BC5CDR_SciBERT_NER
This model is a fine-tuned version of allenai/scibert_scivocab_uncased on the None dataset. It achieves the following results on the evaluation set:
-
Loss: 0.0818
-
Seqeval classification report: precision recall f1-score support
Chemical 0.92 0.94 0.93 7079 Disease 0.98 0.98 0.98 103426
micro avg 0.98 0.97 0.98 110505 macro avg 0.95 0.96 0.95 110505 weighted avg 0.98 0.97 0.98 110505
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 | 0.0891 | precision recall f1-score support |
Chemical 0.91 0.93 0.92 7079
Disease 0.98 0.97 0.97 103426
micro avg 0.97 0.97 0.97 110505 macro avg 0.94 0.95 0.95 110505 weighted avg 0.97 0.97 0.97 110505 | | No log | 2.0 | 286 | 0.0830 | precision recall f1-score support
Chemical 0.93 0.93 0.93 7079
Disease 0.98 0.97 0.98 103426
micro avg 0.98 0.97 0.97 110505 macro avg 0.96 0.95 0.95 110505 weighted avg 0.98 0.97 0.97 110505 | | No log | 3.0 | 429 | 0.0818 | precision recall f1-score support
Chemical 0.92 0.94 0.93 7079
Disease 0.98 0.98 0.98 103426
micro avg 0.98 0.97 0.98 110505 macro avg 0.95 0.96 0.95 110505 weighted avg 0.98 0.97 0.98 110505 |
Framework versions
- Transformers 4.35.2
- Pytorch 2.1.0+cu121
- Datasets 2.15.0
- Tokenizers 0.15.0