Whisper Small Uzbek
This model is a fine-tuned version of openai/whisper-small on the Common Voice 17.0 dataset. It achieves the following results on the evaluation set:
- Loss: 0.3776
- Wer: 35.8660
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: 16
- eval_batch_size: 16
- seed: 42
- optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 1500
- training_steps: 5500
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.913 | 0.2 | 500 | 0.8213 | 62.5843 |
| 0.6404 | 0.4 | 1000 | 0.6082 | 51.8716 |
| 0.5734 | 0.6 | 1500 | 0.5458 | 48.0513 |
| 0.5051 | 0.8 | 2000 | 0.4846 | 43.8649 |
| 0.4407 | 1.0 | 2500 | 0.4483 | 41.3901 |
| 0.3436 | 1.2 | 3000 | 0.4321 | 41.0277 |
| 0.3092 | 1.4 | 3500 | 0.4184 | 40.1141 |
| 0.2861 | 1.6 | 4000 | 0.4091 | 39.9753 |
| 0.289 | 1.8 | 4500 | 0.3811 | 36.7950 |
| 0.2816 | 2.0 | 5000 | 0.3730 | 36.7102 |
| 0.1547 | 2.2 | 5500 | 0.3776 | 35.8660 |
Framework versions
- Transformers 4.47.0
- Pytorch 2.1.1+cu121
- Datasets 3.2.0
- Tokenizers 0.21.0