bert-finetuned-for-medical-ner

This model is a fine-tuned version of google-bert/bert-base-uncased on the PLODv2-filtered dataset. It achieves the following results on the evaluation set:

• Loss: 0.1344
• Accuracy: 0.9402
• Precision: 0.8342
• Recall: 0.9029
• F1: 0.8672

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: 2
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 2000
• num_epochs: 1
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1
0.2039	0.1420	2000	0.1873	0.9244	0.7841	0.8773	0.8281
0.1758	0.2841	4000	0.1613	0.9302	0.8028	0.8920	0.8451
0.1585	0.4261	6000	0.1499	0.9343	0.8309	0.8721	0.8510
0.1613	0.5681	8000	0.1460	0.9358	0.8400	0.8655	0.8526
0.1526	0.7101	10000	0.1402	0.9382	0.8329	0.8930	0.8619
0.1502	0.8522	12000	0.1357	0.9394	0.8346	0.8985	0.8654
0.1486	0.9942	14000	0.1344	0.9402	0.8342	0.9029	0.8672

Framework versions

• Transformers 4.51.3
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1