BERT-Router-large-v2

This model is a fine-tuned version of google-bert/bert-large-uncased on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.3815
• Accuracy: 0.838
• Auc: 0.951

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 1e-06
• train_batch_size: 1024
• eval_batch_size: 1024
• seed: 42
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Auc	Accuracy	Validation Loss
0.3719	1.0	12	0.949	0.835	0.3895
0.3741	2.0	24	0.949	0.835	0.3886
0.3673	3.0	36	0.949	0.836	0.3879
0.3692	4.0	48	0.949	0.836	0.3873
0.3724	5.0	60	0.3866	0.836	0.95
0.3683	6.0	72	0.3859	0.836	0.95
0.3678	7.0	84	0.3853	0.836	0.95
0.3671	8.0	96	0.3847	0.837	0.95
0.3614	9.0	108	0.3842	0.837	0.95
0.3658	10.0	120	0.3838	0.837	0.95
0.3681	11.0	132	0.3834	0.837	0.95
0.3642	12.0	144	0.3831	0.837	0.95
0.3659	13.0	156	0.3827	0.837	0.95
0.3693	14.0	168	0.3823	0.838	0.95
0.3637	15.0	180	0.3820	0.838	0.951
0.3596	16.0	192	0.3819	0.838	0.951
0.3732	17.0	204	0.3817	0.838	0.951
0.3685	18.0	216	0.3816	0.838	0.951
0.3613	19.0	228	0.3815	0.838	0.951
0.3656	20.0	240	0.3815	0.838	0.951

Framework versions

• Transformers 4.46.3
• Pytorch 2.4.1+cu121
• Datasets 3.1.0
• Tokenizers 0.20.3