{ | |
"best_metric": null, | |
"best_model_checkpoint": null, | |
"epoch": 3.0, | |
"global_step": 765, | |
"is_hyper_param_search": false, | |
"is_local_process_zero": true, | |
"is_world_process_zero": true, | |
"log_history": [ | |
{ | |
"epoch": 1.96, | |
"learning_rate": 1.7320261437908496e-05, | |
"loss": 0.301, | |
"step": 500 | |
}, | |
{ | |
"epoch": 3.0, | |
"step": 765, | |
"total_flos": 362767106677248.0, | |
"train_loss": 0.26447707531498926, | |
"train_runtime": 179.164, | |
"train_samples_per_second": 68.1, | |
"train_steps_per_second": 4.27 | |
} | |
], | |
"max_steps": 765, | |
"num_train_epochs": 3, | |
"total_flos": 362767106677248.0, | |
"trial_name": null, | |
"trial_params": null | |
} | |