Uzbek NER model
This model is a fine-tuned version of FacebookAI/xlm-roberta-large on the Uzbek Ner dataset. It achieves the following results on the evaluation set:
- Loss: 0.1421
- Precision: 0.6071
- Recall: 0.6482
- F1: 0.6270
- Accuracy: 0.9486
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: 2e-05
- train_batch_size: 32
- eval_batch_size: 32
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 64
- optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 3
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy |
|---|---|---|---|---|---|---|---|
| 0.1671 | 0.5758 | 150 | 0.1632 | 0.5260 | 0.6425 | 0.5785 | 0.9402 |
| 0.1453 | 1.1497 | 300 | 0.1481 | 0.5935 | 0.6191 | 0.6061 | 0.9467 |
| 0.134 | 1.7255 | 450 | 0.1449 | 0.5936 | 0.6216 | 0.6073 | 0.9480 |
| 0.1273 | 2.2994 | 600 | 0.1413 | 0.6217 | 0.6262 | 0.6239 | 0.9493 |
| 0.1258 | 2.8752 | 750 | 0.1421 | 0.6071 | 0.6482 | 0.6270 | 0.9486 |
Framework versions
- Transformers 4.47.0
- Pytorch 2.1.1+cu121
- Datasets 3.2.0
- Tokenizers 0.21.0