| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 4.878048780487805, | |
| "global_step": 13000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 8.126876876876877e-05, | |
| "loss": 0.7477, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 3.75, | |
| "learning_rate": 6.25e-05, | |
| "loss": 0.3572, | |
| "step": 10000 | |
| } | |
| ], | |
| "max_steps": 26650, | |
| "num_train_epochs": 10, | |
| "total_flos": 2.9029760945985946e+17, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |