BERT-Router-large

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: 0.3906
• Accuracy: 0.833
• Auc: 0.949

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: 5e-06
• train_batch_size: 256
• eval_batch_size: 256
• 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
0.5459	1.0	45	0.893	0.728	0.5640
0.5209	2.0	90	0.904	0.754	0.5379
0.5029	3.0	135	0.911	0.764	0.5176
0.4966	4.0	180	0.916	0.77	0.5019
0.4802	5.0	225	0.92	0.782	0.4860
0.4669	6.0	270	0.925	0.79	0.4712
0.4516	7.0	315	0.928	0.799	0.4597
0.434	8.0	360	0.931	0.803	0.4495
0.429	9.0	405	0.934	0.808	0.4395
0.4187	10.0	450	0.937	0.815	0.4301
0.4008	11.0	495	0.939	0.817	0.4226
0.4088	12.0	540	0.4162	0.82	0.942
0.3915	13.0	585	0.4093	0.824	0.943
0.3877	14.0	630	0.4045	0.827	0.945
0.388	15.0	675	0.4000	0.83	0.946
0.3842	16.0	720	0.3974	0.831	0.947
0.3676	17.0	765	0.3940	0.832	0.948
0.3717	18.0	810	0.3922	0.833	0.948
0.3793	19.0	855	0.3909	0.833	0.949
0.3768	20.0	900	0.3906	0.833	0.949

Framework versions

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