Whisper Small Uzbek
This model is a fine-tuned version of openai/whisper-small on the Uzbek Speech Dataset dataset. It achieves the following results on the evaluation set:
- Loss: 1.1297
- Wer: 58.6128
- Cer: 19.2069
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: 1e-05
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- 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
- lr_scheduler_warmup_steps: 50
- training_steps: 600
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 0.1019 | 18.8 | 150 | 1.0722 | 60.8869 | 20.3323 |
| 0.0043 | 37.5333 | 300 | 1.0996 | 58.6697 | 19.1012 |
| 0.0016 | 56.2667 | 450 | 1.1218 | 58.3854 | 18.9577 |
| 0.0013 | 75.0 | 600 | 1.1297 | 58.6128 | 19.2069 |
Framework versions
- Transformers 5.20.1
- Pytorch 2.11.0+cu130
- Datasets 4.8.5
- Tokenizers 0.23.3