multilingual-roberta-base-kanuri-ner-v1
This model is a fine-tuned version of roberta-base on the Beijuka/Multilingual_PII_NER_dataset dataset. It achieves the following results on the evaluation set:
- Loss: 0.0957
- Precision: 0.9576
- Recall: 0.9302
- F1: 0.9437
- Accuracy: 0.9867
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 | 0.1120 | 0.8716 | 0.8372 | 0.8541 | 0.9691 |
| 0.187 | 2.0 | 602 | 0.0885 | 0.8735 | 0.9206 | 0.8964 | 0.9750 |
| 0.187 | 3.0 | 903 | 0.0975 | 0.8666 | 0.8911 | 0.8787 | 0.9742 |
| 0.0664 | 4.0 | 1204 | 0.0992 | 0.8715 | 0.9194 | 0.8948 | 0.9764 |
| 0.0458 | 5.0 | 1505 | 0.0900 | 0.9008 | 0.9228 | 0.9116 | 0.9767 |
| 0.0458 | 6.0 | 1806 | 0.0900 | 0.9050 | 0.9267 | 0.9157 | 0.9800 |
| 0.0311 | 7.0 | 2107 | 0.1075 | 0.8921 | 0.9328 | 0.9120 | 0.9787 |
| 0.0311 | 8.0 | 2408 | 0.1353 | 0.8920 | 0.9311 | 0.9111 | 0.9791 |
| 0.0215 | 9.0 | 2709 | 0.1167 | 0.9090 | 0.9267 | 0.9177 | 0.9792 |
| 0.0109 | 10.0 | 3010 | 0.1201 | 0.9082 | 0.9289 | 0.9184 | 0.9807 |
| 0.0109 | 11.0 | 3311 | 0.1304 | 0.9110 | 0.9272 | 0.9191 | 0.9810 |
| 0.0064 | 12.0 | 3612 | 0.1823 | 0.8918 | 0.9344 | 0.9126 | 0.9788 |
| 0.0064 | 13.0 | 3913 | 0.1507 | 0.9038 | 0.9289 | 0.9162 | 0.9803 |
| 0.0042 | 14.0 | 4214 | 0.1763 | 0.8990 | 0.935 | 0.9167 | 0.9807 |
Framework versions
- Transformers 4.55.4
- Pytorch 2.8.0+cu126
- Datasets 4.0.0
- Tokenizers 0.21.4