| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 3.0, | |
| "global_step": 804, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 7.5621890547263685e-06, | |
| "loss": 0.3292, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "step": 804, | |
| "total_flos": 2156995090202112.0, | |
| "train_runtime": 678.8041, | |
| "train_samples_per_second": 1.184 | |
| } | |
| ], | |
| "max_steps": 804, | |
| "num_train_epochs": 3, | |
| "total_flos": 2156995090202112.0, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |