| { |
| "best_metric": null, |
| "best_model_checkpoint": null, |
| "epoch": 5.0, |
| "global_step": 575, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 4.35, |
| "learning_rate": 2.6086956521739132e-06, |
| "loss": 0.3249, |
| "step": 500 |
| }, |
| { |
| "epoch": 5.0, |
| "step": 575, |
| "total_flos": 0.0, |
| "train_loss": 0.2963437312582265, |
| "train_runtime": 254.6718, |
| "train_samples_per_second": 72.014, |
| "train_steps_per_second": 2.258 |
| } |
| ], |
| "max_steps": 575, |
| "num_train_epochs": 5, |
| "total_flos": 0.0, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|
