biobert-base-cased-v1.2-finetuned-ner-CRAFT_English
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: 0.1614
- Precision: 0.8585
- Recall: 0.8623
- F1: 0.8604
- Accuracy: 0.9724
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: 3e-05
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 4
Training results
| Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy |
|---|---|---|---|---|---|---|---|
| 0.0725 | 1.0 | 1360 | 0.1242 | 0.8090 | 0.8698 | 0.8383 | 0.9681 |
| 0.0281 | 2.0 | 2720 | 0.1541 | 0.8497 | 0.8549 | 0.8523 | 0.9705 |
| 0.0162 | 3.0 | 4080 | 0.1510 | 0.8390 | 0.8681 | 0.8533 | 0.9711 |
| 0.0053 | 4.0 | 5440 | 0.1614 | 0.8585 | 0.8623 | 0.8604 | 0.9724 |
Framework versions
- Transformers 4.17.0
- Pytorch 1.10.0+cu111
- Datasets 1.18.4
- Tokenizers 0.11.6