| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 1.0, | |
| "global_step": 1882, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0, | |
| "loss": 4.3013, | |
| "step": 1 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0, | |
| "loss": 5.0537, | |
| "step": 2 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0, | |
| "loss": 4.8477, | |
| "step": 3 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0, | |
| "loss": 4.8228, | |
| "step": 4 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0, | |
| "loss": 5.1895, | |
| "step": 5 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 3.5087719298245616e-07, | |
| "loss": 5.0039, | |
| "step": 6 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 3.5087719298245616e-07, | |
| "loss": 4.5347, | |
| "step": 7 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 7.017543859649123e-07, | |
| "loss": 4.8564, | |
| "step": 8 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 7.017543859649123e-07, | |
| "loss": 4.5962, | |
| "step": 9 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.0526315789473685e-06, | |
| "loss": 5.0859, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.4035087719298246e-06, | |
| "loss": 4.3311, | |
| "step": 11 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.7543859649122807e-06, | |
| "loss": 5.0488, | |
| "step": 12 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 2.105263157894737e-06, | |
| "loss": 4.167, | |
| "step": 13 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 2.456140350877193e-06, | |
| "loss": 4.1665, | |
| "step": 14 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 2.8070175438596493e-06, | |
| "loss": 2.9976, | |
| "step": 15 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 3.157894736842105e-06, | |
| "loss": 4.0581, | |
| "step": 16 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 3.5087719298245615e-06, | |
| "loss": 3.6899, | |
| "step": 17 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 3.859649122807018e-06, | |
| "loss": 4.1084, | |
| "step": 18 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 4.210526315789474e-06, | |
| "loss": 3.2878, | |
| "step": 19 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 4.56140350877193e-06, | |
| "loss": 3.7476, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 4.912280701754386e-06, | |
| "loss": 3.3218, | |
| "step": 21 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 5.263157894736842e-06, | |
| "loss": 3.4507, | |
| "step": 22 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 5.6140350877192985e-06, | |
| "loss": 2.7751, | |
| "step": 23 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 5.964912280701755e-06, | |
| "loss": 3.501, | |
| "step": 24 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 6.31578947368421e-06, | |
| "loss": 2.3955, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 6.666666666666667e-06, | |
| "loss": 3.0967, | |
| "step": 26 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 7.017543859649123e-06, | |
| "loss": 3.251, | |
| "step": 27 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 7.368421052631579e-06, | |
| "loss": 2.8652, | |
| "step": 28 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 7.719298245614036e-06, | |
| "loss": 2.2854, | |
| "step": 29 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 8.070175438596492e-06, | |
| "loss": 2.7856, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 8.421052631578948e-06, | |
| "loss": 2.5811, | |
| "step": 31 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 8.771929824561405e-06, | |
| "loss": 2.8745, | |
| "step": 32 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.12280701754386e-06, | |
| "loss": 2.2925, | |
| "step": 33 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.473684210526315e-06, | |
| "loss": 2.1658, | |
| "step": 34 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.824561403508772e-06, | |
| "loss": 2.0569, | |
| "step": 35 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.017543859649123e-05, | |
| "loss": 2.467, | |
| "step": 36 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.0526315789473684e-05, | |
| "loss": 2.2043, | |
| "step": 37 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.0877192982456142e-05, | |
| "loss": 1.7561, | |
| "step": 38 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.1228070175438597e-05, | |
| "loss": 2.1765, | |
| "step": 39 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.1578947368421053e-05, | |
| "loss": 1.9973, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.192982456140351e-05, | |
| "loss": 1.6996, | |
| "step": 41 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.2280701754385966e-05, | |
| "loss": 1.8696, | |
| "step": 42 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.2280701754385966e-05, | |
| "loss": 1.8546, | |
| "step": 43 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.263157894736842e-05, | |
| "loss": 2.0425, | |
| "step": 44 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.2982456140350879e-05, | |
| "loss": 1.5527, | |
| "step": 45 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.3333333333333333e-05, | |
| "loss": 1.3848, | |
| "step": 46 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.3684210526315791e-05, | |
| "loss": 1.7981, | |
| "step": 47 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.4035087719298246e-05, | |
| "loss": 1.7969, | |
| "step": 48 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.4385964912280704e-05, | |
| "loss": 1.7014, | |
| "step": 49 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.4736842105263159e-05, | |
| "loss": 2.1182, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.5087719298245615e-05, | |
| "loss": 1.8892, | |
| "step": 51 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.543859649122807e-05, | |
| "loss": 1.424, | |
| "step": 52 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.578947368421053e-05, | |
| "loss": 1.9663, | |
| "step": 53 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.6140350877192984e-05, | |
| "loss": 1.5938, | |
| "step": 54 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.649122807017544e-05, | |
| "loss": 1.8535, | |
| "step": 55 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.6842105263157896e-05, | |
| "loss": 1.7747, | |
| "step": 56 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.719298245614035e-05, | |
| "loss": 1.4884, | |
| "step": 57 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.754385964912281e-05, | |
| "loss": 1.9126, | |
| "step": 58 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.7894736842105264e-05, | |
| "loss": 1.8411, | |
| "step": 59 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.824561403508772e-05, | |
| "loss": 1.4797, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.8596491228070176e-05, | |
| "loss": 1.9333, | |
| "step": 61 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.894736842105263e-05, | |
| "loss": 1.6064, | |
| "step": 62 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.929824561403509e-05, | |
| "loss": 1.2698, | |
| "step": 63 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.9649122807017544e-05, | |
| "loss": 1.8608, | |
| "step": 64 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2e-05, | |
| "loss": 1.667, | |
| "step": 65 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9999985183558698e-05, | |
| "loss": 1.9678, | |
| "step": 66 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.999994073427869e-05, | |
| "loss": 1.8567, | |
| "step": 67 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9999866652291696e-05, | |
| "loss": 1.3821, | |
| "step": 68 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9999762937817238e-05, | |
| "loss": 1.4658, | |
| "step": 69 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9999629591162658e-05, | |
| "loss": 1.4771, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9999466612723096e-05, | |
| "loss": 1.5078, | |
| "step": 71 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9999274002981504e-05, | |
| "loss": 1.6313, | |
| "step": 72 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9999051762508643e-05, | |
| "loss": 1.5953, | |
| "step": 73 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9998799891963078e-05, | |
| "loss": 1.4343, | |
| "step": 74 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9998518392091163e-05, | |
| "loss": 1.7412, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9998207263727076e-05, | |
| "loss": 1.2452, | |
| "step": 76 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9997866507792773e-05, | |
| "loss": 2.5244, | |
| "step": 77 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.999749612529801e-05, | |
| "loss": 1.4602, | |
| "step": 78 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9997096117340346e-05, | |
| "loss": 1.387, | |
| "step": 79 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9996666485105115e-05, | |
| "loss": 1.8118, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.999620722986544e-05, | |
| "loss": 1.0156, | |
| "step": 81 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9995718352982222e-05, | |
| "loss": 1.8379, | |
| "step": 82 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9995199855904156e-05, | |
| "loss": 1.3518, | |
| "step": 83 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.999465174016769e-05, | |
| "loss": 1.5979, | |
| "step": 84 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.999407400739705e-05, | |
| "loss": 1.5774, | |
| "step": 85 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9993466659304227e-05, | |
| "loss": 1.6187, | |
| "step": 86 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9992829697688965e-05, | |
| "loss": 1.0842, | |
| "step": 87 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.999216312443877e-05, | |
| "loss": 1.2056, | |
| "step": 88 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.999146694152888e-05, | |
| "loss": 1.1965, | |
| "step": 89 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9990741151022302e-05, | |
| "loss": 1.5195, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.998998575506975e-05, | |
| "loss": 1.6423, | |
| "step": 91 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9989200755909683e-05, | |
| "loss": 1.1444, | |
| "step": 92 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9988386155868284e-05, | |
| "loss": 1.679, | |
| "step": 93 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.998754195735944e-05, | |
| "loss": 1.3276, | |
| "step": 94 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9986668162884763e-05, | |
| "loss": 1.6624, | |
| "step": 95 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.998576477503355e-05, | |
| "loss": 1.7776, | |
| "step": 96 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.99848317964828e-05, | |
| "loss": 1.835, | |
| "step": 97 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.998386922999721e-05, | |
| "loss": 2.0483, | |
| "step": 98 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9982877078429127e-05, | |
| "loss": 1.6133, | |
| "step": 99 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9981855344718587e-05, | |
| "loss": 1.7546, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9980804031893285e-05, | |
| "loss": 1.5505, | |
| "step": 101 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9979723143068558e-05, | |
| "loss": 2.0273, | |
| "step": 102 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.99786126814474e-05, | |
| "loss": 1.2808, | |
| "step": 103 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9977472650320418e-05, | |
| "loss": 1.3708, | |
| "step": 104 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.997630305306586e-05, | |
| "loss": 1.3704, | |
| "step": 105 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9975103893149578e-05, | |
| "loss": 1.3005, | |
| "step": 106 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9973875174125027e-05, | |
| "loss": 1.5488, | |
| "step": 107 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9972616899633256e-05, | |
| "loss": 1.8574, | |
| "step": 108 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.99713290734029e-05, | |
| "loss": 1.3904, | |
| "step": 109 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.997001169925015e-05, | |
| "loss": 1.6042, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9968664781078777e-05, | |
| "loss": 1.7639, | |
| "step": 111 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9967288322880077e-05, | |
| "loss": 1.6328, | |
| "step": 112 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9965882328732903e-05, | |
| "loss": 1.6816, | |
| "step": 113 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.996444680280361e-05, | |
| "loss": 1.5764, | |
| "step": 114 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.996298174934608e-05, | |
| "loss": 1.5027, | |
| "step": 115 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.996148717270169e-05, | |
| "loss": 1.3486, | |
| "step": 116 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.99599630772993e-05, | |
| "loss": 1.3203, | |
| "step": 117 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9958409467655245e-05, | |
| "loss": 1.73, | |
| "step": 118 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9956826348373317e-05, | |
| "loss": 1.6848, | |
| "step": 119 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9955213724144754e-05, | |
| "loss": 1.4216, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.995357159974823e-05, | |
| "loss": 1.5259, | |
| "step": 121 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9951899980049827e-05, | |
| "loss": 1.0669, | |
| "step": 122 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9950198870003043e-05, | |
| "loss": 1.5493, | |
| "step": 123 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9948468274648752e-05, | |
| "loss": 1.3342, | |
| "step": 124 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.994670819911521e-05, | |
| "loss": 1.1314, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.994491864861803e-05, | |
| "loss": 1.4163, | |
| "step": 126 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.994309962846016e-05, | |
| "loss": 1.3759, | |
| "step": 127 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9941251144031885e-05, | |
| "loss": 1.5073, | |
| "step": 128 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9939373200810797e-05, | |
| "loss": 1.4451, | |
| "step": 129 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9937465804361783e-05, | |
| "loss": 1.509, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9935528960337006e-05, | |
| "loss": 1.3611, | |
| "step": 131 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9933562674475898e-05, | |
| "loss": 1.5658, | |
| "step": 132 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9931566952605126e-05, | |
| "loss": 1.5679, | |
| "step": 133 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9929541800638593e-05, | |
| "loss": 1.4707, | |
| "step": 134 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9927487224577402e-05, | |
| "loss": 1.9133, | |
| "step": 135 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.992540323050986e-05, | |
| "loss": 1.7588, | |
| "step": 136 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9923289824611446e-05, | |
| "loss": 1.5405, | |
| "step": 137 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9921147013144782e-05, | |
| "loss": 1.4901, | |
| "step": 138 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9918974802459637e-05, | |
| "loss": 1.7483, | |
| "step": 139 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.99167731989929e-05, | |
| "loss": 1.1279, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9914542209268562e-05, | |
| "loss": 1.2559, | |
| "step": 141 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9912281839897678e-05, | |
| "loss": 1.5688, | |
| "step": 142 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.990999209757838e-05, | |
| "loss": 1.1227, | |
| "step": 143 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9907672989095835e-05, | |
| "loss": 1.2604, | |
| "step": 144 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.990532452132223e-05, | |
| "loss": 1.1417, | |
| "step": 145 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.990294670121675e-05, | |
| "loss": 1.5295, | |
| "step": 146 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9900539535825562e-05, | |
| "loss": 1.2983, | |
| "step": 147 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.989810303228179e-05, | |
| "loss": 1.47, | |
| "step": 148 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.98956371978055e-05, | |
| "loss": 1.5767, | |
| "step": 149 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9893142039703662e-05, | |
| "loss": 1.2815, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9890617565370164e-05, | |
| "loss": 1.3386, | |
| "step": 151 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9888063782285734e-05, | |
| "loss": 1.6052, | |
| "step": 152 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9885480698017978e-05, | |
| "loss": 1.3588, | |
| "step": 153 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.988286832022132e-05, | |
| "loss": 1.7864, | |
| "step": 154 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9880226656636977e-05, | |
| "loss": 1.3562, | |
| "step": 155 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9877555715092974e-05, | |
| "loss": 1.2598, | |
| "step": 156 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.987485550350407e-05, | |
| "loss": 1.3306, | |
| "step": 157 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9872126029871776e-05, | |
| "loss": 1.2645, | |
| "step": 158 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9869367302284307e-05, | |
| "loss": 1.4874, | |
| "step": 159 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.986657932891657e-05, | |
| "loss": 1.34, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9863762118030135e-05, | |
| "loss": 1.512, | |
| "step": 161 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9860915677973203e-05, | |
| "loss": 1.4294, | |
| "step": 162 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9858040017180604e-05, | |
| "loss": 1.5164, | |
| "step": 163 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9855135144173747e-05, | |
| "loss": 1.7258, | |
| "step": 164 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9852201067560607e-05, | |
| "loss": 1.2268, | |
| "step": 165 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.98492377960357e-05, | |
| "loss": 1.6624, | |
| "step": 166 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9846245338380053e-05, | |
| "loss": 1.7061, | |
| "step": 167 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.984322370346118e-05, | |
| "loss": 1.8572, | |
| "step": 168 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.984017290023306e-05, | |
| "loss": 1.6533, | |
| "step": 169 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.98370929377361e-05, | |
| "loss": 1.9138, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9833983825097115e-05, | |
| "loss": 1.4189, | |
| "step": 171 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.98308455715293e-05, | |
| "loss": 1.5532, | |
| "step": 172 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.982767818633221e-05, | |
| "loss": 1.4919, | |
| "step": 173 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9824481678891724e-05, | |
| "loss": 1.307, | |
| "step": 174 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.982125605868001e-05, | |
| "loss": 1.8789, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9818001335255503e-05, | |
| "loss": 1.5227, | |
| "step": 176 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9814717518262897e-05, | |
| "loss": 1.389, | |
| "step": 177 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9811404617433085e-05, | |
| "loss": 1.5764, | |
| "step": 178 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.980806264258315e-05, | |
| "loss": 1.418, | |
| "step": 179 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.9804691603616324e-05, | |
| "loss": 1.6702, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.9801291510521965e-05, | |
| "loss": 1.1479, | |
| "step": 181 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.979786237337553e-05, | |
| "loss": 1.236, | |
| "step": 182 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.9794404202338545e-05, | |
| "loss": 1.2885, | |
| "step": 183 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.9790917007658564e-05, | |
| "loss": 1.0085, | |
| "step": 184 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.9787400799669155e-05, | |
| "loss": 1.5703, | |
| "step": 185 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.978385558878985e-05, | |
| "loss": 1.2833, | |
| "step": 186 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.9780281385526127e-05, | |
| "loss": 1.8744, | |
| "step": 187 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.977667820046939e-05, | |
| "loss": 1.3972, | |
| "step": 188 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.9773046044296914e-05, | |
| "loss": 1.4944, | |
| "step": 189 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.976938492777182e-05, | |
| "loss": 2.0579, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.976569486174305e-05, | |
| "loss": 1.2612, | |
| "step": 191 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.976197585714534e-05, | |
| "loss": 1.6089, | |
| "step": 192 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.975822792499917e-05, | |
| "loss": 1.2356, | |
| "step": 193 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.9754451076410742e-05, | |
| "loss": 1.6089, | |
| "step": 194 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.9750645322571952e-05, | |
| "loss": 1.3259, | |
| "step": 195 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.9746810674760337e-05, | |
| "loss": 1.5605, | |
| "step": 196 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.974294714433907e-05, | |
| "loss": 1.4265, | |
| "step": 197 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9739054742756903e-05, | |
| "loss": 1.6106, | |
| "step": 198 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9735133481548145e-05, | |
| "loss": 1.3005, | |
| "step": 199 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.973118337233262e-05, | |
| "loss": 1.3264, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9727204426815647e-05, | |
| "loss": 1.2583, | |
| "step": 201 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9723196656787985e-05, | |
| "loss": 1.4524, | |
| "step": 202 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.971916007412581e-05, | |
| "loss": 1.4526, | |
| "step": 203 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9715094690790684e-05, | |
| "loss": 1.5798, | |
| "step": 204 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9711000518829505e-05, | |
| "loss": 1.4553, | |
| "step": 205 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.970687757037449e-05, | |
| "loss": 1.4746, | |
| "step": 206 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9702725857643118e-05, | |
| "loss": 1.35, | |
| "step": 207 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9698545392938114e-05, | |
| "loss": 1.1138, | |
| "step": 208 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.96943361886474e-05, | |
| "loss": 1.4111, | |
| "step": 209 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9690098257244063e-05, | |
| "loss": 1.3623, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9685831611286312e-05, | |
| "loss": 1.5491, | |
| "step": 211 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.968153626341745e-05, | |
| "loss": 1.1873, | |
| "step": 212 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9677212226365833e-05, | |
| "loss": 1.4312, | |
| "step": 213 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9672859512944827e-05, | |
| "loss": 1.4226, | |
| "step": 214 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9668478136052776e-05, | |
| "loss": 1.3821, | |
| "step": 215 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9664068108672968e-05, | |
| "loss": 1.3643, | |
| "step": 216 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9659629443873576e-05, | |
| "loss": 0.9437, | |
| "step": 217 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.965516215480765e-05, | |
| "loss": 1.4976, | |
| "step": 218 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9650666254713055e-05, | |
| "loss": 1.2327, | |
| "step": 219 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9646141756912437e-05, | |
| "loss": 1.3577, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9641588674813188e-05, | |
| "loss": 1.396, | |
| "step": 221 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.96370070219074e-05, | |
| "loss": 1.402, | |
| "step": 222 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9632396811771837e-05, | |
| "loss": 1.4839, | |
| "step": 223 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.962775805806788e-05, | |
| "loss": 2.3965, | |
| "step": 224 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.962309077454149e-05, | |
| "loss": 1.8308, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.961839497502317e-05, | |
| "loss": 1.554, | |
| "step": 226 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9613670673427937e-05, | |
| "loss": 0.9191, | |
| "step": 227 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.960891788375525e-05, | |
| "loss": 1.2283, | |
| "step": 228 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9604136620088998e-05, | |
| "loss": 1.278, | |
| "step": 229 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9599326896597448e-05, | |
| "loss": 1.3079, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.959448872753319e-05, | |
| "loss": 1.0803, | |
| "step": 231 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9589622127233117e-05, | |
| "loss": 1.4797, | |
| "step": 232 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9584727110118365e-05, | |
| "loss": 1.4531, | |
| "step": 233 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.957980369069429e-05, | |
| "loss": 1.45, | |
| "step": 234 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9574851883550395e-05, | |
| "loss": 1.1499, | |
| "step": 235 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.956987170336031e-05, | |
| "loss": 1.3311, | |
| "step": 236 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9564863164881755e-05, | |
| "loss": 0.9471, | |
| "step": 237 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.955982628295646e-05, | |
| "loss": 1.3564, | |
| "step": 238 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9554761072510165e-05, | |
| "loss": 1.3945, | |
| "step": 239 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9549667548552557e-05, | |
| "loss": 1.4125, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9544545726177197e-05, | |
| "loss": 1.5623, | |
| "step": 241 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9539395620561536e-05, | |
| "loss": 1.3271, | |
| "step": 242 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9534217246966815e-05, | |
| "loss": 1.2179, | |
| "step": 243 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9529010620738053e-05, | |
| "loss": 1.043, | |
| "step": 244 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9523775757303975e-05, | |
| "loss": 1.1204, | |
| "step": 245 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9518512672177e-05, | |
| "loss": 1.5908, | |
| "step": 246 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9513221380953162e-05, | |
| "loss": 1.7395, | |
| "step": 247 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9507901899312084e-05, | |
| "loss": 2.0334, | |
| "step": 248 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.950255424301692e-05, | |
| "loss": 1.3418, | |
| "step": 249 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.949717842791432e-05, | |
| "loss": 1.5706, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.949177446993437e-05, | |
| "loss": 1.218, | |
| "step": 251 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9486342385090565e-05, | |
| "loss": 1.3102, | |
| "step": 252 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9480882189479725e-05, | |
| "loss": 1.3953, | |
| "step": 253 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9475393899281996e-05, | |
| "loss": 1.4675, | |
| "step": 254 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9469877530760753e-05, | |
| "loss": 1.741, | |
| "step": 255 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9464333100262595e-05, | |
| "loss": 1.4646, | |
| "step": 256 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9458760624217263e-05, | |
| "loss": 1.627, | |
| "step": 257 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.945316011913761e-05, | |
| "loss": 1.5132, | |
| "step": 258 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9447531601619548e-05, | |
| "loss": 1.2711, | |
| "step": 259 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9441875088342e-05, | |
| "loss": 1.5598, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9436190596066834e-05, | |
| "loss": 1.541, | |
| "step": 261 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9430478141638853e-05, | |
| "loss": 1.47, | |
| "step": 262 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9424737741985698e-05, | |
| "loss": 1.082, | |
| "step": 263 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9418969414117833e-05, | |
| "loss": 1.7605, | |
| "step": 264 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9413173175128472e-05, | |
| "loss": 1.2251, | |
| "step": 265 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9407349042193544e-05, | |
| "loss": 1.3693, | |
| "step": 266 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.940149703257163e-05, | |
| "loss": 1.4143, | |
| "step": 267 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9395617163603934e-05, | |
| "loss": 1.3105, | |
| "step": 268 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9389709452714183e-05, | |
| "loss": 1.4005, | |
| "step": 269 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9383773917408644e-05, | |
| "loss": 1.0288, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.937781057527601e-05, | |
| "loss": 1.1827, | |
| "step": 271 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9371819443987388e-05, | |
| "loss": 1.4882, | |
| "step": 272 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.936580054129622e-05, | |
| "loss": 1.6135, | |
| "step": 273 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.935975388503826e-05, | |
| "loss": 1.3396, | |
| "step": 274 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9353679493131486e-05, | |
| "loss": 1.1813, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.934757738357607e-05, | |
| "loss": 2.2588, | |
| "step": 276 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.934144757445433e-05, | |
| "loss": 1.4857, | |
| "step": 277 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9335290083930652e-05, | |
| "loss": 1.5806, | |
| "step": 278 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9329104930251453e-05, | |
| "loss": 1.741, | |
| "step": 279 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9322892131745135e-05, | |
| "loss": 1.2673, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9316651706822e-05, | |
| "loss": 1.2403, | |
| "step": 281 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9310383673974237e-05, | |
| "loss": 1.4321, | |
| "step": 282 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.930408805177583e-05, | |
| "loss": 1.6628, | |
| "step": 283 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9297764858882516e-05, | |
| "loss": 1.2108, | |
| "step": 284 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9291414114031744e-05, | |
| "loss": 1.0958, | |
| "step": 285 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.92850358360426e-05, | |
| "loss": 1.2808, | |
| "step": 286 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9278630043815765e-05, | |
| "loss": 1.5178, | |
| "step": 287 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9272196756333443e-05, | |
| "loss": 1.0857, | |
| "step": 288 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.926573599265932e-05, | |
| "loss": 1.4011, | |
| "step": 289 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.92592477719385e-05, | |
| "loss": 1.345, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9252732113397454e-05, | |
| "loss": 0.9194, | |
| "step": 291 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9246189036343956e-05, | |
| "loss": 1.2887, | |
| "step": 292 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.923961856016703e-05, | |
| "loss": 1.3684, | |
| "step": 293 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9233020704336883e-05, | |
| "loss": 1.2866, | |
| "step": 294 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9226395488404875e-05, | |
| "loss": 1.3705, | |
| "step": 295 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.921974293200343e-05, | |
| "loss": 1.6641, | |
| "step": 296 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9213063054845983e-05, | |
| "loss": 0.8669, | |
| "step": 297 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9206355876726937e-05, | |
| "loss": 1.9102, | |
| "step": 298 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9199621417521597e-05, | |
| "loss": 1.4536, | |
| "step": 299 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9192859697186105e-05, | |
| "loss": 1.2253, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9186070735757393e-05, | |
| "loss": 1.2181, | |
| "step": 301 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.91792545533531e-05, | |
| "loss": 1.5474, | |
| "step": 302 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.917241117017155e-05, | |
| "loss": 1.3098, | |
| "step": 303 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9165540606491654e-05, | |
| "loss": 1.369, | |
| "step": 304 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9158642882672873e-05, | |
| "loss": 1.7593, | |
| "step": 305 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9151718019155153e-05, | |
| "loss": 1.5852, | |
| "step": 306 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.914476603645886e-05, | |
| "loss": 1.151, | |
| "step": 307 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9137786955184723e-05, | |
| "loss": 1.1422, | |
| "step": 308 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9130780796013773e-05, | |
| "loss": 1.2618, | |
| "step": 309 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9123747579707275e-05, | |
| "loss": 1.5203, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.911668732710668e-05, | |
| "loss": 1.3743, | |
| "step": 311 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.910960005913355e-05, | |
| "loss": 1.2466, | |
| "step": 312 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.9102485796789502e-05, | |
| "loss": 1.3762, | |
| "step": 313 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.909534456115615e-05, | |
| "loss": 1.2278, | |
| "step": 314 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.908817637339503e-05, | |
| "loss": 1.0023, | |
| "step": 315 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.908098125474755e-05, | |
| "loss": 1.1909, | |
| "step": 316 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.907375922653492e-05, | |
| "loss": 1.1071, | |
| "step": 317 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.9066510310158093e-05, | |
| "loss": 1.2804, | |
| "step": 318 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.90592345270977e-05, | |
| "loss": 1.3889, | |
| "step": 319 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.9051931898913977e-05, | |
| "loss": 1.165, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.904460244724672e-05, | |
| "loss": 1.3802, | |
| "step": 321 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.9037246193815206e-05, | |
| "loss": 1.2313, | |
| "step": 322 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.9029863160418138e-05, | |
| "loss": 0.9822, | |
| "step": 323 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.9022453368933566e-05, | |
| "loss": 1.4841, | |
| "step": 324 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.9015016841318843e-05, | |
| "loss": 1.0494, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.9007553599610543e-05, | |
| "loss": 1.9556, | |
| "step": 326 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.90000636659244e-05, | |
| "loss": 1.3706, | |
| "step": 327 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.899254706245525e-05, | |
| "loss": 1.1641, | |
| "step": 328 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.8985003811476954e-05, | |
| "loss": 1.2551, | |
| "step": 329 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8977433935342338e-05, | |
| "loss": 1.4486, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8969837456483132e-05, | |
| "loss": 1.1564, | |
| "step": 331 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.896221439740989e-05, | |
| "loss": 1.5452, | |
| "step": 332 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.895456478071193e-05, | |
| "loss": 1.5942, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8946888629057275e-05, | |
| "loss": 1.5427, | |
| "step": 334 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8939185965192572e-05, | |
| "loss": 1.3293, | |
| "step": 335 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8931456811943038e-05, | |
| "loss": 1.6553, | |
| "step": 336 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.892370119221238e-05, | |
| "loss": 1.2086, | |
| "step": 337 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8915919128982736e-05, | |
| "loss": 1.3274, | |
| "step": 338 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8908110645314602e-05, | |
| "loss": 1.2142, | |
| "step": 339 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.890027576434677e-05, | |
| "loss": 1.3337, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8892414509296243e-05, | |
| "loss": 1.2729, | |
| "step": 341 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.888452690345819e-05, | |
| "loss": 1.1399, | |
| "step": 342 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.887661297020586e-05, | |
| "loss": 1.125, | |
| "step": 343 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.886867273299052e-05, | |
| "loss": 1.1442, | |
| "step": 344 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8860706215341383e-05, | |
| "loss": 1.3079, | |
| "step": 345 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.885271344086553e-05, | |
| "loss": 1.0145, | |
| "step": 346 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8844694433247865e-05, | |
| "loss": 1.048, | |
| "step": 347 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8836649216251014e-05, | |
| "loss": 1.3777, | |
| "step": 348 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8828577813715274e-05, | |
| "loss": 1.5127, | |
| "step": 349 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8820480249558538e-05, | |
| "loss": 1.5735, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8812356547776223e-05, | |
| "loss": 1.0406, | |
| "step": 351 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8804206732441202e-05, | |
| "loss": 1.416, | |
| "step": 352 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8796030827703723e-05, | |
| "loss": 1.8662, | |
| "step": 353 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.878782885779135e-05, | |
| "loss": 1.5626, | |
| "step": 354 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8779600847008884e-05, | |
| "loss": 1.459, | |
| "step": 355 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8771346819738293e-05, | |
| "loss": 1.6757, | |
| "step": 356 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8763066800438638e-05, | |
| "loss": 1.7344, | |
| "step": 357 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8754760813646005e-05, | |
| "loss": 1.1858, | |
| "step": 358 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8746428883973427e-05, | |
| "loss": 1.2311, | |
| "step": 359 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.873807103611081e-05, | |
| "loss": 1.9961, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.872968729482487e-05, | |
| "loss": 1.0763, | |
| "step": 361 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8721277684959048e-05, | |
| "loss": 1.4128, | |
| "step": 362 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8712842231433443e-05, | |
| "loss": 1.3157, | |
| "step": 363 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8704380959244733e-05, | |
| "loss": 1.7656, | |
| "step": 364 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.869589389346611e-05, | |
| "loss": 1.0778, | |
| "step": 365 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8687381059247195e-05, | |
| "loss": 1.2793, | |
| "step": 366 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.867884248181397e-05, | |
| "loss": 1.3031, | |
| "step": 367 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.86702781864687e-05, | |
| "loss": 1.655, | |
| "step": 368 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.866168819858986e-05, | |
| "loss": 1.2544, | |
| "step": 369 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8653072543632064e-05, | |
| "loss": 1.9014, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8644431247125978e-05, | |
| "loss": 2.3298, | |
| "step": 371 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8635764334678256e-05, | |
| "loss": 1.7388, | |
| "step": 372 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8627071831971455e-05, | |
| "loss": 1.3171, | |
| "step": 373 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.861835376476397e-05, | |
| "loss": 1.1583, | |
| "step": 374 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8609610158889943e-05, | |
| "loss": 1.3706, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8600841040259207e-05, | |
| "loss": 1.2729, | |
| "step": 376 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8592046434857178e-05, | |
| "loss": 1.3982, | |
| "step": 377 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8583226368744816e-05, | |
| "loss": 1.9009, | |
| "step": 378 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8574380868058512e-05, | |
| "loss": 1.344, | |
| "step": 379 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8565509959010037e-05, | |
| "loss": 1.2316, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8556613667886455e-05, | |
| "loss": 1.4287, | |
| "step": 381 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8547692021050036e-05, | |
| "loss": 1.3557, | |
| "step": 382 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.853874504493819e-05, | |
| "loss": 1.3776, | |
| "step": 383 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8529772766063398e-05, | |
| "loss": 1.4821, | |
| "step": 384 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8520775211013094e-05, | |
| "loss": 1.4719, | |
| "step": 385 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8511752406449637e-05, | |
| "loss": 1.3525, | |
| "step": 386 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.850270437911019e-05, | |
| "loss": 1.1719, | |
| "step": 387 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8493631155806676e-05, | |
| "loss": 1.0137, | |
| "step": 388 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8484532763425662e-05, | |
| "loss": 1.7134, | |
| "step": 389 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8475409228928314e-05, | |
| "loss": 1.3694, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.846626057935029e-05, | |
| "loss": 1.5881, | |
| "step": 391 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8457086841801677e-05, | |
| "loss": 1.4768, | |
| "step": 392 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8447888043466906e-05, | |
| "loss": 1.5454, | |
| "step": 393 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8438664211604665e-05, | |
| "loss": 1.2961, | |
| "step": 394 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.842941537354783e-05, | |
| "loss": 1.3921, | |
| "step": 395 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.842014155670337e-05, | |
| "loss": 1.6172, | |
| "step": 396 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8410842788552283e-05, | |
| "loss": 1.751, | |
| "step": 397 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8401519096649497e-05, | |
| "loss": 1.3857, | |
| "step": 398 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8392170508623797e-05, | |
| "loss": 1.4458, | |
| "step": 399 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8382797052177746e-05, | |
| "loss": 1.1118, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8373398755087596e-05, | |
| "loss": 1.3054, | |
| "step": 401 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.836397564520321e-05, | |
| "loss": 1.2959, | |
| "step": 402 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8354527750447985e-05, | |
| "loss": 1.6274, | |
| "step": 403 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8345055098818748e-05, | |
| "loss": 1.5647, | |
| "step": 404 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8335557718385702e-05, | |
| "loss": 1.7551, | |
| "step": 405 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.832603563729232e-05, | |
| "loss": 1.8701, | |
| "step": 406 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8316488883755275e-05, | |
| "loss": 1.5657, | |
| "step": 407 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8306917486064347e-05, | |
| "loss": 1.8792, | |
| "step": 408 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8297321472582354e-05, | |
| "loss": 1.0668, | |
| "step": 409 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8287700871745036e-05, | |
| "loss": 1.0858, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8278055712061023e-05, | |
| "loss": 1.802, | |
| "step": 411 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.826838602211169e-05, | |
| "loss": 1.2771, | |
| "step": 412 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8258691830551127e-05, | |
| "loss": 1.5005, | |
| "step": 413 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.824897316610601e-05, | |
| "loss": 0.9551, | |
| "step": 414 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8239230057575542e-05, | |
| "loss": 1.3989, | |
| "step": 415 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.822946253383137e-05, | |
| "loss": 1.2817, | |
| "step": 416 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8219670623817478e-05, | |
| "loss": 1.4912, | |
| "step": 417 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8209854356550117e-05, | |
| "loss": 1.6382, | |
| "step": 418 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8200013761117715e-05, | |
| "loss": 1.4473, | |
| "step": 419 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.81901488666808e-05, | |
| "loss": 1.1669, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8180259702471895e-05, | |
| "loss": 1.0759, | |
| "step": 421 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.817034629779544e-05, | |
| "loss": 1.1851, | |
| "step": 422 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.816040868202771e-05, | |
| "loss": 1.2006, | |
| "step": 423 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.815044688461673e-05, | |
| "loss": 1.3104, | |
| "step": 424 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.814046093508218e-05, | |
| "loss": 1.4014, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.81304508630153e-05, | |
| "loss": 1.63, | |
| "step": 426 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.8120416698078823e-05, | |
| "loss": 1.4056, | |
| "step": 427 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.811035847000687e-05, | |
| "loss": 0.9681, | |
| "step": 428 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.8100276208604873e-05, | |
| "loss": 1.1392, | |
| "step": 429 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.8090169943749477e-05, | |
| "loss": 1.6406, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.8080039705388455e-05, | |
| "loss": 1.3745, | |
| "step": 431 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.8069885523540633e-05, | |
| "loss": 1.1749, | |
| "step": 432 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.8059707428295764e-05, | |
| "loss": 1.1752, | |
| "step": 433 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.8049505449814493e-05, | |
| "loss": 1.1987, | |
| "step": 434 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.8039279618328215e-05, | |
| "loss": 1.397, | |
| "step": 435 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.802902996413902e-05, | |
| "loss": 1.7351, | |
| "step": 436 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.801875651761958e-05, | |
| "loss": 1.5282, | |
| "step": 437 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.8008459309213087e-05, | |
| "loss": 1.2327, | |
| "step": 438 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.799813836943314e-05, | |
| "loss": 1.038, | |
| "step": 439 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.798779372886365e-05, | |
| "loss": 1.1648, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.7977425418158777e-05, | |
| "loss": 1.2062, | |
| "step": 441 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.7967033468042806e-05, | |
| "loss": 1.0673, | |
| "step": 442 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7956617909310086e-05, | |
| "loss": 1.5444, | |
| "step": 443 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7946178772824917e-05, | |
| "loss": 1.7515, | |
| "step": 444 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7935716089521474e-05, | |
| "loss": 1.0441, | |
| "step": 445 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7925229890403702e-05, | |
| "loss": 1.2168, | |
| "step": 446 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.791472020654523e-05, | |
| "loss": 1.3445, | |
| "step": 447 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7904187069089284e-05, | |
| "loss": 1.3372, | |
| "step": 448 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.789363050924858e-05, | |
| "loss": 1.3062, | |
| "step": 449 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7883050558305255e-05, | |
| "loss": 1.1299, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.787244724761075e-05, | |
| "loss": 1.5092, | |
| "step": 451 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7861820608585733e-05, | |
| "loss": 1.4873, | |
| "step": 452 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7851170672719998e-05, | |
| "loss": 1.2642, | |
| "step": 453 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.784049747157237e-05, | |
| "loss": 1.3093, | |
| "step": 454 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7829801036770628e-05, | |
| "loss": 1.5476, | |
| "step": 455 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7819081400011392e-05, | |
| "loss": 1.6396, | |
| "step": 456 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.780833859306003e-05, | |
| "loss": 1.3862, | |
| "step": 457 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7797572647750575e-05, | |
| "loss": 1.0403, | |
| "step": 458 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7786783595985633e-05, | |
| "loss": 1.8372, | |
| "step": 459 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.777597146973627e-05, | |
| "loss": 1.3606, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7765136301041936e-05, | |
| "loss": 1.7499, | |
| "step": 461 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7754278122010353e-05, | |
| "loss": 1.0402, | |
| "step": 462 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.774339696481744e-05, | |
| "loss": 1.5647, | |
| "step": 463 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7732492861707205e-05, | |
| "loss": 1.4514, | |
| "step": 464 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7721565844991643e-05, | |
| "loss": 1.75, | |
| "step": 465 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.771061594705066e-05, | |
| "loss": 1.4635, | |
| "step": 466 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7699643200331956e-05, | |
| "loss": 1.5381, | |
| "step": 467 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.768864763735094e-05, | |
| "loss": 1.1631, | |
| "step": 468 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.767762929069064e-05, | |
| "loss": 1.4386, | |
| "step": 469 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7666588193001595e-05, | |
| "loss": 1.3824, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.765552437700175e-05, | |
| "loss": 1.4971, | |
| "step": 471 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7644437875476393e-05, | |
| "loss": 1.7865, | |
| "step": 472 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7633328721278017e-05, | |
| "loss": 1.4209, | |
| "step": 473 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.762219694732625e-05, | |
| "loss": 1.0745, | |
| "step": 474 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7611042586607748e-05, | |
| "loss": 1.6687, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7599865672176095e-05, | |
| "loss": 1.377, | |
| "step": 476 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.758866623715171e-05, | |
| "loss": 1.4009, | |
| "step": 477 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7577444314721752e-05, | |
| "loss": 1.2423, | |
| "step": 478 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7566199938140006e-05, | |
| "loss": 1.3231, | |
| "step": 479 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7554933140726803e-05, | |
| "loss": 1.3174, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7543643955868916e-05, | |
| "loss": 1.201, | |
| "step": 481 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.753233241701945e-05, | |
| "loss": 1.0433, | |
| "step": 482 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7520998557697753e-05, | |
| "loss": 1.6056, | |
| "step": 483 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.750964241148932e-05, | |
| "loss": 1.3042, | |
| "step": 484 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7498264012045686e-05, | |
| "loss": 1.3018, | |
| "step": 485 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7486863393084327e-05, | |
| "loss": 1.5397, | |
| "step": 486 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.747544058838857e-05, | |
| "loss": 1.4519, | |
| "step": 487 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.746399563180747e-05, | |
| "loss": 1.4274, | |
| "step": 488 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.745252855725573e-05, | |
| "loss": 1.3577, | |
| "step": 489 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.744103939871361e-05, | |
| "loss": 1.4641, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7429528190226784e-05, | |
| "loss": 1.3225, | |
| "step": 491 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.741799496590629e-05, | |
| "loss": 1.2202, | |
| "step": 492 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.74064397599284e-05, | |
| "loss": 1.3506, | |
| "step": 493 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7394862606534508e-05, | |
| "loss": 1.5345, | |
| "step": 494 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.738326354003107e-05, | |
| "loss": 1.0104, | |
| "step": 495 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.737164259478945e-05, | |
| "loss": 1.2505, | |
| "step": 496 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7359999805245865e-05, | |
| "loss": 1.3712, | |
| "step": 497 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7348335205901262e-05, | |
| "loss": 1.4348, | |
| "step": 498 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7336648831321205e-05, | |
| "loss": 1.0427, | |
| "step": 499 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.732494071613579e-05, | |
| "loss": 1.2288, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7313210895039542e-05, | |
| "loss": 1.1718, | |
| "step": 501 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7301459402791297e-05, | |
| "loss": 1.2524, | |
| "step": 502 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7289686274214116e-05, | |
| "loss": 0.8266, | |
| "step": 503 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7277891544195177e-05, | |
| "loss": 1.3052, | |
| "step": 504 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7266075247685656e-05, | |
| "loss": 1.301, | |
| "step": 505 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7254237419700654e-05, | |
| "loss": 0.8944, | |
| "step": 506 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7242378095319066e-05, | |
| "loss": 1.3367, | |
| "step": 507 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7230497309683484e-05, | |
| "loss": 1.5593, | |
| "step": 508 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7218595098000107e-05, | |
| "loss": 1.4673, | |
| "step": 509 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.720667149553861e-05, | |
| "loss": 1.2711, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7194726537632077e-05, | |
| "loss": 1.3459, | |
| "step": 511 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7182760259676854e-05, | |
| "loss": 1.7144, | |
| "step": 512 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.717077269713247e-05, | |
| "loss": 1.2749, | |
| "step": 513 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7158763885521533e-05, | |
| "loss": 1.3904, | |
| "step": 514 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7146733860429614e-05, | |
| "loss": 0.9453, | |
| "step": 515 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.713468265750514e-05, | |
| "loss": 1.4587, | |
| "step": 516 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.71226103124593e-05, | |
| "loss": 1.5103, | |
| "step": 517 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.7110516861065934e-05, | |
| "loss": 1.0211, | |
| "step": 518 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.709840233916143e-05, | |
| "loss": 1.4265, | |
| "step": 519 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.70862667826446e-05, | |
| "loss": 1.2263, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.70741102274766e-05, | |
| "loss": 1.0671, | |
| "step": 521 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.7061932709680803e-05, | |
| "loss": 1.3446, | |
| "step": 522 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.7049734265342713e-05, | |
| "loss": 1.5065, | |
| "step": 523 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.7037514930609833e-05, | |
| "loss": 1.4187, | |
| "step": 524 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.702527474169157e-05, | |
| "loss": 1.6499, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.7013013734859138e-05, | |
| "loss": 1.0529, | |
| "step": 526 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.7000731946445433e-05, | |
| "loss": 1.3557, | |
| "step": 527 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.698842941284493e-05, | |
| "loss": 1.4429, | |
| "step": 528 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.6976106170513585e-05, | |
| "loss": 1.4391, | |
| "step": 529 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.6963762255968723e-05, | |
| "loss": 1.7664, | |
| "step": 530 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.6951397705788916e-05, | |
| "loss": 1.5029, | |
| "step": 531 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.693901255661389e-05, | |
| "loss": 1.2434, | |
| "step": 532 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.692660684514441e-05, | |
| "loss": 1.655, | |
| "step": 533 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.6914180608142182e-05, | |
| "loss": 1.5623, | |
| "step": 534 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.690173388242972e-05, | |
| "loss": 1.5901, | |
| "step": 535 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.688926670489027e-05, | |
| "loss": 1.697, | |
| "step": 536 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.687677911246766e-05, | |
| "loss": 1.256, | |
| "step": 537 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.686427114216624e-05, | |
| "loss": 1.0255, | |
| "step": 538 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6851742831050725e-05, | |
| "loss": 1.282, | |
| "step": 539 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.683919421624611e-05, | |
| "loss": 1.5226, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6826625334937557e-05, | |
| "loss": 1.2266, | |
| "step": 541 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.681403622437029e-05, | |
| "loss": 1.2521, | |
| "step": 542 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.680142692184947e-05, | |
| "loss": 1.939, | |
| "step": 543 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6788797464740096e-05, | |
| "loss": 1.1821, | |
| "step": 544 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.677614789046689e-05, | |
| "loss": 1.2166, | |
| "step": 545 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6763478236514187e-05, | |
| "loss": 1.3701, | |
| "step": 546 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.675078854042582e-05, | |
| "loss": 1.0472, | |
| "step": 547 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6738078839805023e-05, | |
| "loss": 1.2358, | |
| "step": 548 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6725349172314296e-05, | |
| "loss": 1.4001, | |
| "step": 549 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6712599575675318e-05, | |
| "loss": 1.0444, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6699830087668812e-05, | |
| "loss": 1.1985, | |
| "step": 551 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6687040746134468e-05, | |
| "loss": 1.5686, | |
| "step": 552 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6674231588970772e-05, | |
| "loss": 1.5007, | |
| "step": 553 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.666140265413496e-05, | |
| "loss": 1.5073, | |
| "step": 554 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6648553979642867e-05, | |
| "loss": 1.375, | |
| "step": 555 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.663568560356881e-05, | |
| "loss": 1.3721, | |
| "step": 556 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6622797564045502e-05, | |
| "loss": 2.0796, | |
| "step": 557 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6609889899263918e-05, | |
| "loss": 1.1556, | |
| "step": 558 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.659696264747319e-05, | |
| "loss": 1.0874, | |
| "step": 559 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.658401584698049e-05, | |
| "loss": 1.5211, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.658401584698049e-05, | |
| "loss": 1.501, | |
| "step": 561 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6571049536150922e-05, | |
| "loss": 1.3557, | |
| "step": 562 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6558063753407403e-05, | |
| "loss": 1.0767, | |
| "step": 563 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.654505853723055e-05, | |
| "loss": 1.6049, | |
| "step": 564 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6532033926158564e-05, | |
| "loss": 1.4917, | |
| "step": 565 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6518989958787126e-05, | |
| "loss": 1.5476, | |
| "step": 566 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6505926673769273e-05, | |
| "loss": 1.2092, | |
| "step": 567 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6492844109815277e-05, | |
| "loss": 1.6907, | |
| "step": 568 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6479742305692553e-05, | |
| "loss": 1.6099, | |
| "step": 569 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6466621300225522e-05, | |
| "loss": 1.4248, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6453481132295507e-05, | |
| "loss": 1.2837, | |
| "step": 571 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6440321840840604e-05, | |
| "loss": 1.412, | |
| "step": 572 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6427143464855596e-05, | |
| "loss": 1.3649, | |
| "step": 573 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6413946043391812e-05, | |
| "loss": 1.73, | |
| "step": 574 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6400729615557006e-05, | |
| "loss": 1.0562, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6387494220515276e-05, | |
| "loss": 1.1256, | |
| "step": 576 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.63742398974869e-05, | |
| "loss": 1.5317, | |
| "step": 577 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6360966685748264e-05, | |
| "loss": 1.5579, | |
| "step": 578 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6347674624631723e-05, | |
| "loss": 1.1333, | |
| "step": 579 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.633436375352548e-05, | |
| "loss": 1.4811, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6321034111873487e-05, | |
| "loss": 0.8725, | |
| "step": 581 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6307685739175318e-05, | |
| "loss": 1.4773, | |
| "step": 582 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6294318674986038e-05, | |
| "loss": 1.2795, | |
| "step": 583 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6280932958916126e-05, | |
| "loss": 1.1422, | |
| "step": 584 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6267528630631303e-05, | |
| "loss": 1.2706, | |
| "step": 585 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6254105729852466e-05, | |
| "loss": 0.918, | |
| "step": 586 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6240664296355534e-05, | |
| "loss": 2.1196, | |
| "step": 587 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6227204369971357e-05, | |
| "loss": 1.4177, | |
| "step": 588 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6213725990585572e-05, | |
| "loss": 1.0474, | |
| "step": 589 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6200229198138503e-05, | |
| "loss": 1.6956, | |
| "step": 590 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6186714032625036e-05, | |
| "loss": 1.0042, | |
| "step": 591 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6173180534094502e-05, | |
| "loss": 1.4338, | |
| "step": 592 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.615962874265056e-05, | |
| "loss": 1.0308, | |
| "step": 593 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.6146058698451072e-05, | |
| "loss": 1.0973, | |
| "step": 594 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.6132470441707993e-05, | |
| "loss": 1.1433, | |
| "step": 595 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.6118864012687246e-05, | |
| "loss": 0.9455, | |
| "step": 596 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.6105239451708596e-05, | |
| "loss": 0.9333, | |
| "step": 597 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.6091596799145556e-05, | |
| "loss": 1.3525, | |
| "step": 598 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.6077936095425223e-05, | |
| "loss": 1.4381, | |
| "step": 599 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.606425738102821e-05, | |
| "loss": 1.6763, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.6050560696488493e-05, | |
| "loss": 1.1395, | |
| "step": 601 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.6036846082393288e-05, | |
| "loss": 1.5312, | |
| "step": 602 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.6023113579382958e-05, | |
| "loss": 1.5347, | |
| "step": 603 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.6009363228150864e-05, | |
| "loss": 1.4935, | |
| "step": 604 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.5995595069443257e-05, | |
| "loss": 1.5088, | |
| "step": 605 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.5981809144059167e-05, | |
| "loss": 1.3062, | |
| "step": 606 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.5968005492850258e-05, | |
| "loss": 1.2356, | |
| "step": 607 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.5954184156720733e-05, | |
| "loss": 1.2164, | |
| "step": 608 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.5940345176627197e-05, | |
| "loss": 1.2456, | |
| "step": 609 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.592648859357853e-05, | |
| "loss": 1.2041, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.5912614448635784e-05, | |
| "loss": 0.9447, | |
| "step": 611 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5898722782912053e-05, | |
| "loss": 1.9265, | |
| "step": 612 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5884813637572345e-05, | |
| "loss": 1.4448, | |
| "step": 613 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.587088705383347e-05, | |
| "loss": 1.4138, | |
| "step": 614 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5856943072963904e-05, | |
| "loss": 1.0443, | |
| "step": 615 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5842981736283686e-05, | |
| "loss": 1.2151, | |
| "step": 616 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.582900308516428e-05, | |
| "loss": 1.6422, | |
| "step": 617 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.581500716102846e-05, | |
| "loss": 1.4928, | |
| "step": 618 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5800994005350182e-05, | |
| "loss": 1.3112, | |
| "step": 619 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.578696365965447e-05, | |
| "loss": 1.4049, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5772916165517275e-05, | |
| "loss": 1.4368, | |
| "step": 621 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5758851564565378e-05, | |
| "loss": 1.2131, | |
| "step": 622 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5744769898476244e-05, | |
| "loss": 1.0887, | |
| "step": 623 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5730671208977904e-05, | |
| "loss": 1.6411, | |
| "step": 624 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5716555537848847e-05, | |
| "loss": 1.7678, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5702422926917872e-05, | |
| "loss": 1.3297, | |
| "step": 626 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5688273418063977e-05, | |
| "loss": 1.1743, | |
| "step": 627 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.567410705321624e-05, | |
| "loss": 1.321, | |
| "step": 628 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5659923874353685e-05, | |
| "loss": 1.2682, | |
| "step": 629 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.564572392350515e-05, | |
| "loss": 1.5809, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5631507242749187e-05, | |
| "loss": 1.217, | |
| "step": 631 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5617273874213923e-05, | |
| "loss": 1.2251, | |
| "step": 632 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5603023860076932e-05, | |
| "loss": 1.1614, | |
| "step": 633 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5588757242565108e-05, | |
| "loss": 1.688, | |
| "step": 634 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5574474063954555e-05, | |
| "loss": 1.3516, | |
| "step": 635 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5560174366570448e-05, | |
| "loss": 1.4905, | |
| "step": 636 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5545858192786913e-05, | |
| "loss": 1.0424, | |
| "step": 637 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5531525585026894e-05, | |
| "loss": 1.29, | |
| "step": 638 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.551717658576205e-05, | |
| "loss": 1.2068, | |
| "step": 639 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5502811237512592e-05, | |
| "loss": 1.1237, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5488429582847194e-05, | |
| "loss": 1.2704, | |
| "step": 641 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.547403166438284e-05, | |
| "loss": 1.0896, | |
| "step": 642 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.545961752478472e-05, | |
| "loss": 1.3013, | |
| "step": 643 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.544518720676608e-05, | |
| "loss": 1.3387, | |
| "step": 644 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5430740753088107e-05, | |
| "loss": 0.9934, | |
| "step": 645 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5416278206559816e-05, | |
| "loss": 1.3215, | |
| "step": 646 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.54017996100379e-05, | |
| "loss": 2.1733, | |
| "step": 647 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5387305006426606e-05, | |
| "loss": 1.0956, | |
| "step": 648 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5372794438677628e-05, | |
| "loss": 1.251, | |
| "step": 649 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5358267949789968e-05, | |
| "loss": 1.4861, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5343725582809793e-05, | |
| "loss": 1.3944, | |
| "step": 651 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5329167380830325e-05, | |
| "loss": 1.2812, | |
| "step": 652 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.531459338699172e-05, | |
| "loss": 1.4543, | |
| "step": 653 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.530000364448092e-05, | |
| "loss": 1.6724, | |
| "step": 654 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5285398196531534e-05, | |
| "loss": 1.2543, | |
| "step": 655 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5270777086423724e-05, | |
| "loss": 1.075, | |
| "step": 656 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5256140357484044e-05, | |
| "loss": 1.067, | |
| "step": 657 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5241488053085347e-05, | |
| "loss": 1.1625, | |
| "step": 658 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5226820216646631e-05, | |
| "loss": 1.3367, | |
| "step": 659 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5212136891632926e-05, | |
| "loss": 2.0203, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5197438121555159e-05, | |
| "loss": 1.731, | |
| "step": 661 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5182723949970014e-05, | |
| "loss": 1.5283, | |
| "step": 662 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5167994420479835e-05, | |
| "loss": 2.6377, | |
| "step": 663 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5153249576732453e-05, | |
| "loss": 1.4193, | |
| "step": 664 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5138489462421096e-05, | |
| "loss": 1.4926, | |
| "step": 665 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.512371412128424e-05, | |
| "loss": 1.558, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.510892359710547e-05, | |
| "loss": 1.5416, | |
| "step": 667 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5094117933713381e-05, | |
| "loss": 1.7449, | |
| "step": 668 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.5079297174981421e-05, | |
| "loss": 1.2051, | |
| "step": 669 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.5064461364827768e-05, | |
| "loss": 1.3638, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.5049610547215205e-05, | |
| "loss": 1.1598, | |
| "step": 671 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.5034744766150988e-05, | |
| "loss": 1.4731, | |
| "step": 672 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.5019864065686705e-05, | |
| "loss": 1.1443, | |
| "step": 673 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.5004968489918168e-05, | |
| "loss": 1.0511, | |
| "step": 674 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4990058082985255e-05, | |
| "loss": 1.2859, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4975132889071808e-05, | |
| "loss": 1.4406, | |
| "step": 676 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4960192952405471e-05, | |
| "loss": 1.2043, | |
| "step": 677 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4945238317257589e-05, | |
| "loss": 1.3831, | |
| "step": 678 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4930269027943055e-05, | |
| "loss": 1.2134, | |
| "step": 679 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4915285128820186e-05, | |
| "loss": 1.3292, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4900286664290593e-05, | |
| "loss": 1.4397, | |
| "step": 681 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4885273678799057e-05, | |
| "loss": 1.3098, | |
| "step": 682 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4870246216833375e-05, | |
| "loss": 1.7405, | |
| "step": 683 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4855204322924253e-05, | |
| "loss": 1.0132, | |
| "step": 684 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4840148041645154e-05, | |
| "loss": 1.0487, | |
| "step": 685 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4825077417612187e-05, | |
| "loss": 1.2492, | |
| "step": 686 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4809992495483944e-05, | |
| "loss": 1.4015, | |
| "step": 687 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4794893319961407e-05, | |
| "loss": 1.3253, | |
| "step": 688 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4779779935787786e-05, | |
| "loss": 1.1931, | |
| "step": 689 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4764652387748388e-05, | |
| "loss": 1.1512, | |
| "step": 690 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4749510720670506e-05, | |
| "loss": 0.9799, | |
| "step": 691 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4734354979423255e-05, | |
| "loss": 1.085, | |
| "step": 692 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4719185208917475e-05, | |
| "loss": 1.1534, | |
| "step": 693 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4704001454105564e-05, | |
| "loss": 1.0544, | |
| "step": 694 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4688803759981363e-05, | |
| "loss": 1.223, | |
| "step": 695 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4673592171580025e-05, | |
| "loss": 1.5525, | |
| "step": 696 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4658366733977866e-05, | |
| "loss": 1.1721, | |
| "step": 697 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.464312749229225e-05, | |
| "loss": 1.1598, | |
| "step": 698 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4627874491681445e-05, | |
| "loss": 1.0315, | |
| "step": 699 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4612607777344483e-05, | |
| "loss": 1.4377, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4597327394521044e-05, | |
| "loss": 1.1886, | |
| "step": 701 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4582033388491303e-05, | |
| "loss": 1.1663, | |
| "step": 702 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4566725804575815e-05, | |
| "loss": 1.2335, | |
| "step": 703 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4551404688135357e-05, | |
| "loss": 1.1642, | |
| "step": 704 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4536070084570816e-05, | |
| "loss": 1.2684, | |
| "step": 705 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4520722039323045e-05, | |
| "loss": 1.2324, | |
| "step": 706 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4505360597872721e-05, | |
| "loss": 1.5039, | |
| "step": 707 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4489985805740227e-05, | |
| "loss": 1.5469, | |
| "step": 708 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4474597708485507e-05, | |
| "loss": 1.3518, | |
| "step": 709 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.445919635170792e-05, | |
| "loss": 1.1639, | |
| "step": 710 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4443781781046135e-05, | |
| "loss": 1.2637, | |
| "step": 711 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4428354042177963e-05, | |
| "loss": 1.0478, | |
| "step": 712 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4412913180820241e-05, | |
| "loss": 1.1447, | |
| "step": 713 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4397459242728697e-05, | |
| "loss": 0.9182, | |
| "step": 714 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4381992273697801e-05, | |
| "loss": 1.0979, | |
| "step": 715 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4366512319560642e-05, | |
| "loss": 1.2832, | |
| "step": 716 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4351019426188782e-05, | |
| "loss": 1.1368, | |
| "step": 717 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4335513639492136e-05, | |
| "loss": 1.057, | |
| "step": 718 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4319995005418817e-05, | |
| "loss": 1.1445, | |
| "step": 719 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4304463569955008e-05, | |
| "loss": 1.7061, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4288919379124837e-05, | |
| "loss": 1.4265, | |
| "step": 721 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4273362478990219e-05, | |
| "loss": 1.3456, | |
| "step": 722 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4257792915650728e-05, | |
| "loss": 1.7507, | |
| "step": 723 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4242210735243477e-05, | |
| "loss": 1.5076, | |
| "step": 724 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.4226615983942948e-05, | |
| "loss": 1.2609, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.4211008707960897e-05, | |
| "loss": 1.3202, | |
| "step": 726 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.4195388953546172e-05, | |
| "loss": 1.0796, | |
| "step": 727 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.417975676698461e-05, | |
| "loss": 1.0226, | |
| "step": 728 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.4164112194598891e-05, | |
| "loss": 0.7252, | |
| "step": 729 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.4148455282748389e-05, | |
| "loss": 1.0625, | |
| "step": 730 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.4132786077829044e-05, | |
| "loss": 1.1504, | |
| "step": 731 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.411710462627323e-05, | |
| "loss": 1.2118, | |
| "step": 732 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.410141097454961e-05, | |
| "loss": 1.3751, | |
| "step": 733 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.4085705169162995e-05, | |
| "loss": 1.5261, | |
| "step": 734 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.4069987256654214e-05, | |
| "loss": 1.4089, | |
| "step": 735 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.4054257283599974e-05, | |
| "loss": 1.2002, | |
| "step": 736 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.4038515296612717e-05, | |
| "loss": 1.0856, | |
| "step": 737 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.4022761342340492e-05, | |
| "loss": 1.199, | |
| "step": 738 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.4006995467466805e-05, | |
| "loss": 1.2207, | |
| "step": 739 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.3991217718710489e-05, | |
| "loss": 1.3678, | |
| "step": 740 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.3975428142825562e-05, | |
| "loss": 1.4629, | |
| "step": 741 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.3959626786601084e-05, | |
| "loss": 1.1454, | |
| "step": 742 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.394381369686103e-05, | |
| "loss": 1.2129, | |
| "step": 743 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3927988920464151e-05, | |
| "loss": 1.1326, | |
| "step": 744 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3912152504303809e-05, | |
| "loss": 1.3939, | |
| "step": 745 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3896304495307881e-05, | |
| "loss": 0.9822, | |
| "step": 746 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.388044494043858e-05, | |
| "loss": 2.6819, | |
| "step": 747 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3864573886692338e-05, | |
| "loss": 1.0029, | |
| "step": 748 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3848691381099668e-05, | |
| "loss": 1.6024, | |
| "step": 749 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3832797470725004e-05, | |
| "loss": 1.3887, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3816892202666591e-05, | |
| "loss": 1.4873, | |
| "step": 751 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.380097562405632e-05, | |
| "loss": 1.7628, | |
| "step": 752 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3785047782059606e-05, | |
| "loss": 1.0988, | |
| "step": 753 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.376910872387523e-05, | |
| "loss": 0.9005, | |
| "step": 754 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3753158496735217e-05, | |
| "loss": 1.323, | |
| "step": 755 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.373719714790469e-05, | |
| "loss": 1.7451, | |
| "step": 756 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3721224724681727e-05, | |
| "loss": 1.4352, | |
| "step": 757 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.370524127439722e-05, | |
| "loss": 1.3624, | |
| "step": 758 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3689246844414745e-05, | |
| "loss": 1.3093, | |
| "step": 759 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.36732414821304e-05, | |
| "loss": 1.5457, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3657225234972695e-05, | |
| "loss": 1.0577, | |
| "step": 761 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3641198150402379e-05, | |
| "loss": 1.2852, | |
| "step": 762 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.362516027591233e-05, | |
| "loss": 1.355, | |
| "step": 763 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3609111659027395e-05, | |
| "loss": 1.3281, | |
| "step": 764 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3593052347304244e-05, | |
| "loss": 1.0302, | |
| "step": 765 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3576982388331258e-05, | |
| "loss": 1.1615, | |
| "step": 766 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3560901829728346e-05, | |
| "loss": 1.1178, | |
| "step": 767 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3544810719146844e-05, | |
| "loss": 1.208, | |
| "step": 768 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3528709104269354e-05, | |
| "loss": 1.1903, | |
| "step": 769 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3512597032809596e-05, | |
| "loss": 1.7498, | |
| "step": 770 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3496474552512286e-05, | |
| "loss": 1.2798, | |
| "step": 771 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3480341711152981e-05, | |
| "loss": 1.0945, | |
| "step": 772 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.346419855653794e-05, | |
| "loss": 1.6519, | |
| "step": 773 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3448045136503981e-05, | |
| "loss": 0.9259, | |
| "step": 774 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.343188149891835e-05, | |
| "loss": 1.1141, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3415707691678557e-05, | |
| "loss": 1.2324, | |
| "step": 776 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3399523762712258e-05, | |
| "loss": 2.1128, | |
| "step": 777 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3383329759977104e-05, | |
| "loss": 1.4387, | |
| "step": 778 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3367125731460588e-05, | |
| "loss": 1.5535, | |
| "step": 779 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3350911725179919e-05, | |
| "loss": 0.87, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.333468778918187e-05, | |
| "loss": 1.4635, | |
| "step": 781 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3318453971542643e-05, | |
| "loss": 1.328, | |
| "step": 782 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3302210320367717e-05, | |
| "loss": 0.8982, | |
| "step": 783 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3285956883791715e-05, | |
| "loss": 1.2517, | |
| "step": 784 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3269693709978253e-05, | |
| "loss": 1.2649, | |
| "step": 785 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3253420847119804e-05, | |
| "loss": 1.3037, | |
| "step": 786 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.323713834343755e-05, | |
| "loss": 1.1211, | |
| "step": 787 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3220846247181242e-05, | |
| "loss": 1.4719, | |
| "step": 788 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3204544606629063e-05, | |
| "loss": 1.1603, | |
| "step": 789 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3188233470087469e-05, | |
| "loss": 1.25, | |
| "step": 790 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3171912885891063e-05, | |
| "loss": 1.2263, | |
| "step": 791 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3155582902402433e-05, | |
| "loss": 1.0795, | |
| "step": 792 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3139243568012035e-05, | |
| "loss": 1.2391, | |
| "step": 793 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3122894931138021e-05, | |
| "loss": 1.2001, | |
| "step": 794 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.310653704022612e-05, | |
| "loss": 1.3142, | |
| "step": 795 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3090169943749475e-05, | |
| "loss": 1.3589, | |
| "step": 796 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3073793690208513e-05, | |
| "loss": 1.29, | |
| "step": 797 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3057408328130792e-05, | |
| "loss": 1.283, | |
| "step": 798 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.304101390607086e-05, | |
| "loss": 1.4434, | |
| "step": 799 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.3024610472610122e-05, | |
| "loss": 1.4553, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.3008198076356678e-05, | |
| "loss": 1.0197, | |
| "step": 801 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2991776765945185e-05, | |
| "loss": 1.6511, | |
| "step": 802 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2975346590036724e-05, | |
| "loss": 1.0184, | |
| "step": 803 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2958907597318642e-05, | |
| "loss": 1.0, | |
| "step": 804 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2942459836504409e-05, | |
| "loss": 1.2113, | |
| "step": 805 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2926003356333487e-05, | |
| "loss": 0.9717, | |
| "step": 806 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2909538205571166e-05, | |
| "loss": 1.0708, | |
| "step": 807 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2893064433008435e-05, | |
| "loss": 1.0999, | |
| "step": 808 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2876582087461832e-05, | |
| "loss": 1.3896, | |
| "step": 809 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.28600912177733e-05, | |
| "loss": 1.5399, | |
| "step": 810 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2843591872810039e-05, | |
| "loss": 1.2301, | |
| "step": 811 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2827084101464358e-05, | |
| "loss": 1.6228, | |
| "step": 812 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2810567952653548e-05, | |
| "loss": 1.6653, | |
| "step": 813 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2794043475319717e-05, | |
| "loss": 1.0361, | |
| "step": 814 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2777510718429656e-05, | |
| "loss": 1.2181, | |
| "step": 815 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2760969730974692e-05, | |
| "loss": 1.3923, | |
| "step": 816 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.274442056197053e-05, | |
| "loss": 2.4648, | |
| "step": 817 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2727863260457137e-05, | |
| "loss": 1.1083, | |
| "step": 818 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2711297875498565e-05, | |
| "loss": 1.4048, | |
| "step": 819 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2694724456182827e-05, | |
| "loss": 1.3169, | |
| "step": 820 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2678143051621743e-05, | |
| "loss": 1.243, | |
| "step": 821 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2661553710950796e-05, | |
| "loss": 1.228, | |
| "step": 822 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2644956483328976e-05, | |
| "loss": 1.4551, | |
| "step": 823 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2628351417938662e-05, | |
| "loss": 1.4553, | |
| "step": 824 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2611738563985443e-05, | |
| "loss": 1.2489, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2595117970697998e-05, | |
| "loss": 1.4492, | |
| "step": 826 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2578489687327938e-05, | |
| "loss": 1.3523, | |
| "step": 827 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2561853763149654e-05, | |
| "loss": 1.3286, | |
| "step": 828 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2545210247460186e-05, | |
| "loss": 0.9979, | |
| "step": 829 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2528559189579072e-05, | |
| "loss": 1.3179, | |
| "step": 830 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2511900638848196e-05, | |
| "loss": 1.0013, | |
| "step": 831 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2495234644631643e-05, | |
| "loss": 0.9755, | |
| "step": 832 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2478561256315556e-05, | |
| "loss": 1.1569, | |
| "step": 833 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2461880523307999e-05, | |
| "loss": 1.3511, | |
| "step": 834 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2445192495038788e-05, | |
| "loss": 1.3884, | |
| "step": 835 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2428497220959359e-05, | |
| "loss": 1.0623, | |
| "step": 836 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2411794750542622e-05, | |
| "loss": 1.2485, | |
| "step": 837 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2395085133282814e-05, | |
| "loss": 1.4451, | |
| "step": 838 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2378368418695347e-05, | |
| "loss": 1.6956, | |
| "step": 839 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2361644656316666e-05, | |
| "loss": 1.185, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2344913895704099e-05, | |
| "loss": 1.2, | |
| "step": 841 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2328176186435712e-05, | |
| "loss": 1.0282, | |
| "step": 842 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.231143157811016e-05, | |
| "loss": 1.7632, | |
| "step": 843 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.229468012034655e-05, | |
| "loss": 1.1569, | |
| "step": 844 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2277921862784273e-05, | |
| "loss": 1.9338, | |
| "step": 845 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2261156855082882e-05, | |
| "loss": 1.2981, | |
| "step": 846 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.224438514692193e-05, | |
| "loss": 1.0535, | |
| "step": 847 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2227606788000817e-05, | |
| "loss": 1.3912, | |
| "step": 848 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.221082182803866e-05, | |
| "loss": 1.2954, | |
| "step": 849 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2194030316774128e-05, | |
| "loss": 1.0662, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.217723230396532e-05, | |
| "loss": 1.4008, | |
| "step": 851 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2160427839389581e-05, | |
| "loss": 1.203, | |
| "step": 852 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2143616972843388e-05, | |
| "loss": 0.9495, | |
| "step": 853 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2126799754142187e-05, | |
| "loss": 1.531, | |
| "step": 854 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2109976233120237e-05, | |
| "loss": 1.3204, | |
| "step": 855 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2093146459630488e-05, | |
| "loss": 1.2781, | |
| "step": 856 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.2076310483544405e-05, | |
| "loss": 1.2793, | |
| "step": 857 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.2059468354751838e-05, | |
| "loss": 1.2566, | |
| "step": 858 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.2042620123160876e-05, | |
| "loss": 1.0972, | |
| "step": 859 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.2025765838697679e-05, | |
| "loss": 1.1635, | |
| "step": 860 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.2008905551306356e-05, | |
| "loss": 1.0101, | |
| "step": 861 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1992039310948791e-05, | |
| "loss": 1.2179, | |
| "step": 862 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1975167167604522e-05, | |
| "loss": 0.892, | |
| "step": 863 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1958289171270574e-05, | |
| "loss": 1.2198, | |
| "step": 864 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.194140537196131e-05, | |
| "loss": 1.1333, | |
| "step": 865 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.19245158197083e-05, | |
| "loss": 0.7264, | |
| "step": 866 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1907620564560154e-05, | |
| "loss": 0.996, | |
| "step": 867 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1890719656582381e-05, | |
| "loss": 0.9788, | |
| "step": 868 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.187381314585725e-05, | |
| "loss": 1.1442, | |
| "step": 869 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1856901082483614e-05, | |
| "loss": 1.4034, | |
| "step": 870 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1839983516576802e-05, | |
| "loss": 1.1276, | |
| "step": 871 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1823060498268437e-05, | |
| "loss": 1.1305, | |
| "step": 872 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1806132077706294e-05, | |
| "loss": 0.9137, | |
| "step": 873 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1789198305054173e-05, | |
| "loss": 1.5786, | |
| "step": 874 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1772259230491714e-05, | |
| "loss": 0.9683, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1755314904214284e-05, | |
| "loss": 1.204, | |
| "step": 876 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.17383653764328e-05, | |
| "loss": 1.3901, | |
| "step": 877 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1721410697373608e-05, | |
| "loss": 1.2372, | |
| "step": 878 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1704450917278305e-05, | |
| "loss": 1.24, | |
| "step": 879 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1687486086403603e-05, | |
| "loss": 1.1837, | |
| "step": 880 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1670516255021193e-05, | |
| "loss": 1.3528, | |
| "step": 881 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1653541473417575e-05, | |
| "loss": 1.2477, | |
| "step": 882 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.163656179189392e-05, | |
| "loss": 1.157, | |
| "step": 883 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.161957726076592e-05, | |
| "loss": 1.0137, | |
| "step": 884 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1602587930363634e-05, | |
| "loss": 0.635, | |
| "step": 885 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1585593851031346e-05, | |
| "loss": 1.2363, | |
| "step": 886 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1568595073127415e-05, | |
| "loss": 1.1462, | |
| "step": 887 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1551591647024116e-05, | |
| "loss": 1.1281, | |
| "step": 888 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1534583623107505e-05, | |
| "loss": 1.5206, | |
| "step": 889 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1517571051777256e-05, | |
| "loss": 1.3745, | |
| "step": 890 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1500553983446527e-05, | |
| "loss": 0.9387, | |
| "step": 891 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1483532468541796e-05, | |
| "loss": 1.1343, | |
| "step": 892 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1466506557502713e-05, | |
| "loss": 1.4128, | |
| "step": 893 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1449476300781962e-05, | |
| "loss": 2.0195, | |
| "step": 894 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1432441748845106e-05, | |
| "loss": 1.2207, | |
| "step": 895 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1415402952170434e-05, | |
| "loss": 1.1995, | |
| "step": 896 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1398359961248808e-05, | |
| "loss": 1.6768, | |
| "step": 897 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1381312826583522e-05, | |
| "loss": 1.0055, | |
| "step": 898 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1364261598690154e-05, | |
| "loss": 1.2704, | |
| "step": 899 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1347206328096405e-05, | |
| "loss": 1.2705, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1330147065341962e-05, | |
| "loss": 1.585, | |
| "step": 901 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.131308386097833e-05, | |
| "loss": 1.1731, | |
| "step": 902 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1296016765568709e-05, | |
| "loss": 1.4296, | |
| "step": 903 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1278945829687826e-05, | |
| "loss": 1.1735, | |
| "step": 904 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1261871103921773e-05, | |
| "loss": 1.2449, | |
| "step": 905 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1244792638867895e-05, | |
| "loss": 1.2799, | |
| "step": 906 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.12277104851346e-05, | |
| "loss": 1.0692, | |
| "step": 907 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1210624693341236e-05, | |
| "loss": 1.0469, | |
| "step": 908 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1193535314117933e-05, | |
| "loss": 1.061, | |
| "step": 909 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1176442398105445e-05, | |
| "loss": 1.1091, | |
| "step": 910 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1159345995955007e-05, | |
| "loss": 1.7698, | |
| "step": 911 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1142246158328189e-05, | |
| "loss": 1.3259, | |
| "step": 912 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.112514293589674e-05, | |
| "loss": 1.3638, | |
| "step": 913 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.1108036379342438e-05, | |
| "loss": 1.3446, | |
| "step": 914 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.1090926539356937e-05, | |
| "loss": 0.9806, | |
| "step": 915 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.1073813466641633e-05, | |
| "loss": 1.3291, | |
| "step": 916 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.1056697211907486e-05, | |
| "loss": 1.6324, | |
| "step": 917 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.1039577825874897e-05, | |
| "loss": 1.4233, | |
| "step": 918 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.1022455359273538e-05, | |
| "loss": 0.687, | |
| "step": 919 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.1005329862842218e-05, | |
| "loss": 1.2122, | |
| "step": 920 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0988201387328716e-05, | |
| "loss": 1.468, | |
| "step": 921 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0971069983489644e-05, | |
| "loss": 1.0305, | |
| "step": 922 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.095393570209029e-05, | |
| "loss": 1.1708, | |
| "step": 923 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0936798593904467e-05, | |
| "loss": 0.6168, | |
| "step": 924 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.091965870971437e-05, | |
| "loss": 1.1608, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0902516100310412e-05, | |
| "loss": 2.0864, | |
| "step": 926 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.088537081649109e-05, | |
| "loss": 1.5171, | |
| "step": 927 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0868222909062817e-05, | |
| "loss": 1.3392, | |
| "step": 928 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0851072428839798e-05, | |
| "loss": 1.532, | |
| "step": 929 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0833919426643835e-05, | |
| "loss": 1.4431, | |
| "step": 930 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0816763953304228e-05, | |
| "loss": 1.0113, | |
| "step": 931 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0799606059657583e-05, | |
| "loss": 1.4988, | |
| "step": 932 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.078244579654769e-05, | |
| "loss": 1.7383, | |
| "step": 933 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0765283214825354e-05, | |
| "loss": 1.8447, | |
| "step": 934 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.074811836534825e-05, | |
| "loss": 1.0134, | |
| "step": 935 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0730951298980776e-05, | |
| "loss": 1.3136, | |
| "step": 936 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0713782066593901e-05, | |
| "loss": 1.0061, | |
| "step": 937 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0696610719065005e-05, | |
| "loss": 1.6003, | |
| "step": 938 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0679437307277742e-05, | |
| "loss": 1.1381, | |
| "step": 939 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0662261882121883e-05, | |
| "loss": 1.6621, | |
| "step": 940 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0645084494493166e-05, | |
| "loss": 1.3147, | |
| "step": 941 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0627905195293135e-05, | |
| "loss": 1.4095, | |
| "step": 942 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0610724035429012e-05, | |
| "loss": 1.1195, | |
| "step": 943 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0593541065813525e-05, | |
| "loss": 1.2004, | |
| "step": 944 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0576356337364762e-05, | |
| "loss": 0.7862, | |
| "step": 945 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0559169901006035e-05, | |
| "loss": 1.4277, | |
| "step": 946 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.05419818076657e-05, | |
| "loss": 1.1136, | |
| "step": 947 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.052479210827704e-05, | |
| "loss": 1.3027, | |
| "step": 948 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0507600853778083e-05, | |
| "loss": 1.3463, | |
| "step": 949 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0490408095111477e-05, | |
| "loss": 1.1909, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0473213883224321e-05, | |
| "loss": 0.8979, | |
| "step": 951 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0456018269068015e-05, | |
| "loss": 1.333, | |
| "step": 952 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.043882130359813e-05, | |
| "loss": 1.2893, | |
| "step": 953 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0421623037774229e-05, | |
| "loss": 1.4016, | |
| "step": 954 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0404423522559726e-05, | |
| "loss": 1.1405, | |
| "step": 955 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0387222808921746e-05, | |
| "loss": 1.2434, | |
| "step": 956 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0370020947830963e-05, | |
| "loss": 1.2371, | |
| "step": 957 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0352817990261448e-05, | |
| "loss": 1.0873, | |
| "step": 958 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0335613987190527e-05, | |
| "loss": 1.4753, | |
| "step": 959 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0318408989598616e-05, | |
| "loss": 1.2473, | |
| "step": 960 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0301203048469084e-05, | |
| "loss": 1.4866, | |
| "step": 961 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0283996214788095e-05, | |
| "loss": 1.1008, | |
| "step": 962 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0266788539544455e-05, | |
| "loss": 1.1621, | |
| "step": 963 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0249580073729467e-05, | |
| "loss": 1.358, | |
| "step": 964 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0232370868336778e-05, | |
| "loss": 1.2321, | |
| "step": 965 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0215160974362224e-05, | |
| "loss": 1.3789, | |
| "step": 966 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.019795044280368e-05, | |
| "loss": 1.2107, | |
| "step": 967 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0180739324660909e-05, | |
| "loss": 1.1742, | |
| "step": 968 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0163527670935422e-05, | |
| "loss": 0.983, | |
| "step": 969 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.0146315532630307e-05, | |
| "loss": 1.216, | |
| "step": 970 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.0129102960750092e-05, | |
| "loss": 1.05, | |
| "step": 971 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.0111890006300588e-05, | |
| "loss": 1.0742, | |
| "step": 972 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.0094676720288744e-05, | |
| "loss": 1.1709, | |
| "step": 973 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.0077463153722484e-05, | |
| "loss": 1.2085, | |
| "step": 974 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.006024935761057e-05, | |
| "loss": 1.3473, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.0043035382962443e-05, | |
| "loss": 1.4586, | |
| "step": 976 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.002582128078807e-05, | |
| "loss": 1.2925, | |
| "step": 977 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.0008607102097797e-05, | |
| "loss": 1.4438, | |
| "step": 978 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.991392897902204e-06, | |
| "loss": 0.9645, | |
| "step": 979 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.974178719211933e-06, | |
| "loss": 1.3232, | |
| "step": 980 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.956964617037559e-06, | |
| "loss": 1.6111, | |
| "step": 981 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.939750642389432e-06, | |
| "loss": 1.1509, | |
| "step": 982 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.922536846277521e-06, | |
| "loss": 0.9647, | |
| "step": 983 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.90532327971126e-06, | |
| "loss": 1.2428, | |
| "step": 984 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.888109993699415e-06, | |
| "loss": 1.0573, | |
| "step": 985 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.87089703924991e-06, | |
| "loss": 1.444, | |
| "step": 986 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.853684467369697e-06, | |
| "loss": 1.2001, | |
| "step": 987 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.836472329064582e-06, | |
| "loss": 1.0806, | |
| "step": 988 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.819260675339093e-06, | |
| "loss": 1.6799, | |
| "step": 989 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.802049557196326e-06, | |
| "loss": 1.5151, | |
| "step": 990 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.78483902563778e-06, | |
| "loss": 1.1816, | |
| "step": 991 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.767629131663225e-06, | |
| "loss": 1.0006, | |
| "step": 992 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.750419926270533e-06, | |
| "loss": 1.1964, | |
| "step": 993 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.733211460455549e-06, | |
| "loss": 1.0698, | |
| "step": 994 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.71600378521191e-06, | |
| "loss": 1.1614, | |
| "step": 995 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.69879695153092e-06, | |
| "loss": 1.1968, | |
| "step": 996 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.68159101040139e-06, | |
| "loss": 0.9535, | |
| "step": 997 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.664386012809473e-06, | |
| "loss": 1.3633, | |
| "step": 998 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.647182009738555e-06, | |
| "loss": 1.1453, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.629979052169037e-06, | |
| "loss": 1.2378, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.612777191078257e-06, | |
| "loss": 1.0991, | |
| "step": 1001 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.59557647744028e-06, | |
| "loss": 1.7493, | |
| "step": 1002 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.578376962225775e-06, | |
| "loss": 0.8675, | |
| "step": 1003 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.561178696401874e-06, | |
| "loss": 1.087, | |
| "step": 1004 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.543981730931984e-06, | |
| "loss": 1.0146, | |
| "step": 1005 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.526786116775682e-06, | |
| "loss": 1.735, | |
| "step": 1006 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.509591904888528e-06, | |
| "loss": 1.0151, | |
| "step": 1007 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.492399146221918e-06, | |
| "loss": 1.1523, | |
| "step": 1008 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.475207891722966e-06, | |
| "loss": 1.1028, | |
| "step": 1009 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.458018192334302e-06, | |
| "loss": 1.1841, | |
| "step": 1010 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.440830098993969e-06, | |
| "loss": 1.2302, | |
| "step": 1011 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.423643662635238e-06, | |
| "loss": 0.8669, | |
| "step": 1012 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.406458934186479e-06, | |
| "loss": 1.084, | |
| "step": 1013 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.389275964570995e-06, | |
| "loss": 1.0892, | |
| "step": 1014 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.372094804706867e-06, | |
| "loss": 1.2284, | |
| "step": 1015 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.354915505506839e-06, | |
| "loss": 1.1218, | |
| "step": 1016 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.337738117878119e-06, | |
| "loss": 1.4651, | |
| "step": 1017 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.32056269272226e-06, | |
| "loss": 1.0541, | |
| "step": 1018 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.303389280934997e-06, | |
| "loss": 0.9251, | |
| "step": 1019 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.286217933406102e-06, | |
| "loss": 1.1958, | |
| "step": 1020 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.269048701019226e-06, | |
| "loss": 1.2272, | |
| "step": 1021 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.251881634651752e-06, | |
| "loss": 1.3547, | |
| "step": 1022 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.234716785174648e-06, | |
| "loss": 1.0444, | |
| "step": 1023 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.217554203452312e-06, | |
| "loss": 1.1429, | |
| "step": 1024 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.200393940342419e-06, | |
| "loss": 1.371, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 9.183236046695777e-06, | |
| "loss": 1.3058, | |
| "step": 1026 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 9.166080573356167e-06, | |
| "loss": 1.2865, | |
| "step": 1027 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 9.148927571160207e-06, | |
| "loss": 0.9491, | |
| "step": 1028 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 9.131777090937181e-06, | |
| "loss": 1.3591, | |
| "step": 1029 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 9.114629183508914e-06, | |
| "loss": 0.9855, | |
| "step": 1030 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 9.09748389968959e-06, | |
| "loss": 1.2375, | |
| "step": 1031 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 9.080341290285635e-06, | |
| "loss": 0.9453, | |
| "step": 1032 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 9.063201406095536e-06, | |
| "loss": 1.4829, | |
| "step": 1033 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 9.046064297909713e-06, | |
| "loss": 1.2815, | |
| "step": 1034 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 9.02893001651036e-06, | |
| "loss": 1.2023, | |
| "step": 1035 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 9.011798612671286e-06, | |
| "loss": 1.5374, | |
| "step": 1036 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.994670137157785e-06, | |
| "loss": 0.9001, | |
| "step": 1037 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.97754464072646e-06, | |
| "loss": 1.348, | |
| "step": 1038 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.960422174125107e-06, | |
| "loss": 1.1074, | |
| "step": 1039 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.943302788092519e-06, | |
| "loss": 1.3378, | |
| "step": 1040 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.92618653335837e-06, | |
| "loss": 0.792, | |
| "step": 1041 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.909073460643066e-06, | |
| "loss": 0.9462, | |
| "step": 1042 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.891963620657564e-06, | |
| "loss": 1.3356, | |
| "step": 1043 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.874857064103263e-06, | |
| "loss": 0.5074, | |
| "step": 1044 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.857753841671816e-06, | |
| "loss": 1.3483, | |
| "step": 1045 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.840654004044996e-06, | |
| "loss": 1.2861, | |
| "step": 1046 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.82355760189456e-06, | |
| "loss": 0.9042, | |
| "step": 1047 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.806464685882069e-06, | |
| "loss": 1.72, | |
| "step": 1048 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.789375306658766e-06, | |
| "loss": 1.2408, | |
| "step": 1049 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.772289514865401e-06, | |
| "loss": 1.3529, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.755207361132109e-06, | |
| "loss": 1.1007, | |
| "step": 1051 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.738128896078232e-06, | |
| "loss": 1.3054, | |
| "step": 1052 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.721054170312177e-06, | |
| "loss": 1.3102, | |
| "step": 1053 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.703983234431293e-06, | |
| "loss": 0.9943, | |
| "step": 1054 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.686916139021671e-06, | |
| "loss": 1.2206, | |
| "step": 1055 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.669852934658042e-06, | |
| "loss": 1.3779, | |
| "step": 1056 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.6527936719036e-06, | |
| "loss": 1.4082, | |
| "step": 1057 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.635738401309848e-06, | |
| "loss": 1.3735, | |
| "step": 1058 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.618687173416483e-06, | |
| "loss": 1.7146, | |
| "step": 1059 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.601640038751195e-06, | |
| "loss": 1.361, | |
| "step": 1060 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.58459704782957e-06, | |
| "loss": 1.5331, | |
| "step": 1061 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.567558251154894e-06, | |
| "loss": 1.7787, | |
| "step": 1062 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.55052369921804e-06, | |
| "loss": 1.3635, | |
| "step": 1063 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.53349344249729e-06, | |
| "loss": 0.8132, | |
| "step": 1064 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.516467531458207e-06, | |
| "loss": 0.9005, | |
| "step": 1065 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.499446016553475e-06, | |
| "loss": 1.2839, | |
| "step": 1066 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.482428948222744e-06, | |
| "loss": 1.1522, | |
| "step": 1067 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.465416376892496e-06, | |
| "loss": 1.2395, | |
| "step": 1068 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.448408352975885e-06, | |
| "loss": 1.5293, | |
| "step": 1069 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.431404926872587e-06, | |
| "loss": 1.1669, | |
| "step": 1070 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.414406148968657e-06, | |
| "loss": 1.0824, | |
| "step": 1071 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.397412069636367e-06, | |
| "loss": 1.2506, | |
| "step": 1072 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.380422739234085e-06, | |
| "loss": 1.4546, | |
| "step": 1073 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.363438208106082e-06, | |
| "loss": 1.5023, | |
| "step": 1074 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.346458526582427e-06, | |
| "loss": 1.2305, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.32948374497881e-06, | |
| "loss": 1.1208, | |
| "step": 1076 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.312513913596399e-06, | |
| "loss": 1.641, | |
| "step": 1077 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.295549082721701e-06, | |
| "loss": 1.0588, | |
| "step": 1078 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.278589302626394e-06, | |
| "loss": 1.3009, | |
| "step": 1079 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.261634623567203e-06, | |
| "loss": 1.1765, | |
| "step": 1080 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.24468509578572e-06, | |
| "loss": 1.2001, | |
| "step": 1081 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.22774076950829e-06, | |
| "loss": 1.6414, | |
| "step": 1082 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 8.210801694945834e-06, | |
| "loss": 0.9662, | |
| "step": 1083 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 8.193867922293708e-06, | |
| "loss": 1.1682, | |
| "step": 1084 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 8.176939501731568e-06, | |
| "loss": 1.1984, | |
| "step": 1085 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 8.1600164834232e-06, | |
| "loss": 0.8819, | |
| "step": 1086 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 8.14309891751639e-06, | |
| "loss": 1.3154, | |
| "step": 1087 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 8.126186854142752e-06, | |
| "loss": 1.1461, | |
| "step": 1088 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 8.10928034341762e-06, | |
| "loss": 1.1509, | |
| "step": 1089 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 8.09237943543985e-06, | |
| "loss": 1.1991, | |
| "step": 1090 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 8.075484180291702e-06, | |
| "loss": 1.4072, | |
| "step": 1091 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 8.058594628038692e-06, | |
| "loss": 1.3363, | |
| "step": 1092 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 8.041710828729428e-06, | |
| "loss": 1.4863, | |
| "step": 1093 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 8.024832832395482e-06, | |
| "loss": 0.7885, | |
| "step": 1094 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 8.007960689051214e-06, | |
| "loss": 1.4431, | |
| "step": 1095 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.991094448693648e-06, | |
| "loss": 1.7296, | |
| "step": 1096 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.974234161302326e-06, | |
| "loss": 1.1038, | |
| "step": 1097 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.957379876839126e-06, | |
| "loss": 1.347, | |
| "step": 1098 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.940531645248165e-06, | |
| "loss": 1.3861, | |
| "step": 1099 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.923689516455597e-06, | |
| "loss": 1.2363, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.906853540369514e-06, | |
| "loss": 1.0636, | |
| "step": 1101 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.890023766879768e-06, | |
| "loss": 1.2531, | |
| "step": 1102 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.873200245857814e-06, | |
| "loss": 1.1078, | |
| "step": 1103 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.856383027156615e-06, | |
| "loss": 0.9789, | |
| "step": 1104 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.839572160610419e-06, | |
| "loss": 1.2516, | |
| "step": 1105 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.822767696034683e-06, | |
| "loss": 1.26, | |
| "step": 1106 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.805969683225872e-06, | |
| "loss": 1.4375, | |
| "step": 1107 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.789178171961344e-06, | |
| "loss": 1.2234, | |
| "step": 1108 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.772393211999187e-06, | |
| "loss": 1.2322, | |
| "step": 1109 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.755614853078073e-06, | |
| "loss": 0.9767, | |
| "step": 1110 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.738843144917119e-06, | |
| "loss": 0.8742, | |
| "step": 1111 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.722078137215727e-06, | |
| "loss": 1.2495, | |
| "step": 1112 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.705319879653454e-06, | |
| "loss": 1.0775, | |
| "step": 1113 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.688568421889842e-06, | |
| "loss": 0.962, | |
| "step": 1114 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.671823813564292e-06, | |
| "loss": 1.3313, | |
| "step": 1115 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.655086104295904e-06, | |
| "loss": 1.1252, | |
| "step": 1116 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.638355343683335e-06, | |
| "loss": 1.3293, | |
| "step": 1117 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.621631581304654e-06, | |
| "loss": 1.2411, | |
| "step": 1118 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.604914866717187e-06, | |
| "loss": 1.2231, | |
| "step": 1119 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.5882052494573795e-06, | |
| "loss": 0.8765, | |
| "step": 1120 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.571502779040646e-06, | |
| "loss": 1.3521, | |
| "step": 1121 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.554807504961216e-06, | |
| "loss": 1.3434, | |
| "step": 1122 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.538119476692004e-06, | |
| "loss": 0.9928, | |
| "step": 1123 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.521438743684443e-06, | |
| "loss": 1.2837, | |
| "step": 1124 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.504765355368362e-06, | |
| "loss": 1.3057, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.4880993611518095e-06, | |
| "loss": 1.0222, | |
| "step": 1126 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.471440810420931e-06, | |
| "loss": 1.532, | |
| "step": 1127 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.454789752539817e-06, | |
| "loss": 0.8085, | |
| "step": 1128 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.4381462368503495e-06, | |
| "loss": 1.4207, | |
| "step": 1129 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.421510312672067e-06, | |
| "loss": 1.3245, | |
| "step": 1130 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.404882029302003e-06, | |
| "loss": 0.9618, | |
| "step": 1131 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.388261436014559e-06, | |
| "loss": 1.3616, | |
| "step": 1132 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.371648582061344e-06, | |
| "loss": 1.4818, | |
| "step": 1133 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.355043516671027e-06, | |
| "loss": 1.2877, | |
| "step": 1134 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.3384462890492105e-06, | |
| "loss": 1.098, | |
| "step": 1135 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.321856948378259e-06, | |
| "loss": 1.1381, | |
| "step": 1136 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.305275543817175e-06, | |
| "loss": 0.9509, | |
| "step": 1137 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.288702124501434e-06, | |
| "loss": 1.2222, | |
| "step": 1138 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.272136739542867e-06, | |
| "loss": 1.2975, | |
| "step": 1139 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.255579438029474e-06, | |
| "loss": 1.3058, | |
| "step": 1140 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.239030269025311e-06, | |
| "loss": 1.0585, | |
| "step": 1141 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.2224892815703465e-06, | |
| "loss": 1.5507, | |
| "step": 1142 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.205956524680282e-06, | |
| "loss": 1.2452, | |
| "step": 1143 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.189432047346456e-06, | |
| "loss": 1.3704, | |
| "step": 1144 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.172915898535647e-06, | |
| "loss": 1.2801, | |
| "step": 1145 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.156408127189964e-06, | |
| "loss": 0.9839, | |
| "step": 1146 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.139908782226703e-06, | |
| "loss": 1.0979, | |
| "step": 1147 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.1234179125381675e-06, | |
| "loss": 1.2588, | |
| "step": 1148 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.106935566991568e-06, | |
| "loss": 0.972, | |
| "step": 1149 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.090461794428835e-06, | |
| "loss": 1.3669, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.073996643666516e-06, | |
| "loss": 1.1478, | |
| "step": 1151 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.0575401634955955e-06, | |
| "loss": 0.9991, | |
| "step": 1152 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.041092402681361e-06, | |
| "loss": 1.0133, | |
| "step": 1153 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.024653409963278e-06, | |
| "loss": 0.9839, | |
| "step": 1154 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.008223234054816e-06, | |
| "loss": 1.3931, | |
| "step": 1155 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 6.991801923643324e-06, | |
| "loss": 1.0984, | |
| "step": 1156 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 6.975389527389877e-06, | |
| "loss": 1.1241, | |
| "step": 1157 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.958986093929141e-06, | |
| "loss": 0.8605, | |
| "step": 1158 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.942591671869212e-06, | |
| "loss": 1.1362, | |
| "step": 1159 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.92620630979149e-06, | |
| "loss": 0.9994, | |
| "step": 1160 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.909830056250527e-06, | |
| "loss": 0.6498, | |
| "step": 1161 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.893462959773881e-06, | |
| "loss": 1.2766, | |
| "step": 1162 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.87710506886198e-06, | |
| "loss": 1.3064, | |
| "step": 1163 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.8607564319879695e-06, | |
| "loss": 1.193, | |
| "step": 1164 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.844417097597569e-06, | |
| "loss": 1.9314, | |
| "step": 1165 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.8280871141089415e-06, | |
| "loss": 1.2041, | |
| "step": 1166 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.811766529912531e-06, | |
| "loss": 1.0328, | |
| "step": 1167 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.795455393370938e-06, | |
| "loss": 1.1625, | |
| "step": 1168 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.779153752818757e-06, | |
| "loss": 1.4506, | |
| "step": 1169 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.762861656562453e-06, | |
| "loss": 1.5669, | |
| "step": 1170 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.746579152880201e-06, | |
| "loss": 1.2576, | |
| "step": 1171 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.73030629002175e-06, | |
| "loss": 1.2913, | |
| "step": 1172 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.7140431162082885e-06, | |
| "loss": 1.2849, | |
| "step": 1173 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.697789679632285e-06, | |
| "loss": 1.0459, | |
| "step": 1174 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.68154602845736e-06, | |
| "loss": 1.4825, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.665312210818133e-06, | |
| "loss": 1.0842, | |
| "step": 1176 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.6490882748200855e-06, | |
| "loss": 1.0912, | |
| "step": 1177 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.6328742685394176e-06, | |
| "loss": 1.1294, | |
| "step": 1178 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.6166702400229e-06, | |
| "loss": 0.8629, | |
| "step": 1179 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.600476237287744e-06, | |
| "loss": 1.1915, | |
| "step": 1180 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.584292308321445e-06, | |
| "loss": 1.4719, | |
| "step": 1181 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.568118501081655e-06, | |
| "loss": 2.2937, | |
| "step": 1182 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.551954863496022e-06, | |
| "loss": 1.6372, | |
| "step": 1183 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.535801443462063e-06, | |
| "loss": 1.1743, | |
| "step": 1184 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.519658288847023e-06, | |
| "loss": 0.9791, | |
| "step": 1185 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.503525447487717e-06, | |
| "loss": 1.2975, | |
| "step": 1186 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.487402967190408e-06, | |
| "loss": 1.418, | |
| "step": 1187 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.471290895730649e-06, | |
| "loss": 0.8595, | |
| "step": 1188 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.455189280853159e-06, | |
| "loss": 1.1379, | |
| "step": 1189 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.4390981702716585e-06, | |
| "loss": 1.0599, | |
| "step": 1190 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.423017611668745e-06, | |
| "loss": 0.8749, | |
| "step": 1191 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.406947652695759e-06, | |
| "loss": 2.6394, | |
| "step": 1192 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.390888340972606e-06, | |
| "loss": 1.4756, | |
| "step": 1193 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.374839724087671e-06, | |
| "loss": 1.4912, | |
| "step": 1194 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.358801849597624e-06, | |
| "loss": 1.3335, | |
| "step": 1195 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.342774765027309e-06, | |
| "loss": 1.1232, | |
| "step": 1196 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.326758517869604e-06, | |
| "loss": 1.3688, | |
| "step": 1197 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.3107531555852585e-06, | |
| "loss": 1.0728, | |
| "step": 1198 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.294758725602783e-06, | |
| "loss": 1.426, | |
| "step": 1199 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.278775275318274e-06, | |
| "loss": 1.2897, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.262802852095311e-06, | |
| "loss": 0.972, | |
| "step": 1201 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.246841503264788e-06, | |
| "loss": 1.038, | |
| "step": 1202 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.230891276124773e-06, | |
| "loss": 1.4712, | |
| "step": 1203 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.2149522179403976e-06, | |
| "loss": 1.2784, | |
| "step": 1204 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.19902437594368e-06, | |
| "loss": 1.1786, | |
| "step": 1205 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.183107797333411e-06, | |
| "loss": 1.3757, | |
| "step": 1206 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.167202529274996e-06, | |
| "loss": 1.0862, | |
| "step": 1207 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.1513086189003355e-06, | |
| "loss": 1.2388, | |
| "step": 1208 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.135426113307664e-06, | |
| "loss": 1.0646, | |
| "step": 1209 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.119555059561423e-06, | |
| "loss": 1.7363, | |
| "step": 1210 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.103695504692122e-06, | |
| "loss": 1.1722, | |
| "step": 1211 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.087847495696191e-06, | |
| "loss": 1.4141, | |
| "step": 1212 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.072011079535853e-06, | |
| "loss": 1.1649, | |
| "step": 1213 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 6.0561863031389715e-06, | |
| "loss": 1.3896, | |
| "step": 1214 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 6.040373213398919e-06, | |
| "loss": 1.5479, | |
| "step": 1215 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 6.024571857174443e-06, | |
| "loss": 1.0277, | |
| "step": 1216 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 6.008782281289513e-06, | |
| "loss": 1.4941, | |
| "step": 1217 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.993004532533196e-06, | |
| "loss": 1.3849, | |
| "step": 1218 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.9772386576595075e-06, | |
| "loss": 1.0861, | |
| "step": 1219 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.961484703387284e-06, | |
| "loss": 1.6108, | |
| "step": 1220 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.94574271640003e-06, | |
| "loss": 1.1156, | |
| "step": 1221 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.93001274334579e-06, | |
| "loss": 1.5676, | |
| "step": 1222 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.9142948308370095e-06, | |
| "loss": 1.3209, | |
| "step": 1223 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.898589025450394e-06, | |
| "loss": 1.6685, | |
| "step": 1224 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.882895373726774e-06, | |
| "loss": 0.8633, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.867213922170958e-06, | |
| "loss": 1.0881, | |
| "step": 1226 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.851544717251616e-06, | |
| "loss": 1.263, | |
| "step": 1227 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.8358878054011125e-06, | |
| "loss": 1.2852, | |
| "step": 1228 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.82024323301539e-06, | |
| "loss": 1.3383, | |
| "step": 1229 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.80461104645383e-06, | |
| "loss": 1.1542, | |
| "step": 1230 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.788991292039103e-06, | |
| "loss": 1.511, | |
| "step": 1231 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.773384016057053e-06, | |
| "loss": 1.7708, | |
| "step": 1232 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.757789264756527e-06, | |
| "loss": 1.6067, | |
| "step": 1233 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.742207084349274e-06, | |
| "loss": 1.0046, | |
| "step": 1234 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.726637521009788e-06, | |
| "loss": 1.4104, | |
| "step": 1235 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.711080620875165e-06, | |
| "loss": 1.1908, | |
| "step": 1236 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.695536430044992e-06, | |
| "loss": 1.2358, | |
| "step": 1237 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.680004994581186e-06, | |
| "loss": 1.3168, | |
| "step": 1238 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.6644863605078685e-06, | |
| "loss": 1.3182, | |
| "step": 1239 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.648980573811221e-06, | |
| "loss": 1.057, | |
| "step": 1240 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.633487680439362e-06, | |
| "loss": 0.9918, | |
| "step": 1241 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.618007726302204e-06, | |
| "loss": 1.5173, | |
| "step": 1242 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.602540757271305e-06, | |
| "loss": 1.1489, | |
| "step": 1243 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.587086819179759e-06, | |
| "loss": 1.8193, | |
| "step": 1244 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.571645957822039e-06, | |
| "loss": 1.0527, | |
| "step": 1245 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.5562182189538684e-06, | |
| "loss": 1.0796, | |
| "step": 1246 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.540803648292082e-06, | |
| "loss": 1.1464, | |
| "step": 1247 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.525402291514496e-06, | |
| "loss": 1.2778, | |
| "step": 1248 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.510014194259776e-06, | |
| "loss": 1.854, | |
| "step": 1249 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.494639402127282e-06, | |
| "loss": 1.009, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.479277960676959e-06, | |
| "loss": 1.1156, | |
| "step": 1251 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.46392991542919e-06, | |
| "loss": 0.9073, | |
| "step": 1252 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.448595311864647e-06, | |
| "loss": 0.8607, | |
| "step": 1253 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.433274195424189e-06, | |
| "loss": 1.132, | |
| "step": 1254 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.417966611508697e-06, | |
| "loss": 1.1418, | |
| "step": 1255 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.40267260547896e-06, | |
| "loss": 1.3899, | |
| "step": 1256 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.3873922226555165e-06, | |
| "loss": 1.1235, | |
| "step": 1257 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.372125508318558e-06, | |
| "loss": 1.0985, | |
| "step": 1258 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.356872507707754e-06, | |
| "loss": 1.3009, | |
| "step": 1259 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.341633266022137e-06, | |
| "loss": 1.1115, | |
| "step": 1260 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.32640782841998e-06, | |
| "loss": 1.4194, | |
| "step": 1261 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.311196240018639e-06, | |
| "loss": 0.6877, | |
| "step": 1262 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.2959985458944404e-06, | |
| "loss": 1.2451, | |
| "step": 1263 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.280814791082529e-06, | |
| "loss": 1.0422, | |
| "step": 1264 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.265645020576746e-06, | |
| "loss": 1.6565, | |
| "step": 1265 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.250489279329501e-06, | |
| "loss": 1.0183, | |
| "step": 1266 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.235347612251611e-06, | |
| "loss": 1.4055, | |
| "step": 1267 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.220220064212218e-06, | |
| "loss": 1.277, | |
| "step": 1268 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.205106680038593e-06, | |
| "loss": 1.1342, | |
| "step": 1269 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.190007504516056e-06, | |
| "loss": 0.9867, | |
| "step": 1270 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 5.174922582387819e-06, | |
| "loss": 1.1816, | |
| "step": 1271 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 5.1598519583548474e-06, | |
| "loss": 1.2834, | |
| "step": 1272 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 5.144795677075753e-06, | |
| "loss": 1.5513, | |
| "step": 1273 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 5.129753783166626e-06, | |
| "loss": 1.4084, | |
| "step": 1274 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 5.114726321200947e-06, | |
| "loss": 1.0637, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 5.099713335709409e-06, | |
| "loss": 0.9497, | |
| "step": 1276 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 5.084714871179817e-06, | |
| "loss": 1.0099, | |
| "step": 1277 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 5.0697309720569495e-06, | |
| "loss": 1.3157, | |
| "step": 1278 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 5.0547616827424125e-06, | |
| "loss": 1.1377, | |
| "step": 1279 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 5.039807047594529e-06, | |
| "loss": 1.2195, | |
| "step": 1280 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 5.024867110928193e-06, | |
| "loss": 1.2329, | |
| "step": 1281 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 5.009941917014746e-06, | |
| "loss": 1.2856, | |
| "step": 1282 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.995031510081835e-06, | |
| "loss": 1.2283, | |
| "step": 1283 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.980135934313296e-06, | |
| "loss": 0.522, | |
| "step": 1284 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.965255233849017e-06, | |
| "loss": 0.988, | |
| "step": 1285 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.950389452784796e-06, | |
| "loss": 1.0524, | |
| "step": 1286 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.935538635172233e-06, | |
| "loss": 1.1737, | |
| "step": 1287 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.920702825018579e-06, | |
| "loss": 1.4213, | |
| "step": 1288 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.905882066286621e-06, | |
| "loss": 1.067, | |
| "step": 1289 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.891076402894532e-06, | |
| "loss": 1.3657, | |
| "step": 1290 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.876285878715764e-06, | |
| "loss": 1.2211, | |
| "step": 1291 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.861510537578908e-06, | |
| "loss": 0.9081, | |
| "step": 1292 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.84675042326755e-06, | |
| "loss": 1.3564, | |
| "step": 1293 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.8320055795201685e-06, | |
| "loss": 1.8823, | |
| "step": 1294 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.8172760500299856e-06, | |
| "loss": 1.147, | |
| "step": 1295 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.802561878444846e-06, | |
| "loss": 1.4531, | |
| "step": 1296 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.787863108367076e-06, | |
| "loss": 1.0188, | |
| "step": 1297 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.77317978335337e-06, | |
| "loss": 1.2588, | |
| "step": 1298 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.758511946914658e-06, | |
| "loss": 1.021, | |
| "step": 1299 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.743859642515959e-06, | |
| "loss": 1.7355, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.729222913576279e-06, | |
| "loss": 1.6702, | |
| "step": 1301 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.714601803468469e-06, | |
| "loss": 0.8987, | |
| "step": 1302 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.699996355519085e-06, | |
| "loss": 0.9365, | |
| "step": 1303 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.685406613008283e-06, | |
| "loss": 1.2063, | |
| "step": 1304 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.670832619169676e-06, | |
| "loss": 1.1644, | |
| "step": 1305 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.6562744171902144e-06, | |
| "loss": 1.072, | |
| "step": 1306 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.641732050210032e-06, | |
| "loss": 1.4893, | |
| "step": 1307 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.627205561322372e-06, | |
| "loss": 0.9828, | |
| "step": 1308 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.6126949935734e-06, | |
| "loss": 1.254, | |
| "step": 1309 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.598200389962104e-06, | |
| "loss": 1.016, | |
| "step": 1310 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.583721793440188e-06, | |
| "loss": 1.3418, | |
| "step": 1311 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.569259246911895e-06, | |
| "loss": 1.0654, | |
| "step": 1312 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.554812793233927e-06, | |
| "loss": 1.3479, | |
| "step": 1313 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.54038247521528e-06, | |
| "loss": 1.3589, | |
| "step": 1314 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.5259683356171605e-06, | |
| "loss": 1.3197, | |
| "step": 1315 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.5115704171528105e-06, | |
| "loss": 1.2043, | |
| "step": 1316 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.497188762487409e-06, | |
| "loss": 1.0808, | |
| "step": 1317 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.4828234142379545e-06, | |
| "loss": 1.1379, | |
| "step": 1318 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.468474414973107e-06, | |
| "loss": 0.9905, | |
| "step": 1319 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.454141807213091e-06, | |
| "loss": 1.2524, | |
| "step": 1320 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.439825633429558e-06, | |
| "loss": 1.5481, | |
| "step": 1321 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.425525936045449e-06, | |
| "loss": 1.4647, | |
| "step": 1322 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.411242757434895e-06, | |
| "loss": 1.2137, | |
| "step": 1323 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.396976139923071e-06, | |
| "loss": 1.0253, | |
| "step": 1324 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.382726125786079e-06, | |
| "loss": 2.4722, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.368492757250814e-06, | |
| "loss": 1.1974, | |
| "step": 1326 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.354276076494854e-06, | |
| "loss": 1.146, | |
| "step": 1327 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.3400761256463216e-06, | |
| "loss": 1.0884, | |
| "step": 1328 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.325892946783762e-06, | |
| "loss": 0.9946, | |
| "step": 1329 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.311726581936023e-06, | |
| "loss": 1.2375, | |
| "step": 1330 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.297577073082129e-06, | |
| "loss": 1.1218, | |
| "step": 1331 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.2834444621511565e-06, | |
| "loss": 1.6365, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.269328791022097e-06, | |
| "loss": 1.0503, | |
| "step": 1333 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.25523010152376e-06, | |
| "loss": 0.8755, | |
| "step": 1334 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.2411484354346266e-06, | |
| "loss": 0.8441, | |
| "step": 1335 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.2270838344827285e-06, | |
| "loss": 1.1179, | |
| "step": 1336 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.213036340345533e-06, | |
| "loss": 1.3887, | |
| "step": 1337 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.199005994649818e-06, | |
| "loss": 1.032, | |
| "step": 1338 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.184992838971544e-06, | |
| "loss": 1.0466, | |
| "step": 1339 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.1709969148357225e-06, | |
| "loss": 1.0794, | |
| "step": 1340 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.1570182637163155e-06, | |
| "loss": 1.1957, | |
| "step": 1341 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.143056927036101e-06, | |
| "loss": 1.0605, | |
| "step": 1342 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.1291129461665345e-06, | |
| "loss": 1.2688, | |
| "step": 1343 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.115186362427657e-06, | |
| "loss": 1.2217, | |
| "step": 1344 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.101277217087948e-06, | |
| "loss": 1.0563, | |
| "step": 1345 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 4.087385551364219e-06, | |
| "loss": 1.2766, | |
| "step": 1346 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 4.073511406421474e-06, | |
| "loss": 1.0941, | |
| "step": 1347 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 4.059654823372806e-06, | |
| "loss": 0.8867, | |
| "step": 1348 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 4.0458158432792704e-06, | |
| "loss": 1.3054, | |
| "step": 1349 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 4.031994507149745e-06, | |
| "loss": 1.7527, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 4.018190855940837e-06, | |
| "loss": 1.2487, | |
| "step": 1351 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 4.004404930556749e-06, | |
| "loss": 1.2922, | |
| "step": 1352 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.990636771849142e-06, | |
| "loss": 1.3895, | |
| "step": 1353 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.976886420617045e-06, | |
| "loss": 1.0813, | |
| "step": 1354 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.963153917606714e-06, | |
| "loss": 1.2223, | |
| "step": 1355 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.949439303511512e-06, | |
| "loss": 0.9825, | |
| "step": 1356 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.935742618971789e-06, | |
| "loss": 1.0371, | |
| "step": 1357 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.922063904574779e-06, | |
| "loss": 1.221, | |
| "step": 1358 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.908403200854452e-06, | |
| "loss": 0.8656, | |
| "step": 1359 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.894760548291403e-06, | |
| "loss": 2.1082, | |
| "step": 1360 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.881135987312758e-06, | |
| "loss": 1.3252, | |
| "step": 1361 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.867529558292008e-06, | |
| "loss": 1.2201, | |
| "step": 1362 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.853941301548931e-06, | |
| "loss": 1.0023, | |
| "step": 1363 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.84037125734944e-06, | |
| "loss": 1.3484, | |
| "step": 1364 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.8268194659055e-06, | |
| "loss": 1.1333, | |
| "step": 1365 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.8132859673749688e-06, | |
| "loss": 1.2039, | |
| "step": 1366 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.7997708018614976e-06, | |
| "loss": 1.4542, | |
| "step": 1367 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.78627400941443e-06, | |
| "loss": 0.9183, | |
| "step": 1368 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.772795630028644e-06, | |
| "loss": 1.2688, | |
| "step": 1369 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.7593357036444644e-06, | |
| "loss": 1.0909, | |
| "step": 1370 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.7458942701475385e-06, | |
| "loss": 1.043, | |
| "step": 1371 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.732471369368701e-06, | |
| "loss": 0.9329, | |
| "step": 1372 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.719067041083879e-06, | |
| "loss": 1.1179, | |
| "step": 1373 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.705681325013961e-06, | |
| "loss": 1.0306, | |
| "step": 1374 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.6923142608246876e-06, | |
| "loss": 1.276, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.6789658881265135e-06, | |
| "loss": 0.7154, | |
| "step": 1376 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.665636246474521e-06, | |
| "loss": 1.7302, | |
| "step": 1377 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.652325375368282e-06, | |
| "loss": 0.8818, | |
| "step": 1378 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.6390333142517388e-06, | |
| "loss": 1.1136, | |
| "step": 1379 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.625760102513103e-06, | |
| "loss": 1.2056, | |
| "step": 1380 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.6125057794847286e-06, | |
| "loss": 1.4336, | |
| "step": 1381 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.5992703844429967e-06, | |
| "loss": 1.2463, | |
| "step": 1382 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.586053956608192e-06, | |
| "loss": 1.2368, | |
| "step": 1383 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.572856535144403e-06, | |
| "loss": 1.4221, | |
| "step": 1384 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.5596781591594e-06, | |
| "loss": 1.2025, | |
| "step": 1385 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.546518867704499e-06, | |
| "loss": 0.8099, | |
| "step": 1386 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.53337869977448e-06, | |
| "loss": 1.1111, | |
| "step": 1387 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.520257694307447e-06, | |
| "loss": 0.8453, | |
| "step": 1388 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.507155890184726e-06, | |
| "loss": 0.6997, | |
| "step": 1389 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.4940733262307302e-06, | |
| "loss": 1.134, | |
| "step": 1390 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.4810100412128743e-06, | |
| "loss": 1.2769, | |
| "step": 1391 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.46796607384144e-06, | |
| "loss": 1.304, | |
| "step": 1392 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.4549414627694545e-06, | |
| "loss": 0.8821, | |
| "step": 1393 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.4419362465925987e-06, | |
| "loss": 1.4559, | |
| "step": 1394 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.4289504638490778e-06, | |
| "loss": 1.0525, | |
| "step": 1395 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.4159841530195127e-06, | |
| "loss": 1.0161, | |
| "step": 1396 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.4030373525268132e-06, | |
| "loss": 1.2169, | |
| "step": 1397 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.3901101007360837e-06, | |
| "loss": 0.8925, | |
| "step": 1398 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.377202435954503e-06, | |
| "loss": 0.9165, | |
| "step": 1399 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.364314396431193e-06, | |
| "loss": 1.1256, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.3514460203571365e-06, | |
| "loss": 1.1941, | |
| "step": 1401 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.3385973458650422e-06, | |
| "loss": 1.1324, | |
| "step": 1402 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.3257684110292297e-06, | |
| "loss": 1.2957, | |
| "step": 1403 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.3129592538655366e-06, | |
| "loss": 1.1028, | |
| "step": 1404 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.300169912331186e-06, | |
| "loss": 1.5562, | |
| "step": 1405 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.287400424324687e-06, | |
| "loss": 1.2429, | |
| "step": 1406 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.2746508276857047e-06, | |
| "loss": 1.495, | |
| "step": 1407 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.261921160194981e-06, | |
| "loss": 1.1694, | |
| "step": 1408 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.249211459574184e-06, | |
| "loss": 0.98, | |
| "step": 1409 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.236521763485815e-06, | |
| "loss": 1.3732, | |
| "step": 1410 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.223852109533112e-06, | |
| "loss": 1.0464, | |
| "step": 1411 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.2112025352599043e-06, | |
| "loss": 1.1538, | |
| "step": 1412 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.198573078150533e-06, | |
| "loss": 1.1567, | |
| "step": 1413 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.185963775629709e-06, | |
| "loss": 0.9827, | |
| "step": 1414 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.1733746650624453e-06, | |
| "loss": 1.0913, | |
| "step": 1415 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.1608057837538976e-06, | |
| "loss": 1.264, | |
| "step": 1416 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.148257168949278e-06, | |
| "loss": 0.9103, | |
| "step": 1417 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.135728857833763e-06, | |
| "loss": 0.8977, | |
| "step": 1418 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.12322088753234e-06, | |
| "loss": 1.3052, | |
| "step": 1419 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.1107332951097334e-06, | |
| "loss": 1.0128, | |
| "step": 1420 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 3.098266117570282e-06, | |
| "loss": 0.8707, | |
| "step": 1421 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 3.0858193918578215e-06, | |
| "loss": 1.3741, | |
| "step": 1422 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 3.073393154855592e-06, | |
| "loss": 1.093, | |
| "step": 1423 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 3.060987443386112e-06, | |
| "loss": 1.4836, | |
| "step": 1424 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 3.0486022942110882e-06, | |
| "loss": 1.4055, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 3.0362377440312783e-06, | |
| "loss": 1.1049, | |
| "step": 1426 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 3.023893829486414e-06, | |
| "loss": 1.0234, | |
| "step": 1427 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 3.0115705871550738e-06, | |
| "loss": 1.2646, | |
| "step": 1428 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.9992680535545714e-06, | |
| "loss": 1.1853, | |
| "step": 1429 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.986986265140864e-06, | |
| "loss": 1.3156, | |
| "step": 1430 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.9747252583084297e-06, | |
| "loss": 1.0259, | |
| "step": 1431 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.962485069390171e-06, | |
| "loss": 2.3806, | |
| "step": 1432 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.950265734657288e-06, | |
| "loss": 1.4436, | |
| "step": 1433 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.9380672903191963e-06, | |
| "loss": 1.1807, | |
| "step": 1434 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.9258897725234058e-06, | |
| "loss": 1.2041, | |
| "step": 1435 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.9137332173554043e-06, | |
| "loss": 1.5625, | |
| "step": 1436 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.9015976608385743e-06, | |
| "loss": 1.4534, | |
| "step": 1437 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.889483138934066e-06, | |
| "loss": 1.3853, | |
| "step": 1438 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.877389687540705e-06, | |
| "loss": 1.2054, | |
| "step": 1439 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.8653173424948654e-06, | |
| "loss": 1.159, | |
| "step": 1440 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.853266139570391e-06, | |
| "loss": 1.1483, | |
| "step": 1441 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.8412361144784706e-06, | |
| "loss": 1.1906, | |
| "step": 1442 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.829227302867532e-06, | |
| "loss": 0.9348, | |
| "step": 1443 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.8172397403231486e-06, | |
| "loss": 1.135, | |
| "step": 1444 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.8052734623679233e-06, | |
| "loss": 1.2592, | |
| "step": 1445 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.793328504461391e-06, | |
| "loss": 1.5969, | |
| "step": 1446 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.7814049019998966e-06, | |
| "loss": 1.1038, | |
| "step": 1447 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.7695026903165177e-06, | |
| "loss": 1.3395, | |
| "step": 1448 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.7576219046809394e-06, | |
| "loss": 1.2384, | |
| "step": 1449 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.745762580299346e-06, | |
| "loss": 1.2466, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.733924752314345e-06, | |
| "loss": 1.4849, | |
| "step": 1451 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.7221084558048296e-06, | |
| "loss": 1.4033, | |
| "step": 1452 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.7103137257858867e-06, | |
| "loss": 1.1096, | |
| "step": 1453 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.6985405972087062e-06, | |
| "loss": 1.2427, | |
| "step": 1454 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.686789104960461e-06, | |
| "loss": 1.1127, | |
| "step": 1455 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.6750592838642144e-06, | |
| "loss": 1.1168, | |
| "step": 1456 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.663351168678796e-06, | |
| "loss": 0.8147, | |
| "step": 1457 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.6516647940987407e-06, | |
| "loss": 1.0462, | |
| "step": 1458 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.6400001947541378e-06, | |
| "loss": 1.1788, | |
| "step": 1459 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.628357405210552e-06, | |
| "loss": 1.3608, | |
| "step": 1460 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.616736459968936e-06, | |
| "loss": 1.285, | |
| "step": 1461 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.605137393465493e-06, | |
| "loss": 1.3605, | |
| "step": 1462 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.593560240071602e-06, | |
| "loss": 1.3765, | |
| "step": 1463 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.5820050340937075e-06, | |
| "loss": 0.8442, | |
| "step": 1464 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.5704718097732173e-06, | |
| "loss": 0.9516, | |
| "step": 1465 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.5589606012863968e-06, | |
| "loss": 1.4961, | |
| "step": 1466 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.5474714427442693e-06, | |
| "loss": 1.6093, | |
| "step": 1467 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.536004368192535e-06, | |
| "loss": 1.3179, | |
| "step": 1468 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.5245594116114335e-06, | |
| "loss": 1.0151, | |
| "step": 1469 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.5131366069156716e-06, | |
| "loss": 1.5992, | |
| "step": 1470 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.5017359879543168e-06, | |
| "loss": 1.6814, | |
| "step": 1471 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.4903575885106823e-06, | |
| "loss": 1.2478, | |
| "step": 1472 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.479001442302249e-06, | |
| "loss": 1.1147, | |
| "step": 1473 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.467667582980553e-06, | |
| "loss": 1.1527, | |
| "step": 1474 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.4563560441310874e-06, | |
| "loss": 1.5266, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.4450668592731976e-06, | |
| "loss": 0.9844, | |
| "step": 1476 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.4338000618599955e-06, | |
| "loss": 0.8414, | |
| "step": 1477 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.422555685278253e-06, | |
| "loss": 0.9423, | |
| "step": 1478 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.411333762848291e-06, | |
| "loss": 1.2606, | |
| "step": 1479 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.4001343278239075e-06, | |
| "loss": 0.9369, | |
| "step": 1480 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.3889574133922532e-06, | |
| "loss": 0.9907, | |
| "step": 1481 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.377803052673753e-06, | |
| "loss": 1.4275, | |
| "step": 1482 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.366671278721985e-06, | |
| "loss": 1.7986, | |
| "step": 1483 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.355562124523608e-06, | |
| "loss": 1.3253, | |
| "step": 1484 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.3444756229982525e-06, | |
| "loss": 1.0879, | |
| "step": 1485 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.3334118069984102e-06, | |
| "loss": 1.1261, | |
| "step": 1486 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.3223707093093618e-06, | |
| "loss": 0.9889, | |
| "step": 1487 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.311352362649061e-06, | |
| "loss": 1.5701, | |
| "step": 1488 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.30035679966805e-06, | |
| "loss": 1.571, | |
| "step": 1489 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.2893840529493437e-06, | |
| "loss": 1.1091, | |
| "step": 1490 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.2784341550083577e-06, | |
| "loss": 0.8779, | |
| "step": 1491 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.2675071382927983e-06, | |
| "loss": 1.1149, | |
| "step": 1492 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.256603035182561e-06, | |
| "loss": 1.2928, | |
| "step": 1493 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.2457218779896496e-06, | |
| "loss": 1.311, | |
| "step": 1494 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.2348636989580673e-06, | |
| "loss": 1.1299, | |
| "step": 1495 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.224028530263733e-06, | |
| "loss": 0.9456, | |
| "step": 1496 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.21321640401437e-06, | |
| "loss": 1.0245, | |
| "step": 1497 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.2024273522494254e-06, | |
| "loss": 0.7163, | |
| "step": 1498 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.1916614069399754e-06, | |
| "loss": 1.4607, | |
| "step": 1499 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.1809185999886094e-06, | |
| "loss": 1.0012, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.170198963229372e-06, | |
| "loss": 1.0116, | |
| "step": 1501 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.1595025284276293e-06, | |
| "loss": 1.1841, | |
| "step": 1502 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.1488293272800065e-06, | |
| "loss": 1.0403, | |
| "step": 1503 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.138179391414269e-06, | |
| "loss": 0.9885, | |
| "step": 1504 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.1275527523892505e-06, | |
| "loss": 1.0875, | |
| "step": 1505 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.116949441694748e-06, | |
| "loss": 0.9882, | |
| "step": 1506 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.10636949075142e-06, | |
| "loss": 1.1741, | |
| "step": 1507 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.09581293091072e-06, | |
| "loss": 1.1354, | |
| "step": 1508 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.085279793454774e-06, | |
| "loss": 2.0255, | |
| "step": 1509 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.074770109596298e-06, | |
| "loss": 1.6372, | |
| "step": 1510 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.064283910478527e-06, | |
| "loss": 1.2935, | |
| "step": 1511 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.0538212271750836e-06, | |
| "loss": 0.9198, | |
| "step": 1512 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.043382090689916e-06, | |
| "loss": 1.103, | |
| "step": 1513 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.0329665319571956e-06, | |
| "loss": 0.9265, | |
| "step": 1514 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.0225745818412277e-06, | |
| "loss": 1.3657, | |
| "step": 1515 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 2.012206271136353e-06, | |
| "loss": 1.2462, | |
| "step": 1516 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 2.0018616305668613e-06, | |
| "loss": 1.2153, | |
| "step": 1517 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.991540690786913e-06, | |
| "loss": 1.1732, | |
| "step": 1518 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.9812434823804204e-06, | |
| "loss": 1.0846, | |
| "step": 1519 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.970970035860984e-06, | |
| "loss": 1.0807, | |
| "step": 1520 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.960720381671789e-06, | |
| "loss": 1.2954, | |
| "step": 1521 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.95049455018551e-06, | |
| "loss": 1.7871, | |
| "step": 1522 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.9402925717042365e-06, | |
| "loss": 1.1863, | |
| "step": 1523 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.9301144764593705e-06, | |
| "loss": 1.2092, | |
| "step": 1524 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.9199602946115468e-06, | |
| "loss": 1.2778, | |
| "step": 1525 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.9098300562505266e-06, | |
| "loss": 0.9995, | |
| "step": 1526 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.8997237913951294e-06, | |
| "loss": 0.9391, | |
| "step": 1527 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.8896415299931326e-06, | |
| "loss": 1.9828, | |
| "step": 1528 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.8795833019211795e-06, | |
| "loss": 0.9631, | |
| "step": 1529 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.8695491369847018e-06, | |
| "loss": 1.293, | |
| "step": 1530 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.8595390649178214e-06, | |
| "loss": 1.0909, | |
| "step": 1531 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.8495531153832713e-06, | |
| "loss": 1.5894, | |
| "step": 1532 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.8395913179722925e-06, | |
| "loss": 1.1472, | |
| "step": 1533 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.8296537022045646e-06, | |
| "loss": 1.076, | |
| "step": 1534 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.819740297528111e-06, | |
| "loss": 0.8835, | |
| "step": 1535 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.8098511333192026e-06, | |
| "loss": 1.136, | |
| "step": 1536 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.7999862388822854e-06, | |
| "loss": 1.222, | |
| "step": 1537 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.7901456434498853e-06, | |
| "loss": 1.0618, | |
| "step": 1538 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.7803293761825258e-06, | |
| "loss": 1.0753, | |
| "step": 1539 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.7705374661686326e-06, | |
| "loss": 0.9271, | |
| "step": 1540 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.7607699424244583e-06, | |
| "loss": 1.2874, | |
| "step": 1541 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.751026833893995e-06, | |
| "loss": 1.4094, | |
| "step": 1542 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.7413081694488765e-06, | |
| "loss": 1.4553, | |
| "step": 1543 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.7316139778883102e-06, | |
| "loss": 1.0115, | |
| "step": 1544 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.7219442879389792e-06, | |
| "loss": 0.9066, | |
| "step": 1545 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.712299128254965e-06, | |
| "loss": 1.1272, | |
| "step": 1546 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.7026785274176516e-06, | |
| "loss": 1.0156, | |
| "step": 1547 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.6930825139356543e-06, | |
| "loss": 0.8509, | |
| "step": 1548 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.6835111162447294e-06, | |
| "loss": 1.1371, | |
| "step": 1549 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.6739643627076817e-06, | |
| "loss": 1.2996, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.6644422816143024e-06, | |
| "loss": 1.8409, | |
| "step": 1551 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.6549449011812546e-06, | |
| "loss": 1.0548, | |
| "step": 1552 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.6454722495520202e-06, | |
| "loss": 0.9034, | |
| "step": 1553 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.6360243547967913e-06, | |
| "loss": 0.957, | |
| "step": 1554 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.6266012449124068e-06, | |
| "loss": 1.1157, | |
| "step": 1555 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.6172029478222595e-06, | |
| "loss": 0.8828, | |
| "step": 1556 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.6078294913762048e-06, | |
| "loss": 1.1169, | |
| "step": 1557 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.5984809033505055e-06, | |
| "loss": 1.1997, | |
| "step": 1558 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.5891572114477205e-06, | |
| "loss": 1.0707, | |
| "step": 1559 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.579858443296629e-06, | |
| "loss": 1.0626, | |
| "step": 1560 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.5705846264521728e-06, | |
| "loss": 0.9697, | |
| "step": 1561 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.5613357883953361e-06, | |
| "loss": 1.0522, | |
| "step": 1562 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.5521119565330955e-06, | |
| "loss": 0.8218, | |
| "step": 1563 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.5429131581983236e-06, | |
| "loss": 0.9563, | |
| "step": 1564 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.5337394206497135e-06, | |
| "loss": 0.7686, | |
| "step": 1565 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.5245907710716912e-06, | |
| "loss": 1.0132, | |
| "step": 1566 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.5154672365743384e-06, | |
| "loss": 1.162, | |
| "step": 1567 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.5063688441933278e-06, | |
| "loss": 1.1438, | |
| "step": 1568 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.4972956208898105e-06, | |
| "loss": 0.9092, | |
| "step": 1569 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.488247593550366e-06, | |
| "loss": 1.2915, | |
| "step": 1570 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.4792247889869072e-06, | |
| "loss": 0.8971, | |
| "step": 1571 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.4702272339366064e-06, | |
| "loss": 1.3024, | |
| "step": 1572 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.4612549550618093e-06, | |
| "loss": 1.0593, | |
| "step": 1573 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.452307978949966e-06, | |
| "loss": 1.4166, | |
| "step": 1574 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.4433863321135488e-06, | |
| "loss": 1.7939, | |
| "step": 1575 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.4344900409899643e-06, | |
| "loss": 0.9315, | |
| "step": 1576 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.4256191319414903e-06, | |
| "loss": 1.1056, | |
| "step": 1577 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.4167736312551883e-06, | |
| "loss": 1.6176, | |
| "step": 1578 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.4079535651428234e-06, | |
| "loss": 1.2653, | |
| "step": 1579 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.3991589597407961e-06, | |
| "loss": 1.301, | |
| "step": 1580 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.3903898411100568e-06, | |
| "loss": 1.4731, | |
| "step": 1581 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.3816462352360337e-06, | |
| "loss": 1.22, | |
| "step": 1582 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.3729281680285477e-06, | |
| "loss": 0.543, | |
| "step": 1583 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.364235665321748e-06, | |
| "loss": 1.0205, | |
| "step": 1584 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.355568752874027e-06, | |
| "loss": 1.3539, | |
| "step": 1585 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.3469274563679401e-06, | |
| "loss": 0.9204, | |
| "step": 1586 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.3383118014101415e-06, | |
| "loss": 0.9436, | |
| "step": 1587 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.3297218135313029e-06, | |
| "loss": 1.5491, | |
| "step": 1588 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.3211575181860336e-06, | |
| "loss": 0.8912, | |
| "step": 1589 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.3126189407528067e-06, | |
| "loss": 1.0469, | |
| "step": 1590 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.30410610653389e-06, | |
| "loss": 0.8495, | |
| "step": 1591 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.2956190407552694e-06, | |
| "loss": 1.0691, | |
| "step": 1592 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.2871577685665592e-06, | |
| "loss": 1.1404, | |
| "step": 1593 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.2787223150409546e-06, | |
| "loss": 0.9884, | |
| "step": 1594 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.2703127051751319e-06, | |
| "loss": 1.6191, | |
| "step": 1595 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.261928963889194e-06, | |
| "loss": 1.0682, | |
| "step": 1596 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.2535711160265773e-06, | |
| "loss": 2.0875, | |
| "step": 1597 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.2452391863539982e-06, | |
| "loss": 1.4458, | |
| "step": 1598 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.2369331995613664e-06, | |
| "loss": 1.2345, | |
| "step": 1599 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.2286531802617086e-06, | |
| "loss": 1.4313, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.2203991529911197e-06, | |
| "loss": 1.0363, | |
| "step": 1601 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.2121711422086523e-06, | |
| "loss": 1.4131, | |
| "step": 1602 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.2039691722962788e-06, | |
| "loss": 0.8237, | |
| "step": 1603 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.1957932675588013e-06, | |
| "loss": 0.8777, | |
| "step": 1604 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.1876434522237778e-06, | |
| "loss": 1.3579, | |
| "step": 1605 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.1795197504414657e-06, | |
| "loss": 1.3248, | |
| "step": 1606 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.1714221862847275e-06, | |
| "loss": 0.9655, | |
| "step": 1607 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.16335078374899e-06, | |
| "loss": 1.6265, | |
| "step": 1608 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.1553055667521395e-06, | |
| "loss": 1.1726, | |
| "step": 1609 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.1472865591344706e-06, | |
| "loss": 1.4851, | |
| "step": 1610 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.1392937846586216e-06, | |
| "loss": 0.9188, | |
| "step": 1611 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.131327267009482e-06, | |
| "loss": 1.0537, | |
| "step": 1612 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.1233870297941408e-06, | |
| "loss": 1.3816, | |
| "step": 1613 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.1154730965418114e-06, | |
| "loss": 1.3112, | |
| "step": 1614 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.1075854907037608e-06, | |
| "loss": 1.3673, | |
| "step": 1615 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.0997242356532335e-06, | |
| "loss": 1.0923, | |
| "step": 1616 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.091889354685398e-06, | |
| "loss": 1.2228, | |
| "step": 1617 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.0840808710172656e-06, | |
| "loss": 0.854, | |
| "step": 1618 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.0762988077876214e-06, | |
| "loss": 1.1024, | |
| "step": 1619 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.068543188056964e-06, | |
| "loss": 0.978, | |
| "step": 1620 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.0608140348074292e-06, | |
| "loss": 1.7422, | |
| "step": 1621 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.0531113709427298e-06, | |
| "loss": 1.1754, | |
| "step": 1622 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.0454352192880735e-06, | |
| "loss": 1.5741, | |
| "step": 1623 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.0377856025901134e-06, | |
| "loss": 1.0587, | |
| "step": 1624 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.0301625435168706e-06, | |
| "loss": 0.9442, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.022566064657663e-06, | |
| "loss": 1.032, | |
| "step": 1626 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.0149961885230485e-06, | |
| "loss": 1.3259, | |
| "step": 1627 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.007452937544754e-06, | |
| "loss": 0.9509, | |
| "step": 1628 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 9.99936334075603e-07, | |
| "loss": 1.3113, | |
| "step": 1629 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 9.924464003894607e-07, | |
| "loss": 1.2169, | |
| "step": 1630 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 9.849831586811597e-07, | |
| "loss": 1.6875, | |
| "step": 1631 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 9.775466310664383e-07, | |
| "loss": 0.9775, | |
| "step": 1632 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 9.701368395818667e-07, | |
| "loss": 1.8611, | |
| "step": 1633 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 9.627538061847964e-07, | |
| "loss": 1.2373, | |
| "step": 1634 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 9.553975527532844e-07, | |
| "loss": 0.8123, | |
| "step": 1635 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 9.48068101086026e-07, | |
| "loss": 1.1172, | |
| "step": 1636 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 9.40765472902303e-07, | |
| "loss": 0.8605, | |
| "step": 1637 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 9.334896898419066e-07, | |
| "loss": 1.3319, | |
| "step": 1638 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 9.262407734650803e-07, | |
| "loss": 1.1722, | |
| "step": 1639 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 9.190187452524491e-07, | |
| "loss": 1.0609, | |
| "step": 1640 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 9.118236266049707e-07, | |
| "loss": 1.2035, | |
| "step": 1641 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 9.046554388438533e-07, | |
| "loss": 1.4054, | |
| "step": 1642 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 8.975142032105e-07, | |
| "loss": 1.3263, | |
| "step": 1643 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 8.903999408664532e-07, | |
| "loss": 1.3567, | |
| "step": 1644 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 8.833126728933228e-07, | |
| "loss": 0.9875, | |
| "step": 1645 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 8.762524202927281e-07, | |
| "loss": 1.0922, | |
| "step": 1646 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 8.692192039862302e-07, | |
| "loss": 1.0594, | |
| "step": 1647 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 8.622130448152788e-07, | |
| "loss": 1.088, | |
| "step": 1648 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 8.552339635411422e-07, | |
| "loss": 1.2129, | |
| "step": 1649 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 8.482819808448472e-07, | |
| "loss": 1.4199, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 8.413571173271295e-07, | |
| "loss": 1.3149, | |
| "step": 1651 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 8.344593935083489e-07, | |
| "loss": 1.3691, | |
| "step": 1652 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 8.275888298284517e-07, | |
| "loss": 1.3091, | |
| "step": 1653 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 8.207454466469011e-07, | |
| "loss": 1.0507, | |
| "step": 1654 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 8.139292642426111e-07, | |
| "loss": 1.2271, | |
| "step": 1655 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 8.071403028138969e-07, | |
| "loss": 1.3777, | |
| "step": 1656 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 8.003785824784038e-07, | |
| "loss": 1.2208, | |
| "step": 1657 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.936441232730662e-07, | |
| "loss": 1.1237, | |
| "step": 1658 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.86936945154022e-07, | |
| "loss": 0.8015, | |
| "step": 1659 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.802570679965727e-07, | |
| "loss": 1.1066, | |
| "step": 1660 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.736045115951252e-07, | |
| "loss": 1.3555, | |
| "step": 1661 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.669792956631172e-07, | |
| "loss": 1.5464, | |
| "step": 1662 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.603814398329745e-07, | |
| "loss": 2.7554, | |
| "step": 1663 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.538109636560454e-07, | |
| "loss": 0.8209, | |
| "step": 1664 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.472678866025484e-07, | |
| "loss": 0.9681, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 7.40752228061502e-07, | |
| "loss": 1.3126, | |
| "step": 1666 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 7.342640073406826e-07, | |
| "loss": 1.4536, | |
| "step": 1667 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 7.278032436665606e-07, | |
| "loss": 1.3575, | |
| "step": 1668 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 7.213699561842369e-07, | |
| "loss": 0.975, | |
| "step": 1669 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 7.14964163957399e-07, | |
| "loss": 1.114, | |
| "step": 1670 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 7.08585885968257e-07, | |
| "loss": 1.0768, | |
| "step": 1671 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 7.022351411174866e-07, | |
| "loss": 0.9077, | |
| "step": 1672 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.959119482241738e-07, | |
| "loss": 1.2089, | |
| "step": 1673 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.896163260257627e-07, | |
| "loss": 1.0822, | |
| "step": 1674 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.833482931779989e-07, | |
| "loss": 1.5574, | |
| "step": 1675 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.77107868254867e-07, | |
| "loss": 1.1277, | |
| "step": 1676 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.708950697485461e-07, | |
| "loss": 1.4772, | |
| "step": 1677 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.647099160693515e-07, | |
| "loss": 1.6194, | |
| "step": 1678 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.585524255456721e-07, | |
| "loss": 1.658, | |
| "step": 1679 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.524226164239311e-07, | |
| "loss": 1.4142, | |
| "step": 1680 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.463205068685174e-07, | |
| "loss": 1.0142, | |
| "step": 1681 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.402461149617434e-07, | |
| "loss": 1.1327, | |
| "step": 1682 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.341994587037803e-07, | |
| "loss": 1.2181, | |
| "step": 1683 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.281805560126153e-07, | |
| "loss": 1.4877, | |
| "step": 1684 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 6.221894247239923e-07, | |
| "loss": 1.1664, | |
| "step": 1685 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 6.162260825913591e-07, | |
| "loss": 1.0341, | |
| "step": 1686 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 6.102905472858178e-07, | |
| "loss": 1.3522, | |
| "step": 1687 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 6.043828363960703e-07, | |
| "loss": 1.1774, | |
| "step": 1688 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.985029674283694e-07, | |
| "loss": 1.3983, | |
| "step": 1689 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.926509578064565e-07, | |
| "loss": 1.0116, | |
| "step": 1690 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.868268248715292e-07, | |
| "loss": 1.2556, | |
| "step": 1691 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.810305858821697e-07, | |
| "loss": 1.2278, | |
| "step": 1692 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.75262258014303e-07, | |
| "loss": 1.166, | |
| "step": 1693 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.695218583611494e-07, | |
| "loss": 1.2881, | |
| "step": 1694 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.638094039331665e-07, | |
| "loss": 1.1519, | |
| "step": 1695 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.581249116580068e-07, | |
| "loss": 2.3774, | |
| "step": 1696 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.524683983804546e-07, | |
| "loss": 1.3981, | |
| "step": 1697 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.468398808623932e-07, | |
| "loss": 1.1517, | |
| "step": 1698 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.412393757827405e-07, | |
| "loss": 1.4005, | |
| "step": 1699 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.356668997374071e-07, | |
| "loss": 1.3738, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.301224692392482e-07, | |
| "loss": 1.3712, | |
| "step": 1701 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.24606100718007e-07, | |
| "loss": 0.9336, | |
| "step": 1702 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.191178105202744e-07, | |
| "loss": 1.4121, | |
| "step": 1703 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 5.136576149094386e-07, | |
| "loss": 1.2474, | |
| "step": 1704 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 5.082255300656303e-07, | |
| "loss": 0.9681, | |
| "step": 1705 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 5.028215720856821e-07, | |
| "loss": 1.7571, | |
| "step": 1706 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.974457569830826e-07, | |
| "loss": 1.1774, | |
| "step": 1707 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.920981006879199e-07, | |
| "loss": 1.2262, | |
| "step": 1708 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.867786190468393e-07, | |
| "loss": 1.0053, | |
| "step": 1709 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.814873278230014e-07, | |
| "loss": 1.1084, | |
| "step": 1710 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.762242426960262e-07, | |
| "loss": 1.0565, | |
| "step": 1711 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.709893792619513e-07, | |
| "loss": 1.0879, | |
| "step": 1712 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.6578275303318597e-07, | |
| "loss": 1.3372, | |
| "step": 1713 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.6060437943846516e-07, | |
| "loss": 1.0554, | |
| "step": 1714 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.5545427382280405e-07, | |
| "loss": 0.9502, | |
| "step": 1715 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.503324514474483e-07, | |
| "loss": 1.0466, | |
| "step": 1716 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.4523892748983387e-07, | |
| "loss": 1.4745, | |
| "step": 1717 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.401737170435416e-07, | |
| "loss": 1.0989, | |
| "step": 1718 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.351368351182494e-07, | |
| "loss": 1.01, | |
| "step": 1719 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.3012829663969025e-07, | |
| "loss": 1.46, | |
| "step": 1720 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.2514811644960743e-07, | |
| "loss": 1.2498, | |
| "step": 1721 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.2019630930571264e-07, | |
| "loss": 1.0365, | |
| "step": 1722 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 4.1527288988163386e-07, | |
| "loss": 1.0742, | |
| "step": 1723 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 4.10377872766885e-07, | |
| "loss": 1.2878, | |
| "step": 1724 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 4.055112724668131e-07, | |
| "loss": 0.9735, | |
| "step": 1725 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 4.006731034025546e-07, | |
| "loss": 0.9079, | |
| "step": 1726 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.9586337991100233e-07, | |
| "loss": 1.125, | |
| "step": 1727 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.910821162447531e-07, | |
| "loss": 1.3057, | |
| "step": 1728 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.863293265720669e-07, | |
| "loss": 1.0385, | |
| "step": 1729 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.8160502497683105e-07, | |
| "loss": 1.2158, | |
| "step": 1730 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.769092254585138e-07, | |
| "loss": 0.9371, | |
| "step": 1731 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.722419419321233e-07, | |
| "loss": 1.0679, | |
| "step": 1732 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.676031882281628e-07, | |
| "loss": 1.1014, | |
| "step": 1733 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.629929780925989e-07, | |
| "loss": 1.0851, | |
| "step": 1734 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.584113251868149e-07, | |
| "loss": 1.0845, | |
| "step": 1735 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.538582430875659e-07, | |
| "loss": 1.1143, | |
| "step": 1736 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.493337452869472e-07, | |
| "loss": 1.231, | |
| "step": 1737 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.4483784519235173e-07, | |
| "loss": 0.7744, | |
| "step": 1738 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.40370556126427e-07, | |
| "loss": 1.2645, | |
| "step": 1739 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.3593189132703486e-07, | |
| "loss": 1.0438, | |
| "step": 1740 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 3.3152186394722506e-07, | |
| "loss": 1.4044, | |
| "step": 1741 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 3.271404870551753e-07, | |
| "loss": 1.265, | |
| "step": 1742 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 3.227877736341678e-07, | |
| "loss": 1.6082, | |
| "step": 1743 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 3.1846373658255046e-07, | |
| "loss": 0.9684, | |
| "step": 1744 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 3.1416838871368925e-07, | |
| "loss": 1.4989, | |
| "step": 1745 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 3.099017427559392e-07, | |
| "loss": 1.3258, | |
| "step": 1746 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 3.056638113526e-07, | |
| "loss": 1.1499, | |
| "step": 1747 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 3.014546070618862e-07, | |
| "loss": 0.9586, | |
| "step": 1748 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.972741423568837e-07, | |
| "loss": 1.1364, | |
| "step": 1749 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.9312242962551196e-07, | |
| "loss": 1.7053, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.889994811704966e-07, | |
| "loss": 0.3802, | |
| "step": 1751 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.8490530920931794e-07, | |
| "loss": 1.4187, | |
| "step": 1752 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.8083992587419116e-07, | |
| "loss": 0.7334, | |
| "step": 1753 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.7680334321201763e-07, | |
| "loss": 1.1738, | |
| "step": 1754 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.727955731843546e-07, | |
| "loss": 1.106, | |
| "step": 1755 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.688166276673809e-07, | |
| "loss": 1.3446, | |
| "step": 1756 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.648665184518584e-07, | |
| "loss": 0.7734, | |
| "step": 1757 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.609452572431015e-07, | |
| "loss": 1.1541, | |
| "step": 1758 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.5705285566093417e-07, | |
| "loss": 1.0292, | |
| "step": 1759 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.5318932523966664e-07, | |
| "loss": 1.1058, | |
| "step": 1760 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.493546774280531e-07, | |
| "loss": 0.7595, | |
| "step": 1761 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.455489235892594e-07, | |
| "loss": 1.2739, | |
| "step": 1762 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.4177207500083124e-07, | |
| "loss": 1.0774, | |
| "step": 1763 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.3802414285466035e-07, | |
| "loss": 1.4163, | |
| "step": 1764 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.3430513825695057e-07, | |
| "loss": 1.24, | |
| "step": 1765 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.3061507222818303e-07, | |
| "loss": 1.1718, | |
| "step": 1766 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.2695395570308865e-07, | |
| "loss": 1.3743, | |
| "step": 1767 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.233217995306103e-07, | |
| "loss": 1.2489, | |
| "step": 1768 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.197186144738739e-07, | |
| "loss": 1.28, | |
| "step": 1769 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.1614441121015405e-07, | |
| "loss": 1.2249, | |
| "step": 1770 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.1259920033084746e-07, | |
| "loss": 1.509, | |
| "step": 1771 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.0908299234143614e-07, | |
| "loss": 0.7593, | |
| "step": 1772 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.0559579766145532e-07, | |
| "loss": 1.2822, | |
| "step": 1773 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.0213762662447012e-07, | |
| "loss": 0.9626, | |
| "step": 1774 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.9870848947803777e-07, | |
| "loss": 1.1726, | |
| "step": 1775 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.9530839638367993e-07, | |
| "loss": 1.0725, | |
| "step": 1776 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.919373574168515e-07, | |
| "loss": 0.9667, | |
| "step": 1777 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.8859538256691511e-07, | |
| "loss": 0.8696, | |
| "step": 1778 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.8528248173710462e-07, | |
| "loss": 1.2599, | |
| "step": 1779 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.8199866474449825e-07, | |
| "loss": 1.087, | |
| "step": 1780 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.7874394131999428e-07, | |
| "loss": 1.0607, | |
| "step": 1781 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.7551832110827783e-07, | |
| "loss": 0.9539, | |
| "step": 1782 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.7232181366778734e-07, | |
| "loss": 2.5266, | |
| "step": 1783 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.6915442847069917e-07, | |
| "loss": 1.2703, | |
| "step": 1784 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.660161749028877e-07, | |
| "loss": 1.4539, | |
| "step": 1785 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.6290706226390286e-07, | |
| "loss": 1.0526, | |
| "step": 1786 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.5982709976694043e-07, | |
| "loss": 1.3832, | |
| "step": 1787 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.5677629653881975e-07, | |
| "loss": 1.356, | |
| "step": 1788 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.5375466161994813e-07, | |
| "loss": 0.9634, | |
| "step": 1789 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.5076220396430085e-07, | |
| "loss": 1.1569, | |
| "step": 1790 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.4779893243939358e-07, | |
| "loss": 1.0776, | |
| "step": 1791 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.4486485582625443e-07, | |
| "loss": 1.0148, | |
| "step": 1792 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.4195998281939516e-07, | |
| "loss": 1.0742, | |
| "step": 1793 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.390843220267968e-07, | |
| "loss": 1.191, | |
| "step": 1794 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.362378819698662e-07, | |
| "loss": 1.1591, | |
| "step": 1795 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.334206710834296e-07, | |
| "loss": 1.1677, | |
| "step": 1796 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.3063269771569354e-07, | |
| "loss": 1.3925, | |
| "step": 1797 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.27873970128225e-07, | |
| "loss": 1.6865, | |
| "step": 1798 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.2514449649593252e-07, | |
| "loss": 1.4231, | |
| "step": 1799 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.2244428490702954e-07, | |
| "loss": 0.8359, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.1977334336302437e-07, | |
| "loss": 1.4092, | |
| "step": 1801 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.1713167977868589e-07, | |
| "loss": 0.886, | |
| "step": 1802 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.1451930198202344e-07, | |
| "loss": 1.4358, | |
| "step": 1803 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.11936217714268e-07, | |
| "loss": 1.0712, | |
| "step": 1804 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.0938243462984e-07, | |
| "loss": 1.0885, | |
| "step": 1805 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.068579602963371e-07, | |
| "loss": 1.0312, | |
| "step": 1806 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.0436280219450312e-07, | |
| "loss": 0.9579, | |
| "step": 1807 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.0189696771821133e-07, | |
| "loss": 1.1969, | |
| "step": 1808 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 9.94604641744401e-08, | |
| "loss": 1.36, | |
| "step": 1809 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 9.705329878325176e-08, | |
| "loss": 1.6885, | |
| "step": 1810 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 9.467547867777261e-08, | |
| "loss": 1.2024, | |
| "step": 1811 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 9.232701090416519e-08, | |
| "loss": 0.908, | |
| "step": 1812 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 9.000790242162049e-08, | |
| "loss": 0.9421, | |
| "step": 1813 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 8.77181601023236e-08, | |
| "loss": 1.6077, | |
| "step": 1814 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 8.545779073144134e-08, | |
| "loss": 1.0955, | |
| "step": 1815 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 8.322680100710023e-08, | |
| "loss": 1.1652, | |
| "step": 1816 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 8.10251975403653e-08, | |
| "loss": 1.1426, | |
| "step": 1817 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 7.885298685522235e-08, | |
| "loss": 1.3347, | |
| "step": 1818 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 7.671017538855685e-08, | |
| "loss": 1.2991, | |
| "step": 1819 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 7.459676949013949e-08, | |
| "loss": 1.2213, | |
| "step": 1820 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 7.251277542259849e-08, | |
| "loss": 1.5123, | |
| "step": 1821 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 7.045819936141063e-08, | |
| "loss": 1.4468, | |
| "step": 1822 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 6.843304739487577e-08, | |
| "loss": 1.0222, | |
| "step": 1823 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 6.643732552410464e-08, | |
| "loss": 1.1989, | |
| "step": 1824 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 6.447103966299551e-08, | |
| "loss": 1.1437, | |
| "step": 1825 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 6.253419563821971e-08, | |
| "loss": 0.8972, | |
| "step": 1826 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 6.0626799189204e-08, | |
| "loss": 1.4302, | |
| "step": 1827 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 5.874885596811597e-08, | |
| "loss": 1.2257, | |
| "step": 1828 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 5.690037153984085e-08, | |
| "loss": 1.2238, | |
| "step": 1829 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 5.508135138197146e-08, | |
| "loss": 0.8918, | |
| "step": 1830 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 5.3291800884789356e-08, | |
| "loss": 0.9634, | |
| "step": 1831 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 5.1531725351248175e-08, | |
| "loss": 1.1287, | |
| "step": 1832 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 4.980112999695808e-08, | |
| "loss": 1.7037, | |
| "step": 1833 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 4.810001995017355e-08, | |
| "loss": 1.2876, | |
| "step": 1834 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 4.642840025177342e-08, | |
| "loss": 1.1527, | |
| "step": 1835 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 4.478627585524753e-08, | |
| "loss": 1.0923, | |
| "step": 1836 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 4.317365162668563e-08, | |
| "loss": 0.723, | |
| "step": 1837 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 4.1590532344757406e-08, | |
| "loss": 1.2411, | |
| "step": 1838 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 4.003692270070247e-08, | |
| "loss": 1.1282, | |
| "step": 1839 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 3.8512827298312626e-08, | |
| "loss": 1.0688, | |
| "step": 1840 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 3.701825065392184e-08, | |
| "loss": 1.2012, | |
| "step": 1841 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 3.555319719639405e-08, | |
| "loss": 1.0967, | |
| "step": 1842 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 3.411767126710097e-08, | |
| "loss": 1.1287, | |
| "step": 1843 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 3.271167711992318e-08, | |
| "loss": 0.8544, | |
| "step": 1844 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 3.1335218921224595e-08, | |
| "loss": 1.2084, | |
| "step": 1845 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.998830074984915e-08, | |
| "loss": 1.2543, | |
| "step": 1846 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.867092659710302e-08, | |
| "loss": 1.6738, | |
| "step": 1847 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.738310036674463e-08, | |
| "loss": 0.8047, | |
| "step": 1848 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.6124825874975778e-08, | |
| "loss": 1.4177, | |
| "step": 1849 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.4896106850424984e-08, | |
| "loss": 1.0066, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.369694693414304e-08, | |
| "loss": 1.1952, | |
| "step": 1851 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.252734967958414e-08, | |
| "loss": 1.269, | |
| "step": 1852 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.1387318552603676e-08, | |
| "loss": 1.7168, | |
| "step": 1853 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 2.0276856931442656e-08, | |
| "loss": 1.3433, | |
| "step": 1854 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.9195968106716644e-08, | |
| "loss": 1.4639, | |
| "step": 1855 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.8144655281413515e-08, | |
| "loss": 1.1709, | |
| "step": 1856 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.7122921570874586e-08, | |
| "loss": 1.5776, | |
| "step": 1857 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.6130770002793505e-08, | |
| "loss": 0.9138, | |
| "step": 1858 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.51682035171985e-08, | |
| "loss": 1.1935, | |
| "step": 1859 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.4235224966452354e-08, | |
| "loss": 1.416, | |
| "step": 1860 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.3331837115241331e-08, | |
| "loss": 1.4404, | |
| "step": 1861 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.2458042640561829e-08, | |
| "loss": 1.0931, | |
| "step": 1862 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.1613844131719287e-08, | |
| "loss": 1.0295, | |
| "step": 1863 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.0799244090317073e-08, | |
| "loss": 1.469, | |
| "step": 1864 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.0014244930252048e-08, | |
| "loss": 0.9879, | |
| "step": 1865 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 9.25884897770013e-09, | |
| "loss": 1.2225, | |
| "step": 1866 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 8.533058471118516e-09, | |
| "loss": 0.9188, | |
| "step": 1867 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 7.83687556123458e-09, | |
| "loss": 1.4021, | |
| "step": 1868 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 7.170302311036992e-09, | |
| "loss": 1.3613, | |
| "step": 1869 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 6.533340695776824e-09, | |
| "loss": 0.9288, | |
| "step": 1870 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 5.925992602952013e-09, | |
| "loss": 1.4185, | |
| "step": 1871 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 5.348259832312908e-09, | |
| "loss": 1.0289, | |
| "step": 1872 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 4.800144095846726e-09, | |
| "loss": 1.0648, | |
| "step": 1873 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 4.281647017777557e-09, | |
| "loss": 1.1176, | |
| "step": 1874 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 3.792770134564139e-09, | |
| "loss": 0.9559, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 3.333514894887646e-09, | |
| "loss": 0.812, | |
| "step": 1876 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 2.903882659655022e-09, | |
| "loss": 1.2731, | |
| "step": 1877 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 2.503874701988984e-09, | |
| "loss": 1.1798, | |
| "step": 1878 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 2.1334922072291376e-09, | |
| "loss": 1.3124, | |
| "step": 1879 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.7927362729264208e-09, | |
| "loss": 1.11, | |
| "step": 1880 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.4816079088375567e-09, | |
| "loss": 1.1592, | |
| "step": 1881 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.200108036926162e-09, | |
| "loss": 0.9718, | |
| "step": 1882 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "step": 1882, | |
| "total_flos": 4.523005348535075e+18, | |
| "train_loss": 1.3381117355557481, | |
| "train_runtime": 32683.789, | |
| "train_samples_per_second": 0.23, | |
| "train_steps_per_second": 0.058 | |
| } | |
| ], | |
| "max_steps": 1882, | |
| "num_train_epochs": 1, | |
| "total_flos": 4.523005348535075e+18, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |