| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 0.78125, | |
| "global_step": 75, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0002, | |
| "loss": 4.1673, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0002, | |
| "loss": 2.8311, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 0.0002, | |
| "loss": 2.5688, | |
| "step": 75 | |
| } | |
| ], | |
| "max_steps": 96, | |
| "num_train_epochs": 1, | |
| "total_flos": 60843769036800.0, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |