|
{ |
|
"best_metric": null, |
|
"best_model_checkpoint": null, |
|
"epoch": 3.0, |
|
"global_step": 4446, |
|
"is_hyper_param_search": false, |
|
"is_local_process_zero": true, |
|
"is_world_process_zero": true, |
|
"log_history": [ |
|
{ |
|
"epoch": 0.34, |
|
"learning_rate": 1.7750787224471436e-05, |
|
"loss": 0.2516, |
|
"step": 500 |
|
}, |
|
{ |
|
"epoch": 0.67, |
|
"learning_rate": 1.550157444894287e-05, |
|
"loss": 0.1893, |
|
"step": 1000 |
|
}, |
|
{ |
|
"epoch": 1.0, |
|
"eval_accuracy": 0.9572655742438461, |
|
"eval_f1": 0.5739710789766408, |
|
"eval_loss": 0.11261405795812607, |
|
"eval_precision": 0.5113974231912785, |
|
"eval_recall": 0.6539923954372624, |
|
"eval_runtime": 94.9871, |
|
"eval_samples_per_second": 12.486, |
|
"eval_steps_per_second": 3.127, |
|
"step": 1482 |
|
}, |
|
{ |
|
"epoch": 1.01, |
|
"learning_rate": 1.3252361673414307e-05, |
|
"loss": 0.1719, |
|
"step": 1500 |
|
}, |
|
{ |
|
"epoch": 1.35, |
|
"learning_rate": 1.100314889788574e-05, |
|
"loss": 0.1202, |
|
"step": 2000 |
|
}, |
|
{ |
|
"epoch": 1.69, |
|
"learning_rate": 8.753936122357177e-06, |
|
"loss": 0.1117, |
|
"step": 2500 |
|
}, |
|
{ |
|
"epoch": 2.0, |
|
"eval_accuracy": 0.9754806419788734, |
|
"eval_f1": 0.725097218067604, |
|
"eval_loss": 0.06848359107971191, |
|
"eval_precision": 0.686685552407932, |
|
"eval_recall": 0.7680608365019012, |
|
"eval_runtime": 101.9581, |
|
"eval_samples_per_second": 11.632, |
|
"eval_steps_per_second": 2.913, |
|
"step": 2964 |
|
}, |
|
{ |
|
"epoch": 2.02, |
|
"learning_rate": 6.504723346828611e-06, |
|
"loss": 0.1075, |
|
"step": 3000 |
|
}, |
|
{ |
|
"epoch": 2.36, |
|
"learning_rate": 4.255510571300046e-06, |
|
"loss": 0.0829, |
|
"step": 3500 |
|
}, |
|
{ |
|
"epoch": 2.7, |
|
"learning_rate": 2.00629779577148e-06, |
|
"loss": 0.0889, |
|
"step": 4000 |
|
}, |
|
{ |
|
"epoch": 3.0, |
|
"eval_accuracy": 0.9814961861449987, |
|
"eval_f1": 0.777242624924744, |
|
"eval_loss": 0.054215945303440094, |
|
"eval_precision": 0.7402522935779816, |
|
"eval_recall": 0.8181242078580482, |
|
"eval_runtime": 95.5514, |
|
"eval_samples_per_second": 12.412, |
|
"eval_steps_per_second": 3.108, |
|
"step": 4446 |
|
} |
|
], |
|
"max_steps": 4446, |
|
"num_train_epochs": 3, |
|
"total_flos": 1931548401639798.0, |
|
"trial_name": null, |
|
"trial_params": null |
|
} |
|
|