whisper-small
This model is a fine-tuned version of openai/whisper-small on the fleurs dataset. It achieves the following results on the evaluation set:
- Loss: 0.3925
- Wer Strict(with punctuation): 13.6333
- Wer Ortho: 8.6149
- Punctuation Penalty: 5.0184
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: 4
- total_train_batch_size: 64
- optimizer: Use OptimizerNames.ADAMW_TORCH 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: 20
- training_steps: 300
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer Strict(with punctuation) | Wer Ortho | Punctuation Penalty |
|---|---|---|---|---|---|---|
| 0.8871 | 1.2346 | 50 | 0.3234 | 12.9390 | 7.8269 | 5.1121 |
| 0.2165 | 2.4691 | 100 | 0.3421 | 13.4197 | 8.3745 | 5.0452 |
| 0.0508 | 3.7037 | 150 | 0.3664 | 13.5132 | 8.4279 | 5.0852 |
| 0.0203 | 4.9383 | 200 | 0.3925 | 13.6333 | 8.6149 | 5.0184 |
Framework versions
- Transformers 4.46.0
- Pytorch 2.10.0+cu128
- Datasets 3.5.0
- Tokenizers 0.20.3