Whisper Small En Medimix
This model is a fine-tuned version of openai/whisper-small on the cutsom_whatsapp_audio dataset. It achieves the following results on the evaluation set:
- Loss: 0.5352
- Wer: 12.2137
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: 8
- seed: 42
- 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: linear
- lr_scheduler_warmup_steps: 200
- training_steps: 1000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.0004 | 66.6667 | 200 | 0.4815 | 11.8321 |
| 0.0001 | 133.3333 | 400 | 0.5102 | 11.4504 |
| 0.0001 | 200.0 | 600 | 0.5246 | 11.4504 |
| 0.0001 | 266.6667 | 800 | 0.5323 | 12.2137 |
| 0.0001 | 333.3333 | 1000 | 0.5352 | 12.2137 |
Framework versions
- Transformers 4.47.1
- Pytorch 2.5.1+cu124
- Tokenizers 0.21.0