BC5CDR_PubMedBERT_NER
This model is a fine-tuned version of microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext on the None dataset. It achieves the following results on the evaluation set:
-
Loss: 0.0783
-
Seqeval classification report: precision recall f1-score support
Chemical 0.99 0.98 0.98 103336 Disease 0.76 0.86 0.81 3447
micro avg 0.98 0.98 0.98 106783 macro avg 0.87 0.92 0.89 106783 weighted avg 0.98 0.98 0.98 106783
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
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 3
Training results
| Training Loss | Epoch | Step | Validation Loss | Seqeval classification report |
|---|---|---|---|---|
| No log | 1.0 | 143 | 0.0952 | precision recall f1-score support |
Chemical 0.99 0.97 0.98 103336
Disease 0.68 0.88 0.76 3447
micro avg 0.97 0.97 0.97 106783 macro avg 0.83 0.92 0.87 106783 weighted avg 0.98 0.97 0.97 106783 | | No log | 2.0 | 286 | 0.0804 | precision recall f1-score support
Chemical 0.99 0.98 0.98 103336
Disease 0.75 0.86 0.80 3447
micro avg 0.98 0.97 0.97 106783 macro avg 0.87 0.92 0.89 106783 weighted avg 0.98 0.97 0.98 106783 | | No log | 3.0 | 429 | 0.0783 | precision recall f1-score support
Chemical 0.99 0.98 0.98 103336
Disease 0.76 0.86 0.81 3447
micro avg 0.98 0.98 0.98 106783 macro avg 0.87 0.92 0.89 106783 weighted avg 0.98 0.98 0.98 106783 |
Framework versions
- Transformers 4.35.2
- Pytorch 2.1.0+cu118
- Datasets 2.15.0
- Tokenizers 0.15.0