|
{ |
|
"best_metric": 0.7792222559126744, |
|
"best_model_checkpoint": "output-no-git/EN-OUTPUT/bert-base-cased/TrueTrue-0/3/checkpoint-416", |
|
"epoch": 4.0, |
|
"global_step": 416, |
|
"is_hyper_param_search": false, |
|
"is_local_process_zero": true, |
|
"is_world_process_zero": true, |
|
"log_history": [ |
|
{ |
|
"epoch": 1.0, |
|
"eval_accuracy": 0.7725321650505066, |
|
"eval_f1": 0.7499645670088482, |
|
"eval_f1_binary": 0.8250825082508252, |
|
"eval_f1_weighted": 0.7664066914867057, |
|
"eval_loss": 0.6242344379425049, |
|
"eval_runtime": 1.3874, |
|
"eval_samples_per_second": 335.883, |
|
"eval_steps_per_second": 42.526, |
|
"step": 104 |
|
}, |
|
{ |
|
"epoch": 2.0, |
|
"eval_accuracy": 0.7639485001564026, |
|
"eval_f1": 0.7552193761580044, |
|
"eval_f1_binary": 0.8014440433212996, |
|
"eval_f1_weighted": 0.7653372217602706, |
|
"eval_loss": 0.7207056283950806, |
|
"eval_runtime": 1.3858, |
|
"eval_samples_per_second": 336.264, |
|
"eval_steps_per_second": 42.574, |
|
"step": 208 |
|
}, |
|
{ |
|
"epoch": 3.0, |
|
"eval_accuracy": 0.7875536680221558, |
|
"eval_f1": 0.7770819743422482, |
|
"eval_f1_binary": 0.8253968253968252, |
|
"eval_f1_weighted": 0.7876573280065549, |
|
"eval_loss": 0.9023728966712952, |
|
"eval_runtime": 1.3902, |
|
"eval_samples_per_second": 335.192, |
|
"eval_steps_per_second": 42.439, |
|
"step": 312 |
|
}, |
|
{ |
|
"epoch": 4.0, |
|
"eval_accuracy": 0.7854077219963074, |
|
"eval_f1": 0.7792222559126744, |
|
"eval_f1_binary": 0.8161764705882353, |
|
"eval_f1_weighted": 0.7873109466785696, |
|
"eval_loss": 1.1547576189041138, |
|
"eval_runtime": 1.3832, |
|
"eval_samples_per_second": 336.908, |
|
"eval_steps_per_second": 42.656, |
|
"step": 416 |
|
} |
|
], |
|
"max_steps": 936, |
|
"num_train_epochs": 9, |
|
"total_flos": 875370481182720.0, |
|
"trial_name": null, |
|
"trial_params": null |
|
} |
|
|