{ | |
"best_metric": null, | |
"best_model_checkpoint": null, | |
"epoch": 15.029411764705882, | |
"eval_steps": 500, | |
"global_step": 16, | |
"is_hyper_param_search": false, | |
"is_local_process_zero": true, | |
"is_world_process_zero": true, | |
"log_history": [ | |
{ | |
"epoch": 5.03, | |
"learning_rate": 2.1818181818181818e-05, | |
"loss": 2.3905, | |
"step": 6 | |
}, | |
{ | |
"epoch": 11.03, | |
"learning_rate": 4.3636363636363636e-05, | |
"loss": 2.2885, | |
"step": 12 | |
} | |
], | |
"logging_steps": 6, | |
"max_steps": 544, | |
"num_train_epochs": 16, | |
"save_steps": 500, | |
"total_flos": 4980664864604160.0, | |
"trial_name": null, | |
"trial_params": null | |
} | |