| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 0.5957446808510638, | |
| "eval_steps": 500, | |
| "global_step": 14, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9906859460363307e-05, | |
| "loss": 1.7537, | |
| "step": 1 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.77571129070442e-05, | |
| "loss": 0.7295, | |
| "step": 5 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2034560130526341e-05, | |
| "loss": 0.144, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "eval_loss": 0.08691789209842682, | |
| "eval_runtime": 5.969, | |
| "eval_samples_per_second": 37.694, | |
| "eval_steps_per_second": 1.34, | |
| "step": 14 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "step": 14, | |
| "total_flos": 5613992017920.0, | |
| "train_loss": 0.4192912685019629, | |
| "train_runtime": 308.2018, | |
| "train_samples_per_second": 9.734, | |
| "train_steps_per_second": 0.075 | |
| } | |
| ], | |
| "logging_steps": 5, | |
| "max_steps": 23, | |
| "num_train_epochs": 1, | |
| "save_steps": 500, | |
| "total_flos": 5613992017920.0, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |