|
{ |
|
"best_metric": null, |
|
"best_model_checkpoint": null, |
|
"epoch": 0.8779631255487269, |
|
"global_step": 500, |
|
"is_hyper_param_search": false, |
|
"is_local_process_zero": true, |
|
"is_world_process_zero": true, |
|
"log_history": [ |
|
{ |
|
"epoch": 0.0, |
|
"learning_rate": 1.9988283538371414e-05, |
|
"loss": 2.1598, |
|
"step": 1 |
|
}, |
|
{ |
|
"epoch": 0.88, |
|
"learning_rate": 1.4141769185705918e-05, |
|
"loss": 1.9104, |
|
"step": 500 |
|
} |
|
], |
|
"max_steps": 1707, |
|
"num_train_epochs": 3, |
|
"total_flos": 1.41603259809792e+17, |
|
"trial_name": null, |
|
"trial_params": null |
|
} |
|
|