Whisper Small N- Augmented
This model is a fine-tuned version of openai/whisper-small on the N Demo Final (Augmented) dataset. It achieves the following results on the evaluation set:
- Loss: 0.5186
- Wer: 52.4803
- Cer: 23.2548
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: cosine
- lr_scheduler_warmup_steps: 0.1
- num_epochs: 15
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 1.4115 | 0.9965 | 285 | 0.6066 | 75.3937 | 27.9735 |
| 1.0703 | 1.9930 | 570 | 0.4821 | 64.8031 | 22.3183 |
| 0.8049 | 2.9895 | 855 | 0.4383 | 60.8268 | 23.3196 |
| 0.6565 | 3.9860 | 1140 | 0.4189 | 55.0787 | 19.8473 |
| 0.5014 | 4.9825 | 1425 | 0.4191 | 52.5591 | 18.8171 |
| 0.3444 | 5.9790 | 1710 | 0.4216 | 52.0472 | 19.9769 |
| 0.2355 | 6.9755 | 1995 | 0.4462 | 49.3307 | 18.1975 |
| 0.1876 | 7.9720 | 2280 | 0.4599 | 51.3780 | 19.2061 |
| 0.1180 | 8.9685 | 2565 | 0.4719 | 49.0157 | 17.7941 |
| 0.0884 | 9.9650 | 2850 | 0.4888 | 49.4094 | 18.0895 |
| 0.0712 | 10.9615 | 3135 | 0.5017 | 48.6220 | 17.7509 |
| 0.0525 | 11.9580 | 3420 | 0.5097 | 48.4252 | 17.5420 |
| 0.0477 | 12.9545 | 3705 | 0.5152 | 49.1732 | 18.4353 |
| 0.0469 | 13.9510 | 3990 | 0.5191 | 51.9291 | 23.0891 |
| 0.0473 | 14.9476 | 4275 | 0.5186 | 52.4803 | 23.2548 |
Framework versions
- Transformers 5.2.0
- Pytorch 2.9.0+cu128
- Datasets 4.5.0
- Tokenizers 0.22.2