|
{ |
|
"best_metric": null, |
|
"best_model_checkpoint": null, |
|
"epoch": 10.0, |
|
"eval_steps": 500, |
|
"global_step": 200, |
|
"is_hyper_param_search": false, |
|
"is_local_process_zero": true, |
|
"is_world_process_zero": true, |
|
"log_history": [ |
|
{ |
|
"epoch": 10.0, |
|
"step": 200, |
|
"total_flos": 1115250888906000.0, |
|
"train_loss": 0.705328369140625, |
|
"train_runtime": 5429.659, |
|
"train_samples_per_second": 1.142, |
|
"train_steps_per_second": 0.037 |
|
} |
|
], |
|
"logging_steps": 500, |
|
"max_steps": 200, |
|
"num_input_tokens_seen": 0, |
|
"num_train_epochs": 10, |
|
"save_steps": 500, |
|
"total_flos": 1115250888906000.0, |
|
"train_batch_size": 32, |
|
"trial_name": null, |
|
"trial_params": null |
|
} |
|
|