bert-large-uncased_ner_wikiann
This model is a fine-tuned version of bert-large-uncased on the wikiann dataset. It achieves the following results on the evaluation set:
- Loss: 0.3373
- Precision: 0.8384
- Recall: 0.8609
- F1: 0.8495
- Accuracy: 0.9379
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: cosine
- num_epochs: 5
Training results
| Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy |
|---|---|---|---|---|---|---|---|
| 0.3146 | 1.0 | 1250 | 0.2545 | 0.7956 | 0.8372 | 0.8159 | 0.9285 |
| 0.1973 | 2.0 | 2500 | 0.2438 | 0.8267 | 0.8546 | 0.8404 | 0.9349 |
| 0.1181 | 3.0 | 3750 | 0.2637 | 0.8320 | 0.8588 | 0.8452 | 0.9374 |
| 0.0647 | 4.0 | 5000 | 0.3175 | 0.8389 | 0.8627 | 0.8507 | 0.9387 |
| 0.0443 | 5.0 | 6250 | 0.3373 | 0.8384 | 0.8609 | 0.8495 | 0.9379 |
Framework versions
- Transformers 4.20.1
- Pytorch 1.11.0
- Datasets 2.1.0
- Tokenizers 0.12.1