| { | |
| "best_metric": 0.5912045240402222, | |
| "best_model_checkpoint": "add_BERT_no_pretrain_mrpc/checkpoint-203", | |
| "epoch": 12.0, | |
| "global_step": 348, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 3.9200000000000004e-05, | |
| "loss": 0.6854, | |
| "step": 29 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.6838235294117647, | |
| "eval_combined_score": 0.7480253018237863, | |
| "eval_f1": 0.8122270742358079, | |
| "eval_loss": 0.6710845828056335, | |
| "eval_runtime": 0.7677, | |
| "eval_samples_per_second": 531.445, | |
| "eval_steps_per_second": 5.21, | |
| "step": 29 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 3.8400000000000005e-05, | |
| "loss": 0.6496, | |
| "step": 58 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.6838235294117647, | |
| "eval_combined_score": 0.7480253018237863, | |
| "eval_f1": 0.8122270742358079, | |
| "eval_loss": 0.6802303791046143, | |
| "eval_runtime": 0.7711, | |
| "eval_samples_per_second": 529.088, | |
| "eval_steps_per_second": 5.187, | |
| "step": 58 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 3.76e-05, | |
| "loss": 0.648, | |
| "step": 87 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.6838235294117647, | |
| "eval_combined_score": 0.7480253018237863, | |
| "eval_f1": 0.8122270742358079, | |
| "eval_loss": 0.6246173977851868, | |
| "eval_runtime": 0.7699, | |
| "eval_samples_per_second": 529.951, | |
| "eval_steps_per_second": 5.196, | |
| "step": 87 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 3.680000000000001e-05, | |
| "loss": 0.6363, | |
| "step": 116 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.6838235294117647, | |
| "eval_combined_score": 0.7480253018237863, | |
| "eval_f1": 0.8122270742358079, | |
| "eval_loss": 0.6174120903015137, | |
| "eval_runtime": 0.7694, | |
| "eval_samples_per_second": 530.299, | |
| "eval_steps_per_second": 5.199, | |
| "step": 116 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "learning_rate": 3.6e-05, | |
| "loss": 0.6049, | |
| "step": 145 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.6593137254901961, | |
| "eval_combined_score": 0.7026001899845862, | |
| "eval_f1": 0.7458866544789762, | |
| "eval_loss": 0.6176258325576782, | |
| "eval_runtime": 0.7696, | |
| "eval_samples_per_second": 530.143, | |
| "eval_steps_per_second": 5.197, | |
| "step": 145 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "learning_rate": 3.52e-05, | |
| "loss": 0.5491, | |
| "step": 174 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.6813725490196079, | |
| "eval_combined_score": 0.7381626152038103, | |
| "eval_f1": 0.7949526813880127, | |
| "eval_loss": 0.6037724018096924, | |
| "eval_runtime": 0.7688, | |
| "eval_samples_per_second": 530.695, | |
| "eval_steps_per_second": 5.203, | |
| "step": 174 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "learning_rate": 3.44e-05, | |
| "loss": 0.5601, | |
| "step": 203 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.696078431372549, | |
| "eval_combined_score": 0.7447058823529411, | |
| "eval_f1": 0.7933333333333332, | |
| "eval_loss": 0.5912045240402222, | |
| "eval_runtime": 0.7735, | |
| "eval_samples_per_second": 527.497, | |
| "eval_steps_per_second": 5.172, | |
| "step": 203 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "learning_rate": 3.3600000000000004e-05, | |
| "loss": 0.5505, | |
| "step": 232 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.6715686274509803, | |
| "eval_combined_score": 0.7248571614076094, | |
| "eval_f1": 0.7781456953642384, | |
| "eval_loss": 0.6346378922462463, | |
| "eval_runtime": 0.772, | |
| "eval_samples_per_second": 528.513, | |
| "eval_steps_per_second": 5.182, | |
| "step": 232 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "learning_rate": 3.28e-05, | |
| "loss": 0.5327, | |
| "step": 261 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.6544117647058824, | |
| "eval_combined_score": 0.7037382816524158, | |
| "eval_f1": 0.7530647985989491, | |
| "eval_loss": 0.628277063369751, | |
| "eval_runtime": 0.7697, | |
| "eval_samples_per_second": 530.071, | |
| "eval_steps_per_second": 5.197, | |
| "step": 261 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "learning_rate": 3.2000000000000005e-05, | |
| "loss": 0.529, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_accuracy": 0.6519607843137255, | |
| "eval_combined_score": 0.7044050496911094, | |
| "eval_f1": 0.7568493150684933, | |
| "eval_loss": 0.634149432182312, | |
| "eval_runtime": 0.7681, | |
| "eval_samples_per_second": 531.164, | |
| "eval_steps_per_second": 5.207, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "learning_rate": 3.1200000000000006e-05, | |
| "loss": 0.5337, | |
| "step": 319 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "eval_accuracy": 0.6617647058823529, | |
| "eval_combined_score": 0.7098297213622291, | |
| "eval_f1": 0.7578947368421053, | |
| "eval_loss": 0.6285053491592407, | |
| "eval_runtime": 0.7705, | |
| "eval_samples_per_second": 529.529, | |
| "eval_steps_per_second": 5.191, | |
| "step": 319 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "learning_rate": 3.0400000000000004e-05, | |
| "loss": 0.5383, | |
| "step": 348 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "eval_accuracy": 0.6348039215686274, | |
| "eval_combined_score": 0.6817006857387764, | |
| "eval_f1": 0.7285974499089254, | |
| "eval_loss": 0.632221519947052, | |
| "eval_runtime": 0.7713, | |
| "eval_samples_per_second": 528.973, | |
| "eval_steps_per_second": 5.186, | |
| "step": 348 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "step": 348, | |
| "total_flos": 6286539604623360.0, | |
| "train_loss": 0.5847925564338421, | |
| "train_runtime": 324.1281, | |
| "train_samples_per_second": 565.826, | |
| "train_steps_per_second": 4.474 | |
| } | |
| ], | |
| "max_steps": 1450, | |
| "num_train_epochs": 50, | |
| "total_flos": 6286539604623360.0, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |