multilingual-google-bert/bert-base-multilingual-cased-lumasaba-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: 0.3203
- Precision: 0.9705
- Recall: 0.9529
- F1: 0.9616
- Accuracy: 0.9604
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 | 0.6266 | 0.8401 | 0.8225 | 0.8312 | 0.8062 |
| 1.0576 | 2.0 | 796 | 0.3751 | 0.9033 | 0.8891 | 0.8962 | 0.8859 |
| 0.3626 | 3.0 | 1194 | 0.3664 | 0.9336 | 0.9273 | 0.9305 | 0.9163 |
| 0.1629 | 4.0 | 1592 | 0.4134 | 0.9381 | 0.9303 | 0.9342 | 0.9244 |
| 0.1629 | 5.0 | 1990 | 0.3573 | 0.9497 | 0.9476 | 0.9486 | 0.9417 |
| 0.0925 | 6.0 | 2388 | 0.4060 | 0.9501 | 0.9416 | 0.9458 | 0.9434 |
| 0.0516 | 7.0 | 2786 | 0.3767 | 0.9371 | 0.9483 | 0.9427 | 0.9377 |
| 0.0409 | 8.0 | 3184 | 0.4152 | 0.9450 | 0.9528 | 0.9489 | 0.9409 |
| 0.0389 | 9.0 | 3582 | 0.3901 | 0.9624 | 0.9386 | 0.9503 | 0.9458 |
| 0.0389 | 10.0 | 3980 | 0.4474 | 0.9388 | 0.9536 | 0.9461 | 0.9426 |
| 0.0212 | 11.0 | 4378 | 0.3165 | 0.9591 | 0.9663 | 0.9627 | 0.9547 |
| 0.0167 | 12.0 | 4776 | 0.3941 | 0.9590 | 0.9633 | 0.9611 | 0.9543 |
| 0.0199 | 13.0 | 5174 | 0.4243 | 0.9496 | 0.9588 | 0.9542 | 0.9478 |
| 0.0156 | 14.0 | 5572 | 0.4842 | 0.9539 | 0.9618 | 0.9579 | 0.9494 |
Framework versions
- Transformers 4.55.4
- Pytorch 2.8.0+cu126
- Datasets 4.0.0
- Tokenizers 0.21.4