EYEDOL/whisper-tiny-yoruba1
This model is a fine-tuned version of EYEDOL/whisper-tiny-yoruba on the EYEDOL/naija-voices-yoruba-split_0-4 dataset. It achieves the following results on the evaluation set:
- Loss: 0.8130
- Wer Ortho: 0.7657
- Wer: 0.6755
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 |
|---|---|---|---|---|---|
| 1.5706 | 1.0 | 583 | 0.7699 | 0.8067 | 0.7173 |
| 1.4778 | 2.0 | 1166 | 0.7536 | 0.7432 | 0.6515 |
| 1.3621 | 3.0 | 1749 | 0.7465 | 0.7352 | 0.6440 |
| 1.2687 | 4.0 | 2332 | 0.7418 | 0.7670 | 0.6748 |
| 1.1844 | 5.0 | 2915 | 0.7442 | 0.7682 | 0.6766 |
| 1.1091 | 6.0 | 3498 | 0.7471 | 0.7690 | 0.6733 |
| 1.0363 | 7.0 | 4081 | 0.7512 | 0.7385 | 0.6481 |
| 0.9677 | 8.0 | 4664 | 0.7607 | 0.7673 | 0.6829 |
| 0.9016 | 9.0 | 5247 | 0.7728 | 0.7618 | 0.6768 |
| 0.8397 | 10.0 | 5830 | 0.7889 | 0.7395 | 0.6530 |
| 0.7784 | 11.0 | 6413 | 0.7975 | 0.7635 | 0.6761 |
| 0.7197 | 12.0 | 6996 | 0.8130 | 0.7657 | 0.6755 |
Framework versions
- Transformers 5.0.0
- Pytorch 2.10.0+cu128
- Datasets 4.8.3
- Tokenizers 0.22.2