| { | |
| "best_metric": 0.6158979909555603, | |
| "best_model_checkpoint": "test-glue/checkpoint-1070", | |
| "epoch": 3.0, | |
| "global_step": 1605, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.3769470404984425e-05, | |
| "loss": 0.438, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_loss": 0.44685283303260803, | |
| "eval_matthews_correlation": 0.5729657494988228, | |
| "eval_runtime": 4.6654, | |
| "eval_samples_per_second": 223.559, | |
| "step": 535 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 7.538940809968847e-06, | |
| "loss": 0.248, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_loss": 0.4980889558792114, | |
| "eval_matthews_correlation": 0.6158979909555603, | |
| "eval_runtime": 4.6816, | |
| "eval_samples_per_second": 222.789, | |
| "step": 1070 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "learning_rate": 1.308411214953271e-06, | |
| "loss": 0.147, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_loss": 0.6851685643196106, | |
| "eval_matthews_correlation": 0.6142064528308289, | |
| "eval_runtime": 4.6892, | |
| "eval_samples_per_second": 222.426, | |
| "step": 1605 | |
| } | |
| ], | |
| "max_steps": 1605, | |
| "num_train_epochs": 3, | |
| "total_flos": 1071782270269488.0, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |