|
{ |
|
"best_metric": null, |
|
"best_model_checkpoint": null, |
|
"epoch": 3.9879518072289155, |
|
"global_step": 164, |
|
"is_hyper_param_search": false, |
|
"is_local_process_zero": true, |
|
"is_world_process_zero": true, |
|
"log_history": [ |
|
{ |
|
"epoch": 3.99, |
|
"step": 164, |
|
"total_flos": 3.930845214880973e+16, |
|
"train_loss": 0.5002818223906726, |
|
"train_runtime": 1397.9128, |
|
"train_samples_per_second": 30.36, |
|
"train_steps_per_second": 0.117 |
|
} |
|
], |
|
"max_steps": 164, |
|
"num_train_epochs": 4, |
|
"total_flos": 3.930845214880973e+16, |
|
"trial_name": null, |
|
"trial_params": null |
|
} |
|
|