|
{ |
|
"best_metric": null, |
|
"best_model_checkpoint": null, |
|
"epoch": 18.181818181818183, |
|
"global_step": 200, |
|
"is_hyper_param_search": false, |
|
"is_local_process_zero": true, |
|
"is_world_process_zero": true, |
|
"log_history": [ |
|
{ |
|
"epoch": 9.09, |
|
"learning_rate": 1.6666666666666667e-05, |
|
"loss": 4.228, |
|
"step": 100 |
|
}, |
|
{ |
|
"epoch": 18.18, |
|
"learning_rate": 0.0, |
|
"loss": 1.5397, |
|
"step": 200 |
|
}, |
|
{ |
|
"epoch": 18.18, |
|
"step": 200, |
|
"total_flos": 374838363061440.0, |
|
"train_loss": 2.8838272857666016, |
|
"train_runtime": 79.0949, |
|
"train_samples_per_second": 30.343, |
|
"train_steps_per_second": 2.529 |
|
} |
|
], |
|
"max_steps": 200, |
|
"num_train_epochs": 19, |
|
"total_flos": 374838363061440.0, |
|
"trial_name": null, |
|
"trial_params": null |
|
} |
|
|