| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 3.959322033898305, | |
| "global_step": 292, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.99, | |
| "eval_accuracy": 0.5730886850152905, | |
| "eval_loss": 0.6870243549346924, | |
| "eval_runtime": 9.2341, | |
| "eval_samples_per_second": 354.122, | |
| "eval_steps_per_second": 11.154, | |
| "step": 73 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "eval_accuracy": 0.5957186544342508, | |
| "eval_loss": 0.6825068593025208, | |
| "eval_runtime": 9.2756, | |
| "eval_samples_per_second": 352.537, | |
| "eval_steps_per_second": 11.104, | |
| "step": 147 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.6012232415902141, | |
| "eval_loss": 0.6809478402137756, | |
| "eval_runtime": 9.287, | |
| "eval_samples_per_second": 352.104, | |
| "eval_steps_per_second": 11.091, | |
| "step": 221 | |
| }, | |
| { | |
| "epoch": 3.96, | |
| "eval_accuracy": 0.5957186544342508, | |
| "eval_loss": 0.679547131061554, | |
| "eval_runtime": 9.2851, | |
| "eval_samples_per_second": 352.177, | |
| "eval_steps_per_second": 11.093, | |
| "step": 292 | |
| }, | |
| { | |
| "epoch": 3.96, | |
| "step": 292, | |
| "total_flos": 1.2991635112329216e+16, | |
| "train_loss": 0.6814518758695419, | |
| "train_runtime": 329.9402, | |
| "train_samples_per_second": 114.287, | |
| "train_steps_per_second": 0.885 | |
| } | |
| ], | |
| "max_steps": 292, | |
| "num_train_epochs": 4, | |
| "total_flos": 1.2991635112329216e+16, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |