06, + "loss": 0.9707, + "step": 33035 + }, + { + "epoch": 0.79, + "learning_rate": 2.134802392601867e-06, + "loss": 0.9872, + "step": 33036 + }, + { + "epoch": 0.79, + "learning_rate": 2.1343213984571064e-06, + "loss": 0.9851, + "step": 33037 + }, + { + "epoch": 0.79, + "learning_rate": 2.133840452031564e-06, + "loss": 1.0468, + "step": 33038 + }, + { + "epoch": 0.79, + "learning_rate": 2.1333595533281646e-06, + "loss": 0.9624, + "step": 33039 + }, + { + "epoch": 0.79, + "learning_rate": 2.13287870234982e-06, + "loss": 1.0559, + "step": 33040 + }, + { + "epoch": 0.79, + "learning_rate": 2.1323978990994487e-06, + "loss": 1.1654, + "step": 33041 + }, + { + "epoch": 0.79, + "learning_rate": 2.1319171435799723e-06, + "loss": 0.9643, + "step": 33042 + }, + { + "epoch": 0.79, + "learning_rate": 2.1314364357943007e-06, + "loss": 1.1645, + "step": 33043 + }, + { + "epoch": 0.79, + "learning_rate": 2.130955775745355e-06, + "loss": 0.8657, + "step": 33044 + }, + { + "epoch": 0.79, + "learning_rate": 2.130475163436051e-06, + "loss": 1.3466, + "step": 33045 + }, + { + "epoch": 0.79, + "learning_rate": 2.1299945988693006e-06, + "loss": 1.1748, + "step": 33046 + }, + { + "epoch": 0.79, + "learning_rate": 2.1295140820480244e-06, + "loss": 1.1594, + "step": 33047 + }, + { + "epoch": 0.79, + "learning_rate": 2.129033612975132e-06, + "loss": 1.0244, + "step": 33048 + }, + { + "epoch": 0.79, + "learning_rate": 2.128553191653542e-06, + "loss": 0.9263, + "step": 33049 + }, + { + "epoch": 0.79, + "learning_rate": 2.128072818086171e-06, + "loss": 0.9575, + "step": 33050 + }, + { + "epoch": 0.79, + "learning_rate": 2.1275924922759282e-06, + "loss": 0.8177, + "step": 33051 + }, + { + "epoch": 0.79, + "learning_rate": 2.1271122142257296e-06, + "loss": 0.9541, + "step": 33052 + }, + { + "epoch": 0.79, + "learning_rate": 2.126631983938492e-06, + "loss": 1.1327, + "step": 33053 + }, + { + "epoch": 0.79, + "learning_rate": 2.1261518014171235e-06, + "loss": 0.9941, + "step": 33054 + }, + { + "epoch": 0.79, + "learning_rate": 2.125671666664543e-06, + "loss": 1.1953, + "step": 33055 + }, + { + "epoch": 0.79, + "learning_rate": 2.1251915796836577e-06, + "loss": 0.9499, + "step": 33056 + }, + { + "epoch": 0.79, + "learning_rate": 2.124711540477382e-06, + "loss": 1.0238, + "step": 33057 + }, + { + "epoch": 0.8, + "learning_rate": 2.124231549048632e-06, + "loss": 0.6401, + "step": 33058 + }, + { + "epoch": 0.8, + "learning_rate": 2.123751605400315e-06, + "loss": 0.9774, + "step": 33059 + }, + { + "epoch": 0.8, + "learning_rate": 2.1232717095353437e-06, + "loss": 1.1535, + "step": 33060 + }, + { + "epoch": 0.8, + "learning_rate": 2.1227918614566323e-06, + "loss": 0.9429, + "step": 33061 + }, + { + "epoch": 0.8, + "learning_rate": 2.122312061167088e-06, + "loss": 0.9323, + "step": 33062 + }, + { + "epoch": 0.8, + "learning_rate": 2.1218323086696236e-06, + "loss": 0.8588, + "step": 33063 + }, + { + "epoch": 0.8, + "learning_rate": 2.121352603967152e-06, + "loss": 1.1794, + "step": 33064 + }, + { + "epoch": 0.8, + "learning_rate": 2.1208729470625776e-06, + "loss": 1.0998, + "step": 33065 + }, + { + "epoch": 0.8, + "learning_rate": 2.120393337958817e-06, + "loss": 1.1645, + "step": 33066 + }, + { + "epoch": 0.8, + "learning_rate": 2.119913776658774e-06, + "loss": 1.0223, + "step": 33067 + }, + { + "epoch": 0.8, + "learning_rate": 2.119434263165361e-06, + "loss": 1.1216, + "step": 33068 + }, + { + "epoch": 0.8, + "learning_rate": 2.1189547974814885e-06, + "loss": 1.3163, + "step": 33069 + }, + { + "epoch": 0.8, + "learning_rate": 2.118475379610061e-06, + "loss": 1.04, + "step": 33070 + }, + { + "epoch": 0.8, + "learning_rate": 2.1179960095539897e-06, + "loss": 1.0952, + "step": 33071 + }, + { + "epoch": 0.8, + "learning_rate": 2.1175166873161857e-06, + "loss": 1.1316, + "step": 33072 + }, + { + "epoch": 0.8, + "learning_rate": 2.1170374128995507e-06, + "loss": 0.9457, + "step": 33073 + }, + { + "epoch": 0.8, + "learning_rate": 2.1165581863069985e-06, + "loss": 0.9063, + "step": 33074 + }, + { + "epoch": 0.8, + "learning_rate": 2.116079007541431e-06, + "loss": 1.1909, + "step": 33075 + }, + { + "epoch": 0.8, + "learning_rate": 2.1155998766057574e-06, + "loss": 1.0219, + "step": 33076 + }, + { + "epoch": 0.8, + "learning_rate": 2.115120793502887e-06, + "loss": 1.2445, + "step": 33077 + }, + { + "epoch": 0.8, + "learning_rate": 2.1146417582357226e-06, + "loss": 1.2059, + "step": 33078 + }, + { + "epoch": 0.8, + "learning_rate": 2.114162770807171e-06, + "loss": 1.1471, + "step": 33079 + }, + { + "epoch": 0.8, + "learning_rate": 2.1136838312201423e-06, + "loss": 0.936, + "step": 33080 + }, + { + "epoch": 0.8, + "learning_rate": 2.1132049394775358e-06, + "loss": 1.0706, + "step": 33081 + }, + { + "epoch": 0.8, + "learning_rate": 2.1127260955822627e-06, + "loss": 1.111, + "step": 33082 + }, + { + "epoch": 0.8, + "learning_rate": 2.112247299537222e-06, + "loss": 0.989, + "step": 33083 + }, + { + "epoch": 0.8, + "learning_rate": 2.111768551345321e-06, + "loss": 1.1621, + "step": 33084 + }, + { + "epoch": 0.8, + "learning_rate": 2.111289851009469e-06, + "loss": 1.2806, + "step": 33085 + }, + { + "epoch": 0.8, + "learning_rate": 2.110811198532563e-06, + "loss": 1.286, + "step": 33086 + }, + { + "epoch": 0.8, + "learning_rate": 2.110332593917509e-06, + "loss": 1.0749, + "step": 33087 + }, + { + "epoch": 0.8, + "learning_rate": 2.109854037167215e-06, + "loss": 0.8639, + "step": 33088 + }, + { + "epoch": 0.8, + "learning_rate": 2.1093755282845775e-06, + "loss": 1.1838, + "step": 33089 + }, + { + "epoch": 0.8, + "learning_rate": 2.108897067272505e-06, + "loss": 1.2956, + "step": 33090 + }, + { + "epoch": 0.8, + "learning_rate": 2.1084186541338957e-06, + "loss": 1.2007, + "step": 33091 + }, + { + "epoch": 0.8, + "learning_rate": 2.1079402888716547e-06, + "loss": 0.7624, + "step": 33092 + }, + { + "epoch": 0.8, + "learning_rate": 2.107461971488686e-06, + "loss": 1.0573, + "step": 33093 + }, + { + "epoch": 0.8, + "learning_rate": 2.106983701987886e-06, + "loss": 0.9906, + "step": 33094 + }, + { + "epoch": 0.8, + "learning_rate": 2.1065054803721584e-06, + "loss": 1.2114, + "step": 33095 + }, + { + "epoch": 0.8, + "learning_rate": 2.1060273066444104e-06, + "loss": 1.172, + "step": 33096 + }, + { + "epoch": 0.8, + "learning_rate": 2.1055491808075334e-06, + "loss": 1.0018, + "step": 33097 + }, + { + "epoch": 0.8, + "learning_rate": 2.1050711028644356e-06, + "loss": 1.0044, + "step": 33098 + }, + { + "epoch": 0.8, + "learning_rate": 2.104593072818012e-06, + "loss": 0.8464, + "step": 33099 + }, + { + "epoch": 0.8, + "learning_rate": 2.1041150906711662e-06, + "loss": 1.0685, + "step": 33100 + }, + { + "epoch": 0.8, + "learning_rate": 2.1036371564267976e-06, + "loss": 1.0293, + "step": 33101 + }, + { + "epoch": 0.8, + "learning_rate": 2.1031592700878033e-06, + "loss": 1.0111, + "step": 33102 + }, + { + "epoch": 0.8, + "learning_rate": 2.1026814316570844e-06, + "loss": 1.07, + "step": 33103 + }, + { + "epoch": 0.8, + "learning_rate": 2.1022036411375424e-06, + "loss": 1.0411, + "step": 33104 + }, + { + "epoch": 0.8, + "learning_rate": 2.101725898532071e-06, + "loss": 0.9396, + "step": 33105 + }, + { + "epoch": 0.8, + "learning_rate": 2.10124820384357e-06, + "loss": 1.0184, + "step": 33106 + }, + { + "epoch": 0.8, + "learning_rate": 2.1007705570749413e-06, + "loss": 0.9489, + "step": 33107 + }, + { + "epoch": 0.8, + "learning_rate": 2.1002929582290764e-06, + "loss": 0.9075, + "step": 33108 + }, + { + "epoch": 0.8, + "learning_rate": 2.0998154073088793e-06, + "loss": 1.0136, + "step": 33109 + }, + { + "epoch": 0.8, + "learning_rate": 2.099337904317241e-06, + "loss": 1.2094, + "step": 33110 + }, + { + "epoch": 0.8, + "learning_rate": 2.098860449257061e-06, + "loss": 0.9468, + "step": 33111 + }, + { + "epoch": 0.8, + "learning_rate": 2.098383042131239e-06, + "loss": 1.2477, + "step": 33112 + }, + { + "epoch": 0.8, + "learning_rate": 2.097905682942666e-06, + "loss": 0.8945, + "step": 33113 + }, + { + "epoch": 0.8, + "learning_rate": 2.09742837169424e-06, + "loss": 1.164, + "step": 33114 + }, + { + "epoch": 0.8, + "learning_rate": 2.0969511083888615e-06, + "loss": 1.0665, + "step": 33115 + }, + { + "epoch": 0.8, + "learning_rate": 2.0964738930294173e-06, + "loss": 1.0728, + "step": 33116 + }, + { + "epoch": 0.8, + "learning_rate": 2.0959967256188107e-06, + "loss": 1.1505, + "step": 33117 + }, + { + "epoch": 0.8, + "learning_rate": 2.0955196061599294e-06, + "loss": 0.8377, + "step": 33118 + }, + { + "epoch": 0.8, + "learning_rate": 2.0950425346556723e-06, + "loss": 1.0312, + "step": 33119 + }, + { + "epoch": 0.8, + "learning_rate": 2.0945655111089345e-06, + "loss": 1.1151, + "step": 33120 + }, + { + "epoch": 0.8, + "learning_rate": 2.094088535522606e-06, + "loss": 0.9759, + "step": 33121 + }, + { + "epoch": 0.8, + "learning_rate": 2.093611607899583e-06, + "loss": 1.085, + "step": 33122 + }, + { + "epoch": 0.8, + "learning_rate": 2.0931347282427614e-06, + "loss": 1.148, + "step": 33123 + }, + { + "epoch": 0.8, + "learning_rate": 2.092657896555028e-06, + "loss": 1.0342, + "step": 33124 + }, + { + "epoch": 0.8, + "learning_rate": 2.092181112839283e-06, + "loss": 0.7853, + "step": 33125 + }, + { + "epoch": 0.8, + "learning_rate": 2.0917043770984124e-06, + "loss": 1.0716, + "step": 33126 + }, + { + "epoch": 0.8, + "learning_rate": 2.0912276893353102e-06, + "loss": 1.2441, + "step": 33127 + }, + { + "epoch": 0.8, + "learning_rate": 2.0907510495528725e-06, + "loss": 1.1228, + "step": 33128 + }, + { + "epoch": 0.8, + "learning_rate": 2.0902744577539856e-06, + "loss": 1.062, + "step": 33129 + }, + { + "epoch": 0.8, + "learning_rate": 2.0897979139415426e-06, + "loss": 1.1168, + "step": 33130 + }, + { + "epoch": 0.8, + "learning_rate": 2.0893214181184385e-06, + "loss": 1.2344, + "step": 33131 + }, + { + "epoch": 0.8, + "learning_rate": 2.0888449702875568e-06, + "loss": 0.9243, + "step": 33132 + }, + { + "epoch": 0.8, + "learning_rate": 2.0883685704517954e-06, + "loss": 0.8716, + "step": 33133 + }, + { + "epoch": 0.8, + "learning_rate": 2.087892218614038e-06, + "loss": 1.0662, + "step": 33134 + }, + { + "epoch": 0.8, + "learning_rate": 2.0874159147771778e-06, + "loss": 1.096, + "step": 33135 + }, + { + "epoch": 0.8, + "learning_rate": 2.086939658944106e-06, + "loss": 1.0851, + "step": 33136 + }, + { + "epoch": 0.8, + "learning_rate": 2.0864634511177083e-06, + "loss": 0.9252, + "step": 33137 + }, + { + "epoch": 0.8, + "learning_rate": 2.0859872913008748e-06, + "loss": 1.159, + "step": 33138 + }, + { + "epoch": 0.8, + "learning_rate": 2.085511179496498e-06, + "loss": 0.9643, + "step": 33139 + }, + { + "epoch": 0.8, + "learning_rate": 2.08503511570746e-06, + "loss": 0.6959, + "step": 33140 + }, + { + "epoch": 0.8, + "learning_rate": 2.084559099936655e-06, + "loss": 1.2185, + "step": 33141 + }, + { + "epoch": 0.8, + "learning_rate": 2.084083132186966e-06, + "loss": 1.3814, + "step": 33142 + }, + { + "epoch": 0.8, + "learning_rate": 2.083607212461283e-06, + "loss": 1.0687, + "step": 33143 + }, + { + "epoch": 0.8, + "learning_rate": 2.083131340762494e-06, + "loss": 0.9246, + "step": 33144 + }, + { + "epoch": 0.8, + "learning_rate": 2.0826555170934835e-06, + "loss": 0.9529, + "step": 33145 + }, + { + "epoch": 0.8, + "learning_rate": 2.0821797414571397e-06, + "loss": 1.2123, + "step": 33146 + }, + { + "epoch": 0.8, + "learning_rate": 2.0817040138563515e-06, + "loss": 1.0842, + "step": 33147 + }, + { + "epoch": 0.8, + "learning_rate": 2.081228334294e-06, + "loss": 0.9847, + "step": 33148 + }, + { + "epoch": 0.8, + "learning_rate": 2.0807527027729766e-06, + "loss": 1.2499, + "step": 33149 + }, + { + "epoch": 0.8, + "learning_rate": 2.080277119296161e-06, + "loss": 1.0179, + "step": 33150 + }, + { + "epoch": 0.8, + "learning_rate": 2.079801583866441e-06, + "loss": 1.1153, + "step": 33151 + }, + { + "epoch": 0.8, + "learning_rate": 2.0793260964867047e-06, + "loss": 0.7737, + "step": 33152 + }, + { + "epoch": 0.8, + "learning_rate": 2.078850657159831e-06, + "loss": 0.8575, + "step": 33153 + }, + { + "epoch": 0.8, + "learning_rate": 2.078375265888707e-06, + "loss": 1.0706, + "step": 33154 + }, + { + "epoch": 0.8, + "learning_rate": 2.07789992267622e-06, + "loss": 0.7818, + "step": 33155 + }, + { + "epoch": 0.8, + "learning_rate": 2.0774246275252484e-06, + "loss": 1.1654, + "step": 33156 + }, + { + "epoch": 0.8, + "learning_rate": 2.0769493804386775e-06, + "loss": 1.0054, + "step": 33157 + }, + { + "epoch": 0.8, + "learning_rate": 2.0764741814193946e-06, + "loss": 1.2607, + "step": 33158 + }, + { + "epoch": 0.8, + "learning_rate": 2.075999030470276e-06, + "loss": 0.9919, + "step": 33159 + }, + { + "epoch": 0.8, + "learning_rate": 2.0755239275942097e-06, + "loss": 1.1326, + "step": 33160 + }, + { + "epoch": 0.8, + "learning_rate": 2.0750488727940734e-06, + "loss": 1.0977, + "step": 33161 + }, + { + "epoch": 0.8, + "learning_rate": 2.074573866072751e-06, + "loss": 1.1496, + "step": 33162 + }, + { + "epoch": 0.8, + "learning_rate": 2.0740989074331287e-06, + "loss": 1.0284, + "step": 33163 + }, + { + "epoch": 0.8, + "learning_rate": 2.0736239968780803e-06, + "loss": 0.9398, + "step": 33164 + }, + { + "epoch": 0.8, + "learning_rate": 2.0731491344104915e-06, + "loss": 1.0993, + "step": 33165 + }, + { + "epoch": 0.8, + "learning_rate": 2.072674320033241e-06, + "loss": 1.0608, + "step": 33166 + }, + { + "epoch": 0.8, + "learning_rate": 2.0721995537492123e-06, + "loss": 1.0002, + "step": 33167 + }, + { + "epoch": 0.8, + "learning_rate": 2.0717248355612864e-06, + "loss": 1.0198, + "step": 33168 + }, + { + "epoch": 0.8, + "learning_rate": 2.071250165472338e-06, + "loss": 0.9981, + "step": 33169 + }, + { + "epoch": 0.8, + "learning_rate": 2.070775543485253e-06, + "loss": 1.0558, + "step": 33170 + }, + { + "epoch": 0.8, + "learning_rate": 2.0703009696029053e-06, + "loss": 0.7925, + "step": 33171 + }, + { + "epoch": 0.8, + "learning_rate": 2.069826443828176e-06, + "loss": 0.8555, + "step": 33172 + }, + { + "epoch": 0.8, + "learning_rate": 2.069351966163947e-06, + "loss": 0.9875, + "step": 33173 + }, + { + "epoch": 0.8, + "learning_rate": 2.068877536613091e-06, + "loss": 1.0301, + "step": 33174 + }, + { + "epoch": 0.8, + "learning_rate": 2.06840315517849e-06, + "loss": 1.211, + "step": 33175 + }, + { + "epoch": 0.8, + "learning_rate": 2.0679288218630245e-06, + "loss": 0.9708, + "step": 33176 + }, + { + "epoch": 0.8, + "learning_rate": 2.0674545366695664e-06, + "loss": 1.238, + "step": 33177 + }, + { + "epoch": 0.8, + "learning_rate": 2.066980299600998e-06, + "loss": 0.8254, + "step": 33178 + }, + { + "epoch": 0.8, + "learning_rate": 2.0665061106601914e-06, + "loss": 1.0388, + "step": 33179 + }, + { + "epoch": 0.8, + "learning_rate": 2.0660319698500264e-06, + "loss": 0.8198, + "step": 33180 + }, + { + "epoch": 0.8, + "learning_rate": 2.065557877173382e-06, + "loss": 0.8099, + "step": 33181 + }, + { + "epoch": 0.8, + "learning_rate": 2.065083832633129e-06, + "loss": 1.0973, + "step": 33182 + }, + { + "epoch": 0.8, + "learning_rate": 2.0646098362321465e-06, + "loss": 1.0017, + "step": 33183 + }, + { + "epoch": 0.8, + "learning_rate": 2.064135887973311e-06, + "loss": 0.8218, + "step": 33184 + }, + { + "epoch": 0.8, + "learning_rate": 2.063661987859494e-06, + "loss": 1.2561, + "step": 33185 + }, + { + "epoch": 0.8, + "learning_rate": 2.0631881358935747e-06, + "loss": 1.0478, + "step": 33186 + }, + { + "epoch": 0.8, + "learning_rate": 2.0627143320784273e-06, + "loss": 1.0726, + "step": 33187 + }, + { + "epoch": 0.8, + "learning_rate": 2.062240576416923e-06, + "loss": 1.0672, + "step": 33188 + }, + { + "epoch": 0.8, + "learning_rate": 2.06176686891194e-06, + "loss": 0.8637, + "step": 33189 + }, + { + "epoch": 0.8, + "learning_rate": 2.0612932095663473e-06, + "loss": 0.5538, + "step": 33190 + }, + { + "epoch": 0.8, + "learning_rate": 2.0608195983830228e-06, + "loss": 1.041, + "step": 33191 + }, + { + "epoch": 0.8, + "learning_rate": 2.0603460353648398e-06, + "loss": 0.8487, + "step": 33192 + }, + { + "epoch": 0.8, + "learning_rate": 2.0598725205146676e-06, + "loss": 1.047, + "step": 33193 + }, + { + "epoch": 0.8, + "learning_rate": 2.0593990538353816e-06, + "loss": 1.2057, + "step": 33194 + }, + { + "epoch": 0.8, + "learning_rate": 2.058925635329856e-06, + "loss": 1.0066, + "step": 33195 + }, + { + "epoch": 0.8, + "learning_rate": 2.0584522650009587e-06, + "loss": 1.1999, + "step": 33196 + }, + { + "epoch": 0.8, + "learning_rate": 2.057978942851565e-06, + "loss": 0.849, + "step": 33197 + }, + { + "epoch": 0.8, + "learning_rate": 2.057505668884543e-06, + "loss": 0.9668, + "step": 33198 + }, + { + "epoch": 0.8, + "learning_rate": 2.0570324431027667e-06, + "loss": 1.2174, + "step": 33199 + }, + { + "epoch": 0.8, + "learning_rate": 2.056559265509108e-06, + "loss": 1.0296, + "step": 33200 + }, + { + "epoch": 0.8, + "learning_rate": 2.0560861361064333e-06, + "loss": 0.9683, + "step": 33201 + }, + { + "epoch": 0.8, + "learning_rate": 2.0556130548976148e-06, + "loss": 1.0283, + "step": 33202 + }, + { + "epoch": 0.8, + "learning_rate": 2.0551400218855266e-06, + "loss": 0.6966, + "step": 33203 + }, + { + "epoch": 0.8, + "learning_rate": 2.0546670370730326e-06, + "loss": 1.1388, + "step": 33204 + }, + { + "epoch": 0.8, + "learning_rate": 2.054194100463007e-06, + "loss": 1.1767, + "step": 33205 + }, + { + "epoch": 0.8, + "learning_rate": 2.053721212058315e-06, + "loss": 0.9264, + "step": 33206 + }, + { + "epoch": 0.8, + "learning_rate": 2.0532483718618267e-06, + "loss": 1.0073, + "step": 33207 + }, + { + "epoch": 0.8, + "learning_rate": 2.052775579876415e-06, + "loss": 1.2708, + "step": 33208 + }, + { + "epoch": 0.8, + "learning_rate": 2.0523028361049414e-06, + "loss": 1.2554, + "step": 33209 + }, + { + "epoch": 0.8, + "learning_rate": 2.0518301405502763e-06, + "loss": 1.231, + "step": 33210 + }, + { + "epoch": 0.8, + "learning_rate": 2.0513574932152915e-06, + "loss": 1.3409, + "step": 33211 + }, + { + "epoch": 0.8, + "learning_rate": 2.050884894102848e-06, + "loss": 0.9958, + "step": 33212 + }, + { + "epoch": 0.8, + "learning_rate": 2.0504123432158196e-06, + "loss": 1.1768, + "step": 33213 + }, + { + "epoch": 0.8, + "learning_rate": 2.0499398405570658e-06, + "loss": 0.8507, + "step": 33214 + }, + { + "epoch": 0.8, + "learning_rate": 2.0494673861294576e-06, + "loss": 1.1712, + "step": 33215 + }, + { + "epoch": 0.8, + "learning_rate": 2.0489949799358646e-06, + "loss": 0.936, + "step": 33216 + }, + { + "epoch": 0.8, + "learning_rate": 2.0485226219791444e-06, + "loss": 1.133, + "step": 33217 + }, + { + "epoch": 0.8, + "learning_rate": 2.0480503122621674e-06, + "loss": 0.8131, + "step": 33218 + }, + { + "epoch": 0.8, + "learning_rate": 2.047578050787802e-06, + "loss": 0.8608, + "step": 33219 + }, + { + "epoch": 0.8, + "learning_rate": 2.047105837558907e-06, + "loss": 0.9905, + "step": 33220 + }, + { + "epoch": 0.8, + "learning_rate": 2.046633672578353e-06, + "loss": 1.1037, + "step": 33221 + }, + { + "epoch": 0.8, + "learning_rate": 2.046161555849e-06, + "loss": 0.9208, + "step": 33222 + }, + { + "epoch": 0.8, + "learning_rate": 2.0456894873737133e-06, + "loss": 0.9481, + "step": 33223 + }, + { + "epoch": 0.8, + "learning_rate": 2.0452174671553593e-06, + "loss": 1.0673, + "step": 33224 + }, + { + "epoch": 0.8, + "learning_rate": 2.0447454951967984e-06, + "loss": 0.8063, + "step": 33225 + }, + { + "epoch": 0.8, + "learning_rate": 2.0442735715008955e-06, + "loss": 1.1288, + "step": 33226 + }, + { + "epoch": 0.8, + "learning_rate": 2.0438016960705155e-06, + "loss": 1.2256, + "step": 33227 + }, + { + "epoch": 0.8, + "learning_rate": 2.043329868908517e-06, + "loss": 1.0165, + "step": 33228 + }, + { + "epoch": 0.8, + "learning_rate": 2.0428580900177663e-06, + "loss": 0.9546, + "step": 33229 + }, + { + "epoch": 0.8, + "learning_rate": 2.0423863594011217e-06, + "loss": 0.987, + "step": 33230 + }, + { + "epoch": 0.8, + "learning_rate": 2.0419146770614485e-06, + "loss": 1.2404, + "step": 33231 + }, + { + "epoch": 0.8, + "learning_rate": 2.0414430430016085e-06, + "loss": 1.1191, + "step": 33232 + }, + { + "epoch": 0.8, + "learning_rate": 2.040971457224459e-06, + "loss": 1.0656, + "step": 33233 + }, + { + "epoch": 0.8, + "learning_rate": 2.0404999197328644e-06, + "loss": 1.0763, + "step": 33234 + }, + { + "epoch": 0.8, + "learning_rate": 2.040028430529687e-06, + "loss": 0.8612, + "step": 33235 + }, + { + "epoch": 0.8, + "learning_rate": 2.0395569896177814e-06, + "loss": 1.1062, + "step": 33236 + }, + { + "epoch": 0.8, + "learning_rate": 2.0390855970000113e-06, + "loss": 1.1557, + "step": 33237 + }, + { + "epoch": 0.8, + "learning_rate": 2.03861425267924e-06, + "loss": 0.9265, + "step": 33238 + }, + { + "epoch": 0.8, + "learning_rate": 2.03814295665832e-06, + "loss": 1.0346, + "step": 33239 + }, + { + "epoch": 0.8, + "learning_rate": 2.0376717089401166e-06, + "loss": 1.0599, + "step": 33240 + }, + { + "epoch": 0.8, + "learning_rate": 2.0372005095274828e-06, + "loss": 1.1134, + "step": 33241 + }, + { + "epoch": 0.8, + "learning_rate": 2.0367293584232817e-06, + "loss": 1.0636, + "step": 33242 + }, + { + "epoch": 0.8, + "learning_rate": 2.0362582556303723e-06, + "loss": 1.2881, + "step": 33243 + }, + { + "epoch": 0.8, + "learning_rate": 2.035787201151609e-06, + "loss": 0.9419, + "step": 33244 + }, + { + "epoch": 0.8, + "learning_rate": 2.0353161949898515e-06, + "loss": 1.1785, + "step": 33245 + }, + { + "epoch": 0.8, + "learning_rate": 2.034845237147959e-06, + "loss": 1.1966, + "step": 33246 + }, + { + "epoch": 0.8, + "learning_rate": 2.0343743276287853e-06, + "loss": 1.1785, + "step": 33247 + }, + { + "epoch": 0.8, + "learning_rate": 2.033903466435191e-06, + "loss": 1.0883, + "step": 33248 + }, + { + "epoch": 0.8, + "learning_rate": 2.033432653570028e-06, + "loss": 1.1706, + "step": 33249 + }, + { + "epoch": 0.8, + "learning_rate": 2.0329618890361547e-06, + "loss": 0.9227, + "step": 33250 + }, + { + "epoch": 0.8, + "learning_rate": 2.0324911728364317e-06, + "loss": 1.0967, + "step": 33251 + }, + { + "epoch": 0.8, + "learning_rate": 2.0320205049737063e-06, + "loss": 0.9349, + "step": 33252 + }, + { + "epoch": 0.8, + "learning_rate": 2.0315498854508396e-06, + "loss": 0.9981, + "step": 33253 + }, + { + "epoch": 0.8, + "learning_rate": 2.0310793142706876e-06, + "loss": 1.2007, + "step": 33254 + }, + { + "epoch": 0.8, + "learning_rate": 2.0306087914361007e-06, + "loss": 1.1839, + "step": 33255 + }, + { + "epoch": 0.8, + "learning_rate": 2.0301383169499376e-06, + "loss": 1.3082, + "step": 33256 + }, + { + "epoch": 0.8, + "learning_rate": 2.029667890815049e-06, + "loss": 0.9922, + "step": 33257 + }, + { + "epoch": 0.8, + "learning_rate": 2.02919751303429e-06, + "loss": 1.0477, + "step": 33258 + }, + { + "epoch": 0.8, + "learning_rate": 2.0287271836105182e-06, + "loss": 1.0633, + "step": 33259 + }, + { + "epoch": 0.8, + "learning_rate": 2.0282569025465805e-06, + "loss": 1.1676, + "step": 33260 + }, + { + "epoch": 0.8, + "learning_rate": 2.027786669845333e-06, + "loss": 1.1227, + "step": 33261 + }, + { + "epoch": 0.8, + "learning_rate": 2.0273164855096305e-06, + "loss": 1.1278, + "step": 33262 + }, + { + "epoch": 0.8, + "learning_rate": 2.026846349542322e-06, + "loss": 0.9043, + "step": 33263 + }, + { + "epoch": 0.8, + "learning_rate": 2.0263762619462637e-06, + "loss": 1.1833, + "step": 33264 + }, + { + "epoch": 0.8, + "learning_rate": 2.025906222724302e-06, + "loss": 0.9149, + "step": 33265 + }, + { + "epoch": 0.8, + "learning_rate": 2.025436231879293e-06, + "loss": 1.1283, + "step": 33266 + }, + { + "epoch": 0.8, + "learning_rate": 2.0249662894140874e-06, + "loss": 1.1025, + "step": 33267 + }, + { + "epoch": 0.8, + "learning_rate": 2.024496395331534e-06, + "loss": 1.102, + "step": 33268 + }, + { + "epoch": 0.8, + "learning_rate": 2.0240265496344845e-06, + "loss": 0.9012, + "step": 33269 + }, + { + "epoch": 0.8, + "learning_rate": 2.0235567523257937e-06, + "loss": 0.819, + "step": 33270 + }, + { + "epoch": 0.8, + "learning_rate": 2.023087003408304e-06, + "loss": 1.2162, + "step": 33271 + }, + { + "epoch": 0.8, + "learning_rate": 2.0226173028848727e-06, + "loss": 1.3073, + "step": 33272 + }, + { + "epoch": 0.8, + "learning_rate": 2.0221476507583426e-06, + "loss": 0.8514, + "step": 33273 + }, + { + "epoch": 0.8, + "learning_rate": 2.0216780470315657e-06, + "loss": 1.1699, + "step": 33274 + }, + { + "epoch": 0.8, + "learning_rate": 2.0212084917073947e-06, + "loss": 1.1037, + "step": 33275 + }, + { + "epoch": 0.8, + "learning_rate": 2.0207389847886727e-06, + "loss": 1.1694, + "step": 33276 + }, + { + "epoch": 0.8, + "learning_rate": 2.02026952627825e-06, + "loss": 1.0162, + "step": 33277 + }, + { + "epoch": 0.8, + "learning_rate": 2.019800116178978e-06, + "loss": 1.0585, + "step": 33278 + }, + { + "epoch": 0.8, + "learning_rate": 2.019330754493699e-06, + "loss": 1.1339, + "step": 33279 + }, + { + "epoch": 0.8, + "learning_rate": 2.0188614412252625e-06, + "loss": 1.0314, + "step": 33280 + }, + { + "epoch": 0.8, + "learning_rate": 2.01839217637652e-06, + "loss": 1.0294, + "step": 33281 + }, + { + "epoch": 0.8, + "learning_rate": 2.0179229599503114e-06, + "loss": 0.9764, + "step": 33282 + }, + { + "epoch": 0.8, + "learning_rate": 2.0174537919494896e-06, + "loss": 1.0686, + "step": 33283 + }, + { + "epoch": 0.8, + "learning_rate": 2.016984672376896e-06, + "loss": 1.124, + "step": 33284 + }, + { + "epoch": 0.8, + "learning_rate": 2.016515601235378e-06, + "loss": 1.0528, + "step": 33285 + }, + { + "epoch": 0.8, + "learning_rate": 2.016046578527785e-06, + "loss": 0.9233, + "step": 33286 + }, + { + "epoch": 0.8, + "learning_rate": 2.0155776042569562e-06, + "loss": 0.9208, + "step": 33287 + }, + { + "epoch": 0.8, + "learning_rate": 2.0151086784257403e-06, + "loss": 1.1497, + "step": 33288 + }, + { + "epoch": 0.8, + "learning_rate": 2.014639801036985e-06, + "loss": 1.2061, + "step": 33289 + }, + { + "epoch": 0.8, + "learning_rate": 2.014170972093529e-06, + "loss": 1.0361, + "step": 33290 + }, + { + "epoch": 0.8, + "learning_rate": 2.013702191598221e-06, + "loss": 1.1208, + "step": 33291 + }, + { + "epoch": 0.8, + "learning_rate": 2.0132334595539015e-06, + "loss": 1.0896, + "step": 33292 + }, + { + "epoch": 0.8, + "learning_rate": 2.012764775963416e-06, + "loss": 1.1219, + "step": 33293 + }, + { + "epoch": 0.8, + "learning_rate": 2.0122961408296105e-06, + "loss": 0.8019, + "step": 33294 + }, + { + "epoch": 0.8, + "learning_rate": 2.0118275541553223e-06, + "loss": 1.0589, + "step": 33295 + }, + { + "epoch": 0.8, + "learning_rate": 2.011359015943398e-06, + "loss": 1.1166, + "step": 33296 + }, + { + "epoch": 0.8, + "learning_rate": 2.0108905261966826e-06, + "loss": 0.9863, + "step": 33297 + }, + { + "epoch": 0.8, + "learning_rate": 2.0104220849180113e-06, + "loss": 1.0183, + "step": 33298 + }, + { + "epoch": 0.8, + "learning_rate": 2.0099536921102337e-06, + "loss": 1.0259, + "step": 33299 + }, + { + "epoch": 0.8, + "learning_rate": 2.0094853477761845e-06, + "loss": 1.057, + "step": 33300 + }, + { + "epoch": 0.8, + "learning_rate": 2.009017051918708e-06, + "loss": 1.1581, + "step": 33301 + }, + { + "epoch": 0.8, + "learning_rate": 2.0085488045406488e-06, + "loss": 0.8741, + "step": 33302 + }, + { + "epoch": 0.8, + "learning_rate": 2.008080605644841e-06, + "loss": 0.8774, + "step": 33303 + }, + { + "epoch": 0.8, + "learning_rate": 2.0076124552341293e-06, + "loss": 1.0749, + "step": 33304 + }, + { + "epoch": 0.8, + "learning_rate": 2.0071443533113546e-06, + "loss": 0.9986, + "step": 33305 + }, + { + "epoch": 0.8, + "learning_rate": 2.006676299879353e-06, + "loss": 1.0677, + "step": 33306 + }, + { + "epoch": 0.8, + "learning_rate": 2.006208294940968e-06, + "loss": 1.0607, + "step": 33307 + }, + { + "epoch": 0.8, + "learning_rate": 2.0057403384990347e-06, + "loss": 1.1802, + "step": 33308 + }, + { + "epoch": 0.8, + "learning_rate": 2.0052724305563954e-06, + "loss": 1.0436, + "step": 33309 + }, + { + "epoch": 0.8, + "learning_rate": 2.0048045711158894e-06, + "loss": 1.0366, + "step": 33310 + }, + { + "epoch": 0.8, + "learning_rate": 2.004336760180351e-06, + "loss": 1.0809, + "step": 33311 + }, + { + "epoch": 0.8, + "learning_rate": 2.0038689977526206e-06, + "loss": 1.2639, + "step": 33312 + }, + { + "epoch": 0.8, + "learning_rate": 2.003401283835539e-06, + "loss": 1.1413, + "step": 33313 + }, + { + "epoch": 0.8, + "learning_rate": 2.002933618431938e-06, + "loss": 1.1404, + "step": 33314 + }, + { + "epoch": 0.8, + "learning_rate": 2.0024660015446618e-06, + "loss": 1.1955, + "step": 33315 + }, + { + "epoch": 0.8, + "learning_rate": 2.0019984331765384e-06, + "loss": 0.9079, + "step": 33316 + }, + { + "epoch": 0.8, + "learning_rate": 2.0015309133304106e-06, + "loss": 0.9268, + "step": 33317 + }, + { + "epoch": 0.8, + "learning_rate": 2.0010634420091157e-06, + "loss": 0.9843, + "step": 33318 + }, + { + "epoch": 0.8, + "learning_rate": 2.0005960192154848e-06, + "loss": 1.055, + "step": 33319 + }, + { + "epoch": 0.8, + "learning_rate": 2.000128644952357e-06, + "loss": 1.1143, + "step": 33320 + }, + { + "epoch": 0.8, + "learning_rate": 1.999661319222569e-06, + "loss": 0.848, + "step": 33321 + }, + { + "epoch": 0.8, + "learning_rate": 1.999194042028951e-06, + "loss": 1.2053, + "step": 33322 + }, + { + "epoch": 0.8, + "learning_rate": 1.9987268133743443e-06, + "loss": 0.9532, + "step": 33323 + }, + { + "epoch": 0.8, + "learning_rate": 1.9982596332615765e-06, + "loss": 0.7622, + "step": 33324 + }, + { + "epoch": 0.8, + "learning_rate": 1.9977925016934853e-06, + "loss": 0.9249, + "step": 33325 + }, + { + "epoch": 0.8, + "learning_rate": 1.9973254186729084e-06, + "loss": 1.1228, + "step": 33326 + }, + { + "epoch": 0.8, + "learning_rate": 1.9968583842026722e-06, + "loss": 0.9578, + "step": 33327 + }, + { + "epoch": 0.8, + "learning_rate": 1.9963913982856144e-06, + "loss": 0.7864, + "step": 33328 + }, + { + "epoch": 0.8, + "learning_rate": 1.99592446092457e-06, + "loss": 1.1397, + "step": 33329 + }, + { + "epoch": 0.8, + "learning_rate": 1.9954575721223667e-06, + "loss": 0.9376, + "step": 33330 + }, + { + "epoch": 0.8, + "learning_rate": 1.9949907318818395e-06, + "loss": 0.8914, + "step": 33331 + }, + { + "epoch": 0.8, + "learning_rate": 1.9945239402058203e-06, + "loss": 1.0129, + "step": 33332 + }, + { + "epoch": 0.8, + "learning_rate": 1.9940571970971413e-06, + "loss": 0.9642, + "step": 33333 + }, + { + "epoch": 0.8, + "learning_rate": 1.9935905025586368e-06, + "loss": 0.9777, + "step": 33334 + }, + { + "epoch": 0.8, + "learning_rate": 1.9931238565931333e-06, + "loss": 1.0944, + "step": 33335 + }, + { + "epoch": 0.8, + "learning_rate": 1.9926572592034667e-06, + "loss": 1.0852, + "step": 33336 + }, + { + "epoch": 0.8, + "learning_rate": 1.992190710392462e-06, + "loss": 0.8993, + "step": 33337 + }, + { + "epoch": 0.8, + "learning_rate": 1.991724210162952e-06, + "loss": 1.0192, + "step": 33338 + }, + { + "epoch": 0.8, + "learning_rate": 1.9912577585177715e-06, + "loss": 0.9624, + "step": 33339 + }, + { + "epoch": 0.8, + "learning_rate": 1.9907913554597434e-06, + "loss": 0.8342, + "step": 33340 + }, + { + "epoch": 0.8, + "learning_rate": 1.9903250009916997e-06, + "loss": 1.0895, + "step": 33341 + }, + { + "epoch": 0.8, + "learning_rate": 1.9898586951164734e-06, + "loss": 1.1858, + "step": 33342 + }, + { + "epoch": 0.8, + "learning_rate": 1.989392437836888e-06, + "loss": 1.2217, + "step": 33343 + }, + { + "epoch": 0.8, + "learning_rate": 1.9889262291557764e-06, + "loss": 0.84, + "step": 33344 + }, + { + "epoch": 0.8, + "learning_rate": 1.9884600690759628e-06, + "loss": 1.0212, + "step": 33345 + }, + { + "epoch": 0.8, + "learning_rate": 1.9879939576002773e-06, + "loss": 1.018, + "step": 33346 + }, + { + "epoch": 0.8, + "learning_rate": 1.9875278947315512e-06, + "loss": 1.0797, + "step": 33347 + }, + { + "epoch": 0.8, + "learning_rate": 1.9870618804726048e-06, + "loss": 0.9544, + "step": 33348 + }, + { + "epoch": 0.8, + "learning_rate": 1.9865959148262704e-06, + "loss": 1.0467, + "step": 33349 + }, + { + "epoch": 0.8, + "learning_rate": 1.9861299977953764e-06, + "loss": 1.0248, + "step": 33350 + }, + { + "epoch": 0.8, + "learning_rate": 1.985664129382743e-06, + "loss": 1.1334, + "step": 33351 + }, + { + "epoch": 0.8, + "learning_rate": 1.985198309591204e-06, + "loss": 1.0123, + "step": 33352 + }, + { + "epoch": 0.8, + "learning_rate": 1.984732538423578e-06, + "loss": 1.246, + "step": 33353 + }, + { + "epoch": 0.8, + "learning_rate": 1.984266815882695e-06, + "loss": 0.7985, + "step": 33354 + }, + { + "epoch": 0.8, + "learning_rate": 1.9838011419713822e-06, + "loss": 1.0965, + "step": 33355 + }, + { + "epoch": 0.8, + "learning_rate": 1.9833355166924607e-06, + "loss": 0.9788, + "step": 33356 + }, + { + "epoch": 0.8, + "learning_rate": 1.982869940048756e-06, + "loss": 0.9534, + "step": 33357 + }, + { + "epoch": 0.8, + "learning_rate": 1.9824044120430964e-06, + "loss": 1.0775, + "step": 33358 + }, + { + "epoch": 0.8, + "learning_rate": 1.9819389326783e-06, + "loss": 0.8934, + "step": 33359 + }, + { + "epoch": 0.8, + "learning_rate": 1.981473501957195e-06, + "loss": 1.0622, + "step": 33360 + }, + { + "epoch": 0.8, + "learning_rate": 1.981008119882607e-06, + "loss": 1.1718, + "step": 33361 + }, + { + "epoch": 0.8, + "learning_rate": 1.9805427864573532e-06, + "loss": 0.8934, + "step": 33362 + }, + { + "epoch": 0.8, + "learning_rate": 1.980077501684263e-06, + "loss": 1.0096, + "step": 33363 + }, + { + "epoch": 0.8, + "learning_rate": 1.9796122655661543e-06, + "loss": 1.2372, + "step": 33364 + }, + { + "epoch": 0.8, + "learning_rate": 1.979147078105851e-06, + "loss": 0.9983, + "step": 33365 + }, + { + "epoch": 0.8, + "learning_rate": 1.9786819393061783e-06, + "loss": 1.1947, + "step": 33366 + }, + { + "epoch": 0.8, + "learning_rate": 1.978216849169953e-06, + "loss": 1.1161, + "step": 33367 + }, + { + "epoch": 0.8, + "learning_rate": 1.977751807699999e-06, + "loss": 1.1164, + "step": 33368 + }, + { + "epoch": 0.8, + "learning_rate": 1.97728681489914e-06, + "loss": 0.9002, + "step": 33369 + }, + { + "epoch": 0.8, + "learning_rate": 1.9768218707701924e-06, + "loss": 1.0241, + "step": 33370 + }, + { + "epoch": 0.8, + "learning_rate": 1.976356975315983e-06, + "loss": 1.1065, + "step": 33371 + }, + { + "epoch": 0.8, + "learning_rate": 1.975892128539325e-06, + "loss": 0.9021, + "step": 33372 + }, + { + "epoch": 0.8, + "learning_rate": 1.9754273304430427e-06, + "loss": 0.9474, + "step": 33373 + }, + { + "epoch": 0.8, + "learning_rate": 1.9749625810299577e-06, + "loss": 0.9417, + "step": 33374 + }, + { + "epoch": 0.8, + "learning_rate": 1.9744978803028836e-06, + "loss": 0.673, + "step": 33375 + }, + { + "epoch": 0.8, + "learning_rate": 1.9740332282646448e-06, + "loss": 1.1303, + "step": 33376 + }, + { + "epoch": 0.8, + "learning_rate": 1.9735686249180598e-06, + "loss": 1.1729, + "step": 33377 + }, + { + "epoch": 0.8, + "learning_rate": 1.973104070265944e-06, + "loss": 1.0969, + "step": 33378 + }, + { + "epoch": 0.8, + "learning_rate": 1.9726395643111198e-06, + "loss": 1.0802, + "step": 33379 + }, + { + "epoch": 0.8, + "learning_rate": 1.972175107056401e-06, + "loss": 1.1108, + "step": 33380 + }, + { + "epoch": 0.8, + "learning_rate": 1.9717106985046073e-06, + "loss": 1.3123, + "step": 33381 + }, + { + "epoch": 0.8, + "learning_rate": 1.97124633865856e-06, + "loss": 0.9552, + "step": 33382 + }, + { + "epoch": 0.8, + "learning_rate": 1.9707820275210696e-06, + "loss": 0.8696, + "step": 33383 + }, + { + "epoch": 0.8, + "learning_rate": 1.9703177650949558e-06, + "loss": 1.0773, + "step": 33384 + }, + { + "epoch": 0.8, + "learning_rate": 1.9698535513830386e-06, + "loss": 0.8627, + "step": 33385 + }, + { + "epoch": 0.8, + "learning_rate": 1.9693893863881274e-06, + "loss": 1.2909, + "step": 33386 + }, + { + "epoch": 0.8, + "learning_rate": 1.9689252701130457e-06, + "loss": 1.0417, + "step": 33387 + }, + { + "epoch": 0.8, + "learning_rate": 1.968461202560602e-06, + "loss": 1.0162, + "step": 33388 + }, + { + "epoch": 0.8, + "learning_rate": 1.9679971837336153e-06, + "loss": 1.4401, + "step": 33389 + }, + { + "epoch": 0.8, + "learning_rate": 1.9675332136349023e-06, + "loss": 1.0984, + "step": 33390 + }, + { + "epoch": 0.8, + "learning_rate": 1.9670692922672742e-06, + "loss": 1.1932, + "step": 33391 + }, + { + "epoch": 0.8, + "learning_rate": 1.966605419633547e-06, + "loss": 1.1099, + "step": 33392 + }, + { + "epoch": 0.8, + "learning_rate": 1.9661415957365383e-06, + "loss": 0.9241, + "step": 33393 + }, + { + "epoch": 0.8, + "learning_rate": 1.965677820579055e-06, + "loss": 1.2072, + "step": 33394 + }, + { + "epoch": 0.8, + "learning_rate": 1.9652140941639186e-06, + "loss": 1.1658, + "step": 33395 + }, + { + "epoch": 0.8, + "learning_rate": 1.9647504164939345e-06, + "loss": 1.0116, + "step": 33396 + }, + { + "epoch": 0.8, + "learning_rate": 1.964286787571921e-06, + "loss": 0.8663, + "step": 33397 + }, + { + "epoch": 0.8, + "learning_rate": 1.9638232074006904e-06, + "loss": 0.9286, + "step": 33398 + }, + { + "epoch": 0.8, + "learning_rate": 1.963359675983053e-06, + "loss": 0.9886, + "step": 33399 + }, + { + "epoch": 0.8, + "learning_rate": 1.9628961933218216e-06, + "loss": 1.0937, + "step": 33400 + }, + { + "epoch": 0.8, + "learning_rate": 1.962432759419811e-06, + "loss": 0.9897, + "step": 33401 + }, + { + "epoch": 0.8, + "learning_rate": 1.961969374279827e-06, + "loss": 1.0526, + "step": 33402 + }, + { + "epoch": 0.8, + "learning_rate": 1.9615060379046858e-06, + "loss": 1.085, + "step": 33403 + }, + { + "epoch": 0.8, + "learning_rate": 1.9610427502971987e-06, + "loss": 1.3148, + "step": 33404 + }, + { + "epoch": 0.8, + "learning_rate": 1.9605795114601712e-06, + "loss": 0.8798, + "step": 33405 + }, + { + "epoch": 0.8, + "learning_rate": 1.9601163213964194e-06, + "loss": 1.1564, + "step": 33406 + }, + { + "epoch": 0.8, + "learning_rate": 1.959653180108748e-06, + "loss": 1.0575, + "step": 33407 + }, + { + "epoch": 0.8, + "learning_rate": 1.9591900875999703e-06, + "loss": 0.9418, + "step": 33408 + }, + { + "epoch": 0.8, + "learning_rate": 1.958727043872898e-06, + "loss": 1.1927, + "step": 33409 + }, + { + "epoch": 0.8, + "learning_rate": 1.958264048930334e-06, + "loss": 1.3609, + "step": 33410 + }, + { + "epoch": 0.8, + "learning_rate": 1.957801102775091e-06, + "loss": 0.9169, + "step": 33411 + }, + { + "epoch": 0.8, + "learning_rate": 1.9573382054099786e-06, + "loss": 1.1264, + "step": 33412 + }, + { + "epoch": 0.8, + "learning_rate": 1.956875356837802e-06, + "loss": 1.2814, + "step": 33413 + }, + { + "epoch": 0.8, + "learning_rate": 1.9564125570613723e-06, + "loss": 0.7843, + "step": 33414 + }, + { + "epoch": 0.8, + "learning_rate": 1.9559498060834935e-06, + "loss": 0.9108, + "step": 33415 + }, + { + "epoch": 0.8, + "learning_rate": 1.9554871039069754e-06, + "loss": 0.9113, + "step": 33416 + }, + { + "epoch": 0.8, + "learning_rate": 1.9550244505346284e-06, + "loss": 0.8063, + "step": 33417 + }, + { + "epoch": 0.8, + "learning_rate": 1.9545618459692516e-06, + "loss": 1.0429, + "step": 33418 + }, + { + "epoch": 0.8, + "learning_rate": 1.9540992902136567e-06, + "loss": 1.0247, + "step": 33419 + }, + { + "epoch": 0.8, + "learning_rate": 1.9536367832706503e-06, + "loss": 1.3176, + "step": 33420 + }, + { + "epoch": 0.8, + "learning_rate": 1.9531743251430356e-06, + "loss": 1.1294, + "step": 33421 + }, + { + "epoch": 0.8, + "learning_rate": 1.9527119158336215e-06, + "loss": 1.2174, + "step": 33422 + }, + { + "epoch": 0.8, + "learning_rate": 1.952249555345209e-06, + "loss": 1.1836, + "step": 33423 + }, + { + "epoch": 0.8, + "learning_rate": 1.951787243680605e-06, + "loss": 1.204, + "step": 33424 + }, + { + "epoch": 0.8, + "learning_rate": 1.951324980842618e-06, + "loss": 1.014, + "step": 33425 + }, + { + "epoch": 0.8, + "learning_rate": 1.9508627668340463e-06, + "loss": 0.9447, + "step": 33426 + }, + { + "epoch": 0.8, + "learning_rate": 1.950400601657697e-06, + "loss": 1.0223, + "step": 33427 + }, + { + "epoch": 0.8, + "learning_rate": 1.949938485316376e-06, + "loss": 0.8363, + "step": 33428 + }, + { + "epoch": 0.8, + "learning_rate": 1.9494764178128823e-06, + "loss": 0.959, + "step": 33429 + }, + { + "epoch": 0.8, + "learning_rate": 1.9490143991500245e-06, + "loss": 0.9877, + "step": 33430 + }, + { + "epoch": 0.8, + "learning_rate": 1.9485524293306003e-06, + "loss": 0.9808, + "step": 33431 + }, + { + "epoch": 0.8, + "learning_rate": 1.9480905083574144e-06, + "loss": 0.9802, + "step": 33432 + }, + { + "epoch": 0.8, + "learning_rate": 1.947628636233272e-06, + "loss": 0.9057, + "step": 33433 + }, + { + "epoch": 0.8, + "learning_rate": 1.94716681296097e-06, + "loss": 0.6589, + "step": 33434 + }, + { + "epoch": 0.8, + "learning_rate": 1.946705038543313e-06, + "loss": 0.9869, + "step": 33435 + }, + { + "epoch": 0.8, + "learning_rate": 1.946243312983104e-06, + "loss": 1.186, + "step": 33436 + }, + { + "epoch": 0.8, + "learning_rate": 1.9457816362831406e-06, + "loss": 1.1302, + "step": 33437 + }, + { + "epoch": 0.8, + "learning_rate": 1.9453200084462275e-06, + "loss": 1.1054, + "step": 33438 + }, + { + "epoch": 0.8, + "learning_rate": 1.9448584294751614e-06, + "loss": 0.8014, + "step": 33439 + }, + { + "epoch": 0.8, + "learning_rate": 1.9443968993727447e-06, + "loss": 0.7911, + "step": 33440 + }, + { + "epoch": 0.8, + "learning_rate": 1.9439354181417803e-06, + "loss": 0.8907, + "step": 33441 + }, + { + "epoch": 0.8, + "learning_rate": 1.9434739857850603e-06, + "loss": 1.2424, + "step": 33442 + }, + { + "epoch": 0.8, + "learning_rate": 1.943012602305391e-06, + "loss": 1.1578, + "step": 33443 + }, + { + "epoch": 0.8, + "learning_rate": 1.94255126770557e-06, + "loss": 0.9568, + "step": 33444 + }, + { + "epoch": 0.8, + "learning_rate": 1.942089981988393e-06, + "loss": 1.2091, + "step": 33445 + }, + { + "epoch": 0.8, + "learning_rate": 1.941628745156664e-06, + "loss": 0.6423, + "step": 33446 + }, + { + "epoch": 0.8, + "learning_rate": 1.9411675572131737e-06, + "loss": 1.1238, + "step": 33447 + }, + { + "epoch": 0.8, + "learning_rate": 1.940706418160726e-06, + "loss": 0.9851, + "step": 33448 + }, + { + "epoch": 0.8, + "learning_rate": 1.9402453280021195e-06, + "loss": 0.8631, + "step": 33449 + }, + { + "epoch": 0.8, + "learning_rate": 1.939784286740145e-06, + "loss": 1.1269, + "step": 33450 + }, + { + "epoch": 0.8, + "learning_rate": 1.939323294377604e-06, + "loss": 0.5914, + "step": 33451 + }, + { + "epoch": 0.8, + "learning_rate": 1.9388623509172945e-06, + "loss": 0.9579, + "step": 33452 + }, + { + "epoch": 0.8, + "learning_rate": 1.938401456362009e-06, + "loss": 1.2702, + "step": 33453 + }, + { + "epoch": 0.8, + "learning_rate": 1.937940610714546e-06, + "loss": 1.2231, + "step": 33454 + }, + { + "epoch": 0.8, + "learning_rate": 1.937479813977703e-06, + "loss": 1.0325, + "step": 33455 + }, + { + "epoch": 0.8, + "learning_rate": 1.937019066154272e-06, + "loss": 0.7876, + "step": 33456 + }, + { + "epoch": 0.8, + "learning_rate": 1.9365583672470512e-06, + "loss": 1.2617, + "step": 33457 + }, + { + "epoch": 0.8, + "learning_rate": 1.9360977172588326e-06, + "loss": 1.028, + "step": 33458 + }, + { + "epoch": 0.8, + "learning_rate": 1.9356371161924113e-06, + "loss": 0.9971, + "step": 33459 + }, + { + "epoch": 0.8, + "learning_rate": 1.935176564050586e-06, + "loss": 0.907, + "step": 33460 + }, + { + "epoch": 0.8, + "learning_rate": 1.934716060836145e-06, + "loss": 1.3137, + "step": 33461 + }, + { + "epoch": 0.8, + "learning_rate": 1.9342556065518834e-06, + "loss": 0.9235, + "step": 33462 + }, + { + "epoch": 0.8, + "learning_rate": 1.9337952012006e-06, + "loss": 1.0181, + "step": 33463 + }, + { + "epoch": 0.8, + "learning_rate": 1.9333348447850796e-06, + "loss": 1.1169, + "step": 33464 + }, + { + "epoch": 0.8, + "learning_rate": 1.9328745373081235e-06, + "loss": 1.1843, + "step": 33465 + }, + { + "epoch": 0.8, + "learning_rate": 1.9324142787725176e-06, + "loss": 1.0747, + "step": 33466 + }, + { + "epoch": 0.8, + "learning_rate": 1.931954069181056e-06, + "loss": 1.0229, + "step": 33467 + }, + { + "epoch": 0.8, + "learning_rate": 1.9314939085365335e-06, + "loss": 0.9471, + "step": 33468 + }, + { + "epoch": 0.8, + "learning_rate": 1.931033796841737e-06, + "loss": 0.926, + "step": 33469 + }, + { + "epoch": 0.8, + "learning_rate": 1.930573734099461e-06, + "loss": 0.7663, + "step": 33470 + }, + { + "epoch": 0.8, + "learning_rate": 1.9301137203124986e-06, + "loss": 1.0478, + "step": 33471 + }, + { + "epoch": 0.8, + "learning_rate": 1.929653755483636e-06, + "loss": 1.0815, + "step": 33472 + }, + { + "epoch": 0.8, + "learning_rate": 1.9291938396156675e-06, + "loss": 1.2047, + "step": 33473 + }, + { + "epoch": 0.81, + "learning_rate": 1.9287339727113796e-06, + "loss": 1.0668, + "step": 33474 + }, + { + "epoch": 0.81, + "learning_rate": 1.928274154773564e-06, + "loss": 0.9895, + "step": 33475 + }, + { + "epoch": 0.81, + "learning_rate": 1.9278143858050137e-06, + "loss": 1.1232, + "step": 33476 + }, + { + "epoch": 0.81, + "learning_rate": 1.9273546658085117e-06, + "loss": 0.8791, + "step": 33477 + }, + { + "epoch": 0.81, + "learning_rate": 1.92689499478685e-06, + "loss": 0.8891, + "step": 33478 + }, + { + "epoch": 0.81, + "learning_rate": 1.92643537274282e-06, + "loss": 1.1952, + "step": 33479 + }, + { + "epoch": 0.81, + "learning_rate": 1.925975799679205e-06, + "loss": 1.25, + "step": 33480 + }, + { + "epoch": 0.81, + "learning_rate": 1.925516275598799e-06, + "loss": 1.0633, + "step": 33481 + }, + { + "epoch": 0.81, + "learning_rate": 1.9250568005043835e-06, + "loss": 0.9859, + "step": 33482 + }, + { + "epoch": 0.81, + "learning_rate": 1.924597374398749e-06, + "loss": 1.1151, + "step": 33483 + }, + { + "epoch": 0.81, + "learning_rate": 1.9241379972846852e-06, + "loss": 1.2636, + "step": 33484 + }, + { + "epoch": 0.81, + "learning_rate": 1.9236786691649736e-06, + "loss": 1.1209, + "step": 33485 + }, + { + "epoch": 0.81, + "learning_rate": 1.9232193900424045e-06, + "loss": 1.2253, + "step": 33486 + }, + { + "epoch": 0.81, + "learning_rate": 1.9227601599197664e-06, + "loss": 0.9644, + "step": 33487 + }, + { + "epoch": 0.81, + "learning_rate": 1.9223009787998394e-06, + "loss": 1.0253, + "step": 33488 + }, + { + "epoch": 0.81, + "learning_rate": 1.9218418466854148e-06, + "loss": 1.0605, + "step": 33489 + }, + { + "epoch": 0.81, + "learning_rate": 1.9213827635792737e-06, + "loss": 0.9131, + "step": 33490 + }, + { + "epoch": 0.81, + "learning_rate": 1.920923729484203e-06, + "loss": 0.9192, + "step": 33491 + }, + { + "epoch": 0.81, + "learning_rate": 1.92046474440299e-06, + "loss": 0.9457, + "step": 33492 + }, + { + "epoch": 0.81, + "learning_rate": 1.9200058083384144e-06, + "loss": 1.0601, + "step": 33493 + }, + { + "epoch": 0.81, + "learning_rate": 1.919546921293264e-06, + "loss": 0.9967, + "step": 33494 + }, + { + "epoch": 0.81, + "learning_rate": 1.919088083270323e-06, + "loss": 0.9624, + "step": 33495 + }, + { + "epoch": 0.81, + "learning_rate": 1.918629294272372e-06, + "loss": 0.9571, + "step": 33496 + }, + { + "epoch": 0.81, + "learning_rate": 1.9181705543021958e-06, + "loss": 1.122, + "step": 33497 + }, + { + "epoch": 0.81, + "learning_rate": 1.917711863362581e-06, + "loss": 1.0491, + "step": 33498 + }, + { + "epoch": 0.81, + "learning_rate": 1.917253221456302e-06, + "loss": 0.8412, + "step": 33499 + }, + { + "epoch": 0.81, + "learning_rate": 1.9167946285861515e-06, + "loss": 1.0765, + "step": 33500 + }, + { + "epoch": 0.81, + "learning_rate": 1.916336084754904e-06, + "loss": 1.1718, + "step": 33501 + }, + { + "epoch": 0.81, + "learning_rate": 1.915877589965347e-06, + "loss": 1.29, + "step": 33502 + }, + { + "epoch": 0.81, + "learning_rate": 1.9154191442202564e-06, + "loss": 1.1068, + "step": 33503 + }, + { + "epoch": 0.81, + "learning_rate": 1.9149607475224164e-06, + "loss": 1.0494, + "step": 33504 + }, + { + "epoch": 0.81, + "learning_rate": 1.9145023998746093e-06, + "loss": 0.7885, + "step": 33505 + }, + { + "epoch": 0.81, + "learning_rate": 1.9140441012796128e-06, + "loss": 1.157, + "step": 33506 + }, + { + "epoch": 0.81, + "learning_rate": 1.9135858517402083e-06, + "loss": 0.8779, + "step": 33507 + }, + { + "epoch": 0.81, + "learning_rate": 1.91312765125918e-06, + "loss": 1.1747, + "step": 33508 + }, + { + "epoch": 0.81, + "learning_rate": 1.9126694998393003e-06, + "loss": 1.0506, + "step": 33509 + }, + { + "epoch": 0.81, + "learning_rate": 1.9122113974833558e-06, + "loss": 0.7995, + "step": 33510 + }, + { + "epoch": 0.81, + "learning_rate": 1.9117533441941194e-06, + "loss": 1.148, + "step": 33511 + }, + { + "epoch": 0.81, + "learning_rate": 1.9112953399743737e-06, + "loss": 0.996, + "step": 33512 + }, + { + "epoch": 0.81, + "learning_rate": 1.910837384826899e-06, + "loss": 0.9915, + "step": 33513 + }, + { + "epoch": 0.81, + "learning_rate": 1.910379478754468e-06, + "loss": 1.0708, + "step": 33514 + }, + { + "epoch": 0.81, + "learning_rate": 1.9099216217598628e-06, + "loss": 1.0395, + "step": 33515 + }, + { + "epoch": 0.81, + "learning_rate": 1.909463813845862e-06, + "loss": 1.0944, + "step": 33516 + }, + { + "epoch": 0.81, + "learning_rate": 1.909006055015239e-06, + "loss": 0.8447, + "step": 33517 + }, + { + "epoch": 0.81, + "learning_rate": 1.908548345270775e-06, + "loss": 1.1099, + "step": 33518 + }, + { + "epoch": 0.81, + "learning_rate": 1.9080906846152425e-06, + "loss": 0.9623, + "step": 33519 + }, + { + "epoch": 0.81, + "learning_rate": 1.9076330730514214e-06, + "loss": 1.2712, + "step": 33520 + }, + { + "epoch": 0.81, + "learning_rate": 1.907175510582089e-06, + "loss": 1.0465, + "step": 33521 + }, + { + "epoch": 0.81, + "learning_rate": 1.906717997210017e-06, + "loss": 1.2681, + "step": 33522 + }, + { + "epoch": 0.81, + "learning_rate": 1.9062605329379823e-06, + "loss": 1.009, + "step": 33523 + }, + { + "epoch": 0.81, + "learning_rate": 1.9058031177687652e-06, + "loss": 1.1575, + "step": 33524 + }, + { + "epoch": 0.81, + "learning_rate": 1.9053457517051332e-06, + "loss": 0.7844, + "step": 33525 + }, + { + "epoch": 0.81, + "learning_rate": 1.9048884347498674e-06, + "loss": 1.1073, + "step": 33526 + }, + { + "epoch": 0.81, + "learning_rate": 1.9044311669057369e-06, + "loss": 1.2293, + "step": 33527 + }, + { + "epoch": 0.81, + "learning_rate": 1.9039739481755182e-06, + "loss": 0.7995, + "step": 33528 + }, + { + "epoch": 0.81, + "learning_rate": 1.9035167785619879e-06, + "loss": 0.7742, + "step": 33529 + }, + { + "epoch": 0.81, + "learning_rate": 1.9030596580679139e-06, + "loss": 1.2015, + "step": 33530 + }, + { + "epoch": 0.81, + "learning_rate": 1.902602586696073e-06, + "loss": 0.9712, + "step": 33531 + }, + { + "epoch": 0.81, + "learning_rate": 1.9021455644492403e-06, + "loss": 1.0753, + "step": 33532 + }, + { + "epoch": 0.81, + "learning_rate": 1.9016885913301819e-06, + "loss": 0.8637, + "step": 33533 + }, + { + "epoch": 0.81, + "learning_rate": 1.9012316673416754e-06, + "loss": 1.0191, + "step": 33534 + }, + { + "epoch": 0.81, + "learning_rate": 1.900774792486494e-06, + "loss": 1.1791, + "step": 33535 + }, + { + "epoch": 0.81, + "learning_rate": 1.9003179667674032e-06, + "loss": 0.9908, + "step": 33536 + }, + { + "epoch": 0.81, + "learning_rate": 1.899861190187181e-06, + "loss": 0.8776, + "step": 33537 + }, + { + "epoch": 0.81, + "learning_rate": 1.899404462748593e-06, + "loss": 0.853, + "step": 33538 + }, + { + "epoch": 0.81, + "learning_rate": 1.8989477844544135e-06, + "loss": 1.2067, + "step": 33539 + }, + { + "epoch": 0.81, + "learning_rate": 1.8984911553074148e-06, + "loss": 0.9477, + "step": 33540 + }, + { + "epoch": 0.81, + "learning_rate": 1.8980345753103614e-06, + "loss": 1.0041, + "step": 33541 + }, + { + "epoch": 0.81, + "learning_rate": 1.8975780444660274e-06, + "loss": 0.8531, + "step": 33542 + }, + { + "epoch": 0.81, + "learning_rate": 1.8971215627771833e-06, + "loss": 1.0279, + "step": 33543 + }, + { + "epoch": 0.81, + "learning_rate": 1.8966651302465943e-06, + "loss": 1.0631, + "step": 33544 + }, + { + "epoch": 0.81, + "learning_rate": 1.8962087468770352e-06, + "loss": 0.7577, + "step": 33545 + }, + { + "epoch": 0.81, + "learning_rate": 1.895752412671269e-06, + "loss": 1.0847, + "step": 33546 + }, + { + "epoch": 0.81, + "learning_rate": 1.8952961276320658e-06, + "loss": 1.1097, + "step": 33547 + }, + { + "epoch": 0.81, + "learning_rate": 1.894839891762198e-06, + "loss": 1.2357, + "step": 33548 + }, + { + "epoch": 0.81, + "learning_rate": 1.8943837050644265e-06, + "loss": 1.0241, + "step": 33549 + }, + { + "epoch": 0.81, + "learning_rate": 1.8939275675415236e-06, + "loss": 1.0697, + "step": 33550 + }, + { + "epoch": 0.81, + "learning_rate": 1.8934714791962583e-06, + "loss": 0.9886, + "step": 33551 + }, + { + "epoch": 0.81, + "learning_rate": 1.8930154400313904e-06, + "loss": 1.0407, + "step": 33552 + }, + { + "epoch": 0.81, + "learning_rate": 1.8925594500496958e-06, + "loss": 1.2669, + "step": 33553 + }, + { + "epoch": 0.81, + "learning_rate": 1.892103509253932e-06, + "loss": 0.9411, + "step": 33554 + }, + { + "epoch": 0.81, + "learning_rate": 1.891647617646869e-06, + "loss": 1.208, + "step": 33555 + }, + { + "epoch": 0.81, + "learning_rate": 1.891191775231276e-06, + "loss": 1.0037, + "step": 33556 + }, + { + "epoch": 0.81, + "learning_rate": 1.8907359820099124e-06, + "loss": 1.0435, + "step": 33557 + }, + { + "epoch": 0.81, + "learning_rate": 1.8902802379855456e-06, + "loss": 0.9186, + "step": 33558 + }, + { + "epoch": 0.81, + "learning_rate": 1.8898245431609442e-06, + "loss": 0.9996, + "step": 33559 + }, + { + "epoch": 0.81, + "learning_rate": 1.889368897538868e-06, + "loss": 0.9679, + "step": 33560 + }, + { + "epoch": 0.81, + "learning_rate": 1.888913301122084e-06, + "loss": 1.1972, + "step": 33561 + }, + { + "epoch": 0.81, + "learning_rate": 1.888457753913353e-06, + "loss": 0.8714, + "step": 33562 + }, + { + "epoch": 0.81, + "learning_rate": 1.888002255915441e-06, + "loss": 0.7377, + "step": 33563 + }, + { + "epoch": 0.81, + "learning_rate": 1.8875468071311143e-06, + "loss": 0.996, + "step": 33564 + }, + { + "epoch": 0.81, + "learning_rate": 1.8870914075631297e-06, + "loss": 0.9935, + "step": 33565 + }, + { + "epoch": 0.81, + "learning_rate": 1.886636057214254e-06, + "loss": 0.917, + "step": 33566 + }, + { + "epoch": 0.81, + "learning_rate": 1.8861807560872502e-06, + "loss": 1.2098, + "step": 33567 + }, + { + "epoch": 0.81, + "learning_rate": 1.8857255041848776e-06, + "loss": 0.6804, + "step": 33568 + }, + { + "epoch": 0.81, + "learning_rate": 1.8852703015099027e-06, + "loss": 1.042, + "step": 33569 + }, + { + "epoch": 0.81, + "learning_rate": 1.8848151480650812e-06, + "loss": 1.2478, + "step": 33570 + }, + { + "epoch": 0.81, + "learning_rate": 1.8843600438531772e-06, + "loss": 1.1255, + "step": 33571 + }, + { + "epoch": 0.81, + "learning_rate": 1.8839049888769546e-06, + "loss": 1.1379, + "step": 33572 + }, + { + "epoch": 0.81, + "learning_rate": 1.8834499831391694e-06, + "loss": 0.9084, + "step": 33573 + }, + { + "epoch": 0.81, + "learning_rate": 1.8829950266425833e-06, + "loss": 1.0704, + "step": 33574 + }, + { + "epoch": 0.81, + "learning_rate": 1.8825401193899595e-06, + "loss": 0.9802, + "step": 33575 + }, + { + "epoch": 0.81, + "learning_rate": 1.8820852613840534e-06, + "loss": 1.2132, + "step": 33576 + }, + { + "epoch": 0.81, + "learning_rate": 1.8816304526276263e-06, + "loss": 0.7835, + "step": 33577 + }, + { + "epoch": 0.81, + "learning_rate": 1.8811756931234403e-06, + "loss": 1.0513, + "step": 33578 + }, + { + "epoch": 0.81, + "learning_rate": 1.8807209828742502e-06, + "loss": 1.1917, + "step": 33579 + }, + { + "epoch": 0.81, + "learning_rate": 1.8802663218828177e-06, + "loss": 1.0097, + "step": 33580 + }, + { + "epoch": 0.81, + "learning_rate": 1.8798117101518964e-06, + "loss": 1.0255, + "step": 33581 + }, + { + "epoch": 0.81, + "learning_rate": 1.8793571476842487e-06, + "loss": 1.0888, + "step": 33582 + }, + { + "epoch": 0.81, + "learning_rate": 1.8789026344826321e-06, + "loss": 0.9219, + "step": 33583 + }, + { + "epoch": 0.81, + "learning_rate": 1.8784481705498014e-06, + "loss": 0.9524, + "step": 33584 + }, + { + "epoch": 0.81, + "learning_rate": 1.8779937558885153e-06, + "loss": 1.2767, + "step": 33585 + }, + { + "epoch": 0.81, + "learning_rate": 1.8775393905015337e-06, + "loss": 1.0818, + "step": 33586 + }, + { + "epoch": 0.81, + "learning_rate": 1.877085074391607e-06, + "loss": 1.0632, + "step": 33587 + }, + { + "epoch": 0.81, + "learning_rate": 1.8766308075614969e-06, + "loss": 0.8307, + "step": 33588 + }, + { + "epoch": 0.81, + "learning_rate": 1.876176590013955e-06, + "loss": 1.0032, + "step": 33589 + }, + { + "epoch": 0.81, + "learning_rate": 1.875722421751739e-06, + "loss": 1.2158, + "step": 33590 + }, + { + "epoch": 0.81, + "learning_rate": 1.8752683027776064e-06, + "loss": 1.1032, + "step": 33591 + }, + { + "epoch": 0.81, + "learning_rate": 1.8748142330943064e-06, + "loss": 1.0365, + "step": 33592 + }, + { + "epoch": 0.81, + "learning_rate": 1.874360212704599e-06, + "loss": 0.7286, + "step": 33593 + }, + { + "epoch": 0.81, + "learning_rate": 1.8739062416112385e-06, + "loss": 0.7974, + "step": 33594 + }, + { + "epoch": 0.81, + "learning_rate": 1.8734523198169752e-06, + "loss": 0.9411, + "step": 33595 + }, + { + "epoch": 0.81, + "learning_rate": 1.8729984473245677e-06, + "loss": 0.7369, + "step": 33596 + }, + { + "epoch": 0.81, + "learning_rate": 1.872544624136764e-06, + "loss": 0.9937, + "step": 33597 + }, + { + "epoch": 0.81, + "learning_rate": 1.8720908502563207e-06, + "loss": 1.1061, + "step": 33598 + }, + { + "epoch": 0.81, + "learning_rate": 1.8716371256859932e-06, + "loss": 0.8798, + "step": 33599 + }, + { + "epoch": 0.81, + "learning_rate": 1.8711834504285287e-06, + "loss": 1.0818, + "step": 33600 + }, + { + "epoch": 0.81, + "learning_rate": 1.8707298244866822e-06, + "loss": 1.2288, + "step": 33601 + }, + { + "epoch": 0.81, + "learning_rate": 1.8702762478632075e-06, + "loss": 1.0738, + "step": 33602 + }, + { + "epoch": 0.81, + "learning_rate": 1.8698227205608532e-06, + "loss": 1.1225, + "step": 33603 + }, + { + "epoch": 0.81, + "learning_rate": 1.869369242582374e-06, + "loss": 1.2549, + "step": 33604 + }, + { + "epoch": 0.81, + "learning_rate": 1.8689158139305164e-06, + "loss": 1.003, + "step": 33605 + }, + { + "epoch": 0.81, + "learning_rate": 1.8684624346080349e-06, + "loss": 1.1181, + "step": 33606 + }, + { + "epoch": 0.81, + "learning_rate": 1.8680091046176807e-06, + "loss": 1.1015, + "step": 33607 + }, + { + "epoch": 0.81, + "learning_rate": 1.8675558239622005e-06, + "loss": 0.9483, + "step": 33608 + }, + { + "epoch": 0.81, + "learning_rate": 1.8671025926443464e-06, + "loss": 0.9582, + "step": 33609 + }, + { + "epoch": 0.81, + "learning_rate": 1.8666494106668709e-06, + "loss": 1.0903, + "step": 33610 + }, + { + "epoch": 0.81, + "learning_rate": 1.866196278032517e-06, + "loss": 0.9186, + "step": 33611 + }, + { + "epoch": 0.81, + "learning_rate": 1.8657431947440396e-06, + "loss": 1.0149, + "step": 33612 + }, + { + "epoch": 0.81, + "learning_rate": 1.8652901608041829e-06, + "loss": 0.9105, + "step": 33613 + }, + { + "epoch": 0.81, + "learning_rate": 1.864837176215697e-06, + "loss": 0.9559, + "step": 33614 + }, + { + "epoch": 0.81, + "learning_rate": 1.8643842409813329e-06, + "loss": 1.1628, + "step": 33615 + }, + { + "epoch": 0.81, + "learning_rate": 1.8639313551038329e-06, + "loss": 0.9392, + "step": 33616 + }, + { + "epoch": 0.81, + "learning_rate": 1.8634785185859482e-06, + "loss": 1.0546, + "step": 33617 + }, + { + "epoch": 0.81, + "learning_rate": 1.863025731430429e-06, + "loss": 0.8045, + "step": 33618 + }, + { + "epoch": 0.81, + "learning_rate": 1.862572993640015e-06, + "loss": 1.2965, + "step": 33619 + }, + { + "epoch": 0.81, + "learning_rate": 1.8621203052174596e-06, + "loss": 1.148, + "step": 33620 + }, + { + "epoch": 0.81, + "learning_rate": 1.8616676661655042e-06, + "loss": 1.11, + "step": 33621 + }, + { + "epoch": 0.81, + "learning_rate": 1.8612150764868952e-06, + "loss": 0.9935, + "step": 33622 + }, + { + "epoch": 0.81, + "learning_rate": 1.860762536184384e-06, + "loss": 1.0291, + "step": 33623 + }, + { + "epoch": 0.81, + "learning_rate": 1.8603100452607092e-06, + "loss": 1.1846, + "step": 33624 + }, + { + "epoch": 0.81, + "learning_rate": 1.859857603718619e-06, + "loss": 1.2995, + "step": 33625 + }, + { + "epoch": 0.81, + "learning_rate": 1.8594052115608596e-06, + "loss": 1.0311, + "step": 33626 + }, + { + "epoch": 0.81, + "learning_rate": 1.8589528687901725e-06, + "loss": 1.1358, + "step": 33627 + }, + { + "epoch": 0.81, + "learning_rate": 1.8585005754093021e-06, + "loss": 1.053, + "step": 33628 + }, + { + "epoch": 0.81, + "learning_rate": 1.8580483314209973e-06, + "loss": 0.9938, + "step": 33629 + }, + { + "epoch": 0.81, + "learning_rate": 1.857596136827996e-06, + "loss": 0.913, + "step": 33630 + }, + { + "epoch": 0.81, + "learning_rate": 1.8571439916330457e-06, + "loss": 1.1343, + "step": 33631 + }, + { + "epoch": 0.81, + "learning_rate": 1.8566918958388846e-06, + "loss": 1.0986, + "step": 33632 + }, + { + "epoch": 0.81, + "learning_rate": 1.856239849448259e-06, + "loss": 1.1611, + "step": 33633 + }, + { + "epoch": 0.81, + "learning_rate": 1.8557878524639128e-06, + "loss": 1.0106, + "step": 33634 + }, + { + "epoch": 0.81, + "learning_rate": 1.8553359048885834e-06, + "loss": 0.8872, + "step": 33635 + }, + { + "epoch": 0.81, + "learning_rate": 1.8548840067250162e-06, + "loss": 1.0816, + "step": 33636 + }, + { + "epoch": 0.81, + "learning_rate": 1.854432157975954e-06, + "loss": 1.0618, + "step": 33637 + }, + { + "epoch": 0.81, + "learning_rate": 1.853980358644133e-06, + "loss": 0.909, + "step": 33638 + }, + { + "epoch": 0.81, + "learning_rate": 1.8535286087323001e-06, + "loss": 1.0814, + "step": 33639 + }, + { + "epoch": 0.81, + "learning_rate": 1.85307690824319e-06, + "loss": 1.2147, + "step": 33640 + }, + { + "epoch": 0.81, + "learning_rate": 1.8526252571795455e-06, + "loss": 1.0297, + "step": 33641 + }, + { + "epoch": 0.81, + "learning_rate": 1.8521736555441105e-06, + "loss": 1.1733, + "step": 33642 + }, + { + "epoch": 0.81, + "learning_rate": 1.8517221033396194e-06, + "loss": 1.1797, + "step": 33643 + }, + { + "epoch": 0.81, + "learning_rate": 1.851270600568813e-06, + "loss": 0.8586, + "step": 33644 + }, + { + "epoch": 0.81, + "learning_rate": 1.850819147234434e-06, + "loss": 0.9534, + "step": 33645 + }, + { + "epoch": 0.81, + "learning_rate": 1.8503677433392154e-06, + "loss": 0.954, + "step": 33646 + }, + { + "epoch": 0.81, + "learning_rate": 1.8499163888859018e-06, + "loss": 1.0814, + "step": 33647 + }, + { + "epoch": 0.81, + "learning_rate": 1.8494650838772266e-06, + "loss": 1.1305, + "step": 33648 + }, + { + "epoch": 0.81, + "learning_rate": 1.8490138283159287e-06, + "loss": 1.1818, + "step": 33649 + }, + { + "epoch": 0.81, + "learning_rate": 1.8485626222047492e-06, + "loss": 0.6459, + "step": 33650 + }, + { + "epoch": 0.81, + "learning_rate": 1.8481114655464204e-06, + "loss": 1.3287, + "step": 33651 + }, + { + "epoch": 0.81, + "learning_rate": 1.8476603583436814e-06, + "loss": 1.085, + "step": 33652 + }, + { + "epoch": 0.81, + "learning_rate": 1.847209300599273e-06, + "loss": 1.122, + "step": 33653 + }, + { + "epoch": 0.81, + "learning_rate": 1.8467582923159254e-06, + "loss": 0.791, + "step": 33654 + }, + { + "epoch": 0.81, + "learning_rate": 1.84630733349638e-06, + "loss": 1.216, + "step": 33655 + }, + { + "epoch": 0.81, + "learning_rate": 1.8458564241433674e-06, + "loss": 1.117, + "step": 33656 + }, + { + "epoch": 0.81, + "learning_rate": 1.8454055642596258e-06, + "loss": 0.8321, + "step": 33657 + }, + { + "epoch": 0.81, + "learning_rate": 1.8449547538478929e-06, + "loss": 0.8788, + "step": 33658 + }, + { + "epoch": 0.81, + "learning_rate": 1.8445039929108988e-06, + "loss": 0.9889, + "step": 33659 + }, + { + "epoch": 0.81, + "learning_rate": 1.8440532814513812e-06, + "loss": 1.1483, + "step": 33660 + }, + { + "epoch": 0.81, + "learning_rate": 1.8436026194720768e-06, + "loss": 1.0227, + "step": 33661 + }, + { + "epoch": 0.81, + "learning_rate": 1.8431520069757135e-06, + "loss": 1.0552, + "step": 33662 + }, + { + "epoch": 0.81, + "learning_rate": 1.8427014439650314e-06, + "loss": 1.0589, + "step": 33663 + }, + { + "epoch": 0.81, + "learning_rate": 1.8422509304427583e-06, + "loss": 1.2411, + "step": 33664 + }, + { + "epoch": 0.81, + "learning_rate": 1.8418004664116295e-06, + "loss": 1.0198, + "step": 33665 + }, + { + "epoch": 0.81, + "learning_rate": 1.8413500518743788e-06, + "loss": 0.8855, + "step": 33666 + }, + { + "epoch": 0.81, + "learning_rate": 1.840899686833738e-06, + "loss": 1.0166, + "step": 33667 + }, + { + "epoch": 0.81, + "learning_rate": 1.840449371292442e-06, + "loss": 1.1584, + "step": 33668 + }, + { + "epoch": 0.81, + "learning_rate": 1.8399991052532184e-06, + "loss": 0.9342, + "step": 33669 + }, + { + "epoch": 0.81, + "learning_rate": 1.8395488887188007e-06, + "loss": 0.9722, + "step": 33670 + }, + { + "epoch": 0.81, + "learning_rate": 1.8390987216919232e-06, + "loss": 1.2541, + "step": 33671 + }, + { + "epoch": 0.81, + "learning_rate": 1.8386486041753114e-06, + "loss": 0.8942, + "step": 33672 + }, + { + "epoch": 0.81, + "learning_rate": 1.838198536171699e-06, + "loss": 0.8059, + "step": 33673 + }, + { + "epoch": 0.81, + "learning_rate": 1.8377485176838194e-06, + "loss": 0.904, + "step": 33674 + }, + { + "epoch": 0.81, + "learning_rate": 1.8372985487143956e-06, + "loss": 1.0248, + "step": 33675 + }, + { + "epoch": 0.81, + "learning_rate": 1.8368486292661657e-06, + "loss": 1.1722, + "step": 33676 + }, + { + "epoch": 0.81, + "learning_rate": 1.8363987593418519e-06, + "loss": 1.0399, + "step": 33677 + }, + { + "epoch": 0.81, + "learning_rate": 1.8359489389441865e-06, + "loss": 0.8435, + "step": 33678 + }, + { + "epoch": 0.81, + "learning_rate": 1.8354991680759005e-06, + "loss": 0.8562, + "step": 33679 + }, + { + "epoch": 0.81, + "learning_rate": 1.8350494467397196e-06, + "loss": 0.9881, + "step": 33680 + }, + { + "epoch": 0.81, + "learning_rate": 1.8345997749383715e-06, + "loss": 1.1677, + "step": 33681 + }, + { + "epoch": 0.81, + "learning_rate": 1.8341501526745897e-06, + "loss": 0.9693, + "step": 33682 + }, + { + "epoch": 0.81, + "learning_rate": 1.8337005799510954e-06, + "loss": 0.8535, + "step": 33683 + }, + { + "epoch": 0.81, + "learning_rate": 1.8332510567706208e-06, + "loss": 1.1793, + "step": 33684 + }, + { + "epoch": 0.81, + "learning_rate": 1.8328015831358893e-06, + "loss": 0.9892, + "step": 33685 + }, + { + "epoch": 0.81, + "learning_rate": 1.8323521590496296e-06, + "loss": 1.0146, + "step": 33686 + }, + { + "epoch": 0.81, + "learning_rate": 1.83190278451457e-06, + "loss": 1.1029, + "step": 33687 + }, + { + "epoch": 0.81, + "learning_rate": 1.8314534595334332e-06, + "loss": 1.1096, + "step": 33688 + }, + { + "epoch": 0.81, + "learning_rate": 1.8310041841089465e-06, + "loss": 0.9911, + "step": 33689 + }, + { + "epoch": 0.81, + "learning_rate": 1.8305549582438387e-06, + "loss": 0.8885, + "step": 33690 + }, + { + "epoch": 0.81, + "learning_rate": 1.8301057819408297e-06, + "loss": 1.0547, + "step": 33691 + }, + { + "epoch": 0.81, + "learning_rate": 1.8296566552026495e-06, + "loss": 0.7951, + "step": 33692 + }, + { + "epoch": 0.81, + "learning_rate": 1.8292075780320184e-06, + "loss": 0.8271, + "step": 33693 + }, + { + "epoch": 0.81, + "learning_rate": 1.828758550431663e-06, + "loss": 0.9968, + "step": 33694 + }, + { + "epoch": 0.81, + "learning_rate": 1.82830957240431e-06, + "loss": 0.802, + "step": 33695 + }, + { + "epoch": 0.81, + "learning_rate": 1.827860643952678e-06, + "loss": 1.0219, + "step": 33696 + }, + { + "epoch": 0.81, + "learning_rate": 1.827411765079493e-06, + "loss": 1.2794, + "step": 33697 + }, + { + "epoch": 0.81, + "learning_rate": 1.8269629357874818e-06, + "loss": 1.0855, + "step": 33698 + }, + { + "epoch": 0.81, + "learning_rate": 1.826514156079361e-06, + "loss": 1.0596, + "step": 33699 + }, + { + "epoch": 0.81, + "learning_rate": 1.8260654259578592e-06, + "loss": 1.2824, + "step": 33700 + }, + { + "epoch": 0.81, + "learning_rate": 1.8256167454256934e-06, + "loss": 0.8392, + "step": 33701 + }, + { + "epoch": 0.81, + "learning_rate": 1.8251681144855882e-06, + "loss": 1.0578, + "step": 33702 + }, + { + "epoch": 0.81, + "learning_rate": 1.8247195331402666e-06, + "loss": 1.0587, + "step": 33703 + }, + { + "epoch": 0.81, + "learning_rate": 1.8242710013924469e-06, + "loss": 1.0859, + "step": 33704 + }, + { + "epoch": 0.81, + "learning_rate": 1.8238225192448523e-06, + "loss": 1.2922, + "step": 33705 + }, + { + "epoch": 0.81, + "learning_rate": 1.823374086700206e-06, + "loss": 0.9849, + "step": 33706 + }, + { + "epoch": 0.81, + "learning_rate": 1.8229257037612225e-06, + "loss": 1.0162, + "step": 33707 + }, + { + "epoch": 0.81, + "learning_rate": 1.8224773704306254e-06, + "loss": 0.8271, + "step": 33708 + }, + { + "epoch": 0.81, + "learning_rate": 1.8220290867111379e-06, + "loss": 1.1028, + "step": 33709 + }, + { + "epoch": 0.81, + "learning_rate": 1.8215808526054734e-06, + "loss": 1.1198, + "step": 33710 + }, + { + "epoch": 0.81, + "learning_rate": 1.8211326681163566e-06, + "loss": 0.8944, + "step": 33711 + }, + { + "epoch": 0.81, + "learning_rate": 1.8206845332465017e-06, + "loss": 0.8662, + "step": 33712 + }, + { + "epoch": 0.81, + "learning_rate": 1.8202364479986302e-06, + "loss": 1.1257, + "step": 33713 + }, + { + "epoch": 0.81, + "learning_rate": 1.8197884123754616e-06, + "loss": 0.8752, + "step": 33714 + }, + { + "epoch": 0.81, + "learning_rate": 1.8193404263797109e-06, + "loss": 0.8391, + "step": 33715 + }, + { + "epoch": 0.81, + "learning_rate": 1.8188924900140982e-06, + "loss": 1.2115, + "step": 33716 + }, + { + "epoch": 0.81, + "learning_rate": 1.818444603281342e-06, + "loss": 0.8651, + "step": 33717 + }, + { + "epoch": 0.81, + "learning_rate": 1.8179967661841557e-06, + "loss": 0.6927, + "step": 33718 + }, + { + "epoch": 0.81, + "learning_rate": 1.817548978725261e-06, + "loss": 1.0762, + "step": 33719 + }, + { + "epoch": 0.81, + "learning_rate": 1.8171012409073696e-06, + "loss": 1.1093, + "step": 33720 + }, + { + "epoch": 0.81, + "learning_rate": 1.8166535527331997e-06, + "loss": 1.0644, + "step": 33721 + }, + { + "epoch": 0.81, + "learning_rate": 1.816205914205471e-06, + "loss": 0.7191, + "step": 33722 + }, + { + "epoch": 0.81, + "learning_rate": 1.815758325326894e-06, + "loss": 0.9881, + "step": 33723 + }, + { + "epoch": 0.81, + "learning_rate": 1.8153107861001851e-06, + "loss": 1.1486, + "step": 33724 + }, + { + "epoch": 0.81, + "learning_rate": 1.8148632965280632e-06, + "loss": 1.1169, + "step": 33725 + }, + { + "epoch": 0.81, + "learning_rate": 1.8144158566132386e-06, + "loss": 1.0732, + "step": 33726 + }, + { + "epoch": 0.81, + "learning_rate": 1.8139684663584301e-06, + "loss": 0.9601, + "step": 33727 + }, + { + "epoch": 0.81, + "learning_rate": 1.8135211257663464e-06, + "loss": 1.0638, + "step": 33728 + }, + { + "epoch": 0.81, + "learning_rate": 1.8130738348397048e-06, + "loss": 0.8559, + "step": 33729 + }, + { + "epoch": 0.81, + "learning_rate": 1.8126265935812215e-06, + "loss": 1.1271, + "step": 33730 + }, + { + "epoch": 0.81, + "learning_rate": 1.8121794019936034e-06, + "loss": 0.9496, + "step": 33731 + }, + { + "epoch": 0.81, + "learning_rate": 1.8117322600795672e-06, + "loss": 0.9107, + "step": 33732 + }, + { + "epoch": 0.81, + "learning_rate": 1.8112851678418286e-06, + "loss": 1.1815, + "step": 33733 + }, + { + "epoch": 0.81, + "learning_rate": 1.810838125283093e-06, + "loss": 1.0289, + "step": 33734 + }, + { + "epoch": 0.81, + "learning_rate": 1.8103911324060797e-06, + "loss": 1.0654, + "step": 33735 + }, + { + "epoch": 0.81, + "learning_rate": 1.809944189213495e-06, + "loss": 1.1407, + "step": 33736 + }, + { + "epoch": 0.81, + "learning_rate": 1.8094972957080514e-06, + "loss": 1.1572, + "step": 33737 + }, + { + "epoch": 0.81, + "learning_rate": 1.8090504518924645e-06, + "loss": 1.0742, + "step": 33738 + }, + { + "epoch": 0.81, + "learning_rate": 1.8086036577694388e-06, + "loss": 0.9278, + "step": 33739 + }, + { + "epoch": 0.81, + "learning_rate": 1.8081569133416888e-06, + "loss": 0.9293, + "step": 33740 + }, + { + "epoch": 0.81, + "learning_rate": 1.8077102186119255e-06, + "loss": 0.7601, + "step": 33741 + }, + { + "epoch": 0.81, + "learning_rate": 1.8072635735828547e-06, + "loss": 1.1379, + "step": 33742 + }, + { + "epoch": 0.81, + "learning_rate": 1.806816978257192e-06, + "loss": 1.1235, + "step": 33743 + }, + { + "epoch": 0.81, + "learning_rate": 1.8063704326376408e-06, + "loss": 1.153, + "step": 33744 + }, + { + "epoch": 0.81, + "learning_rate": 1.8059239367269132e-06, + "loss": 1.1509, + "step": 33745 + }, + { + "epoch": 0.81, + "learning_rate": 1.8054774905277206e-06, + "loss": 1.1842, + "step": 33746 + }, + { + "epoch": 0.81, + "learning_rate": 1.805031094042765e-06, + "loss": 0.8185, + "step": 33747 + }, + { + "epoch": 0.81, + "learning_rate": 1.8045847472747592e-06, + "loss": 1.0687, + "step": 33748 + }, + { + "epoch": 0.81, + "learning_rate": 1.8041384502264114e-06, + "loss": 0.9578, + "step": 33749 + }, + { + "epoch": 0.81, + "learning_rate": 1.8036922029004266e-06, + "loss": 0.9348, + "step": 33750 + }, + { + "epoch": 0.81, + "learning_rate": 1.8032460052995126e-06, + "loss": 1.3587, + "step": 33751 + }, + { + "epoch": 0.81, + "learning_rate": 1.8027998574263794e-06, + "loss": 1.0128, + "step": 33752 + }, + { + "epoch": 0.81, + "learning_rate": 1.8023537592837293e-06, + "loss": 1.0022, + "step": 33753 + }, + { + "epoch": 0.81, + "learning_rate": 1.8019077108742734e-06, + "loss": 1.128, + "step": 33754 + }, + { + "epoch": 0.81, + "learning_rate": 1.8014617122007117e-06, + "loss": 1.2245, + "step": 33755 + }, + { + "epoch": 0.81, + "learning_rate": 1.8010157632657544e-06, + "loss": 1.1595, + "step": 33756 + }, + { + "epoch": 0.81, + "learning_rate": 1.800569864072108e-06, + "loss": 1.0391, + "step": 33757 + }, + { + "epoch": 0.81, + "learning_rate": 1.800124014622473e-06, + "loss": 1.005, + "step": 33758 + }, + { + "epoch": 0.81, + "learning_rate": 1.7996782149195568e-06, + "loss": 0.919, + "step": 33759 + }, + { + "epoch": 0.81, + "learning_rate": 1.7992324649660664e-06, + "loss": 1.153, + "step": 33760 + }, + { + "epoch": 0.81, + "learning_rate": 1.7987867647647017e-06, + "loss": 0.6522, + "step": 33761 + }, + { + "epoch": 0.81, + "learning_rate": 1.7983411143181717e-06, + "loss": 1.2103, + "step": 33762 + }, + { + "epoch": 0.81, + "learning_rate": 1.7978955136291732e-06, + "loss": 0.9728, + "step": 33763 + }, + { + "epoch": 0.81, + "learning_rate": 1.7974499627004138e-06, + "loss": 1.1118, + "step": 33764 + }, + { + "epoch": 0.81, + "learning_rate": 1.7970044615345994e-06, + "loss": 0.9068, + "step": 33765 + }, + { + "epoch": 0.81, + "learning_rate": 1.7965590101344254e-06, + "loss": 1.061, + "step": 33766 + }, + { + "epoch": 0.81, + "learning_rate": 1.7961136085025998e-06, + "loss": 1.0521, + "step": 33767 + }, + { + "epoch": 0.81, + "learning_rate": 1.7956682566418249e-06, + "loss": 1.0739, + "step": 33768 + }, + { + "epoch": 0.81, + "learning_rate": 1.7952229545547994e-06, + "loss": 1.1846, + "step": 33769 + }, + { + "epoch": 0.81, + "learning_rate": 1.7947777022442281e-06, + "loss": 1.2535, + "step": 33770 + }, + { + "epoch": 0.81, + "learning_rate": 1.7943324997128076e-06, + "loss": 1.0989, + "step": 33771 + }, + { + "epoch": 0.81, + "learning_rate": 1.7938873469632422e-06, + "loss": 1.3695, + "step": 33772 + }, + { + "epoch": 0.81, + "learning_rate": 1.7934422439982346e-06, + "loss": 0.982, + "step": 33773 + }, + { + "epoch": 0.81, + "learning_rate": 1.7929971908204802e-06, + "loss": 1.2291, + "step": 33774 + }, + { + "epoch": 0.81, + "learning_rate": 1.7925521874326813e-06, + "loss": 1.0879, + "step": 33775 + }, + { + "epoch": 0.81, + "learning_rate": 1.7921072338375411e-06, + "loss": 0.943, + "step": 33776 + }, + { + "epoch": 0.81, + "learning_rate": 1.7916623300377523e-06, + "loss": 1.0958, + "step": 33777 + }, + { + "epoch": 0.81, + "learning_rate": 1.79121747603602e-06, + "loss": 0.9704, + "step": 33778 + }, + { + "epoch": 0.81, + "learning_rate": 1.7907726718350383e-06, + "loss": 1.0595, + "step": 33779 + }, + { + "epoch": 0.81, + "learning_rate": 1.7903279174375065e-06, + "loss": 1.0862, + "step": 33780 + }, + { + "epoch": 0.81, + "learning_rate": 1.7898832128461286e-06, + "loss": 0.9226, + "step": 33781 + }, + { + "epoch": 0.81, + "learning_rate": 1.7894385580635943e-06, + "loss": 1.188, + "step": 33782 + }, + { + "epoch": 0.81, + "learning_rate": 1.788993953092606e-06, + "loss": 1.2191, + "step": 33783 + }, + { + "epoch": 0.81, + "learning_rate": 1.7885493979358616e-06, + "loss": 1.0946, + "step": 33784 + }, + { + "epoch": 0.81, + "learning_rate": 1.7881048925960542e-06, + "loss": 1.1501, + "step": 33785 + }, + { + "epoch": 0.81, + "learning_rate": 1.7876604370758855e-06, + "loss": 0.9908, + "step": 33786 + }, + { + "epoch": 0.81, + "learning_rate": 1.7872160313780474e-06, + "loss": 1.0975, + "step": 33787 + }, + { + "epoch": 0.81, + "learning_rate": 1.7867716755052378e-06, + "loss": 0.8625, + "step": 33788 + }, + { + "epoch": 0.81, + "learning_rate": 1.7863273694601547e-06, + "loss": 1.1824, + "step": 33789 + }, + { + "epoch": 0.81, + "learning_rate": 1.785883113245488e-06, + "loss": 1.0594, + "step": 33790 + }, + { + "epoch": 0.81, + "learning_rate": 1.7854389068639378e-06, + "loss": 0.7525, + "step": 33791 + }, + { + "epoch": 0.81, + "learning_rate": 1.7849947503181996e-06, + "loss": 1.0248, + "step": 33792 + }, + { + "epoch": 0.81, + "learning_rate": 1.7845506436109628e-06, + "loss": 1.1296, + "step": 33793 + }, + { + "epoch": 0.81, + "learning_rate": 1.7841065867449258e-06, + "loss": 0.9921, + "step": 33794 + }, + { + "epoch": 0.81, + "learning_rate": 1.7836625797227837e-06, + "loss": 1.0362, + "step": 33795 + }, + { + "epoch": 0.81, + "learning_rate": 1.783218622547226e-06, + "loss": 1.0019, + "step": 33796 + }, + { + "epoch": 0.81, + "learning_rate": 1.7827747152209506e-06, + "loss": 1.2995, + "step": 33797 + }, + { + "epoch": 0.81, + "learning_rate": 1.7823308577466459e-06, + "loss": 1.2113, + "step": 33798 + }, + { + "epoch": 0.81, + "learning_rate": 1.7818870501270059e-06, + "loss": 1.0697, + "step": 33799 + }, + { + "epoch": 0.81, + "learning_rate": 1.7814432923647274e-06, + "loss": 0.9523, + "step": 33800 + }, + { + "epoch": 0.81, + "learning_rate": 1.7809995844624973e-06, + "loss": 1.2111, + "step": 33801 + }, + { + "epoch": 0.81, + "learning_rate": 1.7805559264230088e-06, + "loss": 0.7922, + "step": 33802 + }, + { + "epoch": 0.81, + "learning_rate": 1.7801123182489567e-06, + "loss": 1.0287, + "step": 33803 + }, + { + "epoch": 0.81, + "learning_rate": 1.7796687599430273e-06, + "loss": 0.9449, + "step": 33804 + }, + { + "epoch": 0.81, + "learning_rate": 1.7792252515079166e-06, + "loss": 0.98, + "step": 33805 + }, + { + "epoch": 0.81, + "learning_rate": 1.7787817929463103e-06, + "loss": 0.8544, + "step": 33806 + }, + { + "epoch": 0.81, + "learning_rate": 1.7783383842609015e-06, + "loss": 1.2076, + "step": 33807 + }, + { + "epoch": 0.81, + "learning_rate": 1.7778950254543815e-06, + "loss": 1.0187, + "step": 33808 + }, + { + "epoch": 0.81, + "learning_rate": 1.7774517165294357e-06, + "loss": 0.9529, + "step": 33809 + }, + { + "epoch": 0.81, + "learning_rate": 1.777008457488757e-06, + "loss": 1.1253, + "step": 33810 + }, + { + "epoch": 0.81, + "learning_rate": 1.776565248335036e-06, + "loss": 1.032, + "step": 33811 + }, + { + "epoch": 0.81, + "learning_rate": 1.7761220890709574e-06, + "loss": 1.1125, + "step": 33812 + }, + { + "epoch": 0.81, + "learning_rate": 1.7756789796992135e-06, + "loss": 1.0474, + "step": 33813 + }, + { + "epoch": 0.81, + "learning_rate": 1.775235920222489e-06, + "loss": 1.1284, + "step": 33814 + }, + { + "epoch": 0.81, + "learning_rate": 1.7747929106434736e-06, + "loss": 0.9825, + "step": 33815 + }, + { + "epoch": 0.81, + "learning_rate": 1.7743499509648565e-06, + "loss": 1.1529, + "step": 33816 + }, + { + "epoch": 0.81, + "learning_rate": 1.773907041189321e-06, + "loss": 0.8531, + "step": 33817 + }, + { + "epoch": 0.81, + "learning_rate": 1.7734641813195574e-06, + "loss": 1.11, + "step": 33818 + }, + { + "epoch": 0.81, + "learning_rate": 1.7730213713582534e-06, + "loss": 1.2221, + "step": 33819 + }, + { + "epoch": 0.81, + "learning_rate": 1.7725786113080924e-06, + "loss": 1.0446, + "step": 33820 + }, + { + "epoch": 0.81, + "learning_rate": 1.772135901171763e-06, + "loss": 0.8875, + "step": 33821 + }, + { + "epoch": 0.81, + "learning_rate": 1.7716932409519472e-06, + "loss": 0.9946, + "step": 33822 + }, + { + "epoch": 0.81, + "learning_rate": 1.7712506306513333e-06, + "loss": 1.1963, + "step": 33823 + }, + { + "epoch": 0.81, + "learning_rate": 1.7708080702726094e-06, + "loss": 0.9871, + "step": 33824 + }, + { + "epoch": 0.81, + "learning_rate": 1.7703655598184544e-06, + "loss": 1.1029, + "step": 33825 + }, + { + "epoch": 0.81, + "learning_rate": 1.7699230992915562e-06, + "loss": 1.0375, + "step": 33826 + }, + { + "epoch": 0.81, + "learning_rate": 1.7694806886946003e-06, + "loss": 0.8951, + "step": 33827 + }, + { + "epoch": 0.81, + "learning_rate": 1.769038328030267e-06, + "loss": 1.1339, + "step": 33828 + }, + { + "epoch": 0.81, + "learning_rate": 1.768596017301245e-06, + "loss": 0.8567, + "step": 33829 + }, + { + "epoch": 0.81, + "learning_rate": 1.7681537565102124e-06, + "loss": 1.1099, + "step": 33830 + }, + { + "epoch": 0.81, + "learning_rate": 1.7677115456598548e-06, + "loss": 0.7441, + "step": 33831 + }, + { + "epoch": 0.81, + "learning_rate": 1.7672693847528544e-06, + "loss": 1.0953, + "step": 33832 + }, + { + "epoch": 0.81, + "learning_rate": 1.7668272737918946e-06, + "loss": 1.4083, + "step": 33833 + }, + { + "epoch": 0.81, + "learning_rate": 1.7663852127796588e-06, + "loss": 0.9675, + "step": 33834 + }, + { + "epoch": 0.81, + "learning_rate": 1.765943201718826e-06, + "loss": 0.9746, + "step": 33835 + }, + { + "epoch": 0.81, + "learning_rate": 1.7655012406120776e-06, + "loss": 1.1764, + "step": 33836 + }, + { + "epoch": 0.81, + "learning_rate": 1.7650593294620999e-06, + "loss": 1.0074, + "step": 33837 + }, + { + "epoch": 0.81, + "learning_rate": 1.7646174682715677e-06, + "loss": 0.98, + "step": 33838 + }, + { + "epoch": 0.81, + "learning_rate": 1.7641756570431634e-06, + "loss": 0.915, + "step": 33839 + }, + { + "epoch": 0.81, + "learning_rate": 1.7637338957795714e-06, + "loss": 0.8501, + "step": 33840 + }, + { + "epoch": 0.81, + "learning_rate": 1.763292184483466e-06, + "loss": 1.1748, + "step": 33841 + }, + { + "epoch": 0.81, + "learning_rate": 1.7628505231575321e-06, + "loss": 0.9933, + "step": 33842 + }, + { + "epoch": 0.81, + "learning_rate": 1.762408911804444e-06, + "loss": 0.9587, + "step": 33843 + }, + { + "epoch": 0.81, + "learning_rate": 1.761967350426883e-06, + "loss": 1.0804, + "step": 33844 + }, + { + "epoch": 0.81, + "learning_rate": 1.7615258390275313e-06, + "loss": 1.2721, + "step": 33845 + }, + { + "epoch": 0.81, + "learning_rate": 1.7610843776090626e-06, + "loss": 0.965, + "step": 33846 + }, + { + "epoch": 0.81, + "learning_rate": 1.7606429661741565e-06, + "loss": 0.9601, + "step": 33847 + }, + { + "epoch": 0.81, + "learning_rate": 1.7602016047254943e-06, + "loss": 1.1097, + "step": 33848 + }, + { + "epoch": 0.81, + "learning_rate": 1.7597602932657476e-06, + "loss": 1.0713, + "step": 33849 + }, + { + "epoch": 0.81, + "learning_rate": 1.7593190317975994e-06, + "loss": 1.0288, + "step": 33850 + }, + { + "epoch": 0.81, + "learning_rate": 1.7588778203237233e-06, + "loss": 1.1392, + "step": 33851 + }, + { + "epoch": 0.81, + "learning_rate": 1.758436658846796e-06, + "loss": 0.9011, + "step": 33852 + }, + { + "epoch": 0.81, + "learning_rate": 1.7579955473694976e-06, + "loss": 1.2621, + "step": 33853 + }, + { + "epoch": 0.81, + "learning_rate": 1.7575544858944982e-06, + "loss": 0.8804, + "step": 33854 + }, + { + "epoch": 0.81, + "learning_rate": 1.7571134744244777e-06, + "loss": 1.2335, + "step": 33855 + }, + { + "epoch": 0.81, + "learning_rate": 1.7566725129621143e-06, + "loss": 1.1049, + "step": 33856 + }, + { + "epoch": 0.81, + "learning_rate": 1.7562316015100756e-06, + "loss": 0.9138, + "step": 33857 + }, + { + "epoch": 0.81, + "learning_rate": 1.755790740071045e-06, + "loss": 0.9572, + "step": 33858 + }, + { + "epoch": 0.81, + "learning_rate": 1.7553499286476895e-06, + "loss": 1.0797, + "step": 33859 + }, + { + "epoch": 0.81, + "learning_rate": 1.7549091672426866e-06, + "loss": 0.9761, + "step": 33860 + }, + { + "epoch": 0.81, + "learning_rate": 1.7544684558587122e-06, + "loss": 0.9919, + "step": 33861 + }, + { + "epoch": 0.81, + "learning_rate": 1.7540277944984362e-06, + "loss": 0.988, + "step": 33862 + }, + { + "epoch": 0.81, + "learning_rate": 1.7535871831645345e-06, + "loss": 0.9169, + "step": 33863 + }, + { + "epoch": 0.81, + "learning_rate": 1.7531466218596826e-06, + "loss": 0.876, + "step": 33864 + }, + { + "epoch": 0.81, + "learning_rate": 1.7527061105865473e-06, + "loss": 1.0923, + "step": 33865 + }, + { + "epoch": 0.81, + "learning_rate": 1.7522656493478063e-06, + "loss": 0.8783, + "step": 33866 + }, + { + "epoch": 0.81, + "learning_rate": 1.7518252381461276e-06, + "loss": 0.6714, + "step": 33867 + }, + { + "epoch": 0.81, + "learning_rate": 1.7513848769841846e-06, + "loss": 1.1225, + "step": 33868 + }, + { + "epoch": 0.81, + "learning_rate": 1.7509445658646517e-06, + "loss": 1.0924, + "step": 33869 + }, + { + "epoch": 0.81, + "learning_rate": 1.750504304790196e-06, + "loss": 1.1335, + "step": 33870 + }, + { + "epoch": 0.81, + "learning_rate": 1.7500640937634906e-06, + "loss": 1.0147, + "step": 33871 + }, + { + "epoch": 0.81, + "learning_rate": 1.7496239327872078e-06, + "loss": 1.0927, + "step": 33872 + }, + { + "epoch": 0.81, + "learning_rate": 1.7491838218640134e-06, + "loss": 1.2309, + "step": 33873 + }, + { + "epoch": 0.81, + "learning_rate": 1.7487437609965796e-06, + "loss": 1.0538, + "step": 33874 + }, + { + "epoch": 0.81, + "learning_rate": 1.74830375018758e-06, + "loss": 0.8491, + "step": 33875 + }, + { + "epoch": 0.81, + "learning_rate": 1.7478637894396777e-06, + "loss": 1.2069, + "step": 33876 + }, + { + "epoch": 0.81, + "learning_rate": 1.7474238787555475e-06, + "loss": 1.0629, + "step": 33877 + }, + { + "epoch": 0.81, + "learning_rate": 1.7469840181378527e-06, + "loss": 1.3327, + "step": 33878 + }, + { + "epoch": 0.81, + "learning_rate": 1.7465442075892646e-06, + "loss": 1.1322, + "step": 33879 + }, + { + "epoch": 0.81, + "learning_rate": 1.7461044471124545e-06, + "loss": 1.2213, + "step": 33880 + }, + { + "epoch": 0.81, + "learning_rate": 1.7456647367100843e-06, + "loss": 1.0277, + "step": 33881 + }, + { + "epoch": 0.81, + "learning_rate": 1.7452250763848244e-06, + "loss": 0.7881, + "step": 33882 + }, + { + "epoch": 0.81, + "learning_rate": 1.744785466139346e-06, + "loss": 1.0273, + "step": 33883 + }, + { + "epoch": 0.81, + "learning_rate": 1.744345905976309e-06, + "loss": 1.1018, + "step": 33884 + }, + { + "epoch": 0.81, + "learning_rate": 1.743906395898386e-06, + "loss": 0.9324, + "step": 33885 + }, + { + "epoch": 0.81, + "learning_rate": 1.743466935908238e-06, + "loss": 1.0239, + "step": 33886 + }, + { + "epoch": 0.81, + "learning_rate": 1.743027526008535e-06, + "loss": 0.9061, + "step": 33887 + }, + { + "epoch": 0.81, + "learning_rate": 1.742588166201944e-06, + "loss": 1.1701, + "step": 33888 + }, + { + "epoch": 0.81, + "learning_rate": 1.742148856491126e-06, + "loss": 0.7759, + "step": 33889 + }, + { + "epoch": 0.82, + "learning_rate": 1.7417095968787478e-06, + "loss": 1.0582, + "step": 33890 + }, + { + "epoch": 0.82, + "learning_rate": 1.7412703873674785e-06, + "loss": 1.0144, + "step": 33891 + }, + { + "epoch": 0.82, + "learning_rate": 1.740831227959976e-06, + "loss": 1.0911, + "step": 33892 + }, + { + "epoch": 0.82, + "learning_rate": 1.7403921186589102e-06, + "loss": 1.0748, + "step": 33893 + }, + { + "epoch": 0.82, + "learning_rate": 1.73995305946694e-06, + "loss": 1.1207, + "step": 33894 + }, + { + "epoch": 0.82, + "learning_rate": 1.7395140503867326e-06, + "loss": 1.1909, + "step": 33895 + }, + { + "epoch": 0.82, + "learning_rate": 1.739075091420952e-06, + "loss": 1.1374, + "step": 33896 + }, + { + "epoch": 0.82, + "learning_rate": 1.7386361825722575e-06, + "loss": 1.0117, + "step": 33897 + }, + { + "epoch": 0.82, + "learning_rate": 1.7381973238433148e-06, + "loss": 1.1697, + "step": 33898 + }, + { + "epoch": 0.82, + "learning_rate": 1.737758515236787e-06, + "loss": 1.1739, + "step": 33899 + }, + { + "epoch": 0.82, + "learning_rate": 1.7373197567553335e-06, + "loss": 1.2196, + "step": 33900 + }, + { + "epoch": 0.82, + "learning_rate": 1.7368810484016197e-06, + "loss": 1.0254, + "step": 33901 + }, + { + "epoch": 0.82, + "learning_rate": 1.7364423901783024e-06, + "loss": 1.2051, + "step": 33902 + }, + { + "epoch": 0.82, + "learning_rate": 1.7360037820880448e-06, + "loss": 0.8288, + "step": 33903 + }, + { + "epoch": 0.82, + "learning_rate": 1.7355652241335119e-06, + "loss": 1.0347, + "step": 33904 + }, + { + "epoch": 0.82, + "learning_rate": 1.7351267163173579e-06, + "loss": 1.363, + "step": 33905 + }, + { + "epoch": 0.82, + "learning_rate": 1.7346882586422465e-06, + "loss": 1.0889, + "step": 33906 + }, + { + "epoch": 0.82, + "learning_rate": 1.7342498511108397e-06, + "loss": 0.8893, + "step": 33907 + }, + { + "epoch": 0.82, + "learning_rate": 1.7338114937257921e-06, + "loss": 1.0408, + "step": 33908 + }, + { + "epoch": 0.82, + "learning_rate": 1.7333731864897673e-06, + "loss": 1.0023, + "step": 33909 + }, + { + "epoch": 0.82, + "learning_rate": 1.732934929405422e-06, + "loss": 1.0449, + "step": 33910 + }, + { + "epoch": 0.82, + "learning_rate": 1.7324967224754151e-06, + "loss": 0.8472, + "step": 33911 + }, + { + "epoch": 0.82, + "learning_rate": 1.7320585657024091e-06, + "loss": 1.1266, + "step": 33912 + }, + { + "epoch": 0.82, + "learning_rate": 1.7316204590890563e-06, + "loss": 1.0629, + "step": 33913 + }, + { + "epoch": 0.82, + "learning_rate": 1.7311824026380176e-06, + "loss": 0.9693, + "step": 33914 + }, + { + "epoch": 0.82, + "learning_rate": 1.730744396351952e-06, + "loss": 0.9026, + "step": 33915 + }, + { + "epoch": 0.82, + "learning_rate": 1.7303064402335135e-06, + "loss": 1.0907, + "step": 33916 + }, + { + "epoch": 0.82, + "learning_rate": 1.7298685342853628e-06, + "loss": 1.0176, + "step": 33917 + }, + { + "epoch": 0.82, + "learning_rate": 1.729430678510151e-06, + "loss": 1.006, + "step": 33918 + }, + { + "epoch": 0.82, + "learning_rate": 1.72899287291054e-06, + "loss": 1.0603, + "step": 33919 + }, + { + "epoch": 0.82, + "learning_rate": 1.7285551174891846e-06, + "loss": 1.1079, + "step": 33920 + }, + { + "epoch": 0.82, + "learning_rate": 1.7281174122487377e-06, + "loss": 1.026, + "step": 33921 + }, + { + "epoch": 0.82, + "learning_rate": 1.727679757191857e-06, + "loss": 1.1677, + "step": 33922 + }, + { + "epoch": 0.82, + "learning_rate": 1.727242152321199e-06, + "loss": 1.1312, + "step": 33923 + }, + { + "epoch": 0.82, + "learning_rate": 1.7268045976394154e-06, + "loss": 0.9075, + "step": 33924 + }, + { + "epoch": 0.82, + "learning_rate": 1.7263670931491628e-06, + "loss": 1.1114, + "step": 33925 + }, + { + "epoch": 0.82, + "learning_rate": 1.7259296388530967e-06, + "loss": 1.1474, + "step": 33926 + }, + { + "epoch": 0.82, + "learning_rate": 1.7254922347538672e-06, + "loss": 0.848, + "step": 33927 + }, + { + "epoch": 0.82, + "learning_rate": 1.7250548808541324e-06, + "loss": 1.1376, + "step": 33928 + }, + { + "epoch": 0.82, + "learning_rate": 1.7246175771565398e-06, + "loss": 0.9597, + "step": 33929 + }, + { + "epoch": 0.82, + "learning_rate": 1.7241803236637478e-06, + "loss": 1.2262, + "step": 33930 + }, + { + "epoch": 0.82, + "learning_rate": 1.7237431203784082e-06, + "loss": 1.0394, + "step": 33931 + }, + { + "epoch": 0.82, + "learning_rate": 1.7233059673031705e-06, + "loss": 1.0219, + "step": 33932 + }, + { + "epoch": 0.82, + "learning_rate": 1.7228688644406877e-06, + "loss": 1.1249, + "step": 33933 + }, + { + "epoch": 0.82, + "learning_rate": 1.7224318117936168e-06, + "loss": 1.2179, + "step": 33934 + }, + { + "epoch": 0.82, + "learning_rate": 1.721994809364601e-06, + "loss": 1.1908, + "step": 33935 + }, + { + "epoch": 0.82, + "learning_rate": 1.7215578571562986e-06, + "loss": 0.9653, + "step": 33936 + }, + { + "epoch": 0.82, + "learning_rate": 1.7211209551713558e-06, + "loss": 1.059, + "step": 33937 + }, + { + "epoch": 0.82, + "learning_rate": 1.7206841034124245e-06, + "loss": 1.1013, + "step": 33938 + }, + { + "epoch": 0.82, + "learning_rate": 1.7202473018821575e-06, + "loss": 1.0548, + "step": 33939 + }, + { + "epoch": 0.82, + "learning_rate": 1.7198105505831997e-06, + "loss": 1.0931, + "step": 33940 + }, + { + "epoch": 0.82, + "learning_rate": 1.7193738495182045e-06, + "loss": 1.2318, + "step": 33941 + }, + { + "epoch": 0.82, + "learning_rate": 1.7189371986898217e-06, + "loss": 1.1162, + "step": 33942 + }, + { + "epoch": 0.82, + "learning_rate": 1.7185005981006974e-06, + "loss": 1.0281, + "step": 33943 + }, + { + "epoch": 0.82, + "learning_rate": 1.718064047753485e-06, + "loss": 1.0989, + "step": 33944 + }, + { + "epoch": 0.82, + "learning_rate": 1.7176275476508275e-06, + "loss": 0.9885, + "step": 33945 + }, + { + "epoch": 0.82, + "learning_rate": 1.7171910977953744e-06, + "loss": 1.0307, + "step": 33946 + }, + { + "epoch": 0.82, + "learning_rate": 1.7167546981897786e-06, + "loss": 1.04, + "step": 33947 + }, + { + "epoch": 0.82, + "learning_rate": 1.7163183488366797e-06, + "loss": 0.8959, + "step": 33948 + }, + { + "epoch": 0.82, + "learning_rate": 1.7158820497387296e-06, + "loss": 1.05, + "step": 33949 + }, + { + "epoch": 0.82, + "learning_rate": 1.7154458008985775e-06, + "loss": 1.2693, + "step": 33950 + }, + { + "epoch": 0.82, + "learning_rate": 1.7150096023188645e-06, + "loss": 1.0288, + "step": 33951 + }, + { + "epoch": 0.82, + "learning_rate": 1.7145734540022418e-06, + "loss": 1.1438, + "step": 33952 + }, + { + "epoch": 0.82, + "learning_rate": 1.7141373559513507e-06, + "loss": 1.1777, + "step": 33953 + }, + { + "epoch": 0.82, + "learning_rate": 1.7137013081688392e-06, + "loss": 0.7913, + "step": 33954 + }, + { + "epoch": 0.82, + "learning_rate": 1.7132653106573549e-06, + "loss": 0.9557, + "step": 33955 + }, + { + "epoch": 0.82, + "learning_rate": 1.7128293634195393e-06, + "loss": 1.0135, + "step": 33956 + }, + { + "epoch": 0.82, + "learning_rate": 1.7123934664580376e-06, + "loss": 0.9662, + "step": 33957 + }, + { + "epoch": 0.82, + "learning_rate": 1.7119576197754985e-06, + "loss": 0.9536, + "step": 33958 + }, + { + "epoch": 0.82, + "learning_rate": 1.71152182337456e-06, + "loss": 1.2051, + "step": 33959 + }, + { + "epoch": 0.82, + "learning_rate": 1.711086077257872e-06, + "loss": 0.976, + "step": 33960 + }, + { + "epoch": 0.82, + "learning_rate": 1.7106503814280728e-06, + "loss": 0.8645, + "step": 33961 + }, + { + "epoch": 0.82, + "learning_rate": 1.7102147358878073e-06, + "loss": 1.1, + "step": 33962 + }, + { + "epoch": 0.82, + "learning_rate": 1.7097791406397223e-06, + "loss": 1.2593, + "step": 33963 + }, + { + "epoch": 0.82, + "learning_rate": 1.7093435956864545e-06, + "loss": 1.0719, + "step": 33964 + }, + { + "epoch": 0.82, + "learning_rate": 1.7089081010306485e-06, + "loss": 1.0022, + "step": 33965 + }, + { + "epoch": 0.82, + "learning_rate": 1.7084726566749499e-06, + "loss": 1.0143, + "step": 33966 + }, + { + "epoch": 0.82, + "learning_rate": 1.7080372626219954e-06, + "loss": 1.1386, + "step": 33967 + }, + { + "epoch": 0.82, + "learning_rate": 1.7076019188744274e-06, + "loss": 0.9682, + "step": 33968 + }, + { + "epoch": 0.82, + "learning_rate": 1.7071666254348918e-06, + "loss": 1.0522, + "step": 33969 + }, + { + "epoch": 0.82, + "learning_rate": 1.7067313823060216e-06, + "loss": 1.2331, + "step": 33970 + }, + { + "epoch": 0.82, + "learning_rate": 1.706296189490465e-06, + "loss": 0.9079, + "step": 33971 + }, + { + "epoch": 0.82, + "learning_rate": 1.7058610469908566e-06, + "loss": 0.8549, + "step": 33972 + }, + { + "epoch": 0.82, + "learning_rate": 1.7054259548098373e-06, + "loss": 1.0795, + "step": 33973 + }, + { + "epoch": 0.82, + "learning_rate": 1.7049909129500508e-06, + "loss": 1.1186, + "step": 33974 + }, + { + "epoch": 0.82, + "learning_rate": 1.7045559214141305e-06, + "loss": 1.2504, + "step": 33975 + }, + { + "epoch": 0.82, + "learning_rate": 1.7041209802047177e-06, + "loss": 1.4005, + "step": 33976 + }, + { + "epoch": 0.82, + "learning_rate": 1.7036860893244555e-06, + "loss": 0.9464, + "step": 33977 + }, + { + "epoch": 0.82, + "learning_rate": 1.7032512487759745e-06, + "loss": 1.1529, + "step": 33978 + }, + { + "epoch": 0.82, + "learning_rate": 1.70281645856192e-06, + "loss": 1.1653, + "step": 33979 + }, + { + "epoch": 0.82, + "learning_rate": 1.7023817186849235e-06, + "loss": 1.0034, + "step": 33980 + }, + { + "epoch": 0.82, + "learning_rate": 1.701947029147626e-06, + "loss": 1.0489, + "step": 33981 + }, + { + "epoch": 0.82, + "learning_rate": 1.7015123899526652e-06, + "loss": 1.0173, + "step": 33982 + }, + { + "epoch": 0.82, + "learning_rate": 1.7010778011026762e-06, + "loss": 1.0355, + "step": 33983 + }, + { + "epoch": 0.82, + "learning_rate": 1.7006432626002945e-06, + "loss": 0.9659, + "step": 33984 + }, + { + "epoch": 0.82, + "learning_rate": 1.700208774448161e-06, + "loss": 0.7954, + "step": 33985 + }, + { + "epoch": 0.82, + "learning_rate": 1.6997743366489072e-06, + "loss": 0.8948, + "step": 33986 + }, + { + "epoch": 0.82, + "learning_rate": 1.699339949205171e-06, + "loss": 1.0296, + "step": 33987 + }, + { + "epoch": 0.82, + "learning_rate": 1.6989056121195858e-06, + "loss": 0.9258, + "step": 33988 + }, + { + "epoch": 0.82, + "learning_rate": 1.698471325394787e-06, + "loss": 0.9292, + "step": 33989 + }, + { + "epoch": 0.82, + "learning_rate": 1.698037089033413e-06, + "loss": 0.956, + "step": 33990 + }, + { + "epoch": 0.82, + "learning_rate": 1.6976029030380914e-06, + "loss": 1.1085, + "step": 33991 + }, + { + "epoch": 0.82, + "learning_rate": 1.6971687674114612e-06, + "loss": 1.0421, + "step": 33992 + }, + { + "epoch": 0.82, + "learning_rate": 1.6967346821561582e-06, + "loss": 0.9842, + "step": 33993 + }, + { + "epoch": 0.82, + "learning_rate": 1.696300647274809e-06, + "loss": 1.0652, + "step": 33994 + }, + { + "epoch": 0.82, + "learning_rate": 1.6958666627700537e-06, + "loss": 1.0881, + "step": 33995 + }, + { + "epoch": 0.82, + "learning_rate": 1.6954327286445205e-06, + "loss": 1.0069, + "step": 33996 + }, + { + "epoch": 0.82, + "learning_rate": 1.6949988449008425e-06, + "loss": 1.0476, + "step": 33997 + }, + { + "epoch": 0.82, + "learning_rate": 1.6945650115416556e-06, + "loss": 1.0793, + "step": 33998 + }, + { + "epoch": 0.82, + "learning_rate": 1.6941312285695843e-06, + "loss": 0.8035, + "step": 33999 + }, + { + "epoch": 0.82, + "learning_rate": 1.693697495987271e-06, + "loss": 0.9466, + "step": 34000 + }, + { + "epoch": 0.82, + "learning_rate": 1.6932638137973378e-06, + "loss": 1.0355, + "step": 34001 + }, + { + "epoch": 0.82, + "learning_rate": 1.6928301820024195e-06, + "loss": 1.0426, + "step": 34002 + }, + { + "epoch": 0.82, + "learning_rate": 1.6923966006051483e-06, + "loss": 0.9701, + "step": 34003 + }, + { + "epoch": 0.82, + "learning_rate": 1.6919630696081513e-06, + "loss": 1.1305, + "step": 34004 + }, + { + "epoch": 0.82, + "learning_rate": 1.6915295890140604e-06, + "loss": 1.0047, + "step": 34005 + }, + { + "epoch": 0.82, + "learning_rate": 1.6910961588255071e-06, + "loss": 0.8907, + "step": 34006 + }, + { + "epoch": 0.82, + "learning_rate": 1.690662779045117e-06, + "loss": 0.7063, + "step": 34007 + }, + { + "epoch": 0.82, + "learning_rate": 1.6902294496755233e-06, + "loss": 1.0666, + "step": 34008 + }, + { + "epoch": 0.82, + "learning_rate": 1.6897961707193512e-06, + "loss": 1.2126, + "step": 34009 + }, + { + "epoch": 0.82, + "learning_rate": 1.6893629421792302e-06, + "loss": 1.0767, + "step": 34010 + }, + { + "epoch": 0.82, + "learning_rate": 1.6889297640577918e-06, + "loss": 1.0452, + "step": 34011 + }, + { + "epoch": 0.82, + "learning_rate": 1.6884966363576593e-06, + "loss": 0.9788, + "step": 34012 + }, + { + "epoch": 0.82, + "learning_rate": 1.6880635590814632e-06, + "loss": 0.9661, + "step": 34013 + }, + { + "epoch": 0.82, + "learning_rate": 1.687630532231833e-06, + "loss": 1.1907, + "step": 34014 + }, + { + "epoch": 0.82, + "learning_rate": 1.6871975558113907e-06, + "loss": 1.1152, + "step": 34015 + }, + { + "epoch": 0.82, + "learning_rate": 1.686764629822767e-06, + "loss": 0.9904, + "step": 34016 + }, + { + "epoch": 0.82, + "learning_rate": 1.6863317542685853e-06, + "loss": 1.0714, + "step": 34017 + }, + { + "epoch": 0.82, + "learning_rate": 1.6858989291514737e-06, + "loss": 1.26, + "step": 34018 + }, + { + "epoch": 0.82, + "learning_rate": 1.6854661544740592e-06, + "loss": 1.0121, + "step": 34019 + }, + { + "epoch": 0.82, + "learning_rate": 1.6850334302389638e-06, + "loss": 1.1057, + "step": 34020 + }, + { + "epoch": 0.82, + "learning_rate": 1.6846007564488154e-06, + "loss": 0.9776, + "step": 34021 + }, + { + "epoch": 0.82, + "learning_rate": 1.6841681331062398e-06, + "loss": 0.8772, + "step": 34022 + }, + { + "epoch": 0.82, + "learning_rate": 1.6837355602138584e-06, + "loss": 1.0656, + "step": 34023 + }, + { + "epoch": 0.82, + "learning_rate": 1.6833030377742988e-06, + "loss": 1.0339, + "step": 34024 + }, + { + "epoch": 0.82, + "learning_rate": 1.6828705657901823e-06, + "loss": 1.2802, + "step": 34025 + }, + { + "epoch": 0.82, + "learning_rate": 1.6824381442641325e-06, + "loss": 1.0578, + "step": 34026 + }, + { + "epoch": 0.82, + "learning_rate": 1.682005773198777e-06, + "loss": 1.1615, + "step": 34027 + }, + { + "epoch": 0.82, + "learning_rate": 1.6815734525967332e-06, + "loss": 0.9043, + "step": 34028 + }, + { + "epoch": 0.82, + "learning_rate": 1.681141182460627e-06, + "loss": 0.9624, + "step": 34029 + }, + { + "epoch": 0.82, + "learning_rate": 1.680708962793084e-06, + "loss": 1.0181, + "step": 34030 + }, + { + "epoch": 0.82, + "learning_rate": 1.680276793596719e-06, + "loss": 0.95, + "step": 34031 + }, + { + "epoch": 0.82, + "learning_rate": 1.6798446748741603e-06, + "loss": 0.9357, + "step": 34032 + }, + { + "epoch": 0.82, + "learning_rate": 1.6794126066280247e-06, + "loss": 0.8869, + "step": 34033 + }, + { + "epoch": 0.82, + "learning_rate": 1.6789805888609368e-06, + "loss": 0.9896, + "step": 34034 + }, + { + "epoch": 0.82, + "learning_rate": 1.6785486215755176e-06, + "loss": 1.0291, + "step": 34035 + }, + { + "epoch": 0.82, + "learning_rate": 1.678116704774384e-06, + "loss": 1.0122, + "step": 34036 + }, + { + "epoch": 0.82, + "learning_rate": 1.6776848384601596e-06, + "loss": 1.0951, + "step": 34037 + }, + { + "epoch": 0.82, + "learning_rate": 1.6772530226354666e-06, + "loss": 1.1913, + "step": 34038 + }, + { + "epoch": 0.82, + "learning_rate": 1.6768212573029186e-06, + "loss": 1.0156, + "step": 34039 + }, + { + "epoch": 0.82, + "learning_rate": 1.676389542465141e-06, + "loss": 1.132, + "step": 34040 + }, + { + "epoch": 0.82, + "learning_rate": 1.6759578781247476e-06, + "loss": 1.0982, + "step": 34041 + }, + { + "epoch": 0.82, + "learning_rate": 1.6755262642843607e-06, + "loss": 0.9471, + "step": 34042 + }, + { + "epoch": 0.82, + "learning_rate": 1.675094700946599e-06, + "loss": 0.923, + "step": 34043 + }, + { + "epoch": 0.82, + "learning_rate": 1.6746631881140775e-06, + "loss": 1.0975, + "step": 34044 + }, + { + "epoch": 0.82, + "learning_rate": 1.674231725789417e-06, + "loss": 0.981, + "step": 34045 + }, + { + "epoch": 0.82, + "learning_rate": 1.6738003139752358e-06, + "loss": 0.9881, + "step": 34046 + }, + { + "epoch": 0.82, + "learning_rate": 1.6733689526741482e-06, + "loss": 1.0855, + "step": 34047 + }, + { + "epoch": 0.82, + "learning_rate": 1.6729376418887722e-06, + "loss": 1.0194, + "step": 34048 + }, + { + "epoch": 0.82, + "learning_rate": 1.6725063816217269e-06, + "loss": 0.959, + "step": 34049 + }, + { + "epoch": 0.82, + "learning_rate": 1.6720751718756244e-06, + "loss": 0.9701, + "step": 34050 + }, + { + "epoch": 0.82, + "learning_rate": 1.6716440126530853e-06, + "loss": 1.0235, + "step": 34051 + }, + { + "epoch": 0.82, + "learning_rate": 1.6712129039567215e-06, + "loss": 0.8507, + "step": 34052 + }, + { + "epoch": 0.82, + "learning_rate": 1.6707818457891489e-06, + "loss": 1.1607, + "step": 34053 + }, + { + "epoch": 0.82, + "learning_rate": 1.6703508381529865e-06, + "loss": 1.1296, + "step": 34054 + }, + { + "epoch": 0.82, + "learning_rate": 1.6699198810508443e-06, + "loss": 1.1974, + "step": 34055 + }, + { + "epoch": 0.82, + "learning_rate": 1.6694889744853381e-06, + "loss": 1.0806, + "step": 34056 + }, + { + "epoch": 0.82, + "learning_rate": 1.6690581184590859e-06, + "loss": 0.9863, + "step": 34057 + }, + { + "epoch": 0.82, + "learning_rate": 1.6686273129746954e-06, + "loss": 1.0415, + "step": 34058 + }, + { + "epoch": 0.82, + "learning_rate": 1.668196558034786e-06, + "loss": 1.4484, + "step": 34059 + }, + { + "epoch": 0.82, + "learning_rate": 1.6677658536419661e-06, + "loss": 1.0164, + "step": 34060 + }, + { + "epoch": 0.82, + "learning_rate": 1.6673351997988507e-06, + "loss": 1.0717, + "step": 34061 + }, + { + "epoch": 0.82, + "learning_rate": 1.6669045965080555e-06, + "loss": 1.1033, + "step": 34062 + }, + { + "epoch": 0.82, + "learning_rate": 1.6664740437721872e-06, + "loss": 1.0144, + "step": 34063 + }, + { + "epoch": 0.82, + "learning_rate": 1.6660435415938615e-06, + "loss": 0.9569, + "step": 34064 + }, + { + "epoch": 0.82, + "learning_rate": 1.6656130899756906e-06, + "loss": 1.0257, + "step": 34065 + }, + { + "epoch": 0.82, + "learning_rate": 1.6651826889202827e-06, + "loss": 1.1991, + "step": 34066 + }, + { + "epoch": 0.82, + "learning_rate": 1.6647523384302534e-06, + "loss": 1.2497, + "step": 34067 + }, + { + "epoch": 0.82, + "learning_rate": 1.6643220385082092e-06, + "loss": 0.888, + "step": 34068 + }, + { + "epoch": 0.82, + "learning_rate": 1.663891789156762e-06, + "loss": 1.067, + "step": 34069 + }, + { + "epoch": 0.82, + "learning_rate": 1.6634615903785255e-06, + "loss": 1.1505, + "step": 34070 + }, + { + "epoch": 0.82, + "learning_rate": 1.663031442176104e-06, + "loss": 1.1397, + "step": 34071 + }, + { + "epoch": 0.82, + "learning_rate": 1.6626013445521094e-06, + "loss": 0.9681, + "step": 34072 + }, + { + "epoch": 0.82, + "learning_rate": 1.6621712975091542e-06, + "loss": 1.163, + "step": 34073 + }, + { + "epoch": 0.82, + "learning_rate": 1.6617413010498419e-06, + "loss": 1.1455, + "step": 34074 + }, + { + "epoch": 0.82, + "learning_rate": 1.6613113551767856e-06, + "loss": 1.0965, + "step": 34075 + }, + { + "epoch": 0.82, + "learning_rate": 1.6608814598925904e-06, + "loss": 0.9273, + "step": 34076 + }, + { + "epoch": 0.82, + "learning_rate": 1.6604516151998663e-06, + "loss": 0.9743, + "step": 34077 + }, + { + "epoch": 0.82, + "learning_rate": 1.660021821101222e-06, + "loss": 1.0627, + "step": 34078 + }, + { + "epoch": 0.82, + "learning_rate": 1.6595920775992613e-06, + "loss": 0.7412, + "step": 34079 + }, + { + "epoch": 0.82, + "learning_rate": 1.6591623846965942e-06, + "loss": 1.2749, + "step": 34080 + }, + { + "epoch": 0.82, + "learning_rate": 1.65873274239583e-06, + "loss": 0.8366, + "step": 34081 + }, + { + "epoch": 0.82, + "learning_rate": 1.6583031506995684e-06, + "loss": 0.9667, + "step": 34082 + }, + { + "epoch": 0.82, + "learning_rate": 1.657873609610423e-06, + "loss": 1.0098, + "step": 34083 + }, + { + "epoch": 0.82, + "learning_rate": 1.6574441191309941e-06, + "loss": 1.0194, + "step": 34084 + }, + { + "epoch": 0.82, + "learning_rate": 1.6570146792638886e-06, + "loss": 1.0861, + "step": 34085 + }, + { + "epoch": 0.82, + "learning_rate": 1.6565852900117153e-06, + "loss": 1.0745, + "step": 34086 + }, + { + "epoch": 0.82, + "learning_rate": 1.6561559513770742e-06, + "loss": 0.8411, + "step": 34087 + }, + { + "epoch": 0.82, + "learning_rate": 1.6557266633625724e-06, + "loss": 0.7791, + "step": 34088 + }, + { + "epoch": 0.82, + "learning_rate": 1.6552974259708166e-06, + "loss": 0.8661, + "step": 34089 + }, + { + "epoch": 0.82, + "learning_rate": 1.6548682392044057e-06, + "loss": 1.1313, + "step": 34090 + }, + { + "epoch": 0.82, + "learning_rate": 1.6544391030659469e-06, + "loss": 1.1226, + "step": 34091 + }, + { + "epoch": 0.82, + "learning_rate": 1.6540100175580454e-06, + "loss": 1.2818, + "step": 34092 + }, + { + "epoch": 0.82, + "learning_rate": 1.6535809826832993e-06, + "loss": 0.9105, + "step": 34093 + }, + { + "epoch": 0.82, + "learning_rate": 1.6531519984443167e-06, + "loss": 1.18, + "step": 34094 + }, + { + "epoch": 0.82, + "learning_rate": 1.6527230648436953e-06, + "loss": 0.9224, + "step": 34095 + }, + { + "epoch": 0.82, + "learning_rate": 1.6522941818840398e-06, + "loss": 0.8821, + "step": 34096 + }, + { + "epoch": 0.82, + "learning_rate": 1.6518653495679538e-06, + "loss": 1.2636, + "step": 34097 + }, + { + "epoch": 0.82, + "learning_rate": 1.6514365678980349e-06, + "loss": 1.2487, + "step": 34098 + }, + { + "epoch": 0.82, + "learning_rate": 1.6510078368768879e-06, + "loss": 1.1948, + "step": 34099 + }, + { + "epoch": 0.82, + "learning_rate": 1.6505791565071139e-06, + "loss": 1.0793, + "step": 34100 + }, + { + "epoch": 0.82, + "learning_rate": 1.6501505267913098e-06, + "loss": 0.9786, + "step": 34101 + }, + { + "epoch": 0.82, + "learning_rate": 1.6497219477320804e-06, + "loss": 1.2284, + "step": 34102 + }, + { + "epoch": 0.82, + "learning_rate": 1.649293419332022e-06, + "loss": 1.1706, + "step": 34103 + }, + { + "epoch": 0.82, + "learning_rate": 1.6488649415937364e-06, + "loss": 1.0163, + "step": 34104 + }, + { + "epoch": 0.82, + "learning_rate": 1.6484365145198255e-06, + "loss": 1.1784, + "step": 34105 + }, + { + "epoch": 0.82, + "learning_rate": 1.648008138112882e-06, + "loss": 1.114, + "step": 34106 + }, + { + "epoch": 0.82, + "learning_rate": 1.6475798123755104e-06, + "loss": 0.7941, + "step": 34107 + }, + { + "epoch": 0.82, + "learning_rate": 1.6471515373103086e-06, + "loss": 0.8837, + "step": 34108 + }, + { + "epoch": 0.82, + "learning_rate": 1.6467233129198724e-06, + "loss": 1.2244, + "step": 34109 + }, + { + "epoch": 0.82, + "learning_rate": 1.6462951392068027e-06, + "loss": 1.1654, + "step": 34110 + }, + { + "epoch": 0.82, + "learning_rate": 1.6458670161736934e-06, + "loss": 0.8124, + "step": 34111 + }, + { + "epoch": 0.82, + "learning_rate": 1.6454389438231433e-06, + "loss": 1.1582, + "step": 34112 + }, + { + "epoch": 0.82, + "learning_rate": 1.6450109221577537e-06, + "loss": 0.7968, + "step": 34113 + }, + { + "epoch": 0.82, + "learning_rate": 1.6445829511801137e-06, + "loss": 1.2667, + "step": 34114 + }, + { + "epoch": 0.82, + "learning_rate": 1.6441550308928244e-06, + "loss": 0.8818, + "step": 34115 + }, + { + "epoch": 0.82, + "learning_rate": 1.643727161298484e-06, + "loss": 1.1756, + "step": 34116 + }, + { + "epoch": 0.82, + "learning_rate": 1.6432993423996824e-06, + "loss": 0.7275, + "step": 34117 + }, + { + "epoch": 0.82, + "learning_rate": 1.642871574199021e-06, + "loss": 0.8927, + "step": 34118 + }, + { + "epoch": 0.82, + "learning_rate": 1.6424438566990885e-06, + "loss": 0.9188, + "step": 34119 + }, + { + "epoch": 0.82, + "learning_rate": 1.6420161899024845e-06, + "loss": 1.0383, + "step": 34120 + }, + { + "epoch": 0.82, + "learning_rate": 1.6415885738118042e-06, + "loss": 1.3086, + "step": 34121 + }, + { + "epoch": 0.82, + "learning_rate": 1.641161008429637e-06, + "loss": 1.1361, + "step": 34122 + }, + { + "epoch": 0.82, + "learning_rate": 1.6407334937585805e-06, + "loss": 0.9986, + "step": 34123 + }, + { + "epoch": 0.82, + "learning_rate": 1.6403060298012309e-06, + "loss": 0.941, + "step": 34124 + }, + { + "epoch": 0.82, + "learning_rate": 1.6398786165601743e-06, + "loss": 1.166, + "step": 34125 + }, + { + "epoch": 0.82, + "learning_rate": 1.6394512540380114e-06, + "loss": 0.7532, + "step": 34126 + }, + { + "epoch": 0.82, + "learning_rate": 1.6390239422373278e-06, + "loss": 0.9548, + "step": 34127 + }, + { + "epoch": 0.82, + "learning_rate": 1.63859668116072e-06, + "loss": 1.1474, + "step": 34128 + }, + { + "epoch": 0.82, + "learning_rate": 1.638169470810782e-06, + "loss": 1.0693, + "step": 34129 + }, + { + "epoch": 0.82, + "learning_rate": 1.6377423111900993e-06, + "loss": 0.9798, + "step": 34130 + }, + { + "epoch": 0.82, + "learning_rate": 1.6373152023012683e-06, + "loss": 1.1854, + "step": 34131 + }, + { + "epoch": 0.82, + "learning_rate": 1.636888144146881e-06, + "loss": 1.2039, + "step": 34132 + }, + { + "epoch": 0.82, + "learning_rate": 1.6364611367295236e-06, + "loss": 0.9302, + "step": 34133 + }, + { + "epoch": 0.82, + "learning_rate": 1.636034180051792e-06, + "loss": 0.8469, + "step": 34134 + }, + { + "epoch": 0.82, + "learning_rate": 1.635607274116271e-06, + "loss": 0.922, + "step": 34135 + }, + { + "epoch": 0.82, + "learning_rate": 1.635180418925554e-06, + "loss": 1.1401, + "step": 34136 + }, + { + "epoch": 0.82, + "learning_rate": 1.6347536144822318e-06, + "loss": 1.1708, + "step": 34137 + }, + { + "epoch": 0.82, + "learning_rate": 1.6343268607888896e-06, + "loss": 0.9729, + "step": 34138 + }, + { + "epoch": 0.82, + "learning_rate": 1.6339001578481184e-06, + "loss": 0.9954, + "step": 34139 + }, + { + "epoch": 0.82, + "learning_rate": 1.6334735056625096e-06, + "loss": 1.0832, + "step": 34140 + }, + { + "epoch": 0.82, + "learning_rate": 1.6330469042346463e-06, + "loss": 0.9323, + "step": 34141 + }, + { + "epoch": 0.82, + "learning_rate": 1.6326203535671191e-06, + "loss": 0.8278, + "step": 34142 + }, + { + "epoch": 0.82, + "learning_rate": 1.6321938536625192e-06, + "loss": 0.9917, + "step": 34143 + }, + { + "epoch": 0.82, + "learning_rate": 1.6317674045234288e-06, + "loss": 1.1041, + "step": 34144 + }, + { + "epoch": 0.82, + "learning_rate": 1.6313410061524393e-06, + "loss": 0.7366, + "step": 34145 + }, + { + "epoch": 0.82, + "learning_rate": 1.6309146585521329e-06, + "loss": 0.8603, + "step": 34146 + }, + { + "epoch": 0.82, + "learning_rate": 1.6304883617250989e-06, + "loss": 1.2754, + "step": 34147 + }, + { + "epoch": 0.82, + "learning_rate": 1.630062115673925e-06, + "loss": 0.9787, + "step": 34148 + }, + { + "epoch": 0.82, + "learning_rate": 1.6296359204011947e-06, + "loss": 1.1249, + "step": 34149 + }, + { + "epoch": 0.82, + "learning_rate": 1.6292097759094927e-06, + "loss": 1.0157, + "step": 34150 + }, + { + "epoch": 0.82, + "learning_rate": 1.6287836822014103e-06, + "loss": 1.1051, + "step": 34151 + }, + { + "epoch": 0.82, + "learning_rate": 1.6283576392795242e-06, + "loss": 0.8566, + "step": 34152 + }, + { + "epoch": 0.82, + "learning_rate": 1.6279316471464268e-06, + "loss": 1.1739, + "step": 34153 + }, + { + "epoch": 0.82, + "learning_rate": 1.627505705804696e-06, + "loss": 1.2153, + "step": 34154 + }, + { + "epoch": 0.82, + "learning_rate": 1.6270798152569189e-06, + "loss": 1.1721, + "step": 34155 + }, + { + "epoch": 0.82, + "learning_rate": 1.626653975505681e-06, + "loss": 1.0111, + "step": 34156 + }, + { + "epoch": 0.82, + "learning_rate": 1.6262281865535612e-06, + "loss": 1.1718, + "step": 34157 + }, + { + "epoch": 0.82, + "learning_rate": 1.6258024484031465e-06, + "loss": 1.2157, + "step": 34158 + }, + { + "epoch": 0.82, + "learning_rate": 1.6253767610570204e-06, + "loss": 1.156, + "step": 34159 + }, + { + "epoch": 0.82, + "learning_rate": 1.6249511245177619e-06, + "loss": 1.2206, + "step": 34160 + }, + { + "epoch": 0.82, + "learning_rate": 1.6245255387879566e-06, + "loss": 0.8768, + "step": 34161 + }, + { + "epoch": 0.82, + "learning_rate": 1.6241000038701827e-06, + "loss": 1.034, + "step": 34162 + }, + { + "epoch": 0.82, + "learning_rate": 1.6236745197670233e-06, + "loss": 0.7762, + "step": 34163 + }, + { + "epoch": 0.82, + "learning_rate": 1.6232490864810634e-06, + "loss": 0.8397, + "step": 34164 + }, + { + "epoch": 0.82, + "learning_rate": 1.6228237040148787e-06, + "loss": 1.2267, + "step": 34165 + }, + { + "epoch": 0.82, + "learning_rate": 1.6223983723710523e-06, + "loss": 1.0424, + "step": 34166 + }, + { + "epoch": 0.82, + "learning_rate": 1.6219730915521647e-06, + "loss": 1.0688, + "step": 34167 + }, + { + "epoch": 0.82, + "learning_rate": 1.6215478615607948e-06, + "loss": 1.0473, + "step": 34168 + }, + { + "epoch": 0.82, + "learning_rate": 1.6211226823995263e-06, + "loss": 0.9368, + "step": 34169 + }, + { + "epoch": 0.82, + "learning_rate": 1.6206975540709325e-06, + "loss": 1.1927, + "step": 34170 + }, + { + "epoch": 0.82, + "learning_rate": 1.6202724765775967e-06, + "loss": 1.0834, + "step": 34171 + }, + { + "epoch": 0.82, + "learning_rate": 1.6198474499220985e-06, + "loss": 1.1543, + "step": 34172 + }, + { + "epoch": 0.82, + "learning_rate": 1.619422474107012e-06, + "loss": 1.3112, + "step": 34173 + }, + { + "epoch": 0.82, + "learning_rate": 1.618997549134922e-06, + "loss": 0.9262, + "step": 34174 + }, + { + "epoch": 0.82, + "learning_rate": 1.6185726750083985e-06, + "loss": 1.03, + "step": 34175 + }, + { + "epoch": 0.82, + "learning_rate": 1.6181478517300243e-06, + "loss": 1.2078, + "step": 34176 + }, + { + "epoch": 0.82, + "learning_rate": 1.6177230793023767e-06, + "loss": 0.8942, + "step": 34177 + }, + { + "epoch": 0.82, + "learning_rate": 1.61729835772803e-06, + "loss": 1.0831, + "step": 34178 + }, + { + "epoch": 0.82, + "learning_rate": 1.6168736870095625e-06, + "loss": 1.0731, + "step": 34179 + }, + { + "epoch": 0.82, + "learning_rate": 1.6164490671495525e-06, + "loss": 1.0638, + "step": 34180 + }, + { + "epoch": 0.82, + "learning_rate": 1.6160244981505724e-06, + "loss": 0.9466, + "step": 34181 + }, + { + "epoch": 0.82, + "learning_rate": 1.615599980015201e-06, + "loss": 0.9829, + "step": 34182 + }, + { + "epoch": 0.82, + "learning_rate": 1.615175512746011e-06, + "loss": 1.1184, + "step": 34183 + }, + { + "epoch": 0.82, + "learning_rate": 1.614751096345578e-06, + "loss": 0.9972, + "step": 34184 + }, + { + "epoch": 0.82, + "learning_rate": 1.6143267308164812e-06, + "loss": 1.1164, + "step": 34185 + }, + { + "epoch": 0.82, + "learning_rate": 1.6139024161612882e-06, + "loss": 1.0444, + "step": 34186 + }, + { + "epoch": 0.82, + "learning_rate": 1.613478152382577e-06, + "loss": 1.1514, + "step": 34187 + }, + { + "epoch": 0.82, + "learning_rate": 1.6130539394829237e-06, + "loss": 0.9421, + "step": 34188 + }, + { + "epoch": 0.82, + "learning_rate": 1.612629777464898e-06, + "loss": 1.1538, + "step": 34189 + }, + { + "epoch": 0.82, + "learning_rate": 1.612205666331076e-06, + "loss": 1.1134, + "step": 34190 + }, + { + "epoch": 0.82, + "learning_rate": 1.6117816060840263e-06, + "loss": 0.9128, + "step": 34191 + }, + { + "epoch": 0.82, + "learning_rate": 1.6113575967263262e-06, + "loss": 1.1518, + "step": 34192 + }, + { + "epoch": 0.82, + "learning_rate": 1.610933638260548e-06, + "loss": 1.0403, + "step": 34193 + }, + { + "epoch": 0.82, + "learning_rate": 1.610509730689259e-06, + "loss": 1.1094, + "step": 34194 + }, + { + "epoch": 0.82, + "learning_rate": 1.6100858740150349e-06, + "loss": 0.8882, + "step": 34195 + }, + { + "epoch": 0.82, + "learning_rate": 1.6096620682404496e-06, + "loss": 1.1086, + "step": 34196 + }, + { + "epoch": 0.82, + "learning_rate": 1.6092383133680677e-06, + "loss": 1.2132, + "step": 34197 + }, + { + "epoch": 0.82, + "learning_rate": 1.6088146094004652e-06, + "loss": 1.038, + "step": 34198 + }, + { + "epoch": 0.82, + "learning_rate": 1.6083909563402088e-06, + "loss": 0.9181, + "step": 34199 + }, + { + "epoch": 0.82, + "learning_rate": 1.607967354189871e-06, + "loss": 0.8833, + "step": 34200 + }, + { + "epoch": 0.82, + "learning_rate": 1.607543802952024e-06, + "loss": 0.9068, + "step": 34201 + }, + { + "epoch": 0.82, + "learning_rate": 1.6071203026292315e-06, + "loss": 0.6716, + "step": 34202 + }, + { + "epoch": 0.82, + "learning_rate": 1.6066968532240668e-06, + "loss": 1.1802, + "step": 34203 + }, + { + "epoch": 0.82, + "learning_rate": 1.606273454739099e-06, + "loss": 1.1183, + "step": 34204 + }, + { + "epoch": 0.82, + "learning_rate": 1.6058501071768951e-06, + "loss": 1.052, + "step": 34205 + }, + { + "epoch": 0.82, + "learning_rate": 1.605426810540026e-06, + "loss": 1.0419, + "step": 34206 + }, + { + "epoch": 0.82, + "learning_rate": 1.6050035648310546e-06, + "loss": 1.037, + "step": 34207 + }, + { + "epoch": 0.82, + "learning_rate": 1.604580370052553e-06, + "loss": 1.0549, + "step": 34208 + }, + { + "epoch": 0.82, + "learning_rate": 1.6041572262070903e-06, + "loss": 1.1545, + "step": 34209 + }, + { + "epoch": 0.82, + "learning_rate": 1.603734133297228e-06, + "loss": 1.1887, + "step": 34210 + }, + { + "epoch": 0.82, + "learning_rate": 1.6033110913255356e-06, + "loss": 1.1529, + "step": 34211 + }, + { + "epoch": 0.82, + "learning_rate": 1.6028881002945817e-06, + "loss": 1.1757, + "step": 34212 + }, + { + "epoch": 0.82, + "learning_rate": 1.6024651602069286e-06, + "loss": 1.0004, + "step": 34213 + }, + { + "epoch": 0.82, + "learning_rate": 1.6020422710651474e-06, + "loss": 1.1278, + "step": 34214 + }, + { + "epoch": 0.82, + "learning_rate": 1.601619432871797e-06, + "loss": 1.0337, + "step": 34215 + }, + { + "epoch": 0.82, + "learning_rate": 1.601196645629447e-06, + "loss": 1.0874, + "step": 34216 + }, + { + "epoch": 0.82, + "learning_rate": 1.6007739093406628e-06, + "loss": 0.9191, + "step": 34217 + }, + { + "epoch": 0.82, + "learning_rate": 1.6003512240080054e-06, + "loss": 1.0937, + "step": 34218 + }, + { + "epoch": 0.82, + "learning_rate": 1.5999285896340422e-06, + "loss": 1.0501, + "step": 34219 + }, + { + "epoch": 0.82, + "learning_rate": 1.5995060062213385e-06, + "loss": 0.9629, + "step": 34220 + }, + { + "epoch": 0.82, + "learning_rate": 1.5990834737724525e-06, + "loss": 1.285, + "step": 34221 + }, + { + "epoch": 0.82, + "learning_rate": 1.598660992289952e-06, + "loss": 0.9216, + "step": 34222 + }, + { + "epoch": 0.82, + "learning_rate": 1.5982385617764019e-06, + "loss": 0.954, + "step": 34223 + }, + { + "epoch": 0.82, + "learning_rate": 1.5978161822343596e-06, + "loss": 1.074, + "step": 34224 + }, + { + "epoch": 0.82, + "learning_rate": 1.5973938536663924e-06, + "loss": 1.1278, + "step": 34225 + }, + { + "epoch": 0.82, + "learning_rate": 1.596971576075058e-06, + "loss": 0.8649, + "step": 34226 + }, + { + "epoch": 0.82, + "learning_rate": 1.5965493494629203e-06, + "loss": 1.1244, + "step": 34227 + }, + { + "epoch": 0.82, + "learning_rate": 1.5961271738325434e-06, + "loss": 1.1594, + "step": 34228 + }, + { + "epoch": 0.82, + "learning_rate": 1.5957050491864845e-06, + "loss": 0.9096, + "step": 34229 + }, + { + "epoch": 0.82, + "learning_rate": 1.5952829755273057e-06, + "loss": 1.1549, + "step": 34230 + }, + { + "epoch": 0.82, + "learning_rate": 1.5948609528575709e-06, + "loss": 1.1806, + "step": 34231 + }, + { + "epoch": 0.82, + "learning_rate": 1.5944389811798344e-06, + "loss": 1.2607, + "step": 34232 + }, + { + "epoch": 0.82, + "learning_rate": 1.5940170604966632e-06, + "loss": 0.9589, + "step": 34233 + }, + { + "epoch": 0.82, + "learning_rate": 1.5935951908106095e-06, + "loss": 1.0516, + "step": 34234 + }, + { + "epoch": 0.82, + "learning_rate": 1.5931733721242382e-06, + "loss": 0.9312, + "step": 34235 + }, + { + "epoch": 0.82, + "learning_rate": 1.5927516044401071e-06, + "loss": 0.9175, + "step": 34236 + }, + { + "epoch": 0.82, + "learning_rate": 1.592329887760774e-06, + "loss": 0.8932, + "step": 34237 + }, + { + "epoch": 0.82, + "learning_rate": 1.591908222088796e-06, + "loss": 0.7541, + "step": 34238 + }, + { + "epoch": 0.82, + "learning_rate": 1.5914866074267366e-06, + "loss": 1.0442, + "step": 34239 + }, + { + "epoch": 0.82, + "learning_rate": 1.591065043777148e-06, + "loss": 0.92, + "step": 34240 + }, + { + "epoch": 0.82, + "learning_rate": 1.5906435311425915e-06, + "loss": 1.1564, + "step": 34241 + }, + { + "epoch": 0.82, + "learning_rate": 1.590222069525621e-06, + "loss": 0.8654, + "step": 34242 + }, + { + "epoch": 0.82, + "learning_rate": 1.5898006589287952e-06, + "loss": 1.047, + "step": 34243 + }, + { + "epoch": 0.82, + "learning_rate": 1.589379299354672e-06, + "loss": 1.0771, + "step": 34244 + }, + { + "epoch": 0.82, + "learning_rate": 1.5889579908058051e-06, + "loss": 1.2327, + "step": 34245 + }, + { + "epoch": 0.82, + "learning_rate": 1.58853673328475e-06, + "loss": 0.8312, + "step": 34246 + }, + { + "epoch": 0.82, + "learning_rate": 1.5881155267940685e-06, + "loss": 1.0898, + "step": 34247 + }, + { + "epoch": 0.82, + "learning_rate": 1.587694371336308e-06, + "loss": 1.0252, + "step": 34248 + }, + { + "epoch": 0.82, + "learning_rate": 1.5872732669140288e-06, + "loss": 1.2537, + "step": 34249 + }, + { + "epoch": 0.82, + "learning_rate": 1.5868522135297827e-06, + "loss": 0.6704, + "step": 34250 + }, + { + "epoch": 0.82, + "learning_rate": 1.5864312111861247e-06, + "loss": 1.0136, + "step": 34251 + }, + { + "epoch": 0.82, + "learning_rate": 1.586010259885612e-06, + "loss": 0.9187, + "step": 34252 + }, + { + "epoch": 0.82, + "learning_rate": 1.5855893596307936e-06, + "loss": 1.0578, + "step": 34253 + }, + { + "epoch": 0.82, + "learning_rate": 1.5851685104242254e-06, + "loss": 0.9431, + "step": 34254 + }, + { + "epoch": 0.82, + "learning_rate": 1.584747712268463e-06, + "loss": 0.9324, + "step": 34255 + }, + { + "epoch": 0.82, + "learning_rate": 1.5843269651660554e-06, + "loss": 0.8453, + "step": 34256 + }, + { + "epoch": 0.82, + "learning_rate": 1.5839062691195572e-06, + "loss": 1.2764, + "step": 34257 + }, + { + "epoch": 0.82, + "learning_rate": 1.583485624131519e-06, + "loss": 1.1412, + "step": 34258 + }, + { + "epoch": 0.82, + "learning_rate": 1.5830650302044936e-06, + "loss": 0.9502, + "step": 34259 + }, + { + "epoch": 0.82, + "learning_rate": 1.5826444873410352e-06, + "loss": 1.2768, + "step": 34260 + }, + { + "epoch": 0.82, + "learning_rate": 1.5822239955436903e-06, + "loss": 0.9717, + "step": 34261 + }, + { + "epoch": 0.82, + "learning_rate": 1.5818035548150124e-06, + "loss": 0.8017, + "step": 34262 + }, + { + "epoch": 0.82, + "learning_rate": 1.5813831651575539e-06, + "loss": 0.8485, + "step": 34263 + }, + { + "epoch": 0.82, + "learning_rate": 1.5809628265738619e-06, + "loss": 0.862, + "step": 34264 + }, + { + "epoch": 0.82, + "learning_rate": 1.5805425390664885e-06, + "loss": 0.9503, + "step": 34265 + }, + { + "epoch": 0.82, + "learning_rate": 1.5801223026379854e-06, + "loss": 1.1181, + "step": 34266 + }, + { + "epoch": 0.82, + "learning_rate": 1.5797021172908966e-06, + "loss": 1.1882, + "step": 34267 + }, + { + "epoch": 0.82, + "learning_rate": 1.5792819830277783e-06, + "loss": 1.0594, + "step": 34268 + }, + { + "epoch": 0.82, + "learning_rate": 1.5788618998511718e-06, + "loss": 1.0351, + "step": 34269 + }, + { + "epoch": 0.82, + "learning_rate": 1.5784418677636304e-06, + "loss": 1.0877, + "step": 34270 + }, + { + "epoch": 0.82, + "learning_rate": 1.5780218867677034e-06, + "loss": 1.0678, + "step": 34271 + }, + { + "epoch": 0.82, + "learning_rate": 1.577601956865934e-06, + "loss": 1.1137, + "step": 34272 + }, + { + "epoch": 0.82, + "learning_rate": 1.5771820780608715e-06, + "loss": 0.9806, + "step": 34273 + }, + { + "epoch": 0.82, + "learning_rate": 1.5767622503550683e-06, + "loss": 1.2136, + "step": 34274 + }, + { + "epoch": 0.82, + "learning_rate": 1.5763424737510635e-06, + "loss": 1.0539, + "step": 34275 + }, + { + "epoch": 0.82, + "learning_rate": 1.5759227482514106e-06, + "loss": 1.1732, + "step": 34276 + }, + { + "epoch": 0.82, + "learning_rate": 1.5755030738586507e-06, + "loss": 1.1422, + "step": 34277 + }, + { + "epoch": 0.82, + "learning_rate": 1.5750834505753322e-06, + "loss": 1.0026, + "step": 34278 + }, + { + "epoch": 0.82, + "learning_rate": 1.5746638784040025e-06, + "loss": 1.1827, + "step": 34279 + }, + { + "epoch": 0.82, + "learning_rate": 1.5742443573472032e-06, + "loss": 1.0973, + "step": 34280 + }, + { + "epoch": 0.82, + "learning_rate": 1.5738248874074824e-06, + "loss": 1.2258, + "step": 34281 + }, + { + "epoch": 0.82, + "learning_rate": 1.5734054685873856e-06, + "loss": 1.0101, + "step": 34282 + }, + { + "epoch": 0.82, + "learning_rate": 1.5729861008894532e-06, + "loss": 0.9874, + "step": 34283 + }, + { + "epoch": 0.82, + "learning_rate": 1.572566784316234e-06, + "loss": 1.0186, + "step": 34284 + }, + { + "epoch": 0.82, + "learning_rate": 1.5721475188702685e-06, + "loss": 1.0453, + "step": 34285 + }, + { + "epoch": 0.82, + "learning_rate": 1.5717283045541009e-06, + "loss": 1.2312, + "step": 34286 + }, + { + "epoch": 0.82, + "learning_rate": 1.5713091413702787e-06, + "loss": 1.0259, + "step": 34287 + }, + { + "epoch": 0.82, + "learning_rate": 1.570890029321338e-06, + "loss": 1.1457, + "step": 34288 + }, + { + "epoch": 0.82, + "learning_rate": 1.570470968409824e-06, + "loss": 1.0123, + "step": 34289 + }, + { + "epoch": 0.82, + "learning_rate": 1.5700519586382834e-06, + "loss": 1.0473, + "step": 34290 + }, + { + "epoch": 0.82, + "learning_rate": 1.569633000009253e-06, + "loss": 1.2145, + "step": 34291 + }, + { + "epoch": 0.82, + "learning_rate": 1.5692140925252776e-06, + "loss": 1.1403, + "step": 34292 + }, + { + "epoch": 0.82, + "learning_rate": 1.5687952361888947e-06, + "loss": 0.8503, + "step": 34293 + }, + { + "epoch": 0.82, + "learning_rate": 1.5683764310026496e-06, + "loss": 1.2112, + "step": 34294 + }, + { + "epoch": 0.82, + "learning_rate": 1.5679576769690818e-06, + "loss": 1.1472, + "step": 34295 + }, + { + "epoch": 0.82, + "learning_rate": 1.5675389740907299e-06, + "loss": 1.0113, + "step": 34296 + }, + { + "epoch": 0.82, + "learning_rate": 1.5671203223701358e-06, + "loss": 1.1534, + "step": 34297 + }, + { + "epoch": 0.82, + "learning_rate": 1.5667017218098412e-06, + "loss": 1.0949, + "step": 34298 + }, + { + "epoch": 0.82, + "learning_rate": 1.5662831724123817e-06, + "loss": 0.9276, + "step": 34299 + }, + { + "epoch": 0.82, + "learning_rate": 1.5658646741803008e-06, + "loss": 1.1774, + "step": 34300 + }, + { + "epoch": 0.82, + "learning_rate": 1.565446227116132e-06, + "loss": 0.6486, + "step": 34301 + }, + { + "epoch": 0.82, + "learning_rate": 1.5650278312224176e-06, + "loss": 0.9343, + "step": 34302 + }, + { + "epoch": 0.82, + "learning_rate": 1.5646094865016982e-06, + "loss": 0.9676, + "step": 34303 + }, + { + "epoch": 0.82, + "learning_rate": 1.5641911929565056e-06, + "loss": 0.729, + "step": 34304 + }, + { + "epoch": 0.82, + "learning_rate": 1.5637729505893807e-06, + "loss": 1.1673, + "step": 34305 + }, + { + "epoch": 0.83, + "learning_rate": 1.5633547594028642e-06, + "loss": 1.2019, + "step": 34306 + }, + { + "epoch": 0.83, + "learning_rate": 1.5629366193994866e-06, + "loss": 1.1246, + "step": 34307 + }, + { + "epoch": 0.83, + "learning_rate": 1.5625185305817892e-06, + "loss": 1.1427, + "step": 34308 + }, + { + "epoch": 0.83, + "learning_rate": 1.5621004929523087e-06, + "loss": 1.2534, + "step": 34309 + }, + { + "epoch": 0.83, + "learning_rate": 1.5616825065135778e-06, + "loss": 1.0197, + "step": 34310 + }, + { + "epoch": 0.83, + "learning_rate": 1.5612645712681362e-06, + "loss": 1.0874, + "step": 34311 + }, + { + "epoch": 0.83, + "learning_rate": 1.5608466872185157e-06, + "loss": 1.0199, + "step": 34312 + }, + { + "epoch": 0.83, + "learning_rate": 1.560428854367253e-06, + "loss": 1.0456, + "step": 34313 + }, + { + "epoch": 0.83, + "learning_rate": 1.560011072716886e-06, + "loss": 1.1516, + "step": 34314 + }, + { + "epoch": 0.83, + "learning_rate": 1.559593342269945e-06, + "loss": 1.1318, + "step": 34315 + }, + { + "epoch": 0.83, + "learning_rate": 1.5591756630289645e-06, + "loss": 1.0397, + "step": 34316 + }, + { + "epoch": 0.83, + "learning_rate": 1.5587580349964826e-06, + "loss": 1.0706, + "step": 34317 + }, + { + "epoch": 0.83, + "learning_rate": 1.5583404581750283e-06, + "loss": 1.062, + "step": 34318 + }, + { + "epoch": 0.83, + "learning_rate": 1.5579229325671387e-06, + "loss": 1.1452, + "step": 34319 + }, + { + "epoch": 0.83, + "learning_rate": 1.5575054581753435e-06, + "loss": 1.0427, + "step": 34320 + }, + { + "epoch": 0.83, + "learning_rate": 1.5570880350021756e-06, + "loss": 1.0395, + "step": 34321 + }, + { + "epoch": 0.83, + "learning_rate": 1.5566706630501715e-06, + "loss": 0.8529, + "step": 34322 + }, + { + "epoch": 0.83, + "learning_rate": 1.5562533423218585e-06, + "loss": 1.1025, + "step": 34323 + }, + { + "epoch": 0.83, + "learning_rate": 1.5558360728197707e-06, + "loss": 0.9417, + "step": 34324 + }, + { + "epoch": 0.83, + "learning_rate": 1.555418854546441e-06, + "loss": 0.8685, + "step": 34325 + }, + { + "epoch": 0.83, + "learning_rate": 1.555001687504396e-06, + "loss": 1.0061, + "step": 34326 + }, + { + "epoch": 0.83, + "learning_rate": 1.5545845716961727e-06, + "loss": 0.9053, + "step": 34327 + }, + { + "epoch": 0.83, + "learning_rate": 1.5541675071242956e-06, + "loss": 1.0141, + "step": 34328 + }, + { + "epoch": 0.83, + "learning_rate": 1.5537504937912973e-06, + "loss": 1.06, + "step": 34329 + }, + { + "epoch": 0.83, + "learning_rate": 1.553333531699711e-06, + "loss": 1.0228, + "step": 34330 + }, + { + "epoch": 0.83, + "learning_rate": 1.5529166208520606e-06, + "loss": 1.2346, + "step": 34331 + }, + { + "epoch": 0.83, + "learning_rate": 1.5524997612508785e-06, + "loss": 1.1652, + "step": 34332 + }, + { + "epoch": 0.83, + "learning_rate": 1.5520829528986936e-06, + "loss": 0.9968, + "step": 34333 + }, + { + "epoch": 0.83, + "learning_rate": 1.551666195798034e-06, + "loss": 1.0238, + "step": 34334 + }, + { + "epoch": 0.83, + "learning_rate": 1.5512494899514307e-06, + "loss": 0.8977, + "step": 34335 + }, + { + "epoch": 0.83, + "learning_rate": 1.5508328353614065e-06, + "loss": 1.119, + "step": 34336 + }, + { + "epoch": 0.83, + "learning_rate": 1.5504162320304962e-06, + "loss": 1.019, + "step": 34337 + }, + { + "epoch": 0.83, + "learning_rate": 1.5499996799612193e-06, + "loss": 0.953, + "step": 34338 + }, + { + "epoch": 0.83, + "learning_rate": 1.5495831791561067e-06, + "loss": 0.8912, + "step": 34339 + }, + { + "epoch": 0.83, + "learning_rate": 1.5491667296176882e-06, + "loss": 1.0549, + "step": 34340 + }, + { + "epoch": 0.83, + "learning_rate": 1.5487503313484853e-06, + "loss": 1.0815, + "step": 34341 + }, + { + "epoch": 0.83, + "learning_rate": 1.5483339843510258e-06, + "loss": 1.0269, + "step": 34342 + }, + { + "epoch": 0.83, + "learning_rate": 1.547917688627839e-06, + "loss": 0.9293, + "step": 34343 + }, + { + "epoch": 0.83, + "learning_rate": 1.547501444181445e-06, + "loss": 1.029, + "step": 34344 + }, + { + "epoch": 0.83, + "learning_rate": 1.5470852510143707e-06, + "loss": 1.5007, + "step": 34345 + }, + { + "epoch": 0.83, + "learning_rate": 1.5466691091291452e-06, + "loss": 0.9866, + "step": 34346 + }, + { + "epoch": 0.83, + "learning_rate": 1.5462530185282876e-06, + "loss": 1.2184, + "step": 34347 + }, + { + "epoch": 0.83, + "learning_rate": 1.545836979214327e-06, + "loss": 1.0952, + "step": 34348 + }, + { + "epoch": 0.83, + "learning_rate": 1.5454209911897823e-06, + "loss": 1.2388, + "step": 34349 + }, + { + "epoch": 0.83, + "learning_rate": 1.5450050544571792e-06, + "loss": 1.0299, + "step": 34350 + }, + { + "epoch": 0.83, + "learning_rate": 1.544589169019044e-06, + "loss": 0.9883, + "step": 34351 + }, + { + "epoch": 0.83, + "learning_rate": 1.5441733348778965e-06, + "loss": 1.056, + "step": 34352 + }, + { + "epoch": 0.83, + "learning_rate": 1.543757552036259e-06, + "loss": 0.9596, + "step": 34353 + }, + { + "epoch": 0.83, + "learning_rate": 1.5433418204966588e-06, + "loss": 0.8732, + "step": 34354 + }, + { + "epoch": 0.83, + "learning_rate": 1.5429261402616115e-06, + "loss": 0.8277, + "step": 34355 + }, + { + "epoch": 0.83, + "learning_rate": 1.5425105113336448e-06, + "loss": 0.8391, + "step": 34356 + }, + { + "epoch": 0.83, + "learning_rate": 1.5420949337152746e-06, + "loss": 1.0891, + "step": 34357 + }, + { + "epoch": 0.83, + "learning_rate": 1.5416794074090258e-06, + "loss": 0.8628, + "step": 34358 + }, + { + "epoch": 0.83, + "learning_rate": 1.5412639324174206e-06, + "loss": 1.0004, + "step": 34359 + }, + { + "epoch": 0.83, + "learning_rate": 1.5408485087429747e-06, + "loss": 1.1342, + "step": 34360 + }, + { + "epoch": 0.83, + "learning_rate": 1.540433136388212e-06, + "loss": 0.8449, + "step": 34361 + }, + { + "epoch": 0.83, + "learning_rate": 1.5400178153556533e-06, + "loss": 0.945, + "step": 34362 + }, + { + "epoch": 0.83, + "learning_rate": 1.5396025456478147e-06, + "loss": 1.2171, + "step": 34363 + }, + { + "epoch": 0.83, + "learning_rate": 1.5391873272672197e-06, + "loss": 0.7184, + "step": 34364 + }, + { + "epoch": 0.83, + "learning_rate": 1.538772160216383e-06, + "loss": 0.9878, + "step": 34365 + }, + { + "epoch": 0.83, + "learning_rate": 1.5383570444978245e-06, + "loss": 0.7579, + "step": 34366 + }, + { + "epoch": 0.83, + "learning_rate": 1.5379419801140671e-06, + "loss": 1.2768, + "step": 34367 + }, + { + "epoch": 0.83, + "learning_rate": 1.5375269670676218e-06, + "loss": 1.0718, + "step": 34368 + }, + { + "epoch": 0.83, + "learning_rate": 1.53711200536101e-06, + "loss": 1.1434, + "step": 34369 + }, + { + "epoch": 0.83, + "learning_rate": 1.5366970949967518e-06, + "loss": 1.0245, + "step": 34370 + }, + { + "epoch": 0.83, + "learning_rate": 1.53628223597736e-06, + "loss": 0.9089, + "step": 34371 + }, + { + "epoch": 0.83, + "learning_rate": 1.5358674283053543e-06, + "loss": 1.0652, + "step": 34372 + }, + { + "epoch": 0.83, + "learning_rate": 1.5354526719832484e-06, + "loss": 1.1218, + "step": 34373 + }, + { + "epoch": 0.83, + "learning_rate": 1.5350379670135596e-06, + "loss": 1.2073, + "step": 34374 + }, + { + "epoch": 0.83, + "learning_rate": 1.5346233133988075e-06, + "loss": 0.9663, + "step": 34375 + }, + { + "epoch": 0.83, + "learning_rate": 1.5342087111415015e-06, + "loss": 1.2783, + "step": 34376 + }, + { + "epoch": 0.83, + "learning_rate": 1.5337941602441609e-06, + "loss": 0.9946, + "step": 34377 + }, + { + "epoch": 0.83, + "learning_rate": 1.533379660709301e-06, + "loss": 1.1717, + "step": 34378 + }, + { + "epoch": 0.83, + "learning_rate": 1.5329652125394333e-06, + "loss": 0.9314, + "step": 34379 + }, + { + "epoch": 0.83, + "learning_rate": 1.5325508157370761e-06, + "loss": 0.9385, + "step": 34380 + }, + { + "epoch": 0.83, + "learning_rate": 1.5321364703047392e-06, + "loss": 1.0545, + "step": 34381 + }, + { + "epoch": 0.83, + "learning_rate": 1.5317221762449398e-06, + "loss": 1.0225, + "step": 34382 + }, + { + "epoch": 0.83, + "learning_rate": 1.531307933560191e-06, + "loss": 1.217, + "step": 34383 + }, + { + "epoch": 0.83, + "learning_rate": 1.5308937422530024e-06, + "loss": 0.9278, + "step": 34384 + }, + { + "epoch": 0.83, + "learning_rate": 1.5304796023258905e-06, + "loss": 1.2718, + "step": 34385 + }, + { + "epoch": 0.83, + "learning_rate": 1.5300655137813692e-06, + "loss": 0.8939, + "step": 34386 + }, + { + "epoch": 0.83, + "learning_rate": 1.529651476621945e-06, + "loss": 1.2083, + "step": 34387 + }, + { + "epoch": 0.83, + "learning_rate": 1.5292374908501328e-06, + "loss": 1.1275, + "step": 34388 + }, + { + "epoch": 0.83, + "learning_rate": 1.5288235564684473e-06, + "loss": 0.8079, + "step": 34389 + }, + { + "epoch": 0.83, + "learning_rate": 1.528409673479394e-06, + "loss": 1.1377, + "step": 34390 + }, + { + "epoch": 0.83, + "learning_rate": 1.527995841885489e-06, + "loss": 1.0946, + "step": 34391 + }, + { + "epoch": 0.83, + "learning_rate": 1.527582061689239e-06, + "loss": 0.8611, + "step": 34392 + }, + { + "epoch": 0.83, + "learning_rate": 1.5271683328931542e-06, + "loss": 1.1141, + "step": 34393 + }, + { + "epoch": 0.83, + "learning_rate": 1.5267546554997493e-06, + "loss": 1.0537, + "step": 34394 + }, + { + "epoch": 0.83, + "learning_rate": 1.5263410295115289e-06, + "loss": 0.9584, + "step": 34395 + }, + { + "epoch": 0.83, + "learning_rate": 1.5259274549310043e-06, + "loss": 0.9734, + "step": 34396 + }, + { + "epoch": 0.83, + "learning_rate": 1.525513931760686e-06, + "loss": 1.2619, + "step": 34397 + }, + { + "epoch": 0.83, + "learning_rate": 1.5251004600030804e-06, + "loss": 0.9117, + "step": 34398 + }, + { + "epoch": 0.83, + "learning_rate": 1.5246870396606983e-06, + "loss": 1.244, + "step": 34399 + }, + { + "epoch": 0.83, + "learning_rate": 1.5242736707360438e-06, + "loss": 1.0656, + "step": 34400 + }, + { + "epoch": 0.83, + "learning_rate": 1.5238603532316288e-06, + "loss": 1.0365, + "step": 34401 + }, + { + "epoch": 0.83, + "learning_rate": 1.5234470871499607e-06, + "loss": 0.9466, + "step": 34402 + }, + { + "epoch": 0.83, + "learning_rate": 1.5230338724935423e-06, + "loss": 1.2334, + "step": 34403 + }, + { + "epoch": 0.83, + "learning_rate": 1.5226207092648849e-06, + "loss": 1.1653, + "step": 34404 + }, + { + "epoch": 0.83, + "learning_rate": 1.522207597466495e-06, + "loss": 1.1848, + "step": 34405 + }, + { + "epoch": 0.83, + "learning_rate": 1.5217945371008746e-06, + "loss": 1.1038, + "step": 34406 + }, + { + "epoch": 0.83, + "learning_rate": 1.5213815281705358e-06, + "loss": 1.1061, + "step": 34407 + }, + { + "epoch": 0.83, + "learning_rate": 1.5209685706779786e-06, + "loss": 0.7408, + "step": 34408 + }, + { + "epoch": 0.83, + "learning_rate": 1.5205556646257113e-06, + "loss": 1.0863, + "step": 34409 + }, + { + "epoch": 0.83, + "learning_rate": 1.5201428100162397e-06, + "loss": 0.8692, + "step": 34410 + }, + { + "epoch": 0.83, + "learning_rate": 1.5197300068520648e-06, + "loss": 0.9878, + "step": 34411 + }, + { + "epoch": 0.83, + "learning_rate": 1.5193172551356939e-06, + "loss": 1.1887, + "step": 34412 + }, + { + "epoch": 0.83, + "learning_rate": 1.5189045548696324e-06, + "loss": 1.1415, + "step": 34413 + }, + { + "epoch": 0.83, + "learning_rate": 1.5184919060563796e-06, + "loss": 1.0674, + "step": 34414 + }, + { + "epoch": 0.83, + "learning_rate": 1.5180793086984435e-06, + "loss": 0.9613, + "step": 34415 + }, + { + "epoch": 0.83, + "learning_rate": 1.5176667627983233e-06, + "loss": 0.8936, + "step": 34416 + }, + { + "epoch": 0.83, + "learning_rate": 1.5172542683585234e-06, + "loss": 0.9029, + "step": 34417 + }, + { + "epoch": 0.83, + "learning_rate": 1.5168418253815498e-06, + "loss": 1.074, + "step": 34418 + }, + { + "epoch": 0.83, + "learning_rate": 1.5164294338698981e-06, + "loss": 0.8715, + "step": 34419 + }, + { + "epoch": 0.83, + "learning_rate": 1.5160170938260742e-06, + "loss": 0.9424, + "step": 34420 + }, + { + "epoch": 0.83, + "learning_rate": 1.5156048052525807e-06, + "loss": 0.8856, + "step": 34421 + }, + { + "epoch": 0.83, + "learning_rate": 1.5151925681519152e-06, + "loss": 1.1608, + "step": 34422 + }, + { + "epoch": 0.83, + "learning_rate": 1.514780382526584e-06, + "loss": 1.1388, + "step": 34423 + }, + { + "epoch": 0.83, + "learning_rate": 1.5143682483790812e-06, + "loss": 1.1289, + "step": 34424 + }, + { + "epoch": 0.83, + "learning_rate": 1.5139561657119095e-06, + "loss": 1.2649, + "step": 34425 + }, + { + "epoch": 0.83, + "learning_rate": 1.5135441345275726e-06, + "loss": 0.7151, + "step": 34426 + }, + { + "epoch": 0.83, + "learning_rate": 1.5131321548285648e-06, + "loss": 0.9449, + "step": 34427 + }, + { + "epoch": 0.83, + "learning_rate": 1.512720226617389e-06, + "loss": 1.1616, + "step": 34428 + }, + { + "epoch": 0.83, + "learning_rate": 1.5123083498965452e-06, + "loss": 1.1687, + "step": 34429 + }, + { + "epoch": 0.83, + "learning_rate": 1.5118965246685269e-06, + "loss": 0.8365, + "step": 34430 + }, + { + "epoch": 0.83, + "learning_rate": 1.5114847509358389e-06, + "loss": 1.0541, + "step": 34431 + }, + { + "epoch": 0.83, + "learning_rate": 1.5110730287009744e-06, + "loss": 1.0092, + "step": 34432 + }, + { + "epoch": 0.83, + "learning_rate": 1.510661357966432e-06, + "loss": 0.7591, + "step": 34433 + }, + { + "epoch": 0.83, + "learning_rate": 1.5102497387347138e-06, + "loss": 1.0914, + "step": 34434 + }, + { + "epoch": 0.83, + "learning_rate": 1.5098381710083098e-06, + "loss": 1.1189, + "step": 34435 + }, + { + "epoch": 0.83, + "learning_rate": 1.5094266547897219e-06, + "loss": 0.9891, + "step": 34436 + }, + { + "epoch": 0.83, + "learning_rate": 1.5090151900814476e-06, + "loss": 1.0148, + "step": 34437 + }, + { + "epoch": 0.83, + "learning_rate": 1.5086037768859784e-06, + "loss": 0.8458, + "step": 34438 + }, + { + "epoch": 0.83, + "learning_rate": 1.5081924152058124e-06, + "loss": 0.9131, + "step": 34439 + }, + { + "epoch": 0.83, + "learning_rate": 1.5077811050434477e-06, + "loss": 0.7611, + "step": 34440 + }, + { + "epoch": 0.83, + "learning_rate": 1.5073698464013764e-06, + "loss": 1.0089, + "step": 34441 + }, + { + "epoch": 0.83, + "learning_rate": 1.5069586392820968e-06, + "loss": 1.0809, + "step": 34442 + }, + { + "epoch": 0.83, + "learning_rate": 1.5065474836880977e-06, + "loss": 1.087, + "step": 34443 + }, + { + "epoch": 0.83, + "learning_rate": 1.5061363796218787e-06, + "loss": 1.0615, + "step": 34444 + }, + { + "epoch": 0.83, + "learning_rate": 1.505725327085934e-06, + "loss": 1.2096, + "step": 34445 + }, + { + "epoch": 0.83, + "learning_rate": 1.5053143260827542e-06, + "loss": 0.9745, + "step": 34446 + }, + { + "epoch": 0.83, + "learning_rate": 1.5049033766148336e-06, + "loss": 0.97, + "step": 34447 + }, + { + "epoch": 0.83, + "learning_rate": 1.5044924786846681e-06, + "loss": 1.0564, + "step": 34448 + }, + { + "epoch": 0.83, + "learning_rate": 1.5040816322947471e-06, + "loss": 0.8943, + "step": 34449 + }, + { + "epoch": 0.83, + "learning_rate": 1.5036708374475662e-06, + "loss": 1.0589, + "step": 34450 + }, + { + "epoch": 0.83, + "learning_rate": 1.5032600941456132e-06, + "loss": 0.9768, + "step": 34451 + }, + { + "epoch": 0.83, + "learning_rate": 1.502849402391383e-06, + "loss": 1.213, + "step": 34452 + }, + { + "epoch": 0.83, + "learning_rate": 1.5024387621873692e-06, + "loss": 0.955, + "step": 34453 + }, + { + "epoch": 0.83, + "learning_rate": 1.5020281735360588e-06, + "loss": 1.0669, + "step": 34454 + }, + { + "epoch": 0.83, + "learning_rate": 1.5016176364399438e-06, + "loss": 1.1007, + "step": 34455 + }, + { + "epoch": 0.83, + "learning_rate": 1.5012071509015192e-06, + "loss": 1.2036, + "step": 34456 + }, + { + "epoch": 0.83, + "learning_rate": 1.5007967169232685e-06, + "loss": 1.0025, + "step": 34457 + }, + { + "epoch": 0.83, + "learning_rate": 1.5003863345076875e-06, + "loss": 0.8529, + "step": 34458 + }, + { + "epoch": 0.83, + "learning_rate": 1.4999760036572608e-06, + "loss": 1.0204, + "step": 34459 + }, + { + "epoch": 0.83, + "learning_rate": 1.4995657243744798e-06, + "loss": 1.0458, + "step": 34460 + }, + { + "epoch": 0.83, + "learning_rate": 1.4991554966618382e-06, + "loss": 0.9609, + "step": 34461 + }, + { + "epoch": 0.83, + "learning_rate": 1.4987453205218172e-06, + "loss": 1.094, + "step": 34462 + }, + { + "epoch": 0.83, + "learning_rate": 1.4983351959569082e-06, + "loss": 1.0815, + "step": 34463 + }, + { + "epoch": 0.83, + "learning_rate": 1.4979251229696034e-06, + "loss": 1.0358, + "step": 34464 + }, + { + "epoch": 0.83, + "learning_rate": 1.4975151015623845e-06, + "loss": 0.8902, + "step": 34465 + }, + { + "epoch": 0.83, + "learning_rate": 1.4971051317377427e-06, + "loss": 0.9663, + "step": 34466 + }, + { + "epoch": 0.83, + "learning_rate": 1.4966952134981626e-06, + "loss": 1.1276, + "step": 34467 + }, + { + "epoch": 0.83, + "learning_rate": 1.4962853468461336e-06, + "loss": 0.936, + "step": 34468 + }, + { + "epoch": 0.83, + "learning_rate": 1.4958755317841422e-06, + "loss": 1.0834, + "step": 34469 + }, + { + "epoch": 0.83, + "learning_rate": 1.495465768314671e-06, + "loss": 1.0615, + "step": 34470 + }, + { + "epoch": 0.83, + "learning_rate": 1.4950560564402095e-06, + "loss": 1.0367, + "step": 34471 + }, + { + "epoch": 0.83, + "learning_rate": 1.494646396163245e-06, + "loss": 1.1716, + "step": 34472 + }, + { + "epoch": 0.83, + "learning_rate": 1.4942367874862574e-06, + "loss": 0.9743, + "step": 34473 + }, + { + "epoch": 0.83, + "learning_rate": 1.4938272304117362e-06, + "loss": 0.9471, + "step": 34474 + }, + { + "epoch": 0.83, + "learning_rate": 1.4934177249421633e-06, + "loss": 0.9439, + "step": 34475 + }, + { + "epoch": 0.83, + "learning_rate": 1.4930082710800232e-06, + "loss": 1.0253, + "step": 34476 + }, + { + "epoch": 0.83, + "learning_rate": 1.4925988688278036e-06, + "loss": 1.1934, + "step": 34477 + }, + { + "epoch": 0.83, + "learning_rate": 1.4921895181879841e-06, + "loss": 0.9894, + "step": 34478 + }, + { + "epoch": 0.83, + "learning_rate": 1.4917802191630492e-06, + "loss": 0.8971, + "step": 34479 + }, + { + "epoch": 0.83, + "learning_rate": 1.4913709717554836e-06, + "loss": 1.0934, + "step": 34480 + }, + { + "epoch": 0.83, + "learning_rate": 1.4909617759677685e-06, + "loss": 1.1445, + "step": 34481 + }, + { + "epoch": 0.83, + "learning_rate": 1.4905526318023856e-06, + "loss": 0.8969, + "step": 34482 + }, + { + "epoch": 0.83, + "learning_rate": 1.4901435392618213e-06, + "loss": 1.1514, + "step": 34483 + }, + { + "epoch": 0.83, + "learning_rate": 1.4897344983485518e-06, + "loss": 1.0926, + "step": 34484 + }, + { + "epoch": 0.83, + "learning_rate": 1.4893255090650637e-06, + "loss": 0.9537, + "step": 34485 + }, + { + "epoch": 0.83, + "learning_rate": 1.4889165714138331e-06, + "loss": 1.0606, + "step": 34486 + }, + { + "epoch": 0.83, + "learning_rate": 1.4885076853973445e-06, + "loss": 1.0918, + "step": 34487 + }, + { + "epoch": 0.83, + "learning_rate": 1.4880988510180794e-06, + "loss": 0.8384, + "step": 34488 + }, + { + "epoch": 0.83, + "learning_rate": 1.4876900682785145e-06, + "loss": 1.0118, + "step": 34489 + }, + { + "epoch": 0.83, + "learning_rate": 1.4872813371811324e-06, + "loss": 0.8589, + "step": 34490 + }, + { + "epoch": 0.83, + "learning_rate": 1.486872657728413e-06, + "loss": 0.9652, + "step": 34491 + }, + { + "epoch": 0.83, + "learning_rate": 1.4864640299228339e-06, + "loss": 0.9734, + "step": 34492 + }, + { + "epoch": 0.83, + "learning_rate": 1.486055453766877e-06, + "loss": 1.156, + "step": 34493 + }, + { + "epoch": 0.83, + "learning_rate": 1.4856469292630161e-06, + "loss": 0.8905, + "step": 34494 + }, + { + "epoch": 0.83, + "learning_rate": 1.4852384564137334e-06, + "loss": 1.1175, + "step": 34495 + }, + { + "epoch": 0.83, + "learning_rate": 1.4848300352215073e-06, + "loss": 1.2034, + "step": 34496 + }, + { + "epoch": 0.83, + "learning_rate": 1.4844216656888134e-06, + "loss": 0.7245, + "step": 34497 + }, + { + "epoch": 0.83, + "learning_rate": 1.4840133478181296e-06, + "loss": 1.0451, + "step": 34498 + }, + { + "epoch": 0.83, + "learning_rate": 1.483605081611934e-06, + "loss": 1.1552, + "step": 34499 + }, + { + "epoch": 0.83, + "learning_rate": 1.4831968670727038e-06, + "loss": 0.9303, + "step": 34500 + }, + { + "epoch": 0.83, + "learning_rate": 1.4827887042029153e-06, + "loss": 1.3193, + "step": 34501 + }, + { + "epoch": 0.83, + "learning_rate": 1.4823805930050439e-06, + "loss": 1.0919, + "step": 34502 + }, + { + "epoch": 0.83, + "learning_rate": 1.4819725334815683e-06, + "loss": 0.8364, + "step": 34503 + }, + { + "epoch": 0.83, + "learning_rate": 1.4815645256349587e-06, + "loss": 0.7775, + "step": 34504 + }, + { + "epoch": 0.83, + "learning_rate": 1.4811565694676933e-06, + "loss": 0.9853, + "step": 34505 + }, + { + "epoch": 0.83, + "learning_rate": 1.48074866498225e-06, + "loss": 0.7447, + "step": 34506 + }, + { + "epoch": 0.83, + "learning_rate": 1.480340812181098e-06, + "loss": 1.0371, + "step": 34507 + }, + { + "epoch": 0.83, + "learning_rate": 1.4799330110667142e-06, + "loss": 1.3125, + "step": 34508 + }, + { + "epoch": 0.83, + "learning_rate": 1.4795252616415756e-06, + "loss": 0.8398, + "step": 34509 + }, + { + "epoch": 0.83, + "learning_rate": 1.4791175639081513e-06, + "loss": 0.996, + "step": 34510 + }, + { + "epoch": 0.83, + "learning_rate": 1.478709917868918e-06, + "loss": 0.9455, + "step": 34511 + }, + { + "epoch": 0.83, + "learning_rate": 1.478302323526345e-06, + "loss": 0.9384, + "step": 34512 + }, + { + "epoch": 0.83, + "learning_rate": 1.477894780882908e-06, + "loss": 1.1235, + "step": 34513 + }, + { + "epoch": 0.83, + "learning_rate": 1.477487289941081e-06, + "loss": 0.9155, + "step": 34514 + }, + { + "epoch": 0.83, + "learning_rate": 1.4770798507033312e-06, + "loss": 1.124, + "step": 34515 + }, + { + "epoch": 0.83, + "learning_rate": 1.476672463172133e-06, + "loss": 1.0061, + "step": 34516 + }, + { + "epoch": 0.83, + "learning_rate": 1.4762651273499618e-06, + "loss": 1.0708, + "step": 34517 + }, + { + "epoch": 0.83, + "learning_rate": 1.475857843239281e-06, + "loss": 1.1829, + "step": 34518 + }, + { + "epoch": 0.83, + "learning_rate": 1.475450610842567e-06, + "loss": 0.9742, + "step": 34519 + }, + { + "epoch": 0.83, + "learning_rate": 1.4750434301622907e-06, + "loss": 1.0651, + "step": 34520 + }, + { + "epoch": 0.83, + "learning_rate": 1.474636301200919e-06, + "loss": 1.0357, + "step": 34521 + }, + { + "epoch": 0.83, + "learning_rate": 1.4742292239609257e-06, + "loss": 0.9969, + "step": 34522 + }, + { + "epoch": 0.83, + "learning_rate": 1.4738221984447753e-06, + "loss": 0.8523, + "step": 34523 + }, + { + "epoch": 0.83, + "learning_rate": 1.4734152246549417e-06, + "loss": 0.885, + "step": 34524 + }, + { + "epoch": 0.83, + "learning_rate": 1.4730083025938935e-06, + "loss": 1.0564, + "step": 34525 + }, + { + "epoch": 0.83, + "learning_rate": 1.4726014322640958e-06, + "loss": 1.0016, + "step": 34526 + }, + { + "epoch": 0.83, + "learning_rate": 1.4721946136680198e-06, + "loss": 1.166, + "step": 34527 + }, + { + "epoch": 0.83, + "learning_rate": 1.471787846808136e-06, + "loss": 1.3549, + "step": 34528 + }, + { + "epoch": 0.83, + "learning_rate": 1.4713811316869065e-06, + "loss": 1.1846, + "step": 34529 + }, + { + "epoch": 0.83, + "learning_rate": 1.470974468306804e-06, + "loss": 1.0457, + "step": 34530 + }, + { + "epoch": 0.83, + "learning_rate": 1.4705678566702909e-06, + "loss": 1.0457, + "step": 34531 + }, + { + "epoch": 0.83, + "learning_rate": 1.4701612967798363e-06, + "loss": 1.3064, + "step": 34532 + }, + { + "epoch": 0.83, + "learning_rate": 1.4697547886379094e-06, + "loss": 1.092, + "step": 34533 + }, + { + "epoch": 0.83, + "learning_rate": 1.4693483322469703e-06, + "loss": 1.0049, + "step": 34534 + }, + { + "epoch": 0.83, + "learning_rate": 1.4689419276094895e-06, + "loss": 1.042, + "step": 34535 + }, + { + "epoch": 0.83, + "learning_rate": 1.468535574727934e-06, + "loss": 0.8775, + "step": 34536 + }, + { + "epoch": 0.83, + "learning_rate": 1.4681292736047637e-06, + "loss": 1.0344, + "step": 34537 + }, + { + "epoch": 0.83, + "learning_rate": 1.4677230242424478e-06, + "loss": 1.0058, + "step": 34538 + }, + { + "epoch": 0.83, + "learning_rate": 1.467316826643448e-06, + "loss": 1.1421, + "step": 34539 + }, + { + "epoch": 0.83, + "learning_rate": 1.4669106808102307e-06, + "loss": 1.1244, + "step": 34540 + }, + { + "epoch": 0.83, + "learning_rate": 1.4665045867452598e-06, + "loss": 1.035, + "step": 34541 + }, + { + "epoch": 0.83, + "learning_rate": 1.4660985444509968e-06, + "loss": 1.074, + "step": 34542 + }, + { + "epoch": 0.83, + "learning_rate": 1.4656925539299073e-06, + "loss": 0.8857, + "step": 34543 + }, + { + "epoch": 0.83, + "learning_rate": 1.4652866151844548e-06, + "loss": 1.0157, + "step": 34544 + }, + { + "epoch": 0.83, + "learning_rate": 1.4648807282170995e-06, + "loss": 1.2905, + "step": 34545 + }, + { + "epoch": 0.83, + "learning_rate": 1.4644748930303077e-06, + "loss": 1.0003, + "step": 34546 + }, + { + "epoch": 0.83, + "learning_rate": 1.4640691096265358e-06, + "loss": 1.0976, + "step": 34547 + }, + { + "epoch": 0.83, + "learning_rate": 1.4636633780082499e-06, + "loss": 1.1373, + "step": 34548 + }, + { + "epoch": 0.83, + "learning_rate": 1.4632576981779124e-06, + "loss": 1.132, + "step": 34549 + }, + { + "epoch": 0.83, + "learning_rate": 1.462852070137979e-06, + "loss": 1.1019, + "step": 34550 + }, + { + "epoch": 0.83, + "learning_rate": 1.4624464938909156e-06, + "loss": 1.0175, + "step": 34551 + }, + { + "epoch": 0.83, + "learning_rate": 1.4620409694391824e-06, + "loss": 0.9258, + "step": 34552 + }, + { + "epoch": 0.83, + "learning_rate": 1.4616354967852365e-06, + "loss": 0.9908, + "step": 34553 + }, + { + "epoch": 0.83, + "learning_rate": 1.4612300759315423e-06, + "loss": 1.0234, + "step": 34554 + }, + { + "epoch": 0.83, + "learning_rate": 1.4608247068805537e-06, + "loss": 1.2063, + "step": 34555 + }, + { + "epoch": 0.83, + "learning_rate": 1.460419389634734e-06, + "loss": 1.1122, + "step": 34556 + }, + { + "epoch": 0.83, + "learning_rate": 1.4600141241965425e-06, + "loss": 1.0061, + "step": 34557 + }, + { + "epoch": 0.83, + "learning_rate": 1.4596089105684352e-06, + "loss": 1.0247, + "step": 34558 + }, + { + "epoch": 0.83, + "learning_rate": 1.459203748752872e-06, + "loss": 0.7795, + "step": 34559 + }, + { + "epoch": 0.83, + "learning_rate": 1.4587986387523123e-06, + "loss": 1.0868, + "step": 34560 + }, + { + "epoch": 0.83, + "learning_rate": 1.4583935805692107e-06, + "loss": 0.9983, + "step": 34561 + }, + { + "epoch": 0.83, + "learning_rate": 1.4579885742060252e-06, + "loss": 1.1373, + "step": 34562 + }, + { + "epoch": 0.83, + "learning_rate": 1.457583619665216e-06, + "loss": 0.7752, + "step": 34563 + }, + { + "epoch": 0.83, + "learning_rate": 1.4571787169492369e-06, + "loss": 0.8789, + "step": 34564 + }, + { + "epoch": 0.83, + "learning_rate": 1.4567738660605457e-06, + "loss": 1.1871, + "step": 34565 + }, + { + "epoch": 0.83, + "learning_rate": 1.456369067001596e-06, + "loss": 1.052, + "step": 34566 + }, + { + "epoch": 0.83, + "learning_rate": 1.4559643197748462e-06, + "loss": 1.2772, + "step": 34567 + }, + { + "epoch": 0.83, + "learning_rate": 1.455559624382754e-06, + "loss": 1.1862, + "step": 34568 + }, + { + "epoch": 0.83, + "learning_rate": 1.4551549808277687e-06, + "loss": 0.825, + "step": 34569 + }, + { + "epoch": 0.83, + "learning_rate": 1.4547503891123483e-06, + "loss": 0.9845, + "step": 34570 + }, + { + "epoch": 0.83, + "learning_rate": 1.4543458492389495e-06, + "loss": 1.0292, + "step": 34571 + }, + { + "epoch": 0.83, + "learning_rate": 1.4539413612100228e-06, + "loss": 0.9697, + "step": 34572 + }, + { + "epoch": 0.83, + "learning_rate": 1.453536925028025e-06, + "loss": 1.101, + "step": 34573 + }, + { + "epoch": 0.83, + "learning_rate": 1.4531325406954066e-06, + "loss": 1.0823, + "step": 34574 + }, + { + "epoch": 0.83, + "learning_rate": 1.4527282082146222e-06, + "loss": 1.0281, + "step": 34575 + }, + { + "epoch": 0.83, + "learning_rate": 1.4523239275881284e-06, + "loss": 0.9967, + "step": 34576 + }, + { + "epoch": 0.83, + "learning_rate": 1.4519196988183726e-06, + "loss": 0.8449, + "step": 34577 + }, + { + "epoch": 0.83, + "learning_rate": 1.4515155219078091e-06, + "loss": 1.0824, + "step": 34578 + }, + { + "epoch": 0.83, + "learning_rate": 1.4511113968588918e-06, + "loss": 0.9096, + "step": 34579 + }, + { + "epoch": 0.83, + "learning_rate": 1.4507073236740698e-06, + "loss": 0.9804, + "step": 34580 + }, + { + "epoch": 0.83, + "learning_rate": 1.4503033023557966e-06, + "loss": 1.1516, + "step": 34581 + }, + { + "epoch": 0.83, + "learning_rate": 1.4498993329065203e-06, + "loss": 1.2186, + "step": 34582 + }, + { + "epoch": 0.83, + "learning_rate": 1.4494954153286944e-06, + "loss": 0.9991, + "step": 34583 + }, + { + "epoch": 0.83, + "learning_rate": 1.4490915496247704e-06, + "loss": 0.8536, + "step": 34584 + }, + { + "epoch": 0.83, + "learning_rate": 1.448687735797194e-06, + "loss": 1.0439, + "step": 34585 + }, + { + "epoch": 0.83, + "learning_rate": 1.4482839738484178e-06, + "loss": 1.1458, + "step": 34586 + }, + { + "epoch": 0.83, + "learning_rate": 1.447880263780894e-06, + "loss": 0.6919, + "step": 34587 + }, + { + "epoch": 0.83, + "learning_rate": 1.4474766055970668e-06, + "loss": 0.8322, + "step": 34588 + }, + { + "epoch": 0.83, + "learning_rate": 1.4470729992993892e-06, + "loss": 0.9703, + "step": 34589 + }, + { + "epoch": 0.83, + "learning_rate": 1.4466694448903063e-06, + "loss": 1.1265, + "step": 34590 + }, + { + "epoch": 0.83, + "learning_rate": 1.4462659423722669e-06, + "loss": 0.9543, + "step": 34591 + }, + { + "epoch": 0.83, + "learning_rate": 1.4458624917477238e-06, + "loss": 1.1196, + "step": 34592 + }, + { + "epoch": 0.83, + "learning_rate": 1.445459093019117e-06, + "loss": 0.9739, + "step": 34593 + }, + { + "epoch": 0.83, + "learning_rate": 1.4450557461888993e-06, + "loss": 0.9937, + "step": 34594 + }, + { + "epoch": 0.83, + "learning_rate": 1.4446524512595172e-06, + "loss": 1.2228, + "step": 34595 + }, + { + "epoch": 0.83, + "learning_rate": 1.4442492082334148e-06, + "loss": 0.8967, + "step": 34596 + }, + { + "epoch": 0.83, + "learning_rate": 1.443846017113042e-06, + "loss": 1.0905, + "step": 34597 + }, + { + "epoch": 0.83, + "learning_rate": 1.4434428779008403e-06, + "loss": 1.134, + "step": 34598 + }, + { + "epoch": 0.83, + "learning_rate": 1.4430397905992588e-06, + "loss": 1.2609, + "step": 34599 + }, + { + "epoch": 0.83, + "learning_rate": 1.4426367552107435e-06, + "loss": 1.0908, + "step": 34600 + }, + { + "epoch": 0.83, + "learning_rate": 1.4422337717377366e-06, + "loss": 1.0331, + "step": 34601 + }, + { + "epoch": 0.83, + "learning_rate": 1.4418308401826842e-06, + "loss": 0.9075, + "step": 34602 + }, + { + "epoch": 0.83, + "learning_rate": 1.4414279605480342e-06, + "loss": 0.9704, + "step": 34603 + }, + { + "epoch": 0.83, + "learning_rate": 1.4410251328362245e-06, + "loss": 1.0605, + "step": 34604 + }, + { + "epoch": 0.83, + "learning_rate": 1.4406223570497025e-06, + "loss": 1.1248, + "step": 34605 + }, + { + "epoch": 0.83, + "learning_rate": 1.4402196331909136e-06, + "loss": 1.0332, + "step": 34606 + }, + { + "epoch": 0.83, + "learning_rate": 1.4398169612622958e-06, + "loss": 1.0481, + "step": 34607 + }, + { + "epoch": 0.83, + "learning_rate": 1.4394143412662987e-06, + "loss": 1.3444, + "step": 34608 + }, + { + "epoch": 0.83, + "learning_rate": 1.4390117732053576e-06, + "loss": 0.7559, + "step": 34609 + }, + { + "epoch": 0.83, + "learning_rate": 1.4386092570819188e-06, + "loss": 1.1324, + "step": 34610 + }, + { + "epoch": 0.83, + "learning_rate": 1.4382067928984266e-06, + "loss": 1.0423, + "step": 34611 + }, + { + "epoch": 0.83, + "learning_rate": 1.4378043806573173e-06, + "loss": 0.8246, + "step": 34612 + }, + { + "epoch": 0.83, + "learning_rate": 1.437402020361035e-06, + "loss": 0.9127, + "step": 34613 + }, + { + "epoch": 0.83, + "learning_rate": 1.4369997120120228e-06, + "loss": 1.1349, + "step": 34614 + }, + { + "epoch": 0.83, + "learning_rate": 1.436597455612717e-06, + "loss": 1.1344, + "step": 34615 + }, + { + "epoch": 0.83, + "learning_rate": 1.4361952511655618e-06, + "loss": 1.0829, + "step": 34616 + }, + { + "epoch": 0.83, + "learning_rate": 1.435793098672994e-06, + "loss": 1.3618, + "step": 34617 + }, + { + "epoch": 0.83, + "learning_rate": 1.4353909981374559e-06, + "loss": 1.1471, + "step": 34618 + }, + { + "epoch": 0.83, + "learning_rate": 1.4349889495613866e-06, + "loss": 1.4386, + "step": 34619 + }, + { + "epoch": 0.83, + "learning_rate": 1.4345869529472235e-06, + "loss": 0.8697, + "step": 34620 + }, + { + "epoch": 0.83, + "learning_rate": 1.4341850082974075e-06, + "loss": 1.2525, + "step": 34621 + }, + { + "epoch": 0.83, + "learning_rate": 1.4337831156143767e-06, + "loss": 0.8072, + "step": 34622 + }, + { + "epoch": 0.83, + "learning_rate": 1.433381274900567e-06, + "loss": 0.8672, + "step": 34623 + }, + { + "epoch": 0.83, + "learning_rate": 1.4329794861584212e-06, + "loss": 1.1187, + "step": 34624 + }, + { + "epoch": 0.83, + "learning_rate": 1.4325777493903714e-06, + "loss": 1.0249, + "step": 34625 + }, + { + "epoch": 0.83, + "learning_rate": 1.432176064598857e-06, + "loss": 0.8224, + "step": 34626 + }, + { + "epoch": 0.83, + "learning_rate": 1.4317744317863169e-06, + "loss": 0.9716, + "step": 34627 + }, + { + "epoch": 0.83, + "learning_rate": 1.4313728509551838e-06, + "loss": 1.1594, + "step": 34628 + }, + { + "epoch": 0.83, + "learning_rate": 1.4309713221078968e-06, + "loss": 1.1647, + "step": 34629 + }, + { + "epoch": 0.83, + "learning_rate": 1.430569845246893e-06, + "loss": 0.8672, + "step": 34630 + }, + { + "epoch": 0.83, + "learning_rate": 1.4301684203746046e-06, + "loss": 1.1763, + "step": 34631 + }, + { + "epoch": 0.83, + "learning_rate": 1.42976704749347e-06, + "loss": 1.0906, + "step": 34632 + }, + { + "epoch": 0.83, + "learning_rate": 1.4293657266059214e-06, + "loss": 1.0684, + "step": 34633 + }, + { + "epoch": 0.83, + "learning_rate": 1.4289644577143947e-06, + "loss": 1.1178, + "step": 34634 + }, + { + "epoch": 0.83, + "learning_rate": 1.4285632408213268e-06, + "loss": 1.2247, + "step": 34635 + }, + { + "epoch": 0.83, + "learning_rate": 1.428162075929148e-06, + "loss": 1.0432, + "step": 34636 + }, + { + "epoch": 0.83, + "learning_rate": 1.4277609630402932e-06, + "loss": 0.9452, + "step": 34637 + }, + { + "epoch": 0.83, + "learning_rate": 1.4273599021571982e-06, + "loss": 1.1424, + "step": 34638 + }, + { + "epoch": 0.83, + "learning_rate": 1.426958893282292e-06, + "loss": 1.1773, + "step": 34639 + }, + { + "epoch": 0.83, + "learning_rate": 1.4265579364180126e-06, + "loss": 1.0313, + "step": 34640 + }, + { + "epoch": 0.83, + "learning_rate": 1.426157031566787e-06, + "loss": 1.1455, + "step": 34641 + }, + { + "epoch": 0.83, + "learning_rate": 1.4257561787310492e-06, + "loss": 1.0898, + "step": 34642 + }, + { + "epoch": 0.83, + "learning_rate": 1.4253553779132357e-06, + "loss": 1.174, + "step": 34643 + }, + { + "epoch": 0.83, + "learning_rate": 1.4249546291157712e-06, + "loss": 1.1061, + "step": 34644 + }, + { + "epoch": 0.83, + "learning_rate": 1.4245539323410894e-06, + "loss": 1.2246, + "step": 34645 + }, + { + "epoch": 0.83, + "learning_rate": 1.4241532875916253e-06, + "loss": 1.1591, + "step": 34646 + }, + { + "epoch": 0.83, + "learning_rate": 1.4237526948698033e-06, + "loss": 0.9667, + "step": 34647 + }, + { + "epoch": 0.83, + "learning_rate": 1.4233521541780583e-06, + "loss": 1.0817, + "step": 34648 + }, + { + "epoch": 0.83, + "learning_rate": 1.4229516655188158e-06, + "loss": 0.7823, + "step": 34649 + }, + { + "epoch": 0.83, + "learning_rate": 1.4225512288945088e-06, + "loss": 0.8233, + "step": 34650 + }, + { + "epoch": 0.83, + "learning_rate": 1.4221508443075681e-06, + "loss": 0.9666, + "step": 34651 + }, + { + "epoch": 0.83, + "learning_rate": 1.4217505117604192e-06, + "loss": 1.0346, + "step": 34652 + }, + { + "epoch": 0.83, + "learning_rate": 1.4213502312554906e-06, + "loss": 1.1309, + "step": 34653 + }, + { + "epoch": 0.83, + "learning_rate": 1.420950002795216e-06, + "loss": 1.1088, + "step": 34654 + }, + { + "epoch": 0.83, + "learning_rate": 1.4205498263820162e-06, + "loss": 1.1884, + "step": 34655 + }, + { + "epoch": 0.83, + "learning_rate": 1.4201497020183241e-06, + "loss": 1.2254, + "step": 34656 + }, + { + "epoch": 0.83, + "learning_rate": 1.4197496297065672e-06, + "loss": 0.9924, + "step": 34657 + }, + { + "epoch": 0.83, + "learning_rate": 1.4193496094491688e-06, + "loss": 1.0304, + "step": 34658 + }, + { + "epoch": 0.83, + "learning_rate": 1.4189496412485593e-06, + "loss": 0.8642, + "step": 34659 + }, + { + "epoch": 0.83, + "learning_rate": 1.4185497251071633e-06, + "loss": 1.0618, + "step": 34660 + }, + { + "epoch": 0.83, + "learning_rate": 1.418149861027407e-06, + "loss": 1.2452, + "step": 34661 + }, + { + "epoch": 0.83, + "learning_rate": 1.417750049011719e-06, + "loss": 1.1473, + "step": 34662 + }, + { + "epoch": 0.83, + "learning_rate": 1.41735028906252e-06, + "loss": 1.0285, + "step": 34663 + }, + { + "epoch": 0.83, + "learning_rate": 1.4169505811822393e-06, + "loss": 1.1056, + "step": 34664 + }, + { + "epoch": 0.83, + "learning_rate": 1.4165509253733023e-06, + "loss": 1.0186, + "step": 34665 + }, + { + "epoch": 0.83, + "learning_rate": 1.4161513216381273e-06, + "loss": 0.8524, + "step": 34666 + }, + { + "epoch": 0.83, + "learning_rate": 1.4157517699791478e-06, + "loss": 0.9016, + "step": 34667 + }, + { + "epoch": 0.83, + "learning_rate": 1.415352270398781e-06, + "loss": 1.0224, + "step": 34668 + }, + { + "epoch": 0.83, + "learning_rate": 1.4149528228994547e-06, + "loss": 0.9602, + "step": 34669 + }, + { + "epoch": 0.83, + "learning_rate": 1.4145534274835882e-06, + "loss": 1.0255, + "step": 34670 + }, + { + "epoch": 0.83, + "learning_rate": 1.414154084153606e-06, + "loss": 1.1556, + "step": 34671 + }, + { + "epoch": 0.83, + "learning_rate": 1.4137547929119343e-06, + "loss": 1.0136, + "step": 34672 + }, + { + "epoch": 0.83, + "learning_rate": 1.413355553760991e-06, + "loss": 0.9759, + "step": 34673 + }, + { + "epoch": 0.83, + "learning_rate": 1.4129563667031987e-06, + "loss": 1.0999, + "step": 34674 + }, + { + "epoch": 0.83, + "learning_rate": 1.412557231740983e-06, + "loss": 0.9789, + "step": 34675 + }, + { + "epoch": 0.83, + "learning_rate": 1.412158148876761e-06, + "loss": 1.0469, + "step": 34676 + }, + { + "epoch": 0.83, + "learning_rate": 1.4117591181129575e-06, + "loss": 0.9225, + "step": 34677 + }, + { + "epoch": 0.83, + "learning_rate": 1.4113601394519894e-06, + "loss": 1.0428, + "step": 34678 + }, + { + "epoch": 0.83, + "learning_rate": 1.410961212896278e-06, + "loss": 0.8678, + "step": 34679 + }, + { + "epoch": 0.83, + "learning_rate": 1.410562338448248e-06, + "loss": 0.914, + "step": 34680 + }, + { + "epoch": 0.83, + "learning_rate": 1.4101635161103134e-06, + "loss": 1.0431, + "step": 34681 + }, + { + "epoch": 0.83, + "learning_rate": 1.4097647458848961e-06, + "loss": 0.954, + "step": 34682 + }, + { + "epoch": 0.83, + "learning_rate": 1.4093660277744182e-06, + "loss": 0.9017, + "step": 34683 + }, + { + "epoch": 0.83, + "learning_rate": 1.4089673617812927e-06, + "loss": 1.1239, + "step": 34684 + }, + { + "epoch": 0.83, + "learning_rate": 1.4085687479079424e-06, + "loss": 1.097, + "step": 34685 + }, + { + "epoch": 0.83, + "learning_rate": 1.4081701861567864e-06, + "loss": 1.1262, + "step": 34686 + }, + { + "epoch": 0.83, + "learning_rate": 1.4077716765302395e-06, + "loss": 0.844, + "step": 34687 + }, + { + "epoch": 0.83, + "learning_rate": 1.4073732190307232e-06, + "loss": 1.1957, + "step": 34688 + }, + { + "epoch": 0.83, + "learning_rate": 1.4069748136606487e-06, + "loss": 1.1744, + "step": 34689 + }, + { + "epoch": 0.83, + "learning_rate": 1.4065764604224386e-06, + "loss": 1.0696, + "step": 34690 + }, + { + "epoch": 0.83, + "learning_rate": 1.4061781593185086e-06, + "loss": 1.1074, + "step": 34691 + }, + { + "epoch": 0.83, + "learning_rate": 1.4057799103512725e-06, + "loss": 1.0765, + "step": 34692 + }, + { + "epoch": 0.83, + "learning_rate": 1.4053817135231485e-06, + "loss": 1.1038, + "step": 34693 + }, + { + "epoch": 0.83, + "learning_rate": 1.4049835688365543e-06, + "loss": 0.7657, + "step": 34694 + }, + { + "epoch": 0.83, + "learning_rate": 1.4045854762939004e-06, + "loss": 0.9274, + "step": 34695 + }, + { + "epoch": 0.83, + "learning_rate": 1.4041874358976081e-06, + "loss": 1.0339, + "step": 34696 + }, + { + "epoch": 0.83, + "learning_rate": 1.4037894476500857e-06, + "loss": 1.3077, + "step": 34697 + }, + { + "epoch": 0.83, + "learning_rate": 1.40339151155375e-06, + "loss": 0.9841, + "step": 34698 + }, + { + "epoch": 0.83, + "learning_rate": 1.4029936276110189e-06, + "loss": 1.012, + "step": 34699 + }, + { + "epoch": 0.83, + "learning_rate": 1.402595795824302e-06, + "loss": 1.0168, + "step": 34700 + }, + { + "epoch": 0.83, + "learning_rate": 1.4021980161960124e-06, + "loss": 1.1155, + "step": 34701 + }, + { + "epoch": 0.83, + "learning_rate": 1.4018002887285687e-06, + "loss": 1.0334, + "step": 34702 + }, + { + "epoch": 0.83, + "learning_rate": 1.4014026134243774e-06, + "loss": 0.8022, + "step": 34703 + }, + { + "epoch": 0.83, + "learning_rate": 1.401004990285857e-06, + "loss": 0.9433, + "step": 34704 + }, + { + "epoch": 0.83, + "learning_rate": 1.400607419315413e-06, + "loss": 1.084, + "step": 34705 + }, + { + "epoch": 0.83, + "learning_rate": 1.4002099005154624e-06, + "loss": 1.0068, + "step": 34706 + }, + { + "epoch": 0.83, + "learning_rate": 1.3998124338884167e-06, + "loss": 0.8421, + "step": 34707 + }, + { + "epoch": 0.83, + "learning_rate": 1.3994150194366851e-06, + "loss": 1.0533, + "step": 34708 + }, + { + "epoch": 0.83, + "learning_rate": 1.3990176571626778e-06, + "loss": 1.03, + "step": 34709 + }, + { + "epoch": 0.83, + "learning_rate": 1.3986203470688109e-06, + "loss": 0.8729, + "step": 34710 + }, + { + "epoch": 0.83, + "learning_rate": 1.3982230891574889e-06, + "loss": 1.1311, + "step": 34711 + }, + { + "epoch": 0.83, + "learning_rate": 1.3978258834311264e-06, + "loss": 1.1242, + "step": 34712 + }, + { + "epoch": 0.83, + "learning_rate": 1.3974287298921284e-06, + "loss": 0.9529, + "step": 34713 + }, + { + "epoch": 0.83, + "learning_rate": 1.3970316285429065e-06, + "loss": 1.0144, + "step": 34714 + }, + { + "epoch": 0.83, + "learning_rate": 1.3966345793858727e-06, + "loss": 0.9005, + "step": 34715 + }, + { + "epoch": 0.83, + "learning_rate": 1.396237582423432e-06, + "loss": 0.9673, + "step": 34716 + }, + { + "epoch": 0.83, + "learning_rate": 1.3958406376579925e-06, + "loss": 0.8659, + "step": 34717 + }, + { + "epoch": 0.83, + "learning_rate": 1.3954437450919679e-06, + "loss": 0.9586, + "step": 34718 + }, + { + "epoch": 0.83, + "learning_rate": 1.395046904727758e-06, + "loss": 1.0934, + "step": 34719 + }, + { + "epoch": 0.83, + "learning_rate": 1.3946501165677783e-06, + "loss": 0.926, + "step": 34720 + }, + { + "epoch": 0.84, + "learning_rate": 1.3942533806144299e-06, + "loss": 1.0151, + "step": 34721 + }, + { + "epoch": 0.84, + "learning_rate": 1.3938566968701217e-06, + "loss": 1.0123, + "step": 34722 + }, + { + "epoch": 0.84, + "learning_rate": 1.3934600653372633e-06, + "loss": 0.8395, + "step": 34723 + }, + { + "epoch": 0.84, + "learning_rate": 1.3930634860182546e-06, + "loss": 1.3782, + "step": 34724 + }, + { + "epoch": 0.84, + "learning_rate": 1.3926669589155063e-06, + "loss": 1.0944, + "step": 34725 + }, + { + "epoch": 0.84, + "learning_rate": 1.392270484031426e-06, + "loss": 1.0142, + "step": 34726 + }, + { + "epoch": 0.84, + "learning_rate": 1.3918740613684123e-06, + "loss": 1.0618, + "step": 34727 + }, + { + "epoch": 0.84, + "learning_rate": 1.3914776909288774e-06, + "loss": 1.0603, + "step": 34728 + }, + { + "epoch": 0.84, + "learning_rate": 1.3910813727152195e-06, + "loss": 1.1015, + "step": 34729 + }, + { + "epoch": 0.84, + "learning_rate": 1.3906851067298466e-06, + "loss": 1.0581, + "step": 34730 + }, + { + "epoch": 0.84, + "learning_rate": 1.3902888929751634e-06, + "loss": 1.1741, + "step": 34731 + }, + { + "epoch": 0.84, + "learning_rate": 1.3898927314535705e-06, + "loss": 1.1292, + "step": 34732 + }, + { + "epoch": 0.84, + "learning_rate": 1.3894966221674733e-06, + "loss": 1.0535, + "step": 34733 + }, + { + "epoch": 0.84, + "learning_rate": 1.3891005651192769e-06, + "loss": 0.834, + "step": 34734 + }, + { + "epoch": 0.84, + "learning_rate": 1.3887045603113803e-06, + "loss": 1.0637, + "step": 34735 + }, + { + "epoch": 0.84, + "learning_rate": 1.3883086077461872e-06, + "loss": 0.7424, + "step": 34736 + }, + { + "epoch": 0.84, + "learning_rate": 1.3879127074261024e-06, + "loss": 0.977, + "step": 34737 + }, + { + "epoch": 0.84, + "learning_rate": 1.3875168593535237e-06, + "loss": 0.9521, + "step": 34738 + }, + { + "epoch": 0.84, + "learning_rate": 1.3871210635308564e-06, + "loss": 1.0704, + "step": 34739 + }, + { + "epoch": 0.84, + "learning_rate": 1.3867253199604969e-06, + "loss": 1.002, + "step": 34740 + }, + { + "epoch": 0.84, + "learning_rate": 1.386329628644849e-06, + "loss": 0.9356, + "step": 34741 + }, + { + "epoch": 0.84, + "learning_rate": 1.3859339895863167e-06, + "loss": 1.1049, + "step": 34742 + }, + { + "epoch": 0.84, + "learning_rate": 1.385538402787293e-06, + "loss": 1.0909, + "step": 34743 + }, + { + "epoch": 0.84, + "learning_rate": 1.3851428682501821e-06, + "loss": 0.9845, + "step": 34744 + }, + { + "epoch": 0.84, + "learning_rate": 1.3847473859773853e-06, + "loss": 0.9087, + "step": 34745 + }, + { + "epoch": 0.84, + "learning_rate": 1.3843519559712981e-06, + "loss": 1.0679, + "step": 34746 + }, + { + "epoch": 0.84, + "learning_rate": 1.3839565782343222e-06, + "loss": 0.9833, + "step": 34747 + }, + { + "epoch": 0.84, + "learning_rate": 1.3835612527688535e-06, + "loss": 1.0038, + "step": 34748 + }, + { + "epoch": 0.84, + "learning_rate": 1.3831659795772922e-06, + "loss": 0.8706, + "step": 34749 + }, + { + "epoch": 0.84, + "learning_rate": 1.3827707586620375e-06, + "loss": 1.0619, + "step": 34750 + }, + { + "epoch": 0.84, + "learning_rate": 1.3823755900254843e-06, + "loss": 0.8923, + "step": 34751 + }, + { + "epoch": 0.84, + "learning_rate": 1.3819804736700315e-06, + "loss": 1.1292, + "step": 34752 + }, + { + "epoch": 0.84, + "learning_rate": 1.3815854095980774e-06, + "loss": 1.0245, + "step": 34753 + }, + { + "epoch": 0.84, + "learning_rate": 1.3811903978120156e-06, + "loss": 1.0615, + "step": 34754 + }, + { + "epoch": 0.84, + "learning_rate": 1.3807954383142475e-06, + "loss": 1.1697, + "step": 34755 + }, + { + "epoch": 0.84, + "learning_rate": 1.3804005311071633e-06, + "loss": 1.1455, + "step": 34756 + }, + { + "epoch": 0.84, + "learning_rate": 1.3800056761931614e-06, + "loss": 1.0768, + "step": 34757 + }, + { + "epoch": 0.84, + "learning_rate": 1.3796108735746395e-06, + "loss": 1.0607, + "step": 34758 + }, + { + "epoch": 0.84, + "learning_rate": 1.3792161232539892e-06, + "loss": 1.0171, + "step": 34759 + }, + { + "epoch": 0.84, + "learning_rate": 1.3788214252336063e-06, + "loss": 0.9527, + "step": 34760 + }, + { + "epoch": 0.84, + "learning_rate": 1.378426779515889e-06, + "loss": 0.9057, + "step": 34761 + }, + { + "epoch": 0.84, + "learning_rate": 1.3780321861032252e-06, + "loss": 1.1211, + "step": 34762 + }, + { + "epoch": 0.84, + "learning_rate": 1.3776376449980156e-06, + "loss": 1.2172, + "step": 34763 + }, + { + "epoch": 0.84, + "learning_rate": 1.3772431562026467e-06, + "loss": 1.1698, + "step": 34764 + }, + { + "epoch": 0.84, + "learning_rate": 1.376848719719517e-06, + "loss": 0.847, + "step": 34765 + }, + { + "epoch": 0.84, + "learning_rate": 1.3764543355510186e-06, + "loss": 1.0448, + "step": 34766 + }, + { + "epoch": 0.84, + "learning_rate": 1.376060003699542e-06, + "loss": 1.1107, + "step": 34767 + }, + { + "epoch": 0.84, + "learning_rate": 1.3756657241674798e-06, + "loss": 0.9675, + "step": 34768 + }, + { + "epoch": 0.84, + "learning_rate": 1.375271496957229e-06, + "loss": 0.8721, + "step": 34769 + }, + { + "epoch": 0.84, + "learning_rate": 1.3748773220711741e-06, + "loss": 1.1431, + "step": 34770 + }, + { + "epoch": 0.84, + "learning_rate": 1.3744831995117113e-06, + "loss": 0.8851, + "step": 34771 + }, + { + "epoch": 0.84, + "learning_rate": 1.3740891292812287e-06, + "loss": 0.9788, + "step": 34772 + }, + { + "epoch": 0.84, + "learning_rate": 1.3736951113821184e-06, + "loss": 0.9125, + "step": 34773 + }, + { + "epoch": 0.84, + "learning_rate": 1.3733011458167734e-06, + "loss": 0.8349, + "step": 34774 + }, + { + "epoch": 0.84, + "learning_rate": 1.3729072325875791e-06, + "loss": 1.0397, + "step": 34775 + }, + { + "epoch": 0.84, + "learning_rate": 1.3725133716969263e-06, + "loss": 0.758, + "step": 34776 + }, + { + "epoch": 0.84, + "learning_rate": 1.3721195631472085e-06, + "loss": 0.9498, + "step": 34777 + }, + { + "epoch": 0.84, + "learning_rate": 1.3717258069408102e-06, + "loss": 1.0582, + "step": 34778 + }, + { + "epoch": 0.84, + "learning_rate": 1.3713321030801208e-06, + "loss": 0.9839, + "step": 34779 + }, + { + "epoch": 0.84, + "learning_rate": 1.3709384515675327e-06, + "loss": 1.0236, + "step": 34780 + }, + { + "epoch": 0.84, + "learning_rate": 1.3705448524054288e-06, + "loss": 1.0413, + "step": 34781 + }, + { + "epoch": 0.84, + "learning_rate": 1.3701513055962024e-06, + "loss": 0.9542, + "step": 34782 + }, + { + "epoch": 0.84, + "learning_rate": 1.369757811142236e-06, + "loss": 1.2505, + "step": 34783 + }, + { + "epoch": 0.84, + "learning_rate": 1.3693643690459179e-06, + "loss": 1.0763, + "step": 34784 + }, + { + "epoch": 0.84, + "learning_rate": 1.3689709793096395e-06, + "loss": 0.8883, + "step": 34785 + }, + { + "epoch": 0.84, + "learning_rate": 1.368577641935781e-06, + "loss": 1.1798, + "step": 34786 + }, + { + "epoch": 0.84, + "learning_rate": 1.3681843569267327e-06, + "loss": 1.0212, + "step": 34787 + }, + { + "epoch": 0.84, + "learning_rate": 1.3677911242848807e-06, + "loss": 1.0471, + "step": 34788 + }, + { + "epoch": 0.84, + "learning_rate": 1.3673979440126085e-06, + "loss": 0.9196, + "step": 34789 + }, + { + "epoch": 0.84, + "learning_rate": 1.367004816112303e-06, + "loss": 1.02, + "step": 34790 + }, + { + "epoch": 0.84, + "learning_rate": 1.3666117405863477e-06, + "loss": 1.0557, + "step": 34791 + }, + { + "epoch": 0.84, + "learning_rate": 1.3662187174371278e-06, + "loss": 1.1971, + "step": 34792 + }, + { + "epoch": 0.84, + "learning_rate": 1.365825746667031e-06, + "loss": 1.1148, + "step": 34793 + }, + { + "epoch": 0.84, + "learning_rate": 1.3654328282784358e-06, + "loss": 1.1639, + "step": 34794 + }, + { + "epoch": 0.84, + "learning_rate": 1.3650399622737286e-06, + "loss": 0.8278, + "step": 34795 + }, + { + "epoch": 0.84, + "learning_rate": 1.3646471486552947e-06, + "loss": 0.9811, + "step": 34796 + }, + { + "epoch": 0.84, + "learning_rate": 1.3642543874255142e-06, + "loss": 0.7506, + "step": 34797 + }, + { + "epoch": 0.84, + "learning_rate": 1.363861678586773e-06, + "loss": 0.7716, + "step": 34798 + }, + { + "epoch": 0.84, + "learning_rate": 1.3634690221414492e-06, + "loss": 1.0542, + "step": 34799 + }, + { + "epoch": 0.84, + "learning_rate": 1.3630764180919276e-06, + "loss": 0.9813, + "step": 34800 + }, + { + "epoch": 0.84, + "learning_rate": 1.3626838664405916e-06, + "loss": 1.2228, + "step": 34801 + }, + { + "epoch": 0.84, + "learning_rate": 1.3622913671898186e-06, + "loss": 0.8246, + "step": 34802 + }, + { + "epoch": 0.84, + "learning_rate": 1.361898920341993e-06, + "loss": 1.1283, + "step": 34803 + }, + { + "epoch": 0.84, + "learning_rate": 1.3615065258994963e-06, + "loss": 1.0517, + "step": 34804 + }, + { + "epoch": 0.84, + "learning_rate": 1.3611141838647057e-06, + "loss": 1.0351, + "step": 34805 + }, + { + "epoch": 0.84, + "learning_rate": 1.3607218942400057e-06, + "loss": 1.3232, + "step": 34806 + }, + { + "epoch": 0.84, + "learning_rate": 1.3603296570277713e-06, + "loss": 0.8905, + "step": 34807 + }, + { + "epoch": 0.84, + "learning_rate": 1.3599374722303848e-06, + "loss": 0.9911, + "step": 34808 + }, + { + "epoch": 0.84, + "learning_rate": 1.3595453398502279e-06, + "loss": 0.974, + "step": 34809 + }, + { + "epoch": 0.84, + "learning_rate": 1.3591532598896751e-06, + "loss": 1.3158, + "step": 34810 + }, + { + "epoch": 0.84, + "learning_rate": 1.3587612323511068e-06, + "loss": 1.0145, + "step": 34811 + }, + { + "epoch": 0.84, + "learning_rate": 1.3583692572369034e-06, + "loss": 1.346, + "step": 34812 + }, + { + "epoch": 0.84, + "learning_rate": 1.3579773345494397e-06, + "loss": 1.0684, + "step": 34813 + }, + { + "epoch": 0.84, + "learning_rate": 1.3575854642910957e-06, + "loss": 0.8142, + "step": 34814 + }, + { + "epoch": 0.84, + "learning_rate": 1.3571936464642466e-06, + "loss": 0.8485, + "step": 34815 + }, + { + "epoch": 0.84, + "learning_rate": 1.3568018810712702e-06, + "loss": 1.0794, + "step": 34816 + }, + { + "epoch": 0.84, + "learning_rate": 1.356410168114547e-06, + "loss": 0.846, + "step": 34817 + }, + { + "epoch": 0.84, + "learning_rate": 1.356018507596447e-06, + "loss": 0.7885, + "step": 34818 + }, + { + "epoch": 0.84, + "learning_rate": 1.35562689951935e-06, + "loss": 0.991, + "step": 34819 + }, + { + "epoch": 0.84, + "learning_rate": 1.3552353438856324e-06, + "loss": 1.0468, + "step": 34820 + }, + { + "epoch": 0.84, + "learning_rate": 1.3548438406976684e-06, + "loss": 1.089, + "step": 34821 + }, + { + "epoch": 0.84, + "learning_rate": 1.3544523899578343e-06, + "loss": 0.9492, + "step": 34822 + }, + { + "epoch": 0.84, + "learning_rate": 1.354060991668502e-06, + "loss": 1.1261, + "step": 34823 + }, + { + "epoch": 0.84, + "learning_rate": 1.3536696458320476e-06, + "loss": 0.9932, + "step": 34824 + }, + { + "epoch": 0.84, + "learning_rate": 1.3532783524508486e-06, + "loss": 0.7671, + "step": 34825 + }, + { + "epoch": 0.84, + "learning_rate": 1.3528871115272735e-06, + "loss": 1.3311, + "step": 34826 + }, + { + "epoch": 0.84, + "learning_rate": 1.352495923063698e-06, + "loss": 1.1283, + "step": 34827 + }, + { + "epoch": 0.84, + "learning_rate": 1.3521047870624982e-06, + "loss": 1.1847, + "step": 34828 + }, + { + "epoch": 0.84, + "learning_rate": 1.351713703526043e-06, + "loss": 0.9331, + "step": 34829 + }, + { + "epoch": 0.84, + "learning_rate": 1.3513226724567053e-06, + "loss": 1.0661, + "step": 34830 + }, + { + "epoch": 0.84, + "learning_rate": 1.3509316938568616e-06, + "loss": 0.9457, + "step": 34831 + }, + { + "epoch": 0.84, + "learning_rate": 1.3505407677288763e-06, + "loss": 0.9273, + "step": 34832 + }, + { + "epoch": 0.84, + "learning_rate": 1.3501498940751311e-06, + "loss": 1.1631, + "step": 34833 + }, + { + "epoch": 0.84, + "learning_rate": 1.349759072897988e-06, + "loss": 1.0769, + "step": 34834 + }, + { + "epoch": 0.84, + "learning_rate": 1.3493683041998252e-06, + "loss": 1.1275, + "step": 34835 + }, + { + "epoch": 0.84, + "learning_rate": 1.348977587983008e-06, + "loss": 1.1652, + "step": 34836 + }, + { + "epoch": 0.84, + "learning_rate": 1.3485869242499083e-06, + "loss": 1.3136, + "step": 34837 + }, + { + "epoch": 0.84, + "learning_rate": 1.3481963130028997e-06, + "loss": 1.3121, + "step": 34838 + }, + { + "epoch": 0.84, + "learning_rate": 1.3478057542443467e-06, + "loss": 1.2637, + "step": 34839 + }, + { + "epoch": 0.84, + "learning_rate": 1.347415247976621e-06, + "loss": 1.0303, + "step": 34840 + }, + { + "epoch": 0.84, + "learning_rate": 1.3470247942020941e-06, + "loss": 0.6979, + "step": 34841 + }, + { + "epoch": 0.84, + "learning_rate": 1.3466343929231295e-06, + "loss": 1.2651, + "step": 34842 + }, + { + "epoch": 0.84, + "learning_rate": 1.346244044142102e-06, + "loss": 0.9735, + "step": 34843 + }, + { + "epoch": 0.84, + "learning_rate": 1.3458537478613731e-06, + "loss": 1.1007, + "step": 34844 + }, + { + "epoch": 0.84, + "learning_rate": 1.3454635040833142e-06, + "loss": 1.0719, + "step": 34845 + }, + { + "epoch": 0.84, + "learning_rate": 1.3450733128102954e-06, + "loss": 0.9604, + "step": 34846 + }, + { + "epoch": 0.84, + "learning_rate": 1.3446831740446786e-06, + "loss": 1.2766, + "step": 34847 + }, + { + "epoch": 0.84, + "learning_rate": 1.3442930877888327e-06, + "loss": 1.1717, + "step": 34848 + }, + { + "epoch": 0.84, + "learning_rate": 1.343903054045127e-06, + "loss": 1.1411, + "step": 34849 + }, + { + "epoch": 0.84, + "learning_rate": 1.343513072815924e-06, + "loss": 1.037, + "step": 34850 + }, + { + "epoch": 0.84, + "learning_rate": 1.343123144103592e-06, + "loss": 0.7946, + "step": 34851 + }, + { + "epoch": 0.84, + "learning_rate": 1.3427332679104943e-06, + "loss": 0.9547, + "step": 34852 + }, + { + "epoch": 0.84, + "learning_rate": 1.342343444238997e-06, + "loss": 1.1799, + "step": 34853 + }, + { + "epoch": 0.84, + "learning_rate": 1.3419536730914683e-06, + "loss": 1.0329, + "step": 34854 + }, + { + "epoch": 0.84, + "learning_rate": 1.3415639544702686e-06, + "loss": 1.0157, + "step": 34855 + }, + { + "epoch": 0.84, + "learning_rate": 1.3411742883777624e-06, + "loss": 1.0574, + "step": 34856 + }, + { + "epoch": 0.84, + "learning_rate": 1.3407846748163179e-06, + "loss": 1.1203, + "step": 34857 + }, + { + "epoch": 0.84, + "learning_rate": 1.3403951137882942e-06, + "loss": 1.141, + "step": 34858 + }, + { + "epoch": 0.84, + "learning_rate": 1.3400056052960564e-06, + "loss": 1.1199, + "step": 34859 + }, + { + "epoch": 0.84, + "learning_rate": 1.339616149341969e-06, + "loss": 1.0767, + "step": 34860 + }, + { + "epoch": 0.84, + "learning_rate": 1.3392267459283915e-06, + "loss": 1.039, + "step": 34861 + }, + { + "epoch": 0.84, + "learning_rate": 1.3388373950576906e-06, + "loss": 0.9209, + "step": 34862 + }, + { + "epoch": 0.84, + "learning_rate": 1.3384480967322244e-06, + "loss": 1.0603, + "step": 34863 + }, + { + "epoch": 0.84, + "learning_rate": 1.3380588509543558e-06, + "loss": 1.1911, + "step": 34864 + }, + { + "epoch": 0.84, + "learning_rate": 1.3376696577264481e-06, + "loss": 1.0473, + "step": 34865 + }, + { + "epoch": 0.84, + "learning_rate": 1.3372805170508595e-06, + "loss": 1.0809, + "step": 34866 + }, + { + "epoch": 0.84, + "learning_rate": 1.336891428929953e-06, + "loss": 1.0411, + "step": 34867 + }, + { + "epoch": 0.84, + "learning_rate": 1.3365023933660903e-06, + "loss": 1.0895, + "step": 34868 + }, + { + "epoch": 0.84, + "learning_rate": 1.336113410361628e-06, + "loss": 0.9033, + "step": 34869 + }, + { + "epoch": 0.84, + "learning_rate": 1.3357244799189295e-06, + "loss": 0.9056, + "step": 34870 + }, + { + "epoch": 0.84, + "learning_rate": 1.3353356020403496e-06, + "loss": 1.2043, + "step": 34871 + }, + { + "epoch": 0.84, + "learning_rate": 1.334946776728252e-06, + "loss": 0.9316, + "step": 34872 + }, + { + "epoch": 0.84, + "learning_rate": 1.3345580039849959e-06, + "loss": 1.1494, + "step": 34873 + }, + { + "epoch": 0.84, + "learning_rate": 1.334169283812936e-06, + "loss": 0.9673, + "step": 34874 + }, + { + "epoch": 0.84, + "learning_rate": 1.3337806162144317e-06, + "loss": 0.8957, + "step": 34875 + }, + { + "epoch": 0.84, + "learning_rate": 1.3333920011918455e-06, + "loss": 0.8195, + "step": 34876 + }, + { + "epoch": 0.84, + "learning_rate": 1.3330034387475288e-06, + "loss": 1.2101, + "step": 34877 + }, + { + "epoch": 0.84, + "learning_rate": 1.332614928883844e-06, + "loss": 1.0812, + "step": 34878 + }, + { + "epoch": 0.84, + "learning_rate": 1.332226471603143e-06, + "loss": 1.2008, + "step": 34879 + }, + { + "epoch": 0.84, + "learning_rate": 1.3318380669077857e-06, + "loss": 0.9112, + "step": 34880 + }, + { + "epoch": 0.84, + "learning_rate": 1.3314497148001304e-06, + "loss": 1.3352, + "step": 34881 + }, + { + "epoch": 0.84, + "learning_rate": 1.3310614152825274e-06, + "loss": 1.069, + "step": 34882 + }, + { + "epoch": 0.84, + "learning_rate": 1.330673168357337e-06, + "loss": 0.8431, + "step": 34883 + }, + { + "epoch": 0.84, + "learning_rate": 1.3302849740269152e-06, + "loss": 1.0354, + "step": 34884 + }, + { + "epoch": 0.84, + "learning_rate": 1.3298968322936122e-06, + "loss": 0.8594, + "step": 34885 + }, + { + "epoch": 0.84, + "learning_rate": 1.3295087431597898e-06, + "loss": 1.2186, + "step": 34886 + }, + { + "epoch": 0.84, + "learning_rate": 1.3291207066277944e-06, + "loss": 1.2157, + "step": 34887 + }, + { + "epoch": 0.84, + "learning_rate": 1.3287327226999846e-06, + "loss": 1.0025, + "step": 34888 + }, + { + "epoch": 0.84, + "learning_rate": 1.3283447913787163e-06, + "loss": 0.8104, + "step": 34889 + }, + { + "epoch": 0.84, + "learning_rate": 1.3279569126663384e-06, + "loss": 1.1166, + "step": 34890 + }, + { + "epoch": 0.84, + "learning_rate": 1.3275690865652057e-06, + "loss": 1.1161, + "step": 34891 + }, + { + "epoch": 0.84, + "learning_rate": 1.3271813130776745e-06, + "loss": 1.1005, + "step": 34892 + }, + { + "epoch": 0.84, + "learning_rate": 1.3267935922060916e-06, + "loss": 1.0497, + "step": 34893 + }, + { + "epoch": 0.84, + "learning_rate": 1.3264059239528137e-06, + "loss": 1.0541, + "step": 34894 + }, + { + "epoch": 0.84, + "learning_rate": 1.3260183083201895e-06, + "loss": 0.8445, + "step": 34895 + }, + { + "epoch": 0.84, + "learning_rate": 1.325630745310571e-06, + "loss": 1.1519, + "step": 34896 + }, + { + "epoch": 0.84, + "learning_rate": 1.3252432349263133e-06, + "loss": 1.0289, + "step": 34897 + }, + { + "epoch": 0.84, + "learning_rate": 1.3248557771697613e-06, + "loss": 1.0044, + "step": 34898 + }, + { + "epoch": 0.84, + "learning_rate": 1.3244683720432705e-06, + "loss": 0.9835, + "step": 34899 + }, + { + "epoch": 0.84, + "learning_rate": 1.3240810195491904e-06, + "loss": 1.1222, + "step": 34900 + }, + { + "epoch": 0.84, + "learning_rate": 1.323693719689868e-06, + "loss": 1.0926, + "step": 34901 + }, + { + "epoch": 0.84, + "learning_rate": 1.3233064724676548e-06, + "loss": 1.0658, + "step": 34902 + }, + { + "epoch": 0.84, + "learning_rate": 1.3229192778849032e-06, + "loss": 0.9683, + "step": 34903 + }, + { + "epoch": 0.84, + "learning_rate": 1.3225321359439568e-06, + "loss": 0.9131, + "step": 34904 + }, + { + "epoch": 0.84, + "learning_rate": 1.3221450466471697e-06, + "loss": 1.0282, + "step": 34905 + }, + { + "epoch": 0.84, + "learning_rate": 1.3217580099968841e-06, + "loss": 1.3062, + "step": 34906 + }, + { + "epoch": 0.84, + "learning_rate": 1.3213710259954526e-06, + "loss": 0.7341, + "step": 34907 + }, + { + "epoch": 0.84, + "learning_rate": 1.3209840946452236e-06, + "loss": 0.8702, + "step": 34908 + }, + { + "epoch": 0.84, + "learning_rate": 1.3205972159485391e-06, + "loss": 1.1237, + "step": 34909 + }, + { + "epoch": 0.84, + "learning_rate": 1.3202103899077512e-06, + "loss": 1.1103, + "step": 34910 + }, + { + "epoch": 0.84, + "learning_rate": 1.3198236165252076e-06, + "loss": 1.0653, + "step": 34911 + }, + { + "epoch": 0.84, + "learning_rate": 1.31943689580325e-06, + "loss": 0.8976, + "step": 34912 + }, + { + "epoch": 0.84, + "learning_rate": 1.3190502277442285e-06, + "loss": 1.127, + "step": 34913 + }, + { + "epoch": 0.84, + "learning_rate": 1.3186636123504858e-06, + "loss": 1.0685, + "step": 34914 + }, + { + "epoch": 0.84, + "learning_rate": 1.3182770496243692e-06, + "loss": 0.9739, + "step": 34915 + }, + { + "epoch": 0.84, + "learning_rate": 1.317890539568225e-06, + "loss": 1.0091, + "step": 34916 + }, + { + "epoch": 0.84, + "learning_rate": 1.3175040821843953e-06, + "loss": 1.087, + "step": 34917 + }, + { + "epoch": 0.84, + "learning_rate": 1.3171176774752258e-06, + "loss": 1.0358, + "step": 34918 + }, + { + "epoch": 0.84, + "learning_rate": 1.3167313254430624e-06, + "loss": 0.9216, + "step": 34919 + }, + { + "epoch": 0.84, + "learning_rate": 1.3163450260902466e-06, + "loss": 1.1262, + "step": 34920 + }, + { + "epoch": 0.84, + "learning_rate": 1.3159587794191253e-06, + "loss": 1.0599, + "step": 34921 + }, + { + "epoch": 0.84, + "learning_rate": 1.3155725854320367e-06, + "loss": 0.9442, + "step": 34922 + }, + { + "epoch": 0.84, + "learning_rate": 1.3151864441313266e-06, + "loss": 0.9878, + "step": 34923 + }, + { + "epoch": 0.84, + "learning_rate": 1.31480035551934e-06, + "loss": 0.7249, + "step": 34924 + }, + { + "epoch": 0.84, + "learning_rate": 1.3144143195984138e-06, + "loss": 1.0382, + "step": 34925 + }, + { + "epoch": 0.84, + "learning_rate": 1.3140283363708938e-06, + "loss": 1.094, + "step": 34926 + }, + { + "epoch": 0.84, + "learning_rate": 1.313642405839123e-06, + "loss": 0.8432, + "step": 34927 + }, + { + "epoch": 0.84, + "learning_rate": 1.3132565280054376e-06, + "loss": 0.9676, + "step": 34928 + }, + { + "epoch": 0.84, + "learning_rate": 1.3128707028721843e-06, + "loss": 0.848, + "step": 34929 + }, + { + "epoch": 0.84, + "learning_rate": 1.3124849304416986e-06, + "loss": 0.9556, + "step": 34930 + }, + { + "epoch": 0.84, + "learning_rate": 1.3120992107163243e-06, + "loss": 0.7761, + "step": 34931 + }, + { + "epoch": 0.84, + "learning_rate": 1.311713543698402e-06, + "loss": 1.089, + "step": 34932 + }, + { + "epoch": 0.84, + "learning_rate": 1.3113279293902681e-06, + "loss": 0.838, + "step": 34933 + }, + { + "epoch": 0.84, + "learning_rate": 1.3109423677942635e-06, + "loss": 1.074, + "step": 34934 + }, + { + "epoch": 0.84, + "learning_rate": 1.3105568589127305e-06, + "loss": 0.9711, + "step": 34935 + }, + { + "epoch": 0.84, + "learning_rate": 1.3101714027480027e-06, + "loss": 1.1341, + "step": 34936 + }, + { + "epoch": 0.84, + "learning_rate": 1.309785999302423e-06, + "loss": 1.0309, + "step": 34937 + }, + { + "epoch": 0.84, + "learning_rate": 1.309400648578326e-06, + "loss": 1.0629, + "step": 34938 + }, + { + "epoch": 0.84, + "learning_rate": 1.3090153505780511e-06, + "loss": 1.1829, + "step": 34939 + }, + { + "epoch": 0.84, + "learning_rate": 1.3086301053039373e-06, + "loss": 1.2375, + "step": 34940 + }, + { + "epoch": 0.84, + "learning_rate": 1.3082449127583186e-06, + "loss": 0.9857, + "step": 34941 + }, + { + "epoch": 0.84, + "learning_rate": 1.307859772943534e-06, + "loss": 0.9266, + "step": 34942 + }, + { + "epoch": 0.84, + "learning_rate": 1.3074746858619214e-06, + "loss": 1.3124, + "step": 34943 + }, + { + "epoch": 0.84, + "learning_rate": 1.3070896515158128e-06, + "loss": 1.4041, + "step": 34944 + }, + { + "epoch": 0.84, + "learning_rate": 1.3067046699075492e-06, + "loss": 1.0381, + "step": 34945 + }, + { + "epoch": 0.84, + "learning_rate": 1.3063197410394612e-06, + "loss": 0.947, + "step": 34946 + }, + { + "epoch": 0.84, + "learning_rate": 1.305934864913886e-06, + "loss": 1.0106, + "step": 34947 + }, + { + "epoch": 0.84, + "learning_rate": 1.3055500415331613e-06, + "loss": 1.0219, + "step": 34948 + }, + { + "epoch": 0.84, + "learning_rate": 1.3051652708996176e-06, + "loss": 1.0404, + "step": 34949 + }, + { + "epoch": 0.84, + "learning_rate": 1.304780553015591e-06, + "loss": 1.1439, + "step": 34950 + }, + { + "epoch": 0.84, + "learning_rate": 1.3043958878834185e-06, + "loss": 0.9551, + "step": 34951 + }, + { + "epoch": 0.84, + "learning_rate": 1.304011275505428e-06, + "loss": 0.9214, + "step": 34952 + }, + { + "epoch": 0.84, + "learning_rate": 1.3036267158839555e-06, + "loss": 1.0787, + "step": 34953 + }, + { + "epoch": 0.84, + "learning_rate": 1.303242209021336e-06, + "loss": 0.9645, + "step": 34954 + }, + { + "epoch": 0.84, + "learning_rate": 1.3028577549198984e-06, + "loss": 1.0664, + "step": 34955 + }, + { + "epoch": 0.84, + "learning_rate": 1.30247335358198e-06, + "loss": 1.0527, + "step": 34956 + }, + { + "epoch": 0.84, + "learning_rate": 1.3020890050099078e-06, + "loss": 1.0034, + "step": 34957 + }, + { + "epoch": 0.84, + "learning_rate": 1.3017047092060143e-06, + "loss": 1.2236, + "step": 34958 + }, + { + "epoch": 0.84, + "learning_rate": 1.3013204661726364e-06, + "loss": 1.1566, + "step": 34959 + }, + { + "epoch": 0.84, + "learning_rate": 1.300936275912098e-06, + "loss": 1.0406, + "step": 34960 + }, + { + "epoch": 0.84, + "learning_rate": 1.3005521384267328e-06, + "loss": 0.93, + "step": 34961 + }, + { + "epoch": 0.84, + "learning_rate": 1.3001680537188744e-06, + "loss": 1.0401, + "step": 34962 + }, + { + "epoch": 0.84, + "learning_rate": 1.2997840217908475e-06, + "loss": 0.9737, + "step": 34963 + }, + { + "epoch": 0.84, + "learning_rate": 1.2994000426449872e-06, + "loss": 0.9882, + "step": 34964 + }, + { + "epoch": 0.84, + "learning_rate": 1.299016116283618e-06, + "loss": 1.2441, + "step": 34965 + }, + { + "epoch": 0.84, + "learning_rate": 1.2986322427090703e-06, + "loss": 0.9946, + "step": 34966 + }, + { + "epoch": 0.84, + "learning_rate": 1.298248421923678e-06, + "loss": 0.9655, + "step": 34967 + }, + { + "epoch": 0.84, + "learning_rate": 1.2978646539297623e-06, + "loss": 0.6402, + "step": 34968 + }, + { + "epoch": 0.84, + "learning_rate": 1.2974809387296549e-06, + "loss": 0.9685, + "step": 34969 + }, + { + "epoch": 0.84, + "learning_rate": 1.2970972763256872e-06, + "loss": 1.095, + "step": 34970 + }, + { + "epoch": 0.84, + "learning_rate": 1.2967136667201796e-06, + "loss": 1.0721, + "step": 34971 + }, + { + "epoch": 0.84, + "learning_rate": 1.2963301099154656e-06, + "loss": 1.0963, + "step": 34972 + }, + { + "epoch": 0.84, + "learning_rate": 1.2959466059138682e-06, + "loss": 0.9823, + "step": 34973 + }, + { + "epoch": 0.84, + "learning_rate": 1.295563154717715e-06, + "loss": 1.128, + "step": 34974 + }, + { + "epoch": 0.84, + "learning_rate": 1.2951797563293344e-06, + "loss": 0.8821, + "step": 34975 + }, + { + "epoch": 0.84, + "learning_rate": 1.294796410751049e-06, + "loss": 0.8544, + "step": 34976 + }, + { + "epoch": 0.84, + "learning_rate": 1.294413117985186e-06, + "loss": 1.2272, + "step": 34977 + }, + { + "epoch": 0.84, + "learning_rate": 1.2940298780340732e-06, + "loss": 0.9775, + "step": 34978 + }, + { + "epoch": 0.84, + "learning_rate": 1.2936466909000322e-06, + "loss": 0.9641, + "step": 34979 + }, + { + "epoch": 0.84, + "learning_rate": 1.293263556585389e-06, + "loss": 1.0227, + "step": 34980 + }, + { + "epoch": 0.84, + "learning_rate": 1.2928804750924674e-06, + "loss": 0.7021, + "step": 34981 + }, + { + "epoch": 0.84, + "learning_rate": 1.2924974464235905e-06, + "loss": 1.0272, + "step": 34982 + }, + { + "epoch": 0.84, + "learning_rate": 1.2921144705810873e-06, + "loss": 0.7273, + "step": 34983 + }, + { + "epoch": 0.84, + "learning_rate": 1.2917315475672743e-06, + "loss": 1.0013, + "step": 34984 + }, + { + "epoch": 0.84, + "learning_rate": 1.2913486773844775e-06, + "loss": 1.1648, + "step": 34985 + }, + { + "epoch": 0.84, + "learning_rate": 1.2909658600350227e-06, + "loss": 0.858, + "step": 34986 + }, + { + "epoch": 0.84, + "learning_rate": 1.2905830955212261e-06, + "loss": 0.9357, + "step": 34987 + }, + { + "epoch": 0.84, + "learning_rate": 1.290200383845417e-06, + "loss": 1.162, + "step": 34988 + }, + { + "epoch": 0.84, + "learning_rate": 1.2898177250099098e-06, + "loss": 0.7348, + "step": 34989 + }, + { + "epoch": 0.84, + "learning_rate": 1.2894351190170306e-06, + "loss": 1.1607, + "step": 34990 + }, + { + "epoch": 0.84, + "learning_rate": 1.2890525658691022e-06, + "loss": 1.0541, + "step": 34991 + }, + { + "epoch": 0.84, + "learning_rate": 1.2886700655684404e-06, + "loss": 1.1075, + "step": 34992 + }, + { + "epoch": 0.84, + "learning_rate": 1.2882876181173676e-06, + "loss": 1.2102, + "step": 34993 + }, + { + "epoch": 0.84, + "learning_rate": 1.2879052235182077e-06, + "loss": 1.0164, + "step": 34994 + }, + { + "epoch": 0.84, + "learning_rate": 1.2875228817732755e-06, + "loss": 0.8752, + "step": 34995 + }, + { + "epoch": 0.84, + "learning_rate": 1.2871405928848924e-06, + "loss": 1.1063, + "step": 34996 + }, + { + "epoch": 0.84, + "learning_rate": 1.2867583568553799e-06, + "loss": 1.1109, + "step": 34997 + }, + { + "epoch": 0.84, + "learning_rate": 1.2863761736870527e-06, + "loss": 0.9756, + "step": 34998 + }, + { + "epoch": 0.84, + "learning_rate": 1.2859940433822327e-06, + "loss": 0.9578, + "step": 34999 + }, + { + "epoch": 0.84, + "learning_rate": 1.2856119659432365e-06, + "loss": 1.227, + "step": 35000 + }, + { + "epoch": 0.84, + "learning_rate": 1.2852299413723846e-06, + "loss": 0.7548, + "step": 35001 + }, + { + "epoch": 0.84, + "learning_rate": 1.2848479696719906e-06, + "loss": 0.9407, + "step": 35002 + }, + { + "epoch": 0.84, + "learning_rate": 1.284466050844374e-06, + "loss": 1.0137, + "step": 35003 + }, + { + "epoch": 0.84, + "learning_rate": 1.284084184891855e-06, + "loss": 1.1732, + "step": 35004 + }, + { + "epoch": 0.84, + "learning_rate": 1.283702371816744e-06, + "loss": 0.942, + "step": 35005 + }, + { + "epoch": 0.84, + "learning_rate": 1.2833206116213615e-06, + "loss": 1.1484, + "step": 35006 + }, + { + "epoch": 0.84, + "learning_rate": 1.2829389043080232e-06, + "loss": 0.9926, + "step": 35007 + }, + { + "epoch": 0.84, + "learning_rate": 1.2825572498790438e-06, + "loss": 0.8932, + "step": 35008 + }, + { + "epoch": 0.84, + "learning_rate": 1.2821756483367398e-06, + "loss": 1.0488, + "step": 35009 + }, + { + "epoch": 0.84, + "learning_rate": 1.2817940996834243e-06, + "loss": 0.9562, + "step": 35010 + }, + { + "epoch": 0.84, + "learning_rate": 1.2814126039214126e-06, + "loss": 0.967, + "step": 35011 + }, + { + "epoch": 0.84, + "learning_rate": 1.2810311610530223e-06, + "loss": 0.9668, + "step": 35012 + }, + { + "epoch": 0.84, + "learning_rate": 1.2806497710805633e-06, + "loss": 0.9923, + "step": 35013 + }, + { + "epoch": 0.84, + "learning_rate": 1.280268434006351e-06, + "loss": 1.0907, + "step": 35014 + }, + { + "epoch": 0.84, + "learning_rate": 1.2798871498327015e-06, + "loss": 0.9784, + "step": 35015 + }, + { + "epoch": 0.84, + "learning_rate": 1.279505918561923e-06, + "loss": 0.9265, + "step": 35016 + }, + { + "epoch": 0.84, + "learning_rate": 1.2791247401963337e-06, + "loss": 1.14, + "step": 35017 + }, + { + "epoch": 0.84, + "learning_rate": 1.2787436147382404e-06, + "loss": 1.2395, + "step": 35018 + }, + { + "epoch": 0.84, + "learning_rate": 1.2783625421899592e-06, + "loss": 0.9978, + "step": 35019 + }, + { + "epoch": 0.84, + "learning_rate": 1.2779815225538028e-06, + "loss": 1.069, + "step": 35020 + }, + { + "epoch": 0.84, + "learning_rate": 1.277600555832078e-06, + "loss": 0.7914, + "step": 35021 + }, + { + "epoch": 0.84, + "learning_rate": 1.2772196420270999e-06, + "loss": 1.1496, + "step": 35022 + }, + { + "epoch": 0.84, + "learning_rate": 1.2768387811411798e-06, + "loss": 1.1027, + "step": 35023 + }, + { + "epoch": 0.84, + "learning_rate": 1.2764579731766257e-06, + "loss": 1.0993, + "step": 35024 + }, + { + "epoch": 0.84, + "learning_rate": 1.2760772181357507e-06, + "loss": 1.0923, + "step": 35025 + }, + { + "epoch": 0.84, + "learning_rate": 1.2756965160208624e-06, + "loss": 1.1972, + "step": 35026 + }, + { + "epoch": 0.84, + "learning_rate": 1.275315866834269e-06, + "loss": 0.9987, + "step": 35027 + }, + { + "epoch": 0.84, + "learning_rate": 1.2749352705782858e-06, + "loss": 0.8251, + "step": 35028 + }, + { + "epoch": 0.84, + "learning_rate": 1.2745547272552161e-06, + "loss": 1.1063, + "step": 35029 + }, + { + "epoch": 0.84, + "learning_rate": 1.2741742368673693e-06, + "loss": 0.8755, + "step": 35030 + }, + { + "epoch": 0.84, + "learning_rate": 1.2737937994170579e-06, + "loss": 1.1163, + "step": 35031 + }, + { + "epoch": 0.84, + "learning_rate": 1.2734134149065846e-06, + "loss": 0.883, + "step": 35032 + }, + { + "epoch": 0.84, + "learning_rate": 1.2730330833382598e-06, + "loss": 1.0919, + "step": 35033 + }, + { + "epoch": 0.84, + "learning_rate": 1.2726528047143927e-06, + "loss": 0.835, + "step": 35034 + }, + { + "epoch": 0.84, + "learning_rate": 1.2722725790372858e-06, + "loss": 1.1367, + "step": 35035 + }, + { + "epoch": 0.84, + "learning_rate": 1.271892406309251e-06, + "loss": 1.212, + "step": 35036 + }, + { + "epoch": 0.84, + "learning_rate": 1.271512286532589e-06, + "loss": 0.8782, + "step": 35037 + }, + { + "epoch": 0.84, + "learning_rate": 1.2711322197096098e-06, + "loss": 1.0568, + "step": 35038 + }, + { + "epoch": 0.84, + "learning_rate": 1.2707522058426192e-06, + "loss": 1.054, + "step": 35039 + }, + { + "epoch": 0.84, + "learning_rate": 1.2703722449339206e-06, + "loss": 1.0182, + "step": 35040 + }, + { + "epoch": 0.84, + "learning_rate": 1.2699923369858202e-06, + "loss": 1.1774, + "step": 35041 + }, + { + "epoch": 0.84, + "learning_rate": 1.269612482000625e-06, + "loss": 0.8865, + "step": 35042 + }, + { + "epoch": 0.84, + "learning_rate": 1.2692326799806343e-06, + "loss": 1.0593, + "step": 35043 + }, + { + "epoch": 0.84, + "learning_rate": 1.2688529309281583e-06, + "loss": 1.0985, + "step": 35044 + }, + { + "epoch": 0.84, + "learning_rate": 1.2684732348454953e-06, + "loss": 1.2541, + "step": 35045 + }, + { + "epoch": 0.84, + "learning_rate": 1.2680935917349524e-06, + "loss": 0.9943, + "step": 35046 + }, + { + "epoch": 0.84, + "learning_rate": 1.2677140015988333e-06, + "loss": 0.9938, + "step": 35047 + }, + { + "epoch": 0.84, + "learning_rate": 1.2673344644394369e-06, + "loss": 0.8433, + "step": 35048 + }, + { + "epoch": 0.84, + "learning_rate": 1.2669549802590687e-06, + "loss": 0.9975, + "step": 35049 + }, + { + "epoch": 0.84, + "learning_rate": 1.2665755490600318e-06, + "loss": 0.9196, + "step": 35050 + }, + { + "epoch": 0.84, + "learning_rate": 1.2661961708446258e-06, + "loss": 0.5706, + "step": 35051 + }, + { + "epoch": 0.84, + "learning_rate": 1.2658168456151543e-06, + "loss": 0.8209, + "step": 35052 + }, + { + "epoch": 0.84, + "learning_rate": 1.2654375733739166e-06, + "loss": 1.0332, + "step": 35053 + }, + { + "epoch": 0.84, + "learning_rate": 1.2650583541232142e-06, + "loss": 1.1789, + "step": 35054 + }, + { + "epoch": 0.84, + "learning_rate": 1.2646791878653508e-06, + "loss": 1.0503, + "step": 35055 + }, + { + "epoch": 0.84, + "learning_rate": 1.2643000746026213e-06, + "loss": 1.1219, + "step": 35056 + }, + { + "epoch": 0.84, + "learning_rate": 1.2639210143373282e-06, + "loss": 1.1445, + "step": 35057 + }, + { + "epoch": 0.84, + "learning_rate": 1.263542007071774e-06, + "loss": 1.0448, + "step": 35058 + }, + { + "epoch": 0.84, + "learning_rate": 1.263163052808254e-06, + "loss": 1.0266, + "step": 35059 + }, + { + "epoch": 0.84, + "learning_rate": 1.2627841515490703e-06, + "loss": 1.278, + "step": 35060 + }, + { + "epoch": 0.84, + "learning_rate": 1.2624053032965177e-06, + "loss": 0.9491, + "step": 35061 + }, + { + "epoch": 0.84, + "learning_rate": 1.262026508052897e-06, + "loss": 1.0045, + "step": 35062 + }, + { + "epoch": 0.84, + "learning_rate": 1.2616477658205084e-06, + "loss": 1.0801, + "step": 35063 + }, + { + "epoch": 0.84, + "learning_rate": 1.2612690766016455e-06, + "loss": 0.9192, + "step": 35064 + }, + { + "epoch": 0.84, + "learning_rate": 1.2608904403986077e-06, + "loss": 0.997, + "step": 35065 + }, + { + "epoch": 0.84, + "learning_rate": 1.2605118572136932e-06, + "loss": 1.0713, + "step": 35066 + }, + { + "epoch": 0.84, + "learning_rate": 1.2601333270491966e-06, + "loss": 1.065, + "step": 35067 + }, + { + "epoch": 0.84, + "learning_rate": 1.2597548499074175e-06, + "loss": 1.0219, + "step": 35068 + }, + { + "epoch": 0.84, + "learning_rate": 1.2593764257906471e-06, + "loss": 1.1523, + "step": 35069 + }, + { + "epoch": 0.84, + "learning_rate": 1.2589980547011837e-06, + "loss": 1.0161, + "step": 35070 + }, + { + "epoch": 0.84, + "learning_rate": 1.2586197366413256e-06, + "loss": 1.0375, + "step": 35071 + }, + { + "epoch": 0.84, + "learning_rate": 1.258241471613364e-06, + "loss": 0.9982, + "step": 35072 + }, + { + "epoch": 0.84, + "learning_rate": 1.257863259619595e-06, + "loss": 1.1002, + "step": 35073 + }, + { + "epoch": 0.84, + "learning_rate": 1.2574851006623145e-06, + "loss": 0.9663, + "step": 35074 + }, + { + "epoch": 0.84, + "learning_rate": 1.2571069947438141e-06, + "loss": 1.0585, + "step": 35075 + }, + { + "epoch": 0.84, + "learning_rate": 1.2567289418663898e-06, + "loss": 1.1151, + "step": 35076 + }, + { + "epoch": 0.84, + "learning_rate": 1.256350942032335e-06, + "loss": 1.0451, + "step": 35077 + }, + { + "epoch": 0.84, + "learning_rate": 1.2559729952439414e-06, + "loss": 1.1638, + "step": 35078 + }, + { + "epoch": 0.84, + "learning_rate": 1.255595101503505e-06, + "loss": 0.7816, + "step": 35079 + }, + { + "epoch": 0.84, + "learning_rate": 1.255217260813314e-06, + "loss": 0.9216, + "step": 35080 + }, + { + "epoch": 0.84, + "learning_rate": 1.254839473175663e-06, + "loss": 1.1045, + "step": 35081 + }, + { + "epoch": 0.84, + "learning_rate": 1.254461738592846e-06, + "loss": 1.0696, + "step": 35082 + }, + { + "epoch": 0.84, + "learning_rate": 1.2540840570671497e-06, + "loss": 1.0518, + "step": 35083 + }, + { + "epoch": 0.84, + "learning_rate": 1.2537064286008693e-06, + "loss": 0.9844, + "step": 35084 + }, + { + "epoch": 0.84, + "learning_rate": 1.2533288531962972e-06, + "loss": 1.0089, + "step": 35085 + }, + { + "epoch": 0.84, + "learning_rate": 1.2529513308557184e-06, + "loss": 0.7479, + "step": 35086 + }, + { + "epoch": 0.84, + "learning_rate": 1.2525738615814286e-06, + "loss": 1.1344, + "step": 35087 + }, + { + "epoch": 0.84, + "learning_rate": 1.2521964453757141e-06, + "loss": 1.084, + "step": 35088 + }, + { + "epoch": 0.84, + "learning_rate": 1.251819082240866e-06, + "loss": 0.9768, + "step": 35089 + }, + { + "epoch": 0.84, + "learning_rate": 1.251441772179176e-06, + "loss": 1.1811, + "step": 35090 + }, + { + "epoch": 0.84, + "learning_rate": 1.251064515192929e-06, + "loss": 0.9895, + "step": 35091 + }, + { + "epoch": 0.84, + "learning_rate": 1.2506873112844154e-06, + "loss": 1.0935, + "step": 35092 + }, + { + "epoch": 0.84, + "learning_rate": 1.2503101604559264e-06, + "loss": 1.2239, + "step": 35093 + }, + { + "epoch": 0.84, + "learning_rate": 1.249933062709745e-06, + "loss": 1.165, + "step": 35094 + }, + { + "epoch": 0.84, + "learning_rate": 1.2495560180481648e-06, + "loss": 0.9939, + "step": 35095 + }, + { + "epoch": 0.84, + "learning_rate": 1.249179026473466e-06, + "loss": 1.2229, + "step": 35096 + }, + { + "epoch": 0.84, + "learning_rate": 1.2488020879879414e-06, + "loss": 1.2356, + "step": 35097 + }, + { + "epoch": 0.84, + "learning_rate": 1.248425202593877e-06, + "loss": 1.2459, + "step": 35098 + }, + { + "epoch": 0.84, + "learning_rate": 1.2480483702935576e-06, + "loss": 0.959, + "step": 35099 + }, + { + "epoch": 0.84, + "learning_rate": 1.247671591089269e-06, + "loss": 1.3435, + "step": 35100 + }, + { + "epoch": 0.84, + "learning_rate": 1.2472948649833005e-06, + "loss": 1.2453, + "step": 35101 + }, + { + "epoch": 0.84, + "learning_rate": 1.246918191977934e-06, + "loss": 0.9698, + "step": 35102 + }, + { + "epoch": 0.84, + "learning_rate": 1.246541572075457e-06, + "loss": 1.1741, + "step": 35103 + }, + { + "epoch": 0.84, + "learning_rate": 1.2461650052781515e-06, + "loss": 1.1474, + "step": 35104 + }, + { + "epoch": 0.84, + "learning_rate": 1.2457884915883045e-06, + "loss": 1.1739, + "step": 35105 + }, + { + "epoch": 0.84, + "learning_rate": 1.245412031008202e-06, + "loss": 1.3881, + "step": 35106 + }, + { + "epoch": 0.84, + "learning_rate": 1.2450356235401217e-06, + "loss": 0.8119, + "step": 35107 + }, + { + "epoch": 0.84, + "learning_rate": 1.2446592691863524e-06, + "loss": 1.0658, + "step": 35108 + }, + { + "epoch": 0.84, + "learning_rate": 1.2442829679491775e-06, + "loss": 0.7349, + "step": 35109 + }, + { + "epoch": 0.84, + "learning_rate": 1.2439067198308762e-06, + "loss": 1.1122, + "step": 35110 + }, + { + "epoch": 0.84, + "learning_rate": 1.243530524833736e-06, + "loss": 1.0028, + "step": 35111 + }, + { + "epoch": 0.84, + "learning_rate": 1.2431543829600334e-06, + "loss": 0.9864, + "step": 35112 + }, + { + "epoch": 0.84, + "learning_rate": 1.2427782942120547e-06, + "loss": 0.9521, + "step": 35113 + }, + { + "epoch": 0.84, + "learning_rate": 1.2424022585920814e-06, + "loss": 1.0715, + "step": 35114 + }, + { + "epoch": 0.84, + "learning_rate": 1.2420262761023916e-06, + "loss": 0.8424, + "step": 35115 + }, + { + "epoch": 0.84, + "learning_rate": 1.241650346745269e-06, + "loss": 1.0921, + "step": 35116 + }, + { + "epoch": 0.84, + "learning_rate": 1.241274470522995e-06, + "loss": 1.1714, + "step": 35117 + }, + { + "epoch": 0.84, + "learning_rate": 1.2408986474378471e-06, + "loss": 1.1908, + "step": 35118 + }, + { + "epoch": 0.84, + "learning_rate": 1.2405228774921085e-06, + "loss": 1.2262, + "step": 35119 + }, + { + "epoch": 0.84, + "learning_rate": 1.2401471606880554e-06, + "loss": 0.6844, + "step": 35120 + }, + { + "epoch": 0.84, + "learning_rate": 1.2397714970279684e-06, + "loss": 0.9607, + "step": 35121 + }, + { + "epoch": 0.84, + "learning_rate": 1.2393958865141308e-06, + "loss": 0.9455, + "step": 35122 + }, + { + "epoch": 0.84, + "learning_rate": 1.2390203291488144e-06, + "loss": 1.1411, + "step": 35123 + }, + { + "epoch": 0.84, + "learning_rate": 1.238644824934301e-06, + "loss": 1.0687, + "step": 35124 + }, + { + "epoch": 0.84, + "learning_rate": 1.2382693738728713e-06, + "loss": 1.0979, + "step": 35125 + }, + { + "epoch": 0.84, + "learning_rate": 1.2378939759667974e-06, + "loss": 1.3683, + "step": 35126 + }, + { + "epoch": 0.84, + "learning_rate": 1.2375186312183606e-06, + "loss": 0.9545, + "step": 35127 + }, + { + "epoch": 0.84, + "learning_rate": 1.2371433396298393e-06, + "loss": 1.1705, + "step": 35128 + }, + { + "epoch": 0.84, + "learning_rate": 1.236768101203506e-06, + "loss": 1.158, + "step": 35129 + }, + { + "epoch": 0.84, + "learning_rate": 1.236392915941641e-06, + "loss": 1.2533, + "step": 35130 + }, + { + "epoch": 0.84, + "learning_rate": 1.236017783846518e-06, + "loss": 1.0599, + "step": 35131 + }, + { + "epoch": 0.84, + "learning_rate": 1.2356427049204122e-06, + "loss": 1.033, + "step": 35132 + }, + { + "epoch": 0.84, + "learning_rate": 1.2352676791656039e-06, + "loss": 1.1458, + "step": 35133 + }, + { + "epoch": 0.84, + "learning_rate": 1.234892706584362e-06, + "loss": 1.1973, + "step": 35134 + }, + { + "epoch": 0.84, + "learning_rate": 1.234517787178965e-06, + "loss": 0.7955, + "step": 35135 + }, + { + "epoch": 0.84, + "learning_rate": 1.2341429209516885e-06, + "loss": 0.9956, + "step": 35136 + }, + { + "epoch": 0.85, + "learning_rate": 1.2337681079048035e-06, + "loss": 1.0238, + "step": 35137 + }, + { + "epoch": 0.85, + "learning_rate": 1.2333933480405868e-06, + "loss": 0.9094, + "step": 35138 + }, + { + "epoch": 0.85, + "learning_rate": 1.2330186413613087e-06, + "loss": 0.8728, + "step": 35139 + }, + { + "epoch": 0.85, + "learning_rate": 1.2326439878692441e-06, + "loss": 1.0579, + "step": 35140 + }, + { + "epoch": 0.85, + "learning_rate": 1.2322693875666691e-06, + "loss": 0.8633, + "step": 35141 + }, + { + "epoch": 0.85, + "learning_rate": 1.2318948404558517e-06, + "loss": 0.8536, + "step": 35142 + }, + { + "epoch": 0.85, + "learning_rate": 1.2315203465390646e-06, + "loss": 0.7533, + "step": 35143 + }, + { + "epoch": 0.85, + "learning_rate": 1.2311459058185838e-06, + "loss": 0.9875, + "step": 35144 + }, + { + "epoch": 0.85, + "learning_rate": 1.2307715182966762e-06, + "loss": 0.8465, + "step": 35145 + }, + { + "epoch": 0.85, + "learning_rate": 1.2303971839756167e-06, + "loss": 0.5379, + "step": 35146 + }, + { + "epoch": 0.85, + "learning_rate": 1.2300229028576737e-06, + "loss": 1.1066, + "step": 35147 + }, + { + "epoch": 0.85, + "learning_rate": 1.2296486749451187e-06, + "loss": 0.8415, + "step": 35148 + }, + { + "epoch": 0.85, + "learning_rate": 1.229274500240224e-06, + "loss": 1.0133, + "step": 35149 + }, + { + "epoch": 0.85, + "learning_rate": 1.228900378745256e-06, + "loss": 0.6386, + "step": 35150 + }, + { + "epoch": 0.85, + "learning_rate": 1.2285263104624856e-06, + "loss": 0.979, + "step": 35151 + }, + { + "epoch": 0.85, + "learning_rate": 1.228152295394186e-06, + "loss": 1.0864, + "step": 35152 + }, + { + "epoch": 0.85, + "learning_rate": 1.2277783335426197e-06, + "loss": 1.0941, + "step": 35153 + }, + { + "epoch": 0.85, + "learning_rate": 1.2274044249100624e-06, + "loss": 0.9725, + "step": 35154 + }, + { + "epoch": 0.85, + "learning_rate": 1.227030569498776e-06, + "loss": 1.1653, + "step": 35155 + }, + { + "epoch": 0.85, + "learning_rate": 1.2266567673110318e-06, + "loss": 0.9069, + "step": 35156 + }, + { + "epoch": 0.85, + "learning_rate": 1.2262830183491004e-06, + "loss": 1.1023, + "step": 35157 + }, + { + "epoch": 0.85, + "learning_rate": 1.225909322615243e-06, + "loss": 1.0376, + "step": 35158 + }, + { + "epoch": 0.85, + "learning_rate": 1.2255356801117302e-06, + "loss": 1.0511, + "step": 35159 + }, + { + "epoch": 0.85, + "learning_rate": 1.2251620908408312e-06, + "loss": 1.0042, + "step": 35160 + }, + { + "epoch": 0.85, + "learning_rate": 1.2247885548048078e-06, + "loss": 0.9077, + "step": 35161 + }, + { + "epoch": 0.85, + "learning_rate": 1.224415072005929e-06, + "loss": 0.9009, + "step": 35162 + }, + { + "epoch": 0.85, + "learning_rate": 1.2240416424464585e-06, + "loss": 0.9566, + "step": 35163 + }, + { + "epoch": 0.85, + "learning_rate": 1.2236682661286636e-06, + "loss": 1.0839, + "step": 35164 + }, + { + "epoch": 0.85, + "learning_rate": 1.2232949430548102e-06, + "loss": 1.0074, + "step": 35165 + }, + { + "epoch": 0.85, + "learning_rate": 1.2229216732271587e-06, + "loss": 1.0814, + "step": 35166 + }, + { + "epoch": 0.85, + "learning_rate": 1.2225484566479806e-06, + "loss": 1.2076, + "step": 35167 + }, + { + "epoch": 0.85, + "learning_rate": 1.222175293319534e-06, + "loss": 1.2028, + "step": 35168 + }, + { + "epoch": 0.85, + "learning_rate": 1.2218021832440864e-06, + "loss": 1.0932, + "step": 35169 + }, + { + "epoch": 0.85, + "learning_rate": 1.2214291264239009e-06, + "loss": 1.0316, + "step": 35170 + }, + { + "epoch": 0.85, + "learning_rate": 1.2210561228612395e-06, + "loss": 0.9817, + "step": 35171 + }, + { + "epoch": 0.85, + "learning_rate": 1.2206831725583645e-06, + "loss": 1.1237, + "step": 35172 + }, + { + "epoch": 0.85, + "learning_rate": 1.220310275517541e-06, + "loss": 1.0195, + "step": 35173 + }, + { + "epoch": 0.85, + "learning_rate": 1.2199374317410294e-06, + "loss": 0.9655, + "step": 35174 + }, + { + "epoch": 0.85, + "learning_rate": 1.2195646412310936e-06, + "loss": 1.0108, + "step": 35175 + }, + { + "epoch": 0.85, + "learning_rate": 1.2191919039899925e-06, + "loss": 0.9156, + "step": 35176 + }, + { + "epoch": 0.85, + "learning_rate": 1.2188192200199878e-06, + "loss": 1.0717, + "step": 35177 + }, + { + "epoch": 0.85, + "learning_rate": 1.2184465893233432e-06, + "loss": 1.0314, + "step": 35178 + }, + { + "epoch": 0.85, + "learning_rate": 1.2180740119023172e-06, + "loss": 1.0874, + "step": 35179 + }, + { + "epoch": 0.85, + "learning_rate": 1.2177014877591687e-06, + "loss": 1.2045, + "step": 35180 + }, + { + "epoch": 0.85, + "learning_rate": 1.2173290168961628e-06, + "loss": 1.3691, + "step": 35181 + }, + { + "epoch": 0.85, + "learning_rate": 1.216956599315553e-06, + "loss": 0.7931, + "step": 35182 + }, + { + "epoch": 0.85, + "learning_rate": 1.2165842350196044e-06, + "loss": 0.9142, + "step": 35183 + }, + { + "epoch": 0.85, + "learning_rate": 1.2162119240105707e-06, + "loss": 0.8732, + "step": 35184 + }, + { + "epoch": 0.85, + "learning_rate": 1.2158396662907135e-06, + "loss": 1.0424, + "step": 35185 + }, + { + "epoch": 0.85, + "learning_rate": 1.215467461862293e-06, + "loss": 1.0098, + "step": 35186 + }, + { + "epoch": 0.85, + "learning_rate": 1.2150953107275631e-06, + "loss": 1.0889, + "step": 35187 + }, + { + "epoch": 0.85, + "learning_rate": 1.2147232128887842e-06, + "loss": 0.9596, + "step": 35188 + }, + { + "epoch": 0.85, + "learning_rate": 1.2143511683482145e-06, + "loss": 0.9033, + "step": 35189 + }, + { + "epoch": 0.85, + "learning_rate": 1.2139791771081078e-06, + "loss": 0.9831, + "step": 35190 + }, + { + "epoch": 0.85, + "learning_rate": 1.2136072391707244e-06, + "loss": 0.9532, + "step": 35191 + }, + { + "epoch": 0.85, + "learning_rate": 1.213235354538318e-06, + "loss": 1.1328, + "step": 35192 + }, + { + "epoch": 0.85, + "learning_rate": 1.2128635232131446e-06, + "loss": 1.1676, + "step": 35193 + }, + { + "epoch": 0.85, + "learning_rate": 1.212491745197465e-06, + "loss": 0.9655, + "step": 35194 + }, + { + "epoch": 0.85, + "learning_rate": 1.212120020493528e-06, + "loss": 1.2263, + "step": 35195 + }, + { + "epoch": 0.85, + "learning_rate": 1.211748349103592e-06, + "loss": 0.896, + "step": 35196 + }, + { + "epoch": 0.85, + "learning_rate": 1.211376731029914e-06, + "loss": 1.0038, + "step": 35197 + }, + { + "epoch": 0.85, + "learning_rate": 1.2110051662747434e-06, + "loss": 1.2437, + "step": 35198 + }, + { + "epoch": 0.85, + "learning_rate": 1.2106336548403374e-06, + "loss": 1.1319, + "step": 35199 + }, + { + "epoch": 0.85, + "learning_rate": 1.2102621967289519e-06, + "loss": 0.959, + "step": 35200 + }, + { + "epoch": 0.85, + "learning_rate": 1.209890791942837e-06, + "loss": 0.9863, + "step": 35201 + }, + { + "epoch": 0.85, + "learning_rate": 1.2095194404842481e-06, + "loss": 1.2697, + "step": 35202 + }, + { + "epoch": 0.85, + "learning_rate": 1.2091481423554353e-06, + "loss": 0.7757, + "step": 35203 + }, + { + "epoch": 0.85, + "learning_rate": 1.2087768975586522e-06, + "loss": 1.1344, + "step": 35204 + }, + { + "epoch": 0.85, + "learning_rate": 1.208405706096155e-06, + "loss": 1.2216, + "step": 35205 + }, + { + "epoch": 0.85, + "learning_rate": 1.2080345679701899e-06, + "loss": 0.936, + "step": 35206 + }, + { + "epoch": 0.85, + "learning_rate": 1.2076634831830115e-06, + "loss": 1.1333, + "step": 35207 + }, + { + "epoch": 0.85, + "learning_rate": 1.2072924517368722e-06, + "loss": 0.9647, + "step": 35208 + }, + { + "epoch": 0.85, + "learning_rate": 1.2069214736340195e-06, + "loss": 1.1527, + "step": 35209 + }, + { + "epoch": 0.85, + "learning_rate": 1.206550548876708e-06, + "loss": 1.0231, + "step": 35210 + }, + { + "epoch": 0.85, + "learning_rate": 1.206179677467184e-06, + "loss": 1.0914, + "step": 35211 + }, + { + "epoch": 0.85, + "learning_rate": 1.2058088594077e-06, + "loss": 1.0678, + "step": 35212 + }, + { + "epoch": 0.85, + "learning_rate": 1.2054380947005062e-06, + "loss": 1.1415, + "step": 35213 + }, + { + "epoch": 0.85, + "learning_rate": 1.2050673833478498e-06, + "loss": 0.8943, + "step": 35214 + }, + { + "epoch": 0.85, + "learning_rate": 1.2046967253519803e-06, + "loss": 1.038, + "step": 35215 + }, + { + "epoch": 0.85, + "learning_rate": 1.204326120715149e-06, + "loss": 1.3071, + "step": 35216 + }, + { + "epoch": 0.85, + "learning_rate": 1.2039555694396011e-06, + "loss": 1.2555, + "step": 35217 + }, + { + "epoch": 0.85, + "learning_rate": 1.2035850715275865e-06, + "loss": 0.9605, + "step": 35218 + }, + { + "epoch": 0.85, + "learning_rate": 1.2032146269813504e-06, + "loss": 1.0309, + "step": 35219 + }, + { + "epoch": 0.85, + "learning_rate": 1.2028442358031433e-06, + "loss": 1.1571, + "step": 35220 + }, + { + "epoch": 0.85, + "learning_rate": 1.2024738979952123e-06, + "loss": 0.9289, + "step": 35221 + }, + { + "epoch": 0.85, + "learning_rate": 1.2021036135597996e-06, + "loss": 1.0236, + "step": 35222 + }, + { + "epoch": 0.85, + "learning_rate": 1.2017333824991561e-06, + "loss": 0.9532, + "step": 35223 + }, + { + "epoch": 0.85, + "learning_rate": 1.2013632048155288e-06, + "loss": 1.0769, + "step": 35224 + }, + { + "epoch": 0.85, + "learning_rate": 1.200993080511159e-06, + "loss": 1.167, + "step": 35225 + }, + { + "epoch": 0.85, + "learning_rate": 1.2006230095882965e-06, + "loss": 1.0858, + "step": 35226 + }, + { + "epoch": 0.85, + "learning_rate": 1.2002529920491835e-06, + "loss": 1.0653, + "step": 35227 + }, + { + "epoch": 0.85, + "learning_rate": 1.1998830278960638e-06, + "loss": 1.0966, + "step": 35228 + }, + { + "epoch": 0.85, + "learning_rate": 1.199513117131188e-06, + "loss": 1.2549, + "step": 35229 + }, + { + "epoch": 0.85, + "learning_rate": 1.199143259756792e-06, + "loss": 1.202, + "step": 35230 + }, + { + "epoch": 0.85, + "learning_rate": 1.1987734557751252e-06, + "loss": 0.9512, + "step": 35231 + }, + { + "epoch": 0.85, + "learning_rate": 1.1984037051884312e-06, + "loss": 1.4664, + "step": 35232 + }, + { + "epoch": 0.85, + "learning_rate": 1.1980340079989504e-06, + "loss": 0.9613, + "step": 35233 + }, + { + "epoch": 0.85, + "learning_rate": 1.197664364208928e-06, + "loss": 1.1934, + "step": 35234 + }, + { + "epoch": 0.85, + "learning_rate": 1.1972947738206031e-06, + "loss": 0.9889, + "step": 35235 + }, + { + "epoch": 0.85, + "learning_rate": 1.1969252368362216e-06, + "loss": 1.0269, + "step": 35236 + }, + { + "epoch": 0.85, + "learning_rate": 1.196555753258025e-06, + "loss": 1.195, + "step": 35237 + }, + { + "epoch": 0.85, + "learning_rate": 1.1961863230882531e-06, + "loss": 0.9473, + "step": 35238 + }, + { + "epoch": 0.85, + "learning_rate": 1.195816946329147e-06, + "loss": 1.1484, + "step": 35239 + }, + { + "epoch": 0.85, + "learning_rate": 1.1954476229829504e-06, + "loss": 1.0133, + "step": 35240 + }, + { + "epoch": 0.85, + "learning_rate": 1.1950783530519006e-06, + "loss": 1.143, + "step": 35241 + }, + { + "epoch": 0.85, + "learning_rate": 1.1947091365382412e-06, + "loss": 1.0856, + "step": 35242 + }, + { + "epoch": 0.85, + "learning_rate": 1.1943399734442084e-06, + "loss": 1.1364, + "step": 35243 + }, + { + "epoch": 0.85, + "learning_rate": 1.1939708637720438e-06, + "loss": 0.8325, + "step": 35244 + }, + { + "epoch": 0.85, + "learning_rate": 1.1936018075239896e-06, + "loss": 1.0134, + "step": 35245 + }, + { + "epoch": 0.85, + "learning_rate": 1.1932328047022789e-06, + "loss": 1.0106, + "step": 35246 + }, + { + "epoch": 0.85, + "learning_rate": 1.1928638553091532e-06, + "loss": 1.1848, + "step": 35247 + }, + { + "epoch": 0.85, + "learning_rate": 1.1924949593468526e-06, + "loss": 1.2179, + "step": 35248 + }, + { + "epoch": 0.85, + "learning_rate": 1.1921261168176124e-06, + "loss": 1.1675, + "step": 35249 + }, + { + "epoch": 0.85, + "learning_rate": 1.1917573277236705e-06, + "loss": 1.0318, + "step": 35250 + }, + { + "epoch": 0.85, + "learning_rate": 1.1913885920672674e-06, + "loss": 1.2906, + "step": 35251 + }, + { + "epoch": 0.85, + "learning_rate": 1.1910199098506358e-06, + "loss": 1.0729, + "step": 35252 + }, + { + "epoch": 0.85, + "learning_rate": 1.1906512810760162e-06, + "loss": 0.9112, + "step": 35253 + }, + { + "epoch": 0.85, + "learning_rate": 1.1902827057456424e-06, + "loss": 1.1734, + "step": 35254 + }, + { + "epoch": 0.85, + "learning_rate": 1.18991418386175e-06, + "loss": 1.0718, + "step": 35255 + }, + { + "epoch": 0.85, + "learning_rate": 1.1895457154265778e-06, + "loss": 0.8866, + "step": 35256 + }, + { + "epoch": 0.85, + "learning_rate": 1.1891773004423579e-06, + "loss": 1.0357, + "step": 35257 + }, + { + "epoch": 0.85, + "learning_rate": 1.1888089389113267e-06, + "loss": 1.0316, + "step": 35258 + }, + { + "epoch": 0.85, + "learning_rate": 1.188440630835721e-06, + "loss": 1.0619, + "step": 35259 + }, + { + "epoch": 0.85, + "learning_rate": 1.1880723762177714e-06, + "loss": 1.1109, + "step": 35260 + }, + { + "epoch": 0.85, + "learning_rate": 1.1877041750597174e-06, + "loss": 0.9549, + "step": 35261 + }, + { + "epoch": 0.85, + "learning_rate": 1.1873360273637858e-06, + "loss": 0.8611, + "step": 35262 + }, + { + "epoch": 0.85, + "learning_rate": 1.1869679331322137e-06, + "loss": 0.8249, + "step": 35263 + }, + { + "epoch": 0.85, + "learning_rate": 1.1865998923672373e-06, + "loss": 0.8615, + "step": 35264 + }, + { + "epoch": 0.85, + "learning_rate": 1.1862319050710836e-06, + "loss": 1.1209, + "step": 35265 + }, + { + "epoch": 0.85, + "learning_rate": 1.1858639712459875e-06, + "loss": 1.2596, + "step": 35266 + }, + { + "epoch": 0.85, + "learning_rate": 1.1854960908941838e-06, + "loss": 1.075, + "step": 35267 + }, + { + "epoch": 0.85, + "learning_rate": 1.1851282640178996e-06, + "loss": 1.1374, + "step": 35268 + }, + { + "epoch": 0.85, + "learning_rate": 1.1847604906193722e-06, + "loss": 1.2729, + "step": 35269 + }, + { + "epoch": 0.85, + "learning_rate": 1.1843927707008262e-06, + "loss": 0.9737, + "step": 35270 + }, + { + "epoch": 0.85, + "learning_rate": 1.1840251042644956e-06, + "loss": 0.9477, + "step": 35271 + }, + { + "epoch": 0.85, + "learning_rate": 1.183657491312613e-06, + "loss": 1.1187, + "step": 35272 + }, + { + "epoch": 0.85, + "learning_rate": 1.1832899318474056e-06, + "loss": 0.9571, + "step": 35273 + }, + { + "epoch": 0.85, + "learning_rate": 1.1829224258711036e-06, + "loss": 0.8412, + "step": 35274 + }, + { + "epoch": 0.85, + "learning_rate": 1.18255497338594e-06, + "loss": 1.09, + "step": 35275 + }, + { + "epoch": 0.85, + "learning_rate": 1.1821875743941392e-06, + "loss": 1.1906, + "step": 35276 + }, + { + "epoch": 0.85, + "learning_rate": 1.181820228897934e-06, + "loss": 1.1307, + "step": 35277 + }, + { + "epoch": 0.85, + "learning_rate": 1.1814529368995498e-06, + "loss": 1.2642, + "step": 35278 + }, + { + "epoch": 0.85, + "learning_rate": 1.1810856984012154e-06, + "loss": 1.0084, + "step": 35279 + }, + { + "epoch": 0.85, + "learning_rate": 1.1807185134051624e-06, + "loss": 1.1727, + "step": 35280 + }, + { + "epoch": 0.85, + "learning_rate": 1.1803513819136136e-06, + "loss": 1.0005, + "step": 35281 + }, + { + "epoch": 0.85, + "learning_rate": 1.1799843039287984e-06, + "loss": 1.2968, + "step": 35282 + }, + { + "epoch": 0.85, + "learning_rate": 1.179617279452946e-06, + "loss": 0.9454, + "step": 35283 + }, + { + "epoch": 0.85, + "learning_rate": 1.179250308488279e-06, + "loss": 1.0304, + "step": 35284 + }, + { + "epoch": 0.85, + "learning_rate": 1.1788833910370279e-06, + "loss": 1.0826, + "step": 35285 + }, + { + "epoch": 0.85, + "learning_rate": 1.1785165271014144e-06, + "loss": 0.718, + "step": 35286 + }, + { + "epoch": 0.85, + "learning_rate": 1.1781497166836652e-06, + "loss": 0.9135, + "step": 35287 + }, + { + "epoch": 0.85, + "learning_rate": 1.1777829597860092e-06, + "loss": 0.9822, + "step": 35288 + }, + { + "epoch": 0.85, + "learning_rate": 1.1774162564106672e-06, + "loss": 1.1704, + "step": 35289 + }, + { + "epoch": 0.85, + "learning_rate": 1.1770496065598647e-06, + "loss": 1.0333, + "step": 35290 + }, + { + "epoch": 0.85, + "learning_rate": 1.1766830102358295e-06, + "loss": 0.8715, + "step": 35291 + }, + { + "epoch": 0.85, + "learning_rate": 1.176316467440781e-06, + "loss": 1.0588, + "step": 35292 + }, + { + "epoch": 0.85, + "learning_rate": 1.1759499781769457e-06, + "loss": 1.0726, + "step": 35293 + }, + { + "epoch": 0.85, + "learning_rate": 1.1755835424465478e-06, + "loss": 1.1426, + "step": 35294 + }, + { + "epoch": 0.85, + "learning_rate": 1.175217160251807e-06, + "loss": 1.0648, + "step": 35295 + }, + { + "epoch": 0.85, + "learning_rate": 1.1748508315949502e-06, + "loss": 0.9854, + "step": 35296 + }, + { + "epoch": 0.85, + "learning_rate": 1.1744845564781948e-06, + "loss": 1.0691, + "step": 35297 + }, + { + "epoch": 0.85, + "learning_rate": 1.1741183349037665e-06, + "loss": 0.9459, + "step": 35298 + }, + { + "epoch": 0.85, + "learning_rate": 1.1737521668738883e-06, + "loss": 0.9242, + "step": 35299 + }, + { + "epoch": 0.85, + "learning_rate": 1.173386052390777e-06, + "loss": 1.024, + "step": 35300 + }, + { + "epoch": 0.85, + "learning_rate": 1.1730199914566565e-06, + "loss": 1.1467, + "step": 35301 + }, + { + "epoch": 0.85, + "learning_rate": 1.1726539840737494e-06, + "loss": 1.2459, + "step": 35302 + }, + { + "epoch": 0.85, + "learning_rate": 1.172288030244273e-06, + "loss": 0.9094, + "step": 35303 + }, + { + "epoch": 0.85, + "learning_rate": 1.1719221299704497e-06, + "loss": 1.1441, + "step": 35304 + }, + { + "epoch": 0.85, + "learning_rate": 1.1715562832544968e-06, + "loss": 0.9854, + "step": 35305 + }, + { + "epoch": 0.85, + "learning_rate": 1.1711904900986358e-06, + "loss": 0.8692, + "step": 35306 + }, + { + "epoch": 0.85, + "learning_rate": 1.1708247505050873e-06, + "loss": 1.2614, + "step": 35307 + }, + { + "epoch": 0.85, + "learning_rate": 1.170459064476066e-06, + "loss": 1.0431, + "step": 35308 + }, + { + "epoch": 0.85, + "learning_rate": 1.1700934320137924e-06, + "loss": 1.0444, + "step": 35309 + }, + { + "epoch": 0.85, + "learning_rate": 1.169727853120488e-06, + "loss": 1.3312, + "step": 35310 + }, + { + "epoch": 0.85, + "learning_rate": 1.1693623277983646e-06, + "loss": 1.1263, + "step": 35311 + }, + { + "epoch": 0.85, + "learning_rate": 1.1689968560496456e-06, + "loss": 1.1585, + "step": 35312 + }, + { + "epoch": 0.85, + "learning_rate": 1.168631437876544e-06, + "loss": 0.8899, + "step": 35313 + }, + { + "epoch": 0.85, + "learning_rate": 1.1682660732812767e-06, + "loss": 1.0619, + "step": 35314 + }, + { + "epoch": 0.85, + "learning_rate": 1.1679007622660654e-06, + "loss": 0.8791, + "step": 35315 + }, + { + "epoch": 0.85, + "learning_rate": 1.1675355048331195e-06, + "loss": 1.215, + "step": 35316 + }, + { + "epoch": 0.85, + "learning_rate": 1.1671703009846591e-06, + "loss": 1.0433, + "step": 35317 + }, + { + "epoch": 0.85, + "learning_rate": 1.1668051507228995e-06, + "loss": 1.1792, + "step": 35318 + }, + { + "epoch": 0.85, + "learning_rate": 1.1664400540500541e-06, + "loss": 1.1755, + "step": 35319 + }, + { + "epoch": 0.85, + "learning_rate": 1.1660750109683415e-06, + "loss": 1.1887, + "step": 35320 + }, + { + "epoch": 0.85, + "learning_rate": 1.1657100214799722e-06, + "loss": 1.2419, + "step": 35321 + }, + { + "epoch": 0.85, + "learning_rate": 1.1653450855871618e-06, + "loss": 1.0682, + "step": 35322 + }, + { + "epoch": 0.85, + "learning_rate": 1.1649802032921266e-06, + "loss": 1.0522, + "step": 35323 + }, + { + "epoch": 0.85, + "learning_rate": 1.1646153745970768e-06, + "loss": 0.9937, + "step": 35324 + }, + { + "epoch": 0.85, + "learning_rate": 1.1642505995042263e-06, + "loss": 0.9992, + "step": 35325 + }, + { + "epoch": 0.85, + "learning_rate": 1.1638858780157925e-06, + "loss": 0.9125, + "step": 35326 + }, + { + "epoch": 0.85, + "learning_rate": 1.163521210133981e-06, + "loss": 1.0908, + "step": 35327 + }, + { + "epoch": 0.85, + "learning_rate": 1.1631565958610115e-06, + "loss": 1.1849, + "step": 35328 + }, + { + "epoch": 0.85, + "learning_rate": 1.1627920351990895e-06, + "loss": 0.9514, + "step": 35329 + }, + { + "epoch": 0.85, + "learning_rate": 1.1624275281504294e-06, + "loss": 0.9501, + "step": 35330 + }, + { + "epoch": 0.85, + "learning_rate": 1.1620630747172456e-06, + "loss": 1.0049, + "step": 35331 + }, + { + "epoch": 0.85, + "learning_rate": 1.1616986749017412e-06, + "loss": 1.0266, + "step": 35332 + }, + { + "epoch": 0.85, + "learning_rate": 1.1613343287061373e-06, + "loss": 0.8078, + "step": 35333 + }, + { + "epoch": 0.85, + "learning_rate": 1.160970036132636e-06, + "loss": 0.9229, + "step": 35334 + }, + { + "epoch": 0.85, + "learning_rate": 1.1606057971834516e-06, + "loss": 1.0867, + "step": 35335 + }, + { + "epoch": 0.85, + "learning_rate": 1.160241611860794e-06, + "loss": 0.8589, + "step": 35336 + }, + { + "epoch": 0.85, + "learning_rate": 1.1598774801668688e-06, + "loss": 1.1202, + "step": 35337 + }, + { + "epoch": 0.85, + "learning_rate": 1.159513402103888e-06, + "loss": 1.0334, + "step": 35338 + }, + { + "epoch": 0.85, + "learning_rate": 1.1591493776740626e-06, + "loss": 1.0877, + "step": 35339 + }, + { + "epoch": 0.85, + "learning_rate": 1.1587854068795957e-06, + "loss": 0.7747, + "step": 35340 + }, + { + "epoch": 0.85, + "learning_rate": 1.1584214897226996e-06, + "loss": 1.0548, + "step": 35341 + }, + { + "epoch": 0.85, + "learning_rate": 1.1580576262055798e-06, + "loss": 0.9208, + "step": 35342 + }, + { + "epoch": 0.85, + "learning_rate": 1.157693816330444e-06, + "loss": 0.827, + "step": 35343 + }, + { + "epoch": 0.85, + "learning_rate": 1.157330060099503e-06, + "loss": 1.0613, + "step": 35344 + }, + { + "epoch": 0.85, + "learning_rate": 1.156966357514957e-06, + "loss": 0.9507, + "step": 35345 + }, + { + "epoch": 0.85, + "learning_rate": 1.1566027085790177e-06, + "loss": 0.9409, + "step": 35346 + }, + { + "epoch": 0.85, + "learning_rate": 1.1562391132938899e-06, + "loss": 1.0755, + "step": 35347 + }, + { + "epoch": 0.85, + "learning_rate": 1.1558755716617786e-06, + "loss": 1.058, + "step": 35348 + }, + { + "epoch": 0.85, + "learning_rate": 1.155512083684892e-06, + "loss": 1.2542, + "step": 35349 + }, + { + "epoch": 0.85, + "learning_rate": 1.1551486493654307e-06, + "loss": 1.048, + "step": 35350 + }, + { + "epoch": 0.85, + "learning_rate": 1.1547852687056016e-06, + "loss": 1.1348, + "step": 35351 + }, + { + "epoch": 0.85, + "learning_rate": 1.154421941707613e-06, + "loss": 0.7683, + "step": 35352 + }, + { + "epoch": 0.85, + "learning_rate": 1.1540586683736632e-06, + "loss": 0.941, + "step": 35353 + }, + { + "epoch": 0.85, + "learning_rate": 1.1536954487059582e-06, + "loss": 1.1594, + "step": 35354 + }, + { + "epoch": 0.85, + "learning_rate": 1.153332282706705e-06, + "loss": 1.0616, + "step": 35355 + }, + { + "epoch": 0.85, + "learning_rate": 1.1529691703781022e-06, + "loss": 1.2148, + "step": 35356 + }, + { + "epoch": 0.85, + "learning_rate": 1.1526061117223552e-06, + "loss": 1.3358, + "step": 35357 + }, + { + "epoch": 0.85, + "learning_rate": 1.152243106741665e-06, + "loss": 1.0557, + "step": 35358 + }, + { + "epoch": 0.85, + "learning_rate": 1.1518801554382341e-06, + "loss": 0.9946, + "step": 35359 + }, + { + "epoch": 0.85, + "learning_rate": 1.1515172578142675e-06, + "loss": 1.0662, + "step": 35360 + }, + { + "epoch": 0.85, + "learning_rate": 1.1511544138719622e-06, + "loss": 0.885, + "step": 35361 + }, + { + "epoch": 0.85, + "learning_rate": 1.150791623613522e-06, + "loss": 1.1064, + "step": 35362 + }, + { + "epoch": 0.85, + "learning_rate": 1.1504288870411496e-06, + "loss": 1.2463, + "step": 35363 + }, + { + "epoch": 0.85, + "learning_rate": 1.150066204157041e-06, + "loss": 1.2386, + "step": 35364 + }, + { + "epoch": 0.85, + "learning_rate": 1.1497035749634011e-06, + "loss": 0.9815, + "step": 35365 + }, + { + "epoch": 0.85, + "learning_rate": 1.1493409994624272e-06, + "loss": 1.1361, + "step": 35366 + }, + { + "epoch": 0.85, + "learning_rate": 1.1489784776563185e-06, + "loss": 1.0597, + "step": 35367 + }, + { + "epoch": 0.85, + "learning_rate": 1.1486160095472776e-06, + "loss": 1.1218, + "step": 35368 + }, + { + "epoch": 0.85, + "learning_rate": 1.1482535951374995e-06, + "loss": 1.1009, + "step": 35369 + }, + { + "epoch": 0.85, + "learning_rate": 1.1478912344291848e-06, + "loss": 1.0793, + "step": 35370 + }, + { + "epoch": 0.85, + "learning_rate": 1.1475289274245338e-06, + "loss": 1.1627, + "step": 35371 + }, + { + "epoch": 0.85, + "learning_rate": 1.1471666741257404e-06, + "loss": 0.9755, + "step": 35372 + }, + { + "epoch": 0.85, + "learning_rate": 1.146804474535005e-06, + "loss": 1.1995, + "step": 35373 + }, + { + "epoch": 0.85, + "learning_rate": 1.1464423286545256e-06, + "loss": 0.9685, + "step": 35374 + }, + { + "epoch": 0.85, + "learning_rate": 1.1460802364864966e-06, + "loss": 1.0585, + "step": 35375 + }, + { + "epoch": 0.85, + "learning_rate": 1.1457181980331188e-06, + "loss": 1.0073, + "step": 35376 + }, + { + "epoch": 0.85, + "learning_rate": 1.1453562132965835e-06, + "loss": 0.8449, + "step": 35377 + }, + { + "epoch": 0.85, + "learning_rate": 1.1449942822790894e-06, + "loss": 1.0942, + "step": 35378 + }, + { + "epoch": 0.85, + "learning_rate": 1.144632404982834e-06, + "loss": 1.0542, + "step": 35379 + }, + { + "epoch": 0.85, + "learning_rate": 1.1442705814100085e-06, + "loss": 0.9912, + "step": 35380 + }, + { + "epoch": 0.85, + "learning_rate": 1.1439088115628105e-06, + "loss": 1.1095, + "step": 35381 + }, + { + "epoch": 0.85, + "learning_rate": 1.143547095443438e-06, + "loss": 0.9654, + "step": 35382 + }, + { + "epoch": 0.85, + "learning_rate": 1.1431854330540791e-06, + "loss": 1.0219, + "step": 35383 + }, + { + "epoch": 0.85, + "learning_rate": 1.1428238243969325e-06, + "loss": 1.0115, + "step": 35384 + }, + { + "epoch": 0.85, + "learning_rate": 1.1424622694741893e-06, + "loss": 1.0063, + "step": 35385 + }, + { + "epoch": 0.85, + "learning_rate": 1.1421007682880436e-06, + "loss": 1.0717, + "step": 35386 + }, + { + "epoch": 0.85, + "learning_rate": 1.1417393208406913e-06, + "loss": 0.8277, + "step": 35387 + }, + { + "epoch": 0.85, + "learning_rate": 1.1413779271343206e-06, + "loss": 1.2365, + "step": 35388 + }, + { + "epoch": 0.85, + "learning_rate": 1.1410165871711266e-06, + "loss": 1.1238, + "step": 35389 + }, + { + "epoch": 0.85, + "learning_rate": 1.1406553009533028e-06, + "loss": 0.9804, + "step": 35390 + }, + { + "epoch": 0.85, + "learning_rate": 1.1402940684830377e-06, + "loss": 1.1287, + "step": 35391 + }, + { + "epoch": 0.85, + "learning_rate": 1.1399328897625262e-06, + "loss": 0.8124, + "step": 35392 + }, + { + "epoch": 0.85, + "learning_rate": 1.1395717647939563e-06, + "loss": 1.2293, + "step": 35393 + }, + { + "epoch": 0.85, + "learning_rate": 1.139210693579521e-06, + "loss": 1.114, + "step": 35394 + }, + { + "epoch": 0.85, + "learning_rate": 1.1388496761214107e-06, + "loss": 1.2004, + "step": 35395 + }, + { + "epoch": 0.85, + "learning_rate": 1.1384887124218147e-06, + "loss": 1.1879, + "step": 35396 + }, + { + "epoch": 0.85, + "learning_rate": 1.1381278024829223e-06, + "loss": 1.1782, + "step": 35397 + }, + { + "epoch": 0.85, + "learning_rate": 1.1377669463069263e-06, + "loss": 1.1458, + "step": 35398 + }, + { + "epoch": 0.85, + "learning_rate": 1.1374061438960127e-06, + "loss": 1.1492, + "step": 35399 + }, + { + "epoch": 0.85, + "learning_rate": 1.1370453952523719e-06, + "loss": 0.7653, + "step": 35400 + }, + { + "epoch": 0.85, + "learning_rate": 1.1366847003781911e-06, + "loss": 0.9285, + "step": 35401 + }, + { + "epoch": 0.85, + "learning_rate": 1.1363240592756586e-06, + "loss": 1.0365, + "step": 35402 + }, + { + "epoch": 0.85, + "learning_rate": 1.1359634719469658e-06, + "loss": 1.0941, + "step": 35403 + }, + { + "epoch": 0.85, + "learning_rate": 1.1356029383942958e-06, + "loss": 1.1749, + "step": 35404 + }, + { + "epoch": 0.85, + "learning_rate": 1.1352424586198373e-06, + "loss": 1.0807, + "step": 35405 + }, + { + "epoch": 0.85, + "learning_rate": 1.1348820326257792e-06, + "loss": 1.0954, + "step": 35406 + }, + { + "epoch": 0.85, + "learning_rate": 1.1345216604143051e-06, + "loss": 1.1647, + "step": 35407 + }, + { + "epoch": 0.85, + "learning_rate": 1.1341613419876052e-06, + "loss": 0.9198, + "step": 35408 + }, + { + "epoch": 0.85, + "learning_rate": 1.1338010773478602e-06, + "loss": 1.0576, + "step": 35409 + }, + { + "epoch": 0.85, + "learning_rate": 1.1334408664972596e-06, + "loss": 1.0747, + "step": 35410 + }, + { + "epoch": 0.85, + "learning_rate": 1.133080709437989e-06, + "loss": 1.0836, + "step": 35411 + }, + { + "epoch": 0.85, + "learning_rate": 1.1327206061722308e-06, + "loss": 0.9275, + "step": 35412 + }, + { + "epoch": 0.85, + "learning_rate": 1.13236055670217e-06, + "loss": 1.1911, + "step": 35413 + }, + { + "epoch": 0.85, + "learning_rate": 1.1320005610299944e-06, + "loss": 0.933, + "step": 35414 + }, + { + "epoch": 0.85, + "learning_rate": 1.1316406191578844e-06, + "loss": 0.9314, + "step": 35415 + }, + { + "epoch": 0.85, + "learning_rate": 1.1312807310880257e-06, + "loss": 0.9537, + "step": 35416 + }, + { + "epoch": 0.85, + "learning_rate": 1.1309208968225993e-06, + "loss": 1.1051, + "step": 35417 + }, + { + "epoch": 0.85, + "learning_rate": 1.1305611163637886e-06, + "loss": 1.1604, + "step": 35418 + }, + { + "epoch": 0.85, + "learning_rate": 1.130201389713781e-06, + "loss": 0.8551, + "step": 35419 + }, + { + "epoch": 0.85, + "learning_rate": 1.1298417168747522e-06, + "loss": 1.183, + "step": 35420 + }, + { + "epoch": 0.85, + "learning_rate": 1.1294820978488874e-06, + "loss": 1.0243, + "step": 35421 + }, + { + "epoch": 0.85, + "learning_rate": 1.1291225326383703e-06, + "loss": 0.9859, + "step": 35422 + }, + { + "epoch": 0.85, + "learning_rate": 1.1287630212453781e-06, + "loss": 0.7502, + "step": 35423 + }, + { + "epoch": 0.85, + "learning_rate": 1.1284035636720948e-06, + "loss": 0.8799, + "step": 35424 + }, + { + "epoch": 0.85, + "learning_rate": 1.1280441599207016e-06, + "loss": 0.9257, + "step": 35425 + }, + { + "epoch": 0.85, + "learning_rate": 1.1276848099933758e-06, + "loss": 1.0867, + "step": 35426 + }, + { + "epoch": 0.85, + "learning_rate": 1.1273255138923012e-06, + "loss": 0.9145, + "step": 35427 + }, + { + "epoch": 0.85, + "learning_rate": 1.1269662716196538e-06, + "loss": 1.1782, + "step": 35428 + }, + { + "epoch": 0.85, + "learning_rate": 1.1266070831776155e-06, + "loss": 1.0564, + "step": 35429 + }, + { + "epoch": 0.85, + "learning_rate": 1.1262479485683663e-06, + "loss": 0.8472, + "step": 35430 + }, + { + "epoch": 0.85, + "learning_rate": 1.1258888677940816e-06, + "loss": 1.0734, + "step": 35431 + }, + { + "epoch": 0.85, + "learning_rate": 1.1255298408569427e-06, + "loss": 0.7522, + "step": 35432 + }, + { + "epoch": 0.85, + "learning_rate": 1.1251708677591278e-06, + "loss": 1.1309, + "step": 35433 + }, + { + "epoch": 0.85, + "learning_rate": 1.124811948502812e-06, + "loss": 1.0661, + "step": 35434 + }, + { + "epoch": 0.85, + "learning_rate": 1.1244530830901768e-06, + "loss": 0.9086, + "step": 35435 + }, + { + "epoch": 0.85, + "learning_rate": 1.1240942715233948e-06, + "loss": 1.1006, + "step": 35436 + }, + { + "epoch": 0.85, + "learning_rate": 1.1237355138046458e-06, + "loss": 0.8357, + "step": 35437 + }, + { + "epoch": 0.85, + "learning_rate": 1.1233768099361065e-06, + "loss": 0.8687, + "step": 35438 + }, + { + "epoch": 0.85, + "learning_rate": 1.1230181599199508e-06, + "loss": 0.9869, + "step": 35439 + }, + { + "epoch": 0.85, + "learning_rate": 1.122659563758356e-06, + "loss": 1.0963, + "step": 35440 + }, + { + "epoch": 0.85, + "learning_rate": 1.1223010214534991e-06, + "loss": 1.0462, + "step": 35441 + }, + { + "epoch": 0.85, + "learning_rate": 1.1219425330075518e-06, + "loss": 0.7217, + "step": 35442 + }, + { + "epoch": 0.85, + "learning_rate": 1.1215840984226933e-06, + "loss": 1.2107, + "step": 35443 + }, + { + "epoch": 0.85, + "learning_rate": 1.1212257177010944e-06, + "loss": 0.9659, + "step": 35444 + }, + { + "epoch": 0.85, + "learning_rate": 1.1208673908449298e-06, + "loss": 1.2887, + "step": 35445 + }, + { + "epoch": 0.85, + "learning_rate": 1.1205091178563753e-06, + "loss": 0.8315, + "step": 35446 + }, + { + "epoch": 0.85, + "learning_rate": 1.1201508987376031e-06, + "loss": 1.0625, + "step": 35447 + }, + { + "epoch": 0.85, + "learning_rate": 1.1197927334907854e-06, + "loss": 1.0246, + "step": 35448 + }, + { + "epoch": 0.85, + "learning_rate": 1.1194346221180995e-06, + "loss": 1.1834, + "step": 35449 + }, + { + "epoch": 0.85, + "learning_rate": 1.1190765646217128e-06, + "loss": 1.0848, + "step": 35450 + }, + { + "epoch": 0.85, + "learning_rate": 1.118718561003801e-06, + "loss": 1.1175, + "step": 35451 + }, + { + "epoch": 0.85, + "learning_rate": 1.1183606112665334e-06, + "loss": 1.1949, + "step": 35452 + }, + { + "epoch": 0.85, + "learning_rate": 1.118002715412082e-06, + "loss": 0.9809, + "step": 35453 + }, + { + "epoch": 0.85, + "learning_rate": 1.1176448734426214e-06, + "loss": 0.7945, + "step": 35454 + }, + { + "epoch": 0.85, + "learning_rate": 1.117287085360318e-06, + "loss": 0.9352, + "step": 35455 + }, + { + "epoch": 0.85, + "learning_rate": 1.116929351167345e-06, + "loss": 1.2911, + "step": 35456 + }, + { + "epoch": 0.85, + "learning_rate": 1.1165716708658737e-06, + "loss": 1.1773, + "step": 35457 + }, + { + "epoch": 0.85, + "learning_rate": 1.1162140444580705e-06, + "loss": 1.3006, + "step": 35458 + }, + { + "epoch": 0.85, + "learning_rate": 1.1158564719461095e-06, + "loss": 1.0834, + "step": 35459 + }, + { + "epoch": 0.85, + "learning_rate": 1.1154989533321558e-06, + "loss": 0.8093, + "step": 35460 + }, + { + "epoch": 0.85, + "learning_rate": 1.1151414886183799e-06, + "loss": 1.0135, + "step": 35461 + }, + { + "epoch": 0.85, + "learning_rate": 1.1147840778069518e-06, + "loss": 1.0292, + "step": 35462 + }, + { + "epoch": 0.85, + "learning_rate": 1.1144267209000382e-06, + "loss": 0.965, + "step": 35463 + }, + { + "epoch": 0.85, + "learning_rate": 1.1140694178998069e-06, + "loss": 1.076, + "step": 35464 + }, + { + "epoch": 0.85, + "learning_rate": 1.1137121688084285e-06, + "loss": 0.7735, + "step": 35465 + }, + { + "epoch": 0.85, + "learning_rate": 1.113354973628067e-06, + "loss": 0.8917, + "step": 35466 + }, + { + "epoch": 0.85, + "learning_rate": 1.1129978323608891e-06, + "loss": 1.1215, + "step": 35467 + }, + { + "epoch": 0.85, + "learning_rate": 1.1126407450090658e-06, + "loss": 1.0854, + "step": 35468 + }, + { + "epoch": 0.85, + "learning_rate": 1.1122837115747587e-06, + "loss": 0.9799, + "step": 35469 + }, + { + "epoch": 0.85, + "learning_rate": 1.111926732060138e-06, + "loss": 0.9404, + "step": 35470 + }, + { + "epoch": 0.85, + "learning_rate": 1.1115698064673642e-06, + "loss": 1.1867, + "step": 35471 + }, + { + "epoch": 0.85, + "learning_rate": 1.1112129347986068e-06, + "loss": 1.0357, + "step": 35472 + }, + { + "epoch": 0.85, + "learning_rate": 1.1108561170560316e-06, + "loss": 0.9055, + "step": 35473 + }, + { + "epoch": 0.85, + "learning_rate": 1.1104993532417996e-06, + "loss": 1.0464, + "step": 35474 + }, + { + "epoch": 0.85, + "learning_rate": 1.1101426433580764e-06, + "loss": 0.9449, + "step": 35475 + }, + { + "epoch": 0.85, + "learning_rate": 1.1097859874070294e-06, + "loss": 0.9331, + "step": 35476 + }, + { + "epoch": 0.85, + "learning_rate": 1.1094293853908167e-06, + "loss": 1.0571, + "step": 35477 + }, + { + "epoch": 0.85, + "learning_rate": 1.1090728373116077e-06, + "loss": 1.046, + "step": 35478 + }, + { + "epoch": 0.85, + "learning_rate": 1.1087163431715597e-06, + "loss": 1.1006, + "step": 35479 + }, + { + "epoch": 0.85, + "learning_rate": 1.1083599029728399e-06, + "loss": 0.9673, + "step": 35480 + }, + { + "epoch": 0.85, + "learning_rate": 1.1080035167176096e-06, + "loss": 1.1142, + "step": 35481 + }, + { + "epoch": 0.85, + "learning_rate": 1.1076471844080283e-06, + "loss": 0.7895, + "step": 35482 + }, + { + "epoch": 0.85, + "learning_rate": 1.1072909060462611e-06, + "loss": 0.9407, + "step": 35483 + }, + { + "epoch": 0.85, + "learning_rate": 1.1069346816344695e-06, + "loss": 0.9188, + "step": 35484 + }, + { + "epoch": 0.85, + "learning_rate": 1.1065785111748117e-06, + "loss": 0.9292, + "step": 35485 + }, + { + "epoch": 0.85, + "learning_rate": 1.1062223946694527e-06, + "loss": 1.0623, + "step": 35486 + }, + { + "epoch": 0.85, + "learning_rate": 1.1058663321205475e-06, + "loss": 1.0914, + "step": 35487 + }, + { + "epoch": 0.85, + "learning_rate": 1.10551032353026e-06, + "loss": 1.0288, + "step": 35488 + }, + { + "epoch": 0.85, + "learning_rate": 1.1051543689007504e-06, + "loss": 1.214, + "step": 35489 + }, + { + "epoch": 0.85, + "learning_rate": 1.1047984682341762e-06, + "loss": 1.0134, + "step": 35490 + }, + { + "epoch": 0.85, + "learning_rate": 1.1044426215326964e-06, + "loss": 0.8756, + "step": 35491 + }, + { + "epoch": 0.85, + "learning_rate": 1.104086828798473e-06, + "loss": 1.3095, + "step": 35492 + }, + { + "epoch": 0.85, + "learning_rate": 1.1037310900336607e-06, + "loss": 1.1082, + "step": 35493 + }, + { + "epoch": 0.85, + "learning_rate": 1.1033754052404211e-06, + "loss": 0.7745, + "step": 35494 + }, + { + "epoch": 0.85, + "learning_rate": 1.103019774420908e-06, + "loss": 1.1202, + "step": 35495 + }, + { + "epoch": 0.85, + "learning_rate": 1.1026641975772823e-06, + "loss": 1.1394, + "step": 35496 + }, + { + "epoch": 0.85, + "learning_rate": 1.1023086747117018e-06, + "loss": 0.9581, + "step": 35497 + }, + { + "epoch": 0.85, + "learning_rate": 1.1019532058263194e-06, + "loss": 0.9357, + "step": 35498 + }, + { + "epoch": 0.85, + "learning_rate": 1.1015977909232934e-06, + "loss": 1.0695, + "step": 35499 + }, + { + "epoch": 0.85, + "learning_rate": 1.1012424300047818e-06, + "loss": 0.8428, + "step": 35500 + }, + { + "epoch": 0.85, + "learning_rate": 1.100887123072938e-06, + "loss": 1.0335, + "step": 35501 + }, + { + "epoch": 0.85, + "learning_rate": 1.1005318701299216e-06, + "loss": 1.0741, + "step": 35502 + }, + { + "epoch": 0.85, + "learning_rate": 1.1001766711778827e-06, + "loss": 1.2884, + "step": 35503 + }, + { + "epoch": 0.85, + "learning_rate": 1.0998215262189794e-06, + "loss": 1.1576, + "step": 35504 + }, + { + "epoch": 0.85, + "learning_rate": 1.0994664352553664e-06, + "loss": 0.7624, + "step": 35505 + }, + { + "epoch": 0.85, + "learning_rate": 1.0991113982891955e-06, + "loss": 0.9387, + "step": 35506 + }, + { + "epoch": 0.85, + "learning_rate": 1.0987564153226238e-06, + "loss": 1.106, + "step": 35507 + }, + { + "epoch": 0.85, + "learning_rate": 1.0984014863578018e-06, + "loss": 1.2866, + "step": 35508 + }, + { + "epoch": 0.85, + "learning_rate": 1.0980466113968835e-06, + "loss": 1.2138, + "step": 35509 + }, + { + "epoch": 0.85, + "learning_rate": 1.0976917904420247e-06, + "loss": 1.0584, + "step": 35510 + }, + { + "epoch": 0.85, + "learning_rate": 1.0973370234953739e-06, + "loss": 0.9926, + "step": 35511 + }, + { + "epoch": 0.85, + "learning_rate": 1.096982310559086e-06, + "loss": 1.0896, + "step": 35512 + }, + { + "epoch": 0.85, + "learning_rate": 1.0966276516353135e-06, + "loss": 0.9602, + "step": 35513 + }, + { + "epoch": 0.85, + "learning_rate": 1.096273046726205e-06, + "loss": 1.1905, + "step": 35514 + }, + { + "epoch": 0.85, + "learning_rate": 1.0959184958339165e-06, + "loss": 0.8723, + "step": 35515 + }, + { + "epoch": 0.85, + "learning_rate": 1.0955639989605925e-06, + "loss": 0.7366, + "step": 35516 + }, + { + "epoch": 0.85, + "learning_rate": 1.0952095561083887e-06, + "loss": 1.0849, + "step": 35517 + }, + { + "epoch": 0.85, + "learning_rate": 1.094855167279456e-06, + "loss": 0.9657, + "step": 35518 + }, + { + "epoch": 0.85, + "learning_rate": 1.0945008324759399e-06, + "loss": 0.9205, + "step": 35519 + }, + { + "epoch": 0.85, + "learning_rate": 1.0941465516999917e-06, + "loss": 1.2879, + "step": 35520 + }, + { + "epoch": 0.85, + "learning_rate": 1.0937923249537651e-06, + "loss": 1.0713, + "step": 35521 + }, + { + "epoch": 0.85, + "learning_rate": 1.093438152239402e-06, + "loss": 1.3135, + "step": 35522 + }, + { + "epoch": 0.85, + "learning_rate": 1.0930840335590576e-06, + "loss": 1.0801, + "step": 35523 + }, + { + "epoch": 0.85, + "learning_rate": 1.0927299689148752e-06, + "loss": 1.1495, + "step": 35524 + }, + { + "epoch": 0.85, + "learning_rate": 1.0923759583090054e-06, + "loss": 1.142, + "step": 35525 + }, + { + "epoch": 0.85, + "learning_rate": 1.0920220017435968e-06, + "loss": 1.1006, + "step": 35526 + }, + { + "epoch": 0.85, + "learning_rate": 1.091668099220794e-06, + "loss": 1.0721, + "step": 35527 + }, + { + "epoch": 0.85, + "learning_rate": 1.0913142507427455e-06, + "loss": 1.1792, + "step": 35528 + }, + { + "epoch": 0.85, + "learning_rate": 1.0909604563115994e-06, + "loss": 1.1397, + "step": 35529 + }, + { + "epoch": 0.85, + "learning_rate": 1.0906067159294987e-06, + "loss": 1.2821, + "step": 35530 + }, + { + "epoch": 0.85, + "learning_rate": 1.0902530295985947e-06, + "loss": 1.1386, + "step": 35531 + }, + { + "epoch": 0.85, + "learning_rate": 1.0898993973210259e-06, + "loss": 1.0142, + "step": 35532 + }, + { + "epoch": 0.85, + "learning_rate": 1.0895458190989428e-06, + "loss": 1.041, + "step": 35533 + }, + { + "epoch": 0.85, + "learning_rate": 1.0891922949344913e-06, + "loss": 1.0957, + "step": 35534 + }, + { + "epoch": 0.85, + "learning_rate": 1.088838824829812e-06, + "loss": 0.7564, + "step": 35535 + }, + { + "epoch": 0.85, + "learning_rate": 1.0884854087870523e-06, + "loss": 0.9024, + "step": 35536 + }, + { + "epoch": 0.85, + "learning_rate": 1.0881320468083568e-06, + "loss": 1.0235, + "step": 35537 + }, + { + "epoch": 0.85, + "learning_rate": 1.087778738895866e-06, + "loss": 0.9892, + "step": 35538 + }, + { + "epoch": 0.85, + "learning_rate": 1.0874254850517274e-06, + "loss": 0.8782, + "step": 35539 + }, + { + "epoch": 0.85, + "learning_rate": 1.087072285278079e-06, + "loss": 0.9947, + "step": 35540 + }, + { + "epoch": 0.85, + "learning_rate": 1.0867191395770683e-06, + "loss": 0.9533, + "step": 35541 + }, + { + "epoch": 0.85, + "learning_rate": 1.0863660479508375e-06, + "loss": 1.0708, + "step": 35542 + }, + { + "epoch": 0.85, + "learning_rate": 1.0860130104015244e-06, + "loss": 1.1954, + "step": 35543 + }, + { + "epoch": 0.85, + "learning_rate": 1.0856600269312745e-06, + "loss": 0.9823, + "step": 35544 + }, + { + "epoch": 0.85, + "learning_rate": 1.0853070975422296e-06, + "loss": 1.194, + "step": 35545 + }, + { + "epoch": 0.85, + "learning_rate": 1.0849542222365295e-06, + "loss": 1.2621, + "step": 35546 + }, + { + "epoch": 0.85, + "learning_rate": 1.084601401016313e-06, + "loss": 1.0064, + "step": 35547 + }, + { + "epoch": 0.85, + "learning_rate": 1.0842486338837267e-06, + "loss": 1.1868, + "step": 35548 + }, + { + "epoch": 0.85, + "learning_rate": 1.083895920840904e-06, + "loss": 1.1593, + "step": 35549 + }, + { + "epoch": 0.85, + "learning_rate": 1.0835432618899887e-06, + "loss": 1.0761, + "step": 35550 + }, + { + "epoch": 0.85, + "learning_rate": 1.0831906570331185e-06, + "loss": 1.1967, + "step": 35551 + }, + { + "epoch": 0.85, + "learning_rate": 1.0828381062724325e-06, + "loss": 0.7813, + "step": 35552 + }, + { + "epoch": 0.86, + "learning_rate": 1.0824856096100723e-06, + "loss": 1.031, + "step": 35553 + }, + { + "epoch": 0.86, + "learning_rate": 1.0821331670481728e-06, + "loss": 1.1899, + "step": 35554 + }, + { + "epoch": 0.86, + "learning_rate": 1.0817807785888735e-06, + "loss": 1.0458, + "step": 35555 + }, + { + "epoch": 0.86, + "learning_rate": 1.0814284442343136e-06, + "loss": 0.716, + "step": 35556 + }, + { + "epoch": 0.86, + "learning_rate": 1.0810761639866286e-06, + "loss": 0.9449, + "step": 35557 + }, + { + "epoch": 0.86, + "learning_rate": 1.0807239378479583e-06, + "loss": 0.792, + "step": 35558 + }, + { + "epoch": 0.86, + "learning_rate": 1.080371765820436e-06, + "loss": 0.972, + "step": 35559 + }, + { + "epoch": 0.86, + "learning_rate": 1.0800196479062008e-06, + "loss": 1.1102, + "step": 35560 + }, + { + "epoch": 0.86, + "learning_rate": 1.07966758410739e-06, + "loss": 0.7621, + "step": 35561 + }, + { + "epoch": 0.86, + "learning_rate": 1.0793155744261352e-06, + "loss": 0.8586, + "step": 35562 + }, + { + "epoch": 0.86, + "learning_rate": 1.0789636188645747e-06, + "loss": 0.9692, + "step": 35563 + }, + { + "epoch": 0.86, + "learning_rate": 1.0786117174248457e-06, + "loss": 0.8069, + "step": 35564 + }, + { + "epoch": 0.86, + "learning_rate": 1.0782598701090797e-06, + "loss": 1.2271, + "step": 35565 + }, + { + "epoch": 0.86, + "learning_rate": 1.0779080769194127e-06, + "loss": 0.871, + "step": 35566 + }, + { + "epoch": 0.86, + "learning_rate": 1.077556337857979e-06, + "loss": 0.9443, + "step": 35567 + }, + { + "epoch": 0.86, + "learning_rate": 1.0772046529269109e-06, + "loss": 1.0338, + "step": 35568 + }, + { + "epoch": 0.86, + "learning_rate": 1.0768530221283457e-06, + "loss": 0.9542, + "step": 35569 + }, + { + "epoch": 0.86, + "learning_rate": 1.076501445464413e-06, + "loss": 0.9747, + "step": 35570 + }, + { + "epoch": 0.86, + "learning_rate": 1.0761499229372464e-06, + "loss": 1.2531, + "step": 35571 + }, + { + "epoch": 0.86, + "learning_rate": 1.0757984545489807e-06, + "loss": 0.9979, + "step": 35572 + }, + { + "epoch": 0.86, + "learning_rate": 1.0754470403017458e-06, + "loss": 0.8837, + "step": 35573 + }, + { + "epoch": 0.86, + "learning_rate": 1.0750956801976753e-06, + "loss": 1.1207, + "step": 35574 + }, + { + "epoch": 0.86, + "learning_rate": 1.0747443742388996e-06, + "loss": 1.032, + "step": 35575 + }, + { + "epoch": 0.86, + "learning_rate": 1.0743931224275495e-06, + "loss": 1.1471, + "step": 35576 + }, + { + "epoch": 0.86, + "learning_rate": 1.0740419247657585e-06, + "loss": 0.8431, + "step": 35577 + }, + { + "epoch": 0.86, + "learning_rate": 1.0736907812556541e-06, + "loss": 0.9189, + "step": 35578 + }, + { + "epoch": 0.86, + "learning_rate": 1.0733396918993688e-06, + "loss": 0.8492, + "step": 35579 + }, + { + "epoch": 0.86, + "learning_rate": 1.0729886566990333e-06, + "loss": 1.1516, + "step": 35580 + }, + { + "epoch": 0.86, + "learning_rate": 1.0726376756567736e-06, + "loss": 1.0631, + "step": 35581 + }, + { + "epoch": 0.86, + "learning_rate": 1.0722867487747235e-06, + "loss": 0.8928, + "step": 35582 + }, + { + "epoch": 0.86, + "learning_rate": 1.071935876055007e-06, + "loss": 1.0212, + "step": 35583 + }, + { + "epoch": 0.86, + "learning_rate": 1.0715850574997566e-06, + "loss": 1.1916, + "step": 35584 + }, + { + "epoch": 0.86, + "learning_rate": 1.071234293111102e-06, + "loss": 1.0258, + "step": 35585 + }, + { + "epoch": 0.86, + "learning_rate": 1.0708835828911656e-06, + "loss": 1.103, + "step": 35586 + }, + { + "epoch": 0.86, + "learning_rate": 1.0705329268420784e-06, + "loss": 1.1631, + "step": 35587 + }, + { + "epoch": 0.86, + "learning_rate": 1.0701823249659703e-06, + "loss": 1.0088, + "step": 35588 + }, + { + "epoch": 0.86, + "learning_rate": 1.0698317772649624e-06, + "loss": 1.1047, + "step": 35589 + }, + { + "epoch": 0.86, + "learning_rate": 1.069481283741186e-06, + "loss": 1.0666, + "step": 35590 + }, + { + "epoch": 0.86, + "learning_rate": 1.0691308443967673e-06, + "loss": 0.986, + "step": 35591 + }, + { + "epoch": 0.86, + "learning_rate": 1.0687804592338292e-06, + "loss": 1.034, + "step": 35592 + }, + { + "epoch": 0.86, + "learning_rate": 1.0684301282545017e-06, + "loss": 1.0892, + "step": 35593 + }, + { + "epoch": 0.86, + "learning_rate": 1.0680798514609048e-06, + "loss": 1.085, + "step": 35594 + }, + { + "epoch": 0.86, + "learning_rate": 1.0677296288551675e-06, + "loss": 0.8357, + "step": 35595 + }, + { + "epoch": 0.86, + "learning_rate": 1.067379460439415e-06, + "loss": 0.9551, + "step": 35596 + }, + { + "epoch": 0.86, + "learning_rate": 1.0670293462157688e-06, + "loss": 1.0113, + "step": 35597 + }, + { + "epoch": 0.86, + "learning_rate": 1.066679286186354e-06, + "loss": 1.1481, + "step": 35598 + }, + { + "epoch": 0.86, + "learning_rate": 1.0663292803532966e-06, + "loss": 0.896, + "step": 35599 + }, + { + "epoch": 0.86, + "learning_rate": 1.065979328718716e-06, + "loss": 0.8534, + "step": 35600 + }, + { + "epoch": 0.86, + "learning_rate": 1.0656294312847404e-06, + "loss": 1.335, + "step": 35601 + }, + { + "epoch": 0.86, + "learning_rate": 1.0652795880534862e-06, + "loss": 1.0752, + "step": 35602 + }, + { + "epoch": 0.86, + "learning_rate": 1.0649297990270802e-06, + "loss": 0.744, + "step": 35603 + }, + { + "epoch": 0.86, + "learning_rate": 1.0645800642076453e-06, + "loss": 1.0779, + "step": 35604 + }, + { + "epoch": 0.86, + "learning_rate": 1.064230383597299e-06, + "loss": 0.9496, + "step": 35605 + }, + { + "epoch": 0.86, + "learning_rate": 1.0638807571981647e-06, + "loss": 0.9772, + "step": 35606 + }, + { + "epoch": 0.86, + "learning_rate": 1.0635311850123675e-06, + "loss": 1.0648, + "step": 35607 + }, + { + "epoch": 0.86, + "learning_rate": 1.0631816670420214e-06, + "loss": 1.2565, + "step": 35608 + }, + { + "epoch": 0.86, + "learning_rate": 1.0628322032892525e-06, + "loss": 0.9726, + "step": 35609 + }, + { + "epoch": 0.86, + "learning_rate": 1.0624827937561755e-06, + "loss": 1.0871, + "step": 35610 + }, + { + "epoch": 0.86, + "learning_rate": 1.0621334384449144e-06, + "loss": 1.0161, + "step": 35611 + }, + { + "epoch": 0.86, + "learning_rate": 1.0617841373575888e-06, + "loss": 0.9651, + "step": 35612 + }, + { + "epoch": 0.86, + "learning_rate": 1.0614348904963146e-06, + "loss": 0.9479, + "step": 35613 + }, + { + "epoch": 0.86, + "learning_rate": 1.0610856978632123e-06, + "loss": 0.9941, + "step": 35614 + }, + { + "epoch": 0.86, + "learning_rate": 1.0607365594604024e-06, + "loss": 1.2139, + "step": 35615 + }, + { + "epoch": 0.86, + "learning_rate": 1.06038747529e-06, + "loss": 1.1536, + "step": 35616 + }, + { + "epoch": 0.86, + "learning_rate": 1.0600384453541257e-06, + "loss": 1.1474, + "step": 35617 + }, + { + "epoch": 0.86, + "learning_rate": 1.059689469654893e-06, + "loss": 0.9529, + "step": 35618 + }, + { + "epoch": 0.86, + "learning_rate": 1.0593405481944208e-06, + "loss": 1.0182, + "step": 35619 + }, + { + "epoch": 0.86, + "learning_rate": 1.0589916809748301e-06, + "loss": 1.0554, + "step": 35620 + }, + { + "epoch": 0.86, + "learning_rate": 1.0586428679982307e-06, + "loss": 1.0352, + "step": 35621 + }, + { + "epoch": 0.86, + "learning_rate": 1.0582941092667421e-06, + "loss": 1.1334, + "step": 35622 + }, + { + "epoch": 0.86, + "learning_rate": 1.0579454047824823e-06, + "loss": 1.1458, + "step": 35623 + }, + { + "epoch": 0.86, + "learning_rate": 1.0575967545475618e-06, + "loss": 0.8899, + "step": 35624 + }, + { + "epoch": 0.86, + "learning_rate": 1.0572481585641004e-06, + "loss": 1.3586, + "step": 35625 + }, + { + "epoch": 0.86, + "learning_rate": 1.0568996168342094e-06, + "loss": 1.1873, + "step": 35626 + }, + { + "epoch": 0.86, + "learning_rate": 1.0565511293600051e-06, + "loss": 0.789, + "step": 35627 + }, + { + "epoch": 0.86, + "learning_rate": 1.0562026961436023e-06, + "loss": 0.7925, + "step": 35628 + }, + { + "epoch": 0.86, + "learning_rate": 1.055854317187114e-06, + "loss": 1.2184, + "step": 35629 + }, + { + "epoch": 0.86, + "learning_rate": 1.0555059924926513e-06, + "loss": 1.0337, + "step": 35630 + }, + { + "epoch": 0.86, + "learning_rate": 1.0551577220623345e-06, + "loss": 0.8643, + "step": 35631 + }, + { + "epoch": 0.86, + "learning_rate": 1.0548095058982678e-06, + "loss": 1.1, + "step": 35632 + }, + { + "epoch": 0.86, + "learning_rate": 1.0544613440025707e-06, + "loss": 1.1057, + "step": 35633 + }, + { + "epoch": 0.86, + "learning_rate": 1.05411323637735e-06, + "loss": 0.727, + "step": 35634 + }, + { + "epoch": 0.86, + "learning_rate": 1.0537651830247209e-06, + "loss": 1.0476, + "step": 35635 + }, + { + "epoch": 0.86, + "learning_rate": 1.0534171839467956e-06, + "loss": 0.9665, + "step": 35636 + }, + { + "epoch": 0.86, + "learning_rate": 1.053069239145682e-06, + "loss": 0.9247, + "step": 35637 + }, + { + "epoch": 0.86, + "learning_rate": 1.052721348623492e-06, + "loss": 1.2148, + "step": 35638 + }, + { + "epoch": 0.86, + "learning_rate": 1.0523735123823409e-06, + "loss": 1.0659, + "step": 35639 + }, + { + "epoch": 0.86, + "learning_rate": 1.052025730424332e-06, + "loss": 1.1809, + "step": 35640 + }, + { + "epoch": 0.86, + "learning_rate": 1.0516780027515783e-06, + "loss": 1.0493, + "step": 35641 + }, + { + "epoch": 0.86, + "learning_rate": 1.0513303293661914e-06, + "loss": 1.0975, + "step": 35642 + }, + { + "epoch": 0.86, + "learning_rate": 1.0509827102702763e-06, + "loss": 1.0635, + "step": 35643 + }, + { + "epoch": 0.86, + "learning_rate": 1.050635145465946e-06, + "loss": 0.9764, + "step": 35644 + }, + { + "epoch": 0.86, + "learning_rate": 1.050287634955306e-06, + "loss": 0.999, + "step": 35645 + }, + { + "epoch": 0.86, + "learning_rate": 1.0499401787404651e-06, + "loss": 1.0124, + "step": 35646 + }, + { + "epoch": 0.86, + "learning_rate": 1.0495927768235347e-06, + "loss": 0.9555, + "step": 35647 + }, + { + "epoch": 0.86, + "learning_rate": 1.0492454292066178e-06, + "loss": 1.027, + "step": 35648 + }, + { + "epoch": 0.86, + "learning_rate": 1.0488981358918227e-06, + "loss": 0.9707, + "step": 35649 + }, + { + "epoch": 0.86, + "learning_rate": 1.0485508968812596e-06, + "loss": 1.0909, + "step": 35650 + }, + { + "epoch": 0.86, + "learning_rate": 1.0482037121770306e-06, + "loss": 1.0191, + "step": 35651 + }, + { + "epoch": 0.86, + "learning_rate": 1.047856581781246e-06, + "loss": 0.8678, + "step": 35652 + }, + { + "epoch": 0.86, + "learning_rate": 1.0475095056960082e-06, + "loss": 0.928, + "step": 35653 + }, + { + "epoch": 0.86, + "learning_rate": 1.047162483923424e-06, + "loss": 1.2505, + "step": 35654 + }, + { + "epoch": 0.86, + "learning_rate": 1.0468155164656013e-06, + "loss": 1.1929, + "step": 35655 + }, + { + "epoch": 0.86, + "learning_rate": 1.0464686033246419e-06, + "loss": 1.0065, + "step": 35656 + }, + { + "epoch": 0.86, + "learning_rate": 1.0461217445026507e-06, + "loss": 0.7854, + "step": 35657 + }, + { + "epoch": 0.86, + "learning_rate": 1.0457749400017336e-06, + "loss": 0.9928, + "step": 35658 + }, + { + "epoch": 0.86, + "learning_rate": 1.0454281898239937e-06, + "loss": 1.265, + "step": 35659 + }, + { + "epoch": 0.86, + "learning_rate": 1.0450814939715358e-06, + "loss": 1.0314, + "step": 35660 + }, + { + "epoch": 0.86, + "learning_rate": 1.0447348524464606e-06, + "loss": 1.0154, + "step": 35661 + }, + { + "epoch": 0.86, + "learning_rate": 1.0443882652508719e-06, + "loss": 1.0113, + "step": 35662 + }, + { + "epoch": 0.86, + "learning_rate": 1.0440417323868756e-06, + "loss": 0.8614, + "step": 35663 + }, + { + "epoch": 0.86, + "learning_rate": 1.0436952538565692e-06, + "loss": 0.7921, + "step": 35664 + }, + { + "epoch": 0.86, + "learning_rate": 1.0433488296620575e-06, + "loss": 0.9699, + "step": 35665 + }, + { + "epoch": 0.86, + "learning_rate": 1.0430024598054412e-06, + "loss": 1.19, + "step": 35666 + }, + { + "epoch": 0.86, + "learning_rate": 1.0426561442888227e-06, + "loss": 1.0345, + "step": 35667 + }, + { + "epoch": 0.86, + "learning_rate": 1.042309883114304e-06, + "loss": 0.9008, + "step": 35668 + }, + { + "epoch": 0.86, + "learning_rate": 1.0419636762839836e-06, + "loss": 1.2275, + "step": 35669 + }, + { + "epoch": 0.86, + "learning_rate": 1.0416175237999615e-06, + "loss": 1.1501, + "step": 35670 + }, + { + "epoch": 0.86, + "learning_rate": 1.0412714256643397e-06, + "loss": 0.8085, + "step": 35671 + }, + { + "epoch": 0.86, + "learning_rate": 1.0409253818792165e-06, + "loss": 1.1542, + "step": 35672 + }, + { + "epoch": 0.86, + "learning_rate": 1.0405793924466934e-06, + "loss": 0.9412, + "step": 35673 + }, + { + "epoch": 0.86, + "learning_rate": 1.0402334573688654e-06, + "loss": 0.8605, + "step": 35674 + }, + { + "epoch": 0.86, + "learning_rate": 1.0398875766478333e-06, + "loss": 1.2685, + "step": 35675 + }, + { + "epoch": 0.86, + "learning_rate": 1.0395417502856985e-06, + "loss": 0.6899, + "step": 35676 + }, + { + "epoch": 0.86, + "learning_rate": 1.039195978284554e-06, + "loss": 0.7113, + "step": 35677 + }, + { + "epoch": 0.86, + "learning_rate": 1.038850260646499e-06, + "loss": 1.1437, + "step": 35678 + }, + { + "epoch": 0.86, + "learning_rate": 1.038504597373634e-06, + "loss": 0.7878, + "step": 35679 + }, + { + "epoch": 0.86, + "learning_rate": 1.0381589884680522e-06, + "loss": 1.1107, + "step": 35680 + }, + { + "epoch": 0.86, + "learning_rate": 1.0378134339318535e-06, + "loss": 0.9643, + "step": 35681 + }, + { + "epoch": 0.86, + "learning_rate": 1.0374679337671311e-06, + "loss": 1.1554, + "step": 35682 + }, + { + "epoch": 0.86, + "learning_rate": 1.0371224879759822e-06, + "loss": 1.0393, + "step": 35683 + }, + { + "epoch": 0.86, + "learning_rate": 1.0367770965605062e-06, + "loss": 1.0757, + "step": 35684 + }, + { + "epoch": 0.86, + "learning_rate": 1.0364317595227913e-06, + "loss": 1.0157, + "step": 35685 + }, + { + "epoch": 0.86, + "learning_rate": 1.0360864768649381e-06, + "loss": 1.1215, + "step": 35686 + }, + { + "epoch": 0.86, + "learning_rate": 1.0357412485890405e-06, + "loss": 1.0048, + "step": 35687 + }, + { + "epoch": 0.86, + "learning_rate": 1.0353960746971914e-06, + "loss": 0.892, + "step": 35688 + }, + { + "epoch": 0.86, + "learning_rate": 1.0350509551914878e-06, + "loss": 1.2939, + "step": 35689 + }, + { + "epoch": 0.86, + "learning_rate": 1.0347058900740182e-06, + "loss": 1.0723, + "step": 35690 + }, + { + "epoch": 0.86, + "learning_rate": 1.0343608793468795e-06, + "loss": 1.1909, + "step": 35691 + }, + { + "epoch": 0.86, + "learning_rate": 1.0340159230121672e-06, + "loss": 0.8959, + "step": 35692 + }, + { + "epoch": 0.86, + "learning_rate": 1.0336710210719692e-06, + "loss": 0.8509, + "step": 35693 + }, + { + "epoch": 0.86, + "learning_rate": 1.0333261735283795e-06, + "loss": 0.9461, + "step": 35694 + }, + { + "epoch": 0.86, + "learning_rate": 1.0329813803834932e-06, + "loss": 1.195, + "step": 35695 + }, + { + "epoch": 0.86, + "learning_rate": 1.0326366416393973e-06, + "loss": 1.1636, + "step": 35696 + }, + { + "epoch": 0.86, + "learning_rate": 1.0322919572981882e-06, + "loss": 1.0907, + "step": 35697 + }, + { + "epoch": 0.86, + "learning_rate": 1.031947327361953e-06, + "loss": 1.1684, + "step": 35698 + }, + { + "epoch": 0.86, + "learning_rate": 1.0316027518327832e-06, + "loss": 1.0177, + "step": 35699 + }, + { + "epoch": 0.86, + "learning_rate": 1.0312582307127717e-06, + "loss": 0.9913, + "step": 35700 + }, + { + "epoch": 0.86, + "learning_rate": 1.0309137640040057e-06, + "loss": 0.9774, + "step": 35701 + }, + { + "epoch": 0.86, + "learning_rate": 1.0305693517085757e-06, + "loss": 0.9005, + "step": 35702 + }, + { + "epoch": 0.86, + "learning_rate": 1.0302249938285736e-06, + "loss": 1.0779, + "step": 35703 + }, + { + "epoch": 0.86, + "learning_rate": 1.0298806903660852e-06, + "loss": 0.9169, + "step": 35704 + }, + { + "epoch": 0.86, + "learning_rate": 1.0295364413232024e-06, + "loss": 0.8754, + "step": 35705 + }, + { + "epoch": 0.86, + "learning_rate": 1.02919224670201e-06, + "loss": 0.9321, + "step": 35706 + }, + { + "epoch": 0.86, + "learning_rate": 1.0288481065045974e-06, + "loss": 1.1098, + "step": 35707 + }, + { + "epoch": 0.86, + "learning_rate": 1.0285040207330566e-06, + "loss": 1.1022, + "step": 35708 + }, + { + "epoch": 0.86, + "learning_rate": 1.028159989389469e-06, + "loss": 0.8207, + "step": 35709 + }, + { + "epoch": 0.86, + "learning_rate": 1.0278160124759251e-06, + "loss": 1.0997, + "step": 35710 + }, + { + "epoch": 0.86, + "learning_rate": 1.0274720899945124e-06, + "loss": 1.0069, + "step": 35711 + }, + { + "epoch": 0.86, + "learning_rate": 1.0271282219473132e-06, + "loss": 0.9101, + "step": 35712 + }, + { + "epoch": 0.86, + "learning_rate": 1.0267844083364197e-06, + "loss": 1.0744, + "step": 35713 + }, + { + "epoch": 0.86, + "learning_rate": 1.0264406491639123e-06, + "loss": 1.1339, + "step": 35714 + }, + { + "epoch": 0.86, + "learning_rate": 1.0260969444318792e-06, + "loss": 0.8388, + "step": 35715 + }, + { + "epoch": 0.86, + "learning_rate": 1.0257532941424065e-06, + "loss": 1.0195, + "step": 35716 + }, + { + "epoch": 0.86, + "learning_rate": 1.0254096982975758e-06, + "loss": 1.1553, + "step": 35717 + }, + { + "epoch": 0.86, + "learning_rate": 1.0250661568994746e-06, + "loss": 0.6944, + "step": 35718 + }, + { + "epoch": 0.86, + "learning_rate": 1.0247226699501866e-06, + "loss": 0.9351, + "step": 35719 + }, + { + "epoch": 0.86, + "learning_rate": 1.0243792374517937e-06, + "loss": 1.0295, + "step": 35720 + }, + { + "epoch": 0.86, + "learning_rate": 1.0240358594063804e-06, + "loss": 1.0438, + "step": 35721 + }, + { + "epoch": 0.86, + "learning_rate": 1.0236925358160321e-06, + "loss": 0.9996, + "step": 35722 + }, + { + "epoch": 0.86, + "learning_rate": 1.0233492666828282e-06, + "loss": 1.0736, + "step": 35723 + }, + { + "epoch": 0.86, + "learning_rate": 1.0230060520088548e-06, + "loss": 1.0763, + "step": 35724 + }, + { + "epoch": 0.86, + "learning_rate": 1.0226628917961901e-06, + "loss": 0.9695, + "step": 35725 + }, + { + "epoch": 0.86, + "learning_rate": 1.0223197860469192e-06, + "loss": 0.8625, + "step": 35726 + }, + { + "epoch": 0.86, + "learning_rate": 1.0219767347631227e-06, + "loss": 0.9706, + "step": 35727 + }, + { + "epoch": 0.86, + "learning_rate": 1.02163373794688e-06, + "loss": 1.1076, + "step": 35728 + }, + { + "epoch": 0.86, + "learning_rate": 1.021290795600275e-06, + "loss": 1.1927, + "step": 35729 + }, + { + "epoch": 0.86, + "learning_rate": 1.020947907725387e-06, + "loss": 1.1241, + "step": 35730 + }, + { + "epoch": 0.86, + "learning_rate": 1.0206050743242946e-06, + "loss": 0.9472, + "step": 35731 + }, + { + "epoch": 0.86, + "learning_rate": 1.0202622953990816e-06, + "loss": 1.0229, + "step": 35732 + }, + { + "epoch": 0.86, + "learning_rate": 1.0199195709518228e-06, + "loss": 1.0412, + "step": 35733 + }, + { + "epoch": 0.86, + "learning_rate": 1.0195769009845992e-06, + "loss": 1.0924, + "step": 35734 + }, + { + "epoch": 0.86, + "learning_rate": 1.019234285499493e-06, + "loss": 1.0571, + "step": 35735 + }, + { + "epoch": 0.86, + "learning_rate": 1.0188917244985775e-06, + "loss": 1.0575, + "step": 35736 + }, + { + "epoch": 0.86, + "learning_rate": 1.018549217983934e-06, + "loss": 0.8505, + "step": 35737 + }, + { + "epoch": 0.86, + "learning_rate": 1.0182067659576412e-06, + "loss": 0.974, + "step": 35738 + }, + { + "epoch": 0.86, + "learning_rate": 1.0178643684217737e-06, + "loss": 0.9863, + "step": 35739 + }, + { + "epoch": 0.86, + "learning_rate": 1.0175220253784123e-06, + "loss": 1.0101, + "step": 35740 + }, + { + "epoch": 0.86, + "learning_rate": 1.01717973682963e-06, + "loss": 0.9847, + "step": 35741 + }, + { + "epoch": 0.86, + "learning_rate": 1.0168375027775045e-06, + "loss": 1.1801, + "step": 35742 + }, + { + "epoch": 0.86, + "learning_rate": 1.016495323224116e-06, + "loss": 1.0126, + "step": 35743 + }, + { + "epoch": 0.86, + "learning_rate": 1.0161531981715345e-06, + "loss": 1.0774, + "step": 35744 + }, + { + "epoch": 0.86, + "learning_rate": 1.0158111276218396e-06, + "loss": 1.3388, + "step": 35745 + }, + { + "epoch": 0.86, + "learning_rate": 1.0154691115771066e-06, + "loss": 1.0293, + "step": 35746 + }, + { + "epoch": 0.86, + "learning_rate": 1.0151271500394066e-06, + "loss": 0.9869, + "step": 35747 + }, + { + "epoch": 0.86, + "learning_rate": 1.0147852430108195e-06, + "loss": 1.2537, + "step": 35748 + }, + { + "epoch": 0.86, + "learning_rate": 1.0144433904934147e-06, + "loss": 0.8588, + "step": 35749 + }, + { + "epoch": 0.86, + "learning_rate": 1.0141015924892693e-06, + "loss": 1.3064, + "step": 35750 + }, + { + "epoch": 0.86, + "learning_rate": 1.0137598490004564e-06, + "loss": 0.9988, + "step": 35751 + }, + { + "epoch": 0.86, + "learning_rate": 1.0134181600290483e-06, + "loss": 1.0296, + "step": 35752 + }, + { + "epoch": 0.86, + "learning_rate": 1.013076525577117e-06, + "loss": 1.046, + "step": 35753 + }, + { + "epoch": 0.86, + "learning_rate": 1.0127349456467394e-06, + "loss": 1.0941, + "step": 35754 + }, + { + "epoch": 0.86, + "learning_rate": 1.0123934202399821e-06, + "loss": 1.1126, + "step": 35755 + }, + { + "epoch": 0.86, + "learning_rate": 1.0120519493589231e-06, + "loss": 1.1511, + "step": 35756 + }, + { + "epoch": 0.86, + "learning_rate": 1.0117105330056287e-06, + "loss": 0.821, + "step": 35757 + }, + { + "epoch": 0.86, + "learning_rate": 1.0113691711821716e-06, + "loss": 0.9962, + "step": 35758 + }, + { + "epoch": 0.86, + "learning_rate": 1.0110278638906256e-06, + "loss": 1.2628, + "step": 35759 + }, + { + "epoch": 0.86, + "learning_rate": 1.0106866111330572e-06, + "loss": 1.0693, + "step": 35760 + }, + { + "epoch": 0.86, + "learning_rate": 1.0103454129115386e-06, + "loss": 1.0363, + "step": 35761 + }, + { + "epoch": 0.86, + "learning_rate": 1.010004269228143e-06, + "loss": 0.629, + "step": 35762 + }, + { + "epoch": 0.86, + "learning_rate": 1.0096631800849344e-06, + "loss": 1.1364, + "step": 35763 + }, + { + "epoch": 0.86, + "learning_rate": 1.009322145483984e-06, + "loss": 1.1692, + "step": 35764 + }, + { + "epoch": 0.86, + "learning_rate": 1.008981165427364e-06, + "loss": 1.0911, + "step": 35765 + }, + { + "epoch": 0.86, + "learning_rate": 1.008640239917138e-06, + "loss": 0.7702, + "step": 35766 + }, + { + "epoch": 0.86, + "learning_rate": 1.0082993689553788e-06, + "loss": 1.167, + "step": 35767 + }, + { + "epoch": 0.86, + "learning_rate": 1.0079585525441504e-06, + "loss": 0.8541, + "step": 35768 + }, + { + "epoch": 0.86, + "learning_rate": 1.0076177906855222e-06, + "loss": 1.3688, + "step": 35769 + }, + { + "epoch": 0.86, + "learning_rate": 1.007277083381565e-06, + "loss": 0.8726, + "step": 35770 + }, + { + "epoch": 0.86, + "learning_rate": 1.0069364306343388e-06, + "loss": 0.9925, + "step": 35771 + }, + { + "epoch": 0.86, + "learning_rate": 1.0065958324459147e-06, + "loss": 0.8785, + "step": 35772 + }, + { + "epoch": 0.86, + "learning_rate": 1.0062552888183607e-06, + "loss": 0.9107, + "step": 35773 + }, + { + "epoch": 0.86, + "learning_rate": 1.0059147997537378e-06, + "loss": 1.061, + "step": 35774 + }, + { + "epoch": 0.86, + "learning_rate": 1.0055743652541173e-06, + "loss": 0.9551, + "step": 35775 + }, + { + "epoch": 0.86, + "learning_rate": 1.0052339853215588e-06, + "loss": 1.0162, + "step": 35776 + }, + { + "epoch": 0.86, + "learning_rate": 1.0048936599581295e-06, + "loss": 1.0408, + "step": 35777 + }, + { + "epoch": 0.86, + "learning_rate": 1.0045533891658966e-06, + "loss": 0.999, + "step": 35778 + }, + { + "epoch": 0.86, + "learning_rate": 1.0042131729469208e-06, + "loss": 0.9227, + "step": 35779 + }, + { + "epoch": 0.86, + "learning_rate": 1.003873011303268e-06, + "loss": 1.0135, + "step": 35780 + }, + { + "epoch": 0.86, + "learning_rate": 1.0035329042370034e-06, + "loss": 1.1674, + "step": 35781 + }, + { + "epoch": 0.86, + "learning_rate": 1.0031928517501866e-06, + "loss": 1.2776, + "step": 35782 + }, + { + "epoch": 0.86, + "learning_rate": 1.0028528538448846e-06, + "loss": 0.9738, + "step": 35783 + }, + { + "epoch": 0.86, + "learning_rate": 1.0025129105231557e-06, + "loss": 1.2701, + "step": 35784 + }, + { + "epoch": 0.86, + "learning_rate": 1.0021730217870651e-06, + "loss": 1.0233, + "step": 35785 + }, + { + "epoch": 0.86, + "learning_rate": 1.0018331876386767e-06, + "loss": 1.203, + "step": 35786 + }, + { + "epoch": 0.86, + "learning_rate": 1.0014934080800477e-06, + "loss": 1.1001, + "step": 35787 + }, + { + "epoch": 0.86, + "learning_rate": 1.0011536831132407e-06, + "loss": 1.0718, + "step": 35788 + }, + { + "epoch": 0.86, + "learning_rate": 1.000814012740321e-06, + "loss": 1.0405, + "step": 35789 + }, + { + "epoch": 0.86, + "learning_rate": 1.0004743969633423e-06, + "loss": 0.7854, + "step": 35790 + }, + { + "epoch": 0.86, + "learning_rate": 1.000134835784372e-06, + "loss": 1.0836, + "step": 35791 + }, + { + "epoch": 0.86, + "learning_rate": 9.99795329205464e-07, + "loss": 0.9504, + "step": 35792 + }, + { + "epoch": 0.86, + "learning_rate": 9.994558772286812e-07, + "loss": 1.3111, + "step": 35793 + }, + { + "epoch": 0.86, + "learning_rate": 9.991164798560848e-07, + "loss": 1.1231, + "step": 35794 + }, + { + "epoch": 0.86, + "learning_rate": 9.987771370897292e-07, + "loss": 1.2211, + "step": 35795 + }, + { + "epoch": 0.86, + "learning_rate": 9.984378489316747e-07, + "loss": 0.8658, + "step": 35796 + }, + { + "epoch": 0.86, + "learning_rate": 9.980986153839833e-07, + "loss": 0.8016, + "step": 35797 + }, + { + "epoch": 0.86, + "learning_rate": 9.977594364487087e-07, + "loss": 0.931, + "step": 35798 + }, + { + "epoch": 0.86, + "learning_rate": 9.974203121279113e-07, + "loss": 1.0377, + "step": 35799 + }, + { + "epoch": 0.86, + "learning_rate": 9.970812424236464e-07, + "loss": 0.8642, + "step": 35800 + }, + { + "epoch": 0.86, + "learning_rate": 9.967422273379711e-07, + "loss": 0.9014, + "step": 35801 + }, + { + "epoch": 0.86, + "learning_rate": 9.96403266872945e-07, + "loss": 1.0893, + "step": 35802 + }, + { + "epoch": 0.86, + "learning_rate": 9.96064361030621e-07, + "loss": 1.0877, + "step": 35803 + }, + { + "epoch": 0.86, + "learning_rate": 9.957255098130559e-07, + "loss": 0.8212, + "step": 35804 + }, + { + "epoch": 0.86, + "learning_rate": 9.953867132223083e-07, + "loss": 0.9928, + "step": 35805 + }, + { + "epoch": 0.86, + "learning_rate": 9.950479712604299e-07, + "loss": 1.0688, + "step": 35806 + }, + { + "epoch": 0.86, + "learning_rate": 9.947092839294769e-07, + "loss": 1.278, + "step": 35807 + }, + { + "epoch": 0.86, + "learning_rate": 9.943706512315078e-07, + "loss": 1.1058, + "step": 35808 + }, + { + "epoch": 0.86, + "learning_rate": 9.940320731685705e-07, + "loss": 0.8831, + "step": 35809 + }, + { + "epoch": 0.86, + "learning_rate": 9.936935497427248e-07, + "loss": 0.844, + "step": 35810 + }, + { + "epoch": 0.86, + "learning_rate": 9.933550809560188e-07, + "loss": 1.041, + "step": 35811 + }, + { + "epoch": 0.86, + "learning_rate": 9.930166668105101e-07, + "loss": 0.9004, + "step": 35812 + }, + { + "epoch": 0.86, + "learning_rate": 9.926783073082525e-07, + "loss": 1.1292, + "step": 35813 + }, + { + "epoch": 0.86, + "learning_rate": 9.923400024512952e-07, + "loss": 0.8917, + "step": 35814 + }, + { + "epoch": 0.86, + "learning_rate": 9.920017522416926e-07, + "loss": 0.9172, + "step": 35815 + }, + { + "epoch": 0.86, + "learning_rate": 9.91663556681498e-07, + "loss": 1.2345, + "step": 35816 + }, + { + "epoch": 0.86, + "learning_rate": 9.913254157727593e-07, + "loss": 1.0735, + "step": 35817 + }, + { + "epoch": 0.86, + "learning_rate": 9.909873295175331e-07, + "loss": 0.8714, + "step": 35818 + }, + { + "epoch": 0.86, + "learning_rate": 9.906492979178651e-07, + "loss": 0.9242, + "step": 35819 + }, + { + "epoch": 0.86, + "learning_rate": 9.903113209758098e-07, + "loss": 1.0095, + "step": 35820 + }, + { + "epoch": 0.86, + "learning_rate": 9.899733986934169e-07, + "loss": 0.9525, + "step": 35821 + }, + { + "epoch": 0.86, + "learning_rate": 9.896355310727356e-07, + "loss": 1.2433, + "step": 35822 + }, + { + "epoch": 0.86, + "learning_rate": 9.892977181158148e-07, + "loss": 1.2799, + "step": 35823 + }, + { + "epoch": 0.86, + "learning_rate": 9.88959959824708e-07, + "loss": 1.0615, + "step": 35824 + }, + { + "epoch": 0.86, + "learning_rate": 9.886222562014602e-07, + "loss": 0.8632, + "step": 35825 + }, + { + "epoch": 0.86, + "learning_rate": 9.882846072481223e-07, + "loss": 0.8627, + "step": 35826 + }, + { + "epoch": 0.86, + "learning_rate": 9.879470129667413e-07, + "loss": 0.8886, + "step": 35827 + }, + { + "epoch": 0.86, + "learning_rate": 9.876094733593656e-07, + "loss": 0.9052, + "step": 35828 + }, + { + "epoch": 0.86, + "learning_rate": 9.872719884280446e-07, + "loss": 1.1553, + "step": 35829 + }, + { + "epoch": 0.86, + "learning_rate": 9.869345581748236e-07, + "loss": 1.1559, + "step": 35830 + }, + { + "epoch": 0.86, + "learning_rate": 9.865971826017495e-07, + "loss": 1.0769, + "step": 35831 + }, + { + "epoch": 0.86, + "learning_rate": 9.862598617108709e-07, + "loss": 1.048, + "step": 35832 + }, + { + "epoch": 0.86, + "learning_rate": 9.85922595504234e-07, + "loss": 0.8871, + "step": 35833 + }, + { + "epoch": 0.86, + "learning_rate": 9.85585383983886e-07, + "loss": 0.8116, + "step": 35834 + }, + { + "epoch": 0.86, + "learning_rate": 9.852482271518692e-07, + "loss": 0.8836, + "step": 35835 + }, + { + "epoch": 0.86, + "learning_rate": 9.849111250102327e-07, + "loss": 1.0755, + "step": 35836 + }, + { + "epoch": 0.86, + "learning_rate": 9.84574077561018e-07, + "loss": 1.2992, + "step": 35837 + }, + { + "epoch": 0.86, + "learning_rate": 9.842370848062721e-07, + "loss": 0.9554, + "step": 35838 + }, + { + "epoch": 0.86, + "learning_rate": 9.839001467480414e-07, + "loss": 0.8203, + "step": 35839 + }, + { + "epoch": 0.86, + "learning_rate": 9.83563263388365e-07, + "loss": 0.8317, + "step": 35840 + }, + { + "epoch": 0.86, + "learning_rate": 9.832264347292898e-07, + "loss": 0.9813, + "step": 35841 + }, + { + "epoch": 0.86, + "learning_rate": 9.828896607728612e-07, + "loss": 0.9485, + "step": 35842 + }, + { + "epoch": 0.86, + "learning_rate": 9.825529415211177e-07, + "loss": 1.0186, + "step": 35843 + }, + { + "epoch": 0.86, + "learning_rate": 9.822162769761056e-07, + "loss": 1.2302, + "step": 35844 + }, + { + "epoch": 0.86, + "learning_rate": 9.818796671398666e-07, + "loss": 0.9504, + "step": 35845 + }, + { + "epoch": 0.86, + "learning_rate": 9.815431120144414e-07, + "loss": 0.8373, + "step": 35846 + }, + { + "epoch": 0.86, + "learning_rate": 9.81206611601875e-07, + "loss": 1.0372, + "step": 35847 + }, + { + "epoch": 0.86, + "learning_rate": 9.808701659042042e-07, + "loss": 1.1077, + "step": 35848 + }, + { + "epoch": 0.86, + "learning_rate": 9.805337749234733e-07, + "loss": 1.0664, + "step": 35849 + }, + { + "epoch": 0.86, + "learning_rate": 9.801974386617241e-07, + "loss": 0.9754, + "step": 35850 + }, + { + "epoch": 0.86, + "learning_rate": 9.798611571209927e-07, + "loss": 0.9383, + "step": 35851 + }, + { + "epoch": 0.86, + "learning_rate": 9.795249303033228e-07, + "loss": 0.9363, + "step": 35852 + }, + { + "epoch": 0.86, + "learning_rate": 9.79188758210755e-07, + "loss": 0.9532, + "step": 35853 + }, + { + "epoch": 0.86, + "learning_rate": 9.788526408453258e-07, + "loss": 1.0454, + "step": 35854 + }, + { + "epoch": 0.86, + "learning_rate": 9.785165782090778e-07, + "loss": 1.0617, + "step": 35855 + }, + { + "epoch": 0.86, + "learning_rate": 9.78180570304046e-07, + "loss": 0.8789, + "step": 35856 + }, + { + "epoch": 0.86, + "learning_rate": 9.778446171322699e-07, + "loss": 1.1366, + "step": 35857 + }, + { + "epoch": 0.86, + "learning_rate": 9.7750871869579e-07, + "loss": 1.278, + "step": 35858 + }, + { + "epoch": 0.86, + "learning_rate": 9.771728749966414e-07, + "loss": 1.0951, + "step": 35859 + }, + { + "epoch": 0.86, + "learning_rate": 9.768370860368614e-07, + "loss": 0.9955, + "step": 35860 + }, + { + "epoch": 0.86, + "learning_rate": 9.765013518184918e-07, + "loss": 0.9212, + "step": 35861 + }, + { + "epoch": 0.86, + "learning_rate": 9.76165672343563e-07, + "loss": 0.9977, + "step": 35862 + }, + { + "epoch": 0.86, + "learning_rate": 9.75830047614117e-07, + "loss": 0.9977, + "step": 35863 + }, + { + "epoch": 0.86, + "learning_rate": 9.754944776321862e-07, + "loss": 0.9416, + "step": 35864 + }, + { + "epoch": 0.86, + "learning_rate": 9.75158962399807e-07, + "loss": 1.1014, + "step": 35865 + }, + { + "epoch": 0.86, + "learning_rate": 9.748235019190165e-07, + "loss": 1.2089, + "step": 35866 + }, + { + "epoch": 0.86, + "learning_rate": 9.744880961918491e-07, + "loss": 0.9873, + "step": 35867 + }, + { + "epoch": 0.86, + "learning_rate": 9.741527452203382e-07, + "loss": 0.9592, + "step": 35868 + }, + { + "epoch": 0.86, + "learning_rate": 9.73817449006521e-07, + "loss": 1.0172, + "step": 35869 + }, + { + "epoch": 0.86, + "learning_rate": 9.734822075524298e-07, + "loss": 0.8268, + "step": 35870 + }, + { + "epoch": 0.86, + "learning_rate": 9.731470208601e-07, + "loss": 1.1277, + "step": 35871 + }, + { + "epoch": 0.86, + "learning_rate": 9.72811888931562e-07, + "loss": 1.1995, + "step": 35872 + }, + { + "epoch": 0.86, + "learning_rate": 9.72476811768851e-07, + "loss": 1.0775, + "step": 35873 + }, + { + "epoch": 0.86, + "learning_rate": 9.721417893740026e-07, + "loss": 1.0681, + "step": 35874 + }, + { + "epoch": 0.86, + "learning_rate": 9.71806821749044e-07, + "loss": 1.1889, + "step": 35875 + }, + { + "epoch": 0.86, + "learning_rate": 9.71471908896009e-07, + "loss": 0.9179, + "step": 35876 + }, + { + "epoch": 0.86, + "learning_rate": 9.711370508169326e-07, + "loss": 1.0945, + "step": 35877 + }, + { + "epoch": 0.86, + "learning_rate": 9.708022475138422e-07, + "loss": 1.0975, + "step": 35878 + }, + { + "epoch": 0.86, + "learning_rate": 9.70467498988773e-07, + "loss": 0.8451, + "step": 35879 + }, + { + "epoch": 0.86, + "learning_rate": 9.70132805243752e-07, + "loss": 0.6685, + "step": 35880 + }, + { + "epoch": 0.86, + "learning_rate": 9.697981662808108e-07, + "loss": 1.1767, + "step": 35881 + }, + { + "epoch": 0.86, + "learning_rate": 9.694635821019815e-07, + "loss": 0.9852, + "step": 35882 + }, + { + "epoch": 0.86, + "learning_rate": 9.69129052709291e-07, + "loss": 1.1848, + "step": 35883 + }, + { + "epoch": 0.86, + "learning_rate": 9.687945781047714e-07, + "loss": 0.7655, + "step": 35884 + }, + { + "epoch": 0.86, + "learning_rate": 9.684601582904519e-07, + "loss": 1.1393, + "step": 35885 + }, + { + "epoch": 0.86, + "learning_rate": 9.681257932683574e-07, + "loss": 0.9459, + "step": 35886 + }, + { + "epoch": 0.86, + "learning_rate": 9.67791483040521e-07, + "loss": 0.9579, + "step": 35887 + }, + { + "epoch": 0.86, + "learning_rate": 9.6745722760897e-07, + "loss": 1.1398, + "step": 35888 + }, + { + "epoch": 0.86, + "learning_rate": 9.671230269757293e-07, + "loss": 1.2045, + "step": 35889 + }, + { + "epoch": 0.86, + "learning_rate": 9.667888811428316e-07, + "loss": 0.9137, + "step": 35890 + }, + { + "epoch": 0.86, + "learning_rate": 9.664547901122978e-07, + "loss": 0.8766, + "step": 35891 + }, + { + "epoch": 0.86, + "learning_rate": 9.661207538861584e-07, + "loss": 1.0221, + "step": 35892 + }, + { + "epoch": 0.86, + "learning_rate": 9.657867724664416e-07, + "loss": 1.0808, + "step": 35893 + }, + { + "epoch": 0.86, + "learning_rate": 9.654528458551694e-07, + "loss": 0.9042, + "step": 35894 + }, + { + "epoch": 0.86, + "learning_rate": 9.651189740543699e-07, + "loss": 0.8469, + "step": 35895 + }, + { + "epoch": 0.86, + "learning_rate": 9.647851570660704e-07, + "loss": 1.06, + "step": 35896 + }, + { + "epoch": 0.86, + "learning_rate": 9.644513948922928e-07, + "loss": 1.0917, + "step": 35897 + }, + { + "epoch": 0.86, + "learning_rate": 9.641176875350645e-07, + "loss": 1.0943, + "step": 35898 + }, + { + "epoch": 0.86, + "learning_rate": 9.637840349964067e-07, + "loss": 1.1815, + "step": 35899 + }, + { + "epoch": 0.86, + "learning_rate": 9.634504372783459e-07, + "loss": 1.1048, + "step": 35900 + }, + { + "epoch": 0.86, + "learning_rate": 9.631168943829083e-07, + "loss": 1.3816, + "step": 35901 + }, + { + "epoch": 0.86, + "learning_rate": 9.627834063121122e-07, + "loss": 0.9402, + "step": 35902 + }, + { + "epoch": 0.86, + "learning_rate": 9.624499730679837e-07, + "loss": 0.7691, + "step": 35903 + }, + { + "epoch": 0.86, + "learning_rate": 9.62116594652547e-07, + "loss": 0.7967, + "step": 35904 + }, + { + "epoch": 0.86, + "learning_rate": 9.617832710678222e-07, + "loss": 1.0958, + "step": 35905 + }, + { + "epoch": 0.86, + "learning_rate": 9.614500023158335e-07, + "loss": 0.8749, + "step": 35906 + }, + { + "epoch": 0.86, + "learning_rate": 9.611167883985994e-07, + "loss": 1.1979, + "step": 35907 + }, + { + "epoch": 0.86, + "learning_rate": 9.60783629318145e-07, + "loss": 1.038, + "step": 35908 + }, + { + "epoch": 0.86, + "learning_rate": 9.604505250764906e-07, + "loss": 0.8752, + "step": 35909 + }, + { + "epoch": 0.86, + "learning_rate": 9.601174756756558e-07, + "loss": 0.9092, + "step": 35910 + }, + { + "epoch": 0.86, + "learning_rate": 9.597844811176616e-07, + "loss": 1.0493, + "step": 35911 + }, + { + "epoch": 0.86, + "learning_rate": 9.594515414045302e-07, + "loss": 0.9796, + "step": 35912 + }, + { + "epoch": 0.86, + "learning_rate": 9.591186565382782e-07, + "loss": 1.1204, + "step": 35913 + }, + { + "epoch": 0.86, + "learning_rate": 9.587858265209283e-07, + "loss": 1.2414, + "step": 35914 + }, + { + "epoch": 0.86, + "learning_rate": 9.584530513544965e-07, + "loss": 0.9355, + "step": 35915 + }, + { + "epoch": 0.86, + "learning_rate": 9.581203310410037e-07, + "loss": 1.103, + "step": 35916 + }, + { + "epoch": 0.86, + "learning_rate": 9.577876655824692e-07, + "loss": 0.9873, + "step": 35917 + }, + { + "epoch": 0.86, + "learning_rate": 9.574550549809092e-07, + "loss": 0.9131, + "step": 35918 + }, + { + "epoch": 0.86, + "learning_rate": 9.57122499238342e-07, + "loss": 0.929, + "step": 35919 + }, + { + "epoch": 0.86, + "learning_rate": 9.567899983567875e-07, + "loss": 1.0569, + "step": 35920 + }, + { + "epoch": 0.86, + "learning_rate": 9.56457552338258e-07, + "loss": 1.2674, + "step": 35921 + }, + { + "epoch": 0.86, + "learning_rate": 9.561251611847766e-07, + "loss": 0.9755, + "step": 35922 + }, + { + "epoch": 0.86, + "learning_rate": 9.557928248983539e-07, + "loss": 0.92, + "step": 35923 + }, + { + "epoch": 0.86, + "learning_rate": 9.554605434810094e-07, + "loss": 1.072, + "step": 35924 + }, + { + "epoch": 0.86, + "learning_rate": 9.551283169347603e-07, + "loss": 1.0659, + "step": 35925 + }, + { + "epoch": 0.86, + "learning_rate": 9.547961452616182e-07, + "loss": 0.93, + "step": 35926 + }, + { + "epoch": 0.86, + "learning_rate": 9.544640284635999e-07, + "loss": 0.9324, + "step": 35927 + }, + { + "epoch": 0.86, + "learning_rate": 9.541319665427228e-07, + "loss": 1.2056, + "step": 35928 + }, + { + "epoch": 0.86, + "learning_rate": 9.537999595009984e-07, + "loss": 1.0095, + "step": 35929 + }, + { + "epoch": 0.86, + "learning_rate": 9.534680073404423e-07, + "loss": 1.1531, + "step": 35930 + }, + { + "epoch": 0.86, + "learning_rate": 9.531361100630665e-07, + "loss": 1.0687, + "step": 35931 + }, + { + "epoch": 0.86, + "learning_rate": 9.528042676708871e-07, + "loss": 1.0103, + "step": 35932 + }, + { + "epoch": 0.86, + "learning_rate": 9.524724801659169e-07, + "loss": 0.6815, + "step": 35933 + }, + { + "epoch": 0.86, + "learning_rate": 9.521407475501665e-07, + "loss": 0.8988, + "step": 35934 + }, + { + "epoch": 0.86, + "learning_rate": 9.518090698256499e-07, + "loss": 0.9303, + "step": 35935 + }, + { + "epoch": 0.86, + "learning_rate": 9.51477446994381e-07, + "loss": 0.9416, + "step": 35936 + }, + { + "epoch": 0.86, + "learning_rate": 9.511458790583683e-07, + "loss": 1.0318, + "step": 35937 + }, + { + "epoch": 0.86, + "learning_rate": 9.508143660196256e-07, + "loss": 0.8781, + "step": 35938 + }, + { + "epoch": 0.86, + "learning_rate": 9.504829078801647e-07, + "loss": 0.9886, + "step": 35939 + }, + { + "epoch": 0.86, + "learning_rate": 9.50151504641994e-07, + "loss": 0.8114, + "step": 35940 + }, + { + "epoch": 0.86, + "learning_rate": 9.498201563071274e-07, + "loss": 0.9451, + "step": 35941 + }, + { + "epoch": 0.86, + "learning_rate": 9.494888628775711e-07, + "loss": 1.1687, + "step": 35942 + }, + { + "epoch": 0.86, + "learning_rate": 9.49157624355338e-07, + "loss": 0.9674, + "step": 35943 + }, + { + "epoch": 0.86, + "learning_rate": 9.488264407424374e-07, + "loss": 0.7066, + "step": 35944 + }, + { + "epoch": 0.86, + "learning_rate": 9.484953120408769e-07, + "loss": 1.006, + "step": 35945 + }, + { + "epoch": 0.86, + "learning_rate": 9.481642382526656e-07, + "loss": 1.1754, + "step": 35946 + }, + { + "epoch": 0.86, + "learning_rate": 9.478332193798145e-07, + "loss": 1.162, + "step": 35947 + }, + { + "epoch": 0.86, + "learning_rate": 9.475022554243285e-07, + "loss": 1.0197, + "step": 35948 + }, + { + "epoch": 0.86, + "learning_rate": 9.471713463882193e-07, + "loss": 0.944, + "step": 35949 + }, + { + "epoch": 0.86, + "learning_rate": 9.468404922734897e-07, + "loss": 1.0839, + "step": 35950 + }, + { + "epoch": 0.86, + "learning_rate": 9.465096930821494e-07, + "loss": 0.9576, + "step": 35951 + }, + { + "epoch": 0.86, + "learning_rate": 9.461789488162076e-07, + "loss": 0.9844, + "step": 35952 + }, + { + "epoch": 0.86, + "learning_rate": 9.458482594776664e-07, + "loss": 1.1875, + "step": 35953 + }, + { + "epoch": 0.86, + "learning_rate": 9.455176250685338e-07, + "loss": 1.3093, + "step": 35954 + }, + { + "epoch": 0.86, + "learning_rate": 9.451870455908185e-07, + "loss": 0.9147, + "step": 35955 + }, + { + "epoch": 0.86, + "learning_rate": 9.44856521046521e-07, + "loss": 1.0734, + "step": 35956 + }, + { + "epoch": 0.86, + "learning_rate": 9.445260514376519e-07, + "loss": 0.9448, + "step": 35957 + }, + { + "epoch": 0.86, + "learning_rate": 9.441956367662108e-07, + "loss": 0.6822, + "step": 35958 + }, + { + "epoch": 0.86, + "learning_rate": 9.438652770342049e-07, + "loss": 0.9777, + "step": 35959 + }, + { + "epoch": 0.86, + "learning_rate": 9.435349722436404e-07, + "loss": 1.2315, + "step": 35960 + }, + { + "epoch": 0.86, + "learning_rate": 9.432047223965157e-07, + "loss": 1.1611, + "step": 35961 + }, + { + "epoch": 0.86, + "learning_rate": 9.428745274948392e-07, + "loss": 1.0761, + "step": 35962 + }, + { + "epoch": 0.86, + "learning_rate": 9.425443875406137e-07, + "loss": 0.9275, + "step": 35963 + }, + { + "epoch": 0.86, + "learning_rate": 9.422143025358399e-07, + "loss": 1.0804, + "step": 35964 + }, + { + "epoch": 0.86, + "learning_rate": 9.418842724825227e-07, + "loss": 0.9049, + "step": 35965 + }, + { + "epoch": 0.86, + "learning_rate": 9.415542973826608e-07, + "loss": 1.0218, + "step": 35966 + }, + { + "epoch": 0.86, + "learning_rate": 9.41224377238259e-07, + "loss": 0.9994, + "step": 35967 + }, + { + "epoch": 0.86, + "learning_rate": 9.408945120513202e-07, + "loss": 1.024, + "step": 35968 + }, + { + "epoch": 0.87, + "learning_rate": 9.405647018238407e-07, + "loss": 0.9943, + "step": 35969 + }, + { + "epoch": 0.87, + "learning_rate": 9.402349465578242e-07, + "loss": 0.8347, + "step": 35970 + }, + { + "epoch": 0.87, + "learning_rate": 9.399052462552738e-07, + "loss": 0.8812, + "step": 35971 + }, + { + "epoch": 0.87, + "learning_rate": 9.395756009181856e-07, + "loss": 1.0146, + "step": 35972 + }, + { + "epoch": 0.87, + "learning_rate": 9.392460105485623e-07, + "loss": 1.1023, + "step": 35973 + }, + { + "epoch": 0.87, + "learning_rate": 9.389164751484014e-07, + "loss": 1.1814, + "step": 35974 + }, + { + "epoch": 0.87, + "learning_rate": 9.385869947197023e-07, + "loss": 1.127, + "step": 35975 + }, + { + "epoch": 0.87, + "learning_rate": 9.382575692644657e-07, + "loss": 0.8983, + "step": 35976 + }, + { + "epoch": 0.87, + "learning_rate": 9.379281987846878e-07, + "loss": 1.1576, + "step": 35977 + }, + { + "epoch": 0.87, + "learning_rate": 9.37598883282368e-07, + "loss": 1.154, + "step": 35978 + }, + { + "epoch": 0.87, + "learning_rate": 9.372696227595057e-07, + "loss": 0.8856, + "step": 35979 + }, + { + "epoch": 0.87, + "learning_rate": 9.369404172180952e-07, + "loss": 1.3047, + "step": 35980 + }, + { + "epoch": 0.87, + "learning_rate": 9.366112666601357e-07, + "loss": 1.0979, + "step": 35981 + }, + { + "epoch": 0.87, + "learning_rate": 9.362821710876257e-07, + "loss": 1.0229, + "step": 35982 + }, + { + "epoch": 0.87, + "learning_rate": 9.359531305025582e-07, + "loss": 1.1044, + "step": 35983 + }, + { + "epoch": 0.87, + "learning_rate": 9.356241449069326e-07, + "loss": 1.2163, + "step": 35984 + }, + { + "epoch": 0.87, + "learning_rate": 9.352952143027406e-07, + "loss": 0.8016, + "step": 35985 + }, + { + "epoch": 0.87, + "learning_rate": 9.349663386919816e-07, + "loss": 1.0771, + "step": 35986 + }, + { + "epoch": 0.87, + "learning_rate": 9.346375180766509e-07, + "loss": 0.8721, + "step": 35987 + }, + { + "epoch": 0.87, + "learning_rate": 9.343087524587402e-07, + "loss": 1.2318, + "step": 35988 + }, + { + "epoch": 0.87, + "learning_rate": 9.339800418402456e-07, + "loss": 1.0927, + "step": 35989 + }, + { + "epoch": 0.87, + "learning_rate": 9.336513862231644e-07, + "loss": 0.7041, + "step": 35990 + }, + { + "epoch": 0.87, + "learning_rate": 9.333227856094851e-07, + "loss": 0.9484, + "step": 35991 + }, + { + "epoch": 0.87, + "learning_rate": 9.329942400012059e-07, + "loss": 1.1213, + "step": 35992 + }, + { + "epoch": 0.87, + "learning_rate": 9.326657494003155e-07, + "loss": 1.0815, + "step": 35993 + }, + { + "epoch": 0.87, + "learning_rate": 9.323373138088098e-07, + "loss": 0.9214, + "step": 35994 + }, + { + "epoch": 0.87, + "learning_rate": 9.320089332286831e-07, + "loss": 0.9922, + "step": 35995 + }, + { + "epoch": 0.87, + "learning_rate": 9.316806076619234e-07, + "loss": 1.1846, + "step": 35996 + }, + { + "epoch": 0.87, + "learning_rate": 9.313523371105237e-07, + "loss": 1.0458, + "step": 35997 + }, + { + "epoch": 0.87, + "learning_rate": 9.310241215764781e-07, + "loss": 1.2186, + "step": 35998 + }, + { + "epoch": 0.87, + "learning_rate": 9.306959610617727e-07, + "loss": 1.2698, + "step": 35999 + }, + { + "epoch": 0.87, + "learning_rate": 9.303678555684059e-07, + "loss": 0.9805, + "step": 36000 + }, + { + "epoch": 0.87, + "learning_rate": 9.300398050983617e-07, + "loss": 0.8552, + "step": 36001 + }, + { + "epoch": 0.87, + "learning_rate": 9.29711809653634e-07, + "loss": 1.2565, + "step": 36002 + }, + { + "epoch": 0.87, + "learning_rate": 9.2938386923621e-07, + "loss": 1.2175, + "step": 36003 + }, + { + "epoch": 0.87, + "learning_rate": 9.290559838480806e-07, + "loss": 1.0154, + "step": 36004 + }, + { + "epoch": 0.87, + "learning_rate": 9.287281534912363e-07, + "loss": 0.9894, + "step": 36005 + }, + { + "epoch": 0.87, + "learning_rate": 9.284003781676632e-07, + "loss": 0.9272, + "step": 36006 + }, + { + "epoch": 0.87, + "learning_rate": 9.280726578793508e-07, + "loss": 1.1748, + "step": 36007 + }, + { + "epoch": 0.87, + "learning_rate": 9.277449926282899e-07, + "loss": 0.9659, + "step": 36008 + }, + { + "epoch": 0.87, + "learning_rate": 9.274173824164634e-07, + "loss": 0.8244, + "step": 36009 + }, + { + "epoch": 0.87, + "learning_rate": 9.27089827245865e-07, + "loss": 0.7475, + "step": 36010 + }, + { + "epoch": 0.87, + "learning_rate": 9.267623271184755e-07, + "loss": 1.1498, + "step": 36011 + }, + { + "epoch": 0.87, + "learning_rate": 9.264348820362857e-07, + "loss": 1.0776, + "step": 36012 + }, + { + "epoch": 0.87, + "learning_rate": 9.261074920012825e-07, + "loss": 1.2882, + "step": 36013 + }, + { + "epoch": 0.87, + "learning_rate": 9.257801570154489e-07, + "loss": 0.9836, + "step": 36014 + }, + { + "epoch": 0.87, + "learning_rate": 9.254528770807736e-07, + "loss": 1.1382, + "step": 36015 + }, + { + "epoch": 0.87, + "learning_rate": 9.251256521992424e-07, + "loss": 1.0793, + "step": 36016 + }, + { + "epoch": 0.87, + "learning_rate": 9.247984823728373e-07, + "loss": 1.1937, + "step": 36017 + }, + { + "epoch": 0.87, + "learning_rate": 9.244713676035455e-07, + "loss": 0.888, + "step": 36018 + }, + { + "epoch": 0.87, + "learning_rate": 9.241443078933532e-07, + "loss": 1.0261, + "step": 36019 + }, + { + "epoch": 0.87, + "learning_rate": 9.23817303244241e-07, + "loss": 1.0202, + "step": 36020 + }, + { + "epoch": 0.87, + "learning_rate": 9.234903536581952e-07, + "loss": 0.8995, + "step": 36021 + }, + { + "epoch": 0.87, + "learning_rate": 9.231634591371974e-07, + "loss": 0.9916, + "step": 36022 + }, + { + "epoch": 0.87, + "learning_rate": 9.228366196832316e-07, + "loss": 0.8031, + "step": 36023 + }, + { + "epoch": 0.87, + "learning_rate": 9.225098352982831e-07, + "loss": 1.043, + "step": 36024 + }, + { + "epoch": 0.87, + "learning_rate": 9.221831059843301e-07, + "loss": 0.9583, + "step": 36025 + }, + { + "epoch": 0.87, + "learning_rate": 9.218564317433576e-07, + "loss": 0.995, + "step": 36026 + }, + { + "epoch": 0.87, + "learning_rate": 9.215298125773486e-07, + "loss": 0.9329, + "step": 36027 + }, + { + "epoch": 0.87, + "learning_rate": 9.212032484882805e-07, + "loss": 0.9375, + "step": 36028 + }, + { + "epoch": 0.87, + "learning_rate": 9.208767394781382e-07, + "loss": 1.0195, + "step": 36029 + }, + { + "epoch": 0.87, + "learning_rate": 9.205502855489001e-07, + "loss": 0.8123, + "step": 36030 + }, + { + "epoch": 0.87, + "learning_rate": 9.202238867025481e-07, + "loss": 0.7065, + "step": 36031 + }, + { + "epoch": 0.87, + "learning_rate": 9.198975429410628e-07, + "loss": 1.1573, + "step": 36032 + }, + { + "epoch": 0.87, + "learning_rate": 9.195712542664225e-07, + "loss": 0.9345, + "step": 36033 + }, + { + "epoch": 0.87, + "learning_rate": 9.192450206806069e-07, + "loss": 1.0761, + "step": 36034 + }, + { + "epoch": 0.87, + "learning_rate": 9.189188421855977e-07, + "loss": 0.9537, + "step": 36035 + }, + { + "epoch": 0.87, + "learning_rate": 9.185927187833709e-07, + "loss": 1.0416, + "step": 36036 + }, + { + "epoch": 0.87, + "learning_rate": 9.182666504759075e-07, + "loss": 1.1274, + "step": 36037 + }, + { + "epoch": 0.87, + "learning_rate": 9.179406372651823e-07, + "loss": 1.1597, + "step": 36038 + }, + { + "epoch": 0.87, + "learning_rate": 9.176146791531737e-07, + "loss": 0.6874, + "step": 36039 + }, + { + "epoch": 0.87, + "learning_rate": 9.172887761418637e-07, + "loss": 0.8866, + "step": 36040 + }, + { + "epoch": 0.87, + "learning_rate": 9.169629282332249e-07, + "loss": 1.0995, + "step": 36041 + }, + { + "epoch": 0.87, + "learning_rate": 9.166371354292348e-07, + "loss": 0.849, + "step": 36042 + }, + { + "epoch": 0.87, + "learning_rate": 9.163113977318727e-07, + "loss": 0.8489, + "step": 36043 + }, + { + "epoch": 0.87, + "learning_rate": 9.159857151431106e-07, + "loss": 1.13, + "step": 36044 + }, + { + "epoch": 0.87, + "learning_rate": 9.15660087664928e-07, + "loss": 1.06, + "step": 36045 + }, + { + "epoch": 0.87, + "learning_rate": 9.153345152992976e-07, + "loss": 1.1944, + "step": 36046 + }, + { + "epoch": 0.87, + "learning_rate": 9.150089980481958e-07, + "loss": 1.1996, + "step": 36047 + }, + { + "epoch": 0.87, + "learning_rate": 9.146835359135997e-07, + "loss": 1.2262, + "step": 36048 + }, + { + "epoch": 0.87, + "learning_rate": 9.14358128897479e-07, + "loss": 1.0694, + "step": 36049 + }, + { + "epoch": 0.87, + "learning_rate": 9.140327770018109e-07, + "loss": 0.9415, + "step": 36050 + }, + { + "epoch": 0.87, + "learning_rate": 9.137074802285706e-07, + "loss": 1.2239, + "step": 36051 + }, + { + "epoch": 0.87, + "learning_rate": 9.133822385797287e-07, + "loss": 1.1452, + "step": 36052 + }, + { + "epoch": 0.87, + "learning_rate": 9.130570520572601e-07, + "loss": 1.2434, + "step": 36053 + }, + { + "epoch": 0.87, + "learning_rate": 9.127319206631358e-07, + "loss": 1.2123, + "step": 36054 + }, + { + "epoch": 0.87, + "learning_rate": 9.124068443993294e-07, + "loss": 1.135, + "step": 36055 + }, + { + "epoch": 0.87, + "learning_rate": 9.120818232678142e-07, + "loss": 1.0088, + "step": 36056 + }, + { + "epoch": 0.87, + "learning_rate": 9.117568572705604e-07, + "loss": 1.0227, + "step": 36057 + }, + { + "epoch": 0.87, + "learning_rate": 9.114319464095389e-07, + "loss": 1.032, + "step": 36058 + }, + { + "epoch": 0.87, + "learning_rate": 9.111070906867248e-07, + "loss": 1.2759, + "step": 36059 + }, + { + "epoch": 0.87, + "learning_rate": 9.107822901040841e-07, + "loss": 1.1054, + "step": 36060 + }, + { + "epoch": 0.87, + "learning_rate": 9.104575446635889e-07, + "loss": 1.1403, + "step": 36061 + }, + { + "epoch": 0.87, + "learning_rate": 9.101328543672117e-07, + "loss": 1.0616, + "step": 36062 + }, + { + "epoch": 0.87, + "learning_rate": 9.098082192169189e-07, + "loss": 0.8962, + "step": 36063 + }, + { + "epoch": 0.87, + "learning_rate": 9.094836392146833e-07, + "loss": 0.81, + "step": 36064 + }, + { + "epoch": 0.87, + "learning_rate": 9.091591143624701e-07, + "loss": 1.0573, + "step": 36065 + }, + { + "epoch": 0.87, + "learning_rate": 9.088346446622498e-07, + "loss": 0.9005, + "step": 36066 + }, + { + "epoch": 0.87, + "learning_rate": 9.085102301159943e-07, + "loss": 1.1176, + "step": 36067 + }, + { + "epoch": 0.87, + "learning_rate": 9.081858707256663e-07, + "loss": 1.0873, + "step": 36068 + }, + { + "epoch": 0.87, + "learning_rate": 9.078615664932366e-07, + "loss": 1.1575, + "step": 36069 + }, + { + "epoch": 0.87, + "learning_rate": 9.075373174206737e-07, + "loss": 0.7387, + "step": 36070 + }, + { + "epoch": 0.87, + "learning_rate": 9.072131235099413e-07, + "loss": 0.9693, + "step": 36071 + }, + { + "epoch": 0.87, + "learning_rate": 9.068889847630091e-07, + "loss": 0.8714, + "step": 36072 + }, + { + "epoch": 0.87, + "learning_rate": 9.065649011818423e-07, + "loss": 0.7001, + "step": 36073 + }, + { + "epoch": 0.87, + "learning_rate": 9.062408727684069e-07, + "loss": 1.1727, + "step": 36074 + }, + { + "epoch": 0.87, + "learning_rate": 9.059168995246703e-07, + "loss": 0.9696, + "step": 36075 + }, + { + "epoch": 0.87, + "learning_rate": 9.055929814525954e-07, + "loss": 1.2188, + "step": 36076 + }, + { + "epoch": 0.87, + "learning_rate": 9.052691185541485e-07, + "loss": 1.0432, + "step": 36077 + }, + { + "epoch": 0.87, + "learning_rate": 9.049453108312967e-07, + "loss": 0.9654, + "step": 36078 + }, + { + "epoch": 0.87, + "learning_rate": 9.046215582860007e-07, + "loss": 0.9069, + "step": 36079 + }, + { + "epoch": 0.87, + "learning_rate": 9.04297860920228e-07, + "loss": 1.0047, + "step": 36080 + }, + { + "epoch": 0.87, + "learning_rate": 9.039742187359391e-07, + "loss": 1.1547, + "step": 36081 + }, + { + "epoch": 0.87, + "learning_rate": 9.03650631735099e-07, + "loss": 1.014, + "step": 36082 + }, + { + "epoch": 0.87, + "learning_rate": 9.03327099919673e-07, + "loss": 0.9942, + "step": 36083 + }, + { + "epoch": 0.87, + "learning_rate": 9.030036232916195e-07, + "loss": 1.1624, + "step": 36084 + }, + { + "epoch": 0.87, + "learning_rate": 9.026802018529035e-07, + "loss": 1.0678, + "step": 36085 + }, + { + "epoch": 0.87, + "learning_rate": 9.023568356054891e-07, + "loss": 1.2043, + "step": 36086 + }, + { + "epoch": 0.87, + "learning_rate": 9.020335245513334e-07, + "loss": 0.9389, + "step": 36087 + }, + { + "epoch": 0.87, + "learning_rate": 9.017102686924028e-07, + "loss": 1.1364, + "step": 36088 + }, + { + "epoch": 0.87, + "learning_rate": 9.013870680306535e-07, + "loss": 1.3316, + "step": 36089 + }, + { + "epoch": 0.87, + "learning_rate": 9.010639225680496e-07, + "loss": 0.9332, + "step": 36090 + }, + { + "epoch": 0.87, + "learning_rate": 9.007408323065514e-07, + "loss": 1.1524, + "step": 36091 + }, + { + "epoch": 0.87, + "learning_rate": 9.004177972481176e-07, + "loss": 1.071, + "step": 36092 + }, + { + "epoch": 0.87, + "learning_rate": 9.000948173947077e-07, + "loss": 1.1089, + "step": 36093 + }, + { + "epoch": 0.87, + "learning_rate": 8.997718927482846e-07, + "loss": 1.0799, + "step": 36094 + }, + { + "epoch": 0.87, + "learning_rate": 8.994490233108022e-07, + "loss": 1.0335, + "step": 36095 + }, + { + "epoch": 0.87, + "learning_rate": 8.991262090842245e-07, + "loss": 0.9676, + "step": 36096 + }, + { + "epoch": 0.87, + "learning_rate": 8.988034500705056e-07, + "loss": 1.247, + "step": 36097 + }, + { + "epoch": 0.87, + "learning_rate": 8.984807462716061e-07, + "loss": 0.8876, + "step": 36098 + }, + { + "epoch": 0.87, + "learning_rate": 8.981580976894844e-07, + "loss": 1.1589, + "step": 36099 + }, + { + "epoch": 0.87, + "learning_rate": 8.978355043260945e-07, + "loss": 0.9288, + "step": 36100 + }, + { + "epoch": 0.87, + "learning_rate": 8.975129661833948e-07, + "loss": 1.1426, + "step": 36101 + }, + { + "epoch": 0.87, + "learning_rate": 8.97190483263346e-07, + "loss": 1.4689, + "step": 36102 + }, + { + "epoch": 0.87, + "learning_rate": 8.968680555678999e-07, + "loss": 0.824, + "step": 36103 + }, + { + "epoch": 0.87, + "learning_rate": 8.965456830990127e-07, + "loss": 1.2877, + "step": 36104 + }, + { + "epoch": 0.87, + "learning_rate": 8.96223365858645e-07, + "loss": 0.9452, + "step": 36105 + }, + { + "epoch": 0.87, + "learning_rate": 8.959011038487453e-07, + "loss": 1.1575, + "step": 36106 + }, + { + "epoch": 0.87, + "learning_rate": 8.955788970712753e-07, + "loss": 1.0307, + "step": 36107 + }, + { + "epoch": 0.87, + "learning_rate": 8.952567455281847e-07, + "loss": 1.2853, + "step": 36108 + }, + { + "epoch": 0.87, + "learning_rate": 8.949346492214295e-07, + "loss": 1.0546, + "step": 36109 + }, + { + "epoch": 0.87, + "learning_rate": 8.94612608152966e-07, + "loss": 1.2208, + "step": 36110 + }, + { + "epoch": 0.87, + "learning_rate": 8.942906223247439e-07, + "loss": 1.0736, + "step": 36111 + }, + { + "epoch": 0.87, + "learning_rate": 8.939686917387191e-07, + "loss": 1.2015, + "step": 36112 + }, + { + "epoch": 0.87, + "learning_rate": 8.93646816396847e-07, + "loss": 0.892, + "step": 36113 + }, + { + "epoch": 0.87, + "learning_rate": 8.933249963010748e-07, + "loss": 1.0872, + "step": 36114 + }, + { + "epoch": 0.87, + "learning_rate": 8.930032314533599e-07, + "loss": 1.0584, + "step": 36115 + }, + { + "epoch": 0.87, + "learning_rate": 8.926815218556506e-07, + "loss": 1.0543, + "step": 36116 + }, + { + "epoch": 0.87, + "learning_rate": 8.92359867509901e-07, + "loss": 1.0931, + "step": 36117 + }, + { + "epoch": 0.87, + "learning_rate": 8.920382684180629e-07, + "loss": 1.0966, + "step": 36118 + }, + { + "epoch": 0.87, + "learning_rate": 8.917167245820846e-07, + "loss": 1.0624, + "step": 36119 + }, + { + "epoch": 0.87, + "learning_rate": 8.91395236003918e-07, + "loss": 1.0221, + "step": 36120 + }, + { + "epoch": 0.87, + "learning_rate": 8.91073802685517e-07, + "loss": 1.1564, + "step": 36121 + }, + { + "epoch": 0.87, + "learning_rate": 8.907524246288257e-07, + "loss": 1.0143, + "step": 36122 + }, + { + "epoch": 0.87, + "learning_rate": 8.904311018357992e-07, + "loss": 1.1364, + "step": 36123 + }, + { + "epoch": 0.87, + "learning_rate": 8.901098343083825e-07, + "loss": 0.7951, + "step": 36124 + }, + { + "epoch": 0.87, + "learning_rate": 8.897886220485274e-07, + "loss": 1.4022, + "step": 36125 + }, + { + "epoch": 0.87, + "learning_rate": 8.894674650581825e-07, + "loss": 1.0874, + "step": 36126 + }, + { + "epoch": 0.87, + "learning_rate": 8.89146363339295e-07, + "loss": 0.8932, + "step": 36127 + }, + { + "epoch": 0.87, + "learning_rate": 8.888253168938121e-07, + "loss": 0.9303, + "step": 36128 + }, + { + "epoch": 0.87, + "learning_rate": 8.885043257236859e-07, + "loss": 1.1427, + "step": 36129 + }, + { + "epoch": 0.87, + "learning_rate": 8.88183389830859e-07, + "loss": 1.0799, + "step": 36130 + }, + { + "epoch": 0.87, + "learning_rate": 8.878625092172811e-07, + "loss": 1.2156, + "step": 36131 + }, + { + "epoch": 0.87, + "learning_rate": 8.875416838848971e-07, + "loss": 1.1508, + "step": 36132 + }, + { + "epoch": 0.87, + "learning_rate": 8.872209138356536e-07, + "loss": 1.1842, + "step": 36133 + }, + { + "epoch": 0.87, + "learning_rate": 8.869001990714999e-07, + "loss": 0.9827, + "step": 36134 + }, + { + "epoch": 0.87, + "learning_rate": 8.865795395943766e-07, + "loss": 1.1384, + "step": 36135 + }, + { + "epoch": 0.87, + "learning_rate": 8.862589354062323e-07, + "loss": 1.0029, + "step": 36136 + }, + { + "epoch": 0.87, + "learning_rate": 8.859383865090132e-07, + "loss": 0.9978, + "step": 36137 + }, + { + "epoch": 0.87, + "learning_rate": 8.856178929046599e-07, + "loss": 1.0673, + "step": 36138 + }, + { + "epoch": 0.87, + "learning_rate": 8.85297454595121e-07, + "loss": 0.7199, + "step": 36139 + }, + { + "epoch": 0.87, + "learning_rate": 8.849770715823369e-07, + "loss": 0.9092, + "step": 36140 + }, + { + "epoch": 0.87, + "learning_rate": 8.84656743868253e-07, + "loss": 1.0569, + "step": 36141 + }, + { + "epoch": 0.87, + "learning_rate": 8.843364714548141e-07, + "loss": 0.867, + "step": 36142 + }, + { + "epoch": 0.87, + "learning_rate": 8.84016254343959e-07, + "loss": 1.0361, + "step": 36143 + }, + { + "epoch": 0.87, + "learning_rate": 8.836960925376348e-07, + "loss": 1.0887, + "step": 36144 + }, + { + "epoch": 0.87, + "learning_rate": 8.833759860377833e-07, + "loss": 1.0228, + "step": 36145 + }, + { + "epoch": 0.87, + "learning_rate": 8.83055934846343e-07, + "loss": 1.0199, + "step": 36146 + }, + { + "epoch": 0.87, + "learning_rate": 8.827359389652601e-07, + "loss": 1.2797, + "step": 36147 + }, + { + "epoch": 0.87, + "learning_rate": 8.82415998396472e-07, + "loss": 1.1629, + "step": 36148 + }, + { + "epoch": 0.87, + "learning_rate": 8.820961131419214e-07, + "loss": 1.2005, + "step": 36149 + }, + { + "epoch": 0.87, + "learning_rate": 8.817762832035504e-07, + "loss": 1.1302, + "step": 36150 + }, + { + "epoch": 0.87, + "learning_rate": 8.81456508583296e-07, + "loss": 1.2366, + "step": 36151 + }, + { + "epoch": 0.87, + "learning_rate": 8.811367892831013e-07, + "loss": 1.1716, + "step": 36152 + }, + { + "epoch": 0.87, + "learning_rate": 8.808171253049058e-07, + "loss": 1.1525, + "step": 36153 + }, + { + "epoch": 0.87, + "learning_rate": 8.804975166506458e-07, + "loss": 0.9452, + "step": 36154 + }, + { + "epoch": 0.87, + "learning_rate": 8.80177963322263e-07, + "loss": 1.1214, + "step": 36155 + }, + { + "epoch": 0.87, + "learning_rate": 8.798584653216968e-07, + "loss": 1.1037, + "step": 36156 + }, + { + "epoch": 0.87, + "learning_rate": 8.795390226508827e-07, + "loss": 1.0852, + "step": 36157 + }, + { + "epoch": 0.87, + "learning_rate": 8.79219635311761e-07, + "loss": 0.938, + "step": 36158 + }, + { + "epoch": 0.87, + "learning_rate": 8.789003033062682e-07, + "loss": 0.9456, + "step": 36159 + }, + { + "epoch": 0.87, + "learning_rate": 8.785810266363404e-07, + "loss": 0.8421, + "step": 36160 + }, + { + "epoch": 0.87, + "learning_rate": 8.782618053039182e-07, + "loss": 1.1573, + "step": 36161 + }, + { + "epoch": 0.87, + "learning_rate": 8.779426393109348e-07, + "loss": 1.2096, + "step": 36162 + }, + { + "epoch": 0.87, + "learning_rate": 8.776235286593271e-07, + "loss": 0.9291, + "step": 36163 + }, + { + "epoch": 0.87, + "learning_rate": 8.773044733510338e-07, + "loss": 0.8789, + "step": 36164 + }, + { + "epoch": 0.87, + "learning_rate": 8.769854733879868e-07, + "loss": 1.0929, + "step": 36165 + }, + { + "epoch": 0.87, + "learning_rate": 8.766665287721232e-07, + "loss": 1.1807, + "step": 36166 + }, + { + "epoch": 0.87, + "learning_rate": 8.763476395053771e-07, + "loss": 1.3571, + "step": 36167 + }, + { + "epoch": 0.87, + "learning_rate": 8.76028805589686e-07, + "loss": 0.9311, + "step": 36168 + }, + { + "epoch": 0.87, + "learning_rate": 8.757100270269803e-07, + "loss": 0.9673, + "step": 36169 + }, + { + "epoch": 0.87, + "learning_rate": 8.753913038191964e-07, + "loss": 0.947, + "step": 36170 + }, + { + "epoch": 0.87, + "learning_rate": 8.750726359682682e-07, + "loss": 0.9871, + "step": 36171 + }, + { + "epoch": 0.87, + "learning_rate": 8.747540234761265e-07, + "loss": 1.1496, + "step": 36172 + }, + { + "epoch": 0.87, + "learning_rate": 8.744354663447052e-07, + "loss": 1.1539, + "step": 36173 + }, + { + "epoch": 0.87, + "learning_rate": 8.741169645759396e-07, + "loss": 1.1355, + "step": 36174 + }, + { + "epoch": 0.87, + "learning_rate": 8.737985181717579e-07, + "loss": 0.876, + "step": 36175 + }, + { + "epoch": 0.87, + "learning_rate": 8.734801271340964e-07, + "loss": 0.9372, + "step": 36176 + }, + { + "epoch": 0.87, + "learning_rate": 8.731617914648816e-07, + "loss": 0.964, + "step": 36177 + }, + { + "epoch": 0.87, + "learning_rate": 8.728435111660482e-07, + "loss": 1.083, + "step": 36178 + }, + { + "epoch": 0.87, + "learning_rate": 8.725252862395284e-07, + "loss": 1.2312, + "step": 36179 + }, + { + "epoch": 0.87, + "learning_rate": 8.722071166872481e-07, + "loss": 0.9081, + "step": 36180 + }, + { + "epoch": 0.87, + "learning_rate": 8.718890025111404e-07, + "loss": 1.1178, + "step": 36181 + }, + { + "epoch": 0.87, + "learning_rate": 8.71570943713137e-07, + "loss": 1.0665, + "step": 36182 + }, + { + "epoch": 0.87, + "learning_rate": 8.712529402951642e-07, + "loss": 0.8577, + "step": 36183 + }, + { + "epoch": 0.87, + "learning_rate": 8.709349922591526e-07, + "loss": 0.8663, + "step": 36184 + }, + { + "epoch": 0.87, + "learning_rate": 8.706170996070317e-07, + "loss": 0.868, + "step": 36185 + }, + { + "epoch": 0.87, + "learning_rate": 8.702992623407292e-07, + "loss": 1.2577, + "step": 36186 + }, + { + "epoch": 0.87, + "learning_rate": 8.699814804621742e-07, + "loss": 0.9633, + "step": 36187 + }, + { + "epoch": 0.87, + "learning_rate": 8.696637539732933e-07, + "loss": 0.8685, + "step": 36188 + }, + { + "epoch": 0.87, + "learning_rate": 8.693460828760148e-07, + "loss": 1.1158, + "step": 36189 + }, + { + "epoch": 0.87, + "learning_rate": 8.690284671722671e-07, + "loss": 1.0906, + "step": 36190 + }, + { + "epoch": 0.87, + "learning_rate": 8.687109068639744e-07, + "loss": 1.0631, + "step": 36191 + }, + { + "epoch": 0.87, + "learning_rate": 8.683934019530649e-07, + "loss": 0.8829, + "step": 36192 + }, + { + "epoch": 0.87, + "learning_rate": 8.680759524414662e-07, + "loss": 1.0021, + "step": 36193 + }, + { + "epoch": 0.87, + "learning_rate": 8.67758558331101e-07, + "loss": 0.9451, + "step": 36194 + }, + { + "epoch": 0.87, + "learning_rate": 8.67441219623898e-07, + "loss": 0.9926, + "step": 36195 + }, + { + "epoch": 0.87, + "learning_rate": 8.671239363217798e-07, + "loss": 1.1294, + "step": 36196 + }, + { + "epoch": 0.87, + "learning_rate": 8.668067084266729e-07, + "loss": 0.7331, + "step": 36197 + }, + { + "epoch": 0.87, + "learning_rate": 8.664895359405023e-07, + "loss": 1.063, + "step": 36198 + }, + { + "epoch": 0.87, + "learning_rate": 8.66172418865191e-07, + "loss": 1.0282, + "step": 36199 + }, + { + "epoch": 0.87, + "learning_rate": 8.658553572026618e-07, + "loss": 0.8947, + "step": 36200 + }, + { + "epoch": 0.87, + "learning_rate": 8.65538350954842e-07, + "loss": 1.0518, + "step": 36201 + }, + { + "epoch": 0.87, + "learning_rate": 8.652214001236503e-07, + "loss": 1.0523, + "step": 36202 + }, + { + "epoch": 0.87, + "learning_rate": 8.649045047110139e-07, + "loss": 1.1959, + "step": 36203 + }, + { + "epoch": 0.87, + "learning_rate": 8.645876647188511e-07, + "loss": 1.0226, + "step": 36204 + }, + { + "epoch": 0.87, + "learning_rate": 8.642708801490862e-07, + "loss": 1.0033, + "step": 36205 + }, + { + "epoch": 0.87, + "learning_rate": 8.639541510036431e-07, + "loss": 1.0478, + "step": 36206 + }, + { + "epoch": 0.87, + "learning_rate": 8.636374772844391e-07, + "loss": 0.8662, + "step": 36207 + }, + { + "epoch": 0.87, + "learning_rate": 8.633208589933984e-07, + "loss": 1.18, + "step": 36208 + }, + { + "epoch": 0.87, + "learning_rate": 8.630042961324426e-07, + "loss": 1.0077, + "step": 36209 + }, + { + "epoch": 0.87, + "learning_rate": 8.626877887034879e-07, + "loss": 1.1058, + "step": 36210 + }, + { + "epoch": 0.87, + "learning_rate": 8.623713367084596e-07, + "loss": 1.089, + "step": 36211 + }, + { + "epoch": 0.87, + "learning_rate": 8.620549401492728e-07, + "loss": 0.994, + "step": 36212 + }, + { + "epoch": 0.87, + "learning_rate": 8.617385990278493e-07, + "loss": 1.1069, + "step": 36213 + }, + { + "epoch": 0.87, + "learning_rate": 8.614223133461109e-07, + "loss": 1.2471, + "step": 36214 + }, + { + "epoch": 0.87, + "learning_rate": 8.611060831059714e-07, + "loss": 0.6573, + "step": 36215 + }, + { + "epoch": 0.87, + "learning_rate": 8.607899083093507e-07, + "loss": 1.0247, + "step": 36216 + }, + { + "epoch": 0.87, + "learning_rate": 8.604737889581716e-07, + "loss": 0.9087, + "step": 36217 + }, + { + "epoch": 0.87, + "learning_rate": 8.601577250543458e-07, + "loss": 0.7286, + "step": 36218 + }, + { + "epoch": 0.87, + "learning_rate": 8.598417165997941e-07, + "loss": 1.1193, + "step": 36219 + }, + { + "epoch": 0.87, + "learning_rate": 8.595257635964316e-07, + "loss": 1.3606, + "step": 36220 + }, + { + "epoch": 0.87, + "learning_rate": 8.592098660461767e-07, + "loss": 1.0318, + "step": 36221 + }, + { + "epoch": 0.87, + "learning_rate": 8.58894023950948e-07, + "loss": 0.9985, + "step": 36222 + }, + { + "epoch": 0.87, + "learning_rate": 8.585782373126561e-07, + "loss": 1.0695, + "step": 36223 + }, + { + "epoch": 0.87, + "learning_rate": 8.582625061332206e-07, + "loss": 0.891, + "step": 36224 + }, + { + "epoch": 0.87, + "learning_rate": 8.579468304145566e-07, + "loss": 1.1599, + "step": 36225 + }, + { + "epoch": 0.87, + "learning_rate": 8.576312101585782e-07, + "loss": 0.9487, + "step": 36226 + }, + { + "epoch": 0.87, + "learning_rate": 8.573156453672027e-07, + "loss": 1.1352, + "step": 36227 + }, + { + "epoch": 0.87, + "learning_rate": 8.570001360423408e-07, + "loss": 0.9979, + "step": 36228 + }, + { + "epoch": 0.87, + "learning_rate": 8.566846821859087e-07, + "loss": 1.1791, + "step": 36229 + }, + { + "epoch": 0.87, + "learning_rate": 8.563692837998217e-07, + "loss": 0.9789, + "step": 36230 + }, + { + "epoch": 0.87, + "learning_rate": 8.560539408859891e-07, + "loss": 1.2899, + "step": 36231 + }, + { + "epoch": 0.87, + "learning_rate": 8.557386534463264e-07, + "loss": 0.8809, + "step": 36232 + }, + { + "epoch": 0.87, + "learning_rate": 8.554234214827495e-07, + "loss": 0.9741, + "step": 36233 + }, + { + "epoch": 0.87, + "learning_rate": 8.551082449971648e-07, + "loss": 0.7471, + "step": 36234 + }, + { + "epoch": 0.87, + "learning_rate": 8.547931239914875e-07, + "loss": 1.0899, + "step": 36235 + }, + { + "epoch": 0.87, + "learning_rate": 8.544780584676315e-07, + "loss": 0.8878, + "step": 36236 + }, + { + "epoch": 0.87, + "learning_rate": 8.541630484275043e-07, + "loss": 1.1535, + "step": 36237 + }, + { + "epoch": 0.87, + "learning_rate": 8.538480938730198e-07, + "loss": 0.8655, + "step": 36238 + }, + { + "epoch": 0.87, + "learning_rate": 8.535331948060855e-07, + "loss": 1.112, + "step": 36239 + }, + { + "epoch": 0.87, + "learning_rate": 8.532183512286152e-07, + "loss": 1.0171, + "step": 36240 + }, + { + "epoch": 0.87, + "learning_rate": 8.529035631425186e-07, + "loss": 0.9204, + "step": 36241 + }, + { + "epoch": 0.87, + "learning_rate": 8.525888305497032e-07, + "loss": 1.2663, + "step": 36242 + }, + { + "epoch": 0.87, + "learning_rate": 8.522741534520795e-07, + "loss": 1.2297, + "step": 36243 + }, + { + "epoch": 0.87, + "learning_rate": 8.519595318515583e-07, + "loss": 1.0888, + "step": 36244 + }, + { + "epoch": 0.87, + "learning_rate": 8.516449657500459e-07, + "loss": 0.8248, + "step": 36245 + }, + { + "epoch": 0.87, + "learning_rate": 8.513304551494528e-07, + "loss": 1.1564, + "step": 36246 + }, + { + "epoch": 0.87, + "learning_rate": 8.510160000516843e-07, + "loss": 1.0107, + "step": 36247 + }, + { + "epoch": 0.87, + "learning_rate": 8.507016004586499e-07, + "loss": 1.0535, + "step": 36248 + }, + { + "epoch": 0.87, + "learning_rate": 8.503872563722581e-07, + "loss": 0.9158, + "step": 36249 + }, + { + "epoch": 0.87, + "learning_rate": 8.50072967794413e-07, + "loss": 0.9512, + "step": 36250 + }, + { + "epoch": 0.87, + "learning_rate": 8.497587347270242e-07, + "loss": 0.924, + "step": 36251 + }, + { + "epoch": 0.87, + "learning_rate": 8.494445571719978e-07, + "loss": 0.9479, + "step": 36252 + }, + { + "epoch": 0.87, + "learning_rate": 8.491304351312369e-07, + "loss": 1.2385, + "step": 36253 + }, + { + "epoch": 0.87, + "learning_rate": 8.488163686066508e-07, + "loss": 1.0458, + "step": 36254 + }, + { + "epoch": 0.87, + "learning_rate": 8.485023576001427e-07, + "loss": 0.9815, + "step": 36255 + }, + { + "epoch": 0.87, + "learning_rate": 8.481884021136178e-07, + "loss": 0.9923, + "step": 36256 + }, + { + "epoch": 0.87, + "learning_rate": 8.47874502148982e-07, + "loss": 0.937, + "step": 36257 + }, + { + "epoch": 0.87, + "learning_rate": 8.475606577081386e-07, + "loss": 1.0471, + "step": 36258 + }, + { + "epoch": 0.87, + "learning_rate": 8.472468687929902e-07, + "loss": 1.2045, + "step": 36259 + }, + { + "epoch": 0.87, + "learning_rate": 8.469331354054456e-07, + "loss": 0.8158, + "step": 36260 + }, + { + "epoch": 0.87, + "learning_rate": 8.466194575474019e-07, + "loss": 1.2286, + "step": 36261 + }, + { + "epoch": 0.87, + "learning_rate": 8.463058352207676e-07, + "loss": 1.0895, + "step": 36262 + }, + { + "epoch": 0.87, + "learning_rate": 8.459922684274402e-07, + "loss": 1.0814, + "step": 36263 + }, + { + "epoch": 0.87, + "learning_rate": 8.456787571693248e-07, + "loss": 1.1163, + "step": 36264 + }, + { + "epoch": 0.87, + "learning_rate": 8.453653014483243e-07, + "loss": 0.6357, + "step": 36265 + }, + { + "epoch": 0.87, + "learning_rate": 8.450519012663383e-07, + "loss": 1.0377, + "step": 36266 + }, + { + "epoch": 0.87, + "learning_rate": 8.447385566252686e-07, + "loss": 0.9831, + "step": 36267 + }, + { + "epoch": 0.87, + "learning_rate": 8.444252675270192e-07, + "loss": 1.0856, + "step": 36268 + }, + { + "epoch": 0.87, + "learning_rate": 8.441120339734854e-07, + "loss": 1.2746, + "step": 36269 + }, + { + "epoch": 0.87, + "learning_rate": 8.437988559665733e-07, + "loss": 0.7067, + "step": 36270 + }, + { + "epoch": 0.87, + "learning_rate": 8.43485733508177e-07, + "loss": 1.1186, + "step": 36271 + }, + { + "epoch": 0.87, + "learning_rate": 8.431726666001994e-07, + "loss": 1.1201, + "step": 36272 + }, + { + "epoch": 0.87, + "learning_rate": 8.428596552445411e-07, + "loss": 0.8768, + "step": 36273 + }, + { + "epoch": 0.87, + "learning_rate": 8.425466994430975e-07, + "loss": 1.3418, + "step": 36274 + }, + { + "epoch": 0.87, + "learning_rate": 8.422337991977691e-07, + "loss": 1.0258, + "step": 36275 + }, + { + "epoch": 0.87, + "learning_rate": 8.419209545104556e-07, + "loss": 1.1402, + "step": 36276 + }, + { + "epoch": 0.87, + "learning_rate": 8.416081653830521e-07, + "loss": 0.9787, + "step": 36277 + }, + { + "epoch": 0.87, + "learning_rate": 8.412954318174582e-07, + "loss": 1.2422, + "step": 36278 + }, + { + "epoch": 0.87, + "learning_rate": 8.409827538155723e-07, + "loss": 0.9479, + "step": 36279 + }, + { + "epoch": 0.87, + "learning_rate": 8.406701313792875e-07, + "loss": 0.8929, + "step": 36280 + }, + { + "epoch": 0.87, + "learning_rate": 8.403575645105045e-07, + "loss": 0.9515, + "step": 36281 + }, + { + "epoch": 0.87, + "learning_rate": 8.400450532111159e-07, + "loss": 0.8145, + "step": 36282 + }, + { + "epoch": 0.87, + "learning_rate": 8.397325974830195e-07, + "loss": 1.0226, + "step": 36283 + }, + { + "epoch": 0.87, + "learning_rate": 8.394201973281125e-07, + "loss": 1.0755, + "step": 36284 + }, + { + "epoch": 0.87, + "learning_rate": 8.391078527482865e-07, + "loss": 0.796, + "step": 36285 + }, + { + "epoch": 0.87, + "learning_rate": 8.387955637454392e-07, + "loss": 0.8288, + "step": 36286 + }, + { + "epoch": 0.87, + "learning_rate": 8.384833303214657e-07, + "loss": 1.1016, + "step": 36287 + }, + { + "epoch": 0.87, + "learning_rate": 8.381711524782566e-07, + "loss": 0.9836, + "step": 36288 + }, + { + "epoch": 0.87, + "learning_rate": 8.378590302177103e-07, + "loss": 1.1355, + "step": 36289 + }, + { + "epoch": 0.87, + "learning_rate": 8.375469635417166e-07, + "loss": 1.0579, + "step": 36290 + }, + { + "epoch": 0.87, + "learning_rate": 8.372349524521694e-07, + "loss": 1.0815, + "step": 36291 + }, + { + "epoch": 0.87, + "learning_rate": 8.369229969509651e-07, + "loss": 1.0644, + "step": 36292 + }, + { + "epoch": 0.87, + "learning_rate": 8.366110970399921e-07, + "loss": 1.3218, + "step": 36293 + }, + { + "epoch": 0.87, + "learning_rate": 8.362992527211445e-07, + "loss": 1.183, + "step": 36294 + }, + { + "epoch": 0.87, + "learning_rate": 8.359874639963151e-07, + "loss": 1.2167, + "step": 36295 + }, + { + "epoch": 0.87, + "learning_rate": 8.356757308673924e-07, + "loss": 0.7535, + "step": 36296 + }, + { + "epoch": 0.87, + "learning_rate": 8.353640533362717e-07, + "loss": 1.2089, + "step": 36297 + }, + { + "epoch": 0.87, + "learning_rate": 8.350524314048392e-07, + "loss": 0.7942, + "step": 36298 + }, + { + "epoch": 0.87, + "learning_rate": 8.347408650749878e-07, + "loss": 0.8263, + "step": 36299 + }, + { + "epoch": 0.87, + "learning_rate": 8.344293543486104e-07, + "loss": 1.283, + "step": 36300 + }, + { + "epoch": 0.87, + "learning_rate": 8.341178992275923e-07, + "loss": 1.058, + "step": 36301 + }, + { + "epoch": 0.87, + "learning_rate": 8.33806499713824e-07, + "loss": 1.0165, + "step": 36302 + }, + { + "epoch": 0.87, + "learning_rate": 8.334951558091986e-07, + "loss": 1.0457, + "step": 36303 + }, + { + "epoch": 0.87, + "learning_rate": 8.331838675156001e-07, + "loss": 0.9992, + "step": 36304 + }, + { + "epoch": 0.87, + "learning_rate": 8.328726348349192e-07, + "loss": 0.8941, + "step": 36305 + }, + { + "epoch": 0.87, + "learning_rate": 8.325614577690433e-07, + "loss": 1.1979, + "step": 36306 + }, + { + "epoch": 0.87, + "learning_rate": 8.322503363198608e-07, + "loss": 0.7424, + "step": 36307 + }, + { + "epoch": 0.87, + "learning_rate": 8.319392704892604e-07, + "loss": 1.1262, + "step": 36308 + }, + { + "epoch": 0.87, + "learning_rate": 8.31628260279127e-07, + "loss": 0.881, + "step": 36309 + }, + { + "epoch": 0.87, + "learning_rate": 8.313173056913471e-07, + "loss": 1.1421, + "step": 36310 + }, + { + "epoch": 0.87, + "learning_rate": 8.310064067278112e-07, + "loss": 0.9607, + "step": 36311 + }, + { + "epoch": 0.87, + "learning_rate": 8.306955633904012e-07, + "loss": 1.0841, + "step": 36312 + }, + { + "epoch": 0.87, + "learning_rate": 8.303847756810058e-07, + "loss": 1.2138, + "step": 36313 + }, + { + "epoch": 0.87, + "learning_rate": 8.300740436015076e-07, + "loss": 1.0005, + "step": 36314 + }, + { + "epoch": 0.87, + "learning_rate": 8.29763367153793e-07, + "loss": 1.0926, + "step": 36315 + }, + { + "epoch": 0.87, + "learning_rate": 8.294527463397484e-07, + "loss": 0.7777, + "step": 36316 + }, + { + "epoch": 0.87, + "learning_rate": 8.291421811612566e-07, + "loss": 0.8259, + "step": 36317 + }, + { + "epoch": 0.87, + "learning_rate": 8.288316716202005e-07, + "loss": 0.9249, + "step": 36318 + }, + { + "epoch": 0.87, + "learning_rate": 8.285212177184676e-07, + "loss": 1.216, + "step": 36319 + }, + { + "epoch": 0.87, + "learning_rate": 8.282108194579386e-07, + "loss": 1.163, + "step": 36320 + }, + { + "epoch": 0.87, + "learning_rate": 8.279004768404964e-07, + "loss": 1.0679, + "step": 36321 + }, + { + "epoch": 0.87, + "learning_rate": 8.27590189868026e-07, + "loss": 1.0569, + "step": 36322 + }, + { + "epoch": 0.87, + "learning_rate": 8.272799585424074e-07, + "loss": 1.1584, + "step": 36323 + }, + { + "epoch": 0.87, + "learning_rate": 8.269697828655244e-07, + "loss": 0.7653, + "step": 36324 + }, + { + "epoch": 0.87, + "learning_rate": 8.266596628392576e-07, + "loss": 0.9455, + "step": 36325 + }, + { + "epoch": 0.87, + "learning_rate": 8.263495984654879e-07, + "loss": 0.9289, + "step": 36326 + }, + { + "epoch": 0.87, + "learning_rate": 8.260395897460993e-07, + "loss": 1.021, + "step": 36327 + }, + { + "epoch": 0.87, + "learning_rate": 8.257296366829693e-07, + "loss": 1.0673, + "step": 36328 + }, + { + "epoch": 0.87, + "learning_rate": 8.254197392779795e-07, + "loss": 1.3103, + "step": 36329 + }, + { + "epoch": 0.87, + "learning_rate": 8.25109897533013e-07, + "loss": 1.0963, + "step": 36330 + }, + { + "epoch": 0.87, + "learning_rate": 8.248001114499438e-07, + "loss": 1.1894, + "step": 36331 + }, + { + "epoch": 0.87, + "learning_rate": 8.244903810306538e-07, + "loss": 1.0074, + "step": 36332 + }, + { + "epoch": 0.87, + "learning_rate": 8.241807062770235e-07, + "loss": 1.0232, + "step": 36333 + }, + { + "epoch": 0.87, + "learning_rate": 8.238710871909317e-07, + "loss": 0.9832, + "step": 36334 + }, + { + "epoch": 0.87, + "learning_rate": 8.235615237742534e-07, + "loss": 1.0385, + "step": 36335 + }, + { + "epoch": 0.87, + "learning_rate": 8.232520160288704e-07, + "loss": 0.9341, + "step": 36336 + }, + { + "epoch": 0.87, + "learning_rate": 8.22942563956659e-07, + "loss": 1.198, + "step": 36337 + }, + { + "epoch": 0.87, + "learning_rate": 8.226331675594957e-07, + "loss": 0.9431, + "step": 36338 + }, + { + "epoch": 0.87, + "learning_rate": 8.223238268392586e-07, + "loss": 0.9116, + "step": 36339 + }, + { + "epoch": 0.87, + "learning_rate": 8.220145417978254e-07, + "loss": 1.2132, + "step": 36340 + }, + { + "epoch": 0.87, + "learning_rate": 8.217053124370689e-07, + "loss": 1.1653, + "step": 36341 + }, + { + "epoch": 0.87, + "learning_rate": 8.213961387588698e-07, + "loss": 1.0708, + "step": 36342 + }, + { + "epoch": 0.87, + "learning_rate": 8.210870207651001e-07, + "loss": 0.9864, + "step": 36343 + }, + { + "epoch": 0.87, + "learning_rate": 8.207779584576358e-07, + "loss": 1.1737, + "step": 36344 + }, + { + "epoch": 0.87, + "learning_rate": 8.204689518383536e-07, + "loss": 1.2447, + "step": 36345 + }, + { + "epoch": 0.87, + "learning_rate": 8.201600009091259e-07, + "loss": 0.9944, + "step": 36346 + }, + { + "epoch": 0.87, + "learning_rate": 8.198511056718294e-07, + "loss": 1.0715, + "step": 36347 + }, + { + "epoch": 0.87, + "learning_rate": 8.19542266128337e-07, + "loss": 1.1692, + "step": 36348 + }, + { + "epoch": 0.87, + "learning_rate": 8.192334822805215e-07, + "loss": 1.0442, + "step": 36349 + }, + { + "epoch": 0.87, + "learning_rate": 8.189247541302592e-07, + "loss": 1.1752, + "step": 36350 + }, + { + "epoch": 0.87, + "learning_rate": 8.186160816794175e-07, + "loss": 1.2429, + "step": 36351 + }, + { + "epoch": 0.87, + "learning_rate": 8.183074649298739e-07, + "loss": 1.1086, + "step": 36352 + }, + { + "epoch": 0.87, + "learning_rate": 8.179989038835001e-07, + "loss": 1.0467, + "step": 36353 + }, + { + "epoch": 0.87, + "learning_rate": 8.17690398542167e-07, + "loss": 1.0013, + "step": 36354 + }, + { + "epoch": 0.87, + "learning_rate": 8.17381948907745e-07, + "loss": 1.1062, + "step": 36355 + }, + { + "epoch": 0.87, + "learning_rate": 8.170735549821085e-07, + "loss": 1.0428, + "step": 36356 + }, + { + "epoch": 0.87, + "learning_rate": 8.167652167671259e-07, + "loss": 1.0432, + "step": 36357 + }, + { + "epoch": 0.87, + "learning_rate": 8.164569342646677e-07, + "loss": 1.1248, + "step": 36358 + }, + { + "epoch": 0.87, + "learning_rate": 8.161487074766072e-07, + "loss": 1.0229, + "step": 36359 + }, + { + "epoch": 0.87, + "learning_rate": 8.158405364048116e-07, + "loss": 0.8136, + "step": 36360 + }, + { + "epoch": 0.87, + "learning_rate": 8.155324210511517e-07, + "loss": 1.0909, + "step": 36361 + }, + { + "epoch": 0.87, + "learning_rate": 8.15224361417496e-07, + "loss": 1.0191, + "step": 36362 + }, + { + "epoch": 0.87, + "learning_rate": 8.149163575057118e-07, + "loss": 1.1241, + "step": 36363 + }, + { + "epoch": 0.87, + "learning_rate": 8.146084093176732e-07, + "loss": 1.1486, + "step": 36364 + }, + { + "epoch": 0.87, + "learning_rate": 8.14300516855242e-07, + "loss": 0.8319, + "step": 36365 + }, + { + "epoch": 0.87, + "learning_rate": 8.139926801202891e-07, + "loss": 1.038, + "step": 36366 + }, + { + "epoch": 0.87, + "learning_rate": 8.136848991146839e-07, + "loss": 1.0626, + "step": 36367 + }, + { + "epoch": 0.87, + "learning_rate": 8.133771738402896e-07, + "loss": 1.1896, + "step": 36368 + }, + { + "epoch": 0.87, + "learning_rate": 8.130695042989767e-07, + "loss": 1.3396, + "step": 36369 + }, + { + "epoch": 0.87, + "learning_rate": 8.127618904926094e-07, + "loss": 1.0206, + "step": 36370 + }, + { + "epoch": 0.87, + "learning_rate": 8.124543324230539e-07, + "loss": 0.9809, + "step": 36371 + }, + { + "epoch": 0.87, + "learning_rate": 8.121468300921786e-07, + "loss": 1.0873, + "step": 36372 + }, + { + "epoch": 0.87, + "learning_rate": 8.118393835018457e-07, + "loss": 1.2557, + "step": 36373 + }, + { + "epoch": 0.87, + "learning_rate": 8.115319926539222e-07, + "loss": 0.8571, + "step": 36374 + }, + { + "epoch": 0.87, + "learning_rate": 8.112246575502758e-07, + "loss": 0.9855, + "step": 36375 + }, + { + "epoch": 0.87, + "learning_rate": 8.109173781927649e-07, + "loss": 0.964, + "step": 36376 + }, + { + "epoch": 0.87, + "learning_rate": 8.10610154583259e-07, + "loss": 0.9906, + "step": 36377 + }, + { + "epoch": 0.87, + "learning_rate": 8.103029867236178e-07, + "loss": 0.9023, + "step": 36378 + }, + { + "epoch": 0.87, + "learning_rate": 8.099958746157066e-07, + "loss": 1.0198, + "step": 36379 + }, + { + "epoch": 0.87, + "learning_rate": 8.096888182613916e-07, + "loss": 1.1193, + "step": 36380 + }, + { + "epoch": 0.87, + "learning_rate": 8.093818176625302e-07, + "loss": 0.9526, + "step": 36381 + }, + { + "epoch": 0.87, + "learning_rate": 8.090748728209885e-07, + "loss": 1.0489, + "step": 36382 + }, + { + "epoch": 0.87, + "learning_rate": 8.087679837386286e-07, + "loss": 0.9502, + "step": 36383 + }, + { + "epoch": 0.87, + "learning_rate": 8.084611504173101e-07, + "loss": 1.1578, + "step": 36384 + }, + { + "epoch": 0.88, + "learning_rate": 8.081543728588969e-07, + "loss": 0.9682, + "step": 36385 + }, + { + "epoch": 0.88, + "learning_rate": 8.078476510652478e-07, + "loss": 1.1264, + "step": 36386 + }, + { + "epoch": 0.88, + "learning_rate": 8.075409850382254e-07, + "loss": 0.8727, + "step": 36387 + }, + { + "epoch": 0.88, + "learning_rate": 8.072343747796907e-07, + "loss": 0.9827, + "step": 36388 + }, + { + "epoch": 0.88, + "learning_rate": 8.06927820291501e-07, + "loss": 1.121, + "step": 36389 + }, + { + "epoch": 0.88, + "learning_rate": 8.066213215755181e-07, + "loss": 1.0019, + "step": 36390 + }, + { + "epoch": 0.88, + "learning_rate": 8.063148786336028e-07, + "loss": 0.9265, + "step": 36391 + }, + { + "epoch": 0.88, + "learning_rate": 8.060084914676114e-07, + "loss": 0.9824, + "step": 36392 + }, + { + "epoch": 0.88, + "learning_rate": 8.057021600794046e-07, + "loss": 1.071, + "step": 36393 + }, + { + "epoch": 0.88, + "learning_rate": 8.053958844708398e-07, + "loss": 0.7842, + "step": 36394 + }, + { + "epoch": 0.88, + "learning_rate": 8.050896646437745e-07, + "loss": 0.8608, + "step": 36395 + }, + { + "epoch": 0.88, + "learning_rate": 8.047835006000692e-07, + "loss": 1.0625, + "step": 36396 + }, + { + "epoch": 0.88, + "learning_rate": 8.044773923415782e-07, + "loss": 1.1832, + "step": 36397 + }, + { + "epoch": 0.88, + "learning_rate": 8.041713398701601e-07, + "loss": 0.8797, + "step": 36398 + }, + { + "epoch": 0.88, + "learning_rate": 8.038653431876731e-07, + "loss": 0.9075, + "step": 36399 + }, + { + "epoch": 0.88, + "learning_rate": 8.035594022959692e-07, + "loss": 0.8234, + "step": 36400 + }, + { + "epoch": 0.88, + "learning_rate": 8.032535171969092e-07, + "loss": 1.0137, + "step": 36401 + }, + { + "epoch": 0.88, + "learning_rate": 8.02947687892347e-07, + "loss": 1.031, + "step": 36402 + }, + { + "epoch": 0.88, + "learning_rate": 8.026419143841369e-07, + "loss": 1.1272, + "step": 36403 + }, + { + "epoch": 0.88, + "learning_rate": 8.023361966741361e-07, + "loss": 1.0363, + "step": 36404 + }, + { + "epoch": 0.88, + "learning_rate": 8.020305347641977e-07, + "loss": 1.034, + "step": 36405 + }, + { + "epoch": 0.88, + "learning_rate": 8.017249286561757e-07, + "loss": 0.8765, + "step": 36406 + }, + { + "epoch": 0.88, + "learning_rate": 8.014193783519264e-07, + "loss": 0.8397, + "step": 36407 + }, + { + "epoch": 0.88, + "learning_rate": 8.011138838533006e-07, + "loss": 0.9498, + "step": 36408 + }, + { + "epoch": 0.88, + "learning_rate": 8.008084451621534e-07, + "loss": 1.0839, + "step": 36409 + }, + { + "epoch": 0.88, + "learning_rate": 8.005030622803389e-07, + "loss": 1.0678, + "step": 36410 + }, + { + "epoch": 0.88, + "learning_rate": 8.001977352097079e-07, + "loss": 0.861, + "step": 36411 + }, + { + "epoch": 0.88, + "learning_rate": 7.998924639521132e-07, + "loss": 0.6852, + "step": 36412 + }, + { + "epoch": 0.88, + "learning_rate": 7.995872485094069e-07, + "loss": 0.9252, + "step": 36413 + }, + { + "epoch": 0.88, + "learning_rate": 7.992820888834396e-07, + "loss": 1.126, + "step": 36414 + }, + { + "epoch": 0.88, + "learning_rate": 7.989769850760665e-07, + "loss": 1.2174, + "step": 36415 + }, + { + "epoch": 0.88, + "learning_rate": 7.986719370891327e-07, + "loss": 1.2022, + "step": 36416 + }, + { + "epoch": 0.88, + "learning_rate": 7.983669449244924e-07, + "loss": 1.0533, + "step": 36417 + }, + { + "epoch": 0.88, + "learning_rate": 7.980620085839985e-07, + "loss": 1.1892, + "step": 36418 + }, + { + "epoch": 0.88, + "learning_rate": 7.977571280694952e-07, + "loss": 0.9919, + "step": 36419 + }, + { + "epoch": 0.88, + "learning_rate": 7.974523033828363e-07, + "loss": 1.0922, + "step": 36420 + }, + { + "epoch": 0.88, + "learning_rate": 7.971475345258684e-07, + "loss": 0.9264, + "step": 36421 + }, + { + "epoch": 0.88, + "learning_rate": 7.96842821500442e-07, + "loss": 1.1513, + "step": 36422 + }, + { + "epoch": 0.88, + "learning_rate": 7.965381643084069e-07, + "loss": 0.8764, + "step": 36423 + }, + { + "epoch": 0.88, + "learning_rate": 7.962335629516083e-07, + "loss": 1.0529, + "step": 36424 + }, + { + "epoch": 0.88, + "learning_rate": 7.959290174318956e-07, + "loss": 1.2534, + "step": 36425 + }, + { + "epoch": 0.88, + "learning_rate": 7.956245277511187e-07, + "loss": 1.0414, + "step": 36426 + }, + { + "epoch": 0.88, + "learning_rate": 7.953200939111205e-07, + "loss": 1.0782, + "step": 36427 + }, + { + "epoch": 0.88, + "learning_rate": 7.950157159137528e-07, + "loss": 1.1419, + "step": 36428 + }, + { + "epoch": 0.88, + "learning_rate": 7.947113937608575e-07, + "loss": 0.8875, + "step": 36429 + }, + { + "epoch": 0.88, + "learning_rate": 7.94407127454283e-07, + "loss": 1.0244, + "step": 36430 + }, + { + "epoch": 0.88, + "learning_rate": 7.941029169958758e-07, + "loss": 1.1021, + "step": 36431 + }, + { + "epoch": 0.88, + "learning_rate": 7.93798762387481e-07, + "loss": 0.9901, + "step": 36432 + }, + { + "epoch": 0.88, + "learning_rate": 7.934946636309426e-07, + "loss": 0.965, + "step": 36433 + }, + { + "epoch": 0.88, + "learning_rate": 7.931906207281082e-07, + "loss": 1.2626, + "step": 36434 + }, + { + "epoch": 0.88, + "learning_rate": 7.928866336808194e-07, + "loss": 0.9639, + "step": 36435 + }, + { + "epoch": 0.88, + "learning_rate": 7.925827024909239e-07, + "loss": 0.9695, + "step": 36436 + }, + { + "epoch": 0.88, + "learning_rate": 7.92278827160261e-07, + "loss": 1.2609, + "step": 36437 + }, + { + "epoch": 0.88, + "learning_rate": 7.919750076906762e-07, + "loss": 1.1257, + "step": 36438 + }, + { + "epoch": 0.88, + "learning_rate": 7.916712440840158e-07, + "loss": 0.7708, + "step": 36439 + }, + { + "epoch": 0.88, + "learning_rate": 7.91367536342118e-07, + "loss": 0.7986, + "step": 36440 + }, + { + "epoch": 0.88, + "learning_rate": 7.910638844668273e-07, + "loss": 0.9832, + "step": 36441 + }, + { + "epoch": 0.88, + "learning_rate": 7.907602884599874e-07, + "loss": 1.16, + "step": 36442 + }, + { + "epoch": 0.88, + "learning_rate": 7.904567483234383e-07, + "loss": 0.9667, + "step": 36443 + }, + { + "epoch": 0.88, + "learning_rate": 7.901532640590226e-07, + "loss": 0.7974, + "step": 36444 + }, + { + "epoch": 0.88, + "learning_rate": 7.89849835668578e-07, + "loss": 1.1988, + "step": 36445 + }, + { + "epoch": 0.88, + "learning_rate": 7.895464631539495e-07, + "loss": 1.0173, + "step": 36446 + }, + { + "epoch": 0.88, + "learning_rate": 7.892431465169781e-07, + "loss": 0.8388, + "step": 36447 + }, + { + "epoch": 0.88, + "learning_rate": 7.889398857594998e-07, + "loss": 1.1916, + "step": 36448 + }, + { + "epoch": 0.88, + "learning_rate": 7.886366808833557e-07, + "loss": 1.0363, + "step": 36449 + }, + { + "epoch": 0.88, + "learning_rate": 7.883335318903895e-07, + "loss": 0.9489, + "step": 36450 + }, + { + "epoch": 0.88, + "learning_rate": 7.880304387824345e-07, + "loss": 0.902, + "step": 36451 + }, + { + "epoch": 0.88, + "learning_rate": 7.877274015613323e-07, + "loss": 1.1505, + "step": 36452 + }, + { + "epoch": 0.88, + "learning_rate": 7.874244202289228e-07, + "loss": 0.9213, + "step": 36453 + }, + { + "epoch": 0.88, + "learning_rate": 7.871214947870398e-07, + "loss": 1.2962, + "step": 36454 + }, + { + "epoch": 0.88, + "learning_rate": 7.868186252375265e-07, + "loss": 1.0062, + "step": 36455 + }, + { + "epoch": 0.88, + "learning_rate": 7.865158115822158e-07, + "loss": 0.7783, + "step": 36456 + }, + { + "epoch": 0.88, + "learning_rate": 7.862130538229462e-07, + "loss": 0.9289, + "step": 36457 + }, + { + "epoch": 0.88, + "learning_rate": 7.859103519615563e-07, + "loss": 0.9775, + "step": 36458 + }, + { + "epoch": 0.88, + "learning_rate": 7.856077059998801e-07, + "loss": 0.9972, + "step": 36459 + }, + { + "epoch": 0.88, + "learning_rate": 7.85305115939754e-07, + "loss": 0.8748, + "step": 36460 + }, + { + "epoch": 0.88, + "learning_rate": 7.850025817830154e-07, + "loss": 0.9789, + "step": 36461 + }, + { + "epoch": 0.88, + "learning_rate": 7.847001035314983e-07, + "loss": 1.1033, + "step": 36462 + }, + { + "epoch": 0.88, + "learning_rate": 7.843976811870391e-07, + "loss": 0.9291, + "step": 36463 + }, + { + "epoch": 0.88, + "learning_rate": 7.840953147514696e-07, + "loss": 1.093, + "step": 36464 + }, + { + "epoch": 0.88, + "learning_rate": 7.837930042266262e-07, + "loss": 1.141, + "step": 36465 + }, + { + "epoch": 0.88, + "learning_rate": 7.834907496143452e-07, + "loss": 0.9862, + "step": 36466 + }, + { + "epoch": 0.88, + "learning_rate": 7.83188550916455e-07, + "loss": 0.9787, + "step": 36467 + }, + { + "epoch": 0.88, + "learning_rate": 7.828864081347931e-07, + "loss": 1.2346, + "step": 36468 + }, + { + "epoch": 0.88, + "learning_rate": 7.825843212711925e-07, + "loss": 1.1488, + "step": 36469 + }, + { + "epoch": 0.88, + "learning_rate": 7.822822903274829e-07, + "loss": 0.9227, + "step": 36470 + }, + { + "epoch": 0.88, + "learning_rate": 7.819803153055006e-07, + "loss": 1.1562, + "step": 36471 + }, + { + "epoch": 0.88, + "learning_rate": 7.81678396207074e-07, + "loss": 0.9995, + "step": 36472 + }, + { + "epoch": 0.88, + "learning_rate": 7.813765330340362e-07, + "loss": 1.0706, + "step": 36473 + }, + { + "epoch": 0.88, + "learning_rate": 7.810747257882201e-07, + "loss": 1.1229, + "step": 36474 + }, + { + "epoch": 0.88, + "learning_rate": 7.807729744714531e-07, + "loss": 1.1283, + "step": 36475 + }, + { + "epoch": 0.88, + "learning_rate": 7.804712790855684e-07, + "loss": 0.9438, + "step": 36476 + }, + { + "epoch": 0.88, + "learning_rate": 7.801696396323977e-07, + "loss": 1.0047, + "step": 36477 + }, + { + "epoch": 0.88, + "learning_rate": 7.798680561137672e-07, + "loss": 1.1274, + "step": 36478 + }, + { + "epoch": 0.88, + "learning_rate": 7.795665285315113e-07, + "loss": 1.1825, + "step": 36479 + }, + { + "epoch": 0.88, + "learning_rate": 7.792650568874538e-07, + "loss": 1.2446, + "step": 36480 + }, + { + "epoch": 0.88, + "learning_rate": 7.789636411834256e-07, + "loss": 1.0033, + "step": 36481 + }, + { + "epoch": 0.88, + "learning_rate": 7.786622814212597e-07, + "loss": 1.0211, + "step": 36482 + }, + { + "epoch": 0.88, + "learning_rate": 7.783609776027778e-07, + "loss": 1.0, + "step": 36483 + }, + { + "epoch": 0.88, + "learning_rate": 7.780597297298109e-07, + "loss": 0.8649, + "step": 36484 + }, + { + "epoch": 0.88, + "learning_rate": 7.777585378041874e-07, + "loss": 0.8479, + "step": 36485 + }, + { + "epoch": 0.88, + "learning_rate": 7.774574018277336e-07, + "loss": 0.7854, + "step": 36486 + }, + { + "epoch": 0.88, + "learning_rate": 7.771563218022771e-07, + "loss": 1.1106, + "step": 36487 + }, + { + "epoch": 0.88, + "learning_rate": 7.768552977296418e-07, + "loss": 1.132, + "step": 36488 + }, + { + "epoch": 0.88, + "learning_rate": 7.765543296116573e-07, + "loss": 0.8606, + "step": 36489 + }, + { + "epoch": 0.88, + "learning_rate": 7.762534174501491e-07, + "loss": 1.1887, + "step": 36490 + }, + { + "epoch": 0.88, + "learning_rate": 7.759525612469399e-07, + "loss": 0.8731, + "step": 36491 + }, + { + "epoch": 0.88, + "learning_rate": 7.756517610038571e-07, + "loss": 1.0406, + "step": 36492 + }, + { + "epoch": 0.88, + "learning_rate": 7.753510167227274e-07, + "loss": 0.9802, + "step": 36493 + }, + { + "epoch": 0.88, + "learning_rate": 7.750503284053712e-07, + "loss": 1.1266, + "step": 36494 + }, + { + "epoch": 0.88, + "learning_rate": 7.747496960536149e-07, + "loss": 1.1374, + "step": 36495 + }, + { + "epoch": 0.88, + "learning_rate": 7.744491196692839e-07, + "loss": 0.8732, + "step": 36496 + }, + { + "epoch": 0.88, + "learning_rate": 7.741485992541986e-07, + "loss": 0.9264, + "step": 36497 + }, + { + "epoch": 0.88, + "learning_rate": 7.738481348101857e-07, + "loss": 1.099, + "step": 36498 + }, + { + "epoch": 0.88, + "learning_rate": 7.735477263390623e-07, + "loss": 0.9716, + "step": 36499 + }, + { + "epoch": 0.88, + "learning_rate": 7.732473738426582e-07, + "loss": 0.8923, + "step": 36500 + }, + { + "epoch": 0.88, + "learning_rate": 7.729470773227899e-07, + "loss": 0.949, + "step": 36501 + }, + { + "epoch": 0.88, + "learning_rate": 7.726468367812823e-07, + "loss": 1.2619, + "step": 36502 + }, + { + "epoch": 0.88, + "learning_rate": 7.723466522199563e-07, + "loss": 0.9064, + "step": 36503 + }, + { + "epoch": 0.88, + "learning_rate": 7.720465236406316e-07, + "loss": 0.9171, + "step": 36504 + }, + { + "epoch": 0.88, + "learning_rate": 7.7174645104513e-07, + "loss": 1.098, + "step": 36505 + }, + { + "epoch": 0.88, + "learning_rate": 7.714464344352746e-07, + "loss": 1.1284, + "step": 36506 + }, + { + "epoch": 0.88, + "learning_rate": 7.711464738128804e-07, + "loss": 1.0831, + "step": 36507 + }, + { + "epoch": 0.88, + "learning_rate": 7.708465691797718e-07, + "loss": 1.1279, + "step": 36508 + }, + { + "epoch": 0.88, + "learning_rate": 7.705467205377648e-07, + "loss": 1.2551, + "step": 36509 + }, + { + "epoch": 0.88, + "learning_rate": 7.702469278886793e-07, + "loss": 0.7879, + "step": 36510 + }, + { + "epoch": 0.88, + "learning_rate": 7.699471912343371e-07, + "loss": 1.1647, + "step": 36511 + }, + { + "epoch": 0.88, + "learning_rate": 7.696475105765522e-07, + "loss": 0.7899, + "step": 36512 + }, + { + "epoch": 0.88, + "learning_rate": 7.693478859171443e-07, + "loss": 1.0613, + "step": 36513 + }, + { + "epoch": 0.88, + "learning_rate": 7.690483172579333e-07, + "loss": 0.9336, + "step": 36514 + }, + { + "epoch": 0.88, + "learning_rate": 7.687488046007341e-07, + "loss": 0.9927, + "step": 36515 + }, + { + "epoch": 0.88, + "learning_rate": 7.684493479473654e-07, + "loss": 0.9291, + "step": 36516 + }, + { + "epoch": 0.88, + "learning_rate": 7.681499472996423e-07, + "loss": 1.451, + "step": 36517 + }, + { + "epoch": 0.88, + "learning_rate": 7.678506026593813e-07, + "loss": 1.1036, + "step": 36518 + }, + { + "epoch": 0.88, + "learning_rate": 7.675513140284008e-07, + "loss": 1.0593, + "step": 36519 + }, + { + "epoch": 0.88, + "learning_rate": 7.672520814085138e-07, + "loss": 1.3445, + "step": 36520 + }, + { + "epoch": 0.88, + "learning_rate": 7.669529048015356e-07, + "loss": 0.9824, + "step": 36521 + }, + { + "epoch": 0.88, + "learning_rate": 7.666537842092858e-07, + "loss": 0.8028, + "step": 36522 + }, + { + "epoch": 0.88, + "learning_rate": 7.66354719633573e-07, + "loss": 0.9152, + "step": 36523 + }, + { + "epoch": 0.88, + "learning_rate": 7.660557110762157e-07, + "loss": 1.0321, + "step": 36524 + }, + { + "epoch": 0.88, + "learning_rate": 7.657567585390247e-07, + "loss": 0.9992, + "step": 36525 + }, + { + "epoch": 0.88, + "learning_rate": 7.654578620238162e-07, + "loss": 1.1582, + "step": 36526 + }, + { + "epoch": 0.88, + "learning_rate": 7.651590215324034e-07, + "loss": 1.2443, + "step": 36527 + }, + { + "epoch": 0.88, + "learning_rate": 7.648602370665981e-07, + "loss": 1.0303, + "step": 36528 + }, + { + "epoch": 0.88, + "learning_rate": 7.645615086282121e-07, + "loss": 1.1531, + "step": 36529 + }, + { + "epoch": 0.88, + "learning_rate": 7.642628362190618e-07, + "loss": 1.2044, + "step": 36530 + }, + { + "epoch": 0.88, + "learning_rate": 7.639642198409547e-07, + "loss": 1.0927, + "step": 36531 + }, + { + "epoch": 0.88, + "learning_rate": 7.636656594957048e-07, + "loss": 0.9661, + "step": 36532 + }, + { + "epoch": 0.88, + "learning_rate": 7.633671551851241e-07, + "loss": 1.1429, + "step": 36533 + }, + { + "epoch": 0.88, + "learning_rate": 7.630687069110199e-07, + "loss": 1.0626, + "step": 36534 + }, + { + "epoch": 0.88, + "learning_rate": 7.627703146752075e-07, + "loss": 0.9346, + "step": 36535 + }, + { + "epoch": 0.88, + "learning_rate": 7.624719784794932e-07, + "loss": 1.0507, + "step": 36536 + }, + { + "epoch": 0.88, + "learning_rate": 7.621736983256877e-07, + "loss": 1.0319, + "step": 36537 + }, + { + "epoch": 0.88, + "learning_rate": 7.618754742156042e-07, + "loss": 0.7863, + "step": 36538 + }, + { + "epoch": 0.88, + "learning_rate": 7.615773061510479e-07, + "loss": 0.9306, + "step": 36539 + }, + { + "epoch": 0.88, + "learning_rate": 7.612791941338282e-07, + "loss": 1.1436, + "step": 36540 + }, + { + "epoch": 0.88, + "learning_rate": 7.609811381657562e-07, + "loss": 1.0136, + "step": 36541 + }, + { + "epoch": 0.88, + "learning_rate": 7.606831382486368e-07, + "loss": 0.9256, + "step": 36542 + }, + { + "epoch": 0.88, + "learning_rate": 7.603851943842822e-07, + "loss": 0.9639, + "step": 36543 + }, + { + "epoch": 0.88, + "learning_rate": 7.600873065744952e-07, + "loss": 1.025, + "step": 36544 + }, + { + "epoch": 0.88, + "learning_rate": 7.597894748210844e-07, + "loss": 1.1091, + "step": 36545 + }, + { + "epoch": 0.88, + "learning_rate": 7.594916991258605e-07, + "loss": 1.1902, + "step": 36546 + }, + { + "epoch": 0.88, + "learning_rate": 7.591939794906245e-07, + "loss": 1.16, + "step": 36547 + }, + { + "epoch": 0.88, + "learning_rate": 7.588963159171847e-07, + "loss": 0.7993, + "step": 36548 + }, + { + "epoch": 0.88, + "learning_rate": 7.585987084073487e-07, + "loss": 0.8957, + "step": 36549 + }, + { + "epoch": 0.88, + "learning_rate": 7.583011569629206e-07, + "loss": 1.0781, + "step": 36550 + }, + { + "epoch": 0.88, + "learning_rate": 7.580036615857056e-07, + "loss": 1.146, + "step": 36551 + }, + { + "epoch": 0.88, + "learning_rate": 7.577062222775078e-07, + "loss": 1.0899, + "step": 36552 + }, + { + "epoch": 0.88, + "learning_rate": 7.574088390401324e-07, + "loss": 1.1723, + "step": 36553 + }, + { + "epoch": 0.88, + "learning_rate": 7.571115118753847e-07, + "loss": 1.1441, + "step": 36554 + }, + { + "epoch": 0.88, + "learning_rate": 7.568142407850664e-07, + "loss": 1.2132, + "step": 36555 + }, + { + "epoch": 0.88, + "learning_rate": 7.565170257709808e-07, + "loss": 1.0109, + "step": 36556 + }, + { + "epoch": 0.88, + "learning_rate": 7.562198668349352e-07, + "loss": 1.0753, + "step": 36557 + }, + { + "epoch": 0.88, + "learning_rate": 7.55922763978727e-07, + "loss": 1.1475, + "step": 36558 + }, + { + "epoch": 0.88, + "learning_rate": 7.556257172041626e-07, + "loss": 0.9544, + "step": 36559 + }, + { + "epoch": 0.88, + "learning_rate": 7.553287265130415e-07, + "loss": 1.0506, + "step": 36560 + }, + { + "epoch": 0.88, + "learning_rate": 7.550317919071659e-07, + "loss": 0.7894, + "step": 36561 + }, + { + "epoch": 0.88, + "learning_rate": 7.547349133883409e-07, + "loss": 1.0477, + "step": 36562 + }, + { + "epoch": 0.88, + "learning_rate": 7.544380909583615e-07, + "loss": 1.0356, + "step": 36563 + }, + { + "epoch": 0.88, + "learning_rate": 7.541413246190321e-07, + "loss": 1.1648, + "step": 36564 + }, + { + "epoch": 0.88, + "learning_rate": 7.538446143721534e-07, + "loss": 0.9352, + "step": 36565 + }, + { + "epoch": 0.88, + "learning_rate": 7.535479602195239e-07, + "loss": 1.0187, + "step": 36566 + }, + { + "epoch": 0.88, + "learning_rate": 7.532513621629455e-07, + "loss": 0.965, + "step": 36567 + }, + { + "epoch": 0.88, + "learning_rate": 7.529548202042137e-07, + "loss": 0.9511, + "step": 36568 + }, + { + "epoch": 0.88, + "learning_rate": 7.52658334345131e-07, + "loss": 1.1077, + "step": 36569 + }, + { + "epoch": 0.88, + "learning_rate": 7.523619045874964e-07, + "loss": 1.0151, + "step": 36570 + }, + { + "epoch": 0.88, + "learning_rate": 7.520655309331049e-07, + "loss": 1.0225, + "step": 36571 + }, + { + "epoch": 0.88, + "learning_rate": 7.517692133837562e-07, + "loss": 0.9604, + "step": 36572 + }, + { + "epoch": 0.88, + "learning_rate": 7.514729519412511e-07, + "loss": 1.1659, + "step": 36573 + }, + { + "epoch": 0.88, + "learning_rate": 7.511767466073827e-07, + "loss": 1.2578, + "step": 36574 + }, + { + "epoch": 0.88, + "learning_rate": 7.508805973839484e-07, + "loss": 1.2558, + "step": 36575 + }, + { + "epoch": 0.88, + "learning_rate": 7.505845042727478e-07, + "loss": 1.0908, + "step": 36576 + }, + { + "epoch": 0.88, + "learning_rate": 7.50288467275575e-07, + "loss": 0.9991, + "step": 36577 + }, + { + "epoch": 0.88, + "learning_rate": 7.499924863942265e-07, + "loss": 0.6703, + "step": 36578 + }, + { + "epoch": 0.88, + "learning_rate": 7.496965616304975e-07, + "loss": 1.0306, + "step": 36579 + }, + { + "epoch": 0.88, + "learning_rate": 7.49400692986183e-07, + "loss": 1.0959, + "step": 36580 + }, + { + "epoch": 0.88, + "learning_rate": 7.491048804630797e-07, + "loss": 1.3205, + "step": 36581 + }, + { + "epoch": 0.88, + "learning_rate": 7.488091240629803e-07, + "loss": 1.1452, + "step": 36582 + }, + { + "epoch": 0.88, + "learning_rate": 7.485134237876801e-07, + "loss": 0.8317, + "step": 36583 + }, + { + "epoch": 0.88, + "learning_rate": 7.482177796389744e-07, + "loss": 1.5121, + "step": 36584 + }, + { + "epoch": 0.88, + "learning_rate": 7.47922191618653e-07, + "loss": 0.9555, + "step": 36585 + }, + { + "epoch": 0.88, + "learning_rate": 7.476266597285131e-07, + "loss": 1.085, + "step": 36586 + }, + { + "epoch": 0.88, + "learning_rate": 7.473311839703457e-07, + "loss": 1.1812, + "step": 36587 + }, + { + "epoch": 0.88, + "learning_rate": 7.470357643459425e-07, + "loss": 1.1577, + "step": 36588 + }, + { + "epoch": 0.88, + "learning_rate": 7.467404008570989e-07, + "loss": 1.0589, + "step": 36589 + }, + { + "epoch": 0.88, + "learning_rate": 7.464450935056023e-07, + "loss": 1.1802, + "step": 36590 + }, + { + "epoch": 0.88, + "learning_rate": 7.461498422932478e-07, + "loss": 1.0057, + "step": 36591 + }, + { + "epoch": 0.88, + "learning_rate": 7.458546472218264e-07, + "loss": 1.0904, + "step": 36592 + }, + { + "epoch": 0.88, + "learning_rate": 7.455595082931266e-07, + "loss": 1.022, + "step": 36593 + }, + { + "epoch": 0.88, + "learning_rate": 7.452644255089425e-07, + "loss": 1.0013, + "step": 36594 + }, + { + "epoch": 0.88, + "learning_rate": 7.449693988710604e-07, + "loss": 0.9205, + "step": 36595 + }, + { + "epoch": 0.88, + "learning_rate": 7.446744283812723e-07, + "loss": 0.9813, + "step": 36596 + }, + { + "epoch": 0.88, + "learning_rate": 7.44379514041369e-07, + "loss": 0.8158, + "step": 36597 + }, + { + "epoch": 0.88, + "learning_rate": 7.440846558531367e-07, + "loss": 1.1379, + "step": 36598 + }, + { + "epoch": 0.88, + "learning_rate": 7.437898538183652e-07, + "loss": 1.0169, + "step": 36599 + }, + { + "epoch": 0.88, + "learning_rate": 7.434951079388463e-07, + "loss": 1.1505, + "step": 36600 + }, + { + "epoch": 0.88, + "learning_rate": 7.432004182163621e-07, + "loss": 1.1156, + "step": 36601 + }, + { + "epoch": 0.88, + "learning_rate": 7.429057846527066e-07, + "loss": 1.0897, + "step": 36602 + }, + { + "epoch": 0.88, + "learning_rate": 7.426112072496616e-07, + "loss": 0.991, + "step": 36603 + }, + { + "epoch": 0.88, + "learning_rate": 7.423166860090169e-07, + "loss": 0.9685, + "step": 36604 + }, + { + "epoch": 0.88, + "learning_rate": 7.420222209325623e-07, + "loss": 0.9061, + "step": 36605 + }, + { + "epoch": 0.88, + "learning_rate": 7.417278120220784e-07, + "loss": 1.1092, + "step": 36606 + }, + { + "epoch": 0.88, + "learning_rate": 7.414334592793548e-07, + "loss": 1.0313, + "step": 36607 + }, + { + "epoch": 0.88, + "learning_rate": 7.411391627061781e-07, + "loss": 0.8873, + "step": 36608 + }, + { + "epoch": 0.88, + "learning_rate": 7.408449223043301e-07, + "loss": 1.0307, + "step": 36609 + }, + { + "epoch": 0.88, + "learning_rate": 7.405507380756005e-07, + "loss": 1.1582, + "step": 36610 + }, + { + "epoch": 0.88, + "learning_rate": 7.402566100217689e-07, + "loss": 0.9567, + "step": 36611 + }, + { + "epoch": 0.88, + "learning_rate": 7.399625381446241e-07, + "loss": 1.0619, + "step": 36612 + }, + { + "epoch": 0.88, + "learning_rate": 7.396685224459488e-07, + "loss": 0.6675, + "step": 36613 + }, + { + "epoch": 0.88, + "learning_rate": 7.393745629275251e-07, + "loss": 1.1235, + "step": 36614 + }, + { + "epoch": 0.88, + "learning_rate": 7.390806595911381e-07, + "loss": 1.1511, + "step": 36615 + }, + { + "epoch": 0.88, + "learning_rate": 7.38786812438571e-07, + "loss": 0.6993, + "step": 36616 + }, + { + "epoch": 0.88, + "learning_rate": 7.384930214716057e-07, + "loss": 1.1406, + "step": 36617 + }, + { + "epoch": 0.88, + "learning_rate": 7.38199286692024e-07, + "loss": 1.2465, + "step": 36618 + }, + { + "epoch": 0.88, + "learning_rate": 7.3790560810161e-07, + "loss": 1.2826, + "step": 36619 + }, + { + "epoch": 0.88, + "learning_rate": 7.376119857021435e-07, + "loss": 0.8964, + "step": 36620 + }, + { + "epoch": 0.88, + "learning_rate": 7.373184194954075e-07, + "loss": 0.9824, + "step": 36621 + }, + { + "epoch": 0.88, + "learning_rate": 7.370249094831806e-07, + "loss": 0.8955, + "step": 36622 + }, + { + "epoch": 0.88, + "learning_rate": 7.367314556672444e-07, + "loss": 0.8109, + "step": 36623 + }, + { + "epoch": 0.88, + "learning_rate": 7.364380580493813e-07, + "loss": 1.0686, + "step": 36624 + }, + { + "epoch": 0.88, + "learning_rate": 7.361447166313674e-07, + "loss": 0.8663, + "step": 36625 + }, + { + "epoch": 0.88, + "learning_rate": 7.358514314149856e-07, + "loss": 0.9619, + "step": 36626 + }, + { + "epoch": 0.88, + "learning_rate": 7.355582024020147e-07, + "loss": 0.9692, + "step": 36627 + }, + { + "epoch": 0.88, + "learning_rate": 7.352650295942321e-07, + "loss": 1.2246, + "step": 36628 + }, + { + "epoch": 0.88, + "learning_rate": 7.349719129934185e-07, + "loss": 0.9275, + "step": 36629 + }, + { + "epoch": 0.88, + "learning_rate": 7.346788526013504e-07, + "loss": 0.9497, + "step": 36630 + }, + { + "epoch": 0.88, + "learning_rate": 7.343858484198052e-07, + "loss": 1.4027, + "step": 36631 + }, + { + "epoch": 0.88, + "learning_rate": 7.340929004505637e-07, + "loss": 1.1727, + "step": 36632 + }, + { + "epoch": 0.88, + "learning_rate": 7.338000086954001e-07, + "loss": 1.1333, + "step": 36633 + }, + { + "epoch": 0.88, + "learning_rate": 7.335071731560917e-07, + "loss": 1.0056, + "step": 36634 + }, + { + "epoch": 0.88, + "learning_rate": 7.332143938344172e-07, + "loss": 0.9464, + "step": 36635 + }, + { + "epoch": 0.88, + "learning_rate": 7.329216707321496e-07, + "loss": 0.8919, + "step": 36636 + }, + { + "epoch": 0.88, + "learning_rate": 7.326290038510686e-07, + "loss": 0.7942, + "step": 36637 + }, + { + "epoch": 0.88, + "learning_rate": 7.323363931929461e-07, + "loss": 1.1998, + "step": 36638 + }, + { + "epoch": 0.88, + "learning_rate": 7.320438387595585e-07, + "loss": 0.9783, + "step": 36639 + }, + { + "epoch": 0.88, + "learning_rate": 7.317513405526821e-07, + "loss": 0.9705, + "step": 36640 + }, + { + "epoch": 0.88, + "learning_rate": 7.31458898574089e-07, + "loss": 0.9343, + "step": 36641 + }, + { + "epoch": 0.88, + "learning_rate": 7.311665128255552e-07, + "loss": 1.078, + "step": 36642 + }, + { + "epoch": 0.88, + "learning_rate": 7.308741833088551e-07, + "loss": 1.0891, + "step": 36643 + }, + { + "epoch": 0.88, + "learning_rate": 7.305819100257594e-07, + "loss": 1.1517, + "step": 36644 + }, + { + "epoch": 0.88, + "learning_rate": 7.302896929780457e-07, + "loss": 0.8723, + "step": 36645 + }, + { + "epoch": 0.88, + "learning_rate": 7.299975321674812e-07, + "loss": 1.0513, + "step": 36646 + }, + { + "epoch": 0.88, + "learning_rate": 7.297054275958415e-07, + "loss": 0.9515, + "step": 36647 + }, + { + "epoch": 0.88, + "learning_rate": 7.294133792648994e-07, + "loss": 1.1702, + "step": 36648 + }, + { + "epoch": 0.88, + "learning_rate": 7.291213871764257e-07, + "loss": 1.031, + "step": 36649 + }, + { + "epoch": 0.88, + "learning_rate": 7.288294513321902e-07, + "loss": 0.8476, + "step": 36650 + }, + { + "epoch": 0.88, + "learning_rate": 7.285375717339683e-07, + "loss": 1.0946, + "step": 36651 + }, + { + "epoch": 0.88, + "learning_rate": 7.28245748383527e-07, + "loss": 1.1951, + "step": 36652 + }, + { + "epoch": 0.88, + "learning_rate": 7.279539812826386e-07, + "loss": 1.0929, + "step": 36653 + }, + { + "epoch": 0.88, + "learning_rate": 7.276622704330716e-07, + "loss": 0.981, + "step": 36654 + }, + { + "epoch": 0.88, + "learning_rate": 7.273706158365956e-07, + "loss": 0.9886, + "step": 36655 + }, + { + "epoch": 0.88, + "learning_rate": 7.270790174949838e-07, + "loss": 0.9849, + "step": 36656 + }, + { + "epoch": 0.88, + "learning_rate": 7.267874754100001e-07, + "loss": 1.1434, + "step": 36657 + }, + { + "epoch": 0.88, + "learning_rate": 7.264959895834157e-07, + "loss": 1.1145, + "step": 36658 + }, + { + "epoch": 0.88, + "learning_rate": 7.262045600170009e-07, + "loss": 0.9969, + "step": 36659 + }, + { + "epoch": 0.88, + "learning_rate": 7.259131867125202e-07, + "loss": 0.9281, + "step": 36660 + }, + { + "epoch": 0.88, + "learning_rate": 7.256218696717432e-07, + "loss": 1.0925, + "step": 36661 + }, + { + "epoch": 0.88, + "learning_rate": 7.253306088964363e-07, + "loss": 0.8617, + "step": 36662 + }, + { + "epoch": 0.88, + "learning_rate": 7.250394043883669e-07, + "loss": 1.0266, + "step": 36663 + }, + { + "epoch": 0.88, + "learning_rate": 7.247482561493036e-07, + "loss": 1.2411, + "step": 36664 + }, + { + "epoch": 0.88, + "learning_rate": 7.244571641810083e-07, + "loss": 1.0584, + "step": 36665 + }, + { + "epoch": 0.88, + "learning_rate": 7.241661284852508e-07, + "loss": 0.8871, + "step": 36666 + }, + { + "epoch": 0.88, + "learning_rate": 7.238751490637952e-07, + "loss": 0.9271, + "step": 36667 + }, + { + "epoch": 0.88, + "learning_rate": 7.235842259184079e-07, + "loss": 1.1622, + "step": 36668 + }, + { + "epoch": 0.88, + "learning_rate": 7.232933590508539e-07, + "loss": 1.081, + "step": 36669 + }, + { + "epoch": 0.88, + "learning_rate": 7.230025484628955e-07, + "loss": 1.1124, + "step": 36670 + }, + { + "epoch": 0.88, + "learning_rate": 7.227117941562988e-07, + "loss": 1.2262, + "step": 36671 + }, + { + "epoch": 0.88, + "learning_rate": 7.224210961328281e-07, + "loss": 0.9244, + "step": 36672 + }, + { + "epoch": 0.88, + "learning_rate": 7.221304543942465e-07, + "loss": 1.1744, + "step": 36673 + }, + { + "epoch": 0.88, + "learning_rate": 7.218398689423178e-07, + "loss": 1.347, + "step": 36674 + }, + { + "epoch": 0.88, + "learning_rate": 7.215493397788031e-07, + "loss": 1.3248, + "step": 36675 + }, + { + "epoch": 0.88, + "learning_rate": 7.212588669054655e-07, + "loss": 1.1086, + "step": 36676 + }, + { + "epoch": 0.88, + "learning_rate": 7.209684503240699e-07, + "loss": 0.9571, + "step": 36677 + }, + { + "epoch": 0.88, + "learning_rate": 7.206780900363741e-07, + "loss": 1.1958, + "step": 36678 + }, + { + "epoch": 0.88, + "learning_rate": 7.203877860441421e-07, + "loss": 0.8073, + "step": 36679 + }, + { + "epoch": 0.88, + "learning_rate": 7.20097538349136e-07, + "loss": 1.1055, + "step": 36680 + }, + { + "epoch": 0.88, + "learning_rate": 7.198073469531141e-07, + "loss": 1.0128, + "step": 36681 + }, + { + "epoch": 0.88, + "learning_rate": 7.195172118578397e-07, + "loss": 1.0225, + "step": 36682 + }, + { + "epoch": 0.88, + "learning_rate": 7.192271330650702e-07, + "loss": 0.8489, + "step": 36683 + }, + { + "epoch": 0.88, + "learning_rate": 7.189371105765664e-07, + "loss": 0.6427, + "step": 36684 + }, + { + "epoch": 0.88, + "learning_rate": 7.18647144394089e-07, + "loss": 0.977, + "step": 36685 + }, + { + "epoch": 0.88, + "learning_rate": 7.183572345193957e-07, + "loss": 1.0061, + "step": 36686 + }, + { + "epoch": 0.88, + "learning_rate": 7.18067380954246e-07, + "loss": 0.8129, + "step": 36687 + }, + { + "epoch": 0.88, + "learning_rate": 7.177775837003987e-07, + "loss": 1.1565, + "step": 36688 + }, + { + "epoch": 0.88, + "learning_rate": 7.174878427596111e-07, + "loss": 0.9373, + "step": 36689 + }, + { + "epoch": 0.88, + "learning_rate": 7.171981581336429e-07, + "loss": 1.0301, + "step": 36690 + }, + { + "epoch": 0.88, + "learning_rate": 7.169085298242473e-07, + "loss": 1.0766, + "step": 36691 + }, + { + "epoch": 0.88, + "learning_rate": 7.16618957833185e-07, + "loss": 1.0844, + "step": 36692 + }, + { + "epoch": 0.88, + "learning_rate": 7.163294421622136e-07, + "loss": 1.0655, + "step": 36693 + }, + { + "epoch": 0.88, + "learning_rate": 7.16039982813086e-07, + "loss": 0.8371, + "step": 36694 + }, + { + "epoch": 0.88, + "learning_rate": 7.157505797875597e-07, + "loss": 1.2135, + "step": 36695 + }, + { + "epoch": 0.88, + "learning_rate": 7.154612330873933e-07, + "loss": 0.9495, + "step": 36696 + }, + { + "epoch": 0.88, + "learning_rate": 7.151719427143367e-07, + "loss": 0.893, + "step": 36697 + }, + { + "epoch": 0.88, + "learning_rate": 7.148827086701493e-07, + "loss": 0.9363, + "step": 36698 + }, + { + "epoch": 0.88, + "learning_rate": 7.145935309565855e-07, + "loss": 1.0501, + "step": 36699 + }, + { + "epoch": 0.88, + "learning_rate": 7.143044095753982e-07, + "loss": 1.015, + "step": 36700 + }, + { + "epoch": 0.88, + "learning_rate": 7.140153445283427e-07, + "loss": 1.0425, + "step": 36701 + }, + { + "epoch": 0.88, + "learning_rate": 7.137263358171698e-07, + "loss": 1.1683, + "step": 36702 + }, + { + "epoch": 0.88, + "learning_rate": 7.134373834436348e-07, + "loss": 1.1761, + "step": 36703 + }, + { + "epoch": 0.88, + "learning_rate": 7.131484874094941e-07, + "loss": 1.0595, + "step": 36704 + }, + { + "epoch": 0.88, + "learning_rate": 7.12859647716494e-07, + "loss": 1.2055, + "step": 36705 + }, + { + "epoch": 0.88, + "learning_rate": 7.125708643663909e-07, + "loss": 0.9519, + "step": 36706 + }, + { + "epoch": 0.88, + "learning_rate": 7.122821373609368e-07, + "loss": 1.1482, + "step": 36707 + }, + { + "epoch": 0.88, + "learning_rate": 7.119934667018802e-07, + "loss": 1.1843, + "step": 36708 + }, + { + "epoch": 0.88, + "learning_rate": 7.117048523909775e-07, + "loss": 0.9776, + "step": 36709 + }, + { + "epoch": 0.88, + "learning_rate": 7.11416294429974e-07, + "loss": 1.2687, + "step": 36710 + }, + { + "epoch": 0.88, + "learning_rate": 7.111277928206239e-07, + "loss": 1.0509, + "step": 36711 + }, + { + "epoch": 0.88, + "learning_rate": 7.108393475646779e-07, + "loss": 0.6683, + "step": 36712 + }, + { + "epoch": 0.88, + "learning_rate": 7.105509586638826e-07, + "loss": 1.0894, + "step": 36713 + }, + { + "epoch": 0.88, + "learning_rate": 7.102626261199907e-07, + "loss": 1.1215, + "step": 36714 + }, + { + "epoch": 0.88, + "learning_rate": 7.099743499347511e-07, + "loss": 1.0338, + "step": 36715 + }, + { + "epoch": 0.88, + "learning_rate": 7.096861301099101e-07, + "loss": 1.0128, + "step": 36716 + }, + { + "epoch": 0.88, + "learning_rate": 7.093979666472206e-07, + "loss": 1.2069, + "step": 36717 + }, + { + "epoch": 0.88, + "learning_rate": 7.091098595484258e-07, + "loss": 1.2128, + "step": 36718 + }, + { + "epoch": 0.88, + "learning_rate": 7.088218088152776e-07, + "loss": 0.7704, + "step": 36719 + }, + { + "epoch": 0.88, + "learning_rate": 7.085338144495224e-07, + "loss": 1.0814, + "step": 36720 + }, + { + "epoch": 0.88, + "learning_rate": 7.082458764529055e-07, + "loss": 1.0241, + "step": 36721 + }, + { + "epoch": 0.88, + "learning_rate": 7.079579948271764e-07, + "loss": 1.1579, + "step": 36722 + }, + { + "epoch": 0.88, + "learning_rate": 7.076701695740806e-07, + "loss": 1.0697, + "step": 36723 + }, + { + "epoch": 0.88, + "learning_rate": 7.073824006953633e-07, + "loss": 1.1578, + "step": 36724 + }, + { + "epoch": 0.88, + "learning_rate": 7.070946881927731e-07, + "loss": 1.0823, + "step": 36725 + }, + { + "epoch": 0.88, + "learning_rate": 7.06807032068052e-07, + "loss": 0.8557, + "step": 36726 + }, + { + "epoch": 0.88, + "learning_rate": 7.065194323229462e-07, + "loss": 0.8495, + "step": 36727 + }, + { + "epoch": 0.88, + "learning_rate": 7.062318889592035e-07, + "loss": 1.0816, + "step": 36728 + }, + { + "epoch": 0.88, + "learning_rate": 7.059444019785633e-07, + "loss": 1.1204, + "step": 36729 + }, + { + "epoch": 0.88, + "learning_rate": 7.056569713827732e-07, + "loss": 0.9428, + "step": 36730 + }, + { + "epoch": 0.88, + "learning_rate": 7.053695971735763e-07, + "loss": 1.0631, + "step": 36731 + }, + { + "epoch": 0.88, + "learning_rate": 7.050822793527157e-07, + "loss": 1.0024, + "step": 36732 + }, + { + "epoch": 0.88, + "learning_rate": 7.047950179219354e-07, + "loss": 1.1041, + "step": 36733 + }, + { + "epoch": 0.88, + "learning_rate": 7.045078128829763e-07, + "loss": 1.0589, + "step": 36734 + }, + { + "epoch": 0.88, + "learning_rate": 7.042206642375815e-07, + "loss": 1.2113, + "step": 36735 + }, + { + "epoch": 0.88, + "learning_rate": 7.039335719874951e-07, + "loss": 1.0323, + "step": 36736 + }, + { + "epoch": 0.88, + "learning_rate": 7.036465361344557e-07, + "loss": 0.7921, + "step": 36737 + }, + { + "epoch": 0.88, + "learning_rate": 7.033595566802065e-07, + "loss": 1.0655, + "step": 36738 + }, + { + "epoch": 0.88, + "learning_rate": 7.030726336264892e-07, + "loss": 1.0453, + "step": 36739 + }, + { + "epoch": 0.88, + "learning_rate": 7.027857669750415e-07, + "loss": 1.0253, + "step": 36740 + }, + { + "epoch": 0.88, + "learning_rate": 7.024989567276075e-07, + "loss": 1.0116, + "step": 36741 + }, + { + "epoch": 0.88, + "learning_rate": 7.022122028859246e-07, + "loss": 0.8859, + "step": 36742 + }, + { + "epoch": 0.88, + "learning_rate": 7.019255054517326e-07, + "loss": 0.9042, + "step": 36743 + }, + { + "epoch": 0.88, + "learning_rate": 7.016388644267736e-07, + "loss": 1.0248, + "step": 36744 + }, + { + "epoch": 0.88, + "learning_rate": 7.013522798127825e-07, + "loss": 0.7765, + "step": 36745 + }, + { + "epoch": 0.88, + "learning_rate": 7.010657516115005e-07, + "loss": 1.1302, + "step": 36746 + }, + { + "epoch": 0.88, + "learning_rate": 7.00779279824666e-07, + "loss": 0.8925, + "step": 36747 + }, + { + "epoch": 0.88, + "learning_rate": 7.004928644540154e-07, + "loss": 1.0892, + "step": 36748 + }, + { + "epoch": 0.88, + "learning_rate": 7.002065055012863e-07, + "loss": 1.0996, + "step": 36749 + }, + { + "epoch": 0.88, + "learning_rate": 6.999202029682195e-07, + "loss": 1.1335, + "step": 36750 + }, + { + "epoch": 0.88, + "learning_rate": 6.99633956856548e-07, + "loss": 1.0039, + "step": 36751 + }, + { + "epoch": 0.88, + "learning_rate": 6.993477671680104e-07, + "loss": 1.0435, + "step": 36752 + }, + { + "epoch": 0.88, + "learning_rate": 6.99061633904341e-07, + "loss": 0.7589, + "step": 36753 + }, + { + "epoch": 0.88, + "learning_rate": 6.987755570672772e-07, + "loss": 1.3112, + "step": 36754 + }, + { + "epoch": 0.88, + "learning_rate": 6.984895366585553e-07, + "loss": 1.0903, + "step": 36755 + }, + { + "epoch": 0.88, + "learning_rate": 6.982035726799086e-07, + "loss": 1.1007, + "step": 36756 + }, + { + "epoch": 0.88, + "learning_rate": 6.979176651330721e-07, + "loss": 0.9905, + "step": 36757 + }, + { + "epoch": 0.88, + "learning_rate": 6.976318140197835e-07, + "loss": 1.0754, + "step": 36758 + }, + { + "epoch": 0.88, + "learning_rate": 6.973460193417724e-07, + "loss": 1.2093, + "step": 36759 + }, + { + "epoch": 0.88, + "learning_rate": 6.970602811007765e-07, + "loss": 0.8408, + "step": 36760 + }, + { + "epoch": 0.88, + "learning_rate": 6.967745992985264e-07, + "loss": 1.053, + "step": 36761 + }, + { + "epoch": 0.88, + "learning_rate": 6.964889739367575e-07, + "loss": 1.0523, + "step": 36762 + }, + { + "epoch": 0.88, + "learning_rate": 6.962034050172017e-07, + "loss": 1.2289, + "step": 36763 + }, + { + "epoch": 0.88, + "learning_rate": 6.95917892541591e-07, + "loss": 1.2042, + "step": 36764 + }, + { + "epoch": 0.88, + "learning_rate": 6.956324365116573e-07, + "loss": 0.9825, + "step": 36765 + }, + { + "epoch": 0.88, + "learning_rate": 6.953470369291349e-07, + "loss": 0.7921, + "step": 36766 + }, + { + "epoch": 0.88, + "learning_rate": 6.950616937957522e-07, + "loss": 0.9689, + "step": 36767 + }, + { + "epoch": 0.88, + "learning_rate": 6.947764071132423e-07, + "loss": 1.2079, + "step": 36768 + }, + { + "epoch": 0.88, + "learning_rate": 6.94491176883334e-07, + "loss": 1.033, + "step": 36769 + }, + { + "epoch": 0.88, + "learning_rate": 6.942060031077591e-07, + "loss": 1.0202, + "step": 36770 + }, + { + "epoch": 0.88, + "learning_rate": 6.939208857882484e-07, + "loss": 0.9897, + "step": 36771 + }, + { + "epoch": 0.88, + "learning_rate": 6.936358249265307e-07, + "loss": 0.8435, + "step": 36772 + }, + { + "epoch": 0.88, + "learning_rate": 6.933508205243345e-07, + "loss": 0.9067, + "step": 36773 + }, + { + "epoch": 0.88, + "learning_rate": 6.930658725833917e-07, + "loss": 1.0314, + "step": 36774 + }, + { + "epoch": 0.88, + "learning_rate": 6.927809811054265e-07, + "loss": 0.9483, + "step": 36775 + }, + { + "epoch": 0.88, + "learning_rate": 6.924961460921731e-07, + "loss": 1.126, + "step": 36776 + }, + { + "epoch": 0.88, + "learning_rate": 6.922113675453534e-07, + "loss": 0.9382, + "step": 36777 + }, + { + "epoch": 0.88, + "learning_rate": 6.919266454666984e-07, + "loss": 1.1724, + "step": 36778 + }, + { + "epoch": 0.88, + "learning_rate": 6.916419798579366e-07, + "loss": 0.9498, + "step": 36779 + }, + { + "epoch": 0.88, + "learning_rate": 6.913573707207921e-07, + "loss": 0.9184, + "step": 36780 + }, + { + "epoch": 0.88, + "learning_rate": 6.910728180569937e-07, + "loss": 1.0742, + "step": 36781 + }, + { + "epoch": 0.88, + "learning_rate": 6.907883218682676e-07, + "loss": 0.961, + "step": 36782 + }, + { + "epoch": 0.88, + "learning_rate": 6.90503882156337e-07, + "loss": 0.872, + "step": 36783 + }, + { + "epoch": 0.88, + "learning_rate": 6.902194989229316e-07, + "loss": 0.9811, + "step": 36784 + }, + { + "epoch": 0.88, + "learning_rate": 6.899351721697733e-07, + "loss": 1.1285, + "step": 36785 + }, + { + "epoch": 0.88, + "learning_rate": 6.896509018985886e-07, + "loss": 0.9161, + "step": 36786 + }, + { + "epoch": 0.88, + "learning_rate": 6.893666881111038e-07, + "loss": 0.965, + "step": 36787 + }, + { + "epoch": 0.88, + "learning_rate": 6.890825308090388e-07, + "loss": 0.93, + "step": 36788 + }, + { + "epoch": 0.88, + "learning_rate": 6.887984299941208e-07, + "loss": 0.9895, + "step": 36789 + }, + { + "epoch": 0.88, + "learning_rate": 6.885143856680731e-07, + "loss": 0.8718, + "step": 36790 + }, + { + "epoch": 0.88, + "learning_rate": 6.882303978326177e-07, + "loss": 1.1074, + "step": 36791 + }, + { + "epoch": 0.88, + "learning_rate": 6.879464664894786e-07, + "loss": 1.0896, + "step": 36792 + }, + { + "epoch": 0.88, + "learning_rate": 6.87662591640379e-07, + "loss": 0.9397, + "step": 36793 + }, + { + "epoch": 0.88, + "learning_rate": 6.873787732870386e-07, + "loss": 1.1907, + "step": 36794 + }, + { + "epoch": 0.88, + "learning_rate": 6.870950114311825e-07, + "loss": 1.1218, + "step": 36795 + }, + { + "epoch": 0.88, + "learning_rate": 6.868113060745285e-07, + "loss": 1.0871, + "step": 36796 + }, + { + "epoch": 0.88, + "learning_rate": 6.865276572187996e-07, + "loss": 1.2109, + "step": 36797 + }, + { + "epoch": 0.88, + "learning_rate": 6.862440648657187e-07, + "loss": 1.1464, + "step": 36798 + }, + { + "epoch": 0.88, + "learning_rate": 6.859605290170024e-07, + "loss": 1.0887, + "step": 36799 + }, + { + "epoch": 0.88, + "learning_rate": 6.856770496743737e-07, + "loss": 1.1268, + "step": 36800 + }, + { + "epoch": 0.89, + "learning_rate": 6.853936268395523e-07, + "loss": 1.2756, + "step": 36801 + }, + { + "epoch": 0.89, + "learning_rate": 6.851102605142545e-07, + "loss": 1.0293, + "step": 36802 + }, + { + "epoch": 0.89, + "learning_rate": 6.848269507002048e-07, + "loss": 0.9757, + "step": 36803 + }, + { + "epoch": 0.89, + "learning_rate": 6.84543697399116e-07, + "loss": 0.9245, + "step": 36804 + }, + { + "epoch": 0.89, + "learning_rate": 6.842605006127101e-07, + "loss": 1.3692, + "step": 36805 + }, + { + "epoch": 0.89, + "learning_rate": 6.839773603427069e-07, + "loss": 1.2134, + "step": 36806 + }, + { + "epoch": 0.89, + "learning_rate": 6.836942765908195e-07, + "loss": 1.1016, + "step": 36807 + }, + { + "epoch": 0.89, + "learning_rate": 6.834112493587675e-07, + "loss": 1.0227, + "step": 36808 + }, + { + "epoch": 0.89, + "learning_rate": 6.831282786482696e-07, + "loss": 1.1587, + "step": 36809 + }, + { + "epoch": 0.89, + "learning_rate": 6.828453644610388e-07, + "loss": 1.0594, + "step": 36810 + }, + { + "epoch": 0.89, + "learning_rate": 6.825625067987962e-07, + "loss": 0.9872, + "step": 36811 + }, + { + "epoch": 0.89, + "learning_rate": 6.822797056632536e-07, + "loss": 0.9145, + "step": 36812 + }, + { + "epoch": 0.89, + "learning_rate": 6.819969610561283e-07, + "loss": 0.9566, + "step": 36813 + }, + { + "epoch": 0.89, + "learning_rate": 6.81714272979137e-07, + "loss": 0.9081, + "step": 36814 + }, + { + "epoch": 0.89, + "learning_rate": 6.814316414339917e-07, + "loss": 0.8102, + "step": 36815 + }, + { + "epoch": 0.89, + "learning_rate": 6.811490664224096e-07, + "loss": 1.1235, + "step": 36816 + }, + { + "epoch": 0.89, + "learning_rate": 6.808665479461052e-07, + "loss": 0.9982, + "step": 36817 + }, + { + "epoch": 0.89, + "learning_rate": 6.805840860067903e-07, + "loss": 1.284, + "step": 36818 + }, + { + "epoch": 0.89, + "learning_rate": 6.803016806061802e-07, + "loss": 1.0663, + "step": 36819 + }, + { + "epoch": 0.89, + "learning_rate": 6.800193317459869e-07, + "loss": 1.0185, + "step": 36820 + }, + { + "epoch": 0.89, + "learning_rate": 6.797370394279246e-07, + "loss": 1.0982, + "step": 36821 + }, + { + "epoch": 0.89, + "learning_rate": 6.794548036537063e-07, + "loss": 1.0384, + "step": 36822 + }, + { + "epoch": 0.89, + "learning_rate": 6.791726244250419e-07, + "loss": 1.0089, + "step": 36823 + }, + { + "epoch": 0.89, + "learning_rate": 6.788905017436442e-07, + "loss": 0.9155, + "step": 36824 + }, + { + "epoch": 0.89, + "learning_rate": 6.786084356112277e-07, + "loss": 1.1182, + "step": 36825 + }, + { + "epoch": 0.89, + "learning_rate": 6.783264260294997e-07, + "loss": 1.0396, + "step": 36826 + }, + { + "epoch": 0.89, + "learning_rate": 6.780444730001745e-07, + "loss": 1.0223, + "step": 36827 + }, + { + "epoch": 0.89, + "learning_rate": 6.777625765249585e-07, + "loss": 1.1321, + "step": 36828 + }, + { + "epoch": 0.89, + "learning_rate": 6.774807366055647e-07, + "loss": 0.7737, + "step": 36829 + }, + { + "epoch": 0.89, + "learning_rate": 6.771989532437039e-07, + "loss": 1.041, + "step": 36830 + }, + { + "epoch": 0.89, + "learning_rate": 6.769172264410828e-07, + "loss": 1.0405, + "step": 36831 + }, + { + "epoch": 0.89, + "learning_rate": 6.76635556199412e-07, + "loss": 1.3864, + "step": 36832 + }, + { + "epoch": 0.89, + "learning_rate": 6.763539425204002e-07, + "loss": 1.2549, + "step": 36833 + }, + { + "epoch": 0.89, + "learning_rate": 6.760723854057561e-07, + "loss": 1.0282, + "step": 36834 + }, + { + "epoch": 0.89, + "learning_rate": 6.757908848571882e-07, + "loss": 0.9572, + "step": 36835 + }, + { + "epoch": 0.89, + "learning_rate": 6.755094408764029e-07, + "loss": 1.1072, + "step": 36836 + }, + { + "epoch": 0.89, + "learning_rate": 6.75228053465109e-07, + "loss": 0.9792, + "step": 36837 + }, + { + "epoch": 0.89, + "learning_rate": 6.749467226250151e-07, + "loss": 1.0343, + "step": 36838 + }, + { + "epoch": 0.89, + "learning_rate": 6.746654483578241e-07, + "loss": 1.1851, + "step": 36839 + }, + { + "epoch": 0.89, + "learning_rate": 6.743842306652459e-07, + "loss": 0.9206, + "step": 36840 + }, + { + "epoch": 0.89, + "learning_rate": 6.741030695489825e-07, + "loss": 1.1307, + "step": 36841 + }, + { + "epoch": 0.89, + "learning_rate": 6.738219650107436e-07, + "loss": 1.1872, + "step": 36842 + }, + { + "epoch": 0.89, + "learning_rate": 6.735409170522345e-07, + "loss": 0.9596, + "step": 36843 + }, + { + "epoch": 0.89, + "learning_rate": 6.732599256751571e-07, + "loss": 1.3117, + "step": 36844 + }, + { + "epoch": 0.89, + "learning_rate": 6.729789908812179e-07, + "loss": 1.1533, + "step": 36845 + }, + { + "epoch": 0.89, + "learning_rate": 6.726981126721222e-07, + "loss": 1.3881, + "step": 36846 + }, + { + "epoch": 0.89, + "learning_rate": 6.72417291049573e-07, + "loss": 0.9221, + "step": 36847 + }, + { + "epoch": 0.89, + "learning_rate": 6.721365260152746e-07, + "loss": 0.9976, + "step": 36848 + }, + { + "epoch": 0.89, + "learning_rate": 6.718558175709289e-07, + "loss": 1.1919, + "step": 36849 + }, + { + "epoch": 0.89, + "learning_rate": 6.71575165718239e-07, + "loss": 1.0186, + "step": 36850 + }, + { + "epoch": 0.89, + "learning_rate": 6.712945704589102e-07, + "loss": 1.1911, + "step": 36851 + }, + { + "epoch": 0.89, + "learning_rate": 6.710140317946424e-07, + "loss": 1.1248, + "step": 36852 + }, + { + "epoch": 0.89, + "learning_rate": 6.707335497271373e-07, + "loss": 0.8791, + "step": 36853 + }, + { + "epoch": 0.89, + "learning_rate": 6.704531242580992e-07, + "loss": 1.0618, + "step": 36854 + }, + { + "epoch": 0.89, + "learning_rate": 6.701727553892256e-07, + "loss": 1.0905, + "step": 36855 + }, + { + "epoch": 0.89, + "learning_rate": 6.69892443122222e-07, + "loss": 1.154, + "step": 36856 + }, + { + "epoch": 0.89, + "learning_rate": 6.696121874587835e-07, + "loss": 1.1484, + "step": 36857 + }, + { + "epoch": 0.89, + "learning_rate": 6.693319884006144e-07, + "loss": 1.2684, + "step": 36858 + }, + { + "epoch": 0.89, + "learning_rate": 6.690518459494144e-07, + "loss": 1.1059, + "step": 36859 + }, + { + "epoch": 0.89, + "learning_rate": 6.687717601068799e-07, + "loss": 1.05, + "step": 36860 + }, + { + "epoch": 0.89, + "learning_rate": 6.684917308747129e-07, + "loss": 1.1232, + "step": 36861 + }, + { + "epoch": 0.89, + "learning_rate": 6.682117582546132e-07, + "loss": 0.968, + "step": 36862 + }, + { + "epoch": 0.89, + "learning_rate": 6.679318422482761e-07, + "loss": 1.2343, + "step": 36863 + }, + { + "epoch": 0.89, + "learning_rate": 6.676519828574036e-07, + "loss": 0.8515, + "step": 36864 + }, + { + "epoch": 0.89, + "learning_rate": 6.673721800836897e-07, + "loss": 1.2776, + "step": 36865 + }, + { + "epoch": 0.89, + "learning_rate": 6.670924339288332e-07, + "loss": 1.0829, + "step": 36866 + }, + { + "epoch": 0.89, + "learning_rate": 6.668127443945338e-07, + "loss": 1.0341, + "step": 36867 + }, + { + "epoch": 0.89, + "learning_rate": 6.665331114824836e-07, + "loss": 0.8119, + "step": 36868 + }, + { + "epoch": 0.89, + "learning_rate": 6.662535351943822e-07, + "loss": 0.9856, + "step": 36869 + }, + { + "epoch": 0.89, + "learning_rate": 6.659740155319272e-07, + "loss": 0.8391, + "step": 36870 + }, + { + "epoch": 0.89, + "learning_rate": 6.656945524968106e-07, + "loss": 1.3376, + "step": 36871 + }, + { + "epoch": 0.89, + "learning_rate": 6.654151460907288e-07, + "loss": 1.1272, + "step": 36872 + }, + { + "epoch": 0.89, + "learning_rate": 6.651357963153804e-07, + "loss": 1.1982, + "step": 36873 + }, + { + "epoch": 0.89, + "learning_rate": 6.648565031724552e-07, + "loss": 0.9427, + "step": 36874 + }, + { + "epoch": 0.89, + "learning_rate": 6.645772666636519e-07, + "loss": 1.1302, + "step": 36875 + }, + { + "epoch": 0.89, + "learning_rate": 6.642980867906601e-07, + "loss": 0.8749, + "step": 36876 + }, + { + "epoch": 0.89, + "learning_rate": 6.640189635551764e-07, + "loss": 1.1978, + "step": 36877 + }, + { + "epoch": 0.89, + "learning_rate": 6.637398969588949e-07, + "loss": 0.8045, + "step": 36878 + }, + { + "epoch": 0.89, + "learning_rate": 6.634608870035064e-07, + "loss": 1.1534, + "step": 36879 + }, + { + "epoch": 0.89, + "learning_rate": 6.631819336907042e-07, + "loss": 1.1042, + "step": 36880 + }, + { + "epoch": 0.89, + "learning_rate": 6.629030370221834e-07, + "loss": 1.2383, + "step": 36881 + }, + { + "epoch": 0.89, + "learning_rate": 6.626241969996305e-07, + "loss": 1.1311, + "step": 36882 + }, + { + "epoch": 0.89, + "learning_rate": 6.623454136247431e-07, + "loss": 0.963, + "step": 36883 + }, + { + "epoch": 0.89, + "learning_rate": 6.620666868992076e-07, + "loss": 0.9977, + "step": 36884 + }, + { + "epoch": 0.89, + "learning_rate": 6.617880168247182e-07, + "loss": 1.0621, + "step": 36885 + }, + { + "epoch": 0.89, + "learning_rate": 6.615094034029645e-07, + "loss": 0.9366, + "step": 36886 + }, + { + "epoch": 0.89, + "learning_rate": 6.612308466356365e-07, + "loss": 1.0291, + "step": 36887 + }, + { + "epoch": 0.89, + "learning_rate": 6.609523465244239e-07, + "loss": 1.0262, + "step": 36888 + }, + { + "epoch": 0.89, + "learning_rate": 6.606739030710174e-07, + "loss": 0.7803, + "step": 36889 + }, + { + "epoch": 0.89, + "learning_rate": 6.603955162771047e-07, + "loss": 0.9219, + "step": 36890 + }, + { + "epoch": 0.89, + "learning_rate": 6.601171861443778e-07, + "loss": 1.0688, + "step": 36891 + }, + { + "epoch": 0.89, + "learning_rate": 6.598389126745209e-07, + "loss": 0.8057, + "step": 36892 + }, + { + "epoch": 0.89, + "learning_rate": 6.595606958692236e-07, + "loss": 1.1353, + "step": 36893 + }, + { + "epoch": 0.89, + "learning_rate": 6.592825357301769e-07, + "loss": 1.0306, + "step": 36894 + }, + { + "epoch": 0.89, + "learning_rate": 6.59004432259065e-07, + "loss": 1.034, + "step": 36895 + }, + { + "epoch": 0.89, + "learning_rate": 6.587263854575754e-07, + "loss": 0.8422, + "step": 36896 + }, + { + "epoch": 0.89, + "learning_rate": 6.584483953273968e-07, + "loss": 1.0791, + "step": 36897 + }, + { + "epoch": 0.89, + "learning_rate": 6.581704618702134e-07, + "loss": 1.0431, + "step": 36898 + }, + { + "epoch": 0.89, + "learning_rate": 6.578925850877138e-07, + "loss": 0.8568, + "step": 36899 + }, + { + "epoch": 0.89, + "learning_rate": 6.576147649815812e-07, + "loss": 1.0556, + "step": 36900 + }, + { + "epoch": 0.89, + "learning_rate": 6.573370015535018e-07, + "loss": 1.2672, + "step": 36901 + }, + { + "epoch": 0.89, + "learning_rate": 6.570592948051635e-07, + "loss": 1.1559, + "step": 36902 + }, + { + "epoch": 0.89, + "learning_rate": 6.56781644738248e-07, + "loss": 0.84, + "step": 36903 + }, + { + "epoch": 0.89, + "learning_rate": 6.565040513544396e-07, + "loss": 1.0168, + "step": 36904 + }, + { + "epoch": 0.89, + "learning_rate": 6.562265146554248e-07, + "loss": 0.9327, + "step": 36905 + }, + { + "epoch": 0.89, + "learning_rate": 6.559490346428843e-07, + "loss": 1.2009, + "step": 36906 + }, + { + "epoch": 0.89, + "learning_rate": 6.556716113185057e-07, + "loss": 1.0541, + "step": 36907 + }, + { + "epoch": 0.89, + "learning_rate": 6.553942446839667e-07, + "loss": 0.85, + "step": 36908 + }, + { + "epoch": 0.89, + "learning_rate": 6.551169347409536e-07, + "loss": 1.0631, + "step": 36909 + }, + { + "epoch": 0.89, + "learning_rate": 6.548396814911495e-07, + "loss": 0.6552, + "step": 36910 + }, + { + "epoch": 0.89, + "learning_rate": 6.54562484936233e-07, + "loss": 1.06, + "step": 36911 + }, + { + "epoch": 0.89, + "learning_rate": 6.542853450778885e-07, + "loss": 1.1952, + "step": 36912 + }, + { + "epoch": 0.89, + "learning_rate": 6.540082619177979e-07, + "loss": 0.9683, + "step": 36913 + }, + { + "epoch": 0.89, + "learning_rate": 6.537312354576387e-07, + "loss": 1.1885, + "step": 36914 + }, + { + "epoch": 0.89, + "learning_rate": 6.534542656990939e-07, + "loss": 0.9992, + "step": 36915 + }, + { + "epoch": 0.89, + "learning_rate": 6.531773526438456e-07, + "loss": 1.2482, + "step": 36916 + }, + { + "epoch": 0.89, + "learning_rate": 6.529004962935703e-07, + "loss": 1.3141, + "step": 36917 + }, + { + "epoch": 0.89, + "learning_rate": 6.5262369664995e-07, + "loss": 1.0267, + "step": 36918 + }, + { + "epoch": 0.89, + "learning_rate": 6.523469537146621e-07, + "loss": 1.0402, + "step": 36919 + }, + { + "epoch": 0.89, + "learning_rate": 6.520702674893864e-07, + "loss": 0.8162, + "step": 36920 + }, + { + "epoch": 0.89, + "learning_rate": 6.517936379758028e-07, + "loss": 0.7695, + "step": 36921 + }, + { + "epoch": 0.89, + "learning_rate": 6.515170651755876e-07, + "loss": 1.1243, + "step": 36922 + }, + { + "epoch": 0.89, + "learning_rate": 6.512405490904184e-07, + "loss": 1.0863, + "step": 36923 + }, + { + "epoch": 0.89, + "learning_rate": 6.50964089721976e-07, + "loss": 1.2735, + "step": 36924 + }, + { + "epoch": 0.89, + "learning_rate": 6.506876870719348e-07, + "loss": 1.1483, + "step": 36925 + }, + { + "epoch": 0.89, + "learning_rate": 6.504113411419722e-07, + "loss": 1.145, + "step": 36926 + }, + { + "epoch": 0.89, + "learning_rate": 6.501350519337646e-07, + "loss": 0.959, + "step": 36927 + }, + { + "epoch": 0.89, + "learning_rate": 6.498588194489874e-07, + "loss": 1.1581, + "step": 36928 + }, + { + "epoch": 0.89, + "learning_rate": 6.495826436893194e-07, + "loss": 1.2618, + "step": 36929 + }, + { + "epoch": 0.89, + "learning_rate": 6.493065246564334e-07, + "loss": 0.9971, + "step": 36930 + }, + { + "epoch": 0.89, + "learning_rate": 6.490304623520038e-07, + "loss": 0.7878, + "step": 36931 + }, + { + "epoch": 0.89, + "learning_rate": 6.487544567777093e-07, + "loss": 1.1375, + "step": 36932 + }, + { + "epoch": 0.89, + "learning_rate": 6.484785079352207e-07, + "loss": 0.9756, + "step": 36933 + }, + { + "epoch": 0.89, + "learning_rate": 6.482026158262144e-07, + "loss": 1.1389, + "step": 36934 + }, + { + "epoch": 0.89, + "learning_rate": 6.479267804523615e-07, + "loss": 1.0355, + "step": 36935 + }, + { + "epoch": 0.89, + "learning_rate": 6.476510018153371e-07, + "loss": 1.0257, + "step": 36936 + }, + { + "epoch": 0.89, + "learning_rate": 6.473752799168164e-07, + "loss": 1.0268, + "step": 36937 + }, + { + "epoch": 0.89, + "learning_rate": 6.470996147584684e-07, + "loss": 1.1196, + "step": 36938 + }, + { + "epoch": 0.89, + "learning_rate": 6.468240063419662e-07, + "loss": 0.7957, + "step": 36939 + }, + { + "epoch": 0.89, + "learning_rate": 6.465484546689849e-07, + "loss": 0.9489, + "step": 36940 + }, + { + "epoch": 0.89, + "learning_rate": 6.462729597411932e-07, + "loss": 0.9776, + "step": 36941 + }, + { + "epoch": 0.89, + "learning_rate": 6.459975215602632e-07, + "loss": 1.324, + "step": 36942 + }, + { + "epoch": 0.89, + "learning_rate": 6.457221401278657e-07, + "loss": 0.9041, + "step": 36943 + }, + { + "epoch": 0.89, + "learning_rate": 6.454468154456717e-07, + "loss": 1.024, + "step": 36944 + }, + { + "epoch": 0.89, + "learning_rate": 6.451715475153531e-07, + "loss": 1.1489, + "step": 36945 + }, + { + "epoch": 0.89, + "learning_rate": 6.448963363385763e-07, + "loss": 0.7832, + "step": 36946 + }, + { + "epoch": 0.89, + "learning_rate": 6.446211819170134e-07, + "loss": 1.1102, + "step": 36947 + }, + { + "epoch": 0.89, + "learning_rate": 6.443460842523341e-07, + "loss": 0.8659, + "step": 36948 + }, + { + "epoch": 0.89, + "learning_rate": 6.440710433462061e-07, + "loss": 1.1528, + "step": 36949 + }, + { + "epoch": 0.89, + "learning_rate": 6.43796059200299e-07, + "loss": 0.8544, + "step": 36950 + }, + { + "epoch": 0.89, + "learning_rate": 6.435211318162793e-07, + "loss": 0.8773, + "step": 36951 + }, + { + "epoch": 0.89, + "learning_rate": 6.432462611958168e-07, + "loss": 0.9897, + "step": 36952 + }, + { + "epoch": 0.89, + "learning_rate": 6.42971447340579e-07, + "loss": 0.8634, + "step": 36953 + }, + { + "epoch": 0.89, + "learning_rate": 6.426966902522313e-07, + "loss": 1.1867, + "step": 36954 + }, + { + "epoch": 0.89, + "learning_rate": 6.424219899324413e-07, + "loss": 1.194, + "step": 36955 + }, + { + "epoch": 0.89, + "learning_rate": 6.421473463828787e-07, + "loss": 0.978, + "step": 36956 + }, + { + "epoch": 0.89, + "learning_rate": 6.418727596052043e-07, + "loss": 0.8163, + "step": 36957 + }, + { + "epoch": 0.89, + "learning_rate": 6.415982296010892e-07, + "loss": 0.9984, + "step": 36958 + }, + { + "epoch": 0.89, + "learning_rate": 6.413237563721941e-07, + "loss": 1.1409, + "step": 36959 + }, + { + "epoch": 0.89, + "learning_rate": 6.410493399201878e-07, + "loss": 1.0668, + "step": 36960 + }, + { + "epoch": 0.89, + "learning_rate": 6.407749802467345e-07, + "loss": 0.986, + "step": 36961 + }, + { + "epoch": 0.89, + "learning_rate": 6.40500677353496e-07, + "loss": 1.1309, + "step": 36962 + }, + { + "epoch": 0.89, + "learning_rate": 6.40226431242138e-07, + "loss": 1.0586, + "step": 36963 + }, + { + "epoch": 0.89, + "learning_rate": 6.399522419143278e-07, + "loss": 0.9601, + "step": 36964 + }, + { + "epoch": 0.89, + "learning_rate": 6.39678109371723e-07, + "loss": 1.0842, + "step": 36965 + }, + { + "epoch": 0.89, + "learning_rate": 6.394040336159901e-07, + "loss": 1.1772, + "step": 36966 + }, + { + "epoch": 0.89, + "learning_rate": 6.391300146487922e-07, + "loss": 1.1787, + "step": 36967 + }, + { + "epoch": 0.89, + "learning_rate": 6.388560524717891e-07, + "loss": 0.9784, + "step": 36968 + }, + { + "epoch": 0.89, + "learning_rate": 6.38582147086646e-07, + "loss": 0.8607, + "step": 36969 + }, + { + "epoch": 0.89, + "learning_rate": 6.383082984950217e-07, + "loss": 0.8969, + "step": 36970 + }, + { + "epoch": 0.89, + "learning_rate": 6.380345066985794e-07, + "loss": 1.3203, + "step": 36971 + }, + { + "epoch": 0.89, + "learning_rate": 6.377607716989809e-07, + "loss": 1.166, + "step": 36972 + }, + { + "epoch": 0.89, + "learning_rate": 6.374870934978839e-07, + "loss": 1.0258, + "step": 36973 + }, + { + "epoch": 0.89, + "learning_rate": 6.372134720969514e-07, + "loss": 0.8947, + "step": 36974 + }, + { + "epoch": 0.89, + "learning_rate": 6.369399074978433e-07, + "loss": 0.938, + "step": 36975 + }, + { + "epoch": 0.89, + "learning_rate": 6.366663997022171e-07, + "loss": 1.2628, + "step": 36976 + }, + { + "epoch": 0.89, + "learning_rate": 6.363929487117349e-07, + "loss": 1.263, + "step": 36977 + }, + { + "epoch": 0.89, + "learning_rate": 6.361195545280518e-07, + "loss": 1.0518, + "step": 36978 + }, + { + "epoch": 0.89, + "learning_rate": 6.358462171528301e-07, + "loss": 1.1008, + "step": 36979 + }, + { + "epoch": 0.89, + "learning_rate": 6.355729365877283e-07, + "loss": 1.2892, + "step": 36980 + }, + { + "epoch": 0.89, + "learning_rate": 6.352997128344007e-07, + "loss": 0.999, + "step": 36981 + }, + { + "epoch": 0.89, + "learning_rate": 6.35026545894507e-07, + "loss": 0.9137, + "step": 36982 + }, + { + "epoch": 0.89, + "learning_rate": 6.34753435769706e-07, + "loss": 1.1114, + "step": 36983 + }, + { + "epoch": 0.89, + "learning_rate": 6.344803824616519e-07, + "loss": 0.9611, + "step": 36984 + }, + { + "epoch": 0.89, + "learning_rate": 6.342073859720044e-07, + "loss": 1.1262, + "step": 36985 + }, + { + "epoch": 0.89, + "learning_rate": 6.339344463024155e-07, + "loss": 0.8953, + "step": 36986 + }, + { + "epoch": 0.89, + "learning_rate": 6.336615634545428e-07, + "loss": 1.2832, + "step": 36987 + }, + { + "epoch": 0.89, + "learning_rate": 6.333887374300451e-07, + "loss": 1.0946, + "step": 36988 + }, + { + "epoch": 0.89, + "learning_rate": 6.33115968230572e-07, + "loss": 1.1138, + "step": 36989 + }, + { + "epoch": 0.89, + "learning_rate": 6.328432558577824e-07, + "loss": 0.7182, + "step": 36990 + }, + { + "epoch": 0.89, + "learning_rate": 6.325706003133303e-07, + "loss": 1.2359, + "step": 36991 + }, + { + "epoch": 0.89, + "learning_rate": 6.322980015988667e-07, + "loss": 0.9039, + "step": 36992 + }, + { + "epoch": 0.89, + "learning_rate": 6.320254597160502e-07, + "loss": 1.0407, + "step": 36993 + }, + { + "epoch": 0.89, + "learning_rate": 6.317529746665308e-07, + "loss": 0.9528, + "step": 36994 + }, + { + "epoch": 0.89, + "learning_rate": 6.314805464519613e-07, + "loss": 1.1254, + "step": 36995 + }, + { + "epoch": 0.89, + "learning_rate": 6.312081750739973e-07, + "loss": 1.0257, + "step": 36996 + }, + { + "epoch": 0.89, + "learning_rate": 6.309358605342886e-07, + "loss": 0.8341, + "step": 36997 + }, + { + "epoch": 0.89, + "learning_rate": 6.306636028344881e-07, + "loss": 1.2797, + "step": 36998 + }, + { + "epoch": 0.89, + "learning_rate": 6.30391401976248e-07, + "loss": 0.9673, + "step": 36999 + }, + { + "epoch": 0.89, + "learning_rate": 6.301192579612192e-07, + "loss": 0.9932, + "step": 37000 + }, + { + "epoch": 0.89, + "learning_rate": 6.298471707910548e-07, + "loss": 0.8575, + "step": 37001 + }, + { + "epoch": 0.89, + "learning_rate": 6.295751404674022e-07, + "loss": 0.9504, + "step": 37002 + }, + { + "epoch": 0.89, + "learning_rate": 6.293031669919125e-07, + "loss": 0.8299, + "step": 37003 + }, + { + "epoch": 0.89, + "learning_rate": 6.290312503662377e-07, + "loss": 0.9364, + "step": 37004 + }, + { + "epoch": 0.89, + "learning_rate": 6.287593905920253e-07, + "loss": 0.7082, + "step": 37005 + }, + { + "epoch": 0.89, + "learning_rate": 6.284875876709274e-07, + "loss": 0.7945, + "step": 37006 + }, + { + "epoch": 0.89, + "learning_rate": 6.28215841604588e-07, + "loss": 1.0453, + "step": 37007 + }, + { + "epoch": 0.89, + "learning_rate": 6.279441523946595e-07, + "loss": 1.0163, + "step": 37008 + }, + { + "epoch": 0.89, + "learning_rate": 6.276725200427914e-07, + "loss": 1.0179, + "step": 37009 + }, + { + "epoch": 0.89, + "learning_rate": 6.274009445506268e-07, + "loss": 1.1531, + "step": 37010 + }, + { + "epoch": 0.89, + "learning_rate": 6.271294259198168e-07, + "loss": 0.8949, + "step": 37011 + }, + { + "epoch": 0.89, + "learning_rate": 6.268579641520101e-07, + "loss": 0.9739, + "step": 37012 + }, + { + "epoch": 0.89, + "learning_rate": 6.265865592488496e-07, + "loss": 1.1783, + "step": 37013 + }, + { + "epoch": 0.89, + "learning_rate": 6.263152112119841e-07, + "loss": 1.0249, + "step": 37014 + }, + { + "epoch": 0.89, + "learning_rate": 6.26043920043059e-07, + "loss": 1.0757, + "step": 37015 + }, + { + "epoch": 0.89, + "learning_rate": 6.257726857437208e-07, + "loss": 1.2668, + "step": 37016 + }, + { + "epoch": 0.89, + "learning_rate": 6.255015083156158e-07, + "loss": 1.1368, + "step": 37017 + }, + { + "epoch": 0.89, + "learning_rate": 6.25230387760386e-07, + "loss": 1.1178, + "step": 37018 + }, + { + "epoch": 0.89, + "learning_rate": 6.249593240796792e-07, + "loss": 0.8749, + "step": 37019 + }, + { + "epoch": 0.89, + "learning_rate": 6.246883172751405e-07, + "loss": 1.1454, + "step": 37020 + }, + { + "epoch": 0.89, + "learning_rate": 6.2441736734841e-07, + "loss": 0.8931, + "step": 37021 + }, + { + "epoch": 0.89, + "learning_rate": 6.241464743011372e-07, + "loss": 1.1174, + "step": 37022 + }, + { + "epoch": 0.89, + "learning_rate": 6.238756381349598e-07, + "loss": 0.8889, + "step": 37023 + }, + { + "epoch": 0.89, + "learning_rate": 6.236048588515242e-07, + "loss": 1.0699, + "step": 37024 + }, + { + "epoch": 0.89, + "learning_rate": 6.233341364524737e-07, + "loss": 1.0684, + "step": 37025 + }, + { + "epoch": 0.89, + "learning_rate": 6.230634709394479e-07, + "loss": 0.9673, + "step": 37026 + }, + { + "epoch": 0.89, + "learning_rate": 6.2279286231409e-07, + "loss": 1.2152, + "step": 37027 + }, + { + "epoch": 0.89, + "learning_rate": 6.225223105780442e-07, + "loss": 0.985, + "step": 37028 + }, + { + "epoch": 0.89, + "learning_rate": 6.222518157329483e-07, + "loss": 1.0776, + "step": 37029 + }, + { + "epoch": 0.89, + "learning_rate": 6.219813777804451e-07, + "loss": 1.071, + "step": 37030 + }, + { + "epoch": 0.89, + "learning_rate": 6.217109967221746e-07, + "loss": 1.0016, + "step": 37031 + }, + { + "epoch": 0.89, + "learning_rate": 6.214406725597776e-07, + "loss": 0.7814, + "step": 37032 + }, + { + "epoch": 0.89, + "learning_rate": 6.211704052948964e-07, + "loss": 0.9182, + "step": 37033 + }, + { + "epoch": 0.89, + "learning_rate": 6.209001949291659e-07, + "loss": 1.0861, + "step": 37034 + }, + { + "epoch": 0.89, + "learning_rate": 6.206300414642275e-07, + "loss": 1.1726, + "step": 37035 + }, + { + "epoch": 0.89, + "learning_rate": 6.203599449017228e-07, + "loss": 1.2562, + "step": 37036 + }, + { + "epoch": 0.89, + "learning_rate": 6.200899052432863e-07, + "loss": 0.7357, + "step": 37037 + }, + { + "epoch": 0.89, + "learning_rate": 6.198199224905587e-07, + "loss": 1.2015, + "step": 37038 + }, + { + "epoch": 0.89, + "learning_rate": 6.195499966451768e-07, + "loss": 0.8441, + "step": 37039 + }, + { + "epoch": 0.89, + "learning_rate": 6.192801277087779e-07, + "loss": 0.9278, + "step": 37040 + }, + { + "epoch": 0.89, + "learning_rate": 6.19010315683003e-07, + "loss": 1.0675, + "step": 37041 + }, + { + "epoch": 0.89, + "learning_rate": 6.18740560569483e-07, + "loss": 1.3004, + "step": 37042 + }, + { + "epoch": 0.89, + "learning_rate": 6.184708623698576e-07, + "loss": 1.0923, + "step": 37043 + }, + { + "epoch": 0.89, + "learning_rate": 6.182012210857647e-07, + "loss": 0.8686, + "step": 37044 + }, + { + "epoch": 0.89, + "learning_rate": 6.179316367188371e-07, + "loss": 1.1878, + "step": 37045 + }, + { + "epoch": 0.89, + "learning_rate": 6.176621092707103e-07, + "loss": 1.0936, + "step": 37046 + }, + { + "epoch": 0.89, + "learning_rate": 6.173926387430229e-07, + "loss": 1.0157, + "step": 37047 + }, + { + "epoch": 0.89, + "learning_rate": 6.17123225137406e-07, + "loss": 0.9403, + "step": 37048 + }, + { + "epoch": 0.89, + "learning_rate": 6.16853868455497e-07, + "loss": 1.1287, + "step": 37049 + }, + { + "epoch": 0.89, + "learning_rate": 6.165845686989269e-07, + "loss": 1.0233, + "step": 37050 + }, + { + "epoch": 0.89, + "learning_rate": 6.16315325869331e-07, + "loss": 0.928, + "step": 37051 + }, + { + "epoch": 0.89, + "learning_rate": 6.160461399683437e-07, + "loss": 1.191, + "step": 37052 + }, + { + "epoch": 0.89, + "learning_rate": 6.157770109975969e-07, + "loss": 1.0865, + "step": 37053 + }, + { + "epoch": 0.89, + "learning_rate": 6.155079389587227e-07, + "loss": 1.1226, + "step": 37054 + }, + { + "epoch": 0.89, + "learning_rate": 6.152389238533563e-07, + "loss": 0.9855, + "step": 37055 + }, + { + "epoch": 0.89, + "learning_rate": 6.149699656831265e-07, + "loss": 1.0509, + "step": 37056 + }, + { + "epoch": 0.89, + "learning_rate": 6.147010644496687e-07, + "loss": 1.0295, + "step": 37057 + }, + { + "epoch": 0.89, + "learning_rate": 6.144322201546093e-07, + "loss": 1.1171, + "step": 37058 + }, + { + "epoch": 0.89, + "learning_rate": 6.141634327995827e-07, + "loss": 0.7385, + "step": 37059 + }, + { + "epoch": 0.89, + "learning_rate": 6.138947023862208e-07, + "loss": 0.8608, + "step": 37060 + }, + { + "epoch": 0.89, + "learning_rate": 6.136260289161488e-07, + "loss": 0.9352, + "step": 37061 + }, + { + "epoch": 0.89, + "learning_rate": 6.133574123910013e-07, + "loss": 1.3294, + "step": 37062 + }, + { + "epoch": 0.89, + "learning_rate": 6.130888528124079e-07, + "loss": 0.9777, + "step": 37063 + }, + { + "epoch": 0.89, + "learning_rate": 6.12820350181994e-07, + "loss": 0.8055, + "step": 37064 + }, + { + "epoch": 0.89, + "learning_rate": 6.125519045013939e-07, + "loss": 0.9067, + "step": 37065 + }, + { + "epoch": 0.89, + "learning_rate": 6.122835157722307e-07, + "loss": 1.005, + "step": 37066 + }, + { + "epoch": 0.89, + "learning_rate": 6.120151839961363e-07, + "loss": 1.0108, + "step": 37067 + }, + { + "epoch": 0.89, + "learning_rate": 6.117469091747386e-07, + "loss": 0.9909, + "step": 37068 + }, + { + "epoch": 0.89, + "learning_rate": 6.114786913096626e-07, + "loss": 1.0423, + "step": 37069 + }, + { + "epoch": 0.89, + "learning_rate": 6.112105304025373e-07, + "loss": 1.0843, + "step": 37070 + }, + { + "epoch": 0.89, + "learning_rate": 6.109424264549901e-07, + "loss": 1.1399, + "step": 37071 + }, + { + "epoch": 0.89, + "learning_rate": 6.106743794686454e-07, + "loss": 0.9919, + "step": 37072 + }, + { + "epoch": 0.89, + "learning_rate": 6.104063894451329e-07, + "loss": 0.908, + "step": 37073 + }, + { + "epoch": 0.89, + "learning_rate": 6.101384563860746e-07, + "loss": 1.0145, + "step": 37074 + }, + { + "epoch": 0.89, + "learning_rate": 6.098705802930971e-07, + "loss": 0.9736, + "step": 37075 + }, + { + "epoch": 0.89, + "learning_rate": 6.09602761167829e-07, + "loss": 0.9501, + "step": 37076 + }, + { + "epoch": 0.89, + "learning_rate": 6.093349990118891e-07, + "loss": 0.772, + "step": 37077 + }, + { + "epoch": 0.89, + "learning_rate": 6.09067293826906e-07, + "loss": 0.9787, + "step": 37078 + }, + { + "epoch": 0.89, + "learning_rate": 6.08799645614504e-07, + "loss": 0.965, + "step": 37079 + }, + { + "epoch": 0.89, + "learning_rate": 6.08532054376303e-07, + "loss": 1.0348, + "step": 37080 + }, + { + "epoch": 0.89, + "learning_rate": 6.082645201139314e-07, + "loss": 0.9925, + "step": 37081 + }, + { + "epoch": 0.89, + "learning_rate": 6.079970428290083e-07, + "loss": 0.9877, + "step": 37082 + }, + { + "epoch": 0.89, + "learning_rate": 6.077296225231588e-07, + "loss": 0.959, + "step": 37083 + }, + { + "epoch": 0.89, + "learning_rate": 6.074622591980051e-07, + "loss": 1.1181, + "step": 37084 + }, + { + "epoch": 0.89, + "learning_rate": 6.071949528551668e-07, + "loss": 0.9871, + "step": 37085 + }, + { + "epoch": 0.89, + "learning_rate": 6.069277034962684e-07, + "loss": 0.9503, + "step": 37086 + }, + { + "epoch": 0.89, + "learning_rate": 6.066605111229318e-07, + "loss": 0.703, + "step": 37087 + }, + { + "epoch": 0.89, + "learning_rate": 6.063933757367734e-07, + "loss": 1.1317, + "step": 37088 + }, + { + "epoch": 0.89, + "learning_rate": 6.061262973394188e-07, + "loss": 1.1703, + "step": 37089 + }, + { + "epoch": 0.89, + "learning_rate": 6.058592759324866e-07, + "loss": 1.2655, + "step": 37090 + }, + { + "epoch": 0.89, + "learning_rate": 6.055923115175955e-07, + "loss": 0.8862, + "step": 37091 + }, + { + "epoch": 0.89, + "learning_rate": 6.053254040963675e-07, + "loss": 1.0733, + "step": 37092 + }, + { + "epoch": 0.89, + "learning_rate": 6.05058553670419e-07, + "loss": 1.1568, + "step": 37093 + }, + { + "epoch": 0.89, + "learning_rate": 6.0479176024137e-07, + "loss": 1.1444, + "step": 37094 + }, + { + "epoch": 0.89, + "learning_rate": 6.045250238108414e-07, + "loss": 0.9489, + "step": 37095 + }, + { + "epoch": 0.89, + "learning_rate": 6.042583443804484e-07, + "loss": 1.1055, + "step": 37096 + }, + { + "epoch": 0.89, + "learning_rate": 6.039917219518099e-07, + "loss": 1.0659, + "step": 37097 + }, + { + "epoch": 0.89, + "learning_rate": 6.037251565265445e-07, + "loss": 0.9246, + "step": 37098 + }, + { + "epoch": 0.89, + "learning_rate": 6.034586481062676e-07, + "loss": 1.2954, + "step": 37099 + }, + { + "epoch": 0.89, + "learning_rate": 6.031921966925979e-07, + "loss": 1.0688, + "step": 37100 + }, + { + "epoch": 0.89, + "learning_rate": 6.029258022871498e-07, + "loss": 1.1549, + "step": 37101 + }, + { + "epoch": 0.89, + "learning_rate": 6.026594648915396e-07, + "loss": 1.0976, + "step": 37102 + }, + { + "epoch": 0.89, + "learning_rate": 6.023931845073871e-07, + "loss": 1.0117, + "step": 37103 + }, + { + "epoch": 0.89, + "learning_rate": 6.021269611363023e-07, + "loss": 0.9221, + "step": 37104 + }, + { + "epoch": 0.89, + "learning_rate": 6.018607947799027e-07, + "loss": 0.9917, + "step": 37105 + }, + { + "epoch": 0.89, + "learning_rate": 6.015946854398047e-07, + "loss": 1.1944, + "step": 37106 + }, + { + "epoch": 0.89, + "learning_rate": 6.013286331176194e-07, + "loss": 1.1267, + "step": 37107 + }, + { + "epoch": 0.89, + "learning_rate": 6.010626378149642e-07, + "loss": 0.7239, + "step": 37108 + }, + { + "epoch": 0.89, + "learning_rate": 6.007966995334491e-07, + "loss": 0.9164, + "step": 37109 + }, + { + "epoch": 0.89, + "learning_rate": 6.005308182746906e-07, + "loss": 0.8596, + "step": 37110 + }, + { + "epoch": 0.89, + "learning_rate": 6.002649940403016e-07, + "loss": 1.0696, + "step": 37111 + }, + { + "epoch": 0.89, + "learning_rate": 5.999992268318933e-07, + "loss": 0.9944, + "step": 37112 + }, + { + "epoch": 0.89, + "learning_rate": 5.997335166510776e-07, + "loss": 1.1374, + "step": 37113 + }, + { + "epoch": 0.89, + "learning_rate": 5.994678634994688e-07, + "loss": 1.1694, + "step": 37114 + }, + { + "epoch": 0.89, + "learning_rate": 5.992022673786769e-07, + "loss": 0.8269, + "step": 37115 + }, + { + "epoch": 0.89, + "learning_rate": 5.989367282903146e-07, + "loss": 1.0486, + "step": 37116 + }, + { + "epoch": 0.89, + "learning_rate": 5.986712462359912e-07, + "loss": 1.0014, + "step": 37117 + }, + { + "epoch": 0.89, + "learning_rate": 5.984058212173172e-07, + "loss": 1.0341, + "step": 37118 + }, + { + "epoch": 0.89, + "learning_rate": 5.981404532359059e-07, + "loss": 0.9979, + "step": 37119 + }, + { + "epoch": 0.89, + "learning_rate": 5.978751422933626e-07, + "loss": 1.168, + "step": 37120 + }, + { + "epoch": 0.89, + "learning_rate": 5.976098883912996e-07, + "loss": 1.2119, + "step": 37121 + }, + { + "epoch": 0.89, + "learning_rate": 5.973446915313275e-07, + "loss": 0.8395, + "step": 37122 + }, + { + "epoch": 0.89, + "learning_rate": 5.970795517150519e-07, + "loss": 1.098, + "step": 37123 + }, + { + "epoch": 0.89, + "learning_rate": 5.968144689440847e-07, + "loss": 0.9896, + "step": 37124 + }, + { + "epoch": 0.89, + "learning_rate": 5.965494432200303e-07, + "loss": 1.1317, + "step": 37125 + }, + { + "epoch": 0.89, + "learning_rate": 5.962844745444996e-07, + "loss": 1.1543, + "step": 37126 + }, + { + "epoch": 0.89, + "learning_rate": 5.960195629191001e-07, + "loss": 1.1668, + "step": 37127 + }, + { + "epoch": 0.89, + "learning_rate": 5.957547083454362e-07, + "loss": 1.0385, + "step": 37128 + }, + { + "epoch": 0.89, + "learning_rate": 5.954899108251167e-07, + "loss": 1.0406, + "step": 37129 + }, + { + "epoch": 0.89, + "learning_rate": 5.952251703597489e-07, + "loss": 0.9583, + "step": 37130 + }, + { + "epoch": 0.89, + "learning_rate": 5.949604869509362e-07, + "loss": 1.0215, + "step": 37131 + }, + { + "epoch": 0.89, + "learning_rate": 5.946958606002883e-07, + "loss": 1.067, + "step": 37132 + }, + { + "epoch": 0.89, + "learning_rate": 5.944312913094064e-07, + "loss": 1.0577, + "step": 37133 + }, + { + "epoch": 0.89, + "learning_rate": 5.941667790798977e-07, + "loss": 1.0465, + "step": 37134 + }, + { + "epoch": 0.89, + "learning_rate": 5.93902323913369e-07, + "loss": 1.1031, + "step": 37135 + }, + { + "epoch": 0.89, + "learning_rate": 5.9363792581142e-07, + "loss": 0.7834, + "step": 37136 + }, + { + "epoch": 0.89, + "learning_rate": 5.933735847756572e-07, + "loss": 1.1412, + "step": 37137 + }, + { + "epoch": 0.89, + "learning_rate": 5.93109300807686e-07, + "loss": 0.9914, + "step": 37138 + }, + { + "epoch": 0.89, + "learning_rate": 5.928450739091063e-07, + "loss": 1.213, + "step": 37139 + }, + { + "epoch": 0.89, + "learning_rate": 5.925809040815233e-07, + "loss": 0.8598, + "step": 37140 + }, + { + "epoch": 0.89, + "learning_rate": 5.923167913265404e-07, + "loss": 0.8537, + "step": 37141 + }, + { + "epoch": 0.89, + "learning_rate": 5.920527356457573e-07, + "loss": 0.9523, + "step": 37142 + }, + { + "epoch": 0.89, + "learning_rate": 5.917887370407793e-07, + "loss": 1.0217, + "step": 37143 + }, + { + "epoch": 0.89, + "learning_rate": 5.915247955132042e-07, + "loss": 0.7718, + "step": 37144 + }, + { + "epoch": 0.89, + "learning_rate": 5.91260911064635e-07, + "loss": 1.2978, + "step": 37145 + }, + { + "epoch": 0.89, + "learning_rate": 5.909970836966738e-07, + "loss": 0.7296, + "step": 37146 + }, + { + "epoch": 0.89, + "learning_rate": 5.907333134109194e-07, + "loss": 0.9599, + "step": 37147 + }, + { + "epoch": 0.89, + "learning_rate": 5.904696002089727e-07, + "loss": 0.9649, + "step": 37148 + }, + { + "epoch": 0.89, + "learning_rate": 5.902059440924346e-07, + "loss": 0.9782, + "step": 37149 + }, + { + "epoch": 0.89, + "learning_rate": 5.899423450629027e-07, + "loss": 0.8382, + "step": 37150 + }, + { + "epoch": 0.89, + "learning_rate": 5.896788031219791e-07, + "loss": 1.2028, + "step": 37151 + }, + { + "epoch": 0.89, + "learning_rate": 5.894153182712592e-07, + "loss": 1.0014, + "step": 37152 + }, + { + "epoch": 0.89, + "learning_rate": 5.891518905123428e-07, + "loss": 1.0625, + "step": 37153 + }, + { + "epoch": 0.89, + "learning_rate": 5.888885198468297e-07, + "loss": 1.0993, + "step": 37154 + }, + { + "epoch": 0.89, + "learning_rate": 5.886252062763142e-07, + "loss": 1.1363, + "step": 37155 + }, + { + "epoch": 0.89, + "learning_rate": 5.883619498023974e-07, + "loss": 1.0168, + "step": 37156 + }, + { + "epoch": 0.89, + "learning_rate": 5.880987504266755e-07, + "loss": 0.9493, + "step": 37157 + }, + { + "epoch": 0.89, + "learning_rate": 5.878356081507431e-07, + "loss": 0.7495, + "step": 37158 + }, + { + "epoch": 0.89, + "learning_rate": 5.875725229761997e-07, + "loss": 1.1493, + "step": 37159 + }, + { + "epoch": 0.89, + "learning_rate": 5.873094949046387e-07, + "loss": 0.919, + "step": 37160 + }, + { + "epoch": 0.89, + "learning_rate": 5.870465239376566e-07, + "loss": 1.1588, + "step": 37161 + }, + { + "epoch": 0.89, + "learning_rate": 5.86783610076851e-07, + "loss": 1.0231, + "step": 37162 + }, + { + "epoch": 0.89, + "learning_rate": 5.865207533238137e-07, + "loss": 1.195, + "step": 37163 + }, + { + "epoch": 0.89, + "learning_rate": 5.862579536801405e-07, + "loss": 1.0343, + "step": 37164 + }, + { + "epoch": 0.89, + "learning_rate": 5.859952111474276e-07, + "loss": 0.7528, + "step": 37165 + }, + { + "epoch": 0.89, + "learning_rate": 5.85732525727264e-07, + "loss": 1.0882, + "step": 37166 + }, + { + "epoch": 0.89, + "learning_rate": 5.854698974212502e-07, + "loss": 0.8955, + "step": 37167 + }, + { + "epoch": 0.89, + "learning_rate": 5.852073262309754e-07, + "loss": 1.003, + "step": 37168 + }, + { + "epoch": 0.89, + "learning_rate": 5.849448121580326e-07, + "loss": 1.2185, + "step": 37169 + }, + { + "epoch": 0.89, + "learning_rate": 5.84682355204016e-07, + "loss": 1.0042, + "step": 37170 + }, + { + "epoch": 0.89, + "learning_rate": 5.844199553705154e-07, + "loss": 1.1608, + "step": 37171 + }, + { + "epoch": 0.89, + "learning_rate": 5.841576126591264e-07, + "loss": 0.9659, + "step": 37172 + }, + { + "epoch": 0.89, + "learning_rate": 5.838953270714376e-07, + "loss": 1.1436, + "step": 37173 + }, + { + "epoch": 0.89, + "learning_rate": 5.8363309860904e-07, + "loss": 0.9612, + "step": 37174 + }, + { + "epoch": 0.89, + "learning_rate": 5.833709272735266e-07, + "loss": 1.0513, + "step": 37175 + }, + { + "epoch": 0.89, + "learning_rate": 5.831088130664863e-07, + "loss": 1.1136, + "step": 37176 + }, + { + "epoch": 0.89, + "learning_rate": 5.828467559895102e-07, + "loss": 1.0366, + "step": 37177 + }, + { + "epoch": 0.89, + "learning_rate": 5.825847560441889e-07, + "loss": 1.1142, + "step": 37178 + }, + { + "epoch": 0.89, + "learning_rate": 5.82322813232108e-07, + "loss": 1.0489, + "step": 37179 + }, + { + "epoch": 0.89, + "learning_rate": 5.820609275548627e-07, + "loss": 0.845, + "step": 37180 + }, + { + "epoch": 0.89, + "learning_rate": 5.817990990140365e-07, + "loss": 1.1677, + "step": 37181 + }, + { + "epoch": 0.89, + "learning_rate": 5.81537327611219e-07, + "loss": 1.1368, + "step": 37182 + }, + { + "epoch": 0.89, + "learning_rate": 5.812756133480013e-07, + "loss": 0.8768, + "step": 37183 + }, + { + "epoch": 0.89, + "learning_rate": 5.810139562259676e-07, + "loss": 1.044, + "step": 37184 + }, + { + "epoch": 0.89, + "learning_rate": 5.807523562467066e-07, + "loss": 1.103, + "step": 37185 + }, + { + "epoch": 0.89, + "learning_rate": 5.80490813411807e-07, + "loss": 1.1891, + "step": 37186 + }, + { + "epoch": 0.89, + "learning_rate": 5.802293277228532e-07, + "loss": 1.0834, + "step": 37187 + }, + { + "epoch": 0.89, + "learning_rate": 5.799678991814328e-07, + "loss": 1.1974, + "step": 37188 + }, + { + "epoch": 0.89, + "learning_rate": 5.797065277891311e-07, + "loss": 0.874, + "step": 37189 + }, + { + "epoch": 0.89, + "learning_rate": 5.794452135475326e-07, + "loss": 1.0911, + "step": 37190 + }, + { + "epoch": 0.89, + "learning_rate": 5.791839564582269e-07, + "loss": 0.7912, + "step": 37191 + }, + { + "epoch": 0.89, + "learning_rate": 5.78922756522794e-07, + "loss": 1.0151, + "step": 37192 + }, + { + "epoch": 0.89, + "learning_rate": 5.786616137428214e-07, + "loss": 1.1912, + "step": 37193 + }, + { + "epoch": 0.89, + "learning_rate": 5.784005281198946e-07, + "loss": 0.7994, + "step": 37194 + }, + { + "epoch": 0.89, + "learning_rate": 5.781394996555934e-07, + "loss": 1.0598, + "step": 37195 + }, + { + "epoch": 0.89, + "learning_rate": 5.778785283515054e-07, + "loss": 0.8431, + "step": 37196 + }, + { + "epoch": 0.89, + "learning_rate": 5.776176142092105e-07, + "loss": 1.159, + "step": 37197 + }, + { + "epoch": 0.89, + "learning_rate": 5.773567572302929e-07, + "loss": 0.9373, + "step": 37198 + }, + { + "epoch": 0.89, + "learning_rate": 5.770959574163382e-07, + "loss": 1.0267, + "step": 37199 + }, + { + "epoch": 0.89, + "learning_rate": 5.768352147689227e-07, + "loss": 1.0149, + "step": 37200 + }, + { + "epoch": 0.89, + "learning_rate": 5.76574529289633e-07, + "loss": 1.1267, + "step": 37201 + }, + { + "epoch": 0.89, + "learning_rate": 5.763139009800501e-07, + "loss": 1.0202, + "step": 37202 + }, + { + "epoch": 0.89, + "learning_rate": 5.760533298417525e-07, + "loss": 0.9682, + "step": 37203 + }, + { + "epoch": 0.89, + "learning_rate": 5.757928158763249e-07, + "loss": 0.8898, + "step": 37204 + }, + { + "epoch": 0.89, + "learning_rate": 5.755323590853434e-07, + "loss": 1.027, + "step": 37205 + }, + { + "epoch": 0.89, + "learning_rate": 5.752719594703904e-07, + "loss": 0.9125, + "step": 37206 + }, + { + "epoch": 0.89, + "learning_rate": 5.750116170330467e-07, + "loss": 1.122, + "step": 37207 + }, + { + "epoch": 0.89, + "learning_rate": 5.747513317748898e-07, + "loss": 0.9033, + "step": 37208 + }, + { + "epoch": 0.89, + "learning_rate": 5.744911036974999e-07, + "loss": 0.949, + "step": 37209 + }, + { + "epoch": 0.89, + "learning_rate": 5.742309328024565e-07, + "loss": 1.0678, + "step": 37210 + }, + { + "epoch": 0.89, + "learning_rate": 5.739708190913351e-07, + "loss": 1.1632, + "step": 37211 + }, + { + "epoch": 0.89, + "learning_rate": 5.737107625657157e-07, + "loss": 0.9506, + "step": 37212 + }, + { + "epoch": 0.89, + "learning_rate": 5.734507632271779e-07, + "loss": 1.1005, + "step": 37213 + }, + { + "epoch": 0.89, + "learning_rate": 5.731908210772952e-07, + "loss": 1.1349, + "step": 37214 + }, + { + "epoch": 0.89, + "learning_rate": 5.729309361176483e-07, + "loss": 1.1344, + "step": 37215 + }, + { + "epoch": 0.9, + "learning_rate": 5.726711083498093e-07, + "loss": 1.1017, + "step": 37216 + }, + { + "epoch": 0.9, + "learning_rate": 5.724113377753581e-07, + "loss": 1.0469, + "step": 37217 + }, + { + "epoch": 0.9, + "learning_rate": 5.721516243958725e-07, + "loss": 1.1125, + "step": 37218 + }, + { + "epoch": 0.9, + "learning_rate": 5.71891968212922e-07, + "loss": 1.2521, + "step": 37219 + }, + { + "epoch": 0.9, + "learning_rate": 5.716323692280868e-07, + "loss": 1.1727, + "step": 37220 + }, + { + "epoch": 0.9, + "learning_rate": 5.713728274429409e-07, + "loss": 0.8706, + "step": 37221 + }, + { + "epoch": 0.9, + "learning_rate": 5.711133428590565e-07, + "loss": 1.0719, + "step": 37222 + }, + { + "epoch": 0.9, + "learning_rate": 5.708539154780124e-07, + "loss": 0.8109, + "step": 37223 + }, + { + "epoch": 0.9, + "learning_rate": 5.705945453013773e-07, + "loss": 0.8373, + "step": 37224 + }, + { + "epoch": 0.9, + "learning_rate": 5.703352323307276e-07, + "loss": 0.9658, + "step": 37225 + }, + { + "epoch": 0.9, + "learning_rate": 5.700759765676378e-07, + "loss": 1.0684, + "step": 37226 + }, + { + "epoch": 0.9, + "learning_rate": 5.698167780136765e-07, + "loss": 0.9833, + "step": 37227 + }, + { + "epoch": 0.9, + "learning_rate": 5.695576366704191e-07, + "loss": 1.0346, + "step": 37228 + }, + { + "epoch": 0.9, + "learning_rate": 5.69298552539439e-07, + "loss": 1.2995, + "step": 37229 + }, + { + "epoch": 0.9, + "learning_rate": 5.690395256223047e-07, + "loss": 1.0629, + "step": 37230 + }, + { + "epoch": 0.9, + "learning_rate": 5.687805559205917e-07, + "loss": 1.0322, + "step": 37231 + }, + { + "epoch": 0.9, + "learning_rate": 5.685216434358664e-07, + "loss": 1.2208, + "step": 37232 + }, + { + "epoch": 0.9, + "learning_rate": 5.682627881697012e-07, + "loss": 1.1726, + "step": 37233 + }, + { + "epoch": 0.9, + "learning_rate": 5.680039901236701e-07, + "loss": 0.9817, + "step": 37234 + }, + { + "epoch": 0.9, + "learning_rate": 5.677452492993374e-07, + "loss": 1.0859, + "step": 37235 + }, + { + "epoch": 0.9, + "learning_rate": 5.674865656982764e-07, + "loss": 1.0867, + "step": 37236 + }, + { + "epoch": 0.9, + "learning_rate": 5.67227939322057e-07, + "loss": 1.2058, + "step": 37237 + }, + { + "epoch": 0.9, + "learning_rate": 5.669693701722456e-07, + "loss": 0.8891, + "step": 37238 + }, + { + "epoch": 0.9, + "learning_rate": 5.667108582504144e-07, + "loss": 1.264, + "step": 37239 + }, + { + "epoch": 0.9, + "learning_rate": 5.664524035581276e-07, + "loss": 1.2907, + "step": 37240 + }, + { + "epoch": 0.9, + "learning_rate": 5.66194006096955e-07, + "loss": 1.3484, + "step": 37241 + }, + { + "epoch": 0.9, + "learning_rate": 5.659356658684656e-07, + "loss": 0.9593, + "step": 37242 + }, + { + "epoch": 0.9, + "learning_rate": 5.656773828742246e-07, + "loss": 0.9092, + "step": 37243 + }, + { + "epoch": 0.9, + "learning_rate": 5.654191571157996e-07, + "loss": 0.9095, + "step": 37244 + }, + { + "epoch": 0.9, + "learning_rate": 5.651609885947584e-07, + "loss": 0.9728, + "step": 37245 + }, + { + "epoch": 0.9, + "learning_rate": 5.649028773126652e-07, + "loss": 1.1608, + "step": 37246 + }, + { + "epoch": 0.9, + "learning_rate": 5.646448232710888e-07, + "loss": 1.0454, + "step": 37247 + }, + { + "epoch": 0.9, + "learning_rate": 5.643868264715912e-07, + "loss": 0.9545, + "step": 37248 + }, + { + "epoch": 0.9, + "learning_rate": 5.641288869157391e-07, + "loss": 0.9793, + "step": 37249 + }, + { + "epoch": 0.9, + "learning_rate": 5.638710046050988e-07, + "loss": 0.908, + "step": 37250 + }, + { + "epoch": 0.9, + "learning_rate": 5.636131795412337e-07, + "loss": 1.0406, + "step": 37251 + }, + { + "epoch": 0.9, + "learning_rate": 5.633554117257057e-07, + "loss": 1.3664, + "step": 37252 + }, + { + "epoch": 0.9, + "learning_rate": 5.630977011600836e-07, + "loss": 1.1142, + "step": 37253 + }, + { + "epoch": 0.9, + "learning_rate": 5.628400478459261e-07, + "loss": 1.1557, + "step": 37254 + }, + { + "epoch": 0.9, + "learning_rate": 5.625824517848e-07, + "loss": 1.0097, + "step": 37255 + }, + { + "epoch": 0.9, + "learning_rate": 5.623249129782649e-07, + "loss": 1.1385, + "step": 37256 + }, + { + "epoch": 0.9, + "learning_rate": 5.620674314278852e-07, + "loss": 0.96, + "step": 37257 + }, + { + "epoch": 0.9, + "learning_rate": 5.618100071352228e-07, + "loss": 0.9625, + "step": 37258 + }, + { + "epoch": 0.9, + "learning_rate": 5.615526401018378e-07, + "loss": 1.1979, + "step": 37259 + }, + { + "epoch": 0.9, + "learning_rate": 5.612953303292934e-07, + "loss": 0.8932, + "step": 37260 + }, + { + "epoch": 0.9, + "learning_rate": 5.610380778191515e-07, + "loss": 1.2219, + "step": 37261 + }, + { + "epoch": 0.9, + "learning_rate": 5.607808825729699e-07, + "loss": 1.0166, + "step": 37262 + }, + { + "epoch": 0.9, + "learning_rate": 5.605237445923106e-07, + "loss": 1.0456, + "step": 37263 + }, + { + "epoch": 0.9, + "learning_rate": 5.602666638787346e-07, + "loss": 0.8577, + "step": 37264 + }, + { + "epoch": 0.9, + "learning_rate": 5.600096404337985e-07, + "loss": 1.1778, + "step": 37265 + }, + { + "epoch": 0.9, + "learning_rate": 5.597526742590653e-07, + "loss": 0.8139, + "step": 37266 + }, + { + "epoch": 0.9, + "learning_rate": 5.594957653560906e-07, + "loss": 1.1316, + "step": 37267 + }, + { + "epoch": 0.9, + "learning_rate": 5.592389137264354e-07, + "loss": 0.957, + "step": 37268 + }, + { + "epoch": 0.9, + "learning_rate": 5.589821193716583e-07, + "loss": 1.2333, + "step": 37269 + }, + { + "epoch": 0.9, + "learning_rate": 5.587253822933136e-07, + "loss": 1.1583, + "step": 37270 + }, + { + "epoch": 0.9, + "learning_rate": 5.584687024929625e-07, + "loss": 0.9345, + "step": 37271 + }, + { + "epoch": 0.9, + "learning_rate": 5.582120799721613e-07, + "loss": 1.1413, + "step": 37272 + }, + { + "epoch": 0.9, + "learning_rate": 5.579555147324656e-07, + "loss": 0.9013, + "step": 37273 + }, + { + "epoch": 0.9, + "learning_rate": 5.57699006775434e-07, + "loss": 1.0478, + "step": 37274 + }, + { + "epoch": 0.9, + "learning_rate": 5.574425561026208e-07, + "loss": 0.829, + "step": 37275 + }, + { + "epoch": 0.9, + "learning_rate": 5.571861627155828e-07, + "loss": 0.6321, + "step": 37276 + }, + { + "epoch": 0.9, + "learning_rate": 5.569298266158763e-07, + "loss": 1.1218, + "step": 37277 + }, + { + "epoch": 0.9, + "learning_rate": 5.566735478050533e-07, + "loss": 1.0264, + "step": 37278 + }, + { + "epoch": 0.9, + "learning_rate": 5.564173262846717e-07, + "loss": 1.0119, + "step": 37279 + }, + { + "epoch": 0.9, + "learning_rate": 5.561611620562857e-07, + "loss": 0.9868, + "step": 37280 + }, + { + "epoch": 0.9, + "learning_rate": 5.559050551214473e-07, + "loss": 1.0605, + "step": 37281 + }, + { + "epoch": 0.9, + "learning_rate": 5.556490054817132e-07, + "loss": 0.966, + "step": 37282 + }, + { + "epoch": 0.9, + "learning_rate": 5.553930131386331e-07, + "loss": 0.9175, + "step": 37283 + }, + { + "epoch": 0.9, + "learning_rate": 5.551370780937627e-07, + "loss": 1.0963, + "step": 37284 + }, + { + "epoch": 0.9, + "learning_rate": 5.548812003486548e-07, + "loss": 1.0955, + "step": 37285 + }, + { + "epoch": 0.9, + "learning_rate": 5.546253799048606e-07, + "loss": 1.0414, + "step": 37286 + }, + { + "epoch": 0.9, + "learning_rate": 5.543696167639323e-07, + "loss": 1.0843, + "step": 37287 + }, + { + "epoch": 0.9, + "learning_rate": 5.541139109274229e-07, + "loss": 1.032, + "step": 37288 + }, + { + "epoch": 0.9, + "learning_rate": 5.538582623968813e-07, + "loss": 1.1421, + "step": 37289 + }, + { + "epoch": 0.9, + "learning_rate": 5.536026711738606e-07, + "loss": 1.0552, + "step": 37290 + }, + { + "epoch": 0.9, + "learning_rate": 5.533471372599098e-07, + "loss": 0.9289, + "step": 37291 + }, + { + "epoch": 0.9, + "learning_rate": 5.530916606565795e-07, + "loss": 0.7861, + "step": 37292 + }, + { + "epoch": 0.9, + "learning_rate": 5.528362413654221e-07, + "loss": 0.967, + "step": 37293 + }, + { + "epoch": 0.9, + "learning_rate": 5.525808793879839e-07, + "loss": 1.0427, + "step": 37294 + }, + { + "epoch": 0.9, + "learning_rate": 5.523255747258139e-07, + "loss": 1.0845, + "step": 37295 + }, + { + "epoch": 0.9, + "learning_rate": 5.520703273804651e-07, + "loss": 1.1, + "step": 37296 + }, + { + "epoch": 0.9, + "learning_rate": 5.518151373534819e-07, + "loss": 0.9342, + "step": 37297 + }, + { + "epoch": 0.9, + "learning_rate": 5.515600046464143e-07, + "loss": 0.9442, + "step": 37298 + }, + { + "epoch": 0.9, + "learning_rate": 5.513049292608097e-07, + "loss": 1.0949, + "step": 37299 + }, + { + "epoch": 0.9, + "learning_rate": 5.510499111982148e-07, + "loss": 1.0208, + "step": 37300 + }, + { + "epoch": 0.9, + "learning_rate": 5.507949504601784e-07, + "loss": 1.0316, + "step": 37301 + }, + { + "epoch": 0.9, + "learning_rate": 5.505400470482458e-07, + "loss": 1.2366, + "step": 37302 + }, + { + "epoch": 0.9, + "learning_rate": 5.502852009639647e-07, + "loss": 1.0872, + "step": 37303 + }, + { + "epoch": 0.9, + "learning_rate": 5.500304122088807e-07, + "loss": 0.9172, + "step": 37304 + }, + { + "epoch": 0.9, + "learning_rate": 5.49775680784539e-07, + "loss": 1.0599, + "step": 37305 + }, + { + "epoch": 0.9, + "learning_rate": 5.49521006692485e-07, + "loss": 1.2681, + "step": 37306 + }, + { + "epoch": 0.9, + "learning_rate": 5.492663899342654e-07, + "loss": 1.0358, + "step": 37307 + }, + { + "epoch": 0.9, + "learning_rate": 5.490118305114223e-07, + "loss": 1.3229, + "step": 37308 + }, + { + "epoch": 0.9, + "learning_rate": 5.487573284255032e-07, + "loss": 1.0998, + "step": 37309 + }, + { + "epoch": 0.9, + "learning_rate": 5.485028836780482e-07, + "loss": 0.9149, + "step": 37310 + }, + { + "epoch": 0.9, + "learning_rate": 5.482484962706036e-07, + "loss": 0.9231, + "step": 37311 + }, + { + "epoch": 0.9, + "learning_rate": 5.479941662047139e-07, + "loss": 0.8944, + "step": 37312 + }, + { + "epoch": 0.9, + "learning_rate": 5.477398934819189e-07, + "loss": 0.9877, + "step": 37313 + }, + { + "epoch": 0.9, + "learning_rate": 5.474856781037618e-07, + "loss": 1.1711, + "step": 37314 + }, + { + "epoch": 0.9, + "learning_rate": 5.472315200717882e-07, + "loss": 1.0283, + "step": 37315 + }, + { + "epoch": 0.9, + "learning_rate": 5.469774193875354e-07, + "loss": 1.1322, + "step": 37316 + }, + { + "epoch": 0.9, + "learning_rate": 5.467233760525493e-07, + "loss": 1.0977, + "step": 37317 + }, + { + "epoch": 0.9, + "learning_rate": 5.464693900683659e-07, + "loss": 1.1213, + "step": 37318 + }, + { + "epoch": 0.9, + "learning_rate": 5.462154614365312e-07, + "loss": 1.0335, + "step": 37319 + }, + { + "epoch": 0.9, + "learning_rate": 5.459615901585836e-07, + "loss": 0.9763, + "step": 37320 + }, + { + "epoch": 0.9, + "learning_rate": 5.457077762360619e-07, + "loss": 0.9799, + "step": 37321 + }, + { + "epoch": 0.9, + "learning_rate": 5.454540196705083e-07, + "loss": 1.1536, + "step": 37322 + }, + { + "epoch": 0.9, + "learning_rate": 5.452003204634615e-07, + "loss": 0.9483, + "step": 37323 + }, + { + "epoch": 0.9, + "learning_rate": 5.449466786164603e-07, + "loss": 0.8858, + "step": 37324 + }, + { + "epoch": 0.9, + "learning_rate": 5.446930941310447e-07, + "loss": 1.0348, + "step": 37325 + }, + { + "epoch": 0.9, + "learning_rate": 5.444395670087499e-07, + "loss": 0.9461, + "step": 37326 + }, + { + "epoch": 0.9, + "learning_rate": 5.441860972511171e-07, + "loss": 1.17, + "step": 37327 + }, + { + "epoch": 0.9, + "learning_rate": 5.439326848596838e-07, + "loss": 0.918, + "step": 37328 + }, + { + "epoch": 0.9, + "learning_rate": 5.436793298359855e-07, + "loss": 0.9315, + "step": 37329 + }, + { + "epoch": 0.9, + "learning_rate": 5.43426032181561e-07, + "loss": 1.0665, + "step": 37330 + }, + { + "epoch": 0.9, + "learning_rate": 5.431727918979469e-07, + "loss": 1.1395, + "step": 37331 + }, + { + "epoch": 0.9, + "learning_rate": 5.429196089866762e-07, + "loss": 1.0734, + "step": 37332 + }, + { + "epoch": 0.9, + "learning_rate": 5.426664834492911e-07, + "loss": 1.1533, + "step": 37333 + }, + { + "epoch": 0.9, + "learning_rate": 5.424134152873228e-07, + "loss": 1.016, + "step": 37334 + }, + { + "epoch": 0.9, + "learning_rate": 5.421604045023088e-07, + "loss": 0.8557, + "step": 37335 + }, + { + "epoch": 0.9, + "learning_rate": 5.419074510957823e-07, + "loss": 1.0881, + "step": 37336 + }, + { + "epoch": 0.9, + "learning_rate": 5.416545550692776e-07, + "loss": 1.2312, + "step": 37337 + }, + { + "epoch": 0.9, + "learning_rate": 5.414017164243324e-07, + "loss": 1.1064, + "step": 37338 + }, + { + "epoch": 0.9, + "learning_rate": 5.411489351624765e-07, + "loss": 0.9888, + "step": 37339 + }, + { + "epoch": 0.9, + "learning_rate": 5.408962112852445e-07, + "loss": 1.0698, + "step": 37340 + }, + { + "epoch": 0.9, + "learning_rate": 5.406435447941716e-07, + "loss": 1.2167, + "step": 37341 + }, + { + "epoch": 0.9, + "learning_rate": 5.403909356907888e-07, + "loss": 0.999, + "step": 37342 + }, + { + "epoch": 0.9, + "learning_rate": 5.401383839766283e-07, + "loss": 0.8696, + "step": 37343 + }, + { + "epoch": 0.9, + "learning_rate": 5.398858896532255e-07, + "loss": 1.0797, + "step": 37344 + }, + { + "epoch": 0.9, + "learning_rate": 5.396334527221081e-07, + "loss": 1.1476, + "step": 37345 + }, + { + "epoch": 0.9, + "learning_rate": 5.393810731848103e-07, + "loss": 0.8632, + "step": 37346 + }, + { + "epoch": 0.9, + "learning_rate": 5.391287510428611e-07, + "loss": 0.8829, + "step": 37347 + }, + { + "epoch": 0.9, + "learning_rate": 5.388764862977924e-07, + "loss": 1.1619, + "step": 37348 + }, + { + "epoch": 0.9, + "learning_rate": 5.386242789511365e-07, + "loss": 0.8992, + "step": 37349 + }, + { + "epoch": 0.9, + "learning_rate": 5.383721290044197e-07, + "loss": 1.0193, + "step": 37350 + }, + { + "epoch": 0.9, + "learning_rate": 5.381200364591732e-07, + "loss": 1.0673, + "step": 37351 + }, + { + "epoch": 0.9, + "learning_rate": 5.378680013169291e-07, + "loss": 0.9091, + "step": 37352 + }, + { + "epoch": 0.9, + "learning_rate": 5.376160235792127e-07, + "loss": 1.1003, + "step": 37353 + }, + { + "epoch": 0.9, + "learning_rate": 5.373641032475552e-07, + "loss": 1.1529, + "step": 37354 + }, + { + "epoch": 0.9, + "learning_rate": 5.37112240323483e-07, + "loss": 1.0497, + "step": 37355 + }, + { + "epoch": 0.9, + "learning_rate": 5.368604348085249e-07, + "loss": 0.9061, + "step": 37356 + }, + { + "epoch": 0.9, + "learning_rate": 5.366086867042108e-07, + "loss": 1.1338, + "step": 37357 + }, + { + "epoch": 0.9, + "learning_rate": 5.363569960120629e-07, + "loss": 1.0727, + "step": 37358 + }, + { + "epoch": 0.9, + "learning_rate": 5.361053627336122e-07, + "loss": 0.9559, + "step": 37359 + }, + { + "epoch": 0.9, + "learning_rate": 5.35853786870385e-07, + "loss": 1.2355, + "step": 37360 + }, + { + "epoch": 0.9, + "learning_rate": 5.35602268423906e-07, + "loss": 0.9141, + "step": 37361 + }, + { + "epoch": 0.9, + "learning_rate": 5.353508073957037e-07, + "loss": 0.8957, + "step": 37362 + }, + { + "epoch": 0.9, + "learning_rate": 5.350994037872992e-07, + "loss": 1.1038, + "step": 37363 + }, + { + "epoch": 0.9, + "learning_rate": 5.348480576002213e-07, + "loss": 1.1625, + "step": 37364 + }, + { + "epoch": 0.9, + "learning_rate": 5.345967688359943e-07, + "loss": 1.2115, + "step": 37365 + }, + { + "epoch": 0.9, + "learning_rate": 5.343455374961415e-07, + "loss": 0.9876, + "step": 37366 + }, + { + "epoch": 0.9, + "learning_rate": 5.340943635821883e-07, + "loss": 1.1086, + "step": 37367 + }, + { + "epoch": 0.9, + "learning_rate": 5.33843247095659e-07, + "loss": 0.9971, + "step": 37368 + }, + { + "epoch": 0.9, + "learning_rate": 5.335921880380746e-07, + "loss": 0.9622, + "step": 37369 + }, + { + "epoch": 0.9, + "learning_rate": 5.333411864109617e-07, + "loss": 0.9577, + "step": 37370 + }, + { + "epoch": 0.9, + "learning_rate": 5.33090242215839e-07, + "loss": 1.1039, + "step": 37371 + }, + { + "epoch": 0.9, + "learning_rate": 5.328393554542311e-07, + "loss": 1.0526, + "step": 37372 + }, + { + "epoch": 0.9, + "learning_rate": 5.32588526127662e-07, + "loss": 0.9254, + "step": 37373 + }, + { + "epoch": 0.9, + "learning_rate": 5.323377542376495e-07, + "loss": 1.1596, + "step": 37374 + }, + { + "epoch": 0.9, + "learning_rate": 5.320870397857169e-07, + "loss": 0.9148, + "step": 37375 + }, + { + "epoch": 0.9, + "learning_rate": 5.318363827733874e-07, + "loss": 1.1134, + "step": 37376 + }, + { + "epoch": 0.9, + "learning_rate": 5.315857832021776e-07, + "loss": 1.2637, + "step": 37377 + }, + { + "epoch": 0.9, + "learning_rate": 5.313352410736117e-07, + "loss": 1.1553, + "step": 37378 + }, + { + "epoch": 0.9, + "learning_rate": 5.310847563892052e-07, + "loss": 0.9765, + "step": 37379 + }, + { + "epoch": 0.9, + "learning_rate": 5.308343291504814e-07, + "loss": 1.1739, + "step": 37380 + }, + { + "epoch": 0.9, + "learning_rate": 5.30583959358959e-07, + "loss": 0.9758, + "step": 37381 + }, + { + "epoch": 0.9, + "learning_rate": 5.303336470161558e-07, + "loss": 0.9833, + "step": 37382 + }, + { + "epoch": 0.9, + "learning_rate": 5.300833921235904e-07, + "loss": 0.931, + "step": 37383 + }, + { + "epoch": 0.9, + "learning_rate": 5.298331946827828e-07, + "loss": 1.0885, + "step": 37384 + }, + { + "epoch": 0.9, + "learning_rate": 5.295830546952485e-07, + "loss": 1.0195, + "step": 37385 + }, + { + "epoch": 0.9, + "learning_rate": 5.293329721625063e-07, + "loss": 0.7876, + "step": 37386 + }, + { + "epoch": 0.9, + "learning_rate": 5.290829470860748e-07, + "loss": 0.9378, + "step": 37387 + }, + { + "epoch": 0.9, + "learning_rate": 5.288329794674673e-07, + "loss": 0.8556, + "step": 37388 + }, + { + "epoch": 0.9, + "learning_rate": 5.28583069308205e-07, + "loss": 1.2619, + "step": 37389 + }, + { + "epoch": 0.9, + "learning_rate": 5.283332166097987e-07, + "loss": 1.1686, + "step": 37390 + }, + { + "epoch": 0.9, + "learning_rate": 5.280834213737673e-07, + "loss": 0.8485, + "step": 37391 + }, + { + "epoch": 0.9, + "learning_rate": 5.278336836016273e-07, + "loss": 0.7748, + "step": 37392 + }, + { + "epoch": 0.9, + "learning_rate": 5.27584003294892e-07, + "loss": 0.995, + "step": 37393 + }, + { + "epoch": 0.9, + "learning_rate": 5.273343804550757e-07, + "loss": 0.7512, + "step": 37394 + }, + { + "epoch": 0.9, + "learning_rate": 5.270848150836949e-07, + "loss": 1.1698, + "step": 37395 + }, + { + "epoch": 0.9, + "learning_rate": 5.268353071822607e-07, + "loss": 1.0577, + "step": 37396 + }, + { + "epoch": 0.9, + "learning_rate": 5.265858567522897e-07, + "loss": 0.9677, + "step": 37397 + }, + { + "epoch": 0.9, + "learning_rate": 5.263364637952939e-07, + "loss": 1.1186, + "step": 37398 + }, + { + "epoch": 0.9, + "learning_rate": 5.260871283127855e-07, + "loss": 0.9251, + "step": 37399 + }, + { + "epoch": 0.9, + "learning_rate": 5.258378503062788e-07, + "loss": 1.0637, + "step": 37400 + }, + { + "epoch": 0.9, + "learning_rate": 5.25588629777285e-07, + "loss": 0.6048, + "step": 37401 + }, + { + "epoch": 0.9, + "learning_rate": 5.253394667273171e-07, + "loss": 1.1879, + "step": 37402 + }, + { + "epoch": 0.9, + "learning_rate": 5.250903611578861e-07, + "loss": 1.067, + "step": 37403 + }, + { + "epoch": 0.9, + "learning_rate": 5.248413130705021e-07, + "loss": 1.0057, + "step": 37404 + }, + { + "epoch": 0.9, + "learning_rate": 5.245923224666794e-07, + "loss": 1.0791, + "step": 37405 + }, + { + "epoch": 0.9, + "learning_rate": 5.243433893479243e-07, + "loss": 1.125, + "step": 37406 + }, + { + "epoch": 0.9, + "learning_rate": 5.240945137157494e-07, + "loss": 1.2294, + "step": 37407 + }, + { + "epoch": 0.9, + "learning_rate": 5.238456955716653e-07, + "loss": 1.0153, + "step": 37408 + }, + { + "epoch": 0.9, + "learning_rate": 5.235969349171788e-07, + "loss": 1.0165, + "step": 37409 + }, + { + "epoch": 0.9, + "learning_rate": 5.233482317538007e-07, + "loss": 1.0972, + "step": 37410 + }, + { + "epoch": 0.9, + "learning_rate": 5.230995860830412e-07, + "loss": 1.0693, + "step": 37411 + }, + { + "epoch": 0.9, + "learning_rate": 5.228509979064067e-07, + "loss": 1.1673, + "step": 37412 + }, + { + "epoch": 0.9, + "learning_rate": 5.22602467225407e-07, + "loss": 1.0167, + "step": 37413 + }, + { + "epoch": 0.9, + "learning_rate": 5.223539940415479e-07, + "loss": 1.0794, + "step": 37414 + }, + { + "epoch": 0.9, + "learning_rate": 5.221055783563378e-07, + "loss": 0.9205, + "step": 37415 + }, + { + "epoch": 0.9, + "learning_rate": 5.218572201712846e-07, + "loss": 0.9889, + "step": 37416 + }, + { + "epoch": 0.9, + "learning_rate": 5.216089194878937e-07, + "loss": 0.9505, + "step": 37417 + }, + { + "epoch": 0.9, + "learning_rate": 5.213606763076706e-07, + "loss": 1.0077, + "step": 37418 + }, + { + "epoch": 0.9, + "learning_rate": 5.211124906321252e-07, + "loss": 0.8913, + "step": 37419 + }, + { + "epoch": 0.9, + "learning_rate": 5.208643624627596e-07, + "loss": 1.0363, + "step": 37420 + }, + { + "epoch": 0.9, + "learning_rate": 5.206162918010816e-07, + "loss": 0.7131, + "step": 37421 + }, + { + "epoch": 0.9, + "learning_rate": 5.203682786485931e-07, + "loss": 1.2168, + "step": 37422 + }, + { + "epoch": 0.9, + "learning_rate": 5.201203230068009e-07, + "loss": 0.9186, + "step": 37423 + }, + { + "epoch": 0.9, + "learning_rate": 5.198724248772091e-07, + "loss": 0.7582, + "step": 37424 + }, + { + "epoch": 0.9, + "learning_rate": 5.196245842613212e-07, + "loss": 1.1125, + "step": 37425 + }, + { + "epoch": 0.9, + "learning_rate": 5.193768011606404e-07, + "loss": 0.838, + "step": 37426 + }, + { + "epoch": 0.9, + "learning_rate": 5.191290755766731e-07, + "loss": 0.8525, + "step": 37427 + }, + { + "epoch": 0.9, + "learning_rate": 5.188814075109172e-07, + "loss": 1.0361, + "step": 37428 + }, + { + "epoch": 0.9, + "learning_rate": 5.186337969648791e-07, + "loss": 0.8775, + "step": 37429 + }, + { + "epoch": 0.9, + "learning_rate": 5.183862439400589e-07, + "loss": 0.9077, + "step": 37430 + }, + { + "epoch": 0.9, + "learning_rate": 5.181387484379585e-07, + "loss": 0.9921, + "step": 37431 + }, + { + "epoch": 0.9, + "learning_rate": 5.178913104600814e-07, + "loss": 1.2002, + "step": 37432 + }, + { + "epoch": 0.9, + "learning_rate": 5.176439300079272e-07, + "loss": 0.9769, + "step": 37433 + }, + { + "epoch": 0.9, + "learning_rate": 5.17396607082995e-07, + "loss": 0.966, + "step": 37434 + }, + { + "epoch": 0.9, + "learning_rate": 5.171493416867901e-07, + "loss": 1.0033, + "step": 37435 + }, + { + "epoch": 0.9, + "learning_rate": 5.16902133820808e-07, + "loss": 1.1365, + "step": 37436 + }, + { + "epoch": 0.9, + "learning_rate": 5.166549834865497e-07, + "loss": 0.8164, + "step": 37437 + }, + { + "epoch": 0.9, + "learning_rate": 5.164078906855163e-07, + "loss": 1.0964, + "step": 37438 + }, + { + "epoch": 0.9, + "learning_rate": 5.161608554192044e-07, + "loss": 1.2482, + "step": 37439 + }, + { + "epoch": 0.9, + "learning_rate": 5.159138776891148e-07, + "loss": 1.0608, + "step": 37440 + }, + { + "epoch": 0.9, + "learning_rate": 5.156669574967432e-07, + "loss": 0.9867, + "step": 37441 + }, + { + "epoch": 0.9, + "learning_rate": 5.154200948435906e-07, + "loss": 0.9628, + "step": 37442 + }, + { + "epoch": 0.9, + "learning_rate": 5.151732897311545e-07, + "loss": 0.9521, + "step": 37443 + }, + { + "epoch": 0.9, + "learning_rate": 5.149265421609284e-07, + "loss": 0.9678, + "step": 37444 + }, + { + "epoch": 0.9, + "learning_rate": 5.146798521344132e-07, + "loss": 1.1525, + "step": 37445 + }, + { + "epoch": 0.9, + "learning_rate": 5.144332196531054e-07, + "loss": 1.231, + "step": 37446 + }, + { + "epoch": 0.9, + "learning_rate": 5.141866447184985e-07, + "loss": 1.0406, + "step": 37447 + }, + { + "epoch": 0.9, + "learning_rate": 5.139401273320921e-07, + "loss": 1.2345, + "step": 37448 + }, + { + "epoch": 0.9, + "learning_rate": 5.136936674953774e-07, + "loss": 1.1687, + "step": 37449 + }, + { + "epoch": 0.9, + "learning_rate": 5.134472652098521e-07, + "loss": 0.9485, + "step": 37450 + }, + { + "epoch": 0.9, + "learning_rate": 5.132009204770127e-07, + "loss": 1.0584, + "step": 37451 + }, + { + "epoch": 0.9, + "learning_rate": 5.129546332983504e-07, + "loss": 1.0877, + "step": 37452 + }, + { + "epoch": 0.9, + "learning_rate": 5.127084036753604e-07, + "loss": 0.8029, + "step": 37453 + }, + { + "epoch": 0.9, + "learning_rate": 5.124622316095384e-07, + "loss": 0.7906, + "step": 37454 + }, + { + "epoch": 0.9, + "learning_rate": 5.122161171023754e-07, + "loss": 1.1493, + "step": 37455 + }, + { + "epoch": 0.9, + "learning_rate": 5.119700601553668e-07, + "loss": 1.0062, + "step": 37456 + }, + { + "epoch": 0.9, + "learning_rate": 5.117240607700024e-07, + "loss": 1.104, + "step": 37457 + }, + { + "epoch": 0.9, + "learning_rate": 5.114781189477768e-07, + "loss": 1.022, + "step": 37458 + }, + { + "epoch": 0.9, + "learning_rate": 5.112322346901832e-07, + "loss": 1.0725, + "step": 37459 + }, + { + "epoch": 0.9, + "learning_rate": 5.109864079987104e-07, + "loss": 1.1533, + "step": 37460 + }, + { + "epoch": 0.9, + "learning_rate": 5.107406388748503e-07, + "loss": 1.2479, + "step": 37461 + }, + { + "epoch": 0.9, + "learning_rate": 5.104949273200965e-07, + "loss": 1.1745, + "step": 37462 + }, + { + "epoch": 0.9, + "learning_rate": 5.102492733359366e-07, + "loss": 1.1146, + "step": 37463 + }, + { + "epoch": 0.9, + "learning_rate": 5.100036769238637e-07, + "loss": 1.2126, + "step": 37464 + }, + { + "epoch": 0.9, + "learning_rate": 5.097581380853655e-07, + "loss": 1.2004, + "step": 37465 + }, + { + "epoch": 0.9, + "learning_rate": 5.09512656821931e-07, + "loss": 0.916, + "step": 37466 + }, + { + "epoch": 0.9, + "learning_rate": 5.092672331350534e-07, + "loss": 1.0654, + "step": 37467 + }, + { + "epoch": 0.9, + "learning_rate": 5.090218670262171e-07, + "loss": 0.8588, + "step": 37468 + }, + { + "epoch": 0.9, + "learning_rate": 5.087765584969129e-07, + "loss": 0.8613, + "step": 37469 + }, + { + "epoch": 0.9, + "learning_rate": 5.085313075486309e-07, + "loss": 1.1744, + "step": 37470 + }, + { + "epoch": 0.9, + "learning_rate": 5.082861141828555e-07, + "loss": 1.0889, + "step": 37471 + }, + { + "epoch": 0.9, + "learning_rate": 5.080409784010765e-07, + "loss": 1.0601, + "step": 37472 + }, + { + "epoch": 0.9, + "learning_rate": 5.077959002047784e-07, + "loss": 1.1262, + "step": 37473 + }, + { + "epoch": 0.9, + "learning_rate": 5.07550879595451e-07, + "loss": 0.9923, + "step": 37474 + }, + { + "epoch": 0.9, + "learning_rate": 5.073059165745797e-07, + "loss": 0.8156, + "step": 37475 + }, + { + "epoch": 0.9, + "learning_rate": 5.070610111436502e-07, + "loss": 1.0516, + "step": 37476 + }, + { + "epoch": 0.9, + "learning_rate": 5.06816163304148e-07, + "loss": 1.2651, + "step": 37477 + }, + { + "epoch": 0.9, + "learning_rate": 5.065713730575617e-07, + "loss": 1.0525, + "step": 37478 + }, + { + "epoch": 0.9, + "learning_rate": 5.063266404053712e-07, + "loss": 1.2723, + "step": 37479 + }, + { + "epoch": 0.9, + "learning_rate": 5.060819653490645e-07, + "loss": 1.2286, + "step": 37480 + }, + { + "epoch": 0.9, + "learning_rate": 5.058373478901269e-07, + "loss": 1.1075, + "step": 37481 + }, + { + "epoch": 0.9, + "learning_rate": 5.055927880300394e-07, + "loss": 0.962, + "step": 37482 + }, + { + "epoch": 0.9, + "learning_rate": 5.053482857702885e-07, + "loss": 0.9114, + "step": 37483 + }, + { + "epoch": 0.9, + "learning_rate": 5.051038411123554e-07, + "loss": 0.9559, + "step": 37484 + }, + { + "epoch": 0.9, + "learning_rate": 5.048594540577234e-07, + "loss": 1.1047, + "step": 37485 + }, + { + "epoch": 0.9, + "learning_rate": 5.046151246078779e-07, + "loss": 1.0163, + "step": 37486 + }, + { + "epoch": 0.9, + "learning_rate": 5.043708527642965e-07, + "loss": 1.019, + "step": 37487 + }, + { + "epoch": 0.9, + "learning_rate": 5.041266385284648e-07, + "loss": 1.0094, + "step": 37488 + }, + { + "epoch": 0.9, + "learning_rate": 5.03882481901864e-07, + "loss": 1.0412, + "step": 37489 + }, + { + "epoch": 0.9, + "learning_rate": 5.036383828859726e-07, + "loss": 1.0638, + "step": 37490 + }, + { + "epoch": 0.9, + "learning_rate": 5.033943414822751e-07, + "loss": 0.8218, + "step": 37491 + }, + { + "epoch": 0.9, + "learning_rate": 5.031503576922491e-07, + "loss": 1.003, + "step": 37492 + }, + { + "epoch": 0.9, + "learning_rate": 5.029064315173771e-07, + "loss": 1.0098, + "step": 37493 + }, + { + "epoch": 0.9, + "learning_rate": 5.026625629591386e-07, + "loss": 0.9527, + "step": 37494 + }, + { + "epoch": 0.9, + "learning_rate": 5.024187520190105e-07, + "loss": 1.1177, + "step": 37495 + }, + { + "epoch": 0.9, + "learning_rate": 5.021749986984748e-07, + "loss": 1.2165, + "step": 37496 + }, + { + "epoch": 0.9, + "learning_rate": 5.019313029990091e-07, + "loss": 1.1671, + "step": 37497 + }, + { + "epoch": 0.9, + "learning_rate": 5.016876649220925e-07, + "loss": 1.1511, + "step": 37498 + }, + { + "epoch": 0.9, + "learning_rate": 5.014440844692014e-07, + "loss": 1.0047, + "step": 37499 + }, + { + "epoch": 0.9, + "learning_rate": 5.012005616418158e-07, + "loss": 1.0544, + "step": 37500 + }, + { + "epoch": 0.9, + "learning_rate": 5.009570964414134e-07, + "loss": 1.0355, + "step": 37501 + }, + { + "epoch": 0.9, + "learning_rate": 5.007136888694685e-07, + "loss": 0.9429, + "step": 37502 + }, + { + "epoch": 0.9, + "learning_rate": 5.00470338927459e-07, + "loss": 0.8246, + "step": 37503 + }, + { + "epoch": 0.9, + "learning_rate": 5.002270466168634e-07, + "loss": 1.0971, + "step": 37504 + }, + { + "epoch": 0.9, + "learning_rate": 4.999838119391543e-07, + "loss": 0.865, + "step": 37505 + }, + { + "epoch": 0.9, + "learning_rate": 4.997406348958078e-07, + "loss": 1.2845, + "step": 37506 + }, + { + "epoch": 0.9, + "learning_rate": 4.994975154883031e-07, + "loss": 1.2127, + "step": 37507 + }, + { + "epoch": 0.9, + "learning_rate": 4.992544537181099e-07, + "loss": 1.0355, + "step": 37508 + }, + { + "epoch": 0.9, + "learning_rate": 4.990114495867071e-07, + "loss": 0.7891, + "step": 37509 + }, + { + "epoch": 0.9, + "learning_rate": 4.987685030955658e-07, + "loss": 1.0207, + "step": 37510 + }, + { + "epoch": 0.9, + "learning_rate": 4.985256142461615e-07, + "loss": 1.1147, + "step": 37511 + }, + { + "epoch": 0.9, + "learning_rate": 4.982827830399684e-07, + "loss": 1.0246, + "step": 37512 + }, + { + "epoch": 0.9, + "learning_rate": 4.980400094784565e-07, + "loss": 0.905, + "step": 37513 + }, + { + "epoch": 0.9, + "learning_rate": 4.977972935631026e-07, + "loss": 0.9974, + "step": 37514 + }, + { + "epoch": 0.9, + "learning_rate": 4.975546352953775e-07, + "loss": 0.8922, + "step": 37515 + }, + { + "epoch": 0.9, + "learning_rate": 4.973120346767523e-07, + "loss": 1.2299, + "step": 37516 + }, + { + "epoch": 0.9, + "learning_rate": 4.970694917086994e-07, + "loss": 0.8739, + "step": 37517 + }, + { + "epoch": 0.9, + "learning_rate": 4.968270063926928e-07, + "loss": 0.9385, + "step": 37518 + }, + { + "epoch": 0.9, + "learning_rate": 4.965845787302004e-07, + "loss": 0.7291, + "step": 37519 + }, + { + "epoch": 0.9, + "learning_rate": 4.963422087226944e-07, + "loss": 1.0695, + "step": 37520 + }, + { + "epoch": 0.9, + "learning_rate": 4.960998963716435e-07, + "loss": 0.7233, + "step": 37521 + }, + { + "epoch": 0.9, + "learning_rate": 4.958576416785199e-07, + "loss": 1.0414, + "step": 37522 + }, + { + "epoch": 0.9, + "learning_rate": 4.956154446447925e-07, + "loss": 0.9872, + "step": 37523 + }, + { + "epoch": 0.9, + "learning_rate": 4.953733052719311e-07, + "loss": 1.1927, + "step": 37524 + }, + { + "epoch": 0.9, + "learning_rate": 4.951312235614025e-07, + "loss": 1.2558, + "step": 37525 + }, + { + "epoch": 0.9, + "learning_rate": 4.948891995146799e-07, + "loss": 1.1248, + "step": 37526 + }, + { + "epoch": 0.9, + "learning_rate": 4.946472331332263e-07, + "loss": 0.9186, + "step": 37527 + }, + { + "epoch": 0.9, + "learning_rate": 4.944053244185143e-07, + "loss": 0.8889, + "step": 37528 + }, + { + "epoch": 0.9, + "learning_rate": 4.941634733720068e-07, + "loss": 1.0061, + "step": 37529 + }, + { + "epoch": 0.9, + "learning_rate": 4.939216799951751e-07, + "loss": 1.0746, + "step": 37530 + }, + { + "epoch": 0.9, + "learning_rate": 4.936799442894846e-07, + "loss": 0.8709, + "step": 37531 + }, + { + "epoch": 0.9, + "learning_rate": 4.934382662564019e-07, + "loss": 0.8286, + "step": 37532 + }, + { + "epoch": 0.9, + "learning_rate": 4.931966458973924e-07, + "loss": 1.169, + "step": 37533 + }, + { + "epoch": 0.9, + "learning_rate": 4.929550832139241e-07, + "loss": 1.2022, + "step": 37534 + }, + { + "epoch": 0.9, + "learning_rate": 4.927135782074588e-07, + "loss": 0.8046, + "step": 37535 + }, + { + "epoch": 0.9, + "learning_rate": 4.924721308794666e-07, + "loss": 1.049, + "step": 37536 + }, + { + "epoch": 0.9, + "learning_rate": 4.922307412314087e-07, + "loss": 1.0468, + "step": 37537 + }, + { + "epoch": 0.9, + "learning_rate": 4.919894092647493e-07, + "loss": 1.2835, + "step": 37538 + }, + { + "epoch": 0.9, + "learning_rate": 4.917481349809561e-07, + "loss": 0.9794, + "step": 37539 + }, + { + "epoch": 0.9, + "learning_rate": 4.91506918381488e-07, + "loss": 1.2036, + "step": 37540 + }, + { + "epoch": 0.9, + "learning_rate": 4.912657594678116e-07, + "loss": 1.0256, + "step": 37541 + }, + { + "epoch": 0.9, + "learning_rate": 4.910246582413902e-07, + "loss": 0.7713, + "step": 37542 + }, + { + "epoch": 0.9, + "learning_rate": 4.907836147036849e-07, + "loss": 1.03, + "step": 37543 + }, + { + "epoch": 0.9, + "learning_rate": 4.905426288561599e-07, + "loss": 0.9947, + "step": 37544 + }, + { + "epoch": 0.9, + "learning_rate": 4.903017007002741e-07, + "loss": 1.0794, + "step": 37545 + }, + { + "epoch": 0.9, + "learning_rate": 4.90060830237491e-07, + "loss": 1.2901, + "step": 37546 + }, + { + "epoch": 0.9, + "learning_rate": 4.898200174692746e-07, + "loss": 1.0165, + "step": 37547 + }, + { + "epoch": 0.9, + "learning_rate": 4.895792623970808e-07, + "loss": 1.0013, + "step": 37548 + }, + { + "epoch": 0.9, + "learning_rate": 4.893385650223725e-07, + "loss": 0.8513, + "step": 37549 + }, + { + "epoch": 0.9, + "learning_rate": 4.890979253466121e-07, + "loss": 0.9796, + "step": 37550 + }, + { + "epoch": 0.9, + "learning_rate": 4.888573433712551e-07, + "loss": 1.1419, + "step": 37551 + }, + { + "epoch": 0.9, + "learning_rate": 4.886168190977658e-07, + "loss": 1.0565, + "step": 37552 + }, + { + "epoch": 0.9, + "learning_rate": 4.883763525275997e-07, + "loss": 1.0709, + "step": 37553 + }, + { + "epoch": 0.9, + "learning_rate": 4.881359436622168e-07, + "loss": 1.0679, + "step": 37554 + }, + { + "epoch": 0.9, + "learning_rate": 4.87895592503077e-07, + "loss": 0.6917, + "step": 37555 + }, + { + "epoch": 0.9, + "learning_rate": 4.876552990516359e-07, + "loss": 1.0876, + "step": 37556 + }, + { + "epoch": 0.9, + "learning_rate": 4.874150633093522e-07, + "loss": 0.963, + "step": 37557 + }, + { + "epoch": 0.9, + "learning_rate": 4.871748852776859e-07, + "loss": 0.9908, + "step": 37558 + }, + { + "epoch": 0.9, + "learning_rate": 4.869347649580902e-07, + "loss": 1.0444, + "step": 37559 + }, + { + "epoch": 0.9, + "learning_rate": 4.866947023520241e-07, + "loss": 0.8649, + "step": 37560 + }, + { + "epoch": 0.9, + "learning_rate": 4.864546974609452e-07, + "loss": 1.0231, + "step": 37561 + }, + { + "epoch": 0.9, + "learning_rate": 4.862147502863057e-07, + "loss": 1.1694, + "step": 37562 + }, + { + "epoch": 0.9, + "learning_rate": 4.859748608295656e-07, + "loss": 1.1445, + "step": 37563 + }, + { + "epoch": 0.9, + "learning_rate": 4.857350290921759e-07, + "loss": 1.0295, + "step": 37564 + }, + { + "epoch": 0.9, + "learning_rate": 4.854952550755954e-07, + "loss": 1.0138, + "step": 37565 + }, + { + "epoch": 0.9, + "learning_rate": 4.852555387812774e-07, + "loss": 1.1994, + "step": 37566 + }, + { + "epoch": 0.9, + "learning_rate": 4.850158802106752e-07, + "loss": 1.0406, + "step": 37567 + }, + { + "epoch": 0.9, + "learning_rate": 4.847762793652433e-07, + "loss": 0.9497, + "step": 37568 + }, + { + "epoch": 0.9, + "learning_rate": 4.845367362464371e-07, + "loss": 1.0532, + "step": 37569 + }, + { + "epoch": 0.9, + "learning_rate": 4.842972508557064e-07, + "loss": 0.8863, + "step": 37570 + }, + { + "epoch": 0.9, + "learning_rate": 4.840578231945081e-07, + "loss": 1.1549, + "step": 37571 + }, + { + "epoch": 0.9, + "learning_rate": 4.838184532642909e-07, + "loss": 0.7871, + "step": 37572 + }, + { + "epoch": 0.9, + "learning_rate": 4.835791410665092e-07, + "loss": 1.2298, + "step": 37573 + }, + { + "epoch": 0.9, + "learning_rate": 4.833398866026162e-07, + "loss": 0.7493, + "step": 37574 + }, + { + "epoch": 0.9, + "learning_rate": 4.831006898740598e-07, + "loss": 1.0605, + "step": 37575 + }, + { + "epoch": 0.9, + "learning_rate": 4.82861550882292e-07, + "loss": 1.2633, + "step": 37576 + }, + { + "epoch": 0.9, + "learning_rate": 4.826224696287674e-07, + "loss": 1.1565, + "step": 37577 + }, + { + "epoch": 0.9, + "learning_rate": 4.823834461149313e-07, + "loss": 0.9113, + "step": 37578 + }, + { + "epoch": 0.9, + "learning_rate": 4.821444803422382e-07, + "loss": 1.0012, + "step": 37579 + }, + { + "epoch": 0.9, + "learning_rate": 4.819055723121346e-07, + "loss": 1.2454, + "step": 37580 + }, + { + "epoch": 0.9, + "learning_rate": 4.816667220260707e-07, + "loss": 1.0476, + "step": 37581 + }, + { + "epoch": 0.9, + "learning_rate": 4.814279294854973e-07, + "loss": 1.0963, + "step": 37582 + }, + { + "epoch": 0.9, + "learning_rate": 4.811891946918612e-07, + "loss": 1.0447, + "step": 37583 + }, + { + "epoch": 0.9, + "learning_rate": 4.8095051764661e-07, + "loss": 0.9585, + "step": 37584 + }, + { + "epoch": 0.9, + "learning_rate": 4.807118983511949e-07, + "loss": 0.9114, + "step": 37585 + }, + { + "epoch": 0.9, + "learning_rate": 4.804733368070602e-07, + "loss": 0.9822, + "step": 37586 + }, + { + "epoch": 0.9, + "learning_rate": 4.802348330156558e-07, + "loss": 1.0633, + "step": 37587 + }, + { + "epoch": 0.9, + "learning_rate": 4.799963869784274e-07, + "loss": 1.0869, + "step": 37588 + }, + { + "epoch": 0.9, + "learning_rate": 4.797579986968204e-07, + "loss": 1.0713, + "step": 37589 + }, + { + "epoch": 0.9, + "learning_rate": 4.795196681722835e-07, + "loss": 0.9014, + "step": 37590 + }, + { + "epoch": 0.9, + "learning_rate": 4.792813954062614e-07, + "loss": 1.287, + "step": 37591 + }, + { + "epoch": 0.9, + "learning_rate": 4.790431804001982e-07, + "loss": 1.1887, + "step": 37592 + }, + { + "epoch": 0.9, + "learning_rate": 4.78805023155543e-07, + "loss": 0.9111, + "step": 37593 + }, + { + "epoch": 0.9, + "learning_rate": 4.785669236737356e-07, + "loss": 1.0127, + "step": 37594 + }, + { + "epoch": 0.9, + "learning_rate": 4.78328881956226e-07, + "loss": 1.0502, + "step": 37595 + }, + { + "epoch": 0.9, + "learning_rate": 4.78090898004453e-07, + "loss": 1.0731, + "step": 37596 + }, + { + "epoch": 0.9, + "learning_rate": 4.778529718198632e-07, + "loss": 0.8395, + "step": 37597 + }, + { + "epoch": 0.9, + "learning_rate": 4.776151034039011e-07, + "loss": 1.0648, + "step": 37598 + }, + { + "epoch": 0.9, + "learning_rate": 4.773772927580067e-07, + "loss": 0.8918, + "step": 37599 + }, + { + "epoch": 0.9, + "learning_rate": 4.771395398836243e-07, + "loss": 0.9258, + "step": 37600 + }, + { + "epoch": 0.9, + "learning_rate": 4.769018447821982e-07, + "loss": 0.993, + "step": 37601 + }, + { + "epoch": 0.9, + "learning_rate": 4.766642074551675e-07, + "loss": 1.1428, + "step": 37602 + }, + { + "epoch": 0.9, + "learning_rate": 4.764266279039742e-07, + "loss": 1.0898, + "step": 37603 + }, + { + "epoch": 0.9, + "learning_rate": 4.7618910613006277e-07, + "loss": 1.3332, + "step": 37604 + }, + { + "epoch": 0.9, + "learning_rate": 4.7595164213486977e-07, + "loss": 1.4998, + "step": 37605 + }, + { + "epoch": 0.9, + "learning_rate": 4.757142359198397e-07, + "loss": 1.204, + "step": 37606 + }, + { + "epoch": 0.9, + "learning_rate": 4.754768874864091e-07, + "loss": 0.8729, + "step": 37607 + }, + { + "epoch": 0.9, + "learning_rate": 4.752395968360213e-07, + "loss": 1.001, + "step": 37608 + }, + { + "epoch": 0.9, + "learning_rate": 4.7500236397011515e-07, + "loss": 0.7799, + "step": 37609 + }, + { + "epoch": 0.9, + "learning_rate": 4.7476518889012725e-07, + "loss": 0.9867, + "step": 37610 + }, + { + "epoch": 0.9, + "learning_rate": 4.745280715974987e-07, + "loss": 1.1201, + "step": 37611 + }, + { + "epoch": 0.9, + "learning_rate": 4.7429101209366947e-07, + "loss": 1.1101, + "step": 37612 + }, + { + "epoch": 0.9, + "learning_rate": 4.7405401038007504e-07, + "loss": 1.0791, + "step": 37613 + }, + { + "epoch": 0.9, + "learning_rate": 4.738170664581554e-07, + "loss": 0.8881, + "step": 37614 + }, + { + "epoch": 0.9, + "learning_rate": 4.7358018032934494e-07, + "loss": 1.0593, + "step": 37615 + }, + { + "epoch": 0.9, + "learning_rate": 4.733433519950836e-07, + "loss": 0.9036, + "step": 37616 + }, + { + "epoch": 0.9, + "learning_rate": 4.7310658145680923e-07, + "loss": 1.1142, + "step": 37617 + }, + { + "epoch": 0.9, + "learning_rate": 4.728698687159539e-07, + "loss": 1.0645, + "step": 37618 + }, + { + "epoch": 0.9, + "learning_rate": 4.726332137739564e-07, + "loss": 1.0075, + "step": 37619 + }, + { + "epoch": 0.9, + "learning_rate": 4.723966166322536e-07, + "loss": 0.8099, + "step": 37620 + }, + { + "epoch": 0.9, + "learning_rate": 4.7216007729227854e-07, + "loss": 0.9872, + "step": 37621 + }, + { + "epoch": 0.9, + "learning_rate": 4.7192359575546797e-07, + "loss": 0.9794, + "step": 37622 + }, + { + "epoch": 0.9, + "learning_rate": 4.716871720232541e-07, + "loss": 1.218, + "step": 37623 + }, + { + "epoch": 0.9, + "learning_rate": 4.714508060970735e-07, + "loss": 1.0679, + "step": 37624 + }, + { + "epoch": 0.9, + "learning_rate": 4.712144979783617e-07, + "loss": 1.0027, + "step": 37625 + }, + { + "epoch": 0.9, + "learning_rate": 4.709782476685476e-07, + "loss": 1.2324, + "step": 37626 + }, + { + "epoch": 0.9, + "learning_rate": 4.707420551690678e-07, + "loss": 1.0122, + "step": 37627 + }, + { + "epoch": 0.9, + "learning_rate": 4.7050592048135555e-07, + "loss": 0.8125, + "step": 37628 + }, + { + "epoch": 0.9, + "learning_rate": 4.7026984360684204e-07, + "loss": 1.0579, + "step": 37629 + }, + { + "epoch": 0.9, + "learning_rate": 4.7003382454696044e-07, + "loss": 1.0572, + "step": 37630 + }, + { + "epoch": 0.9, + "learning_rate": 4.697978633031408e-07, + "loss": 0.9822, + "step": 37631 + }, + { + "epoch": 0.91, + "learning_rate": 4.695619598768153e-07, + "loss": 1.2453, + "step": 37632 + }, + { + "epoch": 0.91, + "learning_rate": 4.693261142694172e-07, + "loss": 0.9723, + "step": 37633 + }, + { + "epoch": 0.91, + "learning_rate": 4.690903264823754e-07, + "loss": 0.7729, + "step": 37634 + }, + { + "epoch": 0.91, + "learning_rate": 4.688545965171198e-07, + "loss": 0.9493, + "step": 37635 + }, + { + "epoch": 0.91, + "learning_rate": 4.686189243750838e-07, + "loss": 1.0109, + "step": 37636 + }, + { + "epoch": 0.91, + "learning_rate": 4.683833100576929e-07, + "loss": 0.9445, + "step": 37637 + }, + { + "epoch": 0.91, + "learning_rate": 4.681477535663792e-07, + "loss": 1.1, + "step": 37638 + }, + { + "epoch": 0.91, + "learning_rate": 4.6791225490257055e-07, + "loss": 0.8858, + "step": 37639 + }, + { + "epoch": 0.91, + "learning_rate": 4.676768140676957e-07, + "loss": 0.9833, + "step": 37640 + }, + { + "epoch": 0.91, + "learning_rate": 4.6744143106318586e-07, + "loss": 1.0842, + "step": 37641 + }, + { + "epoch": 0.91, + "learning_rate": 4.672061058904642e-07, + "loss": 0.7957, + "step": 37642 + }, + { + "epoch": 0.91, + "learning_rate": 4.669708385509608e-07, + "loss": 0.967, + "step": 37643 + }, + { + "epoch": 0.91, + "learning_rate": 4.667356290461056e-07, + "loss": 1.0103, + "step": 37644 + }, + { + "epoch": 0.91, + "learning_rate": 4.665004773773207e-07, + "loss": 0.8141, + "step": 37645 + }, + { + "epoch": 0.91, + "learning_rate": 4.6626538354603623e-07, + "loss": 1.0117, + "step": 37646 + }, + { + "epoch": 0.91, + "learning_rate": 4.6603034755367647e-07, + "loss": 1.1009, + "step": 37647 + }, + { + "epoch": 0.91, + "learning_rate": 4.6579536940166703e-07, + "loss": 1.0604, + "step": 37648 + }, + { + "epoch": 0.91, + "learning_rate": 4.655604490914367e-07, + "loss": 1.2245, + "step": 37649 + }, + { + "epoch": 0.91, + "learning_rate": 4.653255866244066e-07, + "loss": 1.1636, + "step": 37650 + }, + { + "epoch": 0.91, + "learning_rate": 4.650907820020045e-07, + "loss": 0.7997, + "step": 37651 + }, + { + "epoch": 0.91, + "learning_rate": 4.6485603522565483e-07, + "loss": 1.1561, + "step": 37652 + }, + { + "epoch": 0.91, + "learning_rate": 4.646213462967786e-07, + "loss": 1.0886, + "step": 37653 + }, + { + "epoch": 0.91, + "learning_rate": 4.6438671521680246e-07, + "loss": 1.1351, + "step": 37654 + }, + { + "epoch": 0.91, + "learning_rate": 4.6415214198715085e-07, + "loss": 1.2473, + "step": 37655 + }, + { + "epoch": 0.91, + "learning_rate": 4.6391762660924264e-07, + "loss": 0.7776, + "step": 37656 + }, + { + "epoch": 0.91, + "learning_rate": 4.636831690845056e-07, + "loss": 1.1248, + "step": 37657 + }, + { + "epoch": 0.91, + "learning_rate": 4.634487694143586e-07, + "loss": 1.0852, + "step": 37658 + }, + { + "epoch": 0.91, + "learning_rate": 4.6321442760022374e-07, + "loss": 1.0372, + "step": 37659 + }, + { + "epoch": 0.91, + "learning_rate": 4.629801436435255e-07, + "loss": 1.2258, + "step": 37660 + }, + { + "epoch": 0.91, + "learning_rate": 4.6274591754568276e-07, + "loss": 0.9152, + "step": 37661 + }, + { + "epoch": 0.91, + "learning_rate": 4.625117493081166e-07, + "loss": 1.1622, + "step": 37662 + }, + { + "epoch": 0.91, + "learning_rate": 4.6227763893225033e-07, + "loss": 0.9098, + "step": 37663 + }, + { + "epoch": 0.91, + "learning_rate": 4.6204358641949943e-07, + "loss": 0.9536, + "step": 37664 + }, + { + "epoch": 0.91, + "learning_rate": 4.618095917712895e-07, + "loss": 1.0992, + "step": 37665 + }, + { + "epoch": 0.91, + "learning_rate": 4.615756549890338e-07, + "loss": 0.9057, + "step": 37666 + }, + { + "epoch": 0.91, + "learning_rate": 4.6134177607415785e-07, + "loss": 0.9991, + "step": 37667 + }, + { + "epoch": 0.91, + "learning_rate": 4.6110795502807615e-07, + "loss": 1.0443, + "step": 37668 + }, + { + "epoch": 0.91, + "learning_rate": 4.6087419185220973e-07, + "loss": 1.0877, + "step": 37669 + }, + { + "epoch": 0.91, + "learning_rate": 4.606404865479763e-07, + "loss": 1.1052, + "step": 37670 + }, + { + "epoch": 0.91, + "learning_rate": 4.604068391167915e-07, + "loss": 0.8936, + "step": 37671 + }, + { + "epoch": 0.91, + "learning_rate": 4.6017324956007524e-07, + "loss": 1.089, + "step": 37672 + }, + { + "epoch": 0.91, + "learning_rate": 4.599397178792453e-07, + "loss": 0.9324, + "step": 37673 + }, + { + "epoch": 0.91, + "learning_rate": 4.597062440757161e-07, + "loss": 1.076, + "step": 37674 + }, + { + "epoch": 0.91, + "learning_rate": 4.5947282815090645e-07, + "loss": 0.8647, + "step": 37675 + }, + { + "epoch": 0.91, + "learning_rate": 4.592394701062286e-07, + "loss": 1.0973, + "step": 37676 + }, + { + "epoch": 0.91, + "learning_rate": 4.590061699431014e-07, + "loss": 1.0237, + "step": 37677 + }, + { + "epoch": 0.91, + "learning_rate": 4.5877292766294156e-07, + "loss": 1.0494, + "step": 37678 + }, + { + "epoch": 0.91, + "learning_rate": 4.585397432671601e-07, + "loss": 0.9596, + "step": 37679 + }, + { + "epoch": 0.91, + "learning_rate": 4.583066167571737e-07, + "loss": 1.1033, + "step": 37680 + }, + { + "epoch": 0.91, + "learning_rate": 4.5807354813439896e-07, + "loss": 0.9232, + "step": 37681 + }, + { + "epoch": 0.91, + "learning_rate": 4.578405374002459e-07, + "loss": 1.016, + "step": 37682 + }, + { + "epoch": 0.91, + "learning_rate": 4.5760758455612896e-07, + "loss": 0.9615, + "step": 37683 + }, + { + "epoch": 0.91, + "learning_rate": 4.5737468960346475e-07, + "loss": 0.9829, + "step": 37684 + }, + { + "epoch": 0.91, + "learning_rate": 4.571418525436611e-07, + "loss": 0.977, + "step": 37685 + }, + { + "epoch": 0.91, + "learning_rate": 4.5690907337813563e-07, + "loss": 1.1204, + "step": 37686 + }, + { + "epoch": 0.91, + "learning_rate": 4.566763521082962e-07, + "loss": 1.1244, + "step": 37687 + }, + { + "epoch": 0.91, + "learning_rate": 4.5644368873555723e-07, + "loss": 1.0835, + "step": 37688 + }, + { + "epoch": 0.91, + "learning_rate": 4.562110832613309e-07, + "loss": 0.9857, + "step": 37689 + }, + { + "epoch": 0.91, + "learning_rate": 4.55978535687025e-07, + "loss": 0.8207, + "step": 37690 + }, + { + "epoch": 0.91, + "learning_rate": 4.557460460140528e-07, + "loss": 1.1823, + "step": 37691 + }, + { + "epoch": 0.91, + "learning_rate": 4.5551361424382655e-07, + "loss": 1.1303, + "step": 37692 + }, + { + "epoch": 0.91, + "learning_rate": 4.552812403777518e-07, + "loss": 1.0171, + "step": 37693 + }, + { + "epoch": 0.91, + "learning_rate": 4.550489244172418e-07, + "loss": 1.0457, + "step": 37694 + }, + { + "epoch": 0.91, + "learning_rate": 4.548166663637044e-07, + "loss": 0.991, + "step": 37695 + }, + { + "epoch": 0.91, + "learning_rate": 4.5458446621854947e-07, + "loss": 0.9645, + "step": 37696 + }, + { + "epoch": 0.91, + "learning_rate": 4.54352323983186e-07, + "loss": 1.3655, + "step": 37697 + }, + { + "epoch": 0.91, + "learning_rate": 4.541202396590205e-07, + "loss": 0.9401, + "step": 37698 + }, + { + "epoch": 0.91, + "learning_rate": 4.53888213247462e-07, + "loss": 1.1052, + "step": 37699 + }, + { + "epoch": 0.91, + "learning_rate": 4.5365624474991933e-07, + "loss": 1.043, + "step": 37700 + }, + { + "epoch": 0.91, + "learning_rate": 4.5342433416779796e-07, + "loss": 1.1675, + "step": 37701 + }, + { + "epoch": 0.91, + "learning_rate": 4.5319248150250685e-07, + "loss": 0.823, + "step": 37702 + }, + { + "epoch": 0.91, + "learning_rate": 4.5296068675545037e-07, + "loss": 1.1046, + "step": 37703 + }, + { + "epoch": 0.91, + "learning_rate": 4.527289499280352e-07, + "loss": 0.9829, + "step": 37704 + }, + { + "epoch": 0.91, + "learning_rate": 4.524972710216702e-07, + "loss": 1.019, + "step": 37705 + }, + { + "epoch": 0.91, + "learning_rate": 4.5226565003775645e-07, + "loss": 0.979, + "step": 37706 + }, + { + "epoch": 0.91, + "learning_rate": 4.520340869777018e-07, + "loss": 1.1032, + "step": 37707 + }, + { + "epoch": 0.91, + "learning_rate": 4.518025818429106e-07, + "loss": 1.2202, + "step": 37708 + }, + { + "epoch": 0.91, + "learning_rate": 4.515711346347873e-07, + "loss": 0.8711, + "step": 37709 + }, + { + "epoch": 0.91, + "learning_rate": 4.5133974535473633e-07, + "loss": 1.0405, + "step": 37710 + }, + { + "epoch": 0.91, + "learning_rate": 4.5110841400416103e-07, + "loss": 0.7463, + "step": 37711 + }, + { + "epoch": 0.91, + "learning_rate": 4.508771405844636e-07, + "loss": 1.0862, + "step": 37712 + }, + { + "epoch": 0.91, + "learning_rate": 4.506459250970507e-07, + "loss": 1.1549, + "step": 37713 + }, + { + "epoch": 0.91, + "learning_rate": 4.5041476754332124e-07, + "loss": 1.2098, + "step": 37714 + }, + { + "epoch": 0.91, + "learning_rate": 4.501836679246785e-07, + "loss": 0.9056, + "step": 37715 + }, + { + "epoch": 0.91, + "learning_rate": 4.49952626242528e-07, + "loss": 1.1157, + "step": 37716 + }, + { + "epoch": 0.91, + "learning_rate": 4.4972164249826645e-07, + "loss": 1.0514, + "step": 37717 + }, + { + "epoch": 0.91, + "learning_rate": 4.4949071669329825e-07, + "loss": 0.956, + "step": 37718 + }, + { + "epoch": 0.91, + "learning_rate": 4.492598488290223e-07, + "loss": 1.1195, + "step": 37719 + }, + { + "epoch": 0.91, + "learning_rate": 4.490290389068408e-07, + "loss": 1.0355, + "step": 37720 + }, + { + "epoch": 0.91, + "learning_rate": 4.4879828692815485e-07, + "loss": 1.1289, + "step": 37721 + }, + { + "epoch": 0.91, + "learning_rate": 4.485675928943611e-07, + "loss": 0.8285, + "step": 37722 + }, + { + "epoch": 0.91, + "learning_rate": 4.4833695680686185e-07, + "loss": 0.9097, + "step": 37723 + }, + { + "epoch": 0.91, + "learning_rate": 4.4810637866705697e-07, + "loss": 0.9807, + "step": 37724 + }, + { + "epoch": 0.91, + "learning_rate": 4.478758584763421e-07, + "loss": 0.9469, + "step": 37725 + }, + { + "epoch": 0.91, + "learning_rate": 4.4764539623611937e-07, + "loss": 0.8396, + "step": 37726 + }, + { + "epoch": 0.91, + "learning_rate": 4.4741499194778435e-07, + "loss": 1.2305, + "step": 37727 + }, + { + "epoch": 0.91, + "learning_rate": 4.4718464561273487e-07, + "loss": 0.9434, + "step": 37728 + }, + { + "epoch": 0.91, + "learning_rate": 4.4695435723236977e-07, + "loss": 1.0117, + "step": 37729 + }, + { + "epoch": 0.91, + "learning_rate": 4.4672412680808574e-07, + "loss": 1.2225, + "step": 37730 + }, + { + "epoch": 0.91, + "learning_rate": 4.464939543412783e-07, + "loss": 1.0348, + "step": 37731 + }, + { + "epoch": 0.91, + "learning_rate": 4.462638398333474e-07, + "loss": 1.1125, + "step": 37732 + }, + { + "epoch": 0.91, + "learning_rate": 4.4603378328568424e-07, + "loss": 0.9529, + "step": 37733 + }, + { + "epoch": 0.91, + "learning_rate": 4.4580378469968765e-07, + "loss": 1.0301, + "step": 37734 + }, + { + "epoch": 0.91, + "learning_rate": 4.455738440767532e-07, + "loss": 1.0816, + "step": 37735 + }, + { + "epoch": 0.91, + "learning_rate": 4.453439614182731e-07, + "loss": 0.7872, + "step": 37736 + }, + { + "epoch": 0.91, + "learning_rate": 4.451141367256451e-07, + "loss": 1.3172, + "step": 37737 + }, + { + "epoch": 0.91, + "learning_rate": 4.4488437000026144e-07, + "loss": 0.8469, + "step": 37738 + }, + { + "epoch": 0.91, + "learning_rate": 4.4465466124351654e-07, + "loss": 0.8941, + "step": 37739 + }, + { + "epoch": 0.91, + "learning_rate": 4.4442501045680595e-07, + "loss": 1.0935, + "step": 37740 + }, + { + "epoch": 0.91, + "learning_rate": 4.4419541764151974e-07, + "loss": 1.1012, + "step": 37741 + }, + { + "epoch": 0.91, + "learning_rate": 4.439658827990523e-07, + "loss": 1.0856, + "step": 37742 + }, + { + "epoch": 0.91, + "learning_rate": 4.4373640593079695e-07, + "loss": 0.9208, + "step": 37743 + }, + { + "epoch": 0.91, + "learning_rate": 4.4350698703814476e-07, + "loss": 1.1417, + "step": 37744 + }, + { + "epoch": 0.91, + "learning_rate": 4.4327762612248806e-07, + "loss": 1.0786, + "step": 37745 + }, + { + "epoch": 0.91, + "learning_rate": 4.4304832318521783e-07, + "loss": 1.3192, + "step": 37746 + }, + { + "epoch": 0.91, + "learning_rate": 4.428190782277242e-07, + "loss": 0.9336, + "step": 37747 + }, + { + "epoch": 0.91, + "learning_rate": 4.425898912514015e-07, + "loss": 0.968, + "step": 37748 + }, + { + "epoch": 0.91, + "learning_rate": 4.4236076225763646e-07, + "loss": 0.9353, + "step": 37749 + }, + { + "epoch": 0.91, + "learning_rate": 4.421316912478202e-07, + "loss": 0.9806, + "step": 37750 + }, + { + "epoch": 0.91, + "learning_rate": 4.419026782233449e-07, + "loss": 1.1933, + "step": 37751 + }, + { + "epoch": 0.91, + "learning_rate": 4.416737231855961e-07, + "loss": 0.9691, + "step": 37752 + }, + { + "epoch": 0.91, + "learning_rate": 4.4144482613596493e-07, + "loss": 1.0628, + "step": 37753 + }, + { + "epoch": 0.91, + "learning_rate": 4.4121598707583923e-07, + "loss": 1.0744, + "step": 37754 + }, + { + "epoch": 0.91, + "learning_rate": 4.409872060066078e-07, + "loss": 0.9595, + "step": 37755 + }, + { + "epoch": 0.91, + "learning_rate": 4.4075848292966073e-07, + "loss": 0.9654, + "step": 37756 + }, + { + "epoch": 0.91, + "learning_rate": 4.405298178463813e-07, + "loss": 1.1209, + "step": 37757 + }, + { + "epoch": 0.91, + "learning_rate": 4.4030121075815837e-07, + "loss": 1.1153, + "step": 37758 + }, + { + "epoch": 0.91, + "learning_rate": 4.4007266166638196e-07, + "loss": 1.164, + "step": 37759 + }, + { + "epoch": 0.91, + "learning_rate": 4.3984417057243433e-07, + "loss": 1.1567, + "step": 37760 + }, + { + "epoch": 0.91, + "learning_rate": 4.3961573747770436e-07, + "loss": 1.2722, + "step": 37761 + }, + { + "epoch": 0.91, + "learning_rate": 4.3938736238357535e-07, + "loss": 1.0518, + "step": 37762 + }, + { + "epoch": 0.91, + "learning_rate": 4.391590452914352e-07, + "loss": 1.2373, + "step": 37763 + }, + { + "epoch": 0.91, + "learning_rate": 4.389307862026693e-07, + "loss": 1.1338, + "step": 37764 + }, + { + "epoch": 0.91, + "learning_rate": 4.3870258511865993e-07, + "loss": 1.2284, + "step": 37765 + }, + { + "epoch": 0.91, + "learning_rate": 4.384744420407927e-07, + "loss": 0.8985, + "step": 37766 + }, + { + "epoch": 0.91, + "learning_rate": 4.3824635697045424e-07, + "loss": 0.9637, + "step": 37767 + }, + { + "epoch": 0.91, + "learning_rate": 4.380183299090246e-07, + "loss": 1.1028, + "step": 37768 + }, + { + "epoch": 0.91, + "learning_rate": 4.377903608578893e-07, + "loss": 1.063, + "step": 37769 + }, + { + "epoch": 0.91, + "learning_rate": 4.3756244981843055e-07, + "loss": 1.0596, + "step": 37770 + }, + { + "epoch": 0.91, + "learning_rate": 4.373345967920295e-07, + "loss": 0.912, + "step": 37771 + }, + { + "epoch": 0.91, + "learning_rate": 4.371068017800728e-07, + "loss": 1.1076, + "step": 37772 + }, + { + "epoch": 0.91, + "learning_rate": 4.368790647839383e-07, + "loss": 0.9658, + "step": 37773 + }, + { + "epoch": 0.91, + "learning_rate": 4.366513858050092e-07, + "loss": 1.0239, + "step": 37774 + }, + { + "epoch": 0.91, + "learning_rate": 4.3642376484466785e-07, + "loss": 1.0794, + "step": 37775 + }, + { + "epoch": 0.91, + "learning_rate": 4.3619620190429204e-07, + "loss": 1.0471, + "step": 37776 + }, + { + "epoch": 0.91, + "learning_rate": 4.35968696985265e-07, + "loss": 1.0749, + "step": 37777 + }, + { + "epoch": 0.91, + "learning_rate": 4.3574125008896794e-07, + "loss": 0.9524, + "step": 37778 + }, + { + "epoch": 0.91, + "learning_rate": 4.3551386121677753e-07, + "loss": 1.0972, + "step": 37779 + }, + { + "epoch": 0.91, + "learning_rate": 4.352865303700748e-07, + "loss": 1.006, + "step": 37780 + }, + { + "epoch": 0.91, + "learning_rate": 4.3505925755023883e-07, + "loss": 1.1022, + "step": 37781 + }, + { + "epoch": 0.91, + "learning_rate": 4.3483204275864943e-07, + "loss": 1.0165, + "step": 37782 + }, + { + "epoch": 0.91, + "learning_rate": 4.346048859966845e-07, + "loss": 1.1722, + "step": 37783 + }, + { + "epoch": 0.91, + "learning_rate": 4.3437778726571955e-07, + "loss": 1.3085, + "step": 37784 + }, + { + "epoch": 0.91, + "learning_rate": 4.3415074656713574e-07, + "loss": 0.9952, + "step": 37785 + }, + { + "epoch": 0.91, + "learning_rate": 4.3392376390231085e-07, + "loss": 0.9529, + "step": 37786 + }, + { + "epoch": 0.91, + "learning_rate": 4.336968392726182e-07, + "loss": 0.948, + "step": 37787 + }, + { + "epoch": 0.91, + "learning_rate": 4.3346997267943894e-07, + "loss": 0.9021, + "step": 37788 + }, + { + "epoch": 0.91, + "learning_rate": 4.3324316412414525e-07, + "loss": 0.9934, + "step": 37789 + }, + { + "epoch": 0.91, + "learning_rate": 4.33016413608115e-07, + "loss": 0.4762, + "step": 37790 + }, + { + "epoch": 0.91, + "learning_rate": 4.327897211327259e-07, + "loss": 1.0707, + "step": 37791 + }, + { + "epoch": 0.91, + "learning_rate": 4.325630866993491e-07, + "loss": 1.1252, + "step": 37792 + }, + { + "epoch": 0.91, + "learning_rate": 4.3233651030936244e-07, + "loss": 1.0083, + "step": 37793 + }, + { + "epoch": 0.91, + "learning_rate": 4.3210999196414027e-07, + "loss": 1.3598, + "step": 37794 + }, + { + "epoch": 0.91, + "learning_rate": 4.3188353166505493e-07, + "loss": 1.1987, + "step": 37795 + }, + { + "epoch": 0.91, + "learning_rate": 4.3165712941348305e-07, + "loss": 1.0282, + "step": 37796 + }, + { + "epoch": 0.91, + "learning_rate": 4.314307852107957e-07, + "loss": 0.7845, + "step": 37797 + }, + { + "epoch": 0.91, + "learning_rate": 4.3120449905836746e-07, + "loss": 1.0629, + "step": 37798 + }, + { + "epoch": 0.91, + "learning_rate": 4.3097827095757163e-07, + "loss": 0.7875, + "step": 37799 + }, + { + "epoch": 0.91, + "learning_rate": 4.3075210090977814e-07, + "loss": 0.9787, + "step": 37800 + }, + { + "epoch": 0.91, + "learning_rate": 4.3052598891636157e-07, + "loss": 1.0399, + "step": 37801 + }, + { + "epoch": 0.91, + "learning_rate": 4.3029993497869406e-07, + "loss": 1.1192, + "step": 37802 + }, + { + "epoch": 0.91, + "learning_rate": 4.300739390981445e-07, + "loss": 1.0379, + "step": 37803 + }, + { + "epoch": 0.91, + "learning_rate": 4.2984800127608637e-07, + "loss": 0.8611, + "step": 37804 + }, + { + "epoch": 0.91, + "learning_rate": 4.296221215138885e-07, + "loss": 0.9802, + "step": 37805 + }, + { + "epoch": 0.91, + "learning_rate": 4.2939629981292197e-07, + "loss": 1.1576, + "step": 37806 + }, + { + "epoch": 0.91, + "learning_rate": 4.2917053617455905e-07, + "loss": 0.6977, + "step": 37807 + }, + { + "epoch": 0.91, + "learning_rate": 4.289448306001653e-07, + "loss": 0.9774, + "step": 37808 + }, + { + "epoch": 0.91, + "learning_rate": 4.2871918309111194e-07, + "loss": 1.1503, + "step": 37809 + }, + { + "epoch": 0.91, + "learning_rate": 4.2849359364876885e-07, + "loss": 0.998, + "step": 37810 + }, + { + "epoch": 0.91, + "learning_rate": 4.282680622745028e-07, + "loss": 0.8924, + "step": 37811 + }, + { + "epoch": 0.91, + "learning_rate": 4.280425889696849e-07, + "loss": 1.053, + "step": 37812 + }, + { + "epoch": 0.91, + "learning_rate": 4.2781717373567955e-07, + "loss": 0.9589, + "step": 37813 + }, + { + "epoch": 0.91, + "learning_rate": 4.275918165738557e-07, + "loss": 1.0472, + "step": 37814 + }, + { + "epoch": 0.91, + "learning_rate": 4.2736651748558235e-07, + "loss": 1.0234, + "step": 37815 + }, + { + "epoch": 0.91, + "learning_rate": 4.271412764722227e-07, + "loss": 0.9293, + "step": 37816 + }, + { + "epoch": 0.91, + "learning_rate": 4.2691609353514576e-07, + "loss": 1.0224, + "step": 37817 + }, + { + "epoch": 0.91, + "learning_rate": 4.2669096867571813e-07, + "loss": 1.2047, + "step": 37818 + }, + { + "epoch": 0.91, + "learning_rate": 4.2646590189530434e-07, + "loss": 1.2594, + "step": 37819 + }, + { + "epoch": 0.91, + "learning_rate": 4.2624089319526883e-07, + "loss": 0.9776, + "step": 37820 + }, + { + "epoch": 0.91, + "learning_rate": 4.260159425769794e-07, + "loss": 0.8156, + "step": 37821 + }, + { + "epoch": 0.91, + "learning_rate": 4.2579105004179944e-07, + "loss": 0.8316, + "step": 37822 + }, + { + "epoch": 0.91, + "learning_rate": 4.2556621559109337e-07, + "loss": 1.028, + "step": 37823 + }, + { + "epoch": 0.91, + "learning_rate": 4.253414392262234e-07, + "loss": 1.3004, + "step": 37824 + }, + { + "epoch": 0.91, + "learning_rate": 4.2511672094855514e-07, + "loss": 0.7377, + "step": 37825 + }, + { + "epoch": 0.91, + "learning_rate": 4.248920607594531e-07, + "loss": 1.0401, + "step": 37826 + }, + { + "epoch": 0.91, + "learning_rate": 4.2466745866027726e-07, + "loss": 1.1788, + "step": 37827 + }, + { + "epoch": 0.91, + "learning_rate": 4.2444291465239096e-07, + "loss": 0.7982, + "step": 37828 + }, + { + "epoch": 0.91, + "learning_rate": 4.242184287371598e-07, + "loss": 0.9843, + "step": 37829 + }, + { + "epoch": 0.91, + "learning_rate": 4.2399400091594154e-07, + "loss": 1.0399, + "step": 37830 + }, + { + "epoch": 0.91, + "learning_rate": 4.2376963119009963e-07, + "loss": 1.082, + "step": 37831 + }, + { + "epoch": 0.91, + "learning_rate": 4.2354531956099285e-07, + "loss": 1.0863, + "step": 37832 + }, + { + "epoch": 0.91, + "learning_rate": 4.233210660299869e-07, + "loss": 1.0315, + "step": 37833 + }, + { + "epoch": 0.91, + "learning_rate": 4.230968705984384e-07, + "loss": 0.9735, + "step": 37834 + }, + { + "epoch": 0.91, + "learning_rate": 4.228727332677074e-07, + "loss": 0.8909, + "step": 37835 + }, + { + "epoch": 0.91, + "learning_rate": 4.226486540391572e-07, + "loss": 0.8841, + "step": 37836 + }, + { + "epoch": 0.91, + "learning_rate": 4.224246329141435e-07, + "loss": 0.8587, + "step": 37837 + }, + { + "epoch": 0.91, + "learning_rate": 4.222006698940262e-07, + "loss": 0.9442, + "step": 37838 + }, + { + "epoch": 0.91, + "learning_rate": 4.219767649801665e-07, + "loss": 1.2686, + "step": 37839 + }, + { + "epoch": 0.91, + "learning_rate": 4.217529181739188e-07, + "loss": 1.148, + "step": 37840 + }, + { + "epoch": 0.91, + "learning_rate": 4.2152912947664546e-07, + "loss": 1.0052, + "step": 37841 + }, + { + "epoch": 0.91, + "learning_rate": 4.213053988896998e-07, + "loss": 1.1262, + "step": 37842 + }, + { + "epoch": 0.91, + "learning_rate": 4.2108172641444176e-07, + "loss": 1.0876, + "step": 37843 + }, + { + "epoch": 0.91, + "learning_rate": 4.208581120522293e-07, + "loss": 0.98, + "step": 37844 + }, + { + "epoch": 0.91, + "learning_rate": 4.206345558044156e-07, + "loss": 1.0621, + "step": 37845 + }, + { + "epoch": 0.91, + "learning_rate": 4.204110576723597e-07, + "loss": 0.8884, + "step": 37846 + }, + { + "epoch": 0.91, + "learning_rate": 4.201876176574171e-07, + "loss": 1.1766, + "step": 37847 + }, + { + "epoch": 0.91, + "learning_rate": 4.199642357609412e-07, + "loss": 1.0777, + "step": 37848 + }, + { + "epoch": 0.91, + "learning_rate": 4.1974091198429103e-07, + "loss": 0.9454, + "step": 37849 + }, + { + "epoch": 0.91, + "learning_rate": 4.1951764632881864e-07, + "loss": 0.9276, + "step": 37850 + }, + { + "epoch": 0.91, + "learning_rate": 4.192944387958786e-07, + "loss": 1.1603, + "step": 37851 + }, + { + "epoch": 0.91, + "learning_rate": 4.190712893868265e-07, + "loss": 1.0152, + "step": 37852 + }, + { + "epoch": 0.91, + "learning_rate": 4.1884819810301456e-07, + "loss": 1.0792, + "step": 37853 + }, + { + "epoch": 0.91, + "learning_rate": 4.186251649457962e-07, + "loss": 1.1137, + "step": 37854 + }, + { + "epoch": 0.91, + "learning_rate": 4.184021899165269e-07, + "loss": 0.8648, + "step": 37855 + }, + { + "epoch": 0.91, + "learning_rate": 4.181792730165568e-07, + "loss": 1.0553, + "step": 37856 + }, + { + "epoch": 0.91, + "learning_rate": 4.179564142472392e-07, + "loss": 0.8272, + "step": 37857 + }, + { + "epoch": 0.91, + "learning_rate": 4.1773361360992857e-07, + "loss": 1.291, + "step": 37858 + }, + { + "epoch": 0.91, + "learning_rate": 4.175108711059717e-07, + "loss": 0.9676, + "step": 37859 + }, + { + "epoch": 0.91, + "learning_rate": 4.17288186736724e-07, + "loss": 0.9596, + "step": 37860 + }, + { + "epoch": 0.91, + "learning_rate": 4.170655605035334e-07, + "loss": 1.0688, + "step": 37861 + }, + { + "epoch": 0.91, + "learning_rate": 4.1684299240775215e-07, + "loss": 1.1042, + "step": 37862 + }, + { + "epoch": 0.91, + "learning_rate": 4.166204824507314e-07, + "loss": 1.141, + "step": 37863 + }, + { + "epoch": 0.91, + "learning_rate": 4.1639803063381887e-07, + "loss": 0.7564, + "step": 37864 + }, + { + "epoch": 0.91, + "learning_rate": 4.161756369583658e-07, + "loss": 1.0591, + "step": 37865 + }, + { + "epoch": 0.91, + "learning_rate": 4.159533014257211e-07, + "loss": 0.9202, + "step": 37866 + }, + { + "epoch": 0.91, + "learning_rate": 4.1573102403723254e-07, + "loss": 1.0245, + "step": 37867 + }, + { + "epoch": 0.91, + "learning_rate": 4.1550880479425125e-07, + "loss": 0.9966, + "step": 37868 + }, + { + "epoch": 0.91, + "learning_rate": 4.1528664369812287e-07, + "loss": 1.1593, + "step": 37869 + }, + { + "epoch": 0.91, + "learning_rate": 4.150645407501952e-07, + "loss": 0.7694, + "step": 37870 + }, + { + "epoch": 0.91, + "learning_rate": 4.1484249595181714e-07, + "loss": 0.9778, + "step": 37871 + }, + { + "epoch": 0.91, + "learning_rate": 4.1462050930433426e-07, + "loss": 1.1056, + "step": 37872 + }, + { + "epoch": 0.91, + "learning_rate": 4.1439858080909444e-07, + "loss": 0.9018, + "step": 37873 + }, + { + "epoch": 0.91, + "learning_rate": 4.1417671046744545e-07, + "loss": 1.277, + "step": 37874 + }, + { + "epoch": 0.91, + "learning_rate": 4.1395489828072954e-07, + "loss": 1.179, + "step": 37875 + }, + { + "epoch": 0.91, + "learning_rate": 4.1373314425029565e-07, + "loss": 0.9873, + "step": 37876 + }, + { + "epoch": 0.91, + "learning_rate": 4.1351144837748713e-07, + "loss": 1.055, + "step": 37877 + }, + { + "epoch": 0.91, + "learning_rate": 4.132898106636496e-07, + "loss": 0.8161, + "step": 37878 + }, + { + "epoch": 0.91, + "learning_rate": 4.130682311101297e-07, + "loss": 1.1862, + "step": 37879 + }, + { + "epoch": 0.91, + "learning_rate": 4.1284670971826755e-07, + "loss": 1.1009, + "step": 37880 + }, + { + "epoch": 0.91, + "learning_rate": 4.1262524648940984e-07, + "loss": 1.1197, + "step": 37881 + }, + { + "epoch": 0.91, + "learning_rate": 4.1240384142490097e-07, + "loss": 1.0576, + "step": 37882 + }, + { + "epoch": 0.91, + "learning_rate": 4.121824945260822e-07, + "loss": 0.9688, + "step": 37883 + }, + { + "epoch": 0.91, + "learning_rate": 4.1196120579429786e-07, + "loss": 1.1703, + "step": 37884 + }, + { + "epoch": 0.91, + "learning_rate": 4.117399752308882e-07, + "loss": 0.8649, + "step": 37885 + }, + { + "epoch": 0.91, + "learning_rate": 4.1151880283719637e-07, + "loss": 0.9195, + "step": 37886 + }, + { + "epoch": 0.91, + "learning_rate": 4.11297688614567e-07, + "loss": 1.1953, + "step": 37887 + }, + { + "epoch": 0.91, + "learning_rate": 4.110766325643367e-07, + "loss": 0.7973, + "step": 37888 + }, + { + "epoch": 0.91, + "learning_rate": 4.1085563468785005e-07, + "loss": 1.0064, + "step": 37889 + }, + { + "epoch": 0.91, + "learning_rate": 4.1063469498644705e-07, + "loss": 0.9463, + "step": 37890 + }, + { + "epoch": 0.91, + "learning_rate": 4.1041381346146656e-07, + "loss": 1.0906, + "step": 37891 + }, + { + "epoch": 0.91, + "learning_rate": 4.101929901142521e-07, + "loss": 1.0444, + "step": 37892 + }, + { + "epoch": 0.91, + "learning_rate": 4.0997222494613906e-07, + "loss": 1.0329, + "step": 37893 + }, + { + "epoch": 0.91, + "learning_rate": 4.097515179584699e-07, + "loss": 1.1034, + "step": 37894 + }, + { + "epoch": 0.91, + "learning_rate": 4.095308691525823e-07, + "loss": 0.9485, + "step": 37895 + }, + { + "epoch": 0.91, + "learning_rate": 4.093102785298153e-07, + "loss": 1.1885, + "step": 37896 + }, + { + "epoch": 0.91, + "learning_rate": 4.0908974609150555e-07, + "loss": 1.1098, + "step": 37897 + }, + { + "epoch": 0.91, + "learning_rate": 4.0886927183899416e-07, + "loss": 0.7987, + "step": 37898 + }, + { + "epoch": 0.91, + "learning_rate": 4.086488557736157e-07, + "loss": 1.0754, + "step": 37899 + }, + { + "epoch": 0.91, + "learning_rate": 4.08428497896709e-07, + "loss": 1.0443, + "step": 37900 + }, + { + "epoch": 0.91, + "learning_rate": 4.08208198209612e-07, + "loss": 0.8411, + "step": 37901 + }, + { + "epoch": 0.91, + "learning_rate": 4.0798795671365796e-07, + "loss": 1.0284, + "step": 37902 + }, + { + "epoch": 0.91, + "learning_rate": 4.077677734101859e-07, + "loss": 0.9353, + "step": 37903 + }, + { + "epoch": 0.91, + "learning_rate": 4.0754764830052915e-07, + "loss": 1.0674, + "step": 37904 + }, + { + "epoch": 0.91, + "learning_rate": 4.0732758138602446e-07, + "loss": 1.0628, + "step": 37905 + }, + { + "epoch": 0.91, + "learning_rate": 4.071075726680096e-07, + "loss": 1.0033, + "step": 37906 + }, + { + "epoch": 0.91, + "learning_rate": 4.068876221478135e-07, + "loss": 1.0956, + "step": 37907 + }, + { + "epoch": 0.91, + "learning_rate": 4.0666772982677514e-07, + "loss": 1.3172, + "step": 37908 + }, + { + "epoch": 0.91, + "learning_rate": 4.064478957062279e-07, + "loss": 1.2055, + "step": 37909 + }, + { + "epoch": 0.91, + "learning_rate": 4.0622811978750396e-07, + "loss": 1.094, + "step": 37910 + }, + { + "epoch": 0.91, + "learning_rate": 4.060084020719379e-07, + "loss": 1.103, + "step": 37911 + }, + { + "epoch": 0.91, + "learning_rate": 4.05788742560862e-07, + "loss": 0.8489, + "step": 37912 + }, + { + "epoch": 0.91, + "learning_rate": 4.055691412556084e-07, + "loss": 1.0678, + "step": 37913 + }, + { + "epoch": 0.91, + "learning_rate": 4.0534959815751174e-07, + "loss": 0.8515, + "step": 37914 + }, + { + "epoch": 0.91, + "learning_rate": 4.0513011326789975e-07, + "loss": 1.112, + "step": 37915 + }, + { + "epoch": 0.91, + "learning_rate": 4.0491068658810804e-07, + "loss": 1.1915, + "step": 37916 + }, + { + "epoch": 0.91, + "learning_rate": 4.046913181194667e-07, + "loss": 1.0242, + "step": 37917 + }, + { + "epoch": 0.91, + "learning_rate": 4.0447200786330577e-07, + "loss": 1.0794, + "step": 37918 + }, + { + "epoch": 0.91, + "learning_rate": 4.042527558209564e-07, + "loss": 0.9823, + "step": 37919 + }, + { + "epoch": 0.91, + "learning_rate": 4.040335619937474e-07, + "loss": 1.2824, + "step": 37920 + }, + { + "epoch": 0.91, + "learning_rate": 4.0381442638301017e-07, + "loss": 1.2352, + "step": 37921 + }, + { + "epoch": 0.91, + "learning_rate": 4.035953489900746e-07, + "loss": 0.8049, + "step": 37922 + }, + { + "epoch": 0.91, + "learning_rate": 4.033763298162674e-07, + "loss": 1.0606, + "step": 37923 + }, + { + "epoch": 0.91, + "learning_rate": 4.031573688629198e-07, + "loss": 1.1876, + "step": 37924 + }, + { + "epoch": 0.91, + "learning_rate": 4.0293846613135954e-07, + "loss": 1.0912, + "step": 37925 + }, + { + "epoch": 0.91, + "learning_rate": 4.027196216229123e-07, + "loss": 0.9027, + "step": 37926 + }, + { + "epoch": 0.91, + "learning_rate": 4.025008353389104e-07, + "loss": 1.034, + "step": 37927 + }, + { + "epoch": 0.91, + "learning_rate": 4.0228210728067595e-07, + "loss": 1.0543, + "step": 37928 + }, + { + "epoch": 0.91, + "learning_rate": 4.0206343744953914e-07, + "loss": 1.1369, + "step": 37929 + }, + { + "epoch": 0.91, + "learning_rate": 4.018448258468266e-07, + "loss": 1.2933, + "step": 37930 + }, + { + "epoch": 0.91, + "learning_rate": 4.01626272473864e-07, + "loss": 1.1104, + "step": 37931 + }, + { + "epoch": 0.91, + "learning_rate": 4.014077773319758e-07, + "loss": 0.9591, + "step": 37932 + }, + { + "epoch": 0.91, + "learning_rate": 4.011893404224909e-07, + "loss": 1.1323, + "step": 37933 + }, + { + "epoch": 0.91, + "learning_rate": 4.0097096174673055e-07, + "loss": 1.1881, + "step": 37934 + }, + { + "epoch": 0.91, + "learning_rate": 4.0075264130602256e-07, + "loss": 0.8867, + "step": 37935 + }, + { + "epoch": 0.91, + "learning_rate": 4.005343791016902e-07, + "loss": 1.0879, + "step": 37936 + }, + { + "epoch": 0.91, + "learning_rate": 4.0031617513505704e-07, + "loss": 1.1178, + "step": 37937 + }, + { + "epoch": 0.91, + "learning_rate": 4.000980294074497e-07, + "loss": 1.0423, + "step": 37938 + }, + { + "epoch": 0.91, + "learning_rate": 3.998799419201871e-07, + "loss": 1.0223, + "step": 37939 + }, + { + "epoch": 0.91, + "learning_rate": 3.9966191267459596e-07, + "loss": 1.0885, + "step": 37940 + }, + { + "epoch": 0.91, + "learning_rate": 3.994439416719986e-07, + "loss": 0.9196, + "step": 37941 + }, + { + "epoch": 0.91, + "learning_rate": 3.992260289137151e-07, + "loss": 1.0501, + "step": 37942 + }, + { + "epoch": 0.91, + "learning_rate": 3.9900817440107097e-07, + "loss": 0.9572, + "step": 37943 + }, + { + "epoch": 0.91, + "learning_rate": 3.987903781353841e-07, + "loss": 1.0212, + "step": 37944 + }, + { + "epoch": 0.91, + "learning_rate": 3.985726401179779e-07, + "loss": 1.0451, + "step": 37945 + }, + { + "epoch": 0.91, + "learning_rate": 3.983549603501746e-07, + "loss": 1.0983, + "step": 37946 + }, + { + "epoch": 0.91, + "learning_rate": 3.981373388332921e-07, + "loss": 0.7867, + "step": 37947 + }, + { + "epoch": 0.91, + "learning_rate": 3.979197755686515e-07, + "loss": 1.17, + "step": 37948 + }, + { + "epoch": 0.91, + "learning_rate": 3.9770227055757394e-07, + "loss": 1.0091, + "step": 37949 + }, + { + "epoch": 0.91, + "learning_rate": 3.974848238013773e-07, + "loss": 1.1029, + "step": 37950 + }, + { + "epoch": 0.91, + "learning_rate": 3.9726743530138165e-07, + "loss": 1.203, + "step": 37951 + }, + { + "epoch": 0.91, + "learning_rate": 3.9705010505890705e-07, + "loss": 1.1334, + "step": 37952 + }, + { + "epoch": 0.91, + "learning_rate": 3.968328330752702e-07, + "loss": 1.1583, + "step": 37953 + }, + { + "epoch": 0.91, + "learning_rate": 3.9661561935179004e-07, + "loss": 1.0605, + "step": 37954 + }, + { + "epoch": 0.91, + "learning_rate": 3.963984638897833e-07, + "loss": 0.9949, + "step": 37955 + }, + { + "epoch": 0.91, + "learning_rate": 3.9618136669056783e-07, + "loss": 1.2811, + "step": 37956 + }, + { + "epoch": 0.91, + "learning_rate": 3.959643277554626e-07, + "loss": 0.9715, + "step": 37957 + }, + { + "epoch": 0.91, + "learning_rate": 3.9574734708578087e-07, + "loss": 0.9078, + "step": 37958 + }, + { + "epoch": 0.91, + "learning_rate": 3.9553042468284175e-07, + "loss": 1.1655, + "step": 37959 + }, + { + "epoch": 0.91, + "learning_rate": 3.9531356054796076e-07, + "loss": 0.8703, + "step": 37960 + }, + { + "epoch": 0.91, + "learning_rate": 3.9509675468245245e-07, + "loss": 1.2046, + "step": 37961 + }, + { + "epoch": 0.91, + "learning_rate": 3.9488000708763464e-07, + "loss": 1.0375, + "step": 37962 + }, + { + "epoch": 0.91, + "learning_rate": 3.946633177648185e-07, + "loss": 1.0357, + "step": 37963 + }, + { + "epoch": 0.91, + "learning_rate": 3.9444668671532185e-07, + "loss": 1.0335, + "step": 37964 + }, + { + "epoch": 0.91, + "learning_rate": 3.942301139404581e-07, + "loss": 0.7533, + "step": 37965 + }, + { + "epoch": 0.91, + "learning_rate": 3.940135994415395e-07, + "loss": 0.9044, + "step": 37966 + }, + { + "epoch": 0.91, + "learning_rate": 3.937971432198806e-07, + "loss": 1.2506, + "step": 37967 + }, + { + "epoch": 0.91, + "learning_rate": 3.9358074527679704e-07, + "loss": 1.2093, + "step": 37968 + }, + { + "epoch": 0.91, + "learning_rate": 3.933644056135988e-07, + "loss": 1.0891, + "step": 37969 + }, + { + "epoch": 0.91, + "learning_rate": 3.931481242315993e-07, + "loss": 1.0338, + "step": 37970 + }, + { + "epoch": 0.91, + "learning_rate": 3.929319011321087e-07, + "loss": 1.0242, + "step": 37971 + }, + { + "epoch": 0.91, + "learning_rate": 3.927157363164413e-07, + "loss": 1.0784, + "step": 37972 + }, + { + "epoch": 0.91, + "learning_rate": 3.924996297859096e-07, + "loss": 1.1656, + "step": 37973 + }, + { + "epoch": 0.91, + "learning_rate": 3.9228358154182134e-07, + "loss": 1.0174, + "step": 37974 + }, + { + "epoch": 0.91, + "learning_rate": 3.920675915854877e-07, + "loss": 1.0792, + "step": 37975 + }, + { + "epoch": 0.91, + "learning_rate": 3.9185165991822203e-07, + "loss": 0.9861, + "step": 37976 + }, + { + "epoch": 0.91, + "learning_rate": 3.916357865413312e-07, + "loss": 1.1136, + "step": 37977 + }, + { + "epoch": 0.91, + "learning_rate": 3.914199714561273e-07, + "loss": 1.0314, + "step": 37978 + }, + { + "epoch": 0.91, + "learning_rate": 3.9120421466391723e-07, + "loss": 1.0234, + "step": 37979 + }, + { + "epoch": 0.91, + "learning_rate": 3.909885161660099e-07, + "loss": 1.1361, + "step": 37980 + }, + { + "epoch": 0.91, + "learning_rate": 3.907728759637175e-07, + "loss": 0.955, + "step": 37981 + }, + { + "epoch": 0.91, + "learning_rate": 3.9055729405834464e-07, + "loss": 1.0882, + "step": 37982 + }, + { + "epoch": 0.91, + "learning_rate": 3.9034177045120026e-07, + "loss": 1.022, + "step": 37983 + }, + { + "epoch": 0.91, + "learning_rate": 3.9012630514359215e-07, + "loss": 0.8726, + "step": 37984 + }, + { + "epoch": 0.91, + "learning_rate": 3.8991089813682716e-07, + "loss": 0.9941, + "step": 37985 + }, + { + "epoch": 0.91, + "learning_rate": 3.8969554943221413e-07, + "loss": 1.1313, + "step": 37986 + }, + { + "epoch": 0.91, + "learning_rate": 3.8948025903105536e-07, + "loss": 0.9109, + "step": 37987 + }, + { + "epoch": 0.91, + "learning_rate": 3.8926502693465983e-07, + "loss": 1.3563, + "step": 37988 + }, + { + "epoch": 0.91, + "learning_rate": 3.890498531443332e-07, + "loss": 1.0158, + "step": 37989 + }, + { + "epoch": 0.91, + "learning_rate": 3.88834737661381e-07, + "loss": 0.8053, + "step": 37990 + }, + { + "epoch": 0.91, + "learning_rate": 3.8861968048710673e-07, + "loss": 1.2103, + "step": 37991 + }, + { + "epoch": 0.91, + "learning_rate": 3.884046816228171e-07, + "loss": 0.9263, + "step": 37992 + }, + { + "epoch": 0.91, + "learning_rate": 3.881897410698143e-07, + "loss": 1.0672, + "step": 37993 + }, + { + "epoch": 0.91, + "learning_rate": 3.879748588294041e-07, + "loss": 0.9787, + "step": 37994 + }, + { + "epoch": 0.91, + "learning_rate": 3.8776003490288984e-07, + "loss": 1.0793, + "step": 37995 + }, + { + "epoch": 0.91, + "learning_rate": 3.8754526929157377e-07, + "loss": 1.2028, + "step": 37996 + }, + { + "epoch": 0.91, + "learning_rate": 3.8733056199676157e-07, + "loss": 0.9819, + "step": 37997 + }, + { + "epoch": 0.91, + "learning_rate": 3.8711591301975105e-07, + "loss": 1.1784, + "step": 37998 + }, + { + "epoch": 0.91, + "learning_rate": 3.8690132236184896e-07, + "loss": 1.1421, + "step": 37999 + }, + { + "epoch": 0.91, + "learning_rate": 3.866867900243543e-07, + "loss": 0.9885, + "step": 38000 + }, + { + "epoch": 0.91, + "learning_rate": 3.8647231600856925e-07, + "loss": 1.0484, + "step": 38001 + }, + { + "epoch": 0.91, + "learning_rate": 3.8625790031579736e-07, + "loss": 1.123, + "step": 38002 + }, + { + "epoch": 0.91, + "learning_rate": 3.860435429473364e-07, + "loss": 1.0696, + "step": 38003 + }, + { + "epoch": 0.91, + "learning_rate": 3.858292439044875e-07, + "loss": 0.9881, + "step": 38004 + }, + { + "epoch": 0.91, + "learning_rate": 3.8561500318855304e-07, + "loss": 1.11, + "step": 38005 + }, + { + "epoch": 0.91, + "learning_rate": 3.8540082080082863e-07, + "loss": 1.0207, + "step": 38006 + }, + { + "epoch": 0.91, + "learning_rate": 3.8518669674261766e-07, + "loss": 1.1452, + "step": 38007 + }, + { + "epoch": 0.91, + "learning_rate": 3.849726310152169e-07, + "loss": 0.9609, + "step": 38008 + }, + { + "epoch": 0.91, + "learning_rate": 3.847586236199252e-07, + "loss": 1.0369, + "step": 38009 + }, + { + "epoch": 0.91, + "learning_rate": 3.8454467455804165e-07, + "loss": 1.2646, + "step": 38010 + }, + { + "epoch": 0.91, + "learning_rate": 3.843307838308641e-07, + "loss": 1.0349, + "step": 38011 + }, + { + "epoch": 0.91, + "learning_rate": 3.8411695143968917e-07, + "loss": 0.9202, + "step": 38012 + }, + { + "epoch": 0.91, + "learning_rate": 3.8390317738581596e-07, + "loss": 0.9775, + "step": 38013 + }, + { + "epoch": 0.91, + "learning_rate": 3.836894616705389e-07, + "loss": 0.9209, + "step": 38014 + }, + { + "epoch": 0.91, + "learning_rate": 3.8347580429515807e-07, + "loss": 1.1221, + "step": 38015 + }, + { + "epoch": 0.91, + "learning_rate": 3.832622052609658e-07, + "loss": 1.0302, + "step": 38016 + }, + { + "epoch": 0.91, + "learning_rate": 3.8304866456925993e-07, + "loss": 0.971, + "step": 38017 + }, + { + "epoch": 0.91, + "learning_rate": 3.8283518222133717e-07, + "loss": 0.659, + "step": 38018 + }, + { + "epoch": 0.91, + "learning_rate": 3.826217582184899e-07, + "loss": 1.0517, + "step": 38019 + }, + { + "epoch": 0.91, + "learning_rate": 3.8240839256201365e-07, + "loss": 0.8544, + "step": 38020 + }, + { + "epoch": 0.91, + "learning_rate": 3.8219508525320523e-07, + "loss": 0.9952, + "step": 38021 + }, + { + "epoch": 0.91, + "learning_rate": 3.8198183629335586e-07, + "loss": 0.9336, + "step": 38022 + }, + { + "epoch": 0.91, + "learning_rate": 3.817686456837622e-07, + "loss": 0.9818, + "step": 38023 + }, + { + "epoch": 0.91, + "learning_rate": 3.8155551342571316e-07, + "loss": 1.0065, + "step": 38024 + }, + { + "epoch": 0.91, + "learning_rate": 3.8134243952050567e-07, + "loss": 0.9662, + "step": 38025 + }, + { + "epoch": 0.91, + "learning_rate": 3.8112942396943185e-07, + "loss": 1.0445, + "step": 38026 + }, + { + "epoch": 0.91, + "learning_rate": 3.809164667737819e-07, + "loss": 0.8886, + "step": 38027 + }, + { + "epoch": 0.91, + "learning_rate": 3.807035679348492e-07, + "loss": 0.7465, + "step": 38028 + }, + { + "epoch": 0.91, + "learning_rate": 3.804907274539271e-07, + "loss": 1.1713, + "step": 38029 + }, + { + "epoch": 0.91, + "learning_rate": 3.8027794533230354e-07, + "loss": 1.0541, + "step": 38030 + }, + { + "epoch": 0.91, + "learning_rate": 3.800652215712708e-07, + "loss": 1.1194, + "step": 38031 + }, + { + "epoch": 0.91, + "learning_rate": 3.7985255617212113e-07, + "loss": 0.8354, + "step": 38032 + }, + { + "epoch": 0.91, + "learning_rate": 3.7963994913614244e-07, + "loss": 0.8226, + "step": 38033 + }, + { + "epoch": 0.91, + "learning_rate": 3.794274004646259e-07, + "loss": 0.7298, + "step": 38034 + }, + { + "epoch": 0.91, + "learning_rate": 3.792149101588594e-07, + "loss": 0.8573, + "step": 38035 + }, + { + "epoch": 0.91, + "learning_rate": 3.7900247822013294e-07, + "loss": 0.7458, + "step": 38036 + }, + { + "epoch": 0.91, + "learning_rate": 3.7879010464973666e-07, + "loss": 1.2174, + "step": 38037 + }, + { + "epoch": 0.91, + "learning_rate": 3.785777894489562e-07, + "loss": 0.9661, + "step": 38038 + }, + { + "epoch": 0.91, + "learning_rate": 3.783655326190816e-07, + "loss": 0.8661, + "step": 38039 + }, + { + "epoch": 0.91, + "learning_rate": 3.7815333416140187e-07, + "loss": 1.1106, + "step": 38040 + }, + { + "epoch": 0.91, + "learning_rate": 3.7794119407720155e-07, + "loss": 0.9261, + "step": 38041 + }, + { + "epoch": 0.91, + "learning_rate": 3.7772911236776846e-07, + "loss": 1.0383, + "step": 38042 + }, + { + "epoch": 0.91, + "learning_rate": 3.775170890343893e-07, + "loss": 1.159, + "step": 38043 + }, + { + "epoch": 0.91, + "learning_rate": 3.7730512407835095e-07, + "loss": 1.153, + "step": 38044 + }, + { + "epoch": 0.91, + "learning_rate": 3.7709321750094007e-07, + "loss": 0.7425, + "step": 38045 + }, + { + "epoch": 0.91, + "learning_rate": 3.768813693034401e-07, + "loss": 0.9125, + "step": 38046 + }, + { + "epoch": 0.91, + "learning_rate": 3.7666957948713665e-07, + "loss": 1.0717, + "step": 38047 + }, + { + "epoch": 0.92, + "learning_rate": 3.7645784805331654e-07, + "loss": 0.909, + "step": 38048 + }, + { + "epoch": 0.92, + "learning_rate": 3.762461750032631e-07, + "loss": 1.2041, + "step": 38049 + }, + { + "epoch": 0.92, + "learning_rate": 3.760345603382609e-07, + "loss": 1.275, + "step": 38050 + }, + { + "epoch": 0.92, + "learning_rate": 3.758230040595912e-07, + "loss": 1.261, + "step": 38051 + }, + { + "epoch": 0.92, + "learning_rate": 3.756115061685406e-07, + "loss": 0.9112, + "step": 38052 + }, + { + "epoch": 0.92, + "learning_rate": 3.754000666663926e-07, + "loss": 0.8851, + "step": 38053 + }, + { + "epoch": 0.92, + "learning_rate": 3.7518868555442733e-07, + "loss": 1.072, + "step": 38054 + }, + { + "epoch": 0.92, + "learning_rate": 3.749773628339282e-07, + "loss": 0.763, + "step": 38055 + }, + { + "epoch": 0.92, + "learning_rate": 3.747660985061785e-07, + "loss": 0.8694, + "step": 38056 + }, + { + "epoch": 0.92, + "learning_rate": 3.7455489257245736e-07, + "loss": 0.9384, + "step": 38057 + }, + { + "epoch": 0.92, + "learning_rate": 3.743437450340492e-07, + "loss": 0.9085, + "step": 38058 + }, + { + "epoch": 0.92, + "learning_rate": 3.7413265589223203e-07, + "loss": 0.9058, + "step": 38059 + }, + { + "epoch": 0.92, + "learning_rate": 3.73921625148288e-07, + "loss": 0.8748, + "step": 38060 + }, + { + "epoch": 0.92, + "learning_rate": 3.737106528034995e-07, + "loss": 1.0424, + "step": 38061 + }, + { + "epoch": 0.92, + "learning_rate": 3.7349973885914214e-07, + "loss": 0.7529, + "step": 38062 + }, + { + "epoch": 0.92, + "learning_rate": 3.7328888331649714e-07, + "loss": 1.0349, + "step": 38063 + }, + { + "epoch": 0.92, + "learning_rate": 3.730780861768457e-07, + "loss": 0.9778, + "step": 38064 + }, + { + "epoch": 0.92, + "learning_rate": 3.7286734744146456e-07, + "loss": 1.1349, + "step": 38065 + }, + { + "epoch": 0.92, + "learning_rate": 3.7265666711163273e-07, + "loss": 0.8099, + "step": 38066 + }, + { + "epoch": 0.92, + "learning_rate": 3.72446045188628e-07, + "loss": 0.8018, + "step": 38067 + }, + { + "epoch": 0.92, + "learning_rate": 3.7223548167372833e-07, + "loss": 1.1152, + "step": 38068 + }, + { + "epoch": 0.92, + "learning_rate": 3.720249765682127e-07, + "loss": 0.9696, + "step": 38069 + }, + { + "epoch": 0.92, + "learning_rate": 3.718145298733566e-07, + "loss": 1.0516, + "step": 38070 + }, + { + "epoch": 0.92, + "learning_rate": 3.716041415904359e-07, + "loss": 0.9581, + "step": 38071 + }, + { + "epoch": 0.92, + "learning_rate": 3.7139381172072943e-07, + "loss": 0.7849, + "step": 38072 + }, + { + "epoch": 0.92, + "learning_rate": 3.7118354026551174e-07, + "loss": 0.9525, + "step": 38073 + }, + { + "epoch": 0.92, + "learning_rate": 3.709733272260574e-07, + "loss": 1.0777, + "step": 38074 + }, + { + "epoch": 0.92, + "learning_rate": 3.7076317260364426e-07, + "loss": 0.9527, + "step": 38075 + }, + { + "epoch": 0.92, + "learning_rate": 3.7055307639954574e-07, + "loss": 1.2065, + "step": 38076 + }, + { + "epoch": 0.92, + "learning_rate": 3.7034303861503753e-07, + "loss": 0.9201, + "step": 38077 + }, + { + "epoch": 0.92, + "learning_rate": 3.701330592513919e-07, + "loss": 0.917, + "step": 38078 + }, + { + "epoch": 0.92, + "learning_rate": 3.6992313830988337e-07, + "loss": 1.2167, + "step": 38079 + }, + { + "epoch": 0.92, + "learning_rate": 3.697132757917876e-07, + "loss": 1.1829, + "step": 38080 + }, + { + "epoch": 0.92, + "learning_rate": 3.6950347169837476e-07, + "loss": 0.9872, + "step": 38081 + }, + { + "epoch": 0.92, + "learning_rate": 3.692937260309193e-07, + "loss": 0.7935, + "step": 38082 + }, + { + "epoch": 0.92, + "learning_rate": 3.690840387906958e-07, + "loss": 0.7989, + "step": 38083 + }, + { + "epoch": 0.92, + "learning_rate": 3.6887440997897093e-07, + "loss": 1.008, + "step": 38084 + }, + { + "epoch": 0.92, + "learning_rate": 3.686648395970227e-07, + "loss": 1.2068, + "step": 38085 + }, + { + "epoch": 0.92, + "learning_rate": 3.684553276461167e-07, + "loss": 0.7483, + "step": 38086 + }, + { + "epoch": 0.92, + "learning_rate": 3.682458741275274e-07, + "loss": 0.9443, + "step": 38087 + }, + { + "epoch": 0.92, + "learning_rate": 3.6803647904252613e-07, + "loss": 1.2216, + "step": 38088 + }, + { + "epoch": 0.92, + "learning_rate": 3.678271423923807e-07, + "loss": 0.8177, + "step": 38089 + }, + { + "epoch": 0.92, + "learning_rate": 3.676178641783623e-07, + "loss": 0.9358, + "step": 38090 + }, + { + "epoch": 0.92, + "learning_rate": 3.6740864440174215e-07, + "loss": 0.9018, + "step": 38091 + }, + { + "epoch": 0.92, + "learning_rate": 3.6719948306378707e-07, + "loss": 0.8746, + "step": 38092 + }, + { + "epoch": 0.92, + "learning_rate": 3.6699038016576703e-07, + "loss": 1.0425, + "step": 38093 + }, + { + "epoch": 0.92, + "learning_rate": 3.6678133570895004e-07, + "loss": 0.8122, + "step": 38094 + }, + { + "epoch": 0.92, + "learning_rate": 3.66572349694605e-07, + "loss": 0.9144, + "step": 38095 + }, + { + "epoch": 0.92, + "learning_rate": 3.663634221240009e-07, + "loss": 1.0092, + "step": 38096 + }, + { + "epoch": 0.92, + "learning_rate": 3.661545529984023e-07, + "loss": 1.0238, + "step": 38097 + }, + { + "epoch": 0.92, + "learning_rate": 3.659457423190782e-07, + "loss": 1.0031, + "step": 38098 + }, + { + "epoch": 0.92, + "learning_rate": 3.6573699008729646e-07, + "loss": 0.9738, + "step": 38099 + }, + { + "epoch": 0.92, + "learning_rate": 3.655282963043205e-07, + "loss": 0.8597, + "step": 38100 + }, + { + "epoch": 0.92, + "learning_rate": 3.6531966097142045e-07, + "loss": 0.6894, + "step": 38101 + }, + { + "epoch": 0.92, + "learning_rate": 3.651110840898575e-07, + "loss": 1.1067, + "step": 38102 + }, + { + "epoch": 0.92, + "learning_rate": 3.6490256566090067e-07, + "loss": 1.0041, + "step": 38103 + }, + { + "epoch": 0.92, + "learning_rate": 3.6469410568581334e-07, + "loss": 1.1156, + "step": 38104 + }, + { + "epoch": 0.92, + "learning_rate": 3.6448570416586005e-07, + "loss": 0.8991, + "step": 38105 + }, + { + "epoch": 0.92, + "learning_rate": 3.6427736110230537e-07, + "loss": 0.9241, + "step": 38106 + }, + { + "epoch": 0.92, + "learning_rate": 3.64069076496415e-07, + "loss": 1.377, + "step": 38107 + }, + { + "epoch": 0.92, + "learning_rate": 3.6386085034945005e-07, + "loss": 1.0412, + "step": 38108 + }, + { + "epoch": 0.92, + "learning_rate": 3.6365268266267516e-07, + "loss": 1.1172, + "step": 38109 + }, + { + "epoch": 0.92, + "learning_rate": 3.6344457343735265e-07, + "loss": 1.117, + "step": 38110 + }, + { + "epoch": 0.92, + "learning_rate": 3.6323652267474475e-07, + "loss": 0.8367, + "step": 38111 + }, + { + "epoch": 0.92, + "learning_rate": 3.63028530376115e-07, + "loss": 0.9278, + "step": 38112 + }, + { + "epoch": 0.92, + "learning_rate": 3.628205965427234e-07, + "loss": 0.9428, + "step": 38113 + }, + { + "epoch": 0.92, + "learning_rate": 3.6261272117583346e-07, + "loss": 0.8859, + "step": 38114 + }, + { + "epoch": 0.92, + "learning_rate": 3.6240490427670525e-07, + "loss": 1.1626, + "step": 38115 + }, + { + "epoch": 0.92, + "learning_rate": 3.621971458465989e-07, + "loss": 1.0451, + "step": 38116 + }, + { + "epoch": 0.92, + "learning_rate": 3.6198944588677454e-07, + "loss": 0.9926, + "step": 38117 + }, + { + "epoch": 0.92, + "learning_rate": 3.617818043984955e-07, + "loss": 0.8617, + "step": 38118 + }, + { + "epoch": 0.92, + "learning_rate": 3.6157422138301754e-07, + "loss": 1.0438, + "step": 38119 + }, + { + "epoch": 0.92, + "learning_rate": 3.61366696841603e-07, + "loss": 1.141, + "step": 38120 + }, + { + "epoch": 0.92, + "learning_rate": 3.6115923077550853e-07, + "loss": 1.1285, + "step": 38121 + }, + { + "epoch": 0.92, + "learning_rate": 3.609518231859943e-07, + "loss": 0.8479, + "step": 38122 + }, + { + "epoch": 0.92, + "learning_rate": 3.607444740743182e-07, + "loss": 1.1483, + "step": 38123 + }, + { + "epoch": 0.92, + "learning_rate": 3.60537183441737e-07, + "loss": 0.9993, + "step": 38124 + }, + { + "epoch": 0.92, + "learning_rate": 3.6032995128951086e-07, + "loss": 1.1548, + "step": 38125 + }, + { + "epoch": 0.92, + "learning_rate": 3.6012277761889536e-07, + "loss": 0.9076, + "step": 38126 + }, + { + "epoch": 0.92, + "learning_rate": 3.5991566243114617e-07, + "loss": 1.2412, + "step": 38127 + }, + { + "epoch": 0.92, + "learning_rate": 3.5970860572752343e-07, + "loss": 1.1748, + "step": 38128 + }, + { + "epoch": 0.92, + "learning_rate": 3.595016075092783e-07, + "loss": 1.0155, + "step": 38129 + }, + { + "epoch": 0.92, + "learning_rate": 3.59294667777671e-07, + "loss": 0.8568, + "step": 38130 + }, + { + "epoch": 0.92, + "learning_rate": 3.5908778653395595e-07, + "loss": 1.0947, + "step": 38131 + }, + { + "epoch": 0.92, + "learning_rate": 3.588809637793855e-07, + "loss": 0.806, + "step": 38132 + }, + { + "epoch": 0.92, + "learning_rate": 3.5867419951521764e-07, + "loss": 0.8493, + "step": 38133 + }, + { + "epoch": 0.92, + "learning_rate": 3.5846749374270683e-07, + "loss": 0.9043, + "step": 38134 + }, + { + "epoch": 0.92, + "learning_rate": 3.5826084646310434e-07, + "loss": 0.7587, + "step": 38135 + }, + { + "epoch": 0.92, + "learning_rate": 3.580542576776658e-07, + "loss": 1.0263, + "step": 38136 + }, + { + "epoch": 0.92, + "learning_rate": 3.578477273876435e-07, + "loss": 1.0134, + "step": 38137 + }, + { + "epoch": 0.92, + "learning_rate": 3.576412555942921e-07, + "loss": 0.8957, + "step": 38138 + }, + { + "epoch": 0.92, + "learning_rate": 3.574348422988638e-07, + "loss": 1.0107, + "step": 38139 + }, + { + "epoch": 0.92, + "learning_rate": 3.5722848750260886e-07, + "loss": 1.0551, + "step": 38140 + }, + { + "epoch": 0.92, + "learning_rate": 3.570221912067806e-07, + "loss": 0.8732, + "step": 38141 + }, + { + "epoch": 0.92, + "learning_rate": 3.5681595341263144e-07, + "loss": 0.9872, + "step": 38142 + }, + { + "epoch": 0.92, + "learning_rate": 3.566097741214103e-07, + "loss": 0.9998, + "step": 38143 + }, + { + "epoch": 0.92, + "learning_rate": 3.5640365333437065e-07, + "loss": 0.9059, + "step": 38144 + }, + { + "epoch": 0.92, + "learning_rate": 3.5619759105276155e-07, + "loss": 0.7641, + "step": 38145 + }, + { + "epoch": 0.92, + "learning_rate": 3.55991587277833e-07, + "loss": 1.0396, + "step": 38146 + }, + { + "epoch": 0.92, + "learning_rate": 3.557856420108352e-07, + "loss": 1.0059, + "step": 38147 + }, + { + "epoch": 0.92, + "learning_rate": 3.5557975525301713e-07, + "loss": 0.9829, + "step": 38148 + }, + { + "epoch": 0.92, + "learning_rate": 3.553739270056278e-07, + "loss": 1.1418, + "step": 38149 + }, + { + "epoch": 0.92, + "learning_rate": 3.551681572699184e-07, + "loss": 1.023, + "step": 38150 + }, + { + "epoch": 0.92, + "learning_rate": 3.549624460471335e-07, + "loss": 1.1364, + "step": 38151 + }, + { + "epoch": 0.92, + "learning_rate": 3.547567933385232e-07, + "loss": 1.0513, + "step": 38152 + }, + { + "epoch": 0.92, + "learning_rate": 3.545511991453343e-07, + "loss": 0.9998, + "step": 38153 + }, + { + "epoch": 0.92, + "learning_rate": 3.543456634688147e-07, + "loss": 0.9567, + "step": 38154 + }, + { + "epoch": 0.92, + "learning_rate": 3.541401863102112e-07, + "loss": 1.13, + "step": 38155 + }, + { + "epoch": 0.92, + "learning_rate": 3.539347676707705e-07, + "loss": 1.0121, + "step": 38156 + }, + { + "epoch": 0.92, + "learning_rate": 3.5372940755173836e-07, + "loss": 0.8825, + "step": 38157 + }, + { + "epoch": 0.92, + "learning_rate": 3.5352410595436146e-07, + "loss": 1.1065, + "step": 38158 + }, + { + "epoch": 0.92, + "learning_rate": 3.533188628798845e-07, + "loss": 0.9258, + "step": 38159 + }, + { + "epoch": 0.92, + "learning_rate": 3.53113678329553e-07, + "loss": 1.0969, + "step": 38160 + }, + { + "epoch": 0.92, + "learning_rate": 3.529085523046116e-07, + "loss": 1.0295, + "step": 38161 + }, + { + "epoch": 0.92, + "learning_rate": 3.527034848063049e-07, + "loss": 0.7995, + "step": 38162 + }, + { + "epoch": 0.92, + "learning_rate": 3.5249847583587735e-07, + "loss": 1.0398, + "step": 38163 + }, + { + "epoch": 0.92, + "learning_rate": 3.522935253945725e-07, + "loss": 0.9142, + "step": 38164 + }, + { + "epoch": 0.92, + "learning_rate": 3.5208863348363267e-07, + "loss": 1.1661, + "step": 38165 + }, + { + "epoch": 0.92, + "learning_rate": 3.518838001043012e-07, + "loss": 1.168, + "step": 38166 + }, + { + "epoch": 0.92, + "learning_rate": 3.5167902525782173e-07, + "loss": 1.1778, + "step": 38167 + }, + { + "epoch": 0.92, + "learning_rate": 3.5147430894543755e-07, + "loss": 1.1535, + "step": 38168 + }, + { + "epoch": 0.92, + "learning_rate": 3.512696511683888e-07, + "loss": 1.3157, + "step": 38169 + }, + { + "epoch": 0.92, + "learning_rate": 3.5106505192791684e-07, + "loss": 0.9008, + "step": 38170 + }, + { + "epoch": 0.92, + "learning_rate": 3.5086051122526387e-07, + "loss": 1.0939, + "step": 38171 + }, + { + "epoch": 0.92, + "learning_rate": 3.506560290616712e-07, + "loss": 0.9396, + "step": 38172 + }, + { + "epoch": 0.92, + "learning_rate": 3.5045160543837885e-07, + "loss": 1.0627, + "step": 38173 + }, + { + "epoch": 0.92, + "learning_rate": 3.5024724035662596e-07, + "loss": 0.9335, + "step": 38174 + }, + { + "epoch": 0.92, + "learning_rate": 3.5004293381765364e-07, + "loss": 1.2508, + "step": 38175 + }, + { + "epoch": 0.92, + "learning_rate": 3.49838685822701e-07, + "loss": 1.0372, + "step": 38176 + }, + { + "epoch": 0.92, + "learning_rate": 3.4963449637300694e-07, + "loss": 1.0002, + "step": 38177 + }, + { + "epoch": 0.92, + "learning_rate": 3.4943036546981056e-07, + "loss": 1.0016, + "step": 38178 + }, + { + "epoch": 0.92, + "learning_rate": 3.4922629311435083e-07, + "loss": 0.9466, + "step": 38179 + }, + { + "epoch": 0.92, + "learning_rate": 3.490222793078646e-07, + "loss": 0.7517, + "step": 38180 + }, + { + "epoch": 0.92, + "learning_rate": 3.488183240515919e-07, + "loss": 1.0883, + "step": 38181 + }, + { + "epoch": 0.92, + "learning_rate": 3.486144273467662e-07, + "loss": 0.7501, + "step": 38182 + }, + { + "epoch": 0.92, + "learning_rate": 3.4841058919462657e-07, + "loss": 1.2134, + "step": 38183 + }, + { + "epoch": 0.92, + "learning_rate": 3.4820680959641086e-07, + "loss": 1.0839, + "step": 38184 + }, + { + "epoch": 0.92, + "learning_rate": 3.480030885533536e-07, + "loss": 0.9233, + "step": 38185 + }, + { + "epoch": 0.92, + "learning_rate": 3.4779942606669057e-07, + "loss": 1.1231, + "step": 38186 + }, + { + "epoch": 0.92, + "learning_rate": 3.4759582213765964e-07, + "loss": 0.8334, + "step": 38187 + }, + { + "epoch": 0.92, + "learning_rate": 3.473922767674942e-07, + "loss": 1.3944, + "step": 38188 + }, + { + "epoch": 0.92, + "learning_rate": 3.471887899574289e-07, + "loss": 1.2414, + "step": 38189 + }, + { + "epoch": 0.92, + "learning_rate": 3.4698536170869825e-07, + "loss": 1.2695, + "step": 38190 + }, + { + "epoch": 0.92, + "learning_rate": 3.467819920225368e-07, + "loss": 1.3538, + "step": 38191 + }, + { + "epoch": 0.92, + "learning_rate": 3.465786809001803e-07, + "loss": 1.2262, + "step": 38192 + }, + { + "epoch": 0.92, + "learning_rate": 3.4637542834285887e-07, + "loss": 1.1355, + "step": 38193 + }, + { + "epoch": 0.92, + "learning_rate": 3.46172234351807e-07, + "loss": 1.2055, + "step": 38194 + }, + { + "epoch": 0.92, + "learning_rate": 3.4596909892825825e-07, + "loss": 0.9187, + "step": 38195 + }, + { + "epoch": 0.92, + "learning_rate": 3.4576602207344265e-07, + "loss": 0.7768, + "step": 38196 + }, + { + "epoch": 0.92, + "learning_rate": 3.4556300378859484e-07, + "loss": 1.0944, + "step": 38197 + }, + { + "epoch": 0.92, + "learning_rate": 3.453600440749461e-07, + "loss": 1.0549, + "step": 38198 + }, + { + "epoch": 0.92, + "learning_rate": 3.4515714293372636e-07, + "loss": 0.9469, + "step": 38199 + }, + { + "epoch": 0.92, + "learning_rate": 3.4495430036616706e-07, + "loss": 0.9879, + "step": 38200 + }, + { + "epoch": 0.92, + "learning_rate": 3.4475151637349934e-07, + "loss": 1.0328, + "step": 38201 + }, + { + "epoch": 0.92, + "learning_rate": 3.4454879095695226e-07, + "loss": 0.9801, + "step": 38202 + }, + { + "epoch": 0.92, + "learning_rate": 3.443461241177581e-07, + "loss": 1.26, + "step": 38203 + }, + { + "epoch": 0.92, + "learning_rate": 3.4414351585714377e-07, + "loss": 1.0555, + "step": 38204 + }, + { + "epoch": 0.92, + "learning_rate": 3.439409661763393e-07, + "loss": 1.0164, + "step": 38205 + }, + { + "epoch": 0.92, + "learning_rate": 3.4373847507657486e-07, + "loss": 1.19, + "step": 38206 + }, + { + "epoch": 0.92, + "learning_rate": 3.435360425590772e-07, + "loss": 1.204, + "step": 38207 + }, + { + "epoch": 0.92, + "learning_rate": 3.4333366862507544e-07, + "loss": 1.1725, + "step": 38208 + }, + { + "epoch": 0.92, + "learning_rate": 3.4313135327579627e-07, + "loss": 1.0718, + "step": 38209 + }, + { + "epoch": 0.92, + "learning_rate": 3.4292909651246766e-07, + "loss": 1.0934, + "step": 38210 + }, + { + "epoch": 0.92, + "learning_rate": 3.4272689833631747e-07, + "loss": 0.9035, + "step": 38211 + }, + { + "epoch": 0.92, + "learning_rate": 3.425247587485714e-07, + "loss": 1.0097, + "step": 38212 + }, + { + "epoch": 0.92, + "learning_rate": 3.4232267775045513e-07, + "loss": 1.1599, + "step": 38213 + }, + { + "epoch": 0.92, + "learning_rate": 3.421206553431966e-07, + "loss": 1.0135, + "step": 38214 + }, + { + "epoch": 0.92, + "learning_rate": 3.419186915280204e-07, + "loss": 1.0871, + "step": 38215 + }, + { + "epoch": 0.92, + "learning_rate": 3.417167863061521e-07, + "loss": 1.0276, + "step": 38216 + }, + { + "epoch": 0.92, + "learning_rate": 3.4151493967881534e-07, + "loss": 1.063, + "step": 38217 + }, + { + "epoch": 0.92, + "learning_rate": 3.413131516472357e-07, + "loss": 0.9441, + "step": 38218 + }, + { + "epoch": 0.92, + "learning_rate": 3.411114222126377e-07, + "loss": 0.9666, + "step": 38219 + }, + { + "epoch": 0.92, + "learning_rate": 3.409097513762438e-07, + "loss": 1.1502, + "step": 38220 + }, + { + "epoch": 0.92, + "learning_rate": 3.407081391392797e-07, + "loss": 1.2075, + "step": 38221 + }, + { + "epoch": 0.92, + "learning_rate": 3.4050658550296766e-07, + "loss": 0.9195, + "step": 38222 + }, + { + "epoch": 0.92, + "learning_rate": 3.40305090468529e-07, + "loss": 1.1926, + "step": 38223 + }, + { + "epoch": 0.92, + "learning_rate": 3.4010365403718824e-07, + "loss": 1.1104, + "step": 38224 + }, + { + "epoch": 0.92, + "learning_rate": 3.3990227621016557e-07, + "loss": 0.9606, + "step": 38225 + }, + { + "epoch": 0.92, + "learning_rate": 3.397009569886833e-07, + "loss": 1.0618, + "step": 38226 + }, + { + "epoch": 0.92, + "learning_rate": 3.3949969637396383e-07, + "loss": 1.0854, + "step": 38227 + }, + { + "epoch": 0.92, + "learning_rate": 3.392984943672273e-07, + "loss": 0.8705, + "step": 38228 + }, + { + "epoch": 0.92, + "learning_rate": 3.390973509696938e-07, + "loss": 1.0668, + "step": 38229 + }, + { + "epoch": 0.92, + "learning_rate": 3.3889626618258563e-07, + "loss": 1.0554, + "step": 38230 + }, + { + "epoch": 0.92, + "learning_rate": 3.386952400071208e-07, + "loss": 0.9745, + "step": 38231 + }, + { + "epoch": 0.92, + "learning_rate": 3.384942724445195e-07, + "loss": 1.1984, + "step": 38232 + }, + { + "epoch": 0.92, + "learning_rate": 3.382933634960006e-07, + "loss": 1.0513, + "step": 38233 + }, + { + "epoch": 0.92, + "learning_rate": 3.3809251316278326e-07, + "loss": 1.0456, + "step": 38234 + }, + { + "epoch": 0.92, + "learning_rate": 3.3789172144608753e-07, + "loss": 0.7559, + "step": 38235 + }, + { + "epoch": 0.92, + "learning_rate": 3.37690988347128e-07, + "loss": 1.0531, + "step": 38236 + }, + { + "epoch": 0.92, + "learning_rate": 3.3749031386712484e-07, + "loss": 1.0588, + "step": 38237 + }, + { + "epoch": 0.92, + "learning_rate": 3.3728969800729596e-07, + "loss": 1.24, + "step": 38238 + }, + { + "epoch": 0.92, + "learning_rate": 3.37089140768857e-07, + "loss": 1.1356, + "step": 38239 + }, + { + "epoch": 0.92, + "learning_rate": 3.368886421530271e-07, + "loss": 1.0055, + "step": 38240 + }, + { + "epoch": 0.92, + "learning_rate": 3.366882021610185e-07, + "loss": 0.7213, + "step": 38241 + }, + { + "epoch": 0.92, + "learning_rate": 3.364878207940503e-07, + "loss": 0.9912, + "step": 38242 + }, + { + "epoch": 0.92, + "learning_rate": 3.3628749805333817e-07, + "loss": 1.1409, + "step": 38243 + }, + { + "epoch": 0.92, + "learning_rate": 3.360872339400967e-07, + "loss": 0.9079, + "step": 38244 + }, + { + "epoch": 0.92, + "learning_rate": 3.3588702845554045e-07, + "loss": 1.1514, + "step": 38245 + }, + { + "epoch": 0.92, + "learning_rate": 3.3568688160088514e-07, + "loss": 1.1205, + "step": 38246 + }, + { + "epoch": 0.92, + "learning_rate": 3.3548679337734313e-07, + "loss": 0.8994, + "step": 38247 + }, + { + "epoch": 0.92, + "learning_rate": 3.352867637861301e-07, + "loss": 1.0225, + "step": 38248 + }, + { + "epoch": 0.92, + "learning_rate": 3.3508679282845955e-07, + "loss": 1.0474, + "step": 38249 + }, + { + "epoch": 0.92, + "learning_rate": 3.3488688050554274e-07, + "loss": 1.2325, + "step": 38250 + }, + { + "epoch": 0.92, + "learning_rate": 3.3468702681859646e-07, + "loss": 1.1081, + "step": 38251 + }, + { + "epoch": 0.92, + "learning_rate": 3.344872317688275e-07, + "loss": 1.0425, + "step": 38252 + }, + { + "epoch": 0.92, + "learning_rate": 3.3428749535745265e-07, + "loss": 0.9911, + "step": 38253 + }, + { + "epoch": 0.92, + "learning_rate": 3.340878175856821e-07, + "loss": 1.1167, + "step": 38254 + }, + { + "epoch": 0.92, + "learning_rate": 3.338881984547271e-07, + "loss": 1.0275, + "step": 38255 + }, + { + "epoch": 0.92, + "learning_rate": 3.3368863796579776e-07, + "loss": 0.9173, + "step": 38256 + }, + { + "epoch": 0.92, + "learning_rate": 3.3348913612010757e-07, + "loss": 1.1805, + "step": 38257 + }, + { + "epoch": 0.92, + "learning_rate": 3.332896929188634e-07, + "loss": 1.0947, + "step": 38258 + }, + { + "epoch": 0.92, + "learning_rate": 3.330903083632786e-07, + "loss": 0.77, + "step": 38259 + }, + { + "epoch": 0.92, + "learning_rate": 3.328909824545601e-07, + "loss": 1.3288, + "step": 38260 + }, + { + "epoch": 0.92, + "learning_rate": 3.3269171519391795e-07, + "loss": 1.0268, + "step": 38261 + }, + { + "epoch": 0.92, + "learning_rate": 3.324925065825624e-07, + "loss": 0.9055, + "step": 38262 + }, + { + "epoch": 0.92, + "learning_rate": 3.32293356621699e-07, + "loss": 1.0174, + "step": 38263 + }, + { + "epoch": 0.92, + "learning_rate": 3.320942653125381e-07, + "loss": 0.7996, + "step": 38264 + }, + { + "epoch": 0.92, + "learning_rate": 3.3189523265628967e-07, + "loss": 1.1293, + "step": 38265 + }, + { + "epoch": 0.92, + "learning_rate": 3.316962586541561e-07, + "loss": 1.137, + "step": 38266 + }, + { + "epoch": 0.92, + "learning_rate": 3.3149734330734873e-07, + "loss": 1.1935, + "step": 38267 + }, + { + "epoch": 0.92, + "learning_rate": 3.312984866170721e-07, + "loss": 1.1881, + "step": 38268 + }, + { + "epoch": 0.92, + "learning_rate": 3.31099688584533e-07, + "loss": 1.1637, + "step": 38269 + }, + { + "epoch": 0.92, + "learning_rate": 3.309009492109394e-07, + "loss": 0.9917, + "step": 38270 + }, + { + "epoch": 0.92, + "learning_rate": 3.3070226849749367e-07, + "loss": 0.9064, + "step": 38271 + }, + { + "epoch": 0.92, + "learning_rate": 3.305036464454037e-07, + "loss": 1.1545, + "step": 38272 + }, + { + "epoch": 0.92, + "learning_rate": 3.3030508305587405e-07, + "loss": 0.9763, + "step": 38273 + }, + { + "epoch": 0.92, + "learning_rate": 3.301065783301083e-07, + "loss": 1.0771, + "step": 38274 + }, + { + "epoch": 0.92, + "learning_rate": 3.299081322693132e-07, + "loss": 1.1954, + "step": 38275 + }, + { + "epoch": 0.92, + "learning_rate": 3.2970974487468887e-07, + "loss": 1.0329, + "step": 38276 + }, + { + "epoch": 0.92, + "learning_rate": 3.295114161474411e-07, + "loss": 1.2761, + "step": 38277 + }, + { + "epoch": 0.92, + "learning_rate": 3.293131460887733e-07, + "loss": 0.9647, + "step": 38278 + }, + { + "epoch": 0.92, + "learning_rate": 3.291149346998879e-07, + "loss": 1.2084, + "step": 38279 + }, + { + "epoch": 0.92, + "learning_rate": 3.289167819819872e-07, + "loss": 1.0496, + "step": 38280 + }, + { + "epoch": 0.92, + "learning_rate": 3.287186879362747e-07, + "loss": 1.0539, + "step": 38281 + }, + { + "epoch": 0.92, + "learning_rate": 3.285206525639506e-07, + "loss": 0.848, + "step": 38282 + }, + { + "epoch": 0.92, + "learning_rate": 3.283226758662172e-07, + "loss": 0.9255, + "step": 38283 + }, + { + "epoch": 0.92, + "learning_rate": 3.281247578442737e-07, + "loss": 1.1008, + "step": 38284 + }, + { + "epoch": 0.92, + "learning_rate": 3.2792689849932335e-07, + "loss": 0.8498, + "step": 38285 + }, + { + "epoch": 0.92, + "learning_rate": 3.2772909783256535e-07, + "loss": 1.0624, + "step": 38286 + }, + { + "epoch": 0.92, + "learning_rate": 3.2753135584519977e-07, + "loss": 0.8128, + "step": 38287 + }, + { + "epoch": 0.92, + "learning_rate": 3.2733367253842574e-07, + "loss": 1.0975, + "step": 38288 + }, + { + "epoch": 0.92, + "learning_rate": 3.2713604791344553e-07, + "loss": 1.1222, + "step": 38289 + }, + { + "epoch": 0.92, + "learning_rate": 3.269384819714538e-07, + "loss": 1.1272, + "step": 38290 + }, + { + "epoch": 0.92, + "learning_rate": 3.2674097471365073e-07, + "loss": 1.1745, + "step": 38291 + }, + { + "epoch": 0.92, + "learning_rate": 3.2654352614123643e-07, + "loss": 0.933, + "step": 38292 + }, + { + "epoch": 0.92, + "learning_rate": 3.2634613625540544e-07, + "loss": 0.9578, + "step": 38293 + }, + { + "epoch": 0.92, + "learning_rate": 3.261488050573591e-07, + "loss": 0.8441, + "step": 38294 + }, + { + "epoch": 0.92, + "learning_rate": 3.259515325482909e-07, + "loss": 0.9266, + "step": 38295 + }, + { + "epoch": 0.92, + "learning_rate": 3.2575431872939877e-07, + "loss": 1.0927, + "step": 38296 + }, + { + "epoch": 0.92, + "learning_rate": 3.255571636018817e-07, + "loss": 0.8585, + "step": 38297 + }, + { + "epoch": 0.92, + "learning_rate": 3.253600671669321e-07, + "loss": 0.7558, + "step": 38298 + }, + { + "epoch": 0.92, + "learning_rate": 3.2516302942574794e-07, + "loss": 1.1921, + "step": 38299 + }, + { + "epoch": 0.92, + "learning_rate": 3.249660503795238e-07, + "loss": 0.9961, + "step": 38300 + }, + { + "epoch": 0.92, + "learning_rate": 3.247691300294553e-07, + "loss": 0.8234, + "step": 38301 + }, + { + "epoch": 0.92, + "learning_rate": 3.2457226837673714e-07, + "loss": 1.2699, + "step": 38302 + }, + { + "epoch": 0.92, + "learning_rate": 3.2437546542256283e-07, + "loss": 1.0876, + "step": 38303 + }, + { + "epoch": 0.92, + "learning_rate": 3.2417872116812577e-07, + "loss": 1.1289, + "step": 38304 + }, + { + "epoch": 0.92, + "learning_rate": 3.239820356146217e-07, + "loss": 0.8182, + "step": 38305 + }, + { + "epoch": 0.92, + "learning_rate": 3.2378540876324194e-07, + "loss": 1.0495, + "step": 38306 + }, + { + "epoch": 0.92, + "learning_rate": 3.2358884061517994e-07, + "loss": 0.8613, + "step": 38307 + }, + { + "epoch": 0.92, + "learning_rate": 3.233923311716303e-07, + "loss": 0.9093, + "step": 38308 + }, + { + "epoch": 0.92, + "learning_rate": 3.2319588043378094e-07, + "loss": 1.244, + "step": 38309 + }, + { + "epoch": 0.92, + "learning_rate": 3.2299948840282755e-07, + "loss": 0.8958, + "step": 38310 + }, + { + "epoch": 0.92, + "learning_rate": 3.228031550799593e-07, + "loss": 1.0827, + "step": 38311 + }, + { + "epoch": 0.92, + "learning_rate": 3.2260688046636847e-07, + "loss": 0.9826, + "step": 38312 + }, + { + "epoch": 0.92, + "learning_rate": 3.224106645632463e-07, + "loss": 0.9132, + "step": 38313 + }, + { + "epoch": 0.92, + "learning_rate": 3.2221450737178083e-07, + "loss": 1.0548, + "step": 38314 + }, + { + "epoch": 0.92, + "learning_rate": 3.220184088931633e-07, + "loss": 1.0484, + "step": 38315 + }, + { + "epoch": 0.92, + "learning_rate": 3.21822369128586e-07, + "loss": 1.1372, + "step": 38316 + }, + { + "epoch": 0.92, + "learning_rate": 3.216263880792336e-07, + "loss": 0.9768, + "step": 38317 + }, + { + "epoch": 0.92, + "learning_rate": 3.2143046574629853e-07, + "loss": 1.2305, + "step": 38318 + }, + { + "epoch": 0.92, + "learning_rate": 3.212346021309676e-07, + "loss": 0.8669, + "step": 38319 + }, + { + "epoch": 0.92, + "learning_rate": 3.2103879723442976e-07, + "loss": 0.8504, + "step": 38320 + }, + { + "epoch": 0.92, + "learning_rate": 3.2084305105787417e-07, + "loss": 1.0175, + "step": 38321 + }, + { + "epoch": 0.92, + "learning_rate": 3.206473636024865e-07, + "loss": 1.154, + "step": 38322 + }, + { + "epoch": 0.92, + "learning_rate": 3.2045173486945357e-07, + "loss": 1.1052, + "step": 38323 + }, + { + "epoch": 0.92, + "learning_rate": 3.202561648599645e-07, + "loss": 1.1373, + "step": 38324 + }, + { + "epoch": 0.92, + "learning_rate": 3.200606535752038e-07, + "loss": 1.2121, + "step": 38325 + }, + { + "epoch": 0.92, + "learning_rate": 3.198652010163594e-07, + "loss": 1.0878, + "step": 38326 + }, + { + "epoch": 0.92, + "learning_rate": 3.1966980718461384e-07, + "loss": 1.1083, + "step": 38327 + }, + { + "epoch": 0.92, + "learning_rate": 3.19474472081156e-07, + "loss": 1.1273, + "step": 38328 + }, + { + "epoch": 0.92, + "learning_rate": 3.1927919570716947e-07, + "loss": 0.8748, + "step": 38329 + }, + { + "epoch": 0.92, + "learning_rate": 3.1908397806383775e-07, + "loss": 1.0721, + "step": 38330 + }, + { + "epoch": 0.92, + "learning_rate": 3.188888191523476e-07, + "loss": 0.959, + "step": 38331 + }, + { + "epoch": 0.92, + "learning_rate": 3.1869371897388147e-07, + "loss": 0.7663, + "step": 38332 + }, + { + "epoch": 0.92, + "learning_rate": 3.1849867752962283e-07, + "loss": 1.157, + "step": 38333 + }, + { + "epoch": 0.92, + "learning_rate": 3.183036948207574e-07, + "loss": 1.1361, + "step": 38334 + }, + { + "epoch": 0.92, + "learning_rate": 3.1810877084846424e-07, + "loss": 0.9371, + "step": 38335 + }, + { + "epoch": 0.92, + "learning_rate": 3.17913905613928e-07, + "loss": 1.1589, + "step": 38336 + }, + { + "epoch": 0.92, + "learning_rate": 3.177190991183321e-07, + "loss": 1.1536, + "step": 38337 + }, + { + "epoch": 0.92, + "learning_rate": 3.1752435136285563e-07, + "loss": 1.116, + "step": 38338 + }, + { + "epoch": 0.92, + "learning_rate": 3.173296623486832e-07, + "loss": 0.9214, + "step": 38339 + }, + { + "epoch": 0.92, + "learning_rate": 3.1713503207699283e-07, + "loss": 0.8864, + "step": 38340 + }, + { + "epoch": 0.92, + "learning_rate": 3.1694046054896677e-07, + "loss": 1.0837, + "step": 38341 + }, + { + "epoch": 0.92, + "learning_rate": 3.167459477657864e-07, + "loss": 0.9065, + "step": 38342 + }, + { + "epoch": 0.92, + "learning_rate": 3.1655149372862846e-07, + "loss": 0.9319, + "step": 38343 + }, + { + "epoch": 0.92, + "learning_rate": 3.163570984386766e-07, + "loss": 0.9567, + "step": 38344 + }, + { + "epoch": 0.92, + "learning_rate": 3.161627618971086e-07, + "loss": 0.8343, + "step": 38345 + }, + { + "epoch": 0.92, + "learning_rate": 3.159684841051025e-07, + "loss": 0.8391, + "step": 38346 + }, + { + "epoch": 0.92, + "learning_rate": 3.157742650638385e-07, + "loss": 1.064, + "step": 38347 + }, + { + "epoch": 0.92, + "learning_rate": 3.155801047744944e-07, + "loss": 1.1166, + "step": 38348 + }, + { + "epoch": 0.92, + "learning_rate": 3.153860032382461e-07, + "loss": 1.0475, + "step": 38349 + }, + { + "epoch": 0.92, + "learning_rate": 3.1519196045627586e-07, + "loss": 1.1945, + "step": 38350 + }, + { + "epoch": 0.92, + "learning_rate": 3.1499797642975505e-07, + "loss": 1.1493, + "step": 38351 + }, + { + "epoch": 0.92, + "learning_rate": 3.148040511598649e-07, + "loss": 1.0362, + "step": 38352 + }, + { + "epoch": 0.92, + "learning_rate": 3.1461018464778117e-07, + "loss": 0.9857, + "step": 38353 + }, + { + "epoch": 0.92, + "learning_rate": 3.144163768946784e-07, + "loss": 1.2171, + "step": 38354 + }, + { + "epoch": 0.92, + "learning_rate": 3.142226279017335e-07, + "loss": 1.2212, + "step": 38355 + }, + { + "epoch": 0.92, + "learning_rate": 3.140289376701211e-07, + "loss": 1.1126, + "step": 38356 + }, + { + "epoch": 0.92, + "learning_rate": 3.138353062010169e-07, + "loss": 0.8607, + "step": 38357 + }, + { + "epoch": 0.92, + "learning_rate": 3.136417334955966e-07, + "loss": 0.9108, + "step": 38358 + }, + { + "epoch": 0.92, + "learning_rate": 3.1344821955503257e-07, + "loss": 0.971, + "step": 38359 + }, + { + "epoch": 0.92, + "learning_rate": 3.1325476438049953e-07, + "loss": 1.0792, + "step": 38360 + }, + { + "epoch": 0.92, + "learning_rate": 3.130613679731731e-07, + "loss": 0.8857, + "step": 38361 + }, + { + "epoch": 0.92, + "learning_rate": 3.1286803033422355e-07, + "loss": 1.2065, + "step": 38362 + }, + { + "epoch": 0.92, + "learning_rate": 3.126747514648254e-07, + "loss": 1.1129, + "step": 38363 + }, + { + "epoch": 0.92, + "learning_rate": 3.124815313661511e-07, + "loss": 0.8519, + "step": 38364 + }, + { + "epoch": 0.92, + "learning_rate": 3.1228837003937196e-07, + "loss": 1.2729, + "step": 38365 + }, + { + "epoch": 0.92, + "learning_rate": 3.1209526748566146e-07, + "loss": 1.2052, + "step": 38366 + }, + { + "epoch": 0.92, + "learning_rate": 3.1190222370618974e-07, + "loss": 1.2717, + "step": 38367 + }, + { + "epoch": 0.92, + "learning_rate": 3.1170923870212813e-07, + "loss": 1.2688, + "step": 38368 + }, + { + "epoch": 0.92, + "learning_rate": 3.1151631247464787e-07, + "loss": 0.9461, + "step": 38369 + }, + { + "epoch": 0.92, + "learning_rate": 3.113234450249192e-07, + "loss": 1.013, + "step": 38370 + }, + { + "epoch": 0.92, + "learning_rate": 3.111306363541122e-07, + "loss": 1.214, + "step": 38371 + }, + { + "epoch": 0.92, + "learning_rate": 3.109378864633983e-07, + "loss": 0.9921, + "step": 38372 + }, + { + "epoch": 0.92, + "learning_rate": 3.107451953539431e-07, + "loss": 1.0313, + "step": 38373 + }, + { + "epoch": 0.92, + "learning_rate": 3.1055256302692015e-07, + "loss": 1.0506, + "step": 38374 + }, + { + "epoch": 0.92, + "learning_rate": 3.10359989483493e-07, + "loss": 1.1187, + "step": 38375 + }, + { + "epoch": 0.92, + "learning_rate": 3.10167474724834e-07, + "loss": 1.0317, + "step": 38376 + }, + { + "epoch": 0.92, + "learning_rate": 3.0997501875211e-07, + "loss": 1.1186, + "step": 38377 + }, + { + "epoch": 0.92, + "learning_rate": 3.0978262156648784e-07, + "loss": 1.0956, + "step": 38378 + }, + { + "epoch": 0.92, + "learning_rate": 3.095902831691344e-07, + "loss": 0.9276, + "step": 38379 + }, + { + "epoch": 0.92, + "learning_rate": 3.0939800356121985e-07, + "loss": 1.0064, + "step": 38380 + }, + { + "epoch": 0.92, + "learning_rate": 3.0920578274390547e-07, + "loss": 1.0815, + "step": 38381 + }, + { + "epoch": 0.92, + "learning_rate": 3.090136207183625e-07, + "loss": 1.0727, + "step": 38382 + }, + { + "epoch": 0.92, + "learning_rate": 3.088215174857534e-07, + "loss": 1.1197, + "step": 38383 + }, + { + "epoch": 0.92, + "learning_rate": 3.0862947304724387e-07, + "loss": 0.7653, + "step": 38384 + }, + { + "epoch": 0.92, + "learning_rate": 3.084374874040008e-07, + "loss": 0.7582, + "step": 38385 + }, + { + "epoch": 0.92, + "learning_rate": 3.082455605571877e-07, + "loss": 1.1267, + "step": 38386 + }, + { + "epoch": 0.92, + "learning_rate": 3.080536925079691e-07, + "loss": 0.7944, + "step": 38387 + }, + { + "epoch": 0.92, + "learning_rate": 3.078618832575098e-07, + "loss": 0.8739, + "step": 38388 + }, + { + "epoch": 0.92, + "learning_rate": 3.076701328069709e-07, + "loss": 1.1075, + "step": 38389 + }, + { + "epoch": 0.92, + "learning_rate": 3.0747844115751934e-07, + "loss": 1.0322, + "step": 38390 + }, + { + "epoch": 0.92, + "learning_rate": 3.0728680831031535e-07, + "loss": 1.0509, + "step": 38391 + }, + { + "epoch": 0.92, + "learning_rate": 3.070952342665223e-07, + "loss": 1.0278, + "step": 38392 + }, + { + "epoch": 0.92, + "learning_rate": 3.0690371902730386e-07, + "loss": 0.9337, + "step": 38393 + }, + { + "epoch": 0.92, + "learning_rate": 3.0671226259381904e-07, + "loss": 1.2355, + "step": 38394 + }, + { + "epoch": 0.92, + "learning_rate": 3.0652086496723024e-07, + "loss": 0.8847, + "step": 38395 + }, + { + "epoch": 0.92, + "learning_rate": 3.06329526148702e-07, + "loss": 1.0405, + "step": 38396 + }, + { + "epoch": 0.92, + "learning_rate": 3.0613824613939023e-07, + "loss": 1.0808, + "step": 38397 + }, + { + "epoch": 0.92, + "learning_rate": 3.059470249404595e-07, + "loss": 0.9464, + "step": 38398 + }, + { + "epoch": 0.92, + "learning_rate": 3.0575586255306543e-07, + "loss": 1.1465, + "step": 38399 + }, + { + "epoch": 0.92, + "learning_rate": 3.055647589783717e-07, + "loss": 1.1657, + "step": 38400 + }, + { + "epoch": 0.92, + "learning_rate": 3.0537371421753724e-07, + "loss": 1.2434, + "step": 38401 + }, + { + "epoch": 0.92, + "learning_rate": 3.0518272827171904e-07, + "loss": 0.7668, + "step": 38402 + }, + { + "epoch": 0.92, + "learning_rate": 3.0499180114207607e-07, + "loss": 1.1024, + "step": 38403 + }, + { + "epoch": 0.92, + "learning_rate": 3.0480093282976964e-07, + "loss": 1.1367, + "step": 38404 + }, + { + "epoch": 0.92, + "learning_rate": 3.046101233359533e-07, + "loss": 0.8028, + "step": 38405 + }, + { + "epoch": 0.92, + "learning_rate": 3.0441937266178834e-07, + "loss": 0.9203, + "step": 38406 + }, + { + "epoch": 0.92, + "learning_rate": 3.0422868080842936e-07, + "loss": 0.8211, + "step": 38407 + }, + { + "epoch": 0.92, + "learning_rate": 3.040380477770355e-07, + "loss": 1.2148, + "step": 38408 + }, + { + "epoch": 0.92, + "learning_rate": 3.038474735687624e-07, + "loss": 1.0059, + "step": 38409 + }, + { + "epoch": 0.92, + "learning_rate": 3.036569581847648e-07, + "loss": 1.0074, + "step": 38410 + }, + { + "epoch": 0.92, + "learning_rate": 3.034665016261995e-07, + "loss": 1.0865, + "step": 38411 + }, + { + "epoch": 0.92, + "learning_rate": 3.032761038942234e-07, + "loss": 1.1348, + "step": 38412 + }, + { + "epoch": 0.92, + "learning_rate": 3.0308576498998986e-07, + "loss": 0.9687, + "step": 38413 + }, + { + "epoch": 0.92, + "learning_rate": 3.0289548491465373e-07, + "loss": 1.1654, + "step": 38414 + }, + { + "epoch": 0.92, + "learning_rate": 3.027052636693706e-07, + "loss": 1.2012, + "step": 38415 + }, + { + "epoch": 0.92, + "learning_rate": 3.025151012552929e-07, + "loss": 0.8212, + "step": 38416 + }, + { + "epoch": 0.92, + "learning_rate": 3.0232499767357647e-07, + "loss": 1.3297, + "step": 38417 + }, + { + "epoch": 0.92, + "learning_rate": 3.021349529253714e-07, + "loss": 1.2621, + "step": 38418 + }, + { + "epoch": 0.92, + "learning_rate": 3.019449670118324e-07, + "loss": 1.0662, + "step": 38419 + }, + { + "epoch": 0.92, + "learning_rate": 3.0175503993411405e-07, + "loss": 1.1432, + "step": 38420 + }, + { + "epoch": 0.92, + "learning_rate": 3.015651716933654e-07, + "loss": 0.9278, + "step": 38421 + }, + { + "epoch": 0.92, + "learning_rate": 3.0137536229074006e-07, + "loss": 1.1198, + "step": 38422 + }, + { + "epoch": 0.92, + "learning_rate": 3.0118561172738925e-07, + "loss": 1.0076, + "step": 38423 + }, + { + "epoch": 0.92, + "learning_rate": 3.009959200044632e-07, + "loss": 1.1474, + "step": 38424 + }, + { + "epoch": 0.92, + "learning_rate": 3.0080628712311543e-07, + "loss": 1.2182, + "step": 38425 + }, + { + "epoch": 0.92, + "learning_rate": 3.006167130844928e-07, + "loss": 1.0306, + "step": 38426 + }, + { + "epoch": 0.92, + "learning_rate": 3.004271978897477e-07, + "loss": 1.1598, + "step": 38427 + }, + { + "epoch": 0.92, + "learning_rate": 3.0023774154002925e-07, + "loss": 0.9758, + "step": 38428 + }, + { + "epoch": 0.92, + "learning_rate": 3.0004834403648654e-07, + "loss": 1.108, + "step": 38429 + }, + { + "epoch": 0.92, + "learning_rate": 2.998590053802697e-07, + "loss": 1.1104, + "step": 38430 + }, + { + "epoch": 0.92, + "learning_rate": 2.996697255725267e-07, + "loss": 0.9746, + "step": 38431 + }, + { + "epoch": 0.92, + "learning_rate": 2.9948050461440446e-07, + "loss": 1.1198, + "step": 38432 + }, + { + "epoch": 0.92, + "learning_rate": 2.992913425070543e-07, + "loss": 1.0415, + "step": 38433 + }, + { + "epoch": 0.92, + "learning_rate": 2.9910223925162077e-07, + "loss": 1.1683, + "step": 38434 + }, + { + "epoch": 0.92, + "learning_rate": 2.9891319484925076e-07, + "loss": 1.0167, + "step": 38435 + }, + { + "epoch": 0.92, + "learning_rate": 2.987242093010956e-07, + "loss": 0.8319, + "step": 38436 + }, + { + "epoch": 0.92, + "learning_rate": 2.9853528260829656e-07, + "loss": 1.0322, + "step": 38437 + }, + { + "epoch": 0.92, + "learning_rate": 2.9834641477200164e-07, + "loss": 1.1046, + "step": 38438 + }, + { + "epoch": 0.92, + "learning_rate": 2.9815760579335886e-07, + "loss": 0.9938, + "step": 38439 + }, + { + "epoch": 0.92, + "learning_rate": 2.979688556735105e-07, + "loss": 1.1018, + "step": 38440 + }, + { + "epoch": 0.92, + "learning_rate": 2.977801644136047e-07, + "loss": 1.0845, + "step": 38441 + }, + { + "epoch": 0.92, + "learning_rate": 2.9759153201478264e-07, + "loss": 1.0043, + "step": 38442 + }, + { + "epoch": 0.92, + "learning_rate": 2.974029584781912e-07, + "loss": 1.0726, + "step": 38443 + }, + { + "epoch": 0.92, + "learning_rate": 2.972144438049751e-07, + "loss": 1.1169, + "step": 38444 + }, + { + "epoch": 0.92, + "learning_rate": 2.9702598799627444e-07, + "loss": 1.0484, + "step": 38445 + }, + { + "epoch": 0.92, + "learning_rate": 2.968375910532362e-07, + "loss": 1.0557, + "step": 38446 + }, + { + "epoch": 0.92, + "learning_rate": 2.966492529770026e-07, + "loss": 1.0522, + "step": 38447 + }, + { + "epoch": 0.92, + "learning_rate": 2.964609737687141e-07, + "loss": 1.0716, + "step": 38448 + }, + { + "epoch": 0.92, + "learning_rate": 2.9627275342951625e-07, + "loss": 1.15, + "step": 38449 + }, + { + "epoch": 0.92, + "learning_rate": 2.960845919605471e-07, + "loss": 0.9712, + "step": 38450 + }, + { + "epoch": 0.92, + "learning_rate": 2.9589648936295125e-07, + "loss": 1.2241, + "step": 38451 + }, + { + "epoch": 0.92, + "learning_rate": 2.957084456378689e-07, + "loss": 1.231, + "step": 38452 + }, + { + "epoch": 0.92, + "learning_rate": 2.9552046078644036e-07, + "loss": 1.092, + "step": 38453 + }, + { + "epoch": 0.92, + "learning_rate": 2.953325348098068e-07, + "loss": 0.9726, + "step": 38454 + }, + { + "epoch": 0.92, + "learning_rate": 2.9514466770910853e-07, + "loss": 1.084, + "step": 38455 + }, + { + "epoch": 0.92, + "learning_rate": 2.949568594854846e-07, + "loss": 1.0309, + "step": 38456 + }, + { + "epoch": 0.92, + "learning_rate": 2.9476911014007515e-07, + "loss": 1.0836, + "step": 38457 + }, + { + "epoch": 0.92, + "learning_rate": 2.945814196740182e-07, + "loss": 1.0015, + "step": 38458 + }, + { + "epoch": 0.92, + "learning_rate": 2.9439378808845174e-07, + "loss": 1.0983, + "step": 38459 + }, + { + "epoch": 0.92, + "learning_rate": 2.942062153845171e-07, + "loss": 1.147, + "step": 38460 + }, + { + "epoch": 0.92, + "learning_rate": 2.940187015633489e-07, + "loss": 1.1497, + "step": 38461 + }, + { + "epoch": 0.92, + "learning_rate": 2.938312466260873e-07, + "loss": 1.1041, + "step": 38462 + }, + { + "epoch": 0.92, + "learning_rate": 2.9364385057386923e-07, + "loss": 1.061, + "step": 38463 + }, + { + "epoch": 0.93, + "learning_rate": 2.934565134078293e-07, + "loss": 0.9816, + "step": 38464 + }, + { + "epoch": 0.93, + "learning_rate": 2.9326923512910557e-07, + "loss": 1.0207, + "step": 38465 + }, + { + "epoch": 0.93, + "learning_rate": 2.930820157388348e-07, + "loss": 0.961, + "step": 38466 + }, + { + "epoch": 0.93, + "learning_rate": 2.9289485523815163e-07, + "loss": 0.9452, + "step": 38467 + }, + { + "epoch": 0.93, + "learning_rate": 2.9270775362819413e-07, + "loss": 1.1491, + "step": 38468 + }, + { + "epoch": 0.93, + "learning_rate": 2.925207109100936e-07, + "loss": 1.1805, + "step": 38469 + }, + { + "epoch": 0.93, + "learning_rate": 2.9233372708498577e-07, + "loss": 1.2954, + "step": 38470 + }, + { + "epoch": 0.93, + "learning_rate": 2.921468021540086e-07, + "loss": 0.8707, + "step": 38471 + }, + { + "epoch": 0.93, + "learning_rate": 2.9195993611829123e-07, + "loss": 1.0869, + "step": 38472 + }, + { + "epoch": 0.93, + "learning_rate": 2.917731289789694e-07, + "loss": 0.8794, + "step": 38473 + }, + { + "epoch": 0.93, + "learning_rate": 2.9158638073717773e-07, + "loss": 1.1365, + "step": 38474 + }, + { + "epoch": 0.93, + "learning_rate": 2.9139969139404647e-07, + "loss": 0.8253, + "step": 38475 + }, + { + "epoch": 0.93, + "learning_rate": 2.912130609507113e-07, + "loss": 1.0067, + "step": 38476 + }, + { + "epoch": 0.93, + "learning_rate": 2.910264894083015e-07, + "loss": 0.8876, + "step": 38477 + }, + { + "epoch": 0.93, + "learning_rate": 2.908399767679504e-07, + "loss": 0.7737, + "step": 38478 + }, + { + "epoch": 0.93, + "learning_rate": 2.9065352303079054e-07, + "loss": 1.008, + "step": 38479 + }, + { + "epoch": 0.93, + "learning_rate": 2.9046712819795096e-07, + "loss": 1.0943, + "step": 38480 + }, + { + "epoch": 0.93, + "learning_rate": 2.902807922705642e-07, + "loss": 1.1441, + "step": 38481 + }, + { + "epoch": 0.93, + "learning_rate": 2.900945152497592e-07, + "loss": 1.1998, + "step": 38482 + }, + { + "epoch": 0.93, + "learning_rate": 2.8990829713666733e-07, + "loss": 0.9416, + "step": 38483 + }, + { + "epoch": 0.93, + "learning_rate": 2.897221379324189e-07, + "loss": 0.8872, + "step": 38484 + }, + { + "epoch": 0.93, + "learning_rate": 2.8953603763814065e-07, + "loss": 0.985, + "step": 38485 + }, + { + "epoch": 0.93, + "learning_rate": 2.8934999625496287e-07, + "loss": 1.0253, + "step": 38486 + }, + { + "epoch": 0.93, + "learning_rate": 2.891640137840168e-07, + "loss": 0.9664, + "step": 38487 + }, + { + "epoch": 0.93, + "learning_rate": 2.889780902264261e-07, + "loss": 0.9321, + "step": 38488 + }, + { + "epoch": 0.93, + "learning_rate": 2.88792225583322e-07, + "loss": 1.0501, + "step": 38489 + }, + { + "epoch": 0.93, + "learning_rate": 2.886064198558325e-07, + "loss": 1.0615, + "step": 38490 + }, + { + "epoch": 0.93, + "learning_rate": 2.8842067304508226e-07, + "loss": 0.9248, + "step": 38491 + }, + { + "epoch": 0.93, + "learning_rate": 2.882349851522004e-07, + "loss": 0.9084, + "step": 38492 + }, + { + "epoch": 0.93, + "learning_rate": 2.8804935617831153e-07, + "loss": 1.1696, + "step": 38493 + }, + { + "epoch": 0.93, + "learning_rate": 2.878637861245426e-07, + "loss": 1.0237, + "step": 38494 + }, + { + "epoch": 0.93, + "learning_rate": 2.876782749920215e-07, + "loss": 1.168, + "step": 38495 + }, + { + "epoch": 0.93, + "learning_rate": 2.8749282278186965e-07, + "loss": 0.8125, + "step": 38496 + }, + { + "epoch": 0.93, + "learning_rate": 2.8730742949521494e-07, + "loss": 1.1275, + "step": 38497 + }, + { + "epoch": 0.93, + "learning_rate": 2.8712209513318325e-07, + "loss": 1.0305, + "step": 38498 + }, + { + "epoch": 0.93, + "learning_rate": 2.8693681969689355e-07, + "loss": 1.1486, + "step": 38499 + }, + { + "epoch": 0.93, + "learning_rate": 2.867516031874773e-07, + "loss": 1.1003, + "step": 38500 + }, + { + "epoch": 0.93, + "learning_rate": 2.865664456060524e-07, + "loss": 0.9444, + "step": 38501 + }, + { + "epoch": 0.93, + "learning_rate": 2.8638134695374355e-07, + "loss": 0.9489, + "step": 38502 + }, + { + "epoch": 0.93, + "learning_rate": 2.861963072316765e-07, + "loss": 1.0644, + "step": 38503 + }, + { + "epoch": 0.93, + "learning_rate": 2.8601132644096916e-07, + "loss": 1.0623, + "step": 38504 + }, + { + "epoch": 0.93, + "learning_rate": 2.8582640458274855e-07, + "loss": 1.3346, + "step": 38505 + }, + { + "epoch": 0.93, + "learning_rate": 2.856415416581326e-07, + "loss": 0.8272, + "step": 38506 + }, + { + "epoch": 0.93, + "learning_rate": 2.8545673766824487e-07, + "loss": 0.9912, + "step": 38507 + }, + { + "epoch": 0.93, + "learning_rate": 2.8527199261420666e-07, + "loss": 0.782, + "step": 38508 + }, + { + "epoch": 0.93, + "learning_rate": 2.8508730649713824e-07, + "loss": 1.1523, + "step": 38509 + }, + { + "epoch": 0.93, + "learning_rate": 2.8490267931815864e-07, + "loss": 0.8595, + "step": 38510 + }, + { + "epoch": 0.93, + "learning_rate": 2.847181110783914e-07, + "loss": 1.1175, + "step": 38511 + }, + { + "epoch": 0.93, + "learning_rate": 2.845336017789524e-07, + "loss": 0.8724, + "step": 38512 + }, + { + "epoch": 0.93, + "learning_rate": 2.8434915142096507e-07, + "loss": 1.0777, + "step": 38513 + }, + { + "epoch": 0.93, + "learning_rate": 2.841647600055442e-07, + "loss": 1.0326, + "step": 38514 + }, + { + "epoch": 0.93, + "learning_rate": 2.8398042753380984e-07, + "loss": 0.6902, + "step": 38515 + }, + { + "epoch": 0.93, + "learning_rate": 2.837961540068834e-07, + "loss": 0.9887, + "step": 38516 + }, + { + "epoch": 0.93, + "learning_rate": 2.836119394258785e-07, + "loss": 0.8592, + "step": 38517 + }, + { + "epoch": 0.93, + "learning_rate": 2.834277837919142e-07, + "loss": 1.21, + "step": 38518 + }, + { + "epoch": 0.93, + "learning_rate": 2.832436871061095e-07, + "loss": 0.8788, + "step": 38519 + }, + { + "epoch": 0.93, + "learning_rate": 2.830596493695792e-07, + "loss": 1.1119, + "step": 38520 + }, + { + "epoch": 0.93, + "learning_rate": 2.828756705834401e-07, + "loss": 1.3067, + "step": 38521 + }, + { + "epoch": 0.93, + "learning_rate": 2.8269175074880915e-07, + "loss": 1.0194, + "step": 38522 + }, + { + "epoch": 0.93, + "learning_rate": 2.82507889866801e-07, + "loss": 1.0117, + "step": 38523 + }, + { + "epoch": 0.93, + "learning_rate": 2.823240879385325e-07, + "loss": 0.9873, + "step": 38524 + }, + { + "epoch": 0.93, + "learning_rate": 2.8214034496511833e-07, + "loss": 0.8158, + "step": 38525 + }, + { + "epoch": 0.93, + "learning_rate": 2.8195666094767203e-07, + "loss": 0.9666, + "step": 38526 + }, + { + "epoch": 0.93, + "learning_rate": 2.8177303588730943e-07, + "loss": 1.1215, + "step": 38527 + }, + { + "epoch": 0.93, + "learning_rate": 2.815894697851429e-07, + "loss": 1.143, + "step": 38528 + }, + { + "epoch": 0.93, + "learning_rate": 2.8140596264228935e-07, + "loss": 0.9566, + "step": 38529 + }, + { + "epoch": 0.93, + "learning_rate": 2.8122251445985793e-07, + "loss": 0.9878, + "step": 38530 + }, + { + "epoch": 0.93, + "learning_rate": 2.810391252389633e-07, + "loss": 1.0501, + "step": 38531 + }, + { + "epoch": 0.93, + "learning_rate": 2.8085579498072003e-07, + "loss": 1.0016, + "step": 38532 + }, + { + "epoch": 0.93, + "learning_rate": 2.806725236862373e-07, + "loss": 0.9507, + "step": 38533 + }, + { + "epoch": 0.93, + "learning_rate": 2.804893113566276e-07, + "loss": 1.2027, + "step": 38534 + }, + { + "epoch": 0.93, + "learning_rate": 2.8030615799300556e-07, + "loss": 1.0293, + "step": 38535 + }, + { + "epoch": 0.93, + "learning_rate": 2.801230635964769e-07, + "loss": 1.2024, + "step": 38536 + }, + { + "epoch": 0.93, + "learning_rate": 2.799400281681575e-07, + "loss": 1.1795, + "step": 38537 + }, + { + "epoch": 0.93, + "learning_rate": 2.7975705170915526e-07, + "loss": 0.8467, + "step": 38538 + }, + { + "epoch": 0.93, + "learning_rate": 2.795741342205793e-07, + "loss": 0.8408, + "step": 38539 + }, + { + "epoch": 0.93, + "learning_rate": 2.7939127570354216e-07, + "loss": 1.0048, + "step": 38540 + }, + { + "epoch": 0.93, + "learning_rate": 2.792084761591507e-07, + "loss": 0.7963, + "step": 38541 + }, + { + "epoch": 0.93, + "learning_rate": 2.790257355885162e-07, + "loss": 1.1028, + "step": 38542 + }, + { + "epoch": 0.93, + "learning_rate": 2.788430539927456e-07, + "loss": 1.1867, + "step": 38543 + }, + { + "epoch": 0.93, + "learning_rate": 2.7866043137294687e-07, + "loss": 0.9489, + "step": 38544 + }, + { + "epoch": 0.93, + "learning_rate": 2.784778677302291e-07, + "loss": 1.1942, + "step": 38545 + }, + { + "epoch": 0.93, + "learning_rate": 2.782953630657004e-07, + "loss": 1.0001, + "step": 38546 + }, + { + "epoch": 0.93, + "learning_rate": 2.781129173804653e-07, + "loss": 1.1388, + "step": 38547 + }, + { + "epoch": 0.93, + "learning_rate": 2.7793053067563415e-07, + "loss": 0.9138, + "step": 38548 + }, + { + "epoch": 0.93, + "learning_rate": 2.777482029523104e-07, + "loss": 0.9958, + "step": 38549 + }, + { + "epoch": 0.93, + "learning_rate": 2.7756593421160104e-07, + "loss": 1.163, + "step": 38550 + }, + { + "epoch": 0.93, + "learning_rate": 2.7738372445461404e-07, + "loss": 1.197, + "step": 38551 + }, + { + "epoch": 0.93, + "learning_rate": 2.772015736824518e-07, + "loss": 1.0926, + "step": 38552 + }, + { + "epoch": 0.93, + "learning_rate": 2.770194818962213e-07, + "loss": 1.128, + "step": 38553 + }, + { + "epoch": 0.93, + "learning_rate": 2.7683744909702714e-07, + "loss": 1.1367, + "step": 38554 + }, + { + "epoch": 0.93, + "learning_rate": 2.766554752859718e-07, + "loss": 1.038, + "step": 38555 + }, + { + "epoch": 0.93, + "learning_rate": 2.764735604641622e-07, + "loss": 0.8637, + "step": 38556 + }, + { + "epoch": 0.93, + "learning_rate": 2.762917046326985e-07, + "loss": 1.1348, + "step": 38557 + }, + { + "epoch": 0.93, + "learning_rate": 2.7610990779268655e-07, + "loss": 1.018, + "step": 38558 + }, + { + "epoch": 0.93, + "learning_rate": 2.759281699452299e-07, + "loss": 1.1673, + "step": 38559 + }, + { + "epoch": 0.93, + "learning_rate": 2.757464910914287e-07, + "loss": 1.0257, + "step": 38560 + }, + { + "epoch": 0.93, + "learning_rate": 2.755648712323855e-07, + "loss": 1.0743, + "step": 38561 + }, + { + "epoch": 0.93, + "learning_rate": 2.7538331036920494e-07, + "loss": 1.0465, + "step": 38562 + }, + { + "epoch": 0.93, + "learning_rate": 2.7520180850298396e-07, + "loss": 0.9897, + "step": 38563 + }, + { + "epoch": 0.93, + "learning_rate": 2.7502036563482826e-07, + "loss": 1.0414, + "step": 38564 + }, + { + "epoch": 0.93, + "learning_rate": 2.748389817658359e-07, + "loss": 0.8356, + "step": 38565 + }, + { + "epoch": 0.93, + "learning_rate": 2.746576568971071e-07, + "loss": 1.1611, + "step": 38566 + }, + { + "epoch": 0.93, + "learning_rate": 2.7447639102974435e-07, + "loss": 0.8595, + "step": 38567 + }, + { + "epoch": 0.93, + "learning_rate": 2.742951841648445e-07, + "loss": 1.1617, + "step": 38568 + }, + { + "epoch": 0.93, + "learning_rate": 2.741140363035077e-07, + "loss": 1.0712, + "step": 38569 + }, + { + "epoch": 0.93, + "learning_rate": 2.739329474468355e-07, + "loss": 0.7671, + "step": 38570 + }, + { + "epoch": 0.93, + "learning_rate": 2.7375191759592245e-07, + "loss": 1.1366, + "step": 38571 + }, + { + "epoch": 0.93, + "learning_rate": 2.735709467518699e-07, + "loss": 0.9408, + "step": 38572 + }, + { + "epoch": 0.93, + "learning_rate": 2.7339003491577365e-07, + "loss": 1.03, + "step": 38573 + }, + { + "epoch": 0.93, + "learning_rate": 2.7320918208873284e-07, + "loss": 1.0694, + "step": 38574 + }, + { + "epoch": 0.93, + "learning_rate": 2.730283882718443e-07, + "loss": 0.9988, + "step": 38575 + }, + { + "epoch": 0.93, + "learning_rate": 2.728476534662039e-07, + "loss": 0.7872, + "step": 38576 + }, + { + "epoch": 0.93, + "learning_rate": 2.726669776729085e-07, + "loss": 1.1024, + "step": 38577 + }, + { + "epoch": 0.93, + "learning_rate": 2.724863608930561e-07, + "loss": 1.0266, + "step": 38578 + }, + { + "epoch": 0.93, + "learning_rate": 2.7230580312773924e-07, + "loss": 1.0197, + "step": 38579 + }, + { + "epoch": 0.93, + "learning_rate": 2.7212530437805696e-07, + "loss": 0.8732, + "step": 38580 + }, + { + "epoch": 0.93, + "learning_rate": 2.719448646451006e-07, + "loss": 0.9177, + "step": 38581 + }, + { + "epoch": 0.93, + "learning_rate": 2.71764483929966e-07, + "loss": 1.1629, + "step": 38582 + }, + { + "epoch": 0.93, + "learning_rate": 2.715841622337501e-07, + "loss": 1.0768, + "step": 38583 + }, + { + "epoch": 0.93, + "learning_rate": 2.714038995575441e-07, + "loss": 1.1514, + "step": 38584 + }, + { + "epoch": 0.93, + "learning_rate": 2.712236959024417e-07, + "loss": 1.151, + "step": 38585 + }, + { + "epoch": 0.93, + "learning_rate": 2.710435512695375e-07, + "loss": 1.1635, + "step": 38586 + }, + { + "epoch": 0.93, + "learning_rate": 2.7086346565992295e-07, + "loss": 0.7075, + "step": 38587 + }, + { + "epoch": 0.93, + "learning_rate": 2.7068343907469154e-07, + "loss": 1.0201, + "step": 38588 + }, + { + "epoch": 0.93, + "learning_rate": 2.705034715149357e-07, + "loss": 1.0983, + "step": 38589 + }, + { + "epoch": 0.93, + "learning_rate": 2.7032356298174576e-07, + "loss": 1.0752, + "step": 38590 + }, + { + "epoch": 0.93, + "learning_rate": 2.7014371347621524e-07, + "loss": 1.0119, + "step": 38591 + }, + { + "epoch": 0.93, + "learning_rate": 2.699639229994344e-07, + "loss": 1.0448, + "step": 38592 + }, + { + "epoch": 0.93, + "learning_rate": 2.6978419155249236e-07, + "loss": 0.9002, + "step": 38593 + }, + { + "epoch": 0.93, + "learning_rate": 2.6960451913648264e-07, + "loss": 0.9926, + "step": 38594 + }, + { + "epoch": 0.93, + "learning_rate": 2.6942490575249226e-07, + "loss": 0.7518, + "step": 38595 + }, + { + "epoch": 0.93, + "learning_rate": 2.6924535140161244e-07, + "loss": 1.1276, + "step": 38596 + }, + { + "epoch": 0.93, + "learning_rate": 2.6906585608493354e-07, + "loss": 0.9798, + "step": 38597 + }, + { + "epoch": 0.93, + "learning_rate": 2.6888641980354125e-07, + "loss": 0.7228, + "step": 38598 + }, + { + "epoch": 0.93, + "learning_rate": 2.6870704255852806e-07, + "loss": 0.9168, + "step": 38599 + }, + { + "epoch": 0.93, + "learning_rate": 2.6852772435097873e-07, + "loss": 1.0644, + "step": 38600 + }, + { + "epoch": 0.93, + "learning_rate": 2.683484651819834e-07, + "loss": 1.1178, + "step": 38601 + }, + { + "epoch": 0.93, + "learning_rate": 2.681692650526291e-07, + "loss": 1.1109, + "step": 38602 + }, + { + "epoch": 0.93, + "learning_rate": 2.679901239640026e-07, + "loss": 0.9692, + "step": 38603 + }, + { + "epoch": 0.93, + "learning_rate": 2.6781104191719086e-07, + "loss": 1.0421, + "step": 38604 + }, + { + "epoch": 0.93, + "learning_rate": 2.6763201891328086e-07, + "loss": 1.0879, + "step": 38605 + }, + { + "epoch": 0.93, + "learning_rate": 2.6745305495335717e-07, + "loss": 1.1278, + "step": 38606 + }, + { + "epoch": 0.93, + "learning_rate": 2.6727415003850787e-07, + "loss": 0.8391, + "step": 38607 + }, + { + "epoch": 0.93, + "learning_rate": 2.6709530416981657e-07, + "loss": 1.0854, + "step": 38608 + }, + { + "epoch": 0.93, + "learning_rate": 2.669165173483679e-07, + "loss": 1.0839, + "step": 38609 + }, + { + "epoch": 0.93, + "learning_rate": 2.6673778957524876e-07, + "loss": 1.1388, + "step": 38610 + }, + { + "epoch": 0.93, + "learning_rate": 2.6655912085154055e-07, + "loss": 1.0501, + "step": 38611 + }, + { + "epoch": 0.93, + "learning_rate": 2.6638051117832906e-07, + "loss": 1.0493, + "step": 38612 + }, + { + "epoch": 0.93, + "learning_rate": 2.662019605566979e-07, + "loss": 1.012, + "step": 38613 + }, + { + "epoch": 0.93, + "learning_rate": 2.6602346898772944e-07, + "loss": 1.0691, + "step": 38614 + }, + { + "epoch": 0.93, + "learning_rate": 2.658450364725085e-07, + "loss": 0.8728, + "step": 38615 + }, + { + "epoch": 0.93, + "learning_rate": 2.656666630121152e-07, + "loss": 1.2437, + "step": 38616 + }, + { + "epoch": 0.93, + "learning_rate": 2.654883486076321e-07, + "loss": 1.1426, + "step": 38617 + }, + { + "epoch": 0.93, + "learning_rate": 2.6531009326014156e-07, + "loss": 1.2101, + "step": 38618 + }, + { + "epoch": 0.93, + "learning_rate": 2.6513189697072505e-07, + "loss": 1.3142, + "step": 38619 + }, + { + "epoch": 0.93, + "learning_rate": 2.649537597404639e-07, + "loss": 1.1684, + "step": 38620 + }, + { + "epoch": 0.93, + "learning_rate": 2.6477568157043833e-07, + "loss": 0.9502, + "step": 38621 + }, + { + "epoch": 0.93, + "learning_rate": 2.645976624617286e-07, + "loss": 1.0529, + "step": 38622 + }, + { + "epoch": 0.93, + "learning_rate": 2.6441970241541605e-07, + "loss": 1.1334, + "step": 38623 + }, + { + "epoch": 0.93, + "learning_rate": 2.642418014325776e-07, + "loss": 1.2224, + "step": 38624 + }, + { + "epoch": 0.93, + "learning_rate": 2.6406395951429465e-07, + "loss": 1.1895, + "step": 38625 + }, + { + "epoch": 0.93, + "learning_rate": 2.638861766616463e-07, + "loss": 0.9414, + "step": 38626 + }, + { + "epoch": 0.93, + "learning_rate": 2.637084528757095e-07, + "loss": 0.9063, + "step": 38627 + }, + { + "epoch": 0.93, + "learning_rate": 2.635307881575633e-07, + "loss": 0.8444, + "step": 38628 + }, + { + "epoch": 0.93, + "learning_rate": 2.633531825082869e-07, + "loss": 1.0922, + "step": 38629 + }, + { + "epoch": 0.93, + "learning_rate": 2.631756359289561e-07, + "loss": 1.1023, + "step": 38630 + }, + { + "epoch": 0.93, + "learning_rate": 2.6299814842064897e-07, + "loss": 1.2095, + "step": 38631 + }, + { + "epoch": 0.93, + "learning_rate": 2.6282071998444015e-07, + "loss": 1.1961, + "step": 38632 + }, + { + "epoch": 0.93, + "learning_rate": 2.626433506214088e-07, + "loss": 0.7844, + "step": 38633 + }, + { + "epoch": 0.93, + "learning_rate": 2.6246604033262955e-07, + "loss": 0.8592, + "step": 38634 + }, + { + "epoch": 0.93, + "learning_rate": 2.6228878911917833e-07, + "loss": 1.1658, + "step": 38635 + }, + { + "epoch": 0.93, + "learning_rate": 2.6211159698213087e-07, + "loss": 1.2387, + "step": 38636 + }, + { + "epoch": 0.93, + "learning_rate": 2.6193446392256293e-07, + "loss": 1.0467, + "step": 38637 + }, + { + "epoch": 0.93, + "learning_rate": 2.61757389941546e-07, + "loss": 1.0449, + "step": 38638 + }, + { + "epoch": 0.93, + "learning_rate": 2.6158037504015685e-07, + "loss": 1.0601, + "step": 38639 + }, + { + "epoch": 0.93, + "learning_rate": 2.614034192194703e-07, + "loss": 1.014, + "step": 38640 + }, + { + "epoch": 0.93, + "learning_rate": 2.612265224805577e-07, + "loss": 0.7587, + "step": 38641 + }, + { + "epoch": 0.93, + "learning_rate": 2.6104968482449366e-07, + "loss": 1.1127, + "step": 38642 + }, + { + "epoch": 0.93, + "learning_rate": 2.6087290625234963e-07, + "loss": 0.9199, + "step": 38643 + }, + { + "epoch": 0.93, + "learning_rate": 2.6069618676519806e-07, + "loss": 1.0382, + "step": 38644 + }, + { + "epoch": 0.93, + "learning_rate": 2.605195263641136e-07, + "loss": 1.0194, + "step": 38645 + }, + { + "epoch": 0.93, + "learning_rate": 2.6034292505016546e-07, + "loss": 1.1136, + "step": 38646 + }, + { + "epoch": 0.93, + "learning_rate": 2.601663828244261e-07, + "loss": 1.0791, + "step": 38647 + }, + { + "epoch": 0.93, + "learning_rate": 2.599898996879668e-07, + "loss": 1.0162, + "step": 38648 + }, + { + "epoch": 0.93, + "learning_rate": 2.598134756418569e-07, + "loss": 1.1303, + "step": 38649 + }, + { + "epoch": 0.93, + "learning_rate": 2.5963711068716867e-07, + "loss": 1.1895, + "step": 38650 + }, + { + "epoch": 0.93, + "learning_rate": 2.5946080482497025e-07, + "loss": 0.5845, + "step": 38651 + }, + { + "epoch": 0.93, + "learning_rate": 2.592845580563319e-07, + "loss": 0.7951, + "step": 38652 + }, + { + "epoch": 0.93, + "learning_rate": 2.5910837038232384e-07, + "loss": 1.0199, + "step": 38653 + }, + { + "epoch": 0.93, + "learning_rate": 2.5893224180401297e-07, + "loss": 0.9264, + "step": 38654 + }, + { + "epoch": 0.93, + "learning_rate": 2.587561723224685e-07, + "loss": 1.0195, + "step": 38655 + }, + { + "epoch": 0.93, + "learning_rate": 2.5858016193876066e-07, + "loss": 1.0817, + "step": 38656 + }, + { + "epoch": 0.93, + "learning_rate": 2.584042106539553e-07, + "loss": 0.9851, + "step": 38657 + }, + { + "epoch": 0.93, + "learning_rate": 2.5822831846912035e-07, + "loss": 0.8594, + "step": 38658 + }, + { + "epoch": 0.93, + "learning_rate": 2.5805248538532167e-07, + "loss": 1.3325, + "step": 38659 + }, + { + "epoch": 0.93, + "learning_rate": 2.5787671140362847e-07, + "loss": 1.1719, + "step": 38660 + }, + { + "epoch": 0.93, + "learning_rate": 2.5770099652510537e-07, + "loss": 1.0173, + "step": 38661 + }, + { + "epoch": 0.93, + "learning_rate": 2.5752534075081825e-07, + "loss": 1.1654, + "step": 38662 + }, + { + "epoch": 0.93, + "learning_rate": 2.573497440818329e-07, + "loss": 1.081, + "step": 38663 + }, + { + "epoch": 0.93, + "learning_rate": 2.571742065192162e-07, + "loss": 0.9235, + "step": 38664 + }, + { + "epoch": 0.93, + "learning_rate": 2.569987280640318e-07, + "loss": 0.9799, + "step": 38665 + }, + { + "epoch": 0.93, + "learning_rate": 2.568233087173433e-07, + "loss": 1.1334, + "step": 38666 + }, + { + "epoch": 0.93, + "learning_rate": 2.5664794848021647e-07, + "loss": 0.7363, + "step": 38667 + }, + { + "epoch": 0.93, + "learning_rate": 2.564726473537149e-07, + "loss": 1.1571, + "step": 38668 + }, + { + "epoch": 0.93, + "learning_rate": 2.5629740533890333e-07, + "loss": 1.1424, + "step": 38669 + }, + { + "epoch": 0.93, + "learning_rate": 2.5612222243684206e-07, + "loss": 0.9442, + "step": 38670 + }, + { + "epoch": 0.93, + "learning_rate": 2.559470986485957e-07, + "loss": 0.9501, + "step": 38671 + }, + { + "epoch": 0.93, + "learning_rate": 2.5577203397522676e-07, + "loss": 1.0118, + "step": 38672 + }, + { + "epoch": 0.93, + "learning_rate": 2.5559702841779556e-07, + "loss": 1.0728, + "step": 38673 + }, + { + "epoch": 0.93, + "learning_rate": 2.554220819773667e-07, + "loss": 0.7957, + "step": 38674 + }, + { + "epoch": 0.93, + "learning_rate": 2.5524719465499946e-07, + "loss": 1.0927, + "step": 38675 + }, + { + "epoch": 0.93, + "learning_rate": 2.5507236645175515e-07, + "loss": 1.1391, + "step": 38676 + }, + { + "epoch": 0.93, + "learning_rate": 2.5489759736869513e-07, + "loss": 0.8367, + "step": 38677 + }, + { + "epoch": 0.93, + "learning_rate": 2.5472288740687856e-07, + "loss": 0.8402, + "step": 38678 + }, + { + "epoch": 0.93, + "learning_rate": 2.5454823656736794e-07, + "loss": 0.9671, + "step": 38679 + }, + { + "epoch": 0.93, + "learning_rate": 2.543736448512191e-07, + "loss": 1.0803, + "step": 38680 + }, + { + "epoch": 0.93, + "learning_rate": 2.5419911225949224e-07, + "loss": 1.2229, + "step": 38681 + }, + { + "epoch": 0.93, + "learning_rate": 2.540246387932488e-07, + "loss": 1.0217, + "step": 38682 + }, + { + "epoch": 0.93, + "learning_rate": 2.5385022445354456e-07, + "loss": 0.9699, + "step": 38683 + }, + { + "epoch": 0.93, + "learning_rate": 2.536758692414387e-07, + "loss": 1.0237, + "step": 38684 + }, + { + "epoch": 0.93, + "learning_rate": 2.535015731579893e-07, + "loss": 1.2625, + "step": 38685 + }, + { + "epoch": 0.93, + "learning_rate": 2.5332733620425207e-07, + "loss": 1.2367, + "step": 38686 + }, + { + "epoch": 0.93, + "learning_rate": 2.531531583812874e-07, + "loss": 1.0116, + "step": 38687 + }, + { + "epoch": 0.93, + "learning_rate": 2.5297903969014883e-07, + "loss": 1.0092, + "step": 38688 + }, + { + "epoch": 0.93, + "learning_rate": 2.5280498013189325e-07, + "loss": 1.0477, + "step": 38689 + }, + { + "epoch": 0.93, + "learning_rate": 2.5263097970757767e-07, + "loss": 1.1264, + "step": 38690 + }, + { + "epoch": 0.93, + "learning_rate": 2.5245703841825674e-07, + "loss": 0.9145, + "step": 38691 + }, + { + "epoch": 0.93, + "learning_rate": 2.5228315626498636e-07, + "loss": 1.1243, + "step": 38692 + }, + { + "epoch": 0.93, + "learning_rate": 2.5210933324882224e-07, + "loss": 0.9275, + "step": 38693 + }, + { + "epoch": 0.93, + "learning_rate": 2.51935569370817e-07, + "loss": 0.8219, + "step": 38694 + }, + { + "epoch": 0.93, + "learning_rate": 2.5176186463202635e-07, + "loss": 1.1854, + "step": 38695 + }, + { + "epoch": 0.93, + "learning_rate": 2.5158821903350284e-07, + "loss": 1.1075, + "step": 38696 + }, + { + "epoch": 0.93, + "learning_rate": 2.5141463257630117e-07, + "loss": 0.9803, + "step": 38697 + }, + { + "epoch": 0.93, + "learning_rate": 2.5124110526147383e-07, + "loss": 1.1215, + "step": 38698 + }, + { + "epoch": 0.93, + "learning_rate": 2.510676370900733e-07, + "loss": 1.1829, + "step": 38699 + }, + { + "epoch": 0.93, + "learning_rate": 2.508942280631521e-07, + "loss": 0.7221, + "step": 38700 + }, + { + "epoch": 0.93, + "learning_rate": 2.507208781817638e-07, + "loss": 1.1721, + "step": 38701 + }, + { + "epoch": 0.93, + "learning_rate": 2.505475874469576e-07, + "loss": 1.0245, + "step": 38702 + }, + { + "epoch": 0.93, + "learning_rate": 2.503743558597871e-07, + "loss": 1.109, + "step": 38703 + }, + { + "epoch": 0.93, + "learning_rate": 2.502011834213014e-07, + "loss": 0.9888, + "step": 38704 + }, + { + "epoch": 0.93, + "learning_rate": 2.5002807013255193e-07, + "loss": 1.3331, + "step": 38705 + }, + { + "epoch": 0.93, + "learning_rate": 2.4985501599458896e-07, + "loss": 1.1629, + "step": 38706 + }, + { + "epoch": 0.93, + "learning_rate": 2.496820210084616e-07, + "loss": 0.9259, + "step": 38707 + }, + { + "epoch": 0.93, + "learning_rate": 2.4950908517522023e-07, + "loss": 1.1001, + "step": 38708 + }, + { + "epoch": 0.93, + "learning_rate": 2.493362084959139e-07, + "loss": 1.0498, + "step": 38709 + }, + { + "epoch": 0.93, + "learning_rate": 2.491633909715918e-07, + "loss": 1.0192, + "step": 38710 + }, + { + "epoch": 0.93, + "learning_rate": 2.4899063260330094e-07, + "loss": 0.8716, + "step": 38711 + }, + { + "epoch": 0.93, + "learning_rate": 2.488179333920915e-07, + "loss": 1.0751, + "step": 38712 + }, + { + "epoch": 0.93, + "learning_rate": 2.486452933390082e-07, + "loss": 0.9229, + "step": 38713 + }, + { + "epoch": 0.93, + "learning_rate": 2.484727124451025e-07, + "loss": 1.1008, + "step": 38714 + }, + { + "epoch": 0.93, + "learning_rate": 2.4830019071141796e-07, + "loss": 1.032, + "step": 38715 + }, + { + "epoch": 0.93, + "learning_rate": 2.4812772813900155e-07, + "loss": 0.8948, + "step": 38716 + }, + { + "epoch": 0.93, + "learning_rate": 2.479553247289024e-07, + "loss": 1.1129, + "step": 38717 + }, + { + "epoch": 0.93, + "learning_rate": 2.47782980482163e-07, + "loss": 0.8261, + "step": 38718 + }, + { + "epoch": 0.93, + "learning_rate": 2.4761069539983027e-07, + "loss": 1.1486, + "step": 38719 + }, + { + "epoch": 0.93, + "learning_rate": 2.4743846948295124e-07, + "loss": 1.0675, + "step": 38720 + }, + { + "epoch": 0.93, + "learning_rate": 2.4726630273256724e-07, + "loss": 1.0647, + "step": 38721 + }, + { + "epoch": 0.93, + "learning_rate": 2.470941951497263e-07, + "loss": 1.069, + "step": 38722 + }, + { + "epoch": 0.93, + "learning_rate": 2.4692214673546875e-07, + "loss": 1.2493, + "step": 38723 + }, + { + "epoch": 0.93, + "learning_rate": 2.467501574908415e-07, + "loss": 0.813, + "step": 38724 + }, + { + "epoch": 0.93, + "learning_rate": 2.465782274168882e-07, + "loss": 0.9857, + "step": 38725 + }, + { + "epoch": 0.93, + "learning_rate": 2.464063565146491e-07, + "loss": 1.0783, + "step": 38726 + }, + { + "epoch": 0.93, + "learning_rate": 2.4623454478516775e-07, + "loss": 0.9967, + "step": 38727 + }, + { + "epoch": 0.93, + "learning_rate": 2.460627922294889e-07, + "loss": 1.0867, + "step": 38728 + }, + { + "epoch": 0.93, + "learning_rate": 2.458910988486518e-07, + "loss": 1.0743, + "step": 38729 + }, + { + "epoch": 0.93, + "learning_rate": 2.457194646437011e-07, + "loss": 1.0327, + "step": 38730 + }, + { + "epoch": 0.93, + "learning_rate": 2.455478896156738e-07, + "loss": 0.9155, + "step": 38731 + }, + { + "epoch": 0.93, + "learning_rate": 2.453763737656134e-07, + "loss": 0.9581, + "step": 38732 + }, + { + "epoch": 0.93, + "learning_rate": 2.452049170945614e-07, + "loss": 0.9913, + "step": 38733 + }, + { + "epoch": 0.93, + "learning_rate": 2.4503351960355473e-07, + "loss": 1.0816, + "step": 38734 + }, + { + "epoch": 0.93, + "learning_rate": 2.4486218129363584e-07, + "loss": 0.803, + "step": 38735 + }, + { + "epoch": 0.93, + "learning_rate": 2.44690902165845e-07, + "loss": 0.9191, + "step": 38736 + }, + { + "epoch": 0.93, + "learning_rate": 2.4451968222121815e-07, + "loss": 1.1375, + "step": 38737 + }, + { + "epoch": 0.93, + "learning_rate": 2.443485214607966e-07, + "loss": 0.9643, + "step": 38738 + }, + { + "epoch": 0.93, + "learning_rate": 2.4417741988561727e-07, + "loss": 0.9279, + "step": 38739 + }, + { + "epoch": 0.93, + "learning_rate": 2.440063774967194e-07, + "loss": 1.0502, + "step": 38740 + }, + { + "epoch": 0.93, + "learning_rate": 2.438353942951399e-07, + "loss": 0.9802, + "step": 38741 + }, + { + "epoch": 0.93, + "learning_rate": 2.436644702819169e-07, + "loss": 1.0099, + "step": 38742 + }, + { + "epoch": 0.93, + "learning_rate": 2.4349360545808497e-07, + "loss": 0.9691, + "step": 38743 + }, + { + "epoch": 0.93, + "learning_rate": 2.4332279982468454e-07, + "loss": 1.0254, + "step": 38744 + }, + { + "epoch": 0.93, + "learning_rate": 2.431520533827492e-07, + "loss": 1.043, + "step": 38745 + }, + { + "epoch": 0.93, + "learning_rate": 2.429813661333147e-07, + "loss": 1.0983, + "step": 38746 + }, + { + "epoch": 0.93, + "learning_rate": 2.4281073807741804e-07, + "loss": 0.7361, + "step": 38747 + }, + { + "epoch": 0.93, + "learning_rate": 2.426401692160929e-07, + "loss": 1.0384, + "step": 38748 + }, + { + "epoch": 0.93, + "learning_rate": 2.4246965955037507e-07, + "loss": 1.0523, + "step": 38749 + }, + { + "epoch": 0.93, + "learning_rate": 2.422992090812981e-07, + "loss": 1.0699, + "step": 38750 + }, + { + "epoch": 0.93, + "learning_rate": 2.421288178098968e-07, + "loss": 0.9806, + "step": 38751 + }, + { + "epoch": 0.93, + "learning_rate": 2.419584857372059e-07, + "loss": 1.0317, + "step": 38752 + }, + { + "epoch": 0.93, + "learning_rate": 2.417882128642568e-07, + "loss": 0.9175, + "step": 38753 + }, + { + "epoch": 0.93, + "learning_rate": 2.4161799919208416e-07, + "loss": 1.0947, + "step": 38754 + }, + { + "epoch": 0.93, + "learning_rate": 2.4144784472171945e-07, + "loss": 1.1496, + "step": 38755 + }, + { + "epoch": 0.93, + "learning_rate": 2.4127774945419404e-07, + "loss": 0.7466, + "step": 38756 + }, + { + "epoch": 0.93, + "learning_rate": 2.4110771339054263e-07, + "loss": 0.84, + "step": 38757 + }, + { + "epoch": 0.93, + "learning_rate": 2.409377365317944e-07, + "loss": 1.0356, + "step": 38758 + }, + { + "epoch": 0.93, + "learning_rate": 2.4076781887898195e-07, + "loss": 0.8656, + "step": 38759 + }, + { + "epoch": 0.93, + "learning_rate": 2.4059796043313653e-07, + "loss": 0.7942, + "step": 38760 + }, + { + "epoch": 0.93, + "learning_rate": 2.4042816119528635e-07, + "loss": 0.8991, + "step": 38761 + }, + { + "epoch": 0.93, + "learning_rate": 2.4025842116646267e-07, + "loss": 0.9694, + "step": 38762 + }, + { + "epoch": 0.93, + "learning_rate": 2.4008874034769703e-07, + "loss": 0.9426, + "step": 38763 + }, + { + "epoch": 0.93, + "learning_rate": 2.399191187400163e-07, + "loss": 0.9215, + "step": 38764 + }, + { + "epoch": 0.93, + "learning_rate": 2.3974955634445076e-07, + "loss": 0.8249, + "step": 38765 + }, + { + "epoch": 0.93, + "learning_rate": 2.3958005316202847e-07, + "loss": 1.0915, + "step": 38766 + }, + { + "epoch": 0.93, + "learning_rate": 2.394106091937787e-07, + "loss": 0.9589, + "step": 38767 + }, + { + "epoch": 0.93, + "learning_rate": 2.392412244407294e-07, + "loss": 0.9541, + "step": 38768 + }, + { + "epoch": 0.93, + "learning_rate": 2.390718989039065e-07, + "loss": 0.9248, + "step": 38769 + }, + { + "epoch": 0.93, + "learning_rate": 2.389026325843391e-07, + "loss": 1.2431, + "step": 38770 + }, + { + "epoch": 0.93, + "learning_rate": 2.387334254830542e-07, + "loss": 0.8222, + "step": 38771 + }, + { + "epoch": 0.93, + "learning_rate": 2.385642776010766e-07, + "loss": 1.1702, + "step": 38772 + }, + { + "epoch": 0.93, + "learning_rate": 2.3839518893943424e-07, + "loss": 0.6724, + "step": 38773 + }, + { + "epoch": 0.93, + "learning_rate": 2.3822615949915085e-07, + "loss": 1.1312, + "step": 38774 + }, + { + "epoch": 0.93, + "learning_rate": 2.3805718928125443e-07, + "loss": 1.0613, + "step": 38775 + }, + { + "epoch": 0.93, + "learning_rate": 2.3788827828676864e-07, + "loss": 0.7601, + "step": 38776 + }, + { + "epoch": 0.93, + "learning_rate": 2.377194265167182e-07, + "loss": 0.7931, + "step": 38777 + }, + { + "epoch": 0.93, + "learning_rate": 2.3755063397212785e-07, + "loss": 1.2335, + "step": 38778 + }, + { + "epoch": 0.93, + "learning_rate": 2.3738190065402234e-07, + "loss": 1.1577, + "step": 38779 + }, + { + "epoch": 0.93, + "learning_rate": 2.3721322656342416e-07, + "loss": 0.9541, + "step": 38780 + }, + { + "epoch": 0.93, + "learning_rate": 2.3704461170135695e-07, + "loss": 1.1059, + "step": 38781 + }, + { + "epoch": 0.93, + "learning_rate": 2.3687605606884324e-07, + "loss": 1.2635, + "step": 38782 + }, + { + "epoch": 0.93, + "learning_rate": 2.367075596669055e-07, + "loss": 1.0211, + "step": 38783 + }, + { + "epoch": 0.93, + "learning_rate": 2.3653912249656853e-07, + "loss": 1.2079, + "step": 38784 + }, + { + "epoch": 0.93, + "learning_rate": 2.3637074455885033e-07, + "loss": 1.0485, + "step": 38785 + }, + { + "epoch": 0.93, + "learning_rate": 2.362024258547746e-07, + "loss": 1.029, + "step": 38786 + }, + { + "epoch": 0.93, + "learning_rate": 2.360341663853627e-07, + "loss": 1.1489, + "step": 38787 + }, + { + "epoch": 0.93, + "learning_rate": 2.3586596615163492e-07, + "loss": 1.1752, + "step": 38788 + }, + { + "epoch": 0.93, + "learning_rate": 2.356978251546116e-07, + "loss": 1.0052, + "step": 38789 + }, + { + "epoch": 0.93, + "learning_rate": 2.35529743395313e-07, + "loss": 0.9719, + "step": 38790 + }, + { + "epoch": 0.93, + "learning_rate": 2.3536172087475827e-07, + "loss": 0.8539, + "step": 38791 + }, + { + "epoch": 0.93, + "learning_rate": 2.351937575939678e-07, + "loss": 0.9542, + "step": 38792 + }, + { + "epoch": 0.93, + "learning_rate": 2.3502585355395846e-07, + "loss": 0.8581, + "step": 38793 + }, + { + "epoch": 0.93, + "learning_rate": 2.348580087557517e-07, + "loss": 0.9433, + "step": 38794 + }, + { + "epoch": 0.93, + "learning_rate": 2.3469022320036449e-07, + "loss": 1.2009, + "step": 38795 + }, + { + "epoch": 0.93, + "learning_rate": 2.345224968888149e-07, + "loss": 0.9419, + "step": 38796 + }, + { + "epoch": 0.93, + "learning_rate": 2.3435482982211988e-07, + "loss": 0.8235, + "step": 38797 + }, + { + "epoch": 0.93, + "learning_rate": 2.341872220012975e-07, + "loss": 1.2026, + "step": 38798 + }, + { + "epoch": 0.93, + "learning_rate": 2.3401967342736365e-07, + "loss": 0.9626, + "step": 38799 + }, + { + "epoch": 0.93, + "learning_rate": 2.3385218410133526e-07, + "loss": 1.0854, + "step": 38800 + }, + { + "epoch": 0.93, + "learning_rate": 2.336847540242293e-07, + "loss": 1.051, + "step": 38801 + }, + { + "epoch": 0.93, + "learning_rate": 2.335173831970594e-07, + "loss": 0.9644, + "step": 38802 + }, + { + "epoch": 0.93, + "learning_rate": 2.3335007162084365e-07, + "loss": 0.8152, + "step": 38803 + }, + { + "epoch": 0.93, + "learning_rate": 2.331828192965957e-07, + "loss": 1.0423, + "step": 38804 + }, + { + "epoch": 0.93, + "learning_rate": 2.3301562622532913e-07, + "loss": 1.0866, + "step": 38805 + }, + { + "epoch": 0.93, + "learning_rate": 2.3284849240806096e-07, + "loss": 1.033, + "step": 38806 + }, + { + "epoch": 0.93, + "learning_rate": 2.3268141784580255e-07, + "loss": 1.2665, + "step": 38807 + }, + { + "epoch": 0.93, + "learning_rate": 2.325144025395698e-07, + "loss": 1.1239, + "step": 38808 + }, + { + "epoch": 0.93, + "learning_rate": 2.32347446490373e-07, + "loss": 1.0073, + "step": 38809 + }, + { + "epoch": 0.93, + "learning_rate": 2.3218054969922687e-07, + "loss": 0.8454, + "step": 38810 + }, + { + "epoch": 0.93, + "learning_rate": 2.3201371216714508e-07, + "loss": 1.0441, + "step": 38811 + }, + { + "epoch": 0.93, + "learning_rate": 2.318469338951379e-07, + "loss": 0.9938, + "step": 38812 + }, + { + "epoch": 0.93, + "learning_rate": 2.3168021488421788e-07, + "loss": 0.8409, + "step": 38813 + }, + { + "epoch": 0.93, + "learning_rate": 2.315135551353964e-07, + "loss": 1.0147, + "step": 38814 + }, + { + "epoch": 0.93, + "learning_rate": 2.3134695464968383e-07, + "loss": 0.8673, + "step": 38815 + }, + { + "epoch": 0.93, + "learning_rate": 2.3118041342809262e-07, + "loss": 1.0391, + "step": 38816 + }, + { + "epoch": 0.93, + "learning_rate": 2.31013931471632e-07, + "loss": 0.7238, + "step": 38817 + }, + { + "epoch": 0.93, + "learning_rate": 2.3084750878131112e-07, + "loss": 0.9268, + "step": 38818 + }, + { + "epoch": 0.93, + "learning_rate": 2.3068114535814146e-07, + "loss": 0.985, + "step": 38819 + }, + { + "epoch": 0.93, + "learning_rate": 2.3051484120313105e-07, + "loss": 1.1576, + "step": 38820 + }, + { + "epoch": 0.93, + "learning_rate": 2.3034859631728802e-07, + "loss": 1.09, + "step": 38821 + }, + { + "epoch": 0.93, + "learning_rate": 2.3018241070162372e-07, + "loss": 1.2901, + "step": 38822 + }, + { + "epoch": 0.93, + "learning_rate": 2.3001628435714407e-07, + "loss": 0.857, + "step": 38823 + }, + { + "epoch": 0.93, + "learning_rate": 2.2985021728485824e-07, + "loss": 1.2591, + "step": 38824 + }, + { + "epoch": 0.93, + "learning_rate": 2.2968420948577207e-07, + "loss": 1.15, + "step": 38825 + }, + { + "epoch": 0.93, + "learning_rate": 2.2951826096089368e-07, + "loss": 0.8859, + "step": 38826 + }, + { + "epoch": 0.93, + "learning_rate": 2.2935237171123114e-07, + "loss": 1.1768, + "step": 38827 + }, + { + "epoch": 0.93, + "learning_rate": 2.291865417377881e-07, + "loss": 0.9145, + "step": 38828 + }, + { + "epoch": 0.93, + "learning_rate": 2.290207710415726e-07, + "loss": 1.1947, + "step": 38829 + }, + { + "epoch": 0.93, + "learning_rate": 2.2885505962359055e-07, + "loss": 1.1032, + "step": 38830 + }, + { + "epoch": 0.93, + "learning_rate": 2.286894074848456e-07, + "loss": 0.9707, + "step": 38831 + }, + { + "epoch": 0.93, + "learning_rate": 2.2852381462634354e-07, + "loss": 1.1219, + "step": 38832 + }, + { + "epoch": 0.93, + "learning_rate": 2.2835828104908919e-07, + "loss": 1.0229, + "step": 38833 + }, + { + "epoch": 0.93, + "learning_rate": 2.281928067540884e-07, + "loss": 1.1587, + "step": 38834 + }, + { + "epoch": 0.93, + "learning_rate": 2.2802739174234145e-07, + "loss": 1.1882, + "step": 38835 + }, + { + "epoch": 0.93, + "learning_rate": 2.2786203601485424e-07, + "loss": 1.2978, + "step": 38836 + }, + { + "epoch": 0.93, + "learning_rate": 2.2769673957262928e-07, + "loss": 1.2594, + "step": 38837 + }, + { + "epoch": 0.93, + "learning_rate": 2.2753150241667022e-07, + "loss": 0.9981, + "step": 38838 + }, + { + "epoch": 0.93, + "learning_rate": 2.2736632454797847e-07, + "loss": 1.0905, + "step": 38839 + }, + { + "epoch": 0.93, + "learning_rate": 2.272012059675577e-07, + "loss": 1.0708, + "step": 38840 + }, + { + "epoch": 0.93, + "learning_rate": 2.2703614667640705e-07, + "loss": 0.9585, + "step": 38841 + }, + { + "epoch": 0.93, + "learning_rate": 2.2687114667552912e-07, + "loss": 1.1156, + "step": 38842 + }, + { + "epoch": 0.93, + "learning_rate": 2.267062059659264e-07, + "loss": 0.9882, + "step": 38843 + }, + { + "epoch": 0.93, + "learning_rate": 2.2654132454859812e-07, + "loss": 1.199, + "step": 38844 + }, + { + "epoch": 0.93, + "learning_rate": 2.2637650242454457e-07, + "loss": 0.7245, + "step": 38845 + }, + { + "epoch": 0.93, + "learning_rate": 2.2621173959476607e-07, + "loss": 1.0915, + "step": 38846 + }, + { + "epoch": 0.93, + "learning_rate": 2.260470360602607e-07, + "loss": 1.0608, + "step": 38847 + }, + { + "epoch": 0.93, + "learning_rate": 2.2588239182203097e-07, + "loss": 0.8254, + "step": 38848 + }, + { + "epoch": 0.93, + "learning_rate": 2.2571780688107281e-07, + "loss": 1.0934, + "step": 38849 + }, + { + "epoch": 0.93, + "learning_rate": 2.2555328123838537e-07, + "loss": 0.983, + "step": 38850 + }, + { + "epoch": 0.93, + "learning_rate": 2.2538881489496789e-07, + "loss": 1.2674, + "step": 38851 + }, + { + "epoch": 0.93, + "learning_rate": 2.252244078518162e-07, + "loss": 1.0092, + "step": 38852 + }, + { + "epoch": 0.93, + "learning_rate": 2.2506006010992953e-07, + "loss": 1.2815, + "step": 38853 + }, + { + "epoch": 0.93, + "learning_rate": 2.2489577167030486e-07, + "loss": 0.8119, + "step": 38854 + }, + { + "epoch": 0.93, + "learning_rate": 2.2473154253393692e-07, + "loss": 0.8534, + "step": 38855 + }, + { + "epoch": 0.93, + "learning_rate": 2.2456737270182493e-07, + "loss": 1.1576, + "step": 38856 + }, + { + "epoch": 0.93, + "learning_rate": 2.2440326217496254e-07, + "loss": 1.2025, + "step": 38857 + }, + { + "epoch": 0.93, + "learning_rate": 2.2423921095434564e-07, + "loss": 1.1208, + "step": 38858 + }, + { + "epoch": 0.93, + "learning_rate": 2.2407521904097117e-07, + "loss": 1.0052, + "step": 38859 + }, + { + "epoch": 0.93, + "learning_rate": 2.239112864358317e-07, + "loss": 1.1858, + "step": 38860 + }, + { + "epoch": 0.93, + "learning_rate": 2.2374741313992422e-07, + "loss": 1.0447, + "step": 38861 + }, + { + "epoch": 0.93, + "learning_rate": 2.235835991542412e-07, + "loss": 1.2484, + "step": 38862 + }, + { + "epoch": 0.93, + "learning_rate": 2.234198444797764e-07, + "loss": 1.057, + "step": 38863 + }, + { + "epoch": 0.93, + "learning_rate": 2.2325614911752557e-07, + "loss": 0.976, + "step": 38864 + }, + { + "epoch": 0.93, + "learning_rate": 2.23092513068478e-07, + "loss": 1.0013, + "step": 38865 + }, + { + "epoch": 0.93, + "learning_rate": 2.2292893633362844e-07, + "loss": 1.0153, + "step": 38866 + }, + { + "epoch": 0.93, + "learning_rate": 2.2276541891397164e-07, + "loss": 1.1022, + "step": 38867 + }, + { + "epoch": 0.93, + "learning_rate": 2.2260196081049458e-07, + "loss": 1.1076, + "step": 38868 + }, + { + "epoch": 0.93, + "learning_rate": 2.2243856202419422e-07, + "loss": 1.208, + "step": 38869 + }, + { + "epoch": 0.93, + "learning_rate": 2.222752225560576e-07, + "loss": 0.909, + "step": 38870 + }, + { + "epoch": 0.93, + "learning_rate": 2.2211194240707723e-07, + "loss": 0.9293, + "step": 38871 + }, + { + "epoch": 0.93, + "learning_rate": 2.2194872157824453e-07, + "loss": 1.1143, + "step": 38872 + }, + { + "epoch": 0.93, + "learning_rate": 2.2178556007054876e-07, + "loss": 0.9813, + "step": 38873 + }, + { + "epoch": 0.93, + "learning_rate": 2.2162245788497906e-07, + "loss": 1.1123, + "step": 38874 + }, + { + "epoch": 0.93, + "learning_rate": 2.21459415022528e-07, + "loss": 0.8977, + "step": 38875 + }, + { + "epoch": 0.93, + "learning_rate": 2.2129643148418034e-07, + "loss": 1.0627, + "step": 38876 + }, + { + "epoch": 0.93, + "learning_rate": 2.2113350727092975e-07, + "loss": 1.001, + "step": 38877 + }, + { + "epoch": 0.93, + "learning_rate": 2.2097064238375987e-07, + "loss": 1.0933, + "step": 38878 + }, + { + "epoch": 0.93, + "learning_rate": 2.2080783682366102e-07, + "loss": 0.881, + "step": 38879 + }, + { + "epoch": 0.94, + "learning_rate": 2.2064509059162243e-07, + "loss": 1.053, + "step": 38880 + }, + { + "epoch": 0.94, + "learning_rate": 2.204824036886277e-07, + "loss": 0.957, + "step": 38881 + }, + { + "epoch": 0.94, + "learning_rate": 2.2031977611566723e-07, + "loss": 0.8768, + "step": 38882 + }, + { + "epoch": 0.94, + "learning_rate": 2.2015720787372686e-07, + "loss": 1.0009, + "step": 38883 + }, + { + "epoch": 0.94, + "learning_rate": 2.1999469896379133e-07, + "loss": 1.1877, + "step": 38884 + }, + { + "epoch": 0.94, + "learning_rate": 2.1983224938684655e-07, + "loss": 0.9388, + "step": 38885 + }, + { + "epoch": 0.94, + "learning_rate": 2.1966985914388173e-07, + "loss": 0.8698, + "step": 38886 + }, + { + "epoch": 0.94, + "learning_rate": 2.1950752823587717e-07, + "loss": 1.118, + "step": 38887 + }, + { + "epoch": 0.94, + "learning_rate": 2.1934525666382211e-07, + "loss": 0.8205, + "step": 38888 + }, + { + "epoch": 0.94, + "learning_rate": 2.1918304442869686e-07, + "loss": 1.0367, + "step": 38889 + }, + { + "epoch": 0.94, + "learning_rate": 2.190208915314873e-07, + "loss": 1.0077, + "step": 38890 + }, + { + "epoch": 0.94, + "learning_rate": 2.1885879797317933e-07, + "loss": 0.7632, + "step": 38891 + }, + { + "epoch": 0.94, + "learning_rate": 2.1869676375475324e-07, + "loss": 0.8647, + "step": 38892 + }, + { + "epoch": 0.94, + "learning_rate": 2.185347888771927e-07, + "loss": 1.1059, + "step": 38893 + }, + { + "epoch": 0.94, + "learning_rate": 2.183728733414814e-07, + "loss": 0.8794, + "step": 38894 + }, + { + "epoch": 0.94, + "learning_rate": 2.182110171486007e-07, + "loss": 1.1785, + "step": 38895 + }, + { + "epoch": 0.94, + "learning_rate": 2.1804922029953436e-07, + "loss": 0.946, + "step": 38896 + }, + { + "epoch": 0.94, + "learning_rate": 2.178874827952615e-07, + "loss": 1.1083, + "step": 38897 + }, + { + "epoch": 0.94, + "learning_rate": 2.1772580463676364e-07, + "loss": 0.995, + "step": 38898 + }, + { + "epoch": 0.94, + "learning_rate": 2.1756418582502326e-07, + "loss": 0.7825, + "step": 38899 + }, + { + "epoch": 0.94, + "learning_rate": 2.174026263610196e-07, + "loss": 1.1414, + "step": 38900 + }, + { + "epoch": 0.94, + "learning_rate": 2.17241126245733e-07, + "loss": 0.7158, + "step": 38901 + }, + { + "epoch": 0.94, + "learning_rate": 2.1707968548014492e-07, + "loss": 0.9963, + "step": 38902 + }, + { + "epoch": 0.94, + "learning_rate": 2.1691830406523229e-07, + "loss": 1.0144, + "step": 38903 + }, + { + "epoch": 0.94, + "learning_rate": 2.1675698200197547e-07, + "loss": 1.0698, + "step": 38904 + }, + { + "epoch": 0.94, + "learning_rate": 2.165957192913526e-07, + "loss": 1.0642, + "step": 38905 + }, + { + "epoch": 0.94, + "learning_rate": 2.1643451593434172e-07, + "loss": 1.045, + "step": 38906 + }, + { + "epoch": 0.94, + "learning_rate": 2.162733719319221e-07, + "loss": 1.0546, + "step": 38907 + }, + { + "epoch": 0.94, + "learning_rate": 2.1611228728507072e-07, + "loss": 1.2114, + "step": 38908 + }, + { + "epoch": 0.94, + "learning_rate": 2.1595126199476457e-07, + "loss": 1.1867, + "step": 38909 + }, + { + "epoch": 0.94, + "learning_rate": 2.1579029606198065e-07, + "loss": 0.8311, + "step": 38910 + }, + { + "epoch": 0.94, + "learning_rate": 2.1562938948769596e-07, + "loss": 1.1843, + "step": 38911 + }, + { + "epoch": 0.94, + "learning_rate": 2.154685422728864e-07, + "loss": 0.9447, + "step": 38912 + }, + { + "epoch": 0.94, + "learning_rate": 2.153077544185267e-07, + "loss": 1.025, + "step": 38913 + }, + { + "epoch": 0.94, + "learning_rate": 2.1514702592559388e-07, + "loss": 1.2549, + "step": 38914 + }, + { + "epoch": 0.94, + "learning_rate": 2.1498635679506386e-07, + "loss": 0.8921, + "step": 38915 + }, + { + "epoch": 0.94, + "learning_rate": 2.1482574702790804e-07, + "loss": 1.1343, + "step": 38916 + }, + { + "epoch": 0.94, + "learning_rate": 2.1466519662510343e-07, + "loss": 1.049, + "step": 38917 + }, + { + "epoch": 0.94, + "learning_rate": 2.145047055876248e-07, + "loss": 0.9712, + "step": 38918 + }, + { + "epoch": 0.94, + "learning_rate": 2.1434427391644252e-07, + "loss": 1.052, + "step": 38919 + }, + { + "epoch": 0.94, + "learning_rate": 2.1418390161253356e-07, + "loss": 1.0994, + "step": 38920 + }, + { + "epoch": 0.94, + "learning_rate": 2.1402358867686824e-07, + "loss": 0.9569, + "step": 38921 + }, + { + "epoch": 0.94, + "learning_rate": 2.1386333511041912e-07, + "loss": 0.8285, + "step": 38922 + }, + { + "epoch": 0.94, + "learning_rate": 2.13703140914161e-07, + "loss": 1.0273, + "step": 38923 + }, + { + "epoch": 0.94, + "learning_rate": 2.1354300608906198e-07, + "loss": 0.9678, + "step": 38924 + }, + { + "epoch": 0.94, + "learning_rate": 2.1338293063609682e-07, + "loss": 0.9592, + "step": 38925 + }, + { + "epoch": 0.94, + "learning_rate": 2.132229145562359e-07, + "loss": 1.0734, + "step": 38926 + }, + { + "epoch": 0.94, + "learning_rate": 2.130629578504495e-07, + "loss": 0.9859, + "step": 38927 + }, + { + "epoch": 0.94, + "learning_rate": 2.12903060519708e-07, + "loss": 1.0757, + "step": 38928 + }, + { + "epoch": 0.94, + "learning_rate": 2.1274322256498058e-07, + "loss": 1.2003, + "step": 38929 + }, + { + "epoch": 0.94, + "learning_rate": 2.1258344398723875e-07, + "loss": 0.7861, + "step": 38930 + }, + { + "epoch": 0.94, + "learning_rate": 2.1242372478745165e-07, + "loss": 1.1331, + "step": 38931 + }, + { + "epoch": 0.94, + "learning_rate": 2.1226406496658635e-07, + "loss": 1.0276, + "step": 38932 + }, + { + "epoch": 0.94, + "learning_rate": 2.1210446452561319e-07, + "loss": 0.9462, + "step": 38933 + }, + { + "epoch": 0.94, + "learning_rate": 2.119449234655002e-07, + "loss": 0.9023, + "step": 38934 + }, + { + "epoch": 0.94, + "learning_rate": 2.117854417872145e-07, + "loss": 1.0921, + "step": 38935 + }, + { + "epoch": 0.94, + "learning_rate": 2.1162601949172413e-07, + "loss": 0.9302, + "step": 38936 + }, + { + "epoch": 0.94, + "learning_rate": 2.114666565799972e-07, + "loss": 1.2326, + "step": 38937 + }, + { + "epoch": 0.94, + "learning_rate": 2.1130735305299966e-07, + "loss": 1.0827, + "step": 38938 + }, + { + "epoch": 0.94, + "learning_rate": 2.1114810891169846e-07, + "loss": 0.9916, + "step": 38939 + }, + { + "epoch": 0.94, + "learning_rate": 2.109889241570573e-07, + "loss": 0.9555, + "step": 38940 + }, + { + "epoch": 0.94, + "learning_rate": 2.1082979879004428e-07, + "loss": 1.1959, + "step": 38941 + }, + { + "epoch": 0.94, + "learning_rate": 2.1067073281162642e-07, + "loss": 0.8771, + "step": 38942 + }, + { + "epoch": 0.94, + "learning_rate": 2.1051172622276406e-07, + "loss": 1.1464, + "step": 38943 + }, + { + "epoch": 0.94, + "learning_rate": 2.103527790244253e-07, + "loss": 0.9345, + "step": 38944 + }, + { + "epoch": 0.94, + "learning_rate": 2.1019389121757493e-07, + "loss": 1.0074, + "step": 38945 + }, + { + "epoch": 0.94, + "learning_rate": 2.1003506280317332e-07, + "loss": 1.0192, + "step": 38946 + }, + { + "epoch": 0.94, + "learning_rate": 2.0987629378218855e-07, + "loss": 0.7229, + "step": 38947 + }, + { + "epoch": 0.94, + "learning_rate": 2.0971758415557986e-07, + "loss": 0.6265, + "step": 38948 + }, + { + "epoch": 0.94, + "learning_rate": 2.0955893392431202e-07, + "loss": 1.2252, + "step": 38949 + }, + { + "epoch": 0.94, + "learning_rate": 2.0940034308934764e-07, + "loss": 1.0491, + "step": 38950 + }, + { + "epoch": 0.94, + "learning_rate": 2.0924181165164814e-07, + "loss": 1.0145, + "step": 38951 + }, + { + "epoch": 0.94, + "learning_rate": 2.0908333961217496e-07, + "loss": 1.23, + "step": 38952 + }, + { + "epoch": 0.94, + "learning_rate": 2.0892492697189182e-07, + "loss": 0.9947, + "step": 38953 + }, + { + "epoch": 0.94, + "learning_rate": 2.0876657373175678e-07, + "loss": 0.8644, + "step": 38954 + }, + { + "epoch": 0.94, + "learning_rate": 2.0860827989273357e-07, + "loss": 1.0538, + "step": 38955 + }, + { + "epoch": 0.94, + "learning_rate": 2.0845004545577919e-07, + "loss": 1.0557, + "step": 38956 + }, + { + "epoch": 0.94, + "learning_rate": 2.0829187042185505e-07, + "loss": 1.0451, + "step": 38957 + }, + { + "epoch": 0.94, + "learning_rate": 2.0813375479192266e-07, + "loss": 0.9285, + "step": 38958 + }, + { + "epoch": 0.94, + "learning_rate": 2.079756985669379e-07, + "loss": 1.0264, + "step": 38959 + }, + { + "epoch": 0.94, + "learning_rate": 2.0781770174786219e-07, + "loss": 0.9555, + "step": 38960 + }, + { + "epoch": 0.94, + "learning_rate": 2.076597643356537e-07, + "loss": 1.1212, + "step": 38961 + }, + { + "epoch": 0.94, + "learning_rate": 2.0750188633126943e-07, + "loss": 0.9815, + "step": 38962 + }, + { + "epoch": 0.94, + "learning_rate": 2.0734406773566861e-07, + "loss": 0.8859, + "step": 38963 + }, + { + "epoch": 0.94, + "learning_rate": 2.0718630854980715e-07, + "loss": 0.9419, + "step": 38964 + }, + { + "epoch": 0.94, + "learning_rate": 2.0702860877464316e-07, + "loss": 1.1737, + "step": 38965 + }, + { + "epoch": 0.94, + "learning_rate": 2.0687096841113364e-07, + "loss": 1.0183, + "step": 38966 + }, + { + "epoch": 0.94, + "learning_rate": 2.067133874602334e-07, + "loss": 1.0809, + "step": 38967 + }, + { + "epoch": 0.94, + "learning_rate": 2.0655586592289944e-07, + "loss": 1.1553, + "step": 38968 + }, + { + "epoch": 0.94, + "learning_rate": 2.063984038000888e-07, + "loss": 0.9044, + "step": 38969 + }, + { + "epoch": 0.94, + "learning_rate": 2.0624100109275514e-07, + "loss": 1.1925, + "step": 38970 + }, + { + "epoch": 0.94, + "learning_rate": 2.0608365780185325e-07, + "loss": 0.8147, + "step": 38971 + }, + { + "epoch": 0.94, + "learning_rate": 2.059263739283379e-07, + "loss": 1.1376, + "step": 38972 + }, + { + "epoch": 0.94, + "learning_rate": 2.0576914947316397e-07, + "loss": 0.8131, + "step": 38973 + }, + { + "epoch": 0.94, + "learning_rate": 2.0561198443728502e-07, + "loss": 1.0327, + "step": 38974 + }, + { + "epoch": 0.94, + "learning_rate": 2.054548788216537e-07, + "loss": 1.3241, + "step": 38975 + }, + { + "epoch": 0.94, + "learning_rate": 2.052978326272237e-07, + "loss": 0.8841, + "step": 38976 + }, + { + "epoch": 0.94, + "learning_rate": 2.0514084585494976e-07, + "loss": 0.8509, + "step": 38977 + }, + { + "epoch": 0.94, + "learning_rate": 2.0498391850578002e-07, + "loss": 1.0568, + "step": 38978 + }, + { + "epoch": 0.94, + "learning_rate": 2.0482705058067042e-07, + "loss": 1.2337, + "step": 38979 + }, + { + "epoch": 0.94, + "learning_rate": 2.0467024208057128e-07, + "loss": 1.0832, + "step": 38980 + }, + { + "epoch": 0.94, + "learning_rate": 2.0451349300643297e-07, + "loss": 0.9396, + "step": 38981 + }, + { + "epoch": 0.94, + "learning_rate": 2.0435680335920805e-07, + "loss": 0.7801, + "step": 38982 + }, + { + "epoch": 0.94, + "learning_rate": 2.0420017313984464e-07, + "loss": 0.886, + "step": 38983 + }, + { + "epoch": 0.94, + "learning_rate": 2.0404360234929644e-07, + "loss": 1.0107, + "step": 38984 + }, + { + "epoch": 0.94, + "learning_rate": 2.0388709098851155e-07, + "loss": 0.7239, + "step": 38985 + }, + { + "epoch": 0.94, + "learning_rate": 2.0373063905843815e-07, + "loss": 1.0751, + "step": 38986 + }, + { + "epoch": 0.94, + "learning_rate": 2.0357424656002657e-07, + "loss": 1.1001, + "step": 38987 + }, + { + "epoch": 0.94, + "learning_rate": 2.0341791349422713e-07, + "loss": 0.9956, + "step": 38988 + }, + { + "epoch": 0.94, + "learning_rate": 2.032616398619869e-07, + "loss": 1.2059, + "step": 38989 + }, + { + "epoch": 0.94, + "learning_rate": 2.0310542566425396e-07, + "loss": 0.9422, + "step": 38990 + }, + { + "epoch": 0.94, + "learning_rate": 2.0294927090197537e-07, + "loss": 0.5757, + "step": 38991 + }, + { + "epoch": 0.94, + "learning_rate": 2.0279317557609922e-07, + "loss": 0.9601, + "step": 38992 + }, + { + "epoch": 0.94, + "learning_rate": 2.0263713968757258e-07, + "loss": 0.6857, + "step": 38993 + }, + { + "epoch": 0.94, + "learning_rate": 2.0248116323734134e-07, + "loss": 0.9873, + "step": 38994 + }, + { + "epoch": 0.94, + "learning_rate": 2.0232524622635253e-07, + "loss": 0.9821, + "step": 38995 + }, + { + "epoch": 0.94, + "learning_rate": 2.0216938865555312e-07, + "loss": 0.8997, + "step": 38996 + }, + { + "epoch": 0.94, + "learning_rate": 2.0201359052588688e-07, + "loss": 1.2152, + "step": 38997 + }, + { + "epoch": 0.94, + "learning_rate": 2.0185785183829964e-07, + "loss": 0.9204, + "step": 38998 + }, + { + "epoch": 0.94, + "learning_rate": 2.0170217259373405e-07, + "loss": 0.9658, + "step": 38999 + }, + { + "epoch": 0.94, + "learning_rate": 2.0154655279313927e-07, + "loss": 0.9549, + "step": 39000 + }, + { + "epoch": 0.94, + "learning_rate": 2.0139099243745575e-07, + "loss": 0.8675, + "step": 39001 + }, + { + "epoch": 0.94, + "learning_rate": 2.0123549152762823e-07, + "loss": 1.3559, + "step": 39002 + }, + { + "epoch": 0.94, + "learning_rate": 2.0108005006460153e-07, + "loss": 0.9651, + "step": 39003 + }, + { + "epoch": 0.94, + "learning_rate": 2.0092466804931597e-07, + "loss": 0.9663, + "step": 39004 + }, + { + "epoch": 0.94, + "learning_rate": 2.007693454827153e-07, + "loss": 1.0472, + "step": 39005 + }, + { + "epoch": 0.94, + "learning_rate": 2.006140823657432e-07, + "loss": 1.0318, + "step": 39006 + }, + { + "epoch": 0.94, + "learning_rate": 2.0045887869934e-07, + "loss": 0.8903, + "step": 39007 + }, + { + "epoch": 0.94, + "learning_rate": 2.0030373448444719e-07, + "loss": 1.0903, + "step": 39008 + }, + { + "epoch": 0.94, + "learning_rate": 2.001486497220073e-07, + "loss": 1.1861, + "step": 39009 + }, + { + "epoch": 0.94, + "learning_rate": 1.9999362441296077e-07, + "loss": 1.1209, + "step": 39010 + }, + { + "epoch": 0.94, + "learning_rate": 1.9983865855824791e-07, + "loss": 1.0828, + "step": 39011 + }, + { + "epoch": 0.94, + "learning_rate": 1.99683752158808e-07, + "loss": 0.9631, + "step": 39012 + }, + { + "epoch": 0.94, + "learning_rate": 1.9952890521558133e-07, + "loss": 1.3433, + "step": 39013 + }, + { + "epoch": 0.94, + "learning_rate": 1.9937411772950831e-07, + "loss": 0.9788, + "step": 39014 + }, + { + "epoch": 0.94, + "learning_rate": 1.992193897015271e-07, + "loss": 0.9661, + "step": 39015 + }, + { + "epoch": 0.94, + "learning_rate": 1.990647211325758e-07, + "loss": 0.9112, + "step": 39016 + }, + { + "epoch": 0.94, + "learning_rate": 1.9891011202359477e-07, + "loss": 0.8124, + "step": 39017 + }, + { + "epoch": 0.94, + "learning_rate": 1.9875556237551997e-07, + "loss": 1.042, + "step": 39018 + }, + { + "epoch": 0.94, + "learning_rate": 1.986010721892906e-07, + "loss": 0.8525, + "step": 39019 + }, + { + "epoch": 0.94, + "learning_rate": 1.9844664146584258e-07, + "loss": 1.3315, + "step": 39020 + }, + { + "epoch": 0.94, + "learning_rate": 1.9829227020611296e-07, + "loss": 1.2498, + "step": 39021 + }, + { + "epoch": 0.94, + "learning_rate": 1.981379584110399e-07, + "loss": 1.0457, + "step": 39022 + }, + { + "epoch": 0.94, + "learning_rate": 1.9798370608155593e-07, + "loss": 1.006, + "step": 39023 + }, + { + "epoch": 0.94, + "learning_rate": 1.9782951321860033e-07, + "loss": 1.1426, + "step": 39024 + }, + { + "epoch": 0.94, + "learning_rate": 1.9767537982310903e-07, + "loss": 1.086, + "step": 39025 + }, + { + "epoch": 0.94, + "learning_rate": 1.9752130589601348e-07, + "loss": 1.2843, + "step": 39026 + }, + { + "epoch": 0.94, + "learning_rate": 1.9736729143825185e-07, + "loss": 0.8516, + "step": 39027 + }, + { + "epoch": 0.94, + "learning_rate": 1.972133364507567e-07, + "loss": 1.0939, + "step": 39028 + }, + { + "epoch": 0.94, + "learning_rate": 1.9705944093446173e-07, + "loss": 1.0016, + "step": 39029 + }, + { + "epoch": 0.94, + "learning_rate": 1.9690560489030285e-07, + "loss": 0.9042, + "step": 39030 + }, + { + "epoch": 0.94, + "learning_rate": 1.9675182831921048e-07, + "loss": 0.6342, + "step": 39031 + }, + { + "epoch": 0.94, + "learning_rate": 1.9659811122211936e-07, + "loss": 1.2959, + "step": 39032 + }, + { + "epoch": 0.94, + "learning_rate": 1.964444535999621e-07, + "loss": 0.7939, + "step": 39033 + }, + { + "epoch": 0.94, + "learning_rate": 1.962908554536691e-07, + "loss": 1.0266, + "step": 39034 + }, + { + "epoch": 0.94, + "learning_rate": 1.9613731678417515e-07, + "loss": 0.9646, + "step": 39035 + }, + { + "epoch": 0.94, + "learning_rate": 1.9598383759240948e-07, + "loss": 0.7117, + "step": 39036 + }, + { + "epoch": 0.94, + "learning_rate": 1.9583041787930356e-07, + "loss": 1.0649, + "step": 39037 + }, + { + "epoch": 0.94, + "learning_rate": 1.9567705764578894e-07, + "loss": 0.9238, + "step": 39038 + }, + { + "epoch": 0.94, + "learning_rate": 1.955237568927948e-07, + "loss": 0.9235, + "step": 39039 + }, + { + "epoch": 0.94, + "learning_rate": 1.9537051562125152e-07, + "loss": 1.0745, + "step": 39040 + }, + { + "epoch": 0.94, + "learning_rate": 1.9521733383209063e-07, + "loss": 0.8587, + "step": 39041 + }, + { + "epoch": 0.94, + "learning_rate": 1.95064211526238e-07, + "loss": 1.1114, + "step": 39042 + }, + { + "epoch": 0.94, + "learning_rate": 1.9491114870462625e-07, + "loss": 0.6345, + "step": 39043 + }, + { + "epoch": 0.94, + "learning_rate": 1.9475814536818127e-07, + "loss": 0.894, + "step": 39044 + }, + { + "epoch": 0.94, + "learning_rate": 1.9460520151783235e-07, + "loss": 0.9334, + "step": 39045 + }, + { + "epoch": 0.94, + "learning_rate": 1.9445231715450875e-07, + "loss": 1.0389, + "step": 39046 + }, + { + "epoch": 0.94, + "learning_rate": 1.9429949227913415e-07, + "loss": 1.0326, + "step": 39047 + }, + { + "epoch": 0.94, + "learning_rate": 1.941467268926389e-07, + "loss": 0.948, + "step": 39048 + }, + { + "epoch": 0.94, + "learning_rate": 1.939940209959501e-07, + "loss": 1.0181, + "step": 39049 + }, + { + "epoch": 0.94, + "learning_rate": 1.9384137458999252e-07, + "loss": 1.169, + "step": 39050 + }, + { + "epoch": 0.94, + "learning_rate": 1.9368878767569322e-07, + "loss": 1.1165, + "step": 39051 + }, + { + "epoch": 0.94, + "learning_rate": 1.9353626025397588e-07, + "loss": 1.1681, + "step": 39052 + }, + { + "epoch": 0.94, + "learning_rate": 1.9338379232576865e-07, + "loss": 0.9783, + "step": 39053 + }, + { + "epoch": 0.94, + "learning_rate": 1.9323138389199526e-07, + "loss": 1.1235, + "step": 39054 + }, + { + "epoch": 0.94, + "learning_rate": 1.930790349535794e-07, + "loss": 0.9638, + "step": 39055 + }, + { + "epoch": 0.94, + "learning_rate": 1.9292674551144696e-07, + "loss": 1.0199, + "step": 39056 + }, + { + "epoch": 0.94, + "learning_rate": 1.9277451556652173e-07, + "loss": 0.8353, + "step": 39057 + }, + { + "epoch": 0.94, + "learning_rate": 1.9262234511972622e-07, + "loss": 0.7396, + "step": 39058 + }, + { + "epoch": 0.94, + "learning_rate": 1.924702341719842e-07, + "loss": 1.0724, + "step": 39059 + }, + { + "epoch": 0.94, + "learning_rate": 1.923181827242182e-07, + "loss": 0.9569, + "step": 39060 + }, + { + "epoch": 0.94, + "learning_rate": 1.9216619077735088e-07, + "loss": 0.9773, + "step": 39061 + }, + { + "epoch": 0.94, + "learning_rate": 1.9201425833230481e-07, + "loss": 1.1062, + "step": 39062 + }, + { + "epoch": 0.94, + "learning_rate": 1.9186238539000147e-07, + "loss": 1.0649, + "step": 39063 + }, + { + "epoch": 0.94, + "learning_rate": 1.917105719513601e-07, + "loss": 1.1787, + "step": 39064 + }, + { + "epoch": 0.94, + "learning_rate": 1.9155881801730557e-07, + "loss": 1.2323, + "step": 39065 + }, + { + "epoch": 0.94, + "learning_rate": 1.9140712358875602e-07, + "loss": 0.9647, + "step": 39066 + }, + { + "epoch": 0.94, + "learning_rate": 1.912554886666318e-07, + "loss": 0.8748, + "step": 39067 + }, + { + "epoch": 0.94, + "learning_rate": 1.911039132518544e-07, + "loss": 1.1485, + "step": 39068 + }, + { + "epoch": 0.94, + "learning_rate": 1.90952397345342e-07, + "loss": 1.1368, + "step": 39069 + }, + { + "epoch": 0.94, + "learning_rate": 1.9080094094801493e-07, + "loss": 1.1271, + "step": 39070 + }, + { + "epoch": 0.94, + "learning_rate": 1.9064954406079027e-07, + "loss": 1.0633, + "step": 39071 + }, + { + "epoch": 0.94, + "learning_rate": 1.904982066845873e-07, + "loss": 0.8933, + "step": 39072 + }, + { + "epoch": 0.94, + "learning_rate": 1.9034692882032524e-07, + "loss": 1.1852, + "step": 39073 + }, + { + "epoch": 0.94, + "learning_rate": 1.9019571046892006e-07, + "loss": 1.0253, + "step": 39074 + }, + { + "epoch": 0.94, + "learning_rate": 1.900445516312899e-07, + "loss": 0.8928, + "step": 39075 + }, + { + "epoch": 0.94, + "learning_rate": 1.8989345230835289e-07, + "loss": 1.1565, + "step": 39076 + }, + { + "epoch": 0.94, + "learning_rate": 1.8974241250102498e-07, + "loss": 1.1125, + "step": 39077 + }, + { + "epoch": 0.94, + "learning_rate": 1.8959143221022214e-07, + "loss": 1.0286, + "step": 39078 + }, + { + "epoch": 0.94, + "learning_rate": 1.8944051143686027e-07, + "loss": 0.9255, + "step": 39079 + }, + { + "epoch": 0.94, + "learning_rate": 1.892896501818553e-07, + "loss": 1.0082, + "step": 39080 + }, + { + "epoch": 0.94, + "learning_rate": 1.8913884844612208e-07, + "loss": 0.9508, + "step": 39081 + }, + { + "epoch": 0.94, + "learning_rate": 1.8898810623057651e-07, + "loss": 0.9435, + "step": 39082 + }, + { + "epoch": 0.94, + "learning_rate": 1.888374235361312e-07, + "loss": 1.0284, + "step": 39083 + }, + { + "epoch": 0.94, + "learning_rate": 1.8868680036370323e-07, + "loss": 1.185, + "step": 39084 + }, + { + "epoch": 0.94, + "learning_rate": 1.8853623671420296e-07, + "loss": 1.0583, + "step": 39085 + }, + { + "epoch": 0.94, + "learning_rate": 1.8838573258854743e-07, + "loss": 1.2415, + "step": 39086 + }, + { + "epoch": 0.94, + "learning_rate": 1.882352879876459e-07, + "loss": 0.9415, + "step": 39087 + }, + { + "epoch": 0.94, + "learning_rate": 1.8808490291241433e-07, + "loss": 1.1745, + "step": 39088 + }, + { + "epoch": 0.94, + "learning_rate": 1.8793457736376309e-07, + "loss": 1.2199, + "step": 39089 + }, + { + "epoch": 0.94, + "learning_rate": 1.8778431134260476e-07, + "loss": 0.7391, + "step": 39090 + }, + { + "epoch": 0.94, + "learning_rate": 1.876341048498509e-07, + "loss": 0.9724, + "step": 39091 + }, + { + "epoch": 0.94, + "learning_rate": 1.8748395788641404e-07, + "loss": 1.0047, + "step": 39092 + }, + { + "epoch": 0.94, + "learning_rate": 1.873338704532035e-07, + "loss": 1.2146, + "step": 39093 + }, + { + "epoch": 0.94, + "learning_rate": 1.8718384255113075e-07, + "loss": 0.9754, + "step": 39094 + }, + { + "epoch": 0.94, + "learning_rate": 1.8703387418110398e-07, + "loss": 1.1185, + "step": 39095 + }, + { + "epoch": 0.94, + "learning_rate": 1.8688396534403575e-07, + "loss": 1.1247, + "step": 39096 + }, + { + "epoch": 0.94, + "learning_rate": 1.8673411604083425e-07, + "loss": 1.18, + "step": 39097 + }, + { + "epoch": 0.94, + "learning_rate": 1.8658432627240764e-07, + "loss": 1.0821, + "step": 39098 + }, + { + "epoch": 0.94, + "learning_rate": 1.864345960396663e-07, + "loss": 1.0488, + "step": 39099 + }, + { + "epoch": 0.94, + "learning_rate": 1.862849253435184e-07, + "loss": 1.1008, + "step": 39100 + }, + { + "epoch": 0.94, + "learning_rate": 1.8613531418487097e-07, + "loss": 1.0024, + "step": 39101 + }, + { + "epoch": 0.94, + "learning_rate": 1.8598576256463218e-07, + "loss": 0.9367, + "step": 39102 + }, + { + "epoch": 0.94, + "learning_rate": 1.8583627048371023e-07, + "loss": 1.4806, + "step": 39103 + }, + { + "epoch": 0.94, + "learning_rate": 1.856868379430099e-07, + "loss": 1.2239, + "step": 39104 + }, + { + "epoch": 0.94, + "learning_rate": 1.8553746494344048e-07, + "loss": 1.1654, + "step": 39105 + }, + { + "epoch": 0.94, + "learning_rate": 1.8538815148590573e-07, + "loss": 1.0634, + "step": 39106 + }, + { + "epoch": 0.94, + "learning_rate": 1.8523889757131264e-07, + "loss": 1.1071, + "step": 39107 + }, + { + "epoch": 0.94, + "learning_rate": 1.850897032005672e-07, + "loss": 0.8478, + "step": 39108 + }, + { + "epoch": 0.94, + "learning_rate": 1.849405683745731e-07, + "loss": 0.9106, + "step": 39109 + }, + { + "epoch": 0.94, + "learning_rate": 1.8479149309423627e-07, + "loss": 1.1353, + "step": 39110 + }, + { + "epoch": 0.94, + "learning_rate": 1.846424773604616e-07, + "loss": 1.0601, + "step": 39111 + }, + { + "epoch": 0.94, + "learning_rate": 1.8449352117415052e-07, + "loss": 1.0793, + "step": 39112 + }, + { + "epoch": 0.94, + "learning_rate": 1.8434462453621017e-07, + "loss": 0.9205, + "step": 39113 + }, + { + "epoch": 0.94, + "learning_rate": 1.8419578744754085e-07, + "loss": 0.9098, + "step": 39114 + }, + { + "epoch": 0.94, + "learning_rate": 1.8404700990904745e-07, + "loss": 1.0847, + "step": 39115 + }, + { + "epoch": 0.94, + "learning_rate": 1.8389829192163256e-07, + "loss": 1.101, + "step": 39116 + }, + { + "epoch": 0.94, + "learning_rate": 1.8374963348619769e-07, + "loss": 0.9942, + "step": 39117 + }, + { + "epoch": 0.94, + "learning_rate": 1.8360103460364432e-07, + "loss": 0.7002, + "step": 39118 + }, + { + "epoch": 0.94, + "learning_rate": 1.8345249527487507e-07, + "loss": 1.1837, + "step": 39119 + }, + { + "epoch": 0.94, + "learning_rate": 1.8330401550078924e-07, + "loss": 1.0077, + "step": 39120 + }, + { + "epoch": 0.94, + "learning_rate": 1.8315559528229055e-07, + "loss": 0.8678, + "step": 39121 + }, + { + "epoch": 0.94, + "learning_rate": 1.8300723462027714e-07, + "loss": 1.0389, + "step": 39122 + }, + { + "epoch": 0.94, + "learning_rate": 1.8285893351564942e-07, + "loss": 0.905, + "step": 39123 + }, + { + "epoch": 0.94, + "learning_rate": 1.827106919693089e-07, + "loss": 1.028, + "step": 39124 + }, + { + "epoch": 0.94, + "learning_rate": 1.8256250998215152e-07, + "loss": 1.1718, + "step": 39125 + }, + { + "epoch": 0.94, + "learning_rate": 1.8241438755507878e-07, + "loss": 0.8394, + "step": 39126 + }, + { + "epoch": 0.94, + "learning_rate": 1.8226632468898886e-07, + "loss": 1.2613, + "step": 39127 + }, + { + "epoch": 0.94, + "learning_rate": 1.8211832138477993e-07, + "loss": 0.9849, + "step": 39128 + }, + { + "epoch": 0.94, + "learning_rate": 1.8197037764335014e-07, + "loss": 1.0034, + "step": 39129 + }, + { + "epoch": 0.94, + "learning_rate": 1.8182249346559655e-07, + "loss": 1.1511, + "step": 39130 + }, + { + "epoch": 0.94, + "learning_rate": 1.8167466885241624e-07, + "loss": 0.9644, + "step": 39131 + }, + { + "epoch": 0.94, + "learning_rate": 1.815269038047063e-07, + "loss": 1.184, + "step": 39132 + }, + { + "epoch": 0.94, + "learning_rate": 1.813791983233637e-07, + "loss": 0.6956, + "step": 39133 + }, + { + "epoch": 0.94, + "learning_rate": 1.8123155240928336e-07, + "loss": 1.338, + "step": 39134 + }, + { + "epoch": 0.94, + "learning_rate": 1.8108396606336232e-07, + "loss": 1.0913, + "step": 39135 + }, + { + "epoch": 0.94, + "learning_rate": 1.809364392864943e-07, + "loss": 0.8882, + "step": 39136 + }, + { + "epoch": 0.94, + "learning_rate": 1.8078897207957636e-07, + "loss": 0.9259, + "step": 39137 + }, + { + "epoch": 0.94, + "learning_rate": 1.8064156444350223e-07, + "loss": 1.0024, + "step": 39138 + }, + { + "epoch": 0.94, + "learning_rate": 1.8049421637916452e-07, + "loss": 1.0191, + "step": 39139 + }, + { + "epoch": 0.94, + "learning_rate": 1.8034692788746034e-07, + "loss": 1.1459, + "step": 39140 + }, + { + "epoch": 0.94, + "learning_rate": 1.8019969896928114e-07, + "loss": 1.1911, + "step": 39141 + }, + { + "epoch": 0.94, + "learning_rate": 1.8005252962551955e-07, + "loss": 1.1218, + "step": 39142 + }, + { + "epoch": 0.94, + "learning_rate": 1.7990541985707154e-07, + "loss": 0.9733, + "step": 39143 + }, + { + "epoch": 0.94, + "learning_rate": 1.797583696648264e-07, + "loss": 0.993, + "step": 39144 + }, + { + "epoch": 0.94, + "learning_rate": 1.7961137904967785e-07, + "loss": 1.1982, + "step": 39145 + }, + { + "epoch": 0.94, + "learning_rate": 1.7946444801251627e-07, + "loss": 1.1245, + "step": 39146 + }, + { + "epoch": 0.94, + "learning_rate": 1.793175765542332e-07, + "loss": 1.0158, + "step": 39147 + }, + { + "epoch": 0.94, + "learning_rate": 1.7917076467572236e-07, + "loss": 1.1589, + "step": 39148 + }, + { + "epoch": 0.94, + "learning_rate": 1.7902401237787193e-07, + "loss": 0.9641, + "step": 39149 + }, + { + "epoch": 0.94, + "learning_rate": 1.788773196615723e-07, + "loss": 0.8444, + "step": 39150 + }, + { + "epoch": 0.94, + "learning_rate": 1.7873068652771496e-07, + "loss": 0.9372, + "step": 39151 + }, + { + "epoch": 0.94, + "learning_rate": 1.7858411297718704e-07, + "loss": 0.9463, + "step": 39152 + }, + { + "epoch": 0.94, + "learning_rate": 1.7843759901087887e-07, + "loss": 1.0056, + "step": 39153 + }, + { + "epoch": 0.94, + "learning_rate": 1.78291144629682e-07, + "loss": 0.9572, + "step": 39154 + }, + { + "epoch": 0.94, + "learning_rate": 1.7814474983448016e-07, + "loss": 1.1643, + "step": 39155 + }, + { + "epoch": 0.94, + "learning_rate": 1.7799841462616595e-07, + "loss": 1.0681, + "step": 39156 + }, + { + "epoch": 0.94, + "learning_rate": 1.7785213900562316e-07, + "loss": 0.9511, + "step": 39157 + }, + { + "epoch": 0.94, + "learning_rate": 1.777059229737421e-07, + "loss": 0.9544, + "step": 39158 + }, + { + "epoch": 0.94, + "learning_rate": 1.7755976653140884e-07, + "loss": 1.1108, + "step": 39159 + }, + { + "epoch": 0.94, + "learning_rate": 1.7741366967950923e-07, + "loss": 1.1082, + "step": 39160 + }, + { + "epoch": 0.94, + "learning_rate": 1.7726763241893152e-07, + "loss": 1.2036, + "step": 39161 + }, + { + "epoch": 0.94, + "learning_rate": 1.7712165475056053e-07, + "loss": 1.1876, + "step": 39162 + }, + { + "epoch": 0.94, + "learning_rate": 1.7697573667528111e-07, + "loss": 1.2118, + "step": 39163 + }, + { + "epoch": 0.94, + "learning_rate": 1.7682987819398035e-07, + "loss": 1.0594, + "step": 39164 + }, + { + "epoch": 0.94, + "learning_rate": 1.7668407930754196e-07, + "loss": 1.2053, + "step": 39165 + }, + { + "epoch": 0.94, + "learning_rate": 1.7653834001685077e-07, + "loss": 1.1649, + "step": 39166 + }, + { + "epoch": 0.94, + "learning_rate": 1.7639266032278945e-07, + "loss": 0.9982, + "step": 39167 + }, + { + "epoch": 0.94, + "learning_rate": 1.7624704022624395e-07, + "loss": 1.2779, + "step": 39168 + }, + { + "epoch": 0.94, + "learning_rate": 1.76101479728098e-07, + "loss": 0.7904, + "step": 39169 + }, + { + "epoch": 0.94, + "learning_rate": 1.759559788292331e-07, + "loss": 1.1441, + "step": 39170 + }, + { + "epoch": 0.94, + "learning_rate": 1.7581053753053189e-07, + "loss": 0.8545, + "step": 39171 + }, + { + "epoch": 0.94, + "learning_rate": 1.7566515583287925e-07, + "loss": 1.0341, + "step": 39172 + }, + { + "epoch": 0.94, + "learning_rate": 1.7551983373715332e-07, + "loss": 1.024, + "step": 39173 + }, + { + "epoch": 0.94, + "learning_rate": 1.7537457124423896e-07, + "loss": 1.1591, + "step": 39174 + }, + { + "epoch": 0.94, + "learning_rate": 1.7522936835501546e-07, + "loss": 0.9643, + "step": 39175 + }, + { + "epoch": 0.94, + "learning_rate": 1.7508422507036438e-07, + "loss": 0.9347, + "step": 39176 + }, + { + "epoch": 0.94, + "learning_rate": 1.7493914139116607e-07, + "loss": 1.1788, + "step": 39177 + }, + { + "epoch": 0.94, + "learning_rate": 1.747941173183021e-07, + "loss": 1.0867, + "step": 39178 + }, + { + "epoch": 0.94, + "learning_rate": 1.746491528526495e-07, + "loss": 0.9321, + "step": 39179 + }, + { + "epoch": 0.94, + "learning_rate": 1.7450424799509093e-07, + "loss": 0.9761, + "step": 39180 + }, + { + "epoch": 0.94, + "learning_rate": 1.7435940274650343e-07, + "loss": 0.7861, + "step": 39181 + }, + { + "epoch": 0.94, + "learning_rate": 1.7421461710776632e-07, + "loss": 1.0314, + "step": 39182 + }, + { + "epoch": 0.94, + "learning_rate": 1.7406989107975892e-07, + "loss": 1.2234, + "step": 39183 + }, + { + "epoch": 0.94, + "learning_rate": 1.7392522466335714e-07, + "loss": 1.2884, + "step": 39184 + }, + { + "epoch": 0.94, + "learning_rate": 1.7378061785944034e-07, + "loss": 1.0626, + "step": 39185 + }, + { + "epoch": 0.94, + "learning_rate": 1.7363607066888444e-07, + "loss": 0.8879, + "step": 39186 + }, + { + "epoch": 0.94, + "learning_rate": 1.7349158309256765e-07, + "loss": 1.2196, + "step": 39187 + }, + { + "epoch": 0.94, + "learning_rate": 1.73347155131367e-07, + "loss": 1.108, + "step": 39188 + }, + { + "epoch": 0.94, + "learning_rate": 1.7320278678615633e-07, + "loss": 0.7727, + "step": 39189 + }, + { + "epoch": 0.94, + "learning_rate": 1.7305847805781373e-07, + "loss": 1.0668, + "step": 39190 + }, + { + "epoch": 0.94, + "learning_rate": 1.729142289472141e-07, + "loss": 1.253, + "step": 39191 + }, + { + "epoch": 0.94, + "learning_rate": 1.7277003945523117e-07, + "loss": 1.0638, + "step": 39192 + }, + { + "epoch": 0.94, + "learning_rate": 1.7262590958274317e-07, + "loss": 1.2205, + "step": 39193 + }, + { + "epoch": 0.94, + "learning_rate": 1.7248183933062046e-07, + "loss": 1.1909, + "step": 39194 + }, + { + "epoch": 0.94, + "learning_rate": 1.7233782869973903e-07, + "loss": 0.9608, + "step": 39195 + }, + { + "epoch": 0.94, + "learning_rate": 1.7219387769097262e-07, + "loss": 1.1964, + "step": 39196 + }, + { + "epoch": 0.94, + "learning_rate": 1.7204998630519387e-07, + "loss": 1.2305, + "step": 39197 + }, + { + "epoch": 0.94, + "learning_rate": 1.7190615454327654e-07, + "loss": 0.7032, + "step": 39198 + }, + { + "epoch": 0.94, + "learning_rate": 1.7176238240609323e-07, + "loss": 1.1985, + "step": 39199 + }, + { + "epoch": 0.94, + "learning_rate": 1.7161866989451548e-07, + "loss": 0.8959, + "step": 39200 + }, + { + "epoch": 0.94, + "learning_rate": 1.7147501700941594e-07, + "loss": 1.0375, + "step": 39201 + }, + { + "epoch": 0.94, + "learning_rate": 1.7133142375166502e-07, + "loss": 1.1832, + "step": 39202 + }, + { + "epoch": 0.94, + "learning_rate": 1.7118789012213422e-07, + "loss": 0.7713, + "step": 39203 + }, + { + "epoch": 0.94, + "learning_rate": 1.7104441612169508e-07, + "loss": 1.1412, + "step": 39204 + }, + { + "epoch": 0.94, + "learning_rate": 1.7090100175121804e-07, + "loss": 0.7902, + "step": 39205 + }, + { + "epoch": 0.94, + "learning_rate": 1.7075764701157128e-07, + "loss": 1.0373, + "step": 39206 + }, + { + "epoch": 0.94, + "learning_rate": 1.706143519036274e-07, + "loss": 1.0804, + "step": 39207 + }, + { + "epoch": 0.94, + "learning_rate": 1.7047111642825355e-07, + "loss": 1.046, + "step": 39208 + }, + { + "epoch": 0.94, + "learning_rate": 1.7032794058632008e-07, + "loss": 1.0995, + "step": 39209 + }, + { + "epoch": 0.94, + "learning_rate": 1.701848243786941e-07, + "loss": 1.0508, + "step": 39210 + }, + { + "epoch": 0.94, + "learning_rate": 1.7004176780624493e-07, + "loss": 1.0168, + "step": 39211 + }, + { + "epoch": 0.94, + "learning_rate": 1.6989877086984076e-07, + "loss": 1.0888, + "step": 39212 + }, + { + "epoch": 0.94, + "learning_rate": 1.6975583357034752e-07, + "loss": 0.9996, + "step": 39213 + }, + { + "epoch": 0.94, + "learning_rate": 1.6961295590863347e-07, + "loss": 0.9539, + "step": 39214 + }, + { + "epoch": 0.94, + "learning_rate": 1.6947013788556677e-07, + "loss": 1.0975, + "step": 39215 + }, + { + "epoch": 0.94, + "learning_rate": 1.6932737950201117e-07, + "loss": 1.0693, + "step": 39216 + }, + { + "epoch": 0.94, + "learning_rate": 1.6918468075883487e-07, + "loss": 1.1275, + "step": 39217 + }, + { + "epoch": 0.94, + "learning_rate": 1.6904204165690163e-07, + "loss": 1.1534, + "step": 39218 + }, + { + "epoch": 0.94, + "learning_rate": 1.6889946219707854e-07, + "loss": 1.0708, + "step": 39219 + }, + { + "epoch": 0.94, + "learning_rate": 1.6875694238023042e-07, + "loss": 1.0916, + "step": 39220 + }, + { + "epoch": 0.94, + "learning_rate": 1.686144822072211e-07, + "loss": 1.057, + "step": 39221 + }, + { + "epoch": 0.94, + "learning_rate": 1.6847208167891428e-07, + "loss": 0.9571, + "step": 39222 + }, + { + "epoch": 0.94, + "learning_rate": 1.6832974079617593e-07, + "loss": 1.174, + "step": 39223 + }, + { + "epoch": 0.94, + "learning_rate": 1.681874595598687e-07, + "loss": 1.0147, + "step": 39224 + }, + { + "epoch": 0.94, + "learning_rate": 1.6804523797085525e-07, + "loss": 1.1863, + "step": 39225 + }, + { + "epoch": 0.94, + "learning_rate": 1.6790307602999823e-07, + "loss": 1.1283, + "step": 39226 + }, + { + "epoch": 0.94, + "learning_rate": 1.6776097373816136e-07, + "loss": 1.137, + "step": 39227 + }, + { + "epoch": 0.94, + "learning_rate": 1.676189310962062e-07, + "loss": 1.2435, + "step": 39228 + }, + { + "epoch": 0.94, + "learning_rate": 1.6747694810499316e-07, + "loss": 1.0455, + "step": 39229 + }, + { + "epoch": 0.94, + "learning_rate": 1.673350247653849e-07, + "loss": 1.1273, + "step": 39230 + }, + { + "epoch": 0.94, + "learning_rate": 1.671931610782429e-07, + "loss": 0.7669, + "step": 39231 + }, + { + "epoch": 0.94, + "learning_rate": 1.6705135704442654e-07, + "loss": 1.1793, + "step": 39232 + }, + { + "epoch": 0.94, + "learning_rate": 1.6690961266479733e-07, + "loss": 0.9755, + "step": 39233 + }, + { + "epoch": 0.94, + "learning_rate": 1.6676792794021457e-07, + "loss": 1.0102, + "step": 39234 + }, + { + "epoch": 0.94, + "learning_rate": 1.666263028715376e-07, + "loss": 1.1363, + "step": 39235 + }, + { + "epoch": 0.94, + "learning_rate": 1.6648473745962678e-07, + "loss": 1.142, + "step": 39236 + }, + { + "epoch": 0.94, + "learning_rate": 1.6634323170533928e-07, + "loss": 0.7567, + "step": 39237 + }, + { + "epoch": 0.94, + "learning_rate": 1.6620178560953437e-07, + "loss": 1.0644, + "step": 39238 + }, + { + "epoch": 0.94, + "learning_rate": 1.6606039917307026e-07, + "loss": 1.2285, + "step": 39239 + }, + { + "epoch": 0.94, + "learning_rate": 1.6591907239680406e-07, + "loss": 1.1827, + "step": 39240 + }, + { + "epoch": 0.94, + "learning_rate": 1.6577780528159393e-07, + "loss": 1.1342, + "step": 39241 + }, + { + "epoch": 0.94, + "learning_rate": 1.656365978282981e-07, + "loss": 0.7935, + "step": 39242 + }, + { + "epoch": 0.94, + "learning_rate": 1.6549545003777034e-07, + "loss": 0.9612, + "step": 39243 + }, + { + "epoch": 0.94, + "learning_rate": 1.6535436191086884e-07, + "loss": 1.1361, + "step": 39244 + }, + { + "epoch": 0.94, + "learning_rate": 1.6521333344844958e-07, + "loss": 1.0318, + "step": 39245 + }, + { + "epoch": 0.94, + "learning_rate": 1.6507236465136633e-07, + "loss": 1.0074, + "step": 39246 + }, + { + "epoch": 0.94, + "learning_rate": 1.6493145552047728e-07, + "loss": 1.0679, + "step": 39247 + }, + { + "epoch": 0.94, + "learning_rate": 1.647906060566351e-07, + "loss": 1.1819, + "step": 39248 + }, + { + "epoch": 0.94, + "learning_rate": 1.6464981626069464e-07, + "loss": 1.0266, + "step": 39249 + }, + { + "epoch": 0.94, + "learning_rate": 1.645090861335119e-07, + "loss": 1.3162, + "step": 39250 + }, + { + "epoch": 0.94, + "learning_rate": 1.643684156759373e-07, + "loss": 1.1876, + "step": 39251 + }, + { + "epoch": 0.94, + "learning_rate": 1.6422780488882795e-07, + "loss": 1.2736, + "step": 39252 + }, + { + "epoch": 0.94, + "learning_rate": 1.6408725377303313e-07, + "loss": 1.0099, + "step": 39253 + }, + { + "epoch": 0.94, + "learning_rate": 1.6394676232940777e-07, + "loss": 1.1576, + "step": 39254 + }, + { + "epoch": 0.94, + "learning_rate": 1.6380633055880445e-07, + "loss": 0.8806, + "step": 39255 + }, + { + "epoch": 0.94, + "learning_rate": 1.6366595846207366e-07, + "loss": 0.8768, + "step": 39256 + }, + { + "epoch": 0.94, + "learning_rate": 1.6352564604006804e-07, + "loss": 0.9319, + "step": 39257 + }, + { + "epoch": 0.94, + "learning_rate": 1.633853932936391e-07, + "loss": 1.0696, + "step": 39258 + }, + { + "epoch": 0.94, + "learning_rate": 1.6324520022363733e-07, + "loss": 0.9371, + "step": 39259 + }, + { + "epoch": 0.94, + "learning_rate": 1.631050668309131e-07, + "loss": 0.72, + "step": 39260 + }, + { + "epoch": 0.94, + "learning_rate": 1.6296499311631687e-07, + "loss": 0.9316, + "step": 39261 + }, + { + "epoch": 0.94, + "learning_rate": 1.6282497908069684e-07, + "loss": 0.9659, + "step": 39262 + }, + { + "epoch": 0.94, + "learning_rate": 1.6268502472490566e-07, + "loss": 1.1266, + "step": 39263 + }, + { + "epoch": 0.94, + "learning_rate": 1.6254513004978934e-07, + "loss": 1.0501, + "step": 39264 + }, + { + "epoch": 0.94, + "learning_rate": 1.624052950561983e-07, + "loss": 0.9079, + "step": 39265 + }, + { + "epoch": 0.94, + "learning_rate": 1.6226551974498073e-07, + "loss": 1.195, + "step": 39266 + }, + { + "epoch": 0.94, + "learning_rate": 1.6212580411698264e-07, + "loss": 1.0767, + "step": 39267 + }, + { + "epoch": 0.94, + "learning_rate": 1.6198614817305557e-07, + "loss": 0.8017, + "step": 39268 + }, + { + "epoch": 0.94, + "learning_rate": 1.618465519140433e-07, + "loss": 1.0154, + "step": 39269 + }, + { + "epoch": 0.94, + "learning_rate": 1.6170701534079291e-07, + "loss": 1.082, + "step": 39270 + }, + { + "epoch": 0.94, + "learning_rate": 1.6156753845415373e-07, + "loss": 0.9714, + "step": 39271 + }, + { + "epoch": 0.94, + "learning_rate": 1.6142812125496954e-07, + "loss": 1.2037, + "step": 39272 + }, + { + "epoch": 0.94, + "learning_rate": 1.612887637440863e-07, + "loss": 1.0452, + "step": 39273 + }, + { + "epoch": 0.94, + "learning_rate": 1.6114946592235115e-07, + "loss": 1.1439, + "step": 39274 + }, + { + "epoch": 0.94, + "learning_rate": 1.6101022779060672e-07, + "loss": 0.7961, + "step": 39275 + }, + { + "epoch": 0.94, + "learning_rate": 1.6087104934969899e-07, + "loss": 1.0524, + "step": 39276 + }, + { + "epoch": 0.94, + "learning_rate": 1.6073193060047288e-07, + "loss": 0.9989, + "step": 39277 + }, + { + "epoch": 0.94, + "learning_rate": 1.6059287154377212e-07, + "loss": 1.2171, + "step": 39278 + }, + { + "epoch": 0.94, + "learning_rate": 1.604538721804394e-07, + "loss": 1.1638, + "step": 39279 + }, + { + "epoch": 0.94, + "learning_rate": 1.6031493251131846e-07, + "loss": 1.041, + "step": 39280 + }, + { + "epoch": 0.94, + "learning_rate": 1.60176052537252e-07, + "loss": 0.8842, + "step": 39281 + }, + { + "epoch": 0.94, + "learning_rate": 1.6003723225908484e-07, + "loss": 0.9831, + "step": 39282 + }, + { + "epoch": 0.94, + "learning_rate": 1.5989847167765527e-07, + "loss": 0.9773, + "step": 39283 + }, + { + "epoch": 0.94, + "learning_rate": 1.597597707938081e-07, + "loss": 0.9115, + "step": 39284 + }, + { + "epoch": 0.94, + "learning_rate": 1.5962112960838384e-07, + "loss": 1.2254, + "step": 39285 + }, + { + "epoch": 0.94, + "learning_rate": 1.5948254812222397e-07, + "loss": 1.2383, + "step": 39286 + }, + { + "epoch": 0.94, + "learning_rate": 1.5934402633616786e-07, + "loss": 0.8188, + "step": 39287 + }, + { + "epoch": 0.94, + "learning_rate": 1.5920556425105705e-07, + "loss": 1.1494, + "step": 39288 + }, + { + "epoch": 0.94, + "learning_rate": 1.5906716186773198e-07, + "loss": 0.9309, + "step": 39289 + }, + { + "epoch": 0.94, + "learning_rate": 1.5892881918703195e-07, + "loss": 1.3543, + "step": 39290 + }, + { + "epoch": 0.94, + "learning_rate": 1.5879053620979523e-07, + "loss": 1.018, + "step": 39291 + }, + { + "epoch": 0.94, + "learning_rate": 1.586523129368611e-07, + "loss": 1.2744, + "step": 39292 + }, + { + "epoch": 0.94, + "learning_rate": 1.5851414936907007e-07, + "loss": 0.744, + "step": 39293 + }, + { + "epoch": 0.94, + "learning_rate": 1.5837604550725803e-07, + "loss": 1.101, + "step": 39294 + }, + { + "epoch": 0.95, + "learning_rate": 1.5823800135226332e-07, + "loss": 1.016, + "step": 39295 + }, + { + "epoch": 0.95, + "learning_rate": 1.5810001690492405e-07, + "loss": 1.0555, + "step": 39296 + }, + { + "epoch": 0.95, + "learning_rate": 1.5796209216607628e-07, + "loss": 1.0551, + "step": 39297 + }, + { + "epoch": 0.95, + "learning_rate": 1.5782422713655931e-07, + "loss": 1.1747, + "step": 39298 + }, + { + "epoch": 0.95, + "learning_rate": 1.5768642181720584e-07, + "loss": 0.997, + "step": 39299 + }, + { + "epoch": 0.95, + "learning_rate": 1.5754867620885516e-07, + "loss": 1.169, + "step": 39300 + }, + { + "epoch": 0.95, + "learning_rate": 1.574109903123411e-07, + "loss": 1.13, + "step": 39301 + }, + { + "epoch": 0.95, + "learning_rate": 1.572733641284996e-07, + "loss": 1.0259, + "step": 39302 + }, + { + "epoch": 0.95, + "learning_rate": 1.5713579765816555e-07, + "loss": 1.241, + "step": 39303 + }, + { + "epoch": 0.95, + "learning_rate": 1.569982909021728e-07, + "loss": 1.1837, + "step": 39304 + }, + { + "epoch": 0.95, + "learning_rate": 1.5686084386135615e-07, + "loss": 1.0167, + "step": 39305 + }, + { + "epoch": 0.95, + "learning_rate": 1.567234565365505e-07, + "loss": 1.1606, + "step": 39306 + }, + { + "epoch": 0.95, + "learning_rate": 1.565861289285875e-07, + "loss": 1.0247, + "step": 39307 + }, + { + "epoch": 0.95, + "learning_rate": 1.5644886103830082e-07, + "loss": 0.9678, + "step": 39308 + }, + { + "epoch": 0.95, + "learning_rate": 1.5631165286652427e-07, + "loss": 1.2181, + "step": 39309 + }, + { + "epoch": 0.95, + "learning_rate": 1.5617450441408943e-07, + "loss": 1.1602, + "step": 39310 + }, + { + "epoch": 0.95, + "learning_rate": 1.5603741568182894e-07, + "loss": 1.0079, + "step": 39311 + }, + { + "epoch": 0.95, + "learning_rate": 1.5590038667057328e-07, + "loss": 1.1925, + "step": 39312 + }, + { + "epoch": 0.95, + "learning_rate": 1.5576341738115397e-07, + "loss": 0.9822, + "step": 39313 + }, + { + "epoch": 0.95, + "learning_rate": 1.5562650781440368e-07, + "loss": 1.1379, + "step": 39314 + }, + { + "epoch": 0.95, + "learning_rate": 1.5548965797115067e-07, + "loss": 1.246, + "step": 39315 + }, + { + "epoch": 0.95, + "learning_rate": 1.5535286785222758e-07, + "loss": 1.0415, + "step": 39316 + }, + { + "epoch": 0.95, + "learning_rate": 1.5521613745846264e-07, + "loss": 0.9463, + "step": 39317 + }, + { + "epoch": 0.95, + "learning_rate": 1.550794667906852e-07, + "loss": 1.1272, + "step": 39318 + }, + { + "epoch": 0.95, + "learning_rate": 1.5494285584972568e-07, + "loss": 1.0632, + "step": 39319 + }, + { + "epoch": 0.95, + "learning_rate": 1.5480630463641234e-07, + "loss": 1.083, + "step": 39320 + }, + { + "epoch": 0.95, + "learning_rate": 1.5466981315157226e-07, + "loss": 0.9739, + "step": 39321 + }, + { + "epoch": 0.95, + "learning_rate": 1.545333813960359e-07, + "loss": 1.011, + "step": 39322 + }, + { + "epoch": 0.95, + "learning_rate": 1.5439700937062928e-07, + "loss": 0.8285, + "step": 39323 + }, + { + "epoch": 0.95, + "learning_rate": 1.5426069707617952e-07, + "loss": 1.0224, + "step": 39324 + }, + { + "epoch": 0.95, + "learning_rate": 1.5412444451351482e-07, + "loss": 1.1105, + "step": 39325 + }, + { + "epoch": 0.95, + "learning_rate": 1.539882516834612e-07, + "loss": 0.8314, + "step": 39326 + }, + { + "epoch": 0.95, + "learning_rate": 1.5385211858684467e-07, + "loss": 1.0893, + "step": 39327 + }, + { + "epoch": 0.95, + "learning_rate": 1.5371604522449235e-07, + "loss": 1.0008, + "step": 39328 + }, + { + "epoch": 0.95, + "learning_rate": 1.53580031597228e-07, + "loss": 1.1176, + "step": 39329 + }, + { + "epoch": 0.95, + "learning_rate": 1.5344407770587765e-07, + "loss": 1.1399, + "step": 39330 + }, + { + "epoch": 0.95, + "learning_rate": 1.533081835512662e-07, + "loss": 1.1644, + "step": 39331 + }, + { + "epoch": 0.95, + "learning_rate": 1.531723491342174e-07, + "loss": 0.9923, + "step": 39332 + }, + { + "epoch": 0.95, + "learning_rate": 1.5303657445555508e-07, + "loss": 0.8091, + "step": 39333 + }, + { + "epoch": 0.95, + "learning_rate": 1.5290085951610521e-07, + "loss": 0.8736, + "step": 39334 + }, + { + "epoch": 0.95, + "learning_rate": 1.527652043166894e-07, + "loss": 0.954, + "step": 39335 + }, + { + "epoch": 0.95, + "learning_rate": 1.5262960885813026e-07, + "loss": 1.2091, + "step": 39336 + }, + { + "epoch": 0.95, + "learning_rate": 1.5249407314125165e-07, + "loss": 0.9351, + "step": 39337 + }, + { + "epoch": 0.95, + "learning_rate": 1.5235859716687506e-07, + "loss": 1.1225, + "step": 39338 + }, + { + "epoch": 0.95, + "learning_rate": 1.5222318093582322e-07, + "loss": 0.7441, + "step": 39339 + }, + { + "epoch": 0.95, + "learning_rate": 1.5208782444891656e-07, + "loss": 1.1795, + "step": 39340 + }, + { + "epoch": 0.95, + "learning_rate": 1.5195252770697665e-07, + "loss": 1.0935, + "step": 39341 + }, + { + "epoch": 0.95, + "learning_rate": 1.5181729071082397e-07, + "loss": 1.0615, + "step": 39342 + }, + { + "epoch": 0.95, + "learning_rate": 1.5168211346128004e-07, + "loss": 0.8872, + "step": 39343 + }, + { + "epoch": 0.95, + "learning_rate": 1.5154699595916312e-07, + "loss": 1.2537, + "step": 39344 + }, + { + "epoch": 0.95, + "learning_rate": 1.5141193820529476e-07, + "loss": 0.8358, + "step": 39345 + }, + { + "epoch": 0.95, + "learning_rate": 1.5127694020049432e-07, + "loss": 0.7792, + "step": 39346 + }, + { + "epoch": 0.95, + "learning_rate": 1.5114200194557893e-07, + "loss": 0.8712, + "step": 39347 + }, + { + "epoch": 0.95, + "learning_rate": 1.5100712344136904e-07, + "loss": 1.0385, + "step": 39348 + }, + { + "epoch": 0.95, + "learning_rate": 1.5087230468868175e-07, + "loss": 1.0729, + "step": 39349 + }, + { + "epoch": 0.95, + "learning_rate": 1.5073754568833643e-07, + "loss": 0.7136, + "step": 39350 + }, + { + "epoch": 0.95, + "learning_rate": 1.506028464411491e-07, + "loss": 1.1087, + "step": 39351 + }, + { + "epoch": 0.95, + "learning_rate": 1.5046820694793794e-07, + "loss": 1.069, + "step": 39352 + }, + { + "epoch": 0.95, + "learning_rate": 1.5033362720951905e-07, + "loss": 1.3381, + "step": 39353 + }, + { + "epoch": 0.95, + "learning_rate": 1.501991072267095e-07, + "loss": 1.0543, + "step": 39354 + }, + { + "epoch": 0.95, + "learning_rate": 1.500646470003253e-07, + "loss": 0.8087, + "step": 39355 + }, + { + "epoch": 0.95, + "learning_rate": 1.499302465311814e-07, + "loss": 1.1605, + "step": 39356 + }, + { + "epoch": 0.95, + "learning_rate": 1.4979590582009484e-07, + "loss": 0.8142, + "step": 39357 + }, + { + "epoch": 0.95, + "learning_rate": 1.496616248678784e-07, + "loss": 1.1958, + "step": 39358 + }, + { + "epoch": 0.95, + "learning_rate": 1.4952740367534913e-07, + "loss": 1.0626, + "step": 39359 + }, + { + "epoch": 0.95, + "learning_rate": 1.4939324224331974e-07, + "loss": 1.1169, + "step": 39360 + }, + { + "epoch": 0.95, + "learning_rate": 1.4925914057260404e-07, + "loss": 1.0604, + "step": 39361 + }, + { + "epoch": 0.95, + "learning_rate": 1.491250986640169e-07, + "loss": 0.9931, + "step": 39362 + }, + { + "epoch": 0.95, + "learning_rate": 1.4899111651836996e-07, + "loss": 1.1275, + "step": 39363 + }, + { + "epoch": 0.95, + "learning_rate": 1.4885719413647692e-07, + "loss": 0.8811, + "step": 39364 + }, + { + "epoch": 0.95, + "learning_rate": 1.4872333151915053e-07, + "loss": 1.0277, + "step": 39365 + }, + { + "epoch": 0.95, + "learning_rate": 1.4858952866720234e-07, + "loss": 0.7932, + "step": 39366 + }, + { + "epoch": 0.95, + "learning_rate": 1.4845578558144503e-07, + "loss": 0.8451, + "step": 39367 + }, + { + "epoch": 0.95, + "learning_rate": 1.4832210226268794e-07, + "loss": 1.2452, + "step": 39368 + }, + { + "epoch": 0.95, + "learning_rate": 1.481884787117438e-07, + "loss": 1.0342, + "step": 39369 + }, + { + "epoch": 0.95, + "learning_rate": 1.4805491492942415e-07, + "loss": 0.8406, + "step": 39370 + }, + { + "epoch": 0.95, + "learning_rate": 1.479214109165361e-07, + "loss": 0.9532, + "step": 39371 + }, + { + "epoch": 0.95, + "learning_rate": 1.4778796667389238e-07, + "loss": 0.9122, + "step": 39372 + }, + { + "epoch": 0.95, + "learning_rate": 1.4765458220230233e-07, + "loss": 1.2071, + "step": 39373 + }, + { + "epoch": 0.95, + "learning_rate": 1.4752125750257307e-07, + "loss": 0.8559, + "step": 39374 + }, + { + "epoch": 0.95, + "learning_rate": 1.4738799257551616e-07, + "loss": 0.9755, + "step": 39375 + }, + { + "epoch": 0.95, + "learning_rate": 1.4725478742193766e-07, + "loss": 0.9126, + "step": 39376 + }, + { + "epoch": 0.95, + "learning_rate": 1.4712164204264801e-07, + "loss": 1.1444, + "step": 39377 + }, + { + "epoch": 0.95, + "learning_rate": 1.4698855643845324e-07, + "loss": 0.9775, + "step": 39378 + }, + { + "epoch": 0.95, + "learning_rate": 1.4685553061016155e-07, + "loss": 0.9621, + "step": 39379 + }, + { + "epoch": 0.95, + "learning_rate": 1.4672256455857902e-07, + "loss": 0.8619, + "step": 39380 + }, + { + "epoch": 0.95, + "learning_rate": 1.465896582845139e-07, + "loss": 1.1454, + "step": 39381 + }, + { + "epoch": 0.95, + "learning_rate": 1.4645681178877103e-07, + "loss": 1.2061, + "step": 39382 + }, + { + "epoch": 0.95, + "learning_rate": 1.463240250721587e-07, + "loss": 1.0813, + "step": 39383 + }, + { + "epoch": 0.95, + "learning_rate": 1.461912981354785e-07, + "loss": 0.9764, + "step": 39384 + }, + { + "epoch": 0.95, + "learning_rate": 1.4605863097953866e-07, + "loss": 0.8974, + "step": 39385 + }, + { + "epoch": 0.95, + "learning_rate": 1.4592602360514406e-07, + "loss": 1.0383, + "step": 39386 + }, + { + "epoch": 0.95, + "learning_rate": 1.4579347601309745e-07, + "loss": 1.0521, + "step": 39387 + }, + { + "epoch": 0.95, + "learning_rate": 1.4566098820420483e-07, + "loss": 1.1883, + "step": 39388 + }, + { + "epoch": 0.95, + "learning_rate": 1.4552856017926886e-07, + "loss": 1.1858, + "step": 39389 + }, + { + "epoch": 0.95, + "learning_rate": 1.4539619193909338e-07, + "loss": 1.0355, + "step": 39390 + }, + { + "epoch": 0.95, + "learning_rate": 1.4526388348448106e-07, + "loss": 1.229, + "step": 39391 + }, + { + "epoch": 0.95, + "learning_rate": 1.451316348162346e-07, + "loss": 0.875, + "step": 39392 + }, + { + "epoch": 0.95, + "learning_rate": 1.4499944593515557e-07, + "loss": 0.9462, + "step": 39393 + }, + { + "epoch": 0.95, + "learning_rate": 1.448673168420489e-07, + "loss": 0.8819, + "step": 39394 + }, + { + "epoch": 0.95, + "learning_rate": 1.4473524753771285e-07, + "loss": 1.1024, + "step": 39395 + }, + { + "epoch": 0.95, + "learning_rate": 1.4460323802295007e-07, + "loss": 0.9492, + "step": 39396 + }, + { + "epoch": 0.95, + "learning_rate": 1.4447128829856216e-07, + "loss": 1.1173, + "step": 39397 + }, + { + "epoch": 0.95, + "learning_rate": 1.4433939836534738e-07, + "loss": 1.146, + "step": 39398 + }, + { + "epoch": 0.95, + "learning_rate": 1.442075682241073e-07, + "loss": 1.0658, + "step": 39399 + }, + { + "epoch": 0.95, + "learning_rate": 1.4407579787564353e-07, + "loss": 1.0941, + "step": 39400 + }, + { + "epoch": 0.95, + "learning_rate": 1.4394408732075203e-07, + "loss": 0.6883, + "step": 39401 + }, + { + "epoch": 0.95, + "learning_rate": 1.4381243656023446e-07, + "loss": 1.034, + "step": 39402 + }, + { + "epoch": 0.95, + "learning_rate": 1.4368084559488794e-07, + "loss": 0.9392, + "step": 39403 + }, + { + "epoch": 0.95, + "learning_rate": 1.4354931442551067e-07, + "loss": 0.9292, + "step": 39404 + }, + { + "epoch": 0.95, + "learning_rate": 1.4341784305290317e-07, + "loss": 1.0711, + "step": 39405 + }, + { + "epoch": 0.95, + "learning_rate": 1.4328643147785925e-07, + "loss": 1.0134, + "step": 39406 + }, + { + "epoch": 0.95, + "learning_rate": 1.4315507970117825e-07, + "loss": 1.0602, + "step": 39407 + }, + { + "epoch": 0.95, + "learning_rate": 1.4302378772365844e-07, + "loss": 0.8176, + "step": 39408 + }, + { + "epoch": 0.95, + "learning_rate": 1.428925555460936e-07, + "loss": 0.873, + "step": 39409 + }, + { + "epoch": 0.95, + "learning_rate": 1.42761383169282e-07, + "loss": 1.0282, + "step": 39410 + }, + { + "epoch": 0.95, + "learning_rate": 1.4263027059401857e-07, + "loss": 0.9187, + "step": 39411 + }, + { + "epoch": 0.95, + "learning_rate": 1.424992178210971e-07, + "loss": 1.084, + "step": 39412 + }, + { + "epoch": 0.95, + "learning_rate": 1.4236822485131585e-07, + "loss": 1.0306, + "step": 39413 + }, + { + "epoch": 0.95, + "learning_rate": 1.422372916854675e-07, + "loss": 0.9217, + "step": 39414 + }, + { + "epoch": 0.95, + "learning_rate": 1.421064183243459e-07, + "loss": 1.1424, + "step": 39415 + }, + { + "epoch": 0.95, + "learning_rate": 1.4197560476874706e-07, + "loss": 1.0668, + "step": 39416 + }, + { + "epoch": 0.95, + "learning_rate": 1.4184485101946365e-07, + "loss": 0.9712, + "step": 39417 + }, + { + "epoch": 0.95, + "learning_rate": 1.4171415707728843e-07, + "loss": 0.7743, + "step": 39418 + }, + { + "epoch": 0.95, + "learning_rate": 1.4158352294301403e-07, + "loss": 1.0311, + "step": 39419 + }, + { + "epoch": 0.95, + "learning_rate": 1.414529486174332e-07, + "loss": 1.0461, + "step": 39420 + }, + { + "epoch": 0.95, + "learning_rate": 1.4132243410133972e-07, + "loss": 1.2333, + "step": 39421 + }, + { + "epoch": 0.95, + "learning_rate": 1.411919793955241e-07, + "loss": 0.9791, + "step": 39422 + }, + { + "epoch": 0.95, + "learning_rate": 1.4106158450077678e-07, + "loss": 1.4631, + "step": 39423 + }, + { + "epoch": 0.95, + "learning_rate": 1.409312494178905e-07, + "loss": 1.0554, + "step": 39424 + }, + { + "epoch": 0.95, + "learning_rate": 1.408009741476557e-07, + "loss": 0.8203, + "step": 39425 + }, + { + "epoch": 0.95, + "learning_rate": 1.4067075869086178e-07, + "loss": 0.8148, + "step": 39426 + }, + { + "epoch": 0.95, + "learning_rate": 1.405406030482992e-07, + "loss": 0.9895, + "step": 39427 + }, + { + "epoch": 0.95, + "learning_rate": 1.4041050722075734e-07, + "loss": 1.0069, + "step": 39428 + }, + { + "epoch": 0.95, + "learning_rate": 1.4028047120902777e-07, + "loss": 1.0141, + "step": 39429 + }, + { + "epoch": 0.95, + "learning_rate": 1.4015049501389543e-07, + "loss": 1.0402, + "step": 39430 + }, + { + "epoch": 0.95, + "learning_rate": 1.4002057863615082e-07, + "loss": 0.8765, + "step": 39431 + }, + { + "epoch": 0.95, + "learning_rate": 1.3989072207658328e-07, + "loss": 1.0866, + "step": 39432 + }, + { + "epoch": 0.95, + "learning_rate": 1.3976092533597884e-07, + "loss": 1.1904, + "step": 39433 + }, + { + "epoch": 0.95, + "learning_rate": 1.396311884151269e-07, + "loss": 1.0767, + "step": 39434 + }, + { + "epoch": 0.95, + "learning_rate": 1.3950151131481126e-07, + "loss": 0.8902, + "step": 39435 + }, + { + "epoch": 0.95, + "learning_rate": 1.3937189403582129e-07, + "loss": 1.071, + "step": 39436 + }, + { + "epoch": 0.95, + "learning_rate": 1.3924233657894416e-07, + "loss": 0.8939, + "step": 39437 + }, + { + "epoch": 0.95, + "learning_rate": 1.3911283894496253e-07, + "loss": 0.9737, + "step": 39438 + }, + { + "epoch": 0.95, + "learning_rate": 1.3898340113466357e-07, + "loss": 1.073, + "step": 39439 + }, + { + "epoch": 0.95, + "learning_rate": 1.3885402314883445e-07, + "loss": 1.202, + "step": 39440 + }, + { + "epoch": 0.95, + "learning_rate": 1.3872470498825786e-07, + "loss": 0.8817, + "step": 39441 + }, + { + "epoch": 0.95, + "learning_rate": 1.385954466537187e-07, + "loss": 1.0094, + "step": 39442 + }, + { + "epoch": 0.95, + "learning_rate": 1.3846624814600195e-07, + "loss": 0.5957, + "step": 39443 + }, + { + "epoch": 0.95, + "learning_rate": 1.383371094658903e-07, + "loss": 1.1501, + "step": 39444 + }, + { + "epoch": 0.95, + "learning_rate": 1.3820803061416865e-07, + "loss": 1.0955, + "step": 39445 + }, + { + "epoch": 0.95, + "learning_rate": 1.3807901159161862e-07, + "loss": 0.9108, + "step": 39446 + }, + { + "epoch": 0.95, + "learning_rate": 1.3795005239902292e-07, + "loss": 0.8853, + "step": 39447 + }, + { + "epoch": 0.95, + "learning_rate": 1.3782115303716537e-07, + "loss": 1.0121, + "step": 39448 + }, + { + "epoch": 0.95, + "learning_rate": 1.3769231350682755e-07, + "loss": 1.0809, + "step": 39449 + }, + { + "epoch": 0.95, + "learning_rate": 1.3756353380878994e-07, + "loss": 0.9854, + "step": 39450 + }, + { + "epoch": 0.95, + "learning_rate": 1.3743481394383528e-07, + "loss": 1.1151, + "step": 39451 + }, + { + "epoch": 0.95, + "learning_rate": 1.3730615391274293e-07, + "loss": 1.094, + "step": 39452 + }, + { + "epoch": 0.95, + "learning_rate": 1.3717755371629448e-07, + "loss": 1.1925, + "step": 39453 + }, + { + "epoch": 0.95, + "learning_rate": 1.3704901335527044e-07, + "loss": 0.8991, + "step": 39454 + }, + { + "epoch": 0.95, + "learning_rate": 1.3692053283044903e-07, + "loss": 1.0226, + "step": 39455 + }, + { + "epoch": 0.95, + "learning_rate": 1.367921121426119e-07, + "loss": 1.2098, + "step": 39456 + }, + { + "epoch": 0.95, + "learning_rate": 1.366637512925373e-07, + "loss": 1.0378, + "step": 39457 + }, + { + "epoch": 0.95, + "learning_rate": 1.3653545028100235e-07, + "loss": 1.1249, + "step": 39458 + }, + { + "epoch": 0.95, + "learning_rate": 1.3640720910878868e-07, + "loss": 0.871, + "step": 39459 + }, + { + "epoch": 0.95, + "learning_rate": 1.3627902777667123e-07, + "loss": 1.0073, + "step": 39460 + }, + { + "epoch": 0.95, + "learning_rate": 1.3615090628542937e-07, + "loss": 1.0675, + "step": 39461 + }, + { + "epoch": 0.95, + "learning_rate": 1.3602284463583914e-07, + "loss": 1.2315, + "step": 39462 + }, + { + "epoch": 0.95, + "learning_rate": 1.358948428286777e-07, + "loss": 1.2974, + "step": 39463 + }, + { + "epoch": 0.95, + "learning_rate": 1.3576690086472332e-07, + "loss": 0.8674, + "step": 39464 + }, + { + "epoch": 0.95, + "learning_rate": 1.3563901874475094e-07, + "loss": 1.1529, + "step": 39465 + }, + { + "epoch": 0.95, + "learning_rate": 1.3551119646953548e-07, + "loss": 1.3023, + "step": 39466 + }, + { + "epoch": 0.95, + "learning_rate": 1.353834340398541e-07, + "loss": 0.7515, + "step": 39467 + }, + { + "epoch": 0.95, + "learning_rate": 1.3525573145648062e-07, + "loss": 1.0599, + "step": 39468 + }, + { + "epoch": 0.95, + "learning_rate": 1.3512808872019113e-07, + "loss": 0.9484, + "step": 39469 + }, + { + "epoch": 0.95, + "learning_rate": 1.3500050583175828e-07, + "loss": 0.967, + "step": 39470 + }, + { + "epoch": 0.95, + "learning_rate": 1.3487298279195593e-07, + "loss": 0.8585, + "step": 39471 + }, + { + "epoch": 0.95, + "learning_rate": 1.3474551960156125e-07, + "loss": 0.7288, + "step": 39472 + }, + { + "epoch": 0.95, + "learning_rate": 1.346181162613436e-07, + "loss": 1.0811, + "step": 39473 + }, + { + "epoch": 0.95, + "learning_rate": 1.3449077277207678e-07, + "loss": 1.2013, + "step": 39474 + }, + { + "epoch": 0.95, + "learning_rate": 1.3436348913453578e-07, + "loss": 1.1956, + "step": 39475 + }, + { + "epoch": 0.95, + "learning_rate": 1.3423626534948886e-07, + "loss": 1.0515, + "step": 39476 + }, + { + "epoch": 0.95, + "learning_rate": 1.3410910141771204e-07, + "loss": 1.0805, + "step": 39477 + }, + { + "epoch": 0.95, + "learning_rate": 1.3398199733997364e-07, + "loss": 1.3589, + "step": 39478 + }, + { + "epoch": 0.95, + "learning_rate": 1.3385495311704523e-07, + "loss": 1.0088, + "step": 39479 + }, + { + "epoch": 0.95, + "learning_rate": 1.337279687496995e-07, + "loss": 1.0951, + "step": 39480 + }, + { + "epoch": 0.95, + "learning_rate": 1.336010442387048e-07, + "loss": 1.0467, + "step": 39481 + }, + { + "epoch": 0.95, + "learning_rate": 1.3347417958483155e-07, + "loss": 0.8296, + "step": 39482 + }, + { + "epoch": 0.95, + "learning_rate": 1.333473747888503e-07, + "loss": 1.1744, + "step": 39483 + }, + { + "epoch": 0.95, + "learning_rate": 1.332206298515293e-07, + "loss": 0.8887, + "step": 39484 + }, + { + "epoch": 0.95, + "learning_rate": 1.330939447736379e-07, + "loss": 1.0071, + "step": 39485 + }, + { + "epoch": 0.95, + "learning_rate": 1.3296731955594443e-07, + "loss": 1.1183, + "step": 39486 + }, + { + "epoch": 0.95, + "learning_rate": 1.3284075419921826e-07, + "loss": 0.9557, + "step": 39487 + }, + { + "epoch": 0.95, + "learning_rate": 1.3271424870422545e-07, + "loss": 1.0249, + "step": 39488 + }, + { + "epoch": 0.95, + "learning_rate": 1.3258780307173535e-07, + "loss": 1.0957, + "step": 39489 + }, + { + "epoch": 0.95, + "learning_rate": 1.3246141730251293e-07, + "loss": 1.0498, + "step": 39490 + }, + { + "epoch": 0.95, + "learning_rate": 1.3233509139732758e-07, + "loss": 1.1552, + "step": 39491 + }, + { + "epoch": 0.95, + "learning_rate": 1.3220882535694312e-07, + "loss": 1.0513, + "step": 39492 + }, + { + "epoch": 0.95, + "learning_rate": 1.3208261918212672e-07, + "loss": 0.9735, + "step": 39493 + }, + { + "epoch": 0.95, + "learning_rate": 1.3195647287364443e-07, + "loss": 1.0901, + "step": 39494 + }, + { + "epoch": 0.95, + "learning_rate": 1.3183038643226008e-07, + "loss": 1.2345, + "step": 39495 + }, + { + "epoch": 0.95, + "learning_rate": 1.3170435985874085e-07, + "loss": 1.1666, + "step": 39496 + }, + { + "epoch": 0.95, + "learning_rate": 1.3157839315384946e-07, + "loss": 1.156, + "step": 39497 + }, + { + "epoch": 0.95, + "learning_rate": 1.3145248631835084e-07, + "loss": 0.9412, + "step": 39498 + }, + { + "epoch": 0.95, + "learning_rate": 1.3132663935300883e-07, + "loss": 0.7584, + "step": 39499 + }, + { + "epoch": 0.95, + "learning_rate": 1.3120085225858614e-07, + "loss": 1.1046, + "step": 39500 + }, + { + "epoch": 0.95, + "learning_rate": 1.3107512503584774e-07, + "loss": 1.1671, + "step": 39501 + }, + { + "epoch": 0.95, + "learning_rate": 1.3094945768555523e-07, + "loss": 1.0266, + "step": 39502 + }, + { + "epoch": 0.95, + "learning_rate": 1.308238502084702e-07, + "loss": 0.9911, + "step": 39503 + }, + { + "epoch": 0.95, + "learning_rate": 1.3069830260535653e-07, + "loss": 0.9815, + "step": 39504 + }, + { + "epoch": 0.95, + "learning_rate": 1.305728148769736e-07, + "loss": 1.1287, + "step": 39505 + }, + { + "epoch": 0.95, + "learning_rate": 1.3044738702408523e-07, + "loss": 1.0879, + "step": 39506 + }, + { + "epoch": 0.95, + "learning_rate": 1.303220190474508e-07, + "loss": 0.7831, + "step": 39507 + }, + { + "epoch": 0.95, + "learning_rate": 1.301967109478297e-07, + "loss": 1.1137, + "step": 39508 + }, + { + "epoch": 0.95, + "learning_rate": 1.300714627259858e-07, + "loss": 1.1626, + "step": 39509 + }, + { + "epoch": 0.95, + "learning_rate": 1.2994627438267514e-07, + "loss": 0.9287, + "step": 39510 + }, + { + "epoch": 0.95, + "learning_rate": 1.2982114591865936e-07, + "loss": 1.0664, + "step": 39511 + }, + { + "epoch": 0.95, + "learning_rate": 1.296960773346978e-07, + "loss": 1.2752, + "step": 39512 + }, + { + "epoch": 0.95, + "learning_rate": 1.2957106863154766e-07, + "loss": 1.205, + "step": 39513 + }, + { + "epoch": 0.95, + "learning_rate": 1.2944611980996947e-07, + "loss": 0.9798, + "step": 39514 + }, + { + "epoch": 0.95, + "learning_rate": 1.2932123087071812e-07, + "loss": 1.1748, + "step": 39515 + }, + { + "epoch": 0.95, + "learning_rate": 1.2919640181455417e-07, + "loss": 1.1653, + "step": 39516 + }, + { + "epoch": 0.95, + "learning_rate": 1.2907163264223477e-07, + "loss": 0.9531, + "step": 39517 + }, + { + "epoch": 0.95, + "learning_rate": 1.2894692335451376e-07, + "loss": 0.8844, + "step": 39518 + }, + { + "epoch": 0.95, + "learning_rate": 1.2882227395215164e-07, + "loss": 1.0948, + "step": 39519 + }, + { + "epoch": 0.95, + "learning_rate": 1.2869768443590224e-07, + "loss": 1.04, + "step": 39520 + }, + { + "epoch": 0.95, + "learning_rate": 1.2857315480652277e-07, + "loss": 1.1197, + "step": 39521 + }, + { + "epoch": 0.95, + "learning_rate": 1.2844868506476815e-07, + "loss": 1.0161, + "step": 39522 + }, + { + "epoch": 0.95, + "learning_rate": 1.2832427521139224e-07, + "loss": 1.1125, + "step": 39523 + }, + { + "epoch": 0.95, + "learning_rate": 1.2819992524715107e-07, + "loss": 1.2014, + "step": 39524 + }, + { + "epoch": 0.95, + "learning_rate": 1.2807563517280074e-07, + "loss": 0.9546, + "step": 39525 + }, + { + "epoch": 0.95, + "learning_rate": 1.2795140498909174e-07, + "loss": 0.8332, + "step": 39526 + }, + { + "epoch": 0.95, + "learning_rate": 1.2782723469678015e-07, + "loss": 0.931, + "step": 39527 + }, + { + "epoch": 0.95, + "learning_rate": 1.277031242966187e-07, + "loss": 1.1768, + "step": 39528 + }, + { + "epoch": 0.95, + "learning_rate": 1.2757907378936007e-07, + "loss": 1.2664, + "step": 39529 + }, + { + "epoch": 0.95, + "learning_rate": 1.2745508317575704e-07, + "loss": 0.9584, + "step": 39530 + }, + { + "epoch": 0.95, + "learning_rate": 1.2733115245656236e-07, + "loss": 0.9563, + "step": 39531 + }, + { + "epoch": 0.95, + "learning_rate": 1.2720728163252648e-07, + "loss": 0.6672, + "step": 39532 + }, + { + "epoch": 0.95, + "learning_rate": 1.270834707044033e-07, + "loss": 1.0438, + "step": 39533 + }, + { + "epoch": 0.95, + "learning_rate": 1.2695971967294108e-07, + "loss": 1.0232, + "step": 39534 + }, + { + "epoch": 0.95, + "learning_rate": 1.2683602853889142e-07, + "loss": 1.1519, + "step": 39535 + }, + { + "epoch": 0.95, + "learning_rate": 1.267123973030071e-07, + "loss": 1.142, + "step": 39536 + }, + { + "epoch": 0.95, + "learning_rate": 1.2658882596603528e-07, + "loss": 1.0739, + "step": 39537 + }, + { + "epoch": 0.95, + "learning_rate": 1.2646531452872645e-07, + "loss": 1.2347, + "step": 39538 + }, + { + "epoch": 0.95, + "learning_rate": 1.2634186299183115e-07, + "loss": 0.8478, + "step": 39539 + }, + { + "epoch": 0.95, + "learning_rate": 1.2621847135609656e-07, + "loss": 1.055, + "step": 39540 + }, + { + "epoch": 0.95, + "learning_rate": 1.2609513962227203e-07, + "loss": 1.1822, + "step": 39541 + }, + { + "epoch": 0.95, + "learning_rate": 1.2597186779110593e-07, + "loss": 1.0923, + "step": 39542 + }, + { + "epoch": 0.95, + "learning_rate": 1.2584865586334648e-07, + "loss": 0.9733, + "step": 39543 + }, + { + "epoch": 0.95, + "learning_rate": 1.257255038397398e-07, + "loss": 0.8577, + "step": 39544 + }, + { + "epoch": 0.95, + "learning_rate": 1.2560241172103415e-07, + "loss": 1.0398, + "step": 39545 + }, + { + "epoch": 0.95, + "learning_rate": 1.2547937950797673e-07, + "loss": 1.2171, + "step": 39546 + }, + { + "epoch": 0.95, + "learning_rate": 1.253564072013136e-07, + "loss": 1.0424, + "step": 39547 + }, + { + "epoch": 0.95, + "learning_rate": 1.252334948017886e-07, + "loss": 1.1349, + "step": 39548 + }, + { + "epoch": 0.95, + "learning_rate": 1.2511064231015113e-07, + "loss": 1.1912, + "step": 39549 + }, + { + "epoch": 0.95, + "learning_rate": 1.2498784972714395e-07, + "loss": 1.0691, + "step": 39550 + }, + { + "epoch": 0.95, + "learning_rate": 1.248651170535131e-07, + "loss": 1.0383, + "step": 39551 + }, + { + "epoch": 0.95, + "learning_rate": 1.247424442900025e-07, + "loss": 1.0487, + "step": 39552 + }, + { + "epoch": 0.95, + "learning_rate": 1.24619831437357e-07, + "loss": 1.1232, + "step": 39553 + }, + { + "epoch": 0.95, + "learning_rate": 1.2449727849631944e-07, + "loss": 0.9871, + "step": 39554 + }, + { + "epoch": 0.95, + "learning_rate": 1.243747854676347e-07, + "loss": 1.0232, + "step": 39555 + }, + { + "epoch": 0.95, + "learning_rate": 1.2425235235204447e-07, + "loss": 1.166, + "step": 39556 + }, + { + "epoch": 0.95, + "learning_rate": 1.2412997915029368e-07, + "loss": 0.7533, + "step": 39557 + }, + { + "epoch": 0.95, + "learning_rate": 1.2400766586312174e-07, + "loss": 1.0519, + "step": 39558 + }, + { + "epoch": 0.95, + "learning_rate": 1.2388541249127362e-07, + "loss": 0.9351, + "step": 39559 + }, + { + "epoch": 0.95, + "learning_rate": 1.2376321903548872e-07, + "loss": 1.0292, + "step": 39560 + }, + { + "epoch": 0.95, + "learning_rate": 1.2364108549650978e-07, + "loss": 0.8385, + "step": 39561 + }, + { + "epoch": 0.95, + "learning_rate": 1.235190118750773e-07, + "loss": 1.0059, + "step": 39562 + }, + { + "epoch": 0.95, + "learning_rate": 1.2339699817193186e-07, + "loss": 1.1977, + "step": 39563 + }, + { + "epoch": 0.95, + "learning_rate": 1.232750443878139e-07, + "loss": 1.0374, + "step": 39564 + }, + { + "epoch": 0.95, + "learning_rate": 1.2315315052346287e-07, + "loss": 0.8082, + "step": 39565 + }, + { + "epoch": 0.95, + "learning_rate": 1.2303131657961932e-07, + "loss": 1.0716, + "step": 39566 + }, + { + "epoch": 0.95, + "learning_rate": 1.2290954255702037e-07, + "loss": 1.0101, + "step": 39567 + }, + { + "epoch": 0.95, + "learning_rate": 1.227878284564066e-07, + "loss": 1.0931, + "step": 39568 + }, + { + "epoch": 0.95, + "learning_rate": 1.2266617427851512e-07, + "loss": 0.8445, + "step": 39569 + }, + { + "epoch": 0.95, + "learning_rate": 1.2254458002408544e-07, + "loss": 1.0198, + "step": 39570 + }, + { + "epoch": 0.95, + "learning_rate": 1.2242304569385465e-07, + "loss": 1.1485, + "step": 39571 + }, + { + "epoch": 0.95, + "learning_rate": 1.223015712885589e-07, + "loss": 0.8134, + "step": 39572 + }, + { + "epoch": 0.95, + "learning_rate": 1.2218015680893646e-07, + "loss": 0.9997, + "step": 39573 + }, + { + "epoch": 0.95, + "learning_rate": 1.2205880225572453e-07, + "loss": 0.9685, + "step": 39574 + }, + { + "epoch": 0.95, + "learning_rate": 1.2193750762965695e-07, + "loss": 1.0652, + "step": 39575 + }, + { + "epoch": 0.95, + "learning_rate": 1.2181627293147312e-07, + "loss": 1.3135, + "step": 39576 + }, + { + "epoch": 0.95, + "learning_rate": 1.2169509816190473e-07, + "loss": 1.3012, + "step": 39577 + }, + { + "epoch": 0.95, + "learning_rate": 1.2157398332168892e-07, + "loss": 1.1611, + "step": 39578 + }, + { + "epoch": 0.95, + "learning_rate": 1.2145292841156066e-07, + "loss": 1.2328, + "step": 39579 + }, + { + "epoch": 0.95, + "learning_rate": 1.213319334322538e-07, + "loss": 1.0428, + "step": 39580 + }, + { + "epoch": 0.95, + "learning_rate": 1.2121099838450222e-07, + "loss": 0.9434, + "step": 39581 + }, + { + "epoch": 0.95, + "learning_rate": 1.210901232690398e-07, + "loss": 0.9002, + "step": 39582 + }, + { + "epoch": 0.95, + "learning_rate": 1.2096930808660034e-07, + "loss": 1.062, + "step": 39583 + }, + { + "epoch": 0.95, + "learning_rate": 1.2084855283791664e-07, + "loss": 1.1014, + "step": 39584 + }, + { + "epoch": 0.95, + "learning_rate": 1.207278575237203e-07, + "loss": 1.1507, + "step": 39585 + }, + { + "epoch": 0.95, + "learning_rate": 1.2060722214474406e-07, + "loss": 1.2253, + "step": 39586 + }, + { + "epoch": 0.95, + "learning_rate": 1.2048664670172182e-07, + "loss": 0.8729, + "step": 39587 + }, + { + "epoch": 0.95, + "learning_rate": 1.2036613119538075e-07, + "loss": 1.03, + "step": 39588 + }, + { + "epoch": 0.95, + "learning_rate": 1.2024567562645584e-07, + "loss": 0.9485, + "step": 39589 + }, + { + "epoch": 0.95, + "learning_rate": 1.2012527999567757e-07, + "loss": 0.9072, + "step": 39590 + }, + { + "epoch": 0.95, + "learning_rate": 1.200049443037743e-07, + "loss": 0.9732, + "step": 39591 + }, + { + "epoch": 0.95, + "learning_rate": 1.198846685514776e-07, + "loss": 1.2905, + "step": 39592 + }, + { + "epoch": 0.95, + "learning_rate": 1.1976445273951586e-07, + "loss": 0.9186, + "step": 39593 + }, + { + "epoch": 0.95, + "learning_rate": 1.1964429686861844e-07, + "loss": 0.8809, + "step": 39594 + }, + { + "epoch": 0.95, + "learning_rate": 1.19524200939517e-07, + "loss": 0.9867, + "step": 39595 + }, + { + "epoch": 0.95, + "learning_rate": 1.194041649529376e-07, + "loss": 1.0155, + "step": 39596 + }, + { + "epoch": 0.95, + "learning_rate": 1.1928418890960858e-07, + "loss": 1.0554, + "step": 39597 + }, + { + "epoch": 0.95, + "learning_rate": 1.1916427281025933e-07, + "loss": 1.0765, + "step": 39598 + }, + { + "epoch": 0.95, + "learning_rate": 1.1904441665561483e-07, + "loss": 1.111, + "step": 39599 + }, + { + "epoch": 0.95, + "learning_rate": 1.1892462044640452e-07, + "loss": 1.1195, + "step": 39600 + }, + { + "epoch": 0.95, + "learning_rate": 1.1880488418335446e-07, + "loss": 0.9331, + "step": 39601 + }, + { + "epoch": 0.95, + "learning_rate": 1.1868520786719073e-07, + "loss": 1.1375, + "step": 39602 + }, + { + "epoch": 0.95, + "learning_rate": 1.1856559149864056e-07, + "loss": 1.0485, + "step": 39603 + }, + { + "epoch": 0.95, + "learning_rate": 1.1844603507842667e-07, + "loss": 0.9067, + "step": 39604 + }, + { + "epoch": 0.95, + "learning_rate": 1.1832653860727739e-07, + "loss": 1.1808, + "step": 39605 + }, + { + "epoch": 0.95, + "learning_rate": 1.1820710208591768e-07, + "loss": 1.0405, + "step": 39606 + }, + { + "epoch": 0.95, + "learning_rate": 1.180877255150703e-07, + "loss": 0.872, + "step": 39607 + }, + { + "epoch": 0.95, + "learning_rate": 1.1796840889546024e-07, + "loss": 1.2918, + "step": 39608 + }, + { + "epoch": 0.95, + "learning_rate": 1.1784915222781134e-07, + "loss": 1.0269, + "step": 39609 + }, + { + "epoch": 0.95, + "learning_rate": 1.177299555128475e-07, + "loss": 1.0327, + "step": 39610 + }, + { + "epoch": 0.95, + "learning_rate": 1.1761081875129143e-07, + "loss": 1.0118, + "step": 39611 + }, + { + "epoch": 0.95, + "learning_rate": 1.1749174194386592e-07, + "loss": 0.6919, + "step": 39612 + }, + { + "epoch": 0.95, + "learning_rate": 1.1737272509129371e-07, + "loss": 1.0199, + "step": 39613 + }, + { + "epoch": 0.95, + "learning_rate": 1.1725376819429757e-07, + "loss": 1.1399, + "step": 39614 + }, + { + "epoch": 0.95, + "learning_rate": 1.171348712535969e-07, + "loss": 0.5613, + "step": 39615 + }, + { + "epoch": 0.95, + "learning_rate": 1.1701603426991448e-07, + "loss": 1.0352, + "step": 39616 + }, + { + "epoch": 0.95, + "learning_rate": 1.1689725724397194e-07, + "loss": 1.0867, + "step": 39617 + }, + { + "epoch": 0.95, + "learning_rate": 1.1677854017648871e-07, + "loss": 1.0807, + "step": 39618 + }, + { + "epoch": 0.95, + "learning_rate": 1.1665988306818532e-07, + "loss": 1.0646, + "step": 39619 + }, + { + "epoch": 0.95, + "learning_rate": 1.165412859197812e-07, + "loss": 1.1423, + "step": 39620 + }, + { + "epoch": 0.95, + "learning_rate": 1.1642274873199688e-07, + "loss": 1.1863, + "step": 39621 + }, + { + "epoch": 0.95, + "learning_rate": 1.163042715055518e-07, + "loss": 1.173, + "step": 39622 + }, + { + "epoch": 0.95, + "learning_rate": 1.1618585424116202e-07, + "loss": 1.0769, + "step": 39623 + }, + { + "epoch": 0.95, + "learning_rate": 1.1606749693954922e-07, + "loss": 1.0241, + "step": 39624 + }, + { + "epoch": 0.95, + "learning_rate": 1.1594919960143058e-07, + "loss": 1.232, + "step": 39625 + }, + { + "epoch": 0.95, + "learning_rate": 1.1583096222752222e-07, + "loss": 0.884, + "step": 39626 + }, + { + "epoch": 0.95, + "learning_rate": 1.1571278481854242e-07, + "loss": 1.409, + "step": 39627 + }, + { + "epoch": 0.95, + "learning_rate": 1.1559466737520842e-07, + "loss": 0.8926, + "step": 39628 + }, + { + "epoch": 0.95, + "learning_rate": 1.1547660989823628e-07, + "loss": 1.0737, + "step": 39629 + }, + { + "epoch": 0.95, + "learning_rate": 1.1535861238834433e-07, + "loss": 1.1331, + "step": 39630 + }, + { + "epoch": 0.95, + "learning_rate": 1.1524067484624535e-07, + "loss": 1.0745, + "step": 39631 + }, + { + "epoch": 0.95, + "learning_rate": 1.151227972726554e-07, + "loss": 0.9052, + "step": 39632 + }, + { + "epoch": 0.95, + "learning_rate": 1.1500497966829172e-07, + "loss": 1.1391, + "step": 39633 + }, + { + "epoch": 0.95, + "learning_rate": 1.1488722203386704e-07, + "loss": 0.8954, + "step": 39634 + }, + { + "epoch": 0.95, + "learning_rate": 1.1476952437009748e-07, + "loss": 1.0927, + "step": 39635 + }, + { + "epoch": 0.95, + "learning_rate": 1.1465188667769467e-07, + "loss": 0.9507, + "step": 39636 + }, + { + "epoch": 0.95, + "learning_rate": 1.145343089573736e-07, + "loss": 1.2161, + "step": 39637 + }, + { + "epoch": 0.95, + "learning_rate": 1.1441679120984927e-07, + "loss": 1.0358, + "step": 39638 + }, + { + "epoch": 0.95, + "learning_rate": 1.142993334358311e-07, + "loss": 0.974, + "step": 39639 + }, + { + "epoch": 0.95, + "learning_rate": 1.1418193563603519e-07, + "loss": 1.1108, + "step": 39640 + }, + { + "epoch": 0.95, + "learning_rate": 1.1406459781117097e-07, + "loss": 0.8431, + "step": 39641 + }, + { + "epoch": 0.95, + "learning_rate": 1.1394731996195229e-07, + "loss": 0.8571, + "step": 39642 + }, + { + "epoch": 0.95, + "learning_rate": 1.1383010208909084e-07, + "loss": 0.8991, + "step": 39643 + }, + { + "epoch": 0.95, + "learning_rate": 1.1371294419329493e-07, + "loss": 1.1856, + "step": 39644 + }, + { + "epoch": 0.95, + "learning_rate": 1.1359584627527731e-07, + "loss": 1.0414, + "step": 39645 + }, + { + "epoch": 0.95, + "learning_rate": 1.1347880833574964e-07, + "loss": 0.8791, + "step": 39646 + }, + { + "epoch": 0.95, + "learning_rate": 1.1336183037541915e-07, + "loss": 0.9509, + "step": 39647 + }, + { + "epoch": 0.95, + "learning_rate": 1.1324491239499746e-07, + "loss": 1.0302, + "step": 39648 + }, + { + "epoch": 0.95, + "learning_rate": 1.1312805439519404e-07, + "loss": 0.9859, + "step": 39649 + }, + { + "epoch": 0.95, + "learning_rate": 1.1301125637671717e-07, + "loss": 1.1882, + "step": 39650 + }, + { + "epoch": 0.95, + "learning_rate": 1.128945183402752e-07, + "loss": 0.8165, + "step": 39651 + }, + { + "epoch": 0.95, + "learning_rate": 1.1277784028657645e-07, + "loss": 0.8657, + "step": 39652 + }, + { + "epoch": 0.95, + "learning_rate": 1.1266122221632814e-07, + "loss": 1.0388, + "step": 39653 + }, + { + "epoch": 0.95, + "learning_rate": 1.1254466413023967e-07, + "loss": 0.8524, + "step": 39654 + }, + { + "epoch": 0.95, + "learning_rate": 1.1242816602901608e-07, + "loss": 0.9577, + "step": 39655 + }, + { + "epoch": 0.95, + "learning_rate": 1.1231172791336564e-07, + "loss": 1.2255, + "step": 39656 + }, + { + "epoch": 0.95, + "learning_rate": 1.121953497839945e-07, + "loss": 1.1315, + "step": 39657 + }, + { + "epoch": 0.95, + "learning_rate": 1.1207903164160872e-07, + "loss": 1.1642, + "step": 39658 + }, + { + "epoch": 0.95, + "learning_rate": 1.1196277348691331e-07, + "loss": 1.0659, + "step": 39659 + }, + { + "epoch": 0.95, + "learning_rate": 1.1184657532061328e-07, + "loss": 1.1279, + "step": 39660 + }, + { + "epoch": 0.95, + "learning_rate": 1.1173043714341469e-07, + "loss": 0.9856, + "step": 39661 + }, + { + "epoch": 0.95, + "learning_rate": 1.1161435895602147e-07, + "loss": 0.9014, + "step": 39662 + }, + { + "epoch": 0.95, + "learning_rate": 1.1149834075913857e-07, + "loss": 0.7991, + "step": 39663 + }, + { + "epoch": 0.95, + "learning_rate": 1.1138238255346878e-07, + "loss": 0.8818, + "step": 39664 + }, + { + "epoch": 0.95, + "learning_rate": 1.1126648433971709e-07, + "loss": 1.2048, + "step": 39665 + }, + { + "epoch": 0.95, + "learning_rate": 1.1115064611858406e-07, + "loss": 0.9664, + "step": 39666 + }, + { + "epoch": 0.95, + "learning_rate": 1.1103486789077466e-07, + "loss": 1.179, + "step": 39667 + }, + { + "epoch": 0.95, + "learning_rate": 1.1091914965699168e-07, + "loss": 1.1245, + "step": 39668 + }, + { + "epoch": 0.95, + "learning_rate": 1.1080349141793456e-07, + "loss": 0.8099, + "step": 39669 + }, + { + "epoch": 0.95, + "learning_rate": 1.1068789317430828e-07, + "loss": 0.9653, + "step": 39670 + }, + { + "epoch": 0.95, + "learning_rate": 1.1057235492681117e-07, + "loss": 1.2333, + "step": 39671 + }, + { + "epoch": 0.95, + "learning_rate": 1.1045687667614602e-07, + "loss": 0.9938, + "step": 39672 + }, + { + "epoch": 0.95, + "learning_rate": 1.1034145842301336e-07, + "loss": 1.0603, + "step": 39673 + }, + { + "epoch": 0.95, + "learning_rate": 1.1022610016811153e-07, + "loss": 1.0335, + "step": 39674 + }, + { + "epoch": 0.95, + "learning_rate": 1.1011080191214329e-07, + "loss": 0.8687, + "step": 39675 + }, + { + "epoch": 0.95, + "learning_rate": 1.0999556365580477e-07, + "loss": 1.0536, + "step": 39676 + }, + { + "epoch": 0.95, + "learning_rate": 1.0988038539979761e-07, + "loss": 1.0645, + "step": 39677 + }, + { + "epoch": 0.95, + "learning_rate": 1.0976526714482128e-07, + "loss": 1.1925, + "step": 39678 + }, + { + "epoch": 0.95, + "learning_rate": 1.0965020889157074e-07, + "loss": 1.1156, + "step": 39679 + }, + { + "epoch": 0.95, + "learning_rate": 1.0953521064074768e-07, + "loss": 0.9295, + "step": 39680 + }, + { + "epoch": 0.95, + "learning_rate": 1.0942027239304709e-07, + "loss": 0.9663, + "step": 39681 + }, + { + "epoch": 0.95, + "learning_rate": 1.093053941491673e-07, + "loss": 1.0629, + "step": 39682 + }, + { + "epoch": 0.95, + "learning_rate": 1.0919057590980553e-07, + "loss": 0.9995, + "step": 39683 + }, + { + "epoch": 0.95, + "learning_rate": 1.090758176756579e-07, + "loss": 0.9523, + "step": 39684 + }, + { + "epoch": 0.95, + "learning_rate": 1.089611194474205e-07, + "loss": 1.0553, + "step": 39685 + }, + { + "epoch": 0.95, + "learning_rate": 1.0884648122579056e-07, + "loss": 1.042, + "step": 39686 + }, + { + "epoch": 0.95, + "learning_rate": 1.0873190301146197e-07, + "loss": 0.9845, + "step": 39687 + }, + { + "epoch": 0.95, + "learning_rate": 1.0861738480513084e-07, + "loss": 1.1675, + "step": 39688 + }, + { + "epoch": 0.95, + "learning_rate": 1.0850292660748995e-07, + "loss": 1.1101, + "step": 39689 + }, + { + "epoch": 0.95, + "learning_rate": 1.0838852841923541e-07, + "loss": 1.0625, + "step": 39690 + }, + { + "epoch": 0.95, + "learning_rate": 1.0827419024106222e-07, + "loss": 0.9166, + "step": 39691 + }, + { + "epoch": 0.95, + "learning_rate": 1.0815991207366206e-07, + "loss": 0.8114, + "step": 39692 + }, + { + "epoch": 0.95, + "learning_rate": 1.0804569391772879e-07, + "loss": 0.8354, + "step": 39693 + }, + { + "epoch": 0.95, + "learning_rate": 1.0793153577395743e-07, + "loss": 1.0014, + "step": 39694 + }, + { + "epoch": 0.95, + "learning_rate": 1.0781743764303631e-07, + "loss": 1.3163, + "step": 39695 + }, + { + "epoch": 0.95, + "learning_rate": 1.0770339952566044e-07, + "loss": 1.1004, + "step": 39696 + }, + { + "epoch": 0.95, + "learning_rate": 1.075894214225226e-07, + "loss": 0.8881, + "step": 39697 + }, + { + "epoch": 0.95, + "learning_rate": 1.0747550333431223e-07, + "loss": 0.9691, + "step": 39698 + }, + { + "epoch": 0.95, + "learning_rate": 1.0736164526172099e-07, + "loss": 1.0063, + "step": 39699 + }, + { + "epoch": 0.95, + "learning_rate": 1.0724784720543946e-07, + "loss": 0.9075, + "step": 39700 + }, + { + "epoch": 0.95, + "learning_rate": 1.0713410916615819e-07, + "loss": 1.0584, + "step": 39701 + }, + { + "epoch": 0.95, + "learning_rate": 1.0702043114456883e-07, + "loss": 1.109, + "step": 39702 + }, + { + "epoch": 0.95, + "learning_rate": 1.0690681314135753e-07, + "loss": 0.8801, + "step": 39703 + }, + { + "epoch": 0.95, + "learning_rate": 1.0679325515721595e-07, + "loss": 1.1724, + "step": 39704 + }, + { + "epoch": 0.95, + "learning_rate": 1.0667975719283353e-07, + "loss": 0.9821, + "step": 39705 + }, + { + "epoch": 0.95, + "learning_rate": 1.0656631924889749e-07, + "loss": 0.7825, + "step": 39706 + }, + { + "epoch": 0.95, + "learning_rate": 1.0645294132609729e-07, + "loss": 0.8729, + "step": 39707 + }, + { + "epoch": 0.95, + "learning_rate": 1.0633962342511905e-07, + "loss": 0.8614, + "step": 39708 + }, + { + "epoch": 0.95, + "learning_rate": 1.0622636554665111e-07, + "loss": 0.8007, + "step": 39709 + }, + { + "epoch": 0.95, + "learning_rate": 1.061131676913818e-07, + "loss": 1.0967, + "step": 39710 + }, + { + "epoch": 0.96, + "learning_rate": 1.0600002985999613e-07, + "loss": 1.1286, + "step": 39711 + }, + { + "epoch": 0.96, + "learning_rate": 1.0588695205318022e-07, + "loss": 1.0747, + "step": 39712 + }, + { + "epoch": 0.96, + "learning_rate": 1.0577393427162241e-07, + "loss": 0.9054, + "step": 39713 + }, + { + "epoch": 0.96, + "learning_rate": 1.0566097651600548e-07, + "loss": 0.8788, + "step": 39714 + }, + { + "epoch": 0.96, + "learning_rate": 1.0554807878701778e-07, + "loss": 1.1813, + "step": 39715 + }, + { + "epoch": 0.96, + "learning_rate": 1.0543524108534098e-07, + "loss": 1.0064, + "step": 39716 + }, + { + "epoch": 0.96, + "learning_rate": 1.053224634116612e-07, + "loss": 1.112, + "step": 39717 + }, + { + "epoch": 0.96, + "learning_rate": 1.0520974576666343e-07, + "loss": 0.8087, + "step": 39718 + }, + { + "epoch": 0.96, + "learning_rate": 1.0509708815103048e-07, + "loss": 0.7832, + "step": 39719 + }, + { + "epoch": 0.96, + "learning_rate": 1.0498449056544624e-07, + "loss": 1.1045, + "step": 39720 + }, + { + "epoch": 0.96, + "learning_rate": 1.048719530105935e-07, + "loss": 1.0389, + "step": 39721 + }, + { + "epoch": 0.96, + "learning_rate": 1.0475947548715503e-07, + "loss": 0.9686, + "step": 39722 + }, + { + "epoch": 0.96, + "learning_rate": 1.0464705799581365e-07, + "loss": 1.0958, + "step": 39723 + }, + { + "epoch": 0.96, + "learning_rate": 1.0453470053725101e-07, + "loss": 1.0353, + "step": 39724 + }, + { + "epoch": 0.96, + "learning_rate": 1.0442240311214879e-07, + "loss": 1.2793, + "step": 39725 + }, + { + "epoch": 0.96, + "learning_rate": 1.0431016572118868e-07, + "loss": 0.9877, + "step": 39726 + }, + { + "epoch": 0.96, + "learning_rate": 1.0419798836505013e-07, + "loss": 0.8854, + "step": 39727 + }, + { + "epoch": 0.96, + "learning_rate": 1.0408587104441481e-07, + "loss": 0.9558, + "step": 39728 + }, + { + "epoch": 0.96, + "learning_rate": 1.0397381375996329e-07, + "loss": 1.1458, + "step": 39729 + }, + { + "epoch": 0.96, + "learning_rate": 1.0386181651237504e-07, + "loss": 0.9422, + "step": 39730 + }, + { + "epoch": 0.96, + "learning_rate": 1.037498793023295e-07, + "loss": 0.9977, + "step": 39731 + }, + { + "epoch": 0.96, + "learning_rate": 1.0363800213050501e-07, + "loss": 1.2299, + "step": 39732 + }, + { + "epoch": 0.96, + "learning_rate": 1.0352618499758216e-07, + "loss": 0.8135, + "step": 39733 + }, + { + "epoch": 0.96, + "learning_rate": 1.0341442790423706e-07, + "loss": 0.8794, + "step": 39734 + }, + { + "epoch": 0.96, + "learning_rate": 1.0330273085114917e-07, + "loss": 1.0955, + "step": 39735 + }, + { + "epoch": 0.96, + "learning_rate": 1.0319109383899572e-07, + "loss": 0.7339, + "step": 39736 + }, + { + "epoch": 0.96, + "learning_rate": 1.0307951686845507e-07, + "loss": 1.1977, + "step": 39737 + }, + { + "epoch": 0.96, + "learning_rate": 1.0296799994020224e-07, + "loss": 0.9955, + "step": 39738 + }, + { + "epoch": 0.96, + "learning_rate": 1.0285654305491443e-07, + "loss": 0.9718, + "step": 39739 + }, + { + "epoch": 0.96, + "learning_rate": 1.027451462132678e-07, + "loss": 0.8924, + "step": 39740 + }, + { + "epoch": 0.96, + "learning_rate": 1.0263380941593848e-07, + "loss": 0.9275, + "step": 39741 + }, + { + "epoch": 0.96, + "learning_rate": 1.0252253266360257e-07, + "loss": 1.0208, + "step": 39742 + }, + { + "epoch": 0.96, + "learning_rate": 1.0241131595693398e-07, + "loss": 1.1021, + "step": 39743 + }, + { + "epoch": 0.96, + "learning_rate": 1.0230015929660775e-07, + "loss": 1.0272, + "step": 39744 + }, + { + "epoch": 0.96, + "learning_rate": 1.0218906268329887e-07, + "loss": 0.9092, + "step": 39745 + }, + { + "epoch": 0.96, + "learning_rate": 1.0207802611768014e-07, + "loss": 1.2196, + "step": 39746 + }, + { + "epoch": 0.96, + "learning_rate": 1.0196704960042547e-07, + "loss": 0.9864, + "step": 39747 + }, + { + "epoch": 0.96, + "learning_rate": 1.0185613313220988e-07, + "loss": 0.8686, + "step": 39748 + }, + { + "epoch": 0.96, + "learning_rate": 1.0174527671370394e-07, + "loss": 1.1513, + "step": 39749 + }, + { + "epoch": 0.96, + "learning_rate": 1.0163448034558266e-07, + "loss": 1.06, + "step": 39750 + }, + { + "epoch": 0.96, + "learning_rate": 1.015237440285155e-07, + "loss": 1.0394, + "step": 39751 + }, + { + "epoch": 0.96, + "learning_rate": 1.0141306776317527e-07, + "loss": 1.0314, + "step": 39752 + }, + { + "epoch": 0.96, + "learning_rate": 1.0130245155023477e-07, + "loss": 1.0795, + "step": 39753 + }, + { + "epoch": 0.96, + "learning_rate": 1.0119189539036345e-07, + "loss": 1.0358, + "step": 39754 + }, + { + "epoch": 0.96, + "learning_rate": 1.0108139928423188e-07, + "loss": 1.0046, + "step": 39755 + }, + { + "epoch": 0.96, + "learning_rate": 1.0097096323251177e-07, + "loss": 1.0115, + "step": 39756 + }, + { + "epoch": 0.96, + "learning_rate": 1.0086058723587255e-07, + "loss": 0.9075, + "step": 39757 + }, + { + "epoch": 0.96, + "learning_rate": 1.0075027129498371e-07, + "loss": 0.8109, + "step": 39758 + }, + { + "epoch": 0.96, + "learning_rate": 1.006400154105136e-07, + "loss": 0.9359, + "step": 39759 + }, + { + "epoch": 0.96, + "learning_rate": 1.005298195831328e-07, + "loss": 0.8395, + "step": 39760 + }, + { + "epoch": 0.96, + "learning_rate": 1.0041968381350964e-07, + "loss": 1.0857, + "step": 39761 + }, + { + "epoch": 0.96, + "learning_rate": 1.0030960810231139e-07, + "loss": 1.0764, + "step": 39762 + }, + { + "epoch": 0.96, + "learning_rate": 1.0019959245020528e-07, + "loss": 0.9341, + "step": 39763 + }, + { + "epoch": 0.96, + "learning_rate": 1.0008963685786078e-07, + "loss": 0.9811, + "step": 39764 + }, + { + "epoch": 0.96, + "learning_rate": 9.99797413259429e-08, + "loss": 1.0615, + "step": 39765 + }, + { + "epoch": 0.96, + "learning_rate": 9.986990585512001e-08, + "loss": 1.0961, + "step": 39766 + }, + { + "epoch": 0.96, + "learning_rate": 9.976013044605825e-08, + "loss": 0.8128, + "step": 39767 + }, + { + "epoch": 0.96, + "learning_rate": 9.965041509942153e-08, + "loss": 1.0155, + "step": 39768 + }, + { + "epoch": 0.96, + "learning_rate": 9.954075981587818e-08, + "loss": 1.0088, + "step": 39769 + }, + { + "epoch": 0.96, + "learning_rate": 9.943116459609214e-08, + "loss": 1.0415, + "step": 39770 + }, + { + "epoch": 0.96, + "learning_rate": 9.932162944072843e-08, + "loss": 0.8076, + "step": 39771 + }, + { + "epoch": 0.96, + "learning_rate": 9.921215435045206e-08, + "loss": 1.1516, + "step": 39772 + }, + { + "epoch": 0.96, + "learning_rate": 9.910273932592584e-08, + "loss": 1.0341, + "step": 39773 + }, + { + "epoch": 0.96, + "learning_rate": 9.89933843678148e-08, + "loss": 0.8315, + "step": 39774 + }, + { + "epoch": 0.96, + "learning_rate": 9.888408947678174e-08, + "loss": 0.8884, + "step": 39775 + }, + { + "epoch": 0.96, + "learning_rate": 9.877485465349057e-08, + "loss": 0.9087, + "step": 39776 + }, + { + "epoch": 0.96, + "learning_rate": 9.866567989860298e-08, + "loss": 0.9626, + "step": 39777 + }, + { + "epoch": 0.96, + "learning_rate": 9.855656521278289e-08, + "loss": 1.1316, + "step": 39778 + }, + { + "epoch": 0.96, + "learning_rate": 9.844751059668977e-08, + "loss": 1.0583, + "step": 39779 + }, + { + "epoch": 0.96, + "learning_rate": 9.833851605098754e-08, + "loss": 0.7673, + "step": 39780 + }, + { + "epoch": 0.96, + "learning_rate": 9.822958157633566e-08, + "loss": 1.0427, + "step": 39781 + }, + { + "epoch": 0.96, + "learning_rate": 9.812070717339694e-08, + "loss": 0.994, + "step": 39782 + }, + { + "epoch": 0.96, + "learning_rate": 9.801189284282974e-08, + "loss": 0.8899, + "step": 39783 + }, + { + "epoch": 0.96, + "learning_rate": 9.790313858529577e-08, + "loss": 1.1223, + "step": 39784 + }, + { + "epoch": 0.96, + "learning_rate": 9.779444440145447e-08, + "loss": 0.8635, + "step": 39785 + }, + { + "epoch": 0.96, + "learning_rate": 9.768581029196534e-08, + "loss": 1.0367, + "step": 39786 + }, + { + "epoch": 0.96, + "learning_rate": 9.757723625748672e-08, + "loss": 0.9769, + "step": 39787 + }, + { + "epoch": 0.96, + "learning_rate": 9.746872229867809e-08, + "loss": 1.3278, + "step": 39788 + }, + { + "epoch": 0.96, + "learning_rate": 9.736026841619783e-08, + "loss": 0.8524, + "step": 39789 + }, + { + "epoch": 0.96, + "learning_rate": 9.725187461070207e-08, + "loss": 1.1203, + "step": 39790 + }, + { + "epoch": 0.96, + "learning_rate": 9.714354088285139e-08, + "loss": 1.124, + "step": 39791 + }, + { + "epoch": 0.96, + "learning_rate": 9.703526723330192e-08, + "loss": 1.0522, + "step": 39792 + }, + { + "epoch": 0.96, + "learning_rate": 9.692705366270983e-08, + "loss": 1.2286, + "step": 39793 + }, + { + "epoch": 0.96, + "learning_rate": 9.681890017173123e-08, + "loss": 0.8923, + "step": 39794 + }, + { + "epoch": 0.96, + "learning_rate": 9.671080676102229e-08, + "loss": 0.6262, + "step": 39795 + }, + { + "epoch": 0.96, + "learning_rate": 9.660277343124135e-08, + "loss": 0.9337, + "step": 39796 + }, + { + "epoch": 0.96, + "learning_rate": 9.649480018304125e-08, + "loss": 0.7373, + "step": 39797 + }, + { + "epoch": 0.96, + "learning_rate": 9.638688701707699e-08, + "loss": 0.8527, + "step": 39798 + }, + { + "epoch": 0.96, + "learning_rate": 9.627903393400473e-08, + "loss": 1.2764, + "step": 39799 + }, + { + "epoch": 0.96, + "learning_rate": 9.617124093447838e-08, + "loss": 1.0804, + "step": 39800 + }, + { + "epoch": 0.96, + "learning_rate": 9.606350801915187e-08, + "loss": 0.7399, + "step": 39801 + }, + { + "epoch": 0.96, + "learning_rate": 9.59558351886769e-08, + "loss": 1.0106, + "step": 39802 + }, + { + "epoch": 0.96, + "learning_rate": 9.584822244370962e-08, + "loss": 1.0552, + "step": 39803 + }, + { + "epoch": 0.96, + "learning_rate": 9.57406697849017e-08, + "loss": 1.0729, + "step": 39804 + }, + { + "epoch": 0.96, + "learning_rate": 9.563317721290488e-08, + "loss": 1.1626, + "step": 39805 + }, + { + "epoch": 0.96, + "learning_rate": 9.552574472837195e-08, + "loss": 0.8257, + "step": 39806 + }, + { + "epoch": 0.96, + "learning_rate": 9.541837233195461e-08, + "loss": 1.0989, + "step": 39807 + }, + { + "epoch": 0.96, + "learning_rate": 9.531106002430457e-08, + "loss": 0.8437, + "step": 39808 + }, + { + "epoch": 0.96, + "learning_rate": 9.520380780607242e-08, + "loss": 1.0757, + "step": 39809 + }, + { + "epoch": 0.96, + "learning_rate": 9.509661567790874e-08, + "loss": 0.7163, + "step": 39810 + }, + { + "epoch": 0.96, + "learning_rate": 9.498948364046412e-08, + "loss": 1.0969, + "step": 39811 + }, + { + "epoch": 0.96, + "learning_rate": 9.488241169439028e-08, + "loss": 1.0133, + "step": 39812 + }, + { + "epoch": 0.96, + "learning_rate": 9.477539984033334e-08, + "loss": 0.997, + "step": 39813 + }, + { + "epoch": 0.96, + "learning_rate": 9.466844807894393e-08, + "loss": 1.1061, + "step": 39814 + }, + { + "epoch": 0.96, + "learning_rate": 9.45615564108715e-08, + "loss": 0.8652, + "step": 39815 + }, + { + "epoch": 0.96, + "learning_rate": 9.445472483676443e-08, + "loss": 0.9741, + "step": 39816 + }, + { + "epoch": 0.96, + "learning_rate": 9.434795335727109e-08, + "loss": 1.2529, + "step": 39817 + }, + { + "epoch": 0.96, + "learning_rate": 9.424124197303764e-08, + "loss": 0.9577, + "step": 39818 + }, + { + "epoch": 0.96, + "learning_rate": 9.413459068471353e-08, + "loss": 0.9159, + "step": 39819 + }, + { + "epoch": 0.96, + "learning_rate": 9.402799949294494e-08, + "loss": 1.0651, + "step": 39820 + }, + { + "epoch": 0.96, + "learning_rate": 9.392146839837802e-08, + "loss": 1.0171, + "step": 39821 + }, + { + "epoch": 0.96, + "learning_rate": 9.381499740165889e-08, + "loss": 1.146, + "step": 39822 + }, + { + "epoch": 0.96, + "learning_rate": 9.370858650343595e-08, + "loss": 0.8965, + "step": 39823 + }, + { + "epoch": 0.96, + "learning_rate": 9.36022357043509e-08, + "loss": 0.9969, + "step": 39824 + }, + { + "epoch": 0.96, + "learning_rate": 9.34959450050521e-08, + "loss": 1.0045, + "step": 39825 + }, + { + "epoch": 0.96, + "learning_rate": 9.338971440618238e-08, + "loss": 1.1306, + "step": 39826 + }, + { + "epoch": 0.96, + "learning_rate": 9.32835439083879e-08, + "loss": 1.2513, + "step": 39827 + }, + { + "epoch": 0.96, + "learning_rate": 9.317743351231256e-08, + "loss": 1.237, + "step": 39828 + }, + { + "epoch": 0.96, + "learning_rate": 9.30713832185981e-08, + "loss": 1.2125, + "step": 39829 + }, + { + "epoch": 0.96, + "learning_rate": 9.296539302788953e-08, + "loss": 0.8781, + "step": 39830 + }, + { + "epoch": 0.96, + "learning_rate": 9.285946294083081e-08, + "loss": 1.2056, + "step": 39831 + }, + { + "epoch": 0.96, + "learning_rate": 9.275359295806142e-08, + "loss": 0.7544, + "step": 39832 + }, + { + "epoch": 0.96, + "learning_rate": 9.264778308022748e-08, + "loss": 0.9617, + "step": 39833 + }, + { + "epoch": 0.96, + "learning_rate": 9.25420333079674e-08, + "loss": 0.9063, + "step": 39834 + }, + { + "epoch": 0.96, + "learning_rate": 9.24363436419251e-08, + "loss": 1.1202, + "step": 39835 + }, + { + "epoch": 0.96, + "learning_rate": 9.23307140827423e-08, + "loss": 0.8178, + "step": 39836 + }, + { + "epoch": 0.96, + "learning_rate": 9.222514463105736e-08, + "loss": 0.8994, + "step": 39837 + }, + { + "epoch": 0.96, + "learning_rate": 9.211963528751311e-08, + "loss": 1.1669, + "step": 39838 + }, + { + "epoch": 0.96, + "learning_rate": 9.201418605274792e-08, + "loss": 0.9626, + "step": 39839 + }, + { + "epoch": 0.96, + "learning_rate": 9.190879692740129e-08, + "loss": 0.8294, + "step": 39840 + }, + { + "epoch": 0.96, + "learning_rate": 9.180346791211491e-08, + "loss": 1.0227, + "step": 39841 + }, + { + "epoch": 0.96, + "learning_rate": 9.169819900752497e-08, + "loss": 1.142, + "step": 39842 + }, + { + "epoch": 0.96, + "learning_rate": 9.159299021427204e-08, + "loss": 1.1056, + "step": 39843 + }, + { + "epoch": 0.96, + "learning_rate": 9.14878415329945e-08, + "loss": 0.8388, + "step": 39844 + }, + { + "epoch": 0.96, + "learning_rate": 9.138275296432964e-08, + "loss": 1.401, + "step": 39845 + }, + { + "epoch": 0.96, + "learning_rate": 9.12777245089147e-08, + "loss": 1.0895, + "step": 39846 + }, + { + "epoch": 0.96, + "learning_rate": 9.117275616738697e-08, + "loss": 0.9553, + "step": 39847 + }, + { + "epoch": 0.96, + "learning_rate": 9.106784794038259e-08, + "loss": 1.073, + "step": 39848 + }, + { + "epoch": 0.96, + "learning_rate": 9.096299982853996e-08, + "loss": 1.1112, + "step": 39849 + }, + { + "epoch": 0.96, + "learning_rate": 9.08582118324941e-08, + "loss": 0.779, + "step": 39850 + }, + { + "epoch": 0.96, + "learning_rate": 9.075348395288009e-08, + "loss": 1.112, + "step": 39851 + }, + { + "epoch": 0.96, + "learning_rate": 9.064881619033406e-08, + "loss": 0.9973, + "step": 39852 + }, + { + "epoch": 0.96, + "learning_rate": 9.054420854549106e-08, + "loss": 1.3439, + "step": 39853 + }, + { + "epoch": 0.96, + "learning_rate": 9.043966101898505e-08, + "loss": 0.8196, + "step": 39854 + }, + { + "epoch": 0.96, + "learning_rate": 9.033517361145105e-08, + "loss": 0.9314, + "step": 39855 + }, + { + "epoch": 0.96, + "learning_rate": 9.02307463235219e-08, + "loss": 1.223, + "step": 39856 + }, + { + "epoch": 0.96, + "learning_rate": 9.012637915583378e-08, + "loss": 1.1715, + "step": 39857 + }, + { + "epoch": 0.96, + "learning_rate": 9.002207210901615e-08, + "loss": 1.1098, + "step": 39858 + }, + { + "epoch": 0.96, + "learning_rate": 8.991782518370407e-08, + "loss": 1.1807, + "step": 39859 + }, + { + "epoch": 0.96, + "learning_rate": 8.981363838053037e-08, + "loss": 0.982, + "step": 39860 + }, + { + "epoch": 0.96, + "learning_rate": 8.970951170012566e-08, + "loss": 0.9697, + "step": 39861 + }, + { + "epoch": 0.96, + "learning_rate": 8.960544514312275e-08, + "loss": 0.8984, + "step": 39862 + }, + { + "epoch": 0.96, + "learning_rate": 8.950143871015227e-08, + "loss": 1.1205, + "step": 39863 + }, + { + "epoch": 0.96, + "learning_rate": 8.93974924018448e-08, + "loss": 1.1793, + "step": 39864 + }, + { + "epoch": 0.96, + "learning_rate": 8.929360621883209e-08, + "loss": 1.0129, + "step": 39865 + }, + { + "epoch": 0.96, + "learning_rate": 8.918978016174473e-08, + "loss": 1.1492, + "step": 39866 + }, + { + "epoch": 0.96, + "learning_rate": 8.908601423121e-08, + "loss": 1.4068, + "step": 39867 + }, + { + "epoch": 0.96, + "learning_rate": 8.898230842786071e-08, + "loss": 0.9815, + "step": 39868 + }, + { + "epoch": 0.96, + "learning_rate": 8.887866275232416e-08, + "loss": 0.8341, + "step": 39869 + }, + { + "epoch": 0.96, + "learning_rate": 8.877507720522982e-08, + "loss": 1.0481, + "step": 39870 + }, + { + "epoch": 0.96, + "learning_rate": 8.86715517872061e-08, + "loss": 1.0285, + "step": 39871 + }, + { + "epoch": 0.96, + "learning_rate": 8.856808649887916e-08, + "loss": 1.0146, + "step": 39872 + }, + { + "epoch": 0.96, + "learning_rate": 8.84646813408807e-08, + "loss": 1.0481, + "step": 39873 + }, + { + "epoch": 0.96, + "learning_rate": 8.836133631383359e-08, + "loss": 0.9359, + "step": 39874 + }, + { + "epoch": 0.96, + "learning_rate": 8.82580514183673e-08, + "loss": 0.8497, + "step": 39875 + }, + { + "epoch": 0.96, + "learning_rate": 8.81548266551091e-08, + "loss": 0.8742, + "step": 39876 + }, + { + "epoch": 0.96, + "learning_rate": 8.805166202468296e-08, + "loss": 1.1058, + "step": 39877 + }, + { + "epoch": 0.96, + "learning_rate": 8.794855752771613e-08, + "loss": 1.0648, + "step": 39878 + }, + { + "epoch": 0.96, + "learning_rate": 8.784551316483481e-08, + "loss": 1.0964, + "step": 39879 + }, + { + "epoch": 0.96, + "learning_rate": 8.77425289366629e-08, + "loss": 1.1888, + "step": 39880 + }, + { + "epoch": 0.96, + "learning_rate": 8.76396048438255e-08, + "loss": 1.035, + "step": 39881 + }, + { + "epoch": 0.96, + "learning_rate": 8.753674088694652e-08, + "loss": 0.8984, + "step": 39882 + }, + { + "epoch": 0.96, + "learning_rate": 8.743393706665105e-08, + "loss": 1.0958, + "step": 39883 + }, + { + "epoch": 0.96, + "learning_rate": 8.733119338356299e-08, + "loss": 0.9967, + "step": 39884 + }, + { + "epoch": 0.96, + "learning_rate": 8.722850983830411e-08, + "loss": 0.9196, + "step": 39885 + }, + { + "epoch": 0.96, + "learning_rate": 8.712588643149834e-08, + "loss": 1.1738, + "step": 39886 + }, + { + "epoch": 0.96, + "learning_rate": 8.702332316376849e-08, + "loss": 0.9336, + "step": 39887 + }, + { + "epoch": 0.96, + "learning_rate": 8.692082003573632e-08, + "loss": 0.9846, + "step": 39888 + }, + { + "epoch": 0.96, + "learning_rate": 8.681837704802353e-08, + "loss": 0.9869, + "step": 39889 + }, + { + "epoch": 0.96, + "learning_rate": 8.671599420125188e-08, + "loss": 1.0186, + "step": 39890 + }, + { + "epoch": 0.96, + "learning_rate": 8.661367149604306e-08, + "loss": 1.1011, + "step": 39891 + }, + { + "epoch": 0.96, + "learning_rate": 8.651140893301768e-08, + "loss": 1.0487, + "step": 39892 + }, + { + "epoch": 0.96, + "learning_rate": 8.640920651279417e-08, + "loss": 0.8215, + "step": 39893 + }, + { + "epoch": 0.96, + "learning_rate": 8.630706423599533e-08, + "loss": 0.9712, + "step": 39894 + }, + { + "epoch": 0.96, + "learning_rate": 8.620498210323958e-08, + "loss": 1.1481, + "step": 39895 + }, + { + "epoch": 0.96, + "learning_rate": 8.61029601151464e-08, + "loss": 1.0939, + "step": 39896 + }, + { + "epoch": 0.96, + "learning_rate": 8.600099827233533e-08, + "loss": 1.0226, + "step": 39897 + }, + { + "epoch": 0.96, + "learning_rate": 8.58990965754225e-08, + "loss": 1.0322, + "step": 39898 + }, + { + "epoch": 0.96, + "learning_rate": 8.579725502502967e-08, + "loss": 1.0785, + "step": 39899 + }, + { + "epoch": 0.96, + "learning_rate": 8.569547362177189e-08, + "loss": 1.254, + "step": 39900 + }, + { + "epoch": 0.96, + "learning_rate": 8.559375236626755e-08, + "loss": 0.7077, + "step": 39901 + }, + { + "epoch": 0.96, + "learning_rate": 8.549209125913505e-08, + "loss": 0.9879, + "step": 39902 + }, + { + "epoch": 0.96, + "learning_rate": 8.539049030098944e-08, + "loss": 1.0551, + "step": 39903 + }, + { + "epoch": 0.96, + "learning_rate": 8.528894949244692e-08, + "loss": 1.0303, + "step": 39904 + }, + { + "epoch": 0.96, + "learning_rate": 8.518746883412587e-08, + "loss": 0.8913, + "step": 39905 + }, + { + "epoch": 0.96, + "learning_rate": 8.508604832663913e-08, + "loss": 1.0162, + "step": 39906 + }, + { + "epoch": 0.96, + "learning_rate": 8.498468797060288e-08, + "loss": 0.9606, + "step": 39907 + }, + { + "epoch": 0.96, + "learning_rate": 8.48833877666333e-08, + "loss": 1.1186, + "step": 39908 + }, + { + "epoch": 0.96, + "learning_rate": 8.478214771534321e-08, + "loss": 0.9893, + "step": 39909 + }, + { + "epoch": 0.96, + "learning_rate": 8.468096781734769e-08, + "loss": 0.9575, + "step": 39910 + }, + { + "epoch": 0.96, + "learning_rate": 8.457984807326181e-08, + "loss": 1.2164, + "step": 39911 + }, + { + "epoch": 0.96, + "learning_rate": 8.447878848369618e-08, + "loss": 1.0969, + "step": 39912 + }, + { + "epoch": 0.96, + "learning_rate": 8.437778904926586e-08, + "loss": 1.008, + "step": 39913 + }, + { + "epoch": 0.96, + "learning_rate": 8.42768497705837e-08, + "loss": 0.8256, + "step": 39914 + }, + { + "epoch": 0.96, + "learning_rate": 8.417597064826144e-08, + "loss": 1.0111, + "step": 39915 + }, + { + "epoch": 0.96, + "learning_rate": 8.407515168291192e-08, + "loss": 0.9212, + "step": 39916 + }, + { + "epoch": 0.96, + "learning_rate": 8.397439287514463e-08, + "loss": 1.1747, + "step": 39917 + }, + { + "epoch": 0.96, + "learning_rate": 8.387369422557246e-08, + "loss": 1.1611, + "step": 39918 + }, + { + "epoch": 0.96, + "learning_rate": 8.37730557348071e-08, + "loss": 0.9819, + "step": 39919 + }, + { + "epoch": 0.96, + "learning_rate": 8.367247740345808e-08, + "loss": 1.1189, + "step": 39920 + }, + { + "epoch": 0.96, + "learning_rate": 8.35719592321349e-08, + "loss": 1.0635, + "step": 39921 + }, + { + "epoch": 0.96, + "learning_rate": 8.34715012214482e-08, + "loss": 1.2406, + "step": 39922 + }, + { + "epoch": 0.96, + "learning_rate": 8.337110337200749e-08, + "loss": 1.2277, + "step": 39923 + }, + { + "epoch": 0.96, + "learning_rate": 8.327076568442339e-08, + "loss": 1.1049, + "step": 39924 + }, + { + "epoch": 0.96, + "learning_rate": 8.317048815930095e-08, + "loss": 1.2332, + "step": 39925 + }, + { + "epoch": 0.96, + "learning_rate": 8.307027079725082e-08, + "loss": 1.0479, + "step": 39926 + }, + { + "epoch": 0.96, + "learning_rate": 8.29701135988814e-08, + "loss": 1.0113, + "step": 39927 + }, + { + "epoch": 0.96, + "learning_rate": 8.287001656479999e-08, + "loss": 0.9444, + "step": 39928 + }, + { + "epoch": 0.96, + "learning_rate": 8.276997969561273e-08, + "loss": 1.1165, + "step": 39929 + }, + { + "epoch": 0.96, + "learning_rate": 8.267000299192806e-08, + "loss": 1.0837, + "step": 39930 + }, + { + "epoch": 0.96, + "learning_rate": 8.257008645435105e-08, + "loss": 1.2034, + "step": 39931 + }, + { + "epoch": 0.96, + "learning_rate": 8.247023008348898e-08, + "loss": 0.9281, + "step": 39932 + }, + { + "epoch": 0.96, + "learning_rate": 8.237043387994693e-08, + "loss": 0.9314, + "step": 39933 + }, + { + "epoch": 0.96, + "learning_rate": 8.227069784433107e-08, + "loss": 1.0368, + "step": 39934 + }, + { + "epoch": 0.96, + "learning_rate": 8.217102197724647e-08, + "loss": 0.9735, + "step": 39935 + }, + { + "epoch": 0.96, + "learning_rate": 8.2071406279296e-08, + "loss": 0.8475, + "step": 39936 + }, + { + "epoch": 0.96, + "learning_rate": 8.197185075108693e-08, + "loss": 1.0773, + "step": 39937 + }, + { + "epoch": 0.96, + "learning_rate": 8.187235539322103e-08, + "loss": 1.2262, + "step": 39938 + }, + { + "epoch": 0.96, + "learning_rate": 8.177292020630335e-08, + "loss": 1.1899, + "step": 39939 + }, + { + "epoch": 0.96, + "learning_rate": 8.167354519093562e-08, + "loss": 1.1065, + "step": 39940 + }, + { + "epoch": 0.96, + "learning_rate": 8.157423034772182e-08, + "loss": 0.8103, + "step": 39941 + }, + { + "epoch": 0.96, + "learning_rate": 8.147497567726482e-08, + "loss": 0.9064, + "step": 39942 + }, + { + "epoch": 0.96, + "learning_rate": 8.137578118016632e-08, + "loss": 1.0823, + "step": 39943 + }, + { + "epoch": 0.96, + "learning_rate": 8.127664685702696e-08, + "loss": 1.0599, + "step": 39944 + }, + { + "epoch": 0.96, + "learning_rate": 8.117757270844851e-08, + "loss": 1.1006, + "step": 39945 + }, + { + "epoch": 0.96, + "learning_rate": 8.10785587350349e-08, + "loss": 0.9713, + "step": 39946 + }, + { + "epoch": 0.96, + "learning_rate": 8.097960493738344e-08, + "loss": 1.1229, + "step": 39947 + }, + { + "epoch": 0.96, + "learning_rate": 8.088071131609587e-08, + "loss": 1.0425, + "step": 39948 + }, + { + "epoch": 0.96, + "learning_rate": 8.07818778717706e-08, + "loss": 1.1015, + "step": 39949 + }, + { + "epoch": 0.96, + "learning_rate": 8.068310460500939e-08, + "loss": 1.2409, + "step": 39950 + }, + { + "epoch": 0.96, + "learning_rate": 8.058439151641173e-08, + "loss": 1.0871, + "step": 39951 + }, + { + "epoch": 0.96, + "learning_rate": 8.048573860657383e-08, + "loss": 0.9736, + "step": 39952 + }, + { + "epoch": 0.96, + "learning_rate": 8.03871458760963e-08, + "loss": 1.0918, + "step": 39953 + }, + { + "epoch": 0.96, + "learning_rate": 8.028861332557759e-08, + "loss": 1.0072, + "step": 39954 + }, + { + "epoch": 0.96, + "learning_rate": 8.019014095561272e-08, + "loss": 0.845, + "step": 39955 + }, + { + "epoch": 0.96, + "learning_rate": 8.009172876680238e-08, + "loss": 1.0868, + "step": 39956 + }, + { + "epoch": 0.96, + "learning_rate": 7.999337675974162e-08, + "loss": 0.9174, + "step": 39957 + }, + { + "epoch": 0.96, + "learning_rate": 7.989508493502773e-08, + "loss": 0.8519, + "step": 39958 + }, + { + "epoch": 0.96, + "learning_rate": 7.979685329325803e-08, + "loss": 1.0309, + "step": 39959 + }, + { + "epoch": 0.96, + "learning_rate": 7.969868183502649e-08, + "loss": 1.1369, + "step": 39960 + }, + { + "epoch": 0.96, + "learning_rate": 7.960057056092929e-08, + "loss": 1.2684, + "step": 39961 + }, + { + "epoch": 0.96, + "learning_rate": 7.950251947156263e-08, + "loss": 0.9993, + "step": 39962 + }, + { + "epoch": 0.96, + "learning_rate": 7.940452856752045e-08, + "loss": 0.9383, + "step": 39963 + }, + { + "epoch": 0.96, + "learning_rate": 7.930659784939787e-08, + "loss": 1.1874, + "step": 39964 + }, + { + "epoch": 0.96, + "learning_rate": 7.920872731778884e-08, + "loss": 1.0479, + "step": 39965 + }, + { + "epoch": 0.96, + "learning_rate": 7.911091697328621e-08, + "loss": 1.1573, + "step": 39966 + }, + { + "epoch": 0.96, + "learning_rate": 7.901316681648508e-08, + "loss": 1.1439, + "step": 39967 + }, + { + "epoch": 0.96, + "learning_rate": 7.891547684797719e-08, + "loss": 1.1976, + "step": 39968 + }, + { + "epoch": 0.96, + "learning_rate": 7.881784706835539e-08, + "loss": 1.2446, + "step": 39969 + }, + { + "epoch": 0.96, + "learning_rate": 7.872027747821142e-08, + "loss": 0.9435, + "step": 39970 + }, + { + "epoch": 0.96, + "learning_rate": 7.862276807813929e-08, + "loss": 1.0498, + "step": 39971 + }, + { + "epoch": 0.96, + "learning_rate": 7.852531886872738e-08, + "loss": 0.9955, + "step": 39972 + }, + { + "epoch": 0.96, + "learning_rate": 7.842792985057079e-08, + "loss": 0.7901, + "step": 39973 + }, + { + "epoch": 0.96, + "learning_rate": 7.833060102425682e-08, + "loss": 1.2601, + "step": 39974 + }, + { + "epoch": 0.96, + "learning_rate": 7.823333239037722e-08, + "loss": 1.1898, + "step": 39975 + }, + { + "epoch": 0.96, + "learning_rate": 7.813612394952264e-08, + "loss": 1.2326, + "step": 39976 + }, + { + "epoch": 0.96, + "learning_rate": 7.803897570228258e-08, + "loss": 1.088, + "step": 39977 + }, + { + "epoch": 0.96, + "learning_rate": 7.79418876492466e-08, + "loss": 0.961, + "step": 39978 + }, + { + "epoch": 0.96, + "learning_rate": 7.784485979100309e-08, + "loss": 0.777, + "step": 39979 + }, + { + "epoch": 0.96, + "learning_rate": 7.774789212814049e-08, + "loss": 0.9527, + "step": 39980 + }, + { + "epoch": 0.96, + "learning_rate": 7.765098466124832e-08, + "loss": 1.0039, + "step": 39981 + }, + { + "epoch": 0.96, + "learning_rate": 7.755413739091277e-08, + "loss": 0.8255, + "step": 39982 + }, + { + "epoch": 0.96, + "learning_rate": 7.745735031772338e-08, + "loss": 1.06, + "step": 39983 + }, + { + "epoch": 0.96, + "learning_rate": 7.736062344226636e-08, + "loss": 1.1041, + "step": 39984 + }, + { + "epoch": 0.96, + "learning_rate": 7.726395676512787e-08, + "loss": 0.923, + "step": 39985 + }, + { + "epoch": 0.96, + "learning_rate": 7.716735028689526e-08, + "loss": 0.9629, + "step": 39986 + }, + { + "epoch": 0.96, + "learning_rate": 7.707080400815358e-08, + "loss": 0.9679, + "step": 39987 + }, + { + "epoch": 0.96, + "learning_rate": 7.697431792949017e-08, + "loss": 0.9275, + "step": 39988 + }, + { + "epoch": 0.96, + "learning_rate": 7.68778920514901e-08, + "loss": 1.0789, + "step": 39989 + }, + { + "epoch": 0.96, + "learning_rate": 7.678152637473624e-08, + "loss": 0.9215, + "step": 39990 + }, + { + "epoch": 0.96, + "learning_rate": 7.668522089981589e-08, + "loss": 1.1002, + "step": 39991 + }, + { + "epoch": 0.96, + "learning_rate": 7.658897562731194e-08, + "loss": 1.0615, + "step": 39992 + }, + { + "epoch": 0.96, + "learning_rate": 7.649279055780833e-08, + "loss": 1.1058, + "step": 39993 + }, + { + "epoch": 0.96, + "learning_rate": 7.639666569188908e-08, + "loss": 1.1768, + "step": 39994 + }, + { + "epoch": 0.96, + "learning_rate": 7.630060103013703e-08, + "loss": 1.0029, + "step": 39995 + }, + { + "epoch": 0.96, + "learning_rate": 7.620459657313505e-08, + "loss": 0.9598, + "step": 39996 + }, + { + "epoch": 0.96, + "learning_rate": 7.610865232146603e-08, + "loss": 1.047, + "step": 39997 + }, + { + "epoch": 0.96, + "learning_rate": 7.601276827571057e-08, + "loss": 1.0342, + "step": 39998 + }, + { + "epoch": 0.96, + "learning_rate": 7.591694443645159e-08, + "loss": 0.9484, + "step": 39999 + }, + { + "epoch": 0.96, + "learning_rate": 7.582118080427081e-08, + "loss": 0.9694, + "step": 40000 + }, + { + "epoch": 0.96, + "learning_rate": 7.572547737974777e-08, + "loss": 0.8101, + "step": 40001 + }, + { + "epoch": 0.96, + "learning_rate": 7.562983416346426e-08, + "loss": 1.0826, + "step": 40002 + }, + { + "epoch": 0.96, + "learning_rate": 7.553425115600088e-08, + "loss": 0.984, + "step": 40003 + }, + { + "epoch": 0.96, + "learning_rate": 7.543872835793608e-08, + "loss": 0.9585, + "step": 40004 + }, + { + "epoch": 0.96, + "learning_rate": 7.534326576985052e-08, + "loss": 0.8464, + "step": 40005 + }, + { + "epoch": 0.96, + "learning_rate": 7.52478633923226e-08, + "loss": 1.1299, + "step": 40006 + }, + { + "epoch": 0.96, + "learning_rate": 7.515252122593186e-08, + "loss": 1.1045, + "step": 40007 + }, + { + "epoch": 0.96, + "learning_rate": 7.505723927125563e-08, + "loss": 1.0746, + "step": 40008 + }, + { + "epoch": 0.96, + "learning_rate": 7.496201752887344e-08, + "loss": 1.0262, + "step": 40009 + }, + { + "epoch": 0.96, + "learning_rate": 7.48668559993615e-08, + "loss": 1.0832, + "step": 40010 + }, + { + "epoch": 0.96, + "learning_rate": 7.477175468329822e-08, + "loss": 0.8897, + "step": 40011 + }, + { + "epoch": 0.96, + "learning_rate": 7.467671358126093e-08, + "loss": 1.1215, + "step": 40012 + }, + { + "epoch": 0.96, + "learning_rate": 7.458173269382363e-08, + "loss": 1.0843, + "step": 40013 + }, + { + "epoch": 0.96, + "learning_rate": 7.448681202156582e-08, + "loss": 0.9835, + "step": 40014 + }, + { + "epoch": 0.96, + "learning_rate": 7.439195156506152e-08, + "loss": 1.2561, + "step": 40015 + }, + { + "epoch": 0.96, + "learning_rate": 7.42971513248858e-08, + "loss": 1.1224, + "step": 40016 + }, + { + "epoch": 0.96, + "learning_rate": 7.420241130161598e-08, + "loss": 1.0627, + "step": 40017 + }, + { + "epoch": 0.96, + "learning_rate": 7.410773149582496e-08, + "loss": 1.1347, + "step": 40018 + }, + { + "epoch": 0.96, + "learning_rate": 7.401311190808669e-08, + "loss": 1.2113, + "step": 40019 + }, + { + "epoch": 0.96, + "learning_rate": 7.391855253897739e-08, + "loss": 1.1246, + "step": 40020 + }, + { + "epoch": 0.96, + "learning_rate": 7.382405338906884e-08, + "loss": 1.0342, + "step": 40021 + }, + { + "epoch": 0.96, + "learning_rate": 7.372961445893501e-08, + "loss": 0.9433, + "step": 40022 + }, + { + "epoch": 0.96, + "learning_rate": 7.363523574914876e-08, + "loss": 0.9948, + "step": 40023 + }, + { + "epoch": 0.96, + "learning_rate": 7.354091726028301e-08, + "loss": 1.106, + "step": 40024 + }, + { + "epoch": 0.96, + "learning_rate": 7.344665899290948e-08, + "loss": 0.7258, + "step": 40025 + }, + { + "epoch": 0.96, + "learning_rate": 7.335246094759996e-08, + "loss": 0.8876, + "step": 40026 + }, + { + "epoch": 0.96, + "learning_rate": 7.32583231249262e-08, + "loss": 0.8945, + "step": 40027 + }, + { + "epoch": 0.96, + "learning_rate": 7.316424552545998e-08, + "loss": 0.9291, + "step": 40028 + }, + { + "epoch": 0.96, + "learning_rate": 7.307022814977083e-08, + "loss": 1.2329, + "step": 40029 + }, + { + "epoch": 0.96, + "learning_rate": 7.297627099842941e-08, + "loss": 0.8633, + "step": 40030 + }, + { + "epoch": 0.96, + "learning_rate": 7.288237407200637e-08, + "loss": 0.9941, + "step": 40031 + }, + { + "epoch": 0.96, + "learning_rate": 7.278853737107128e-08, + "loss": 1.1423, + "step": 40032 + }, + { + "epoch": 0.96, + "learning_rate": 7.269476089619142e-08, + "loss": 0.8853, + "step": 40033 + }, + { + "epoch": 0.96, + "learning_rate": 7.260104464793971e-08, + "loss": 0.8182, + "step": 40034 + }, + { + "epoch": 0.96, + "learning_rate": 7.250738862688123e-08, + "loss": 0.9033, + "step": 40035 + }, + { + "epoch": 0.96, + "learning_rate": 7.241379283358552e-08, + "loss": 1.3461, + "step": 40036 + }, + { + "epoch": 0.96, + "learning_rate": 7.232025726861991e-08, + "loss": 0.99, + "step": 40037 + }, + { + "epoch": 0.96, + "learning_rate": 7.222678193255172e-08, + "loss": 1.1195, + "step": 40038 + }, + { + "epoch": 0.96, + "learning_rate": 7.21333668259494e-08, + "loss": 0.9136, + "step": 40039 + }, + { + "epoch": 0.96, + "learning_rate": 7.204001194937915e-08, + "loss": 1.1016, + "step": 40040 + }, + { + "epoch": 0.96, + "learning_rate": 7.194671730340607e-08, + "loss": 1.0982, + "step": 40041 + }, + { + "epoch": 0.96, + "learning_rate": 7.18534828885975e-08, + "loss": 0.8666, + "step": 40042 + }, + { + "epoch": 0.96, + "learning_rate": 7.176030870551854e-08, + "loss": 1.0117, + "step": 40043 + }, + { + "epoch": 0.96, + "learning_rate": 7.166719475473316e-08, + "loss": 0.9543, + "step": 40044 + }, + { + "epoch": 0.96, + "learning_rate": 7.157414103680981e-08, + "loss": 0.993, + "step": 40045 + }, + { + "epoch": 0.96, + "learning_rate": 7.148114755230918e-08, + "loss": 1.008, + "step": 40046 + }, + { + "epoch": 0.96, + "learning_rate": 7.138821430179854e-08, + "loss": 0.9131, + "step": 40047 + }, + { + "epoch": 0.96, + "learning_rate": 7.129534128583971e-08, + "loss": 0.9427, + "step": 40048 + }, + { + "epoch": 0.96, + "learning_rate": 7.120252850499554e-08, + "loss": 0.9391, + "step": 40049 + }, + { + "epoch": 0.96, + "learning_rate": 7.110977595983226e-08, + "loss": 1.2942, + "step": 40050 + }, + { + "epoch": 0.96, + "learning_rate": 7.101708365090831e-08, + "loss": 1.0796, + "step": 40051 + }, + { + "epoch": 0.96, + "learning_rate": 7.092445157878991e-08, + "loss": 1.0077, + "step": 40052 + }, + { + "epoch": 0.96, + "learning_rate": 7.08318797440366e-08, + "loss": 1.0177, + "step": 40053 + }, + { + "epoch": 0.96, + "learning_rate": 7.073936814721128e-08, + "loss": 0.8537, + "step": 40054 + }, + { + "epoch": 0.96, + "learning_rate": 7.064691678887459e-08, + "loss": 1.1011, + "step": 40055 + }, + { + "epoch": 0.96, + "learning_rate": 7.05545256695872e-08, + "loss": 1.0362, + "step": 40056 + }, + { + "epoch": 0.96, + "learning_rate": 7.04621947899109e-08, + "loss": 0.9093, + "step": 40057 + }, + { + "epoch": 0.96, + "learning_rate": 7.036992415040411e-08, + "loss": 1.0908, + "step": 40058 + }, + { + "epoch": 0.96, + "learning_rate": 7.02777137516275e-08, + "loss": 1.0652, + "step": 40059 + }, + { + "epoch": 0.96, + "learning_rate": 7.018556359413953e-08, + "loss": 0.7752, + "step": 40060 + }, + { + "epoch": 0.96, + "learning_rate": 7.009347367850194e-08, + "loss": 0.8887, + "step": 40061 + }, + { + "epoch": 0.96, + "learning_rate": 7.000144400526987e-08, + "loss": 1.0414, + "step": 40062 + }, + { + "epoch": 0.96, + "learning_rate": 6.990947457500508e-08, + "loss": 0.7827, + "step": 40063 + }, + { + "epoch": 0.96, + "learning_rate": 6.981756538826268e-08, + "loss": 1.1661, + "step": 40064 + }, + { + "epoch": 0.96, + "learning_rate": 6.972571644560111e-08, + "loss": 0.9409, + "step": 40065 + }, + { + "epoch": 0.96, + "learning_rate": 6.963392774757883e-08, + "loss": 1.1714, + "step": 40066 + }, + { + "epoch": 0.96, + "learning_rate": 6.954219929475092e-08, + "loss": 0.8731, + "step": 40067 + }, + { + "epoch": 0.96, + "learning_rate": 6.945053108767475e-08, + "loss": 1.1025, + "step": 40068 + }, + { + "epoch": 0.96, + "learning_rate": 6.93589231269065e-08, + "loss": 1.1186, + "step": 40069 + }, + { + "epoch": 0.96, + "learning_rate": 6.926737541300243e-08, + "loss": 0.9834, + "step": 40070 + }, + { + "epoch": 0.96, + "learning_rate": 6.917588794651653e-08, + "loss": 0.958, + "step": 40071 + }, + { + "epoch": 0.96, + "learning_rate": 6.9084460728005e-08, + "loss": 1.0134, + "step": 40072 + }, + { + "epoch": 0.96, + "learning_rate": 6.899309375802188e-08, + "loss": 0.9067, + "step": 40073 + }, + { + "epoch": 0.96, + "learning_rate": 6.890178703712225e-08, + "loss": 1.0431, + "step": 40074 + }, + { + "epoch": 0.96, + "learning_rate": 6.8810540565859e-08, + "loss": 0.9107, + "step": 40075 + }, + { + "epoch": 0.96, + "learning_rate": 6.871935434478727e-08, + "loss": 1.0252, + "step": 40076 + }, + { + "epoch": 0.96, + "learning_rate": 6.862822837445882e-08, + "loss": 0.8771, + "step": 40077 + }, + { + "epoch": 0.96, + "learning_rate": 6.853716265542654e-08, + "loss": 0.9124, + "step": 40078 + }, + { + "epoch": 0.96, + "learning_rate": 6.844615718824332e-08, + "loss": 1.0166, + "step": 40079 + }, + { + "epoch": 0.96, + "learning_rate": 6.835521197346096e-08, + "loss": 1.0399, + "step": 40080 + }, + { + "epoch": 0.96, + "learning_rate": 6.826432701163122e-08, + "loss": 1.3135, + "step": 40081 + }, + { + "epoch": 0.96, + "learning_rate": 6.817350230330699e-08, + "loss": 0.8225, + "step": 40082 + }, + { + "epoch": 0.96, + "learning_rate": 6.808273784903674e-08, + "loss": 0.9834, + "step": 40083 + }, + { + "epoch": 0.96, + "learning_rate": 6.799203364937334e-08, + "loss": 1.1175, + "step": 40084 + }, + { + "epoch": 0.96, + "learning_rate": 6.790138970486526e-08, + "loss": 0.888, + "step": 40085 + }, + { + "epoch": 0.96, + "learning_rate": 6.781080601606427e-08, + "loss": 1.0124, + "step": 40086 + }, + { + "epoch": 0.96, + "learning_rate": 6.77202825835177e-08, + "loss": 0.9374, + "step": 40087 + }, + { + "epoch": 0.96, + "learning_rate": 6.762981940777735e-08, + "loss": 0.9094, + "step": 40088 + }, + { + "epoch": 0.96, + "learning_rate": 6.753941648938944e-08, + "loss": 1.0335, + "step": 40089 + }, + { + "epoch": 0.96, + "learning_rate": 6.744907382890464e-08, + "loss": 1.0885, + "step": 40090 + }, + { + "epoch": 0.96, + "learning_rate": 6.735879142686919e-08, + "loss": 0.8255, + "step": 40091 + }, + { + "epoch": 0.96, + "learning_rate": 6.726856928383152e-08, + "loss": 1.1422, + "step": 40092 + }, + { + "epoch": 0.96, + "learning_rate": 6.717840740034009e-08, + "loss": 1.0408, + "step": 40093 + }, + { + "epoch": 0.96, + "learning_rate": 6.708830577694004e-08, + "loss": 1.0301, + "step": 40094 + }, + { + "epoch": 0.96, + "learning_rate": 6.699826441417978e-08, + "loss": 0.9486, + "step": 40095 + }, + { + "epoch": 0.96, + "learning_rate": 6.690828331260446e-08, + "loss": 1.2118, + "step": 40096 + }, + { + "epoch": 0.96, + "learning_rate": 6.68183624727592e-08, + "loss": 0.9709, + "step": 40097 + }, + { + "epoch": 0.96, + "learning_rate": 6.672850189519131e-08, + "loss": 0.9807, + "step": 40098 + }, + { + "epoch": 0.96, + "learning_rate": 6.663870158044483e-08, + "loss": 1.0631, + "step": 40099 + }, + { + "epoch": 0.96, + "learning_rate": 6.654896152906487e-08, + "loss": 1.2281, + "step": 40100 + }, + { + "epoch": 0.96, + "learning_rate": 6.645928174159655e-08, + "loss": 0.9833, + "step": 40101 + }, + { + "epoch": 0.96, + "learning_rate": 6.636966221858276e-08, + "loss": 0.8474, + "step": 40102 + }, + { + "epoch": 0.96, + "learning_rate": 6.628010296056753e-08, + "loss": 1.3357, + "step": 40103 + }, + { + "epoch": 0.96, + "learning_rate": 6.619060396809484e-08, + "loss": 1.0936, + "step": 40104 + }, + { + "epoch": 0.96, + "learning_rate": 6.610116524170651e-08, + "loss": 1.2469, + "step": 40105 + }, + { + "epoch": 0.96, + "learning_rate": 6.601178678194653e-08, + "loss": 0.8045, + "step": 40106 + }, + { + "epoch": 0.96, + "learning_rate": 6.592246858935558e-08, + "loss": 0.9278, + "step": 40107 + }, + { + "epoch": 0.96, + "learning_rate": 6.583321066447657e-08, + "loss": 1.1996, + "step": 40108 + }, + { + "epoch": 0.96, + "learning_rate": 6.574401300785238e-08, + "loss": 1.1673, + "step": 40109 + }, + { + "epoch": 0.96, + "learning_rate": 6.565487562002037e-08, + "loss": 1.0557, + "step": 40110 + }, + { + "epoch": 0.96, + "learning_rate": 6.556579850152456e-08, + "loss": 1.0773, + "step": 40111 + }, + { + "epoch": 0.96, + "learning_rate": 6.54767816529045e-08, + "loss": 0.616, + "step": 40112 + }, + { + "epoch": 0.96, + "learning_rate": 6.538782507469976e-08, + "loss": 1.1722, + "step": 40113 + }, + { + "epoch": 0.96, + "learning_rate": 6.529892876745104e-08, + "loss": 1.1582, + "step": 40114 + }, + { + "epoch": 0.96, + "learning_rate": 6.521009273169565e-08, + "loss": 1.0387, + "step": 40115 + }, + { + "epoch": 0.96, + "learning_rate": 6.51213169679743e-08, + "loss": 1.1936, + "step": 40116 + }, + { + "epoch": 0.96, + "learning_rate": 6.503260147682544e-08, + "loss": 0.9984, + "step": 40117 + }, + { + "epoch": 0.96, + "learning_rate": 6.494394625878641e-08, + "loss": 1.3039, + "step": 40118 + }, + { + "epoch": 0.96, + "learning_rate": 6.485535131439569e-08, + "loss": 1.1555, + "step": 40119 + }, + { + "epoch": 0.96, + "learning_rate": 6.476681664419171e-08, + "loss": 1.1041, + "step": 40120 + }, + { + "epoch": 0.96, + "learning_rate": 6.467834224870961e-08, + "loss": 1.1354, + "step": 40121 + }, + { + "epoch": 0.96, + "learning_rate": 6.458992812848675e-08, + "loss": 1.1329, + "step": 40122 + }, + { + "epoch": 0.96, + "learning_rate": 6.450157428406045e-08, + "loss": 1.2449, + "step": 40123 + }, + { + "epoch": 0.96, + "learning_rate": 6.441328071596587e-08, + "loss": 0.9802, + "step": 40124 + }, + { + "epoch": 0.96, + "learning_rate": 6.432504742473922e-08, + "loss": 0.8968, + "step": 40125 + }, + { + "epoch": 0.96, + "learning_rate": 6.423687441091453e-08, + "loss": 1.0553, + "step": 40126 + }, + { + "epoch": 0.97, + "learning_rate": 6.414876167502915e-08, + "loss": 1.141, + "step": 40127 + }, + { + "epoch": 0.97, + "learning_rate": 6.406070921761487e-08, + "loss": 0.8483, + "step": 40128 + }, + { + "epoch": 0.97, + "learning_rate": 6.397271703920793e-08, + "loss": 1.0304, + "step": 40129 + }, + { + "epoch": 0.97, + "learning_rate": 6.388478514034124e-08, + "loss": 0.9848, + "step": 40130 + }, + { + "epoch": 0.97, + "learning_rate": 6.379691352154883e-08, + "loss": 1.1093, + "step": 40131 + }, + { + "epoch": 0.97, + "learning_rate": 6.370910218336245e-08, + "loss": 1.0641, + "step": 40132 + }, + { + "epoch": 0.97, + "learning_rate": 6.362135112631728e-08, + "loss": 1.1604, + "step": 40133 + }, + { + "epoch": 0.97, + "learning_rate": 6.353366035094288e-08, + "loss": 0.8293, + "step": 40134 + }, + { + "epoch": 0.97, + "learning_rate": 6.344602985777326e-08, + "loss": 1.0331, + "step": 40135 + }, + { + "epoch": 0.97, + "learning_rate": 6.335845964733911e-08, + "loss": 1.3143, + "step": 40136 + }, + { + "epoch": 0.97, + "learning_rate": 6.327094972017111e-08, + "loss": 0.9532, + "step": 40137 + }, + { + "epoch": 0.97, + "learning_rate": 6.318350007680219e-08, + "loss": 1.1818, + "step": 40138 + }, + { + "epoch": 0.97, + "learning_rate": 6.309611071776189e-08, + "loss": 0.9751, + "step": 40139 + }, + { + "epoch": 0.97, + "learning_rate": 6.300878164357871e-08, + "loss": 1.2511, + "step": 40140 + }, + { + "epoch": 0.97, + "learning_rate": 6.292151285478554e-08, + "loss": 1.1942, + "step": 40141 + }, + { + "epoch": 0.97, + "learning_rate": 6.283430435190973e-08, + "loss": 1.0511, + "step": 40142 + }, + { + "epoch": 0.97, + "learning_rate": 6.274715613547977e-08, + "loss": 1.2436, + "step": 40143 + }, + { + "epoch": 0.97, + "learning_rate": 6.266006820602633e-08, + "loss": 1.1313, + "step": 40144 + }, + { + "epoch": 0.97, + "learning_rate": 6.257304056407676e-08, + "loss": 0.9701, + "step": 40145 + }, + { + "epoch": 0.97, + "learning_rate": 6.248607321015954e-08, + "loss": 1.0487, + "step": 40146 + }, + { + "epoch": 0.97, + "learning_rate": 6.23991661448009e-08, + "loss": 0.8824, + "step": 40147 + }, + { + "epoch": 0.97, + "learning_rate": 6.231231936853044e-08, + "loss": 1.1281, + "step": 40148 + }, + { + "epoch": 0.97, + "learning_rate": 6.222553288187217e-08, + "loss": 0.9455, + "step": 40149 + }, + { + "epoch": 0.97, + "learning_rate": 6.213880668535455e-08, + "loss": 0.9273, + "step": 40150 + }, + { + "epoch": 0.97, + "learning_rate": 6.205214077950384e-08, + "loss": 1.1722, + "step": 40151 + }, + { + "epoch": 0.97, + "learning_rate": 6.196553516484515e-08, + "loss": 1.012, + "step": 40152 + }, + { + "epoch": 0.97, + "learning_rate": 6.187898984190365e-08, + "loss": 1.2634, + "step": 40153 + }, + { + "epoch": 0.97, + "learning_rate": 6.179250481120446e-08, + "loss": 1.0008, + "step": 40154 + }, + { + "epoch": 0.97, + "learning_rate": 6.170608007327273e-08, + "loss": 0.7887, + "step": 40155 + }, + { + "epoch": 0.97, + "learning_rate": 6.161971562863245e-08, + "loss": 1.0776, + "step": 40156 + }, + { + "epoch": 0.97, + "learning_rate": 6.153341147780878e-08, + "loss": 0.9541, + "step": 40157 + }, + { + "epoch": 0.97, + "learning_rate": 6.144716762132353e-08, + "loss": 0.5323, + "step": 40158 + }, + { + "epoch": 0.97, + "learning_rate": 6.13609840596996e-08, + "loss": 0.8674, + "step": 40159 + }, + { + "epoch": 0.97, + "learning_rate": 6.127486079346212e-08, + "loss": 0.9369, + "step": 40160 + }, + { + "epoch": 0.97, + "learning_rate": 6.11887978231318e-08, + "loss": 0.9554, + "step": 40161 + }, + { + "epoch": 0.97, + "learning_rate": 6.110279514923156e-08, + "loss": 1.2281, + "step": 40162 + }, + { + "epoch": 0.97, + "learning_rate": 6.101685277228319e-08, + "loss": 0.8492, + "step": 40163 + }, + { + "epoch": 0.97, + "learning_rate": 6.09309706928074e-08, + "loss": 1.1617, + "step": 40164 + }, + { + "epoch": 0.97, + "learning_rate": 6.084514891132598e-08, + "loss": 1.0179, + "step": 40165 + }, + { + "epoch": 0.97, + "learning_rate": 6.075938742835741e-08, + "loss": 0.9267, + "step": 40166 + }, + { + "epoch": 0.97, + "learning_rate": 6.067368624442571e-08, + "loss": 0.9706, + "step": 40167 + }, + { + "epoch": 0.97, + "learning_rate": 6.058804536004826e-08, + "loss": 0.8473, + "step": 40168 + }, + { + "epoch": 0.97, + "learning_rate": 6.050246477574573e-08, + "loss": 1.1933, + "step": 40169 + }, + { + "epoch": 0.97, + "learning_rate": 6.041694449203661e-08, + "loss": 1.0901, + "step": 40170 + }, + { + "epoch": 0.97, + "learning_rate": 6.033148450943937e-08, + "loss": 1.1502, + "step": 40171 + }, + { + "epoch": 0.97, + "learning_rate": 6.02460848284736e-08, + "loss": 1.1834, + "step": 40172 + }, + { + "epoch": 0.97, + "learning_rate": 6.016074544965667e-08, + "loss": 1.0899, + "step": 40173 + }, + { + "epoch": 0.97, + "learning_rate": 6.007546637350704e-08, + "loss": 0.9163, + "step": 40174 + }, + { + "epoch": 0.97, + "learning_rate": 5.999024760054095e-08, + "loss": 1.1745, + "step": 40175 + }, + { + "epoch": 0.97, + "learning_rate": 5.99050891312758e-08, + "loss": 0.9062, + "step": 40176 + }, + { + "epoch": 0.97, + "learning_rate": 5.981999096622892e-08, + "loss": 0.9618, + "step": 40177 + }, + { + "epoch": 0.97, + "learning_rate": 5.97349531059166e-08, + "loss": 1.1523, + "step": 40178 + }, + { + "epoch": 0.97, + "learning_rate": 5.964997555085284e-08, + "loss": 0.7613, + "step": 40179 + }, + { + "epoch": 0.97, + "learning_rate": 5.9565058301555015e-08, + "loss": 0.7923, + "step": 40180 + }, + { + "epoch": 0.97, + "learning_rate": 5.948020135853827e-08, + "loss": 0.9935, + "step": 40181 + }, + { + "epoch": 0.97, + "learning_rate": 5.939540472231664e-08, + "loss": 0.9029, + "step": 40182 + }, + { + "epoch": 0.97, + "learning_rate": 5.931066839340527e-08, + "loss": 0.9469, + "step": 40183 + }, + { + "epoch": 0.97, + "learning_rate": 5.9225992372318184e-08, + "loss": 1.0015, + "step": 40184 + }, + { + "epoch": 0.97, + "learning_rate": 5.914137665956832e-08, + "loss": 1.0163, + "step": 40185 + }, + { + "epoch": 0.97, + "learning_rate": 5.905682125566969e-08, + "loss": 1.0342, + "step": 40186 + }, + { + "epoch": 0.97, + "learning_rate": 5.897232616113413e-08, + "loss": 1.072, + "step": 40187 + }, + { + "epoch": 0.97, + "learning_rate": 5.888789137647566e-08, + "loss": 1.141, + "step": 40188 + }, + { + "epoch": 0.97, + "learning_rate": 5.8803516902207204e-08, + "loss": 0.9292, + "step": 40189 + }, + { + "epoch": 0.97, + "learning_rate": 5.8719202738838355e-08, + "loss": 0.9231, + "step": 40190 + }, + { + "epoch": 0.97, + "learning_rate": 5.8634948886882034e-08, + "loss": 1.0313, + "step": 40191 + }, + { + "epoch": 0.97, + "learning_rate": 5.855075534684895e-08, + "loss": 0.9111, + "step": 40192 + }, + { + "epoch": 0.97, + "learning_rate": 5.846662211925092e-08, + "loss": 1.0535, + "step": 40193 + }, + { + "epoch": 0.97, + "learning_rate": 5.838254920459641e-08, + "loss": 0.8541, + "step": 40194 + }, + { + "epoch": 0.97, + "learning_rate": 5.829853660339724e-08, + "loss": 1.1519, + "step": 40195 + }, + { + "epoch": 0.97, + "learning_rate": 5.821458431616189e-08, + "loss": 0.9871, + "step": 40196 + }, + { + "epoch": 0.97, + "learning_rate": 5.8130692343401074e-08, + "loss": 0.6721, + "step": 40197 + }, + { + "epoch": 0.97, + "learning_rate": 5.8046860685622156e-08, + "loss": 1.154, + "step": 40198 + }, + { + "epoch": 0.97, + "learning_rate": 5.7963089343333615e-08, + "loss": 1.0975, + "step": 40199 + }, + { + "epoch": 0.97, + "learning_rate": 5.787937831704615e-08, + "loss": 1.1823, + "step": 40200 + }, + { + "epoch": 0.97, + "learning_rate": 5.779572760726493e-08, + "loss": 0.9016, + "step": 40201 + }, + { + "epoch": 0.97, + "learning_rate": 5.771213721449953e-08, + "loss": 1.0756, + "step": 40202 + }, + { + "epoch": 0.97, + "learning_rate": 5.762860713925511e-08, + "loss": 1.1614, + "step": 40203 + }, + { + "epoch": 0.97, + "learning_rate": 5.754513738203904e-08, + "loss": 1.054, + "step": 40204 + }, + { + "epoch": 0.97, + "learning_rate": 5.746172794335869e-08, + "loss": 0.9362, + "step": 40205 + }, + { + "epoch": 0.97, + "learning_rate": 5.737837882371921e-08, + "loss": 1.0808, + "step": 40206 + }, + { + "epoch": 0.97, + "learning_rate": 5.729509002362688e-08, + "loss": 1.0742, + "step": 40207 + }, + { + "epoch": 0.97, + "learning_rate": 5.721186154358571e-08, + "loss": 1.0414, + "step": 40208 + }, + { + "epoch": 0.97, + "learning_rate": 5.7128693384101985e-08, + "loss": 0.9782, + "step": 40209 + }, + { + "epoch": 0.97, + "learning_rate": 5.704558554567974e-08, + "loss": 1.1712, + "step": 40210 + }, + { + "epoch": 0.97, + "learning_rate": 5.6962538028824124e-08, + "loss": 0.875, + "step": 40211 + }, + { + "epoch": 0.97, + "learning_rate": 5.687955083403695e-08, + "loss": 0.9094, + "step": 40212 + }, + { + "epoch": 0.97, + "learning_rate": 5.679662396182228e-08, + "loss": 1.087, + "step": 40213 + }, + { + "epoch": 0.97, + "learning_rate": 5.671375741268526e-08, + "loss": 1.0113, + "step": 40214 + }, + { + "epoch": 0.97, + "learning_rate": 5.6630951187125473e-08, + "loss": 0.9915, + "step": 40215 + }, + { + "epoch": 0.97, + "learning_rate": 5.6548205285648084e-08, + "loss": 0.8411, + "step": 40216 + }, + { + "epoch": 0.97, + "learning_rate": 5.6465519708752694e-08, + "loss": 0.8904, + "step": 40217 + }, + { + "epoch": 0.97, + "learning_rate": 5.638289445694334e-08, + "loss": 0.7699, + "step": 40218 + }, + { + "epoch": 0.97, + "learning_rate": 5.630032953071962e-08, + "loss": 0.9057, + "step": 40219 + }, + { + "epoch": 0.97, + "learning_rate": 5.6217824930582257e-08, + "loss": 1.0609, + "step": 40220 + }, + { + "epoch": 0.97, + "learning_rate": 5.613538065703417e-08, + "loss": 1.0958, + "step": 40221 + }, + { + "epoch": 0.97, + "learning_rate": 5.605299671057163e-08, + "loss": 1.0522, + "step": 40222 + }, + { + "epoch": 0.97, + "learning_rate": 5.597067309169757e-08, + "loss": 1.0054, + "step": 40223 + }, + { + "epoch": 0.97, + "learning_rate": 5.588840980091159e-08, + "loss": 0.9055, + "step": 40224 + }, + { + "epoch": 0.97, + "learning_rate": 5.5806206838709965e-08, + "loss": 1.1091, + "step": 40225 + }, + { + "epoch": 0.97, + "learning_rate": 5.5724064205593396e-08, + "loss": 0.9786, + "step": 40226 + }, + { + "epoch": 0.97, + "learning_rate": 5.564198190206038e-08, + "loss": 1.1118, + "step": 40227 + }, + { + "epoch": 0.97, + "learning_rate": 5.55599599286083e-08, + "loss": 1.1672, + "step": 40228 + }, + { + "epoch": 0.97, + "learning_rate": 5.5477998285735634e-08, + "loss": 1.0265, + "step": 40229 + }, + { + "epoch": 0.97, + "learning_rate": 5.5396096973937554e-08, + "loss": 0.9443, + "step": 40230 + }, + { + "epoch": 0.97, + "learning_rate": 5.531425599371254e-08, + "loss": 1.134, + "step": 40231 + }, + { + "epoch": 0.97, + "learning_rate": 5.523247534555687e-08, + "loss": 1.0235, + "step": 40232 + }, + { + "epoch": 0.97, + "learning_rate": 5.515075502996681e-08, + "loss": 1.0156, + "step": 40233 + }, + { + "epoch": 0.97, + "learning_rate": 5.506909504743751e-08, + "loss": 1.1124, + "step": 40234 + }, + { + "epoch": 0.97, + "learning_rate": 5.498749539846526e-08, + "loss": 0.9477, + "step": 40235 + }, + { + "epoch": 0.97, + "learning_rate": 5.49059560835441e-08, + "loss": 1.096, + "step": 40236 + }, + { + "epoch": 0.97, + "learning_rate": 5.482447710317029e-08, + "loss": 0.9426, + "step": 40237 + }, + { + "epoch": 0.97, + "learning_rate": 5.474305845783678e-08, + "loss": 0.8993, + "step": 40238 + }, + { + "epoch": 0.97, + "learning_rate": 5.466170014803762e-08, + "loss": 0.8964, + "step": 40239 + }, + { + "epoch": 0.97, + "learning_rate": 5.458040217426686e-08, + "loss": 1.0191, + "step": 40240 + }, + { + "epoch": 0.97, + "learning_rate": 5.449916453701631e-08, + "loss": 1.1199, + "step": 40241 + }, + { + "epoch": 0.97, + "learning_rate": 5.4417987236781156e-08, + "loss": 1.2221, + "step": 40242 + }, + { + "epoch": 0.97, + "learning_rate": 5.433687027405321e-08, + "loss": 1.1887, + "step": 40243 + }, + { + "epoch": 0.97, + "learning_rate": 5.42558136493232e-08, + "loss": 0.8088, + "step": 40244 + }, + { + "epoch": 0.97, + "learning_rate": 5.417481736308405e-08, + "loss": 1.0567, + "step": 40245 + }, + { + "epoch": 0.97, + "learning_rate": 5.40938814158265e-08, + "loss": 0.9147, + "step": 40246 + }, + { + "epoch": 0.97, + "learning_rate": 5.401300580804236e-08, + "loss": 1.0488, + "step": 40247 + }, + { + "epoch": 0.97, + "learning_rate": 5.3932190540222364e-08, + "loss": 1.0502, + "step": 40248 + }, + { + "epoch": 0.97, + "learning_rate": 5.3851435612855e-08, + "loss": 1.1707, + "step": 40249 + }, + { + "epoch": 0.97, + "learning_rate": 5.377074102643209e-08, + "loss": 1.0824, + "step": 40250 + }, + { + "epoch": 0.97, + "learning_rate": 5.369010678144326e-08, + "loss": 0.9119, + "step": 40251 + }, + { + "epoch": 0.97, + "learning_rate": 5.3609532878377e-08, + "loss": 1.0848, + "step": 40252 + }, + { + "epoch": 0.97, + "learning_rate": 5.3529019317721806e-08, + "loss": 1.2424, + "step": 40253 + }, + { + "epoch": 0.97, + "learning_rate": 5.344856609996618e-08, + "loss": 1.0534, + "step": 40254 + }, + { + "epoch": 0.97, + "learning_rate": 5.336817322559862e-08, + "loss": 0.8442, + "step": 40255 + }, + { + "epoch": 0.97, + "learning_rate": 5.328784069510762e-08, + "loss": 0.9474, + "step": 40256 + }, + { + "epoch": 0.97, + "learning_rate": 5.320756850897835e-08, + "loss": 0.989, + "step": 40257 + }, + { + "epoch": 0.97, + "learning_rate": 5.31273566676993e-08, + "loss": 0.8476, + "step": 40258 + }, + { + "epoch": 0.97, + "learning_rate": 5.3047205171757874e-08, + "loss": 0.7981, + "step": 40259 + }, + { + "epoch": 0.97, + "learning_rate": 5.2967114021639234e-08, + "loss": 0.7188, + "step": 40260 + }, + { + "epoch": 0.97, + "learning_rate": 5.288708321782854e-08, + "loss": 1.3829, + "step": 40261 + }, + { + "epoch": 0.97, + "learning_rate": 5.2807112760813186e-08, + "loss": 0.9433, + "step": 40262 + }, + { + "epoch": 0.97, + "learning_rate": 5.2727202651077226e-08, + "loss": 1.0479, + "step": 40263 + }, + { + "epoch": 0.97, + "learning_rate": 5.264735288910472e-08, + "loss": 1.244, + "step": 40264 + }, + { + "epoch": 0.97, + "learning_rate": 5.256756347538194e-08, + "loss": 0.9858, + "step": 40265 + }, + { + "epoch": 0.97, + "learning_rate": 5.2487834410391844e-08, + "loss": 1.1256, + "step": 40266 + }, + { + "epoch": 0.97, + "learning_rate": 5.2408165694617374e-08, + "loss": 1.0615, + "step": 40267 + }, + { + "epoch": 0.97, + "learning_rate": 5.232855732854369e-08, + "loss": 0.7851, + "step": 40268 + }, + { + "epoch": 0.97, + "learning_rate": 5.2249009312652646e-08, + "loss": 1.0849, + "step": 40269 + }, + { + "epoch": 0.97, + "learning_rate": 5.216952164742717e-08, + "loss": 1.1481, + "step": 40270 + }, + { + "epoch": 0.97, + "learning_rate": 5.2090094333348e-08, + "loss": 1.1249, + "step": 40271 + }, + { + "epoch": 0.97, + "learning_rate": 5.201072737090029e-08, + "loss": 1.0411, + "step": 40272 + }, + { + "epoch": 0.97, + "learning_rate": 5.193142076056257e-08, + "loss": 0.9405, + "step": 40273 + }, + { + "epoch": 0.97, + "learning_rate": 5.1852174502816654e-08, + "loss": 0.9578, + "step": 40274 + }, + { + "epoch": 0.97, + "learning_rate": 5.177298859814439e-08, + "loss": 0.9939, + "step": 40275 + }, + { + "epoch": 0.97, + "learning_rate": 5.1693863047025395e-08, + "loss": 1.2741, + "step": 40276 + }, + { + "epoch": 0.97, + "learning_rate": 5.161479784993928e-08, + "loss": 1.1936, + "step": 40277 + }, + { + "epoch": 0.97, + "learning_rate": 5.1535793007366774e-08, + "loss": 0.8214, + "step": 40278 + }, + { + "epoch": 0.97, + "learning_rate": 5.1456848519785275e-08, + "loss": 0.9542, + "step": 40279 + }, + { + "epoch": 0.97, + "learning_rate": 5.137796438767662e-08, + "loss": 0.9473, + "step": 40280 + }, + { + "epoch": 0.97, + "learning_rate": 5.1299140611517084e-08, + "loss": 1.016, + "step": 40281 + }, + { + "epoch": 0.97, + "learning_rate": 5.122037719178519e-08, + "loss": 0.9265, + "step": 40282 + }, + { + "epoch": 0.97, + "learning_rate": 5.114167412895943e-08, + "loss": 0.8171, + "step": 40283 + }, + { + "epoch": 0.97, + "learning_rate": 5.106303142351721e-08, + "loss": 0.9088, + "step": 40284 + }, + { + "epoch": 0.97, + "learning_rate": 5.098444907593592e-08, + "loss": 0.8074, + "step": 40285 + }, + { + "epoch": 0.97, + "learning_rate": 5.090592708669184e-08, + "loss": 1.098, + "step": 40286 + }, + { + "epoch": 0.97, + "learning_rate": 5.082746545626016e-08, + "loss": 0.9279, + "step": 40287 + }, + { + "epoch": 0.97, + "learning_rate": 5.074906418511938e-08, + "loss": 1.0705, + "step": 40288 + }, + { + "epoch": 0.97, + "learning_rate": 5.067072327374245e-08, + "loss": 1.2852, + "step": 40289 + }, + { + "epoch": 0.97, + "learning_rate": 5.0592442722606775e-08, + "loss": 0.8229, + "step": 40290 + }, + { + "epoch": 0.97, + "learning_rate": 5.051422253218641e-08, + "loss": 0.627, + "step": 40291 + }, + { + "epoch": 0.97, + "learning_rate": 5.0436062702956536e-08, + "loss": 1.2111, + "step": 40292 + }, + { + "epoch": 0.97, + "learning_rate": 5.0357963235390103e-08, + "loss": 0.9476, + "step": 40293 + }, + { + "epoch": 0.97, + "learning_rate": 5.0279924129962296e-08, + "loss": 1.036, + "step": 40294 + }, + { + "epoch": 0.97, + "learning_rate": 5.020194538714496e-08, + "loss": 0.9218, + "step": 40295 + }, + { + "epoch": 0.97, + "learning_rate": 5.0124027007412145e-08, + "loss": 0.9002, + "step": 40296 + }, + { + "epoch": 0.97, + "learning_rate": 5.0046168991236823e-08, + "loss": 0.9445, + "step": 40297 + }, + { + "epoch": 0.97, + "learning_rate": 4.9968371339090824e-08, + "loss": 1.0028, + "step": 40298 + }, + { + "epoch": 0.97, + "learning_rate": 4.9890634051447116e-08, + "loss": 1.0377, + "step": 40299 + }, + { + "epoch": 0.97, + "learning_rate": 4.981295712877643e-08, + "loss": 1.0408, + "step": 40300 + }, + { + "epoch": 0.97, + "learning_rate": 4.9735340571549494e-08, + "loss": 0.9262, + "step": 40301 + }, + { + "epoch": 0.97, + "learning_rate": 4.965778438023816e-08, + "loss": 0.9365, + "step": 40302 + }, + { + "epoch": 0.97, + "learning_rate": 4.958028855531205e-08, + "loss": 0.9217, + "step": 40303 + }, + { + "epoch": 0.97, + "learning_rate": 4.9502853097241896e-08, + "loss": 1.177, + "step": 40304 + }, + { + "epoch": 0.97, + "learning_rate": 4.942547800649844e-08, + "loss": 0.7957, + "step": 40305 + }, + { + "epoch": 0.97, + "learning_rate": 4.934816328354908e-08, + "loss": 1.1058, + "step": 40306 + }, + { + "epoch": 0.97, + "learning_rate": 4.9270908928864544e-08, + "loss": 0.935, + "step": 40307 + }, + { + "epoch": 0.97, + "learning_rate": 4.919371494291225e-08, + "loss": 1.0579, + "step": 40308 + }, + { + "epoch": 0.97, + "learning_rate": 4.911658132616071e-08, + "loss": 1.2472, + "step": 40309 + }, + { + "epoch": 0.97, + "learning_rate": 4.9039508079079535e-08, + "loss": 0.9822, + "step": 40310 + }, + { + "epoch": 0.97, + "learning_rate": 4.896249520213392e-08, + "loss": 0.7073, + "step": 40311 + }, + { + "epoch": 0.97, + "learning_rate": 4.8885542695792374e-08, + "loss": 0.9807, + "step": 40312 + }, + { + "epoch": 0.97, + "learning_rate": 4.880865056052231e-08, + "loss": 0.9572, + "step": 40313 + }, + { + "epoch": 0.97, + "learning_rate": 4.8731818796788896e-08, + "loss": 0.7368, + "step": 40314 + }, + { + "epoch": 0.97, + "learning_rate": 4.865504740505955e-08, + "loss": 1.1567, + "step": 40315 + }, + { + "epoch": 0.97, + "learning_rate": 4.8578336385799454e-08, + "loss": 0.9307, + "step": 40316 + }, + { + "epoch": 0.97, + "learning_rate": 4.850168573947267e-08, + "loss": 1.1059, + "step": 40317 + }, + { + "epoch": 0.97, + "learning_rate": 4.842509546654661e-08, + "loss": 0.9238, + "step": 40318 + }, + { + "epoch": 0.97, + "learning_rate": 4.834856556748424e-08, + "loss": 0.933, + "step": 40319 + }, + { + "epoch": 0.97, + "learning_rate": 4.827209604274963e-08, + "loss": 1.1588, + "step": 40320 + }, + { + "epoch": 0.97, + "learning_rate": 4.819568689280796e-08, + "loss": 1.1797, + "step": 40321 + }, + { + "epoch": 0.97, + "learning_rate": 4.8119338118122196e-08, + "loss": 1.1089, + "step": 40322 + }, + { + "epoch": 0.97, + "learning_rate": 4.804304971915641e-08, + "loss": 1.0137, + "step": 40323 + }, + { + "epoch": 0.97, + "learning_rate": 4.7966821696372455e-08, + "loss": 1.0472, + "step": 40324 + }, + { + "epoch": 0.97, + "learning_rate": 4.789065405023219e-08, + "loss": 0.9024, + "step": 40325 + }, + { + "epoch": 0.97, + "learning_rate": 4.7814546781198565e-08, + "loss": 1.1326, + "step": 40326 + }, + { + "epoch": 0.97, + "learning_rate": 4.773849988973345e-08, + "loss": 1.0591, + "step": 40327 + }, + { + "epoch": 0.97, + "learning_rate": 4.766251337629757e-08, + "loss": 1.1081, + "step": 40328 + }, + { + "epoch": 0.97, + "learning_rate": 4.7586587241352786e-08, + "loss": 1.191, + "step": 40329 + }, + { + "epoch": 0.97, + "learning_rate": 4.7510721485358734e-08, + "loss": 1.0034, + "step": 40330 + }, + { + "epoch": 0.97, + "learning_rate": 4.743491610877615e-08, + "loss": 0.9422, + "step": 40331 + }, + { + "epoch": 0.97, + "learning_rate": 4.7359171112064675e-08, + "loss": 1.2058, + "step": 40332 + }, + { + "epoch": 0.97, + "learning_rate": 4.728348649568504e-08, + "loss": 1.2464, + "step": 40333 + }, + { + "epoch": 0.97, + "learning_rate": 4.720786226009466e-08, + "loss": 0.9284, + "step": 40334 + }, + { + "epoch": 0.97, + "learning_rate": 4.713229840575428e-08, + "loss": 0.982, + "step": 40335 + }, + { + "epoch": 0.97, + "learning_rate": 4.705679493312021e-08, + "loss": 0.9581, + "step": 40336 + }, + { + "epoch": 0.97, + "learning_rate": 4.6981351842652065e-08, + "loss": 1.0586, + "step": 40337 + }, + { + "epoch": 0.97, + "learning_rate": 4.6905969134806164e-08, + "loss": 0.9038, + "step": 40338 + }, + { + "epoch": 0.97, + "learning_rate": 4.683064681004101e-08, + "loss": 1.0217, + "step": 40339 + }, + { + "epoch": 0.97, + "learning_rate": 4.6755384868812925e-08, + "loss": 0.819, + "step": 40340 + }, + { + "epoch": 0.97, + "learning_rate": 4.668018331157931e-08, + "loss": 0.8437, + "step": 40341 + }, + { + "epoch": 0.97, + "learning_rate": 4.6605042138796466e-08, + "loss": 0.9411, + "step": 40342 + }, + { + "epoch": 0.97, + "learning_rate": 4.6529961350918475e-08, + "loss": 1.0273, + "step": 40343 + }, + { + "epoch": 0.97, + "learning_rate": 4.645494094840275e-08, + "loss": 1.1358, + "step": 40344 + }, + { + "epoch": 0.97, + "learning_rate": 4.637998093170337e-08, + "loss": 1.0527, + "step": 40345 + }, + { + "epoch": 0.97, + "learning_rate": 4.6305081301274425e-08, + "loss": 1.1662, + "step": 40346 + }, + { + "epoch": 0.97, + "learning_rate": 4.6230242057573315e-08, + "loss": 0.9845, + "step": 40347 + }, + { + "epoch": 0.97, + "learning_rate": 4.61554632010508e-08, + "loss": 0.9197, + "step": 40348 + }, + { + "epoch": 0.97, + "learning_rate": 4.608074473216095e-08, + "loss": 0.95, + "step": 40349 + }, + { + "epoch": 0.97, + "learning_rate": 4.600608665136008e-08, + "loss": 1.1381, + "step": 40350 + }, + { + "epoch": 0.97, + "learning_rate": 4.593148895909783e-08, + "loss": 1.2122, + "step": 40351 + }, + { + "epoch": 0.97, + "learning_rate": 4.585695165582715e-08, + "loss": 1.1754, + "step": 40352 + }, + { + "epoch": 0.97, + "learning_rate": 4.578247474200215e-08, + "loss": 1.1436, + "step": 40353 + }, + { + "epoch": 0.97, + "learning_rate": 4.570805821807245e-08, + "loss": 1.0475, + "step": 40354 + }, + { + "epoch": 0.97, + "learning_rate": 4.563370208449214e-08, + "loss": 1.2455, + "step": 40355 + }, + { + "epoch": 0.97, + "learning_rate": 4.555940634170974e-08, + "loss": 0.945, + "step": 40356 + }, + { + "epoch": 0.97, + "learning_rate": 4.548517099017713e-08, + "loss": 1.0826, + "step": 40357 + }, + { + "epoch": 0.97, + "learning_rate": 4.541099603034394e-08, + "loss": 1.1047, + "step": 40358 + }, + { + "epoch": 0.97, + "learning_rate": 4.5336881462660906e-08, + "loss": 1.0854, + "step": 40359 + }, + { + "epoch": 0.97, + "learning_rate": 4.5262827287578804e-08, + "loss": 1.0285, + "step": 40360 + }, + { + "epoch": 0.97, + "learning_rate": 4.5188833505545035e-08, + "loss": 0.9966, + "step": 40361 + }, + { + "epoch": 0.97, + "learning_rate": 4.5114900117008144e-08, + "loss": 1.0685, + "step": 40362 + }, + { + "epoch": 0.97, + "learning_rate": 4.504102712241887e-08, + "loss": 1.1284, + "step": 40363 + }, + { + "epoch": 0.97, + "learning_rate": 4.496721452222352e-08, + "loss": 1.0614, + "step": 40364 + }, + { + "epoch": 0.97, + "learning_rate": 4.489346231687064e-08, + "loss": 1.1519, + "step": 40365 + }, + { + "epoch": 0.97, + "learning_rate": 4.481977050680764e-08, + "loss": 1.1227, + "step": 40366 + }, + { + "epoch": 0.97, + "learning_rate": 4.474613909248082e-08, + "loss": 0.8882, + "step": 40367 + }, + { + "epoch": 0.97, + "learning_rate": 4.467256807433873e-08, + "loss": 0.9852, + "step": 40368 + }, + { + "epoch": 0.97, + "learning_rate": 4.4599057452825444e-08, + "loss": 1.1169, + "step": 40369 + }, + { + "epoch": 0.97, + "learning_rate": 4.452560722838839e-08, + "loss": 1.0702, + "step": 40370 + }, + { + "epoch": 0.97, + "learning_rate": 4.4452217401472765e-08, + "loss": 0.8951, + "step": 40371 + }, + { + "epoch": 0.97, + "learning_rate": 4.437888797252377e-08, + "loss": 0.9366, + "step": 40372 + }, + { + "epoch": 0.97, + "learning_rate": 4.4305618941986595e-08, + "loss": 0.9222, + "step": 40373 + }, + { + "epoch": 0.97, + "learning_rate": 4.423241031030534e-08, + "loss": 1.0094, + "step": 40374 + }, + { + "epoch": 0.97, + "learning_rate": 4.415926207792409e-08, + "loss": 1.2452, + "step": 40375 + }, + { + "epoch": 0.97, + "learning_rate": 4.408617424528805e-08, + "loss": 0.8954, + "step": 40376 + }, + { + "epoch": 0.97, + "learning_rate": 4.401314681283797e-08, + "loss": 1.1487, + "step": 40377 + }, + { + "epoch": 0.97, + "learning_rate": 4.394017978101905e-08, + "loss": 1.0112, + "step": 40378 + }, + { + "epoch": 0.97, + "learning_rate": 4.386727315027317e-08, + "loss": 0.9574, + "step": 40379 + }, + { + "epoch": 0.97, + "learning_rate": 4.3794426921042185e-08, + "loss": 1.0266, + "step": 40380 + }, + { + "epoch": 0.97, + "learning_rate": 4.372164109376908e-08, + "loss": 1.2992, + "step": 40381 + }, + { + "epoch": 0.97, + "learning_rate": 4.364891566889573e-08, + "loss": 1.23, + "step": 40382 + }, + { + "epoch": 0.97, + "learning_rate": 4.3576250646860665e-08, + "loss": 0.8943, + "step": 40383 + }, + { + "epoch": 0.97, + "learning_rate": 4.350364602810797e-08, + "loss": 1.0253, + "step": 40384 + }, + { + "epoch": 0.97, + "learning_rate": 4.343110181307619e-08, + "loss": 0.8319, + "step": 40385 + }, + { + "epoch": 0.97, + "learning_rate": 4.3358618002206086e-08, + "loss": 0.9568, + "step": 40386 + }, + { + "epoch": 0.97, + "learning_rate": 4.328619459593841e-08, + "loss": 1.0842, + "step": 40387 + }, + { + "epoch": 0.97, + "learning_rate": 4.321383159470949e-08, + "loss": 1.0645, + "step": 40388 + }, + { + "epoch": 0.97, + "learning_rate": 4.314152899896118e-08, + "loss": 1.1446, + "step": 40389 + }, + { + "epoch": 0.97, + "learning_rate": 4.3069286809132026e-08, + "loss": 1.1748, + "step": 40390 + }, + { + "epoch": 0.97, + "learning_rate": 4.299710502565835e-08, + "loss": 0.8479, + "step": 40391 + }, + { + "epoch": 0.97, + "learning_rate": 4.292498364897979e-08, + "loss": 0.894, + "step": 40392 + }, + { + "epoch": 0.97, + "learning_rate": 4.285292267953267e-08, + "loss": 1.0301, + "step": 40393 + }, + { + "epoch": 0.97, + "learning_rate": 4.278092211775553e-08, + "loss": 0.948, + "step": 40394 + }, + { + "epoch": 0.97, + "learning_rate": 4.2708981964083576e-08, + "loss": 1.2036, + "step": 40395 + }, + { + "epoch": 0.97, + "learning_rate": 4.263710221895534e-08, + "loss": 1.0032, + "step": 40396 + }, + { + "epoch": 0.97, + "learning_rate": 4.2565282882804924e-08, + "loss": 0.9286, + "step": 40397 + }, + { + "epoch": 0.97, + "learning_rate": 4.2493523956068646e-08, + "loss": 0.767, + "step": 40398 + }, + { + "epoch": 0.97, + "learning_rate": 4.2421825439182826e-08, + "loss": 1.0696, + "step": 40399 + }, + { + "epoch": 0.97, + "learning_rate": 4.235018733258045e-08, + "loss": 1.3439, + "step": 40400 + }, + { + "epoch": 0.97, + "learning_rate": 4.2278609636697834e-08, + "loss": 0.9719, + "step": 40401 + }, + { + "epoch": 0.97, + "learning_rate": 4.220709235196907e-08, + "loss": 0.7461, + "step": 40402 + }, + { + "epoch": 0.97, + "learning_rate": 4.213563547882715e-08, + "loss": 1.0514, + "step": 40403 + }, + { + "epoch": 0.97, + "learning_rate": 4.206423901770618e-08, + "loss": 0.8889, + "step": 40404 + }, + { + "epoch": 0.97, + "learning_rate": 4.199290296904024e-08, + "loss": 1.21, + "step": 40405 + }, + { + "epoch": 0.97, + "learning_rate": 4.1921627333260104e-08, + "loss": 0.9511, + "step": 40406 + }, + { + "epoch": 0.97, + "learning_rate": 4.1850412110798765e-08, + "loss": 0.7985, + "step": 40407 + }, + { + "epoch": 0.97, + "learning_rate": 4.1779257302089204e-08, + "loss": 1.0525, + "step": 40408 + }, + { + "epoch": 0.97, + "learning_rate": 4.17081629075633e-08, + "loss": 1.1958, + "step": 40409 + }, + { + "epoch": 0.97, + "learning_rate": 4.16371289276507e-08, + "loss": 0.9286, + "step": 40410 + }, + { + "epoch": 0.97, + "learning_rate": 4.15661553627833e-08, + "loss": 0.9396, + "step": 40411 + }, + { + "epoch": 0.97, + "learning_rate": 4.149524221339185e-08, + "loss": 0.9843, + "step": 40412 + }, + { + "epoch": 0.97, + "learning_rate": 4.1424389479906014e-08, + "loss": 1.3048, + "step": 40413 + }, + { + "epoch": 0.97, + "learning_rate": 4.1353597162756555e-08, + "loss": 0.6882, + "step": 40414 + }, + { + "epoch": 0.97, + "learning_rate": 4.1282865262372015e-08, + "loss": 1.076, + "step": 40415 + }, + { + "epoch": 0.97, + "learning_rate": 4.1212193779182065e-08, + "loss": 1.1145, + "step": 40416 + }, + { + "epoch": 0.97, + "learning_rate": 4.114158271361635e-08, + "loss": 0.9195, + "step": 40417 + }, + { + "epoch": 0.97, + "learning_rate": 4.107103206610119e-08, + "loss": 1.1494, + "step": 40418 + }, + { + "epoch": 0.97, + "learning_rate": 4.100054183706514e-08, + "loss": 1.154, + "step": 40419 + }, + { + "epoch": 0.97, + "learning_rate": 4.093011202693675e-08, + "loss": 0.8182, + "step": 40420 + }, + { + "epoch": 0.97, + "learning_rate": 4.0859742636142344e-08, + "loss": 1.051, + "step": 40421 + }, + { + "epoch": 0.97, + "learning_rate": 4.078943366511046e-08, + "loss": 0.984, + "step": 40422 + }, + { + "epoch": 0.97, + "learning_rate": 4.071918511426631e-08, + "loss": 0.9964, + "step": 40423 + }, + { + "epoch": 0.97, + "learning_rate": 4.064899698403513e-08, + "loss": 0.8492, + "step": 40424 + }, + { + "epoch": 0.97, + "learning_rate": 4.0578869274845446e-08, + "loss": 0.9732, + "step": 40425 + }, + { + "epoch": 0.97, + "learning_rate": 4.050880198712026e-08, + "loss": 1.1346, + "step": 40426 + }, + { + "epoch": 0.97, + "learning_rate": 4.043879512128479e-08, + "loss": 1.1359, + "step": 40427 + }, + { + "epoch": 0.97, + "learning_rate": 4.036884867776647e-08, + "loss": 1.0847, + "step": 40428 + }, + { + "epoch": 0.97, + "learning_rate": 4.0298962656986074e-08, + "loss": 1.0471, + "step": 40429 + }, + { + "epoch": 0.97, + "learning_rate": 4.0229137059369925e-08, + "loss": 1.1299, + "step": 40430 + }, + { + "epoch": 0.97, + "learning_rate": 4.015937188534102e-08, + "loss": 1.1775, + "step": 40431 + }, + { + "epoch": 0.97, + "learning_rate": 4.008966713532236e-08, + "loss": 1.0149, + "step": 40432 + }, + { + "epoch": 0.97, + "learning_rate": 4.002002280973693e-08, + "loss": 0.7429, + "step": 40433 + }, + { + "epoch": 0.97, + "learning_rate": 3.995043890900663e-08, + "loss": 1.1224, + "step": 40434 + }, + { + "epoch": 0.97, + "learning_rate": 3.988091543355554e-08, + "loss": 1.0615, + "step": 40435 + }, + { + "epoch": 0.97, + "learning_rate": 3.981145238380335e-08, + "loss": 0.931, + "step": 40436 + }, + { + "epoch": 0.97, + "learning_rate": 3.974204976017193e-08, + "loss": 0.9121, + "step": 40437 + }, + { + "epoch": 0.97, + "learning_rate": 3.967270756308428e-08, + "loss": 1.2409, + "step": 40438 + }, + { + "epoch": 0.97, + "learning_rate": 3.960342579295784e-08, + "loss": 1.1453, + "step": 40439 + }, + { + "epoch": 0.97, + "learning_rate": 3.953420445021561e-08, + "loss": 1.2489, + "step": 40440 + }, + { + "epoch": 0.97, + "learning_rate": 3.946504353527614e-08, + "loss": 1.246, + "step": 40441 + }, + { + "epoch": 0.97, + "learning_rate": 3.939594304855909e-08, + "loss": 1.0314, + "step": 40442 + }, + { + "epoch": 0.97, + "learning_rate": 3.932690299048414e-08, + "loss": 1.0878, + "step": 40443 + }, + { + "epoch": 0.97, + "learning_rate": 3.9257923361469826e-08, + "loss": 0.8631, + "step": 40444 + }, + { + "epoch": 0.97, + "learning_rate": 3.918900416193583e-08, + "loss": 0.9269, + "step": 40445 + }, + { + "epoch": 0.97, + "learning_rate": 3.912014539229847e-08, + "loss": 0.9845, + "step": 40446 + }, + { + "epoch": 0.97, + "learning_rate": 3.9051347052976304e-08, + "loss": 0.8261, + "step": 40447 + }, + { + "epoch": 0.97, + "learning_rate": 3.8982609144385676e-08, + "loss": 1.1007, + "step": 40448 + }, + { + "epoch": 0.97, + "learning_rate": 3.891393166694623e-08, + "loss": 0.8554, + "step": 40449 + }, + { + "epoch": 0.97, + "learning_rate": 3.884531462107211e-08, + "loss": 1.0493, + "step": 40450 + }, + { + "epoch": 0.97, + "learning_rate": 3.877675800718073e-08, + "loss": 1.156, + "step": 40451 + }, + { + "epoch": 0.97, + "learning_rate": 3.870826182568732e-08, + "loss": 1.2831, + "step": 40452 + }, + { + "epoch": 0.97, + "learning_rate": 3.8639826077008223e-08, + "loss": 1.0608, + "step": 40453 + }, + { + "epoch": 0.97, + "learning_rate": 3.857145076155755e-08, + "loss": 1.2483, + "step": 40454 + }, + { + "epoch": 0.97, + "learning_rate": 3.850313587975163e-08, + "loss": 1.121, + "step": 40455 + }, + { + "epoch": 0.97, + "learning_rate": 3.843488143200347e-08, + "loss": 0.912, + "step": 40456 + }, + { + "epoch": 0.97, + "learning_rate": 3.836668741872829e-08, + "loss": 1.0128, + "step": 40457 + }, + { + "epoch": 0.97, + "learning_rate": 3.8298553840339095e-08, + "loss": 1.0722, + "step": 40458 + }, + { + "epoch": 0.97, + "learning_rate": 3.823048069724999e-08, + "loss": 1.3346, + "step": 40459 + }, + { + "epoch": 0.97, + "learning_rate": 3.8162467989872885e-08, + "loss": 0.9629, + "step": 40460 + }, + { + "epoch": 0.97, + "learning_rate": 3.809451571862077e-08, + "loss": 0.8986, + "step": 40461 + }, + { + "epoch": 0.97, + "learning_rate": 3.8026623883906656e-08, + "loss": 1.0372, + "step": 40462 + }, + { + "epoch": 0.97, + "learning_rate": 3.7958792486141314e-08, + "loss": 0.9528, + "step": 40463 + }, + { + "epoch": 0.97, + "learning_rate": 3.789102152573665e-08, + "loss": 0.9849, + "step": 40464 + }, + { + "epoch": 0.97, + "learning_rate": 3.7823311003104545e-08, + "loss": 1.0232, + "step": 40465 + }, + { + "epoch": 0.97, + "learning_rate": 3.775566091865468e-08, + "loss": 1.0296, + "step": 40466 + }, + { + "epoch": 0.97, + "learning_rate": 3.768807127279783e-08, + "loss": 0.9935, + "step": 40467 + }, + { + "epoch": 0.97, + "learning_rate": 3.762054206594479e-08, + "loss": 0.8536, + "step": 40468 + }, + { + "epoch": 0.97, + "learning_rate": 3.7553073298504104e-08, + "loss": 1.0378, + "step": 40469 + }, + { + "epoch": 0.97, + "learning_rate": 3.748566497088546e-08, + "loss": 1.2027, + "step": 40470 + }, + { + "epoch": 0.97, + "learning_rate": 3.7418317083498526e-08, + "loss": 1.0643, + "step": 40471 + }, + { + "epoch": 0.97, + "learning_rate": 3.7351029636750745e-08, + "loss": 1.0806, + "step": 40472 + }, + { + "epoch": 0.97, + "learning_rate": 3.72838026310518e-08, + "loss": 1.1822, + "step": 40473 + }, + { + "epoch": 0.97, + "learning_rate": 3.721663606680692e-08, + "loss": 0.9436, + "step": 40474 + }, + { + "epoch": 0.97, + "learning_rate": 3.714952994442689e-08, + "loss": 0.999, + "step": 40475 + }, + { + "epoch": 0.97, + "learning_rate": 3.708248426431693e-08, + "loss": 0.7887, + "step": 40476 + }, + { + "epoch": 0.97, + "learning_rate": 3.70154990268834e-08, + "loss": 1.0377, + "step": 40477 + }, + { + "epoch": 0.97, + "learning_rate": 3.694857423253373e-08, + "loss": 1.1549, + "step": 40478 + }, + { + "epoch": 0.97, + "learning_rate": 3.6881709881674274e-08, + "loss": 1.0635, + "step": 40479 + }, + { + "epoch": 0.97, + "learning_rate": 3.681490597471027e-08, + "loss": 0.7268, + "step": 40480 + }, + { + "epoch": 0.97, + "learning_rate": 3.674816251204583e-08, + "loss": 1.2162, + "step": 40481 + }, + { + "epoch": 0.97, + "learning_rate": 3.668147949408729e-08, + "loss": 1.0678, + "step": 40482 + }, + { + "epoch": 0.97, + "learning_rate": 3.6614856921237673e-08, + "loss": 0.8186, + "step": 40483 + }, + { + "epoch": 0.97, + "learning_rate": 3.6548294793903314e-08, + "loss": 1.3217, + "step": 40484 + }, + { + "epoch": 0.97, + "learning_rate": 3.648179311248612e-08, + "loss": 0.8809, + "step": 40485 + }, + { + "epoch": 0.97, + "learning_rate": 3.641535187739131e-08, + "loss": 1.1245, + "step": 40486 + }, + { + "epoch": 0.97, + "learning_rate": 3.6348971089020804e-08, + "loss": 0.9513, + "step": 40487 + }, + { + "epoch": 0.97, + "learning_rate": 3.6282650747776484e-08, + "loss": 0.8598, + "step": 40488 + }, + { + "epoch": 0.97, + "learning_rate": 3.621639085406248e-08, + "loss": 0.9491, + "step": 40489 + }, + { + "epoch": 0.97, + "learning_rate": 3.6150191408279576e-08, + "loss": 0.9609, + "step": 40490 + }, + { + "epoch": 0.97, + "learning_rate": 3.6084052410830796e-08, + "loss": 0.5943, + "step": 40491 + }, + { + "epoch": 0.97, + "learning_rate": 3.6017973862115805e-08, + "loss": 1.2665, + "step": 40492 + }, + { + "epoch": 0.97, + "learning_rate": 3.595195576253541e-08, + "loss": 1.3487, + "step": 40493 + }, + { + "epoch": 0.97, + "learning_rate": 3.5885998112491495e-08, + "loss": 0.8582, + "step": 40494 + }, + { + "epoch": 0.97, + "learning_rate": 3.582010091238375e-08, + "loss": 1.0433, + "step": 40495 + }, + { + "epoch": 0.97, + "learning_rate": 3.575426416261074e-08, + "loss": 1.0361, + "step": 40496 + }, + { + "epoch": 0.97, + "learning_rate": 3.568848786357437e-08, + "loss": 1.2808, + "step": 40497 + }, + { + "epoch": 0.97, + "learning_rate": 3.562277201567099e-08, + "loss": 0.947, + "step": 40498 + }, + { + "epoch": 0.97, + "learning_rate": 3.555711661930139e-08, + "loss": 1.296, + "step": 40499 + }, + { + "epoch": 0.97, + "learning_rate": 3.5491521674861914e-08, + "loss": 1.0176, + "step": 40500 + }, + { + "epoch": 0.97, + "learning_rate": 3.542598718275225e-08, + "loss": 0.8724, + "step": 40501 + }, + { + "epoch": 0.97, + "learning_rate": 3.5360513143368745e-08, + "loss": 0.9621, + "step": 40502 + }, + { + "epoch": 0.97, + "learning_rate": 3.529509955710997e-08, + "loss": 1.0098, + "step": 40503 + }, + { + "epoch": 0.97, + "learning_rate": 3.522974642437116e-08, + "loss": 1.0302, + "step": 40504 + }, + { + "epoch": 0.97, + "learning_rate": 3.516445374555089e-08, + "loss": 1.1727, + "step": 40505 + }, + { + "epoch": 0.97, + "learning_rate": 3.509922152104328e-08, + "loss": 1.146, + "step": 40506 + }, + { + "epoch": 0.97, + "learning_rate": 3.503404975124469e-08, + "loss": 0.9971, + "step": 40507 + }, + { + "epoch": 0.97, + "learning_rate": 3.496893843655036e-08, + "loss": 1.1978, + "step": 40508 + }, + { + "epoch": 0.97, + "learning_rate": 3.4903887577355524e-08, + "loss": 1.0201, + "step": 40509 + }, + { + "epoch": 0.97, + "learning_rate": 3.483889717405542e-08, + "loss": 1.014, + "step": 40510 + }, + { + "epoch": 0.97, + "learning_rate": 3.477396722704418e-08, + "loss": 1.0442, + "step": 40511 + }, + { + "epoch": 0.97, + "learning_rate": 3.4709097736714825e-08, + "loss": 1.0071, + "step": 40512 + }, + { + "epoch": 0.97, + "learning_rate": 3.464428870346148e-08, + "loss": 1.2138, + "step": 40513 + }, + { + "epoch": 0.97, + "learning_rate": 3.457954012767606e-08, + "loss": 1.0949, + "step": 40514 + }, + { + "epoch": 0.97, + "learning_rate": 3.45148520097538e-08, + "loss": 1.1002, + "step": 40515 + }, + { + "epoch": 0.97, + "learning_rate": 3.445022435008549e-08, + "loss": 0.8265, + "step": 40516 + }, + { + "epoch": 0.97, + "learning_rate": 3.438565714906417e-08, + "loss": 0.7891, + "step": 40517 + }, + { + "epoch": 0.97, + "learning_rate": 3.4321150407080615e-08, + "loss": 0.8898, + "step": 40518 + }, + { + "epoch": 0.97, + "learning_rate": 3.425670412452675e-08, + "loss": 1.152, + "step": 40519 + }, + { + "epoch": 0.97, + "learning_rate": 3.419231830179337e-08, + "loss": 1.0186, + "step": 40520 + }, + { + "epoch": 0.97, + "learning_rate": 3.412799293927127e-08, + "loss": 0.9796, + "step": 40521 + }, + { + "epoch": 0.97, + "learning_rate": 3.406372803735014e-08, + "loss": 1.0797, + "step": 40522 + }, + { + "epoch": 0.97, + "learning_rate": 3.399952359641967e-08, + "loss": 0.8404, + "step": 40523 + }, + { + "epoch": 0.97, + "learning_rate": 3.3935379616870654e-08, + "loss": 0.9443, + "step": 40524 + }, + { + "epoch": 0.97, + "learning_rate": 3.387129609909168e-08, + "loss": 1.1616, + "step": 40525 + }, + { + "epoch": 0.97, + "learning_rate": 3.380727304347131e-08, + "loss": 0.9689, + "step": 40526 + }, + { + "epoch": 0.97, + "learning_rate": 3.3743310450398136e-08, + "loss": 1.0589, + "step": 40527 + }, + { + "epoch": 0.97, + "learning_rate": 3.367940832025962e-08, + "loss": 0.8252, + "step": 40528 + }, + { + "epoch": 0.97, + "learning_rate": 3.3615566653444343e-08, + "loss": 0.8321, + "step": 40529 + }, + { + "epoch": 0.97, + "learning_rate": 3.355178545033866e-08, + "loss": 0.706, + "step": 40530 + }, + { + "epoch": 0.97, + "learning_rate": 3.348806471133004e-08, + "loss": 1.2121, + "step": 40531 + }, + { + "epoch": 0.97, + "learning_rate": 3.342440443680595e-08, + "loss": 0.8854, + "step": 40532 + }, + { + "epoch": 0.97, + "learning_rate": 3.336080462715052e-08, + "loss": 0.9043, + "step": 40533 + }, + { + "epoch": 0.97, + "learning_rate": 3.3297265282751237e-08, + "loss": 1.1161, + "step": 40534 + }, + { + "epoch": 0.97, + "learning_rate": 3.3233786403993327e-08, + "loss": 1.1094, + "step": 40535 + }, + { + "epoch": 0.97, + "learning_rate": 3.317036799126205e-08, + "loss": 1.3515, + "step": 40536 + }, + { + "epoch": 0.97, + "learning_rate": 3.3107010044941546e-08, + "loss": 1.1219, + "step": 40537 + }, + { + "epoch": 0.97, + "learning_rate": 3.304371256541594e-08, + "loss": 0.7979, + "step": 40538 + }, + { + "epoch": 0.97, + "learning_rate": 3.298047555307049e-08, + "loss": 1.1085, + "step": 40539 + }, + { + "epoch": 0.97, + "learning_rate": 3.291729900828822e-08, + "loss": 1.2896, + "step": 40540 + }, + { + "epoch": 0.97, + "learning_rate": 3.285418293145104e-08, + "loss": 0.7933, + "step": 40541 + }, + { + "epoch": 0.97, + "learning_rate": 3.2791127322944205e-08, + "loss": 0.7552, + "step": 40542 + }, + { + "epoch": 0.98, + "learning_rate": 3.272813218314852e-08, + "loss": 0.9645, + "step": 40543 + }, + { + "epoch": 0.98, + "learning_rate": 3.266519751244812e-08, + "loss": 1.0143, + "step": 40544 + }, + { + "epoch": 0.98, + "learning_rate": 3.2602323311222705e-08, + "loss": 1.0006, + "step": 40545 + }, + { + "epoch": 0.98, + "learning_rate": 3.253950957985419e-08, + "loss": 1.0552, + "step": 40546 + }, + { + "epoch": 0.98, + "learning_rate": 3.247675631872338e-08, + "loss": 1.1653, + "step": 40547 + }, + { + "epoch": 0.98, + "learning_rate": 3.2414063528213305e-08, + "loss": 1.0628, + "step": 40548 + }, + { + "epoch": 0.98, + "learning_rate": 3.235143120870143e-08, + "loss": 1.103, + "step": 40549 + }, + { + "epoch": 0.98, + "learning_rate": 3.228885936056858e-08, + "loss": 0.6254, + "step": 40550 + }, + { + "epoch": 0.98, + "learning_rate": 3.222634798419555e-08, + "loss": 0.8821, + "step": 40551 + }, + { + "epoch": 0.98, + "learning_rate": 3.216389707995982e-08, + "loss": 0.9416, + "step": 40552 + }, + { + "epoch": 0.98, + "learning_rate": 3.210150664824219e-08, + "loss": 0.6652, + "step": 40553 + }, + { + "epoch": 0.98, + "learning_rate": 3.2039176689420137e-08, + "loss": 0.6624, + "step": 40554 + }, + { + "epoch": 0.98, + "learning_rate": 3.1976907203871145e-08, + "loss": 1.0364, + "step": 40555 + }, + { + "epoch": 0.98, + "learning_rate": 3.191469819197379e-08, + "loss": 1.1479, + "step": 40556 + }, + { + "epoch": 0.98, + "learning_rate": 3.1852549654105556e-08, + "loss": 1.0199, + "step": 40557 + }, + { + "epoch": 0.98, + "learning_rate": 3.1790461590642804e-08, + "loss": 1.0156, + "step": 40558 + }, + { + "epoch": 0.98, + "learning_rate": 3.1728434001963016e-08, + "loss": 1.0249, + "step": 40559 + }, + { + "epoch": 0.98, + "learning_rate": 3.166646688844144e-08, + "loss": 1.1615, + "step": 40560 + }, + { + "epoch": 0.98, + "learning_rate": 3.160456025045555e-08, + "loss": 0.8571, + "step": 40561 + }, + { + "epoch": 0.98, + "learning_rate": 3.15427140883795e-08, + "loss": 0.8663, + "step": 40562 + }, + { + "epoch": 0.98, + "learning_rate": 3.148092840258965e-08, + "loss": 1.1456, + "step": 40563 + }, + { + "epoch": 0.98, + "learning_rate": 3.141920319346015e-08, + "loss": 0.8878, + "step": 40564 + }, + { + "epoch": 0.98, + "learning_rate": 3.1357538461365135e-08, + "loss": 0.9585, + "step": 40565 + }, + { + "epoch": 0.98, + "learning_rate": 3.129593420667987e-08, + "loss": 0.9331, + "step": 40566 + }, + { + "epoch": 0.98, + "learning_rate": 3.123439042977627e-08, + "loss": 0.9215, + "step": 40567 + }, + { + "epoch": 0.98, + "learning_rate": 3.117290713102961e-08, + "loss": 1.1123, + "step": 40568 + }, + { + "epoch": 0.98, + "learning_rate": 3.11114843108129e-08, + "loss": 1.0369, + "step": 40569 + }, + { + "epoch": 0.98, + "learning_rate": 3.105012196949697e-08, + "loss": 0.8233, + "step": 40570 + }, + { + "epoch": 0.98, + "learning_rate": 3.0988820107454854e-08, + "loss": 1.2503, + "step": 40571 + }, + { + "epoch": 0.98, + "learning_rate": 3.092757872505958e-08, + "loss": 0.858, + "step": 40572 + }, + { + "epoch": 0.98, + "learning_rate": 3.086639782268086e-08, + "loss": 0.9932, + "step": 40573 + }, + { + "epoch": 0.98, + "learning_rate": 3.0805277400691724e-08, + "loss": 0.8274, + "step": 40574 + }, + { + "epoch": 0.98, + "learning_rate": 3.074421745946188e-08, + "loss": 1.0835, + "step": 40575 + }, + { + "epoch": 0.98, + "learning_rate": 3.068321799936103e-08, + "loss": 1.0041, + "step": 40576 + }, + { + "epoch": 0.98, + "learning_rate": 3.062227902076109e-08, + "loss": 1.072, + "step": 40577 + }, + { + "epoch": 0.98, + "learning_rate": 3.056140052403067e-08, + "loss": 0.952, + "step": 40578 + }, + { + "epoch": 0.98, + "learning_rate": 3.050058250953947e-08, + "loss": 0.9598, + "step": 40579 + }, + { + "epoch": 0.98, + "learning_rate": 3.043982497765607e-08, + "loss": 1.3081, + "step": 40580 + }, + { + "epoch": 0.98, + "learning_rate": 3.037912792874909e-08, + "loss": 1.0857, + "step": 40581 + }, + { + "epoch": 0.98, + "learning_rate": 3.0318491363187096e-08, + "loss": 0.9231, + "step": 40582 + }, + { + "epoch": 0.98, + "learning_rate": 3.0257915281337594e-08, + "loss": 1.0479, + "step": 40583 + }, + { + "epoch": 0.98, + "learning_rate": 3.019739968356805e-08, + "loss": 1.1619, + "step": 40584 + }, + { + "epoch": 0.98, + "learning_rate": 3.0136944570247074e-08, + "loss": 0.9642, + "step": 40585 + }, + { + "epoch": 0.98, + "learning_rate": 3.007654994173881e-08, + "loss": 0.8663, + "step": 40586 + }, + { + "epoch": 0.98, + "learning_rate": 3.0016215798411854e-08, + "loss": 1.0237, + "step": 40587 + }, + { + "epoch": 0.98, + "learning_rate": 2.9955942140631466e-08, + "loss": 1.0866, + "step": 40588 + }, + { + "epoch": 0.98, + "learning_rate": 2.9895728968762915e-08, + "loss": 1.0535, + "step": 40589 + }, + { + "epoch": 0.98, + "learning_rate": 2.983557628317257e-08, + "loss": 0.8513, + "step": 40590 + }, + { + "epoch": 0.98, + "learning_rate": 2.9775484084223483e-08, + "loss": 1.1388, + "step": 40591 + }, + { + "epoch": 0.98, + "learning_rate": 2.9715452372282015e-08, + "loss": 0.7426, + "step": 40592 + }, + { + "epoch": 0.98, + "learning_rate": 2.965548114771233e-08, + "loss": 1.1796, + "step": 40593 + }, + { + "epoch": 0.98, + "learning_rate": 2.959557041087635e-08, + "loss": 1.1367, + "step": 40594 + }, + { + "epoch": 0.98, + "learning_rate": 2.953572016213935e-08, + "loss": 1.0626, + "step": 40595 + }, + { + "epoch": 0.98, + "learning_rate": 2.9475930401864362e-08, + "loss": 0.9961, + "step": 40596 + }, + { + "epoch": 0.98, + "learning_rate": 2.9416201130414435e-08, + "loss": 1.06, + "step": 40597 + }, + { + "epoch": 0.98, + "learning_rate": 2.935653234814928e-08, + "loss": 1.0836, + "step": 40598 + }, + { + "epoch": 0.98, + "learning_rate": 2.929692405543416e-08, + "loss": 0.966, + "step": 40599 + }, + { + "epoch": 0.98, + "learning_rate": 2.923737625262879e-08, + "loss": 1.0053, + "step": 40600 + }, + { + "epoch": 0.98, + "learning_rate": 2.91778889400951e-08, + "loss": 1.0811, + "step": 40601 + }, + { + "epoch": 0.98, + "learning_rate": 2.9118462118193913e-08, + "loss": 1.1295, + "step": 40602 + }, + { + "epoch": 0.98, + "learning_rate": 2.9059095787286052e-08, + "loss": 1.0044, + "step": 40603 + }, + { + "epoch": 0.98, + "learning_rate": 2.899978994773123e-08, + "loss": 1.0671, + "step": 40604 + }, + { + "epoch": 0.98, + "learning_rate": 2.8940544599889152e-08, + "loss": 0.9429, + "step": 40605 + }, + { + "epoch": 0.98, + "learning_rate": 2.888135974411954e-08, + "loss": 1.1207, + "step": 40606 + }, + { + "epoch": 0.98, + "learning_rate": 2.8822235380780995e-08, + "loss": 1.1217, + "step": 40607 + }, + { + "epoch": 0.98, + "learning_rate": 2.8763171510233224e-08, + "loss": 0.9898, + "step": 40608 + }, + { + "epoch": 0.98, + "learning_rate": 2.870416813283372e-08, + "loss": 0.9875, + "step": 40609 + }, + { + "epoch": 0.98, + "learning_rate": 2.8645225248939977e-08, + "loss": 1.0709, + "step": 40610 + }, + { + "epoch": 0.98, + "learning_rate": 2.8586342858910598e-08, + "loss": 1.0591, + "step": 40611 + }, + { + "epoch": 0.98, + "learning_rate": 2.852752096310307e-08, + "loss": 1.1998, + "step": 40612 + }, + { + "epoch": 0.98, + "learning_rate": 2.846875956187378e-08, + "loss": 0.6818, + "step": 40613 + }, + { + "epoch": 0.98, + "learning_rate": 2.8410058655578e-08, + "loss": 0.7792, + "step": 40614 + }, + { + "epoch": 0.98, + "learning_rate": 2.8351418244573214e-08, + "loss": 1.0606, + "step": 40615 + }, + { + "epoch": 0.98, + "learning_rate": 2.829283832921581e-08, + "loss": 1.1438, + "step": 40616 + }, + { + "epoch": 0.98, + "learning_rate": 2.823431890985884e-08, + "loss": 1.0234, + "step": 40617 + }, + { + "epoch": 0.98, + "learning_rate": 2.8175859986859788e-08, + "loss": 1.0915, + "step": 40618 + }, + { + "epoch": 0.98, + "learning_rate": 2.8117461560571713e-08, + "loss": 1.2553, + "step": 40619 + }, + { + "epoch": 0.98, + "learning_rate": 2.8059123631349884e-08, + "loss": 0.7677, + "step": 40620 + }, + { + "epoch": 0.98, + "learning_rate": 2.800084619954846e-08, + "loss": 1.1931, + "step": 40621 + }, + { + "epoch": 0.98, + "learning_rate": 2.7942629265519383e-08, + "loss": 1.1457, + "step": 40622 + }, + { + "epoch": 0.98, + "learning_rate": 2.7884472829616817e-08, + "loss": 1.0576, + "step": 40623 + }, + { + "epoch": 0.98, + "learning_rate": 2.7826376892193807e-08, + "loss": 0.7666, + "step": 40624 + }, + { + "epoch": 0.98, + "learning_rate": 2.77683414536023e-08, + "loss": 0.9753, + "step": 40625 + }, + { + "epoch": 0.98, + "learning_rate": 2.7710366514195343e-08, + "loss": 0.9049, + "step": 40626 + }, + { + "epoch": 0.98, + "learning_rate": 2.7652452074322655e-08, + "loss": 1.0333, + "step": 40627 + }, + { + "epoch": 0.98, + "learning_rate": 2.75945981343384e-08, + "loss": 1.1657, + "step": 40628 + }, + { + "epoch": 0.98, + "learning_rate": 2.753680469459119e-08, + "loss": 1.3428, + "step": 40629 + }, + { + "epoch": 0.98, + "learning_rate": 2.7479071755431853e-08, + "loss": 1.1357, + "step": 40630 + }, + { + "epoch": 0.98, + "learning_rate": 2.7421399317212326e-08, + "loss": 1.0717, + "step": 40631 + }, + { + "epoch": 0.98, + "learning_rate": 2.7363787380281226e-08, + "loss": 1.3719, + "step": 40632 + }, + { + "epoch": 0.98, + "learning_rate": 2.730623594498827e-08, + "loss": 1.0819, + "step": 40633 + }, + { + "epoch": 0.98, + "learning_rate": 2.7248745011683175e-08, + "loss": 1.0867, + "step": 40634 + }, + { + "epoch": 0.98, + "learning_rate": 2.7191314580714555e-08, + "loss": 1.1317, + "step": 40635 + }, + { + "epoch": 0.98, + "learning_rate": 2.7133944652429912e-08, + "loss": 1.0659, + "step": 40636 + }, + { + "epoch": 0.98, + "learning_rate": 2.707663522717674e-08, + "loss": 0.6873, + "step": 40637 + }, + { + "epoch": 0.98, + "learning_rate": 2.7019386305304763e-08, + "loss": 1.1101, + "step": 40638 + }, + { + "epoch": 0.98, + "learning_rate": 2.6962197887161477e-08, + "loss": 0.7887, + "step": 40639 + }, + { + "epoch": 0.98, + "learning_rate": 2.6905069973091058e-08, + "loss": 1.1731, + "step": 40640 + }, + { + "epoch": 0.98, + "learning_rate": 2.6848002563442112e-08, + "loss": 1.0872, + "step": 40641 + }, + { + "epoch": 0.98, + "learning_rate": 2.6790995658561026e-08, + "loss": 1.1305, + "step": 40642 + }, + { + "epoch": 0.98, + "learning_rate": 2.6734049258793082e-08, + "loss": 1.1914, + "step": 40643 + }, + { + "epoch": 0.98, + "learning_rate": 2.667716336448356e-08, + "loss": 0.9601, + "step": 40644 + }, + { + "epoch": 0.98, + "learning_rate": 2.662033797597774e-08, + "loss": 0.7373, + "step": 40645 + }, + { + "epoch": 0.98, + "learning_rate": 2.65635730936209e-08, + "loss": 1.0367, + "step": 40646 + }, + { + "epoch": 0.98, + "learning_rate": 2.6506868717756097e-08, + "loss": 1.0161, + "step": 40647 + }, + { + "epoch": 0.98, + "learning_rate": 2.6450224848728612e-08, + "loss": 1.0908, + "step": 40648 + }, + { + "epoch": 0.98, + "learning_rate": 2.6393641486882614e-08, + "loss": 1.1065, + "step": 40649 + }, + { + "epoch": 0.98, + "learning_rate": 2.6337118632558945e-08, + "loss": 0.8684, + "step": 40650 + }, + { + "epoch": 0.98, + "learning_rate": 2.628065628610288e-08, + "loss": 1.0943, + "step": 40651 + }, + { + "epoch": 0.98, + "learning_rate": 2.6224254447856368e-08, + "loss": 1.058, + "step": 40652 + }, + { + "epoch": 0.98, + "learning_rate": 2.616791311816136e-08, + "loss": 1.0336, + "step": 40653 + }, + { + "epoch": 0.98, + "learning_rate": 2.6111632297358693e-08, + "loss": 1.1978, + "step": 40654 + }, + { + "epoch": 0.98, + "learning_rate": 2.6055411985791422e-08, + "loss": 0.9826, + "step": 40655 + }, + { + "epoch": 0.98, + "learning_rate": 2.5999252183799285e-08, + "loss": 0.9296, + "step": 40656 + }, + { + "epoch": 0.98, + "learning_rate": 2.594315289172422e-08, + "loss": 1.0782, + "step": 40657 + }, + { + "epoch": 0.98, + "learning_rate": 2.5887114109907073e-08, + "loss": 1.0484, + "step": 40658 + }, + { + "epoch": 0.98, + "learning_rate": 2.583113583868535e-08, + "loss": 0.9827, + "step": 40659 + }, + { + "epoch": 0.98, + "learning_rate": 2.5775218078399888e-08, + "loss": 1.1663, + "step": 40660 + }, + { + "epoch": 0.98, + "learning_rate": 2.5719360829390415e-08, + "loss": 1.0915, + "step": 40661 + }, + { + "epoch": 0.98, + "learning_rate": 2.566356409199555e-08, + "loss": 1.0595, + "step": 40662 + }, + { + "epoch": 0.98, + "learning_rate": 2.5607827866553915e-08, + "loss": 0.8371, + "step": 40663 + }, + { + "epoch": 0.98, + "learning_rate": 2.555215215340412e-08, + "loss": 0.9641, + "step": 40664 + }, + { + "epoch": 0.98, + "learning_rate": 2.5496536952882566e-08, + "loss": 0.9642, + "step": 40665 + }, + { + "epoch": 0.98, + "learning_rate": 2.5440982265326764e-08, + "loss": 0.6797, + "step": 40666 + }, + { + "epoch": 0.98, + "learning_rate": 2.5385488091075327e-08, + "loss": 1.3357, + "step": 40667 + }, + { + "epoch": 0.98, + "learning_rate": 2.5330054430463547e-08, + "loss": 0.7842, + "step": 40668 + }, + { + "epoch": 0.98, + "learning_rate": 2.5274681283827817e-08, + "loss": 1.0919, + "step": 40669 + }, + { + "epoch": 0.98, + "learning_rate": 2.5219368651504537e-08, + "loss": 0.7209, + "step": 40670 + }, + { + "epoch": 0.98, + "learning_rate": 2.5164116533830107e-08, + "loss": 1.1325, + "step": 40671 + }, + { + "epoch": 0.98, + "learning_rate": 2.510892493113759e-08, + "loss": 1.0993, + "step": 40672 + }, + { + "epoch": 0.98, + "learning_rate": 2.5053793843763384e-08, + "loss": 1.0505, + "step": 40673 + }, + { + "epoch": 0.98, + "learning_rate": 2.499872327204167e-08, + "loss": 1.2686, + "step": 40674 + }, + { + "epoch": 0.98, + "learning_rate": 2.494371321630662e-08, + "loss": 1.1245, + "step": 40675 + }, + { + "epoch": 0.98, + "learning_rate": 2.4888763676891305e-08, + "loss": 1.0769, + "step": 40676 + }, + { + "epoch": 0.98, + "learning_rate": 2.4833874654129898e-08, + "loss": 1.0836, + "step": 40677 + }, + { + "epoch": 0.98, + "learning_rate": 2.4779046148354358e-08, + "loss": 0.8538, + "step": 40678 + }, + { + "epoch": 0.98, + "learning_rate": 2.4724278159898863e-08, + "loss": 1.1133, + "step": 40679 + }, + { + "epoch": 0.98, + "learning_rate": 2.4669570689094258e-08, + "loss": 0.9508, + "step": 40680 + }, + { + "epoch": 0.98, + "learning_rate": 2.461492373627361e-08, + "loss": 1.0792, + "step": 40681 + }, + { + "epoch": 0.98, + "learning_rate": 2.4560337301767767e-08, + "loss": 0.894, + "step": 40682 + }, + { + "epoch": 0.98, + "learning_rate": 2.4505811385907573e-08, + "loss": 1.0248, + "step": 40683 + }, + { + "epoch": 0.98, + "learning_rate": 2.4451345989024987e-08, + "loss": 0.8632, + "step": 40684 + }, + { + "epoch": 0.98, + "learning_rate": 2.4396941111449744e-08, + "loss": 1.1887, + "step": 40685 + }, + { + "epoch": 0.98, + "learning_rate": 2.4342596753510473e-08, + "loss": 1.1366, + "step": 40686 + }, + { + "epoch": 0.98, + "learning_rate": 2.4288312915539125e-08, + "loss": 1.0242, + "step": 40687 + }, + { + "epoch": 0.98, + "learning_rate": 2.423408959786433e-08, + "loss": 1.1743, + "step": 40688 + }, + { + "epoch": 0.98, + "learning_rate": 2.4179926800814712e-08, + "loss": 0.871, + "step": 40689 + }, + { + "epoch": 0.98, + "learning_rate": 2.41258245247189e-08, + "loss": 1.1192, + "step": 40690 + }, + { + "epoch": 0.98, + "learning_rate": 2.4071782769905517e-08, + "loss": 0.8444, + "step": 40691 + }, + { + "epoch": 0.98, + "learning_rate": 2.4017801536702078e-08, + "loss": 0.9362, + "step": 40692 + }, + { + "epoch": 0.98, + "learning_rate": 2.3963880825436102e-08, + "loss": 0.9967, + "step": 40693 + }, + { + "epoch": 0.98, + "learning_rate": 2.3910020636433994e-08, + "loss": 1.2605, + "step": 40694 + }, + { + "epoch": 0.98, + "learning_rate": 2.385622097002438e-08, + "loss": 1.0235, + "step": 40695 + }, + { + "epoch": 0.98, + "learning_rate": 2.3802481826532554e-08, + "loss": 0.9817, + "step": 40696 + }, + { + "epoch": 0.98, + "learning_rate": 2.374880320628381e-08, + "loss": 1.1449, + "step": 40697 + }, + { + "epoch": 0.98, + "learning_rate": 2.369518510960456e-08, + "loss": 1.2056, + "step": 40698 + }, + { + "epoch": 0.98, + "learning_rate": 2.364162753682009e-08, + "loss": 1.1066, + "step": 40699 + }, + { + "epoch": 0.98, + "learning_rate": 2.35881304882557e-08, + "loss": 1.0155, + "step": 40700 + }, + { + "epoch": 0.98, + "learning_rate": 2.3534693964235576e-08, + "loss": 1.2144, + "step": 40701 + }, + { + "epoch": 0.98, + "learning_rate": 2.3481317965083905e-08, + "loss": 0.9648, + "step": 40702 + }, + { + "epoch": 0.98, + "learning_rate": 2.342800249112376e-08, + "loss": 0.9699, + "step": 40703 + }, + { + "epoch": 0.98, + "learning_rate": 2.337474754267932e-08, + "loss": 0.9293, + "step": 40704 + }, + { + "epoch": 0.98, + "learning_rate": 2.332155312007478e-08, + "loss": 1.0885, + "step": 40705 + }, + { + "epoch": 0.98, + "learning_rate": 2.326841922363099e-08, + "loss": 1.0114, + "step": 40706 + }, + { + "epoch": 0.98, + "learning_rate": 2.3215345853671023e-08, + "loss": 0.9997, + "step": 40707 + }, + { + "epoch": 0.98, + "learning_rate": 2.3162333010516847e-08, + "loss": 0.7362, + "step": 40708 + }, + { + "epoch": 0.98, + "learning_rate": 2.3109380694489315e-08, + "loss": 1.3205, + "step": 40709 + }, + { + "epoch": 0.98, + "learning_rate": 2.30564889059115e-08, + "loss": 0.8763, + "step": 40710 + }, + { + "epoch": 0.98, + "learning_rate": 2.300365764510315e-08, + "loss": 1.0345, + "step": 40711 + }, + { + "epoch": 0.98, + "learning_rate": 2.2950886912385117e-08, + "loss": 0.7938, + "step": 40712 + }, + { + "epoch": 0.98, + "learning_rate": 2.2898176708076035e-08, + "loss": 0.8315, + "step": 40713 + }, + { + "epoch": 0.98, + "learning_rate": 2.284552703249787e-08, + "loss": 1.1155, + "step": 40714 + }, + { + "epoch": 0.98, + "learning_rate": 2.2792937885969256e-08, + "loss": 1.1341, + "step": 40715 + }, + { + "epoch": 0.98, + "learning_rate": 2.2740409268808826e-08, + "loss": 0.8822, + "step": 40716 + }, + { + "epoch": 0.98, + "learning_rate": 2.2687941181336326e-08, + "loss": 1.3239, + "step": 40717 + }, + { + "epoch": 0.98, + "learning_rate": 2.2635533623868166e-08, + "loss": 0.942, + "step": 40718 + }, + { + "epoch": 0.98, + "learning_rate": 2.2583186596724094e-08, + "loss": 0.9968, + "step": 40719 + }, + { + "epoch": 0.98, + "learning_rate": 2.253090010022052e-08, + "loss": 0.8707, + "step": 40720 + }, + { + "epoch": 0.98, + "learning_rate": 2.247867413467497e-08, + "loss": 0.8593, + "step": 40721 + }, + { + "epoch": 0.98, + "learning_rate": 2.242650870040497e-08, + "loss": 1.1778, + "step": 40722 + }, + { + "epoch": 0.98, + "learning_rate": 2.237440379772693e-08, + "loss": 0.9311, + "step": 40723 + }, + { + "epoch": 0.98, + "learning_rate": 2.2322359426955042e-08, + "loss": 0.838, + "step": 40724 + }, + { + "epoch": 0.98, + "learning_rate": 2.2270375588407945e-08, + "loss": 1.183, + "step": 40725 + }, + { + "epoch": 0.98, + "learning_rate": 2.2218452282398716e-08, + "loss": 0.9229, + "step": 40726 + }, + { + "epoch": 0.98, + "learning_rate": 2.2166589509243776e-08, + "loss": 0.9955, + "step": 40727 + }, + { + "epoch": 0.98, + "learning_rate": 2.2114787269256198e-08, + "loss": 0.9972, + "step": 40728 + }, + { + "epoch": 0.98, + "learning_rate": 2.2063045562752406e-08, + "loss": 0.979, + "step": 40729 + }, + { + "epoch": 0.98, + "learning_rate": 2.2011364390044367e-08, + "loss": 1.1819, + "step": 40730 + }, + { + "epoch": 0.98, + "learning_rate": 2.1959743751447384e-08, + "loss": 1.0784, + "step": 40731 + }, + { + "epoch": 0.98, + "learning_rate": 2.190818364727232e-08, + "loss": 0.9812, + "step": 40732 + }, + { + "epoch": 0.98, + "learning_rate": 2.1856684077834477e-08, + "loss": 0.8134, + "step": 40733 + }, + { + "epoch": 0.98, + "learning_rate": 2.180524504344472e-08, + "loss": 1.0849, + "step": 40734 + }, + { + "epoch": 0.98, + "learning_rate": 2.1753866544416134e-08, + "loss": 1.0995, + "step": 40735 + }, + { + "epoch": 0.98, + "learning_rate": 2.1702548581058468e-08, + "loss": 0.9146, + "step": 40736 + }, + { + "epoch": 0.98, + "learning_rate": 2.1651291153684804e-08, + "loss": 1.199, + "step": 40737 + }, + { + "epoch": 0.98, + "learning_rate": 2.160009426260712e-08, + "loss": 1.0907, + "step": 40738 + }, + { + "epoch": 0.98, + "learning_rate": 2.154895790813294e-08, + "loss": 1.0386, + "step": 40739 + }, + { + "epoch": 0.98, + "learning_rate": 2.149788209057424e-08, + "loss": 1.0343, + "step": 40740 + }, + { + "epoch": 0.98, + "learning_rate": 2.1446866810241883e-08, + "loss": 1.1231, + "step": 40741 + }, + { + "epoch": 0.98, + "learning_rate": 2.1395912067443404e-08, + "loss": 0.9378, + "step": 40742 + }, + { + "epoch": 0.98, + "learning_rate": 2.134501786248966e-08, + "loss": 1.1008, + "step": 40743 + }, + { + "epoch": 0.98, + "learning_rate": 2.1294184195688183e-08, + "loss": 1.1082, + "step": 40744 + }, + { + "epoch": 0.98, + "learning_rate": 2.124341106734651e-08, + "loss": 1.074, + "step": 40745 + }, + { + "epoch": 0.98, + "learning_rate": 2.1192698477775498e-08, + "loss": 0.681, + "step": 40746 + }, + { + "epoch": 0.98, + "learning_rate": 2.114204642728046e-08, + "loss": 0.7163, + "step": 40747 + }, + { + "epoch": 0.98, + "learning_rate": 2.109145491617004e-08, + "loss": 1.0212, + "step": 40748 + }, + { + "epoch": 0.98, + "learning_rate": 2.1040923944750657e-08, + "loss": 0.9776, + "step": 40749 + }, + { + "epoch": 0.98, + "learning_rate": 2.0990453513327625e-08, + "loss": 0.9254, + "step": 40750 + }, + { + "epoch": 0.98, + "learning_rate": 2.0940043622209583e-08, + "loss": 1.0655, + "step": 40751 + }, + { + "epoch": 0.98, + "learning_rate": 2.0889694271700732e-08, + "loss": 0.9744, + "step": 40752 + }, + { + "epoch": 0.98, + "learning_rate": 2.0839405462106387e-08, + "loss": 0.9291, + "step": 40753 + }, + { + "epoch": 0.98, + "learning_rate": 2.0789177193732967e-08, + "loss": 1.1149, + "step": 40754 + }, + { + "epoch": 0.98, + "learning_rate": 2.0739009466883565e-08, + "loss": 1.0039, + "step": 40755 + }, + { + "epoch": 0.98, + "learning_rate": 2.0688902281863487e-08, + "loss": 1.2283, + "step": 40756 + }, + { + "epoch": 0.98, + "learning_rate": 2.063885563897694e-08, + "loss": 1.078, + "step": 40757 + }, + { + "epoch": 0.98, + "learning_rate": 2.0588869538527012e-08, + "loss": 1.2424, + "step": 40758 + }, + { + "epoch": 0.98, + "learning_rate": 2.0538943980816793e-08, + "loss": 0.9646, + "step": 40759 + }, + { + "epoch": 0.98, + "learning_rate": 2.0489078966149378e-08, + "loss": 1.3169, + "step": 40760 + }, + { + "epoch": 0.98, + "learning_rate": 2.0439274494827853e-08, + "loss": 1.0345, + "step": 40761 + }, + { + "epoch": 0.98, + "learning_rate": 2.03895305671542e-08, + "loss": 1.186, + "step": 40762 + }, + { + "epoch": 0.98, + "learning_rate": 2.0339847183429296e-08, + "loss": 0.7535, + "step": 40763 + }, + { + "epoch": 0.98, + "learning_rate": 2.0290224343955113e-08, + "loss": 0.9057, + "step": 40764 + }, + { + "epoch": 0.98, + "learning_rate": 2.024066204903252e-08, + "loss": 0.9656, + "step": 40765 + }, + { + "epoch": 0.98, + "learning_rate": 2.0191160298963508e-08, + "loss": 0.9591, + "step": 40766 + }, + { + "epoch": 0.98, + "learning_rate": 2.014171909404672e-08, + "loss": 1.0377, + "step": 40767 + }, + { + "epoch": 0.98, + "learning_rate": 2.0092338434583026e-08, + "loss": 1.2401, + "step": 40768 + }, + { + "epoch": 0.98, + "learning_rate": 2.0043018320871078e-08, + "loss": 1.1114, + "step": 40769 + }, + { + "epoch": 0.98, + "learning_rate": 1.9993758753211746e-08, + "loss": 0.8562, + "step": 40770 + }, + { + "epoch": 0.98, + "learning_rate": 1.994455973190146e-08, + "loss": 1.3473, + "step": 40771 + }, + { + "epoch": 0.98, + "learning_rate": 1.9895421257241086e-08, + "loss": 0.9104, + "step": 40772 + }, + { + "epoch": 0.98, + "learning_rate": 1.9846343329527062e-08, + "loss": 1.0999, + "step": 40773 + }, + { + "epoch": 0.98, + "learning_rate": 1.9797325949058032e-08, + "loss": 1.1683, + "step": 40774 + }, + { + "epoch": 0.98, + "learning_rate": 1.9748369116131537e-08, + "loss": 1.1851, + "step": 40775 + }, + { + "epoch": 0.98, + "learning_rate": 1.9699472831044007e-08, + "loss": 1.2308, + "step": 40776 + }, + { + "epoch": 0.98, + "learning_rate": 1.9650637094091873e-08, + "loss": 1.0798, + "step": 40777 + }, + { + "epoch": 0.98, + "learning_rate": 1.9601861905572673e-08, + "loss": 1.0944, + "step": 40778 + }, + { + "epoch": 0.98, + "learning_rate": 1.9553147265780615e-08, + "loss": 1.0787, + "step": 40779 + }, + { + "epoch": 0.98, + "learning_rate": 1.950449317501213e-08, + "loss": 0.8808, + "step": 40780 + }, + { + "epoch": 0.98, + "learning_rate": 1.9455899633562536e-08, + "loss": 0.7212, + "step": 40781 + }, + { + "epoch": 0.98, + "learning_rate": 1.9407366641727153e-08, + "loss": 0.8797, + "step": 40782 + }, + { + "epoch": 0.98, + "learning_rate": 1.9358894199797974e-08, + "loss": 1.0925, + "step": 40783 + }, + { + "epoch": 0.98, + "learning_rate": 1.931048230807253e-08, + "loss": 1.07, + "step": 40784 + }, + { + "epoch": 0.98, + "learning_rate": 1.9262130966841707e-08, + "loss": 0.9756, + "step": 40785 + }, + { + "epoch": 0.98, + "learning_rate": 1.9213840176400823e-08, + "loss": 0.796, + "step": 40786 + }, + { + "epoch": 0.98, + "learning_rate": 1.916560993704075e-08, + "loss": 1.0144, + "step": 40787 + }, + { + "epoch": 0.98, + "learning_rate": 1.9117440249055708e-08, + "loss": 0.8203, + "step": 40788 + }, + { + "epoch": 0.98, + "learning_rate": 1.9069331112738788e-08, + "loss": 1.0953, + "step": 40789 + }, + { + "epoch": 0.98, + "learning_rate": 1.902128252837865e-08, + "loss": 1.0073, + "step": 40790 + }, + { + "epoch": 0.98, + "learning_rate": 1.89732944962695e-08, + "loss": 1.0476, + "step": 40791 + }, + { + "epoch": 0.98, + "learning_rate": 1.8925367016702223e-08, + "loss": 0.9142, + "step": 40792 + }, + { + "epoch": 0.98, + "learning_rate": 1.8877500089966584e-08, + "loss": 0.9617, + "step": 40793 + }, + { + "epoch": 0.98, + "learning_rate": 1.8829693716353457e-08, + "loss": 1.1605, + "step": 40794 + }, + { + "epoch": 0.98, + "learning_rate": 1.8781947896152618e-08, + "loss": 0.9578, + "step": 40795 + }, + { + "epoch": 0.98, + "learning_rate": 1.873426262965383e-08, + "loss": 1.1633, + "step": 40796 + }, + { + "epoch": 0.98, + "learning_rate": 1.8686637917147976e-08, + "loss": 0.9513, + "step": 40797 + }, + { + "epoch": 0.98, + "learning_rate": 1.863907375892149e-08, + "loss": 1.1325, + "step": 40798 + }, + { + "epoch": 0.98, + "learning_rate": 1.859157015526303e-08, + "loss": 0.9419, + "step": 40799 + }, + { + "epoch": 0.98, + "learning_rate": 1.8544127106463473e-08, + "loss": 0.993, + "step": 40800 + }, + { + "epoch": 0.98, + "learning_rate": 1.849674461280815e-08, + "loss": 1.001, + "step": 40801 + }, + { + "epoch": 0.98, + "learning_rate": 1.8449422674585715e-08, + "loss": 1.0565, + "step": 40802 + }, + { + "epoch": 0.98, + "learning_rate": 1.84021612920815e-08, + "loss": 0.8601, + "step": 40803 + }, + { + "epoch": 0.98, + "learning_rate": 1.835496046558527e-08, + "loss": 0.9924, + "step": 40804 + }, + { + "epoch": 0.98, + "learning_rate": 1.8307820195380133e-08, + "loss": 1.2449, + "step": 40805 + }, + { + "epoch": 0.98, + "learning_rate": 1.8260740481754747e-08, + "loss": 1.1108, + "step": 40806 + }, + { + "epoch": 0.98, + "learning_rate": 1.8213721324993327e-08, + "loss": 1.0925, + "step": 40807 + }, + { + "epoch": 0.98, + "learning_rate": 1.8166762725381205e-08, + "loss": 1.0624, + "step": 40808 + }, + { + "epoch": 0.98, + "learning_rate": 1.81198646832037e-08, + "loss": 0.8958, + "step": 40809 + }, + { + "epoch": 0.98, + "learning_rate": 1.8073027198745042e-08, + "loss": 1.1019, + "step": 40810 + }, + { + "epoch": 0.98, + "learning_rate": 1.8026250272288325e-08, + "loss": 1.0974, + "step": 40811 + }, + { + "epoch": 0.98, + "learning_rate": 1.7979533904119993e-08, + "loss": 1.037, + "step": 40812 + }, + { + "epoch": 0.98, + "learning_rate": 1.7932878094520933e-08, + "loss": 1.1412, + "step": 40813 + }, + { + "epoch": 0.98, + "learning_rate": 1.788628284377647e-08, + "loss": 1.1001, + "step": 40814 + }, + { + "epoch": 0.98, + "learning_rate": 1.783974815216638e-08, + "loss": 1.258, + "step": 40815 + }, + { + "epoch": 0.98, + "learning_rate": 1.7793274019975992e-08, + "loss": 0.9561, + "step": 40816 + }, + { + "epoch": 0.98, + "learning_rate": 1.7746860447485082e-08, + "loss": 1.1152, + "step": 40817 + }, + { + "epoch": 0.98, + "learning_rate": 1.7700507434975644e-08, + "loss": 0.8959, + "step": 40818 + }, + { + "epoch": 0.98, + "learning_rate": 1.765421498272968e-08, + "loss": 0.8525, + "step": 40819 + }, + { + "epoch": 0.98, + "learning_rate": 1.760798309102807e-08, + "loss": 0.8266, + "step": 40820 + }, + { + "epoch": 0.98, + "learning_rate": 1.7561811760150592e-08, + "loss": 0.9245, + "step": 40821 + }, + { + "epoch": 0.98, + "learning_rate": 1.7515700990377028e-08, + "loss": 0.746, + "step": 40822 + }, + { + "epoch": 0.98, + "learning_rate": 1.7469650781988256e-08, + "loss": 1.258, + "step": 40823 + }, + { + "epoch": 0.98, + "learning_rate": 1.7423661135262948e-08, + "loss": 1.0312, + "step": 40824 + }, + { + "epoch": 0.98, + "learning_rate": 1.7377732050479768e-08, + "loss": 0.96, + "step": 40825 + }, + { + "epoch": 0.98, + "learning_rate": 1.733186352791849e-08, + "loss": 1.057, + "step": 40826 + }, + { + "epoch": 0.98, + "learning_rate": 1.7286055567856676e-08, + "loss": 1.0419, + "step": 40827 + }, + { + "epoch": 0.98, + "learning_rate": 1.7240308170571875e-08, + "loss": 0.9838, + "step": 40828 + }, + { + "epoch": 0.98, + "learning_rate": 1.7194621336341644e-08, + "loss": 0.9033, + "step": 40829 + }, + { + "epoch": 0.98, + "learning_rate": 1.7148995065444653e-08, + "loss": 1.3792, + "step": 40830 + }, + { + "epoch": 0.98, + "learning_rate": 1.710342935815623e-08, + "loss": 1.2609, + "step": 40831 + }, + { + "epoch": 0.98, + "learning_rate": 1.705792421475283e-08, + "loss": 1.0325, + "step": 40832 + }, + { + "epoch": 0.98, + "learning_rate": 1.701247963551089e-08, + "loss": 0.9805, + "step": 40833 + }, + { + "epoch": 0.98, + "learning_rate": 1.6967095620705753e-08, + "loss": 1.0348, + "step": 40834 + }, + { + "epoch": 0.98, + "learning_rate": 1.692177217061386e-08, + "loss": 0.8522, + "step": 40835 + }, + { + "epoch": 0.98, + "learning_rate": 1.6876509285509434e-08, + "loss": 0.9179, + "step": 40836 + }, + { + "epoch": 0.98, + "learning_rate": 1.6831306965666706e-08, + "loss": 0.974, + "step": 40837 + }, + { + "epoch": 0.98, + "learning_rate": 1.6786165211359895e-08, + "loss": 1.1254, + "step": 40838 + }, + { + "epoch": 0.98, + "learning_rate": 1.674108402286434e-08, + "loss": 1.0535, + "step": 40839 + }, + { + "epoch": 0.98, + "learning_rate": 1.6696063400450936e-08, + "loss": 0.9932, + "step": 40840 + }, + { + "epoch": 0.98, + "learning_rate": 1.6651103344395015e-08, + "loss": 1.1533, + "step": 40841 + }, + { + "epoch": 0.98, + "learning_rate": 1.6606203854969694e-08, + "loss": 1.0447, + "step": 40842 + }, + { + "epoch": 0.98, + "learning_rate": 1.6561364932444757e-08, + "loss": 0.9372, + "step": 40843 + }, + { + "epoch": 0.98, + "learning_rate": 1.6516586577094428e-08, + "loss": 1.0756, + "step": 40844 + }, + { + "epoch": 0.98, + "learning_rate": 1.6471868789189604e-08, + "loss": 0.897, + "step": 40845 + }, + { + "epoch": 0.98, + "learning_rate": 1.6427211569001178e-08, + "loss": 0.7478, + "step": 40846 + }, + { + "epoch": 0.98, + "learning_rate": 1.6382614916802265e-08, + "loss": 1.0798, + "step": 40847 + }, + { + "epoch": 0.98, + "learning_rate": 1.6338078832860427e-08, + "loss": 0.9244, + "step": 40848 + }, + { + "epoch": 0.98, + "learning_rate": 1.629360331744656e-08, + "loss": 1.0314, + "step": 40849 + }, + { + "epoch": 0.98, + "learning_rate": 1.6249188370832668e-08, + "loss": 0.9556, + "step": 40850 + }, + { + "epoch": 0.98, + "learning_rate": 1.6204833993285208e-08, + "loss": 1.1015, + "step": 40851 + }, + { + "epoch": 0.98, + "learning_rate": 1.6160540185075068e-08, + "loss": 0.9537, + "step": 40852 + }, + { + "epoch": 0.98, + "learning_rate": 1.6116306946470928e-08, + "loss": 1.1293, + "step": 40853 + }, + { + "epoch": 0.98, + "learning_rate": 1.6072134277740346e-08, + "loss": 1.1987, + "step": 40854 + }, + { + "epoch": 0.98, + "learning_rate": 1.6028022179152e-08, + "loss": 0.7903, + "step": 40855 + }, + { + "epoch": 0.98, + "learning_rate": 1.5983970650973457e-08, + "loss": 1.0955, + "step": 40856 + }, + { + "epoch": 0.98, + "learning_rate": 1.5939979693472275e-08, + "loss": 1.1045, + "step": 40857 + }, + { + "epoch": 0.98, + "learning_rate": 1.5896049306913797e-08, + "loss": 1.0064, + "step": 40858 + }, + { + "epoch": 0.98, + "learning_rate": 1.58521794915667e-08, + "loss": 1.076, + "step": 40859 + }, + { + "epoch": 0.98, + "learning_rate": 1.5808370247695215e-08, + "loss": 1.162, + "step": 40860 + }, + { + "epoch": 0.98, + "learning_rate": 1.5764621575565797e-08, + "loss": 1.0611, + "step": 40861 + }, + { + "epoch": 0.98, + "learning_rate": 1.572093347544379e-08, + "loss": 1.0788, + "step": 40862 + }, + { + "epoch": 0.98, + "learning_rate": 1.5677305947595643e-08, + "loss": 1.1509, + "step": 40863 + }, + { + "epoch": 0.98, + "learning_rate": 1.5633738992283377e-08, + "loss": 1.1065, + "step": 40864 + }, + { + "epoch": 0.98, + "learning_rate": 1.5590232609773437e-08, + "loss": 0.9828, + "step": 40865 + }, + { + "epoch": 0.98, + "learning_rate": 1.554678680033006e-08, + "loss": 1.0155, + "step": 40866 + }, + { + "epoch": 0.98, + "learning_rate": 1.5503401564214147e-08, + "loss": 1.1661, + "step": 40867 + }, + { + "epoch": 0.98, + "learning_rate": 1.5460076901691046e-08, + "loss": 1.083, + "step": 40868 + }, + { + "epoch": 0.98, + "learning_rate": 1.5416812813023874e-08, + "loss": 1.1704, + "step": 40869 + }, + { + "epoch": 0.98, + "learning_rate": 1.5373609298473536e-08, + "loss": 0.9148, + "step": 40870 + }, + { + "epoch": 0.98, + "learning_rate": 1.5330466358303154e-08, + "loss": 1.1026, + "step": 40871 + }, + { + "epoch": 0.98, + "learning_rate": 1.528738399277474e-08, + "loss": 1.1413, + "step": 40872 + }, + { + "epoch": 0.98, + "learning_rate": 1.5244362202149198e-08, + "loss": 1.1211, + "step": 40873 + }, + { + "epoch": 0.98, + "learning_rate": 1.5201400986687432e-08, + "loss": 1.0617, + "step": 40874 + }, + { + "epoch": 0.98, + "learning_rate": 1.515850034665034e-08, + "loss": 0.964, + "step": 40875 + }, + { + "epoch": 0.98, + "learning_rate": 1.5115660282297717e-08, + "loss": 1.2595, + "step": 40876 + }, + { + "epoch": 0.98, + "learning_rate": 1.5072880793889356e-08, + "loss": 1.0553, + "step": 40877 + }, + { + "epoch": 0.98, + "learning_rate": 1.5030161881686155e-08, + "loss": 1.143, + "step": 40878 + }, + { + "epoch": 0.98, + "learning_rate": 1.498750354594569e-08, + "loss": 1.0406, + "step": 40879 + }, + { + "epoch": 0.98, + "learning_rate": 1.4944905786927756e-08, + "loss": 1.0952, + "step": 40880 + }, + { + "epoch": 0.98, + "learning_rate": 1.4902368604891028e-08, + "loss": 1.2469, + "step": 40881 + }, + { + "epoch": 0.98, + "learning_rate": 1.4859892000091968e-08, + "loss": 0.8348, + "step": 40882 + }, + { + "epoch": 0.98, + "learning_rate": 1.4817475972789264e-08, + "loss": 1.1221, + "step": 40883 + }, + { + "epoch": 0.98, + "learning_rate": 1.4775120523241593e-08, + "loss": 0.9634, + "step": 40884 + }, + { + "epoch": 0.98, + "learning_rate": 1.4732825651703198e-08, + "loss": 0.7876, + "step": 40885 + }, + { + "epoch": 0.98, + "learning_rate": 1.469059135843276e-08, + "loss": 1.1595, + "step": 40886 + }, + { + "epoch": 0.98, + "learning_rate": 1.464841764368563e-08, + "loss": 0.8553, + "step": 40887 + }, + { + "epoch": 0.98, + "learning_rate": 1.4606304507718271e-08, + "loss": 1.2403, + "step": 40888 + }, + { + "epoch": 0.98, + "learning_rate": 1.4564251950784924e-08, + "loss": 1.0478, + "step": 40889 + }, + { + "epoch": 0.98, + "learning_rate": 1.452225997314205e-08, + "loss": 1.1585, + "step": 40890 + }, + { + "epoch": 0.98, + "learning_rate": 1.448032857504389e-08, + "loss": 1.1587, + "step": 40891 + }, + { + "epoch": 0.98, + "learning_rate": 1.4438457756744683e-08, + "loss": 1.1029, + "step": 40892 + }, + { + "epoch": 0.98, + "learning_rate": 1.4396647518498675e-08, + "loss": 0.9784, + "step": 40893 + }, + { + "epoch": 0.98, + "learning_rate": 1.4354897860558992e-08, + "loss": 1.1307, + "step": 40894 + }, + { + "epoch": 0.98, + "learning_rate": 1.4313208783179877e-08, + "loss": 0.9668, + "step": 40895 + }, + { + "epoch": 0.98, + "learning_rate": 1.4271580286614461e-08, + "loss": 1.1664, + "step": 40896 + }, + { + "epoch": 0.98, + "learning_rate": 1.4230012371113655e-08, + "loss": 1.0237, + "step": 40897 + }, + { + "epoch": 0.98, + "learning_rate": 1.4188505036930588e-08, + "loss": 1.0787, + "step": 40898 + }, + { + "epoch": 0.98, + "learning_rate": 1.4147058284317284e-08, + "loss": 0.8381, + "step": 40899 + }, + { + "epoch": 0.98, + "learning_rate": 1.4105672113524648e-08, + "loss": 1.0119, + "step": 40900 + }, + { + "epoch": 0.98, + "learning_rate": 1.4064346524804705e-08, + "loss": 0.8144, + "step": 40901 + }, + { + "epoch": 0.98, + "learning_rate": 1.4023081518406145e-08, + "loss": 1.1355, + "step": 40902 + }, + { + "epoch": 0.98, + "learning_rate": 1.3981877094582097e-08, + "loss": 1.0084, + "step": 40903 + }, + { + "epoch": 0.98, + "learning_rate": 1.3940733253581251e-08, + "loss": 1.0605, + "step": 40904 + }, + { + "epoch": 0.98, + "learning_rate": 1.3899649995652299e-08, + "loss": 0.9803, + "step": 40905 + }, + { + "epoch": 0.98, + "learning_rate": 1.3858627321045037e-08, + "loss": 0.8666, + "step": 40906 + }, + { + "epoch": 0.98, + "learning_rate": 1.381766523000927e-08, + "loss": 1.0383, + "step": 40907 + }, + { + "epoch": 0.98, + "learning_rate": 1.3776763722792574e-08, + "loss": 1.0275, + "step": 40908 + }, + { + "epoch": 0.98, + "learning_rate": 1.3735922799643642e-08, + "loss": 1.1177, + "step": 40909 + }, + { + "epoch": 0.98, + "learning_rate": 1.369514246081005e-08, + "loss": 0.8388, + "step": 40910 + }, + { + "epoch": 0.98, + "learning_rate": 1.3654422706539382e-08, + "loss": 1.1323, + "step": 40911 + }, + { + "epoch": 0.98, + "learning_rate": 1.3613763537078106e-08, + "loss": 1.0154, + "step": 40912 + }, + { + "epoch": 0.98, + "learning_rate": 1.357316495267269e-08, + "loss": 0.9372, + "step": 40913 + }, + { + "epoch": 0.98, + "learning_rate": 1.3532626953569606e-08, + "loss": 0.9599, + "step": 40914 + }, + { + "epoch": 0.98, + "learning_rate": 1.3492149540016431e-08, + "loss": 1.0641, + "step": 40915 + }, + { + "epoch": 0.98, + "learning_rate": 1.3451732712256305e-08, + "loss": 1.13, + "step": 40916 + }, + { + "epoch": 0.98, + "learning_rate": 1.3411376470535697e-08, + "loss": 1.0998, + "step": 40917 + }, + { + "epoch": 0.98, + "learning_rate": 1.3371080815099969e-08, + "loss": 1.2263, + "step": 40918 + }, + { + "epoch": 0.98, + "learning_rate": 1.3330845746192256e-08, + "loss": 1.092, + "step": 40919 + }, + { + "epoch": 0.98, + "learning_rate": 1.329067126405792e-08, + "loss": 1.1719, + "step": 40920 + }, + { + "epoch": 0.98, + "learning_rate": 1.3250557368938988e-08, + "loss": 1.054, + "step": 40921 + }, + { + "epoch": 0.98, + "learning_rate": 1.321050406108082e-08, + "loss": 1.0225, + "step": 40922 + }, + { + "epoch": 0.98, + "learning_rate": 1.3170511340725444e-08, + "loss": 1.0899, + "step": 40923 + }, + { + "epoch": 0.98, + "learning_rate": 1.313057920811489e-08, + "loss": 1.079, + "step": 40924 + }, + { + "epoch": 0.98, + "learning_rate": 1.3090707663492297e-08, + "loss": 1.2823, + "step": 40925 + }, + { + "epoch": 0.98, + "learning_rate": 1.3050896707099692e-08, + "loss": 1.0953, + "step": 40926 + }, + { + "epoch": 0.98, + "learning_rate": 1.3011146339177994e-08, + "loss": 1.1536, + "step": 40927 + }, + { + "epoch": 0.98, + "learning_rate": 1.297145655996812e-08, + "loss": 1.4468, + "step": 40928 + }, + { + "epoch": 0.98, + "learning_rate": 1.2931827369712103e-08, + "loss": 1.2091, + "step": 40929 + }, + { + "epoch": 0.98, + "learning_rate": 1.2892258768649746e-08, + "loss": 1.0776, + "step": 40930 + }, + { + "epoch": 0.98, + "learning_rate": 1.285275075701975e-08, + "loss": 1.2015, + "step": 40931 + }, + { + "epoch": 0.98, + "learning_rate": 1.2813303335064143e-08, + "loss": 0.957, + "step": 40932 + }, + { + "epoch": 0.98, + "learning_rate": 1.2773916503020512e-08, + "loss": 1.1793, + "step": 40933 + }, + { + "epoch": 0.98, + "learning_rate": 1.2734590261128665e-08, + "loss": 1.0083, + "step": 40934 + }, + { + "epoch": 0.98, + "learning_rate": 1.2695324609627302e-08, + "loss": 0.995, + "step": 40935 + }, + { + "epoch": 0.98, + "learning_rate": 1.2656119548752898e-08, + "loss": 0.9025, + "step": 40936 + }, + { + "epoch": 0.98, + "learning_rate": 1.2616975078746373e-08, + "loss": 0.8128, + "step": 40937 + }, + { + "epoch": 0.98, + "learning_rate": 1.2577891199841986e-08, + "loss": 1.3082, + "step": 40938 + }, + { + "epoch": 0.98, + "learning_rate": 1.2538867912279539e-08, + "loss": 0.9422, + "step": 40939 + }, + { + "epoch": 0.98, + "learning_rate": 1.2499905216293295e-08, + "loss": 0.9726, + "step": 40940 + }, + { + "epoch": 0.98, + "learning_rate": 1.246100311212195e-08, + "loss": 0.8396, + "step": 40941 + }, + { + "epoch": 0.98, + "learning_rate": 1.242216159999976e-08, + "loss": 0.9282, + "step": 40942 + }, + { + "epoch": 0.98, + "learning_rate": 1.2383380680164314e-08, + "loss": 0.9244, + "step": 40943 + }, + { + "epoch": 0.98, + "learning_rate": 1.2344660352848758e-08, + "loss": 1.1501, + "step": 40944 + }, + { + "epoch": 0.98, + "learning_rate": 1.230600061828846e-08, + "loss": 0.9264, + "step": 40945 + }, + { + "epoch": 0.98, + "learning_rate": 1.2267401476718787e-08, + "loss": 1.0079, + "step": 40946 + }, + { + "epoch": 0.98, + "learning_rate": 1.2228862928373997e-08, + "loss": 0.8989, + "step": 40947 + }, + { + "epoch": 0.98, + "learning_rate": 1.2190384973487234e-08, + "loss": 1.0864, + "step": 40948 + }, + { + "epoch": 0.98, + "learning_rate": 1.2151967612291649e-08, + "loss": 0.9974, + "step": 40949 + }, + { + "epoch": 0.98, + "learning_rate": 1.2113610845021495e-08, + "loss": 0.9967, + "step": 40950 + }, + { + "epoch": 0.98, + "learning_rate": 1.207531467190881e-08, + "loss": 1.0236, + "step": 40951 + }, + { + "epoch": 0.98, + "learning_rate": 1.2037079093185632e-08, + "loss": 1.1931, + "step": 40952 + }, + { + "epoch": 0.98, + "learning_rate": 1.1998904109083998e-08, + "loss": 0.9802, + "step": 40953 + }, + { + "epoch": 0.98, + "learning_rate": 1.1960789719835941e-08, + "loss": 1.225, + "step": 40954 + }, + { + "epoch": 0.98, + "learning_rate": 1.1922735925672391e-08, + "loss": 0.9488, + "step": 40955 + }, + { + "epoch": 0.98, + "learning_rate": 1.1884742726824272e-08, + "loss": 1.1925, + "step": 40956 + }, + { + "epoch": 0.98, + "learning_rate": 1.1846810123521401e-08, + "loss": 1.0575, + "step": 40957 + }, + { + "epoch": 0.98, + "learning_rate": 1.1808938115995815e-08, + "loss": 0.9108, + "step": 40958 + }, + { + "epoch": 0.99, + "learning_rate": 1.177112670447511e-08, + "loss": 0.9404, + "step": 40959 + }, + { + "epoch": 0.99, + "learning_rate": 1.1733375889189102e-08, + "loss": 1.2929, + "step": 40960 + }, + { + "epoch": 0.99, + "learning_rate": 1.1695685670368718e-08, + "loss": 1.0419, + "step": 40961 + }, + { + "epoch": 0.99, + "learning_rate": 1.1658056048240441e-08, + "loss": 1.0071, + "step": 40962 + }, + { + "epoch": 0.99, + "learning_rate": 1.1620487023032979e-08, + "loss": 1.0149, + "step": 40963 + }, + { + "epoch": 0.99, + "learning_rate": 1.1582978594975037e-08, + "loss": 0.928, + "step": 40964 + }, + { + "epoch": 0.99, + "learning_rate": 1.154553076429421e-08, + "loss": 1.0465, + "step": 40965 + }, + { + "epoch": 0.99, + "learning_rate": 1.1508143531216986e-08, + "loss": 1.0382, + "step": 40966 + }, + { + "epoch": 0.99, + "learning_rate": 1.1470816895969849e-08, + "loss": 1.0012, + "step": 40967 + }, + { + "epoch": 0.99, + "learning_rate": 1.1433550858780396e-08, + "loss": 0.9815, + "step": 40968 + }, + { + "epoch": 0.99, + "learning_rate": 1.1396345419874e-08, + "loss": 0.9977, + "step": 40969 + }, + { + "epoch": 0.99, + "learning_rate": 1.1359200579476038e-08, + "loss": 0.9999, + "step": 40970 + }, + { + "epoch": 0.99, + "learning_rate": 1.1322116337812994e-08, + "loss": 0.6502, + "step": 40971 + }, + { + "epoch": 0.99, + "learning_rate": 1.1285092695109134e-08, + "loss": 0.8626, + "step": 40972 + }, + { + "epoch": 0.99, + "learning_rate": 1.1248129651588724e-08, + "loss": 1.1772, + "step": 40973 + }, + { + "epoch": 0.99, + "learning_rate": 1.1211227207477139e-08, + "loss": 1.1826, + "step": 40974 + }, + { + "epoch": 0.99, + "learning_rate": 1.1174385362996421e-08, + "loss": 0.9395, + "step": 40975 + }, + { + "epoch": 0.99, + "learning_rate": 1.1137604118371947e-08, + "loss": 0.96, + "step": 40976 + }, + { + "epoch": 0.99, + "learning_rate": 1.1100883473825763e-08, + "loss": 0.806, + "step": 40977 + }, + { + "epoch": 0.99, + "learning_rate": 1.1064223429581022e-08, + "loss": 0.621, + "step": 40978 + }, + { + "epoch": 0.99, + "learning_rate": 1.102762398586088e-08, + "loss": 1.1695, + "step": 40979 + }, + { + "epoch": 0.99, + "learning_rate": 1.099108514288627e-08, + "loss": 1.0033, + "step": 40980 + }, + { + "epoch": 0.99, + "learning_rate": 1.0954606900878128e-08, + "loss": 1.0873, + "step": 40981 + }, + { + "epoch": 0.99, + "learning_rate": 1.0918189260059609e-08, + "loss": 0.867, + "step": 40982 + }, + { + "epoch": 0.99, + "learning_rate": 1.0881832220650535e-08, + "loss": 1.1322, + "step": 40983 + }, + { + "epoch": 0.99, + "learning_rate": 1.0845535782872952e-08, + "loss": 0.9294, + "step": 40984 + }, + { + "epoch": 0.99, + "learning_rate": 1.0809299946944463e-08, + "loss": 0.898, + "step": 40985 + }, + { + "epoch": 0.99, + "learning_rate": 1.0773124713087112e-08, + "loss": 0.9692, + "step": 40986 + }, + { + "epoch": 0.99, + "learning_rate": 1.0737010081519616e-08, + "loss": 1.134, + "step": 40987 + }, + { + "epoch": 0.99, + "learning_rate": 1.0700956052460686e-08, + "loss": 1.0702, + "step": 40988 + }, + { + "epoch": 0.99, + "learning_rate": 1.0664962626130149e-08, + "loss": 0.778, + "step": 40989 + }, + { + "epoch": 0.99, + "learning_rate": 1.0629029802744494e-08, + "loss": 1.1683, + "step": 40990 + }, + { + "epoch": 0.99, + "learning_rate": 1.059315758252355e-08, + "loss": 1.0102, + "step": 40991 + }, + { + "epoch": 0.99, + "learning_rate": 1.0557345965683808e-08, + "loss": 1.061, + "step": 40992 + }, + { + "epoch": 0.99, + "learning_rate": 1.052159495244398e-08, + "loss": 1.1372, + "step": 40993 + }, + { + "epoch": 0.99, + "learning_rate": 1.0485904543019453e-08, + "loss": 0.917, + "step": 40994 + }, + { + "epoch": 0.99, + "learning_rate": 1.045027473762672e-08, + "loss": 0.9838, + "step": 40995 + }, + { + "epoch": 0.99, + "learning_rate": 1.0414705536483383e-08, + "loss": 1.1218, + "step": 40996 + }, + { + "epoch": 0.99, + "learning_rate": 1.0379196939803716e-08, + "loss": 0.9832, + "step": 40997 + }, + { + "epoch": 0.99, + "learning_rate": 1.0343748947804211e-08, + "loss": 0.9771, + "step": 40998 + }, + { + "epoch": 0.99, + "learning_rate": 1.0308361560699143e-08, + "loss": 1.0947, + "step": 40999 + }, + { + "epoch": 0.99, + "learning_rate": 1.0273034778703894e-08, + "loss": 0.8098, + "step": 41000 + }, + { + "epoch": 0.99, + "learning_rate": 1.0237768602031628e-08, + "loss": 1.0321, + "step": 41001 + }, + { + "epoch": 0.99, + "learning_rate": 1.0202563030897727e-08, + "loss": 0.9663, + "step": 41002 + }, + { + "epoch": 0.99, + "learning_rate": 1.0167418065515355e-08, + "loss": 1.0808, + "step": 41003 + }, + { + "epoch": 0.99, + "learning_rate": 1.0132333706096564e-08, + "loss": 1.2164, + "step": 41004 + }, + { + "epoch": 0.99, + "learning_rate": 1.0097309952855627e-08, + "loss": 1.0677, + "step": 41005 + }, + { + "epoch": 0.99, + "learning_rate": 1.0062346806004597e-08, + "loss": 1.0538, + "step": 41006 + }, + { + "epoch": 0.99, + "learning_rate": 1.0027444265755525e-08, + "loss": 1.1396, + "step": 41007 + }, + { + "epoch": 0.99, + "learning_rate": 9.992602332320467e-09, + "loss": 0.7792, + "step": 41008 + }, + { + "epoch": 0.99, + "learning_rate": 9.957821005910362e-09, + "loss": 1.2443, + "step": 41009 + }, + { + "epoch": 0.99, + "learning_rate": 9.923100286736153e-09, + "loss": 0.9964, + "step": 41010 + }, + { + "epoch": 0.99, + "learning_rate": 9.888440175008785e-09, + "loss": 0.8614, + "step": 41011 + }, + { + "epoch": 0.99, + "learning_rate": 9.853840670938087e-09, + "loss": 0.9569, + "step": 41012 + }, + { + "epoch": 0.99, + "learning_rate": 9.819301774735001e-09, + "loss": 1.0474, + "step": 41013 + }, + { + "epoch": 0.99, + "learning_rate": 9.784823486608253e-09, + "loss": 0.9159, + "step": 41014 + }, + { + "epoch": 0.99, + "learning_rate": 9.75040580676656e-09, + "loss": 0.977, + "step": 41015 + }, + { + "epoch": 0.99, + "learning_rate": 9.716048735419758e-09, + "loss": 1.0934, + "step": 41016 + }, + { + "epoch": 0.99, + "learning_rate": 9.681752272775457e-09, + "loss": 0.9779, + "step": 41017 + }, + { + "epoch": 0.99, + "learning_rate": 9.64751641904238e-09, + "loss": 0.9499, + "step": 41018 + }, + { + "epoch": 0.99, + "learning_rate": 9.613341174428138e-09, + "loss": 0.8608, + "step": 41019 + }, + { + "epoch": 0.99, + "learning_rate": 9.57922653913923e-09, + "loss": 0.8934, + "step": 41020 + }, + { + "epoch": 0.99, + "learning_rate": 9.545172513383272e-09, + "loss": 1.0062, + "step": 41021 + }, + { + "epoch": 0.99, + "learning_rate": 9.511179097367872e-09, + "loss": 0.7852, + "step": 41022 + }, + { + "epoch": 0.99, + "learning_rate": 9.477246291297316e-09, + "loss": 1.2722, + "step": 41023 + }, + { + "epoch": 0.99, + "learning_rate": 9.44337409537921e-09, + "loss": 0.9931, + "step": 41024 + }, + { + "epoch": 0.99, + "learning_rate": 9.409562509817837e-09, + "loss": 0.8514, + "step": 41025 + }, + { + "epoch": 0.99, + "learning_rate": 9.375811534818591e-09, + "loss": 0.9722, + "step": 41026 + }, + { + "epoch": 0.99, + "learning_rate": 9.34212117058575e-09, + "loss": 1.2014, + "step": 41027 + }, + { + "epoch": 0.99, + "learning_rate": 9.308491417324705e-09, + "loss": 1.1278, + "step": 41028 + }, + { + "epoch": 0.99, + "learning_rate": 9.27492227523974e-09, + "loss": 1.1074, + "step": 41029 + }, + { + "epoch": 0.99, + "learning_rate": 9.241413744532912e-09, + "loss": 0.8906, + "step": 41030 + }, + { + "epoch": 0.99, + "learning_rate": 9.207965825409615e-09, + "loss": 0.8576, + "step": 41031 + }, + { + "epoch": 0.99, + "learning_rate": 9.174578518070799e-09, + "loss": 0.7825, + "step": 41032 + }, + { + "epoch": 0.99, + "learning_rate": 9.141251822719632e-09, + "loss": 0.8754, + "step": 41033 + }, + { + "epoch": 0.99, + "learning_rate": 9.107985739559289e-09, + "loss": 0.835, + "step": 41034 + }, + { + "epoch": 0.99, + "learning_rate": 9.074780268790717e-09, + "loss": 1.3659, + "step": 41035 + }, + { + "epoch": 0.99, + "learning_rate": 9.041635410614868e-09, + "loss": 0.9465, + "step": 41036 + }, + { + "epoch": 0.99, + "learning_rate": 9.008551165233803e-09, + "loss": 1.0924, + "step": 41037 + }, + { + "epoch": 0.99, + "learning_rate": 8.97552753284736e-09, + "loss": 1.0107, + "step": 41038 + }, + { + "epoch": 0.99, + "learning_rate": 8.942564513656493e-09, + "loss": 0.9529, + "step": 41039 + }, + { + "epoch": 0.99, + "learning_rate": 8.90966210786215e-09, + "loss": 0.9806, + "step": 41040 + }, + { + "epoch": 0.99, + "learning_rate": 8.87682031566084e-09, + "loss": 1.0623, + "step": 41041 + }, + { + "epoch": 0.99, + "learning_rate": 8.844039137255734e-09, + "loss": 0.852, + "step": 41042 + }, + { + "epoch": 0.99, + "learning_rate": 8.811318572842232e-09, + "loss": 1.0615, + "step": 41043 + }, + { + "epoch": 0.99, + "learning_rate": 8.778658622621283e-09, + "loss": 1.1159, + "step": 41044 + }, + { + "epoch": 0.99, + "learning_rate": 8.74605928679051e-09, + "loss": 0.9461, + "step": 41045 + }, + { + "epoch": 0.99, + "learning_rate": 8.713520565546418e-09, + "loss": 0.958, + "step": 41046 + }, + { + "epoch": 0.99, + "learning_rate": 8.68104245908774e-09, + "loss": 1.3025, + "step": 41047 + }, + { + "epoch": 0.99, + "learning_rate": 8.648624967610985e-09, + "loss": 0.9704, + "step": 41048 + }, + { + "epoch": 0.99, + "learning_rate": 8.616268091312663e-09, + "loss": 1.1207, + "step": 41049 + }, + { + "epoch": 0.99, + "learning_rate": 8.583971830389281e-09, + "loss": 1.1416, + "step": 41050 + }, + { + "epoch": 0.99, + "learning_rate": 8.551736185037351e-09, + "loss": 1.0996, + "step": 41051 + }, + { + "epoch": 0.99, + "learning_rate": 8.519561155451161e-09, + "loss": 1.0392, + "step": 41052 + }, + { + "epoch": 0.99, + "learning_rate": 8.487446741827222e-09, + "loss": 1.2252, + "step": 41053 + }, + { + "epoch": 0.99, + "learning_rate": 8.45539294435982e-09, + "loss": 1.101, + "step": 41054 + }, + { + "epoch": 0.99, + "learning_rate": 8.423399763243246e-09, + "loss": 0.9134, + "step": 41055 + }, + { + "epoch": 0.99, + "learning_rate": 8.391467198671788e-09, + "loss": 0.9386, + "step": 41056 + }, + { + "epoch": 0.99, + "learning_rate": 8.359595250839736e-09, + "loss": 1.0902, + "step": 41057 + }, + { + "epoch": 0.99, + "learning_rate": 8.327783919939158e-09, + "loss": 0.9002, + "step": 41058 + }, + { + "epoch": 0.99, + "learning_rate": 8.296033206163234e-09, + "loss": 1.1797, + "step": 41059 + }, + { + "epoch": 0.99, + "learning_rate": 8.264343109706252e-09, + "loss": 0.8363, + "step": 41060 + }, + { + "epoch": 0.99, + "learning_rate": 8.23271363075806e-09, + "loss": 0.8925, + "step": 41061 + }, + { + "epoch": 0.99, + "learning_rate": 8.201144769512948e-09, + "loss": 0.9012, + "step": 41062 + }, + { + "epoch": 0.99, + "learning_rate": 8.169636526160763e-09, + "loss": 1.1318, + "step": 41063 + }, + { + "epoch": 0.99, + "learning_rate": 8.138188900892464e-09, + "loss": 1.0063, + "step": 41064 + }, + { + "epoch": 0.99, + "learning_rate": 8.106801893900118e-09, + "loss": 1.0077, + "step": 41065 + }, + { + "epoch": 0.99, + "learning_rate": 8.075475505373576e-09, + "loss": 0.6557, + "step": 41066 + }, + { + "epoch": 0.99, + "learning_rate": 8.044209735503794e-09, + "loss": 1.0641, + "step": 41067 + }, + { + "epoch": 0.99, + "learning_rate": 8.013004584478401e-09, + "loss": 0.7221, + "step": 41068 + }, + { + "epoch": 0.99, + "learning_rate": 7.981860052488355e-09, + "loss": 1.2107, + "step": 41069 + }, + { + "epoch": 0.99, + "learning_rate": 7.950776139722394e-09, + "loss": 1.094, + "step": 41070 + }, + { + "epoch": 0.99, + "learning_rate": 7.919752846368145e-09, + "loss": 1.1308, + "step": 41071 + }, + { + "epoch": 0.99, + "learning_rate": 7.888790172614347e-09, + "loss": 0.9264, + "step": 41072 + }, + { + "epoch": 0.99, + "learning_rate": 7.857888118649737e-09, + "loss": 1.0074, + "step": 41073 + }, + { + "epoch": 0.99, + "learning_rate": 7.827046684659723e-09, + "loss": 1.1106, + "step": 41074 + }, + { + "epoch": 0.99, + "learning_rate": 7.796265870834152e-09, + "loss": 0.9669, + "step": 41075 + }, + { + "epoch": 0.99, + "learning_rate": 7.765545677357323e-09, + "loss": 0.9808, + "step": 41076 + }, + { + "epoch": 0.99, + "learning_rate": 7.734886104416861e-09, + "loss": 1.0889, + "step": 41077 + }, + { + "epoch": 0.99, + "learning_rate": 7.704287152199285e-09, + "loss": 0.9665, + "step": 41078 + }, + { + "epoch": 0.99, + "learning_rate": 7.673748820888894e-09, + "loss": 1.2367, + "step": 41079 + }, + { + "epoch": 0.99, + "learning_rate": 7.643271110671091e-09, + "loss": 0.9456, + "step": 41080 + }, + { + "epoch": 0.99, + "learning_rate": 7.612854021731287e-09, + "loss": 1.1251, + "step": 41081 + }, + { + "epoch": 0.99, + "learning_rate": 7.582497554254887e-09, + "loss": 1.1264, + "step": 41082 + }, + { + "epoch": 0.99, + "learning_rate": 7.552201708423968e-09, + "loss": 1.0954, + "step": 41083 + }, + { + "epoch": 0.99, + "learning_rate": 7.521966484423936e-09, + "loss": 0.8946, + "step": 41084 + }, + { + "epoch": 0.99, + "learning_rate": 7.491791882437983e-09, + "loss": 1.0588, + "step": 41085 + }, + { + "epoch": 0.99, + "learning_rate": 7.46167790264929e-09, + "loss": 1.0042, + "step": 41086 + }, + { + "epoch": 0.99, + "learning_rate": 7.431624545239935e-09, + "loss": 0.9786, + "step": 41087 + }, + { + "epoch": 0.99, + "learning_rate": 7.401631810391996e-09, + "loss": 0.9785, + "step": 41088 + }, + { + "epoch": 0.99, + "learning_rate": 7.371699698288659e-09, + "loss": 0.9638, + "step": 41089 + }, + { + "epoch": 0.99, + "learning_rate": 7.34182820910978e-09, + "loss": 1.1905, + "step": 41090 + }, + { + "epoch": 0.99, + "learning_rate": 7.3120173430385465e-09, + "loss": 0.8035, + "step": 41091 + }, + { + "epoch": 0.99, + "learning_rate": 7.282267100254814e-09, + "loss": 1.424, + "step": 41092 + }, + { + "epoch": 0.99, + "learning_rate": 7.252577480938438e-09, + "loss": 0.6693, + "step": 41093 + }, + { + "epoch": 0.99, + "learning_rate": 7.222948485271497e-09, + "loss": 0.8097, + "step": 41094 + }, + { + "epoch": 0.99, + "learning_rate": 7.193380113431625e-09, + "loss": 1.1115, + "step": 41095 + }, + { + "epoch": 0.99, + "learning_rate": 7.163872365598679e-09, + "loss": 0.9799, + "step": 41096 + }, + { + "epoch": 0.99, + "learning_rate": 7.134425241952514e-09, + "loss": 1.1503, + "step": 41097 + }, + { + "epoch": 0.99, + "learning_rate": 7.105038742670767e-09, + "loss": 1.0087, + "step": 41098 + }, + { + "epoch": 0.99, + "learning_rate": 7.075712867933293e-09, + "loss": 1.2239, + "step": 41099 + }, + { + "epoch": 0.99, + "learning_rate": 7.046447617916618e-09, + "loss": 1.0109, + "step": 41100 + }, + { + "epoch": 0.99, + "learning_rate": 7.0172429927983784e-09, + "loss": 0.967, + "step": 41101 + }, + { + "epoch": 0.99, + "learning_rate": 6.988098992756209e-09, + "loss": 1.1179, + "step": 41102 + }, + { + "epoch": 0.99, + "learning_rate": 6.9590156179666355e-09, + "loss": 1.0495, + "step": 41103 + }, + { + "epoch": 0.99, + "learning_rate": 6.929992868606184e-09, + "loss": 0.7652, + "step": 41104 + }, + { + "epoch": 0.99, + "learning_rate": 6.901030744851378e-09, + "loss": 1.1026, + "step": 41105 + }, + { + "epoch": 0.99, + "learning_rate": 6.872129246877635e-09, + "loss": 1.1367, + "step": 41106 + }, + { + "epoch": 0.99, + "learning_rate": 6.843288374859258e-09, + "loss": 1.2009, + "step": 41107 + }, + { + "epoch": 0.99, + "learning_rate": 6.814508128972774e-09, + "loss": 1.2927, + "step": 41108 + }, + { + "epoch": 0.99, + "learning_rate": 6.785788509392488e-09, + "loss": 1.1543, + "step": 41109 + }, + { + "epoch": 0.99, + "learning_rate": 6.757129516291594e-09, + "loss": 0.9626, + "step": 41110 + }, + { + "epoch": 0.99, + "learning_rate": 6.728531149844397e-09, + "loss": 1.0024, + "step": 41111 + }, + { + "epoch": 0.99, + "learning_rate": 6.699993410225203e-09, + "loss": 1.0918, + "step": 41112 + }, + { + "epoch": 0.99, + "learning_rate": 6.671516297606095e-09, + "loss": 1.2037, + "step": 41113 + }, + { + "epoch": 0.99, + "learning_rate": 6.64309981216138e-09, + "loss": 1.1242, + "step": 41114 + }, + { + "epoch": 0.99, + "learning_rate": 6.61474395406092e-09, + "loss": 0.9559, + "step": 41115 + }, + { + "epoch": 0.99, + "learning_rate": 6.586448723479022e-09, + "loss": 1.0941, + "step": 41116 + }, + { + "epoch": 0.99, + "learning_rate": 6.558214120586659e-09, + "loss": 1.1315, + "step": 41117 + }, + { + "epoch": 0.99, + "learning_rate": 6.530040145554806e-09, + "loss": 0.9282, + "step": 41118 + }, + { + "epoch": 0.99, + "learning_rate": 6.5019267985544364e-09, + "loss": 0.8451, + "step": 41119 + }, + { + "epoch": 0.99, + "learning_rate": 6.473874079756526e-09, + "loss": 1.0986, + "step": 41120 + }, + { + "epoch": 0.99, + "learning_rate": 6.4458819893309375e-09, + "loss": 1.1072, + "step": 41121 + }, + { + "epoch": 0.99, + "learning_rate": 6.417950527447536e-09, + "loss": 1.0352, + "step": 41122 + }, + { + "epoch": 0.99, + "learning_rate": 6.390079694276186e-09, + "loss": 1.0398, + "step": 41123 + }, + { + "epoch": 0.99, + "learning_rate": 6.3622694899845294e-09, + "loss": 1.2248, + "step": 41124 + }, + { + "epoch": 0.99, + "learning_rate": 6.3345199147435425e-09, + "loss": 1.1088, + "step": 41125 + }, + { + "epoch": 0.99, + "learning_rate": 6.306830968719757e-09, + "loss": 0.9281, + "step": 41126 + }, + { + "epoch": 0.99, + "learning_rate": 6.279202652081928e-09, + "loss": 0.851, + "step": 41127 + }, + { + "epoch": 0.99, + "learning_rate": 6.251634964997699e-09, + "loss": 1.0794, + "step": 41128 + }, + { + "epoch": 0.99, + "learning_rate": 6.2241279076336034e-09, + "loss": 0.9522, + "step": 41129 + }, + { + "epoch": 0.99, + "learning_rate": 6.196681480157285e-09, + "loss": 1.029, + "step": 41130 + }, + { + "epoch": 0.99, + "learning_rate": 6.169295682735277e-09, + "loss": 1.0088, + "step": 41131 + }, + { + "epoch": 0.99, + "learning_rate": 6.141970515533002e-09, + "loss": 1.1449, + "step": 41132 + }, + { + "epoch": 0.99, + "learning_rate": 6.114705978716995e-09, + "loss": 1.0447, + "step": 41133 + }, + { + "epoch": 0.99, + "learning_rate": 6.0875020724515675e-09, + "loss": 1.1159, + "step": 41134 + }, + { + "epoch": 0.99, + "learning_rate": 6.060358796903254e-09, + "loss": 0.9837, + "step": 41135 + }, + { + "epoch": 0.99, + "learning_rate": 6.033276152236367e-09, + "loss": 1.1065, + "step": 41136 + }, + { + "epoch": 0.99, + "learning_rate": 6.00625413861411e-09, + "loss": 1.0919, + "step": 41137 + }, + { + "epoch": 0.99, + "learning_rate": 5.979292756201904e-09, + "loss": 0.9772, + "step": 41138 + }, + { + "epoch": 0.99, + "learning_rate": 5.952392005161845e-09, + "loss": 1.0255, + "step": 41139 + }, + { + "epoch": 0.99, + "learning_rate": 5.925551885659353e-09, + "loss": 0.8585, + "step": 41140 + }, + { + "epoch": 0.99, + "learning_rate": 5.8987723978543024e-09, + "loss": 0.9241, + "step": 41141 + }, + { + "epoch": 0.99, + "learning_rate": 5.872053541911005e-09, + "loss": 0.9743, + "step": 41142 + }, + { + "epoch": 0.99, + "learning_rate": 5.845395317991554e-09, + "loss": 1.0543, + "step": 41143 + }, + { + "epoch": 0.99, + "learning_rate": 5.818797726256931e-09, + "loss": 1.1147, + "step": 41144 + }, + { + "epoch": 0.99, + "learning_rate": 5.792260766869229e-09, + "loss": 1.2192, + "step": 41145 + }, + { + "epoch": 0.99, + "learning_rate": 5.76578443998832e-09, + "loss": 1.0031, + "step": 41146 + }, + { + "epoch": 0.99, + "learning_rate": 5.739368745776297e-09, + "loss": 0.8901, + "step": 41147 + }, + { + "epoch": 0.99, + "learning_rate": 5.7130136843930315e-09, + "loss": 1.1323, + "step": 41148 + }, + { + "epoch": 0.99, + "learning_rate": 5.686719255997286e-09, + "loss": 1.0949, + "step": 41149 + }, + { + "epoch": 0.99, + "learning_rate": 5.6604854607500425e-09, + "loss": 1.1283, + "step": 41150 + }, + { + "epoch": 0.99, + "learning_rate": 5.634312298808953e-09, + "loss": 0.9528, + "step": 41151 + }, + { + "epoch": 0.99, + "learning_rate": 5.608199770334999e-09, + "loss": 0.9702, + "step": 41152 + }, + { + "epoch": 0.99, + "learning_rate": 5.582147875484722e-09, + "loss": 1.0035, + "step": 41153 + }, + { + "epoch": 0.99, + "learning_rate": 5.556156614415775e-09, + "loss": 1.182, + "step": 41154 + }, + { + "epoch": 0.99, + "learning_rate": 5.53022598728803e-09, + "loss": 1.2658, + "step": 41155 + }, + { + "epoch": 0.99, + "learning_rate": 5.504355994256916e-09, + "loss": 1.13, + "step": 41156 + }, + { + "epoch": 0.99, + "learning_rate": 5.478546635480087e-09, + "loss": 1.2334, + "step": 41157 + }, + { + "epoch": 0.99, + "learning_rate": 5.4527979111140824e-09, + "loss": 1.1647, + "step": 41158 + }, + { + "epoch": 0.99, + "learning_rate": 5.427109821315446e-09, + "loss": 1.2329, + "step": 41159 + }, + { + "epoch": 0.99, + "learning_rate": 5.4014823662396074e-09, + "loss": 0.8732, + "step": 41160 + }, + { + "epoch": 0.99, + "learning_rate": 5.375915546041999e-09, + "loss": 1.0626, + "step": 41161 + }, + { + "epoch": 0.99, + "learning_rate": 5.35040936087694e-09, + "loss": 0.9646, + "step": 41162 + }, + { + "epoch": 0.99, + "learning_rate": 5.324963810900974e-09, + "loss": 1.3141, + "step": 41163 + }, + { + "epoch": 0.99, + "learning_rate": 5.2995788962684204e-09, + "loss": 1.0925, + "step": 41164 + }, + { + "epoch": 0.99, + "learning_rate": 5.27425461713138e-09, + "loss": 1.1717, + "step": 41165 + }, + { + "epoch": 0.99, + "learning_rate": 5.2489909736441744e-09, + "loss": 0.8139, + "step": 41166 + }, + { + "epoch": 0.99, + "learning_rate": 5.2237879659611245e-09, + "loss": 1.3419, + "step": 41167 + }, + { + "epoch": 0.99, + "learning_rate": 5.1986455942332205e-09, + "loss": 1.3039, + "step": 41168 + }, + { + "epoch": 0.99, + "learning_rate": 5.173563858615893e-09, + "loss": 1.0075, + "step": 41169 + }, + { + "epoch": 0.99, + "learning_rate": 5.148542759259023e-09, + "loss": 1.1995, + "step": 41170 + }, + { + "epoch": 0.99, + "learning_rate": 5.123582296314711e-09, + "loss": 1.0299, + "step": 41171 + }, + { + "epoch": 0.99, + "learning_rate": 5.0986824699350565e-09, + "loss": 1.0909, + "step": 41172 + }, + { + "epoch": 0.99, + "learning_rate": 5.073843280269941e-09, + "loss": 0.8007, + "step": 41173 + }, + { + "epoch": 0.99, + "learning_rate": 5.049064727471464e-09, + "loss": 0.9766, + "step": 41174 + }, + { + "epoch": 0.99, + "learning_rate": 5.024346811689507e-09, + "loss": 1.1054, + "step": 41175 + }, + { + "epoch": 0.99, + "learning_rate": 4.999689533073948e-09, + "loss": 0.9812, + "step": 41176 + }, + { + "epoch": 0.99, + "learning_rate": 4.975092891774669e-09, + "loss": 1.2735, + "step": 41177 + }, + { + "epoch": 0.99, + "learning_rate": 4.950556887939329e-09, + "loss": 1.0967, + "step": 41178 + }, + { + "epoch": 0.99, + "learning_rate": 4.9260815217189175e-09, + "loss": 1.1259, + "step": 41179 + }, + { + "epoch": 0.99, + "learning_rate": 4.901666793261095e-09, + "loss": 0.9994, + "step": 41180 + }, + { + "epoch": 0.99, + "learning_rate": 4.877312702713521e-09, + "loss": 0.8225, + "step": 41181 + }, + { + "epoch": 0.99, + "learning_rate": 4.853019250223856e-09, + "loss": 1.1996, + "step": 41182 + }, + { + "epoch": 0.99, + "learning_rate": 4.828786435939759e-09, + "loss": 0.987, + "step": 41183 + }, + { + "epoch": 0.99, + "learning_rate": 4.804614260008889e-09, + "loss": 1.0284, + "step": 41184 + }, + { + "epoch": 0.99, + "learning_rate": 4.780502722577796e-09, + "loss": 1.0939, + "step": 41185 + }, + { + "epoch": 0.99, + "learning_rate": 4.756451823791919e-09, + "loss": 1.1861, + "step": 41186 + }, + { + "epoch": 0.99, + "learning_rate": 4.732461563797807e-09, + "loss": 1.0112, + "step": 41187 + }, + { + "epoch": 0.99, + "learning_rate": 4.708531942740902e-09, + "loss": 1.0561, + "step": 41188 + }, + { + "epoch": 0.99, + "learning_rate": 4.6846629607666395e-09, + "loss": 0.8267, + "step": 41189 + }, + { + "epoch": 0.99, + "learning_rate": 4.660854618019351e-09, + "loss": 0.8271, + "step": 41190 + }, + { + "epoch": 0.99, + "learning_rate": 4.637106914642253e-09, + "loss": 1.2805, + "step": 41191 + }, + { + "epoch": 0.99, + "learning_rate": 4.613419850781897e-09, + "loss": 0.5884, + "step": 41192 + }, + { + "epoch": 0.99, + "learning_rate": 4.5897934265815015e-09, + "loss": 0.8622, + "step": 41193 + }, + { + "epoch": 0.99, + "learning_rate": 4.566227642182064e-09, + "loss": 1.0014, + "step": 41194 + }, + { + "epoch": 0.99, + "learning_rate": 4.542722497730134e-09, + "loss": 1.0256, + "step": 41195 + }, + { + "epoch": 0.99, + "learning_rate": 4.5192779933656e-09, + "loss": 1.1088, + "step": 41196 + }, + { + "epoch": 0.99, + "learning_rate": 4.495894129230571e-09, + "loss": 1.1125, + "step": 41197 + }, + { + "epoch": 0.99, + "learning_rate": 4.4725709054693756e-09, + "loss": 0.9888, + "step": 41198 + }, + { + "epoch": 0.99, + "learning_rate": 4.4493083222207915e-09, + "loss": 1.0355, + "step": 41199 + }, + { + "epoch": 0.99, + "learning_rate": 4.4261063796280365e-09, + "loss": 0.8743, + "step": 41200 + }, + { + "epoch": 0.99, + "learning_rate": 4.402965077831001e-09, + "loss": 1.0339, + "step": 41201 + }, + { + "epoch": 0.99, + "learning_rate": 4.379884416969571e-09, + "loss": 1.0505, + "step": 41202 + }, + { + "epoch": 0.99, + "learning_rate": 4.356864397184746e-09, + "loss": 0.8279, + "step": 41203 + }, + { + "epoch": 0.99, + "learning_rate": 4.333905018615303e-09, + "loss": 1.0504, + "step": 41204 + }, + { + "epoch": 0.99, + "learning_rate": 4.311006281401131e-09, + "loss": 1.1351, + "step": 41205 + }, + { + "epoch": 0.99, + "learning_rate": 4.288168185681007e-09, + "loss": 1.0349, + "step": 41206 + }, + { + "epoch": 0.99, + "learning_rate": 4.2653907315926e-09, + "loss": 1.1286, + "step": 41207 + }, + { + "epoch": 0.99, + "learning_rate": 4.242673919275797e-09, + "loss": 0.8773, + "step": 41208 + }, + { + "epoch": 0.99, + "learning_rate": 4.220017748867155e-09, + "loss": 0.8067, + "step": 41209 + }, + { + "epoch": 0.99, + "learning_rate": 4.197422220505453e-09, + "loss": 0.7654, + "step": 41210 + }, + { + "epoch": 0.99, + "learning_rate": 4.174887334325028e-09, + "loss": 0.9798, + "step": 41211 + }, + { + "epoch": 0.99, + "learning_rate": 4.152413090465768e-09, + "loss": 1.1798, + "step": 41212 + }, + { + "epoch": 0.99, + "learning_rate": 4.129999489063119e-09, + "loss": 1.1353, + "step": 41213 + }, + { + "epoch": 0.99, + "learning_rate": 4.107646530251419e-09, + "loss": 1.0323, + "step": 41214 + }, + { + "epoch": 0.99, + "learning_rate": 4.085354214167225e-09, + "loss": 0.8728, + "step": 41215 + }, + { + "epoch": 0.99, + "learning_rate": 4.063122540947095e-09, + "loss": 0.8568, + "step": 41216 + }, + { + "epoch": 0.99, + "learning_rate": 4.040951510724256e-09, + "loss": 0.8111, + "step": 41217 + }, + { + "epoch": 0.99, + "learning_rate": 4.018841123634154e-09, + "loss": 1.1251, + "step": 41218 + }, + { + "epoch": 0.99, + "learning_rate": 3.996791379810017e-09, + "loss": 1.0518, + "step": 41219 + }, + { + "epoch": 0.99, + "learning_rate": 3.974802279386181e-09, + "loss": 1.1574, + "step": 41220 + }, + { + "epoch": 0.99, + "learning_rate": 3.9528738224969835e-09, + "loss": 0.9501, + "step": 41221 + }, + { + "epoch": 0.99, + "learning_rate": 3.93100600927343e-09, + "loss": 1.0507, + "step": 41222 + }, + { + "epoch": 0.99, + "learning_rate": 3.90919883984986e-09, + "loss": 0.957, + "step": 41223 + }, + { + "epoch": 0.99, + "learning_rate": 3.887452314357276e-09, + "loss": 0.8947, + "step": 41224 + }, + { + "epoch": 0.99, + "learning_rate": 3.8657664329289076e-09, + "loss": 1.1076, + "step": 41225 + }, + { + "epoch": 0.99, + "learning_rate": 3.8441411956946505e-09, + "loss": 0.9827, + "step": 41226 + }, + { + "epoch": 0.99, + "learning_rate": 3.822576602787731e-09, + "loss": 0.9512, + "step": 41227 + }, + { + "epoch": 0.99, + "learning_rate": 3.801072654338045e-09, + "loss": 1.0772, + "step": 41228 + }, + { + "epoch": 0.99, + "learning_rate": 3.779629350475489e-09, + "loss": 0.9822, + "step": 41229 + }, + { + "epoch": 0.99, + "learning_rate": 3.758246691331069e-09, + "loss": 1.3503, + "step": 41230 + }, + { + "epoch": 0.99, + "learning_rate": 3.736924677034681e-09, + "loss": 0.9992, + "step": 41231 + }, + { + "epoch": 0.99, + "learning_rate": 3.7156633077140012e-09, + "loss": 1.2723, + "step": 41232 + }, + { + "epoch": 0.99, + "learning_rate": 3.6944625834989256e-09, + "loss": 0.7285, + "step": 41233 + }, + { + "epoch": 0.99, + "learning_rate": 3.67332250451935e-09, + "loss": 1.2021, + "step": 41234 + }, + { + "epoch": 0.99, + "learning_rate": 3.65224307090184e-09, + "loss": 1.2341, + "step": 41235 + }, + { + "epoch": 0.99, + "learning_rate": 3.631224282775181e-09, + "loss": 1.0922, + "step": 41236 + }, + { + "epoch": 0.99, + "learning_rate": 3.610266140265939e-09, + "loss": 0.974, + "step": 41237 + }, + { + "epoch": 0.99, + "learning_rate": 3.5893686435028995e-09, + "loss": 1.0286, + "step": 41238 + }, + { + "epoch": 0.99, + "learning_rate": 3.568531792611518e-09, + "loss": 1.0796, + "step": 41239 + }, + { + "epoch": 0.99, + "learning_rate": 3.54775558771836e-09, + "loss": 1.2818, + "step": 41240 + }, + { + "epoch": 0.99, + "learning_rate": 3.5270400289499908e-09, + "loss": 1.0531, + "step": 41241 + }, + { + "epoch": 0.99, + "learning_rate": 3.506385116430755e-09, + "loss": 0.8898, + "step": 41242 + }, + { + "epoch": 0.99, + "learning_rate": 3.4857908502883288e-09, + "loss": 1.1137, + "step": 41243 + }, + { + "epoch": 0.99, + "learning_rate": 3.465257230644836e-09, + "loss": 0.8995, + "step": 41244 + }, + { + "epoch": 0.99, + "learning_rate": 3.444784257626843e-09, + "loss": 1.1128, + "step": 41245 + }, + { + "epoch": 0.99, + "learning_rate": 3.424371931358694e-09, + "loss": 1.0606, + "step": 41246 + }, + { + "epoch": 0.99, + "learning_rate": 3.4040202519636247e-09, + "loss": 1.051, + "step": 41247 + }, + { + "epoch": 0.99, + "learning_rate": 3.3837292195637583e-09, + "loss": 1.0241, + "step": 41248 + }, + { + "epoch": 0.99, + "learning_rate": 3.363498834284551e-09, + "loss": 1.0042, + "step": 41249 + }, + { + "epoch": 0.99, + "learning_rate": 3.3433290962470166e-09, + "loss": 0.9403, + "step": 41250 + }, + { + "epoch": 0.99, + "learning_rate": 3.32322000557439e-09, + "loss": 1.1909, + "step": 41251 + }, + { + "epoch": 0.99, + "learning_rate": 3.303171562388796e-09, + "loss": 1.0173, + "step": 41252 + }, + { + "epoch": 0.99, + "learning_rate": 3.283183766811249e-09, + "loss": 1.1988, + "step": 41253 + }, + { + "epoch": 0.99, + "learning_rate": 3.2632566189627622e-09, + "loss": 1.0805, + "step": 41254 + }, + { + "epoch": 0.99, + "learning_rate": 3.2433901189654614e-09, + "loss": 1.0098, + "step": 41255 + }, + { + "epoch": 0.99, + "learning_rate": 3.22358426693814e-09, + "loss": 1.1002, + "step": 41256 + }, + { + "epoch": 0.99, + "learning_rate": 3.2038390630029224e-09, + "loss": 1.1262, + "step": 41257 + }, + { + "epoch": 0.99, + "learning_rate": 3.1841545072786027e-09, + "loss": 1.0587, + "step": 41258 + }, + { + "epoch": 0.99, + "learning_rate": 3.164530599883975e-09, + "loss": 0.8571, + "step": 41259 + }, + { + "epoch": 0.99, + "learning_rate": 3.1449673409389425e-09, + "loss": 1.0138, + "step": 41260 + }, + { + "epoch": 0.99, + "learning_rate": 3.1254647305623e-09, + "loss": 1.0987, + "step": 41261 + }, + { + "epoch": 0.99, + "learning_rate": 3.1060227688717303e-09, + "loss": 1.0926, + "step": 41262 + }, + { + "epoch": 0.99, + "learning_rate": 3.0866414559860282e-09, + "loss": 1.0855, + "step": 41263 + }, + { + "epoch": 0.99, + "learning_rate": 3.0673207920217664e-09, + "loss": 0.9138, + "step": 41264 + }, + { + "epoch": 0.99, + "learning_rate": 3.048060777096629e-09, + "loss": 1.0007, + "step": 41265 + }, + { + "epoch": 0.99, + "learning_rate": 3.0288614113282987e-09, + "loss": 0.9177, + "step": 41266 + }, + { + "epoch": 0.99, + "learning_rate": 3.0097226948311298e-09, + "loss": 0.9564, + "step": 41267 + }, + { + "epoch": 0.99, + "learning_rate": 2.9906446277239154e-09, + "loss": 1.1795, + "step": 41268 + }, + { + "epoch": 0.99, + "learning_rate": 2.9716272101198984e-09, + "loss": 1.1204, + "step": 41269 + }, + { + "epoch": 0.99, + "learning_rate": 2.9526704421356523e-09, + "loss": 0.9965, + "step": 41270 + }, + { + "epoch": 0.99, + "learning_rate": 2.93377432388664e-09, + "loss": 0.9475, + "step": 41271 + }, + { + "epoch": 0.99, + "learning_rate": 2.9149388554861045e-09, + "loss": 0.938, + "step": 41272 + }, + { + "epoch": 0.99, + "learning_rate": 2.8961640370506196e-09, + "loss": 1.0952, + "step": 41273 + }, + { + "epoch": 0.99, + "learning_rate": 2.8774498686912066e-09, + "loss": 1.1093, + "step": 41274 + }, + { + "epoch": 0.99, + "learning_rate": 2.8587963505233296e-09, + "loss": 1.0981, + "step": 41275 + }, + { + "epoch": 0.99, + "learning_rate": 2.840203482660231e-09, + "loss": 1.0177, + "step": 41276 + }, + { + "epoch": 0.99, + "learning_rate": 2.8216712652140433e-09, + "loss": 0.918, + "step": 41277 + }, + { + "epoch": 0.99, + "learning_rate": 2.803199698296899e-09, + "loss": 1.0964, + "step": 41278 + }, + { + "epoch": 0.99, + "learning_rate": 2.7847887820209307e-09, + "loss": 1.3216, + "step": 41279 + }, + { + "epoch": 0.99, + "learning_rate": 2.766438516498271e-09, + "loss": 1.1992, + "step": 41280 + }, + { + "epoch": 0.99, + "learning_rate": 2.7481489018410525e-09, + "loss": 0.972, + "step": 41281 + }, + { + "epoch": 0.99, + "learning_rate": 2.7299199381580766e-09, + "loss": 1.0222, + "step": 41282 + }, + { + "epoch": 0.99, + "learning_rate": 2.7117516255625866e-09, + "loss": 1.159, + "step": 41283 + }, + { + "epoch": 0.99, + "learning_rate": 2.6936439641622737e-09, + "loss": 1.0513, + "step": 41284 + }, + { + "epoch": 0.99, + "learning_rate": 2.6755969540681603e-09, + "loss": 0.854, + "step": 41285 + }, + { + "epoch": 0.99, + "learning_rate": 2.657610595390159e-09, + "loss": 1.0042, + "step": 41286 + }, + { + "epoch": 0.99, + "learning_rate": 2.6396848882370707e-09, + "loss": 0.796, + "step": 41287 + }, + { + "epoch": 0.99, + "learning_rate": 2.621819832716588e-09, + "loss": 1.0255, + "step": 41288 + }, + { + "epoch": 0.99, + "learning_rate": 2.604015428938622e-09, + "loss": 0.8841, + "step": 41289 + }, + { + "epoch": 0.99, + "learning_rate": 2.5862716770108654e-09, + "loss": 0.8937, + "step": 41290 + }, + { + "epoch": 0.99, + "learning_rate": 2.568588577039899e-09, + "loss": 0.9671, + "step": 41291 + }, + { + "epoch": 0.99, + "learning_rate": 2.5509661291345245e-09, + "loss": 0.8994, + "step": 41292 + }, + { + "epoch": 0.99, + "learning_rate": 2.5334043334002136e-09, + "loss": 1.1093, + "step": 41293 + }, + { + "epoch": 0.99, + "learning_rate": 2.5159031899446574e-09, + "loss": 1.216, + "step": 41294 + }, + { + "epoch": 0.99, + "learning_rate": 2.498462698873327e-09, + "loss": 0.8423, + "step": 41295 + }, + { + "epoch": 0.99, + "learning_rate": 2.4810828602916947e-09, + "loss": 1.0095, + "step": 41296 + }, + { + "epoch": 0.99, + "learning_rate": 2.463763674306341e-09, + "loss": 1.095, + "step": 41297 + }, + { + "epoch": 0.99, + "learning_rate": 2.4465051410227368e-09, + "loss": 1.0559, + "step": 41298 + }, + { + "epoch": 0.99, + "learning_rate": 2.4293072605430233e-09, + "loss": 1.1052, + "step": 41299 + }, + { + "epoch": 0.99, + "learning_rate": 2.412170032973782e-09, + "loss": 1.0072, + "step": 41300 + }, + { + "epoch": 0.99, + "learning_rate": 2.3950934584182627e-09, + "loss": 0.9685, + "step": 41301 + }, + { + "epoch": 0.99, + "learning_rate": 2.3780775369808275e-09, + "loss": 1.1163, + "step": 41302 + }, + { + "epoch": 0.99, + "learning_rate": 2.361122268763616e-09, + "loss": 1.0071, + "step": 41303 + }, + { + "epoch": 0.99, + "learning_rate": 2.344227653868769e-09, + "loss": 1.1108, + "step": 41304 + }, + { + "epoch": 0.99, + "learning_rate": 2.327393692401758e-09, + "loss": 1.1871, + "step": 41305 + }, + { + "epoch": 0.99, + "learning_rate": 2.310620384461393e-09, + "loss": 0.9898, + "step": 41306 + }, + { + "epoch": 0.99, + "learning_rate": 2.2939077301509237e-09, + "loss": 0.9985, + "step": 41307 + }, + { + "epoch": 0.99, + "learning_rate": 2.277255729572492e-09, + "loss": 0.8795, + "step": 41308 + }, + { + "epoch": 0.99, + "learning_rate": 2.2606643828260167e-09, + "loss": 0.9972, + "step": 41309 + }, + { + "epoch": 0.99, + "learning_rate": 2.2441336900125288e-09, + "loss": 0.7298, + "step": 41310 + }, + { + "epoch": 0.99, + "learning_rate": 2.2276636512330584e-09, + "loss": 0.7435, + "step": 41311 + }, + { + "epoch": 0.99, + "learning_rate": 2.2112542665853054e-09, + "loss": 1.1002, + "step": 41312 + }, + { + "epoch": 0.99, + "learning_rate": 2.19490553617141e-09, + "loss": 1.1872, + "step": 41313 + }, + { + "epoch": 0.99, + "learning_rate": 2.178617460087962e-09, + "loss": 0.6261, + "step": 41314 + }, + { + "epoch": 0.99, + "learning_rate": 2.162390038437101e-09, + "loss": 1.0814, + "step": 41315 + }, + { + "epoch": 0.99, + "learning_rate": 2.1462232713143072e-09, + "loss": 1.0963, + "step": 41316 + }, + { + "epoch": 0.99, + "learning_rate": 2.13011715881839e-09, + "loss": 1.0007, + "step": 41317 + }, + { + "epoch": 0.99, + "learning_rate": 2.1140717010481595e-09, + "loss": 1.1721, + "step": 41318 + }, + { + "epoch": 0.99, + "learning_rate": 2.0980868981002043e-09, + "loss": 1.0147, + "step": 41319 + }, + { + "epoch": 0.99, + "learning_rate": 2.082162750071115e-09, + "loss": 0.9935, + "step": 41320 + }, + { + "epoch": 0.99, + "learning_rate": 2.06629925705748e-09, + "loss": 1.2044, + "step": 41321 + }, + { + "epoch": 0.99, + "learning_rate": 2.0504964191569996e-09, + "loss": 0.9695, + "step": 41322 + }, + { + "epoch": 0.99, + "learning_rate": 2.0347542364629324e-09, + "loss": 0.5762, + "step": 41323 + }, + { + "epoch": 0.99, + "learning_rate": 2.019072709074088e-09, + "loss": 0.6606, + "step": 41324 + }, + { + "epoch": 0.99, + "learning_rate": 2.003451837082615e-09, + "loss": 0.775, + "step": 41325 + }, + { + "epoch": 0.99, + "learning_rate": 1.987891620585103e-09, + "loss": 0.9113, + "step": 41326 + }, + { + "epoch": 0.99, + "learning_rate": 1.9723920596759207e-09, + "loss": 1.0908, + "step": 41327 + }, + { + "epoch": 0.99, + "learning_rate": 1.9569531544472163e-09, + "loss": 1.0125, + "step": 41328 + }, + { + "epoch": 0.99, + "learning_rate": 1.94157490499447e-09, + "loss": 0.9949, + "step": 41329 + }, + { + "epoch": 0.99, + "learning_rate": 1.9262573114109395e-09, + "loss": 0.9357, + "step": 41330 + }, + { + "epoch": 0.99, + "learning_rate": 1.9110003737887737e-09, + "loss": 1.1641, + "step": 41331 + }, + { + "epoch": 0.99, + "learning_rate": 1.8958040922212316e-09, + "loss": 1.1646, + "step": 41332 + }, + { + "epoch": 0.99, + "learning_rate": 1.8806684668004616e-09, + "loss": 1.3031, + "step": 41333 + }, + { + "epoch": 0.99, + "learning_rate": 1.8655934976175017e-09, + "loss": 1.0158, + "step": 41334 + }, + { + "epoch": 0.99, + "learning_rate": 1.8505791847645005e-09, + "loss": 0.9272, + "step": 41335 + }, + { + "epoch": 0.99, + "learning_rate": 1.8356255283313862e-09, + "loss": 0.9379, + "step": 41336 + }, + { + "epoch": 0.99, + "learning_rate": 1.8207325284103072e-09, + "loss": 0.8924, + "step": 41337 + }, + { + "epoch": 0.99, + "learning_rate": 1.8059001850911917e-09, + "loss": 0.9943, + "step": 41338 + }, + { + "epoch": 0.99, + "learning_rate": 1.7911284984639677e-09, + "loss": 1.2261, + "step": 41339 + }, + { + "epoch": 0.99, + "learning_rate": 1.7764174686185631e-09, + "loss": 1.1896, + "step": 41340 + }, + { + "epoch": 0.99, + "learning_rate": 1.761767095643796e-09, + "loss": 1.1493, + "step": 41341 + }, + { + "epoch": 0.99, + "learning_rate": 1.747177379628484e-09, + "loss": 1.2193, + "step": 41342 + }, + { + "epoch": 0.99, + "learning_rate": 1.7326483206614452e-09, + "loss": 0.9285, + "step": 41343 + }, + { + "epoch": 0.99, + "learning_rate": 1.7181799188314974e-09, + "loss": 1.0397, + "step": 41344 + }, + { + "epoch": 0.99, + "learning_rate": 1.7037721742241276e-09, + "loss": 0.9802, + "step": 41345 + }, + { + "epoch": 0.99, + "learning_rate": 1.689425086929264e-09, + "loss": 0.9102, + "step": 41346 + }, + { + "epoch": 0.99, + "learning_rate": 1.6751386570323935e-09, + "loss": 1.1472, + "step": 41347 + }, + { + "epoch": 0.99, + "learning_rate": 1.6609128846212242e-09, + "loss": 0.8937, + "step": 41348 + }, + { + "epoch": 0.99, + "learning_rate": 1.6467477697801326e-09, + "loss": 1.172, + "step": 41349 + }, + { + "epoch": 0.99, + "learning_rate": 1.6326433125979368e-09, + "loss": 0.9792, + "step": 41350 + }, + { + "epoch": 0.99, + "learning_rate": 1.6185995131579035e-09, + "loss": 0.9142, + "step": 41351 + }, + { + "epoch": 0.99, + "learning_rate": 1.60461637154663e-09, + "loss": 0.8457, + "step": 41352 + }, + { + "epoch": 0.99, + "learning_rate": 1.5906938878484935e-09, + "loss": 0.7454, + "step": 41353 + }, + { + "epoch": 0.99, + "learning_rate": 1.5768320621478705e-09, + "loss": 1.0174, + "step": 41354 + }, + { + "epoch": 0.99, + "learning_rate": 1.5630308945280282e-09, + "loss": 0.9201, + "step": 41355 + }, + { + "epoch": 0.99, + "learning_rate": 1.5492903850744534e-09, + "loss": 0.9317, + "step": 41356 + }, + { + "epoch": 0.99, + "learning_rate": 1.5356105338693029e-09, + "loss": 1.1595, + "step": 41357 + }, + { + "epoch": 0.99, + "learning_rate": 1.5219913409947329e-09, + "loss": 1.1262, + "step": 41358 + }, + { + "epoch": 0.99, + "learning_rate": 1.5084328065351206e-09, + "loss": 0.9892, + "step": 41359 + }, + { + "epoch": 0.99, + "learning_rate": 1.4949349305726224e-09, + "loss": 0.9485, + "step": 41360 + }, + { + "epoch": 0.99, + "learning_rate": 1.4814977131882846e-09, + "loss": 0.9394, + "step": 41361 + }, + { + "epoch": 0.99, + "learning_rate": 1.4681211544642638e-09, + "loss": 1.1563, + "step": 41362 + }, + { + "epoch": 0.99, + "learning_rate": 1.4548052544804958e-09, + "loss": 0.9952, + "step": 41363 + }, + { + "epoch": 0.99, + "learning_rate": 1.4415500133191374e-09, + "loss": 0.9217, + "step": 41364 + }, + { + "epoch": 0.99, + "learning_rate": 1.4283554310601243e-09, + "loss": 1.041, + "step": 41365 + }, + { + "epoch": 0.99, + "learning_rate": 1.415221507783393e-09, + "loss": 1.0538, + "step": 41366 + }, + { + "epoch": 0.99, + "learning_rate": 1.4021482435688794e-09, + "loss": 0.8005, + "step": 41367 + }, + { + "epoch": 0.99, + "learning_rate": 1.3891356384954092e-09, + "loss": 1.2584, + "step": 41368 + }, + { + "epoch": 0.99, + "learning_rate": 1.376183692641808e-09, + "loss": 1.2506, + "step": 41369 + }, + { + "epoch": 0.99, + "learning_rate": 1.3632924060880126e-09, + "loss": 1.1184, + "step": 41370 + }, + { + "epoch": 0.99, + "learning_rate": 1.3504617789106277e-09, + "loss": 1.0885, + "step": 41371 + }, + { + "epoch": 0.99, + "learning_rate": 1.3376918111884796e-09, + "loss": 0.9797, + "step": 41372 + }, + { + "epoch": 0.99, + "learning_rate": 1.3249825029981733e-09, + "loss": 1.1297, + "step": 41373 + }, + { + "epoch": 0.99, + "learning_rate": 1.312333854418535e-09, + "loss": 1.0089, + "step": 41374 + }, + { + "epoch": 1.0, + "learning_rate": 1.2997458655239493e-09, + "loss": 0.9343, + "step": 41375 + }, + { + "epoch": 1.0, + "learning_rate": 1.287218536393242e-09, + "loss": 1.0166, + "step": 41376 + }, + { + "epoch": 1.0, + "learning_rate": 1.2747518670996883e-09, + "loss": 1.065, + "step": 41377 + }, + { + "epoch": 1.0, + "learning_rate": 1.2623458577210035e-09, + "loss": 1.064, + "step": 41378 + }, + { + "epoch": 1.0, + "learning_rate": 1.2500005083315726e-09, + "loss": 1.1107, + "step": 41379 + }, + { + "epoch": 1.0, + "learning_rate": 1.2377158190068906e-09, + "loss": 1.0253, + "step": 41380 + }, + { + "epoch": 1.0, + "learning_rate": 1.2254917898213426e-09, + "loss": 0.8234, + "step": 41381 + }, + { + "epoch": 1.0, + "learning_rate": 1.2133284208493135e-09, + "loss": 0.872, + "step": 41382 + }, + { + "epoch": 1.0, + "learning_rate": 1.2012257121629678e-09, + "loss": 1.1425, + "step": 41383 + }, + { + "epoch": 1.0, + "learning_rate": 1.1891836638378008e-09, + "loss": 1.1954, + "step": 41384 + }, + { + "epoch": 1.0, + "learning_rate": 1.1772022759448664e-09, + "loss": 1.1486, + "step": 41385 + }, + { + "epoch": 1.0, + "learning_rate": 1.1652815485585501e-09, + "loss": 0.8051, + "step": 41386 + }, + { + "epoch": 1.0, + "learning_rate": 1.153421481751016e-09, + "loss": 1.09, + "step": 41387 + }, + { + "epoch": 1.0, + "learning_rate": 1.1416220755933183e-09, + "loss": 1.0132, + "step": 41388 + }, + { + "epoch": 1.0, + "learning_rate": 1.129883330158732e-09, + "loss": 0.8584, + "step": 41389 + }, + { + "epoch": 1.0, + "learning_rate": 1.1182052455160908e-09, + "loss": 1.0229, + "step": 41390 + }, + { + "epoch": 1.0, + "learning_rate": 1.1065878217375593e-09, + "loss": 0.8936, + "step": 41391 + }, + { + "epoch": 1.0, + "learning_rate": 1.0950310588930812e-09, + "loss": 0.6163, + "step": 41392 + }, + { + "epoch": 1.0, + "learning_rate": 1.0835349570537113e-09, + "loss": 1.1596, + "step": 41393 + }, + { + "epoch": 1.0, + "learning_rate": 1.072099516288283e-09, + "loss": 1.2389, + "step": 41394 + }, + { + "epoch": 1.0, + "learning_rate": 1.0607247366678508e-09, + "loss": 1.0869, + "step": 41395 + }, + { + "epoch": 1.0, + "learning_rate": 1.049410618259028e-09, + "loss": 1.0455, + "step": 41396 + }, + { + "epoch": 1.0, + "learning_rate": 1.0381571611306484e-09, + "loss": 1.1773, + "step": 41397 + }, + { + "epoch": 1.0, + "learning_rate": 1.0269643653526563e-09, + "loss": 0.8931, + "step": 41398 + }, + { + "epoch": 1.0, + "learning_rate": 1.015832230992775e-09, + "loss": 0.8427, + "step": 41399 + }, + { + "epoch": 1.0, + "learning_rate": 1.0047607581176178e-09, + "loss": 1.0437, + "step": 41400 + }, + { + "epoch": 1.0, + "learning_rate": 9.937499467949086e-10, + "loss": 1.1769, + "step": 41401 + }, + { + "epoch": 1.0, + "learning_rate": 9.827997970901503e-10, + "loss": 0.8467, + "step": 41402 + }, + { + "epoch": 1.0, + "learning_rate": 9.71910309072177e-10, + "loss": 0.7655, + "step": 41403 + }, + { + "epoch": 1.0, + "learning_rate": 9.610814828042715e-10, + "loss": 0.951, + "step": 41404 + }, + { + "epoch": 1.0, + "learning_rate": 9.503133183541568e-10, + "loss": 0.9218, + "step": 41405 + }, + { + "epoch": 1.0, + "learning_rate": 9.396058157873367e-10, + "loss": 1.142, + "step": 41406 + }, + { + "epoch": 1.0, + "learning_rate": 9.289589751670935e-10, + "loss": 1.1197, + "step": 41407 + }, + { + "epoch": 1.0, + "learning_rate": 9.183727965600409e-10, + "loss": 0.8511, + "step": 41408 + }, + { + "epoch": 1.0, + "learning_rate": 9.078472800283511e-10, + "loss": 1.2269, + "step": 41409 + }, + { + "epoch": 1.0, + "learning_rate": 8.973824256364172e-10, + "loss": 0.946, + "step": 41410 + }, + { + "epoch": 1.0, + "learning_rate": 8.869782334486321e-10, + "loss": 0.6748, + "step": 41411 + }, + { + "epoch": 1.0, + "learning_rate": 8.766347035282785e-10, + "loss": 0.9571, + "step": 41412 + }, + { + "epoch": 1.0, + "learning_rate": 8.663518359364187e-10, + "loss": 1.0681, + "step": 41413 + }, + { + "epoch": 1.0, + "learning_rate": 8.561296307374456e-10, + "loss": 0.8932, + "step": 41414 + }, + { + "epoch": 1.0, + "learning_rate": 8.459680879913112e-10, + "loss": 1.099, + "step": 41415 + }, + { + "epoch": 1.0, + "learning_rate": 8.358672077612983e-10, + "loss": 0.8147, + "step": 41416 + }, + { + "epoch": 1.0, + "learning_rate": 8.258269901084692e-10, + "loss": 1.1581, + "step": 41417 + }, + { + "epoch": 1.0, + "learning_rate": 8.158474350927758e-10, + "loss": 1.1392, + "step": 41418 + }, + { + "epoch": 1.0, + "learning_rate": 8.059285427752805e-10, + "loss": 0.9713, + "step": 41419 + }, + { + "epoch": 1.0, + "learning_rate": 7.960703132170456e-10, + "loss": 1.0262, + "step": 41420 + }, + { + "epoch": 1.0, + "learning_rate": 7.862727464769126e-10, + "loss": 0.8767, + "step": 41421 + }, + { + "epoch": 1.0, + "learning_rate": 7.765358426137237e-10, + "loss": 0.9871, + "step": 41422 + }, + { + "epoch": 1.0, + "learning_rate": 7.668596016874308e-10, + "loss": 0.9416, + "step": 41423 + }, + { + "epoch": 1.0, + "learning_rate": 7.572440237579859e-10, + "loss": 1.1289, + "step": 41424 + }, + { + "epoch": 1.0, + "learning_rate": 7.476891088809002e-10, + "loss": 1.1526, + "step": 41425 + }, + { + "epoch": 1.0, + "learning_rate": 7.381948571161257e-10, + "loss": 1.0559, + "step": 41426 + }, + { + "epoch": 1.0, + "learning_rate": 7.28761268521394e-10, + "loss": 1.0889, + "step": 41427 + }, + { + "epoch": 1.0, + "learning_rate": 7.193883431522163e-10, + "loss": 1.0022, + "step": 41428 + }, + { + "epoch": 1.0, + "learning_rate": 7.100760810674345e-10, + "loss": 0.9325, + "step": 41429 + }, + { + "epoch": 1.0, + "learning_rate": 7.008244823214494e-10, + "loss": 1.1498, + "step": 41430 + }, + { + "epoch": 1.0, + "learning_rate": 6.916335469731029e-10, + "loss": 0.9961, + "step": 41431 + }, + { + "epoch": 1.0, + "learning_rate": 6.825032750756855e-10, + "loss": 0.9179, + "step": 41432 + }, + { + "epoch": 1.0, + "learning_rate": 6.734336666858188e-10, + "loss": 1.1471, + "step": 41433 + }, + { + "epoch": 1.0, + "learning_rate": 6.644247218579036e-10, + "loss": 1.0705, + "step": 41434 + }, + { + "epoch": 1.0, + "learning_rate": 6.554764406474512e-10, + "loss": 1.1402, + "step": 41435 + }, + { + "epoch": 1.0, + "learning_rate": 6.465888231077522e-10, + "loss": 1.29, + "step": 41436 + }, + { + "epoch": 1.0, + "learning_rate": 6.377618692932075e-10, + "loss": 1.1157, + "step": 41437 + }, + { + "epoch": 1.0, + "learning_rate": 6.289955792582181e-10, + "loss": 1.1841, + "step": 41438 + }, + { + "epoch": 1.0, + "learning_rate": 6.202899530538542e-10, + "loss": 1.1656, + "step": 41439 + }, + { + "epoch": 1.0, + "learning_rate": 6.11644990735627e-10, + "loss": 1.1209, + "step": 41440 + }, + { + "epoch": 1.0, + "learning_rate": 6.030606923534965e-10, + "loss": 0.9956, + "step": 41441 + }, + { + "epoch": 1.0, + "learning_rate": 5.945370579607534e-10, + "loss": 1.1481, + "step": 41442 + }, + { + "epoch": 1.0, + "learning_rate": 5.860740876095783e-10, + "loss": 0.9992, + "step": 41443 + }, + { + "epoch": 1.0, + "learning_rate": 5.77671781349931e-10, + "loss": 1.1601, + "step": 41444 + }, + { + "epoch": 1.0, + "learning_rate": 5.693301392339922e-10, + "loss": 1.2101, + "step": 41445 + }, + { + "epoch": 1.0, + "learning_rate": 5.610491613117219e-10, + "loss": 1.1645, + "step": 41446 + }, + { + "epoch": 1.0, + "learning_rate": 5.5282884763308e-10, + "loss": 0.9314, + "step": 41447 + }, + { + "epoch": 1.0, + "learning_rate": 5.44669198249137e-10, + "loss": 1.1648, + "step": 41448 + }, + { + "epoch": 1.0, + "learning_rate": 5.365702132076322e-10, + "loss": 1.1454, + "step": 41449 + }, + { + "epoch": 1.0, + "learning_rate": 5.285318925596361e-10, + "loss": 1.0645, + "step": 41450 + }, + { + "epoch": 1.0, + "learning_rate": 5.205542363528882e-10, + "loss": 1.0945, + "step": 41451 + }, + { + "epoch": 1.0, + "learning_rate": 5.126372446362383e-10, + "loss": 1.0107, + "step": 41452 + }, + { + "epoch": 1.0, + "learning_rate": 5.047809174574258e-10, + "loss": 0.9557, + "step": 41453 + }, + { + "epoch": 1.0, + "learning_rate": 4.969852548630804e-10, + "loss": 0.9398, + "step": 41454 + }, + { + "epoch": 1.0, + "learning_rate": 4.892502569020518e-10, + "loss": 0.6204, + "step": 41455 + }, + { + "epoch": 1.0, + "learning_rate": 4.815759236209694e-10, + "loss": 1.1443, + "step": 41456 + }, + { + "epoch": 1.0, + "learning_rate": 4.739622550664624e-10, + "loss": 1.1428, + "step": 41457 + }, + { + "epoch": 1.0, + "learning_rate": 4.664092512840502e-10, + "loss": 0.7627, + "step": 41458 + }, + { + "epoch": 1.0, + "learning_rate": 4.5891691231925163e-10, + "loss": 1.136, + "step": 41459 + }, + { + "epoch": 1.0, + "learning_rate": 4.514852382186963e-10, + "loss": 0.795, + "step": 41460 + }, + { + "epoch": 1.0, + "learning_rate": 4.44114229026793e-10, + "loss": 1.1138, + "step": 41461 + }, + { + "epoch": 1.0, + "learning_rate": 4.36803884789061e-10, + "loss": 0.9027, + "step": 41462 + }, + { + "epoch": 1.0, + "learning_rate": 4.295542055487989e-10, + "loss": 1.0402, + "step": 41463 + }, + { + "epoch": 1.0, + "learning_rate": 4.2236519135041566e-10, + "loss": 1.2932, + "step": 41464 + }, + { + "epoch": 1.0, + "learning_rate": 4.1523684223720994e-10, + "loss": 0.9518, + "step": 41465 + }, + { + "epoch": 1.0, + "learning_rate": 4.081691582524805e-10, + "loss": 1.2546, + "step": 41466 + }, + { + "epoch": 1.0, + "learning_rate": 4.01162139439526e-10, + "loss": 1.2091, + "step": 41467 + }, + { + "epoch": 1.0, + "learning_rate": 3.942157858416451e-10, + "loss": 1.1734, + "step": 41468 + }, + { + "epoch": 1.0, + "learning_rate": 3.873300974988059e-10, + "loss": 0.8707, + "step": 41469 + }, + { + "epoch": 1.0, + "learning_rate": 3.805050744543071e-10, + "loss": 1.015, + "step": 41470 + }, + { + "epoch": 1.0, + "learning_rate": 3.7374071674922687e-10, + "loss": 0.8991, + "step": 41471 + }, + { + "epoch": 1.0, + "learning_rate": 3.670370244246435e-10, + "loss": 1.0342, + "step": 41472 + }, + { + "epoch": 1.0, + "learning_rate": 3.603939975216353e-10, + "loss": 1.0044, + "step": 41473 + }, + { + "epoch": 1.0, + "learning_rate": 3.5381163607906e-10, + "loss": 0.9509, + "step": 41474 + }, + { + "epoch": 1.0, + "learning_rate": 3.472899401391061e-10, + "loss": 1.3754, + "step": 41475 + }, + { + "epoch": 1.0, + "learning_rate": 3.408289097395212e-10, + "loss": 0.9622, + "step": 41476 + }, + { + "epoch": 1.0, + "learning_rate": 3.344285449202733e-10, + "loss": 0.88, + "step": 41477 + }, + { + "epoch": 1.0, + "learning_rate": 3.2808884571911006e-10, + "loss": 0.8774, + "step": 41478 + }, + { + "epoch": 1.0, + "learning_rate": 3.218098121759994e-10, + "loss": 1.2658, + "step": 41479 + }, + { + "epoch": 1.0, + "learning_rate": 3.1559144432868893e-10, + "loss": 1.0182, + "step": 41480 + }, + { + "epoch": 1.0, + "learning_rate": 3.0943374221381604e-10, + "loss": 0.8119, + "step": 41481 + }, + { + "epoch": 1.0, + "learning_rate": 3.033367058702386e-10, + "loss": 0.769, + "step": 41482 + }, + { + "epoch": 1.0, + "learning_rate": 2.973003353334836e-10, + "loss": 0.932, + "step": 41483 + }, + { + "epoch": 1.0, + "learning_rate": 2.913246306412987e-10, + "loss": 1.1531, + "step": 41484 + }, + { + "epoch": 1.0, + "learning_rate": 2.8540959183032124e-10, + "loss": 1.0088, + "step": 41485 + }, + { + "epoch": 1.0, + "learning_rate": 2.7955521893496817e-10, + "loss": 1.2959, + "step": 41486 + }, + { + "epoch": 1.0, + "learning_rate": 2.737615119918768e-10, + "loss": 0.8095, + "step": 41487 + }, + { + "epoch": 1.0, + "learning_rate": 2.6802847103546413e-10, + "loss": 0.9594, + "step": 41488 + }, + { + "epoch": 1.0, + "learning_rate": 2.62356096100147e-10, + "loss": 1.0112, + "step": 41489 + }, + { + "epoch": 1.0, + "learning_rate": 2.567443872225628e-10, + "loss": 0.985, + "step": 41490 + }, + { + "epoch": 1.0, + "learning_rate": 2.511933444337977e-10, + "loss": 1.0349, + "step": 41491 + }, + { + "epoch": 1.0, + "learning_rate": 2.4570296777048917e-10, + "loss": 1.0469, + "step": 41492 + }, + { + "epoch": 1.0, + "learning_rate": 2.402732572626132e-10, + "loss": 1.2486, + "step": 41493 + }, + { + "epoch": 1.0, + "learning_rate": 2.349042129468071e-10, + "loss": 1.0461, + "step": 41494 + }, + { + "epoch": 1.0, + "learning_rate": 2.295958348519367e-10, + "loss": 0.9731, + "step": 41495 + }, + { + "epoch": 1.0, + "learning_rate": 2.2434812301352915e-10, + "loss": 1.3519, + "step": 41496 + }, + { + "epoch": 1.0, + "learning_rate": 2.1916107746045023e-10, + "loss": 1.0384, + "step": 41497 + }, + { + "epoch": 1.0, + "learning_rate": 2.1403469822711686e-10, + "loss": 1.1883, + "step": 41498 + }, + { + "epoch": 1.0, + "learning_rate": 2.0896898534239485e-10, + "loss": 0.962, + "step": 41499 + }, + { + "epoch": 1.0, + "learning_rate": 2.0396393883737042e-10, + "loss": 0.9344, + "step": 41500 + }, + { + "epoch": 1.0, + "learning_rate": 1.9901955874312983e-10, + "loss": 0.9971, + "step": 41501 + }, + { + "epoch": 1.0, + "learning_rate": 1.941358450896491e-10, + "loss": 0.7803, + "step": 41502 + }, + { + "epoch": 1.0, + "learning_rate": 1.8931279790579404e-10, + "loss": 1.2344, + "step": 41503 + }, + { + "epoch": 1.0, + "learning_rate": 1.8455041722154066e-10, + "loss": 0.9412, + "step": 41504 + }, + { + "epoch": 1.0, + "learning_rate": 1.7984870306575474e-10, + "loss": 0.9901, + "step": 41505 + }, + { + "epoch": 1.0, + "learning_rate": 1.752076554661919e-10, + "loss": 1.0614, + "step": 41506 + }, + { + "epoch": 1.0, + "learning_rate": 1.706272744517179e-10, + "loss": 1.0042, + "step": 41507 + }, + { + "epoch": 1.0, + "learning_rate": 1.6610756005008832e-10, + "loss": 1.1075, + "step": 41508 + }, + { + "epoch": 1.0, + "learning_rate": 1.6164851228905876e-10, + "loss": 1.2975, + "step": 41509 + }, + { + "epoch": 1.0, + "learning_rate": 1.5725013119416432e-10, + "loss": 1.0674, + "step": 41510 + }, + { + "epoch": 1.0, + "learning_rate": 1.5291241679427084e-10, + "loss": 1.1373, + "step": 41511 + }, + { + "epoch": 1.0, + "learning_rate": 1.4863536911380317e-10, + "loss": 1.0689, + "step": 41512 + }, + { + "epoch": 1.0, + "learning_rate": 1.444189881794067e-10, + "loss": 1.1474, + "step": 41513 + }, + { + "epoch": 1.0, + "learning_rate": 1.4026327401661654e-10, + "loss": 0.904, + "step": 41514 + }, + { + "epoch": 1.0, + "learning_rate": 1.3616822665096786e-10, + "loss": 1.057, + "step": 41515 + }, + { + "epoch": 1.0, + "learning_rate": 1.3213384610799574e-10, + "loss": 1.183, + "step": 41516 + }, + { + "epoch": 1.0, + "learning_rate": 1.2816013240990465e-10, + "loss": 1.0708, + "step": 41517 + }, + { + "epoch": 1.0, + "learning_rate": 1.2424708558222976e-10, + "loss": 0.7212, + "step": 41518 + }, + { + "epoch": 1.0, + "learning_rate": 1.2039470564939593e-10, + "loss": 0.862, + "step": 41519 + }, + { + "epoch": 1.0, + "learning_rate": 1.1660299263360764e-10, + "loss": 0.9562, + "step": 41520 + }, + { + "epoch": 1.0, + "learning_rate": 1.1287194655817957e-10, + "loss": 1.012, + "step": 41521 + }, + { + "epoch": 1.0, + "learning_rate": 1.092015674464264e-10, + "loss": 0.9195, + "step": 41522 + }, + { + "epoch": 1.0, + "learning_rate": 1.0559185531944238e-10, + "loss": 1.1774, + "step": 41523 + }, + { + "epoch": 1.0, + "learning_rate": 1.0204281019943196e-10, + "loss": 1.097, + "step": 41524 + }, + { + "epoch": 1.0, + "learning_rate": 9.855443210859961e-11, + "loss": 0.9109, + "step": 41525 + }, + { + "epoch": 1.0, + "learning_rate": 9.512672106803955e-11, + "loss": 0.9565, + "step": 41526 + }, + { + "epoch": 1.0, + "learning_rate": 9.175967709773581e-11, + "loss": 1.091, + "step": 41527 + }, + { + "epoch": 1.0, + "learning_rate": 8.845330021878262e-11, + "loss": 0.8673, + "step": 41528 + }, + { + "epoch": 1.0, + "learning_rate": 8.5207590451164e-11, + "loss": 0.9941, + "step": 41529 + }, + { + "epoch": 1.0, + "learning_rate": 8.202254781486397e-11, + "loss": 1.1167, + "step": 41530 + }, + { + "epoch": 1.0, + "learning_rate": 7.889817232764607e-11, + "loss": 1.1941, + "step": 41531 + }, + { + "epoch": 1.0, + "learning_rate": 7.583446401060457e-11, + "loss": 0.8772, + "step": 41532 + }, + { + "epoch": 1.0, + "learning_rate": 7.283142288150303e-11, + "loss": 1.0203, + "step": 41533 + }, + { + "epoch": 1.0, + "learning_rate": 6.9889048958105e-11, + "loss": 1.0602, + "step": 41534 + }, + { + "epoch": 1.0, + "learning_rate": 6.700734225817407e-11, + "loss": 1.0593, + "step": 41535 + }, + { + "epoch": 1.0, + "learning_rate": 6.41863027994738e-11, + "loss": 1.0298, + "step": 41536 + }, + { + "epoch": 1.0, + "learning_rate": 6.142593059976775e-11, + "loss": 1.1832, + "step": 41537 + }, + { + "epoch": 1.0, + "learning_rate": 5.872622567459907e-11, + "loss": 1.0627, + "step": 41538 + }, + { + "epoch": 1.0, + "learning_rate": 5.6087188041731297e-11, + "loss": 0.8041, + "step": 41539 + }, + { + "epoch": 1.0, + "learning_rate": 5.350881771670757e-11, + "loss": 1.2391, + "step": 41540 + }, + { + "epoch": 1.0, + "learning_rate": 5.099111471396079e-11, + "loss": 0.7615, + "step": 41541 + }, + { + "epoch": 1.0, + "learning_rate": 4.853407905014429e-11, + "loss": 0.8625, + "step": 41542 + }, + { + "epoch": 1.0, + "learning_rate": 4.613771073969098e-11, + "loss": 1.0885, + "step": 41543 + }, + { + "epoch": 1.0, + "learning_rate": 4.380200979703375e-11, + "loss": 1.2792, + "step": 41544 + }, + { + "epoch": 1.0, + "learning_rate": 4.152697623660551e-11, + "loss": 1.0528, + "step": 41545 + }, + { + "epoch": 1.0, + "learning_rate": 3.931261007283915e-11, + "loss": 0.8557, + "step": 41546 + }, + { + "epoch": 1.0, + "learning_rate": 3.715891131794713e-11, + "loss": 1.0195, + "step": 41547 + }, + { + "epoch": 1.0, + "learning_rate": 3.506587998525213e-11, + "loss": 0.9632, + "step": 41548 + }, + { + "epoch": 1.0, + "learning_rate": 3.303351608807681e-11, + "loss": 1.173, + "step": 41549 + }, + { + "epoch": 1.0, + "learning_rate": 3.106181963752342e-11, + "loss": 0.9971, + "step": 41550 + }, + { + "epoch": 1.0, + "learning_rate": 2.9150790646914616e-11, + "loss": 0.7065, + "step": 41551 + }, + { + "epoch": 1.0, + "learning_rate": 2.7300429127352647e-11, + "loss": 1.0712, + "step": 41552 + }, + { + "epoch": 1.0, + "learning_rate": 2.551073508993973e-11, + "loss": 0.999, + "step": 41553 + }, + { + "epoch": 1.0, + "learning_rate": 2.3781708545778102e-11, + "loss": 0.956, + "step": 41554 + }, + { + "epoch": 1.0, + "learning_rate": 2.211334950485977e-11, + "loss": 0.954, + "step": 41555 + }, + { + "epoch": 1.0, + "learning_rate": 2.0505657978286964e-11, + "loss": 0.9652, + "step": 41556 + }, + { + "epoch": 1.0, + "learning_rate": 1.895863397383124e-11, + "loss": 0.9891, + "step": 41557 + }, + { + "epoch": 1.0, + "learning_rate": 1.747227750370506e-11, + "loss": 1.1865, + "step": 41558 + }, + { + "epoch": 1.0, + "learning_rate": 1.6046588574569754e-11, + "loss": 1.1262, + "step": 41559 + }, + { + "epoch": 1.0, + "learning_rate": 1.4681567196417333e-11, + "loss": 1.0285, + "step": 41560 + }, + { + "epoch": 1.0, + "learning_rate": 1.337721337701936e-11, + "loss": 0.8237, + "step": 41561 + }, + { + "epoch": 1.0, + "learning_rate": 1.2133527124147392e-11, + "loss": 0.8778, + "step": 41562 + }, + { + "epoch": 1.0, + "learning_rate": 1.0950508445572993e-11, + "loss": 1.0439, + "step": 41563 + }, + { + "epoch": 1.0, + "learning_rate": 9.828157347957501e-12, + "loss": 1.0872, + "step": 41564 + }, + { + "epoch": 1.0, + "learning_rate": 8.766473839072477e-12, + "loss": 1.1261, + "step": 41565 + }, + { + "epoch": 1.0, + "learning_rate": 7.765457924469033e-12, + "loss": 1.2951, + "step": 41566 + }, + { + "epoch": 1.0, + "learning_rate": 6.825109610808511e-12, + "loss": 1.2564, + "step": 41567 + }, + { + "epoch": 1.0, + "learning_rate": 5.945428903642025e-12, + "loss": 0.9172, + "step": 41568 + }, + { + "epoch": 1.0, + "learning_rate": 5.1264158085206904e-12, + "loss": 1.0609, + "step": 41569 + }, + { + "epoch": 1.0, + "learning_rate": 4.368070329885399e-12, + "loss": 0.8316, + "step": 41570 + }, + { + "epoch": 1.0, + "learning_rate": 3.6703924721770425e-12, + "loss": 0.967, + "step": 41571 + }, + { + "epoch": 1.0, + "learning_rate": 3.0333822409467362e-12, + "loss": 0.8384, + "step": 41572 + }, + { + "epoch": 1.0, + "learning_rate": 2.4570396384149264e-12, + "loss": 0.9949, + "step": 41573 + }, + { + "epoch": 1.0, + "learning_rate": 1.9413646690225053e-12, + "loss": 0.9741, + "step": 41574 + }, + { + "epoch": 1.0, + "learning_rate": 1.4863573361001416e-12, + "loss": 1.2461, + "step": 41575 + }, + { + "epoch": 1.0, + "learning_rate": 1.0920176418682815e-12, + "loss": 0.9918, + "step": 41576 + }, + { + "epoch": 1.0, + "learning_rate": 7.583455885473712e-13, + "loss": 0.886, + "step": 41577 + }, + { + "epoch": 1.0, + "learning_rate": 4.853411794680796e-13, + "loss": 1.056, + "step": 41578 + }, + { + "epoch": 1.0, + "learning_rate": 2.730044146304067e-13, + "loss": 0.8618, + "step": 41579 + }, + { + "epoch": 1.0, + "learning_rate": 1.2133529514457565e-13, + "loss": 0.9806, + "step": 41580 + }, + { + "epoch": 1.0, + "learning_rate": 3.033382434125542e-14, + "loss": 1.1274, + "step": 41581 + }, + { + "epoch": 1.0, + "learning_rate": 0.0, + "loss": 1.0317, + "step": 41582 + }, + { + "epoch": 1.0, + "step": 41582, + "total_flos": 6.067200514775319e+18, + "train_loss": 1.0930786170733737, + "train_runtime": 309940.5488, + "train_samples_per_second": 2.147, + "train_steps_per_second": 0.134 + } + ], + "logging_steps": 1.0, + "max_steps": 41582, + "num_input_tokens_seen": 0, + "num_train_epochs": 1, + "save_steps": 24000, + "total_flos": 6.067200514775319e+18, + "train_batch_size": 4, + "trial_name": null, + "trial_params": null +}