|
{ |
|
"best_metric": 59.61538461538461, |
|
"best_model_checkpoint": "outputs/bitfit/t5-base/superglue-wsc.fixed/checkpoint-300", |
|
"epoch": 20.0, |
|
"global_step": 360, |
|
"is_hyper_param_search": false, |
|
"is_local_process_zero": true, |
|
"is_world_process_zero": true, |
|
"log_history": [ |
|
{ |
|
"epoch": 5.56, |
|
"eval_accuracy": 40.38461538461539, |
|
"eval_average_metrics": 40.38461538461539, |
|
"eval_loss": 0.33406707644462585, |
|
"eval_runtime": 0.2912, |
|
"eval_samples_per_second": 178.549, |
|
"step": 100 |
|
}, |
|
{ |
|
"epoch": 11.11, |
|
"eval_accuracy": 40.38461538461539, |
|
"eval_average_metrics": 40.38461538461539, |
|
"eval_loss": 0.27627798914909363, |
|
"eval_runtime": 0.3043, |
|
"eval_samples_per_second": 170.86, |
|
"step": 200 |
|
}, |
|
{ |
|
"epoch": 16.67, |
|
"eval_accuracy": 59.61538461538461, |
|
"eval_average_metrics": 59.61538461538461, |
|
"eval_loss": 0.2665896713733673, |
|
"eval_runtime": 0.2321, |
|
"eval_samples_per_second": 224.02, |
|
"step": 300 |
|
}, |
|
{ |
|
"epoch": 20.0, |
|
"step": 360, |
|
"total_flos": 1502077285006848.0, |
|
"train_loss": 0.33580604129367403, |
|
"train_runtime": 150.2227, |
|
"train_samples_per_second": 73.757, |
|
"train_steps_per_second": 2.396 |
|
} |
|
], |
|
"max_steps": 360, |
|
"num_train_epochs": 20, |
|
"total_flos": 1502077285006848.0, |
|
"trial_name": null, |
|
"trial_params": null |
|
} |
|
|