| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 3.0, | |
| "global_step": 201, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.8178331735378715, | |
| "eval_loss": 0.4154491126537323, | |
| "eval_runtime": 1.4755, | |
| "eval_samples_per_second": 706.864, | |
| "eval_steps_per_second": 11.521, | |
| "step": 67 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.8235858101629914, | |
| "eval_loss": 0.43939003348350525, | |
| "eval_runtime": 1.4773, | |
| "eval_samples_per_second": 706.022, | |
| "eval_steps_per_second": 11.508, | |
| "step": 134 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.835091083413231, | |
| "eval_loss": 0.4842393100261688, | |
| "eval_runtime": 1.486, | |
| "eval_samples_per_second": 701.894, | |
| "eval_steps_per_second": 11.44, | |
| "step": 201 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "step": 201, | |
| "total_flos": 3374793951575040.0, | |
| "train_loss": 0.3470354602111513, | |
| "train_runtime": 116.3712, | |
| "train_samples_per_second": 220.441, | |
| "train_steps_per_second": 1.727 | |
| } | |
| ], | |
| "max_steps": 201, | |
| "num_train_epochs": 3, | |
| "total_flos": 3374793951575040.0, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |