| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 1.0, | |
| "eval_steps": 500, | |
| "global_step": 1269, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 5.128205128205128e-08, | |
| "loss": 0.1219, | |
| "step": 1 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 1.0256410256410256e-07, | |
| "loss": 0.1897, | |
| "step": 2 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 1.5384615384615385e-07, | |
| "loss": 0.1738, | |
| "step": 3 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 2.0512820512820512e-07, | |
| "loss": 0.1381, | |
| "step": 4 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 2.5641025641025636e-07, | |
| "loss": 0.2269, | |
| "step": 5 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 3.076923076923077e-07, | |
| "loss": 0.166, | |
| "step": 6 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 3.5897435897435896e-07, | |
| "loss": 0.1792, | |
| "step": 7 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 4.1025641025641024e-07, | |
| "loss": 0.1033, | |
| "step": 8 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 4.6153846153846156e-07, | |
| "loss": 0.1667, | |
| "step": 9 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 5.128205128205127e-07, | |
| "loss": 0.1539, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 5.641025641025641e-07, | |
| "loss": 0.1112, | |
| "step": 11 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 6.153846153846154e-07, | |
| "loss": 0.1822, | |
| "step": 12 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 6.666666666666666e-07, | |
| "loss": 0.1066, | |
| "step": 13 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 7.179487179487179e-07, | |
| "loss": 0.1413, | |
| "step": 14 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 7.692307692307693e-07, | |
| "loss": 0.0961, | |
| "step": 15 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 8.205128205128205e-07, | |
| "loss": 0.1488, | |
| "step": 16 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 8.717948717948718e-07, | |
| "loss": 0.1815, | |
| "step": 17 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.230769230769231e-07, | |
| "loss": 0.2033, | |
| "step": 18 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.743589743589742e-07, | |
| "loss": 0.1526, | |
| "step": 19 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.0256410256410255e-06, | |
| "loss": 0.197, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.0769230769230769e-06, | |
| "loss": 0.111, | |
| "step": 21 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.1282051282051281e-06, | |
| "loss": 0.1298, | |
| "step": 22 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.1794871794871795e-06, | |
| "loss": 0.117, | |
| "step": 23 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.2307692307692308e-06, | |
| "loss": 0.1328, | |
| "step": 24 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.2820512820512822e-06, | |
| "loss": 0.1367, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.3333333333333332e-06, | |
| "loss": 0.1189, | |
| "step": 26 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.3846153846153844e-06, | |
| "loss": 0.106, | |
| "step": 27 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.4358974358974359e-06, | |
| "loss": 0.1067, | |
| "step": 28 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.487179487179487e-06, | |
| "loss": 0.1524, | |
| "step": 29 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.5384615384615385e-06, | |
| "loss": 0.1196, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.5897435897435895e-06, | |
| "loss": 0.1157, | |
| "step": 31 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.641025641025641e-06, | |
| "loss": 0.1074, | |
| "step": 32 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.6923076923076922e-06, | |
| "loss": 0.1111, | |
| "step": 33 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.7435897435897436e-06, | |
| "loss": 0.0999, | |
| "step": 34 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.7948717948717948e-06, | |
| "loss": 0.1539, | |
| "step": 35 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.8461538461538462e-06, | |
| "loss": 0.0695, | |
| "step": 36 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.8974358974358973e-06, | |
| "loss": 0.1428, | |
| "step": 37 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.9487179487179485e-06, | |
| "loss": 0.1263, | |
| "step": 38 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2e-06, | |
| "loss": 0.1415, | |
| "step": 39 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.9999967381852615e-06, | |
| "loss": 0.0958, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.999986952762325e-06, | |
| "loss": 0.0565, | |
| "step": 41 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.9999706437950266e-06, | |
| "loss": 0.1093, | |
| "step": 42 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.999947811389761e-06, | |
| "loss": 0.0964, | |
| "step": 43 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.9999184556954774e-06, | |
| "loss": 0.1149, | |
| "step": 44 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9998825769036817e-06, | |
| "loss": 0.1131, | |
| "step": 45 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9998401752484335e-06, | |
| "loss": 0.0708, | |
| "step": 46 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.999791251006346e-06, | |
| "loss": 0.1138, | |
| "step": 47 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9997358044965828e-06, | |
| "loss": 0.1655, | |
| "step": 48 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.999673836080856e-06, | |
| "loss": 0.077, | |
| "step": 49 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9996053461634253e-06, | |
| "loss": 0.1785, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.999530335191093e-06, | |
| "loss": 0.0901, | |
| "step": 51 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.999448803653203e-06, | |
| "loss": 0.0857, | |
| "step": 52 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.999360752081637e-06, | |
| "loss": 0.0835, | |
| "step": 53 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.99926618105081e-06, | |
| "loss": 0.0944, | |
| "step": 54 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9991650911776694e-06, | |
| "loss": 0.1283, | |
| "step": 55 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9990574831216874e-06, | |
| "loss": 0.1002, | |
| "step": 56 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.99894335758486e-06, | |
| "loss": 0.1172, | |
| "step": 57 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.998822715311699e-06, | |
| "loss": 0.1096, | |
| "step": 58 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9986955570892297e-06, | |
| "loss": 0.1424, | |
| "step": 59 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.998561883746986e-06, | |
| "loss": 0.0823, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9984216961570036e-06, | |
| "loss": 0.1253, | |
| "step": 61 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.998274995233813e-06, | |
| "loss": 0.0871, | |
| "step": 62 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.998121781934438e-06, | |
| "loss": 0.1174, | |
| "step": 63 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9979620572583845e-06, | |
| "loss": 0.0737, | |
| "step": 64 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9977958222476373e-06, | |
| "loss": 0.1288, | |
| "step": 65 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9976230779866523e-06, | |
| "loss": 0.2142, | |
| "step": 66 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9974438256023486e-06, | |
| "loss": 0.0687, | |
| "step": 67 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9972580662641032e-06, | |
| "loss": 0.0577, | |
| "step": 68 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9970658011837403e-06, | |
| "loss": 0.0603, | |
| "step": 69 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9968670316155264e-06, | |
| "loss": 0.085, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.996661758856161e-06, | |
| "loss": 0.1204, | |
| "step": 71 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9964499842447664e-06, | |
| "loss": 0.0952, | |
| "step": 72 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9962317091628825e-06, | |
| "loss": 0.1838, | |
| "step": 73 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9960069350344546e-06, | |
| "loss": 0.1564, | |
| "step": 74 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9957756633258264e-06, | |
| "loss": 0.1184, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9955378955457286e-06, | |
| "loss": 0.0735, | |
| "step": 76 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9952936332452697e-06, | |
| "loss": 0.0792, | |
| "step": 77 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9950428780179273e-06, | |
| "loss": 0.1148, | |
| "step": 78 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9947856314995347e-06, | |
| "loss": 0.0689, | |
| "step": 79 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.994521895368273e-06, | |
| "loss": 0.0984, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9942516713446596e-06, | |
| "loss": 0.0627, | |
| "step": 81 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9939749611915358e-06, | |
| "loss": 0.1009, | |
| "step": 82 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.993691766714055e-06, | |
| "loss": 0.1338, | |
| "step": 83 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9934020897596747e-06, | |
| "loss": 0.0815, | |
| "step": 84 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9931059322181393e-06, | |
| "loss": 0.0743, | |
| "step": 85 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9928032960214705e-06, | |
| "loss": 0.2224, | |
| "step": 86 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9924941831439547e-06, | |
| "loss": 0.1491, | |
| "step": 87 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9921785956021304e-06, | |
| "loss": 0.0695, | |
| "step": 88 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9918565354547736e-06, | |
| "loss": 0.0841, | |
| "step": 89 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.991528004802885e-06, | |
| "loss": 0.1242, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9911930057896774e-06, | |
| "loss": 0.1025, | |
| "step": 91 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9908515406005593e-06, | |
| "loss": 0.1144, | |
| "step": 92 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9905036114631245e-06, | |
| "loss": 0.114, | |
| "step": 93 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9901492206471324e-06, | |
| "loss": 0.0397, | |
| "step": 94 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.989788370464498e-06, | |
| "loss": 0.1077, | |
| "step": 95 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.989421063269274e-06, | |
| "loss": 0.1196, | |
| "step": 96 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9890473014576374e-06, | |
| "loss": 0.1062, | |
| "step": 97 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9886670874678703e-06, | |
| "loss": 0.0443, | |
| "step": 98 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9882804237803485e-06, | |
| "loss": 0.1749, | |
| "step": 99 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.987887312917523e-06, | |
| "loss": 0.1189, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9874877574439024e-06, | |
| "loss": 0.0986, | |
| "step": 101 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.98708175996604e-06, | |
| "loss": 0.1567, | |
| "step": 102 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9866693231325117e-06, | |
| "loss": 0.0902, | |
| "step": 103 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9862504496339034e-06, | |
| "loss": 0.0995, | |
| "step": 104 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.98582514220279e-06, | |
| "loss": 0.1066, | |
| "step": 105 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.98539340361372e-06, | |
| "loss": 0.1264, | |
| "step": 106 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.984955236683196e-06, | |
| "loss": 0.0841, | |
| "step": 107 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9845106442696558e-06, | |
| "loss": 0.1025, | |
| "step": 108 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.984059629273457e-06, | |
| "loss": 0.1147, | |
| "step": 109 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9836021946368537e-06, | |
| "loss": 0.0909, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9831383433439796e-06, | |
| "loss": 0.057, | |
| "step": 111 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.982668078420829e-06, | |
| "loss": 0.0843, | |
| "step": 112 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.982191402935236e-06, | |
| "loss": 0.1366, | |
| "step": 113 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.981708319996855e-06, | |
| "loss": 0.1024, | |
| "step": 114 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9812188327571398e-06, | |
| "loss": 0.1093, | |
| "step": 115 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9807229444093238e-06, | |
| "loss": 0.0816, | |
| "step": 116 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.980220658188399e-06, | |
| "loss": 0.0944, | |
| "step": 117 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.979711977371095e-06, | |
| "loss": 0.106, | |
| "step": 118 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.979196905275856e-06, | |
| "loss": 0.0827, | |
| "step": 119 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9786754452628224e-06, | |
| "loss": 0.1383, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.9781476007338054e-06, | |
| "loss": 0.0958, | |
| "step": 121 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.977613375132268e-06, | |
| "loss": 0.0881, | |
| "step": 122 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.977072771943299e-06, | |
| "loss": 0.0998, | |
| "step": 123 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.9765257946935944e-06, | |
| "loss": 0.1385, | |
| "step": 124 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.97597244695143e-06, | |
| "loss": 0.0821, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.9754127323266424e-06, | |
| "loss": 0.1069, | |
| "step": 126 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.974846654470602e-06, | |
| "loss": 0.0661, | |
| "step": 127 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.974274217076191e-06, | |
| "loss": 0.1304, | |
| "step": 128 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.973695423877779e-06, | |
| "loss": 0.112, | |
| "step": 129 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.973110278651198e-06, | |
| "loss": 0.1044, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.972518785213719e-06, | |
| "loss": 0.1122, | |
| "step": 131 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.971920947424026e-06, | |
| "loss": 0.103, | |
| "step": 132 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.971316769182191e-06, | |
| "loss": 0.1221, | |
| "step": 133 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9707062544296495e-06, | |
| "loss": 0.0897, | |
| "step": 134 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9700894071491733e-06, | |
| "loss": 0.0974, | |
| "step": 135 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9694662313648446e-06, | |
| "loss": 0.1191, | |
| "step": 136 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9688367311420326e-06, | |
| "loss": 0.0973, | |
| "step": 137 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9682009105873633e-06, | |
| "loss": 0.0772, | |
| "step": 138 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9675587738486934e-06, | |
| "loss": 0.1041, | |
| "step": 139 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.966910325115086e-06, | |
| "loss": 0.0947, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9662555686167805e-06, | |
| "loss": 0.1203, | |
| "step": 141 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9655945086251652e-06, | |
| "loss": 0.104, | |
| "step": 142 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9649271494527507e-06, | |
| "loss": 0.0385, | |
| "step": 143 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.964253495453141e-06, | |
| "loss": 0.0826, | |
| "step": 144 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.963573551021005e-06, | |
| "loss": 0.1079, | |
| "step": 145 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9628873205920485e-06, | |
| "loss": 0.0905, | |
| "step": 146 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9621948086429842e-06, | |
| "loss": 0.1214, | |
| "step": 147 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.961496019691504e-06, | |
| "loss": 0.094, | |
| "step": 148 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9607909582962476e-06, | |
| "loss": 0.0839, | |
| "step": 149 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9600796290567743e-06, | |
| "loss": 0.1209, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9593620366135337e-06, | |
| "loss": 0.0733, | |
| "step": 151 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.958638185647831e-06, | |
| "loss": 0.0754, | |
| "step": 152 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9579080808818032e-06, | |
| "loss": 0.1302, | |
| "step": 153 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9571717270783824e-06, | |
| "loss": 0.1193, | |
| "step": 154 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.956429129041269e-06, | |
| "loss": 0.1008, | |
| "step": 155 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.955680291614896e-06, | |
| "loss": 0.066, | |
| "step": 156 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9549252196844024e-06, | |
| "loss": 0.134, | |
| "step": 157 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9541639181755976e-06, | |
| "loss": 0.1128, | |
| "step": 158 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9533963920549303e-06, | |
| "loss": 0.1061, | |
| "step": 159 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9526226463294568e-06, | |
| "loss": 0.085, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.951842686046807e-06, | |
| "loss": 0.0726, | |
| "step": 161 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9510565162951534e-06, | |
| "loss": 0.0521, | |
| "step": 162 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.950264142203176e-06, | |
| "loss": 0.0997, | |
| "step": 163 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9494655689400293e-06, | |
| "loss": 0.0865, | |
| "step": 164 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.94866080171531e-06, | |
| "loss": 0.0632, | |
| "step": 165 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.947849845779021e-06, | |
| "loss": 0.0946, | |
| "step": 166 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9470327064215382e-06, | |
| "loss": 0.0749, | |
| "step": 167 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9462093889735766e-06, | |
| "loss": 0.1071, | |
| "step": 168 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.945379898806153e-06, | |
| "loss": 0.0767, | |
| "step": 169 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9445442413305555e-06, | |
| "loss": 0.0776, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9437024219983028e-06, | |
| "loss": 0.0928, | |
| "step": 171 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9428544463011124e-06, | |
| "loss": 0.0503, | |
| "step": 172 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9420003197708637e-06, | |
| "loss": 0.073, | |
| "step": 173 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9411400479795615e-06, | |
| "loss": 0.084, | |
| "step": 174 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9402736365393007e-06, | |
| "loss": 0.0698, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9394010911022283e-06, | |
| "loss": 0.066, | |
| "step": 176 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.938522417360507e-06, | |
| "loss": 0.0547, | |
| "step": 177 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9376376210462788e-06, | |
| "loss": 0.0711, | |
| "step": 178 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9367467079316276e-06, | |
| "loss": 0.0332, | |
| "step": 179 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9358496838285406e-06, | |
| "loss": 0.0376, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9349465545888696e-06, | |
| "loss": 0.0495, | |
| "step": 181 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.934037326104296e-06, | |
| "loss": 0.0319, | |
| "step": 182 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.933122004306289e-06, | |
| "loss": 0.0232, | |
| "step": 183 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9322005951660697e-06, | |
| "loss": 0.0151, | |
| "step": 184 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.931273104694569e-06, | |
| "loss": 0.0231, | |
| "step": 185 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9303395389423915e-06, | |
| "loss": 0.0915, | |
| "step": 186 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9293999039997745e-06, | |
| "loss": 0.0108, | |
| "step": 187 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9284542059965474e-06, | |
| "loss": 0.0387, | |
| "step": 188 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.927502451102095e-06, | |
| "loss": 0.0243, | |
| "step": 189 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9265446455253118e-06, | |
| "loss": 0.0784, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9255807955145676e-06, | |
| "loss": 0.0681, | |
| "step": 191 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9246109073576627e-06, | |
| "loss": 0.031, | |
| "step": 192 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9236349873817875e-06, | |
| "loss": 0.0173, | |
| "step": 193 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.922653041953483e-06, | |
| "loss": 0.0866, | |
| "step": 194 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9216650774785973e-06, | |
| "loss": 0.118, | |
| "step": 195 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.920671100402244e-06, | |
| "loss": 0.0613, | |
| "step": 196 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9196711172087616e-06, | |
| "loss": 0.073, | |
| "step": 197 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.91866513442167e-06, | |
| "loss": 0.0247, | |
| "step": 198 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.917653158603628e-06, | |
| "loss": 0.0113, | |
| "step": 199 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.916635196356391e-06, | |
| "loss": 0.0069, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.915611254320767e-06, | |
| "loss": 0.0134, | |
| "step": 201 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.914581339176576e-06, | |
| "loss": 0.0238, | |
| "step": 202 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9135454576426007e-06, | |
| "loss": 0.0345, | |
| "step": 203 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.91250361647655e-06, | |
| "loss": 0.0069, | |
| "step": 204 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9114558224750087e-06, | |
| "loss": 0.0113, | |
| "step": 205 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9104020824733975e-06, | |
| "loss": 0.0144, | |
| "step": 206 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9093424033459247e-06, | |
| "loss": 0.0085, | |
| "step": 207 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.908276792005545e-06, | |
| "loss": 0.049, | |
| "step": 208 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.907205255403912e-06, | |
| "loss": 0.014, | |
| "step": 209 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.9061278005313329e-06, | |
| "loss": 0.031, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.9050444344167248e-06, | |
| "loss": 0.0393, | |
| "step": 211 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.903955164127566e-06, | |
| "loss": 0.0037, | |
| "step": 212 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.9028599967698532e-06, | |
| "loss": 0.0462, | |
| "step": 213 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.9017589394880512e-06, | |
| "loss": 0.0205, | |
| "step": 214 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.900651999465051e-06, | |
| "loss": 0.0245, | |
| "step": 215 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.8995391839221187e-06, | |
| "loss": 0.0188, | |
| "step": 216 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.8984205001188505e-06, | |
| "loss": 0.0037, | |
| "step": 217 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.897295955353125e-06, | |
| "loss": 0.0446, | |
| "step": 218 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.8961655569610556e-06, | |
| "loss": 0.0152, | |
| "step": 219 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.8950293123169425e-06, | |
| "loss": 0.0521, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.893887228833225e-06, | |
| "loss": 0.0933, | |
| "step": 221 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.8927393139604323e-06, | |
| "loss": 0.0073, | |
| "step": 222 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8915855751871362e-06, | |
| "loss": 0.0162, | |
| "step": 223 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8904260200399004e-06, | |
| "loss": 0.0399, | |
| "step": 224 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8892606560832333e-06, | |
| "loss": 0.0122, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8880894909195377e-06, | |
| "loss": 0.0029, | |
| "step": 226 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.886912532189061e-06, | |
| "loss": 0.0242, | |
| "step": 227 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8857297875698452e-06, | |
| "loss": 0.0178, | |
| "step": 228 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8845412647776793e-06, | |
| "loss": 0.0378, | |
| "step": 229 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8833469715660449e-06, | |
| "loss": 0.0206, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8821469157260684e-06, | |
| "loss": 0.013, | |
| "step": 231 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8809411050864691e-06, | |
| "loss": 0.0251, | |
| "step": 232 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8797295475135098e-06, | |
| "loss": 0.0226, | |
| "step": 233 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8785122509109423e-06, | |
| "loss": 0.0307, | |
| "step": 234 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.877289223219959e-06, | |
| "loss": 0.0418, | |
| "step": 235 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8760604724191395e-06, | |
| "loss": 0.0015, | |
| "step": 236 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8748260065243984e-06, | |
| "loss": 0.0091, | |
| "step": 237 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.873585833588934e-06, | |
| "loss": 0.0144, | |
| "step": 238 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.872339961703175e-06, | |
| "loss": 0.031, | |
| "step": 239 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8710883989947278e-06, | |
| "loss": 0.0049, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8698311536283243e-06, | |
| "loss": 0.0238, | |
| "step": 241 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8685682338057664e-06, | |
| "loss": 0.1654, | |
| "step": 242 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8672996477658764e-06, | |
| "loss": 0.0323, | |
| "step": 243 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8660254037844386e-06, | |
| "loss": 0.0072, | |
| "step": 244 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.864745510174149e-06, | |
| "loss": 0.016, | |
| "step": 245 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.863459975284559e-06, | |
| "loss": 0.0356, | |
| "step": 246 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8621688075020226e-06, | |
| "loss": 0.0252, | |
| "step": 247 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8608720152496389e-06, | |
| "loss": 0.03, | |
| "step": 248 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.859569606987201e-06, | |
| "loss": 0.0327, | |
| "step": 249 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8582615912111375e-06, | |
| "loss": 0.015, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8569479764544587e-06, | |
| "loss": 0.029, | |
| "step": 251 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8556287712867005e-06, | |
| "loss": 0.0181, | |
| "step": 252 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8543039843138685e-06, | |
| "loss": 0.0133, | |
| "step": 253 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8529736241783823e-06, | |
| "loss": 0.0228, | |
| "step": 254 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8516376995590183e-06, | |
| "loss": 0.0115, | |
| "step": 255 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.850296219170854e-06, | |
| "loss": 0.0316, | |
| "step": 256 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.84894919176521e-06, | |
| "loss": 0.0173, | |
| "step": 257 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8475966261295945e-06, | |
| "loss": 0.0081, | |
| "step": 258 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.846238531087644e-06, | |
| "loss": 0.016, | |
| "step": 259 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8448749154990674e-06, | |
| "loss": 0.0165, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8435057882595881e-06, | |
| "loss": 0.0068, | |
| "step": 261 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8421311583008846e-06, | |
| "loss": 0.0139, | |
| "step": 262 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.840751034590533e-06, | |
| "loss": 0.0128, | |
| "step": 263 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.83936542613195e-06, | |
| "loss": 0.0273, | |
| "step": 264 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8379743419643307e-06, | |
| "loss": 0.0688, | |
| "step": 265 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8365777911625934e-06, | |
| "loss": 0.0123, | |
| "step": 266 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.835175782837318e-06, | |
| "loss": 0.0266, | |
| "step": 267 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8337683261346871e-06, | |
| "loss": 0.0137, | |
| "step": 268 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.832355430236427e-06, | |
| "loss": 0.0107, | |
| "step": 269 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.830937104359747e-06, | |
| "loss": 0.0027, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8295133577572797e-06, | |
| "loss": 0.0585, | |
| "step": 271 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.82808419971702e-06, | |
| "loss": 0.0366, | |
| "step": 272 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.826649639562266e-06, | |
| "loss": 0.0228, | |
| "step": 273 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8252096866515556e-06, | |
| "loss": 0.0039, | |
| "step": 274 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8237643503786094e-06, | |
| "loss": 0.0393, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8223136401722647e-06, | |
| "loss": 0.0273, | |
| "step": 276 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8208575654964178e-06, | |
| "loss": 0.0162, | |
| "step": 277 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8193961358499602e-06, | |
| "loss": 0.0166, | |
| "step": 278 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8179293607667177e-06, | |
| "loss": 0.0102, | |
| "step": 279 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.816457249815387e-06, | |
| "loss": 0.0677, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8149798125994749e-06, | |
| "loss": 0.0113, | |
| "step": 281 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8134970587572343e-06, | |
| "loss": 0.0135, | |
| "step": 282 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8120089979616019e-06, | |
| "loss": 0.0285, | |
| "step": 283 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8105156399201346e-06, | |
| "loss": 0.0157, | |
| "step": 284 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8090169943749474e-06, | |
| "loss": 0.0056, | |
| "step": 285 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.8075130711026483e-06, | |
| "loss": 0.021, | |
| "step": 286 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.8060038799142757e-06, | |
| "loss": 0.0425, | |
| "step": 287 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.8044894306552334e-06, | |
| "loss": 0.0112, | |
| "step": 288 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.8029697332052275e-06, | |
| "loss": 0.0097, | |
| "step": 289 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.801444797478201e-06, | |
| "loss": 0.0077, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.7999146334222694e-06, | |
| "loss": 0.0537, | |
| "step": 291 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.7983792510196565e-06, | |
| "loss": 0.0077, | |
| "step": 292 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.7968386602866276e-06, | |
| "loss": 0.0044, | |
| "step": 293 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.7952928712734265e-06, | |
| "loss": 0.0307, | |
| "step": 294 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.7937418940642072e-06, | |
| "loss": 0.0323, | |
| "step": 295 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.792185738776971e-06, | |
| "loss": 0.0207, | |
| "step": 296 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.790624415563498e-06, | |
| "loss": 0.0095, | |
| "step": 297 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.7890579346092825e-06, | |
| "loss": 0.0007, | |
| "step": 298 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7874863061334654e-06, | |
| "loss": 0.0179, | |
| "step": 299 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7859095403887695e-06, | |
| "loss": 0.0027, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7843276476614294e-06, | |
| "loss": 0.0255, | |
| "step": 301 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.782740638271128e-06, | |
| "loss": 0.0106, | |
| "step": 302 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7811485225709252e-06, | |
| "loss": 0.0046, | |
| "step": 303 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.779551310947195e-06, | |
| "loss": 0.0289, | |
| "step": 304 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7779490138195543e-06, | |
| "loss": 0.0238, | |
| "step": 305 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.776341641640795e-06, | |
| "loss": 0.0138, | |
| "step": 306 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7747292048968186e-06, | |
| "loss": 0.0075, | |
| "step": 307 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.773111714106564e-06, | |
| "loss": 0.0398, | |
| "step": 308 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.771489179821943e-06, | |
| "loss": 0.0064, | |
| "step": 309 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7698616126277667e-06, | |
| "loss": 0.0258, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7682290231416817e-06, | |
| "loss": 0.0166, | |
| "step": 311 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.766591422014096e-06, | |
| "loss": 0.0076, | |
| "step": 312 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7649488199281128e-06, | |
| "loss": 0.0067, | |
| "step": 313 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7633012275994599e-06, | |
| "loss": 0.0083, | |
| "step": 314 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7616486557764186e-06, | |
| "loss": 0.004, | |
| "step": 315 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7599911152397558e-06, | |
| "loss": 0.0007, | |
| "step": 316 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7583286168026508e-06, | |
| "loss": 0.0565, | |
| "step": 317 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7566611713106285e-06, | |
| "loss": 0.0353, | |
| "step": 318 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.754988789641485e-06, | |
| "loss": 0.0386, | |
| "step": 319 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7533114827052184e-06, | |
| "loss": 0.0016, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7516292614439584e-06, | |
| "loss": 0.014, | |
| "step": 321 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7499421368318924e-06, | |
| "loss": 0.0813, | |
| "step": 322 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7482501198751964e-06, | |
| "loss": 0.0405, | |
| "step": 323 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7465532216119624e-06, | |
| "loss": 0.0054, | |
| "step": 324 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7448514531121262e-06, | |
| "loss": 0.021, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.743144825477394e-06, | |
| "loss": 0.013, | |
| "step": 326 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.741433349841173e-06, | |
| "loss": 0.0356, | |
| "step": 327 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7397170373684959e-06, | |
| "loss": 0.0305, | |
| "step": 328 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7379958992559492e-06, | |
| "loss": 0.0369, | |
| "step": 329 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7362699467316002e-06, | |
| "loss": 0.0144, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7345391910549236e-06, | |
| "loss": 0.0138, | |
| "step": 331 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7328036435167286e-06, | |
| "loss": 0.0223, | |
| "step": 332 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.731063315439084e-06, | |
| "loss": 0.0175, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7293182181752449e-06, | |
| "loss": 0.0066, | |
| "step": 334 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.72756836310958e-06, | |
| "loss": 0.0133, | |
| "step": 335 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7258137616574948e-06, | |
| "loss": 0.0086, | |
| "step": 336 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7240544252653594e-06, | |
| "loss": 0.0107, | |
| "step": 337 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7222903654104327e-06, | |
| "loss": 0.0831, | |
| "step": 338 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7205215936007869e-06, | |
| "loss": 0.0265, | |
| "step": 339 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7187481213752347e-06, | |
| "loss": 0.0031, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7169699603032513e-06, | |
| "loss": 0.0122, | |
| "step": 341 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7151871219849006e-06, | |
| "loss": 0.0413, | |
| "step": 342 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7133996180507598e-06, | |
| "loss": 0.0171, | |
| "step": 343 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7116074601618415e-06, | |
| "loss": 0.0627, | |
| "step": 344 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7098106600095203e-06, | |
| "loss": 0.0255, | |
| "step": 345 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7080092293154542e-06, | |
| "loss": 0.0072, | |
| "step": 346 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7062031798315099e-06, | |
| "loss": 0.0204, | |
| "step": 347 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7043925233396852e-06, | |
| "loss": 0.0057, | |
| "step": 348 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.702577271652032e-06, | |
| "loss": 0.0158, | |
| "step": 349 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.7007574366105795e-06, | |
| "loss": 0.0221, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.6989330300872572e-06, | |
| "loss": 0.0108, | |
| "step": 351 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.6971040639838174e-06, | |
| "loss": 0.0298, | |
| "step": 352 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.6952705502317575e-06, | |
| "loss": 0.0117, | |
| "step": 353 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.6934325007922417e-06, | |
| "loss": 0.0174, | |
| "step": 354 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.6915899276560234e-06, | |
| "loss": 0.0037, | |
| "step": 355 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.689742842843367e-06, | |
| "loss": 0.0192, | |
| "step": 356 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.6878912584039697e-06, | |
| "loss": 0.0225, | |
| "step": 357 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.6860351864168823e-06, | |
| "loss": 0.007, | |
| "step": 358 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.6841746389904303e-06, | |
| "loss": 0.0053, | |
| "step": 359 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.6823096282621363e-06, | |
| "loss": 0.013, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.6804401663986387e-06, | |
| "loss": 0.0248, | |
| "step": 361 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6785662655956145e-06, | |
| "loss": 0.0078, | |
| "step": 362 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.676687938077698e-06, | |
| "loss": 0.0189, | |
| "step": 363 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.674805196098402e-06, | |
| "loss": 0.0277, | |
| "step": 364 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6729180519400376e-06, | |
| "loss": 0.0098, | |
| "step": 365 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6710265179136338e-06, | |
| "loss": 0.0145, | |
| "step": 366 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.669130606358858e-06, | |
| "loss": 0.164, | |
| "step": 367 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6672303296439349e-06, | |
| "loss": 0.0054, | |
| "step": 368 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.665325700165565e-06, | |
| "loss": 0.0033, | |
| "step": 369 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6634167303488464e-06, | |
| "loss": 0.0065, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6615034326471898e-06, | |
| "loss": 0.0101, | |
| "step": 371 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.659585819542241e-06, | |
| "loss": 0.0208, | |
| "step": 372 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6576639035437973e-06, | |
| "loss": 0.0768, | |
| "step": 373 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6557376971897266e-06, | |
| "loss": 0.0121, | |
| "step": 374 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.653807213045885e-06, | |
| "loss": 0.0155, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6518724637060367e-06, | |
| "loss": 0.0194, | |
| "step": 376 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6499334617917687e-06, | |
| "loss": 0.0126, | |
| "step": 377 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6479902199524114e-06, | |
| "loss": 0.0245, | |
| "step": 378 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6460427508649543e-06, | |
| "loss": 0.0199, | |
| "step": 379 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6440910672339644e-06, | |
| "loss": 0.0243, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6421351817915023e-06, | |
| "loss": 0.0101, | |
| "step": 381 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6401751072970407e-06, | |
| "loss": 0.0183, | |
| "step": 382 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6382108565373783e-06, | |
| "loss": 0.0061, | |
| "step": 383 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6362424423265597e-06, | |
| "loss": 0.0081, | |
| "step": 384 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.63426987750579e-06, | |
| "loss": 0.0211, | |
| "step": 385 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.632293174943351e-06, | |
| "loss": 0.0217, | |
| "step": 386 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.630312347534518e-06, | |
| "loss": 0.0156, | |
| "step": 387 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6283274082014749e-06, | |
| "loss": 0.0202, | |
| "step": 388 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6263383698932305e-06, | |
| "loss": 0.0285, | |
| "step": 389 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.624345245585534e-06, | |
| "loss": 0.0078, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6223480482807892e-06, | |
| "loss": 0.0442, | |
| "step": 391 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.620346791007972e-06, | |
| "loss": 0.0073, | |
| "step": 392 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.618341486822543e-06, | |
| "loss": 0.0128, | |
| "step": 393 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6163321488063635e-06, | |
| "loss": 0.0117, | |
| "step": 394 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.614318790067611e-06, | |
| "loss": 0.0026, | |
| "step": 395 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6123014237406911e-06, | |
| "loss": 0.0684, | |
| "step": 396 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6102800629861545e-06, | |
| "loss": 0.0252, | |
| "step": 397 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6082547209906099e-06, | |
| "loss": 0.0156, | |
| "step": 398 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.606225410966638e-06, | |
| "loss": 0.0231, | |
| "step": 399 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.6041921461527053e-06, | |
| "loss": 0.0241, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.6021549398130786e-06, | |
| "loss": 0.0016, | |
| "step": 401 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.6001138052377367e-06, | |
| "loss": 0.0245, | |
| "step": 402 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.598068755742285e-06, | |
| "loss": 0.0144, | |
| "step": 403 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.5960198046678697e-06, | |
| "loss": 0.0116, | |
| "step": 404 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.593966965381088e-06, | |
| "loss": 0.014, | |
| "step": 405 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.591910251273902e-06, | |
| "loss": 0.0026, | |
| "step": 406 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.5898496757635535e-06, | |
| "loss": 0.0026, | |
| "step": 407 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.587785252292473e-06, | |
| "loss": 0.0047, | |
| "step": 408 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.5857169943281947e-06, | |
| "loss": 0.008, | |
| "step": 409 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.583644915363267e-06, | |
| "loss": 0.0453, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.5815690289151656e-06, | |
| "loss": 0.0588, | |
| "step": 411 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.5794893485262044e-06, | |
| "loss": 0.0035, | |
| "step": 412 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.577405887763447e-06, | |
| "loss": 0.0077, | |
| "step": 413 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5753186602186206e-06, | |
| "loss": 0.031, | |
| "step": 414 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5732276795080236e-06, | |
| "loss": 0.01, | |
| "step": 415 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5711329592724396e-06, | |
| "loss": 0.0242, | |
| "step": 416 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5690345131770473e-06, | |
| "loss": 0.0115, | |
| "step": 417 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5669323549113313e-06, | |
| "loss": 0.0036, | |
| "step": 418 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5648264981889934e-06, | |
| "loss": 0.0032, | |
| "step": 419 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5627169567478627e-06, | |
| "loss": 0.0098, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5606037443498057e-06, | |
| "loss": 0.123, | |
| "step": 421 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5584868747806372e-06, | |
| "loss": 0.0278, | |
| "step": 422 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.55636636185003e-06, | |
| "loss": 0.0055, | |
| "step": 423 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5542422193914248e-06, | |
| "loss": 0.0088, | |
| "step": 424 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5521144612619394e-06, | |
| "loss": 0.0047, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.54998310134228e-06, | |
| "loss": 0.0024, | |
| "step": 426 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5478481535366493e-06, | |
| "loss": 0.0025, | |
| "step": 427 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.545709631772655e-06, | |
| "loss": 0.0215, | |
| "step": 428 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5435675500012212e-06, | |
| "loss": 0.0261, | |
| "step": 429 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5414219221964952e-06, | |
| "loss": 0.0132, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5392727623557584e-06, | |
| "loss": 0.0112, | |
| "step": 431 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.537120084499333e-06, | |
| "loss": 0.0031, | |
| "step": 432 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5349639026704914e-06, | |
| "loss": 0.0099, | |
| "step": 433 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5328042309353654e-06, | |
| "loss": 0.012, | |
| "step": 434 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5306410833828535e-06, | |
| "loss": 0.0027, | |
| "step": 435 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.528474474124528e-06, | |
| "loss": 0.0278, | |
| "step": 436 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5263044172945455e-06, | |
| "loss": 0.0194, | |
| "step": 437 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.524130927049552e-06, | |
| "loss": 0.014, | |
| "step": 438 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5219540175685937e-06, | |
| "loss": 0.017, | |
| "step": 439 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5197737030530203e-06, | |
| "loss": 0.0014, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.517589997726396e-06, | |
| "loss": 0.0184, | |
| "step": 441 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5154029158344065e-06, | |
| "loss": 0.0053, | |
| "step": 442 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5132124716447625e-06, | |
| "loss": 0.0056, | |
| "step": 443 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5110186794471103e-06, | |
| "loss": 0.0113, | |
| "step": 444 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5088215535529378e-06, | |
| "loss": 0.0134, | |
| "step": 445 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.50662110829548e-06, | |
| "loss": 0.0391, | |
| "step": 446 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5044173580296265e-06, | |
| "loss": 0.0018, | |
| "step": 447 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5022103171318274e-06, | |
| "loss": 0.0351, | |
| "step": 448 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5e-06, | |
| "loss": 0.02, | |
| "step": 449 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.4977864210534339e-06, | |
| "loss": 0.0468, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4955695947326985e-06, | |
| "loss": 0.0079, | |
| "step": 451 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4933495354995472e-06, | |
| "loss": 0.0162, | |
| "step": 452 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.491126257836823e-06, | |
| "loss": 0.0137, | |
| "step": 453 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4888997762483669e-06, | |
| "loss": 0.0002, | |
| "step": 454 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4866701052589193e-06, | |
| "loss": 0.0026, | |
| "step": 455 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4844372594140268e-06, | |
| "loss": 0.0236, | |
| "step": 456 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4822012532799496e-06, | |
| "loss": 0.003, | |
| "step": 457 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4799621014435626e-06, | |
| "loss": 0.0037, | |
| "step": 458 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4777198185122628e-06, | |
| "loss": 0.0024, | |
| "step": 459 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4754744191138734e-06, | |
| "loss": 0.0029, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4732259178965478e-06, | |
| "loss": 0.0195, | |
| "step": 461 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.470974329528675e-06, | |
| "loss": 0.0293, | |
| "step": 462 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4687196686987831e-06, | |
| "loss": 0.0593, | |
| "step": 463 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4664619501154443e-06, | |
| "loss": 0.0198, | |
| "step": 464 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4642011885071779e-06, | |
| "loss": 0.0061, | |
| "step": 465 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4619373986223546e-06, | |
| "loss": 0.0072, | |
| "step": 466 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4596705952291016e-06, | |
| "loss": 0.0364, | |
| "step": 467 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4574007931152035e-06, | |
| "loss": 0.0148, | |
| "step": 468 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4551280070880087e-06, | |
| "loss": 0.0055, | |
| "step": 469 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4528522519743314e-06, | |
| "loss": 0.0197, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4505735426203543e-06, | |
| "loss": 0.0047, | |
| "step": 471 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.448291893891533e-06, | |
| "loss": 0.0117, | |
| "step": 472 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4460073206724983e-06, | |
| "loss": 0.0662, | |
| "step": 473 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4437198378669597e-06, | |
| "loss": 0.0098, | |
| "step": 474 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4414294603976074e-06, | |
| "loss": 0.008, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.439136203206015e-06, | |
| "loss": 0.0054, | |
| "step": 476 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4368400812525433e-06, | |
| "loss": 0.0121, | |
| "step": 477 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4345411095162405e-06, | |
| "loss": 0.0029, | |
| "step": 478 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4322393029947469e-06, | |
| "loss": 0.0131, | |
| "step": 479 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4299346767041953e-06, | |
| "loss": 0.0386, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4276272456791135e-06, | |
| "loss": 0.0162, | |
| "step": 481 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4253170249723268e-06, | |
| "loss": 0.0112, | |
| "step": 482 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4230040296548586e-06, | |
| "loss": 0.014, | |
| "step": 483 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.420688274815834e-06, | |
| "loss": 0.0007, | |
| "step": 484 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4183697755623793e-06, | |
| "loss": 0.0337, | |
| "step": 485 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4160485470195242e-06, | |
| "loss": 0.0035, | |
| "step": 486 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.413724604330104e-06, | |
| "loss": 0.0074, | |
| "step": 487 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4113979626546599e-06, | |
| "loss": 0.0193, | |
| "step": 488 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.40906863717134e-06, | |
| "loss": 0.0013, | |
| "step": 489 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.4067366430758004e-06, | |
| "loss": 0.0046, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.4044019955811063e-06, | |
| "loss": 0.008, | |
| "step": 491 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.4020647099176337e-06, | |
| "loss": 0.0018, | |
| "step": 492 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.3997248013329678e-06, | |
| "loss": 0.0313, | |
| "step": 493 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.3973822850918054e-06, | |
| "loss": 0.0032, | |
| "step": 494 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.3950371764758542e-06, | |
| "loss": 0.001, | |
| "step": 495 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.392689490783734e-06, | |
| "loss": 0.0239, | |
| "step": 496 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.3903392433308763e-06, | |
| "loss": 0.0297, | |
| "step": 497 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.387986449449425e-06, | |
| "loss": 0.0376, | |
| "step": 498 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.3856311244881357e-06, | |
| "loss": 0.0117, | |
| "step": 499 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.383273283812275e-06, | |
| "loss": 0.0211, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.3809129428035227e-06, | |
| "loss": 0.0341, | |
| "step": 501 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3785501168598686e-06, | |
| "loss": 0.0665, | |
| "step": 502 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3761848213955133e-06, | |
| "loss": 0.0196, | |
| "step": 503 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3738170718407686e-06, | |
| "loss": 0.0028, | |
| "step": 504 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3714468836419553e-06, | |
| "loss": 0.0135, | |
| "step": 505 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3690742722613025e-06, | |
| "loss": 0.0158, | |
| "step": 506 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.366699253176848e-06, | |
| "loss": 0.0595, | |
| "step": 507 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3643218418823366e-06, | |
| "loss": 0.0362, | |
| "step": 508 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3619420538871177e-06, | |
| "loss": 0.0473, | |
| "step": 509 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3595599047160477e-06, | |
| "loss": 0.0253, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3571754099093847e-06, | |
| "loss": 0.0054, | |
| "step": 511 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3547885850226888e-06, | |
| "loss": 0.0014, | |
| "step": 512 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3523994456267218e-06, | |
| "loss": 0.0421, | |
| "step": 513 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3500080073073435e-06, | |
| "loss": 0.0068, | |
| "step": 514 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3476142856654118e-06, | |
| "loss": 0.0062, | |
| "step": 515 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.345218296316679e-06, | |
| "loss": 0.0335, | |
| "step": 516 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.342820054891693e-06, | |
| "loss": 0.015, | |
| "step": 517 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3404195770356917e-06, | |
| "loss": 0.0092, | |
| "step": 518 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3380168784085026e-06, | |
| "loss": 0.0229, | |
| "step": 519 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.335611974684442e-06, | |
| "loss": 0.0197, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.333204881552211e-06, | |
| "loss": 0.0446, | |
| "step": 521 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3307956147147923e-06, | |
| "loss": 0.0164, | |
| "step": 522 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.328384189889351e-06, | |
| "loss": 0.0195, | |
| "step": 523 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3259706228071284e-06, | |
| "loss": 0.0103, | |
| "step": 524 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3235549292133424e-06, | |
| "loss": 0.0033, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3211371248670828e-06, | |
| "loss": 0.0207, | |
| "step": 526 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3187172255412095e-06, | |
| "loss": 0.0084, | |
| "step": 527 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3162952470222486e-06, | |
| "loss": 0.0054, | |
| "step": 528 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3138712051102908e-06, | |
| "loss": 0.0096, | |
| "step": 529 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3114451156188875e-06, | |
| "loss": 0.0045, | |
| "step": 530 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3090169943749473e-06, | |
| "loss": 0.0073, | |
| "step": 531 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.306586857218634e-06, | |
| "loss": 0.004, | |
| "step": 532 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3041547200032617e-06, | |
| "loss": 0.0143, | |
| "step": 533 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3017205985951924e-06, | |
| "loss": 0.0037, | |
| "step": 534 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.2992845088737322e-06, | |
| "loss": 0.0515, | |
| "step": 535 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.296846466731028e-06, | |
| "loss": 0.0052, | |
| "step": 536 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.2944064880719632e-06, | |
| "loss": 0.025, | |
| "step": 537 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.291964588814055e-06, | |
| "loss": 0.032, | |
| "step": 538 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.2895207848873487e-06, | |
| "loss": 0.0585, | |
| "step": 539 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2870750922343158e-06, | |
| "loss": 0.0151, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2846275268097494e-06, | |
| "loss": 0.0088, | |
| "step": 541 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.282178104580659e-06, | |
| "loss": 0.0021, | |
| "step": 542 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.279726841526168e-06, | |
| "loss": 0.0067, | |
| "step": 543 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2772737536374078e-06, | |
| "loss": 0.0004, | |
| "step": 544 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2748188569174155e-06, | |
| "loss": 0.0026, | |
| "step": 545 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2723621673810274e-06, | |
| "loss": 0.0358, | |
| "step": 546 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2699037010547758e-06, | |
| "loss": 0.035, | |
| "step": 547 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.267443473976784e-06, | |
| "loss": 0.0104, | |
| "step": 548 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.264981502196662e-06, | |
| "loss": 0.0225, | |
| "step": 549 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.262517801775401e-06, | |
| "loss": 0.0135, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2600523887852706e-06, | |
| "loss": 0.0166, | |
| "step": 551 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2575852793097111e-06, | |
| "loss": 0.0009, | |
| "step": 552 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2551164894432304e-06, | |
| "loss": 0.0034, | |
| "step": 553 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2526460352912994e-06, | |
| "loss": 0.004, | |
| "step": 554 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2501739329702451e-06, | |
| "loss": 0.0091, | |
| "step": 555 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2477001986071475e-06, | |
| "loss": 0.0169, | |
| "step": 556 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.245224848339733e-06, | |
| "loss": 0.0047, | |
| "step": 557 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2427478983162693e-06, | |
| "loss": 0.0162, | |
| "step": 558 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2402693646954605e-06, | |
| "loss": 0.0713, | |
| "step": 559 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.237789263646342e-06, | |
| "loss": 0.0528, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.235307611348174e-06, | |
| "loss": 0.0033, | |
| "step": 561 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.232824423990336e-06, | |
| "loss": 0.0013, | |
| "step": 562 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2303397177722233e-06, | |
| "loss": 0.0013, | |
| "step": 563 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2278535089031377e-06, | |
| "loss": 0.0149, | |
| "step": 564 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.225365813602185e-06, | |
| "loss": 0.009, | |
| "step": 565 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2228766480981677e-06, | |
| "loss": 0.0029, | |
| "step": 566 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.220386028629479e-06, | |
| "loss": 0.0064, | |
| "step": 567 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2178939714439978e-06, | |
| "loss": 0.0084, | |
| "step": 568 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2154004927989813e-06, | |
| "loss": 0.0075, | |
| "step": 569 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2129056089609609e-06, | |
| "loss": 0.0365, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.210409336205634e-06, | |
| "loss": 0.0213, | |
| "step": 571 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.207911690817759e-06, | |
| "loss": 0.028, | |
| "step": 572 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2054126890910496e-06, | |
| "loss": 0.0038, | |
| "step": 573 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2029123473280666e-06, | |
| "loss": 0.0267, | |
| "step": 574 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2004106818401134e-06, | |
| "loss": 0.0216, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.1979077089471285e-06, | |
| "loss": 0.0008, | |
| "step": 576 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.19540344497758e-06, | |
| "loss": 0.0373, | |
| "step": 577 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1928979062683577e-06, | |
| "loss": 0.0076, | |
| "step": 578 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1903911091646684e-06, | |
| "loss": 0.0195, | |
| "step": 579 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.187883070019927e-06, | |
| "loss": 0.0081, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1853738051956518e-06, | |
| "loss": 0.0452, | |
| "step": 581 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1828633310613567e-06, | |
| "loss": 0.011, | |
| "step": 582 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1803516639944451e-06, | |
| "loss": 0.0082, | |
| "step": 583 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1778388203801018e-06, | |
| "loss": 0.0164, | |
| "step": 584 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1753248166111879e-06, | |
| "loss": 0.0086, | |
| "step": 585 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1728096690881321e-06, | |
| "loss": 0.0068, | |
| "step": 586 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.170293394218825e-06, | |
| "loss": 0.0097, | |
| "step": 587 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1677760084185123e-06, | |
| "loss": 0.0105, | |
| "step": 588 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1652575281096848e-06, | |
| "loss": 0.0232, | |
| "step": 589 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.162737969721976e-06, | |
| "loss": 0.0213, | |
| "step": 590 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.160217349692051e-06, | |
| "loss": 0.0091, | |
| "step": 591 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1576956844635008e-06, | |
| "loss": 0.0151, | |
| "step": 592 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.155172990486735e-06, | |
| "loss": 0.0074, | |
| "step": 593 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1526492842188744e-06, | |
| "loss": 0.0304, | |
| "step": 594 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1501245821236438e-06, | |
| "loss": 0.0083, | |
| "step": 595 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1475989006712641e-06, | |
| "loss": 0.0375, | |
| "step": 596 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1450722563383448e-06, | |
| "loss": 0.0102, | |
| "step": 597 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1425446656077781e-06, | |
| "loss": 0.0107, | |
| "step": 598 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1400161449686292e-06, | |
| "loss": 0.0199, | |
| "step": 599 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1374867109160293e-06, | |
| "loss": 0.0223, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1349563799510698e-06, | |
| "loss": 0.0214, | |
| "step": 601 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1324251685806919e-06, | |
| "loss": 0.0013, | |
| "step": 602 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1298930933175804e-06, | |
| "loss": 0.0096, | |
| "step": 603 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1273601706800569e-06, | |
| "loss": 0.014, | |
| "step": 604 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1248264171919694e-06, | |
| "loss": 0.003, | |
| "step": 605 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1222918493825874e-06, | |
| "loss": 0.01, | |
| "step": 606 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1197564837864922e-06, | |
| "loss": 0.0006, | |
| "step": 607 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1172203369434689e-06, | |
| "loss": 0.0255, | |
| "step": 608 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1146834253984005e-06, | |
| "loss": 0.0093, | |
| "step": 609 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1121457657011576e-06, | |
| "loss": 0.0107, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1096073744064917e-06, | |
| "loss": 0.0036, | |
| "step": 611 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1070682680739274e-06, | |
| "loss": 0.0117, | |
| "step": 612 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1045284632676535e-06, | |
| "loss": 0.0094, | |
| "step": 613 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1019879765564153e-06, | |
| "loss": 0.0065, | |
| "step": 614 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.099446824513407e-06, | |
| "loss": 0.0122, | |
| "step": 615 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.096905023716163e-06, | |
| "loss": 0.0036, | |
| "step": 616 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0943625907464498e-06, | |
| "loss": 0.0546, | |
| "step": 617 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.091819542190158e-06, | |
| "loss": 0.0063, | |
| "step": 618 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0892758946371942e-06, | |
| "loss": 0.0017, | |
| "step": 619 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.086731664681373e-06, | |
| "loss": 0.0057, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0841868689203072e-06, | |
| "loss": 0.0177, | |
| "step": 621 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0816415239553014e-06, | |
| "loss": 0.0443, | |
| "step": 622 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0790956463912437e-06, | |
| "loss": 0.0341, | |
| "step": 623 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.076549252836496e-06, | |
| "loss": 0.0049, | |
| "step": 624 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0740023599027855e-06, | |
| "loss": 0.0092, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0714549842050987e-06, | |
| "loss": 0.0023, | |
| "step": 626 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0689071423615707e-06, | |
| "loss": 0.0402, | |
| "step": 627 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0663588509933777e-06, | |
| "loss": 0.032, | |
| "step": 628 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0638101267246282e-06, | |
| "loss": 0.0604, | |
| "step": 629 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0612609861822555e-06, | |
| "loss": 0.0036, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.058711445995907e-06, | |
| "loss": 0.0054, | |
| "step": 631 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.056161522797839e-06, | |
| "loss": 0.0334, | |
| "step": 632 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0536112332228056e-06, | |
| "loss": 0.0209, | |
| "step": 633 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0510605939079504e-06, | |
| "loss": 0.0115, | |
| "step": 634 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0485096214927e-06, | |
| "loss": 0.0817, | |
| "step": 635 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0459583326186532e-06, | |
| "loss": 0.0116, | |
| "step": 636 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0434067439294724e-06, | |
| "loss": 0.0134, | |
| "step": 637 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0408548720707777e-06, | |
| "loss": 0.0154, | |
| "step": 638 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0383027336900353e-06, | |
| "loss": 0.0349, | |
| "step": 639 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0357503454364503e-06, | |
| "loss": 0.0086, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0331977239608583e-06, | |
| "loss": 0.007, | |
| "step": 641 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0306448859156153e-06, | |
| "loss": 0.0188, | |
| "step": 642 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0280918479544913e-06, | |
| "loss": 0.0251, | |
| "step": 643 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0255386267325602e-06, | |
| "loss": 0.0093, | |
| "step": 644 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0229852389060904e-06, | |
| "loss": 0.0234, | |
| "step": 645 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.020431701132439e-06, | |
| "loss": 0.0314, | |
| "step": 646 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0178780300699395e-06, | |
| "loss": 0.015, | |
| "step": 647 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0153242423777963e-06, | |
| "loss": 0.0076, | |
| "step": 648 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0127703547159738e-06, | |
| "loss": 0.0131, | |
| "step": 649 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0102163837450888e-06, | |
| "loss": 0.0319, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0076623461263016e-06, | |
| "loss": 0.0252, | |
| "step": 651 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0051082585212073e-06, | |
| "loss": 0.0241, | |
| "step": 652 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0025541375917272e-06, | |
| "loss": 0.0018, | |
| "step": 653 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1e-06, | |
| "loss": 0.0117, | |
| "step": 654 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.974458624082725e-07, | |
| "loss": 0.0113, | |
| "step": 655 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.948917414787928e-07, | |
| "loss": 0.0402, | |
| "step": 656 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.923376538736983e-07, | |
| "loss": 0.0098, | |
| "step": 657 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.89783616254911e-07, | |
| "loss": 0.0119, | |
| "step": 658 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.872296452840264e-07, | |
| "loss": 0.0069, | |
| "step": 659 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.846757576222036e-07, | |
| "loss": 0.0054, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.821219699300604e-07, | |
| "loss": 0.0399, | |
| "step": 661 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.79568298867561e-07, | |
| "loss": 0.0212, | |
| "step": 662 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.770147610939097e-07, | |
| "loss": 0.0277, | |
| "step": 663 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.7446137326744e-07, | |
| "loss": 0.0447, | |
| "step": 664 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.719081520455088e-07, | |
| "loss": 0.0169, | |
| "step": 665 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.693551140843846e-07, | |
| "loss": 0.0279, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.668022760391416e-07, | |
| "loss": 0.0121, | |
| "step": 667 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.642496545635496e-07, | |
| "loss": 0.0011, | |
| "step": 668 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.616972663099646e-07, | |
| "loss": 0.0041, | |
| "step": 669 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.59145127929222e-07, | |
| "loss": 0.011, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.565932560705277e-07, | |
| "loss": 0.0136, | |
| "step": 671 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.540416673813469e-07, | |
| "loss": 0.0697, | |
| "step": 672 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.514903785072996e-07, | |
| "loss": 0.0127, | |
| "step": 673 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.489394060920494e-07, | |
| "loss": 0.006, | |
| "step": 674 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.463887667771945e-07, | |
| "loss": 0.0081, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.438384772021607e-07, | |
| "loss": 0.0059, | |
| "step": 676 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.41288554004093e-07, | |
| "loss": 0.012, | |
| "step": 677 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.387390138177446e-07, | |
| "loss": 0.0131, | |
| "step": 678 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.361898732753715e-07, | |
| "loss": 0.001, | |
| "step": 679 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.336411490066223e-07, | |
| "loss": 0.0399, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.310928576384293e-07, | |
| "loss": 0.0109, | |
| "step": 681 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.285450157949014e-07, | |
| "loss": 0.0063, | |
| "step": 682 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.259976400972146e-07, | |
| "loss": 0.0049, | |
| "step": 683 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.234507471635042e-07, | |
| "loss": 0.0187, | |
| "step": 684 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.209043536087565e-07, | |
| "loss": 0.0027, | |
| "step": 685 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.183584760446987e-07, | |
| "loss": 0.0035, | |
| "step": 686 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.15813131079693e-07, | |
| "loss": 0.0057, | |
| "step": 687 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.132683353186275e-07, | |
| "loss": 0.0089, | |
| "step": 688 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.107241053628057e-07, | |
| "loss": 0.0045, | |
| "step": 689 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.081804578098418e-07, | |
| "loss": 0.0113, | |
| "step": 690 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.056374092535503e-07, | |
| "loss": 0.0084, | |
| "step": 691 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 9.03094976283837e-07, | |
| "loss": 0.0046, | |
| "step": 692 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 9.005531754865928e-07, | |
| "loss": 0.0171, | |
| "step": 693 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.980120234435848e-07, | |
| "loss": 0.019, | |
| "step": 694 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.954715367323466e-07, | |
| "loss": 0.0004, | |
| "step": 695 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.929317319260725e-07, | |
| "loss": 0.0021, | |
| "step": 696 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.903926255935085e-07, | |
| "loss": 0.0119, | |
| "step": 697 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.878542342988427e-07, | |
| "loss": 0.0062, | |
| "step": 698 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.853165746015995e-07, | |
| "loss": 0.0217, | |
| "step": 699 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.827796630565312e-07, | |
| "loss": 0.0123, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.802435162135079e-07, | |
| "loss": 0.0152, | |
| "step": 701 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.777081506174126e-07, | |
| "loss": 0.0031, | |
| "step": 702 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.751735828080307e-07, | |
| "loss": 0.0031, | |
| "step": 703 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.726398293199433e-07, | |
| "loss": 0.0024, | |
| "step": 704 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.701069066824194e-07, | |
| "loss": 0.0216, | |
| "step": 705 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.675748314193084e-07, | |
| "loss": 0.0079, | |
| "step": 706 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.650436200489303e-07, | |
| "loss": 0.0198, | |
| "step": 707 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.625132890839704e-07, | |
| "loss": 0.0044, | |
| "step": 708 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.599838550313711e-07, | |
| "loss": 0.0003, | |
| "step": 709 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.574553343922219e-07, | |
| "loss": 0.0487, | |
| "step": 710 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.54927743661655e-07, | |
| "loss": 0.0051, | |
| "step": 711 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.524010993287363e-07, | |
| "loss": 0.032, | |
| "step": 712 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.498754178763563e-07, | |
| "loss": 0.0017, | |
| "step": 713 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.473507157811254e-07, | |
| "loss": 0.0352, | |
| "step": 714 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.448270095132651e-07, | |
| "loss": 0.0032, | |
| "step": 715 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.423043155364993e-07, | |
| "loss": 0.0745, | |
| "step": 716 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.397826503079488e-07, | |
| "loss": 0.0065, | |
| "step": 717 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.372620302780239e-07, | |
| "loss": 0.0182, | |
| "step": 718 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.347424718903151e-07, | |
| "loss": 0.0021, | |
| "step": 719 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.32223991581488e-07, | |
| "loss": 0.0205, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.297066057811749e-07, | |
| "loss": 0.0089, | |
| "step": 721 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.27190330911868e-07, | |
| "loss": 0.0244, | |
| "step": 722 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.246751833888122e-07, | |
| "loss": 0.0229, | |
| "step": 723 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.221611796198983e-07, | |
| "loss": 0.0338, | |
| "step": 724 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.19648336005555e-07, | |
| "loss": 0.047, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.171366689386431e-07, | |
| "loss": 0.0125, | |
| "step": 726 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.146261948043484e-07, | |
| "loss": 0.0476, | |
| "step": 727 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.121169299800729e-07, | |
| "loss": 0.0118, | |
| "step": 728 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.096088908353315e-07, | |
| "loss": 0.042, | |
| "step": 729 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 8.071020937316422e-07, | |
| "loss": 0.034, | |
| "step": 730 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 8.0459655502242e-07, | |
| "loss": 0.0144, | |
| "step": 731 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 8.020922910528716e-07, | |
| "loss": 0.0332, | |
| "step": 732 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.995893181598866e-07, | |
| "loss": 0.0216, | |
| "step": 733 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.970876526719332e-07, | |
| "loss": 0.0134, | |
| "step": 734 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.945873109089502e-07, | |
| "loss": 0.002, | |
| "step": 735 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.920883091822408e-07, | |
| "loss": 0.0096, | |
| "step": 736 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.89590663794366e-07, | |
| "loss": 0.0256, | |
| "step": 737 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.870943910390391e-07, | |
| "loss": 0.0168, | |
| "step": 738 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.845995072010187e-07, | |
| "loss": 0.0161, | |
| "step": 739 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.821060285560023e-07, | |
| "loss": 0.0248, | |
| "step": 740 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.796139713705213e-07, | |
| "loss": 0.0099, | |
| "step": 741 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.771233519018325e-07, | |
| "loss": 0.0127, | |
| "step": 742 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.746341863978149e-07, | |
| "loss": 0.0261, | |
| "step": 743 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.721464910968626e-07, | |
| "loss": 0.0055, | |
| "step": 744 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.696602822277769e-07, | |
| "loss": 0.0135, | |
| "step": 745 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.671755760096638e-07, | |
| "loss": 0.0045, | |
| "step": 746 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.646923886518262e-07, | |
| "loss": 0.0089, | |
| "step": 747 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.622107363536579e-07, | |
| "loss": 0.0128, | |
| "step": 748 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.597306353045392e-07, | |
| "loss": 0.0133, | |
| "step": 749 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.572521016837309e-07, | |
| "loss": 0.032, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.54775151660267e-07, | |
| "loss": 0.0237, | |
| "step": 751 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.522998013928522e-07, | |
| "loss": 0.0179, | |
| "step": 752 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.498260670297549e-07, | |
| "loss": 0.0225, | |
| "step": 753 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.473539647087006e-07, | |
| "loss": 0.0047, | |
| "step": 754 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.448835105567694e-07, | |
| "loss": 0.0161, | |
| "step": 755 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.424147206902891e-07, | |
| "loss": 0.0075, | |
| "step": 756 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.399476112147293e-07, | |
| "loss": 0.0048, | |
| "step": 757 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.37482198224599e-07, | |
| "loss": 0.0132, | |
| "step": 758 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.350184978033385e-07, | |
| "loss": 0.0427, | |
| "step": 759 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.325565260232163e-07, | |
| "loss": 0.0277, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.300962989452242e-07, | |
| "loss": 0.0107, | |
| "step": 761 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.276378326189729e-07, | |
| "loss": 0.0182, | |
| "step": 762 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.251811430825845e-07, | |
| "loss": 0.0077, | |
| "step": 763 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.227262463625919e-07, | |
| "loss": 0.0494, | |
| "step": 764 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.202731584738322e-07, | |
| "loss": 0.0257, | |
| "step": 765 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.178218954193409e-07, | |
| "loss": 0.0091, | |
| "step": 766 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.153724731902505e-07, | |
| "loss": 0.0017, | |
| "step": 767 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.129249077656842e-07, | |
| "loss": 0.0076, | |
| "step": 768 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.104792151126514e-07, | |
| "loss": 0.0107, | |
| "step": 769 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.08035411185945e-07, | |
| "loss": 0.0013, | |
| "step": 770 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.055935119280369e-07, | |
| "loss": 0.0076, | |
| "step": 771 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.031535332689721e-07, | |
| "loss": 0.0157, | |
| "step": 772 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.007154911262677e-07, | |
| "loss": 0.0178, | |
| "step": 773 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 6.982794014048077e-07, | |
| "loss": 0.0245, | |
| "step": 774 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 6.958452799967385e-07, | |
| "loss": 0.037, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 6.934131427813659e-07, | |
| "loss": 0.0012, | |
| "step": 776 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 6.909830056250526e-07, | |
| "loss": 0.0059, | |
| "step": 777 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 6.885548843811124e-07, | |
| "loss": 0.0101, | |
| "step": 778 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 6.861287948897089e-07, | |
| "loss": 0.0094, | |
| "step": 779 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 6.837047529777515e-07, | |
| "loss": 0.0152, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.812827744587906e-07, | |
| "loss": 0.0122, | |
| "step": 781 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.788628751329169e-07, | |
| "loss": 0.0151, | |
| "step": 782 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.764450707866576e-07, | |
| "loss": 0.0309, | |
| "step": 783 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.740293771928716e-07, | |
| "loss": 0.0117, | |
| "step": 784 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.716158101106495e-07, | |
| "loss": 0.0217, | |
| "step": 785 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.692043852852079e-07, | |
| "loss": 0.01, | |
| "step": 786 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.667951184477891e-07, | |
| "loss": 0.0051, | |
| "step": 787 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.64388025315558e-07, | |
| "loss": 0.0244, | |
| "step": 788 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.619831215914973e-07, | |
| "loss": 0.0095, | |
| "step": 789 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.595804229643085e-07, | |
| "loss": 0.0243, | |
| "step": 790 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.571799451083068e-07, | |
| "loss": 0.0075, | |
| "step": 791 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.547817036833207e-07, | |
| "loss": 0.0023, | |
| "step": 792 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.523857143345883e-07, | |
| "loss": 0.0157, | |
| "step": 793 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.499919926926565e-07, | |
| "loss": 0.0073, | |
| "step": 794 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.476005543732782e-07, | |
| "loss": 0.004, | |
| "step": 795 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.452114149773109e-07, | |
| "loss": 0.013, | |
| "step": 796 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.428245900906155e-07, | |
| "loss": 0.0027, | |
| "step": 797 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.404400952839521e-07, | |
| "loss": 0.0399, | |
| "step": 798 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.380579461128819e-07, | |
| "loss": 0.0016, | |
| "step": 799 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.356781581176638e-07, | |
| "loss": 0.0089, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.33300746823152e-07, | |
| "loss": 0.0164, | |
| "step": 801 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.309257277386974e-07, | |
| "loss": 0.0114, | |
| "step": 802 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.285531163580448e-07, | |
| "loss": 0.0087, | |
| "step": 803 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.261829281592312e-07, | |
| "loss": 0.0034, | |
| "step": 804 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.238151786044865e-07, | |
| "loss": 0.0126, | |
| "step": 805 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.214498831401315e-07, | |
| "loss": 0.0374, | |
| "step": 806 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.190870571964772e-07, | |
| "loss": 0.0041, | |
| "step": 807 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.167267161877247e-07, | |
| "loss": 0.0004, | |
| "step": 808 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.143688755118645e-07, | |
| "loss": 0.0088, | |
| "step": 809 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.120135505505747e-07, | |
| "loss": 0.0102, | |
| "step": 810 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.096607566691234e-07, | |
| "loss": 0.0079, | |
| "step": 811 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.07310509216266e-07, | |
| "loss": 0.0131, | |
| "step": 812 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.049628235241458e-07, | |
| "loss": 0.0117, | |
| "step": 813 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.026177149081948e-07, | |
| "loss": 0.0037, | |
| "step": 814 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.002751986670322e-07, | |
| "loss": 0.0391, | |
| "step": 815 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 5.979352900823663e-07, | |
| "loss": 0.0054, | |
| "step": 816 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 5.955980044188934e-07, | |
| "loss": 0.002, | |
| "step": 817 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 5.932633569241999e-07, | |
| "loss": 0.0004, | |
| "step": 818 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.9093136282866e-07, | |
| "loss": 0.002, | |
| "step": 819 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.886020373453397e-07, | |
| "loss": 0.0034, | |
| "step": 820 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.862753956698958e-07, | |
| "loss": 0.0252, | |
| "step": 821 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.839514529804757e-07, | |
| "loss": 0.0084, | |
| "step": 822 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.816302244376205e-07, | |
| "loss": 0.0002, | |
| "step": 823 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.793117251841658e-07, | |
| "loss": 0.0089, | |
| "step": 824 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.769959703451413e-07, | |
| "loss": 0.0239, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.746829750276729e-07, | |
| "loss": 0.0141, | |
| "step": 826 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.723727543208866e-07, | |
| "loss": 0.0069, | |
| "step": 827 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.700653232958046e-07, | |
| "loss": 0.0126, | |
| "step": 828 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.677606970052529e-07, | |
| "loss": 0.0006, | |
| "step": 829 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.654588904837593e-07, | |
| "loss": 0.0088, | |
| "step": 830 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.63159918747457e-07, | |
| "loss": 0.0432, | |
| "step": 831 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.608637967939846e-07, | |
| "loss": 0.0148, | |
| "step": 832 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.58570539602393e-07, | |
| "loss": 0.0131, | |
| "step": 833 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.562801621330402e-07, | |
| "loss": 0.0572, | |
| "step": 834 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.53992679327502e-07, | |
| "loss": 0.0112, | |
| "step": 835 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.517081061084671e-07, | |
| "loss": 0.0034, | |
| "step": 836 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.49426457379646e-07, | |
| "loss": 0.0038, | |
| "step": 837 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.471477480256689e-07, | |
| "loss": 0.0066, | |
| "step": 838 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.448719929119915e-07, | |
| "loss": 0.0001, | |
| "step": 839 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.425992068847964e-07, | |
| "loss": 0.0584, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.403294047708986e-07, | |
| "loss": 0.0059, | |
| "step": 841 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.380626013776455e-07, | |
| "loss": 0.006, | |
| "step": 842 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.35798811492822e-07, | |
| "loss": 0.0036, | |
| "step": 843 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.335380498845559e-07, | |
| "loss": 0.0022, | |
| "step": 844 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.312803313012167e-07, | |
| "loss": 0.0087, | |
| "step": 845 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.29025670471325e-07, | |
| "loss": 0.015, | |
| "step": 846 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.267740821034524e-07, | |
| "loss": 0.0011, | |
| "step": 847 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.245255808861268e-07, | |
| "loss": 0.0014, | |
| "step": 848 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.222801814877369e-07, | |
| "loss": 0.011, | |
| "step": 849 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.200378985564376e-07, | |
| "loss": 0.0095, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.177987467200503e-07, | |
| "loss": 0.0066, | |
| "step": 851 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.155627405859731e-07, | |
| "loss": 0.054, | |
| "step": 852 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.133298947410811e-07, | |
| "loss": 0.0298, | |
| "step": 853 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.111002237516334e-07, | |
| "loss": 0.0048, | |
| "step": 854 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.088737421631767e-07, | |
| "loss": 0.0441, | |
| "step": 855 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.066504645004534e-07, | |
| "loss": 0.0039, | |
| "step": 856 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 5.044304052673014e-07, | |
| "loss": 0.0208, | |
| "step": 857 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 5.02213578946566e-07, | |
| "loss": 0.0183, | |
| "step": 858 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 5.000000000000002e-07, | |
| "loss": 0.0069, | |
| "step": 859 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.977896828681723e-07, | |
| "loss": 0.0003, | |
| "step": 860 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.955826419703734e-07, | |
| "loss": 0.0053, | |
| "step": 861 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.9337889170452e-07, | |
| "loss": 0.0208, | |
| "step": 862 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.911784464470623e-07, | |
| "loss": 0.0169, | |
| "step": 863 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.889813205528894e-07, | |
| "loss": 0.0186, | |
| "step": 864 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.867875283552377e-07, | |
| "loss": 0.0168, | |
| "step": 865 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.845970841655933e-07, | |
| "loss": 0.0122, | |
| "step": 866 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.824100022736035e-07, | |
| "loss": 0.01, | |
| "step": 867 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.802262969469798e-07, | |
| "loss": 0.0004, | |
| "step": 868 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.780459824314066e-07, | |
| "loss": 0.0002, | |
| "step": 869 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.7586907295044774e-07, | |
| "loss": 0.0051, | |
| "step": 870 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.7369558270545496e-07, | |
| "loss": 0.0126, | |
| "step": 871 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.7152552587547203e-07, | |
| "loss": 0.0122, | |
| "step": 872 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.6935891661714656e-07, | |
| "loss": 0.003, | |
| "step": 873 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.6719576906463444e-07, | |
| "loss": 0.0016, | |
| "step": 874 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.6503609732950854e-07, | |
| "loss": 0.038, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.6287991550066686e-07, | |
| "loss": 0.0348, | |
| "step": 876 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.6072723764424146e-07, | |
| "loss": 0.0119, | |
| "step": 877 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.5857807780350465e-07, | |
| "loss": 0.0051, | |
| "step": 878 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.5643244999877896e-07, | |
| "loss": 0.001, | |
| "step": 879 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.542903682273451e-07, | |
| "loss": 0.0168, | |
| "step": 880 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.521518464633505e-07, | |
| "loss": 0.0283, | |
| "step": 881 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.5001689865771963e-07, | |
| "loss": 0.0036, | |
| "step": 882 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.4788553873806045e-07, | |
| "loss": 0.0085, | |
| "step": 883 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.4575778060857537e-07, | |
| "loss": 0.0028, | |
| "step": 884 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.4363363814997e-07, | |
| "loss": 0.0004, | |
| "step": 885 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.415131252193629e-07, | |
| "loss": 0.0045, | |
| "step": 886 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.3939625565019413e-07, | |
| "loss": 0.0135, | |
| "step": 887 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.372830432521376e-07, | |
| "loss": 0.011, | |
| "step": 888 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.3517350181100656e-07, | |
| "loss": 0.0085, | |
| "step": 889 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.3306764508866877e-07, | |
| "loss": 0.0076, | |
| "step": 890 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.3096548682295295e-07, | |
| "loss": 0.0253, | |
| "step": 891 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.2886704072756065e-07, | |
| "loss": 0.0073, | |
| "step": 892 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.267723204919763e-07, | |
| "loss": 0.0174, | |
| "step": 893 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.246813397813794e-07, | |
| "loss": 0.0259, | |
| "step": 894 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.225941122365528e-07, | |
| "loss": 0.0157, | |
| "step": 895 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.2051065147379596e-07, | |
| "loss": 0.0176, | |
| "step": 896 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.184309710848346e-07, | |
| "loss": 0.0039, | |
| "step": 897 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.1635508463673273e-07, | |
| "loss": 0.0117, | |
| "step": 898 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.142830056718052e-07, | |
| "loss": 0.0016, | |
| "step": 899 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.1221474770752696e-07, | |
| "loss": 0.0225, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.1015032423644667e-07, | |
| "loss": 0.0078, | |
| "step": 901 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.080897487260977e-07, | |
| "loss": 0.0008, | |
| "step": 902 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.0603303461891245e-07, | |
| "loss": 0.056, | |
| "step": 903 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.039801953321301e-07, | |
| "loss": 0.0027, | |
| "step": 904 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.0193124425771476e-07, | |
| "loss": 0.0001, | |
| "step": 905 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 3.998861947622635e-07, | |
| "loss": 0.0268, | |
| "step": 906 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 3.9784506018692165e-07, | |
| "loss": 0.0631, | |
| "step": 907 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.958078538472944e-07, | |
| "loss": 0.0128, | |
| "step": 908 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.9377458903336223e-07, | |
| "loss": 0.0107, | |
| "step": 909 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.9174527900939e-07, | |
| "loss": 0.046, | |
| "step": 910 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.8971993701384556e-07, | |
| "loss": 0.0143, | |
| "step": 911 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.87698576259309e-07, | |
| "loss": 0.0353, | |
| "step": 912 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.856812099323893e-07, | |
| "loss": 0.0332, | |
| "step": 913 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.836678511936362e-07, | |
| "loss": 0.0085, | |
| "step": 914 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.81658513177457e-07, | |
| "loss": 0.0114, | |
| "step": 915 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.7965320899202813e-07, | |
| "loss": 0.0291, | |
| "step": 916 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.7765195171921094e-07, | |
| "loss": 0.0042, | |
| "step": 917 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.7565475441446634e-07, | |
| "loss": 0.0139, | |
| "step": 918 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.736616301067693e-07, | |
| "loss": 0.0019, | |
| "step": 919 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.716725917985249e-07, | |
| "loss": 0.0061, | |
| "step": 920 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.6968765246548196e-07, | |
| "loss": 0.0022, | |
| "step": 921 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.6770682505664895e-07, | |
| "loss": 0.0234, | |
| "step": 922 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.6573012249420974e-07, | |
| "loss": 0.0173, | |
| "step": 923 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.6375755767344043e-07, | |
| "loss": 0.0027, | |
| "step": 924 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.617891434626216e-07, | |
| "loss": 0.0216, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.598248927029593e-07, | |
| "loss": 0.004, | |
| "step": 926 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.5786481820849743e-07, | |
| "loss": 0.0093, | |
| "step": 927 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.5590893276603564e-07, | |
| "loss": 0.0132, | |
| "step": 928 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.5395724913504543e-07, | |
| "loss": 0.0254, | |
| "step": 929 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.5200978004758884e-07, | |
| "loss": 0.0866, | |
| "step": 930 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.500665382082313e-07, | |
| "loss": 0.0045, | |
| "step": 931 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.481275362939634e-07, | |
| "loss": 0.0188, | |
| "step": 932 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.461927869541149e-07, | |
| "loss": 0.0123, | |
| "step": 933 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.442623028102737e-07, | |
| "loss": 0.0054, | |
| "step": 934 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.4233609645620253e-07, | |
| "loss": 0.0154, | |
| "step": 935 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.4041418045775893e-07, | |
| "loss": 0.0203, | |
| "step": 936 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.384965673528102e-07, | |
| "loss": 0.0049, | |
| "step": 937 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.365832696511537e-07, | |
| "loss": 0.0479, | |
| "step": 938 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.3467429983443476e-07, | |
| "loss": 0.0221, | |
| "step": 939 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.3276967035606497e-07, | |
| "loss": 0.0169, | |
| "step": 940 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.308693936411421e-07, | |
| "loss": 0.0035, | |
| "step": 941 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.289734820863661e-07, | |
| "loss": 0.0217, | |
| "step": 942 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.270819480599625e-07, | |
| "loss": 0.0374, | |
| "step": 943 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.2519480390159804e-07, | |
| "loss": 0.006, | |
| "step": 944 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.233120619223021e-07, | |
| "loss": 0.0062, | |
| "step": 945 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.2143373440438537e-07, | |
| "loss": 0.0141, | |
| "step": 946 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.1955983360136154e-07, | |
| "loss": 0.0195, | |
| "step": 947 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.1769037173786374e-07, | |
| "loss": 0.0192, | |
| "step": 948 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.1582536100956967e-07, | |
| "loss": 0.0155, | |
| "step": 949 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.139648135831179e-07, | |
| "loss": 0.0054, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.121087415960304e-07, | |
| "loss": 0.0336, | |
| "step": 951 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.102571571566328e-07, | |
| "loss": 0.0065, | |
| "step": 952 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.084100723439765e-07, | |
| "loss": 0.017, | |
| "step": 953 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.065674992077584e-07, | |
| "loss": 0.0088, | |
| "step": 954 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.047294497682422e-07, | |
| "loss": 0.0029, | |
| "step": 955 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.028959360161827e-07, | |
| "loss": 0.0049, | |
| "step": 956 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.0106696991274274e-07, | |
| "loss": 0.0053, | |
| "step": 957 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 2.9924256338942065e-07, | |
| "loss": 0.0067, | |
| "step": 958 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.974227283479681e-07, | |
| "loss": 0.0102, | |
| "step": 959 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.9560747666031474e-07, | |
| "loss": 0.0042, | |
| "step": 960 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.9379682016848973e-07, | |
| "loss": 0.0132, | |
| "step": 961 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.919907706845458e-07, | |
| "loss": 0.0049, | |
| "step": 962 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.901893399904797e-07, | |
| "loss": 0.0025, | |
| "step": 963 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.883925398381585e-07, | |
| "loss": 0.0117, | |
| "step": 964 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.8660038194924043e-07, | |
| "loss": 0.0143, | |
| "step": 965 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.8481287801509945e-07, | |
| "loss": 0.001, | |
| "step": 966 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.8303003969674865e-07, | |
| "loss": 0.0099, | |
| "step": 967 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.812518786247656e-07, | |
| "loss": 0.0137, | |
| "step": 968 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.7947840639921303e-07, | |
| "loss": 0.0048, | |
| "step": 969 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.7770963458956755e-07, | |
| "loss": 0.0122, | |
| "step": 970 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.7594557473464064e-07, | |
| "loss": 0.0408, | |
| "step": 971 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.74186238342505e-07, | |
| "loss": 0.0212, | |
| "step": 972 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.7243163689042006e-07, | |
| "loss": 0.0035, | |
| "step": 973 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.706817818247551e-07, | |
| "loss": 0.0238, | |
| "step": 974 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.6893668456091624e-07, | |
| "loss": 0.0213, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.6719635648327123e-07, | |
| "loss": 0.0175, | |
| "step": 976 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.654608089450764e-07, | |
| "loss": 0.0164, | |
| "step": 977 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.637300532683997e-07, | |
| "loss": 0.0101, | |
| "step": 978 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.620041007440508e-07, | |
| "loss": 0.0137, | |
| "step": 979 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.602829626315041e-07, | |
| "loss": 0.0478, | |
| "step": 980 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.585666501588268e-07, | |
| "loss": 0.0038, | |
| "step": 981 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.568551745226056e-07, | |
| "loss": 0.0433, | |
| "step": 982 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.5514854688787404e-07, | |
| "loss": 0.012, | |
| "step": 983 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.534467783880373e-07, | |
| "loss": 0.0156, | |
| "step": 984 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.517498801248035e-07, | |
| "loss": 0.0045, | |
| "step": 985 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.5005786316810775e-07, | |
| "loss": 0.0045, | |
| "step": 986 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.4837073855604184e-07, | |
| "loss": 0.0098, | |
| "step": 987 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.4668851729478156e-07, | |
| "loss": 0.0068, | |
| "step": 988 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.450112103585149e-07, | |
| "loss": 0.0055, | |
| "step": 989 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.433388286893714e-07, | |
| "loss": 0.0116, | |
| "step": 990 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.4167138319734905e-07, | |
| "loss": 0.0174, | |
| "step": 991 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.400088847602445e-07, | |
| "loss": 0.0097, | |
| "step": 992 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.3835134422358117e-07, | |
| "loss": 0.0165, | |
| "step": 993 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.3669877240054037e-07, | |
| "loss": 0.0034, | |
| "step": 994 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.3505118007188707e-07, | |
| "loss": 0.0052, | |
| "step": 995 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.3340857798590408e-07, | |
| "loss": 0.001, | |
| "step": 996 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.3177097685831848e-07, | |
| "loss": 0.0005, | |
| "step": 997 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.3013838737223323e-07, | |
| "loss": 0.0092, | |
| "step": 998 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.28510820178057e-07, | |
| "loss": 0.0142, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.268882858934359e-07, | |
| "loss": 0.0029, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.2527079510318137e-07, | |
| "loss": 0.0122, | |
| "step": 1001 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.2365835835920487e-07, | |
| "loss": 0.0046, | |
| "step": 1002 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.2205098618044582e-07, | |
| "loss": 0.0064, | |
| "step": 1003 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.2044868905280501e-07, | |
| "loss": 0.0002, | |
| "step": 1004 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.188514774290746e-07, | |
| "loss": 0.001, | |
| "step": 1005 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.1725936172887226e-07, | |
| "loss": 0.0239, | |
| "step": 1006 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.156723523385705e-07, | |
| "loss": 0.0309, | |
| "step": 1007 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.1409045961123063e-07, | |
| "loss": 0.0089, | |
| "step": 1008 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.125136938665345e-07, | |
| "loss": 0.0041, | |
| "step": 1009 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.1094206539071757e-07, | |
| "loss": 0.0129, | |
| "step": 1010 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.0937558443650205e-07, | |
| "loss": 0.0035, | |
| "step": 1011 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.0781426122302915e-07, | |
| "loss": 0.0115, | |
| "step": 1012 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.0625810593579285e-07, | |
| "loss": 0.0051, | |
| "step": 1013 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.0470712872657348e-07, | |
| "loss": 0.0034, | |
| "step": 1014 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.0316133971337235e-07, | |
| "loss": 0.0584, | |
| "step": 1015 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.0162074898034343e-07, | |
| "loss": 0.0052, | |
| "step": 1016 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.0008536657773045e-07, | |
| "loss": 0.0003, | |
| "step": 1017 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.9855520252179902e-07, | |
| "loss": 0.0323, | |
| "step": 1018 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.9703026679477252e-07, | |
| "loss": 0.0057, | |
| "step": 1019 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.955105693447665e-07, | |
| "loss": 0.0217, | |
| "step": 1020 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.9399612008572452e-07, | |
| "loss": 0.0018, | |
| "step": 1021 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.9248692889735163e-07, | |
| "loss": 0.0285, | |
| "step": 1022 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.9098300562505264e-07, | |
| "loss": 0.038, | |
| "step": 1023 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.8948436007986545e-07, | |
| "loss": 0.028, | |
| "step": 1024 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.8799100203839835e-07, | |
| "loss": 0.0149, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.8650294124276556e-07, | |
| "loss": 0.0059, | |
| "step": 1026 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.8502018740052494e-07, | |
| "loss": 0.0298, | |
| "step": 1027 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.8354275018461296e-07, | |
| "loss": 0.0041, | |
| "step": 1028 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.8207063923328235e-07, | |
| "loss": 0.0026, | |
| "step": 1029 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.8060386415003982e-07, | |
| "loss": 0.0099, | |
| "step": 1030 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.7914243450358212e-07, | |
| "loss": 0.0474, | |
| "step": 1031 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.7768635982773529e-07, | |
| "loss": 0.0142, | |
| "step": 1032 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.7623564962139058e-07, | |
| "loss": 0.0034, | |
| "step": 1033 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.747903133484442e-07, | |
| "loss": 0.0099, | |
| "step": 1034 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.7335036043773398e-07, | |
| "loss": 0.0141, | |
| "step": 1035 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.7191580028298003e-07, | |
| "loss": 0.0096, | |
| "step": 1036 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.7048664224272024e-07, | |
| "loss": 0.029, | |
| "step": 1037 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.6906289564025278e-07, | |
| "loss": 0.0393, | |
| "step": 1038 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.6764456976357277e-07, | |
| "loss": 0.0133, | |
| "step": 1039 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.6623167386531277e-07, | |
| "loss": 0.008, | |
| "step": 1040 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.6482421716268214e-07, | |
| "loss": 0.028, | |
| "step": 1041 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.6342220883740666e-07, | |
| "loss": 0.0028, | |
| "step": 1042 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.6202565803566914e-07, | |
| "loss": 0.0046, | |
| "step": 1043 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.6063457386805003e-07, | |
| "loss": 0.0134, | |
| "step": 1044 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.5924896540946674e-07, | |
| "loss": 0.004, | |
| "step": 1045 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.578688416991154e-07, | |
| "loss": 0.0013, | |
| "step": 1046 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.5649421174041188e-07, | |
| "loss": 0.0013, | |
| "step": 1047 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.5512508450093232e-07, | |
| "loss": 0.0205, | |
| "step": 1048 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.5376146891235597e-07, | |
| "loss": 0.0043, | |
| "step": 1049 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.5240337387040557e-07, | |
| "loss": 0.0025, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.510508082347899e-07, | |
| "loss": 0.0107, | |
| "step": 1051 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.497037808291458e-07, | |
| "loss": 0.001, | |
| "step": 1052 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.4836230044098162e-07, | |
| "loss": 0.0006, | |
| "step": 1053 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.4702637582161758e-07, | |
| "loss": 0.0058, | |
| "step": 1054 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.4569601568613143e-07, | |
| "loss": 0.0096, | |
| "step": 1055 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.4437122871329954e-07, | |
| "loss": 0.0067, | |
| "step": 1056 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.4305202354554134e-07, | |
| "loss": 0.0011, | |
| "step": 1057 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.4173840878886234e-07, | |
| "loss": 0.0178, | |
| "step": 1058 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.4043039301279903e-07, | |
| "loss": 0.0375, | |
| "step": 1059 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.391279847503609e-07, | |
| "loss": 0.0014, | |
| "step": 1060 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.378311924979776e-07, | |
| "loss": 0.0001, | |
| "step": 1061 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.365400247154409e-07, | |
| "loss": 0.0079, | |
| "step": 1062 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.352544898258511e-07, | |
| "loss": 0.0067, | |
| "step": 1063 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.3397459621556128e-07, | |
| "loss": 0.0025, | |
| "step": 1064 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.327003522341237e-07, | |
| "loss": 0.0034, | |
| "step": 1065 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.3143176619423345e-07, | |
| "loss": 0.0236, | |
| "step": 1066 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.3016884637167568e-07, | |
| "loss": 0.0194, | |
| "step": 1067 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.2891160100527222e-07, | |
| "loss": 0.0055, | |
| "step": 1068 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.2766003829682504e-07, | |
| "loss": 0.0162, | |
| "step": 1069 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.2641416641106607e-07, | |
| "loss": 0.0256, | |
| "step": 1070 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.2517399347560165e-07, | |
| "loss": 0.0051, | |
| "step": 1071 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.2393952758086056e-07, | |
| "loss": 0.0015, | |
| "step": 1072 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.2271077678004082e-07, | |
| "loss": 0.0034, | |
| "step": 1073 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.2148774908905778e-07, | |
| "loss": 0.0054, | |
| "step": 1074 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.2027045248649015e-07, | |
| "loss": 0.0305, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.1905889491353072e-07, | |
| "loss": 0.0117, | |
| "step": 1076 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.1785308427393181e-07, | |
| "loss": 0.0307, | |
| "step": 1077 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.1665302843395519e-07, | |
| "loss": 0.0429, | |
| "step": 1078 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.1545873522232052e-07, | |
| "loss": 0.0311, | |
| "step": 1079 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.1427021243015466e-07, | |
| "loss": 0.0172, | |
| "step": 1080 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.1308746781093926e-07, | |
| "loss": 0.0175, | |
| "step": 1081 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.1191050908046241e-07, | |
| "loss": 0.0044, | |
| "step": 1082 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.1073934391676665e-07, | |
| "loss": 0.0083, | |
| "step": 1083 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.0957397996009931e-07, | |
| "loss": 0.0027, | |
| "step": 1084 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.0841442481286367e-07, | |
| "loss": 0.0024, | |
| "step": 1085 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.0726068603956739e-07, | |
| "loss": 0.0119, | |
| "step": 1086 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.0611277116677487e-07, | |
| "loss": 0.0105, | |
| "step": 1087 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.0497068768305717e-07, | |
| "loss": 0.0074, | |
| "step": 1088 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.038344430389445e-07, | |
| "loss": 0.0133, | |
| "step": 1089 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.0270404464687499e-07, | |
| "loss": 0.0006, | |
| "step": 1090 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.0157949988114967e-07, | |
| "loss": 0.0157, | |
| "step": 1091 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.0046081607788126e-07, | |
| "loss": 0.0009, | |
| "step": 1092 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 9.934800053494886e-08, | |
| "loss": 0.0241, | |
| "step": 1093 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 9.824106051194858e-08, | |
| "loss": 0.0283, | |
| "step": 1094 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 9.714000323014704e-08, | |
| "loss": 0.031, | |
| "step": 1095 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 9.604483587243383e-08, | |
| "loss": 0.0244, | |
| "step": 1096 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 9.495556558327544e-08, | |
| "loss": 0.0015, | |
| "step": 1097 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 9.387219946866698e-08, | |
| "loss": 0.0131, | |
| "step": 1098 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 9.279474459608804e-08, | |
| "loss": 0.0134, | |
| "step": 1099 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 9.172320799445499e-08, | |
| "loss": 0.0285, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 9.065759665407513e-08, | |
| "loss": 0.0067, | |
| "step": 1101 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 8.959791752660262e-08, | |
| "loss": 0.043, | |
| "step": 1102 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 8.854417752499111e-08, | |
| "loss": 0.011, | |
| "step": 1103 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 8.749638352345001e-08, | |
| "loss": 0.0111, | |
| "step": 1104 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 8.645454235739902e-08, | |
| "loss": 0.0199, | |
| "step": 1105 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 8.54186608234243e-08, | |
| "loss": 0.0032, | |
| "step": 1106 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 8.43887456792326e-08, | |
| "loss": 0.0149, | |
| "step": 1107 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 8.33648036436091e-08, | |
| "loss": 0.0214, | |
| "step": 1108 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 8.234684139637204e-08, | |
| "loss": 0.0014, | |
| "step": 1109 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 8.133486557833013e-08, | |
| "loss": 0.0517, | |
| "step": 1110 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 8.032888279123828e-08, | |
| "loss": 0.0017, | |
| "step": 1111 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.932889959775613e-08, | |
| "loss": 0.0067, | |
| "step": 1112 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.833492252140283e-08, | |
| "loss": 0.0091, | |
| "step": 1113 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.734695804651692e-08, | |
| "loss": 0.0048, | |
| "step": 1114 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.636501261821238e-08, | |
| "loss": 0.0073, | |
| "step": 1115 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.53890926423375e-08, | |
| "loss": 0.0028, | |
| "step": 1116 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.441920448543237e-08, | |
| "loss": 0.0373, | |
| "step": 1117 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.34553544746882e-08, | |
| "loss": 0.0125, | |
| "step": 1118 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.249754889790538e-08, | |
| "loss": 0.0013, | |
| "step": 1119 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.154579400345239e-08, | |
| "loss": 0.0031, | |
| "step": 1120 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.060009600022566e-08, | |
| "loss": 0.0189, | |
| "step": 1121 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 6.966046105760825e-08, | |
| "loss": 0.002, | |
| "step": 1122 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 6.872689530543085e-08, | |
| "loss": 0.0329, | |
| "step": 1123 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.779940483393032e-08, | |
| "loss": 0.0276, | |
| "step": 1124 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.687799569371078e-08, | |
| "loss": 0.0091, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.596267389570387e-08, | |
| "loss": 0.0092, | |
| "step": 1126 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.505344541113045e-08, | |
| "loss": 0.0021, | |
| "step": 1127 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.415031617145949e-08, | |
| "loss": 0.0035, | |
| "step": 1128 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.325329206837216e-08, | |
| "loss": 0.0035, | |
| "step": 1129 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.236237895372099e-08, | |
| "loss": 0.0071, | |
| "step": 1130 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.147758263949321e-08, | |
| "loss": 0.0095, | |
| "step": 1131 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.059890889777198e-08, | |
| "loss": 0.0008, | |
| "step": 1132 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 5.972636346069948e-08, | |
| "loss": 0.0055, | |
| "step": 1133 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 5.885995202043847e-08, | |
| "loss": 0.0006, | |
| "step": 1134 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 5.799968022913637e-08, | |
| "loss": 0.0405, | |
| "step": 1135 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.714555369888774e-08, | |
| "loss": 0.0102, | |
| "step": 1136 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.629757800169732e-08, | |
| "loss": 0.0117, | |
| "step": 1137 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.545575866944441e-08, | |
| "loss": 0.0034, | |
| "step": 1138 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.462010119384664e-08, | |
| "loss": 0.0176, | |
| "step": 1139 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.379061102642357e-08, | |
| "loss": 0.018, | |
| "step": 1140 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.296729357846175e-08, | |
| "loss": 0.0107, | |
| "step": 1141 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.2150154220979014e-08, | |
| "loss": 0.0015, | |
| "step": 1142 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.133919828468991e-08, | |
| "loss": 0.0537, | |
| "step": 1143 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.053443105997068e-08, | |
| "loss": 0.0045, | |
| "step": 1144 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 4.973585779682399e-08, | |
| "loss": 0.004, | |
| "step": 1145 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 4.8943483704846465e-08, | |
| "loss": 0.0114, | |
| "step": 1146 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 4.815731395319278e-08, | |
| "loss": 0.0023, | |
| "step": 1147 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 4.737735367054319e-08, | |
| "loss": 0.0183, | |
| "step": 1148 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.6603607945069456e-08, | |
| "loss": 0.015, | |
| "step": 1149 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.583608182440235e-08, | |
| "loss": 0.0019, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.5074780315597304e-08, | |
| "loss": 0.0227, | |
| "step": 1151 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.43197083851039e-08, | |
| "loss": 0.0121, | |
| "step": 1152 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.357087095873135e-08, | |
| "loss": 0.0043, | |
| "step": 1153 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.282827292161761e-08, | |
| "loss": 0.0044, | |
| "step": 1154 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.209191911819687e-08, | |
| "loss": 0.0135, | |
| "step": 1155 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.136181435216901e-08, | |
| "loss": 0.0147, | |
| "step": 1156 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.063796338646663e-08, | |
| "loss": 0.0173, | |
| "step": 1157 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 3.992037094322531e-08, | |
| "loss": 0.0362, | |
| "step": 1158 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 3.920904170375239e-08, | |
| "loss": 0.0277, | |
| "step": 1159 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 3.8503980308495996e-08, | |
| "loss": 0.0431, | |
| "step": 1160 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 3.780519135701565e-08, | |
| "loss": 0.0393, | |
| "step": 1161 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.71126794079516e-08, | |
| "loss": 0.0084, | |
| "step": 1162 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.642644897899505e-08, | |
| "loss": 0.0168, | |
| "step": 1163 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.574650454685901e-08, | |
| "loss": 0.0032, | |
| "step": 1164 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.507285054724929e-08, | |
| "loss": 0.008, | |
| "step": 1165 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.4405491374834615e-08, | |
| "loss": 0.036, | |
| "step": 1166 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.374443138321936e-08, | |
| "loss": 0.0271, | |
| "step": 1167 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.308967488491365e-08, | |
| "loss": 0.0513, | |
| "step": 1168 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.24412261513064e-08, | |
| "loss": 0.0062, | |
| "step": 1169 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.1799089412636757e-08, | |
| "loss": 0.0303, | |
| "step": 1170 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.1163268857967274e-08, | |
| "loss": 0.0141, | |
| "step": 1171 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.053376863515511e-08, | |
| "loss": 0.0009, | |
| "step": 1172 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 2.991059285082698e-08, | |
| "loss": 0.0043, | |
| "step": 1173 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.929374557035036e-08, | |
| "loss": 0.0028, | |
| "step": 1174 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.868323081780877e-08, | |
| "loss": 0.0401, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.807905257597376e-08, | |
| "loss": 0.0189, | |
| "step": 1176 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.748121478628074e-08, | |
| "loss": 0.0148, | |
| "step": 1177 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.6889721348801853e-08, | |
| "loss": 0.014, | |
| "step": 1178 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.6304576122221034e-08, | |
| "loss": 0.0054, | |
| "step": 1179 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.5725782923808892e-08, | |
| "loss": 0.0202, | |
| "step": 1180 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.5153345529397853e-08, | |
| "loss": 0.0043, | |
| "step": 1181 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.4587267673357394e-08, | |
| "loss": 0.0281, | |
| "step": 1182 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.4027553048569738e-08, | |
| "loss": 0.0086, | |
| "step": 1183 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.3474205306405647e-08, | |
| "loss": 0.0046, | |
| "step": 1184 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.2927228056700777e-08, | |
| "loss": 0.0136, | |
| "step": 1185 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.238662486773213e-08, | |
| "loss": 0.0067, | |
| "step": 1186 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.185239926619431e-08, | |
| "loss": 0.0039, | |
| "step": 1187 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.1324554737177645e-08, | |
| "loss": 0.0008, | |
| "step": 1188 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.0803094724143877e-08, | |
| "loss": 0.0422, | |
| "step": 1189 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.0288022628905165e-08, | |
| "loss": 0.0073, | |
| "step": 1190 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.9779341811600792e-08, | |
| "loss": 0.0446, | |
| "step": 1191 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.927705559067616e-08, | |
| "loss": 0.0047, | |
| "step": 1192 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.8781167242860276e-08, | |
| "loss": 0.0017, | |
| "step": 1193 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.8291680003145073e-08, | |
| "loss": 0.0039, | |
| "step": 1194 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.7808597064764008e-08, | |
| "loss": 0.0469, | |
| "step": 1195 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.733192157917096e-08, | |
| "loss": 0.0144, | |
| "step": 1196 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.686165665602046e-08, | |
| "loss": 0.0464, | |
| "step": 1197 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.6397805363146387e-08, | |
| "loss": 0.0091, | |
| "step": 1198 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.5940370726542863e-08, | |
| "loss": 0.0133, | |
| "step": 1199 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.5489355730343933e-08, | |
| "loss": 0.0042, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.5044763316804266e-08, | |
| "loss": 0.0086, | |
| "step": 1201 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.4606596386279924e-08, | |
| "loss": 0.0226, | |
| "step": 1202 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.4174857797209948e-08, | |
| "loss": 0.0034, | |
| "step": 1203 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.3749550366096596e-08, | |
| "loss": 0.005, | |
| "step": 1204 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.3330676867488123e-08, | |
| "loss": 0.0016, | |
| "step": 1205 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.2918240033960026e-08, | |
| "loss": 0.0238, | |
| "step": 1206 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.2512242556097397e-08, | |
| "loss": 0.0421, | |
| "step": 1207 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.2112687082477147e-08, | |
| "loss": 0.016, | |
| "step": 1208 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.1719576219651584e-08, | |
| "loss": 0.0199, | |
| "step": 1209 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.1332912532129756e-08, | |
| "loss": 0.0009, | |
| "step": 1210 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.0952698542362803e-08, | |
| "loss": 0.0063, | |
| "step": 1211 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.057893673072563e-08, | |
| "loss": 0.0341, | |
| "step": 1212 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.021162953550181e-08, | |
| "loss": 0.024, | |
| "step": 1213 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 9.85077935286749e-09, | |
| "loss": 0.0275, | |
| "step": 1214 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 9.496388536875622e-09, | |
| "loss": 0.0059, | |
| "step": 1215 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 9.148459399440422e-09, | |
| "loss": 0.0136, | |
| "step": 1216 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 8.806994210322604e-09, | |
| "loss": 0.0068, | |
| "step": 1217 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 8.471995197114834e-09, | |
| "loss": 0.0133, | |
| "step": 1218 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 8.143464545226297e-09, | |
| "loss": 0.035, | |
| "step": 1219 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 7.82140439786927e-09, | |
| "loss": 0.0149, | |
| "step": 1220 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 7.505816856045011e-09, | |
| "loss": 0.0135, | |
| "step": 1221 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 7.19670397852945e-09, | |
| "loss": 0.0745, | |
| "step": 1222 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 6.894067781860635e-09, | |
| "loss": 0.0369, | |
| "step": 1223 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 6.5979102403249664e-09, | |
| "loss": 0.0212, | |
| "step": 1224 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 6.308233285944431e-09, | |
| "loss": 0.0147, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 6.025038808464167e-09, | |
| "loss": 0.0084, | |
| "step": 1226 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 5.74832865534014e-09, | |
| "loss": 0.0149, | |
| "step": 1227 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 5.47810463172671e-09, | |
| "loss": 0.0067, | |
| "step": 1228 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 5.214368500465305e-09, | |
| "loss": 0.0037, | |
| "step": 1229 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 4.957121982072765e-09, | |
| "loss": 0.0088, | |
| "step": 1230 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 4.706366754730129e-09, | |
| "loss": 0.0159, | |
| "step": 1231 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 4.46210445427142e-09, | |
| "loss": 0.0006, | |
| "step": 1232 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 4.224336674173545e-09, | |
| "loss": 0.003, | |
| "step": 1233 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 3.99306496554519e-09, | |
| "loss": 0.0075, | |
| "step": 1234 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 3.768290837117604e-09, | |
| "loss": 0.008, | |
| "step": 1235 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 3.550015755233726e-09, | |
| "loss": 0.0028, | |
| "step": 1236 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 3.3382411438392934e-09, | |
| "loss": 0.0037, | |
| "step": 1237 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 3.132968384473411e-09, | |
| "loss": 0.0059, | |
| "step": 1238 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.934198816259559e-09, | |
| "loss": 0.0129, | |
| "step": 1239 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.7419337358967064e-09, | |
| "loss": 0.004, | |
| "step": 1240 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.5561743976511007e-09, | |
| "loss": 0.0256, | |
| "step": 1241 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.376922013347715e-09, | |
| "loss": 0.0142, | |
| "step": 1242 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.2041777523627015e-09, | |
| "loss": 0.0101, | |
| "step": 1243 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.037942741615617e-09, | |
| "loss": 0.0166, | |
| "step": 1244 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.878218065562209e-09, | |
| "loss": 0.0317, | |
| "step": 1245 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.725004766186866e-09, | |
| "loss": 0.0178, | |
| "step": 1246 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.5783038429965089e-09, | |
| "loss": 0.0208, | |
| "step": 1247 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.4381162530135993e-09, | |
| "loss": 0.0208, | |
| "step": 1248 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.3044429107700316e-09, | |
| "loss": 0.005, | |
| "step": 1249 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.1772846883011388e-09, | |
| "loss": 0.001, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.056642415140141e-09, | |
| "loss": 0.0046, | |
| "step": 1251 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 9.425168783123716e-10, | |
| "loss": 0.0017, | |
| "step": 1252 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 8.349088223306156e-10, | |
| "loss": 0.0006, | |
| "step": 1253 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 7.338189491900015e-10, | |
| "loss": 0.0002, | |
| "step": 1254 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 6.392479183633393e-10, | |
| "loss": 0.0092, | |
| "step": 1255 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 5.511963467970116e-10, | |
| "loss": 0.0044, | |
| "step": 1256 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 4.696648089068667e-10, | |
| "loss": 0.0082, | |
| "step": 1257 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 3.946538365744434e-10, | |
| "loss": 0.0107, | |
| "step": 1258 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 3.261639191436405e-10, | |
| "loss": 0.008, | |
| "step": 1259 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 2.6419550341705286e-10, | |
| "loss": 0.0007, | |
| "step": 1260 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 2.0874899365386222e-10, | |
| "loss": 0.0113, | |
| "step": 1261 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.598247515663953e-10, | |
| "loss": 0.0474, | |
| "step": 1262 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.174230963184586e-10, | |
| "loss": 0.0137, | |
| "step": 1263 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 8.15443045226738e-11, | |
| "loss": 0.0256, | |
| "step": 1264 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 5.2188610238923515e-11, | |
| "loss": 0.0168, | |
| "step": 1265 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 2.935620497301894e-11, | |
| "loss": 0.0039, | |
| "step": 1266 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.304723767492355e-11, | |
| "loss": 0.0112, | |
| "step": 1267 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 3.2618147384200213e-12, | |
| "loss": 0.004, | |
| "step": 1268 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 0.0, | |
| "loss": 0.0198, | |
| "step": 1269 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "step": 1269, | |
| "total_flos": 59418315325440.0, | |
| "train_loss": 0.02994988417396055, | |
| "train_runtime": 8676.8672, | |
| "train_samples_per_second": 9.36, | |
| "train_steps_per_second": 0.146 | |
| } | |
| ], | |
| "logging_steps": 1.0, | |
| "max_steps": 1269, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 1, | |
| "save_steps": 50000, | |
| "total_flos": 59418315325440.0, | |
| "train_batch_size": 16, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |