| { |
| "best_metric": null, |
| "best_model_checkpoint": null, |
| "epoch": 3.0, |
| "global_step": 1875, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 0.0, |
| "learning_rate": 3.5087719298245616e-07, |
| "loss": 0.773, |
| "step": 1 |
| }, |
| { |
| "epoch": 0.0, |
| "learning_rate": 7.017543859649123e-07, |
| "loss": 0.7527, |
| "step": 2 |
| }, |
| { |
| "epoch": 0.0, |
| "learning_rate": 1.0526315789473685e-06, |
| "loss": 0.7731, |
| "step": 3 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 1.4035087719298246e-06, |
| "loss": 0.7133, |
| "step": 4 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 1.7543859649122807e-06, |
| "loss": 0.6992, |
| "step": 5 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 2.105263157894737e-06, |
| "loss": 0.7263, |
| "step": 6 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 2.456140350877193e-06, |
| "loss": 0.7031, |
| "step": 7 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 2.8070175438596493e-06, |
| "loss": 0.6531, |
| "step": 8 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 3.157894736842105e-06, |
| "loss": 0.6998, |
| "step": 9 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 3.5087719298245615e-06, |
| "loss": 0.664, |
| "step": 10 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 3.859649122807018e-06, |
| "loss": 0.6692, |
| "step": 11 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 4.210526315789474e-06, |
| "loss": 0.6297, |
| "step": 12 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 4.56140350877193e-06, |
| "loss": 0.6186, |
| "step": 13 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 4.912280701754386e-06, |
| "loss": 0.6456, |
| "step": 14 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 5.263157894736842e-06, |
| "loss": 0.649, |
| "step": 15 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 5.6140350877192985e-06, |
| "loss": 0.6055, |
| "step": 16 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 5.964912280701755e-06, |
| "loss": 0.6142, |
| "step": 17 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 6.31578947368421e-06, |
| "loss": 0.6494, |
| "step": 18 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 6.666666666666667e-06, |
| "loss": 0.6174, |
| "step": 19 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 7.017543859649123e-06, |
| "loss": 0.6076, |
| "step": 20 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 7.368421052631579e-06, |
| "loss": 0.6156, |
| "step": 21 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 7.719298245614036e-06, |
| "loss": 0.6239, |
| "step": 22 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 8.070175438596492e-06, |
| "loss": 0.6183, |
| "step": 23 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 8.421052631578948e-06, |
| "loss": 0.6405, |
| "step": 24 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 8.771929824561405e-06, |
| "loss": 0.6309, |
| "step": 25 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 9.12280701754386e-06, |
| "loss": 0.6392, |
| "step": 26 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 9.473684210526315e-06, |
| "loss": 0.6344, |
| "step": 27 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 9.824561403508772e-06, |
| "loss": 0.5858, |
| "step": 28 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 1.017543859649123e-05, |
| "loss": 0.5697, |
| "step": 29 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 1.0526315789473684e-05, |
| "loss": 0.6234, |
| "step": 30 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 1.0877192982456142e-05, |
| "loss": 0.6187, |
| "step": 31 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 1.1228070175438597e-05, |
| "loss": 0.6317, |
| "step": 32 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 1.1578947368421053e-05, |
| "loss": 0.5855, |
| "step": 33 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 1.192982456140351e-05, |
| "loss": 0.6229, |
| "step": 34 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 1.2280701754385966e-05, |
| "loss": 0.6083, |
| "step": 35 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 1.263157894736842e-05, |
| "loss": 0.6165, |
| "step": 36 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 1.2982456140350879e-05, |
| "loss": 0.6194, |
| "step": 37 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 1.3333333333333333e-05, |
| "loss": 0.6257, |
| "step": 38 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 1.3684210526315791e-05, |
| "loss": 0.5928, |
| "step": 39 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 1.4035087719298246e-05, |
| "loss": 0.6103, |
| "step": 40 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 1.4385964912280704e-05, |
| "loss": 0.5957, |
| "step": 41 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 1.4736842105263159e-05, |
| "loss": 0.5797, |
| "step": 42 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 1.5087719298245615e-05, |
| "loss": 0.6047, |
| "step": 43 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 1.543859649122807e-05, |
| "loss": 0.5812, |
| "step": 44 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 1.578947368421053e-05, |
| "loss": 0.6284, |
| "step": 45 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 1.6140350877192984e-05, |
| "loss": 0.6083, |
| "step": 46 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 1.649122807017544e-05, |
| "loss": 0.6245, |
| "step": 47 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 1.6842105263157896e-05, |
| "loss": 0.6259, |
| "step": 48 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 1.719298245614035e-05, |
| "loss": 0.6268, |
| "step": 49 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 1.754385964912281e-05, |
| "loss": 0.6378, |
| "step": 50 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 1.7894736842105264e-05, |
| "loss": 0.6391, |
| "step": 51 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 1.824561403508772e-05, |
| "loss": 0.5615, |
| "step": 52 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 1.8596491228070176e-05, |
| "loss": 0.608, |
| "step": 53 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 1.894736842105263e-05, |
| "loss": 0.603, |
| "step": 54 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 1.929824561403509e-05, |
| "loss": 0.6075, |
| "step": 55 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 1.9649122807017544e-05, |
| "loss": 0.6005, |
| "step": 56 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 2e-05, |
| "loss": 0.6525, |
| "step": 57 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 1.9999985069241058e-05, |
| "loss": 0.6043, |
| "step": 58 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 1.9999940277008807e-05, |
| "loss": 0.6376, |
| "step": 59 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 1.9999865623437014e-05, |
| "loss": 0.621, |
| "step": 60 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 1.99997611087486e-05, |
| "loss": 0.5908, |
| "step": 61 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 1.9999626733255662e-05, |
| "loss": 0.5732, |
| "step": 62 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 1.9999462497359468e-05, |
| "loss": 0.5849, |
| "step": 63 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 1.9999268401550445e-05, |
| "loss": 0.5908, |
| "step": 64 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 1.9999044446408203e-05, |
| "loss": 0.6015, |
| "step": 65 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 1.9998790632601496e-05, |
| "loss": 0.6159, |
| "step": 66 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 1.9998506960888258e-05, |
| "loss": 0.5821, |
| "step": 67 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 1.999819343211557e-05, |
| "loss": 0.5705, |
| "step": 68 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 1.999785004721968e-05, |
| "loss": 0.5952, |
| "step": 69 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 1.9997476807225987e-05, |
| "loss": 0.5967, |
| "step": 70 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 1.999707371324904e-05, |
| "loss": 0.6182, |
| "step": 71 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 1.9996640766492542e-05, |
| "loss": 0.6011, |
| "step": 72 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 1.9996177968249336e-05, |
| "loss": 0.607, |
| "step": 73 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 1.999568531990141e-05, |
| "loss": 0.6333, |
| "step": 74 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 1.999516282291988e-05, |
| "loss": 0.5477, |
| "step": 75 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 1.9994610478865012e-05, |
| "loss": 0.5655, |
| "step": 76 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 1.999402828938618e-05, |
| "loss": 0.5533, |
| "step": 77 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 1.9993416256221894e-05, |
| "loss": 0.6268, |
| "step": 78 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 1.999277438119978e-05, |
| "loss": 0.6118, |
| "step": 79 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 1.9992102666236567e-05, |
| "loss": 0.5912, |
| "step": 80 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 1.9991401113338103e-05, |
| "loss": 0.5751, |
| "step": 81 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 1.9990669724599336e-05, |
| "loss": 0.5513, |
| "step": 82 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 1.9989908502204295e-05, |
| "loss": 0.6103, |
| "step": 83 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 1.998911744842611e-05, |
| "loss": 0.5854, |
| "step": 84 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 1.9988296565626988e-05, |
| "loss": 0.6099, |
| "step": 85 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 1.9987445856258208e-05, |
| "loss": 0.5671, |
| "step": 86 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 1.9986565322860117e-05, |
| "loss": 0.6284, |
| "step": 87 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 1.9985654968062122e-05, |
| "loss": 0.5725, |
| "step": 88 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 1.9984714794582682e-05, |
| "loss": 0.5973, |
| "step": 89 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 1.9983744805229296e-05, |
| "loss": 0.5854, |
| "step": 90 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 1.99827450028985e-05, |
| "loss": 0.594, |
| "step": 91 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 1.998171539057586e-05, |
| "loss": 0.555, |
| "step": 92 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 1.9980655971335944e-05, |
| "loss": 0.5764, |
| "step": 93 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 1.9979566748342348e-05, |
| "loss": 0.6018, |
| "step": 94 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 1.9978447724847655e-05, |
| "loss": 0.5636, |
| "step": 95 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 1.9977298904193438e-05, |
| "loss": 0.599, |
| "step": 96 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 1.9976120289810247e-05, |
| "loss": 0.5518, |
| "step": 97 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 1.997491188521761e-05, |
| "loss": 0.5843, |
| "step": 98 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 1.9973673694024002e-05, |
| "loss": 0.6028, |
| "step": 99 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 1.997240571992685e-05, |
| "loss": 0.5699, |
| "step": 100 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 1.9971107966712518e-05, |
| "loss": 0.5529, |
| "step": 101 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 1.9969780438256295e-05, |
| "loss": 0.6122, |
| "step": 102 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 1.9968423138522382e-05, |
| "loss": 0.563, |
| "step": 103 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 1.9967036071563878e-05, |
| "loss": 0.5992, |
| "step": 104 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 1.996561924152278e-05, |
| "loss": 0.552, |
| "step": 105 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 1.996417265262996e-05, |
| "loss": 0.5787, |
| "step": 106 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 1.9962696309205146e-05, |
| "loss": 0.5775, |
| "step": 107 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 1.996119021565693e-05, |
| "loss": 0.5694, |
| "step": 108 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 1.995965437648273e-05, |
| "loss": 0.6013, |
| "step": 109 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 1.9958088796268794e-05, |
| "loss": 0.5958, |
| "step": 110 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 1.995649347969019e-05, |
| "loss": 0.6041, |
| "step": 111 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 1.9954868431510764e-05, |
| "loss": 0.5839, |
| "step": 112 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 1.995321365658317e-05, |
| "loss": 0.5498, |
| "step": 113 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 1.9951529159848805e-05, |
| "loss": 0.6324, |
| "step": 114 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 1.994981494633784e-05, |
| "loss": 0.5455, |
| "step": 115 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 1.9948071021169176e-05, |
| "loss": 0.5634, |
| "step": 116 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 1.9946297389550433e-05, |
| "loss": 0.5358, |
| "step": 117 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 1.9944494056777945e-05, |
| "loss": 0.5429, |
| "step": 118 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 1.9942661028236746e-05, |
| "loss": 0.5533, |
| "step": 119 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 1.9940798309400527e-05, |
| "loss": 0.5397, |
| "step": 120 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 1.9938905905831657e-05, |
| "loss": 0.6056, |
| "step": 121 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 1.9936983823181132e-05, |
| "loss": 0.5595, |
| "step": 122 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 1.993503206718859e-05, |
| "loss": 0.5762, |
| "step": 123 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 1.993305064368227e-05, |
| "loss": 0.5569, |
| "step": 124 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 1.9931039558578997e-05, |
| "loss": 0.541, |
| "step": 125 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 1.9928998817884185e-05, |
| "loss": 0.5461, |
| "step": 126 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 1.9926928427691788e-05, |
| "loss": 0.6072, |
| "step": 127 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 1.9924828394184308e-05, |
| "loss": 0.5319, |
| "step": 128 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 1.992269872363277e-05, |
| "loss": 0.5326, |
| "step": 129 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 1.992053942239668e-05, |
| "loss": 0.517, |
| "step": 130 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 1.991835049692405e-05, |
| "loss": 0.575, |
| "step": 131 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 1.9916131953751342e-05, |
| "loss": 0.5749, |
| "step": 132 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 1.991388379950346e-05, |
| "loss": 0.5426, |
| "step": 133 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 1.9911606040893742e-05, |
| "loss": 0.5931, |
| "step": 134 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 1.9909298684723905e-05, |
| "loss": 0.5735, |
| "step": 135 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 1.990696173788408e-05, |
| "loss": 0.5537, |
| "step": 136 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 1.9904595207352736e-05, |
| "loss": 0.5579, |
| "step": 137 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 1.9902199100196697e-05, |
| "loss": 0.5514, |
| "step": 138 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 1.9899773423571102e-05, |
| "loss": 0.5389, |
| "step": 139 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 1.9897318184719386e-05, |
| "loss": 0.5376, |
| "step": 140 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 1.9894833390973266e-05, |
| "loss": 0.5683, |
| "step": 141 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 1.989231904975272e-05, |
| "loss": 0.5769, |
| "step": 142 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 1.9889775168565942e-05, |
| "loss": 0.5338, |
| "step": 143 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 1.9887201755009358e-05, |
| "loss": 0.5489, |
| "step": 144 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 1.9884598816767563e-05, |
| "loss": 0.5508, |
| "step": 145 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 1.988196636161333e-05, |
| "loss": 0.5834, |
| "step": 146 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 1.987930439740757e-05, |
| "loss": 0.5435, |
| "step": 147 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 1.987661293209931e-05, |
| "loss": 0.5958, |
| "step": 148 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 1.9873891973725673e-05, |
| "loss": 0.5975, |
| "step": 149 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 1.9871141530411854e-05, |
| "loss": 0.5965, |
| "step": 150 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 1.98683616103711e-05, |
| "loss": 0.5372, |
| "step": 151 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 1.986555222190467e-05, |
| "loss": 0.5906, |
| "step": 152 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 1.986271337340182e-05, |
| "loss": 0.6062, |
| "step": 153 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 1.9859845073339788e-05, |
| "loss": 0.543, |
| "step": 154 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 1.9856947330283752e-05, |
| "loss": 0.5955, |
| "step": 155 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 1.9854020152886816e-05, |
| "loss": 0.5422, |
| "step": 156 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 1.985106354988997e-05, |
| "loss": 0.5347, |
| "step": 157 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 1.9848077530122083e-05, |
| "loss": 0.5436, |
| "step": 158 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 1.984506210249986e-05, |
| "loss": 0.5881, |
| "step": 159 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 1.984201727602783e-05, |
| "loss": 0.5587, |
| "step": 160 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 1.9838943059798305e-05, |
| "loss": 0.5641, |
| "step": 161 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 1.983583946299136e-05, |
| "loss": 0.5364, |
| "step": 162 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 1.9832706494874812e-05, |
| "loss": 0.5941, |
| "step": 163 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 1.9829544164804172e-05, |
| "loss": 0.557, |
| "step": 164 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 1.982635248222264e-05, |
| "loss": 0.5392, |
| "step": 165 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 1.9823131456661064e-05, |
| "loss": 0.5805, |
| "step": 166 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 1.9819881097737917e-05, |
| "loss": 0.556, |
| "step": 167 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 1.9816601415159266e-05, |
| "loss": 0.5742, |
| "step": 168 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 1.9813292418718734e-05, |
| "loss": 0.594, |
| "step": 169 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 1.980995411829749e-05, |
| "loss": 0.5713, |
| "step": 170 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 1.9806586523864212e-05, |
| "loss": 0.5845, |
| "step": 171 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 1.980318964547504e-05, |
| "loss": 0.5503, |
| "step": 172 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 1.9799763493273572e-05, |
| "loss": 0.5413, |
| "step": 173 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 1.9796308077490817e-05, |
| "loss": 0.5874, |
| "step": 174 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 1.9792823408445173e-05, |
| "loss": 0.5752, |
| "step": 175 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 1.978930949654239e-05, |
| "loss": 0.5462, |
| "step": 176 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 1.978576635227554e-05, |
| "loss": 0.6067, |
| "step": 177 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 1.9782193986224997e-05, |
| "loss": 0.5772, |
| "step": 178 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 1.9778592409058376e-05, |
| "loss": 0.5606, |
| "step": 179 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 1.9774961631530543e-05, |
| "loss": 0.5489, |
| "step": 180 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 1.9771301664483548e-05, |
| "loss": 0.5681, |
| "step": 181 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 1.976761251884661e-05, |
| "loss": 0.506, |
| "step": 182 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 1.976389420563607e-05, |
| "loss": 0.5206, |
| "step": 183 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 1.9760146735955388e-05, |
| "loss": 0.5639, |
| "step": 184 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 1.975637012099507e-05, |
| "loss": 0.5671, |
| "step": 185 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 1.9752564372032655e-05, |
| "loss": 0.5009, |
| "step": 186 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 1.97487295004327e-05, |
| "loss": 0.5492, |
| "step": 187 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 1.974486551764671e-05, |
| "loss": 0.5626, |
| "step": 188 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 1.9740972435213114e-05, |
| "loss": 0.5598, |
| "step": 189 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 1.973705026475726e-05, |
| "loss": 0.558, |
| "step": 190 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 1.9733099017991342e-05, |
| "loss": 0.5754, |
| "step": 191 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 1.9729118706714377e-05, |
| "loss": 0.5846, |
| "step": 192 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 1.972510934281218e-05, |
| "loss": 0.5762, |
| "step": 193 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 1.9721070938257326e-05, |
| "loss": 0.5648, |
| "step": 194 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 1.9717003505109097e-05, |
| "loss": 0.5152, |
| "step": 195 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 1.971290705551347e-05, |
| "loss": 0.5816, |
| "step": 196 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 1.9708781601703066e-05, |
| "loss": 0.5619, |
| "step": 197 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 1.970462715599711e-05, |
| "loss": 0.5928, |
| "step": 198 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 1.9700443730801412e-05, |
| "loss": 0.581, |
| "step": 199 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 1.9696231338608317e-05, |
| "loss": 0.5672, |
| "step": 200 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 1.9691989991996663e-05, |
| "loss": 0.595, |
| "step": 201 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 1.9687719703631757e-05, |
| "loss": 0.5978, |
| "step": 202 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 1.9683420486265328e-05, |
| "loss": 0.5407, |
| "step": 203 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 1.967909235273549e-05, |
| "loss": 0.6042, |
| "step": 204 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 1.967473531596671e-05, |
| "loss": 0.5583, |
| "step": 205 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 1.9670349388969758e-05, |
| "loss": 0.5723, |
| "step": 206 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 1.966593458484168e-05, |
| "loss": 0.5413, |
| "step": 207 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 1.9661490916765752e-05, |
| "loss": 0.5199, |
| "step": 208 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 1.9657018398011435e-05, |
| "loss": 0.5305, |
| "step": 209 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 1.9652517041934357e-05, |
| "loss": 0.5518, |
| "step": 210 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 1.9647986861976246e-05, |
| "loss": 0.5427, |
| "step": 211 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 1.9643427871664912e-05, |
| "loss": 0.583, |
| "step": 212 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 1.9638840084614182e-05, |
| "loss": 0.5996, |
| "step": 213 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 1.963422351452389e-05, |
| "loss": 0.5326, |
| "step": 214 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 1.9629578175179823e-05, |
| "loss": 0.5448, |
| "step": 215 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 1.9624904080453656e-05, |
| "loss": 0.5672, |
| "step": 216 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 1.9620201244302952e-05, |
| "loss": 0.4976, |
| "step": 217 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 1.9615469680771097e-05, |
| "loss": 0.5992, |
| "step": 218 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 1.9610709403987248e-05, |
| "loss": 0.5428, |
| "step": 219 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 1.960592042816632e-05, |
| "loss": 0.5318, |
| "step": 220 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 1.9601102767608924e-05, |
| "loss": 0.5493, |
| "step": 221 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 1.9596256436701324e-05, |
| "loss": 0.5564, |
| "step": 222 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 1.95913814499154e-05, |
| "loss": 0.5432, |
| "step": 223 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 1.9586477821808597e-05, |
| "loss": 0.5481, |
| "step": 224 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 1.95815455670239e-05, |
| "loss": 0.5502, |
| "step": 225 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 1.957658470028977e-05, |
| "loss": 0.5441, |
| "step": 226 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 1.9571595236420103e-05, |
| "loss": 0.5763, |
| "step": 227 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 1.95665771903142e-05, |
| "loss": 0.5385, |
| "step": 228 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 1.9561530576956703e-05, |
| "loss": 0.5697, |
| "step": 229 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 1.9556455411417575e-05, |
| "loss": 0.5381, |
| "step": 230 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 1.955135170885202e-05, |
| "loss": 0.5137, |
| "step": 231 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 1.9546219484500475e-05, |
| "loss": 0.5421, |
| "step": 232 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 1.9541058753688538e-05, |
| "loss": 0.5157, |
| "step": 233 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 1.9535869531826938e-05, |
| "loss": 0.5521, |
| "step": 234 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 1.9530651834411477e-05, |
| "loss": 0.5172, |
| "step": 235 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 1.952540567702299e-05, |
| "loss": 0.5602, |
| "step": 236 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 1.95201310753273e-05, |
| "loss": 0.5446, |
| "step": 237 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 1.951482804507517e-05, |
| "loss": 0.5629, |
| "step": 238 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 1.9509496602102253e-05, |
| "loss": 0.5554, |
| "step": 239 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 1.9504136762329046e-05, |
| "loss": 0.5577, |
| "step": 240 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 1.9498748541760845e-05, |
| "loss": 0.5156, |
| "step": 241 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 1.949333195648769e-05, |
| "loss": 0.5695, |
| "step": 242 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 1.9487887022684336e-05, |
| "loss": 0.5294, |
| "step": 243 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 1.9482413756610175e-05, |
| "loss": 0.5401, |
| "step": 244 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 1.947691217460921e-05, |
| "loss": 0.5339, |
| "step": 245 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 1.9471382293110004e-05, |
| "loss": 0.5386, |
| "step": 246 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 1.946582412862562e-05, |
| "loss": 0.5673, |
| "step": 247 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 1.9460237697753577e-05, |
| "loss": 0.5934, |
| "step": 248 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 1.9454623017175814e-05, |
| "loss": 0.5308, |
| "step": 249 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 1.9448980103658613e-05, |
| "loss": 0.535, |
| "step": 250 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 1.9443308974052574e-05, |
| "loss": 0.5205, |
| "step": 251 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 1.943760964529255e-05, |
| "loss": 0.5622, |
| "step": 252 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 1.9431882134397596e-05, |
| "loss": 0.5574, |
| "step": 253 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 1.9426126458470936e-05, |
| "loss": 0.5185, |
| "step": 254 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 1.9420342634699893e-05, |
| "loss": 0.5185, |
| "step": 255 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 1.9414530680355837e-05, |
| "loss": 0.5536, |
| "step": 256 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 1.9408690612794146e-05, |
| "loss": 0.4975, |
| "step": 257 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 1.9402822449454154e-05, |
| "loss": 0.536, |
| "step": 258 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 1.9396926207859085e-05, |
| "loss": 0.5463, |
| "step": 259 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 1.939100190561601e-05, |
| "loss": 0.563, |
| "step": 260 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 1.9385049560415794e-05, |
| "loss": 0.5377, |
| "step": 261 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 1.9379069190033042e-05, |
| "loss": 0.5848, |
| "step": 262 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 1.9373060812326053e-05, |
| "loss": 0.5423, |
| "step": 263 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 1.936702444523675e-05, |
| "loss": 0.5108, |
| "step": 264 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 1.9360960106790645e-05, |
| "loss": 0.5033, |
| "step": 265 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 1.9354867815096772e-05, |
| "loss": 0.5594, |
| "step": 266 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 1.9348747588347637e-05, |
| "loss": 0.5897, |
| "step": 267 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 1.9342599444819167e-05, |
| "loss": 0.553, |
| "step": 268 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 1.9336423402870655e-05, |
| "loss": 0.529, |
| "step": 269 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 1.9330219480944693e-05, |
| "loss": 0.526, |
| "step": 270 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 1.932398769756714e-05, |
| "loss": 0.5681, |
| "step": 271 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 1.931772807134704e-05, |
| "loss": 0.5699, |
| "step": 272 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 1.9311440620976597e-05, |
| "loss": 0.5258, |
| "step": 273 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 1.9305125365231087e-05, |
| "loss": 0.5319, |
| "step": 274 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 1.9298782322968817e-05, |
| "loss": 0.5743, |
| "step": 275 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 1.929241151313108e-05, |
| "loss": 0.5628, |
| "step": 276 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 1.9286012954742078e-05, |
| "loss": 0.5102, |
| "step": 277 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 1.9279586666908886e-05, |
| "loss": 0.5523, |
| "step": 278 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 1.9273132668821363e-05, |
| "loss": 0.5358, |
| "step": 279 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 1.9266650979752137e-05, |
| "loss": 0.5939, |
| "step": 280 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 1.9260141619056507e-05, |
| "loss": 0.5581, |
| "step": 281 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 1.925360460617242e-05, |
| "loss": 0.5084, |
| "step": 282 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 1.924703996062038e-05, |
| "loss": 0.5425, |
| "step": 283 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 1.9240447702003422e-05, |
| "loss": 0.5214, |
| "step": 284 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 1.9233827850007028e-05, |
| "loss": 0.5398, |
| "step": 285 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 1.9227180424399082e-05, |
| "loss": 0.5856, |
| "step": 286 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 1.9220505445029803e-05, |
| "loss": 0.5809, |
| "step": 287 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 1.9213802931831697e-05, |
| "loss": 0.5514, |
| "step": 288 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 1.9207072904819484e-05, |
| "loss": 0.5397, |
| "step": 289 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 1.9200315384090045e-05, |
| "loss": 0.574, |
| "step": 290 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 1.9193530389822364e-05, |
| "loss": 0.5257, |
| "step": 291 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 1.9186717942277466e-05, |
| "loss": 0.5955, |
| "step": 292 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 1.9179878061798347e-05, |
| "loss": 0.5419, |
| "step": 293 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 1.9173010768809934e-05, |
| "loss": 0.5228, |
| "step": 294 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 1.9166116083819002e-05, |
| "loss": 0.5172, |
| "step": 295 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 1.915919402741413e-05, |
| "loss": 0.5355, |
| "step": 296 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 1.915224462026563e-05, |
| "loss": 0.5261, |
| "step": 297 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 1.9145267883125483e-05, |
| "loss": 0.4926, |
| "step": 298 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 1.913826383682729e-05, |
| "loss": 0.5202, |
| "step": 299 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 1.913123250228619e-05, |
| "loss": 0.5514, |
| "step": 300 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 1.912417390049882e-05, |
| "loss": 0.537, |
| "step": 301 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 1.9117088052543233e-05, |
| "loss": 0.5408, |
| "step": 302 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 1.9109974979578852e-05, |
| "loss": 0.5239, |
| "step": 303 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 1.9102834702846387e-05, |
| "loss": 0.5597, |
| "step": 304 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 1.909566724366779e-05, |
| "loss": 0.5799, |
| "step": 305 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 1.9088472623446182e-05, |
| "loss": 0.5489, |
| "step": 306 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 1.9081250863665794e-05, |
| "loss": 0.5824, |
| "step": 307 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 1.9074001985891893e-05, |
| "loss": 0.5498, |
| "step": 308 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 1.9066726011770725e-05, |
| "loss": 0.5518, |
| "step": 309 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 1.9059422963029464e-05, |
| "loss": 0.5501, |
| "step": 310 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 1.905209286147611e-05, |
| "loss": 0.5595, |
| "step": 311 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 1.9044735728999472e-05, |
| "loss": 0.53, |
| "step": 312 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 1.903735158756905e-05, |
| "loss": 0.5362, |
| "step": 313 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 1.902994045923502e-05, |
| "loss": 0.5774, |
| "step": 314 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 1.9022502366128136e-05, |
| "loss": 0.5099, |
| "step": 315 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 1.901503733045967e-05, |
| "loss": 0.5133, |
| "step": 316 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 1.9007545374521354e-05, |
| "loss": 0.5357, |
| "step": 317 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 1.90000265206853e-05, |
| "loss": 0.5183, |
| "step": 318 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 1.8992480791403957e-05, |
| "loss": 0.5619, |
| "step": 319 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 1.898490820921001e-05, |
| "loss": 0.5447, |
| "step": 320 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 1.897730879671634e-05, |
| "loss": 0.5781, |
| "step": 321 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 1.8969682576615947e-05, |
| "loss": 0.5516, |
| "step": 322 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 1.8962029571681887e-05, |
| "loss": 0.533, |
| "step": 323 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 1.8954349804767185e-05, |
| "loss": 0.5184, |
| "step": 324 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 1.8946643298804794e-05, |
| "loss": 0.545, |
| "step": 325 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 1.8938910076807514e-05, |
| "loss": 0.543, |
| "step": 326 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 1.8931150161867917e-05, |
| "loss": 0.5196, |
| "step": 327 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 1.892336357715829e-05, |
| "loss": 0.5132, |
| "step": 328 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 1.891555034593055e-05, |
| "loss": 0.5825, |
| "step": 329 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 1.8907710491516197e-05, |
| "loss": 0.5452, |
| "step": 330 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 1.8899844037326227e-05, |
| "loss": 0.5496, |
| "step": 331 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 1.889195100685106e-05, |
| "loss": 0.5434, |
| "step": 332 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 1.8884031423660492e-05, |
| "loss": 0.5586, |
| "step": 333 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 1.8876085311403592e-05, |
| "loss": 0.5219, |
| "step": 334 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 1.8868112693808664e-05, |
| "loss": 0.5569, |
| "step": 335 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 1.8860113594683148e-05, |
| "loss": 0.5616, |
| "step": 336 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 1.8852088037913577e-05, |
| "loss": 0.5028, |
| "step": 337 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 1.884403604746547e-05, |
| "loss": 0.564, |
| "step": 338 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 1.8835957647383304e-05, |
| "loss": 0.5429, |
| "step": 339 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 1.8827852861790398e-05, |
| "loss": 0.5158, |
| "step": 340 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 1.8819721714888878e-05, |
| "loss": 0.5281, |
| "step": 341 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 1.8811564230959585e-05, |
| "loss": 0.5143, |
| "step": 342 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 1.8803380434362e-05, |
| "loss": 0.5381, |
| "step": 343 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 1.879517034953418e-05, |
| "loss": 0.5232, |
| "step": 344 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 1.878693400099269e-05, |
| "loss": 0.5464, |
| "step": 345 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 1.8778671413332513e-05, |
| "loss": 0.5214, |
| "step": 346 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 1.877038261122699e-05, |
| "loss": 0.5415, |
| "step": 347 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 1.8762067619427745e-05, |
| "loss": 0.5561, |
| "step": 348 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 1.87537264627646e-05, |
| "loss": 0.5316, |
| "step": 349 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 1.8745359166145526e-05, |
| "loss": 0.5153, |
| "step": 350 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 1.8736965754556527e-05, |
| "loss": 0.513, |
| "step": 351 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 1.8728546253061614e-05, |
| "loss": 0.5381, |
| "step": 352 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 1.8720100686802693e-05, |
| "loss": 0.5461, |
| "step": 353 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 1.8711629080999506e-05, |
| "loss": 0.5239, |
| "step": 354 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 1.8703131460949555e-05, |
| "loss": 0.5534, |
| "step": 355 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 1.869460785202802e-05, |
| "loss": 0.5271, |
| "step": 356 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 1.86860582796877e-05, |
| "loss": 0.5038, |
| "step": 357 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 1.8677482769458905e-05, |
| "loss": 0.5594, |
| "step": 358 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 1.866888134694942e-05, |
| "loss": 0.5604, |
| "step": 359 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 1.866025403784439e-05, |
| "loss": 0.5339, |
| "step": 360 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 1.865160086790627e-05, |
| "loss": 0.5203, |
| "step": 361 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 1.8642921862974742e-05, |
| "loss": 0.5102, |
| "step": 362 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 1.8634217048966638e-05, |
| "loss": 0.557, |
| "step": 363 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 1.8625486451875843e-05, |
| "loss": 0.5213, |
| "step": 364 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 1.861673009777325e-05, |
| "loss": 0.5415, |
| "step": 365 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 1.8607948012806664e-05, |
| "loss": 0.5273, |
| "step": 366 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 1.8599140223200716e-05, |
| "loss": 0.5467, |
| "step": 367 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 1.859030675525681e-05, |
| "loss": 0.5508, |
| "step": 368 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 1.858144763535302e-05, |
| "loss": 0.5331, |
| "step": 369 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 1.857256288994402e-05, |
| "loss": 0.5672, |
| "step": 370 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 1.8563652545561014e-05, |
| "loss": 0.5578, |
| "step": 371 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 1.855471662881164e-05, |
| "loss": 0.4993, |
| "step": 372 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 1.8545755166379898e-05, |
| "loss": 0.5233, |
| "step": 373 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 1.8536768185026085e-05, |
| "loss": 0.5473, |
| "step": 374 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 1.852775571158668e-05, |
| "loss": 0.531, |
| "step": 375 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 1.85187177729743e-05, |
| "loss": 0.5131, |
| "step": 376 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 1.850965439617761e-05, |
| "loss": 0.5521, |
| "step": 377 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 1.8500565608261215e-05, |
| "loss": 0.5244, |
| "step": 378 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 1.8491451436365628e-05, |
| "loss": 0.5155, |
| "step": 379 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 1.848231190770714e-05, |
| "loss": 0.5239, |
| "step": 380 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 1.8473147049577777e-05, |
| "loss": 0.5411, |
| "step": 381 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 1.8463956889345195e-05, |
| "loss": 0.5505, |
| "step": 382 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 1.8454741454452604e-05, |
| "loss": 0.5618, |
| "step": 383 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 1.8445500772418697e-05, |
| "loss": 0.5266, |
| "step": 384 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 1.843623487083755e-05, |
| "loss": 0.5103, |
| "step": 385 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 1.842694377737855e-05, |
| "loss": 0.5195, |
| "step": 386 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 1.8417627519786317e-05, |
| "loss": 0.5502, |
| "step": 387 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 1.8408286125880605e-05, |
| "loss": 0.5308, |
| "step": 388 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 1.839891962355624e-05, |
| "loss": 0.5232, |
| "step": 389 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 1.8389528040783014e-05, |
| "loss": 0.521, |
| "step": 390 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 1.838011140560562e-05, |
| "loss": 0.5215, |
| "step": 391 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 1.8370669746143566e-05, |
| "loss": 0.5396, |
| "step": 392 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 1.836120309059107e-05, |
| "loss": 0.5212, |
| "step": 393 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 1.835171146721701e-05, |
| "loss": 0.5368, |
| "step": 394 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 1.8342194904364815e-05, |
| "loss": 0.5249, |
| "step": 395 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 1.8332653430452375e-05, |
| "loss": 0.5023, |
| "step": 396 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 1.8323087073971996e-05, |
| "loss": 0.514, |
| "step": 397 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 1.831349586349026e-05, |
| "loss": 0.5262, |
| "step": 398 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 1.8303879827647977e-05, |
| "loss": 0.5248, |
| "step": 399 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 1.8294238995160093e-05, |
| "loss": 0.5219, |
| "step": 400 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 1.8284573394815596e-05, |
| "loss": 0.5015, |
| "step": 401 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 1.8274883055477436e-05, |
| "loss": 0.5267, |
| "step": 402 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 1.826516800608244e-05, |
| "loss": 0.5488, |
| "step": 403 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 1.8255428275641212e-05, |
| "loss": 0.5442, |
| "step": 404 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 1.8245663893238075e-05, |
| "loss": 0.5294, |
| "step": 405 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 1.823587488803095e-05, |
| "loss": 0.5304, |
| "step": 406 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 1.8226061289251297e-05, |
| "loss": 0.5353, |
| "step": 407 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 1.821622312620401e-05, |
| "loss": 0.5206, |
| "step": 408 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 1.8206360428267332e-05, |
| "loss": 0.5482, |
| "step": 409 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 1.8196473224892784e-05, |
| "loss": 0.5639, |
| "step": 410 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 1.8186561545605055e-05, |
| "loss": 0.5894, |
| "step": 411 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 1.817662542000192e-05, |
| "loss": 0.556, |
| "step": 412 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 1.816666487775416e-05, |
| "loss": 0.5398, |
| "step": 413 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 1.815667994860547e-05, |
| "loss": 0.5663, |
| "step": 414 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 1.8146670662372353e-05, |
| "loss": 0.5735, |
| "step": 415 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 1.813663704894407e-05, |
| "loss": 0.5579, |
| "step": 416 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 1.8126579138282502e-05, |
| "loss": 0.5371, |
| "step": 417 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 1.8116496960422108e-05, |
| "loss": 0.5493, |
| "step": 418 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 1.8106390545469797e-05, |
| "loss": 0.5184, |
| "step": 419 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 1.809625992360485e-05, |
| "loss": 0.5472, |
| "step": 420 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 1.8086105125078858e-05, |
| "loss": 0.5135, |
| "step": 421 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 1.8075926180215576e-05, |
| "loss": 0.5222, |
| "step": 422 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 1.8065723119410885e-05, |
| "loss": 0.5065, |
| "step": 423 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 1.805549597313267e-05, |
| "loss": 0.5089, |
| "step": 424 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 1.804524477192075e-05, |
| "loss": 0.5475, |
| "step": 425 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 1.803496954638676e-05, |
| "loss": 0.5491, |
| "step": 426 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 1.8024670327214084e-05, |
| "loss": 0.5439, |
| "step": 427 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 1.8014347145157757e-05, |
| "loss": 0.5322, |
| "step": 428 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 1.8004000031044363e-05, |
| "loss": 0.5415, |
| "step": 429 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 1.799362901577196e-05, |
| "loss": 0.5242, |
| "step": 430 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 1.798323413030997e-05, |
| "loss": 0.5489, |
| "step": 431 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 1.7972815405699105e-05, |
| "loss": 0.5321, |
| "step": 432 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 1.796237287305125e-05, |
| "loss": 0.5466, |
| "step": 433 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 1.7951906563549397e-05, |
| "loss": 0.5374, |
| "step": 434 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 1.7941416508447537e-05, |
| "loss": 0.5242, |
| "step": 435 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 1.793090273907056e-05, |
| "loss": 0.5247, |
| "step": 436 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 1.792036528681418e-05, |
| "loss": 0.5018, |
| "step": 437 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 1.7909804183144837e-05, |
| "loss": 0.5693, |
| "step": 438 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 1.789921945959958e-05, |
| "loss": 0.5148, |
| "step": 439 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 1.7888611147786003e-05, |
| "loss": 0.5647, |
| "step": 440 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 1.7877979279382135e-05, |
| "loss": 0.549, |
| "step": 441 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 1.786732388613635e-05, |
| "loss": 0.5518, |
| "step": 442 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 1.7856644999867264e-05, |
| "loss": 0.5708, |
| "step": 443 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 1.784594265246366e-05, |
| "loss": 0.5007, |
| "step": 444 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 1.783521687588437e-05, |
| "loss": 0.54, |
| "step": 445 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 1.782446770215819e-05, |
| "loss": 0.5319, |
| "step": 446 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 1.781369516338378e-05, |
| "loss": 0.5789, |
| "step": 447 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 1.7802899291729585e-05, |
| "loss": 0.493, |
| "step": 448 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 1.779208011943371e-05, |
| "loss": 0.5259, |
| "step": 449 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 1.7781237678803845e-05, |
| "loss": 0.4941, |
| "step": 450 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 1.777037200221717e-05, |
| "loss": 0.5241, |
| "step": 451 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 1.775948312212024e-05, |
| "loss": 0.5295, |
| "step": 452 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 1.77485710710289e-05, |
| "loss": 0.5477, |
| "step": 453 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 1.7737635881528198e-05, |
| "loss": 0.5369, |
| "step": 454 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 1.7726677586272263e-05, |
| "loss": 0.5818, |
| "step": 455 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 1.7715696217984233e-05, |
| "loss": 0.5075, |
| "step": 456 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 1.7704691809456142e-05, |
| "loss": 0.5317, |
| "step": 457 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 1.7693664393548822e-05, |
| "loss": 0.5202, |
| "step": 458 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 1.7682614003191807e-05, |
| "loss": 0.5721, |
| "step": 459 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 1.7671540671383245e-05, |
| "loss": 0.5158, |
| "step": 460 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 1.766044443118978e-05, |
| "loss": 0.5517, |
| "step": 461 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 1.764932531574648e-05, |
| "loss": 0.4769, |
| "step": 462 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 1.76381833582567e-05, |
| "loss": 0.5741, |
| "step": 463 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 1.762701859199202e-05, |
| "loss": 0.5667, |
| "step": 464 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 1.761583105029213e-05, |
| "loss": 0.5399, |
| "step": 465 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 1.7604620766564725e-05, |
| "loss": 0.5869, |
| "step": 466 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 1.7593387774285412e-05, |
| "loss": 0.5291, |
| "step": 467 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 1.7582132106997615e-05, |
| "loss": 0.5199, |
| "step": 468 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 1.7570853798312462e-05, |
| "loss": 0.5346, |
| "step": 469 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 1.7559552881908698e-05, |
| "loss": 0.5291, |
| "step": 470 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 1.7548229391532572e-05, |
| "loss": 0.5305, |
| "step": 471 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 1.7536883360997743e-05, |
| "loss": 0.5343, |
| "step": 472 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 1.7525514824185187e-05, |
| "loss": 0.5118, |
| "step": 473 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 1.7514123815043073e-05, |
| "loss": 0.5352, |
| "step": 474 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 1.750271036758669e-05, |
| "loss": 0.5526, |
| "step": 475 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 1.749127451589832e-05, |
| "loss": 0.5391, |
| "step": 476 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 1.747981629412715e-05, |
| "loss": 0.5446, |
| "step": 477 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 1.7468335736489177e-05, |
| "loss": 0.5105, |
| "step": 478 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 1.7456832877267083e-05, |
| "loss": 0.509, |
| "step": 479 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 1.7445307750810153e-05, |
| "loss": 0.5702, |
| "step": 480 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 1.7433760391534166e-05, |
| "loss": 0.5632, |
| "step": 481 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 1.7422190833921284e-05, |
| "loss": 0.5361, |
| "step": 482 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 1.741059911251997e-05, |
| "loss": 0.5088, |
| "step": 483 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 1.7398985261944857e-05, |
| "loss": 0.5285, |
| "step": 484 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 1.7387349316876668e-05, |
| "loss": 0.5445, |
| "step": 485 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 1.7375691312062102e-05, |
| "loss": 0.5355, |
| "step": 486 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 1.7364011282313732e-05, |
| "loss": 0.5251, |
| "step": 487 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 1.7352309262509894e-05, |
| "loss": 0.5276, |
| "step": 488 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 1.7340585287594605e-05, |
| "loss": 0.5141, |
| "step": 489 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 1.7328839392577422e-05, |
| "loss": 0.5456, |
| "step": 490 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 1.731707161253338e-05, |
| "loss": 0.5172, |
| "step": 491 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 1.730528198260285e-05, |
| "loss": 0.5523, |
| "step": 492 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 1.7293470537991463e-05, |
| "loss": 0.5068, |
| "step": 493 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 1.728163731396998e-05, |
| "loss": 0.536, |
| "step": 494 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 1.7269782345874204e-05, |
| "loss": 0.5346, |
| "step": 495 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 1.7257905669104874e-05, |
| "loss": 0.5307, |
| "step": 496 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 1.7246007319127547e-05, |
| "loss": 0.5427, |
| "step": 497 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 1.72340873314725e-05, |
| "loss": 0.5374, |
| "step": 498 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 1.7222145741734625e-05, |
| "loss": 0.5415, |
| "step": 499 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 1.721018258557333e-05, |
| "loss": 0.5126, |
| "step": 500 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 1.7198197898712402e-05, |
| "loss": 0.5324, |
| "step": 501 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 1.7186191716939946e-05, |
| "loss": 0.5127, |
| "step": 502 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 1.717416407610824e-05, |
| "loss": 0.5529, |
| "step": 503 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 1.7162115012133643e-05, |
| "loss": 0.5143, |
| "step": 504 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 1.7150044560996488e-05, |
| "loss": 0.504, |
| "step": 505 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 1.713795275874098e-05, |
| "loss": 0.5664, |
| "step": 506 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 1.7125839641475074e-05, |
| "loss": 0.5243, |
| "step": 507 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 1.711370524537037e-05, |
| "loss": 0.5007, |
| "step": 508 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 1.7101549606662025e-05, |
| "loss": 0.5262, |
| "step": 509 |
| }, |
| { |
| "epoch": 0.82, |
| "learning_rate": 1.7089372761648617e-05, |
| "loss": 0.5368, |
| "step": 510 |
| }, |
| { |
| "epoch": 0.82, |
| "learning_rate": 1.7077174746692054e-05, |
| "loss": 0.4999, |
| "step": 511 |
| }, |
| { |
| "epoch": 0.82, |
| "learning_rate": 1.7064955598217463e-05, |
| "loss": 0.5435, |
| "step": 512 |
| }, |
| { |
| "epoch": 0.82, |
| "learning_rate": 1.7052715352713076e-05, |
| "loss": 0.5009, |
| "step": 513 |
| }, |
| { |
| "epoch": 0.82, |
| "learning_rate": 1.7040454046730118e-05, |
| "loss": 0.5084, |
| "step": 514 |
| }, |
| { |
| "epoch": 0.82, |
| "learning_rate": 1.7028171716882714e-05, |
| "loss": 0.4916, |
| "step": 515 |
| }, |
| { |
| "epoch": 0.83, |
| "learning_rate": 1.7015868399847768e-05, |
| "loss": 0.5263, |
| "step": 516 |
| }, |
| { |
| "epoch": 0.83, |
| "learning_rate": 1.7003544132364847e-05, |
| "loss": 0.5584, |
| "step": 517 |
| }, |
| { |
| "epoch": 0.83, |
| "learning_rate": 1.6991198951236088e-05, |
| "loss": 0.5203, |
| "step": 518 |
| }, |
| { |
| "epoch": 0.83, |
| "learning_rate": 1.6978832893326074e-05, |
| "loss": 0.5287, |
| "step": 519 |
| }, |
| { |
| "epoch": 0.83, |
| "learning_rate": 1.696644599556173e-05, |
| "loss": 0.4963, |
| "step": 520 |
| }, |
| { |
| "epoch": 0.83, |
| "learning_rate": 1.6954038294932215e-05, |
| "loss": 0.5203, |
| "step": 521 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 1.6941609828488806e-05, |
| "loss": 0.5368, |
| "step": 522 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 1.692916063334479e-05, |
| "loss": 0.5074, |
| "step": 523 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 1.691669074667535e-05, |
| "loss": 0.5746, |
| "step": 524 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 1.690420020571747e-05, |
| "loss": 0.5415, |
| "step": 525 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 1.689168904776979e-05, |
| "loss": 0.5364, |
| "step": 526 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 1.6879157310192537e-05, |
| "loss": 0.569, |
| "step": 527 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 1.686660503040737e-05, |
| "loss": 0.5123, |
| "step": 528 |
| }, |
| { |
| "epoch": 0.85, |
| "learning_rate": 1.685403224589731e-05, |
| "loss": 0.5165, |
| "step": 529 |
| }, |
| { |
| "epoch": 0.85, |
| "learning_rate": 1.6841438994206597e-05, |
| "loss": 0.5026, |
| "step": 530 |
| }, |
| { |
| "epoch": 0.85, |
| "learning_rate": 1.6828825312940594e-05, |
| "loss": 0.5151, |
| "step": 531 |
| }, |
| { |
| "epoch": 0.85, |
| "learning_rate": 1.6816191239765668e-05, |
| "loss": 0.5475, |
| "step": 532 |
| }, |
| { |
| "epoch": 0.85, |
| "learning_rate": 1.6803536812409077e-05, |
| "loss": 0.5346, |
| "step": 533 |
| }, |
| { |
| "epoch": 0.85, |
| "learning_rate": 1.6790862068658863e-05, |
| "loss": 0.5105, |
| "step": 534 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 1.6778167046363735e-05, |
| "loss": 0.5191, |
| "step": 535 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 1.6765451783432953e-05, |
| "loss": 0.533, |
| "step": 536 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 1.675271631783623e-05, |
| "loss": 0.5268, |
| "step": 537 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 1.6739960687603592e-05, |
| "loss": 0.532, |
| "step": 538 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 1.672718493082529e-05, |
| "loss": 0.5264, |
| "step": 539 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 1.671438908565167e-05, |
| "loss": 0.5355, |
| "step": 540 |
| }, |
| { |
| "epoch": 0.87, |
| "learning_rate": 1.6701573190293076e-05, |
| "loss": 0.5017, |
| "step": 541 |
| }, |
| { |
| "epoch": 0.87, |
| "learning_rate": 1.6688737283019708e-05, |
| "loss": 0.4852, |
| "step": 542 |
| }, |
| { |
| "epoch": 0.87, |
| "learning_rate": 1.667588140216154e-05, |
| "loss": 0.522, |
| "step": 543 |
| }, |
| { |
| "epoch": 0.87, |
| "learning_rate": 1.6663005586108175e-05, |
| "loss": 0.5113, |
| "step": 544 |
| }, |
| { |
| "epoch": 0.87, |
| "learning_rate": 1.6650109873308763e-05, |
| "loss": 0.5307, |
| "step": 545 |
| }, |
| { |
| "epoch": 0.87, |
| "learning_rate": 1.663719430227186e-05, |
| "loss": 0.5167, |
| "step": 546 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 1.6624258911565312e-05, |
| "loss": 0.5402, |
| "step": 547 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 1.661130373981617e-05, |
| "loss": 0.5653, |
| "step": 548 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 1.6598328825710536e-05, |
| "loss": 0.5722, |
| "step": 549 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 1.6585334207993475e-05, |
| "loss": 0.5139, |
| "step": 550 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 1.6572319925468892e-05, |
| "loss": 0.514, |
| "step": 551 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 1.65592860169994e-05, |
| "loss": 0.5645, |
| "step": 552 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 1.654623252150624e-05, |
| "loss": 0.5548, |
| "step": 553 |
| }, |
| { |
| "epoch": 0.89, |
| "learning_rate": 1.6533159477969122e-05, |
| "loss": 0.5154, |
| "step": 554 |
| }, |
| { |
| "epoch": 0.89, |
| "learning_rate": 1.6520066925426146e-05, |
| "loss": 0.5345, |
| "step": 555 |
| }, |
| { |
| "epoch": 0.89, |
| "learning_rate": 1.6506954902973657e-05, |
| "loss": 0.5602, |
| "step": 556 |
| }, |
| { |
| "epoch": 0.89, |
| "learning_rate": 1.6493823449766137e-05, |
| "loss": 0.504, |
| "step": 557 |
| }, |
| { |
| "epoch": 0.89, |
| "learning_rate": 1.648067260501611e-05, |
| "loss": 0.5022, |
| "step": 558 |
| }, |
| { |
| "epoch": 0.89, |
| "learning_rate": 1.6467502407993995e-05, |
| "loss": 0.5361, |
| "step": 559 |
| }, |
| { |
| "epoch": 0.9, |
| "learning_rate": 1.6454312898027992e-05, |
| "loss": 0.5284, |
| "step": 560 |
| }, |
| { |
| "epoch": 0.9, |
| "learning_rate": 1.644110411450398e-05, |
| "loss": 0.5022, |
| "step": 561 |
| }, |
| { |
| "epoch": 0.9, |
| "learning_rate": 1.6427876096865394e-05, |
| "loss": 0.4995, |
| "step": 562 |
| }, |
| { |
| "epoch": 0.9, |
| "learning_rate": 1.6414628884613106e-05, |
| "loss": 0.5703, |
| "step": 563 |
| }, |
| { |
| "epoch": 0.9, |
| "learning_rate": 1.6401362517305296e-05, |
| "loss": 0.5416, |
| "step": 564 |
| }, |
| { |
| "epoch": 0.9, |
| "learning_rate": 1.6388077034557355e-05, |
| "loss": 0.5412, |
| "step": 565 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 1.637477247604175e-05, |
| "loss": 0.5487, |
| "step": 566 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 1.6361448881487913e-05, |
| "loss": 0.5353, |
| "step": 567 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 1.6348106290682117e-05, |
| "loss": 0.5305, |
| "step": 568 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 1.6334744743467366e-05, |
| "loss": 0.4976, |
| "step": 569 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 1.6321364279743267e-05, |
| "loss": 0.5199, |
| "step": 570 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 1.6307964939465914e-05, |
| "loss": 0.5241, |
| "step": 571 |
| }, |
| { |
| "epoch": 0.92, |
| "learning_rate": 1.6294546762647775e-05, |
| "loss": 0.5196, |
| "step": 572 |
| }, |
| { |
| "epoch": 0.92, |
| "learning_rate": 1.628110978935756e-05, |
| "loss": 0.4969, |
| "step": 573 |
| }, |
| { |
| "epoch": 0.92, |
| "learning_rate": 1.626765405972011e-05, |
| "loss": 0.5256, |
| "step": 574 |
| }, |
| { |
| "epoch": 0.92, |
| "learning_rate": 1.625417961391628e-05, |
| "loss": 0.5346, |
| "step": 575 |
| }, |
| { |
| "epoch": 0.92, |
| "learning_rate": 1.6240686492182806e-05, |
| "loss": 0.5145, |
| "step": 576 |
| }, |
| { |
| "epoch": 0.92, |
| "learning_rate": 1.62271747348122e-05, |
| "loss": 0.5431, |
| "step": 577 |
| }, |
| { |
| "epoch": 0.92, |
| "learning_rate": 1.621364438215262e-05, |
| "loss": 0.5255, |
| "step": 578 |
| }, |
| { |
| "epoch": 0.93, |
| "learning_rate": 1.6200095474607753e-05, |
| "loss": 0.5511, |
| "step": 579 |
| }, |
| { |
| "epoch": 0.93, |
| "learning_rate": 1.6186528052636692e-05, |
| "loss": 0.539, |
| "step": 580 |
| }, |
| { |
| "epoch": 0.93, |
| "learning_rate": 1.6172942156753822e-05, |
| "loss": 0.5566, |
| "step": 581 |
| }, |
| { |
| "epoch": 0.93, |
| "learning_rate": 1.6159337827528686e-05, |
| "loss": 0.5197, |
| "step": 582 |
| }, |
| { |
| "epoch": 0.93, |
| "learning_rate": 1.614571510558588e-05, |
| "loss": 0.5419, |
| "step": 583 |
| }, |
| { |
| "epoch": 0.93, |
| "learning_rate": 1.6132074031604917e-05, |
| "loss": 0.5347, |
| "step": 584 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 1.6118414646320115e-05, |
| "loss": 0.5488, |
| "step": 585 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 1.6104736990520468e-05, |
| "loss": 0.5262, |
| "step": 586 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 1.6091041105049542e-05, |
| "loss": 0.4871, |
| "step": 587 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 1.6077327030805318e-05, |
| "loss": 0.4873, |
| "step": 588 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 1.6063594808740112e-05, |
| "loss": 0.5156, |
| "step": 589 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 1.604984447986042e-05, |
| "loss": 0.524, |
| "step": 590 |
| }, |
| { |
| "epoch": 0.95, |
| "learning_rate": 1.6036076085226813e-05, |
| "loss": 0.5355, |
| "step": 591 |
| }, |
| { |
| "epoch": 0.95, |
| "learning_rate": 1.602228966595381e-05, |
| "loss": 0.5181, |
| "step": 592 |
| }, |
| { |
| "epoch": 0.95, |
| "learning_rate": 1.6008485263209742e-05, |
| "loss": 0.5098, |
| "step": 593 |
| }, |
| { |
| "epoch": 0.95, |
| "learning_rate": 1.599466291821666e-05, |
| "loss": 0.5164, |
| "step": 594 |
| }, |
| { |
| "epoch": 0.95, |
| "learning_rate": 1.598082267225018e-05, |
| "loss": 0.555, |
| "step": 595 |
| }, |
| { |
| "epoch": 0.95, |
| "learning_rate": 1.596696456663938e-05, |
| "loss": 0.5361, |
| "step": 596 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 1.595308864276666e-05, |
| "loss": 0.5352, |
| "step": 597 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 1.5939194942067647e-05, |
| "loss": 0.5321, |
| "step": 598 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 1.592528350603103e-05, |
| "loss": 0.5668, |
| "step": 599 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 1.5911354376198468e-05, |
| "loss": 0.5406, |
| "step": 600 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 1.5897407594164468e-05, |
| "loss": 0.5328, |
| "step": 601 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 1.5883443201576225e-05, |
| "loss": 0.5194, |
| "step": 602 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 1.586946124013354e-05, |
| "loss": 0.5382, |
| "step": 603 |
| }, |
| { |
| "epoch": 0.97, |
| "learning_rate": 1.585546175158868e-05, |
| "loss": 0.5389, |
| "step": 604 |
| }, |
| { |
| "epoch": 0.97, |
| "learning_rate": 1.5841444777746232e-05, |
| "loss": 0.5091, |
| "step": 605 |
| }, |
| { |
| "epoch": 0.97, |
| "learning_rate": 1.582741036046301e-05, |
| "loss": 0.5195, |
| "step": 606 |
| }, |
| { |
| "epoch": 0.97, |
| "learning_rate": 1.5813358541647915e-05, |
| "loss": 0.5129, |
| "step": 607 |
| }, |
| { |
| "epoch": 0.97, |
| "learning_rate": 1.5799289363261815e-05, |
| "loss": 0.5357, |
| "step": 608 |
| }, |
| { |
| "epoch": 0.97, |
| "learning_rate": 1.578520286731741e-05, |
| "loss": 0.5363, |
| "step": 609 |
| }, |
| { |
| "epoch": 0.98, |
| "learning_rate": 1.5771099095879108e-05, |
| "loss": 0.5564, |
| "step": 610 |
| }, |
| { |
| "epoch": 0.98, |
| "learning_rate": 1.575697809106292e-05, |
| "loss": 0.5361, |
| "step": 611 |
| }, |
| { |
| "epoch": 0.98, |
| "learning_rate": 1.5742839895036305e-05, |
| "loss": 0.5414, |
| "step": 612 |
| }, |
| { |
| "epoch": 0.98, |
| "learning_rate": 1.5728684550018066e-05, |
| "loss": 0.5389, |
| "step": 613 |
| }, |
| { |
| "epoch": 0.98, |
| "learning_rate": 1.571451209827821e-05, |
| "loss": 0.4878, |
| "step": 614 |
| }, |
| { |
| "epoch": 0.98, |
| "learning_rate": 1.570032258213783e-05, |
| "loss": 0.5199, |
| "step": 615 |
| }, |
| { |
| "epoch": 0.99, |
| "learning_rate": 1.5686116043968975e-05, |
| "loss": 0.524, |
| "step": 616 |
| }, |
| { |
| "epoch": 0.99, |
| "learning_rate": 1.5671892526194515e-05, |
| "loss": 0.5213, |
| "step": 617 |
| }, |
| { |
| "epoch": 0.99, |
| "learning_rate": 1.565765207128805e-05, |
| "loss": 0.5296, |
| "step": 618 |
| }, |
| { |
| "epoch": 0.99, |
| "learning_rate": 1.564339472177373e-05, |
| "loss": 0.5252, |
| "step": 619 |
| }, |
| { |
| "epoch": 0.99, |
| "learning_rate": 1.5629120520226163e-05, |
| "loss": 0.5294, |
| "step": 620 |
| }, |
| { |
| "epoch": 0.99, |
| "learning_rate": 1.561482950927029e-05, |
| "loss": 0.5446, |
| "step": 621 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 1.560052173158123e-05, |
| "loss": 0.5176, |
| "step": 622 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 1.5586197229884185e-05, |
| "loss": 0.527, |
| "step": 623 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 1.5571856046954284e-05, |
| "loss": 0.5231, |
| "step": 624 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 1.5557498225616488e-05, |
| "loss": 0.5215, |
| "step": 625 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 1.5543123808745418e-05, |
| "loss": 0.4602, |
| "step": 626 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 1.5528732839265272e-05, |
| "loss": 0.4252, |
| "step": 627 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 1.5514325360149668e-05, |
| "loss": 0.4617, |
| "step": 628 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 1.549990141442153e-05, |
| "loss": 0.463, |
| "step": 629 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 1.5485461045152937e-05, |
| "loss": 0.4323, |
| "step": 630 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 1.5471004295465034e-05, |
| "loss": 0.4598, |
| "step": 631 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 1.5456531208527868e-05, |
| "loss": 0.4401, |
| "step": 632 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 1.5442041827560274e-05, |
| "loss": 0.457, |
| "step": 633 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 1.542753619582974e-05, |
| "loss": 0.4742, |
| "step": 634 |
| }, |
| { |
| "epoch": 1.02, |
| "learning_rate": 1.5413014356652287e-05, |
| "loss": 0.423, |
| "step": 635 |
| }, |
| { |
| "epoch": 1.02, |
| "learning_rate": 1.5398476353392323e-05, |
| "loss": 0.4665, |
| "step": 636 |
| }, |
| { |
| "epoch": 1.02, |
| "learning_rate": 1.538392222946255e-05, |
| "loss": 0.4347, |
| "step": 637 |
| }, |
| { |
| "epoch": 1.02, |
| "learning_rate": 1.5369352028323773e-05, |
| "loss": 0.4784, |
| "step": 638 |
| }, |
| { |
| "epoch": 1.02, |
| "learning_rate": 1.5354765793484834e-05, |
| "loss": 0.4647, |
| "step": 639 |
| }, |
| { |
| "epoch": 1.02, |
| "learning_rate": 1.534016356850244e-05, |
| "loss": 0.4541, |
| "step": 640 |
| }, |
| { |
| "epoch": 1.03, |
| "learning_rate": 1.5325545396981053e-05, |
| "loss": 0.4706, |
| "step": 641 |
| }, |
| { |
| "epoch": 1.03, |
| "learning_rate": 1.531091132257275e-05, |
| "loss": 0.4632, |
| "step": 642 |
| }, |
| { |
| "epoch": 1.03, |
| "learning_rate": 1.5296261388977107e-05, |
| "loss": 0.4521, |
| "step": 643 |
| }, |
| { |
| "epoch": 1.03, |
| "learning_rate": 1.528159563994104e-05, |
| "loss": 0.4687, |
| "step": 644 |
| }, |
| { |
| "epoch": 1.03, |
| "learning_rate": 1.52669141192587e-05, |
| "loss": 0.4488, |
| "step": 645 |
| }, |
| { |
| "epoch": 1.03, |
| "learning_rate": 1.5252216870771345e-05, |
| "loss": 0.4784, |
| "step": 646 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 1.5237503938367186e-05, |
| "loss": 0.4386, |
| "step": 647 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 1.5222775365981272e-05, |
| "loss": 0.4474, |
| "step": 648 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 1.5208031197595357e-05, |
| "loss": 0.4449, |
| "step": 649 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 1.5193271477237761e-05, |
| "loss": 0.4572, |
| "step": 650 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 1.5178496248983254e-05, |
| "loss": 0.464, |
| "step": 651 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 1.5163705556952912e-05, |
| "loss": 0.4575, |
| "step": 652 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 1.5148899445313983e-05, |
| "loss": 0.4702, |
| "step": 653 |
| }, |
| { |
| "epoch": 1.05, |
| "learning_rate": 1.5134077958279764e-05, |
| "loss": 0.4449, |
| "step": 654 |
| }, |
| { |
| "epoch": 1.05, |
| "learning_rate": 1.5119241140109466e-05, |
| "loss": 0.4476, |
| "step": 655 |
| }, |
| { |
| "epoch": 1.05, |
| "learning_rate": 1.5104389035108078e-05, |
| "loss": 0.4599, |
| "step": 656 |
| }, |
| { |
| "epoch": 1.05, |
| "learning_rate": 1.5089521687626243e-05, |
| "loss": 0.4783, |
| "step": 657 |
| }, |
| { |
| "epoch": 1.05, |
| "learning_rate": 1.5074639142060119e-05, |
| "loss": 0.456, |
| "step": 658 |
| }, |
| { |
| "epoch": 1.05, |
| "learning_rate": 1.505974144285124e-05, |
| "loss": 0.4319, |
| "step": 659 |
| }, |
| { |
| "epoch": 1.06, |
| "learning_rate": 1.50448286344864e-05, |
| "loss": 0.4476, |
| "step": 660 |
| }, |
| { |
| "epoch": 1.06, |
| "learning_rate": 1.5029900761497507e-05, |
| "loss": 0.4692, |
| "step": 661 |
| }, |
| { |
| "epoch": 1.06, |
| "learning_rate": 1.501495786846146e-05, |
| "loss": 0.4639, |
| "step": 662 |
| }, |
| { |
| "epoch": 1.06, |
| "learning_rate": 1.5000000000000002e-05, |
| "loss": 0.4525, |
| "step": 663 |
| }, |
| { |
| "epoch": 1.06, |
| "learning_rate": 1.4985027200779599e-05, |
| "loss": 0.479, |
| "step": 664 |
| }, |
| { |
| "epoch": 1.06, |
| "learning_rate": 1.4970039515511303e-05, |
| "loss": 0.464, |
| "step": 665 |
| }, |
| { |
| "epoch": 1.07, |
| "learning_rate": 1.4955036988950617e-05, |
| "loss": 0.455, |
| "step": 666 |
| }, |
| { |
| "epoch": 1.07, |
| "learning_rate": 1.4940019665897363e-05, |
| "loss": 0.4406, |
| "step": 667 |
| }, |
| { |
| "epoch": 1.07, |
| "learning_rate": 1.4924987591195548e-05, |
| "loss": 0.4372, |
| "step": 668 |
| }, |
| { |
| "epoch": 1.07, |
| "learning_rate": 1.4909940809733223e-05, |
| "loss": 0.4632, |
| "step": 669 |
| }, |
| { |
| "epoch": 1.07, |
| "learning_rate": 1.489487936644237e-05, |
| "loss": 0.4378, |
| "step": 670 |
| }, |
| { |
| "epoch": 1.07, |
| "learning_rate": 1.4879803306298736e-05, |
| "loss": 0.445, |
| "step": 671 |
| }, |
| { |
| "epoch": 1.08, |
| "learning_rate": 1.4864712674321733e-05, |
| "loss": 0.4371, |
| "step": 672 |
| }, |
| { |
| "epoch": 1.08, |
| "learning_rate": 1.4849607515574276e-05, |
| "loss": 0.425, |
| "step": 673 |
| }, |
| { |
| "epoch": 1.08, |
| "learning_rate": 1.4834487875162657e-05, |
| "loss": 0.4426, |
| "step": 674 |
| }, |
| { |
| "epoch": 1.08, |
| "learning_rate": 1.4819353798236427e-05, |
| "loss": 0.4427, |
| "step": 675 |
| }, |
| { |
| "epoch": 1.08, |
| "learning_rate": 1.4804205329988226e-05, |
| "loss": 0.4662, |
| "step": 676 |
| }, |
| { |
| "epoch": 1.08, |
| "learning_rate": 1.4789042515653687e-05, |
| "loss": 0.4678, |
| "step": 677 |
| }, |
| { |
| "epoch": 1.08, |
| "learning_rate": 1.477386540051127e-05, |
| "loss": 0.4439, |
| "step": 678 |
| }, |
| { |
| "epoch": 1.09, |
| "learning_rate": 1.4758674029882152e-05, |
| "loss": 0.4533, |
| "step": 679 |
| }, |
| { |
| "epoch": 1.09, |
| "learning_rate": 1.4743468449130065e-05, |
| "loss": 0.4471, |
| "step": 680 |
| }, |
| { |
| "epoch": 1.09, |
| "learning_rate": 1.4728248703661183e-05, |
| "loss": 0.4469, |
| "step": 681 |
| }, |
| { |
| "epoch": 1.09, |
| "learning_rate": 1.4713014838923975e-05, |
| "loss": 0.4788, |
| "step": 682 |
| }, |
| { |
| "epoch": 1.09, |
| "learning_rate": 1.4697766900409076e-05, |
| "loss": 0.4593, |
| "step": 683 |
| }, |
| { |
| "epoch": 1.09, |
| "learning_rate": 1.4682504933649144e-05, |
| "loss": 0.4921, |
| "step": 684 |
| }, |
| { |
| "epoch": 1.1, |
| "learning_rate": 1.466722898421873e-05, |
| "loss": 0.45, |
| "step": 685 |
| }, |
| { |
| "epoch": 1.1, |
| "learning_rate": 1.4651939097734132e-05, |
| "loss": 0.454, |
| "step": 686 |
| }, |
| { |
| "epoch": 1.1, |
| "learning_rate": 1.4636635319853274e-05, |
| "loss": 0.4566, |
| "step": 687 |
| }, |
| { |
| "epoch": 1.1, |
| "learning_rate": 1.4621317696275563e-05, |
| "loss": 0.4568, |
| "step": 688 |
| }, |
| { |
| "epoch": 1.1, |
| "learning_rate": 1.4605986272741748e-05, |
| "loss": 0.4702, |
| "step": 689 |
| }, |
| { |
| "epoch": 1.1, |
| "learning_rate": 1.4590641095033786e-05, |
| "loss": 0.4553, |
| "step": 690 |
| }, |
| { |
| "epoch": 1.11, |
| "learning_rate": 1.4575282208974704e-05, |
| "loss": 0.4461, |
| "step": 691 |
| }, |
| { |
| "epoch": 1.11, |
| "learning_rate": 1.4559909660428469e-05, |
| "loss": 0.4728, |
| "step": 692 |
| }, |
| { |
| "epoch": 1.11, |
| "learning_rate": 1.4544523495299843e-05, |
| "loss": 0.4565, |
| "step": 693 |
| }, |
| { |
| "epoch": 1.11, |
| "learning_rate": 1.4529123759534253e-05, |
| "loss": 0.4779, |
| "step": 694 |
| }, |
| { |
| "epoch": 1.11, |
| "learning_rate": 1.4513710499117648e-05, |
| "loss": 0.4699, |
| "step": 695 |
| }, |
| { |
| "epoch": 1.11, |
| "learning_rate": 1.4498283760076362e-05, |
| "loss": 0.4603, |
| "step": 696 |
| }, |
| { |
| "epoch": 1.12, |
| "learning_rate": 1.4482843588476976e-05, |
| "loss": 0.4428, |
| "step": 697 |
| }, |
| { |
| "epoch": 1.12, |
| "learning_rate": 1.4467390030426187e-05, |
| "loss": 0.473, |
| "step": 698 |
| }, |
| { |
| "epoch": 1.12, |
| "learning_rate": 1.445192313207067e-05, |
| "loss": 0.4584, |
| "step": 699 |
| }, |
| { |
| "epoch": 1.12, |
| "learning_rate": 1.443644293959693e-05, |
| "loss": 0.4376, |
| "step": 700 |
| }, |
| { |
| "epoch": 1.12, |
| "learning_rate": 1.4420949499231172e-05, |
| "loss": 0.4468, |
| "step": 701 |
| }, |
| { |
| "epoch": 1.12, |
| "learning_rate": 1.4405442857239151e-05, |
| "loss": 0.4407, |
| "step": 702 |
| }, |
| { |
| "epoch": 1.12, |
| "learning_rate": 1.4389923059926064e-05, |
| "loss": 0.432, |
| "step": 703 |
| }, |
| { |
| "epoch": 1.13, |
| "learning_rate": 1.437439015363638e-05, |
| "loss": 0.4939, |
| "step": 704 |
| }, |
| { |
| "epoch": 1.13, |
| "learning_rate": 1.4358844184753713e-05, |
| "loss": 0.4575, |
| "step": 705 |
| }, |
| { |
| "epoch": 1.13, |
| "learning_rate": 1.4343285199700685e-05, |
| "loss": 0.4712, |
| "step": 706 |
| }, |
| { |
| "epoch": 1.13, |
| "learning_rate": 1.432771324493879e-05, |
| "loss": 0.4164, |
| "step": 707 |
| }, |
| { |
| "epoch": 1.13, |
| "learning_rate": 1.4312128366968244e-05, |
| "loss": 0.4266, |
| "step": 708 |
| }, |
| { |
| "epoch": 1.13, |
| "learning_rate": 1.4296530612327864e-05, |
| "loss": 0.4279, |
| "step": 709 |
| }, |
| { |
| "epoch": 1.14, |
| "learning_rate": 1.428092002759491e-05, |
| "loss": 0.4587, |
| "step": 710 |
| }, |
| { |
| "epoch": 1.14, |
| "learning_rate": 1.4265296659384956e-05, |
| "loss": 0.4249, |
| "step": 711 |
| }, |
| { |
| "epoch": 1.14, |
| "learning_rate": 1.4249660554351752e-05, |
| "loss": 0.4775, |
| "step": 712 |
| }, |
| { |
| "epoch": 1.14, |
| "learning_rate": 1.4234011759187084e-05, |
| "loss": 0.4924, |
| "step": 713 |
| }, |
| { |
| "epoch": 1.14, |
| "learning_rate": 1.4218350320620625e-05, |
| "loss": 0.4824, |
| "step": 714 |
| }, |
| { |
| "epoch": 1.14, |
| "learning_rate": 1.4202676285419811e-05, |
| "loss": 0.4556, |
| "step": 715 |
| }, |
| { |
| "epoch": 1.15, |
| "learning_rate": 1.4186989700389689e-05, |
| "loss": 0.4425, |
| "step": 716 |
| }, |
| { |
| "epoch": 1.15, |
| "learning_rate": 1.4171290612372781e-05, |
| "loss": 0.4635, |
| "step": 717 |
| }, |
| { |
| "epoch": 1.15, |
| "learning_rate": 1.4155579068248951e-05, |
| "loss": 0.4551, |
| "step": 718 |
| }, |
| { |
| "epoch": 1.15, |
| "learning_rate": 1.4139855114935253e-05, |
| "loss": 0.4745, |
| "step": 719 |
| }, |
| { |
| "epoch": 1.15, |
| "learning_rate": 1.4124118799385797e-05, |
| "loss": 0.452, |
| "step": 720 |
| }, |
| { |
| "epoch": 1.15, |
| "learning_rate": 1.410837016859161e-05, |
| "loss": 0.4689, |
| "step": 721 |
| }, |
| { |
| "epoch": 1.16, |
| "learning_rate": 1.4092609269580498e-05, |
| "loss": 0.4422, |
| "step": 722 |
| }, |
| { |
| "epoch": 1.16, |
| "learning_rate": 1.4076836149416889e-05, |
| "loss": 0.454, |
| "step": 723 |
| }, |
| { |
| "epoch": 1.16, |
| "learning_rate": 1.4061050855201723e-05, |
| "loss": 0.447, |
| "step": 724 |
| }, |
| { |
| "epoch": 1.16, |
| "learning_rate": 1.4045253434072278e-05, |
| "loss": 0.4757, |
| "step": 725 |
| }, |
| { |
| "epoch": 1.16, |
| "learning_rate": 1.4029443933202059e-05, |
| "loss": 0.4748, |
| "step": 726 |
| }, |
| { |
| "epoch": 1.16, |
| "learning_rate": 1.4013622399800628e-05, |
| "loss": 0.432, |
| "step": 727 |
| }, |
| { |
| "epoch": 1.16, |
| "learning_rate": 1.399778888111349e-05, |
| "loss": 0.4319, |
| "step": 728 |
| }, |
| { |
| "epoch": 1.17, |
| "learning_rate": 1.3981943424421932e-05, |
| "loss": 0.434, |
| "step": 729 |
| }, |
| { |
| "epoch": 1.17, |
| "learning_rate": 1.3966086077042891e-05, |
| "loss": 0.4334, |
| "step": 730 |
| }, |
| { |
| "epoch": 1.17, |
| "learning_rate": 1.3950216886328818e-05, |
| "loss": 0.4674, |
| "step": 731 |
| }, |
| { |
| "epoch": 1.17, |
| "learning_rate": 1.3934335899667526e-05, |
| "loss": 0.4335, |
| "step": 732 |
| }, |
| { |
| "epoch": 1.17, |
| "learning_rate": 1.3918443164482048e-05, |
| "loss": 0.4507, |
| "step": 733 |
| }, |
| { |
| "epoch": 1.17, |
| "learning_rate": 1.3902538728230502e-05, |
| "loss": 0.4602, |
| "step": 734 |
| }, |
| { |
| "epoch": 1.18, |
| "learning_rate": 1.3886622638405953e-05, |
| "loss": 0.441, |
| "step": 735 |
| }, |
| { |
| "epoch": 1.18, |
| "learning_rate": 1.387069494253626e-05, |
| "loss": 0.4386, |
| "step": 736 |
| }, |
| { |
| "epoch": 1.18, |
| "learning_rate": 1.3854755688183941e-05, |
| "loss": 0.4482, |
| "step": 737 |
| }, |
| { |
| "epoch": 1.18, |
| "learning_rate": 1.3838804922946027e-05, |
| "loss": 0.4478, |
| "step": 738 |
| }, |
| { |
| "epoch": 1.18, |
| "learning_rate": 1.3822842694453923e-05, |
| "loss": 0.4745, |
| "step": 739 |
| }, |
| { |
| "epoch": 1.18, |
| "learning_rate": 1.380686905037327e-05, |
| "loss": 0.4612, |
| "step": 740 |
| }, |
| { |
| "epoch": 1.19, |
| "learning_rate": 1.3790884038403796e-05, |
| "loss": 0.4518, |
| "step": 741 |
| }, |
| { |
| "epoch": 1.19, |
| "learning_rate": 1.3774887706279165e-05, |
| "loss": 0.4623, |
| "step": 742 |
| }, |
| { |
| "epoch": 1.19, |
| "learning_rate": 1.375888010176686e-05, |
| "loss": 0.4597, |
| "step": 743 |
| }, |
| { |
| "epoch": 1.19, |
| "learning_rate": 1.374286127266801e-05, |
| "loss": 0.4711, |
| "step": 744 |
| }, |
| { |
| "epoch": 1.19, |
| "learning_rate": 1.3726831266817278e-05, |
| "loss": 0.4621, |
| "step": 745 |
| }, |
| { |
| "epoch": 1.19, |
| "learning_rate": 1.3710790132082693e-05, |
| "loss": 0.4535, |
| "step": 746 |
| }, |
| { |
| "epoch": 1.2, |
| "learning_rate": 1.3694737916365517e-05, |
| "loss": 0.4907, |
| "step": 747 |
| }, |
| { |
| "epoch": 1.2, |
| "learning_rate": 1.3678674667600102e-05, |
| "loss": 0.4597, |
| "step": 748 |
| }, |
| { |
| "epoch": 1.2, |
| "learning_rate": 1.3662600433753746e-05, |
| "loss": 0.469, |
| "step": 749 |
| }, |
| { |
| "epoch": 1.2, |
| "learning_rate": 1.3646515262826551e-05, |
| "loss": 0.4583, |
| "step": 750 |
| }, |
| { |
| "epoch": 1.2, |
| "learning_rate": 1.3630419202851287e-05, |
| "loss": 0.4281, |
| "step": 751 |
| }, |
| { |
| "epoch": 1.2, |
| "learning_rate": 1.3614312301893222e-05, |
| "loss": 0.4344, |
| "step": 752 |
| }, |
| { |
| "epoch": 1.2, |
| "learning_rate": 1.3598194608050011e-05, |
| "loss": 0.4765, |
| "step": 753 |
| }, |
| { |
| "epoch": 1.21, |
| "learning_rate": 1.3582066169451535e-05, |
| "loss": 0.4554, |
| "step": 754 |
| }, |
| { |
| "epoch": 1.21, |
| "learning_rate": 1.3565927034259757e-05, |
| "loss": 0.4546, |
| "step": 755 |
| }, |
| { |
| "epoch": 1.21, |
| "learning_rate": 1.354977725066859e-05, |
| "loss": 0.4536, |
| "step": 756 |
| }, |
| { |
| "epoch": 1.21, |
| "learning_rate": 1.3533616866903736e-05, |
| "loss": 0.4275, |
| "step": 757 |
| }, |
| { |
| "epoch": 1.21, |
| "learning_rate": 1.351744593122255e-05, |
| "loss": 0.4532, |
| "step": 758 |
| }, |
| { |
| "epoch": 1.21, |
| "learning_rate": 1.3501264491913909e-05, |
| "loss": 0.4544, |
| "step": 759 |
| }, |
| { |
| "epoch": 1.22, |
| "learning_rate": 1.3485072597298038e-05, |
| "loss": 0.4215, |
| "step": 760 |
| }, |
| { |
| "epoch": 1.22, |
| "learning_rate": 1.3468870295726399e-05, |
| "loss": 0.4275, |
| "step": 761 |
| }, |
| { |
| "epoch": 1.22, |
| "learning_rate": 1.3452657635581521e-05, |
| "loss": 0.4339, |
| "step": 762 |
| }, |
| { |
| "epoch": 1.22, |
| "learning_rate": 1.3436434665276865e-05, |
| "loss": 0.4307, |
| "step": 763 |
| }, |
| { |
| "epoch": 1.22, |
| "learning_rate": 1.342020143325669e-05, |
| "loss": 0.4071, |
| "step": 764 |
| }, |
| { |
| "epoch": 1.22, |
| "learning_rate": 1.3403957987995884e-05, |
| "loss": 0.4491, |
| "step": 765 |
| }, |
| { |
| "epoch": 1.23, |
| "learning_rate": 1.3387704377999842e-05, |
| "loss": 0.4646, |
| "step": 766 |
| }, |
| { |
| "epoch": 1.23, |
| "learning_rate": 1.3371440651804313e-05, |
| "loss": 0.4807, |
| "step": 767 |
| }, |
| { |
| "epoch": 1.23, |
| "learning_rate": 1.335516685797525e-05, |
| "loss": 0.4576, |
| "step": 768 |
| }, |
| { |
| "epoch": 1.23, |
| "learning_rate": 1.3338883045108674e-05, |
| "loss": 0.4761, |
| "step": 769 |
| }, |
| { |
| "epoch": 1.23, |
| "learning_rate": 1.3322589261830517e-05, |
| "loss": 0.4531, |
| "step": 770 |
| }, |
| { |
| "epoch": 1.23, |
| "learning_rate": 1.3306285556796494e-05, |
| "loss": 0.4383, |
| "step": 771 |
| }, |
| { |
| "epoch": 1.24, |
| "learning_rate": 1.328997197869194e-05, |
| "loss": 0.4565, |
| "step": 772 |
| }, |
| { |
| "epoch": 1.24, |
| "learning_rate": 1.327364857623168e-05, |
| "loss": 0.456, |
| "step": 773 |
| }, |
| { |
| "epoch": 1.24, |
| "learning_rate": 1.3257315398159865e-05, |
| "loss": 0.48, |
| "step": 774 |
| }, |
| { |
| "epoch": 1.24, |
| "learning_rate": 1.3240972493249846e-05, |
| "loss": 0.4235, |
| "step": 775 |
| }, |
| { |
| "epoch": 1.24, |
| "learning_rate": 1.3224619910304019e-05, |
| "loss": 0.453, |
| "step": 776 |
| }, |
| { |
| "epoch": 1.24, |
| "learning_rate": 1.3208257698153677e-05, |
| "loss": 0.4564, |
| "step": 777 |
| }, |
| { |
| "epoch": 1.24, |
| "learning_rate": 1.3191885905658873e-05, |
| "loss": 0.4446, |
| "step": 778 |
| }, |
| { |
| "epoch": 1.25, |
| "learning_rate": 1.3175504581708261e-05, |
| "loss": 0.457, |
| "step": 779 |
| }, |
| { |
| "epoch": 1.25, |
| "learning_rate": 1.3159113775218963e-05, |
| "loss": 0.464, |
| "step": 780 |
| }, |
| { |
| "epoch": 1.25, |
| "learning_rate": 1.3142713535136413e-05, |
| "loss": 0.4641, |
| "step": 781 |
| }, |
| { |
| "epoch": 1.25, |
| "learning_rate": 1.3126303910434215e-05, |
| "loss": 0.4705, |
| "step": 782 |
| }, |
| { |
| "epoch": 1.25, |
| "learning_rate": 1.3109884950114007e-05, |
| "loss": 0.4573, |
| "step": 783 |
| }, |
| { |
| "epoch": 1.25, |
| "learning_rate": 1.309345670320529e-05, |
| "loss": 0.444, |
| "step": 784 |
| }, |
| { |
| "epoch": 1.26, |
| "learning_rate": 1.3077019218765306e-05, |
| "loss": 0.4503, |
| "step": 785 |
| }, |
| { |
| "epoch": 1.26, |
| "learning_rate": 1.3060572545878875e-05, |
| "loss": 0.4484, |
| "step": 786 |
| }, |
| { |
| "epoch": 1.26, |
| "learning_rate": 1.3044116733658261e-05, |
| "loss": 0.4614, |
| "step": 787 |
| }, |
| { |
| "epoch": 1.26, |
| "learning_rate": 1.302765183124302e-05, |
| "loss": 0.4261, |
| "step": 788 |
| }, |
| { |
| "epoch": 1.26, |
| "learning_rate": 1.3011177887799846e-05, |
| "loss": 0.4595, |
| "step": 789 |
| }, |
| { |
| "epoch": 1.26, |
| "learning_rate": 1.2994694952522435e-05, |
| "loss": 0.4501, |
| "step": 790 |
| }, |
| { |
| "epoch": 1.27, |
| "learning_rate": 1.2978203074631335e-05, |
| "loss": 0.4238, |
| "step": 791 |
| }, |
| { |
| "epoch": 1.27, |
| "learning_rate": 1.2961702303373795e-05, |
| "loss": 0.4719, |
| "step": 792 |
| }, |
| { |
| "epoch": 1.27, |
| "learning_rate": 1.2945192688023625e-05, |
| "loss": 0.4619, |
| "step": 793 |
| }, |
| { |
| "epoch": 1.27, |
| "learning_rate": 1.2928674277881041e-05, |
| "loss": 0.4668, |
| "step": 794 |
| }, |
| { |
| "epoch": 1.27, |
| "learning_rate": 1.2912147122272523e-05, |
| "loss": 0.459, |
| "step": 795 |
| }, |
| { |
| "epoch": 1.27, |
| "learning_rate": 1.2895611270550666e-05, |
| "loss": 0.4452, |
| "step": 796 |
| }, |
| { |
| "epoch": 1.28, |
| "learning_rate": 1.287906677209403e-05, |
| "loss": 0.4598, |
| "step": 797 |
| }, |
| { |
| "epoch": 1.28, |
| "learning_rate": 1.2862513676307009e-05, |
| "loss": 0.4743, |
| "step": 798 |
| }, |
| { |
| "epoch": 1.28, |
| "learning_rate": 1.2845952032619651e-05, |
| "loss": 0.4635, |
| "step": 799 |
| }, |
| { |
| "epoch": 1.28, |
| "learning_rate": 1.2829381890487536e-05, |
| "loss": 0.4494, |
| "step": 800 |
| }, |
| { |
| "epoch": 1.28, |
| "learning_rate": 1.2812803299391629e-05, |
| "loss": 0.4439, |
| "step": 801 |
| }, |
| { |
| "epoch": 1.28, |
| "learning_rate": 1.2796216308838116e-05, |
| "loss": 0.4632, |
| "step": 802 |
| }, |
| { |
| "epoch": 1.28, |
| "learning_rate": 1.2779620968358276e-05, |
| "loss": 0.4375, |
| "step": 803 |
| }, |
| { |
| "epoch": 1.29, |
| "learning_rate": 1.2763017327508304e-05, |
| "loss": 0.4364, |
| "step": 804 |
| }, |
| { |
| "epoch": 1.29, |
| "learning_rate": 1.2746405435869198e-05, |
| "loss": 0.4277, |
| "step": 805 |
| }, |
| { |
| "epoch": 1.29, |
| "learning_rate": 1.2729785343046587e-05, |
| "loss": 0.4558, |
| "step": 806 |
| }, |
| { |
| "epoch": 1.29, |
| "learning_rate": 1.271315709867059e-05, |
| "loss": 0.439, |
| "step": 807 |
| }, |
| { |
| "epoch": 1.29, |
| "learning_rate": 1.2696520752395671e-05, |
| "loss": 0.4449, |
| "step": 808 |
| }, |
| { |
| "epoch": 1.29, |
| "learning_rate": 1.2679876353900482e-05, |
| "loss": 0.4631, |
| "step": 809 |
| }, |
| { |
| "epoch": 1.3, |
| "learning_rate": 1.2663223952887724e-05, |
| "loss": 0.4335, |
| "step": 810 |
| }, |
| { |
| "epoch": 1.3, |
| "learning_rate": 1.2646563599083997e-05, |
| "loss": 0.4638, |
| "step": 811 |
| }, |
| { |
| "epoch": 1.3, |
| "learning_rate": 1.2629895342239643e-05, |
| "loss": 0.4528, |
| "step": 812 |
| }, |
| { |
| "epoch": 1.3, |
| "learning_rate": 1.2613219232128608e-05, |
| "loss": 0.4932, |
| "step": 813 |
| }, |
| { |
| "epoch": 1.3, |
| "learning_rate": 1.2596535318548288e-05, |
| "loss": 0.4403, |
| "step": 814 |
| }, |
| { |
| "epoch": 1.3, |
| "learning_rate": 1.2579843651319382e-05, |
| "loss": 0.4786, |
| "step": 815 |
| }, |
| { |
| "epoch": 1.31, |
| "learning_rate": 1.2563144280285742e-05, |
| "loss": 0.4497, |
| "step": 816 |
| }, |
| { |
| "epoch": 1.31, |
| "learning_rate": 1.2546437255314223e-05, |
| "loss": 0.46, |
| "step": 817 |
| }, |
| { |
| "epoch": 1.31, |
| "learning_rate": 1.252972262629454e-05, |
| "loss": 0.4617, |
| "step": 818 |
| }, |
| { |
| "epoch": 1.31, |
| "learning_rate": 1.2513000443139112e-05, |
| "loss": 0.4654, |
| "step": 819 |
| }, |
| { |
| "epoch": 1.31, |
| "learning_rate": 1.2496270755782913e-05, |
| "loss": 0.4681, |
| "step": 820 |
| }, |
| { |
| "epoch": 1.31, |
| "learning_rate": 1.2479533614183334e-05, |
| "loss": 0.4353, |
| "step": 821 |
| }, |
| { |
| "epoch": 1.32, |
| "learning_rate": 1.2462789068320016e-05, |
| "loss": 0.4631, |
| "step": 822 |
| }, |
| { |
| "epoch": 1.32, |
| "learning_rate": 1.2446037168194716e-05, |
| "loss": 0.4433, |
| "step": 823 |
| }, |
| { |
| "epoch": 1.32, |
| "learning_rate": 1.2429277963831147e-05, |
| "loss": 0.4663, |
| "step": 824 |
| }, |
| { |
| "epoch": 1.32, |
| "learning_rate": 1.2412511505274845e-05, |
| "loss": 0.4758, |
| "step": 825 |
| }, |
| { |
| "epoch": 1.32, |
| "learning_rate": 1.2395737842592997e-05, |
| "loss": 0.435, |
| "step": 826 |
| }, |
| { |
| "epoch": 1.32, |
| "learning_rate": 1.23789570258743e-05, |
| "loss": 0.4498, |
| "step": 827 |
| }, |
| { |
| "epoch": 1.32, |
| "learning_rate": 1.2362169105228828e-05, |
| "loss": 0.4576, |
| "step": 828 |
| }, |
| { |
| "epoch": 1.33, |
| "learning_rate": 1.2345374130787855e-05, |
| "loss": 0.4352, |
| "step": 829 |
| }, |
| { |
| "epoch": 1.33, |
| "learning_rate": 1.2328572152703726e-05, |
| "loss": 0.4837, |
| "step": 830 |
| }, |
| { |
| "epoch": 1.33, |
| "learning_rate": 1.23117632211497e-05, |
| "loss": 0.477, |
| "step": 831 |
| }, |
| { |
| "epoch": 1.33, |
| "learning_rate": 1.2294947386319793e-05, |
| "loss": 0.4589, |
| "step": 832 |
| }, |
| { |
| "epoch": 1.33, |
| "learning_rate": 1.2278124698428643e-05, |
| "loss": 0.4583, |
| "step": 833 |
| }, |
| { |
| "epoch": 1.33, |
| "learning_rate": 1.2261295207711347e-05, |
| "loss": 0.4439, |
| "step": 834 |
| }, |
| { |
| "epoch": 1.34, |
| "learning_rate": 1.2244458964423328e-05, |
| "loss": 0.4562, |
| "step": 835 |
| }, |
| { |
| "epoch": 1.34, |
| "learning_rate": 1.2227616018840154e-05, |
| "loss": 0.4396, |
| "step": 836 |
| }, |
| { |
| "epoch": 1.34, |
| "learning_rate": 1.221076642125742e-05, |
| "loss": 0.4482, |
| "step": 837 |
| }, |
| { |
| "epoch": 1.34, |
| "learning_rate": 1.2193910221990582e-05, |
| "loss": 0.4275, |
| "step": 838 |
| }, |
| { |
| "epoch": 1.34, |
| "learning_rate": 1.2177047471374808e-05, |
| "loss": 0.4577, |
| "step": 839 |
| }, |
| { |
| "epoch": 1.34, |
| "learning_rate": 1.2160178219764838e-05, |
| "loss": 0.4611, |
| "step": 840 |
| }, |
| { |
| "epoch": 1.35, |
| "learning_rate": 1.214330251753481e-05, |
| "loss": 0.4262, |
| "step": 841 |
| }, |
| { |
| "epoch": 1.35, |
| "learning_rate": 1.2126420415078133e-05, |
| "loss": 0.4393, |
| "step": 842 |
| }, |
| { |
| "epoch": 1.35, |
| "learning_rate": 1.2109531962807333e-05, |
| "loss": 0.449, |
| "step": 843 |
| }, |
| { |
| "epoch": 1.35, |
| "learning_rate": 1.2092637211153885e-05, |
| "loss": 0.4458, |
| "step": 844 |
| }, |
| { |
| "epoch": 1.35, |
| "learning_rate": 1.207573621056809e-05, |
| "loss": 0.4532, |
| "step": 845 |
| }, |
| { |
| "epoch": 1.35, |
| "learning_rate": 1.2058829011518896e-05, |
| "loss": 0.4618, |
| "step": 846 |
| }, |
| { |
| "epoch": 1.36, |
| "learning_rate": 1.2041915664493763e-05, |
| "loss": 0.4582, |
| "step": 847 |
| }, |
| { |
| "epoch": 1.36, |
| "learning_rate": 1.2024996219998517e-05, |
| "loss": 0.4583, |
| "step": 848 |
| }, |
| { |
| "epoch": 1.36, |
| "learning_rate": 1.2008070728557186e-05, |
| "loss": 0.4611, |
| "step": 849 |
| }, |
| { |
| "epoch": 1.36, |
| "learning_rate": 1.1991139240711857e-05, |
| "loss": 0.4451, |
| "step": 850 |
| }, |
| { |
| "epoch": 1.36, |
| "learning_rate": 1.1974201807022525e-05, |
| "loss": 0.4352, |
| "step": 851 |
| }, |
| { |
| "epoch": 1.36, |
| "learning_rate": 1.195725847806693e-05, |
| "loss": 0.429, |
| "step": 852 |
| }, |
| { |
| "epoch": 1.36, |
| "learning_rate": 1.1940309304440434e-05, |
| "loss": 0.4381, |
| "step": 853 |
| }, |
| { |
| "epoch": 1.37, |
| "learning_rate": 1.1923354336755835e-05, |
| "loss": 0.4227, |
| "step": 854 |
| }, |
| { |
| "epoch": 1.37, |
| "learning_rate": 1.1906393625643244e-05, |
| "loss": 0.457, |
| "step": 855 |
| }, |
| { |
| "epoch": 1.37, |
| "learning_rate": 1.1889427221749916e-05, |
| "loss": 0.4511, |
| "step": 856 |
| }, |
| { |
| "epoch": 1.37, |
| "learning_rate": 1.1872455175740111e-05, |
| "loss": 0.4788, |
| "step": 857 |
| }, |
| { |
| "epoch": 1.37, |
| "learning_rate": 1.1855477538294934e-05, |
| "loss": 0.4739, |
| "step": 858 |
| }, |
| { |
| "epoch": 1.37, |
| "learning_rate": 1.1838494360112185e-05, |
| "loss": 0.4658, |
| "step": 859 |
| }, |
| { |
| "epoch": 1.38, |
| "learning_rate": 1.1821505691906216e-05, |
| "loss": 0.4791, |
| "step": 860 |
| }, |
| { |
| "epoch": 1.38, |
| "learning_rate": 1.1804511584407763e-05, |
| "loss": 0.4498, |
| "step": 861 |
| }, |
| { |
| "epoch": 1.38, |
| "learning_rate": 1.1787512088363817e-05, |
| "loss": 0.4648, |
| "step": 862 |
| }, |
| { |
| "epoch": 1.38, |
| "learning_rate": 1.1770507254537454e-05, |
| "loss": 0.4245, |
| "step": 863 |
| }, |
| { |
| "epoch": 1.38, |
| "learning_rate": 1.1753497133707678e-05, |
| "loss": 0.4551, |
| "step": 864 |
| }, |
| { |
| "epoch": 1.38, |
| "learning_rate": 1.1736481776669307e-05, |
| "loss": 0.4545, |
| "step": 865 |
| }, |
| { |
| "epoch": 1.39, |
| "learning_rate": 1.1719461234232765e-05, |
| "loss": 0.4729, |
| "step": 866 |
| }, |
| { |
| "epoch": 1.39, |
| "learning_rate": 1.1702435557223988e-05, |
| "loss": 0.4325, |
| "step": 867 |
| }, |
| { |
| "epoch": 1.39, |
| "learning_rate": 1.1685404796484226e-05, |
| "loss": 0.4684, |
| "step": 868 |
| }, |
| { |
| "epoch": 1.39, |
| "learning_rate": 1.1668369002869912e-05, |
| "loss": 0.4481, |
| "step": 869 |
| }, |
| { |
| "epoch": 1.39, |
| "learning_rate": 1.1651328227252516e-05, |
| "loss": 0.4544, |
| "step": 870 |
| }, |
| { |
| "epoch": 1.39, |
| "learning_rate": 1.1634282520518382e-05, |
| "loss": 0.466, |
| "step": 871 |
| }, |
| { |
| "epoch": 1.4, |
| "learning_rate": 1.1617231933568579e-05, |
| "loss": 0.4349, |
| "step": 872 |
| }, |
| { |
| "epoch": 1.4, |
| "learning_rate": 1.1600176517318742e-05, |
| "loss": 0.4451, |
| "step": 873 |
| }, |
| { |
| "epoch": 1.4, |
| "learning_rate": 1.1583116322698936e-05, |
| "loss": 0.4418, |
| "step": 874 |
| }, |
| { |
| "epoch": 1.4, |
| "learning_rate": 1.1566051400653486e-05, |
| "loss": 0.49, |
| "step": 875 |
| }, |
| { |
| "epoch": 1.4, |
| "learning_rate": 1.1548981802140849e-05, |
| "loss": 0.45, |
| "step": 876 |
| }, |
| { |
| "epoch": 1.4, |
| "learning_rate": 1.153190757813343e-05, |
| "loss": 0.4777, |
| "step": 877 |
| }, |
| { |
| "epoch": 1.4, |
| "learning_rate": 1.151482877961746e-05, |
| "loss": 0.4674, |
| "step": 878 |
| }, |
| { |
| "epoch": 1.41, |
| "learning_rate": 1.1497745457592817e-05, |
| "loss": 0.4496, |
| "step": 879 |
| }, |
| { |
| "epoch": 1.41, |
| "learning_rate": 1.1480657663072896e-05, |
| "loss": 0.4408, |
| "step": 880 |
| }, |
| { |
| "epoch": 1.41, |
| "learning_rate": 1.1463565447084446e-05, |
| "loss": 0.4636, |
| "step": 881 |
| }, |
| { |
| "epoch": 1.41, |
| "learning_rate": 1.1446468860667422e-05, |
| "loss": 0.4443, |
| "step": 882 |
| }, |
| { |
| "epoch": 1.41, |
| "learning_rate": 1.142936795487482e-05, |
| "loss": 0.4543, |
| "step": 883 |
| }, |
| { |
| "epoch": 1.41, |
| "learning_rate": 1.141226278077254e-05, |
| "loss": 0.4392, |
| "step": 884 |
| }, |
| { |
| "epoch": 1.42, |
| "learning_rate": 1.1395153389439232e-05, |
| "loss": 0.4638, |
| "step": 885 |
| }, |
| { |
| "epoch": 1.42, |
| "learning_rate": 1.1378039831966134e-05, |
| "loss": 0.4399, |
| "step": 886 |
| }, |
| { |
| "epoch": 1.42, |
| "learning_rate": 1.1360922159456929e-05, |
| "loss": 0.4775, |
| "step": 887 |
| }, |
| { |
| "epoch": 1.42, |
| "learning_rate": 1.1343800423027583e-05, |
| "loss": 0.4597, |
| "step": 888 |
| }, |
| { |
| "epoch": 1.42, |
| "learning_rate": 1.1326674673806195e-05, |
| "loss": 0.4852, |
| "step": 889 |
| }, |
| { |
| "epoch": 1.42, |
| "learning_rate": 1.1309544962932861e-05, |
| "loss": 0.4314, |
| "step": 890 |
| }, |
| { |
| "epoch": 1.43, |
| "learning_rate": 1.129241134155949e-05, |
| "loss": 0.4545, |
| "step": 891 |
| }, |
| { |
| "epoch": 1.43, |
| "learning_rate": 1.1275273860849684e-05, |
| "loss": 0.4925, |
| "step": 892 |
| }, |
| { |
| "epoch": 1.43, |
| "learning_rate": 1.1258132571978555e-05, |
| "loss": 0.4676, |
| "step": 893 |
| }, |
| { |
| "epoch": 1.43, |
| "learning_rate": 1.1240987526132595e-05, |
| "loss": 0.4504, |
| "step": 894 |
| }, |
| { |
| "epoch": 1.43, |
| "learning_rate": 1.1223838774509515e-05, |
| "loss": 0.454, |
| "step": 895 |
| }, |
| { |
| "epoch": 1.43, |
| "learning_rate": 1.1206686368318087e-05, |
| "loss": 0.4496, |
| "step": 896 |
| }, |
| { |
| "epoch": 1.44, |
| "learning_rate": 1.1189530358778005e-05, |
| "loss": 0.4316, |
| "step": 897 |
| }, |
| { |
| "epoch": 1.44, |
| "learning_rate": 1.1172370797119711e-05, |
| "loss": 0.4721, |
| "step": 898 |
| }, |
| { |
| "epoch": 1.44, |
| "learning_rate": 1.1155207734584264e-05, |
| "loss": 0.4797, |
| "step": 899 |
| }, |
| { |
| "epoch": 1.44, |
| "learning_rate": 1.1138041222423177e-05, |
| "loss": 0.4445, |
| "step": 900 |
| }, |
| { |
| "epoch": 1.44, |
| "learning_rate": 1.1120871311898254e-05, |
| "loss": 0.4544, |
| "step": 901 |
| }, |
| { |
| "epoch": 1.44, |
| "learning_rate": 1.110369805428146e-05, |
| "loss": 0.4599, |
| "step": 902 |
| }, |
| { |
| "epoch": 1.44, |
| "learning_rate": 1.1086521500854746e-05, |
| "loss": 0.4835, |
| "step": 903 |
| }, |
| { |
| "epoch": 1.45, |
| "learning_rate": 1.106934170290991e-05, |
| "loss": 0.4391, |
| "step": 904 |
| }, |
| { |
| "epoch": 1.45, |
| "learning_rate": 1.1052158711748435e-05, |
| "loss": 0.4197, |
| "step": 905 |
| }, |
| { |
| "epoch": 1.45, |
| "learning_rate": 1.1034972578681338e-05, |
| "loss": 0.4371, |
| "step": 906 |
| }, |
| { |
| "epoch": 1.45, |
| "learning_rate": 1.1017783355029027e-05, |
| "loss": 0.45, |
| "step": 907 |
| }, |
| { |
| "epoch": 1.45, |
| "learning_rate": 1.1000591092121126e-05, |
| "loss": 0.4482, |
| "step": 908 |
| }, |
| { |
| "epoch": 1.45, |
| "learning_rate": 1.0983395841296349e-05, |
| "loss": 0.4301, |
| "step": 909 |
| }, |
| { |
| "epoch": 1.46, |
| "learning_rate": 1.0966197653902319e-05, |
| "loss": 0.4712, |
| "step": 910 |
| }, |
| { |
| "epoch": 1.46, |
| "learning_rate": 1.0948996581295437e-05, |
| "loss": 0.4448, |
| "step": 911 |
| }, |
| { |
| "epoch": 1.46, |
| "learning_rate": 1.0931792674840718e-05, |
| "loss": 0.4346, |
| "step": 912 |
| }, |
| { |
| "epoch": 1.46, |
| "learning_rate": 1.0914585985911632e-05, |
| "loss": 0.4374, |
| "step": 913 |
| }, |
| { |
| "epoch": 1.46, |
| "learning_rate": 1.0897376565889972e-05, |
| "loss": 0.4553, |
| "step": 914 |
| }, |
| { |
| "epoch": 1.46, |
| "learning_rate": 1.0880164466165675e-05, |
| "loss": 0.4481, |
| "step": 915 |
| }, |
| { |
| "epoch": 1.47, |
| "learning_rate": 1.0862949738136682e-05, |
| "loss": 0.4307, |
| "step": 916 |
| }, |
| { |
| "epoch": 1.47, |
| "learning_rate": 1.084573243320878e-05, |
| "loss": 0.412, |
| "step": 917 |
| }, |
| { |
| "epoch": 1.47, |
| "learning_rate": 1.0828512602795462e-05, |
| "loss": 0.4764, |
| "step": 918 |
| }, |
| { |
| "epoch": 1.47, |
| "learning_rate": 1.0811290298317755e-05, |
| "loss": 0.4558, |
| "step": 919 |
| }, |
| { |
| "epoch": 1.47, |
| "learning_rate": 1.0794065571204073e-05, |
| "loss": 0.4359, |
| "step": 920 |
| }, |
| { |
| "epoch": 1.47, |
| "learning_rate": 1.0776838472890065e-05, |
| "loss": 0.4471, |
| "step": 921 |
| }, |
| { |
| "epoch": 1.48, |
| "learning_rate": 1.0759609054818459e-05, |
| "loss": 0.4861, |
| "step": 922 |
| }, |
| { |
| "epoch": 1.48, |
| "learning_rate": 1.0742377368438915e-05, |
| "loss": 0.4861, |
| "step": 923 |
| }, |
| { |
| "epoch": 1.48, |
| "learning_rate": 1.0725143465207868e-05, |
| "loss": 0.4759, |
| "step": 924 |
| }, |
| { |
| "epoch": 1.48, |
| "learning_rate": 1.0707907396588362e-05, |
| "loss": 0.4615, |
| "step": 925 |
| }, |
| { |
| "epoch": 1.48, |
| "learning_rate": 1.069066921404992e-05, |
| "loss": 0.4533, |
| "step": 926 |
| }, |
| { |
| "epoch": 1.48, |
| "learning_rate": 1.0673428969068365e-05, |
| "loss": 0.4574, |
| "step": 927 |
| }, |
| { |
| "epoch": 1.48, |
| "learning_rate": 1.065618671312569e-05, |
| "loss": 0.4622, |
| "step": 928 |
| }, |
| { |
| "epoch": 1.49, |
| "learning_rate": 1.063894249770989e-05, |
| "loss": 0.4415, |
| "step": 929 |
| }, |
| { |
| "epoch": 1.49, |
| "learning_rate": 1.0621696374314807e-05, |
| "loss": 0.4544, |
| "step": 930 |
| }, |
| { |
| "epoch": 1.49, |
| "learning_rate": 1.0604448394439983e-05, |
| "loss": 0.4697, |
| "step": 931 |
| }, |
| { |
| "epoch": 1.49, |
| "learning_rate": 1.0587198609590505e-05, |
| "loss": 0.4509, |
| "step": 932 |
| }, |
| { |
| "epoch": 1.49, |
| "learning_rate": 1.0569947071276847e-05, |
| "loss": 0.4053, |
| "step": 933 |
| }, |
| { |
| "epoch": 1.49, |
| "learning_rate": 1.0552693831014726e-05, |
| "loss": 0.4883, |
| "step": 934 |
| }, |
| { |
| "epoch": 1.5, |
| "learning_rate": 1.053543894032493e-05, |
| "loss": 0.4478, |
| "step": 935 |
| }, |
| { |
| "epoch": 1.5, |
| "learning_rate": 1.0518182450733185e-05, |
| "loss": 0.42, |
| "step": 936 |
| }, |
| { |
| "epoch": 1.5, |
| "learning_rate": 1.0500924413769988e-05, |
| "loss": 0.4294, |
| "step": 937 |
| }, |
| { |
| "epoch": 1.5, |
| "learning_rate": 1.0483664880970456e-05, |
| "loss": 0.4371, |
| "step": 938 |
| }, |
| { |
| "epoch": 1.5, |
| "learning_rate": 1.0466403903874176e-05, |
| "loss": 0.4412, |
| "step": 939 |
| }, |
| { |
| "epoch": 1.5, |
| "learning_rate": 1.0449141534025044e-05, |
| "loss": 0.4406, |
| "step": 940 |
| }, |
| { |
| "epoch": 1.51, |
| "learning_rate": 1.0431877822971118e-05, |
| "loss": 0.4185, |
| "step": 941 |
| }, |
| { |
| "epoch": 1.51, |
| "learning_rate": 1.0414612822264457e-05, |
| "loss": 0.4692, |
| "step": 942 |
| }, |
| { |
| "epoch": 1.51, |
| "learning_rate": 1.0397346583460972e-05, |
| "loss": 0.4496, |
| "step": 943 |
| }, |
| { |
| "epoch": 1.51, |
| "learning_rate": 1.038007915812028e-05, |
| "loss": 0.4509, |
| "step": 944 |
| }, |
| { |
| "epoch": 1.51, |
| "learning_rate": 1.0362810597805526e-05, |
| "loss": 0.48, |
| "step": 945 |
| }, |
| { |
| "epoch": 1.51, |
| "learning_rate": 1.034554095408326e-05, |
| "loss": 0.4355, |
| "step": 946 |
| }, |
| { |
| "epoch": 1.52, |
| "learning_rate": 1.0328270278523256e-05, |
| "loss": 0.4604, |
| "step": 947 |
| }, |
| { |
| "epoch": 1.52, |
| "learning_rate": 1.031099862269837e-05, |
| "loss": 0.4429, |
| "step": 948 |
| }, |
| { |
| "epoch": 1.52, |
| "learning_rate": 1.0293726038184393e-05, |
| "loss": 0.4355, |
| "step": 949 |
| }, |
| { |
| "epoch": 1.52, |
| "learning_rate": 1.0276452576559878e-05, |
| "loss": 0.4407, |
| "step": 950 |
| }, |
| { |
| "epoch": 1.52, |
| "learning_rate": 1.0259178289406011e-05, |
| "loss": 0.4104, |
| "step": 951 |
| }, |
| { |
| "epoch": 1.52, |
| "learning_rate": 1.024190322830643e-05, |
| "loss": 0.4515, |
| "step": 952 |
| }, |
| { |
| "epoch": 1.52, |
| "learning_rate": 1.022462744484709e-05, |
| "loss": 0.4678, |
| "step": 953 |
| }, |
| { |
| "epoch": 1.53, |
| "learning_rate": 1.0207350990616107e-05, |
| "loss": 0.4622, |
| "step": 954 |
| }, |
| { |
| "epoch": 1.53, |
| "learning_rate": 1.019007391720359e-05, |
| "loss": 0.4388, |
| "step": 955 |
| }, |
| { |
| "epoch": 1.53, |
| "learning_rate": 1.0172796276201504e-05, |
| "loss": 0.4674, |
| "step": 956 |
| }, |
| { |
| "epoch": 1.53, |
| "learning_rate": 1.0155518119203511e-05, |
| "loss": 0.4226, |
| "step": 957 |
| }, |
| { |
| "epoch": 1.53, |
| "learning_rate": 1.0138239497804804e-05, |
| "loss": 0.4632, |
| "step": 958 |
| }, |
| { |
| "epoch": 1.53, |
| "learning_rate": 1.0120960463601977e-05, |
| "loss": 0.4762, |
| "step": 959 |
| }, |
| { |
| "epoch": 1.54, |
| "learning_rate": 1.0103681068192845e-05, |
| "loss": 0.4275, |
| "step": 960 |
| }, |
| { |
| "epoch": 1.54, |
| "learning_rate": 1.0086401363176306e-05, |
| "loss": 0.4611, |
| "step": 961 |
| }, |
| { |
| "epoch": 1.54, |
| "learning_rate": 1.0069121400152182e-05, |
| "loss": 0.4293, |
| "step": 962 |
| }, |
| { |
| "epoch": 1.54, |
| "learning_rate": 1.0051841230721065e-05, |
| "loss": 0.4575, |
| "step": 963 |
| }, |
| { |
| "epoch": 1.54, |
| "learning_rate": 1.0034560906484161e-05, |
| "loss": 0.4673, |
| "step": 964 |
| }, |
| { |
| "epoch": 1.54, |
| "learning_rate": 1.0017280479043148e-05, |
| "loss": 0.4511, |
| "step": 965 |
| }, |
| { |
| "epoch": 1.55, |
| "learning_rate": 1e-05, |
| "loss": 0.4193, |
| "step": 966 |
| }, |
| { |
| "epoch": 1.55, |
| "learning_rate": 9.982719520956856e-06, |
| "loss": 0.4537, |
| "step": 967 |
| }, |
| { |
| "epoch": 1.55, |
| "learning_rate": 9.965439093515842e-06, |
| "loss": 0.4771, |
| "step": 968 |
| }, |
| { |
| "epoch": 1.55, |
| "learning_rate": 9.948158769278939e-06, |
| "loss": 0.4805, |
| "step": 969 |
| }, |
| { |
| "epoch": 1.55, |
| "learning_rate": 9.930878599847822e-06, |
| "loss": 0.4376, |
| "step": 970 |
| }, |
| { |
| "epoch": 1.55, |
| "learning_rate": 9.913598636823694e-06, |
| "loss": 0.4488, |
| "step": 971 |
| }, |
| { |
| "epoch": 1.56, |
| "learning_rate": 9.896318931807155e-06, |
| "loss": 0.4498, |
| "step": 972 |
| }, |
| { |
| "epoch": 1.56, |
| "learning_rate": 9.879039536398023e-06, |
| "loss": 0.429, |
| "step": 973 |
| }, |
| { |
| "epoch": 1.56, |
| "learning_rate": 9.861760502195197e-06, |
| "loss": 0.4573, |
| "step": 974 |
| }, |
| { |
| "epoch": 1.56, |
| "learning_rate": 9.844481880796492e-06, |
| "loss": 0.4525, |
| "step": 975 |
| }, |
| { |
| "epoch": 1.56, |
| "learning_rate": 9.827203723798498e-06, |
| "loss": 0.4609, |
| "step": 976 |
| }, |
| { |
| "epoch": 1.56, |
| "learning_rate": 9.809926082796415e-06, |
| "loss": 0.4524, |
| "step": 977 |
| }, |
| { |
| "epoch": 1.56, |
| "learning_rate": 9.7926490093839e-06, |
| "loss": 0.4229, |
| "step": 978 |
| }, |
| { |
| "epoch": 1.57, |
| "learning_rate": 9.775372555152912e-06, |
| "loss": 0.4325, |
| "step": 979 |
| }, |
| { |
| "epoch": 1.57, |
| "learning_rate": 9.758096771693574e-06, |
| "loss": 0.4542, |
| "step": 980 |
| }, |
| { |
| "epoch": 1.57, |
| "learning_rate": 9.740821710593989e-06, |
| "loss": 0.4875, |
| "step": 981 |
| }, |
| { |
| "epoch": 1.57, |
| "learning_rate": 9.723547423440122e-06, |
| "loss": 0.4473, |
| "step": 982 |
| }, |
| { |
| "epoch": 1.57, |
| "learning_rate": 9.70627396181561e-06, |
| "loss": 0.4639, |
| "step": 983 |
| }, |
| { |
| "epoch": 1.57, |
| "learning_rate": 9.689001377301634e-06, |
| "loss": 0.4372, |
| "step": 984 |
| }, |
| { |
| "epoch": 1.58, |
| "learning_rate": 9.671729721476747e-06, |
| "loss": 0.4525, |
| "step": 985 |
| }, |
| { |
| "epoch": 1.58, |
| "learning_rate": 9.654459045916743e-06, |
| "loss": 0.4376, |
| "step": 986 |
| }, |
| { |
| "epoch": 1.58, |
| "learning_rate": 9.637189402194477e-06, |
| "loss": 0.4408, |
| "step": 987 |
| }, |
| { |
| "epoch": 1.58, |
| "learning_rate": 9.619920841879726e-06, |
| "loss": 0.4157, |
| "step": 988 |
| }, |
| { |
| "epoch": 1.58, |
| "learning_rate": 9.602653416539031e-06, |
| "loss": 0.4464, |
| "step": 989 |
| }, |
| { |
| "epoch": 1.58, |
| "learning_rate": 9.585387177735548e-06, |
| "loss": 0.4296, |
| "step": 990 |
| }, |
| { |
| "epoch": 1.59, |
| "learning_rate": 9.568122177028884e-06, |
| "loss": 0.4772, |
| "step": 991 |
| }, |
| { |
| "epoch": 1.59, |
| "learning_rate": 9.550858465974958e-06, |
| "loss": 0.4389, |
| "step": 992 |
| }, |
| { |
| "epoch": 1.59, |
| "learning_rate": 9.533596096125826e-06, |
| "loss": 0.4226, |
| "step": 993 |
| }, |
| { |
| "epoch": 1.59, |
| "learning_rate": 9.516335119029547e-06, |
| "loss": 0.4391, |
| "step": 994 |
| }, |
| { |
| "epoch": 1.59, |
| "learning_rate": 9.499075586230014e-06, |
| "loss": 0.4583, |
| "step": 995 |
| }, |
| { |
| "epoch": 1.59, |
| "learning_rate": 9.481817549266817e-06, |
| "loss": 0.4488, |
| "step": 996 |
| }, |
| { |
| "epoch": 1.6, |
| "learning_rate": 9.464561059675073e-06, |
| "loss": 0.4373, |
| "step": 997 |
| }, |
| { |
| "epoch": 1.6, |
| "learning_rate": 9.44730616898528e-06, |
| "loss": 0.4602, |
| "step": 998 |
| }, |
| { |
| "epoch": 1.6, |
| "learning_rate": 9.430052928723153e-06, |
| "loss": 0.4742, |
| "step": 999 |
| }, |
| { |
| "epoch": 1.6, |
| "learning_rate": 9.412801390409496e-06, |
| "loss": 0.4604, |
| "step": 1000 |
| }, |
| { |
| "epoch": 1.6, |
| "learning_rate": 9.395551605560018e-06, |
| "loss": 0.445, |
| "step": 1001 |
| }, |
| { |
| "epoch": 1.6, |
| "learning_rate": 9.378303625685196e-06, |
| "loss": 0.418, |
| "step": 1002 |
| }, |
| { |
| "epoch": 1.6, |
| "learning_rate": 9.361057502290112e-06, |
| "loss": 0.4528, |
| "step": 1003 |
| }, |
| { |
| "epoch": 1.61, |
| "learning_rate": 9.343813286874312e-06, |
| "loss": 0.4566, |
| "step": 1004 |
| }, |
| { |
| "epoch": 1.61, |
| "learning_rate": 9.326571030931636e-06, |
| "loss": 0.4409, |
| "step": 1005 |
| }, |
| { |
| "epoch": 1.61, |
| "learning_rate": 9.309330785950086e-06, |
| "loss": 0.4453, |
| "step": 1006 |
| }, |
| { |
| "epoch": 1.61, |
| "learning_rate": 9.292092603411642e-06, |
| "loss": 0.4504, |
| "step": 1007 |
| }, |
| { |
| "epoch": 1.61, |
| "learning_rate": 9.274856534792138e-06, |
| "loss": 0.4388, |
| "step": 1008 |
| }, |
| { |
| "epoch": 1.61, |
| "learning_rate": 9.257622631561085e-06, |
| "loss": 0.454, |
| "step": 1009 |
| }, |
| { |
| "epoch": 1.62, |
| "learning_rate": 9.240390945181543e-06, |
| "loss": 0.4553, |
| "step": 1010 |
| }, |
| { |
| "epoch": 1.62, |
| "learning_rate": 9.223161527109938e-06, |
| "loss": 0.4528, |
| "step": 1011 |
| }, |
| { |
| "epoch": 1.62, |
| "learning_rate": 9.205934428795929e-06, |
| "loss": 0.453, |
| "step": 1012 |
| }, |
| { |
| "epoch": 1.62, |
| "learning_rate": 9.188709701682246e-06, |
| "loss": 0.444, |
| "step": 1013 |
| }, |
| { |
| "epoch": 1.62, |
| "learning_rate": 9.17148739720454e-06, |
| "loss": 0.4687, |
| "step": 1014 |
| }, |
| { |
| "epoch": 1.62, |
| "learning_rate": 9.154267566791224e-06, |
| "loss": 0.4444, |
| "step": 1015 |
| }, |
| { |
| "epoch": 1.63, |
| "learning_rate": 9.137050261863323e-06, |
| "loss": 0.4438, |
| "step": 1016 |
| }, |
| { |
| "epoch": 1.63, |
| "learning_rate": 9.119835533834332e-06, |
| "loss": 0.4173, |
| "step": 1017 |
| }, |
| { |
| "epoch": 1.63, |
| "learning_rate": 9.102623434110028e-06, |
| "loss": 0.467, |
| "step": 1018 |
| }, |
| { |
| "epoch": 1.63, |
| "learning_rate": 9.085414014088368e-06, |
| "loss": 0.4464, |
| "step": 1019 |
| }, |
| { |
| "epoch": 1.63, |
| "learning_rate": 9.068207325159285e-06, |
| "loss": 0.4456, |
| "step": 1020 |
| }, |
| { |
| "epoch": 1.63, |
| "learning_rate": 9.051003418704566e-06, |
| "loss": 0.4552, |
| "step": 1021 |
| }, |
| { |
| "epoch": 1.64, |
| "learning_rate": 9.033802346097683e-06, |
| "loss": 0.4578, |
| "step": 1022 |
| }, |
| { |
| "epoch": 1.64, |
| "learning_rate": 9.016604158703654e-06, |
| "loss": 0.4654, |
| "step": 1023 |
| }, |
| { |
| "epoch": 1.64, |
| "learning_rate": 8.999408907878877e-06, |
| "loss": 0.4604, |
| "step": 1024 |
| }, |
| { |
| "epoch": 1.64, |
| "learning_rate": 8.982216644970978e-06, |
| "loss": 0.4709, |
| "step": 1025 |
| }, |
| { |
| "epoch": 1.64, |
| "learning_rate": 8.965027421318666e-06, |
| "loss": 0.4277, |
| "step": 1026 |
| }, |
| { |
| "epoch": 1.64, |
| "learning_rate": 8.947841288251568e-06, |
| "loss": 0.4175, |
| "step": 1027 |
| }, |
| { |
| "epoch": 1.64, |
| "learning_rate": 8.930658297090092e-06, |
| "loss": 0.4867, |
| "step": 1028 |
| }, |
| { |
| "epoch": 1.65, |
| "learning_rate": 8.913478499145255e-06, |
| "loss": 0.4325, |
| "step": 1029 |
| }, |
| { |
| "epoch": 1.65, |
| "learning_rate": 8.896301945718541e-06, |
| "loss": 0.4527, |
| "step": 1030 |
| }, |
| { |
| "epoch": 1.65, |
| "learning_rate": 8.879128688101749e-06, |
| "loss": 0.4382, |
| "step": 1031 |
| }, |
| { |
| "epoch": 1.65, |
| "learning_rate": 8.861958777576826e-06, |
| "loss": 0.4337, |
| "step": 1032 |
| }, |
| { |
| "epoch": 1.65, |
| "learning_rate": 8.844792265415738e-06, |
| "loss": 0.4489, |
| "step": 1033 |
| }, |
| { |
| "epoch": 1.65, |
| "learning_rate": 8.827629202880294e-06, |
| "loss": 0.4516, |
| "step": 1034 |
| }, |
| { |
| "epoch": 1.66, |
| "learning_rate": 8.810469641222001e-06, |
| "loss": 0.4569, |
| "step": 1035 |
| }, |
| { |
| "epoch": 1.66, |
| "learning_rate": 8.793313631681915e-06, |
| "loss": 0.4424, |
| "step": 1036 |
| }, |
| { |
| "epoch": 1.66, |
| "learning_rate": 8.776161225490488e-06, |
| "loss": 0.4606, |
| "step": 1037 |
| }, |
| { |
| "epoch": 1.66, |
| "learning_rate": 8.759012473867407e-06, |
| "loss": 0.4562, |
| "step": 1038 |
| }, |
| { |
| "epoch": 1.66, |
| "learning_rate": 8.741867428021447e-06, |
| "loss": 0.4817, |
| "step": 1039 |
| }, |
| { |
| "epoch": 1.66, |
| "learning_rate": 8.72472613915032e-06, |
| "loss": 0.4307, |
| "step": 1040 |
| }, |
| { |
| "epoch": 1.67, |
| "learning_rate": 8.707588658440511e-06, |
| "loss": 0.4381, |
| "step": 1041 |
| }, |
| { |
| "epoch": 1.67, |
| "learning_rate": 8.690455037067142e-06, |
| "loss": 0.4648, |
| "step": 1042 |
| }, |
| { |
| "epoch": 1.67, |
| "learning_rate": 8.673325326193806e-06, |
| "loss": 0.4318, |
| "step": 1043 |
| }, |
| { |
| "epoch": 1.67, |
| "learning_rate": 8.656199576972424e-06, |
| "loss": 0.4435, |
| "step": 1044 |
| }, |
| { |
| "epoch": 1.67, |
| "learning_rate": 8.639077840543078e-06, |
| "loss": 0.4284, |
| "step": 1045 |
| }, |
| { |
| "epoch": 1.67, |
| "learning_rate": 8.621960168033868e-06, |
| "loss": 0.4595, |
| "step": 1046 |
| }, |
| { |
| "epoch": 1.68, |
| "learning_rate": 8.604846610560771e-06, |
| "loss": 0.486, |
| "step": 1047 |
| }, |
| { |
| "epoch": 1.68, |
| "learning_rate": 8.587737219227462e-06, |
| "loss": 0.4411, |
| "step": 1048 |
| }, |
| { |
| "epoch": 1.68, |
| "learning_rate": 8.570632045125185e-06, |
| "loss": 0.4579, |
| "step": 1049 |
| }, |
| { |
| "epoch": 1.68, |
| "learning_rate": 8.553531139332583e-06, |
| "loss": 0.4426, |
| "step": 1050 |
| }, |
| { |
| "epoch": 1.68, |
| "learning_rate": 8.536434552915555e-06, |
| "loss": 0.4188, |
| "step": 1051 |
| }, |
| { |
| "epoch": 1.68, |
| "learning_rate": 8.519342336927106e-06, |
| "loss": 0.4459, |
| "step": 1052 |
| }, |
| { |
| "epoch": 1.68, |
| "learning_rate": 8.502254542407186e-06, |
| "loss": 0.4735, |
| "step": 1053 |
| }, |
| { |
| "epoch": 1.69, |
| "learning_rate": 8.485171220382545e-06, |
| "loss": 0.4516, |
| "step": 1054 |
| }, |
| { |
| "epoch": 1.69, |
| "learning_rate": 8.468092421866575e-06, |
| "loss": 0.4252, |
| "step": 1055 |
| }, |
| { |
| "epoch": 1.69, |
| "learning_rate": 8.451018197859153e-06, |
| "loss": 0.4601, |
| "step": 1056 |
| }, |
| { |
| "epoch": 1.69, |
| "learning_rate": 8.433948599346516e-06, |
| "loss": 0.452, |
| "step": 1057 |
| }, |
| { |
| "epoch": 1.69, |
| "learning_rate": 8.41688367730107e-06, |
| "loss": 0.4544, |
| "step": 1058 |
| }, |
| { |
| "epoch": 1.69, |
| "learning_rate": 8.399823482681263e-06, |
| "loss": 0.4464, |
| "step": 1059 |
| }, |
| { |
| "epoch": 1.7, |
| "learning_rate": 8.382768066431427e-06, |
| "loss": 0.4349, |
| "step": 1060 |
| }, |
| { |
| "epoch": 1.7, |
| "learning_rate": 8.36571747948162e-06, |
| "loss": 0.4434, |
| "step": 1061 |
| }, |
| { |
| "epoch": 1.7, |
| "learning_rate": 8.348671772747488e-06, |
| "loss": 0.5116, |
| "step": 1062 |
| }, |
| { |
| "epoch": 1.7, |
| "learning_rate": 8.331630997130091e-06, |
| "loss": 0.447, |
| "step": 1063 |
| }, |
| { |
| "epoch": 1.7, |
| "learning_rate": 8.314595203515781e-06, |
| "loss": 0.409, |
| "step": 1064 |
| }, |
| { |
| "epoch": 1.7, |
| "learning_rate": 8.297564442776014e-06, |
| "loss": 0.4349, |
| "step": 1065 |
| }, |
| { |
| "epoch": 1.71, |
| "learning_rate": 8.280538765767236e-06, |
| "loss": 0.477, |
| "step": 1066 |
| }, |
| { |
| "epoch": 1.71, |
| "learning_rate": 8.263518223330698e-06, |
| "loss": 0.4823, |
| "step": 1067 |
| }, |
| { |
| "epoch": 1.71, |
| "learning_rate": 8.246502866292324e-06, |
| "loss": 0.4413, |
| "step": 1068 |
| }, |
| { |
| "epoch": 1.71, |
| "learning_rate": 8.229492745462551e-06, |
| "loss": 0.4635, |
| "step": 1069 |
| }, |
| { |
| "epoch": 1.71, |
| "learning_rate": 8.212487911636185e-06, |
| "loss": 0.4245, |
| "step": 1070 |
| }, |
| { |
| "epoch": 1.71, |
| "learning_rate": 8.195488415592238e-06, |
| "loss": 0.442, |
| "step": 1071 |
| }, |
| { |
| "epoch": 1.72, |
| "learning_rate": 8.17849430809379e-06, |
| "loss": 0.4419, |
| "step": 1072 |
| }, |
| { |
| "epoch": 1.72, |
| "learning_rate": 8.161505639887818e-06, |
| "loss": 0.4561, |
| "step": 1073 |
| }, |
| { |
| "epoch": 1.72, |
| "learning_rate": 8.144522461705067e-06, |
| "loss": 0.4337, |
| "step": 1074 |
| }, |
| { |
| "epoch": 1.72, |
| "learning_rate": 8.12754482425989e-06, |
| "loss": 0.4017, |
| "step": 1075 |
| }, |
| { |
| "epoch": 1.72, |
| "learning_rate": 8.110572778250086e-06, |
| "loss": 0.4576, |
| "step": 1076 |
| }, |
| { |
| "epoch": 1.72, |
| "learning_rate": 8.09360637435676e-06, |
| "loss": 0.4394, |
| "step": 1077 |
| }, |
| { |
| "epoch": 1.72, |
| "learning_rate": 8.076645663244168e-06, |
| "loss": 0.4563, |
| "step": 1078 |
| }, |
| { |
| "epoch": 1.73, |
| "learning_rate": 8.05969069555957e-06, |
| "loss": 0.4284, |
| "step": 1079 |
| }, |
| { |
| "epoch": 1.73, |
| "learning_rate": 8.042741521933071e-06, |
| "loss": 0.4655, |
| "step": 1080 |
| }, |
| { |
| "epoch": 1.73, |
| "learning_rate": 8.025798192977482e-06, |
| "loss": 0.4211, |
| "step": 1081 |
| }, |
| { |
| "epoch": 1.73, |
| "learning_rate": 8.008860759288148e-06, |
| "loss": 0.4656, |
| "step": 1082 |
| }, |
| { |
| "epoch": 1.73, |
| "learning_rate": 7.991929271442817e-06, |
| "loss": 0.4751, |
| "step": 1083 |
| }, |
| { |
| "epoch": 1.73, |
| "learning_rate": 7.975003780001486e-06, |
| "loss": 0.4312, |
| "step": 1084 |
| }, |
| { |
| "epoch": 1.74, |
| "learning_rate": 7.958084335506239e-06, |
| "loss": 0.4688, |
| "step": 1085 |
| }, |
| { |
| "epoch": 1.74, |
| "learning_rate": 7.941170988481108e-06, |
| "loss": 0.4507, |
| "step": 1086 |
| }, |
| { |
| "epoch": 1.74, |
| "learning_rate": 7.924263789431913e-06, |
| "loss": 0.4394, |
| "step": 1087 |
| }, |
| { |
| "epoch": 1.74, |
| "learning_rate": 7.907362788846116e-06, |
| "loss": 0.4354, |
| "step": 1088 |
| }, |
| { |
| "epoch": 1.74, |
| "learning_rate": 7.89046803719267e-06, |
| "loss": 0.4294, |
| "step": 1089 |
| }, |
| { |
| "epoch": 1.74, |
| "learning_rate": 7.873579584921869e-06, |
| "loss": 0.4723, |
| "step": 1090 |
| }, |
| { |
| "epoch": 1.75, |
| "learning_rate": 7.856697482465195e-06, |
| "loss": 0.4418, |
| "step": 1091 |
| }, |
| { |
| "epoch": 1.75, |
| "learning_rate": 7.839821780235168e-06, |
| "loss": 0.4395, |
| "step": 1092 |
| }, |
| { |
| "epoch": 1.75, |
| "learning_rate": 7.822952528625192e-06, |
| "loss": 0.4178, |
| "step": 1093 |
| }, |
| { |
| "epoch": 1.75, |
| "learning_rate": 7.806089778009421e-06, |
| "loss": 0.4579, |
| "step": 1094 |
| }, |
| { |
| "epoch": 1.75, |
| "learning_rate": 7.789233578742583e-06, |
| "loss": 0.4418, |
| "step": 1095 |
| }, |
| { |
| "epoch": 1.75, |
| "learning_rate": 7.77238398115985e-06, |
| "loss": 0.4317, |
| "step": 1096 |
| }, |
| { |
| "epoch": 1.76, |
| "learning_rate": 7.755541035576677e-06, |
| "loss": 0.4223, |
| "step": 1097 |
| }, |
| { |
| "epoch": 1.76, |
| "learning_rate": 7.738704792288654e-06, |
| "loss": 0.4624, |
| "step": 1098 |
| }, |
| { |
| "epoch": 1.76, |
| "learning_rate": 7.721875301571359e-06, |
| "loss": 0.463, |
| "step": 1099 |
| }, |
| { |
| "epoch": 1.76, |
| "learning_rate": 7.705052613680212e-06, |
| "loss": 0.4574, |
| "step": 1100 |
| }, |
| { |
| "epoch": 1.76, |
| "learning_rate": 7.688236778850307e-06, |
| "loss": 0.4321, |
| "step": 1101 |
| }, |
| { |
| "epoch": 1.76, |
| "learning_rate": 7.671427847296274e-06, |
| "loss": 0.4406, |
| "step": 1102 |
| }, |
| { |
| "epoch": 1.76, |
| "learning_rate": 7.654625869212147e-06, |
| "loss": 0.4582, |
| "step": 1103 |
| }, |
| { |
| "epoch": 1.77, |
| "learning_rate": 7.637830894771176e-06, |
| "loss": 0.4259, |
| "step": 1104 |
| }, |
| { |
| "epoch": 1.77, |
| "learning_rate": 7.621042974125701e-06, |
| "loss": 0.4351, |
| "step": 1105 |
| }, |
| { |
| "epoch": 1.77, |
| "learning_rate": 7.604262157407008e-06, |
| "loss": 0.417, |
| "step": 1106 |
| }, |
| { |
| "epoch": 1.77, |
| "learning_rate": 7.587488494725157e-06, |
| "loss": 0.4149, |
| "step": 1107 |
| }, |
| { |
| "epoch": 1.77, |
| "learning_rate": 7.570722036168855e-06, |
| "loss": 0.4563, |
| "step": 1108 |
| }, |
| { |
| "epoch": 1.77, |
| "learning_rate": 7.553962831805291e-06, |
| "loss": 0.4813, |
| "step": 1109 |
| }, |
| { |
| "epoch": 1.78, |
| "learning_rate": 7.537210931679988e-06, |
| "loss": 0.4256, |
| "step": 1110 |
| }, |
| { |
| "epoch": 1.78, |
| "learning_rate": 7.520466385816672e-06, |
| "loss": 0.4664, |
| "step": 1111 |
| }, |
| { |
| "epoch": 1.78, |
| "learning_rate": 7.5037292442170865e-06, |
| "loss": 0.4494, |
| "step": 1112 |
| }, |
| { |
| "epoch": 1.78, |
| "learning_rate": 7.48699955686089e-06, |
| "loss": 0.4694, |
| "step": 1113 |
| }, |
| { |
| "epoch": 1.78, |
| "learning_rate": 7.470277373705461e-06, |
| "loss": 0.4512, |
| "step": 1114 |
| }, |
| { |
| "epoch": 1.78, |
| "learning_rate": 7.453562744685779e-06, |
| "loss": 0.4643, |
| "step": 1115 |
| }, |
| { |
| "epoch": 1.79, |
| "learning_rate": 7.4368557197142596e-06, |
| "loss": 0.44, |
| "step": 1116 |
| }, |
| { |
| "epoch": 1.79, |
| "learning_rate": 7.420156348680621e-06, |
| "loss": 0.4749, |
| "step": 1117 |
| }, |
| { |
| "epoch": 1.79, |
| "learning_rate": 7.4034646814517155e-06, |
| "loss": 0.4445, |
| "step": 1118 |
| }, |
| { |
| "epoch": 1.79, |
| "learning_rate": 7.3867807678713965e-06, |
| "loss": 0.4526, |
| "step": 1119 |
| }, |
| { |
| "epoch": 1.79, |
| "learning_rate": 7.3701046577603605e-06, |
| "loss": 0.4375, |
| "step": 1120 |
| }, |
| { |
| "epoch": 1.79, |
| "learning_rate": 7.353436400916006e-06, |
| "loss": 0.4095, |
| "step": 1121 |
| }, |
| { |
| "epoch": 1.8, |
| "learning_rate": 7.336776047112277e-06, |
| "loss": 0.4484, |
| "step": 1122 |
| }, |
| { |
| "epoch": 1.8, |
| "learning_rate": 7.32012364609952e-06, |
| "loss": 0.4567, |
| "step": 1123 |
| }, |
| { |
| "epoch": 1.8, |
| "learning_rate": 7.303479247604333e-06, |
| "loss": 0.4522, |
| "step": 1124 |
| }, |
| { |
| "epoch": 1.8, |
| "learning_rate": 7.286842901329413e-06, |
| "loss": 0.4174, |
| "step": 1125 |
| }, |
| { |
| "epoch": 1.8, |
| "learning_rate": 7.270214656953415e-06, |
| "loss": 0.4281, |
| "step": 1126 |
| }, |
| { |
| "epoch": 1.8, |
| "learning_rate": 7.253594564130804e-06, |
| "loss": 0.4207, |
| "step": 1127 |
| }, |
| { |
| "epoch": 1.8, |
| "learning_rate": 7.236982672491699e-06, |
| "loss": 0.4654, |
| "step": 1128 |
| }, |
| { |
| "epoch": 1.81, |
| "learning_rate": 7.22037903164173e-06, |
| "loss": 0.4379, |
| "step": 1129 |
| }, |
| { |
| "epoch": 1.81, |
| "learning_rate": 7.203783691161883e-06, |
| "loss": 0.4494, |
| "step": 1130 |
| }, |
| { |
| "epoch": 1.81, |
| "learning_rate": 7.187196700608373e-06, |
| "loss": 0.4502, |
| "step": 1131 |
| }, |
| { |
| "epoch": 1.81, |
| "learning_rate": 7.170618109512465e-06, |
| "loss": 0.391, |
| "step": 1132 |
| }, |
| { |
| "epoch": 1.81, |
| "learning_rate": 7.154047967380353e-06, |
| "loss": 0.4679, |
| "step": 1133 |
| }, |
| { |
| "epoch": 1.81, |
| "learning_rate": 7.137486323692994e-06, |
| "loss": 0.4506, |
| "step": 1134 |
| }, |
| { |
| "epoch": 1.82, |
| "learning_rate": 7.120933227905971e-06, |
| "loss": 0.4317, |
| "step": 1135 |
| }, |
| { |
| "epoch": 1.82, |
| "learning_rate": 7.104388729449338e-06, |
| "loss": 0.4451, |
| "step": 1136 |
| }, |
| { |
| "epoch": 1.82, |
| "learning_rate": 7.0878528777274814e-06, |
| "loss": 0.4593, |
| "step": 1137 |
| }, |
| { |
| "epoch": 1.82, |
| "learning_rate": 7.0713257221189635e-06, |
| "loss": 0.4339, |
| "step": 1138 |
| }, |
| { |
| "epoch": 1.82, |
| "learning_rate": 7.05480731197638e-06, |
| "loss": 0.4392, |
| "step": 1139 |
| }, |
| { |
| "epoch": 1.82, |
| "learning_rate": 7.0382976966262065e-06, |
| "loss": 0.4344, |
| "step": 1140 |
| }, |
| { |
| "epoch": 1.83, |
| "learning_rate": 7.021796925368667e-06, |
| "loss": 0.4409, |
| "step": 1141 |
| }, |
| { |
| "epoch": 1.83, |
| "learning_rate": 7.005305047477566e-06, |
| "loss": 0.4597, |
| "step": 1142 |
| }, |
| { |
| "epoch": 1.83, |
| "learning_rate": 6.988822112200157e-06, |
| "loss": 0.4594, |
| "step": 1143 |
| }, |
| { |
| "epoch": 1.83, |
| "learning_rate": 6.9723481687569836e-06, |
| "loss": 0.4615, |
| "step": 1144 |
| }, |
| { |
| "epoch": 1.83, |
| "learning_rate": 6.955883266341741e-06, |
| "loss": 0.4488, |
| "step": 1145 |
| }, |
| { |
| "epoch": 1.83, |
| "learning_rate": 6.939427454121128e-06, |
| "loss": 0.4408, |
| "step": 1146 |
| }, |
| { |
| "epoch": 1.84, |
| "learning_rate": 6.9229807812346985e-06, |
| "loss": 0.4727, |
| "step": 1147 |
| }, |
| { |
| "epoch": 1.84, |
| "learning_rate": 6.9065432967947145e-06, |
| "loss": 0.4519, |
| "step": 1148 |
| }, |
| { |
| "epoch": 1.84, |
| "learning_rate": 6.890115049885995e-06, |
| "loss": 0.4502, |
| "step": 1149 |
| }, |
| { |
| "epoch": 1.84, |
| "learning_rate": 6.8736960895657854e-06, |
| "loss": 0.4276, |
| "step": 1150 |
| }, |
| { |
| "epoch": 1.84, |
| "learning_rate": 6.85728646486359e-06, |
| "loss": 0.4585, |
| "step": 1151 |
| }, |
| { |
| "epoch": 1.84, |
| "learning_rate": 6.840886224781039e-06, |
| "loss": 0.4563, |
| "step": 1152 |
| }, |
| { |
| "epoch": 1.84, |
| "learning_rate": 6.824495418291741e-06, |
| "loss": 0.4215, |
| "step": 1153 |
| }, |
| { |
| "epoch": 1.85, |
| "learning_rate": 6.8081140943411296e-06, |
| "loss": 0.4497, |
| "step": 1154 |
| }, |
| { |
| "epoch": 1.85, |
| "learning_rate": 6.791742301846325e-06, |
| "loss": 0.4537, |
| "step": 1155 |
| }, |
| { |
| "epoch": 1.85, |
| "learning_rate": 6.775380089695986e-06, |
| "loss": 0.4659, |
| "step": 1156 |
| }, |
| { |
| "epoch": 1.85, |
| "learning_rate": 6.759027506750159e-06, |
| "loss": 0.4451, |
| "step": 1157 |
| }, |
| { |
| "epoch": 1.85, |
| "learning_rate": 6.742684601840142e-06, |
| "loss": 0.4604, |
| "step": 1158 |
| }, |
| { |
| "epoch": 1.85, |
| "learning_rate": 6.726351423768323e-06, |
| "loss": 0.4596, |
| "step": 1159 |
| }, |
| { |
| "epoch": 1.86, |
| "learning_rate": 6.710028021308061e-06, |
| "loss": 0.4571, |
| "step": 1160 |
| }, |
| { |
| "epoch": 1.86, |
| "learning_rate": 6.693714443203507e-06, |
| "loss": 0.4494, |
| "step": 1161 |
| }, |
| { |
| "epoch": 1.86, |
| "learning_rate": 6.677410738169485e-06, |
| "loss": 0.4349, |
| "step": 1162 |
| }, |
| { |
| "epoch": 1.86, |
| "learning_rate": 6.661116954891329e-06, |
| "loss": 0.43, |
| "step": 1163 |
| }, |
| { |
| "epoch": 1.86, |
| "learning_rate": 6.644833142024752e-06, |
| "loss": 0.448, |
| "step": 1164 |
| }, |
| { |
| "epoch": 1.86, |
| "learning_rate": 6.62855934819569e-06, |
| "loss": 0.437, |
| "step": 1165 |
| }, |
| { |
| "epoch": 1.87, |
| "learning_rate": 6.612295622000162e-06, |
| "loss": 0.4481, |
| "step": 1166 |
| }, |
| { |
| "epoch": 1.87, |
| "learning_rate": 6.59604201200412e-06, |
| "loss": 0.4049, |
| "step": 1167 |
| }, |
| { |
| "epoch": 1.87, |
| "learning_rate": 6.579798566743314e-06, |
| "loss": 0.4411, |
| "step": 1168 |
| }, |
| { |
| "epoch": 1.87, |
| "learning_rate": 6.563565334723134e-06, |
| "loss": 0.4457, |
| "step": 1169 |
| }, |
| { |
| "epoch": 1.87, |
| "learning_rate": 6.547342364418482e-06, |
| "loss": 0.44, |
| "step": 1170 |
| }, |
| { |
| "epoch": 1.87, |
| "learning_rate": 6.5311297042736046e-06, |
| "loss": 0.4714, |
| "step": 1171 |
| }, |
| { |
| "epoch": 1.88, |
| "learning_rate": 6.514927402701965e-06, |
| "loss": 0.4487, |
| "step": 1172 |
| }, |
| { |
| "epoch": 1.88, |
| "learning_rate": 6.498735508086094e-06, |
| "loss": 0.4459, |
| "step": 1173 |
| }, |
| { |
| "epoch": 1.88, |
| "learning_rate": 6.482554068777451e-06, |
| "loss": 0.4505, |
| "step": 1174 |
| }, |
| { |
| "epoch": 1.88, |
| "learning_rate": 6.466383133096268e-06, |
| "loss": 0.4385, |
| "step": 1175 |
| }, |
| { |
| "epoch": 1.88, |
| "learning_rate": 6.450222749331414e-06, |
| "loss": 0.429, |
| "step": 1176 |
| }, |
| { |
| "epoch": 1.88, |
| "learning_rate": 6.4340729657402424e-06, |
| "loss": 0.4632, |
| "step": 1177 |
| }, |
| { |
| "epoch": 1.88, |
| "learning_rate": 6.4179338305484675e-06, |
| "loss": 0.4454, |
| "step": 1178 |
| }, |
| { |
| "epoch": 1.89, |
| "learning_rate": 6.40180539194999e-06, |
| "loss": 0.4443, |
| "step": 1179 |
| }, |
| { |
| "epoch": 1.89, |
| "learning_rate": 6.385687698106781e-06, |
| "loss": 0.4409, |
| "step": 1180 |
| }, |
| { |
| "epoch": 1.89, |
| "learning_rate": 6.3695807971487175e-06, |
| "loss": 0.4482, |
| "step": 1181 |
| }, |
| { |
| "epoch": 1.89, |
| "learning_rate": 6.35348473717345e-06, |
| "loss": 0.4258, |
| "step": 1182 |
| }, |
| { |
| "epoch": 1.89, |
| "learning_rate": 6.337399566246257e-06, |
| "loss": 0.4182, |
| "step": 1183 |
| }, |
| { |
| "epoch": 1.89, |
| "learning_rate": 6.321325332399904e-06, |
| "loss": 0.4315, |
| "step": 1184 |
| }, |
| { |
| "epoch": 1.9, |
| "learning_rate": 6.305262083634488e-06, |
| "loss": 0.4443, |
| "step": 1185 |
| }, |
| { |
| "epoch": 1.9, |
| "learning_rate": 6.289209867917312e-06, |
| "loss": 0.4495, |
| "step": 1186 |
| }, |
| { |
| "epoch": 1.9, |
| "learning_rate": 6.2731687331827214e-06, |
| "loss": 0.4299, |
| "step": 1187 |
| }, |
| { |
| "epoch": 1.9, |
| "learning_rate": 6.2571387273319905e-06, |
| "loss": 0.4601, |
| "step": 1188 |
| }, |
| { |
| "epoch": 1.9, |
| "learning_rate": 6.2411198982331435e-06, |
| "loss": 0.4327, |
| "step": 1189 |
| }, |
| { |
| "epoch": 1.9, |
| "learning_rate": 6.225112293720836e-06, |
| "loss": 0.4178, |
| "step": 1190 |
| }, |
| { |
| "epoch": 1.91, |
| "learning_rate": 6.209115961596208e-06, |
| "loss": 0.4473, |
| "step": 1191 |
| }, |
| { |
| "epoch": 1.91, |
| "learning_rate": 6.193130949626731e-06, |
| "loss": 0.4255, |
| "step": 1192 |
| }, |
| { |
| "epoch": 1.91, |
| "learning_rate": 6.177157305546077e-06, |
| "loss": 0.4509, |
| "step": 1193 |
| }, |
| { |
| "epoch": 1.91, |
| "learning_rate": 6.1611950770539766e-06, |
| "loss": 0.4751, |
| "step": 1194 |
| }, |
| { |
| "epoch": 1.91, |
| "learning_rate": 6.145244311816063e-06, |
| "loss": 0.4525, |
| "step": 1195 |
| }, |
| { |
| "epoch": 1.91, |
| "learning_rate": 6.129305057463741e-06, |
| "loss": 0.4418, |
| "step": 1196 |
| }, |
| { |
| "epoch": 1.92, |
| "learning_rate": 6.113377361594048e-06, |
| "loss": 0.4385, |
| "step": 1197 |
| }, |
| { |
| "epoch": 1.92, |
| "learning_rate": 6.0974612717695e-06, |
| "loss": 0.4043, |
| "step": 1198 |
| }, |
| { |
| "epoch": 1.92, |
| "learning_rate": 6.081556835517955e-06, |
| "loss": 0.4526, |
| "step": 1199 |
| }, |
| { |
| "epoch": 1.92, |
| "learning_rate": 6.065664100332478e-06, |
| "loss": 0.5019, |
| "step": 1200 |
| }, |
| { |
| "epoch": 1.92, |
| "learning_rate": 6.049783113671184e-06, |
| "loss": 0.4453, |
| "step": 1201 |
| }, |
| { |
| "epoch": 1.92, |
| "learning_rate": 6.033913922957112e-06, |
| "loss": 0.437, |
| "step": 1202 |
| }, |
| { |
| "epoch": 1.92, |
| "learning_rate": 6.018056575578075e-06, |
| "loss": 0.4379, |
| "step": 1203 |
| }, |
| { |
| "epoch": 1.93, |
| "learning_rate": 6.002211118886514e-06, |
| "loss": 0.4299, |
| "step": 1204 |
| }, |
| { |
| "epoch": 1.93, |
| "learning_rate": 5.986377600199371e-06, |
| "loss": 0.4589, |
| "step": 1205 |
| }, |
| { |
| "epoch": 1.93, |
| "learning_rate": 5.970556066797941e-06, |
| "loss": 0.4402, |
| "step": 1206 |
| }, |
| { |
| "epoch": 1.93, |
| "learning_rate": 5.9547465659277215e-06, |
| "loss": 0.48, |
| "step": 1207 |
| }, |
| { |
| "epoch": 1.93, |
| "learning_rate": 5.93894914479828e-06, |
| "loss": 0.4107, |
| "step": 1208 |
| }, |
| { |
| "epoch": 1.93, |
| "learning_rate": 5.923163850583114e-06, |
| "loss": 0.4379, |
| "step": 1209 |
| }, |
| { |
| "epoch": 1.94, |
| "learning_rate": 5.907390730419506e-06, |
| "loss": 0.4533, |
| "step": 1210 |
| }, |
| { |
| "epoch": 1.94, |
| "learning_rate": 5.891629831408392e-06, |
| "loss": 0.4855, |
| "step": 1211 |
| }, |
| { |
| "epoch": 1.94, |
| "learning_rate": 5.875881200614208e-06, |
| "loss": 0.4239, |
| "step": 1212 |
| }, |
| { |
| "epoch": 1.94, |
| "learning_rate": 5.8601448850647515e-06, |
| "loss": 0.4434, |
| "step": 1213 |
| }, |
| { |
| "epoch": 1.94, |
| "learning_rate": 5.8444209317510515e-06, |
| "loss": 0.4327, |
| "step": 1214 |
| }, |
| { |
| "epoch": 1.94, |
| "learning_rate": 5.828709387627219e-06, |
| "loss": 0.4485, |
| "step": 1215 |
| }, |
| { |
| "epoch": 1.95, |
| "learning_rate": 5.813010299610313e-06, |
| "loss": 0.4418, |
| "step": 1216 |
| }, |
| { |
| "epoch": 1.95, |
| "learning_rate": 5.797323714580192e-06, |
| "loss": 0.4414, |
| "step": 1217 |
| }, |
| { |
| "epoch": 1.95, |
| "learning_rate": 5.781649679379379e-06, |
| "loss": 0.4472, |
| "step": 1218 |
| }, |
| { |
| "epoch": 1.95, |
| "learning_rate": 5.7659882408129204e-06, |
| "loss": 0.4336, |
| "step": 1219 |
| }, |
| { |
| "epoch": 1.95, |
| "learning_rate": 5.750339445648252e-06, |
| "loss": 0.4337, |
| "step": 1220 |
| }, |
| { |
| "epoch": 1.95, |
| "learning_rate": 5.7347033406150494e-06, |
| "loss": 0.4526, |
| "step": 1221 |
| }, |
| { |
| "epoch": 1.96, |
| "learning_rate": 5.7190799724050924e-06, |
| "loss": 0.4367, |
| "step": 1222 |
| }, |
| { |
| "epoch": 1.96, |
| "learning_rate": 5.703469387672138e-06, |
| "loss": 0.4313, |
| "step": 1223 |
| }, |
| { |
| "epoch": 1.96, |
| "learning_rate": 5.687871633031754e-06, |
| "loss": 0.4525, |
| "step": 1224 |
| }, |
| { |
| "epoch": 1.96, |
| "learning_rate": 5.672286755061212e-06, |
| "loss": 0.4355, |
| "step": 1225 |
| }, |
| { |
| "epoch": 1.96, |
| "learning_rate": 5.656714800299317e-06, |
| "loss": 0.4318, |
| "step": 1226 |
| }, |
| { |
| "epoch": 1.96, |
| "learning_rate": 5.64115581524629e-06, |
| "loss": 0.4624, |
| "step": 1227 |
| }, |
| { |
| "epoch": 1.96, |
| "learning_rate": 5.625609846363622e-06, |
| "loss": 0.4117, |
| "step": 1228 |
| }, |
| { |
| "epoch": 1.97, |
| "learning_rate": 5.610076940073939e-06, |
| "loss": 0.4771, |
| "step": 1229 |
| }, |
| { |
| "epoch": 1.97, |
| "learning_rate": 5.594557142760853e-06, |
| "loss": 0.4592, |
| "step": 1230 |
| }, |
| { |
| "epoch": 1.97, |
| "learning_rate": 5.579050500768837e-06, |
| "loss": 0.4518, |
| "step": 1231 |
| }, |
| { |
| "epoch": 1.97, |
| "learning_rate": 5.563557060403071e-06, |
| "loss": 0.4525, |
| "step": 1232 |
| }, |
| { |
| "epoch": 1.97, |
| "learning_rate": 5.548076867929331e-06, |
| "loss": 0.439, |
| "step": 1233 |
| }, |
| { |
| "epoch": 1.97, |
| "learning_rate": 5.53260996957381e-06, |
| "loss": 0.4141, |
| "step": 1234 |
| }, |
| { |
| "epoch": 1.98, |
| "learning_rate": 5.517156411523026e-06, |
| "loss": 0.439, |
| "step": 1235 |
| }, |
| { |
| "epoch": 1.98, |
| "learning_rate": 5.501716239923642e-06, |
| "loss": 0.4404, |
| "step": 1236 |
| }, |
| { |
| "epoch": 1.98, |
| "learning_rate": 5.486289500882355e-06, |
| "loss": 0.4588, |
| "step": 1237 |
| }, |
| { |
| "epoch": 1.98, |
| "learning_rate": 5.47087624046575e-06, |
| "loss": 0.4488, |
| "step": 1238 |
| }, |
| { |
| "epoch": 1.98, |
| "learning_rate": 5.455476504700161e-06, |
| "loss": 0.4412, |
| "step": 1239 |
| }, |
| { |
| "epoch": 1.98, |
| "learning_rate": 5.440090339571537e-06, |
| "loss": 0.4132, |
| "step": 1240 |
| }, |
| { |
| "epoch": 1.99, |
| "learning_rate": 5.424717791025302e-06, |
| "loss": 0.4451, |
| "step": 1241 |
| }, |
| { |
| "epoch": 1.99, |
| "learning_rate": 5.4093589049662175e-06, |
| "loss": 0.4456, |
| "step": 1242 |
| }, |
| { |
| "epoch": 1.99, |
| "learning_rate": 5.3940137272582534e-06, |
| "loss": 0.4503, |
| "step": 1243 |
| }, |
| { |
| "epoch": 1.99, |
| "learning_rate": 5.378682303724435e-06, |
| "loss": 0.4151, |
| "step": 1244 |
| }, |
| { |
| "epoch": 1.99, |
| "learning_rate": 5.3633646801467255e-06, |
| "loss": 0.4488, |
| "step": 1245 |
| }, |
| { |
| "epoch": 1.99, |
| "learning_rate": 5.348060902265871e-06, |
| "loss": 0.4536, |
| "step": 1246 |
| }, |
| { |
| "epoch": 2.0, |
| "learning_rate": 5.332771015781275e-06, |
| "loss": 0.4289, |
| "step": 1247 |
| }, |
| { |
| "epoch": 2.0, |
| "learning_rate": 5.31749506635086e-06, |
| "loss": 0.4533, |
| "step": 1248 |
| }, |
| { |
| "epoch": 2.0, |
| "learning_rate": 5.302233099590928e-06, |
| "loss": 0.4432, |
| "step": 1249 |
| }, |
| { |
| "epoch": 2.0, |
| "learning_rate": 5.286985161076029e-06, |
| "loss": 0.4484, |
| "step": 1250 |
| }, |
| { |
| "epoch": 2.0, |
| "learning_rate": 5.271751296338823e-06, |
| "loss": 0.3634, |
| "step": 1251 |
| }, |
| { |
| "epoch": 2.0, |
| "learning_rate": 5.2565315508699374e-06, |
| "loss": 0.3413, |
| "step": 1252 |
| }, |
| { |
| "epoch": 2.0, |
| "learning_rate": 5.241325970117851e-06, |
| "loss": 0.3738, |
| "step": 1253 |
| }, |
| { |
| "epoch": 2.01, |
| "learning_rate": 5.226134599488728e-06, |
| "loss": 0.3488, |
| "step": 1254 |
| }, |
| { |
| "epoch": 2.01, |
| "learning_rate": 5.210957484346314e-06, |
| "loss": 0.3397, |
| "step": 1255 |
| }, |
| { |
| "epoch": 2.01, |
| "learning_rate": 5.195794670011775e-06, |
| "loss": 0.3376, |
| "step": 1256 |
| }, |
| { |
| "epoch": 2.01, |
| "learning_rate": 5.1806462017635775e-06, |
| "loss": 0.331, |
| "step": 1257 |
| }, |
| { |
| "epoch": 2.01, |
| "learning_rate": 5.165512124837344e-06, |
| "loss": 0.3429, |
| "step": 1258 |
| }, |
| { |
| "epoch": 2.01, |
| "learning_rate": 5.150392484425728e-06, |
| "loss": 0.3554, |
| "step": 1259 |
| }, |
| { |
| "epoch": 2.02, |
| "learning_rate": 5.135287325678271e-06, |
| "loss": 0.3557, |
| "step": 1260 |
| }, |
| { |
| "epoch": 2.02, |
| "learning_rate": 5.120196693701267e-06, |
| "loss": 0.3591, |
| "step": 1261 |
| }, |
| { |
| "epoch": 2.02, |
| "learning_rate": 5.105120633557634e-06, |
| "loss": 0.3375, |
| "step": 1262 |
| }, |
| { |
| "epoch": 2.02, |
| "learning_rate": 5.090059190266779e-06, |
| "loss": 0.3412, |
| "step": 1263 |
| }, |
| { |
| "epoch": 2.02, |
| "learning_rate": 5.075012408804458e-06, |
| "loss": 0.3375, |
| "step": 1264 |
| }, |
| { |
| "epoch": 2.02, |
| "learning_rate": 5.059980334102637e-06, |
| "loss": 0.3576, |
| "step": 1265 |
| }, |
| { |
| "epoch": 2.03, |
| "learning_rate": 5.044963011049384e-06, |
| "loss": 0.3386, |
| "step": 1266 |
| }, |
| { |
| "epoch": 2.03, |
| "learning_rate": 5.0299604844886985e-06, |
| "loss": 0.3239, |
| "step": 1267 |
| }, |
| { |
| "epoch": 2.03, |
| "learning_rate": 5.0149727992204034e-06, |
| "loss": 0.3163, |
| "step": 1268 |
| }, |
| { |
| "epoch": 2.03, |
| "learning_rate": 5.000000000000003e-06, |
| "loss": 0.3516, |
| "step": 1269 |
| }, |
| { |
| "epoch": 2.03, |
| "learning_rate": 4.985042131538545e-06, |
| "loss": 0.3512, |
| "step": 1270 |
| }, |
| { |
| "epoch": 2.03, |
| "learning_rate": 4.970099238502494e-06, |
| "loss": 0.3335, |
| "step": 1271 |
| }, |
| { |
| "epoch": 2.04, |
| "learning_rate": 4.955171365513603e-06, |
| "loss": 0.3572, |
| "step": 1272 |
| }, |
| { |
| "epoch": 2.04, |
| "learning_rate": 4.940258557148765e-06, |
| "loss": 0.3246, |
| "step": 1273 |
| }, |
| { |
| "epoch": 2.04, |
| "learning_rate": 4.925360857939886e-06, |
| "loss": 0.3342, |
| "step": 1274 |
| }, |
| { |
| "epoch": 2.04, |
| "learning_rate": 4.910478312373757e-06, |
| "loss": 0.3475, |
| "step": 1275 |
| }, |
| { |
| "epoch": 2.04, |
| "learning_rate": 4.895610964891923e-06, |
| "loss": 0.3521, |
| "step": 1276 |
| }, |
| { |
| "epoch": 2.04, |
| "learning_rate": 4.8807588598905364e-06, |
| "loss": 0.3199, |
| "step": 1277 |
| }, |
| { |
| "epoch": 2.04, |
| "learning_rate": 4.865922041720239e-06, |
| "loss": 0.3312, |
| "step": 1278 |
| }, |
| { |
| "epoch": 2.05, |
| "learning_rate": 4.8511005546860214e-06, |
| "loss": 0.3336, |
| "step": 1279 |
| }, |
| { |
| "epoch": 2.05, |
| "learning_rate": 4.836294443047088e-06, |
| "loss": 0.3443, |
| "step": 1280 |
| }, |
| { |
| "epoch": 2.05, |
| "learning_rate": 4.821503751016746e-06, |
| "loss": 0.3575, |
| "step": 1281 |
| }, |
| { |
| "epoch": 2.05, |
| "learning_rate": 4.806728522762241e-06, |
| "loss": 0.3476, |
| "step": 1282 |
| }, |
| { |
| "epoch": 2.05, |
| "learning_rate": 4.791968802404648e-06, |
| "loss": 0.3536, |
| "step": 1283 |
| }, |
| { |
| "epoch": 2.05, |
| "learning_rate": 4.777224634018732e-06, |
| "loss": 0.3399, |
| "step": 1284 |
| }, |
| { |
| "epoch": 2.06, |
| "learning_rate": 4.762496061632814e-06, |
| "loss": 0.3323, |
| "step": 1285 |
| }, |
| { |
| "epoch": 2.06, |
| "learning_rate": 4.7477831292286555e-06, |
| "loss": 0.3524, |
| "step": 1286 |
| }, |
| { |
| "epoch": 2.06, |
| "learning_rate": 4.733085880741301e-06, |
| "loss": 0.3366, |
| "step": 1287 |
| }, |
| { |
| "epoch": 2.06, |
| "learning_rate": 4.7184043600589655e-06, |
| "loss": 0.3678, |
| "step": 1288 |
| }, |
| { |
| "epoch": 2.06, |
| "learning_rate": 4.703738611022899e-06, |
| "loss": 0.343, |
| "step": 1289 |
| }, |
| { |
| "epoch": 2.06, |
| "learning_rate": 4.689088677427249e-06, |
| "loss": 0.3182, |
| "step": 1290 |
| }, |
| { |
| "epoch": 2.07, |
| "learning_rate": 4.674454603018949e-06, |
| "loss": 0.3395, |
| "step": 1291 |
| }, |
| { |
| "epoch": 2.07, |
| "learning_rate": 4.659836431497563e-06, |
| "loss": 0.3523, |
| "step": 1292 |
| }, |
| { |
| "epoch": 2.07, |
| "learning_rate": 4.645234206515171e-06, |
| "loss": 0.3357, |
| "step": 1293 |
| }, |
| { |
| "epoch": 2.07, |
| "learning_rate": 4.630647971676232e-06, |
| "loss": 0.3256, |
| "step": 1294 |
| }, |
| { |
| "epoch": 2.07, |
| "learning_rate": 4.616077770537453e-06, |
| "loss": 0.3656, |
| "step": 1295 |
| }, |
| { |
| "epoch": 2.07, |
| "learning_rate": 4.601523646607675e-06, |
| "loss": 0.3368, |
| "step": 1296 |
| }, |
| { |
| "epoch": 2.08, |
| "learning_rate": 4.586985643347716e-06, |
| "loss": 0.3209, |
| "step": 1297 |
| }, |
| { |
| "epoch": 2.08, |
| "learning_rate": 4.572463804170263e-06, |
| "loss": 0.3571, |
| "step": 1298 |
| }, |
| { |
| "epoch": 2.08, |
| "learning_rate": 4.557958172439726e-06, |
| "loss": 0.3437, |
| "step": 1299 |
| }, |
| { |
| "epoch": 2.08, |
| "learning_rate": 4.543468791472131e-06, |
| "loss": 0.3474, |
| "step": 1300 |
| }, |
| { |
| "epoch": 2.08, |
| "learning_rate": 4.5289957045349655e-06, |
| "loss": 0.338, |
| "step": 1301 |
| }, |
| { |
| "epoch": 2.08, |
| "learning_rate": 4.5145389548470645e-06, |
| "loss": 0.3305, |
| "step": 1302 |
| }, |
| { |
| "epoch": 2.08, |
| "learning_rate": 4.500098585578475e-06, |
| "loss": 0.3243, |
| "step": 1303 |
| }, |
| { |
| "epoch": 2.09, |
| "learning_rate": 4.485674639850334e-06, |
| "loss": 0.3412, |
| "step": 1304 |
| }, |
| { |
| "epoch": 2.09, |
| "learning_rate": 4.471267160734731e-06, |
| "loss": 0.3241, |
| "step": 1305 |
| }, |
| { |
| "epoch": 2.09, |
| "learning_rate": 4.456876191254582e-06, |
| "loss": 0.3099, |
| "step": 1306 |
| }, |
| { |
| "epoch": 2.09, |
| "learning_rate": 4.4425017743835155e-06, |
| "loss": 0.3414, |
| "step": 1307 |
| }, |
| { |
| "epoch": 2.09, |
| "learning_rate": 4.4281439530457174e-06, |
| "loss": 0.3437, |
| "step": 1308 |
| }, |
| { |
| "epoch": 2.09, |
| "learning_rate": 4.413802770115816e-06, |
| "loss": 0.3396, |
| "step": 1309 |
| }, |
| { |
| "epoch": 2.1, |
| "learning_rate": 4.399478268418771e-06, |
| "loss": 0.3288, |
| "step": 1310 |
| }, |
| { |
| "epoch": 2.1, |
| "learning_rate": 4.385170490729712e-06, |
| "loss": 0.3215, |
| "step": 1311 |
| }, |
| { |
| "epoch": 2.1, |
| "learning_rate": 4.370879479773837e-06, |
| "loss": 0.3286, |
| "step": 1312 |
| }, |
| { |
| "epoch": 2.1, |
| "learning_rate": 4.356605278226274e-06, |
| "loss": 0.3369, |
| "step": 1313 |
| }, |
| { |
| "epoch": 2.1, |
| "learning_rate": 4.342347928711953e-06, |
| "loss": 0.3633, |
| "step": 1314 |
| }, |
| { |
| "epoch": 2.1, |
| "learning_rate": 4.328107473805487e-06, |
| "loss": 0.3335, |
| "step": 1315 |
| }, |
| { |
| "epoch": 2.11, |
| "learning_rate": 4.313883956031031e-06, |
| "loss": 0.3411, |
| "step": 1316 |
| }, |
| { |
| "epoch": 2.11, |
| "learning_rate": 4.299677417862174e-06, |
| "loss": 0.3255, |
| "step": 1317 |
| }, |
| { |
| "epoch": 2.11, |
| "learning_rate": 4.28548790172179e-06, |
| "loss": 0.3102, |
| "step": 1318 |
| }, |
| { |
| "epoch": 2.11, |
| "learning_rate": 4.2713154499819345e-06, |
| "loss": 0.3386, |
| "step": 1319 |
| }, |
| { |
| "epoch": 2.11, |
| "learning_rate": 4.257160104963695e-06, |
| "loss": 0.3615, |
| "step": 1320 |
| }, |
| { |
| "epoch": 2.11, |
| "learning_rate": 4.243021908937083e-06, |
| "loss": 0.3537, |
| "step": 1321 |
| }, |
| { |
| "epoch": 2.12, |
| "learning_rate": 4.228900904120895e-06, |
| "loss": 0.3493, |
| "step": 1322 |
| }, |
| { |
| "epoch": 2.12, |
| "learning_rate": 4.214797132682597e-06, |
| "loss": 0.3388, |
| "step": 1323 |
| }, |
| { |
| "epoch": 2.12, |
| "learning_rate": 4.200710636738189e-06, |
| "loss": 0.3484, |
| "step": 1324 |
| }, |
| { |
| "epoch": 2.12, |
| "learning_rate": 4.186641458352088e-06, |
| "loss": 0.3386, |
| "step": 1325 |
| }, |
| { |
| "epoch": 2.12, |
| "learning_rate": 4.172589639536992e-06, |
| "loss": 0.3445, |
| "step": 1326 |
| }, |
| { |
| "epoch": 2.12, |
| "learning_rate": 4.158555222253772e-06, |
| "loss": 0.3405, |
| "step": 1327 |
| }, |
| { |
| "epoch": 2.12, |
| "learning_rate": 4.144538248411321e-06, |
| "loss": 0.3397, |
| "step": 1328 |
| }, |
| { |
| "epoch": 2.13, |
| "learning_rate": 4.130538759866457e-06, |
| "loss": 0.3329, |
| "step": 1329 |
| }, |
| { |
| "epoch": 2.13, |
| "learning_rate": 4.116556798423776e-06, |
| "loss": 0.3512, |
| "step": 1330 |
| }, |
| { |
| "epoch": 2.13, |
| "learning_rate": 4.102592405835536e-06, |
| "loss": 0.3262, |
| "step": 1331 |
| }, |
| { |
| "epoch": 2.13, |
| "learning_rate": 4.088645623801534e-06, |
| "loss": 0.3191, |
| "step": 1332 |
| }, |
| { |
| "epoch": 2.13, |
| "learning_rate": 4.074716493968976e-06, |
| "loss": 0.3481, |
| "step": 1333 |
| }, |
| { |
| "epoch": 2.13, |
| "learning_rate": 4.060805057932359e-06, |
| "loss": 0.329, |
| "step": 1334 |
| }, |
| { |
| "epoch": 2.14, |
| "learning_rate": 4.046911357233343e-06, |
| "loss": 0.3397, |
| "step": 1335 |
| }, |
| { |
| "epoch": 2.14, |
| "learning_rate": 4.033035433360624e-06, |
| "loss": 0.3198, |
| "step": 1336 |
| }, |
| { |
| "epoch": 2.14, |
| "learning_rate": 4.019177327749822e-06, |
| "loss": 0.3308, |
| "step": 1337 |
| }, |
| { |
| "epoch": 2.14, |
| "learning_rate": 4.00533708178334e-06, |
| "loss": 0.3244, |
| "step": 1338 |
| }, |
| { |
| "epoch": 2.14, |
| "learning_rate": 3.991514736790259e-06, |
| "loss": 0.3103, |
| "step": 1339 |
| }, |
| { |
| "epoch": 2.14, |
| "learning_rate": 3.977710334046193e-06, |
| "loss": 0.3348, |
| "step": 1340 |
| }, |
| { |
| "epoch": 2.15, |
| "learning_rate": 3.9639239147731865e-06, |
| "loss": 0.3552, |
| "step": 1341 |
| }, |
| { |
| "epoch": 2.15, |
| "learning_rate": 3.950155520139581e-06, |
| "loss": 0.3439, |
| "step": 1342 |
| }, |
| { |
| "epoch": 2.15, |
| "learning_rate": 3.936405191259891e-06, |
| "loss": 0.3185, |
| "step": 1343 |
| }, |
| { |
| "epoch": 2.15, |
| "learning_rate": 3.9226729691946865e-06, |
| "loss": 0.3579, |
| "step": 1344 |
| }, |
| { |
| "epoch": 2.15, |
| "learning_rate": 3.908958894950465e-06, |
| "loss": 0.3448, |
| "step": 1345 |
| }, |
| { |
| "epoch": 2.15, |
| "learning_rate": 3.895263009479534e-06, |
| "loss": 0.3281, |
| "step": 1346 |
| }, |
| { |
| "epoch": 2.16, |
| "learning_rate": 3.881585353679891e-06, |
| "loss": 0.3281, |
| "step": 1347 |
| }, |
| { |
| "epoch": 2.16, |
| "learning_rate": 3.867925968395085e-06, |
| "loss": 0.3346, |
| "step": 1348 |
| }, |
| { |
| "epoch": 2.16, |
| "learning_rate": 3.854284894414122e-06, |
| "loss": 0.3315, |
| "step": 1349 |
| }, |
| { |
| "epoch": 2.16, |
| "learning_rate": 3.840662172471315e-06, |
| "loss": 0.3525, |
| "step": 1350 |
| }, |
| { |
| "epoch": 2.16, |
| "learning_rate": 3.827057843246181e-06, |
| "loss": 0.3271, |
| "step": 1351 |
| }, |
| { |
| "epoch": 2.16, |
| "learning_rate": 3.8134719473633098e-06, |
| "loss": 0.3521, |
| "step": 1352 |
| }, |
| { |
| "epoch": 2.16, |
| "learning_rate": 3.799904525392251e-06, |
| "loss": 0.3297, |
| "step": 1353 |
| }, |
| { |
| "epoch": 2.17, |
| "learning_rate": 3.786355617847385e-06, |
| "loss": 0.334, |
| "step": 1354 |
| }, |
| { |
| "epoch": 2.17, |
| "learning_rate": 3.7728252651878018e-06, |
| "loss": 0.341, |
| "step": 1355 |
| }, |
| { |
| "epoch": 2.17, |
| "learning_rate": 3.759313507817196e-06, |
| "loss": 0.359, |
| "step": 1356 |
| }, |
| { |
| "epoch": 2.17, |
| "learning_rate": 3.745820386083724e-06, |
| "loss": 0.3282, |
| "step": 1357 |
| }, |
| { |
| "epoch": 2.17, |
| "learning_rate": 3.7323459402798936e-06, |
| "loss": 0.3487, |
| "step": 1358 |
| }, |
| { |
| "epoch": 2.17, |
| "learning_rate": 3.718890210642442e-06, |
| "loss": 0.3197, |
| "step": 1359 |
| }, |
| { |
| "epoch": 2.18, |
| "learning_rate": 3.705453237352227e-06, |
| "loss": 0.322, |
| "step": 1360 |
| }, |
| { |
| "epoch": 2.18, |
| "learning_rate": 3.6920350605340883e-06, |
| "loss": 0.3509, |
| "step": 1361 |
| }, |
| { |
| "epoch": 2.18, |
| "learning_rate": 3.6786357202567367e-06, |
| "loss": 0.3232, |
| "step": 1362 |
| }, |
| { |
| "epoch": 2.18, |
| "learning_rate": 3.6652552565326382e-06, |
| "loss": 0.311, |
| "step": 1363 |
| }, |
| { |
| "epoch": 2.18, |
| "learning_rate": 3.6518937093178873e-06, |
| "loss": 0.3238, |
| "step": 1364 |
| }, |
| { |
| "epoch": 2.18, |
| "learning_rate": 3.638551118512089e-06, |
| "loss": 0.3374, |
| "step": 1365 |
| }, |
| { |
| "epoch": 2.19, |
| "learning_rate": 3.6252275239582522e-06, |
| "loss": 0.3329, |
| "step": 1366 |
| }, |
| { |
| "epoch": 2.19, |
| "learning_rate": 3.611922965442648e-06, |
| "loss": 0.3201, |
| "step": 1367 |
| }, |
| { |
| "epoch": 2.19, |
| "learning_rate": 3.5986374826947067e-06, |
| "loss": 0.3364, |
| "step": 1368 |
| }, |
| { |
| "epoch": 2.19, |
| "learning_rate": 3.5853711153868962e-06, |
| "loss": 0.3194, |
| "step": 1369 |
| }, |
| { |
| "epoch": 2.19, |
| "learning_rate": 3.5721239031346067e-06, |
| "loss": 0.3495, |
| "step": 1370 |
| }, |
| { |
| "epoch": 2.19, |
| "learning_rate": 3.558895885496023e-06, |
| "loss": 0.3215, |
| "step": 1371 |
| }, |
| { |
| "epoch": 2.2, |
| "learning_rate": 3.545687101972013e-06, |
| "loss": 0.3311, |
| "step": 1372 |
| }, |
| { |
| "epoch": 2.2, |
| "learning_rate": 3.53249759200601e-06, |
| "loss": 0.325, |
| "step": 1373 |
| }, |
| { |
| "epoch": 2.2, |
| "learning_rate": 3.519327394983888e-06, |
| "loss": 0.3246, |
| "step": 1374 |
| }, |
| { |
| "epoch": 2.2, |
| "learning_rate": 3.506176550233863e-06, |
| "loss": 0.3285, |
| "step": 1375 |
| }, |
| { |
| "epoch": 2.2, |
| "learning_rate": 3.4930450970263485e-06, |
| "loss": 0.3488, |
| "step": 1376 |
| }, |
| { |
| "epoch": 2.2, |
| "learning_rate": 3.479933074573858e-06, |
| "loss": 0.3272, |
| "step": 1377 |
| }, |
| { |
| "epoch": 2.2, |
| "learning_rate": 3.4668405220308797e-06, |
| "loss": 0.3279, |
| "step": 1378 |
| }, |
| { |
| "epoch": 2.21, |
| "learning_rate": 3.453767478493761e-06, |
| "loss": 0.3676, |
| "step": 1379 |
| }, |
| { |
| "epoch": 2.21, |
| "learning_rate": 3.440713983000601e-06, |
| "loss": 0.3239, |
| "step": 1380 |
| }, |
| { |
| "epoch": 2.21, |
| "learning_rate": 3.4276800745311135e-06, |
| "loss": 0.3382, |
| "step": 1381 |
| }, |
| { |
| "epoch": 2.21, |
| "learning_rate": 3.4146657920065286e-06, |
| "loss": 0.322, |
| "step": 1382 |
| }, |
| { |
| "epoch": 2.21, |
| "learning_rate": 3.401671174289469e-06, |
| "loss": 0.3417, |
| "step": 1383 |
| }, |
| { |
| "epoch": 2.21, |
| "learning_rate": 3.3886962601838327e-06, |
| "loss": 0.3516, |
| "step": 1384 |
| }, |
| { |
| "epoch": 2.22, |
| "learning_rate": 3.37574108843469e-06, |
| "loss": 0.3361, |
| "step": 1385 |
| }, |
| { |
| "epoch": 2.22, |
| "learning_rate": 3.3628056977281456e-06, |
| "loss": 0.3605, |
| "step": 1386 |
| }, |
| { |
| "epoch": 2.22, |
| "learning_rate": 3.3498901266912397e-06, |
| "loss": 0.3311, |
| "step": 1387 |
| }, |
| { |
| "epoch": 2.22, |
| "learning_rate": 3.3369944138918286e-06, |
| "loss": 0.3319, |
| "step": 1388 |
| }, |
| { |
| "epoch": 2.22, |
| "learning_rate": 3.3241185978384636e-06, |
| "loss": 0.3445, |
| "step": 1389 |
| }, |
| { |
| "epoch": 2.22, |
| "learning_rate": 3.3112627169802948e-06, |
| "loss": 0.3407, |
| "step": 1390 |
| }, |
| { |
| "epoch": 2.23, |
| "learning_rate": 3.2984268097069284e-06, |
| "loss": 0.3208, |
| "step": 1391 |
| }, |
| { |
| "epoch": 2.23, |
| "learning_rate": 3.2856109143483316e-06, |
| "loss": 0.3327, |
| "step": 1392 |
| }, |
| { |
| "epoch": 2.23, |
| "learning_rate": 3.2728150691747117e-06, |
| "loss": 0.3356, |
| "step": 1393 |
| }, |
| { |
| "epoch": 2.23, |
| "learning_rate": 3.2600393123964114e-06, |
| "loss": 0.3217, |
| "step": 1394 |
| }, |
| { |
| "epoch": 2.23, |
| "learning_rate": 3.2472836821637744e-06, |
| "loss": 0.341, |
| "step": 1395 |
| }, |
| { |
| "epoch": 2.23, |
| "learning_rate": 3.2345482165670493e-06, |
| "loss": 0.3597, |
| "step": 1396 |
| }, |
| { |
| "epoch": 2.24, |
| "learning_rate": 3.22183295363627e-06, |
| "loss": 0.3255, |
| "step": 1397 |
| }, |
| { |
| "epoch": 2.24, |
| "learning_rate": 3.209137931341143e-06, |
| "loss": 0.3266, |
| "step": 1398 |
| }, |
| { |
| "epoch": 2.24, |
| "learning_rate": 3.196463187590929e-06, |
| "loss": 0.3441, |
| "step": 1399 |
| }, |
| { |
| "epoch": 2.24, |
| "learning_rate": 3.183808760234335e-06, |
| "loss": 0.3407, |
| "step": 1400 |
| }, |
| { |
| "epoch": 2.24, |
| "learning_rate": 3.1711746870594083e-06, |
| "loss": 0.3436, |
| "step": 1401 |
| }, |
| { |
| "epoch": 2.24, |
| "learning_rate": 3.1585610057934022e-06, |
| "loss": 0.3333, |
| "step": 1402 |
| }, |
| { |
| "epoch": 2.24, |
| "learning_rate": 3.145967754102691e-06, |
| "loss": 0.3487, |
| "step": 1403 |
| }, |
| { |
| "epoch": 2.25, |
| "learning_rate": 3.1333949695926323e-06, |
| "loss": 0.3219, |
| "step": 1404 |
| }, |
| { |
| "epoch": 2.25, |
| "learning_rate": 3.1208426898074685e-06, |
| "loss": 0.3244, |
| "step": 1405 |
| }, |
| { |
| "epoch": 2.25, |
| "learning_rate": 3.1083109522302124e-06, |
| "loss": 0.3192, |
| "step": 1406 |
| }, |
| { |
| "epoch": 2.25, |
| "learning_rate": 3.0957997942825337e-06, |
| "loss": 0.3328, |
| "step": 1407 |
| }, |
| { |
| "epoch": 2.25, |
| "learning_rate": 3.083309253324651e-06, |
| "loss": 0.3389, |
| "step": 1408 |
| }, |
| { |
| "epoch": 2.25, |
| "learning_rate": 3.070839366655215e-06, |
| "loss": 0.3262, |
| "step": 1409 |
| }, |
| { |
| "epoch": 2.26, |
| "learning_rate": 3.0583901715111965e-06, |
| "loss": 0.3604, |
| "step": 1410 |
| }, |
| { |
| "epoch": 2.26, |
| "learning_rate": 3.045961705067787e-06, |
| "loss": 0.3392, |
| "step": 1411 |
| }, |
| { |
| "epoch": 2.26, |
| "learning_rate": 3.0335540044382693e-06, |
| "loss": 0.3255, |
| "step": 1412 |
| }, |
| { |
| "epoch": 2.26, |
| "learning_rate": 3.021167106673928e-06, |
| "loss": 0.3358, |
| "step": 1413 |
| }, |
| { |
| "epoch": 2.26, |
| "learning_rate": 3.008801048763914e-06, |
| "loss": 0.3463, |
| "step": 1414 |
| }, |
| { |
| "epoch": 2.26, |
| "learning_rate": 2.996455867635155e-06, |
| "loss": 0.3384, |
| "step": 1415 |
| }, |
| { |
| "epoch": 2.27, |
| "learning_rate": 2.9841316001522345e-06, |
| "loss": 0.3229, |
| "step": 1416 |
| }, |
| { |
| "epoch": 2.27, |
| "learning_rate": 2.9718282831172885e-06, |
| "loss": 0.3485, |
| "step": 1417 |
| }, |
| { |
| "epoch": 2.27, |
| "learning_rate": 2.9595459532698854e-06, |
| "loss": 0.3325, |
| "step": 1418 |
| }, |
| { |
| "epoch": 2.27, |
| "learning_rate": 2.94728464728693e-06, |
| "loss": 0.3433, |
| "step": 1419 |
| }, |
| { |
| "epoch": 2.27, |
| "learning_rate": 2.9350444017825385e-06, |
| "loss": 0.3513, |
| "step": 1420 |
| }, |
| { |
| "epoch": 2.27, |
| "learning_rate": 2.922825253307947e-06, |
| "loss": 0.329, |
| "step": 1421 |
| }, |
| { |
| "epoch": 2.28, |
| "learning_rate": 2.910627238351383e-06, |
| "loss": 0.3426, |
| "step": 1422 |
| }, |
| { |
| "epoch": 2.28, |
| "learning_rate": 2.898450393337977e-06, |
| "loss": 0.3283, |
| "step": 1423 |
| }, |
| { |
| "epoch": 2.28, |
| "learning_rate": 2.886294754629632e-06, |
| "loss": 0.3261, |
| "step": 1424 |
| }, |
| { |
| "epoch": 2.28, |
| "learning_rate": 2.8741603585249312e-06, |
| "loss": 0.3445, |
| "step": 1425 |
| }, |
| { |
| "epoch": 2.28, |
| "learning_rate": 2.8620472412590227e-06, |
| "loss": 0.3452, |
| "step": 1426 |
| }, |
| { |
| "epoch": 2.28, |
| "learning_rate": 2.8499554390035144e-06, |
| "loss": 0.3199, |
| "step": 1427 |
| }, |
| { |
| "epoch": 2.28, |
| "learning_rate": 2.837884987866363e-06, |
| "loss": 0.3301, |
| "step": 1428 |
| }, |
| { |
| "epoch": 2.29, |
| "learning_rate": 2.8258359238917665e-06, |
| "loss": 0.3511, |
| "step": 1429 |
| }, |
| { |
| "epoch": 2.29, |
| "learning_rate": 2.8138082830600556e-06, |
| "loss": 0.3434, |
| "step": 1430 |
| }, |
| { |
| "epoch": 2.29, |
| "learning_rate": 2.8018021012875994e-06, |
| "loss": 0.3643, |
| "step": 1431 |
| }, |
| { |
| "epoch": 2.29, |
| "learning_rate": 2.789817414426673e-06, |
| "loss": 0.3538, |
| "step": 1432 |
| }, |
| { |
| "epoch": 2.29, |
| "learning_rate": 2.7778542582653746e-06, |
| "loss": 0.3239, |
| "step": 1433 |
| }, |
| { |
| "epoch": 2.29, |
| "learning_rate": 2.7659126685275028e-06, |
| "loss": 0.3456, |
| "step": 1434 |
| }, |
| { |
| "epoch": 2.3, |
| "learning_rate": 2.753992680872457e-06, |
| "loss": 0.3022, |
| "step": 1435 |
| }, |
| { |
| "epoch": 2.3, |
| "learning_rate": 2.7420943308951287e-06, |
| "loss": 0.3428, |
| "step": 1436 |
| }, |
| { |
| "epoch": 2.3, |
| "learning_rate": 2.7302176541257984e-06, |
| "loss": 0.3214, |
| "step": 1437 |
| }, |
| { |
| "epoch": 2.3, |
| "learning_rate": 2.718362686030025e-06, |
| "loss": 0.3533, |
| "step": 1438 |
| }, |
| { |
| "epoch": 2.3, |
| "learning_rate": 2.7065294620085425e-06, |
| "loss": 0.3447, |
| "step": 1439 |
| }, |
| { |
| "epoch": 2.3, |
| "learning_rate": 2.694718017397151e-06, |
| "loss": 0.3527, |
| "step": 1440 |
| }, |
| { |
| "epoch": 2.31, |
| "learning_rate": 2.6829283874666236e-06, |
| "loss": 0.3156, |
| "step": 1441 |
| }, |
| { |
| "epoch": 2.31, |
| "learning_rate": 2.6711606074225783e-06, |
| "loss": 0.3412, |
| "step": 1442 |
| }, |
| { |
| "epoch": 2.31, |
| "learning_rate": 2.6594147124053983e-06, |
| "loss": 0.3273, |
| "step": 1443 |
| }, |
| { |
| "epoch": 2.31, |
| "learning_rate": 2.6476907374901062e-06, |
| "loss": 0.3315, |
| "step": 1444 |
| }, |
| { |
| "epoch": 2.31, |
| "learning_rate": 2.635988717686272e-06, |
| "loss": 0.3472, |
| "step": 1445 |
| }, |
| { |
| "epoch": 2.31, |
| "learning_rate": 2.6243086879379e-06, |
| "loss": 0.3473, |
| "step": 1446 |
| }, |
| { |
| "epoch": 2.32, |
| "learning_rate": 2.6126506831233343e-06, |
| "loss": 0.3285, |
| "step": 1447 |
| }, |
| { |
| "epoch": 2.32, |
| "learning_rate": 2.6010147380551474e-06, |
| "loss": 0.3403, |
| "step": 1448 |
| }, |
| { |
| "epoch": 2.32, |
| "learning_rate": 2.5894008874800323e-06, |
| "loss": 0.3374, |
| "step": 1449 |
| }, |
| { |
| "epoch": 2.32, |
| "learning_rate": 2.577809166078716e-06, |
| "loss": 0.3459, |
| "step": 1450 |
| }, |
| { |
| "epoch": 2.32, |
| "learning_rate": 2.5662396084658383e-06, |
| "loss": 0.3353, |
| "step": 1451 |
| }, |
| { |
| "epoch": 2.32, |
| "learning_rate": 2.5546922491898497e-06, |
| "loss": 0.3338, |
| "step": 1452 |
| }, |
| { |
| "epoch": 2.32, |
| "learning_rate": 2.543167122732918e-06, |
| "loss": 0.3319, |
| "step": 1453 |
| }, |
| { |
| "epoch": 2.33, |
| "learning_rate": 2.5316642635108247e-06, |
| "loss": 0.3385, |
| "step": 1454 |
| }, |
| { |
| "epoch": 2.33, |
| "learning_rate": 2.5201837058728506e-06, |
| "loss": 0.355, |
| "step": 1455 |
| }, |
| { |
| "epoch": 2.33, |
| "learning_rate": 2.508725484101684e-06, |
| "loss": 0.3194, |
| "step": 1456 |
| }, |
| { |
| "epoch": 2.33, |
| "learning_rate": 2.4972896324133143e-06, |
| "loss": 0.3259, |
| "step": 1457 |
| }, |
| { |
| "epoch": 2.33, |
| "learning_rate": 2.485876184956928e-06, |
| "loss": 0.3235, |
| "step": 1458 |
| }, |
| { |
| "epoch": 2.33, |
| "learning_rate": 2.474485175814816e-06, |
| "loss": 0.3444, |
| "step": 1459 |
| }, |
| { |
| "epoch": 2.34, |
| "learning_rate": 2.4631166390022574e-06, |
| "loss": 0.3124, |
| "step": 1460 |
| }, |
| { |
| "epoch": 2.34, |
| "learning_rate": 2.451770608467432e-06, |
| "loss": 0.356, |
| "step": 1461 |
| }, |
| { |
| "epoch": 2.34, |
| "learning_rate": 2.440447118091306e-06, |
| "loss": 0.3521, |
| "step": 1462 |
| }, |
| { |
| "epoch": 2.34, |
| "learning_rate": 2.429146201687538e-06, |
| "loss": 0.3366, |
| "step": 1463 |
| }, |
| { |
| "epoch": 2.34, |
| "learning_rate": 2.417867893002387e-06, |
| "loss": 0.345, |
| "step": 1464 |
| }, |
| { |
| "epoch": 2.34, |
| "learning_rate": 2.4066122257145898e-06, |
| "loss": 0.344, |
| "step": 1465 |
| }, |
| { |
| "epoch": 2.35, |
| "learning_rate": 2.3953792334352787e-06, |
| "loss": 0.3312, |
| "step": 1466 |
| }, |
| { |
| "epoch": 2.35, |
| "learning_rate": 2.3841689497078746e-06, |
| "loss": 0.3263, |
| "step": 1467 |
| }, |
| { |
| "epoch": 2.35, |
| "learning_rate": 2.3729814080079815e-06, |
| "loss": 0.3322, |
| "step": 1468 |
| }, |
| { |
| "epoch": 2.35, |
| "learning_rate": 2.361816641743303e-06, |
| "loss": 0.3275, |
| "step": 1469 |
| }, |
| { |
| "epoch": 2.35, |
| "learning_rate": 2.3506746842535244e-06, |
| "loss": 0.3624, |
| "step": 1470 |
| }, |
| { |
| "epoch": 2.35, |
| "learning_rate": 2.339555568810221e-06, |
| "loss": 0.3337, |
| "step": 1471 |
| }, |
| { |
| "epoch": 2.36, |
| "learning_rate": 2.328459328616759e-06, |
| "loss": 0.3617, |
| "step": 1472 |
| }, |
| { |
| "epoch": 2.36, |
| "learning_rate": 2.317385996808195e-06, |
| "loss": 0.3349, |
| "step": 1473 |
| }, |
| { |
| "epoch": 2.36, |
| "learning_rate": 2.306335606451181e-06, |
| "loss": 0.3142, |
| "step": 1474 |
| }, |
| { |
| "epoch": 2.36, |
| "learning_rate": 2.295308190543859e-06, |
| "loss": 0.331, |
| "step": 1475 |
| }, |
| { |
| "epoch": 2.36, |
| "learning_rate": 2.2843037820157678e-06, |
| "loss": 0.3355, |
| "step": 1476 |
| }, |
| { |
| "epoch": 2.36, |
| "learning_rate": 2.2733224137277366e-06, |
| "loss": 0.3422, |
| "step": 1477 |
| }, |
| { |
| "epoch": 2.36, |
| "learning_rate": 2.2623641184718048e-06, |
| "loss": 0.3366, |
| "step": 1478 |
| }, |
| { |
| "epoch": 2.37, |
| "learning_rate": 2.251428928971102e-06, |
| "loss": 0.3431, |
| "step": 1479 |
| }, |
| { |
| "epoch": 2.37, |
| "learning_rate": 2.240516877879765e-06, |
| "loss": 0.3151, |
| "step": 1480 |
| }, |
| { |
| "epoch": 2.37, |
| "learning_rate": 2.229627997782834e-06, |
| "loss": 0.333, |
| "step": 1481 |
| }, |
| { |
| "epoch": 2.37, |
| "learning_rate": 2.218762321196156e-06, |
| "loss": 0.3419, |
| "step": 1482 |
| }, |
| { |
| "epoch": 2.37, |
| "learning_rate": 2.2079198805662917e-06, |
| "loss": 0.3062, |
| "step": 1483 |
| }, |
| { |
| "epoch": 2.37, |
| "learning_rate": 2.1971007082704167e-06, |
| "loss": 0.3406, |
| "step": 1484 |
| }, |
| { |
| "epoch": 2.38, |
| "learning_rate": 2.186304836616221e-06, |
| "loss": 0.3432, |
| "step": 1485 |
| }, |
| { |
| "epoch": 2.38, |
| "learning_rate": 2.1755322978418134e-06, |
| "loss": 0.3472, |
| "step": 1486 |
| }, |
| { |
| "epoch": 2.38, |
| "learning_rate": 2.1647831241156304e-06, |
| "loss": 0.3362, |
| "step": 1487 |
| }, |
| { |
| "epoch": 2.38, |
| "learning_rate": 2.1540573475363402e-06, |
| "loss": 0.3224, |
| "step": 1488 |
| }, |
| { |
| "epoch": 2.38, |
| "learning_rate": 2.1433550001327376e-06, |
| "loss": 0.3294, |
| "step": 1489 |
| }, |
| { |
| "epoch": 2.38, |
| "learning_rate": 2.1326761138636555e-06, |
| "loss": 0.3253, |
| "step": 1490 |
| }, |
| { |
| "epoch": 2.39, |
| "learning_rate": 2.122020720617869e-06, |
| "loss": 0.3398, |
| "step": 1491 |
| }, |
| { |
| "epoch": 2.39, |
| "learning_rate": 2.111388852214001e-06, |
| "loss": 0.3276, |
| "step": 1492 |
| }, |
| { |
| "epoch": 2.39, |
| "learning_rate": 2.1007805404004247e-06, |
| "loss": 0.3405, |
| "step": 1493 |
| }, |
| { |
| "epoch": 2.39, |
| "learning_rate": 2.090195816855164e-06, |
| "loss": 0.3497, |
| "step": 1494 |
| }, |
| { |
| "epoch": 2.39, |
| "learning_rate": 2.0796347131858187e-06, |
| "loss": 0.3339, |
| "step": 1495 |
| }, |
| { |
| "epoch": 2.39, |
| "learning_rate": 2.069097260929439e-06, |
| "loss": 0.3392, |
| "step": 1496 |
| }, |
| { |
| "epoch": 2.4, |
| "learning_rate": 2.058583491552465e-06, |
| "loss": 0.3429, |
| "step": 1497 |
| }, |
| { |
| "epoch": 2.4, |
| "learning_rate": 2.048093436450603e-06, |
| "loss": 0.323, |
| "step": 1498 |
| }, |
| { |
| "epoch": 2.4, |
| "learning_rate": 2.037627126948751e-06, |
| "loss": 0.318, |
| "step": 1499 |
| }, |
| { |
| "epoch": 2.4, |
| "learning_rate": 2.0271845943008984e-06, |
| "loss": 0.3261, |
| "step": 1500 |
| }, |
| { |
| "epoch": 2.4, |
| "learning_rate": 2.0167658696900317e-06, |
| "loss": 0.3259, |
| "step": 1501 |
| }, |
| { |
| "epoch": 2.4, |
| "learning_rate": 2.006370984228043e-06, |
| "loss": 0.3392, |
| "step": 1502 |
| }, |
| { |
| "epoch": 2.4, |
| "learning_rate": 1.9959999689556407e-06, |
| "loss": 0.3363, |
| "step": 1503 |
| }, |
| { |
| "epoch": 2.41, |
| "learning_rate": 1.985652854842247e-06, |
| "loss": 0.3418, |
| "step": 1504 |
| }, |
| { |
| "epoch": 2.41, |
| "learning_rate": 1.9753296727859195e-06, |
| "loss": 0.3227, |
| "step": 1505 |
| }, |
| { |
| "epoch": 2.41, |
| "learning_rate": 1.9650304536132426e-06, |
| "loss": 0.342, |
| "step": 1506 |
| }, |
| { |
| "epoch": 2.41, |
| "learning_rate": 1.9547552280792528e-06, |
| "loss": 0.3445, |
| "step": 1507 |
| }, |
| { |
| "epoch": 2.41, |
| "learning_rate": 1.9445040268673297e-06, |
| "loss": 0.3428, |
| "step": 1508 |
| }, |
| { |
| "epoch": 2.41, |
| "learning_rate": 1.9342768805891176e-06, |
| "loss": 0.3352, |
| "step": 1509 |
| }, |
| { |
| "epoch": 2.42, |
| "learning_rate": 1.924073819784428e-06, |
| "loss": 0.3474, |
| "step": 1510 |
| }, |
| { |
| "epoch": 2.42, |
| "learning_rate": 1.9138948749211473e-06, |
| "loss": 0.3349, |
| "step": 1511 |
| }, |
| { |
| "epoch": 2.42, |
| "learning_rate": 1.9037400763951508e-06, |
| "loss": 0.3078, |
| "step": 1512 |
| }, |
| { |
| "epoch": 2.42, |
| "learning_rate": 1.8936094545302098e-06, |
| "loss": 0.3253, |
| "step": 1513 |
| }, |
| { |
| "epoch": 2.42, |
| "learning_rate": 1.8835030395778941e-06, |
| "loss": 0.2942, |
| "step": 1514 |
| }, |
| { |
| "epoch": 2.42, |
| "learning_rate": 1.8734208617174986e-06, |
| "loss": 0.3213, |
| "step": 1515 |
| }, |
| { |
| "epoch": 2.43, |
| "learning_rate": 1.8633629510559315e-06, |
| "loss": 0.3154, |
| "step": 1516 |
| }, |
| { |
| "epoch": 2.43, |
| "learning_rate": 1.8533293376276473e-06, |
| "loss": 0.344, |
| "step": 1517 |
| }, |
| { |
| "epoch": 2.43, |
| "learning_rate": 1.8433200513945338e-06, |
| "loss": 0.337, |
| "step": 1518 |
| }, |
| { |
| "epoch": 2.43, |
| "learning_rate": 1.8333351222458407e-06, |
| "loss": 0.3246, |
| "step": 1519 |
| }, |
| { |
| "epoch": 2.43, |
| "learning_rate": 1.8233745799980818e-06, |
| "loss": 0.329, |
| "step": 1520 |
| }, |
| { |
| "epoch": 2.43, |
| "learning_rate": 1.813438454394948e-06, |
| "loss": 0.3434, |
| "step": 1521 |
| }, |
| { |
| "epoch": 2.44, |
| "learning_rate": 1.8035267751072172e-06, |
| "loss": 0.3377, |
| "step": 1522 |
| }, |
| { |
| "epoch": 2.44, |
| "learning_rate": 1.7936395717326705e-06, |
| "loss": 0.3467, |
| "step": 1523 |
| }, |
| { |
| "epoch": 2.44, |
| "learning_rate": 1.7837768737959937e-06, |
| "loss": 0.3507, |
| "step": 1524 |
| }, |
| { |
| "epoch": 2.44, |
| "learning_rate": 1.773938710748706e-06, |
| "loss": 0.3353, |
| "step": 1525 |
| }, |
| { |
| "epoch": 2.44, |
| "learning_rate": 1.7641251119690505e-06, |
| "loss": 0.3136, |
| "step": 1526 |
| }, |
| { |
| "epoch": 2.44, |
| "learning_rate": 1.7543361067619269e-06, |
| "loss": 0.3346, |
| "step": 1527 |
| }, |
| { |
| "epoch": 2.44, |
| "learning_rate": 1.7445717243587889e-06, |
| "loss": 0.3293, |
| "step": 1528 |
| }, |
| { |
| "epoch": 2.45, |
| "learning_rate": 1.734831993917564e-06, |
| "loss": 0.3286, |
| "step": 1529 |
| }, |
| { |
| "epoch": 2.45, |
| "learning_rate": 1.7251169445225658e-06, |
| "loss": 0.3214, |
| "step": 1530 |
| }, |
| { |
| "epoch": 2.45, |
| "learning_rate": 1.715426605184407e-06, |
| "loss": 0.3257, |
| "step": 1531 |
| }, |
| { |
| "epoch": 2.45, |
| "learning_rate": 1.705761004839911e-06, |
| "loss": 0.3357, |
| "step": 1532 |
| }, |
| { |
| "epoch": 2.45, |
| "learning_rate": 1.6961201723520248e-06, |
| "loss": 0.3118, |
| "step": 1533 |
| }, |
| { |
| "epoch": 2.45, |
| "learning_rate": 1.6865041365097434e-06, |
| "loss": 0.3253, |
| "step": 1534 |
| }, |
| { |
| "epoch": 2.46, |
| "learning_rate": 1.676912926028007e-06, |
| "loss": 0.3197, |
| "step": 1535 |
| }, |
| { |
| "epoch": 2.46, |
| "learning_rate": 1.6673465695476233e-06, |
| "loss": 0.3178, |
| "step": 1536 |
| }, |
| { |
| "epoch": 2.46, |
| "learning_rate": 1.6578050956351887e-06, |
| "loss": 0.3238, |
| "step": 1537 |
| }, |
| { |
| "epoch": 2.46, |
| "learning_rate": 1.6482885327829912e-06, |
| "loss": 0.3268, |
| "step": 1538 |
| }, |
| { |
| "epoch": 2.46, |
| "learning_rate": 1.6387969094089318e-06, |
| "loss": 0.3201, |
| "step": 1539 |
| }, |
| { |
| "epoch": 2.46, |
| "learning_rate": 1.6293302538564381e-06, |
| "loss": 0.3434, |
| "step": 1540 |
| }, |
| { |
| "epoch": 2.47, |
| "learning_rate": 1.619888594394382e-06, |
| "loss": 0.3392, |
| "step": 1541 |
| }, |
| { |
| "epoch": 2.47, |
| "learning_rate": 1.6104719592169905e-06, |
| "loss": 0.3445, |
| "step": 1542 |
| }, |
| { |
| "epoch": 2.47, |
| "learning_rate": 1.6010803764437633e-06, |
| "loss": 0.3271, |
| "step": 1543 |
| }, |
| { |
| "epoch": 2.47, |
| "learning_rate": 1.5917138741193972e-06, |
| "loss": 0.3154, |
| "step": 1544 |
| }, |
| { |
| "epoch": 2.47, |
| "learning_rate": 1.5823724802136863e-06, |
| "loss": 0.31, |
| "step": 1545 |
| }, |
| { |
| "epoch": 2.47, |
| "learning_rate": 1.5730562226214529e-06, |
| "loss": 0.3213, |
| "step": 1546 |
| }, |
| { |
| "epoch": 2.48, |
| "learning_rate": 1.5637651291624522e-06, |
| "loss": 0.3453, |
| "step": 1547 |
| }, |
| { |
| "epoch": 2.48, |
| "learning_rate": 1.5544992275813053e-06, |
| "loss": 0.3267, |
| "step": 1548 |
| }, |
| { |
| "epoch": 2.48, |
| "learning_rate": 1.545258545547398e-06, |
| "loss": 0.3458, |
| "step": 1549 |
| }, |
| { |
| "epoch": 2.48, |
| "learning_rate": 1.536043110654809e-06, |
| "loss": 0.3379, |
| "step": 1550 |
| }, |
| { |
| "epoch": 2.48, |
| "learning_rate": 1.5268529504222262e-06, |
| "loss": 0.3427, |
| "step": 1551 |
| }, |
| { |
| "epoch": 2.48, |
| "learning_rate": 1.5176880922928615e-06, |
| "loss": 0.3187, |
| "step": 1552 |
| }, |
| { |
| "epoch": 2.48, |
| "learning_rate": 1.5085485636343755e-06, |
| "loss": 0.3376, |
| "step": 1553 |
| }, |
| { |
| "epoch": 2.49, |
| "learning_rate": 1.4994343917387854e-06, |
| "loss": 0.3069, |
| "step": 1554 |
| }, |
| { |
| "epoch": 2.49, |
| "learning_rate": 1.4903456038223941e-06, |
| "loss": 0.3193, |
| "step": 1555 |
| }, |
| { |
| "epoch": 2.49, |
| "learning_rate": 1.481282227025701e-06, |
| "loss": 0.3317, |
| "step": 1556 |
| }, |
| { |
| "epoch": 2.49, |
| "learning_rate": 1.4722442884133214e-06, |
| "loss": 0.3131, |
| "step": 1557 |
| }, |
| { |
| "epoch": 2.49, |
| "learning_rate": 1.4632318149739177e-06, |
| "loss": 0.3239, |
| "step": 1558 |
| }, |
| { |
| "epoch": 2.49, |
| "learning_rate": 1.4542448336201021e-06, |
| "loss": 0.3206, |
| "step": 1559 |
| }, |
| { |
| "epoch": 2.5, |
| "learning_rate": 1.4452833711883629e-06, |
| "loss": 0.3331, |
| "step": 1560 |
| }, |
| { |
| "epoch": 2.5, |
| "learning_rate": 1.4363474544389876e-06, |
| "loss": 0.3528, |
| "step": 1561 |
| }, |
| { |
| "epoch": 2.5, |
| "learning_rate": 1.4274371100559792e-06, |
| "loss": 0.3088, |
| "step": 1562 |
| }, |
| { |
| "epoch": 2.5, |
| "learning_rate": 1.4185523646469822e-06, |
| "loss": 0.3451, |
| "step": 1563 |
| }, |
| { |
| "epoch": 2.5, |
| "learning_rate": 1.409693244743192e-06, |
| "loss": 0.3345, |
| "step": 1564 |
| }, |
| { |
| "epoch": 2.5, |
| "learning_rate": 1.4008597767992872e-06, |
| "loss": 0.3453, |
| "step": 1565 |
| }, |
| { |
| "epoch": 2.51, |
| "learning_rate": 1.3920519871933425e-06, |
| "loss": 0.3431, |
| "step": 1566 |
| }, |
| { |
| "epoch": 2.51, |
| "learning_rate": 1.3832699022267516e-06, |
| "loss": 0.3166, |
| "step": 1567 |
| }, |
| { |
| "epoch": 2.51, |
| "learning_rate": 1.3745135481241602e-06, |
| "loss": 0.3147, |
| "step": 1568 |
| }, |
| { |
| "epoch": 2.51, |
| "learning_rate": 1.3657829510333653e-06, |
| "loss": 0.3591, |
| "step": 1569 |
| }, |
| { |
| "epoch": 2.51, |
| "learning_rate": 1.3570781370252584e-06, |
| "loss": 0.3243, |
| "step": 1570 |
| }, |
| { |
| "epoch": 2.51, |
| "learning_rate": 1.3483991320937307e-06, |
| "loss": 0.3596, |
| "step": 1571 |
| }, |
| { |
| "epoch": 2.52, |
| "learning_rate": 1.339745962155613e-06, |
| "loss": 0.3299, |
| "step": 1572 |
| }, |
| { |
| "epoch": 2.52, |
| "learning_rate": 1.3311186530505838e-06, |
| "loss": 0.3526, |
| "step": 1573 |
| }, |
| { |
| "epoch": 2.52, |
| "learning_rate": 1.322517230541096e-06, |
| "loss": 0.3458, |
| "step": 1574 |
| }, |
| { |
| "epoch": 2.52, |
| "learning_rate": 1.313941720312303e-06, |
| "loss": 0.3446, |
| "step": 1575 |
| }, |
| { |
| "epoch": 2.52, |
| "learning_rate": 1.30539214797198e-06, |
| "loss": 0.3338, |
| "step": 1576 |
| }, |
| { |
| "epoch": 2.52, |
| "learning_rate": 1.2968685390504465e-06, |
| "loss": 0.3159, |
| "step": 1577 |
| }, |
| { |
| "epoch": 2.52, |
| "learning_rate": 1.2883709190004956e-06, |
| "loss": 0.3471, |
| "step": 1578 |
| }, |
| { |
| "epoch": 2.53, |
| "learning_rate": 1.2798993131973093e-06, |
| "loss": 0.346, |
| "step": 1579 |
| }, |
| { |
| "epoch": 2.53, |
| "learning_rate": 1.2714537469383858e-06, |
| "loss": 0.342, |
| "step": 1580 |
| }, |
| { |
| "epoch": 2.53, |
| "learning_rate": 1.263034245443473e-06, |
| "loss": 0.3333, |
| "step": 1581 |
| }, |
| { |
| "epoch": 2.53, |
| "learning_rate": 1.254640833854477e-06, |
| "loss": 0.3443, |
| "step": 1582 |
| }, |
| { |
| "epoch": 2.53, |
| "learning_rate": 1.2462735372353996e-06, |
| "loss": 0.3239, |
| "step": 1583 |
| }, |
| { |
| "epoch": 2.53, |
| "learning_rate": 1.2379323805722575e-06, |
| "loss": 0.3357, |
| "step": 1584 |
| }, |
| { |
| "epoch": 2.54, |
| "learning_rate": 1.2296173887730122e-06, |
| "loss": 0.3264, |
| "step": 1585 |
| }, |
| { |
| "epoch": 2.54, |
| "learning_rate": 1.2213285866674908e-06, |
| "loss": 0.3504, |
| "step": 1586 |
| }, |
| { |
| "epoch": 2.54, |
| "learning_rate": 1.2130659990073146e-06, |
| "loss": 0.3069, |
| "step": 1587 |
| }, |
| { |
| "epoch": 2.54, |
| "learning_rate": 1.2048296504658208e-06, |
| "loss": 0.3284, |
| "step": 1588 |
| }, |
| { |
| "epoch": 2.54, |
| "learning_rate": 1.196619565638003e-06, |
| "loss": 0.3108, |
| "step": 1589 |
| }, |
| { |
| "epoch": 2.54, |
| "learning_rate": 1.1884357690404157e-06, |
| "loss": 0.3245, |
| "step": 1590 |
| }, |
| { |
| "epoch": 2.55, |
| "learning_rate": 1.1802782851111206e-06, |
| "loss": 0.3679, |
| "step": 1591 |
| }, |
| { |
| "epoch": 2.55, |
| "learning_rate": 1.1721471382096028e-06, |
| "loss": 0.3242, |
| "step": 1592 |
| }, |
| { |
| "epoch": 2.55, |
| "learning_rate": 1.1640423526166987e-06, |
| "loss": 0.3102, |
| "step": 1593 |
| }, |
| { |
| "epoch": 2.55, |
| "learning_rate": 1.1559639525345313e-06, |
| "loss": 0.3087, |
| "step": 1594 |
| }, |
| { |
| "epoch": 2.55, |
| "learning_rate": 1.1479119620864277e-06, |
| "loss": 0.3276, |
| "step": 1595 |
| }, |
| { |
| "epoch": 2.55, |
| "learning_rate": 1.1398864053168534e-06, |
| "loss": 0.3395, |
| "step": 1596 |
| }, |
| { |
| "epoch": 2.56, |
| "learning_rate": 1.1318873061913405e-06, |
| "loss": 0.3206, |
| "step": 1597 |
| }, |
| { |
| "epoch": 2.56, |
| "learning_rate": 1.123914688596409e-06, |
| "loss": 0.3445, |
| "step": 1598 |
| }, |
| { |
| "epoch": 2.56, |
| "learning_rate": 1.1159685763395113e-06, |
| "loss": 0.3271, |
| "step": 1599 |
| }, |
| { |
| "epoch": 2.56, |
| "learning_rate": 1.108048993148939e-06, |
| "loss": 0.3436, |
| "step": 1600 |
| }, |
| { |
| "epoch": 2.56, |
| "learning_rate": 1.1001559626737757e-06, |
| "loss": 0.3329, |
| "step": 1601 |
| }, |
| { |
| "epoch": 2.56, |
| "learning_rate": 1.0922895084838036e-06, |
| "loss": 0.3176, |
| "step": 1602 |
| }, |
| { |
| "epoch": 2.56, |
| "learning_rate": 1.0844496540694515e-06, |
| "loss": 0.3466, |
| "step": 1603 |
| }, |
| { |
| "epoch": 2.57, |
| "learning_rate": 1.0766364228417148e-06, |
| "loss": 0.3505, |
| "step": 1604 |
| }, |
| { |
| "epoch": 2.57, |
| "learning_rate": 1.0688498381320855e-06, |
| "loss": 0.3113, |
| "step": 1605 |
| }, |
| { |
| "epoch": 2.57, |
| "learning_rate": 1.0610899231924887e-06, |
| "loss": 0.3554, |
| "step": 1606 |
| }, |
| { |
| "epoch": 2.57, |
| "learning_rate": 1.0533567011952094e-06, |
| "loss": 0.3243, |
| "step": 1607 |
| }, |
| { |
| "epoch": 2.57, |
| "learning_rate": 1.0456501952328191e-06, |
| "loss": 0.3305, |
| "step": 1608 |
| }, |
| { |
| "epoch": 2.57, |
| "learning_rate": 1.037970428318118e-06, |
| "loss": 0.338, |
| "step": 1609 |
| }, |
| { |
| "epoch": 2.58, |
| "learning_rate": 1.0303174233840529e-06, |
| "loss": 0.339, |
| "step": 1610 |
| }, |
| { |
| "epoch": 2.58, |
| "learning_rate": 1.022691203283661e-06, |
| "loss": 0.3116, |
| "step": 1611 |
| }, |
| { |
| "epoch": 2.58, |
| "learning_rate": 1.0150917907899926e-06, |
| "loss": 0.3402, |
| "step": 1612 |
| }, |
| { |
| "epoch": 2.58, |
| "learning_rate": 1.0075192085960451e-06, |
| "loss": 0.3503, |
| "step": 1613 |
| }, |
| { |
| "epoch": 2.58, |
| "learning_rate": 9.999734793146998e-07, |
| "loss": 0.3374, |
| "step": 1614 |
| }, |
| { |
| "epoch": 2.58, |
| "learning_rate": 9.924546254786493e-07, |
| "loss": 0.3266, |
| "step": 1615 |
| }, |
| { |
| "epoch": 2.59, |
| "learning_rate": 9.849626695403326e-07, |
| "loss": 0.3289, |
| "step": 1616 |
| }, |
| { |
| "epoch": 2.59, |
| "learning_rate": 9.77497633871868e-07, |
| "loss": 0.3314, |
| "step": 1617 |
| }, |
| { |
| "epoch": 2.59, |
| "learning_rate": 9.700595407649805e-07, |
| "loss": 0.3399, |
| "step": 1618 |
| }, |
| { |
| "epoch": 2.59, |
| "learning_rate": 9.62648412430951e-07, |
| "loss": 0.3206, |
| "step": 1619 |
| }, |
| { |
| "epoch": 2.59, |
| "learning_rate": 9.5526427100053e-07, |
| "loss": 0.3319, |
| "step": 1620 |
| }, |
| { |
| "epoch": 2.59, |
| "learning_rate": 9.479071385238892e-07, |
| "loss": 0.3407, |
| "step": 1621 |
| }, |
| { |
| "epoch": 2.6, |
| "learning_rate": 9.40577036970538e-07, |
| "loss": 0.3739, |
| "step": 1622 |
| }, |
| { |
| "epoch": 2.6, |
| "learning_rate": 9.332739882292752e-07, |
| "loss": 0.3135, |
| "step": 1623 |
| }, |
| { |
| "epoch": 2.6, |
| "learning_rate": 9.259980141081115e-07, |
| "loss": 0.3197, |
| "step": 1624 |
| }, |
| { |
| "epoch": 2.6, |
| "learning_rate": 9.187491363342094e-07, |
| "loss": 0.325, |
| "step": 1625 |
| }, |
| { |
| "epoch": 2.6, |
| "learning_rate": 9.115273765538202e-07, |
| "loss": 0.3289, |
| "step": 1626 |
| }, |
| { |
| "epoch": 2.6, |
| "learning_rate": 9.043327563322113e-07, |
| "loss": 0.332, |
| "step": 1627 |
| }, |
| { |
| "epoch": 2.6, |
| "learning_rate": 8.971652971536149e-07, |
| "loss": 0.3281, |
| "step": 1628 |
| }, |
| { |
| "epoch": 2.61, |
| "learning_rate": 8.900250204211513e-07, |
| "loss": 0.3212, |
| "step": 1629 |
| }, |
| { |
| "epoch": 2.61, |
| "learning_rate": 8.829119474567672e-07, |
| "loss": 0.3448, |
| "step": 1630 |
| }, |
| { |
| "epoch": 2.61, |
| "learning_rate": 8.758260995011825e-07, |
| "loss": 0.322, |
| "step": 1631 |
| }, |
| { |
| "epoch": 2.61, |
| "learning_rate": 8.687674977138116e-07, |
| "loss": 0.3425, |
| "step": 1632 |
| }, |
| { |
| "epoch": 2.61, |
| "learning_rate": 8.617361631727139e-07, |
| "loss": 0.3321, |
| "step": 1633 |
| }, |
| { |
| "epoch": 2.61, |
| "learning_rate": 8.547321168745192e-07, |
| "loss": 0.3408, |
| "step": 1634 |
| }, |
| { |
| "epoch": 2.62, |
| "learning_rate": 8.477553797343729e-07, |
| "loss": 0.3122, |
| "step": 1635 |
| }, |
| { |
| "epoch": 2.62, |
| "learning_rate": 8.40805972585872e-07, |
| "loss": 0.331, |
| "step": 1636 |
| }, |
| { |
| "epoch": 2.62, |
| "learning_rate": 8.338839161809997e-07, |
| "loss": 0.3057, |
| "step": 1637 |
| }, |
| { |
| "epoch": 2.62, |
| "learning_rate": 8.269892311900696e-07, |
| "loss": 0.3205, |
| "step": 1638 |
| }, |
| { |
| "epoch": 2.62, |
| "learning_rate": 8.201219382016556e-07, |
| "loss": 0.3448, |
| "step": 1639 |
| }, |
| { |
| "epoch": 2.62, |
| "learning_rate": 8.132820577225386e-07, |
| "loss": 0.3316, |
| "step": 1640 |
| }, |
| { |
| "epoch": 2.63, |
| "learning_rate": 8.06469610177636e-07, |
| "loss": 0.3055, |
| "step": 1641 |
| }, |
| { |
| "epoch": 2.63, |
| "learning_rate": 7.996846159099558e-07, |
| "loss": 0.3145, |
| "step": 1642 |
| }, |
| { |
| "epoch": 2.63, |
| "learning_rate": 7.92927095180518e-07, |
| "loss": 0.3273, |
| "step": 1643 |
| }, |
| { |
| "epoch": 2.63, |
| "learning_rate": 7.861970681683051e-07, |
| "loss": 0.3164, |
| "step": 1644 |
| }, |
| { |
| "epoch": 2.63, |
| "learning_rate": 7.794945549701993e-07, |
| "loss": 0.3531, |
| "step": 1645 |
| }, |
| { |
| "epoch": 2.63, |
| "learning_rate": 7.728195756009204e-07, |
| "loss": 0.3173, |
| "step": 1646 |
| }, |
| { |
| "epoch": 2.64, |
| "learning_rate": 7.661721499929753e-07, |
| "loss": 0.327, |
| "step": 1647 |
| }, |
| { |
| "epoch": 2.64, |
| "learning_rate": 7.595522979965819e-07, |
| "loss": 0.338, |
| "step": 1648 |
| }, |
| { |
| "epoch": 2.64, |
| "learning_rate": 7.529600393796232e-07, |
| "loss": 0.3183, |
| "step": 1649 |
| }, |
| { |
| "epoch": 2.64, |
| "learning_rate": 7.463953938275859e-07, |
| "loss": 0.3222, |
| "step": 1650 |
| }, |
| { |
| "epoch": 2.64, |
| "learning_rate": 7.398583809434944e-07, |
| "loss": 0.3233, |
| "step": 1651 |
| }, |
| { |
| "epoch": 2.64, |
| "learning_rate": 7.333490202478666e-07, |
| "loss": 0.31, |
| "step": 1652 |
| }, |
| { |
| "epoch": 2.64, |
| "learning_rate": 7.268673311786378e-07, |
| "loss": 0.3295, |
| "step": 1653 |
| }, |
| { |
| "epoch": 2.65, |
| "learning_rate": 7.204133330911179e-07, |
| "loss": 0.3356, |
| "step": 1654 |
| }, |
| { |
| "epoch": 2.65, |
| "learning_rate": 7.1398704525792e-07, |
| "loss": 0.3246, |
| "step": 1655 |
| }, |
| { |
| "epoch": 2.65, |
| "learning_rate": 7.07588486868922e-07, |
| "loss": 0.3248, |
| "step": 1656 |
| }, |
| { |
| "epoch": 2.65, |
| "learning_rate": 7.012176770311863e-07, |
| "loss": 0.3253, |
| "step": 1657 |
| }, |
| { |
| "epoch": 2.65, |
| "learning_rate": 6.948746347689184e-07, |
| "loss": 0.3224, |
| "step": 1658 |
| }, |
| { |
| "epoch": 2.65, |
| "learning_rate": 6.885593790234057e-07, |
| "loss": 0.3147, |
| "step": 1659 |
| }, |
| { |
| "epoch": 2.66, |
| "learning_rate": 6.8227192865296e-07, |
| "loss": 0.3228, |
| "step": 1660 |
| }, |
| { |
| "epoch": 2.66, |
| "learning_rate": 6.760123024328624e-07, |
| "loss": 0.3368, |
| "step": 1661 |
| }, |
| { |
| "epoch": 2.66, |
| "learning_rate": 6.697805190553086e-07, |
| "loss": 0.3258, |
| "step": 1662 |
| }, |
| { |
| "epoch": 2.66, |
| "learning_rate": 6.635765971293484e-07, |
| "loss": 0.346, |
| "step": 1663 |
| }, |
| { |
| "epoch": 2.66, |
| "learning_rate": 6.574005551808338e-07, |
| "loss": 0.3448, |
| "step": 1664 |
| }, |
| { |
| "epoch": 2.66, |
| "learning_rate": 6.512524116523633e-07, |
| "loss": 0.3009, |
| "step": 1665 |
| }, |
| { |
| "epoch": 2.67, |
| "learning_rate": 6.451321849032289e-07, |
| "loss": 0.3015, |
| "step": 1666 |
| }, |
| { |
| "epoch": 2.67, |
| "learning_rate": 6.390398932093555e-07, |
| "loss": 0.35, |
| "step": 1667 |
| }, |
| { |
| "epoch": 2.67, |
| "learning_rate": 6.329755547632499e-07, |
| "loss": 0.3271, |
| "step": 1668 |
| }, |
| { |
| "epoch": 2.67, |
| "learning_rate": 6.269391876739494e-07, |
| "loss": 0.3212, |
| "step": 1669 |
| }, |
| { |
| "epoch": 2.67, |
| "learning_rate": 6.209308099669598e-07, |
| "loss": 0.3549, |
| "step": 1670 |
| }, |
| { |
| "epoch": 2.67, |
| "learning_rate": 6.149504395842087e-07, |
| "loss": 0.3544, |
| "step": 1671 |
| }, |
| { |
| "epoch": 2.68, |
| "learning_rate": 6.089980943839924e-07, |
| "loss": 0.3378, |
| "step": 1672 |
| }, |
| { |
| "epoch": 2.68, |
| "learning_rate": 6.030737921409169e-07, |
| "loss": 0.3266, |
| "step": 1673 |
| }, |
| { |
| "epoch": 2.68, |
| "learning_rate": 5.971775505458444e-07, |
| "loss": 0.3247, |
| "step": 1674 |
| }, |
| { |
| "epoch": 2.68, |
| "learning_rate": 5.913093872058528e-07, |
| "loss": 0.334, |
| "step": 1675 |
| }, |
| { |
| "epoch": 2.68, |
| "learning_rate": 5.854693196441641e-07, |
| "loss": 0.3101, |
| "step": 1676 |
| }, |
| { |
| "epoch": 2.68, |
| "learning_rate": 5.796573653001091e-07, |
| "loss": 0.3363, |
| "step": 1677 |
| }, |
| { |
| "epoch": 2.68, |
| "learning_rate": 5.738735415290642e-07, |
| "loss": 0.3268, |
| "step": 1678 |
| }, |
| { |
| "epoch": 2.69, |
| "learning_rate": 5.681178656024055e-07, |
| "loss": 0.3453, |
| "step": 1679 |
| }, |
| { |
| "epoch": 2.69, |
| "learning_rate": 5.62390354707455e-07, |
| "loss": 0.3344, |
| "step": 1680 |
| }, |
| { |
| "epoch": 2.69, |
| "learning_rate": 5.56691025947429e-07, |
| "loss": 0.3373, |
| "step": 1681 |
| }, |
| { |
| "epoch": 2.69, |
| "learning_rate": 5.510198963413882e-07, |
| "loss": 0.3209, |
| "step": 1682 |
| }, |
| { |
| "epoch": 2.69, |
| "learning_rate": 5.453769828241872e-07, |
| "loss": 0.3085, |
| "step": 1683 |
| }, |
| { |
| "epoch": 2.69, |
| "learning_rate": 5.397623022464227e-07, |
| "loss": 0.3433, |
| "step": 1684 |
| }, |
| { |
| "epoch": 2.7, |
| "learning_rate": 5.341758713743828e-07, |
| "loss": 0.2998, |
| "step": 1685 |
| }, |
| { |
| "epoch": 2.7, |
| "learning_rate": 5.286177068899989e-07, |
| "loss": 0.3296, |
| "step": 1686 |
| }, |
| { |
| "epoch": 2.7, |
| "learning_rate": 5.230878253907911e-07, |
| "loss": 0.3326, |
| "step": 1687 |
| }, |
| { |
| "epoch": 2.7, |
| "learning_rate": 5.175862433898282e-07, |
| "loss": 0.3605, |
| "step": 1688 |
| }, |
| { |
| "epoch": 2.7, |
| "learning_rate": 5.121129773156663e-07, |
| "loss": 0.3498, |
| "step": 1689 |
| }, |
| { |
| "epoch": 2.7, |
| "learning_rate": 5.066680435123106e-07, |
| "loss": 0.3224, |
| "step": 1690 |
| }, |
| { |
| "epoch": 2.71, |
| "learning_rate": 5.012514582391592e-07, |
| "loss": 0.3338, |
| "step": 1691 |
| }, |
| { |
| "epoch": 2.71, |
| "learning_rate": 4.95863237670956e-07, |
| "loss": 0.2988, |
| "step": 1692 |
| }, |
| { |
| "epoch": 2.71, |
| "learning_rate": 4.905033978977492e-07, |
| "loss": 0.326, |
| "step": 1693 |
| }, |
| { |
| "epoch": 2.71, |
| "learning_rate": 4.851719549248301e-07, |
| "loss": 0.3182, |
| "step": 1694 |
| }, |
| { |
| "epoch": 2.71, |
| "learning_rate": 4.798689246727006e-07, |
| "loss": 0.334, |
| "step": 1695 |
| }, |
| { |
| "epoch": 2.71, |
| "learning_rate": 4.7459432297701224e-07, |
| "loss": 0.3202, |
| "step": 1696 |
| }, |
| { |
| "epoch": 2.72, |
| "learning_rate": 4.693481655885257e-07, |
| "loss": 0.3216, |
| "step": 1697 |
| }, |
| { |
| "epoch": 2.72, |
| "learning_rate": 4.6413046817306404e-07, |
| "loss": 0.3273, |
| "step": 1698 |
| }, |
| { |
| "epoch": 2.72, |
| "learning_rate": 4.58941246311464e-07, |
| "loss": 0.3579, |
| "step": 1699 |
| }, |
| { |
| "epoch": 2.72, |
| "learning_rate": 4.5378051549952783e-07, |
| "loss": 0.3118, |
| "step": 1700 |
| }, |
| { |
| "epoch": 2.72, |
| "learning_rate": 4.4864829114798394e-07, |
| "loss": 0.3562, |
| "step": 1701 |
| }, |
| { |
| "epoch": 2.72, |
| "learning_rate": 4.4354458858242857e-07, |
| "loss": 0.3396, |
| "step": 1702 |
| }, |
| { |
| "epoch": 2.72, |
| "learning_rate": 4.384694230432984e-07, |
| "loss": 0.3257, |
| "step": 1703 |
| }, |
| { |
| "epoch": 2.73, |
| "learning_rate": 4.3342280968580287e-07, |
| "loss": 0.3479, |
| "step": 1704 |
| }, |
| { |
| "epoch": 2.73, |
| "learning_rate": 4.2840476357989825e-07, |
| "loss": 0.3165, |
| "step": 1705 |
| }, |
| { |
| "epoch": 2.73, |
| "learning_rate": 4.2341529971023253e-07, |
| "loss": 0.3319, |
| "step": 1706 |
| }, |
| { |
| "epoch": 2.73, |
| "learning_rate": 4.184544329761009e-07, |
| "loss": 0.3213, |
| "step": 1707 |
| }, |
| { |
| "epoch": 2.73, |
| "learning_rate": 4.1352217819140337e-07, |
| "loss": 0.3226, |
| "step": 1708 |
| }, |
| { |
| "epoch": 2.73, |
| "learning_rate": 4.0861855008460403e-07, |
| "loss": 0.3259, |
| "step": 1709 |
| }, |
| { |
| "epoch": 2.74, |
| "learning_rate": 4.037435632986786e-07, |
| "loss": 0.3339, |
| "step": 1710 |
| }, |
| { |
| "epoch": 2.74, |
| "learning_rate": 3.988972323910778e-07, |
| "loss": 0.3475, |
| "step": 1711 |
| }, |
| { |
| "epoch": 2.74, |
| "learning_rate": 3.9407957183368093e-07, |
| "loss": 0.3162, |
| "step": 1712 |
| }, |
| { |
| "epoch": 2.74, |
| "learning_rate": 3.8929059601275463e-07, |
| "loss": 0.3382, |
| "step": 1713 |
| }, |
| { |
| "epoch": 2.74, |
| "learning_rate": 3.845303192289074e-07, |
| "loss": 0.3162, |
| "step": 1714 |
| }, |
| { |
| "epoch": 2.74, |
| "learning_rate": 3.797987556970495e-07, |
| "loss": 0.3469, |
| "step": 1715 |
| }, |
| { |
| "epoch": 2.75, |
| "learning_rate": 3.750959195463466e-07, |
| "loss": 0.3506, |
| "step": 1716 |
| }, |
| { |
| "epoch": 2.75, |
| "learning_rate": 3.7042182482018074e-07, |
| "loss": 0.3216, |
| "step": 1717 |
| }, |
| { |
| "epoch": 2.75, |
| "learning_rate": 3.6577648547611033e-07, |
| "loss": 0.3294, |
| "step": 1718 |
| }, |
| { |
| "epoch": 2.75, |
| "learning_rate": 3.611599153858214e-07, |
| "loss": 0.3531, |
| "step": 1719 |
| }, |
| { |
| "epoch": 2.75, |
| "learning_rate": 3.5657212833509313e-07, |
| "loss": 0.3156, |
| "step": 1720 |
| }, |
| { |
| "epoch": 2.75, |
| "learning_rate": 3.520131380237546e-07, |
| "loss": 0.3474, |
| "step": 1721 |
| }, |
| { |
| "epoch": 2.76, |
| "learning_rate": 3.474829580656436e-07, |
| "loss": 0.3292, |
| "step": 1722 |
| }, |
| { |
| "epoch": 2.76, |
| "learning_rate": 3.429816019885657e-07, |
| "loss": 0.3366, |
| "step": 1723 |
| }, |
| { |
| "epoch": 2.76, |
| "learning_rate": 3.385090832342497e-07, |
| "loss": 0.3264, |
| "step": 1724 |
| }, |
| { |
| "epoch": 2.76, |
| "learning_rate": 3.3406541515832e-07, |
| "loss": 0.3175, |
| "step": 1725 |
| }, |
| { |
| "epoch": 2.76, |
| "learning_rate": 3.296506110302422e-07, |
| "loss": 0.3161, |
| "step": 1726 |
| }, |
| { |
| "epoch": 2.76, |
| "learning_rate": 3.252646840332918e-07, |
| "loss": 0.3449, |
| "step": 1727 |
| }, |
| { |
| "epoch": 2.76, |
| "learning_rate": 3.209076472645112e-07, |
| "loss": 0.3245, |
| "step": 1728 |
| }, |
| { |
| "epoch": 2.77, |
| "learning_rate": 3.16579513734675e-07, |
| "loss": 0.326, |
| "step": 1729 |
| }, |
| { |
| "epoch": 2.77, |
| "learning_rate": 3.1228029636824477e-07, |
| "loss": 0.3248, |
| "step": 1730 |
| }, |
| { |
| "epoch": 2.77, |
| "learning_rate": 3.080100080033388e-07, |
| "loss": 0.3171, |
| "step": 1731 |
| }, |
| { |
| "epoch": 2.77, |
| "learning_rate": 3.037686613916857e-07, |
| "loss": 0.3514, |
| "step": 1732 |
| }, |
| { |
| "epoch": 2.77, |
| "learning_rate": 2.995562691985898e-07, |
| "loss": 0.3189, |
| "step": 1733 |
| }, |
| { |
| "epoch": 2.77, |
| "learning_rate": 2.9537284400289354e-07, |
| "loss": 0.3255, |
| "step": 1734 |
| }, |
| { |
| "epoch": 2.78, |
| "learning_rate": 2.9121839829693857e-07, |
| "loss": 0.3294, |
| "step": 1735 |
| }, |
| { |
| "epoch": 2.78, |
| "learning_rate": 2.8709294448653223e-07, |
| "loss": 0.3198, |
| "step": 1736 |
| }, |
| { |
| "epoch": 2.78, |
| "learning_rate": 2.829964948909048e-07, |
| "loss": 0.3585, |
| "step": 1737 |
| }, |
| { |
| "epoch": 2.78, |
| "learning_rate": 2.7892906174267653e-07, |
| "loss": 0.3301, |
| "step": 1738 |
| }, |
| { |
| "epoch": 2.78, |
| "learning_rate": 2.748906571878207e-07, |
| "loss": 0.3364, |
| "step": 1739 |
| }, |
| { |
| "epoch": 2.78, |
| "learning_rate": 2.708812932856253e-07, |
| "loss": 0.3227, |
| "step": 1740 |
| }, |
| { |
| "epoch": 2.79, |
| "learning_rate": 2.6690098200866097e-07, |
| "loss": 0.3425, |
| "step": 1741 |
| }, |
| { |
| "epoch": 2.79, |
| "learning_rate": 2.6294973524274127e-07, |
| "loss": 0.3325, |
| "step": 1742 |
| }, |
| { |
| "epoch": 2.79, |
| "learning_rate": 2.5902756478688674e-07, |
| "loss": 0.325, |
| "step": 1743 |
| }, |
| { |
| "epoch": 2.79, |
| "learning_rate": 2.551344823532964e-07, |
| "loss": 0.322, |
| "step": 1744 |
| }, |
| { |
| "epoch": 2.79, |
| "learning_rate": 2.5127049956730207e-07, |
| "loss": 0.3332, |
| "step": 1745 |
| }, |
| { |
| "epoch": 2.79, |
| "learning_rate": 2.474356279673462e-07, |
| "loss": 0.3324, |
| "step": 1746 |
| }, |
| { |
| "epoch": 2.8, |
| "learning_rate": 2.436298790049363e-07, |
| "loss": 0.3193, |
| "step": 1747 |
| }, |
| { |
| "epoch": 2.8, |
| "learning_rate": 2.398532640446161e-07, |
| "loss": 0.3277, |
| "step": 1748 |
| }, |
| { |
| "epoch": 2.8, |
| "learning_rate": 2.3610579436392999e-07, |
| "loss": 0.3404, |
| "step": 1749 |
| }, |
| { |
| "epoch": 2.8, |
| "learning_rate": 2.3238748115339327e-07, |
| "loss": 0.3335, |
| "step": 1750 |
| }, |
| { |
| "epoch": 2.8, |
| "learning_rate": 2.2869833551645293e-07, |
| "loss": 0.3273, |
| "step": 1751 |
| }, |
| { |
| "epoch": 2.8, |
| "learning_rate": 2.2503836846945792e-07, |
| "loss": 0.3303, |
| "step": 1752 |
| }, |
| { |
| "epoch": 2.8, |
| "learning_rate": 2.2140759094162468e-07, |
| "loss": 0.307, |
| "step": 1753 |
| }, |
| { |
| "epoch": 2.81, |
| "learning_rate": 2.178060137750071e-07, |
| "loss": 0.3182, |
| "step": 1754 |
| }, |
| { |
| "epoch": 2.81, |
| "learning_rate": 2.1423364772445886e-07, |
| "loss": 0.3256, |
| "step": 1755 |
| }, |
| { |
| "epoch": 2.81, |
| "learning_rate": 2.106905034576112e-07, |
| "loss": 0.3482, |
| "step": 1756 |
| }, |
| { |
| "epoch": 2.81, |
| "learning_rate": 2.071765915548274e-07, |
| "loss": 0.3164, |
| "step": 1757 |
| }, |
| { |
| "epoch": 2.81, |
| "learning_rate": 2.036919225091827e-07, |
| "loss": 0.3077, |
| "step": 1758 |
| }, |
| { |
| "epoch": 2.81, |
| "learning_rate": 2.002365067264289e-07, |
| "loss": 0.3266, |
| "step": 1759 |
| }, |
| { |
| "epoch": 2.82, |
| "learning_rate": 1.9681035452496112e-07, |
| "loss": 0.3257, |
| "step": 1760 |
| }, |
| { |
| "epoch": 2.82, |
| "learning_rate": 1.9341347613579086e-07, |
| "loss": 0.3353, |
| "step": 1761 |
| }, |
| { |
| "epoch": 2.82, |
| "learning_rate": 1.900458817025097e-07, |
| "loss": 0.327, |
| "step": 1762 |
| }, |
| { |
| "epoch": 2.82, |
| "learning_rate": 1.867075812812691e-07, |
| "loss": 0.3248, |
| "step": 1763 |
| }, |
| { |
| "epoch": 2.82, |
| "learning_rate": 1.8339858484073935e-07, |
| "loss": 0.3284, |
| "step": 1764 |
| }, |
| { |
| "epoch": 2.82, |
| "learning_rate": 1.8011890226208527e-07, |
| "loss": 0.3199, |
| "step": 1765 |
| }, |
| { |
| "epoch": 2.83, |
| "learning_rate": 1.7686854333893833e-07, |
| "loss": 0.3425, |
| "step": 1766 |
| }, |
| { |
| "epoch": 2.83, |
| "learning_rate": 1.7364751777736334e-07, |
| "loss": 0.3357, |
| "step": 1767 |
| }, |
| { |
| "epoch": 2.83, |
| "learning_rate": 1.7045583519583075e-07, |
| "loss": 0.3323, |
| "step": 1768 |
| }, |
| { |
| "epoch": 2.83, |
| "learning_rate": 1.6729350512519006e-07, |
| "loss": 0.3492, |
| "step": 1769 |
| }, |
| { |
| "epoch": 2.83, |
| "learning_rate": 1.6416053700863965e-07, |
| "loss": 0.3324, |
| "step": 1770 |
| }, |
| { |
| "epoch": 2.83, |
| "learning_rate": 1.6105694020169594e-07, |
| "loss": 0.3145, |
| "step": 1771 |
| }, |
| { |
| "epoch": 2.84, |
| "learning_rate": 1.5798272397217097e-07, |
| "loss": 0.3287, |
| "step": 1772 |
| }, |
| { |
| "epoch": 2.84, |
| "learning_rate": 1.5493789750014032e-07, |
| "loss": 0.3157, |
| "step": 1773 |
| }, |
| { |
| "epoch": 2.84, |
| "learning_rate": 1.519224698779198e-07, |
| "loss": 0.3407, |
| "step": 1774 |
| }, |
| { |
| "epoch": 2.84, |
| "learning_rate": 1.489364501100332e-07, |
| "loss": 0.3245, |
| "step": 1775 |
| }, |
| { |
| "epoch": 2.84, |
| "learning_rate": 1.459798471131868e-07, |
| "loss": 0.3031, |
| "step": 1776 |
| }, |
| { |
| "epoch": 2.84, |
| "learning_rate": 1.430526697162482e-07, |
| "loss": 0.3368, |
| "step": 1777 |
| }, |
| { |
| "epoch": 2.84, |
| "learning_rate": 1.4015492666021313e-07, |
| "loss": 0.3037, |
| "step": 1778 |
| }, |
| { |
| "epoch": 2.85, |
| "learning_rate": 1.3728662659818205e-07, |
| "loss": 0.3293, |
| "step": 1779 |
| }, |
| { |
| "epoch": 2.85, |
| "learning_rate": 1.344477780953346e-07, |
| "loss": 0.3317, |
| "step": 1780 |
| }, |
| { |
| "epoch": 2.85, |
| "learning_rate": 1.3163838962890196e-07, |
| "loss": 0.3408, |
| "step": 1781 |
| }, |
| { |
| "epoch": 2.85, |
| "learning_rate": 1.2885846958814673e-07, |
| "loss": 0.3301, |
| "step": 1782 |
| }, |
| { |
| "epoch": 2.85, |
| "learning_rate": 1.2610802627432972e-07, |
| "loss": 0.3196, |
| "step": 1783 |
| }, |
| { |
| "epoch": 2.85, |
| "learning_rate": 1.2338706790069433e-07, |
| "loss": 0.3213, |
| "step": 1784 |
| }, |
| { |
| "epoch": 2.86, |
| "learning_rate": 1.206956025924333e-07, |
| "loss": 0.3526, |
| "step": 1785 |
| }, |
| { |
| "epoch": 2.86, |
| "learning_rate": 1.1803363838667092e-07, |
| "loss": 0.3262, |
| "step": 1786 |
| }, |
| { |
| "epoch": 2.86, |
| "learning_rate": 1.1540118323243866e-07, |
| "loss": 0.3079, |
| "step": 1787 |
| }, |
| { |
| "epoch": 2.86, |
| "learning_rate": 1.1279824499064396e-07, |
| "loss": 0.328, |
| "step": 1788 |
| }, |
| { |
| "epoch": 2.86, |
| "learning_rate": 1.1022483143405705e-07, |
| "loss": 0.3181, |
| "step": 1789 |
| }, |
| { |
| "epoch": 2.86, |
| "learning_rate": 1.0768095024728309e-07, |
| "loss": 0.3284, |
| "step": 1790 |
| }, |
| { |
| "epoch": 2.87, |
| "learning_rate": 1.0516660902673448e-07, |
| "loss": 0.3029, |
| "step": 1791 |
| }, |
| { |
| "epoch": 2.87, |
| "learning_rate": 1.0268181528061749e-07, |
| "loss": 0.3403, |
| "step": 1792 |
| }, |
| { |
| "epoch": 2.87, |
| "learning_rate": 1.0022657642890232e-07, |
| "loss": 0.3203, |
| "step": 1793 |
| }, |
| { |
| "epoch": 2.87, |
| "learning_rate": 9.780089980330643e-08, |
| "loss": 0.3414, |
| "step": 1794 |
| }, |
| { |
| "epoch": 2.87, |
| "learning_rate": 9.540479264726676e-08, |
| "loss": 0.3215, |
| "step": 1795 |
| }, |
| { |
| "epoch": 2.87, |
| "learning_rate": 9.303826211592316e-08, |
| "loss": 0.3396, |
| "step": 1796 |
| }, |
| { |
| "epoch": 2.88, |
| "learning_rate": 9.070131527609604e-08, |
| "loss": 0.3284, |
| "step": 1797 |
| }, |
| { |
| "epoch": 2.88, |
| "learning_rate": 8.839395910626214e-08, |
| "loss": 0.3185, |
| "step": 1798 |
| }, |
| { |
| "epoch": 2.88, |
| "learning_rate": 8.61162004965388e-08, |
| "loss": 0.3079, |
| "step": 1799 |
| }, |
| { |
| "epoch": 2.88, |
| "learning_rate": 8.386804624865851e-08, |
| "loss": 0.3274, |
| "step": 1800 |
| }, |
| { |
| "epoch": 2.88, |
| "learning_rate": 8.16495030759501e-08, |
| "loss": 0.3205, |
| "step": 1801 |
| }, |
| { |
| "epoch": 2.88, |
| "learning_rate": 7.946057760332193e-08, |
| "loss": 0.3236, |
| "step": 1802 |
| }, |
| { |
| "epoch": 2.88, |
| "learning_rate": 7.730127636723539e-08, |
| "loss": 0.321, |
| "step": 1803 |
| }, |
| { |
| "epoch": 2.89, |
| "learning_rate": 7.517160581569371e-08, |
| "loss": 0.3397, |
| "step": 1804 |
| }, |
| { |
| "epoch": 2.89, |
| "learning_rate": 7.307157230821426e-08, |
| "loss": 0.3155, |
| "step": 1805 |
| }, |
| { |
| "epoch": 2.89, |
| "learning_rate": 7.100118211581852e-08, |
| "loss": 0.3525, |
| "step": 1806 |
| }, |
| { |
| "epoch": 2.89, |
| "learning_rate": 6.896044142100433e-08, |
| "loss": 0.3094, |
| "step": 1807 |
| }, |
| { |
| "epoch": 2.89, |
| "learning_rate": 6.694935631773259e-08, |
| "loss": 0.3171, |
| "step": 1808 |
| }, |
| { |
| "epoch": 2.89, |
| "learning_rate": 6.496793281141056e-08, |
| "loss": 0.322, |
| "step": 1809 |
| }, |
| { |
| "epoch": 2.9, |
| "learning_rate": 6.301617681886863e-08, |
| "loss": 0.3158, |
| "step": 1810 |
| }, |
| { |
| "epoch": 2.9, |
| "learning_rate": 6.109409416834689e-08, |
| "loss": 0.3263, |
| "step": 1811 |
| }, |
| { |
| "epoch": 2.9, |
| "learning_rate": 5.920169059947412e-08, |
| "loss": 0.3165, |
| "step": 1812 |
| }, |
| { |
| "epoch": 2.9, |
| "learning_rate": 5.7338971763256646e-08, |
| "loss": 0.3454, |
| "step": 1813 |
| }, |
| { |
| "epoch": 2.9, |
| "learning_rate": 5.5505943222055046e-08, |
| "loss": 0.3344, |
| "step": 1814 |
| }, |
| { |
| "epoch": 2.9, |
| "learning_rate": 5.37026104495697e-08, |
| "loss": 0.311, |
| "step": 1815 |
| }, |
| { |
| "epoch": 2.91, |
| "learning_rate": 5.192897883082748e-08, |
| "loss": 0.3108, |
| "step": 1816 |
| }, |
| { |
| "epoch": 2.91, |
| "learning_rate": 5.0185053662161756e-08, |
| "loss": 0.3206, |
| "step": 1817 |
| }, |
| { |
| "epoch": 2.91, |
| "learning_rate": 4.8470840151195745e-08, |
| "loss": 0.3613, |
| "step": 1818 |
| }, |
| { |
| "epoch": 2.91, |
| "learning_rate": 4.678634341683252e-08, |
| "loss": 0.3364, |
| "step": 1819 |
| }, |
| { |
| "epoch": 2.91, |
| "learning_rate": 4.513156848923616e-08, |
| "loss": 0.3185, |
| "step": 1820 |
| }, |
| { |
| "epoch": 2.91, |
| "learning_rate": 4.350652030981395e-08, |
| "loss": 0.3149, |
| "step": 1821 |
| }, |
| { |
| "epoch": 2.92, |
| "learning_rate": 4.19112037312075e-08, |
| "loss": 0.3315, |
| "step": 1822 |
| }, |
| { |
| "epoch": 2.92, |
| "learning_rate": 4.0345623517273894e-08, |
| "loss": 0.3421, |
| "step": 1823 |
| }, |
| { |
| "epoch": 2.92, |
| "learning_rate": 3.8809784343072364e-08, |
| "loss": 0.3109, |
| "step": 1824 |
| }, |
| { |
| "epoch": 2.92, |
| "learning_rate": 3.7303690794854296e-08, |
| "loss": 0.3475, |
| "step": 1825 |
| }, |
| { |
| "epoch": 2.92, |
| "learning_rate": 3.582734737004101e-08, |
| "loss": 0.3181, |
| "step": 1826 |
| }, |
| { |
| "epoch": 2.92, |
| "learning_rate": 3.438075847721933e-08, |
| "loss": 0.3258, |
| "step": 1827 |
| }, |
| { |
| "epoch": 2.92, |
| "learning_rate": 3.2963928436122726e-08, |
| "loss": 0.3253, |
| "step": 1828 |
| }, |
| { |
| "epoch": 2.93, |
| "learning_rate": 3.157686147762129e-08, |
| "loss": 0.3158, |
| "step": 1829 |
| }, |
| { |
| "epoch": 2.93, |
| "learning_rate": 3.0219561743707326e-08, |
| "loss": 0.3133, |
| "step": 1830 |
| }, |
| { |
| "epoch": 2.93, |
| "learning_rate": 2.8892033287484245e-08, |
| "loss": 0.3035, |
| "step": 1831 |
| }, |
| { |
| "epoch": 2.93, |
| "learning_rate": 2.7594280073152123e-08, |
| "loss": 0.337, |
| "step": 1832 |
| }, |
| { |
| "epoch": 2.93, |
| "learning_rate": 2.6326305976001054e-08, |
| "loss": 0.3242, |
| "step": 1833 |
| }, |
| { |
| "epoch": 2.93, |
| "learning_rate": 2.508811478239226e-08, |
| "loss": 0.3442, |
| "step": 1834 |
| }, |
| { |
| "epoch": 2.94, |
| "learning_rate": 2.3879710189753657e-08, |
| "loss": 0.3129, |
| "step": 1835 |
| }, |
| { |
| "epoch": 2.94, |
| "learning_rate": 2.2701095806565432e-08, |
| "loss": 0.3345, |
| "step": 1836 |
| }, |
| { |
| "epoch": 2.94, |
| "learning_rate": 2.1552275152346702e-08, |
| "loss": 0.3464, |
| "step": 1837 |
| }, |
| { |
| "epoch": 2.94, |
| "learning_rate": 2.0433251657653307e-08, |
| "loss": 0.323, |
| "step": 1838 |
| }, |
| { |
| "epoch": 2.94, |
| "learning_rate": 1.9344028664056715e-08, |
| "loss": 0.351, |
| "step": 1839 |
| }, |
| { |
| "epoch": 2.94, |
| "learning_rate": 1.8284609424142897e-08, |
| "loss": 0.3225, |
| "step": 1840 |
| }, |
| { |
| "epoch": 2.95, |
| "learning_rate": 1.7254997101500137e-08, |
| "loss": 0.3295, |
| "step": 1841 |
| }, |
| { |
| "epoch": 2.95, |
| "learning_rate": 1.6255194770704586e-08, |
| "loss": 0.3221, |
| "step": 1842 |
| }, |
| { |
| "epoch": 2.95, |
| "learning_rate": 1.528520541731915e-08, |
| "loss": 0.3303, |
| "step": 1843 |
| }, |
| { |
| "epoch": 2.95, |
| "learning_rate": 1.4345031937879061e-08, |
| "loss": 0.3178, |
| "step": 1844 |
| }, |
| { |
| "epoch": 2.95, |
| "learning_rate": 1.3434677139885222e-08, |
| "loss": 0.3437, |
| "step": 1845 |
| }, |
| { |
| "epoch": 2.95, |
| "learning_rate": 1.2554143741795311e-08, |
| "loss": 0.322, |
| "step": 1846 |
| }, |
| { |
| "epoch": 2.96, |
| "learning_rate": 1.170343437301491e-08, |
| "loss": 0.317, |
| "step": 1847 |
| }, |
| { |
| "epoch": 2.96, |
| "learning_rate": 1.0882551573891953e-08, |
| "loss": 0.3074, |
| "step": 1848 |
| }, |
| { |
| "epoch": 2.96, |
| "learning_rate": 1.0091497795706728e-08, |
| "loss": 0.3327, |
| "step": 1849 |
| }, |
| { |
| "epoch": 2.96, |
| "learning_rate": 9.330275400666334e-09, |
| "loss": 0.309, |
| "step": 1850 |
| }, |
| { |
| "epoch": 2.96, |
| "learning_rate": 8.59888666189579e-09, |
| "loss": 0.3029, |
| "step": 1851 |
| }, |
| { |
| "epoch": 2.96, |
| "learning_rate": 7.897333763433601e-09, |
| "loss": 0.3427, |
| "step": 1852 |
| }, |
| { |
| "epoch": 2.96, |
| "learning_rate": 7.225618800222878e-09, |
| "loss": 0.3179, |
| "step": 1853 |
| }, |
| { |
| "epoch": 2.97, |
| "learning_rate": 6.583743778106888e-09, |
| "loss": 0.3371, |
| "step": 1854 |
| }, |
| { |
| "epoch": 2.97, |
| "learning_rate": 5.971710613821291e-09, |
| "loss": 0.3216, |
| "step": 1855 |
| }, |
| { |
| "epoch": 2.97, |
| "learning_rate": 5.3895211349896946e-09, |
| "loss": 0.3416, |
| "step": 1856 |
| }, |
| { |
| "epoch": 2.97, |
| "learning_rate": 4.837177080119215e-09, |
| "loss": 0.3226, |
| "step": 1857 |
| }, |
| { |
| "epoch": 2.97, |
| "learning_rate": 4.314680098592705e-09, |
| "loss": 0.3293, |
| "step": 1858 |
| }, |
| { |
| "epoch": 2.97, |
| "learning_rate": 3.8220317506654226e-09, |
| "loss": 0.3373, |
| "step": 1859 |
| }, |
| { |
| "epoch": 2.98, |
| "learning_rate": 3.3592335074594805e-09, |
| "loss": 0.3316, |
| "step": 1860 |
| }, |
| { |
| "epoch": 2.98, |
| "learning_rate": 2.9262867509605164e-09, |
| "loss": 0.3476, |
| "step": 1861 |
| }, |
| { |
| "epoch": 2.98, |
| "learning_rate": 2.5231927740154705e-09, |
| "loss": 0.3487, |
| "step": 1862 |
| }, |
| { |
| "epoch": 2.98, |
| "learning_rate": 2.149952780321485e-09, |
| "loss": 0.3266, |
| "step": 1863 |
| }, |
| { |
| "epoch": 2.98, |
| "learning_rate": 1.8065678844314538e-09, |
| "loss": 0.3206, |
| "step": 1864 |
| }, |
| { |
| "epoch": 2.98, |
| "learning_rate": 1.4930391117451427e-09, |
| "loss": 0.3481, |
| "step": 1865 |
| }, |
| { |
| "epoch": 2.99, |
| "learning_rate": 1.209367398504746e-09, |
| "loss": 0.3225, |
| "step": 1866 |
| }, |
| { |
| "epoch": 2.99, |
| "learning_rate": 9.555535917993297e-10, |
| "loss": 0.3197, |
| "step": 1867 |
| }, |
| { |
| "epoch": 2.99, |
| "learning_rate": 7.315984495548378e-10, |
| "loss": 0.3163, |
| "step": 1868 |
| }, |
| { |
| "epoch": 2.99, |
| "learning_rate": 5.375026405352035e-10, |
| "loss": 0.3077, |
| "step": 1869 |
| }, |
| { |
| "epoch": 2.99, |
| "learning_rate": 3.732667443390181e-10, |
| "loss": 0.318, |
| "step": 1870 |
| }, |
| { |
| "epoch": 2.99, |
| "learning_rate": 2.388912514017516e-10, |
| "loss": 0.3293, |
| "step": 1871 |
| }, |
| { |
| "epoch": 3.0, |
| "learning_rate": 1.3437656298687096e-10, |
| "loss": 0.3146, |
| "step": 1872 |
| }, |
| { |
| "epoch": 3.0, |
| "learning_rate": 5.972299119250124e-11, |
| "loss": 0.3115, |
| "step": 1873 |
| }, |
| { |
| "epoch": 3.0, |
| "learning_rate": 1.4930758944764477e-11, |
| "loss": 0.3086, |
| "step": 1874 |
| }, |
| { |
| "epoch": 3.0, |
| "learning_rate": 0.0, |
| "loss": 0.3199, |
| "step": 1875 |
| }, |
| { |
| "epoch": 3.0, |
| "step": 1875, |
| "total_flos": 5.1437414787946906e+17, |
| "train_loss": 0.44529419000943504, |
| "train_runtime": 5883.7163, |
| "train_samples_per_second": 10.198, |
| "train_steps_per_second": 0.319 |
| } |
| ], |
| "max_steps": 1875, |
| "num_train_epochs": 3, |
| "total_flos": 5.1437414787946906e+17, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|