multilingual-google-bert/bert-base-multilingual-cased-kanuri-ner-v1
This model is a fine-tuned version of google-bert/bert-base-multilingual-cased on the Beijuka/Multilingual_PII_NER_dataset dataset. It achieves the following results on the evaluation set:
- Loss: 0.1582
- Precision: 0.9257
- Recall: 0.9368
- F1: 0.9312
- Accuracy: 0.9764
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 | 301 | 0.1495 | 0.8764 | 0.8390 | 0.8573 | 0.9552 |
| 0.2147 | 2.0 | 602 | 0.1440 | 0.8558 | 0.9372 | 0.8947 | 0.9603 |
| 0.2147 | 3.0 | 903 | 0.1117 | 0.8943 | 0.9403 | 0.9168 | 0.9686 |
| 0.0879 | 4.0 | 1204 | 0.1529 | 0.875 | 0.9506 | 0.9112 | 0.9670 |
| 0.0566 | 5.0 | 1505 | 0.1413 | 0.9036 | 0.9306 | 0.9169 | 0.9694 |
| 0.0566 | 6.0 | 1806 | 0.1745 | 0.8961 | 0.9228 | 0.9093 | 0.9668 |
| 0.0349 | 7.0 | 2107 | 0.1738 | 0.9021 | 0.9342 | 0.9179 | 0.9699 |
| 0.0349 | 8.0 | 2408 | 0.2014 | 0.8914 | 0.9372 | 0.9137 | 0.9679 |
| 0.0201 | 9.0 | 2709 | 0.1736 | 0.9071 | 0.9444 | 0.9254 | 0.9727 |
| 0.0115 | 10.0 | 3010 | 0.1846 | 0.8935 | 0.9280 | 0.9104 | 0.9675 |
| 0.0115 | 11.0 | 3311 | 0.2373 | 0.8895 | 0.9403 | 0.9142 | 0.9681 |
| 0.0073 | 12.0 | 3612 | 0.2186 | 0.9027 | 0.9501 | 0.9258 | 0.9721 |
| 0.0073 | 13.0 | 3913 | 0.2325 | 0.9015 | 0.9372 | 0.9190 | 0.9686 |
| 0.0035 | 14.0 | 4214 | 0.2364 | 0.9028 | 0.9465 | 0.9242 | 0.9710 |
| 0.0016 | 15.0 | 4515 | 0.2366 | 0.9004 | 0.9486 | 0.9238 | 0.9708 |
Framework versions
- Transformers 4.55.4
- Pytorch 2.8.0+cu126
- Datasets 4.0.0
- Tokenizers 0.21.4