whisper-small-yoruba
This model is a fine-tuned version of openai/whisper-small on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.1768
- Wer: 0.1126
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: 32
- eval_batch_size: 16
- 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
- lr_scheduler_warmup_steps: 500
- num_epochs: 8
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.3379 | 0.8989 | 2000 | 0.3092 | 0.1718 |
| 0.1838 | 1.7978 | 4000 | 0.2349 | 0.1526 |
| 0.1078 | 2.6966 | 6000 | 0.2006 | 0.1321 |
| 0.0410 | 3.5955 | 8000 | 0.1795 | 0.1268 |
| 0.0165 | 4.4944 | 10000 | 0.1808 | 0.1233 |
| 0.0071 | 5.3933 | 12000 | 0.1715 | 0.1178 |
| 0.0029 | 6.2921 | 14000 | 0.1719 | 0.1148 |
| 0.0018 | 7.1910 | 16000 | 0.1762 | 0.1123 |
| 0.0015 | 8.0 | 17800 | 0.1768 | 0.1126 |
Framework versions
- Transformers 5.12.0
- Pytorch 2.8.0+cu128
- Datasets 5.0.0
- Tokenizers 0.22.2