|
{ |
|
"best_metric": null, |
|
"best_model_checkpoint": null, |
|
"epoch": 0.9596928982725528, |
|
"global_step": 1500, |
|
"is_hyper_param_search": false, |
|
"is_local_process_zero": true, |
|
"is_world_process_zero": true, |
|
"log_history": [ |
|
{ |
|
"epoch": 0.32, |
|
"learning_rate": 1.872040946896993e-05, |
|
"loss": 0.3165, |
|
"step": 500 |
|
}, |
|
{ |
|
"epoch": 0.64, |
|
"learning_rate": 1.744081893793986e-05, |
|
"loss": 0.2508, |
|
"step": 1000 |
|
}, |
|
{ |
|
"epoch": 0.96, |
|
"learning_rate": 1.616122840690979e-05, |
|
"loss": 0.2268, |
|
"step": 1500 |
|
} |
|
], |
|
"max_steps": 7815, |
|
"num_train_epochs": 5, |
|
"total_flos": 3150811589944704.0, |
|
"trial_name": null, |
|
"trial_params": null |
|
} |
|
|