conan1024hao/cjkbert-base
113
1{2 "best_metric": null,3 "best_model_checkpoint": null,4 "epoch": 19.999995595043544,5 "global_step": 2270160,6 "is_hyper_param_search": false,7 "is_local_process_zero": true,8 "is_world_process_zero": true,9 "log_history": [10 {11 "epoch": 1.0,12 "learning_rate": 4.666798815942489e-05,13 "loss": 2.7208,14 "step": 11350815 },16 {17 "epoch": 2.0,18 "learning_rate": 4.3335917585838296e-05,19 "loss": 2.1526,20 "step": 22701621 },22 {23 "epoch": 3.0,24 "learning_rate": 4.000384701225171e-05,25 "loss": 2.0051,26 "step": 34052427 },28 {29 "epoch": 4.0,30 "learning_rate": 3.667171770565364e-05,31 "loss": 1.9189,32 "step": 45403233 },34 {35 "epoch": 5.0,36 "learning_rate": 3.3339647132067057e-05,37 "loss": 1.8591,38 "step": 56754039 },40 {41 "epoch": 6.0,42 "learning_rate": 3.000757655848046e-05,43 "loss": 1.8134,44 "step": 68104845 },46 {47 "epoch": 7.0,48 "learning_rate": 2.6675505984893874e-05,49 "loss": 1.7768,50 "step": 79455651 },52 {53 "epoch": 8.0,54 "learning_rate": 2.3343376678295804e-05,55 "loss": 1.7453,56 "step": 90806457 },58 {59 "epoch": 9.0,60 "learning_rate": 2.001136483772069e-05,61 "loss": 1.7178,62 "step": 102157263 },64 {65 "epoch": 10.0,66 "learning_rate": 1.6679294264134097e-05,67 "loss": 1.6944,68 "step": 113508069 },70 {71 "epoch": 11.0,72 "learning_rate": 1.334719432404177e-05,73 "loss": 1.6729,74 "step": 124858875 },76 {77 "epoch": 12.0,78 "learning_rate": 1.0015123750455182e-05,79 "loss": 1.6535,80 "step": 136209681 },82 {83 "epoch": 13.0,84 "learning_rate": 6.683023810362853e-06,85 "loss": 1.6363,86 "step": 147560487 },88 {89 "epoch": 14.0,90 "learning_rate": 3.350953236776263e-06,91 "loss": 1.6216,92 "step": 158911293 },94 {95 "epoch": 15.0,96 "learning_rate": 1.8882663189672388e-08,97 "loss": 1.6102,98 "step": 170262099 },100 {101 "epoch": 16.0,102 "learning_rate": 1.0015065017443704e-05,103 "loss": 1.6294,104 "step": 1816128105 },106 {107 "epoch": 17.0,108 "learning_rate": 7.515968037495155e-06,109 "loss": 1.6191,110 "step": 1929636111 },112 {113 "epoch": 18.0,114 "learning_rate": 5.016915107305212e-06,115 "loss": 1.6079,116 "step": 2043144117 },118 {119 "epoch": 19.0,120 "learning_rate": 2.517884201994573e-06,121 "loss": 1.5972,122 "step": 2156652123 },124 {125 "epoch": 20.0,126 "learning_rate": 1.8831271804630513e-08,127 "loss": 1.5893,128 "step": 2270160129 },130 {131 "epoch": 20.0,132 "step": 2270160,133 "total_flos": 3.805740641757115e+19,134 "train_loss": 0.4021474394976566,135 "train_runtime": 51134.1112,136 "train_samples_per_second": 11365.472,137 "train_steps_per_second": 44.396138 }139 ],140 "max_steps": 2270160,141 "num_train_epochs": 20,142 "total_flos": 3.805740641757115e+19,143 "trial_name": null,144 "trial_params": null145}146 