| { | |
| "best_metric": 6.38665771484375, | |
| "best_model_checkpoint": "/content/drive/MyDrive/AI\\ Camp/words-detector/checkpoint-800", | |
| "epoch": 3.0, | |
| "global_step": 876, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 4.9942922374429226e-05, | |
| "loss": 17.8686, | |
| "step": 2 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 4.9942922374429226e-05, | |
| "loss": 18.2644, | |
| "step": 4 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 4.9942922374429226e-05, | |
| "loss": 17.8102, | |
| "step": 6 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 4.9942922374429226e-05, | |
| "loss": 18.0033, | |
| "step": 8 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 4.9828767123287674e-05, | |
| "loss": 15.0071, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 4.971461187214612e-05, | |
| "loss": 10.0967, | |
| "step": 12 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 4.960045662100457e-05, | |
| "loss": 9.4344, | |
| "step": 14 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 4.948630136986301e-05, | |
| "loss": 9.3036, | |
| "step": 16 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 4.937214611872146e-05, | |
| "loss": 8.8597, | |
| "step": 18 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 4.9257990867579914e-05, | |
| "loss": 8.7451, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 4.914383561643836e-05, | |
| "loss": 9.0271, | |
| "step": 22 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 4.9029680365296804e-05, | |
| "loss": 8.3687, | |
| "step": 24 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 4.891552511415525e-05, | |
| "loss": 8.7046, | |
| "step": 26 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 4.88013698630137e-05, | |
| "loss": 8.2987, | |
| "step": 28 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 4.868721461187215e-05, | |
| "loss": 8.3789, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 4.8573059360730596e-05, | |
| "loss": 7.7792, | |
| "step": 32 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 4.8458904109589044e-05, | |
| "loss": 8.1005, | |
| "step": 34 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 4.834474885844749e-05, | |
| "loss": 8.0337, | |
| "step": 36 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 4.823059360730594e-05, | |
| "loss": 7.8095, | |
| "step": 38 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 4.811643835616438e-05, | |
| "loss": 7.5034, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 4.8002283105022836e-05, | |
| "loss": 7.6029, | |
| "step": 42 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 4.7888127853881284e-05, | |
| "loss": 7.7257, | |
| "step": 44 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 4.777397260273973e-05, | |
| "loss": 7.4601, | |
| "step": 46 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 4.765981735159817e-05, | |
| "loss": 7.7843, | |
| "step": 48 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 4.754566210045662e-05, | |
| "loss": 7.7357, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 4.743150684931507e-05, | |
| "loss": 7.1895, | |
| "step": 52 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 4.7317351598173523e-05, | |
| "loss": 7.2758, | |
| "step": 54 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 4.7203196347031965e-05, | |
| "loss": 7.818, | |
| "step": 56 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 4.708904109589041e-05, | |
| "loss": 7.4382, | |
| "step": 58 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 4.697488584474886e-05, | |
| "loss": 7.4169, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 4.686073059360731e-05, | |
| "loss": 7.5099, | |
| "step": 62 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 4.674657534246576e-05, | |
| "loss": 7.309, | |
| "step": 64 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 4.6632420091324205e-05, | |
| "loss": 7.4911, | |
| "step": 66 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 4.651826484018265e-05, | |
| "loss": 7.595, | |
| "step": 68 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 4.64041095890411e-05, | |
| "loss": 7.3767, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 4.628995433789954e-05, | |
| "loss": 7.3467, | |
| "step": 72 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 4.617579908675799e-05, | |
| "loss": 7.4988, | |
| "step": 74 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 4.6061643835616445e-05, | |
| "loss": 7.2302, | |
| "step": 76 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 4.594748858447489e-05, | |
| "loss": 7.0727, | |
| "step": 78 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 4.5833333333333334e-05, | |
| "loss": 7.5031, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 4.571917808219178e-05, | |
| "loss": 7.1357, | |
| "step": 82 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 4.560502283105023e-05, | |
| "loss": 7.1202, | |
| "step": 84 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 4.549086757990868e-05, | |
| "loss": 7.2538, | |
| "step": 86 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 4.5376712328767126e-05, | |
| "loss": 7.071, | |
| "step": 88 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 4.5262557077625574e-05, | |
| "loss": 7.5201, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 4.514840182648402e-05, | |
| "loss": 7.5082, | |
| "step": 92 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 4.503424657534247e-05, | |
| "loss": 7.1858, | |
| "step": 94 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 4.492009132420091e-05, | |
| "loss": 7.2599, | |
| "step": 96 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 4.4805936073059366e-05, | |
| "loss": 7.3367, | |
| "step": 98 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 4.4691780821917814e-05, | |
| "loss": 6.9713, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "eval_cer": 0.7946802397175929, | |
| "eval_loss": 7.137301445007324, | |
| "eval_runtime": 208.567, | |
| "eval_samples_per_second": 2.795, | |
| "eval_steps_per_second": 0.35, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 4.457762557077626e-05, | |
| "loss": 6.8404, | |
| "step": 102 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 4.44634703196347e-05, | |
| "loss": 7.0133, | |
| "step": 104 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 4.434931506849315e-05, | |
| "loss": 7.2514, | |
| "step": 106 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 4.42351598173516e-05, | |
| "loss": 7.2344, | |
| "step": 108 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 4.412100456621005e-05, | |
| "loss": 7.1263, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 4.4006849315068495e-05, | |
| "loss": 7.1429, | |
| "step": 112 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 4.389269406392694e-05, | |
| "loss": 7.422, | |
| "step": 114 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 4.377853881278539e-05, | |
| "loss": 7.2147, | |
| "step": 116 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 4.366438356164384e-05, | |
| "loss": 6.9941, | |
| "step": 118 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 4.355022831050228e-05, | |
| "loss": 7.0812, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 4.3436073059360735e-05, | |
| "loss": 7.4399, | |
| "step": 122 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 4.332191780821918e-05, | |
| "loss": 6.9418, | |
| "step": 124 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 4.320776255707763e-05, | |
| "loss": 7.0305, | |
| "step": 126 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 4.309360730593607e-05, | |
| "loss": 7.262, | |
| "step": 128 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 4.297945205479452e-05, | |
| "loss": 7.092, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 4.286529680365297e-05, | |
| "loss": 6.9877, | |
| "step": 132 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 4.275114155251142e-05, | |
| "loss": 6.825, | |
| "step": 134 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 4.2636986301369864e-05, | |
| "loss": 6.6798, | |
| "step": 136 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 4.252283105022831e-05, | |
| "loss": 6.9598, | |
| "step": 138 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 4.240867579908676e-05, | |
| "loss": 6.6182, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 4.229452054794521e-05, | |
| "loss": 7.0775, | |
| "step": 142 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 4.2180365296803656e-05, | |
| "loss": 7.0996, | |
| "step": 144 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 4.2066210045662104e-05, | |
| "loss": 6.9547, | |
| "step": 146 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.195205479452055e-05, | |
| "loss": 6.8326, | |
| "step": 148 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.1837899543379e-05, | |
| "loss": 7.1501, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.172374429223744e-05, | |
| "loss": 7.0974, | |
| "step": 152 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.160958904109589e-05, | |
| "loss": 6.8626, | |
| "step": 154 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.1495433789954344e-05, | |
| "loss": 6.4989, | |
| "step": 156 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.138127853881279e-05, | |
| "loss": 6.7234, | |
| "step": 158 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.126712328767123e-05, | |
| "loss": 6.7773, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.115296803652968e-05, | |
| "loss": 6.9259, | |
| "step": 162 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 4.103881278538813e-05, | |
| "loss": 7.0744, | |
| "step": 164 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 4.092465753424658e-05, | |
| "loss": 6.8039, | |
| "step": 166 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 4.0810502283105025e-05, | |
| "loss": 6.8271, | |
| "step": 168 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 4.069634703196347e-05, | |
| "loss": 6.7427, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 4.058219178082192e-05, | |
| "loss": 6.5537, | |
| "step": 172 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 4.046803652968037e-05, | |
| "loss": 6.6001, | |
| "step": 174 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 4.035388127853881e-05, | |
| "loss": 6.6144, | |
| "step": 176 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 4.0239726027397265e-05, | |
| "loss": 6.8667, | |
| "step": 178 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 4.012557077625571e-05, | |
| "loss": 7.0148, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 4.001141552511416e-05, | |
| "loss": 6.7102, | |
| "step": 182 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 3.98972602739726e-05, | |
| "loss": 6.8212, | |
| "step": 184 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 3.978310502283105e-05, | |
| "loss": 6.6284, | |
| "step": 186 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 3.96689497716895e-05, | |
| "loss": 6.5087, | |
| "step": 188 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 3.9554794520547946e-05, | |
| "loss": 7.1345, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 3.9440639269406394e-05, | |
| "loss": 7.0863, | |
| "step": 192 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 3.932648401826484e-05, | |
| "loss": 6.8418, | |
| "step": 194 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 3.921232876712329e-05, | |
| "loss": 7.0545, | |
| "step": 196 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 3.909817351598174e-05, | |
| "loss": 7.1057, | |
| "step": 198 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 3.8984018264840186e-05, | |
| "loss": 6.6551, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "eval_cer": 0.8836712913553896, | |
| "eval_loss": 6.700944900512695, | |
| "eval_runtime": 40.0981, | |
| "eval_samples_per_second": 14.539, | |
| "eval_steps_per_second": 1.821, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 3.8869863013698634e-05, | |
| "loss": 6.7477, | |
| "step": 202 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 3.875570776255708e-05, | |
| "loss": 6.9312, | |
| "step": 204 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 3.864155251141553e-05, | |
| "loss": 6.6784, | |
| "step": 206 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 3.852739726027397e-05, | |
| "loss": 6.9644, | |
| "step": 208 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.841324200913242e-05, | |
| "loss": 6.5752, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.829908675799087e-05, | |
| "loss": 7.0168, | |
| "step": 212 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.818493150684932e-05, | |
| "loss": 6.7026, | |
| "step": 214 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.8070776255707764e-05, | |
| "loss": 6.8185, | |
| "step": 216 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.795662100456621e-05, | |
| "loss": 6.7238, | |
| "step": 218 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.784246575342466e-05, | |
| "loss": 6.7133, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 3.772831050228311e-05, | |
| "loss": 6.771, | |
| "step": 222 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 3.7614155251141555e-05, | |
| "loss": 6.8149, | |
| "step": 224 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 3.7500000000000003e-05, | |
| "loss": 6.6512, | |
| "step": 226 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 3.738584474885845e-05, | |
| "loss": 6.213, | |
| "step": 228 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 3.727168949771689e-05, | |
| "loss": 6.7743, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 3.715753424657534e-05, | |
| "loss": 6.7843, | |
| "step": 232 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 3.704337899543379e-05, | |
| "loss": 6.6052, | |
| "step": 234 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.6929223744292243e-05, | |
| "loss": 6.5319, | |
| "step": 236 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.6815068493150685e-05, | |
| "loss": 6.4869, | |
| "step": 238 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.670091324200913e-05, | |
| "loss": 6.7245, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.658675799086758e-05, | |
| "loss": 6.8325, | |
| "step": 242 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 3.647260273972603e-05, | |
| "loss": 6.6019, | |
| "step": 244 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 3.635844748858448e-05, | |
| "loss": 6.2614, | |
| "step": 246 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 3.6244292237442925e-05, | |
| "loss": 7.0083, | |
| "step": 248 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 3.613013698630137e-05, | |
| "loss": 6.4853, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 3.601598173515982e-05, | |
| "loss": 6.7758, | |
| "step": 252 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 3.590182648401826e-05, | |
| "loss": 6.7792, | |
| "step": 254 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 3.578767123287671e-05, | |
| "loss": 6.6991, | |
| "step": 256 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 3.5673515981735165e-05, | |
| "loss": 6.5657, | |
| "step": 258 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 3.555936073059361e-05, | |
| "loss": 6.7322, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 3.5445205479452054e-05, | |
| "loss": 6.4798, | |
| "step": 262 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 3.53310502283105e-05, | |
| "loss": 6.7884, | |
| "step": 264 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 3.521689497716895e-05, | |
| "loss": 6.424, | |
| "step": 266 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.51027397260274e-05, | |
| "loss": 6.6925, | |
| "step": 268 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.4988584474885846e-05, | |
| "loss": 7.0121, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 3.4874429223744294e-05, | |
| "loss": 6.8289, | |
| "step": 272 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 3.476027397260274e-05, | |
| "loss": 6.7504, | |
| "step": 274 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 3.464611872146119e-05, | |
| "loss": 6.6898, | |
| "step": 276 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 3.453196347031963e-05, | |
| "loss": 6.7379, | |
| "step": 278 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 3.4417808219178086e-05, | |
| "loss": 6.4448, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 3.4303652968036534e-05, | |
| "loss": 6.678, | |
| "step": 282 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 3.418949771689498e-05, | |
| "loss": 6.4519, | |
| "step": 284 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 3.407534246575342e-05, | |
| "loss": 6.5809, | |
| "step": 286 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 3.396118721461187e-05, | |
| "loss": 6.5462, | |
| "step": 288 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 3.384703196347032e-05, | |
| "loss": 6.4562, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 3.373287671232877e-05, | |
| "loss": 6.5566, | |
| "step": 292 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 3.3618721461187215e-05, | |
| "loss": 6.4929, | |
| "step": 294 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 3.350456621004566e-05, | |
| "loss": 6.5228, | |
| "step": 296 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 3.339041095890411e-05, | |
| "loss": 6.6275, | |
| "step": 298 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 3.327625570776256e-05, | |
| "loss": 6.2089, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "eval_cer": 0.7978408997619243, | |
| "eval_loss": 6.618570327758789, | |
| "eval_runtime": 34.5444, | |
| "eval_samples_per_second": 16.877, | |
| "eval_steps_per_second": 2.113, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 3.316210045662101e-05, | |
| "loss": 6.3685, | |
| "step": 302 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 3.3047945205479455e-05, | |
| "loss": 6.1731, | |
| "step": 304 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 3.29337899543379e-05, | |
| "loss": 6.359, | |
| "step": 306 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 3.281963470319635e-05, | |
| "loss": 6.2789, | |
| "step": 308 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 3.270547945205479e-05, | |
| "loss": 6.2938, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 3.259132420091324e-05, | |
| "loss": 6.3706, | |
| "step": 312 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 3.247716894977169e-05, | |
| "loss": 6.4628, | |
| "step": 314 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 3.236301369863014e-05, | |
| "loss": 5.9578, | |
| "step": 316 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 3.2248858447488584e-05, | |
| "loss": 6.3769, | |
| "step": 318 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 3.213470319634703e-05, | |
| "loss": 6.7954, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 3.202054794520548e-05, | |
| "loss": 6.7632, | |
| "step": 322 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 3.190639269406393e-05, | |
| "loss": 6.3459, | |
| "step": 324 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 3.1792237442922376e-05, | |
| "loss": 6.4654, | |
| "step": 326 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 3.1678082191780824e-05, | |
| "loss": 6.6141, | |
| "step": 328 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 3.156392694063927e-05, | |
| "loss": 6.2931, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 3.144977168949772e-05, | |
| "loss": 5.9657, | |
| "step": 332 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 3.133561643835616e-05, | |
| "loss": 6.6218, | |
| "step": 334 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 3.122146118721461e-05, | |
| "loss": 6.2199, | |
| "step": 336 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 3.1107305936073064e-05, | |
| "loss": 6.5139, | |
| "step": 338 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 3.099315068493151e-05, | |
| "loss": 6.5002, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 3.087899543378995e-05, | |
| "loss": 6.3457, | |
| "step": 342 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 3.07648401826484e-05, | |
| "loss": 6.1393, | |
| "step": 344 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 3.065068493150685e-05, | |
| "loss": 6.0003, | |
| "step": 346 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 3.05365296803653e-05, | |
| "loss": 6.237, | |
| "step": 348 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 3.0422374429223742e-05, | |
| "loss": 6.4162, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 3.0308219178082193e-05, | |
| "loss": 6.4468, | |
| "step": 352 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 3.019406392694064e-05, | |
| "loss": 6.6031, | |
| "step": 354 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 3.007990867579909e-05, | |
| "loss": 6.3145, | |
| "step": 356 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 2.9965753424657534e-05, | |
| "loss": 6.3375, | |
| "step": 358 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 2.9851598173515982e-05, | |
| "loss": 6.4769, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 2.973744292237443e-05, | |
| "loss": 6.0631, | |
| "step": 362 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 2.962328767123288e-05, | |
| "loss": 6.2347, | |
| "step": 364 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 2.9509132420091322e-05, | |
| "loss": 6.0945, | |
| "step": 366 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 2.9394977168949774e-05, | |
| "loss": 6.2858, | |
| "step": 368 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 2.9280821917808222e-05, | |
| "loss": 6.3601, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 2.916666666666667e-05, | |
| "loss": 6.3449, | |
| "step": 372 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 2.9052511415525114e-05, | |
| "loss": 6.3585, | |
| "step": 374 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 2.8938356164383562e-05, | |
| "loss": 6.0235, | |
| "step": 376 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 2.882420091324201e-05, | |
| "loss": 6.1927, | |
| "step": 378 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 2.871004566210046e-05, | |
| "loss": 6.4136, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 2.8595890410958903e-05, | |
| "loss": 6.4073, | |
| "step": 382 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 2.848173515981735e-05, | |
| "loss": 6.4599, | |
| "step": 384 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 2.8367579908675802e-05, | |
| "loss": 6.1341, | |
| "step": 386 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 2.825342465753425e-05, | |
| "loss": 6.0782, | |
| "step": 388 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 2.813926940639269e-05, | |
| "loss": 6.1504, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 2.8025114155251143e-05, | |
| "loss": 6.3133, | |
| "step": 392 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 2.791095890410959e-05, | |
| "loss": 6.4356, | |
| "step": 394 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 2.779680365296804e-05, | |
| "loss": 6.4857, | |
| "step": 396 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 2.7682648401826484e-05, | |
| "loss": 6.4011, | |
| "step": 398 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 2.756849315068493e-05, | |
| "loss": 6.1637, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "eval_cer": 0.9765208110992529, | |
| "eval_loss": 6.528327941894531, | |
| "eval_runtime": 37.1998, | |
| "eval_samples_per_second": 15.672, | |
| "eval_steps_per_second": 1.962, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 2.745433789954338e-05, | |
| "loss": 6.3984, | |
| "step": 402 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 2.734018264840183e-05, | |
| "loss": 6.0221, | |
| "step": 404 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 2.7226027397260272e-05, | |
| "loss": 6.2472, | |
| "step": 406 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 2.7111872146118723e-05, | |
| "loss": 6.333, | |
| "step": 408 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 2.699771689497717e-05, | |
| "loss": 6.2516, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 2.688356164383562e-05, | |
| "loss": 6.2314, | |
| "step": 412 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 2.6769406392694064e-05, | |
| "loss": 6.1203, | |
| "step": 414 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 2.6655251141552512e-05, | |
| "loss": 6.311, | |
| "step": 416 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 2.654109589041096e-05, | |
| "loss": 6.1741, | |
| "step": 418 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 2.6426940639269408e-05, | |
| "loss": 6.0226, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 2.6312785388127853e-05, | |
| "loss": 6.1573, | |
| "step": 422 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 2.61986301369863e-05, | |
| "loss": 6.3298, | |
| "step": 424 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 2.6084474885844752e-05, | |
| "loss": 6.3197, | |
| "step": 426 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 2.59703196347032e-05, | |
| "loss": 6.3265, | |
| "step": 428 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 2.5856164383561645e-05, | |
| "loss": 6.5904, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 2.5742009132420093e-05, | |
| "loss": 6.0295, | |
| "step": 432 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 2.562785388127854e-05, | |
| "loss": 6.2887, | |
| "step": 434 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 2.551369863013699e-05, | |
| "loss": 6.2315, | |
| "step": 436 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 2.5399543378995433e-05, | |
| "loss": 6.1131, | |
| "step": 438 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 2.528538812785388e-05, | |
| "loss": 6.0897, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 2.517123287671233e-05, | |
| "loss": 6.2543, | |
| "step": 442 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 2.505707762557078e-05, | |
| "loss": 6.2021, | |
| "step": 444 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 2.4942922374429225e-05, | |
| "loss": 6.2414, | |
| "step": 446 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 2.4828767123287673e-05, | |
| "loss": 6.3754, | |
| "step": 448 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 2.471461187214612e-05, | |
| "loss": 6.4199, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 2.4600456621004566e-05, | |
| "loss": 6.1184, | |
| "step": 452 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 2.4486301369863017e-05, | |
| "loss": 6.308, | |
| "step": 454 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 2.4372146118721462e-05, | |
| "loss": 6.4365, | |
| "step": 456 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 2.425799086757991e-05, | |
| "loss": 6.0473, | |
| "step": 458 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 2.4143835616438358e-05, | |
| "loss": 6.055, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 2.4029680365296806e-05, | |
| "loss": 6.0705, | |
| "step": 462 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 2.391552511415525e-05, | |
| "loss": 6.0425, | |
| "step": 464 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 2.3801369863013702e-05, | |
| "loss": 5.9382, | |
| "step": 466 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 2.3687214611872146e-05, | |
| "loss": 6.0501, | |
| "step": 468 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 2.3573059360730594e-05, | |
| "loss": 6.1394, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 2.3458904109589042e-05, | |
| "loss": 6.4409, | |
| "step": 472 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 2.334474885844749e-05, | |
| "loss": 6.1642, | |
| "step": 474 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 2.3230593607305935e-05, | |
| "loss": 6.1099, | |
| "step": 476 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 2.3116438356164386e-05, | |
| "loss": 6.1933, | |
| "step": 478 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 2.300228310502283e-05, | |
| "loss": 6.2047, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 2.288812785388128e-05, | |
| "loss": 6.0044, | |
| "step": 482 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 2.2773972602739727e-05, | |
| "loss": 6.0892, | |
| "step": 484 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 2.2659817351598175e-05, | |
| "loss": 6.245, | |
| "step": 486 | |
| }, | |
| { | |
| "epoch": 1.67, | |
| "learning_rate": 2.2545662100456623e-05, | |
| "loss": 6.1838, | |
| "step": 488 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 2.243150684931507e-05, | |
| "loss": 5.7336, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 2.2317351598173515e-05, | |
| "loss": 6.1649, | |
| "step": 492 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 2.2203196347031967e-05, | |
| "loss": 6.001, | |
| "step": 494 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 2.208904109589041e-05, | |
| "loss": 6.6607, | |
| "step": 496 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 2.197488584474886e-05, | |
| "loss": 5.9693, | |
| "step": 498 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 2.1860730593607307e-05, | |
| "loss": 6.2159, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "eval_cer": 0.8505459321894754, | |
| "eval_loss": 6.468786716461182, | |
| "eval_runtime": 38.2571, | |
| "eval_samples_per_second": 15.239, | |
| "eval_steps_per_second": 1.908, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 2.1746575342465755e-05, | |
| "loss": 6.5233, | |
| "step": 502 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 2.16324200913242e-05, | |
| "loss": 6.0243, | |
| "step": 504 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 2.151826484018265e-05, | |
| "loss": 6.0025, | |
| "step": 506 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 2.1404109589041096e-05, | |
| "loss": 6.19, | |
| "step": 508 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 2.1289954337899544e-05, | |
| "loss": 6.1283, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 2.1175799086757992e-05, | |
| "loss": 5.9461, | |
| "step": 512 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 2.106164383561644e-05, | |
| "loss": 6.163, | |
| "step": 514 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 2.0947488584474888e-05, | |
| "loss": 5.9075, | |
| "step": 516 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 2.0833333333333336e-05, | |
| "loss": 6.061, | |
| "step": 518 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 2.071917808219178e-05, | |
| "loss": 5.8001, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 2.060502283105023e-05, | |
| "loss": 6.3543, | |
| "step": 522 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 2.0490867579908677e-05, | |
| "loss": 6.1044, | |
| "step": 524 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 2.0376712328767125e-05, | |
| "loss": 6.0392, | |
| "step": 526 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 2.0262557077625573e-05, | |
| "loss": 6.3613, | |
| "step": 528 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 2.014840182648402e-05, | |
| "loss": 6.1969, | |
| "step": 530 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 2.0034246575342465e-05, | |
| "loss": 6.2186, | |
| "step": 532 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 1.9920091324200917e-05, | |
| "loss": 6.5516, | |
| "step": 534 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 1.980593607305936e-05, | |
| "loss": 6.2319, | |
| "step": 536 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 1.969178082191781e-05, | |
| "loss": 6.1139, | |
| "step": 538 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 1.9577625570776257e-05, | |
| "loss": 6.6036, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 1.9463470319634705e-05, | |
| "loss": 6.3099, | |
| "step": 542 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 1.934931506849315e-05, | |
| "loss": 6.2986, | |
| "step": 544 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 1.92351598173516e-05, | |
| "loss": 6.0614, | |
| "step": 546 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 1.9121004566210046e-05, | |
| "loss": 5.8707, | |
| "step": 548 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 1.9006849315068494e-05, | |
| "loss": 6.4741, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 1.8892694063926942e-05, | |
| "loss": 5.9351, | |
| "step": 552 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 1.877853881278539e-05, | |
| "loss": 6.3108, | |
| "step": 554 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 1.8664383561643838e-05, | |
| "loss": 5.8719, | |
| "step": 556 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 1.8550228310502286e-05, | |
| "loss": 5.9978, | |
| "step": 558 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 1.843607305936073e-05, | |
| "loss": 5.9883, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 1.832191780821918e-05, | |
| "loss": 5.7932, | |
| "step": 562 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 1.8207762557077626e-05, | |
| "loss": 6.1634, | |
| "step": 564 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 1.8093607305936074e-05, | |
| "loss": 6.2713, | |
| "step": 566 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 1.7979452054794522e-05, | |
| "loss": 6.2679, | |
| "step": 568 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 1.786529680365297e-05, | |
| "loss": 6.2969, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 1.7751141552511415e-05, | |
| "loss": 6.0192, | |
| "step": 572 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 1.7636986301369866e-05, | |
| "loss": 5.9912, | |
| "step": 574 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 1.752283105022831e-05, | |
| "loss": 6.2804, | |
| "step": 576 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 1.740867579908676e-05, | |
| "loss": 6.1972, | |
| "step": 578 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 1.7294520547945207e-05, | |
| "loss": 6.0236, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 1.7180365296803655e-05, | |
| "loss": 5.9499, | |
| "step": 582 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 1.70662100456621e-05, | |
| "loss": 6.1688, | |
| "step": 584 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 1.695205479452055e-05, | |
| "loss": 5.9289, | |
| "step": 586 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 1.6837899543378995e-05, | |
| "loss": 5.5866, | |
| "step": 588 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 1.6723744292237443e-05, | |
| "loss": 6.2259, | |
| "step": 590 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 1.660958904109589e-05, | |
| "loss": 5.9349, | |
| "step": 592 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 1.649543378995434e-05, | |
| "loss": 6.0916, | |
| "step": 594 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 1.6381278538812787e-05, | |
| "loss": 5.8322, | |
| "step": 596 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 1.6267123287671235e-05, | |
| "loss": 6.1475, | |
| "step": 598 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 1.615296803652968e-05, | |
| "loss": 6.1342, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "eval_cer": 0.8386831951399721, | |
| "eval_loss": 6.42148494720459, | |
| "eval_runtime": 43.5835, | |
| "eval_samples_per_second": 13.377, | |
| "eval_steps_per_second": 1.675, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 1.603881278538813e-05, | |
| "loss": 5.7958, | |
| "step": 602 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 1.5924657534246576e-05, | |
| "loss": 6.3319, | |
| "step": 604 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 1.5810502283105024e-05, | |
| "loss": 6.181, | |
| "step": 606 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 1.5696347031963472e-05, | |
| "loss": 6.1645, | |
| "step": 608 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 1.558219178082192e-05, | |
| "loss": 5.8628, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 1.5468036529680365e-05, | |
| "loss": 6.1039, | |
| "step": 612 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 1.5353881278538816e-05, | |
| "loss": 6.196, | |
| "step": 614 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 1.523972602739726e-05, | |
| "loss": 6.0791, | |
| "step": 616 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 1.512557077625571e-05, | |
| "loss": 6.2871, | |
| "step": 618 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 1.5011415525114157e-05, | |
| "loss": 5.8731, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 1.4897260273972605e-05, | |
| "loss": 6.1108, | |
| "step": 622 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 1.4783105022831051e-05, | |
| "loss": 5.9225, | |
| "step": 624 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 1.4668949771689499e-05, | |
| "loss": 6.1894, | |
| "step": 626 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 1.4554794520547945e-05, | |
| "loss": 5.7433, | |
| "step": 628 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 1.4440639269406395e-05, | |
| "loss": 6.0826, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 1.4326484018264841e-05, | |
| "loss": 6.0097, | |
| "step": 632 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 1.421232876712329e-05, | |
| "loss": 5.8454, | |
| "step": 634 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 1.4098173515981735e-05, | |
| "loss": 6.1941, | |
| "step": 636 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 1.3984018264840185e-05, | |
| "loss": 5.8484, | |
| "step": 638 | |
| }, | |
| { | |
| "epoch": 2.19, | |
| "learning_rate": 1.3869863013698631e-05, | |
| "loss": 6.328, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 1.375570776255708e-05, | |
| "loss": 6.3417, | |
| "step": 642 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 1.3641552511415526e-05, | |
| "loss": 6.1993, | |
| "step": 644 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 1.3527397260273974e-05, | |
| "loss": 6.0958, | |
| "step": 646 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 1.341324200913242e-05, | |
| "loss": 6.0639, | |
| "step": 648 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 1.329908675799087e-05, | |
| "loss": 6.0299, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 1.3184931506849316e-05, | |
| "loss": 6.1234, | |
| "step": 652 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 1.3070776255707764e-05, | |
| "loss": 6.0303, | |
| "step": 654 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 1.295662100456621e-05, | |
| "loss": 6.0098, | |
| "step": 656 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 1.284246575342466e-05, | |
| "loss": 5.9058, | |
| "step": 658 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 1.2728310502283106e-05, | |
| "loss": 6.2086, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 1.2614155251141554e-05, | |
| "loss": 6.1667, | |
| "step": 662 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 1.25e-05, | |
| "loss": 6.26, | |
| "step": 664 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 1.2385844748858449e-05, | |
| "loss": 5.698, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 1.2271689497716895e-05, | |
| "loss": 6.0933, | |
| "step": 668 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 1.2157534246575343e-05, | |
| "loss": 6.0448, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 1.2043378995433791e-05, | |
| "loss": 6.1017, | |
| "step": 672 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 1.1929223744292239e-05, | |
| "loss": 6.2529, | |
| "step": 674 | |
| }, | |
| { | |
| "epoch": 2.32, | |
| "learning_rate": 1.1815068493150685e-05, | |
| "loss": 5.9125, | |
| "step": 676 | |
| }, | |
| { | |
| "epoch": 2.32, | |
| "learning_rate": 1.1700913242009133e-05, | |
| "loss": 6.0414, | |
| "step": 678 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 1.1586757990867581e-05, | |
| "loss": 6.1012, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 1.1472602739726027e-05, | |
| "loss": 6.1578, | |
| "step": 682 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 1.1358447488584475e-05, | |
| "loss": 6.1924, | |
| "step": 684 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "learning_rate": 1.1244292237442923e-05, | |
| "loss": 5.9418, | |
| "step": 686 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 1.113013698630137e-05, | |
| "loss": 6.0778, | |
| "step": 688 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 1.1015981735159818e-05, | |
| "loss": 6.1784, | |
| "step": 690 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 1.0901826484018266e-05, | |
| "loss": 6.2346, | |
| "step": 692 | |
| }, | |
| { | |
| "epoch": 2.38, | |
| "learning_rate": 1.0787671232876714e-05, | |
| "loss": 6.0849, | |
| "step": 694 | |
| }, | |
| { | |
| "epoch": 2.38, | |
| "learning_rate": 1.067351598173516e-05, | |
| "loss": 6.0025, | |
| "step": 696 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 1.0559360730593608e-05, | |
| "loss": 6.2221, | |
| "step": 698 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 1.0445205479452056e-05, | |
| "loss": 5.8295, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "eval_cer": 0.7918069124045645, | |
| "eval_loss": 6.3999199867248535, | |
| "eval_runtime": 40.7886, | |
| "eval_samples_per_second": 14.293, | |
| "eval_steps_per_second": 1.79, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 1.0331050228310502e-05, | |
| "loss": 5.8848, | |
| "step": 702 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 1.021689497716895e-05, | |
| "loss": 6.0501, | |
| "step": 704 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "learning_rate": 1.0102739726027398e-05, | |
| "loss": 6.1242, | |
| "step": 706 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "learning_rate": 9.988584474885845e-06, | |
| "loss": 5.9263, | |
| "step": 708 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 9.874429223744293e-06, | |
| "loss": 5.897, | |
| "step": 710 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 9.76027397260274e-06, | |
| "loss": 6.3139, | |
| "step": 712 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "learning_rate": 9.646118721461189e-06, | |
| "loss": 5.9507, | |
| "step": 714 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "learning_rate": 9.531963470319635e-06, | |
| "loss": 5.9616, | |
| "step": 716 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 9.417808219178083e-06, | |
| "loss": 6.0359, | |
| "step": 718 | |
| }, | |
| { | |
| "epoch": 2.47, | |
| "learning_rate": 9.303652968036531e-06, | |
| "loss": 5.8949, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 2.47, | |
| "learning_rate": 9.189497716894977e-06, | |
| "loss": 6.1793, | |
| "step": 722 | |
| }, | |
| { | |
| "epoch": 2.48, | |
| "learning_rate": 9.075342465753425e-06, | |
| "loss": 5.9515, | |
| "step": 724 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 8.961187214611873e-06, | |
| "loss": 5.8504, | |
| "step": 726 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 8.84703196347032e-06, | |
| "loss": 6.0409, | |
| "step": 728 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 8.732876712328767e-06, | |
| "loss": 6.2968, | |
| "step": 730 | |
| }, | |
| { | |
| "epoch": 2.51, | |
| "learning_rate": 8.618721461187215e-06, | |
| "loss": 5.8559, | |
| "step": 732 | |
| }, | |
| { | |
| "epoch": 2.51, | |
| "learning_rate": 8.504566210045663e-06, | |
| "loss": 5.7911, | |
| "step": 734 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "learning_rate": 8.39041095890411e-06, | |
| "loss": 6.1524, | |
| "step": 736 | |
| }, | |
| { | |
| "epoch": 2.53, | |
| "learning_rate": 8.276255707762558e-06, | |
| "loss": 5.7204, | |
| "step": 738 | |
| }, | |
| { | |
| "epoch": 2.53, | |
| "learning_rate": 8.162100456621006e-06, | |
| "loss": 6.1016, | |
| "step": 740 | |
| }, | |
| { | |
| "epoch": 2.54, | |
| "learning_rate": 8.047945205479452e-06, | |
| "loss": 5.9837, | |
| "step": 742 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 7.9337899543379e-06, | |
| "loss": 5.8908, | |
| "step": 744 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 7.819634703196348e-06, | |
| "loss": 5.843, | |
| "step": 746 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "learning_rate": 7.705479452054794e-06, | |
| "loss": 5.9089, | |
| "step": 748 | |
| }, | |
| { | |
| "epoch": 2.57, | |
| "learning_rate": 7.591324200913242e-06, | |
| "loss": 6.0063, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "learning_rate": 7.47716894977169e-06, | |
| "loss": 5.9461, | |
| "step": 752 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "learning_rate": 7.3630136986301374e-06, | |
| "loss": 5.6393, | |
| "step": 754 | |
| }, | |
| { | |
| "epoch": 2.59, | |
| "learning_rate": 7.2488584474885854e-06, | |
| "loss": 6.0984, | |
| "step": 756 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "learning_rate": 7.134703196347033e-06, | |
| "loss": 5.7783, | |
| "step": 758 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "learning_rate": 7.02054794520548e-06, | |
| "loss": 5.8534, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 2.61, | |
| "learning_rate": 6.906392694063928e-06, | |
| "loss": 5.9853, | |
| "step": 762 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "learning_rate": 6.792237442922375e-06, | |
| "loss": 6.2198, | |
| "step": 764 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "learning_rate": 6.678082191780823e-06, | |
| "loss": 6.1982, | |
| "step": 766 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "learning_rate": 6.56392694063927e-06, | |
| "loss": 5.9753, | |
| "step": 768 | |
| }, | |
| { | |
| "epoch": 2.64, | |
| "learning_rate": 6.449771689497717e-06, | |
| "loss": 5.806, | |
| "step": 770 | |
| }, | |
| { | |
| "epoch": 2.64, | |
| "learning_rate": 6.335616438356165e-06, | |
| "loss": 5.8563, | |
| "step": 772 | |
| }, | |
| { | |
| "epoch": 2.65, | |
| "learning_rate": 6.2214611872146115e-06, | |
| "loss": 6.0158, | |
| "step": 774 | |
| }, | |
| { | |
| "epoch": 2.66, | |
| "learning_rate": 6.1073059360730594e-06, | |
| "loss": 6.2569, | |
| "step": 776 | |
| }, | |
| { | |
| "epoch": 2.66, | |
| "learning_rate": 5.993150684931507e-06, | |
| "loss": 5.9379, | |
| "step": 778 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 5.878995433789955e-06, | |
| "loss": 5.914, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 2.68, | |
| "learning_rate": 5.764840182648402e-06, | |
| "loss": 5.847, | |
| "step": 782 | |
| }, | |
| { | |
| "epoch": 2.68, | |
| "learning_rate": 5.650684931506849e-06, | |
| "loss": 5.9925, | |
| "step": 784 | |
| }, | |
| { | |
| "epoch": 2.69, | |
| "learning_rate": 5.536529680365297e-06, | |
| "loss": 5.7515, | |
| "step": 786 | |
| }, | |
| { | |
| "epoch": 2.7, | |
| "learning_rate": 5.422374429223744e-06, | |
| "loss": 6.0283, | |
| "step": 788 | |
| }, | |
| { | |
| "epoch": 2.71, | |
| "learning_rate": 5.308219178082192e-06, | |
| "loss": 5.8207, | |
| "step": 790 | |
| }, | |
| { | |
| "epoch": 2.71, | |
| "learning_rate": 5.194063926940639e-06, | |
| "loss": 5.9529, | |
| "step": 792 | |
| }, | |
| { | |
| "epoch": 2.72, | |
| "learning_rate": 5.079908675799086e-06, | |
| "loss": 6.3816, | |
| "step": 794 | |
| }, | |
| { | |
| "epoch": 2.73, | |
| "learning_rate": 4.965753424657534e-06, | |
| "loss": 5.921, | |
| "step": 796 | |
| }, | |
| { | |
| "epoch": 2.73, | |
| "learning_rate": 4.8515981735159814e-06, | |
| "loss": 5.9369, | |
| "step": 798 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "learning_rate": 4.7374429223744294e-06, | |
| "loss": 6.0542, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "eval_cer": 0.7485838601100074, | |
| "eval_loss": 6.38665771484375, | |
| "eval_runtime": 43.0264, | |
| "eval_samples_per_second": 13.55, | |
| "eval_steps_per_second": 1.697, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 4.623287671232877e-06, | |
| "loss": 5.6539, | |
| "step": 802 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 4.509132420091325e-06, | |
| "loss": 6.0837, | |
| "step": 804 | |
| }, | |
| { | |
| "epoch": 2.76, | |
| "learning_rate": 4.394977168949772e-06, | |
| "loss": 5.5705, | |
| "step": 806 | |
| }, | |
| { | |
| "epoch": 2.77, | |
| "learning_rate": 4.280821917808219e-06, | |
| "loss": 5.797, | |
| "step": 808 | |
| }, | |
| { | |
| "epoch": 2.77, | |
| "learning_rate": 4.166666666666667e-06, | |
| "loss": 6.2038, | |
| "step": 810 | |
| }, | |
| { | |
| "epoch": 2.78, | |
| "learning_rate": 4.052511415525114e-06, | |
| "loss": 5.8418, | |
| "step": 812 | |
| }, | |
| { | |
| "epoch": 2.79, | |
| "learning_rate": 3.938356164383562e-06, | |
| "loss": 5.9495, | |
| "step": 814 | |
| }, | |
| { | |
| "epoch": 2.79, | |
| "learning_rate": 3.824200913242009e-06, | |
| "loss": 6.2109, | |
| "step": 816 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "learning_rate": 3.7100456621004567e-06, | |
| "loss": 5.8276, | |
| "step": 818 | |
| }, | |
| { | |
| "epoch": 2.81, | |
| "learning_rate": 3.5958904109589043e-06, | |
| "loss": 5.8859, | |
| "step": 820 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "learning_rate": 3.4817351598173514e-06, | |
| "loss": 5.9916, | |
| "step": 822 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "learning_rate": 3.367579908675799e-06, | |
| "loss": 6.2421, | |
| "step": 824 | |
| }, | |
| { | |
| "epoch": 2.83, | |
| "learning_rate": 3.2534246575342466e-06, | |
| "loss": 5.9774, | |
| "step": 826 | |
| }, | |
| { | |
| "epoch": 2.84, | |
| "learning_rate": 3.139269406392694e-06, | |
| "loss": 5.6032, | |
| "step": 828 | |
| }, | |
| { | |
| "epoch": 2.84, | |
| "learning_rate": 3.0251141552511417e-06, | |
| "loss": 6.0698, | |
| "step": 830 | |
| }, | |
| { | |
| "epoch": 2.85, | |
| "learning_rate": 2.910958904109589e-06, | |
| "loss": 6.1411, | |
| "step": 832 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "learning_rate": 2.7968036529680364e-06, | |
| "loss": 6.0411, | |
| "step": 834 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "learning_rate": 2.682648401826484e-06, | |
| "loss": 5.9259, | |
| "step": 836 | |
| }, | |
| { | |
| "epoch": 2.87, | |
| "learning_rate": 2.5684931506849316e-06, | |
| "loss": 6.1342, | |
| "step": 838 | |
| }, | |
| { | |
| "epoch": 2.88, | |
| "learning_rate": 2.454337899543379e-06, | |
| "loss": 5.9385, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 2.88, | |
| "learning_rate": 2.3401826484018263e-06, | |
| "loss": 6.1046, | |
| "step": 842 | |
| }, | |
| { | |
| "epoch": 2.89, | |
| "learning_rate": 2.226027397260274e-06, | |
| "loss": 5.5534, | |
| "step": 844 | |
| }, | |
| { | |
| "epoch": 2.9, | |
| "learning_rate": 2.1118721461187214e-06, | |
| "loss": 6.0064, | |
| "step": 846 | |
| }, | |
| { | |
| "epoch": 2.9, | |
| "learning_rate": 1.997716894977169e-06, | |
| "loss": 5.9357, | |
| "step": 848 | |
| }, | |
| { | |
| "epoch": 2.91, | |
| "learning_rate": 1.8835616438356164e-06, | |
| "loss": 5.591, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 1.769406392694064e-06, | |
| "loss": 6.0179, | |
| "step": 852 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 1.6552511415525115e-06, | |
| "loss": 6.0487, | |
| "step": 854 | |
| }, | |
| { | |
| "epoch": 2.93, | |
| "learning_rate": 1.5410958904109589e-06, | |
| "loss": 6.0967, | |
| "step": 856 | |
| }, | |
| { | |
| "epoch": 2.94, | |
| "learning_rate": 1.4269406392694064e-06, | |
| "loss": 6.1703, | |
| "step": 858 | |
| }, | |
| { | |
| "epoch": 2.95, | |
| "learning_rate": 1.3127853881278538e-06, | |
| "loss": 5.8776, | |
| "step": 860 | |
| }, | |
| { | |
| "epoch": 2.95, | |
| "learning_rate": 1.1986301369863014e-06, | |
| "loss": 5.8851, | |
| "step": 862 | |
| }, | |
| { | |
| "epoch": 2.96, | |
| "learning_rate": 1.084474885844749e-06, | |
| "loss": 5.7484, | |
| "step": 864 | |
| }, | |
| { | |
| "epoch": 2.97, | |
| "learning_rate": 9.703196347031963e-07, | |
| "loss": 6.1194, | |
| "step": 866 | |
| }, | |
| { | |
| "epoch": 2.97, | |
| "learning_rate": 8.561643835616439e-07, | |
| "loss": 5.8983, | |
| "step": 868 | |
| }, | |
| { | |
| "epoch": 2.98, | |
| "learning_rate": 7.420091324200913e-07, | |
| "loss": 6.2007, | |
| "step": 870 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "learning_rate": 6.278538812785388e-07, | |
| "loss": 5.9643, | |
| "step": 872 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "learning_rate": 5.136986301369863e-07, | |
| "loss": 5.7904, | |
| "step": 874 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 3.9954337899543377e-07, | |
| "loss": 5.9321, | |
| "step": 876 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "step": 876, | |
| "total_flos": 8.368943131860664e+17, | |
| "train_loss": 6.569673863720132, | |
| "train_runtime": 1471.7945, | |
| "train_samples_per_second": 4.753, | |
| "train_steps_per_second": 0.595 | |
| } | |
| ], | |
| "max_steps": 876, | |
| "num_train_epochs": 3, | |
| "total_flos": 8.368943131860664e+17, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |