multilingual-microsoft/deberta-v3-base-lumasaba-ner-v1
This model is a fine-tuned version of microsoft/deberta-v3-base on the Beijuka/Multilingual_PII_NER_dataset dataset. It achieves the following results on the evaluation set:
- Loss: 0.3746
- Precision: 0.9802
- Recall: 0.9459
- F1: 0.9627
- Accuracy: 0.9529
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: 5e-05
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- num_epochs: 20
Training results
| Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy |
|---|---|---|---|---|---|---|---|
| No log | 1.0 | 398 | 0.7058 | 0.7983 | 0.7704 | 0.7841 | 0.7637 |
| 1.0932 | 2.0 | 796 | 0.4247 | 0.8727 | 0.8933 | 0.8829 | 0.8807 |
| 0.3981 | 3.0 | 1194 | 0.4242 | 0.8830 | 0.9218 | 0.9020 | 0.9055 |
| 0.2187 | 4.0 | 1592 | 0.4187 | 0.9194 | 0.9194 | 0.9194 | 0.9190 |
| 0.2187 | 5.0 | 1990 | 0.3810 | 0.9433 | 0.9487 | 0.9460 | 0.9383 |
| 0.108 | 6.0 | 2388 | 0.4557 | 0.9701 | 0.9251 | 0.9471 | 0.9338 |
| 0.0769 | 7.0 | 2786 | 0.4815 | 0.9330 | 0.9406 | 0.9367 | 0.9293 |
| 0.0401 | 8.0 | 3184 | 0.4978 | 0.9602 | 0.9430 | 0.9515 | 0.9401 |
| 0.0384 | 9.0 | 3582 | 0.5352 | 0.9437 | 0.9422 | 0.9430 | 0.9356 |
| 0.0384 | 10.0 | 3980 | 0.5006 | 0.9436 | 0.9536 | 0.9486 | 0.9374 |
| 0.0181 | 11.0 | 4378 | 0.5544 | 0.9481 | 0.9528 | 0.9504 | 0.9388 |
Framework versions
- Transformers 4.55.4
- Pytorch 2.8.0+cu126
- Datasets 4.0.0
- Tokenizers 0.21.4