|
{ |
|
"best_metric": 0.008413420058786869, |
|
"best_model_checkpoint": "/kaggle/working/BERT-BASE-threatening/checkpoint-8936", |
|
"epoch": 2.0, |
|
"eval_steps": 500, |
|
"global_step": 8936, |
|
"is_hyper_param_search": false, |
|
"is_local_process_zero": true, |
|
"is_world_process_zero": true, |
|
"log_history": [ |
|
{ |
|
"epoch": 0.11, |
|
"learning_rate": 1.8880931065353626e-05, |
|
"loss": 0.0302, |
|
"step": 500 |
|
}, |
|
{ |
|
"epoch": 0.22, |
|
"learning_rate": 1.7761862130707253e-05, |
|
"loss": 0.0143, |
|
"step": 1000 |
|
}, |
|
{ |
|
"epoch": 0.34, |
|
"learning_rate": 1.6642793196060877e-05, |
|
"loss": 0.0094, |
|
"step": 1500 |
|
}, |
|
{ |
|
"epoch": 0.45, |
|
"learning_rate": 1.5523724261414504e-05, |
|
"loss": 0.0101, |
|
"step": 2000 |
|
}, |
|
{ |
|
"epoch": 0.56, |
|
"learning_rate": 1.4404655326768131e-05, |
|
"loss": 0.0096, |
|
"step": 2500 |
|
}, |
|
{ |
|
"epoch": 0.67, |
|
"learning_rate": 1.3285586392121757e-05, |
|
"loss": 0.0137, |
|
"step": 3000 |
|
}, |
|
{ |
|
"epoch": 0.78, |
|
"learning_rate": 1.2166517457475382e-05, |
|
"loss": 0.0114, |
|
"step": 3500 |
|
}, |
|
{ |
|
"epoch": 0.9, |
|
"learning_rate": 1.1047448522829008e-05, |
|
"loss": 0.0098, |
|
"step": 4000 |
|
}, |
|
{ |
|
"epoch": 1.0, |
|
"eval_accuracy": 0.9969225604297225, |
|
"eval_loss": 0.010326598770916462, |
|
"eval_runtime": 220.3979, |
|
"eval_samples_per_second": 81.09, |
|
"eval_steps_per_second": 5.068, |
|
"step": 4468 |
|
}, |
|
{ |
|
"epoch": 1.01, |
|
"learning_rate": 9.928379588182634e-06, |
|
"loss": 0.0085, |
|
"step": 4500 |
|
}, |
|
{ |
|
"epoch": 1.12, |
|
"learning_rate": 8.809310653536257e-06, |
|
"loss": 0.006, |
|
"step": 5000 |
|
}, |
|
{ |
|
"epoch": 1.23, |
|
"learning_rate": 7.690241718889885e-06, |
|
"loss": 0.0069, |
|
"step": 5500 |
|
}, |
|
{ |
|
"epoch": 1.34, |
|
"learning_rate": 6.57117278424351e-06, |
|
"loss": 0.0059, |
|
"step": 6000 |
|
}, |
|
{ |
|
"epoch": 1.45, |
|
"learning_rate": 5.452103849597136e-06, |
|
"loss": 0.0053, |
|
"step": 6500 |
|
}, |
|
{ |
|
"epoch": 1.57, |
|
"learning_rate": 4.3330349149507615e-06, |
|
"loss": 0.0027, |
|
"step": 7000 |
|
}, |
|
{ |
|
"epoch": 1.68, |
|
"learning_rate": 3.213965980304387e-06, |
|
"loss": 0.0047, |
|
"step": 7500 |
|
}, |
|
{ |
|
"epoch": 1.79, |
|
"learning_rate": 2.0948970456580127e-06, |
|
"loss": 0.0051, |
|
"step": 8000 |
|
}, |
|
{ |
|
"epoch": 1.9, |
|
"learning_rate": 9.758281110116385e-07, |
|
"loss": 0.0064, |
|
"step": 8500 |
|
}, |
|
{ |
|
"epoch": 2.0, |
|
"eval_accuracy": 0.9977059086839749, |
|
"eval_loss": 0.008413420058786869, |
|
"eval_runtime": 220.3659, |
|
"eval_samples_per_second": 81.101, |
|
"eval_steps_per_second": 5.069, |
|
"step": 8936 |
|
} |
|
], |
|
"logging_steps": 500, |
|
"max_steps": 8936, |
|
"num_train_epochs": 2, |
|
"save_steps": 500, |
|
"total_flos": 2.592649627027338e+16, |
|
"trial_name": null, |
|
"trial_params": null |
|
} |
|
|