multilingual-Davlan/afro-xlmr-base-luganda-ner-v1
This model is a fine-tuned version of Davlan/afro-xlmr-base on the Beijuka/Multilingual_PII_NER_dataset dataset. It achieves the following results on the evaluation set:
- Loss: 0.3035
- Precision: 0.8380
- Recall: 0.8267
- F1: 0.8323
- Accuracy: 0.9570
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 | 261 | 0.3247 | 0.6104 | 0.5574 | 0.5827 | 0.9102 |
| 0.469 | 2.0 | 522 | 0.2311 | 0.7543 | 0.7239 | 0.7388 | 0.9381 |
| 0.469 | 3.0 | 783 | 0.2044 | 0.8267 | 0.7249 | 0.7725 | 0.9464 |
| 0.1333 | 4.0 | 1044 | 0.1721 | 0.8119 | 0.8035 | 0.8077 | 0.9577 |
| 0.1333 | 5.0 | 1305 | 0.2471 | 0.7674 | 0.8087 | 0.7875 | 0.9467 |
| 0.0634 | 6.0 | 1566 | 0.2065 | 0.8111 | 0.8304 | 0.8206 | 0.9585 |
| 0.0634 | 7.0 | 1827 | 0.2330 | 0.8451 | 0.8066 | 0.8254 | 0.9595 |
| 0.0328 | 8.0 | 2088 | 0.2440 | 0.8341 | 0.8004 | 0.8169 | 0.9580 |
| 0.0328 | 9.0 | 2349 | 0.2671 | 0.8383 | 0.8201 | 0.8291 | 0.9563 |
| 0.0191 | 10.0 | 2610 | 0.2515 | 0.8337 | 0.8294 | 0.8315 | 0.9590 |
| 0.0191 | 11.0 | 2871 | 0.2451 | 0.8477 | 0.8118 | 0.8294 | 0.9604 |
| 0.0095 | 12.0 | 3132 | 0.2637 | 0.8550 | 0.8232 | 0.8388 | 0.9624 |
| 0.0095 | 13.0 | 3393 | 0.2708 | 0.8425 | 0.8242 | 0.8332 | 0.9624 |
| 0.0068 | 14.0 | 3654 | 0.2922 | 0.8488 | 0.8128 | 0.8304 | 0.9594 |
| 0.0068 | 15.0 | 3915 | 0.2966 | 0.8303 | 0.8046 | 0.8172 | 0.9594 |
Framework versions
- Transformers 4.55.4
- Pytorch 2.8.0+cu126
- Datasets 4.0.0
- Tokenizers 0.21.4