| { | |
| "best_metric": 0.8072445019404917, | |
| "best_model_checkpoint": "final_models/glue_models/structroberta_s1_final//finetune/cola/checkpoint-400", | |
| "epoch": 10.0, | |
| "global_step": 690, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 5.8, | |
| "eval_accuracy": 0.707556426525116, | |
| "eval_f1": 0.8072445019404917, | |
| "eval_loss": 1.117124319076538, | |
| "eval_mcc": 0.23284228747571145, | |
| "eval_runtime": 1.9992, | |
| "eval_samples_per_second": 509.691, | |
| "eval_steps_per_second": 64.024, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 7.25, | |
| "learning_rate": 1.3768115942028985e-05, | |
| "loss": 0.2812, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "step": 690, | |
| "total_flos": 6812785759703040.0, | |
| "train_loss": 0.21529039092685864, | |
| "train_runtime": 379.9226, | |
| "train_samples_per_second": 214.886, | |
| "train_steps_per_second": 1.816 | |
| } | |
| ], | |
| "max_steps": 690, | |
| "num_train_epochs": 10, | |
| "total_flos": 6812785759703040.0, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |