| { | |
| "best_metric": 0.0, | |
| "best_model_checkpoint": "outputs/bitfit/t5-base/cola/checkpoint-86", | |
| "epoch": 20.0, | |
| "global_step": 1720, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "eval_average_metrics": 0.0, | |
| "eval_loss": 0.27982431650161743, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 0.8066, | |
| "eval_samples_per_second": 645.909, | |
| "step": 86 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_average_metrics": 0.0, | |
| "eval_loss": 0.27295124530792236, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 0.8088, | |
| "eval_samples_per_second": 644.148, | |
| "step": 172 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_average_metrics": 0.0, | |
| "eval_loss": 0.27825212478637695, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 0.7356, | |
| "eval_samples_per_second": 708.276, | |
| "step": 258 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_average_metrics": 0.0, | |
| "eval_loss": 0.2720172107219696, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 0.7952, | |
| "eval_samples_per_second": 655.185, | |
| "step": 344 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_average_metrics": 0.0, | |
| "eval_loss": 0.2690984904766083, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 0.7358, | |
| "eval_samples_per_second": 708.105, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 5.81, | |
| "learning_rate": 0.002127906976744186, | |
| "loss": 0.3, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_average_metrics": 0.0, | |
| "eval_loss": 0.2694063186645508, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 0.7511, | |
| "eval_samples_per_second": 693.659, | |
| "step": 516 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_average_metrics": 0.0, | |
| "eval_loss": 0.28033140301704407, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 0.7448, | |
| "eval_samples_per_second": 699.547, | |
| "step": 602 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_average_metrics": 0.0, | |
| "eval_loss": 0.2737719714641571, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 0.7024, | |
| "eval_samples_per_second": 741.703, | |
| "step": 688 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_average_metrics": 0.0, | |
| "eval_loss": 0.26990658044815063, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 0.7557, | |
| "eval_samples_per_second": 689.411, | |
| "step": 774 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_average_metrics": 0.0, | |
| "eval_loss": 0.269914448261261, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 0.7515, | |
| "eval_samples_per_second": 693.272, | |
| "step": 860 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "eval_average_metrics": 0.0, | |
| "eval_loss": 0.2703316807746887, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 0.7544, | |
| "eval_samples_per_second": 690.634, | |
| "step": 946 | |
| }, | |
| { | |
| "epoch": 11.63, | |
| "learning_rate": 0.0012558139534883722, | |
| "loss": 0.2698, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "eval_average_metrics": 0.0, | |
| "eval_loss": 0.2690633535385132, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 0.8328, | |
| "eval_samples_per_second": 625.617, | |
| "step": 1032 | |
| }, | |
| { | |
| "epoch": 13.0, | |
| "eval_average_metrics": 0.0, | |
| "eval_loss": 0.2688796818256378, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 0.7677, | |
| "eval_samples_per_second": 678.682, | |
| "step": 1118 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "eval_average_metrics": 0.0, | |
| "eval_loss": 0.26915279030799866, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 0.7677, | |
| "eval_samples_per_second": 678.671, | |
| "step": 1204 | |
| }, | |
| { | |
| "epoch": 15.0, | |
| "eval_average_metrics": 0.0, | |
| "eval_loss": 0.27187806367874146, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 0.7511, | |
| "eval_samples_per_second": 693.632, | |
| "step": 1290 | |
| }, | |
| { | |
| "epoch": 16.0, | |
| "eval_average_metrics": 0.0, | |
| "eval_loss": 0.27496567368507385, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 0.7145, | |
| "eval_samples_per_second": 729.175, | |
| "step": 1376 | |
| }, | |
| { | |
| "epoch": 17.0, | |
| "eval_average_metrics": 0.0, | |
| "eval_loss": 0.26937127113342285, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 0.7488, | |
| "eval_samples_per_second": 695.759, | |
| "step": 1462 | |
| }, | |
| { | |
| "epoch": 17.44, | |
| "learning_rate": 0.0003837209302325582, | |
| "loss": 0.2686, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 18.0, | |
| "eval_average_metrics": 0.0, | |
| "eval_loss": 0.26900675892829895, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 0.7556, | |
| "eval_samples_per_second": 689.484, | |
| "step": 1548 | |
| }, | |
| { | |
| "epoch": 19.0, | |
| "eval_average_metrics": 0.0, | |
| "eval_loss": 0.2690284252166748, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 0.7419, | |
| "eval_samples_per_second": 702.253, | |
| "step": 1634 | |
| }, | |
| { | |
| "epoch": 20.0, | |
| "eval_average_metrics": 0.0, | |
| "eval_loss": 0.26897749304771423, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 0.7455, | |
| "eval_samples_per_second": 698.863, | |
| "step": 1720 | |
| }, | |
| { | |
| "epoch": 20.0, | |
| "step": 1720, | |
| "total_flos": 6646762904679216.0, | |
| "train_loss": 0.2779904210290243, | |
| "train_runtime": 435.9382, | |
| "train_samples_per_second": 392.303, | |
| "train_steps_per_second": 3.946 | |
| } | |
| ], | |
| "max_steps": 1720, | |
| "num_train_epochs": 20, | |
| "total_flos": 6646762904679216.0, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |