distilbert-base-uncased_ner_wikiann
This model is a fine-tuned version of distilbert-base-uncased on the wikiann dataset. It achieves the following results on the evaluation set:
- Loss: 0.2834
- Precision: 0.8139
- Recall: 0.8367
- F1: 0.8251
- Accuracy: 0.9300
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 |
|---|---|---|---|---|---|---|---|
| 0.3325 | 1.0 | 1250 | 0.2657 | 0.7732 | 0.8175 | 0.7947 | 0.9214 |
| 0.2242 | 2.0 | 2500 | 0.2505 | 0.7942 | 0.8289 | 0.8111 | 0.9262 |
| 0.158 | 3.0 | 3750 | 0.2539 | 0.8099 | 0.8367 | 0.8231 | 0.9294 |
| 0.1155 | 4.0 | 5000 | 0.2804 | 0.8172 | 0.8373 | 0.8271 | 0.9302 |
| 0.1047 | 5.0 | 6250 | 0.2834 | 0.8139 | 0.8367 | 0.8251 | 0.9300 |
Framework versions
- Transformers 4.20.1
- Pytorch 1.11.0
- Datasets 2.1.0
- Tokenizers 0.12.1