Biobert_combo_1-6lakh-8-8-2

This model is a fine-tuned version of dmis-lab/biobert-base-cased-v1.2 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.5221
• Accuracy: 0.62
• Auc: 0.602
• Precision: 0.586
• Recall: 0.773
• F1: 0.666
• F1-macro: 0.612
• F1-micro: 0.62
• F1-weighted: 0.611

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: 2e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 2
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Auc	Precision	Recall	F1	F1-macro	F1-micro	F1-weighted
0.3173	0.1935	2000	1.4050	0.584	0.572	0.562	0.703	0.624	0.58	0.584	0.579
0.3027	0.3870	4000	1.6222	0.603	0.585	0.574	0.744	0.648	0.596	0.603	0.595
0.4245	0.5805	6000	1.3241	0.605	0.571	0.569	0.81	0.668	0.59	0.605	0.589
0.4055	0.7740	8000	1.2442	0.607	0.593	0.572	0.787	0.663	0.595	0.607	0.594
0.3941	0.9675	10000	1.2942	0.61	0.595	0.573	0.813	0.672	0.596	0.61	0.595
0.3418	1.1610	12000	1.4656	0.614	0.601	0.581	0.77	0.662	0.606	0.614	0.606
0.3277	1.3545	14000	1.4494	0.615	0.609	0.589	0.717	0.647	0.612	0.615	0.611
0.3228	1.5480	16000	1.4825	0.616	0.598	0.585	0.756	0.659	0.61	0.616	0.609
0.3185	1.7415	18000	1.5090	0.618	0.602	0.588	0.742	0.656	0.614	0.618	0.613
0.3098	1.9350	20000	1.5221	0.62	0.602	0.586	0.773	0.666	0.612	0.62	0.611

Framework versions

• Transformers 4.55.2
• Pytorch 2.8.0+cu126
• Datasets 4.0.0
• Tokenizers 0.21.4