<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->
BioBERT
This model is a fine-tuned version of dmis-lab/biobert-v1.1 on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.9310
- Accuracy: 0.79
- Precision: 0.8730
- Recall: 0.8088
- F1: 0.8397
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: 16
- eval_batch_size: 16
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy | Precision | Recall | F1 |
|---|---|---|---|---|---|---|---|
| 0.6755 | 1.0 | 50 | 0.6608 | 0.47 | 0.9412 | 0.2353 | 0.3765 |
| 0.6097 | 2.0 | 100 | 0.5785 | 0.68 | 0.8214 | 0.6765 | 0.7419 |
| 0.5123 | 3.0 | 150 | 0.5240 | 0.7 | 0.9318 | 0.6029 | 0.7321 |
| 0.3547 | 4.0 | 200 | 0.4475 | 0.8 | 0.9138 | 0.7794 | 0.8413 |
| 0.2413 | 5.0 | 250 | 0.5033 | 0.81 | 0.9153 | 0.7941 | 0.8504 |
| 0.1398 | 6.0 | 300 | 0.6918 | 0.8 | 0.8636 | 0.8382 | 0.8507 |
| 0.0995 | 7.0 | 350 | 0.7694 | 0.79 | 0.8730 | 0.8088 | 0.8397 |
| 0.0636 | 8.0 | 400 | 0.8876 | 0.84 | 0.8824 | 0.8824 | 0.8824 |
| 0.0512 | 9.0 | 450 | 0.9019 | 0.8 | 0.875 | 0.8235 | 0.8485 |
| 0.0359 | 10.0 | 500 | 0.9310 | 0.79 | 0.8730 | 0.8088 | 0.8397 |
Framework versions
- Transformers 4.21.2
- Pytorch 1.12.1+cu113
- Datasets 2.4.0
- Tokenizers 0.12.1
