| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 0.47619047619047616, | |
| "global_step": 500, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.6825396825396828e-05, | |
| "loss": 4.4123, | |
| "step": 500 | |
| } | |
| ], | |
| "max_steps": 3150, | |
| "num_train_epochs": 3, | |
| "total_flos": 87360049152000.0, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |