|
{ |
|
"best_metric": 0.6573426573426573, |
|
"best_model_checkpoint": "OS_OUT/harish/PT-UP-xlmR-FalseFalse-0_0_BEST/FalseFalse-0/0/checkpoint-12", |
|
"epoch": 6.0, |
|
"global_step": 12, |
|
"is_hyper_param_search": false, |
|
"is_local_process_zero": true, |
|
"is_world_process_zero": true, |
|
"log_history": [ |
|
{ |
|
"epoch": 1.0, |
|
"eval_accuracy": 0.5311355590820312, |
|
"eval_f1": 0.6049382716049383, |
|
"eval_loss": 1.011836290359497, |
|
"eval_runtime": 0.7598, |
|
"eval_samples_per_second": 359.298, |
|
"eval_steps_per_second": 46.064, |
|
"step": 2 |
|
}, |
|
{ |
|
"epoch": 2.0, |
|
"eval_accuracy": 0.6043956279754639, |
|
"eval_f1": 0.568, |
|
"eval_loss": 0.671142041683197, |
|
"eval_runtime": 0.7664, |
|
"eval_samples_per_second": 356.204, |
|
"eval_steps_per_second": 45.667, |
|
"step": 4 |
|
}, |
|
{ |
|
"epoch": 3.0, |
|
"eval_accuracy": 0.6190476417541504, |
|
"eval_f1": 0.5478260869565218, |
|
"eval_loss": 0.6358144879341125, |
|
"eval_runtime": 0.7706, |
|
"eval_samples_per_second": 354.265, |
|
"eval_steps_per_second": 45.419, |
|
"step": 6 |
|
}, |
|
{ |
|
"epoch": 4.0, |
|
"eval_accuracy": 0.6263736486434937, |
|
"eval_f1": 0.6015625, |
|
"eval_loss": 0.6275585293769836, |
|
"eval_runtime": 0.7711, |
|
"eval_samples_per_second": 354.038, |
|
"eval_steps_per_second": 45.39, |
|
"step": 8 |
|
}, |
|
{ |
|
"epoch": 5.0, |
|
"eval_accuracy": 0.6410256624221802, |
|
"eval_f1": 0.65, |
|
"eval_loss": 0.6581979990005493, |
|
"eval_runtime": 0.7584, |
|
"eval_samples_per_second": 359.969, |
|
"eval_steps_per_second": 46.15, |
|
"step": 10 |
|
}, |
|
{ |
|
"epoch": 6.0, |
|
"eval_accuracy": 0.6410256624221802, |
|
"eval_f1": 0.6573426573426573, |
|
"eval_loss": 0.760171115398407, |
|
"eval_runtime": 0.7597, |
|
"eval_samples_per_second": 359.348, |
|
"eval_steps_per_second": 46.07, |
|
"step": 12 |
|
} |
|
], |
|
"max_steps": 18, |
|
"num_train_epochs": 9, |
|
"total_flos": 20917328901120.0, |
|
"trial_name": null, |
|
"trial_params": null |
|
} |
|
|