Whisper Small Bisaya
This model is a fine-tuned version of openai/whisper-small on the Bisaya Speech Corpus (private) dataset. It achieves the following results on the evaluation set:
- Loss: 0.7840
- Wer: 44.5199
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: 16
- eval_batch_size: 16
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- 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: 5
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 5.8351 | 1.0 | 12 | 1.0625 | 61.7408 |
| 5.7016 | 2.0 | 24 | 0.9944 | 56.6508 |
| 5.1602 | 3.0 | 36 | 0.9169 | 51.2894 |
| 4.8167 | 4.0 | 48 | 0.8472 | 48.4221 |
| 4.2978 | 5.0 | 60 | 0.7840 | 44.5199 |
Framework versions
- Transformers 5.15.1
- Pytorch 2.13.0+cu130
- Datasets 3.6.0
- Tokenizers 0.22.2