microsoft-deberta-v3-large_ner_wnut_17
This model is a fine-tuned version of microsoft/deberta-v3-large on the wnut_17 dataset. It achieves the following results on the evaluation set:
- Loss: 0.2199
- Precision: 0.7671
- Recall: 0.6184
- F1: 0.6848
- Accuracy: 0.9667
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 |
|---|---|---|---|---|---|---|---|
| No log | 1.0 | 213 | 0.1751 | 0.6884 | 0.5682 | 0.6225 | 0.9601 |
| No log | 2.0 | 426 | 0.1702 | 0.7351 | 0.6208 | 0.6732 | 0.9655 |
| 0.1003 | 3.0 | 639 | 0.1954 | 0.7360 | 0.6136 | 0.6693 | 0.9656 |
| 0.1003 | 4.0 | 852 | 0.2113 | 0.7595 | 0.6232 | 0.6846 | 0.9669 |
| 0.015 | 5.0 | 1065 | 0.2199 | 0.7671 | 0.6184 | 0.6848 | 0.9667 |
Framework versions
- Transformers 4.20.1
- Pytorch 1.11.0
- Datasets 2.1.0
- Tokenizers 0.12.1