| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 20.0, | |
| "global_step": 1280, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 7.81, | |
| "learning_rate": 3.0468750000000002e-05, | |
| "loss": 0.132, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 15.62, | |
| "learning_rate": 1.09375e-05, | |
| "loss": 0.0102, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 20.0, | |
| "step": 1280, | |
| "total_flos": 13844892618547200, | |
| "train_runtime": 841.737, | |
| "train_samples_per_second": 1.521 | |
| } | |
| ], | |
| "max_steps": 1280, | |
| "num_train_epochs": 20, | |
| "total_flos": 13844892618547200, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |