EYEDOL/whisper-tiny-hausa
This model is a fine-tuned version of EYEDOL/whisper-tiny-hausa on the EYEDOL/naija-voices-hausa-split_0-1 dataset. It achieves the following results on the evaluation set:
- Loss: 0.7462
- Wer Ortho: 0.5629
- Wer: 0.5106
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: constant_with_warmup
- lr_scheduler_warmup_steps: 0.1
- num_epochs: 12
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer Ortho | Wer |
|---|---|---|---|---|---|
| 1.4906 | 1.0 | 665 | 0.7049 | 0.5690 | 0.5170 |
| 1.3724 | 2.0 | 1330 | 0.6859 | 0.5655 | 0.5104 |
| 1.2209 | 3.0 | 1995 | 0.6762 | 0.5457 | 0.4862 |
| 1.1009 | 4.0 | 2660 | 0.6690 | 0.5565 | 0.4982 |
| 0.9961 | 5.0 | 3325 | 0.6663 | 0.5415 | 0.4836 |
| 0.9033 | 6.0 | 3990 | 0.6694 | 0.5448 | 0.4877 |
| 0.8164 | 7.0 | 4655 | 0.6788 | 0.5634 | 0.5001 |
| 0.7366 | 8.0 | 5320 | 0.6853 | 0.5382 | 0.4875 |
| 0.6609 | 9.0 | 5985 | 0.7018 | 0.5691 | 0.5168 |
| 0.5888 | 10.0 | 6650 | 0.7114 | 0.5560 | 0.5005 |
| 0.5208 | 11.0 | 7315 | 0.7292 | 0.5602 | 0.5055 |
| 0.4565 | 12.0 | 7980 | 0.7462 | 0.5629 | 0.5106 |
Framework versions
- Transformers 5.0.0
- Pytorch 2.10.0+cu128
- Datasets 4.8.3
- Tokenizers 0.22.2