| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 3.998694942903752, | |
| "eval_steps": 500, | |
| "global_step": 3064, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 4.347826086956521e-08, | |
| "loss": 1.726, | |
| "step": 2 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 8.695652173913042e-08, | |
| "loss": 1.7175, | |
| "step": 4 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.3043478260869563e-07, | |
| "loss": 1.7023, | |
| "step": 6 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.7391304347826085e-07, | |
| "loss": 1.6936, | |
| "step": 8 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 2.1739130434782607e-07, | |
| "loss": 1.7156, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 2.6086956521739126e-07, | |
| "loss": 1.7237, | |
| "step": 12 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 3.043478260869565e-07, | |
| "loss": 1.7174, | |
| "step": 14 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 3.478260869565217e-07, | |
| "loss": 1.7304, | |
| "step": 16 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 3.9130434782608694e-07, | |
| "loss": 1.7269, | |
| "step": 18 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 4.3478260869565214e-07, | |
| "loss": 1.7134, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 4.782608695652174e-07, | |
| "loss": 1.6357, | |
| "step": 22 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 5.217391304347825e-07, | |
| "loss": 1.708, | |
| "step": 24 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 5.652173913043477e-07, | |
| "loss": 1.6731, | |
| "step": 26 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 6.08695652173913e-07, | |
| "loss": 1.6733, | |
| "step": 28 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 6.521739130434782e-07, | |
| "loss": 1.7246, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 6.956521739130434e-07, | |
| "loss": 1.6945, | |
| "step": 32 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 7.391304347826086e-07, | |
| "loss": 1.6814, | |
| "step": 34 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 7.826086956521739e-07, | |
| "loss": 1.7222, | |
| "step": 36 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 8.260869565217391e-07, | |
| "loss": 1.6781, | |
| "step": 38 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 8.695652173913043e-07, | |
| "loss": 1.6882, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.130434782608695e-07, | |
| "loss": 1.7144, | |
| "step": 42 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.565217391304349e-07, | |
| "loss": 1.7113, | |
| "step": 44 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1e-06, | |
| "loss": 1.7489, | |
| "step": 46 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.043478260869565e-06, | |
| "loss": 1.7169, | |
| "step": 48 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.0869565217391303e-06, | |
| "loss": 1.7396, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.1304347826086954e-06, | |
| "loss": 1.6864, | |
| "step": 52 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.173913043478261e-06, | |
| "loss": 1.76, | |
| "step": 54 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.217391304347826e-06, | |
| "loss": 1.7147, | |
| "step": 56 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.2608695652173913e-06, | |
| "loss": 1.7296, | |
| "step": 58 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.3043478260869564e-06, | |
| "loss": 1.7195, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.3478260869565217e-06, | |
| "loss": 1.7456, | |
| "step": 62 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.3913043478260868e-06, | |
| "loss": 1.8055, | |
| "step": 64 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.434782608695652e-06, | |
| "loss": 1.6711, | |
| "step": 66 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.4782608695652172e-06, | |
| "loss": 1.6789, | |
| "step": 68 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.5217391304347827e-06, | |
| "loss": 1.6871, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.5652173913043478e-06, | |
| "loss": 1.7404, | |
| "step": 72 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.608695652173913e-06, | |
| "loss": 1.7063, | |
| "step": 74 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.6521739130434782e-06, | |
| "loss": 1.7236, | |
| "step": 76 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.6956521739130435e-06, | |
| "loss": 1.7363, | |
| "step": 78 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.7391304347826085e-06, | |
| "loss": 1.6989, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.7826086956521738e-06, | |
| "loss": 1.7187, | |
| "step": 82 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.826086956521739e-06, | |
| "loss": 1.7102, | |
| "step": 84 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.8695652173913044e-06, | |
| "loss": 1.7201, | |
| "step": 86 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9130434782608697e-06, | |
| "loss": 1.7161, | |
| "step": 88 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9565217391304346e-06, | |
| "loss": 1.6945, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2e-06, | |
| "loss": 1.6973, | |
| "step": 92 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9999977652344436e-06, | |
| "loss": 1.6626, | |
| "step": 94 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9999910609477626e-06, | |
| "loss": 1.7519, | |
| "step": 96 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9999798871699223e-06, | |
| "loss": 1.7536, | |
| "step": 98 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9999642439508647e-06, | |
| "loss": 1.7231, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9999441313605068e-06, | |
| "loss": 1.6999, | |
| "step": 102 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9999195494887424e-06, | |
| "loss": 1.7409, | |
| "step": 104 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.999890498445442e-06, | |
| "loss": 1.6538, | |
| "step": 106 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.999856978360449e-06, | |
| "loss": 1.7398, | |
| "step": 108 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9998189893835827e-06, | |
| "loss": 1.6626, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.999776531684637e-06, | |
| "loss": 1.6812, | |
| "step": 112 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9997296054533767e-06, | |
| "loss": 1.6895, | |
| "step": 114 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.99967821089954e-06, | |
| "loss": 1.7202, | |
| "step": 116 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9996223482528377e-06, | |
| "loss": 1.7228, | |
| "step": 118 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9995620177629485e-06, | |
| "loss": 1.7067, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.999497219699522e-06, | |
| "loss": 1.6683, | |
| "step": 122 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.999427954352175e-06, | |
| "loss": 1.7136, | |
| "step": 124 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9993542220304907e-06, | |
| "loss": 1.6608, | |
| "step": 126 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.9992760230640188e-06, | |
| "loss": 1.7298, | |
| "step": 128 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.9991933578022713e-06, | |
| "loss": 1.7399, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.9991062266147236e-06, | |
| "loss": 1.7385, | |
| "step": 132 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.999014629890811e-06, | |
| "loss": 1.7144, | |
| "step": 134 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.998918568039928e-06, | |
| "loss": 1.6974, | |
| "step": 136 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.998818041491426e-06, | |
| "loss": 1.6432, | |
| "step": 138 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.998713050694612e-06, | |
| "loss": 1.7132, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.9986035961187444e-06, | |
| "loss": 1.7044, | |
| "step": 142 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.9984896782530355e-06, | |
| "loss": 1.6433, | |
| "step": 144 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.9983712976066435e-06, | |
| "loss": 1.7169, | |
| "step": 146 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.9982484547086748e-06, | |
| "loss": 1.7329, | |
| "step": 148 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.9981211501081795e-06, | |
| "loss": 1.6872, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.9979893843741495e-06, | |
| "loss": 1.655, | |
| "step": 152 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.9978531580955158e-06, | |
| "loss": 1.6696, | |
| "step": 154 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.9977124718811463e-06, | |
| "loss": 1.693, | |
| "step": 156 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.9975673263598416e-06, | |
| "loss": 1.7035, | |
| "step": 158 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.9974177221803353e-06, | |
| "loss": 1.6509, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.997263660011287e-06, | |
| "loss": 1.6743, | |
| "step": 162 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.9971051405412828e-06, | |
| "loss": 1.7133, | |
| "step": 164 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.9969421644788302e-06, | |
| "loss": 1.7181, | |
| "step": 166 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.996774732552356e-06, | |
| "loss": 1.6975, | |
| "step": 168 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.996602845510202e-06, | |
| "loss": 1.7002, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.996426504120623e-06, | |
| "loss": 1.6724, | |
| "step": 172 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.9962457091717822e-06, | |
| "loss": 1.7257, | |
| "step": 174 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.9960604614717485e-06, | |
| "loss": 1.7096, | |
| "step": 176 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.995870761848492e-06, | |
| "loss": 1.7282, | |
| "step": 178 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.995676611149881e-06, | |
| "loss": 1.6936, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.9954780102436785e-06, | |
| "loss": 1.7012, | |
| "step": 182 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.995274960017537e-06, | |
| "loss": 1.7187, | |
| "step": 184 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.9950674613789958e-06, | |
| "loss": 1.654, | |
| "step": 186 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.994855515255477e-06, | |
| "loss": 1.6977, | |
| "step": 188 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.9946391225942798e-06, | |
| "loss": 1.7194, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.994418284362578e-06, | |
| "loss": 1.7294, | |
| "step": 192 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.994193001547416e-06, | |
| "loss": 1.7039, | |
| "step": 194 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.993963275155701e-06, | |
| "loss": 1.7086, | |
| "step": 196 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.9937291062142028e-06, | |
| "loss": 1.7105, | |
| "step": 198 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.9934904957695468e-06, | |
| "loss": 1.6925, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.9932474448882093e-06, | |
| "loss": 1.6923, | |
| "step": 202 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.9929999546565148e-06, | |
| "loss": 1.7375, | |
| "step": 204 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.9927480261806276e-06, | |
| "loss": 1.6615, | |
| "step": 206 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.9924916605865506e-06, | |
| "loss": 1.6564, | |
| "step": 208 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.992230859020117e-06, | |
| "loss": 1.695, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.9919656226469885e-06, | |
| "loss": 1.6644, | |
| "step": 212 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.9916959526526466e-06, | |
| "loss": 1.6498, | |
| "step": 214 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.9914218502423895e-06, | |
| "loss": 1.7261, | |
| "step": 216 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.9911433166413275e-06, | |
| "loss": 1.7343, | |
| "step": 218 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.9908603530943742e-06, | |
| "loss": 1.7154, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.9905729608662447e-06, | |
| "loss": 1.7613, | |
| "step": 222 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.9902811412414467e-06, | |
| "loss": 1.6935, | |
| "step": 224 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.9899848955242778e-06, | |
| "loss": 1.6729, | |
| "step": 226 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.9896842250388174e-06, | |
| "loss": 1.6515, | |
| "step": 228 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.9893791311289216e-06, | |
| "loss": 1.6641, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.9890696151582166e-06, | |
| "loss": 1.7039, | |
| "step": 232 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.9887556785100937e-06, | |
| "loss": 1.6975, | |
| "step": 234 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.9884373225877028e-06, | |
| "loss": 1.7449, | |
| "step": 236 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.988114548813946e-06, | |
| "loss": 1.6453, | |
| "step": 238 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.98778735863147e-06, | |
| "loss": 1.6413, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.9874557535026623e-06, | |
| "loss": 1.7127, | |
| "step": 242 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.987119734909641e-06, | |
| "loss": 1.6847, | |
| "step": 244 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.986779304354253e-06, | |
| "loss": 1.6642, | |
| "step": 246 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.9864344633580628e-06, | |
| "loss": 1.6848, | |
| "step": 248 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.9860852134623482e-06, | |
| "loss": 1.6777, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.985731556228092e-06, | |
| "loss": 1.7287, | |
| "step": 252 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.985373493235976e-06, | |
| "loss": 1.7026, | |
| "step": 254 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.985011026086375e-06, | |
| "loss": 1.6671, | |
| "step": 256 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.984644156399346e-06, | |
| "loss": 1.6576, | |
| "step": 258 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.984272885814625e-06, | |
| "loss": 1.7098, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.9838972159916176e-06, | |
| "loss": 1.677, | |
| "step": 262 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.9835171486093916e-06, | |
| "loss": 1.6591, | |
| "step": 264 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.98313268536667e-06, | |
| "loss": 1.7512, | |
| "step": 266 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.9827438279818234e-06, | |
| "loss": 1.6488, | |
| "step": 268 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.9823505781928613e-06, | |
| "loss": 1.6573, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.9819529377574263e-06, | |
| "loss": 1.6737, | |
| "step": 272 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.981550908452785e-06, | |
| "loss": 1.6792, | |
| "step": 274 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.9811444920758196e-06, | |
| "loss": 1.6532, | |
| "step": 276 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.9807336904430206e-06, | |
| "loss": 1.6684, | |
| "step": 278 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.9803185053904792e-06, | |
| "loss": 1.6911, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.9798989387738773e-06, | |
| "loss": 1.6826, | |
| "step": 282 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.9794749924684814e-06, | |
| "loss": 1.6848, | |
| "step": 284 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.979046668369132e-06, | |
| "loss": 1.6931, | |
| "step": 286 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.9786139683902377e-06, | |
| "loss": 1.6803, | |
| "step": 288 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.978176894465765e-06, | |
| "loss": 1.7187, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.977735448549228e-06, | |
| "loss": 1.6719, | |
| "step": 292 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.977289632613684e-06, | |
| "loss": 1.6764, | |
| "step": 294 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.9768394486517205e-06, | |
| "loss": 1.7275, | |
| "step": 296 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.9763848986754494e-06, | |
| "loss": 1.6961, | |
| "step": 298 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.9759259847164957e-06, | |
| "loss": 1.7397, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.975462708825989e-06, | |
| "loss": 1.6787, | |
| "step": 302 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.9749950730745562e-06, | |
| "loss": 1.7241, | |
| "step": 304 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.97452307955231e-06, | |
| "loss": 1.6624, | |
| "step": 306 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.974046730368839e-06, | |
| "loss": 1.7419, | |
| "step": 308 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.973566027653202e-06, | |
| "loss": 1.7145, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.9730809735539135e-06, | |
| "loss": 1.6548, | |
| "step": 312 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.9725915702389388e-06, | |
| "loss": 1.6818, | |
| "step": 314 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.9720978198956806e-06, | |
| "loss": 1.6362, | |
| "step": 316 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.971599724730972e-06, | |
| "loss": 1.6918, | |
| "step": 318 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.9710972869710646e-06, | |
| "loss": 1.6339, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.9705905088616194e-06, | |
| "loss": 1.6672, | |
| "step": 322 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.9700793926676972e-06, | |
| "loss": 1.6967, | |
| "step": 324 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.9695639406737476e-06, | |
| "loss": 1.7376, | |
| "step": 326 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.9690441551835994e-06, | |
| "loss": 1.6528, | |
| "step": 328 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.9685200385204496e-06, | |
| "loss": 1.6948, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.967991593026855e-06, | |
| "loss": 1.718, | |
| "step": 332 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.9674588210647182e-06, | |
| "loss": 1.7174, | |
| "step": 334 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.96692172501528e-06, | |
| "loss": 1.6923, | |
| "step": 336 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.966380307279109e-06, | |
| "loss": 1.646, | |
| "step": 338 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.965834570276087e-06, | |
| "loss": 1.6976, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.965284516445404e-06, | |
| "loss": 1.6921, | |
| "step": 342 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.964730148245542e-06, | |
| "loss": 1.6751, | |
| "step": 344 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.9641714681542667e-06, | |
| "loss": 1.6766, | |
| "step": 346 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.963608478668617e-06, | |
| "loss": 1.7215, | |
| "step": 348 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.963041182304891e-06, | |
| "loss": 1.6842, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.962469581598638e-06, | |
| "loss": 1.6541, | |
| "step": 352 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.961893679104645e-06, | |
| "loss": 1.6841, | |
| "step": 354 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.9613134773969254e-06, | |
| "loss": 1.7075, | |
| "step": 356 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.9607289790687104e-06, | |
| "loss": 1.7045, | |
| "step": 358 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.960140186732432e-06, | |
| "loss": 1.679, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.9595471030197163e-06, | |
| "loss": 1.7257, | |
| "step": 362 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.95894973058137e-06, | |
| "loss": 1.691, | |
| "step": 364 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.958348072087367e-06, | |
| "loss": 1.6648, | |
| "step": 366 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.9577421302268393e-06, | |
| "loss": 1.697, | |
| "step": 368 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.957131907708062e-06, | |
| "loss": 1.7119, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.9565174072584445e-06, | |
| "loss": 1.6618, | |
| "step": 372 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.9558986316245153e-06, | |
| "loss": 1.7103, | |
| "step": 374 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.9552755835719113e-06, | |
| "loss": 1.6766, | |
| "step": 376 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.954648265885366e-06, | |
| "loss": 1.6481, | |
| "step": 378 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.9540166813686936e-06, | |
| "loss": 1.6343, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.9533808328447825e-06, | |
| "loss": 1.6898, | |
| "step": 382 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.9527407231555762e-06, | |
| "loss": 1.7079, | |
| "step": 384 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.9520963551620657e-06, | |
| "loss": 1.656, | |
| "step": 386 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.951447731744273e-06, | |
| "loss": 1.7079, | |
| "step": 388 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.9507948558012416e-06, | |
| "loss": 1.6665, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.9501377302510204e-06, | |
| "loss": 1.7302, | |
| "step": 392 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.949476358030652e-06, | |
| "loss": 1.6232, | |
| "step": 394 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.948810742096161e-06, | |
| "loss": 1.6972, | |
| "step": 396 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.9481408854225377e-06, | |
| "loss": 1.7178, | |
| "step": 398 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.947466791003728e-06, | |
| "loss": 1.6614, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.9467884618526175e-06, | |
| "loss": 1.6718, | |
| "step": 402 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 1.94610590100102e-06, | |
| "loss": 1.7231, | |
| "step": 404 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 1.9454191114996614e-06, | |
| "loss": 1.6678, | |
| "step": 406 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 1.9447280964181696e-06, | |
| "loss": 1.6821, | |
| "step": 408 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.944032858845058e-06, | |
| "loss": 1.7083, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.9433334018877117e-06, | |
| "loss": 1.6879, | |
| "step": 412 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.9426297286723765e-06, | |
| "loss": 1.6622, | |
| "step": 414 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.9419218423441413e-06, | |
| "loss": 1.739, | |
| "step": 416 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 1.9412097460669255e-06, | |
| "loss": 1.6973, | |
| "step": 418 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 1.940493443023466e-06, | |
| "loss": 1.6889, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 1.939772936415302e-06, | |
| "loss": 1.6539, | |
| "step": 422 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 1.93904822946276e-06, | |
| "loss": 1.6705, | |
| "step": 424 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.9383193254049396e-06, | |
| "loss": 1.7566, | |
| "step": 426 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.9375862274997007e-06, | |
| "loss": 1.606, | |
| "step": 428 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.936848939023647e-06, | |
| "loss": 1.7073, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.9361074632721124e-06, | |
| "loss": 1.6675, | |
| "step": 432 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.935361803559146e-06, | |
| "loss": 1.7374, | |
| "step": 434 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.934611963217497e-06, | |
| "loss": 1.6816, | |
| "step": 436 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.9338579455985995e-06, | |
| "loss": 1.6804, | |
| "step": 438 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.9330997540725597e-06, | |
| "loss": 1.6717, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.932337392028138e-06, | |
| "loss": 1.6788, | |
| "step": 442 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.9315708628727343e-06, | |
| "loss": 1.6859, | |
| "step": 444 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.930800170032376e-06, | |
| "loss": 1.655, | |
| "step": 446 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.9300253169516975e-06, | |
| "loss": 1.6668, | |
| "step": 448 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 1.9292463070939292e-06, | |
| "loss": 1.6822, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 1.9284631439408805e-06, | |
| "loss": 1.7345, | |
| "step": 452 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 1.9276758309929225e-06, | |
| "loss": 1.7079, | |
| "step": 454 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.9268843717689753e-06, | |
| "loss": 1.6829, | |
| "step": 456 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.926088769806491e-06, | |
| "loss": 1.7123, | |
| "step": 458 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.9252890286614366e-06, | |
| "loss": 1.6493, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.9244851519082798e-06, | |
| "loss": 1.6968, | |
| "step": 462 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.923677143139974e-06, | |
| "loss": 1.6282, | |
| "step": 464 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.9228650059679387e-06, | |
| "loss": 1.6724, | |
| "step": 466 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.922048744022046e-06, | |
| "loss": 1.6737, | |
| "step": 468 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.9212283609506046e-06, | |
| "loss": 1.6696, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.920403860420342e-06, | |
| "loss": 1.6488, | |
| "step": 472 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.919575246116389e-06, | |
| "loss": 1.6218, | |
| "step": 474 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.918742521742263e-06, | |
| "loss": 1.6535, | |
| "step": 476 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.9179056910198514e-06, | |
| "loss": 1.6459, | |
| "step": 478 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.9170647576893952e-06, | |
| "loss": 1.6519, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.916219725509472e-06, | |
| "loss": 1.6875, | |
| "step": 482 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.91537059825698e-06, | |
| "loss": 1.6848, | |
| "step": 484 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.9145173797271186e-06, | |
| "loss": 1.6329, | |
| "step": 486 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.913660073733376e-06, | |
| "loss": 1.6982, | |
| "step": 488 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.912798684107507e-06, | |
| "loss": 1.6685, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.9119332146995203e-06, | |
| "loss": 1.7009, | |
| "step": 492 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.9110636693776573e-06, | |
| "loss": 1.6551, | |
| "step": 494 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.9101900520283784e-06, | |
| "loss": 1.6704, | |
| "step": 496 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.9093123665563434e-06, | |
| "loss": 1.6666, | |
| "step": 498 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.9084306168843946e-06, | |
| "loss": 1.6699, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.90754480695354e-06, | |
| "loss": 1.6998, | |
| "step": 502 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.906654940722935e-06, | |
| "loss": 1.6992, | |
| "step": 504 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.9057610221698634e-06, | |
| "loss": 1.7065, | |
| "step": 506 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.9048630552897226e-06, | |
| "loss": 1.6426, | |
| "step": 508 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.9039610440960034e-06, | |
| "loss": 1.6601, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.903054992620273e-06, | |
| "loss": 1.6394, | |
| "step": 512 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.9021449049121565e-06, | |
| "loss": 1.6503, | |
| "step": 514 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.9012307850393193e-06, | |
| "loss": 1.6865, | |
| "step": 516 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.900312637087449e-06, | |
| "loss": 1.6529, | |
| "step": 518 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.8993904651602356e-06, | |
| "loss": 1.6776, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.8984642733793554e-06, | |
| "loss": 1.6201, | |
| "step": 522 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.8975340658844522e-06, | |
| "loss": 1.6725, | |
| "step": 524 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.8965998468331165e-06, | |
| "loss": 1.6393, | |
| "step": 526 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.8956616204008698e-06, | |
| "loss": 1.6464, | |
| "step": 528 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.894719390781144e-06, | |
| "loss": 1.677, | |
| "step": 530 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.893773162185264e-06, | |
| "loss": 1.6843, | |
| "step": 532 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.892822938842428e-06, | |
| "loss": 1.63, | |
| "step": 534 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.8918687249996885e-06, | |
| "loss": 1.6523, | |
| "step": 536 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.8909105249219342e-06, | |
| "loss": 1.6744, | |
| "step": 538 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.8899483428918701e-06, | |
| "loss": 1.6678, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 1.8889821832099986e-06, | |
| "loss": 1.6631, | |
| "step": 542 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 1.8880120501946008e-06, | |
| "loss": 1.6328, | |
| "step": 544 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 1.887037948181716e-06, | |
| "loss": 1.6832, | |
| "step": 546 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 1.8860598815251239e-06, | |
| "loss": 1.7224, | |
| "step": 548 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 1.8850778545963234e-06, | |
| "loss": 1.7022, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 1.8840918717845142e-06, | |
| "loss": 1.6811, | |
| "step": 552 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 1.8831019374965779e-06, | |
| "loss": 1.6761, | |
| "step": 554 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 1.8821080561570562e-06, | |
| "loss": 1.6503, | |
| "step": 556 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 1.8811102322081327e-06, | |
| "loss": 1.744, | |
| "step": 558 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 1.8801084701096124e-06, | |
| "loss": 1.6645, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 1.8791027743389028e-06, | |
| "loss": 1.6282, | |
| "step": 562 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 1.8780931493909916e-06, | |
| "loss": 1.6378, | |
| "step": 564 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 1.8770795997784292e-06, | |
| "loss": 1.6921, | |
| "step": 566 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 1.8760621300313076e-06, | |
| "loss": 1.6503, | |
| "step": 568 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 1.8750407446972388e-06, | |
| "loss": 1.6626, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 1.8740154483413368e-06, | |
| "loss": 1.7116, | |
| "step": 572 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 1.8729862455461953e-06, | |
| "loss": 1.6613, | |
| "step": 574 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 1.8719531409118688e-06, | |
| "loss": 1.6898, | |
| "step": 576 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 1.8709161390558497e-06, | |
| "loss": 1.6868, | |
| "step": 578 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 1.8698752446130505e-06, | |
| "loss": 1.6382, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 1.8688304622357815e-06, | |
| "loss": 1.6634, | |
| "step": 582 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 1.8677817965937294e-06, | |
| "loss": 1.6494, | |
| "step": 584 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 1.8667292523739389e-06, | |
| "loss": 1.672, | |
| "step": 586 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 1.8656728342807884e-06, | |
| "loss": 1.6626, | |
| "step": 588 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 1.8646125470359714e-06, | |
| "loss": 1.6826, | |
| "step": 590 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 1.8635483953784752e-06, | |
| "loss": 1.6692, | |
| "step": 592 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 1.8624803840645585e-06, | |
| "loss": 1.6642, | |
| "step": 594 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 1.8614085178677308e-06, | |
| "loss": 1.6654, | |
| "step": 596 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 1.8603328015787317e-06, | |
| "loss": 1.684, | |
| "step": 598 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 1.859253240005509e-06, | |
| "loss": 1.6485, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 1.8581698379731962e-06, | |
| "loss": 1.6645, | |
| "step": 602 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 1.8570826003240928e-06, | |
| "loss": 1.6559, | |
| "step": 604 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 1.8559915319176405e-06, | |
| "loss": 1.6521, | |
| "step": 606 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 1.8548966376304045e-06, | |
| "loss": 1.656, | |
| "step": 608 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.853797922356048e-06, | |
| "loss": 1.6939, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.852695391005314e-06, | |
| "loss": 1.677, | |
| "step": 612 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.851589048506e-06, | |
| "loss": 1.6756, | |
| "step": 614 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.8504788998029386e-06, | |
| "loss": 1.7516, | |
| "step": 616 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.8493649498579735e-06, | |
| "loss": 1.6878, | |
| "step": 618 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.848247203649939e-06, | |
| "loss": 1.7089, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.8471256661746366e-06, | |
| "loss": 1.6735, | |
| "step": 622 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.8460003424448127e-06, | |
| "loss": 1.6585, | |
| "step": 624 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.844871237490137e-06, | |
| "loss": 1.6943, | |
| "step": 626 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.8437383563571788e-06, | |
| "loss": 1.6922, | |
| "step": 628 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.842601704109386e-06, | |
| "loss": 1.6556, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.8414612858270613e-06, | |
| "loss": 1.6795, | |
| "step": 632 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.8403171066073391e-06, | |
| "loss": 1.6719, | |
| "step": 634 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.8391691715641646e-06, | |
| "loss": 1.6462, | |
| "step": 636 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.8380174858282684e-06, | |
| "loss": 1.7138, | |
| "step": 638 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.8368620545471468e-06, | |
| "loss": 1.6404, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.8357028828850352e-06, | |
| "loss": 1.6931, | |
| "step": 642 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.8345399760228874e-06, | |
| "loss": 1.6342, | |
| "step": 644 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.8333733391583518e-06, | |
| "loss": 1.6657, | |
| "step": 646 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.8322029775057489e-06, | |
| "loss": 1.6783, | |
| "step": 648 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.8310288962960453e-06, | |
| "loss": 1.6747, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.8298511007768344e-06, | |
| "loss": 1.6914, | |
| "step": 652 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.8286695962123099e-06, | |
| "loss": 1.7069, | |
| "step": 654 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.8274843878832424e-06, | |
| "loss": 1.6734, | |
| "step": 656 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.826295481086958e-06, | |
| "loss": 1.6667, | |
| "step": 658 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.8251028811373126e-06, | |
| "loss": 1.7158, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.823906593364669e-06, | |
| "loss": 1.671, | |
| "step": 662 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.8227066231158723e-06, | |
| "loss": 1.7136, | |
| "step": 664 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.8215029757542269e-06, | |
| "loss": 1.6296, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.8202956566594718e-06, | |
| "loss": 1.7197, | |
| "step": 668 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.819084671227758e-06, | |
| "loss": 1.6423, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.8178700248716222e-06, | |
| "loss": 1.643, | |
| "step": 672 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.8166517230199635e-06, | |
| "loss": 1.6546, | |
| "step": 674 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.815429771118021e-06, | |
| "loss": 1.6814, | |
| "step": 676 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.8142041746273463e-06, | |
| "loss": 1.6313, | |
| "step": 678 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.8129749390257806e-06, | |
| "loss": 1.6718, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.8117420698074316e-06, | |
| "loss": 1.7085, | |
| "step": 682 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.8105055724826453e-06, | |
| "loss": 1.6729, | |
| "step": 684 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.8092654525779865e-06, | |
| "loss": 1.6683, | |
| "step": 686 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.8080217156362083e-06, | |
| "loss": 1.6816, | |
| "step": 688 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.8067743672162328e-06, | |
| "loss": 1.658, | |
| "step": 690 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.8055234128931216e-06, | |
| "loss": 1.6674, | |
| "step": 692 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.8042688582580547e-06, | |
| "loss": 1.6738, | |
| "step": 694 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.8030107089183027e-06, | |
| "loss": 1.6648, | |
| "step": 696 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.8017489704972033e-06, | |
| "loss": 1.6757, | |
| "step": 698 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.8004836486341358e-06, | |
| "loss": 1.6839, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.7992147489844954e-06, | |
| "loss": 1.6858, | |
| "step": 702 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.7979422772196684e-06, | |
| "loss": 1.7143, | |
| "step": 704 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.7966662390270076e-06, | |
| "loss": 1.6553, | |
| "step": 706 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.7953866401098047e-06, | |
| "loss": 1.7045, | |
| "step": 708 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.7941034861872674e-06, | |
| "loss": 1.6441, | |
| "step": 710 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.7928167829944915e-06, | |
| "loss": 1.6946, | |
| "step": 712 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.7915265362824377e-06, | |
| "loss": 1.6954, | |
| "step": 714 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.7902327518179032e-06, | |
| "loss": 1.6722, | |
| "step": 716 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.7889354353834982e-06, | |
| "loss": 1.6733, | |
| "step": 718 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.7876345927776188e-06, | |
| "loss": 1.7008, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.7863302298144216e-06, | |
| "loss": 1.6631, | |
| "step": 722 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.785022352323797e-06, | |
| "loss": 1.6355, | |
| "step": 724 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.7837109661513448e-06, | |
| "loss": 1.6214, | |
| "step": 726 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.782396077158346e-06, | |
| "loss": 1.7022, | |
| "step": 728 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.7810776912217378e-06, | |
| "loss": 1.7172, | |
| "step": 730 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.779755814234087e-06, | |
| "loss": 1.7013, | |
| "step": 732 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.7784304521035645e-06, | |
| "loss": 1.6754, | |
| "step": 734 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.7771016107539167e-06, | |
| "loss": 1.6523, | |
| "step": 736 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.775769296124442e-06, | |
| "loss": 1.6637, | |
| "step": 738 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.7744335141699626e-06, | |
| "loss": 1.6844, | |
| "step": 740 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.7730942708607964e-06, | |
| "loss": 1.6568, | |
| "step": 742 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.7717515721827334e-06, | |
| "loss": 1.7084, | |
| "step": 744 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.7704054241370072e-06, | |
| "loss": 1.6698, | |
| "step": 746 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.7690558327402685e-06, | |
| "loss": 1.7337, | |
| "step": 748 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.767702804024558e-06, | |
| "loss": 1.7196, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.7663463440372795e-06, | |
| "loss": 1.6206, | |
| "step": 752 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.7649864588411734e-06, | |
| "loss": 1.6425, | |
| "step": 754 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.7636231545142884e-06, | |
| "loss": 1.6834, | |
| "step": 756 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.762256437149956e-06, | |
| "loss": 1.6772, | |
| "step": 758 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.7608863128567614e-06, | |
| "loss": 1.6882, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.7595127877585188e-06, | |
| "loss": 1.7602, | |
| "step": 762 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.7581358679942406e-06, | |
| "loss": 1.663, | |
| "step": 764 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.7567555597181127e-06, | |
| "loss": 1.6718, | |
| "step": 766 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.7553718690994659e-06, | |
| "loss": 1.6516, | |
| "step": 768 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.7539848023227486e-06, | |
| "loss": 1.5969, | |
| "step": 770 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 1.7525943655874986e-06, | |
| "loss": 1.6172, | |
| "step": 772 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 1.7512005651083163e-06, | |
| "loss": 1.6425, | |
| "step": 774 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 1.7498034071148364e-06, | |
| "loss": 1.6208, | |
| "step": 776 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 1.7484028978517e-06, | |
| "loss": 1.6898, | |
| "step": 778 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 1.7469990435785269e-06, | |
| "loss": 1.646, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 1.7455918505698873e-06, | |
| "loss": 1.6232, | |
| "step": 782 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 1.7441813251152739e-06, | |
| "loss": 1.6009, | |
| "step": 784 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 1.7427674735190746e-06, | |
| "loss": 1.6495, | |
| "step": 786 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 1.741350302100543e-06, | |
| "loss": 1.6192, | |
| "step": 788 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 1.7399298171937703e-06, | |
| "loss": 1.6069, | |
| "step": 790 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 1.7385060251476587e-06, | |
| "loss": 1.6259, | |
| "step": 792 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 1.7370789323258905e-06, | |
| "loss": 1.6499, | |
| "step": 794 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 1.735648545106902e-06, | |
| "loss": 1.6528, | |
| "step": 796 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 1.734214869883853e-06, | |
| "loss": 1.6183, | |
| "step": 798 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 1.7327779130645994e-06, | |
| "loss": 1.587, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 1.7313376810716652e-06, | |
| "loss": 1.6316, | |
| "step": 802 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 1.7298941803422108e-06, | |
| "loss": 1.6709, | |
| "step": 804 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 1.7284474173280088e-06, | |
| "loss": 1.6394, | |
| "step": 806 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 1.7269973984954112e-06, | |
| "loss": 1.6625, | |
| "step": 808 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 1.7255441303253217e-06, | |
| "loss": 1.6603, | |
| "step": 810 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 1.7240876193131682e-06, | |
| "loss": 1.7061, | |
| "step": 812 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 1.7226278719688719e-06, | |
| "loss": 1.7066, | |
| "step": 814 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 1.7211648948168187e-06, | |
| "loss": 1.6273, | |
| "step": 816 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 1.719698694395831e-06, | |
| "loss": 1.6186, | |
| "step": 818 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 1.7182292772591367e-06, | |
| "loss": 1.6412, | |
| "step": 820 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 1.7167566499743416e-06, | |
| "loss": 1.6594, | |
| "step": 822 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 1.7152808191233993e-06, | |
| "loss": 1.6368, | |
| "step": 824 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 1.7138017913025814e-06, | |
| "loss": 1.6097, | |
| "step": 826 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 1.7123195731224488e-06, | |
| "loss": 1.6532, | |
| "step": 828 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 1.7108341712078223e-06, | |
| "loss": 1.614, | |
| "step": 830 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 1.7093455921977514e-06, | |
| "loss": 1.6505, | |
| "step": 832 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 1.7078538427454865e-06, | |
| "loss": 1.6498, | |
| "step": 834 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 1.7063589295184482e-06, | |
| "loss": 1.6582, | |
| "step": 836 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 1.7048608591981974e-06, | |
| "loss": 1.6596, | |
| "step": 838 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 1.7033596384804066e-06, | |
| "loss": 1.6303, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 1.701855274074828e-06, | |
| "loss": 1.6446, | |
| "step": 842 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 1.7003477727052654e-06, | |
| "loss": 1.6726, | |
| "step": 844 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 1.6988371411095428e-06, | |
| "loss": 1.6308, | |
| "step": 846 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 1.6973233860394753e-06, | |
| "loss": 1.6503, | |
| "step": 848 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 1.6958065142608381e-06, | |
| "loss": 1.6022, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 1.6942865325533373e-06, | |
| "loss": 1.6028, | |
| "step": 852 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 1.6927634477105778e-06, | |
| "loss": 1.6536, | |
| "step": 854 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 1.6912372665400353e-06, | |
| "loss": 1.6423, | |
| "step": 856 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 1.6897079958630234e-06, | |
| "loss": 1.6583, | |
| "step": 858 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 1.6881756425146654e-06, | |
| "loss": 1.6365, | |
| "step": 860 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 1.6866402133438623e-06, | |
| "loss": 1.6567, | |
| "step": 862 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 1.6851017152132619e-06, | |
| "loss": 1.624, | |
| "step": 864 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 1.6835601549992305e-06, | |
| "loss": 1.657, | |
| "step": 866 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 1.6820155395918186e-06, | |
| "loss": 1.6555, | |
| "step": 868 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 1.6804678758947333e-06, | |
| "loss": 1.6531, | |
| "step": 870 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 1.6789171708253052e-06, | |
| "loss": 1.6555, | |
| "step": 872 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 1.6773634313144593e-06, | |
| "loss": 1.7117, | |
| "step": 874 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 1.6758066643066826e-06, | |
| "loss": 1.6367, | |
| "step": 876 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 1.6742468767599933e-06, | |
| "loss": 1.6645, | |
| "step": 878 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 1.6726840756459107e-06, | |
| "loss": 1.6659, | |
| "step": 880 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 1.6711182679494232e-06, | |
| "loss": 1.6681, | |
| "step": 882 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 1.6695494606689567e-06, | |
| "loss": 1.6421, | |
| "step": 884 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 1.6679776608163441e-06, | |
| "loss": 1.6319, | |
| "step": 886 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 1.6664028754167938e-06, | |
| "loss": 1.6776, | |
| "step": 888 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 1.6648251115088583e-06, | |
| "loss": 1.6227, | |
| "step": 890 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 1.6632443761444024e-06, | |
| "loss": 1.6288, | |
| "step": 892 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 1.6616606763885717e-06, | |
| "loss": 1.6702, | |
| "step": 894 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 1.6600740193197622e-06, | |
| "loss": 1.6661, | |
| "step": 896 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 1.6584844120295865e-06, | |
| "loss": 1.5971, | |
| "step": 898 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 1.656891861622844e-06, | |
| "loss": 1.6587, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 1.655296375217488e-06, | |
| "loss": 1.6125, | |
| "step": 902 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 1.6536979599445953e-06, | |
| "loss": 1.6308, | |
| "step": 904 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 1.6520966229483322e-06, | |
| "loss": 1.622, | |
| "step": 906 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 1.650492371385924e-06, | |
| "loss": 1.673, | |
| "step": 908 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 1.6488852124276232e-06, | |
| "loss": 1.6179, | |
| "step": 910 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 1.6472751532566775e-06, | |
| "loss": 1.5717, | |
| "step": 912 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 1.6456622010692955e-06, | |
| "loss": 1.661, | |
| "step": 914 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 1.6440463630746172e-06, | |
| "loss": 1.6027, | |
| "step": 916 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 1.6424276464946813e-06, | |
| "loss": 1.6565, | |
| "step": 918 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 1.640806058564392e-06, | |
| "loss": 1.6329, | |
| "step": 920 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 1.6391816065314864e-06, | |
| "loss": 1.6458, | |
| "step": 922 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 1.6375542976565038e-06, | |
| "loss": 1.6731, | |
| "step": 924 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 1.6359241392127517e-06, | |
| "loss": 1.6625, | |
| "step": 926 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 1.634291138486274e-06, | |
| "loss": 1.7185, | |
| "step": 928 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 1.6326553027758183e-06, | |
| "loss": 1.6164, | |
| "step": 930 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 1.6310166393928033e-06, | |
| "loss": 1.6475, | |
| "step": 932 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 1.6293751556612858e-06, | |
| "loss": 1.6214, | |
| "step": 934 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 1.6277308589179286e-06, | |
| "loss": 1.7042, | |
| "step": 936 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 1.6260837565119668e-06, | |
| "loss": 1.6537, | |
| "step": 938 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 1.6244338558051761e-06, | |
| "loss": 1.6571, | |
| "step": 940 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 1.6227811641718393e-06, | |
| "loss": 1.6417, | |
| "step": 942 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 1.6211256889987129e-06, | |
| "loss": 1.6794, | |
| "step": 944 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 1.6194674376849941e-06, | |
| "loss": 1.6424, | |
| "step": 946 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 1.6178064176422895e-06, | |
| "loss": 1.6647, | |
| "step": 948 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 1.6161426362945796e-06, | |
| "loss": 1.6308, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 1.6144761010781866e-06, | |
| "loss": 1.6578, | |
| "step": 952 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 1.6128068194417424e-06, | |
| "loss": 1.6666, | |
| "step": 954 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 1.6111347988461522e-06, | |
| "loss": 1.6561, | |
| "step": 956 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 1.6094600467645643e-06, | |
| "loss": 1.6719, | |
| "step": 958 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 1.6077825706823356e-06, | |
| "loss": 1.6735, | |
| "step": 960 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 1.6061023780969975e-06, | |
| "loss": 1.6408, | |
| "step": 962 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 1.604419476518223e-06, | |
| "loss": 1.6715, | |
| "step": 964 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 1.6027338734677929e-06, | |
| "loss": 1.6253, | |
| "step": 966 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 1.6010455764795627e-06, | |
| "loss": 1.6585, | |
| "step": 968 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 1.5993545930994286e-06, | |
| "loss": 1.6267, | |
| "step": 970 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 1.5976609308852927e-06, | |
| "loss": 1.6103, | |
| "step": 972 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 1.5959645974070314e-06, | |
| "loss": 1.6536, | |
| "step": 974 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 1.5942656002464596e-06, | |
| "loss": 1.646, | |
| "step": 976 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 1.5925639469972986e-06, | |
| "loss": 1.6503, | |
| "step": 978 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 1.59085964526514e-06, | |
| "loss": 1.6461, | |
| "step": 980 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 1.5891527026674138e-06, | |
| "loss": 1.6556, | |
| "step": 982 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 1.587443126833353e-06, | |
| "loss": 1.6606, | |
| "step": 984 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 1.58573092540396e-06, | |
| "loss": 1.643, | |
| "step": 986 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 1.5840161060319717e-06, | |
| "loss": 1.6463, | |
| "step": 988 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 1.5822986763818271e-06, | |
| "loss": 1.6311, | |
| "step": 990 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 1.5805786441296318e-06, | |
| "loss": 1.6517, | |
| "step": 992 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 1.5788560169631234e-06, | |
| "loss": 1.6295, | |
| "step": 994 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 1.577130802581637e-06, | |
| "loss": 1.6462, | |
| "step": 996 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 1.5754030086960723e-06, | |
| "loss": 1.6615, | |
| "step": 998 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 1.5736726430288578e-06, | |
| "loss": 1.6252, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 1.571939713313917e-06, | |
| "loss": 1.6489, | |
| "step": 1002 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 1.5702042272966328e-06, | |
| "loss": 1.667, | |
| "step": 1004 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 1.5684661927338144e-06, | |
| "loss": 1.6357, | |
| "step": 1006 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 1.5667256173936605e-06, | |
| "loss": 1.6542, | |
| "step": 1008 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 1.5649825090557276e-06, | |
| "loss": 1.6409, | |
| "step": 1010 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 1.5632368755108925e-06, | |
| "loss": 1.6148, | |
| "step": 1012 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 1.5614887245613181e-06, | |
| "loss": 1.636, | |
| "step": 1014 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 1.55973806402042e-06, | |
| "loss": 1.6294, | |
| "step": 1016 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 1.5579849017128294e-06, | |
| "loss": 1.6545, | |
| "step": 1018 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 1.5562292454743606e-06, | |
| "loss": 1.6635, | |
| "step": 1020 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 1.5544711031519732e-06, | |
| "loss": 1.6447, | |
| "step": 1022 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 1.552710482603739e-06, | |
| "loss": 1.662, | |
| "step": 1024 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 1.5509473916988067e-06, | |
| "loss": 1.6113, | |
| "step": 1026 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 1.5491818383173656e-06, | |
| "loss": 1.6634, | |
| "step": 1028 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 1.5474138303506115e-06, | |
| "loss": 1.6267, | |
| "step": 1030 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 1.5456433757007112e-06, | |
| "loss": 1.6948, | |
| "step": 1032 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 1.5438704822807669e-06, | |
| "loss": 1.5969, | |
| "step": 1034 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 1.5420951580147805e-06, | |
| "loss": 1.6356, | |
| "step": 1036 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 1.5403174108376194e-06, | |
| "loss": 1.5955, | |
| "step": 1038 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 1.5385372486949795e-06, | |
| "loss": 1.6533, | |
| "step": 1040 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 1.5367546795433514e-06, | |
| "loss": 1.6796, | |
| "step": 1042 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 1.5349697113499833e-06, | |
| "loss": 1.6064, | |
| "step": 1044 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 1.533182352092846e-06, | |
| "loss": 1.5981, | |
| "step": 1046 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 1.5313926097605969e-06, | |
| "loss": 1.6574, | |
| "step": 1048 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 1.5296004923525457e-06, | |
| "loss": 1.6784, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 1.5278060078786164e-06, | |
| "loss": 1.6619, | |
| "step": 1052 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 1.5260091643593134e-06, | |
| "loss": 1.6148, | |
| "step": 1054 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 1.524209969825685e-06, | |
| "loss": 1.6591, | |
| "step": 1056 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 1.5224084323192864e-06, | |
| "loss": 1.5984, | |
| "step": 1058 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 1.520604559892146e-06, | |
| "loss": 1.6718, | |
| "step": 1060 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 1.5187983606067284e-06, | |
| "loss": 1.6287, | |
| "step": 1062 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 1.5169898425358962e-06, | |
| "loss": 1.6369, | |
| "step": 1064 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 1.5151790137628782e-06, | |
| "loss": 1.6217, | |
| "step": 1066 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 1.5133658823812289e-06, | |
| "loss": 1.6222, | |
| "step": 1068 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 1.5115504564947968e-06, | |
| "loss": 1.6274, | |
| "step": 1070 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 1.5097327442176836e-06, | |
| "loss": 1.6423, | |
| "step": 1072 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 1.5079127536742105e-06, | |
| "loss": 1.638, | |
| "step": 1074 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 1.5060904929988823e-06, | |
| "loss": 1.5695, | |
| "step": 1076 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 1.50426597033635e-06, | |
| "loss": 1.6434, | |
| "step": 1078 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 1.5024391938413741e-06, | |
| "loss": 1.6651, | |
| "step": 1080 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 1.5006101716787894e-06, | |
| "loss": 1.5832, | |
| "step": 1082 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 1.4987789120234663e-06, | |
| "loss": 1.6371, | |
| "step": 1084 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 1.496945423060278e-06, | |
| "loss": 1.6155, | |
| "step": 1086 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 1.49510971298406e-06, | |
| "loss": 1.6308, | |
| "step": 1088 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 1.4932717899995755e-06, | |
| "loss": 1.6545, | |
| "step": 1090 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 1.4914316623214788e-06, | |
| "loss": 1.6111, | |
| "step": 1092 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 1.4895893381742772e-06, | |
| "loss": 1.6386, | |
| "step": 1094 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 1.4877448257922964e-06, | |
| "loss": 1.6522, | |
| "step": 1096 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 1.4858981334196417e-06, | |
| "loss": 1.6319, | |
| "step": 1098 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 1.4840492693101619e-06, | |
| "loss": 1.6326, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 1.4821982417274127e-06, | |
| "loss": 1.6525, | |
| "step": 1102 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 1.4803450589446196e-06, | |
| "loss": 1.6357, | |
| "step": 1104 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 1.478489729244641e-06, | |
| "loss": 1.6227, | |
| "step": 1106 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 1.4766322609199302e-06, | |
| "loss": 1.5378, | |
| "step": 1108 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 1.4747726622724996e-06, | |
| "loss": 1.6475, | |
| "step": 1110 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 1.472910941613884e-06, | |
| "loss": 1.6425, | |
| "step": 1112 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 1.4710471072651013e-06, | |
| "loss": 1.6182, | |
| "step": 1114 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 1.4691811675566167e-06, | |
| "loss": 1.655, | |
| "step": 1116 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 1.4673131308283062e-06, | |
| "loss": 1.6137, | |
| "step": 1118 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 1.4654430054294182e-06, | |
| "loss": 1.6654, | |
| "step": 1120 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 1.4635707997185365e-06, | |
| "loss": 1.6104, | |
| "step": 1122 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 1.4616965220635425e-06, | |
| "loss": 1.6268, | |
| "step": 1124 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 1.459820180841578e-06, | |
| "loss": 1.6756, | |
| "step": 1126 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 1.4579417844390091e-06, | |
| "loss": 1.5819, | |
| "step": 1128 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 1.456061341251387e-06, | |
| "loss": 1.6191, | |
| "step": 1130 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 1.454178859683411e-06, | |
| "loss": 1.6041, | |
| "step": 1132 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 1.4522943481488905e-06, | |
| "loss": 1.7297, | |
| "step": 1134 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 1.450407815070709e-06, | |
| "loss": 1.5986, | |
| "step": 1136 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 1.4485192688807849e-06, | |
| "loss": 1.6505, | |
| "step": 1138 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 1.4466287180200334e-06, | |
| "loss": 1.6483, | |
| "step": 1140 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 1.444736170938331e-06, | |
| "loss": 1.6244, | |
| "step": 1142 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 1.4428416360944758e-06, | |
| "loss": 1.6216, | |
| "step": 1144 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 1.44094512195615e-06, | |
| "loss": 1.6686, | |
| "step": 1146 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 1.4390466369998825e-06, | |
| "loss": 1.6131, | |
| "step": 1148 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 1.437146189711011e-06, | |
| "loss": 1.6217, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 1.4352437885836439e-06, | |
| "loss": 1.5977, | |
| "step": 1152 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 1.433339442120622e-06, | |
| "loss": 1.6495, | |
| "step": 1154 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 1.4314331588334812e-06, | |
| "loss": 1.625, | |
| "step": 1156 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 1.4295249472424138e-06, | |
| "loss": 1.6697, | |
| "step": 1158 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 1.4276148158762312e-06, | |
| "loss": 1.6801, | |
| "step": 1160 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 1.4257027732723247e-06, | |
| "loss": 1.6695, | |
| "step": 1162 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 1.4237888279766282e-06, | |
| "loss": 1.6368, | |
| "step": 1164 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 1.4218729885435795e-06, | |
| "loss": 1.6403, | |
| "step": 1166 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 1.4199552635360831e-06, | |
| "loss": 1.6456, | |
| "step": 1168 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 1.418035661525471e-06, | |
| "loss": 1.6798, | |
| "step": 1170 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 1.4161141910914628e-06, | |
| "loss": 1.6831, | |
| "step": 1172 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 1.414190860822131e-06, | |
| "loss": 1.6382, | |
| "step": 1174 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 1.4122656793138604e-06, | |
| "loss": 1.6395, | |
| "step": 1176 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 1.4103386551713092e-06, | |
| "loss": 1.5895, | |
| "step": 1178 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 1.4084097970073724e-06, | |
| "loss": 1.6587, | |
| "step": 1180 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 1.4064791134431409e-06, | |
| "loss": 1.5971, | |
| "step": 1182 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 1.4045466131078648e-06, | |
| "loss": 1.5755, | |
| "step": 1184 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 1.4026123046389153e-06, | |
| "loss": 1.6285, | |
| "step": 1186 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 1.4006761966817436e-06, | |
| "loss": 1.6766, | |
| "step": 1188 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 1.398738297889845e-06, | |
| "loss": 1.6323, | |
| "step": 1190 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 1.396798616924718e-06, | |
| "loss": 1.6665, | |
| "step": 1192 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 1.394857162455827e-06, | |
| "loss": 1.6308, | |
| "step": 1194 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 1.3929139431605633e-06, | |
| "loss": 1.6035, | |
| "step": 1196 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 1.390968967724206e-06, | |
| "loss": 1.6314, | |
| "step": 1198 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 1.3890222448398835e-06, | |
| "loss": 1.6227, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 1.3870737832085341e-06, | |
| "loss": 1.626, | |
| "step": 1202 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 1.3851235915388678e-06, | |
| "loss": 1.6649, | |
| "step": 1204 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 1.3831716785473266e-06, | |
| "loss": 1.6319, | |
| "step": 1206 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 1.3812180529580468e-06, | |
| "loss": 1.7266, | |
| "step": 1208 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 1.3792627235028187e-06, | |
| "loss": 1.6233, | |
| "step": 1210 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 1.3773056989210478e-06, | |
| "loss": 1.6475, | |
| "step": 1212 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 1.3753469879597166e-06, | |
| "loss": 1.617, | |
| "step": 1214 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 1.3733865993733446e-06, | |
| "loss": 1.61, | |
| "step": 1216 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 1.3714245419239494e-06, | |
| "loss": 1.658, | |
| "step": 1218 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 1.3694608243810087e-06, | |
| "loss": 1.6522, | |
| "step": 1220 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 1.367495455521418e-06, | |
| "loss": 1.6231, | |
| "step": 1222 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 1.3655284441294552e-06, | |
| "loss": 1.5785, | |
| "step": 1224 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 1.3635597989967384e-06, | |
| "loss": 1.6403, | |
| "step": 1226 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 1.361589528922189e-06, | |
| "loss": 1.658, | |
| "step": 1228 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 1.3596176427119902e-06, | |
| "loss": 1.6734, | |
| "step": 1230 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 1.3576441491795482e-06, | |
| "loss": 1.6276, | |
| "step": 1232 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 1.3556690571454543e-06, | |
| "loss": 1.6124, | |
| "step": 1234 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 1.3536923754374436e-06, | |
| "loss": 1.6217, | |
| "step": 1236 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 1.3517141128903566e-06, | |
| "loss": 1.6518, | |
| "step": 1238 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 1.3497342783460996e-06, | |
| "loss": 1.662, | |
| "step": 1240 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 1.3477528806536039e-06, | |
| "loss": 1.6347, | |
| "step": 1242 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 1.3457699286687885e-06, | |
| "loss": 1.5949, | |
| "step": 1244 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 1.3437854312545191e-06, | |
| "loss": 1.6731, | |
| "step": 1246 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 1.3417993972805687e-06, | |
| "loss": 1.6209, | |
| "step": 1248 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 1.3398118356235782e-06, | |
| "loss": 1.6307, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 1.3378227551670154e-06, | |
| "loss": 1.6018, | |
| "step": 1252 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 1.335832164801138e-06, | |
| "loss": 1.6766, | |
| "step": 1254 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 1.3338400734229513e-06, | |
| "loss": 1.6397, | |
| "step": 1256 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 1.33184648993617e-06, | |
| "loss": 1.6589, | |
| "step": 1258 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 1.3298514232511769e-06, | |
| "loss": 1.6371, | |
| "step": 1260 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 1.3278548822849851e-06, | |
| "loss": 1.5834, | |
| "step": 1262 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 1.3258568759611963e-06, | |
| "loss": 1.6901, | |
| "step": 1264 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 1.3238574132099625e-06, | |
| "loss": 1.6112, | |
| "step": 1266 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 1.321856502967944e-06, | |
| "loss": 1.6388, | |
| "step": 1268 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 1.3198541541782718e-06, | |
| "loss": 1.6094, | |
| "step": 1270 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 1.3178503757905058e-06, | |
| "loss": 1.6612, | |
| "step": 1272 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 1.315845176760596e-06, | |
| "loss": 1.6111, | |
| "step": 1274 | |
| }, | |
| { | |
| "epoch": 1.67, | |
| "learning_rate": 1.3138385660508418e-06, | |
| "loss": 1.6964, | |
| "step": 1276 | |
| }, | |
| { | |
| "epoch": 1.67, | |
| "learning_rate": 1.3118305526298526e-06, | |
| "loss": 1.6161, | |
| "step": 1278 | |
| }, | |
| { | |
| "epoch": 1.67, | |
| "learning_rate": 1.3098211454725062e-06, | |
| "loss": 1.6516, | |
| "step": 1280 | |
| }, | |
| { | |
| "epoch": 1.67, | |
| "learning_rate": 1.3078103535599108e-06, | |
| "loss": 1.6194, | |
| "step": 1282 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 1.3057981858793633e-06, | |
| "loss": 1.6381, | |
| "step": 1284 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 1.3037846514243095e-06, | |
| "loss": 1.6109, | |
| "step": 1286 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 1.3017697591943054e-06, | |
| "loss": 1.6274, | |
| "step": 1288 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 1.299753518194973e-06, | |
| "loss": 1.635, | |
| "step": 1290 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 1.297735937437965e-06, | |
| "loss": 1.6735, | |
| "step": 1292 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 1.2957170259409214e-06, | |
| "loss": 1.6424, | |
| "step": 1294 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 1.2936967927274296e-06, | |
| "loss": 1.6002, | |
| "step": 1296 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 1.2916752468269854e-06, | |
| "loss": 1.636, | |
| "step": 1298 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 1.2896523972749502e-06, | |
| "loss": 1.5958, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 1.2876282531125138e-06, | |
| "loss": 1.5757, | |
| "step": 1302 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 1.2856028233866508e-06, | |
| "loss": 1.6604, | |
| "step": 1304 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 1.2835761171500832e-06, | |
| "loss": 1.6398, | |
| "step": 1306 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 1.281548143461237e-06, | |
| "loss": 1.6934, | |
| "step": 1308 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 1.279518911384204e-06, | |
| "loss": 1.6705, | |
| "step": 1310 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 1.2774884299887e-06, | |
| "loss": 1.5953, | |
| "step": 1312 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 1.2754567083500245e-06, | |
| "loss": 1.614, | |
| "step": 1314 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 1.2734237555490208e-06, | |
| "loss": 1.6368, | |
| "step": 1316 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 1.271389580672035e-06, | |
| "loss": 1.6018, | |
| "step": 1318 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 1.2693541928108747e-06, | |
| "loss": 1.6416, | |
| "step": 1320 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 1.2673176010627689e-06, | |
| "loss": 1.6012, | |
| "step": 1322 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 1.2652798145303284e-06, | |
| "loss": 1.6251, | |
| "step": 1324 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 1.2632408423215032e-06, | |
| "loss": 1.6255, | |
| "step": 1326 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 1.261200693549543e-06, | |
| "loss": 1.6101, | |
| "step": 1328 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 1.2591593773329567e-06, | |
| "loss": 1.6254, | |
| "step": 1330 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 1.25711690279547e-06, | |
| "loss": 1.599, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 1.2550732790659866e-06, | |
| "loss": 1.6217, | |
| "step": 1334 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 1.2530285152785465e-06, | |
| "loss": 1.6199, | |
| "step": 1336 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 1.2509826205722851e-06, | |
| "loss": 1.6429, | |
| "step": 1338 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 1.248935604091392e-06, | |
| "loss": 1.6596, | |
| "step": 1340 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 1.2468874749850713e-06, | |
| "loss": 1.6165, | |
| "step": 1342 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 1.2448382424074999e-06, | |
| "loss": 1.6128, | |
| "step": 1344 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 1.2427879155177865e-06, | |
| "loss": 1.6414, | |
| "step": 1346 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 1.240736503479931e-06, | |
| "loss": 1.6124, | |
| "step": 1348 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 1.238684015462783e-06, | |
| "loss": 1.5882, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 1.236630460640002e-06, | |
| "loss": 1.6906, | |
| "step": 1352 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 1.2345758481900148e-06, | |
| "loss": 1.5847, | |
| "step": 1354 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 1.2325201872959759e-06, | |
| "loss": 1.6506, | |
| "step": 1356 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 1.2304634871457254e-06, | |
| "loss": 1.6201, | |
| "step": 1358 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 1.228405756931749e-06, | |
| "loss": 1.6168, | |
| "step": 1360 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 1.2263470058511353e-06, | |
| "loss": 1.6581, | |
| "step": 1362 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 1.2242872431055372e-06, | |
| "loss": 1.6472, | |
| "step": 1364 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 1.2222264779011274e-06, | |
| "loss": 1.596, | |
| "step": 1366 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 1.2201647194485607e-06, | |
| "loss": 1.6063, | |
| "step": 1368 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 1.2181019769629304e-06, | |
| "loss": 1.6184, | |
| "step": 1370 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 1.2160382596637285e-06, | |
| "loss": 1.6643, | |
| "step": 1372 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 1.213973576774803e-06, | |
| "loss": 1.6594, | |
| "step": 1374 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 1.2119079375243192e-06, | |
| "loss": 1.5792, | |
| "step": 1376 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 1.2098413511447158e-06, | |
| "loss": 1.6062, | |
| "step": 1378 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 1.2077738268726644e-06, | |
| "loss": 1.6273, | |
| "step": 1380 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 1.2057053739490295e-06, | |
| "loss": 1.6365, | |
| "step": 1382 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 1.2036360016188254e-06, | |
| "loss": 1.6119, | |
| "step": 1384 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 1.2015657191311767e-06, | |
| "loss": 1.6355, | |
| "step": 1386 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 1.1994945357392752e-06, | |
| "loss": 1.6216, | |
| "step": 1388 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 1.1974224607003392e-06, | |
| "loss": 1.6438, | |
| "step": 1390 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 1.1953495032755726e-06, | |
| "loss": 1.6323, | |
| "step": 1392 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 1.1932756727301235e-06, | |
| "loss": 1.6625, | |
| "step": 1394 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 1.1912009783330416e-06, | |
| "loss": 1.6583, | |
| "step": 1396 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 1.1891254293572387e-06, | |
| "loss": 1.6348, | |
| "step": 1398 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 1.1870490350794442e-06, | |
| "loss": 1.6305, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 1.1849718047801685e-06, | |
| "loss": 1.636, | |
| "step": 1402 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 1.182893747743656e-06, | |
| "loss": 1.6382, | |
| "step": 1404 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 1.180814873257848e-06, | |
| "loss": 1.6281, | |
| "step": 1406 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 1.1787351906143382e-06, | |
| "loss": 1.692, | |
| "step": 1408 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 1.1766547091083328e-06, | |
| "loss": 1.6823, | |
| "step": 1410 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 1.174573438038609e-06, | |
| "loss": 1.6749, | |
| "step": 1412 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 1.1724913867074723e-06, | |
| "loss": 1.6037, | |
| "step": 1414 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 1.1704085644207162e-06, | |
| "loss": 1.5969, | |
| "step": 1416 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 1.1683249804875794e-06, | |
| "loss": 1.6382, | |
| "step": 1418 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 1.1662406442207052e-06, | |
| "loss": 1.6633, | |
| "step": 1420 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 1.1641555649360997e-06, | |
| "loss": 1.6059, | |
| "step": 1422 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 1.1620697519530892e-06, | |
| "loss": 1.6296, | |
| "step": 1424 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 1.15998321459428e-06, | |
| "loss": 1.6399, | |
| "step": 1426 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 1.1578959621855158e-06, | |
| "loss": 1.6216, | |
| "step": 1428 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 1.1558080040558358e-06, | |
| "loss": 1.5693, | |
| "step": 1430 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 1.153719349537434e-06, | |
| "loss": 1.6424, | |
| "step": 1432 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 1.1516300079656167e-06, | |
| "loss": 1.6442, | |
| "step": 1434 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 1.1495399886787618e-06, | |
| "loss": 1.6205, | |
| "step": 1436 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 1.1474493010182743e-06, | |
| "loss": 1.6661, | |
| "step": 1438 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 1.1453579543285489e-06, | |
| "loss": 1.5974, | |
| "step": 1440 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 1.1432659579569232e-06, | |
| "loss": 1.6562, | |
| "step": 1442 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 1.1411733212536416e-06, | |
| "loss": 1.6273, | |
| "step": 1444 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 1.1390800535718088e-06, | |
| "loss": 1.6656, | |
| "step": 1446 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 1.1369861642673489e-06, | |
| "loss": 1.5945, | |
| "step": 1448 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 1.1348916626989658e-06, | |
| "loss": 1.6511, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 1.1327965582280994e-06, | |
| "loss": 1.6789, | |
| "step": 1452 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 1.1307008602188843e-06, | |
| "loss": 1.6192, | |
| "step": 1454 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 1.1286045780381084e-06, | |
| "loss": 1.6453, | |
| "step": 1456 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 1.1265077210551693e-06, | |
| "loss": 1.6298, | |
| "step": 1458 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 1.124410298642035e-06, | |
| "loss": 1.6565, | |
| "step": 1460 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 1.1223123201732e-06, | |
| "loss": 1.6615, | |
| "step": 1462 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 1.120213795025645e-06, | |
| "loss": 1.6985, | |
| "step": 1464 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 1.1181147325787924e-06, | |
| "loss": 1.6486, | |
| "step": 1466 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 1.1160151422144682e-06, | |
| "loss": 1.6277, | |
| "step": 1468 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 1.113915033316856e-06, | |
| "loss": 1.6892, | |
| "step": 1470 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 1.1118144152724583e-06, | |
| "loss": 1.578, | |
| "step": 1472 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 1.1097132974700527e-06, | |
| "loss": 1.6344, | |
| "step": 1474 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 1.1076116893006505e-06, | |
| "loss": 1.6645, | |
| "step": 1476 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 1.1055096001574549e-06, | |
| "loss": 1.6057, | |
| "step": 1478 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 1.1034070394358187e-06, | |
| "loss": 1.569, | |
| "step": 1480 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 1.1013040165332023e-06, | |
| "loss": 1.6646, | |
| "step": 1482 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 1.0992005408491324e-06, | |
| "loss": 1.6215, | |
| "step": 1484 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 1.0970966217851586e-06, | |
| "loss": 1.6169, | |
| "step": 1486 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 1.094992268744813e-06, | |
| "loss": 1.6557, | |
| "step": 1488 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 1.092887491133567e-06, | |
| "loss": 1.654, | |
| "step": 1490 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 1.0907822983587887e-06, | |
| "loss": 1.6597, | |
| "step": 1492 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 1.0886766998297036e-06, | |
| "loss": 1.6098, | |
| "step": 1494 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 1.08657070495735e-06, | |
| "loss": 1.6635, | |
| "step": 1496 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 1.084464323154537e-06, | |
| "loss": 1.6338, | |
| "step": 1498 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 1.0823575638358033e-06, | |
| "loss": 1.6412, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 1.0802504364173762e-06, | |
| "loss": 1.5538, | |
| "step": 1502 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 1.0781429503171264e-06, | |
| "loss": 1.5697, | |
| "step": 1504 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 1.0760351149545295e-06, | |
| "loss": 1.6507, | |
| "step": 1506 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 1.0739269397506198e-06, | |
| "loss": 1.6406, | |
| "step": 1508 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 1.0718184341279532e-06, | |
| "loss": 1.5975, | |
| "step": 1510 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 1.0697096075105609e-06, | |
| "loss": 1.6288, | |
| "step": 1512 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 1.0676004693239089e-06, | |
| "loss": 1.6503, | |
| "step": 1514 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 1.0654910289948561e-06, | |
| "loss": 1.653, | |
| "step": 1516 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 1.0633812959516115e-06, | |
| "loss": 1.6504, | |
| "step": 1518 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 1.0612712796236926e-06, | |
| "loss": 1.6386, | |
| "step": 1520 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 1.0591609894418833e-06, | |
| "loss": 1.6573, | |
| "step": 1522 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 1.057050434838191e-06, | |
| "loss": 1.602, | |
| "step": 1524 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 1.054939625245805e-06, | |
| "loss": 1.6156, | |
| "step": 1526 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 1.052828570099055e-06, | |
| "loss": 1.6396, | |
| "step": 1528 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 1.0507172788333669e-06, | |
| "loss": 1.6265, | |
| "step": 1530 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 1.0486057608852235e-06, | |
| "loss": 1.5756, | |
| "step": 1532 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 1.0464940256921194e-06, | |
| "loss": 1.6054, | |
| "step": 1534 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 1.0443820826925207e-06, | |
| "loss": 1.6173, | |
| "step": 1536 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 1.0422699413258227e-06, | |
| "loss": 1.6201, | |
| "step": 1538 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 1.0401576110323068e-06, | |
| "loss": 1.583, | |
| "step": 1540 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 1.0380451012530988e-06, | |
| "loss": 1.6072, | |
| "step": 1542 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 1.0359324214301274e-06, | |
| "loss": 1.6032, | |
| "step": 1544 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 1.0338195810060804e-06, | |
| "loss": 1.6334, | |
| "step": 1546 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 1.0317065894243638e-06, | |
| "loss": 1.6142, | |
| "step": 1548 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 1.029593456129059e-06, | |
| "loss": 1.6001, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 1.0274801905648817e-06, | |
| "loss": 1.6233, | |
| "step": 1552 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 1.0253668021771377e-06, | |
| "loss": 1.6106, | |
| "step": 1554 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 1.0232533004116822e-06, | |
| "loss": 1.6218, | |
| "step": 1556 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 1.0211396947148766e-06, | |
| "loss": 1.6109, | |
| "step": 1558 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 1.0190259945335477e-06, | |
| "loss": 1.5625, | |
| "step": 1560 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 1.0169122093149447e-06, | |
| "loss": 1.5947, | |
| "step": 1562 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 1.0147983485066961e-06, | |
| "loss": 1.611, | |
| "step": 1564 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 1.012684421556768e-06, | |
| "loss": 1.6111, | |
| "step": 1566 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 1.0105704379134233e-06, | |
| "loss": 1.5724, | |
| "step": 1568 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 1.0084564070251775e-06, | |
| "loss": 1.5952, | |
| "step": 1570 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 1.0063423383407574e-06, | |
| "loss": 1.6118, | |
| "step": 1572 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 1.0042282413090585e-06, | |
| "loss": 1.5657, | |
| "step": 1574 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 1.002114125379104e-06, | |
| "loss": 1.626, | |
| "step": 1576 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 1e-06, | |
| "loss": 1.5821, | |
| "step": 1578 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 9.978858746208963e-07, | |
| "loss": 1.5955, | |
| "step": 1580 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 9.957717586909414e-07, | |
| "loss": 1.5962, | |
| "step": 1582 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 9.936576616592427e-07, | |
| "loss": 1.5815, | |
| "step": 1584 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 9.915435929748224e-07, | |
| "loss": 1.5756, | |
| "step": 1586 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 9.894295620865766e-07, | |
| "loss": 1.6007, | |
| "step": 1588 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 9.87315578443232e-07, | |
| "loss": 1.5676, | |
| "step": 1590 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 9.85201651493304e-07, | |
| "loss": 1.6159, | |
| "step": 1592 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 9.830877906850554e-07, | |
| "loss": 1.6068, | |
| "step": 1594 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 9.80974005466452e-07, | |
| "loss": 1.5853, | |
| "step": 1596 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 9.788603052851235e-07, | |
| "loss": 1.5878, | |
| "step": 1598 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 9.767466995883182e-07, | |
| "loss": 1.6147, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 9.746331978228622e-07, | |
| "loss": 1.5736, | |
| "step": 1602 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 9.725198094351183e-07, | |
| "loss": 1.6053, | |
| "step": 1604 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 9.704065438709408e-07, | |
| "loss": 1.6418, | |
| "step": 1606 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 9.68293410575636e-07, | |
| "loss": 1.5614, | |
| "step": 1608 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 9.6618041899392e-07, | |
| "loss": 1.5847, | |
| "step": 1610 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 9.640675785698725e-07, | |
| "loss": 1.5962, | |
| "step": 1612 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 9.619548987469013e-07, | |
| "loss": 1.5906, | |
| "step": 1614 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 9.598423889676931e-07, | |
| "loss": 1.6022, | |
| "step": 1616 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 9.577300586741774e-07, | |
| "loss": 1.6196, | |
| "step": 1618 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 9.556179173074796e-07, | |
| "loss": 1.6029, | |
| "step": 1620 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 9.535059743078808e-07, | |
| "loss": 1.6369, | |
| "step": 1622 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 9.513942391147766e-07, | |
| "loss": 1.5692, | |
| "step": 1624 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 9.492827211666327e-07, | |
| "loss": 1.6079, | |
| "step": 1626 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 9.471714299009451e-07, | |
| "loss": 1.6144, | |
| "step": 1628 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 9.450603747541951e-07, | |
| "loss": 1.5912, | |
| "step": 1630 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 9.42949565161809e-07, | |
| "loss": 1.6431, | |
| "step": 1632 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 9.408390105581167e-07, | |
| "loss": 1.5686, | |
| "step": 1634 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 9.38728720376307e-07, | |
| "loss": 1.6041, | |
| "step": 1636 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 9.366187040483885e-07, | |
| "loss": 1.5625, | |
| "step": 1638 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 9.34508971005144e-07, | |
| "loss": 1.6035, | |
| "step": 1640 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 9.323995306760907e-07, | |
| "loss": 1.5617, | |
| "step": 1642 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 9.30290392489439e-07, | |
| "loss": 1.569, | |
| "step": 1644 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 9.281815658720465e-07, | |
| "loss": 1.5936, | |
| "step": 1646 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 9.260730602493802e-07, | |
| "loss": 1.6216, | |
| "step": 1648 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 9.239648850454709e-07, | |
| "loss": 1.6273, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 9.218570496828733e-07, | |
| "loss": 1.6103, | |
| "step": 1652 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 9.197495635826238e-07, | |
| "loss": 1.6061, | |
| "step": 1654 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 9.176424361641965e-07, | |
| "loss": 1.5982, | |
| "step": 1656 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 9.155356768454631e-07, | |
| "loss": 1.6264, | |
| "step": 1658 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 9.134292950426502e-07, | |
| "loss": 1.6516, | |
| "step": 1660 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 9.113233001702963e-07, | |
| "loss": 1.6002, | |
| "step": 1662 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 9.092177016412115e-07, | |
| "loss": 1.5742, | |
| "step": 1664 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 9.071125088664333e-07, | |
| "loss": 1.6369, | |
| "step": 1666 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 9.050077312551869e-07, | |
| "loss": 1.6052, | |
| "step": 1668 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 9.029033782148415e-07, | |
| "loss": 1.625, | |
| "step": 1670 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 9.007994591508675e-07, | |
| "loss": 1.5875, | |
| "step": 1672 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 8.986959834667975e-07, | |
| "loss": 1.5897, | |
| "step": 1674 | |
| }, | |
| { | |
| "epoch": 2.19, | |
| "learning_rate": 8.965929605641811e-07, | |
| "loss": 1.5835, | |
| "step": 1676 | |
| }, | |
| { | |
| "epoch": 2.19, | |
| "learning_rate": 8.944903998425451e-07, | |
| "loss": 1.5926, | |
| "step": 1678 | |
| }, | |
| { | |
| "epoch": 2.19, | |
| "learning_rate": 8.923883106993498e-07, | |
| "loss": 1.572, | |
| "step": 1680 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 8.902867025299474e-07, | |
| "loss": 1.6465, | |
| "step": 1682 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 8.881855847275417e-07, | |
| "loss": 1.623, | |
| "step": 1684 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 8.860849666831438e-07, | |
| "loss": 1.6077, | |
| "step": 1686 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 8.839848577855318e-07, | |
| "loss": 1.5757, | |
| "step": 1688 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 8.818852674212074e-07, | |
| "loss": 1.618, | |
| "step": 1690 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 8.79786204974355e-07, | |
| "loss": 1.6239, | |
| "step": 1692 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 8.776876798267999e-07, | |
| "loss": 1.5725, | |
| "step": 1694 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 8.75589701357965e-07, | |
| "loss": 1.6088, | |
| "step": 1696 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 8.734922789448309e-07, | |
| "loss": 1.6064, | |
| "step": 1698 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 8.713954219618918e-07, | |
| "loss": 1.5796, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 8.692991397811155e-07, | |
| "loss": 1.6071, | |
| "step": 1702 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 8.672034417719007e-07, | |
| "loss": 1.6032, | |
| "step": 1704 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 8.651083373010344e-07, | |
| "loss": 1.6373, | |
| "step": 1706 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 8.63013835732651e-07, | |
| "loss": 1.5926, | |
| "step": 1708 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 8.609199464281916e-07, | |
| "loss": 1.5944, | |
| "step": 1710 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 8.58826678746358e-07, | |
| "loss": 1.6278, | |
| "step": 1712 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 8.567340420430766e-07, | |
| "loss": 1.5819, | |
| "step": 1714 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 8.546420456714516e-07, | |
| "loss": 1.5929, | |
| "step": 1716 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 8.525506989817256e-07, | |
| "loss": 1.6337, | |
| "step": 1718 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 8.504600113212386e-07, | |
| "loss": 1.6449, | |
| "step": 1720 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 8.48369992034383e-07, | |
| "loss": 1.5987, | |
| "step": 1722 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 8.462806504625659e-07, | |
| "loss": 1.6369, | |
| "step": 1724 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 8.441919959441644e-07, | |
| "loss": 1.6061, | |
| "step": 1726 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 8.421040378144843e-07, | |
| "loss": 1.646, | |
| "step": 1728 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 8.400167854057202e-07, | |
| "loss": 1.6509, | |
| "step": 1730 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 8.379302480469107e-07, | |
| "loss": 1.6275, | |
| "step": 1732 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 8.358444350639002e-07, | |
| "loss": 1.6423, | |
| "step": 1734 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 8.337593557792949e-07, | |
| "loss": 1.6447, | |
| "step": 1736 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 8.316750195124206e-07, | |
| "loss": 1.5823, | |
| "step": 1738 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 8.29591435579284e-07, | |
| "loss": 1.616, | |
| "step": 1740 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 8.275086132925275e-07, | |
| "loss": 1.646, | |
| "step": 1742 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 8.25426561961391e-07, | |
| "loss": 1.6423, | |
| "step": 1744 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 8.233452908916674e-07, | |
| "loss": 1.5761, | |
| "step": 1746 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 8.21264809385662e-07, | |
| "loss": 1.6237, | |
| "step": 1748 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 8.191851267421521e-07, | |
| "loss": 1.5361, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 8.171062522563436e-07, | |
| "loss": 1.6289, | |
| "step": 1752 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 8.150281952198315e-07, | |
| "loss": 1.6148, | |
| "step": 1754 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 8.129509649205557e-07, | |
| "loss": 1.5721, | |
| "step": 1756 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 8.108745706427615e-07, | |
| "loss": 1.5964, | |
| "step": 1758 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 8.087990216669584e-07, | |
| "loss": 1.6142, | |
| "step": 1760 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 8.067243272698765e-07, | |
| "loss": 1.6425, | |
| "step": 1762 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 8.046504967244275e-07, | |
| "loss": 1.5647, | |
| "step": 1764 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 8.02577539299661e-07, | |
| "loss": 1.6374, | |
| "step": 1766 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 8.005054642607247e-07, | |
| "loss": 1.6104, | |
| "step": 1768 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 7.984342808688234e-07, | |
| "loss": 1.597, | |
| "step": 1770 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 7.963639983811743e-07, | |
| "loss": 1.6116, | |
| "step": 1772 | |
| }, | |
| { | |
| "epoch": 2.32, | |
| "learning_rate": 7.942946260509704e-07, | |
| "loss": 1.6451, | |
| "step": 1774 | |
| }, | |
| { | |
| "epoch": 2.32, | |
| "learning_rate": 7.922261731273357e-07, | |
| "loss": 1.5678, | |
| "step": 1776 | |
| }, | |
| { | |
| "epoch": 2.32, | |
| "learning_rate": 7.901586488552843e-07, | |
| "loss": 1.6032, | |
| "step": 1778 | |
| }, | |
| { | |
| "epoch": 2.32, | |
| "learning_rate": 7.880920624756808e-07, | |
| "loss": 1.5844, | |
| "step": 1780 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 7.860264232251967e-07, | |
| "loss": 1.6054, | |
| "step": 1782 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 7.839617403362716e-07, | |
| "loss": 1.6125, | |
| "step": 1784 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 7.818980230370699e-07, | |
| "loss": 1.6237, | |
| "step": 1786 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 7.798352805514394e-07, | |
| "loss": 1.625, | |
| "step": 1788 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 7.777735220988728e-07, | |
| "loss": 1.641, | |
| "step": 1790 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 7.757127568944628e-07, | |
| "loss": 1.6097, | |
| "step": 1792 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 7.736529941488646e-07, | |
| "loss": 1.6077, | |
| "step": 1794 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 7.715942430682514e-07, | |
| "loss": 1.578, | |
| "step": 1796 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "learning_rate": 7.695365128542745e-07, | |
| "loss": 1.5961, | |
| "step": 1798 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "learning_rate": 7.674798127040241e-07, | |
| "loss": 1.6232, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "learning_rate": 7.654241518099851e-07, | |
| "loss": 1.6184, | |
| "step": 1802 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "learning_rate": 7.633695393599981e-07, | |
| "loss": 1.6431, | |
| "step": 1804 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 7.61315984537217e-07, | |
| "loss": 1.6183, | |
| "step": 1806 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 7.592634965200689e-07, | |
| "loss": 1.6466, | |
| "step": 1808 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 7.572120844822136e-07, | |
| "loss": 1.6053, | |
| "step": 1810 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 7.551617575925e-07, | |
| "loss": 1.6241, | |
| "step": 1812 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 7.531125250149288e-07, | |
| "loss": 1.5994, | |
| "step": 1814 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 7.510643959086082e-07, | |
| "loss": 1.6053, | |
| "step": 1816 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 7.49017379427715e-07, | |
| "loss": 1.6266, | |
| "step": 1818 | |
| }, | |
| { | |
| "epoch": 2.38, | |
| "learning_rate": 7.469714847214537e-07, | |
| "loss": 1.6137, | |
| "step": 1820 | |
| }, | |
| { | |
| "epoch": 2.38, | |
| "learning_rate": 7.449267209340133e-07, | |
| "loss": 1.6115, | |
| "step": 1822 | |
| }, | |
| { | |
| "epoch": 2.38, | |
| "learning_rate": 7.428830972045299e-07, | |
| "loss": 1.5982, | |
| "step": 1824 | |
| }, | |
| { | |
| "epoch": 2.38, | |
| "learning_rate": 7.408406226670435e-07, | |
| "loss": 1.6685, | |
| "step": 1826 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 7.387993064504567e-07, | |
| "loss": 1.6131, | |
| "step": 1828 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 7.36759157678497e-07, | |
| "loss": 1.594, | |
| "step": 1830 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 7.347201854696716e-07, | |
| "loss": 1.6526, | |
| "step": 1832 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 7.32682398937231e-07, | |
| "loss": 1.6181, | |
| "step": 1834 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 7.306458071891257e-07, | |
| "loss": 1.5931, | |
| "step": 1836 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 7.286104193279651e-07, | |
| "loss": 1.6306, | |
| "step": 1838 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 7.26576244450979e-07, | |
| "loss": 1.6434, | |
| "step": 1840 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 7.245432916499754e-07, | |
| "loss": 1.6254, | |
| "step": 1842 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 7.225115700113002e-07, | |
| "loss": 1.583, | |
| "step": 1844 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 7.204810886157963e-07, | |
| "loss": 1.5963, | |
| "step": 1846 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 7.18451856538763e-07, | |
| "loss": 1.6276, | |
| "step": 1848 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 7.164238828499167e-07, | |
| "loss": 1.5814, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "learning_rate": 7.143971766133489e-07, | |
| "loss": 1.6451, | |
| "step": 1852 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "learning_rate": 7.123717468874863e-07, | |
| "loss": 1.5674, | |
| "step": 1854 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "learning_rate": 7.103476027250499e-07, | |
| "loss": 1.6422, | |
| "step": 1856 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "learning_rate": 7.083247531730147e-07, | |
| "loss": 1.6395, | |
| "step": 1858 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 7.063032072725703e-07, | |
| "loss": 1.6274, | |
| "step": 1860 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 7.042829740590786e-07, | |
| "loss": 1.5986, | |
| "step": 1862 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 7.022640625620351e-07, | |
| "loss": 1.5921, | |
| "step": 1864 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 7.002464818050271e-07, | |
| "loss": 1.6184, | |
| "step": 1866 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 6.982302408056947e-07, | |
| "loss": 1.5867, | |
| "step": 1868 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 6.962153485756903e-07, | |
| "loss": 1.5978, | |
| "step": 1870 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 6.942018141206367e-07, | |
| "loss": 1.6097, | |
| "step": 1872 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "learning_rate": 6.921896464400891e-07, | |
| "loss": 1.6044, | |
| "step": 1874 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "learning_rate": 6.90178854527494e-07, | |
| "loss": 1.5581, | |
| "step": 1876 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "learning_rate": 6.881694473701476e-07, | |
| "loss": 1.5993, | |
| "step": 1878 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "learning_rate": 6.861614339491584e-07, | |
| "loss": 1.6154, | |
| "step": 1880 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 6.84154823239404e-07, | |
| "loss": 1.629, | |
| "step": 1882 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 6.821496242094942e-07, | |
| "loss": 1.6388, | |
| "step": 1884 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 6.801458458217285e-07, | |
| "loss": 1.6005, | |
| "step": 1886 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 6.781434970320561e-07, | |
| "loss": 1.5789, | |
| "step": 1888 | |
| }, | |
| { | |
| "epoch": 2.47, | |
| "learning_rate": 6.761425867900377e-07, | |
| "loss": 1.6105, | |
| "step": 1890 | |
| }, | |
| { | |
| "epoch": 2.47, | |
| "learning_rate": 6.741431240388034e-07, | |
| "loss": 1.6177, | |
| "step": 1892 | |
| }, | |
| { | |
| "epoch": 2.47, | |
| "learning_rate": 6.72145117715015e-07, | |
| "loss": 1.6137, | |
| "step": 1894 | |
| }, | |
| { | |
| "epoch": 2.47, | |
| "learning_rate": 6.701485767488234e-07, | |
| "loss": 1.5801, | |
| "step": 1896 | |
| }, | |
| { | |
| "epoch": 2.48, | |
| "learning_rate": 6.681535100638302e-07, | |
| "loss": 1.6135, | |
| "step": 1898 | |
| }, | |
| { | |
| "epoch": 2.48, | |
| "learning_rate": 6.661599265770488e-07, | |
| "loss": 1.6488, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 2.48, | |
| "learning_rate": 6.641678351988618e-07, | |
| "loss": 1.6106, | |
| "step": 1902 | |
| }, | |
| { | |
| "epoch": 2.48, | |
| "learning_rate": 6.621772448329845e-07, | |
| "loss": 1.6269, | |
| "step": 1904 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 6.60188164376422e-07, | |
| "loss": 1.5986, | |
| "step": 1906 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 6.582006027194309e-07, | |
| "loss": 1.6331, | |
| "step": 1908 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 6.562145687454808e-07, | |
| "loss": 1.6468, | |
| "step": 1910 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 6.542300713312112e-07, | |
| "loss": 1.5422, | |
| "step": 1912 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 6.522471193463964e-07, | |
| "loss": 1.5916, | |
| "step": 1914 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 6.502657216539006e-07, | |
| "loss": 1.6554, | |
| "step": 1916 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 6.482858871096432e-07, | |
| "loss": 1.6019, | |
| "step": 1918 | |
| }, | |
| { | |
| "epoch": 2.51, | |
| "learning_rate": 6.463076245625564e-07, | |
| "loss": 1.6108, | |
| "step": 1920 | |
| }, | |
| { | |
| "epoch": 2.51, | |
| "learning_rate": 6.443309428545456e-07, | |
| "loss": 1.626, | |
| "step": 1922 | |
| }, | |
| { | |
| "epoch": 2.51, | |
| "learning_rate": 6.423558508204516e-07, | |
| "loss": 1.6034, | |
| "step": 1924 | |
| }, | |
| { | |
| "epoch": 2.51, | |
| "learning_rate": 6.4038235728801e-07, | |
| "loss": 1.6722, | |
| "step": 1926 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "learning_rate": 6.384104710778109e-07, | |
| "loss": 1.5998, | |
| "step": 1928 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "learning_rate": 6.364402010032616e-07, | |
| "loss": 1.61, | |
| "step": 1930 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "learning_rate": 6.344715558705449e-07, | |
| "loss": 1.5932, | |
| "step": 1932 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "learning_rate": 6.32504544478582e-07, | |
| "loss": 1.6136, | |
| "step": 1934 | |
| }, | |
| { | |
| "epoch": 2.53, | |
| "learning_rate": 6.305391756189916e-07, | |
| "loss": 1.542, | |
| "step": 1936 | |
| }, | |
| { | |
| "epoch": 2.53, | |
| "learning_rate": 6.285754580760502e-07, | |
| "loss": 1.6062, | |
| "step": 1938 | |
| }, | |
| { | |
| "epoch": 2.53, | |
| "learning_rate": 6.266134006266554e-07, | |
| "loss": 1.5812, | |
| "step": 1940 | |
| }, | |
| { | |
| "epoch": 2.53, | |
| "learning_rate": 6.246530120402832e-07, | |
| "loss": 1.5957, | |
| "step": 1942 | |
| }, | |
| { | |
| "epoch": 2.54, | |
| "learning_rate": 6.226943010789522e-07, | |
| "loss": 1.5808, | |
| "step": 1944 | |
| }, | |
| { | |
| "epoch": 2.54, | |
| "learning_rate": 6.207372764971815e-07, | |
| "loss": 1.5974, | |
| "step": 1946 | |
| }, | |
| { | |
| "epoch": 2.54, | |
| "learning_rate": 6.187819470419531e-07, | |
| "loss": 1.5691, | |
| "step": 1948 | |
| }, | |
| { | |
| "epoch": 2.54, | |
| "learning_rate": 6.168283214526732e-07, | |
| "loss": 1.6269, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 6.148764084611325e-07, | |
| "loss": 1.6148, | |
| "step": 1952 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 6.129262167914659e-07, | |
| "loss": 1.6568, | |
| "step": 1954 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 6.109777551601165e-07, | |
| "loss": 1.6273, | |
| "step": 1956 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "learning_rate": 6.090310322757936e-07, | |
| "loss": 1.6499, | |
| "step": 1958 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "learning_rate": 6.070860568394367e-07, | |
| "loss": 1.5942, | |
| "step": 1960 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "learning_rate": 6.051428375441734e-07, | |
| "loss": 1.6227, | |
| "step": 1962 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "learning_rate": 6.032013830752822e-07, | |
| "loss": 1.6561, | |
| "step": 1964 | |
| }, | |
| { | |
| "epoch": 2.57, | |
| "learning_rate": 6.012617021101551e-07, | |
| "loss": 1.5966, | |
| "step": 1966 | |
| }, | |
| { | |
| "epoch": 2.57, | |
| "learning_rate": 5.993238033182562e-07, | |
| "loss": 1.6152, | |
| "step": 1968 | |
| }, | |
| { | |
| "epoch": 2.57, | |
| "learning_rate": 5.973876953610848e-07, | |
| "loss": 1.6279, | |
| "step": 1970 | |
| }, | |
| { | |
| "epoch": 2.57, | |
| "learning_rate": 5.954533868921352e-07, | |
| "loss": 1.5849, | |
| "step": 1972 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "learning_rate": 5.935208865568591e-07, | |
| "loss": 1.5829, | |
| "step": 1974 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "learning_rate": 5.915902029926279e-07, | |
| "loss": 1.5645, | |
| "step": 1976 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "learning_rate": 5.896613448286905e-07, | |
| "loss": 1.6053, | |
| "step": 1978 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "learning_rate": 5.877343206861396e-07, | |
| "loss": 1.5773, | |
| "step": 1980 | |
| }, | |
| { | |
| "epoch": 2.59, | |
| "learning_rate": 5.85809139177869e-07, | |
| "loss": 1.6491, | |
| "step": 1982 | |
| }, | |
| { | |
| "epoch": 2.59, | |
| "learning_rate": 5.838858089085372e-07, | |
| "loss": 1.5725, | |
| "step": 1984 | |
| }, | |
| { | |
| "epoch": 2.59, | |
| "learning_rate": 5.819643384745292e-07, | |
| "loss": 1.6302, | |
| "step": 1986 | |
| }, | |
| { | |
| "epoch": 2.59, | |
| "learning_rate": 5.800447364639166e-07, | |
| "loss": 1.6063, | |
| "step": 1988 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "learning_rate": 5.781270114564203e-07, | |
| "loss": 1.6031, | |
| "step": 1990 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "learning_rate": 5.76211172023372e-07, | |
| "loss": 1.6047, | |
| "step": 1992 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "learning_rate": 5.742972267276756e-07, | |
| "loss": 1.6326, | |
| "step": 1994 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "learning_rate": 5.72385184123769e-07, | |
| "loss": 1.5856, | |
| "step": 1996 | |
| }, | |
| { | |
| "epoch": 2.61, | |
| "learning_rate": 5.704750527575863e-07, | |
| "loss": 1.6225, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 2.61, | |
| "learning_rate": 5.68566841166519e-07, | |
| "loss": 1.5513, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 2.61, | |
| "learning_rate": 5.666605578793781e-07, | |
| "loss": 1.6497, | |
| "step": 2002 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "learning_rate": 5.657081420126779e-07, | |
| "loss": 1.6067, | |
| "step": 2004 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "learning_rate": 5.647562114163562e-07, | |
| "loss": 1.622, | |
| "step": 2006 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "learning_rate": 5.62853810288989e-07, | |
| "loss": 1.5964, | |
| "step": 2008 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "learning_rate": 5.609533630001176e-07, | |
| "loss": 1.5819, | |
| "step": 2010 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "learning_rate": 5.590548780438501e-07, | |
| "loss": 1.5751, | |
| "step": 2012 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "learning_rate": 5.571583639055243e-07, | |
| "loss": 1.6159, | |
| "step": 2014 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "learning_rate": 5.55263829061669e-07, | |
| "loss": 1.5722, | |
| "step": 2016 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "learning_rate": 5.533712819799666e-07, | |
| "loss": 1.5887, | |
| "step": 2018 | |
| }, | |
| { | |
| "epoch": 2.64, | |
| "learning_rate": 5.514807311192154e-07, | |
| "loss": 1.6163, | |
| "step": 2020 | |
| }, | |
| { | |
| "epoch": 2.64, | |
| "learning_rate": 5.495921849292911e-07, | |
| "loss": 1.5551, | |
| "step": 2022 | |
| }, | |
| { | |
| "epoch": 2.64, | |
| "learning_rate": 5.477056518511095e-07, | |
| "loss": 1.6046, | |
| "step": 2024 | |
| }, | |
| { | |
| "epoch": 2.64, | |
| "learning_rate": 5.458211403165892e-07, | |
| "loss": 1.5995, | |
| "step": 2026 | |
| }, | |
| { | |
| "epoch": 2.65, | |
| "learning_rate": 5.43938658748613e-07, | |
| "loss": 1.5944, | |
| "step": 2028 | |
| }, | |
| { | |
| "epoch": 2.65, | |
| "learning_rate": 5.420582155609906e-07, | |
| "loss": 1.574, | |
| "step": 2030 | |
| }, | |
| { | |
| "epoch": 2.65, | |
| "learning_rate": 5.401798191584221e-07, | |
| "loss": 1.6547, | |
| "step": 2032 | |
| }, | |
| { | |
| "epoch": 2.65, | |
| "learning_rate": 5.383034779364579e-07, | |
| "loss": 1.623, | |
| "step": 2034 | |
| }, | |
| { | |
| "epoch": 2.66, | |
| "learning_rate": 5.364292002814637e-07, | |
| "loss": 1.6148, | |
| "step": 2036 | |
| }, | |
| { | |
| "epoch": 2.66, | |
| "learning_rate": 5.345569945705817e-07, | |
| "loss": 1.5472, | |
| "step": 2038 | |
| }, | |
| { | |
| "epoch": 2.66, | |
| "learning_rate": 5.326868691716934e-07, | |
| "loss": 1.6321, | |
| "step": 2040 | |
| }, | |
| { | |
| "epoch": 2.66, | |
| "learning_rate": 5.308188324433834e-07, | |
| "loss": 1.5996, | |
| "step": 2042 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 5.289528927348991e-07, | |
| "loss": 1.6188, | |
| "step": 2044 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 5.270890583861159e-07, | |
| "loss": 1.5869, | |
| "step": 2046 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 5.252273377275003e-07, | |
| "loss": 1.6318, | |
| "step": 2048 | |
| }, | |
| { | |
| "epoch": 2.68, | |
| "learning_rate": 5.233677390800697e-07, | |
| "loss": 1.5828, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 2.68, | |
| "learning_rate": 5.215102707553594e-07, | |
| "loss": 1.6117, | |
| "step": 2052 | |
| }, | |
| { | |
| "epoch": 2.68, | |
| "learning_rate": 5.196549410553805e-07, | |
| "loss": 1.6299, | |
| "step": 2054 | |
| }, | |
| { | |
| "epoch": 2.68, | |
| "learning_rate": 5.178017582725871e-07, | |
| "loss": 1.6214, | |
| "step": 2056 | |
| }, | |
| { | |
| "epoch": 2.69, | |
| "learning_rate": 5.159507306898385e-07, | |
| "loss": 1.6021, | |
| "step": 2058 | |
| }, | |
| { | |
| "epoch": 2.69, | |
| "learning_rate": 5.141018665803583e-07, | |
| "loss": 1.5559, | |
| "step": 2060 | |
| }, | |
| { | |
| "epoch": 2.69, | |
| "learning_rate": 5.122551742077035e-07, | |
| "loss": 1.6264, | |
| "step": 2062 | |
| }, | |
| { | |
| "epoch": 2.69, | |
| "learning_rate": 5.104106618257229e-07, | |
| "loss": 1.5952, | |
| "step": 2064 | |
| }, | |
| { | |
| "epoch": 2.7, | |
| "learning_rate": 5.085683376785212e-07, | |
| "loss": 1.6336, | |
| "step": 2066 | |
| }, | |
| { | |
| "epoch": 2.7, | |
| "learning_rate": 5.067282100004247e-07, | |
| "loss": 1.6012, | |
| "step": 2068 | |
| }, | |
| { | |
| "epoch": 2.7, | |
| "learning_rate": 5.048902870159398e-07, | |
| "loss": 1.6388, | |
| "step": 2070 | |
| }, | |
| { | |
| "epoch": 2.7, | |
| "learning_rate": 5.030545769397217e-07, | |
| "loss": 1.5987, | |
| "step": 2072 | |
| }, | |
| { | |
| "epoch": 2.71, | |
| "learning_rate": 5.012210879765338e-07, | |
| "loss": 1.6562, | |
| "step": 2074 | |
| }, | |
| { | |
| "epoch": 2.71, | |
| "learning_rate": 4.993898283212108e-07, | |
| "loss": 1.5812, | |
| "step": 2076 | |
| }, | |
| { | |
| "epoch": 2.71, | |
| "learning_rate": 4.975608061586257e-07, | |
| "loss": 1.6145, | |
| "step": 2078 | |
| }, | |
| { | |
| "epoch": 2.71, | |
| "learning_rate": 4.9573402966365e-07, | |
| "loss": 1.6089, | |
| "step": 2080 | |
| }, | |
| { | |
| "epoch": 2.72, | |
| "learning_rate": 4.939095070011176e-07, | |
| "loss": 1.6022, | |
| "step": 2082 | |
| }, | |
| { | |
| "epoch": 2.72, | |
| "learning_rate": 4.920872463257899e-07, | |
| "loss": 1.6055, | |
| "step": 2084 | |
| }, | |
| { | |
| "epoch": 2.72, | |
| "learning_rate": 4.902672557823167e-07, | |
| "loss": 1.6268, | |
| "step": 2086 | |
| }, | |
| { | |
| "epoch": 2.72, | |
| "learning_rate": 4.884495435052033e-07, | |
| "loss": 1.5951, | |
| "step": 2088 | |
| }, | |
| { | |
| "epoch": 2.73, | |
| "learning_rate": 4.866341176187708e-07, | |
| "loss": 1.6209, | |
| "step": 2090 | |
| }, | |
| { | |
| "epoch": 2.73, | |
| "learning_rate": 4.84820986237122e-07, | |
| "loss": 1.5918, | |
| "step": 2092 | |
| }, | |
| { | |
| "epoch": 2.73, | |
| "learning_rate": 4.830101574641038e-07, | |
| "loss": 1.6082, | |
| "step": 2094 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "learning_rate": 4.812016393932717e-07, | |
| "loss": 1.6074, | |
| "step": 2096 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "learning_rate": 4.793954401078538e-07, | |
| "loss": 1.6342, | |
| "step": 2098 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "learning_rate": 4.775915676807135e-07, | |
| "loss": 1.6049, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "learning_rate": 4.757900301743152e-07, | |
| "loss": 1.5615, | |
| "step": 2102 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 4.7399083564068655e-07, | |
| "loss": 1.5625, | |
| "step": 2104 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 4.7219399212138366e-07, | |
| "loss": 1.5922, | |
| "step": 2106 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 4.703995076474544e-07, | |
| "loss": 1.6264, | |
| "step": 2108 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 4.686073902394031e-07, | |
| "loss": 1.5971, | |
| "step": 2110 | |
| }, | |
| { | |
| "epoch": 2.76, | |
| "learning_rate": 4.6681764790715416e-07, | |
| "loss": 1.5968, | |
| "step": 2112 | |
| }, | |
| { | |
| "epoch": 2.76, | |
| "learning_rate": 4.650302886500167e-07, | |
| "loss": 1.6098, | |
| "step": 2114 | |
| }, | |
| { | |
| "epoch": 2.76, | |
| "learning_rate": 4.632453204566485e-07, | |
| "loss": 1.606, | |
| "step": 2116 | |
| }, | |
| { | |
| "epoch": 2.76, | |
| "learning_rate": 4.6146275130502046e-07, | |
| "loss": 1.5781, | |
| "step": 2118 | |
| }, | |
| { | |
| "epoch": 2.77, | |
| "learning_rate": 4.596825891623808e-07, | |
| "loss": 1.5659, | |
| "step": 2120 | |
| }, | |
| { | |
| "epoch": 2.77, | |
| "learning_rate": 4.579048419852196e-07, | |
| "loss": 1.6135, | |
| "step": 2122 | |
| }, | |
| { | |
| "epoch": 2.77, | |
| "learning_rate": 4.5612951771923335e-07, | |
| "loss": 1.5918, | |
| "step": 2124 | |
| }, | |
| { | |
| "epoch": 2.77, | |
| "learning_rate": 4.543566242992889e-07, | |
| "loss": 1.6348, | |
| "step": 2126 | |
| }, | |
| { | |
| "epoch": 2.78, | |
| "learning_rate": 4.525861696493886e-07, | |
| "loss": 1.5651, | |
| "step": 2128 | |
| }, | |
| { | |
| "epoch": 2.78, | |
| "learning_rate": 4.508181616826342e-07, | |
| "loss": 1.5775, | |
| "step": 2130 | |
| }, | |
| { | |
| "epoch": 2.78, | |
| "learning_rate": 4.4905260830119343e-07, | |
| "loss": 1.5981, | |
| "step": 2132 | |
| }, | |
| { | |
| "epoch": 2.78, | |
| "learning_rate": 4.47289517396261e-07, | |
| "loss": 1.5891, | |
| "step": 2134 | |
| }, | |
| { | |
| "epoch": 2.79, | |
| "learning_rate": 4.4552889684802695e-07, | |
| "loss": 1.5942, | |
| "step": 2136 | |
| }, | |
| { | |
| "epoch": 2.79, | |
| "learning_rate": 4.437707545256395e-07, | |
| "loss": 1.5945, | |
| "step": 2138 | |
| }, | |
| { | |
| "epoch": 2.79, | |
| "learning_rate": 4.420150982871702e-07, | |
| "loss": 1.598, | |
| "step": 2140 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "learning_rate": 4.402619359795803e-07, | |
| "loss": 1.5905, | |
| "step": 2142 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "learning_rate": 4.3851127543868206e-07, | |
| "loss": 1.591, | |
| "step": 2144 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "learning_rate": 4.367631244891075e-07, | |
| "loss": 1.5987, | |
| "step": 2146 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "learning_rate": 4.350174909442724e-07, | |
| "loss": 1.5921, | |
| "step": 2148 | |
| }, | |
| { | |
| "epoch": 2.81, | |
| "learning_rate": 4.332743826063392e-07, | |
| "loss": 1.6046, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 2.81, | |
| "learning_rate": 4.315338072661859e-07, | |
| "loss": 1.6006, | |
| "step": 2152 | |
| }, | |
| { | |
| "epoch": 2.81, | |
| "learning_rate": 4.2979577270336717e-07, | |
| "loss": 1.6061, | |
| "step": 2154 | |
| }, | |
| { | |
| "epoch": 2.81, | |
| "learning_rate": 4.280602866860827e-07, | |
| "loss": 1.6212, | |
| "step": 2156 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "learning_rate": 4.263273569711422e-07, | |
| "loss": 1.6164, | |
| "step": 2158 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "learning_rate": 4.2459699130392757e-07, | |
| "loss": 1.6501, | |
| "step": 2160 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "learning_rate": 4.2286919741836305e-07, | |
| "loss": 1.5785, | |
| "step": 2162 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "learning_rate": 4.2114398303687703e-07, | |
| "loss": 1.6141, | |
| "step": 2164 | |
| }, | |
| { | |
| "epoch": 2.83, | |
| "learning_rate": 4.19421355870368e-07, | |
| "loss": 1.5799, | |
| "step": 2166 | |
| }, | |
| { | |
| "epoch": 2.83, | |
| "learning_rate": 4.1770132361817303e-07, | |
| "loss": 1.5517, | |
| "step": 2168 | |
| }, | |
| { | |
| "epoch": 2.83, | |
| "learning_rate": 4.1598389396802825e-07, | |
| "loss": 1.5986, | |
| "step": 2170 | |
| }, | |
| { | |
| "epoch": 2.83, | |
| "learning_rate": 4.142690745960401e-07, | |
| "loss": 1.6096, | |
| "step": 2172 | |
| }, | |
| { | |
| "epoch": 2.84, | |
| "learning_rate": 4.1255687316664724e-07, | |
| "loss": 1.6133, | |
| "step": 2174 | |
| }, | |
| { | |
| "epoch": 2.84, | |
| "learning_rate": 4.1084729733258584e-07, | |
| "loss": 1.6272, | |
| "step": 2176 | |
| }, | |
| { | |
| "epoch": 2.84, | |
| "learning_rate": 4.091403547348596e-07, | |
| "loss": 1.6366, | |
| "step": 2178 | |
| }, | |
| { | |
| "epoch": 2.85, | |
| "learning_rate": 4.074360530027012e-07, | |
| "loss": 1.6148, | |
| "step": 2180 | |
| }, | |
| { | |
| "epoch": 2.85, | |
| "learning_rate": 4.057343997535402e-07, | |
| "loss": 1.5835, | |
| "step": 2182 | |
| }, | |
| { | |
| "epoch": 2.85, | |
| "learning_rate": 4.0403540259296897e-07, | |
| "loss": 1.6169, | |
| "step": 2184 | |
| }, | |
| { | |
| "epoch": 2.85, | |
| "learning_rate": 4.0233906911470737e-07, | |
| "loss": 1.6529, | |
| "step": 2186 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "learning_rate": 4.0064540690057145e-07, | |
| "loss": 1.6738, | |
| "step": 2188 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "learning_rate": 3.9895442352043707e-07, | |
| "loss": 1.6212, | |
| "step": 2190 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "learning_rate": 3.9726612653220693e-07, | |
| "loss": 1.6021, | |
| "step": 2192 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "learning_rate": 3.955805234817769e-07, | |
| "loss": 1.6696, | |
| "step": 2194 | |
| }, | |
| { | |
| "epoch": 2.87, | |
| "learning_rate": 3.9389762190300245e-07, | |
| "loss": 1.617, | |
| "step": 2196 | |
| }, | |
| { | |
| "epoch": 2.87, | |
| "learning_rate": 3.9221742931766435e-07, | |
| "loss": 1.5683, | |
| "step": 2198 | |
| }, | |
| { | |
| "epoch": 2.87, | |
| "learning_rate": 3.90539953235436e-07, | |
| "loss": 1.6077, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 2.87, | |
| "learning_rate": 3.888652011538479e-07, | |
| "loss": 1.5631, | |
| "step": 2202 | |
| }, | |
| { | |
| "epoch": 2.88, | |
| "learning_rate": 3.8719318055825777e-07, | |
| "loss": 1.641, | |
| "step": 2204 | |
| }, | |
| { | |
| "epoch": 2.88, | |
| "learning_rate": 3.8552389892181323e-07, | |
| "loss": 1.6098, | |
| "step": 2206 | |
| }, | |
| { | |
| "epoch": 2.88, | |
| "learning_rate": 3.8385736370542054e-07, | |
| "loss": 1.6761, | |
| "step": 2208 | |
| }, | |
| { | |
| "epoch": 2.88, | |
| "learning_rate": 3.821935823577106e-07, | |
| "loss": 1.6118, | |
| "step": 2210 | |
| }, | |
| { | |
| "epoch": 2.89, | |
| "learning_rate": 3.8053256231500595e-07, | |
| "loss": 1.595, | |
| "step": 2212 | |
| }, | |
| { | |
| "epoch": 2.89, | |
| "learning_rate": 3.788743110012874e-07, | |
| "loss": 1.6574, | |
| "step": 2214 | |
| }, | |
| { | |
| "epoch": 2.89, | |
| "learning_rate": 3.7721883582816073e-07, | |
| "loss": 1.5992, | |
| "step": 2216 | |
| }, | |
| { | |
| "epoch": 2.89, | |
| "learning_rate": 3.755661441948238e-07, | |
| "loss": 1.5584, | |
| "step": 2218 | |
| }, | |
| { | |
| "epoch": 2.9, | |
| "learning_rate": 3.739162434880332e-07, | |
| "loss": 1.6191, | |
| "step": 2220 | |
| }, | |
| { | |
| "epoch": 2.9, | |
| "learning_rate": 3.722691410820715e-07, | |
| "loss": 1.6008, | |
| "step": 2222 | |
| }, | |
| { | |
| "epoch": 2.9, | |
| "learning_rate": 3.706248443387142e-07, | |
| "loss": 1.6182, | |
| "step": 2224 | |
| }, | |
| { | |
| "epoch": 2.91, | |
| "learning_rate": 3.689833606071967e-07, | |
| "loss": 1.5821, | |
| "step": 2226 | |
| }, | |
| { | |
| "epoch": 2.91, | |
| "learning_rate": 3.673446972241817e-07, | |
| "loss": 1.6215, | |
| "step": 2228 | |
| }, | |
| { | |
| "epoch": 2.91, | |
| "learning_rate": 3.65708861513726e-07, | |
| "loss": 1.616, | |
| "step": 2230 | |
| }, | |
| { | |
| "epoch": 2.91, | |
| "learning_rate": 3.640758607872484e-07, | |
| "loss": 1.5608, | |
| "step": 2232 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 3.6244570234349636e-07, | |
| "loss": 1.5848, | |
| "step": 2234 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 3.6081839346851374e-07, | |
| "loss": 1.5598, | |
| "step": 2236 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 3.591939414356081e-07, | |
| "loss": 1.6165, | |
| "step": 2238 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 3.575723535053186e-07, | |
| "loss": 1.6312, | |
| "step": 2240 | |
| }, | |
| { | |
| "epoch": 2.93, | |
| "learning_rate": 3.559536369253827e-07, | |
| "loss": 1.5831, | |
| "step": 2242 | |
| }, | |
| { | |
| "epoch": 2.93, | |
| "learning_rate": 3.543377989307047e-07, | |
| "loss": 1.6081, | |
| "step": 2244 | |
| }, | |
| { | |
| "epoch": 2.93, | |
| "learning_rate": 3.5272484674332223e-07, | |
| "loss": 1.5999, | |
| "step": 2246 | |
| }, | |
| { | |
| "epoch": 2.93, | |
| "learning_rate": 3.5111478757237656e-07, | |
| "loss": 1.5825, | |
| "step": 2248 | |
| }, | |
| { | |
| "epoch": 2.94, | |
| "learning_rate": 3.495076286140761e-07, | |
| "loss": 1.5792, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 2.94, | |
| "learning_rate": 3.4790337705166806e-07, | |
| "loss": 1.6666, | |
| "step": 2252 | |
| }, | |
| { | |
| "epoch": 2.94, | |
| "learning_rate": 3.4630204005540486e-07, | |
| "loss": 1.5955, | |
| "step": 2254 | |
| }, | |
| { | |
| "epoch": 2.94, | |
| "learning_rate": 3.4470362478251167e-07, | |
| "loss": 1.6012, | |
| "step": 2256 | |
| }, | |
| { | |
| "epoch": 2.95, | |
| "learning_rate": 3.431081383771562e-07, | |
| "loss": 1.6122, | |
| "step": 2258 | |
| }, | |
| { | |
| "epoch": 2.95, | |
| "learning_rate": 3.415155879704137e-07, | |
| "loss": 1.6524, | |
| "step": 2260 | |
| }, | |
| { | |
| "epoch": 2.95, | |
| "learning_rate": 3.399259806802376e-07, | |
| "loss": 1.6138, | |
| "step": 2262 | |
| }, | |
| { | |
| "epoch": 2.95, | |
| "learning_rate": 3.383393236114283e-07, | |
| "loss": 1.6209, | |
| "step": 2264 | |
| }, | |
| { | |
| "epoch": 2.96, | |
| "learning_rate": 3.367556238555974e-07, | |
| "loss": 1.5978, | |
| "step": 2266 | |
| }, | |
| { | |
| "epoch": 2.96, | |
| "learning_rate": 3.3517488849114184e-07, | |
| "loss": 1.6429, | |
| "step": 2268 | |
| }, | |
| { | |
| "epoch": 2.96, | |
| "learning_rate": 3.3359712458320634e-07, | |
| "loss": 1.6401, | |
| "step": 2270 | |
| }, | |
| { | |
| "epoch": 2.97, | |
| "learning_rate": 3.3202233918365585e-07, | |
| "loss": 1.6164, | |
| "step": 2272 | |
| }, | |
| { | |
| "epoch": 2.97, | |
| "learning_rate": 3.304505393310436e-07, | |
| "loss": 1.5974, | |
| "step": 2274 | |
| }, | |
| { | |
| "epoch": 2.97, | |
| "learning_rate": 3.2888173205057666e-07, | |
| "loss": 1.6189, | |
| "step": 2276 | |
| }, | |
| { | |
| "epoch": 2.97, | |
| "learning_rate": 3.273159243540891e-07, | |
| "loss": 1.6144, | |
| "step": 2278 | |
| }, | |
| { | |
| "epoch": 2.98, | |
| "learning_rate": 3.257531232400069e-07, | |
| "loss": 1.6307, | |
| "step": 2280 | |
| }, | |
| { | |
| "epoch": 2.98, | |
| "learning_rate": 3.241933356933174e-07, | |
| "loss": 1.5666, | |
| "step": 2282 | |
| }, | |
| { | |
| "epoch": 2.98, | |
| "learning_rate": 3.226365686855409e-07, | |
| "loss": 1.5699, | |
| "step": 2284 | |
| }, | |
| { | |
| "epoch": 2.98, | |
| "learning_rate": 3.210828291746946e-07, | |
| "loss": 1.6517, | |
| "step": 2286 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "learning_rate": 3.1953212410526675e-07, | |
| "loss": 1.593, | |
| "step": 2288 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "learning_rate": 3.179844604081817e-07, | |
| "loss": 1.609, | |
| "step": 2290 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "learning_rate": 3.164398450007695e-07, | |
| "loss": 1.6534, | |
| "step": 2292 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "learning_rate": 3.14898284786738e-07, | |
| "loss": 1.6303, | |
| "step": 2294 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 3.1335978665613783e-07, | |
| "loss": 1.6058, | |
| "step": 2296 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 3.1182435748533455e-07, | |
| "loss": 1.6229, | |
| "step": 2298 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 3.1029200413697685e-07, | |
| "loss": 1.5656, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 3.087627334599647e-07, | |
| "loss": 1.5789, | |
| "step": 2302 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 3.0723655228942203e-07, | |
| "loss": 1.534, | |
| "step": 2304 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 3.057134674466627e-07, | |
| "loss": 1.6436, | |
| "step": 2306 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 3.0419348573916173e-07, | |
| "loss": 1.5725, | |
| "step": 2308 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 3.0267661396052467e-07, | |
| "loss": 1.5541, | |
| "step": 2310 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 3.011628588904572e-07, | |
| "loss": 1.5688, | |
| "step": 2312 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 2.9965222729473474e-07, | |
| "loss": 1.5797, | |
| "step": 2314 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 2.9814472592517193e-07, | |
| "loss": 1.6164, | |
| "step": 2316 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 2.966403615195934e-07, | |
| "loss": 1.5926, | |
| "step": 2318 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 2.951391408018025e-07, | |
| "loss": 1.5881, | |
| "step": 2320 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 2.9364107048155183e-07, | |
| "loss": 1.565, | |
| "step": 2322 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 2.921461572545135e-07, | |
| "loss": 1.5748, | |
| "step": 2324 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 2.906544078022486e-07, | |
| "loss": 1.5936, | |
| "step": 2326 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 2.8916582879217776e-07, | |
| "loss": 1.5673, | |
| "step": 2328 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 2.876804268775511e-07, | |
| "loss": 1.6123, | |
| "step": 2330 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 2.8619820869741873e-07, | |
| "loss": 1.5789, | |
| "step": 2332 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 2.8471918087660087e-07, | |
| "loss": 1.6252, | |
| "step": 2334 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 2.8324335002565844e-07, | |
| "loss": 1.5652, | |
| "step": 2336 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 2.8177072274086344e-07, | |
| "loss": 1.5618, | |
| "step": 2338 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 2.803013056041691e-07, | |
| "loss": 1.58, | |
| "step": 2340 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "learning_rate": 2.788351051831813e-07, | |
| "loss": 1.5702, | |
| "step": 2342 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "learning_rate": 2.773721280311282e-07, | |
| "loss": 1.6083, | |
| "step": 2344 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "learning_rate": 2.7591238068683153e-07, | |
| "loss": 1.5572, | |
| "step": 2346 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "learning_rate": 2.744558696746784e-07, | |
| "loss": 1.6111, | |
| "step": 2348 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 2.730026015045891e-07, | |
| "loss": 1.5847, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 2.715525826719912e-07, | |
| "loss": 1.6328, | |
| "step": 2352 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 2.701058196577891e-07, | |
| "loss": 1.5917, | |
| "step": 2354 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 2.686623189283348e-07, | |
| "loss": 1.5531, | |
| "step": 2356 | |
| }, | |
| { | |
| "epoch": 3.08, | |
| "learning_rate": 2.672220869354005e-07, | |
| "loss": 1.5591, | |
| "step": 2358 | |
| }, | |
| { | |
| "epoch": 3.08, | |
| "learning_rate": 2.6578513011614713e-07, | |
| "loss": 1.6146, | |
| "step": 2360 | |
| }, | |
| { | |
| "epoch": 3.08, | |
| "learning_rate": 2.643514548930977e-07, | |
| "loss": 1.5839, | |
| "step": 2362 | |
| }, | |
| { | |
| "epoch": 3.09, | |
| "learning_rate": 2.629210676741095e-07, | |
| "loss": 1.5523, | |
| "step": 2364 | |
| }, | |
| { | |
| "epoch": 3.09, | |
| "learning_rate": 2.6149397485234115e-07, | |
| "loss": 1.6192, | |
| "step": 2366 | |
| }, | |
| { | |
| "epoch": 3.09, | |
| "learning_rate": 2.600701828062297e-07, | |
| "loss": 1.6184, | |
| "step": 2368 | |
| }, | |
| { | |
| "epoch": 3.09, | |
| "learning_rate": 2.5864969789945737e-07, | |
| "loss": 1.6061, | |
| "step": 2370 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "learning_rate": 2.572325264809252e-07, | |
| "loss": 1.5894, | |
| "step": 2372 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "learning_rate": 2.558186748847262e-07, | |
| "loss": 1.5659, | |
| "step": 2374 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "learning_rate": 2.5440814943011255e-07, | |
| "loss": 1.5714, | |
| "step": 2376 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "learning_rate": 2.530009564214729e-07, | |
| "loss": 1.5887, | |
| "step": 2378 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 2.515971021483e-07, | |
| "loss": 1.5637, | |
| "step": 2380 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 2.501965928851633e-07, | |
| "loss": 1.6213, | |
| "step": 2382 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 2.4879943489168367e-07, | |
| "loss": 1.6037, | |
| "step": 2384 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 2.474056344125012e-07, | |
| "loss": 1.6078, | |
| "step": 2386 | |
| }, | |
| { | |
| "epoch": 3.12, | |
| "learning_rate": 2.460151976772513e-07, | |
| "loss": 1.5884, | |
| "step": 2388 | |
| }, | |
| { | |
| "epoch": 3.12, | |
| "learning_rate": 2.446281309005341e-07, | |
| "loss": 1.6303, | |
| "step": 2390 | |
| }, | |
| { | |
| "epoch": 3.12, | |
| "learning_rate": 2.4324444028188707e-07, | |
| "loss": 1.5694, | |
| "step": 2392 | |
| }, | |
| { | |
| "epoch": 3.12, | |
| "learning_rate": 2.4186413200575917e-07, | |
| "loss": 1.617, | |
| "step": 2394 | |
| }, | |
| { | |
| "epoch": 3.13, | |
| "learning_rate": 2.4048721224148105e-07, | |
| "loss": 1.6069, | |
| "step": 2396 | |
| }, | |
| { | |
| "epoch": 3.13, | |
| "learning_rate": 2.391136871432383e-07, | |
| "loss": 1.5479, | |
| "step": 2398 | |
| }, | |
| { | |
| "epoch": 3.13, | |
| "learning_rate": 2.3774356285004437e-07, | |
| "loss": 1.6079, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 3.13, | |
| "learning_rate": 2.363768454857117e-07, | |
| "loss": 1.5811, | |
| "step": 2402 | |
| }, | |
| { | |
| "epoch": 3.14, | |
| "learning_rate": 2.3501354115882666e-07, | |
| "loss": 1.6253, | |
| "step": 2404 | |
| }, | |
| { | |
| "epoch": 3.14, | |
| "learning_rate": 2.3365365596272036e-07, | |
| "loss": 1.6034, | |
| "step": 2406 | |
| }, | |
| { | |
| "epoch": 3.14, | |
| "learning_rate": 2.3229719597544195e-07, | |
| "loss": 1.6277, | |
| "step": 2408 | |
| }, | |
| { | |
| "epoch": 3.15, | |
| "learning_rate": 2.3094416725973133e-07, | |
| "loss": 1.6208, | |
| "step": 2410 | |
| }, | |
| { | |
| "epoch": 3.15, | |
| "learning_rate": 2.295945758629927e-07, | |
| "loss": 1.6101, | |
| "step": 2412 | |
| }, | |
| { | |
| "epoch": 3.15, | |
| "learning_rate": 2.282484278172666e-07, | |
| "loss": 1.6151, | |
| "step": 2414 | |
| }, | |
| { | |
| "epoch": 3.15, | |
| "learning_rate": 2.269057291392037e-07, | |
| "loss": 1.5795, | |
| "step": 2416 | |
| }, | |
| { | |
| "epoch": 3.16, | |
| "learning_rate": 2.2556648583003745e-07, | |
| "loss": 1.6193, | |
| "step": 2418 | |
| }, | |
| { | |
| "epoch": 3.16, | |
| "learning_rate": 2.2423070387555765e-07, | |
| "loss": 1.5931, | |
| "step": 2420 | |
| }, | |
| { | |
| "epoch": 3.16, | |
| "learning_rate": 2.2289838924608305e-07, | |
| "loss": 1.5982, | |
| "step": 2422 | |
| }, | |
| { | |
| "epoch": 3.16, | |
| "learning_rate": 2.2156954789643568e-07, | |
| "loss": 1.6249, | |
| "step": 2424 | |
| }, | |
| { | |
| "epoch": 3.17, | |
| "learning_rate": 2.2024418576591298e-07, | |
| "loss": 1.592, | |
| "step": 2426 | |
| }, | |
| { | |
| "epoch": 3.17, | |
| "learning_rate": 2.1892230877826235e-07, | |
| "loss": 1.5805, | |
| "step": 2428 | |
| }, | |
| { | |
| "epoch": 3.17, | |
| "learning_rate": 2.1760392284165408e-07, | |
| "loss": 1.5838, | |
| "step": 2430 | |
| }, | |
| { | |
| "epoch": 3.17, | |
| "learning_rate": 2.1628903384865515e-07, | |
| "loss": 1.5925, | |
| "step": 2432 | |
| }, | |
| { | |
| "epoch": 3.18, | |
| "learning_rate": 2.1497764767620286e-07, | |
| "loss": 1.5871, | |
| "step": 2434 | |
| }, | |
| { | |
| "epoch": 3.18, | |
| "learning_rate": 2.1366977018557853e-07, | |
| "loss": 1.5532, | |
| "step": 2436 | |
| }, | |
| { | |
| "epoch": 3.18, | |
| "learning_rate": 2.1236540722238117e-07, | |
| "loss": 1.5955, | |
| "step": 2438 | |
| }, | |
| { | |
| "epoch": 3.18, | |
| "learning_rate": 2.1106456461650178e-07, | |
| "loss": 1.5791, | |
| "step": 2440 | |
| }, | |
| { | |
| "epoch": 3.19, | |
| "learning_rate": 2.0976724818209679e-07, | |
| "loss": 1.5949, | |
| "step": 2442 | |
| }, | |
| { | |
| "epoch": 3.19, | |
| "learning_rate": 2.0847346371756235e-07, | |
| "loss": 1.5967, | |
| "step": 2444 | |
| }, | |
| { | |
| "epoch": 3.19, | |
| "learning_rate": 2.0718321700550813e-07, | |
| "loss": 1.5934, | |
| "step": 2446 | |
| }, | |
| { | |
| "epoch": 3.19, | |
| "learning_rate": 2.0589651381273265e-07, | |
| "loss": 1.6194, | |
| "step": 2448 | |
| }, | |
| { | |
| "epoch": 3.2, | |
| "learning_rate": 2.0461335989019524e-07, | |
| "loss": 1.5485, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 3.2, | |
| "learning_rate": 2.0333376097299248e-07, | |
| "loss": 1.6498, | |
| "step": 2452 | |
| }, | |
| { | |
| "epoch": 3.2, | |
| "learning_rate": 2.020577227803315e-07, | |
| "loss": 1.5994, | |
| "step": 2454 | |
| }, | |
| { | |
| "epoch": 3.21, | |
| "learning_rate": 2.007852510155047e-07, | |
| "loss": 1.5968, | |
| "step": 2456 | |
| }, | |
| { | |
| "epoch": 3.21, | |
| "learning_rate": 1.9951635136586431e-07, | |
| "loss": 1.5915, | |
| "step": 2458 | |
| }, | |
| { | |
| "epoch": 3.21, | |
| "learning_rate": 1.982510295027967e-07, | |
| "loss": 1.5243, | |
| "step": 2460 | |
| }, | |
| { | |
| "epoch": 3.21, | |
| "learning_rate": 1.9698929108169716e-07, | |
| "loss": 1.6405, | |
| "step": 2462 | |
| }, | |
| { | |
| "epoch": 3.22, | |
| "learning_rate": 1.957311417419455e-07, | |
| "loss": 1.5397, | |
| "step": 2464 | |
| }, | |
| { | |
| "epoch": 3.22, | |
| "learning_rate": 1.9447658710687852e-07, | |
| "loss": 1.6, | |
| "step": 2466 | |
| }, | |
| { | |
| "epoch": 3.22, | |
| "learning_rate": 1.9322563278376748e-07, | |
| "loss": 1.5785, | |
| "step": 2468 | |
| }, | |
| { | |
| "epoch": 3.22, | |
| "learning_rate": 1.919782843637917e-07, | |
| "loss": 1.5701, | |
| "step": 2470 | |
| }, | |
| { | |
| "epoch": 3.23, | |
| "learning_rate": 1.9073454742201356e-07, | |
| "loss": 1.5759, | |
| "step": 2472 | |
| }, | |
| { | |
| "epoch": 3.23, | |
| "learning_rate": 1.8949442751735468e-07, | |
| "loss": 1.5871, | |
| "step": 2474 | |
| }, | |
| { | |
| "epoch": 3.23, | |
| "learning_rate": 1.8825793019256876e-07, | |
| "loss": 1.5798, | |
| "step": 2476 | |
| }, | |
| { | |
| "epoch": 3.23, | |
| "learning_rate": 1.870250609742191e-07, | |
| "loss": 1.5643, | |
| "step": 2478 | |
| }, | |
| { | |
| "epoch": 3.24, | |
| "learning_rate": 1.8579582537265392e-07, | |
| "loss": 1.567, | |
| "step": 2480 | |
| }, | |
| { | |
| "epoch": 3.24, | |
| "learning_rate": 1.8457022888197872e-07, | |
| "loss": 1.5608, | |
| "step": 2482 | |
| }, | |
| { | |
| "epoch": 3.24, | |
| "learning_rate": 1.8334827698003642e-07, | |
| "loss": 1.6361, | |
| "step": 2484 | |
| }, | |
| { | |
| "epoch": 3.24, | |
| "learning_rate": 1.8212997512837813e-07, | |
| "loss": 1.581, | |
| "step": 2486 | |
| }, | |
| { | |
| "epoch": 3.25, | |
| "learning_rate": 1.8091532877224192e-07, | |
| "loss": 1.5575, | |
| "step": 2488 | |
| }, | |
| { | |
| "epoch": 3.25, | |
| "learning_rate": 1.797043433405282e-07, | |
| "loss": 1.5808, | |
| "step": 2490 | |
| }, | |
| { | |
| "epoch": 3.25, | |
| "learning_rate": 1.7849702424577317e-07, | |
| "loss": 1.5217, | |
| "step": 2492 | |
| }, | |
| { | |
| "epoch": 3.25, | |
| "learning_rate": 1.772933768841277e-07, | |
| "loss": 1.5584, | |
| "step": 2494 | |
| }, | |
| { | |
| "epoch": 3.26, | |
| "learning_rate": 1.7609340663533112e-07, | |
| "loss": 1.5996, | |
| "step": 2496 | |
| }, | |
| { | |
| "epoch": 3.26, | |
| "learning_rate": 1.748971188626871e-07, | |
| "loss": 1.5791, | |
| "step": 2498 | |
| }, | |
| { | |
| "epoch": 3.26, | |
| "learning_rate": 1.7370451891304204e-07, | |
| "loss": 1.6198, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 3.27, | |
| "learning_rate": 1.725156121167576e-07, | |
| "loss": 1.6167, | |
| "step": 2502 | |
| }, | |
| { | |
| "epoch": 3.27, | |
| "learning_rate": 1.7133040378769037e-07, | |
| "loss": 1.559, | |
| "step": 2504 | |
| }, | |
| { | |
| "epoch": 3.27, | |
| "learning_rate": 1.7014889922316579e-07, | |
| "loss": 1.6096, | |
| "step": 2506 | |
| }, | |
| { | |
| "epoch": 3.27, | |
| "learning_rate": 1.689711037039545e-07, | |
| "loss": 1.5766, | |
| "step": 2508 | |
| }, | |
| { | |
| "epoch": 3.28, | |
| "learning_rate": 1.677970224942512e-07, | |
| "loss": 1.6213, | |
| "step": 2510 | |
| }, | |
| { | |
| "epoch": 3.28, | |
| "learning_rate": 1.666266608416479e-07, | |
| "loss": 1.5691, | |
| "step": 2512 | |
| }, | |
| { | |
| "epoch": 3.28, | |
| "learning_rate": 1.6546002397711246e-07, | |
| "loss": 1.5159, | |
| "step": 2514 | |
| }, | |
| { | |
| "epoch": 3.28, | |
| "learning_rate": 1.6429711711496496e-07, | |
| "loss": 1.5878, | |
| "step": 2516 | |
| }, | |
| { | |
| "epoch": 3.29, | |
| "learning_rate": 1.631379454528531e-07, | |
| "loss": 1.5404, | |
| "step": 2518 | |
| }, | |
| { | |
| "epoch": 3.29, | |
| "learning_rate": 1.6198251417173136e-07, | |
| "loss": 1.5686, | |
| "step": 2520 | |
| }, | |
| { | |
| "epoch": 3.29, | |
| "learning_rate": 1.608308284358355e-07, | |
| "loss": 1.5627, | |
| "step": 2522 | |
| }, | |
| { | |
| "epoch": 3.29, | |
| "learning_rate": 1.5968289339266083e-07, | |
| "loss": 1.5893, | |
| "step": 2524 | |
| }, | |
| { | |
| "epoch": 3.3, | |
| "learning_rate": 1.5853871417293873e-07, | |
| "loss": 1.6329, | |
| "step": 2526 | |
| }, | |
| { | |
| "epoch": 3.3, | |
| "learning_rate": 1.5739829589061383e-07, | |
| "loss": 1.6152, | |
| "step": 2528 | |
| }, | |
| { | |
| "epoch": 3.3, | |
| "learning_rate": 1.56261643642821e-07, | |
| "loss": 1.5956, | |
| "step": 2530 | |
| }, | |
| { | |
| "epoch": 3.3, | |
| "learning_rate": 1.5512876250986306e-07, | |
| "loss": 1.6358, | |
| "step": 2532 | |
| }, | |
| { | |
| "epoch": 3.31, | |
| "learning_rate": 1.539996575551872e-07, | |
| "loss": 1.6428, | |
| "step": 2534 | |
| }, | |
| { | |
| "epoch": 3.31, | |
| "learning_rate": 1.5287433382536342e-07, | |
| "loss": 1.5845, | |
| "step": 2536 | |
| }, | |
| { | |
| "epoch": 3.31, | |
| "learning_rate": 1.51752796350061e-07, | |
| "loss": 1.5707, | |
| "step": 2538 | |
| }, | |
| { | |
| "epoch": 3.31, | |
| "learning_rate": 1.5063505014202648e-07, | |
| "loss": 1.5925, | |
| "step": 2540 | |
| }, | |
| { | |
| "epoch": 3.32, | |
| "learning_rate": 1.4952110019706155e-07, | |
| "loss": 1.5684, | |
| "step": 2542 | |
| }, | |
| { | |
| "epoch": 3.32, | |
| "learning_rate": 1.4841095149399996e-07, | |
| "loss": 1.6393, | |
| "step": 2544 | |
| }, | |
| { | |
| "epoch": 3.32, | |
| "learning_rate": 1.47304608994686e-07, | |
| "loss": 1.5949, | |
| "step": 2546 | |
| }, | |
| { | |
| "epoch": 3.33, | |
| "learning_rate": 1.4620207764395176e-07, | |
| "loss": 1.5885, | |
| "step": 2548 | |
| }, | |
| { | |
| "epoch": 3.33, | |
| "learning_rate": 1.4510336236959554e-07, | |
| "loss": 1.5891, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 3.33, | |
| "learning_rate": 1.4400846808235945e-07, | |
| "loss": 1.5994, | |
| "step": 2552 | |
| }, | |
| { | |
| "epoch": 3.33, | |
| "learning_rate": 1.4291739967590744e-07, | |
| "loss": 1.5805, | |
| "step": 2554 | |
| }, | |
| { | |
| "epoch": 3.34, | |
| "learning_rate": 1.4183016202680375e-07, | |
| "loss": 1.5854, | |
| "step": 2556 | |
| }, | |
| { | |
| "epoch": 3.34, | |
| "learning_rate": 1.4074675999449094e-07, | |
| "loss": 1.5896, | |
| "step": 2558 | |
| }, | |
| { | |
| "epoch": 3.34, | |
| "learning_rate": 1.3966719842126806e-07, | |
| "loss": 1.6235, | |
| "step": 2560 | |
| }, | |
| { | |
| "epoch": 3.34, | |
| "learning_rate": 1.3859148213226901e-07, | |
| "loss": 1.5652, | |
| "step": 2562 | |
| }, | |
| { | |
| "epoch": 3.35, | |
| "learning_rate": 1.375196159354417e-07, | |
| "loss": 1.5949, | |
| "step": 2564 | |
| }, | |
| { | |
| "epoch": 3.35, | |
| "learning_rate": 1.3645160462152493e-07, | |
| "loss": 1.5713, | |
| "step": 2566 | |
| }, | |
| { | |
| "epoch": 3.35, | |
| "learning_rate": 1.3538745296402865e-07, | |
| "loss": 1.5817, | |
| "step": 2568 | |
| }, | |
| { | |
| "epoch": 3.35, | |
| "learning_rate": 1.3432716571921176e-07, | |
| "loss": 1.5355, | |
| "step": 2570 | |
| }, | |
| { | |
| "epoch": 3.36, | |
| "learning_rate": 1.3327074762606094e-07, | |
| "loss": 1.6249, | |
| "step": 2572 | |
| }, | |
| { | |
| "epoch": 3.36, | |
| "learning_rate": 1.3221820340627043e-07, | |
| "loss": 1.6183, | |
| "step": 2574 | |
| }, | |
| { | |
| "epoch": 3.36, | |
| "learning_rate": 1.3116953776421869e-07, | |
| "loss": 1.5645, | |
| "step": 2576 | |
| }, | |
| { | |
| "epoch": 3.36, | |
| "learning_rate": 1.3012475538694934e-07, | |
| "loss": 1.5974, | |
| "step": 2578 | |
| }, | |
| { | |
| "epoch": 3.37, | |
| "learning_rate": 1.2908386094415048e-07, | |
| "loss": 1.5769, | |
| "step": 2580 | |
| }, | |
| { | |
| "epoch": 3.37, | |
| "learning_rate": 1.2804685908813118e-07, | |
| "loss": 1.6138, | |
| "step": 2582 | |
| }, | |
| { | |
| "epoch": 3.37, | |
| "learning_rate": 1.2701375445380458e-07, | |
| "loss": 1.5989, | |
| "step": 2584 | |
| }, | |
| { | |
| "epoch": 3.37, | |
| "learning_rate": 1.2598455165866318e-07, | |
| "loss": 1.6144, | |
| "step": 2586 | |
| }, | |
| { | |
| "epoch": 3.38, | |
| "learning_rate": 1.2495925530276097e-07, | |
| "loss": 1.6266, | |
| "step": 2588 | |
| }, | |
| { | |
| "epoch": 3.38, | |
| "learning_rate": 1.239378699686926e-07, | |
| "loss": 1.589, | |
| "step": 2590 | |
| }, | |
| { | |
| "epoch": 3.38, | |
| "learning_rate": 1.2292040022157057e-07, | |
| "loss": 1.5466, | |
| "step": 2592 | |
| }, | |
| { | |
| "epoch": 3.39, | |
| "learning_rate": 1.219068506090084e-07, | |
| "loss": 1.6047, | |
| "step": 2594 | |
| }, | |
| { | |
| "epoch": 3.39, | |
| "learning_rate": 1.208972256610975e-07, | |
| "loss": 1.5564, | |
| "step": 2596 | |
| }, | |
| { | |
| "epoch": 3.39, | |
| "learning_rate": 1.1989152989038741e-07, | |
| "loss": 1.5896, | |
| "step": 2598 | |
| }, | |
| { | |
| "epoch": 3.39, | |
| "learning_rate": 1.1888976779186743e-07, | |
| "loss": 1.5955, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 3.4, | |
| "learning_rate": 1.1789194384294376e-07, | |
| "loss": 1.6147, | |
| "step": 2602 | |
| }, | |
| { | |
| "epoch": 3.4, | |
| "learning_rate": 1.1689806250342194e-07, | |
| "loss": 1.5934, | |
| "step": 2604 | |
| }, | |
| { | |
| "epoch": 3.4, | |
| "learning_rate": 1.1590812821548579e-07, | |
| "loss": 1.5618, | |
| "step": 2606 | |
| }, | |
| { | |
| "epoch": 3.4, | |
| "learning_rate": 1.1492214540367672e-07, | |
| "loss": 1.5586, | |
| "step": 2608 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "learning_rate": 1.1394011847487616e-07, | |
| "loss": 1.6067, | |
| "step": 2610 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "learning_rate": 1.1296205181828378e-07, | |
| "loss": 1.5777, | |
| "step": 2612 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "learning_rate": 1.1198794980539905e-07, | |
| "loss": 1.5578, | |
| "step": 2614 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "learning_rate": 1.1101781679000133e-07, | |
| "loss": 1.583, | |
| "step": 2616 | |
| }, | |
| { | |
| "epoch": 3.42, | |
| "learning_rate": 1.100516571081298e-07, | |
| "loss": 1.589, | |
| "step": 2618 | |
| }, | |
| { | |
| "epoch": 3.42, | |
| "learning_rate": 1.0908947507806565e-07, | |
| "loss": 1.5534, | |
| "step": 2620 | |
| }, | |
| { | |
| "epoch": 3.42, | |
| "learning_rate": 1.0813127500031139e-07, | |
| "loss": 1.5654, | |
| "step": 2622 | |
| }, | |
| { | |
| "epoch": 3.42, | |
| "learning_rate": 1.0717706115757207e-07, | |
| "loss": 1.653, | |
| "step": 2624 | |
| }, | |
| { | |
| "epoch": 3.43, | |
| "learning_rate": 1.0622683781473596e-07, | |
| "loss": 1.5528, | |
| "step": 2626 | |
| }, | |
| { | |
| "epoch": 3.43, | |
| "learning_rate": 1.0528060921885607e-07, | |
| "loss": 1.5619, | |
| "step": 2628 | |
| }, | |
| { | |
| "epoch": 3.43, | |
| "learning_rate": 1.0433837959913039e-07, | |
| "loss": 1.5973, | |
| "step": 2630 | |
| }, | |
| { | |
| "epoch": 3.43, | |
| "learning_rate": 1.0340015316688355e-07, | |
| "loss": 1.6352, | |
| "step": 2632 | |
| }, | |
| { | |
| "epoch": 3.44, | |
| "learning_rate": 1.0246593411554793e-07, | |
| "loss": 1.598, | |
| "step": 2634 | |
| }, | |
| { | |
| "epoch": 3.44, | |
| "learning_rate": 1.0153572662064447e-07, | |
| "loss": 1.5738, | |
| "step": 2636 | |
| }, | |
| { | |
| "epoch": 3.44, | |
| "learning_rate": 1.0060953483976453e-07, | |
| "loss": 1.5879, | |
| "step": 2638 | |
| }, | |
| { | |
| "epoch": 3.45, | |
| "learning_rate": 9.96873629125512e-08, | |
| "loss": 1.5915, | |
| "step": 2640 | |
| }, | |
| { | |
| "epoch": 3.45, | |
| "learning_rate": 9.876921496068059e-08, | |
| "loss": 1.6072, | |
| "step": 2642 | |
| }, | |
| { | |
| "epoch": 3.45, | |
| "learning_rate": 9.785509508784328e-08, | |
| "loss": 1.5988, | |
| "step": 2644 | |
| }, | |
| { | |
| "epoch": 3.45, | |
| "learning_rate": 9.694500737972688e-08, | |
| "loss": 1.5733, | |
| "step": 2646 | |
| }, | |
| { | |
| "epoch": 3.46, | |
| "learning_rate": 9.603895590399647e-08, | |
| "loss": 1.5686, | |
| "step": 2648 | |
| }, | |
| { | |
| "epoch": 3.46, | |
| "learning_rate": 9.513694471027733e-08, | |
| "loss": 1.5856, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 3.46, | |
| "learning_rate": 9.423897783013657e-08, | |
| "loss": 1.5943, | |
| "step": 2652 | |
| }, | |
| { | |
| "epoch": 3.46, | |
| "learning_rate": 9.334505927706515e-08, | |
| "loss": 1.6321, | |
| "step": 2654 | |
| }, | |
| { | |
| "epoch": 3.47, | |
| "learning_rate": 9.245519304645977e-08, | |
| "loss": 1.606, | |
| "step": 2656 | |
| }, | |
| { | |
| "epoch": 3.47, | |
| "learning_rate": 9.156938311560524e-08, | |
| "loss": 1.5451, | |
| "step": 2658 | |
| }, | |
| { | |
| "epoch": 3.47, | |
| "learning_rate": 9.068763344365682e-08, | |
| "loss": 1.5991, | |
| "step": 2660 | |
| }, | |
| { | |
| "epoch": 3.47, | |
| "learning_rate": 8.980994797162156e-08, | |
| "loss": 1.6528, | |
| "step": 2662 | |
| }, | |
| { | |
| "epoch": 3.48, | |
| "learning_rate": 8.893633062234285e-08, | |
| "loss": 1.5927, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 3.48, | |
| "learning_rate": 8.806678530047983e-08, | |
| "loss": 1.618, | |
| "step": 2666 | |
| }, | |
| { | |
| "epoch": 3.48, | |
| "learning_rate": 8.72013158924928e-08, | |
| "loss": 1.6322, | |
| "step": 2668 | |
| }, | |
| { | |
| "epoch": 3.48, | |
| "learning_rate": 8.6339926266624e-08, | |
| "loss": 1.5926, | |
| "step": 2670 | |
| }, | |
| { | |
| "epoch": 3.49, | |
| "learning_rate": 8.5482620272881e-08, | |
| "loss": 1.5543, | |
| "step": 2672 | |
| }, | |
| { | |
| "epoch": 3.49, | |
| "learning_rate": 8.462940174302025e-08, | |
| "loss": 1.594, | |
| "step": 2674 | |
| }, | |
| { | |
| "epoch": 3.49, | |
| "learning_rate": 8.378027449052782e-08, | |
| "loss": 1.5964, | |
| "step": 2676 | |
| }, | |
| { | |
| "epoch": 3.49, | |
| "learning_rate": 8.293524231060466e-08, | |
| "loss": 1.6331, | |
| "step": 2678 | |
| }, | |
| { | |
| "epoch": 3.5, | |
| "learning_rate": 8.209430898014869e-08, | |
| "loss": 1.5641, | |
| "step": 2680 | |
| }, | |
| { | |
| "epoch": 3.5, | |
| "learning_rate": 8.125747825773688e-08, | |
| "loss": 1.6072, | |
| "step": 2682 | |
| }, | |
| { | |
| "epoch": 3.5, | |
| "learning_rate": 8.042475388361103e-08, | |
| "loss": 1.54, | |
| "step": 2684 | |
| }, | |
| { | |
| "epoch": 3.51, | |
| "learning_rate": 7.959613957965794e-08, | |
| "loss": 1.6152, | |
| "step": 2686 | |
| }, | |
| { | |
| "epoch": 3.51, | |
| "learning_rate": 7.877163904939522e-08, | |
| "loss": 1.6052, | |
| "step": 2688 | |
| }, | |
| { | |
| "epoch": 3.51, | |
| "learning_rate": 7.795125597795405e-08, | |
| "loss": 1.6275, | |
| "step": 2690 | |
| }, | |
| { | |
| "epoch": 3.51, | |
| "learning_rate": 7.713499403206136e-08, | |
| "loss": 1.5732, | |
| "step": 2692 | |
| }, | |
| { | |
| "epoch": 3.52, | |
| "learning_rate": 7.632285686002592e-08, | |
| "loss": 1.5795, | |
| "step": 2694 | |
| }, | |
| { | |
| "epoch": 3.52, | |
| "learning_rate": 7.551484809172004e-08, | |
| "loss": 1.5856, | |
| "step": 2696 | |
| }, | |
| { | |
| "epoch": 3.52, | |
| "learning_rate": 7.471097133856352e-08, | |
| "loss": 1.5807, | |
| "step": 2698 | |
| }, | |
| { | |
| "epoch": 3.52, | |
| "learning_rate": 7.391123019350932e-08, | |
| "loss": 1.6053, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 3.53, | |
| "learning_rate": 7.311562823102468e-08, | |
| "loss": 1.5852, | |
| "step": 2702 | |
| }, | |
| { | |
| "epoch": 3.53, | |
| "learning_rate": 7.232416900707738e-08, | |
| "loss": 1.5762, | |
| "step": 2704 | |
| }, | |
| { | |
| "epoch": 3.53, | |
| "learning_rate": 7.153685605911964e-08, | |
| "loss": 1.5632, | |
| "step": 2706 | |
| }, | |
| { | |
| "epoch": 3.53, | |
| "learning_rate": 7.075369290607048e-08, | |
| "loss": 1.591, | |
| "step": 2708 | |
| }, | |
| { | |
| "epoch": 3.54, | |
| "learning_rate": 6.997468304830246e-08, | |
| "loss": 1.6077, | |
| "step": 2710 | |
| }, | |
| { | |
| "epoch": 3.54, | |
| "learning_rate": 6.91998299676243e-08, | |
| "loss": 1.6017, | |
| "step": 2712 | |
| }, | |
| { | |
| "epoch": 3.54, | |
| "learning_rate": 6.84291371272655e-08, | |
| "loss": 1.6394, | |
| "step": 2714 | |
| }, | |
| { | |
| "epoch": 3.54, | |
| "learning_rate": 6.766260797186241e-08, | |
| "loss": 1.5944, | |
| "step": 2716 | |
| }, | |
| { | |
| "epoch": 3.55, | |
| "learning_rate": 6.690024592744026e-08, | |
| "loss": 1.6005, | |
| "step": 2718 | |
| }, | |
| { | |
| "epoch": 3.55, | |
| "learning_rate": 6.614205440140041e-08, | |
| "loss": 1.5622, | |
| "step": 2720 | |
| }, | |
| { | |
| "epoch": 3.55, | |
| "learning_rate": 6.538803678250337e-08, | |
| "loss": 1.599, | |
| "step": 2722 | |
| }, | |
| { | |
| "epoch": 3.55, | |
| "learning_rate": 6.463819644085411e-08, | |
| "loss": 1.6174, | |
| "step": 2724 | |
| }, | |
| { | |
| "epoch": 3.56, | |
| "learning_rate": 6.389253672788752e-08, | |
| "loss": 1.5911, | |
| "step": 2726 | |
| }, | |
| { | |
| "epoch": 3.56, | |
| "learning_rate": 6.315106097635303e-08, | |
| "loss": 1.5908, | |
| "step": 2728 | |
| }, | |
| { | |
| "epoch": 3.56, | |
| "learning_rate": 6.241377250029933e-08, | |
| "loss": 1.5892, | |
| "step": 2730 | |
| }, | |
| { | |
| "epoch": 3.57, | |
| "learning_rate": 6.168067459506066e-08, | |
| "loss": 1.6196, | |
| "step": 2732 | |
| }, | |
| { | |
| "epoch": 3.57, | |
| "learning_rate": 6.09517705372401e-08, | |
| "loss": 1.5886, | |
| "step": 2734 | |
| }, | |
| { | |
| "epoch": 3.57, | |
| "learning_rate": 6.022706358469776e-08, | |
| "loss": 1.6137, | |
| "step": 2736 | |
| }, | |
| { | |
| "epoch": 3.57, | |
| "learning_rate": 5.950655697653362e-08, | |
| "loss": 1.5554, | |
| "step": 2738 | |
| }, | |
| { | |
| "epoch": 3.58, | |
| "learning_rate": 5.879025393307435e-08, | |
| "loss": 1.6075, | |
| "step": 2740 | |
| }, | |
| { | |
| "epoch": 3.58, | |
| "learning_rate": 5.8078157655858775e-08, | |
| "loss": 1.6098, | |
| "step": 2742 | |
| }, | |
| { | |
| "epoch": 3.58, | |
| "learning_rate": 5.73702713276234e-08, | |
| "loss": 1.5819, | |
| "step": 2744 | |
| }, | |
| { | |
| "epoch": 3.58, | |
| "learning_rate": 5.666659811228802e-08, | |
| "loss": 1.6254, | |
| "step": 2746 | |
| }, | |
| { | |
| "epoch": 3.59, | |
| "learning_rate": 5.596714115494217e-08, | |
| "loss": 1.6061, | |
| "step": 2748 | |
| }, | |
| { | |
| "epoch": 3.59, | |
| "learning_rate": 5.527190358183031e-08, | |
| "loss": 1.6087, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 3.59, | |
| "learning_rate": 5.4580888500338486e-08, | |
| "loss": 1.5698, | |
| "step": 2752 | |
| }, | |
| { | |
| "epoch": 3.59, | |
| "learning_rate": 5.389409899898012e-08, | |
| "loss": 1.6054, | |
| "step": 2754 | |
| }, | |
| { | |
| "epoch": 3.6, | |
| "learning_rate": 5.3211538147382216e-08, | |
| "loss": 1.5646, | |
| "step": 2756 | |
| }, | |
| { | |
| "epoch": 3.6, | |
| "learning_rate": 5.253320899627178e-08, | |
| "loss": 1.6152, | |
| "step": 2758 | |
| }, | |
| { | |
| "epoch": 3.6, | |
| "learning_rate": 5.185911457746206e-08, | |
| "loss": 1.6431, | |
| "step": 2760 | |
| }, | |
| { | |
| "epoch": 3.6, | |
| "learning_rate": 5.118925790383899e-08, | |
| "loss": 1.5424, | |
| "step": 2762 | |
| }, | |
| { | |
| "epoch": 3.61, | |
| "learning_rate": 5.052364196934777e-08, | |
| "loss": 1.6433, | |
| "step": 2764 | |
| }, | |
| { | |
| "epoch": 3.61, | |
| "learning_rate": 4.986226974897967e-08, | |
| "loss": 1.5994, | |
| "step": 2766 | |
| }, | |
| { | |
| "epoch": 3.61, | |
| "learning_rate": 4.920514419875821e-08, | |
| "loss": 1.5944, | |
| "step": 2768 | |
| }, | |
| { | |
| "epoch": 3.62, | |
| "learning_rate": 4.855226825572667e-08, | |
| "loss": 1.5903, | |
| "step": 2770 | |
| }, | |
| { | |
| "epoch": 3.62, | |
| "learning_rate": 4.79036448379343e-08, | |
| "loss": 1.5756, | |
| "step": 2772 | |
| }, | |
| { | |
| "epoch": 3.62, | |
| "learning_rate": 4.725927684442366e-08, | |
| "loss": 1.6207, | |
| "step": 2774 | |
| }, | |
| { | |
| "epoch": 3.62, | |
| "learning_rate": 4.661916715521763e-08, | |
| "loss": 1.6053, | |
| "step": 2776 | |
| }, | |
| { | |
| "epoch": 3.63, | |
| "learning_rate": 4.5983318631306114e-08, | |
| "loss": 1.5439, | |
| "step": 2778 | |
| }, | |
| { | |
| "epoch": 3.63, | |
| "learning_rate": 4.535173411463422e-08, | |
| "loss": 1.5701, | |
| "step": 2780 | |
| }, | |
| { | |
| "epoch": 3.63, | |
| "learning_rate": 4.472441642808844e-08, | |
| "loss": 1.5878, | |
| "step": 2782 | |
| }, | |
| { | |
| "epoch": 3.63, | |
| "learning_rate": 4.4101368375484614e-08, | |
| "loss": 1.61, | |
| "step": 2784 | |
| }, | |
| { | |
| "epoch": 3.64, | |
| "learning_rate": 4.348259274155541e-08, | |
| "loss": 1.5514, | |
| "step": 2786 | |
| }, | |
| { | |
| "epoch": 3.64, | |
| "learning_rate": 4.2868092291937773e-08, | |
| "loss": 1.621, | |
| "step": 2788 | |
| }, | |
| { | |
| "epoch": 3.64, | |
| "learning_rate": 4.225786977316092e-08, | |
| "loss": 1.6143, | |
| "step": 2790 | |
| }, | |
| { | |
| "epoch": 3.64, | |
| "learning_rate": 4.1651927912632944e-08, | |
| "loss": 1.5332, | |
| "step": 2792 | |
| }, | |
| { | |
| "epoch": 3.65, | |
| "learning_rate": 4.105026941862988e-08, | |
| "loss": 1.5658, | |
| "step": 2794 | |
| }, | |
| { | |
| "epoch": 3.65, | |
| "learning_rate": 4.0452896980283424e-08, | |
| "loss": 1.5495, | |
| "step": 2796 | |
| }, | |
| { | |
| "epoch": 3.65, | |
| "learning_rate": 3.985981326756793e-08, | |
| "loss": 1.6149, | |
| "step": 2798 | |
| }, | |
| { | |
| "epoch": 3.65, | |
| "learning_rate": 3.927102093128976e-08, | |
| "loss": 1.6166, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 3.66, | |
| "learning_rate": 3.868652260307437e-08, | |
| "loss": 1.6043, | |
| "step": 2802 | |
| }, | |
| { | |
| "epoch": 3.66, | |
| "learning_rate": 3.810632089535526e-08, | |
| "loss": 1.6229, | |
| "step": 2804 | |
| }, | |
| { | |
| "epoch": 3.66, | |
| "learning_rate": 3.7530418401362175e-08, | |
| "loss": 1.5762, | |
| "step": 2806 | |
| }, | |
| { | |
| "epoch": 3.66, | |
| "learning_rate": 3.6958817695109004e-08, | |
| "loss": 1.6508, | |
| "step": 2808 | |
| }, | |
| { | |
| "epoch": 3.67, | |
| "learning_rate": 3.639152133138312e-08, | |
| "loss": 1.5482, | |
| "step": 2810 | |
| }, | |
| { | |
| "epoch": 3.67, | |
| "learning_rate": 3.5828531845733204e-08, | |
| "loss": 1.5399, | |
| "step": 2812 | |
| }, | |
| { | |
| "epoch": 3.67, | |
| "learning_rate": 3.5269851754457957e-08, | |
| "loss": 1.6152, | |
| "step": 2814 | |
| }, | |
| { | |
| "epoch": 3.68, | |
| "learning_rate": 3.471548355459597e-08, | |
| "loss": 1.5888, | |
| "step": 2816 | |
| }, | |
| { | |
| "epoch": 3.68, | |
| "learning_rate": 3.4165429723912675e-08, | |
| "loss": 1.5982, | |
| "step": 2818 | |
| }, | |
| { | |
| "epoch": 3.68, | |
| "learning_rate": 3.361969272089116e-08, | |
| "loss": 1.5847, | |
| "step": 2820 | |
| }, | |
| { | |
| "epoch": 3.68, | |
| "learning_rate": 3.307827498471982e-08, | |
| "loss": 1.5995, | |
| "step": 2822 | |
| }, | |
| { | |
| "epoch": 3.69, | |
| "learning_rate": 3.254117893528185e-08, | |
| "loss": 1.6057, | |
| "step": 2824 | |
| }, | |
| { | |
| "epoch": 3.69, | |
| "learning_rate": 3.2008406973145e-08, | |
| "loss": 1.6209, | |
| "step": 2826 | |
| }, | |
| { | |
| "epoch": 3.69, | |
| "learning_rate": 3.1479961479550124e-08, | |
| "loss": 1.5954, | |
| "step": 2828 | |
| }, | |
| { | |
| "epoch": 3.69, | |
| "learning_rate": 3.0955844816400675e-08, | |
| "loss": 1.6317, | |
| "step": 2830 | |
| }, | |
| { | |
| "epoch": 3.7, | |
| "learning_rate": 3.043605932625259e-08, | |
| "loss": 1.5995, | |
| "step": 2832 | |
| }, | |
| { | |
| "epoch": 3.7, | |
| "learning_rate": 2.992060733230284e-08, | |
| "loss": 1.5874, | |
| "step": 2834 | |
| }, | |
| { | |
| "epoch": 3.7, | |
| "learning_rate": 2.940949113838065e-08, | |
| "loss": 1.6047, | |
| "step": 2836 | |
| }, | |
| { | |
| "epoch": 3.7, | |
| "learning_rate": 2.8902713028935544e-08, | |
| "loss": 1.6055, | |
| "step": 2838 | |
| }, | |
| { | |
| "epoch": 3.71, | |
| "learning_rate": 2.8400275269028108e-08, | |
| "loss": 1.5751, | |
| "step": 2840 | |
| }, | |
| { | |
| "epoch": 3.71, | |
| "learning_rate": 2.7902180104319438e-08, | |
| "loss": 1.6235, | |
| "step": 2842 | |
| }, | |
| { | |
| "epoch": 3.71, | |
| "learning_rate": 2.740842976106139e-08, | |
| "loss": 1.6023, | |
| "step": 2844 | |
| }, | |
| { | |
| "epoch": 3.71, | |
| "learning_rate": 2.691902644608657e-08, | |
| "loss": 1.6301, | |
| "step": 2846 | |
| }, | |
| { | |
| "epoch": 3.72, | |
| "learning_rate": 2.643397234679823e-08, | |
| "loss": 1.5737, | |
| "step": 2848 | |
| }, | |
| { | |
| "epoch": 3.72, | |
| "learning_rate": 2.5953269631160844e-08, | |
| "loss": 1.593, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 3.72, | |
| "learning_rate": 2.547692044769012e-08, | |
| "loss": 1.5591, | |
| "step": 2852 | |
| }, | |
| { | |
| "epoch": 3.72, | |
| "learning_rate": 2.5004926925443536e-08, | |
| "loss": 1.6058, | |
| "step": 2854 | |
| }, | |
| { | |
| "epoch": 3.73, | |
| "learning_rate": 2.4537291174010822e-08, | |
| "loss": 1.623, | |
| "step": 2856 | |
| }, | |
| { | |
| "epoch": 3.73, | |
| "learning_rate": 2.40740152835045e-08, | |
| "loss": 1.6441, | |
| "step": 2858 | |
| }, | |
| { | |
| "epoch": 3.73, | |
| "learning_rate": 2.3615101324550692e-08, | |
| "loss": 1.5893, | |
| "step": 2860 | |
| }, | |
| { | |
| "epoch": 3.74, | |
| "learning_rate": 2.3160551348279434e-08, | |
| "loss": 1.5671, | |
| "step": 2862 | |
| }, | |
| { | |
| "epoch": 3.74, | |
| "learning_rate": 2.2710367386316154e-08, | |
| "loss": 1.5523, | |
| "step": 2864 | |
| }, | |
| { | |
| "epoch": 3.74, | |
| "learning_rate": 2.2264551450771996e-08, | |
| "loss": 1.642, | |
| "step": 2866 | |
| }, | |
| { | |
| "epoch": 3.74, | |
| "learning_rate": 2.1823105534235164e-08, | |
| "loss": 1.5946, | |
| "step": 2868 | |
| }, | |
| { | |
| "epoch": 3.75, | |
| "learning_rate": 2.1386031609761933e-08, | |
| "loss": 1.5919, | |
| "step": 2870 | |
| }, | |
| { | |
| "epoch": 3.75, | |
| "learning_rate": 2.0953331630867765e-08, | |
| "loss": 1.565, | |
| "step": 2872 | |
| }, | |
| { | |
| "epoch": 3.75, | |
| "learning_rate": 2.0525007531518755e-08, | |
| "loss": 1.5913, | |
| "step": 2874 | |
| }, | |
| { | |
| "epoch": 3.75, | |
| "learning_rate": 2.010106122612265e-08, | |
| "loss": 1.5931, | |
| "step": 2876 | |
| }, | |
| { | |
| "epoch": 3.76, | |
| "learning_rate": 1.968149460952073e-08, | |
| "loss": 1.5929, | |
| "step": 2878 | |
| }, | |
| { | |
| "epoch": 3.76, | |
| "learning_rate": 1.9266309556979165e-08, | |
| "loss": 1.6019, | |
| "step": 2880 | |
| }, | |
| { | |
| "epoch": 3.76, | |
| "learning_rate": 1.8855507924180336e-08, | |
| "loss": 1.5926, | |
| "step": 2882 | |
| }, | |
| { | |
| "epoch": 3.76, | |
| "learning_rate": 1.8449091547214966e-08, | |
| "loss": 1.5855, | |
| "step": 2884 | |
| }, | |
| { | |
| "epoch": 3.77, | |
| "learning_rate": 1.8047062242573573e-08, | |
| "loss": 1.6388, | |
| "step": 2886 | |
| }, | |
| { | |
| "epoch": 3.77, | |
| "learning_rate": 1.7649421807138688e-08, | |
| "loss": 1.5874, | |
| "step": 2888 | |
| }, | |
| { | |
| "epoch": 3.77, | |
| "learning_rate": 1.725617201817686e-08, | |
| "loss": 1.6315, | |
| "step": 2890 | |
| }, | |
| { | |
| "epoch": 3.77, | |
| "learning_rate": 1.686731463333002e-08, | |
| "loss": 1.5194, | |
| "step": 2892 | |
| }, | |
| { | |
| "epoch": 3.78, | |
| "learning_rate": 1.6482851390608233e-08, | |
| "loss": 1.6289, | |
| "step": 2894 | |
| }, | |
| { | |
| "epoch": 3.78, | |
| "learning_rate": 1.6102784008382274e-08, | |
| "loss": 1.5996, | |
| "step": 2896 | |
| }, | |
| { | |
| "epoch": 3.78, | |
| "learning_rate": 1.5727114185374758e-08, | |
| "loss": 1.6209, | |
| "step": 2898 | |
| }, | |
| { | |
| "epoch": 3.78, | |
| "learning_rate": 1.5355843600653895e-08, | |
| "loss": 1.6088, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 3.79, | |
| "learning_rate": 1.498897391362508e-08, | |
| "loss": 1.6058, | |
| "step": 2902 | |
| }, | |
| { | |
| "epoch": 3.79, | |
| "learning_rate": 1.4626506764023661e-08, | |
| "loss": 1.6141, | |
| "step": 2904 | |
| }, | |
| { | |
| "epoch": 3.79, | |
| "learning_rate": 1.4268443771908056e-08, | |
| "loss": 1.5968, | |
| "step": 2906 | |
| }, | |
| { | |
| "epoch": 3.8, | |
| "learning_rate": 1.3914786537651768e-08, | |
| "loss": 1.6004, | |
| "step": 2908 | |
| }, | |
| { | |
| "epoch": 3.8, | |
| "learning_rate": 1.3565536641936826e-08, | |
| "loss": 1.6031, | |
| "step": 2910 | |
| }, | |
| { | |
| "epoch": 3.8, | |
| "learning_rate": 1.3220695645746681e-08, | |
| "loss": 1.5953, | |
| "step": 2912 | |
| }, | |
| { | |
| "epoch": 3.8, | |
| "learning_rate": 1.2880265090358666e-08, | |
| "loss": 1.6611, | |
| "step": 2914 | |
| }, | |
| { | |
| "epoch": 3.81, | |
| "learning_rate": 1.2544246497337984e-08, | |
| "loss": 1.5924, | |
| "step": 2916 | |
| }, | |
| { | |
| "epoch": 3.81, | |
| "learning_rate": 1.221264136852984e-08, | |
| "loss": 1.6318, | |
| "step": 2918 | |
| }, | |
| { | |
| "epoch": 3.81, | |
| "learning_rate": 1.1885451186053886e-08, | |
| "loss": 1.5874, | |
| "step": 2920 | |
| }, | |
| { | |
| "epoch": 3.81, | |
| "learning_rate": 1.1562677412296995e-08, | |
| "loss": 1.6377, | |
| "step": 2922 | |
| }, | |
| { | |
| "epoch": 3.82, | |
| "learning_rate": 1.1244321489906283e-08, | |
| "loss": 1.5846, | |
| "step": 2924 | |
| }, | |
| { | |
| "epoch": 3.82, | |
| "learning_rate": 1.0930384841783546e-08, | |
| "loss": 1.5597, | |
| "step": 2926 | |
| }, | |
| { | |
| "epoch": 3.82, | |
| "learning_rate": 1.0620868871078493e-08, | |
| "loss": 1.6085, | |
| "step": 2928 | |
| }, | |
| { | |
| "epoch": 3.82, | |
| "learning_rate": 1.0315774961182411e-08, | |
| "loss": 1.558, | |
| "step": 2930 | |
| }, | |
| { | |
| "epoch": 3.83, | |
| "learning_rate": 1.0015104475721848e-08, | |
| "loss": 1.5542, | |
| "step": 2932 | |
| }, | |
| { | |
| "epoch": 3.83, | |
| "learning_rate": 9.71885875855294e-09, | |
| "loss": 1.6161, | |
| "step": 2934 | |
| }, | |
| { | |
| "epoch": 3.83, | |
| "learning_rate": 9.4270391337552e-09, | |
| "loss": 1.6006, | |
| "step": 2936 | |
| }, | |
| { | |
| "epoch": 3.83, | |
| "learning_rate": 9.139646905625519e-09, | |
| "loss": 1.5944, | |
| "step": 2938 | |
| }, | |
| { | |
| "epoch": 3.84, | |
| "learning_rate": 8.856683358672402e-09, | |
| "loss": 1.6155, | |
| "step": 2940 | |
| }, | |
| { | |
| "epoch": 3.84, | |
| "learning_rate": 8.578149757610175e-09, | |
| "loss": 1.6357, | |
| "step": 2942 | |
| }, | |
| { | |
| "epoch": 3.84, | |
| "learning_rate": 8.30404734735346e-09, | |
| "loss": 1.584, | |
| "step": 2944 | |
| }, | |
| { | |
| "epoch": 3.84, | |
| "learning_rate": 8.0343773530116e-09, | |
| "loss": 1.5639, | |
| "step": 2946 | |
| }, | |
| { | |
| "epoch": 3.85, | |
| "learning_rate": 7.769140979882905e-09, | |
| "loss": 1.6243, | |
| "step": 2948 | |
| }, | |
| { | |
| "epoch": 3.85, | |
| "learning_rate": 7.508339413449527e-09, | |
| "loss": 1.6321, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 3.85, | |
| "learning_rate": 7.25197381937237e-09, | |
| "loss": 1.6286, | |
| "step": 2952 | |
| }, | |
| { | |
| "epoch": 3.86, | |
| "learning_rate": 7.000045343485306e-09, | |
| "loss": 1.5902, | |
| "step": 2954 | |
| }, | |
| { | |
| "epoch": 3.86, | |
| "learning_rate": 6.752555111790515e-09, | |
| "loss": 1.5719, | |
| "step": 2956 | |
| }, | |
| { | |
| "epoch": 3.86, | |
| "learning_rate": 6.509504230453377e-09, | |
| "loss": 1.6002, | |
| "step": 2958 | |
| }, | |
| { | |
| "epoch": 3.86, | |
| "learning_rate": 6.27089378579726e-09, | |
| "loss": 1.6154, | |
| "step": 2960 | |
| }, | |
| { | |
| "epoch": 3.87, | |
| "learning_rate": 6.0367248442990684e-09, | |
| "loss": 1.5909, | |
| "step": 2962 | |
| }, | |
| { | |
| "epoch": 3.87, | |
| "learning_rate": 5.8069984525840335e-09, | |
| "loss": 1.6277, | |
| "step": 2964 | |
| }, | |
| { | |
| "epoch": 3.87, | |
| "learning_rate": 5.5817156374214916e-09, | |
| "loss": 1.6109, | |
| "step": 2966 | |
| }, | |
| { | |
| "epoch": 3.87, | |
| "learning_rate": 5.36087740572011e-09, | |
| "loss": 1.6023, | |
| "step": 2968 | |
| }, | |
| { | |
| "epoch": 3.88, | |
| "learning_rate": 5.144484744523003e-09, | |
| "loss": 1.6188, | |
| "step": 2970 | |
| }, | |
| { | |
| "epoch": 3.88, | |
| "learning_rate": 4.9325386210040675e-09, | |
| "loss": 1.62, | |
| "step": 2972 | |
| }, | |
| { | |
| "epoch": 3.88, | |
| "learning_rate": 4.7250399824629865e-09, | |
| "loss": 1.5778, | |
| "step": 2974 | |
| }, | |
| { | |
| "epoch": 3.88, | |
| "learning_rate": 4.521989756321565e-09, | |
| "loss": 1.6444, | |
| "step": 2976 | |
| }, | |
| { | |
| "epoch": 3.89, | |
| "learning_rate": 4.323388850118848e-09, | |
| "loss": 1.579, | |
| "step": 2978 | |
| }, | |
| { | |
| "epoch": 3.89, | |
| "learning_rate": 4.129238151508008e-09, | |
| "loss": 1.6021, | |
| "step": 2980 | |
| }, | |
| { | |
| "epoch": 3.89, | |
| "learning_rate": 3.939538528251462e-09, | |
| "loss": 1.572, | |
| "step": 2982 | |
| }, | |
| { | |
| "epoch": 3.89, | |
| "learning_rate": 3.7542908282176545e-09, | |
| "loss": 1.5995, | |
| "step": 2984 | |
| }, | |
| { | |
| "epoch": 3.9, | |
| "learning_rate": 3.573495879376942e-09, | |
| "loss": 1.6413, | |
| "step": 2986 | |
| }, | |
| { | |
| "epoch": 3.9, | |
| "learning_rate": 3.3971544897980488e-09, | |
| "loss": 1.5797, | |
| "step": 2988 | |
| }, | |
| { | |
| "epoch": 3.9, | |
| "learning_rate": 3.2252674476440643e-09, | |
| "loss": 1.6453, | |
| "step": 2990 | |
| }, | |
| { | |
| "epoch": 3.9, | |
| "learning_rate": 3.0578355211697824e-09, | |
| "loss": 1.5531, | |
| "step": 2992 | |
| }, | |
| { | |
| "epoch": 3.91, | |
| "learning_rate": 2.894859458717036e-09, | |
| "loss": 1.5881, | |
| "step": 2994 | |
| }, | |
| { | |
| "epoch": 3.91, | |
| "learning_rate": 2.7363399887128104e-09, | |
| "loss": 1.5912, | |
| "step": 2996 | |
| }, | |
| { | |
| "epoch": 3.91, | |
| "learning_rate": 2.58227781966458e-09, | |
| "loss": 1.6239, | |
| "step": 2998 | |
| }, | |
| { | |
| "epoch": 3.92, | |
| "learning_rate": 2.4326736401579784e-09, | |
| "loss": 1.6034, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 3.92, | |
| "learning_rate": 2.2875281188536875e-09, | |
| "loss": 1.6309, | |
| "step": 3002 | |
| }, | |
| { | |
| "epoch": 3.92, | |
| "learning_rate": 2.1468419044839982e-09, | |
| "loss": 1.5689, | |
| "step": 3004 | |
| }, | |
| { | |
| "epoch": 3.92, | |
| "learning_rate": 2.010615625850365e-09, | |
| "loss": 1.6326, | |
| "step": 3006 | |
| }, | |
| { | |
| "epoch": 3.93, | |
| "learning_rate": 1.878849891820411e-09, | |
| "loss": 1.5711, | |
| "step": 3008 | |
| }, | |
| { | |
| "epoch": 3.93, | |
| "learning_rate": 1.7515452913250405e-09, | |
| "loss": 1.5815, | |
| "step": 3010 | |
| }, | |
| { | |
| "epoch": 3.93, | |
| "learning_rate": 1.689566094741912e-09, | |
| "loss": 1.5682, | |
| "step": 3012 | |
| }, | |
| { | |
| "epoch": 3.93, | |
| "learning_rate": 1.6287023933564403e-09, | |
| "loss": 1.5606, | |
| "step": 3014 | |
| }, | |
| { | |
| "epoch": 3.94, | |
| "learning_rate": 1.5103217469644158e-09, | |
| "loss": 1.5411, | |
| "step": 3016 | |
| }, | |
| { | |
| "epoch": 3.94, | |
| "learning_rate": 1.3964038812551705e-09, | |
| "loss": 1.6349, | |
| "step": 3018 | |
| }, | |
| { | |
| "epoch": 3.94, | |
| "learning_rate": 1.2869493053880853e-09, | |
| "loss": 1.6072, | |
| "step": 3020 | |
| }, | |
| { | |
| "epoch": 3.94, | |
| "learning_rate": 1.1819585085737215e-09, | |
| "loss": 1.599, | |
| "step": 3022 | |
| }, | |
| { | |
| "epoch": 3.95, | |
| "learning_rate": 1.0814319600718213e-09, | |
| "loss": 1.6435, | |
| "step": 3024 | |
| }, | |
| { | |
| "epoch": 3.95, | |
| "learning_rate": 9.853701091888656e-10, | |
| "loss": 1.6176, | |
| "step": 3026 | |
| }, | |
| { | |
| "epoch": 3.95, | |
| "learning_rate": 8.937733852764085e-10, | |
| "loss": 1.5915, | |
| "step": 3028 | |
| }, | |
| { | |
| "epoch": 3.95, | |
| "learning_rate": 8.066421977286352e-10, | |
| "loss": 1.5667, | |
| "step": 3030 | |
| }, | |
| { | |
| "epoch": 3.96, | |
| "learning_rate": 7.23976935981141e-10, | |
| "loss": 1.5865, | |
| "step": 3032 | |
| }, | |
| { | |
| "epoch": 3.96, | |
| "learning_rate": 6.457779695090426e-10, | |
| "loss": 1.594, | |
| "step": 3034 | |
| }, | |
| { | |
| "epoch": 3.96, | |
| "learning_rate": 5.720456478249813e-10, | |
| "loss": 1.6078, | |
| "step": 3036 | |
| }, | |
| { | |
| "epoch": 3.96, | |
| "learning_rate": 5.027803004779008e-10, | |
| "loss": 1.6193, | |
| "step": 3038 | |
| }, | |
| { | |
| "epoch": 3.97, | |
| "learning_rate": 4.379822370512709e-10, | |
| "loss": 1.5808, | |
| "step": 3040 | |
| }, | |
| { | |
| "epoch": 3.97, | |
| "learning_rate": 3.776517471621998e-10, | |
| "loss": 1.5668, | |
| "step": 3042 | |
| }, | |
| { | |
| "epoch": 3.97, | |
| "learning_rate": 3.2178910045965734e-10, | |
| "loss": 1.5583, | |
| "step": 3044 | |
| }, | |
| { | |
| "epoch": 3.98, | |
| "learning_rate": 2.703945466233648e-10, | |
| "loss": 1.6006, | |
| "step": 3046 | |
| }, | |
| { | |
| "epoch": 3.98, | |
| "learning_rate": 2.2346831536312892e-10, | |
| "loss": 1.6075, | |
| "step": 3048 | |
| }, | |
| { | |
| "epoch": 3.98, | |
| "learning_rate": 1.8101061641695447e-10, | |
| "loss": 1.6492, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 3.98, | |
| "learning_rate": 1.4302163955093316e-10, | |
| "loss": 1.5577, | |
| "step": 3052 | |
| }, | |
| { | |
| "epoch": 3.99, | |
| "learning_rate": 1.0950155455802245e-10, | |
| "loss": 1.5638, | |
| "step": 3054 | |
| }, | |
| { | |
| "epoch": 3.99, | |
| "learning_rate": 8.045051125726843e-11, | |
| "loss": 1.5856, | |
| "step": 3056 | |
| }, | |
| { | |
| "epoch": 3.99, | |
| "learning_rate": 5.5868639493250603e-11, | |
| "loss": 1.5727, | |
| "step": 3058 | |
| }, | |
| { | |
| "epoch": 3.99, | |
| "learning_rate": 3.5756049135304836e-11, | |
| "loss": 1.6079, | |
| "step": 3060 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 2.0112830077301247e-11, | |
| "loss": 1.5869, | |
| "step": 3062 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 8.939052237089128e-12, | |
| "loss": 1.5579, | |
| "step": 3064 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "step": 3064, | |
| "total_flos": 1.7885944480792576e+17, | |
| "train_loss": 1.6310015863133474, | |
| "train_runtime": 32557.3578, | |
| "train_samples_per_second": 6.024, | |
| "train_steps_per_second": 0.094 | |
| } | |
| ], | |
| "logging_steps": 2, | |
| "max_steps": 3064, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 4, | |
| "save_steps": 4000, | |
| "total_flos": 1.7885944480792576e+17, | |
| "train_batch_size": 4, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |