bert-base-medmentions

This model is a fine-tuned version of bert-base-uncased on the Ben10x/MedMentions-NER dataset. It achieves the following results on the evaluation set:

• Loss: 1.5156
• Precision: 0.5821
• Recall: 0.6344
• F1: 0.6071
• Accuracy: 0.8689

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-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Use OptimizerNames.ADAFACTOR and the args are: No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 15.0
• label_smoothing_factor: 0.2

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
1.5686	1.0	2911	1.5440	0.5246	0.6123	0.5650	0.8550
1.4792	2.0	5822	1.5156	0.5821	0.6344	0.6071	0.8689
1.4111	3.0	8733	1.5191	0.5865	0.6494	0.6163	0.8714
1.356	4.0	11644	1.5293	0.6236	0.6403	0.6318	0.8777
1.3182	5.0	14555	1.5433	0.6283	0.6426	0.6354	0.8789
1.2919	6.0	17466	1.5671	0.6242	0.6628	0.6429	0.8794
1.2743	7.0	20377	1.5697	0.6356	0.6574	0.6463	0.8809
1.2633	8.0	23288	1.5806	0.6364	0.6699	0.6528	0.8813
1.2542	9.0	26199	1.5942	0.6278	0.6734	0.6498	0.8808
1.2457	10.0	29110	1.6076	0.6372	0.6634	0.6500	0.8814
1.2398	11.0	32021	1.6077	0.6414	0.6696	0.6552	0.8835
1.2377	12.0	34932	1.6135	0.6478	0.6759	0.6615	0.8847
1.2349	13.0	37843	1.6195	0.6433	0.6756	0.6590	0.8839
1.2328	14.0	40754	1.6228	0.6462	0.6726	0.6592	0.8845
1.231	15.0	43665	1.6247	0.6473	0.6735	0.6601	0.8847

Framework versions

• Transformers 4.50.3
• Pytorch 2.6.0+cu124
• Datasets 3.5.0
• Tokenizers 0.21.1