Whisper Small ES-CL - Roberto Castro-Vexler
This model is a fine-tuned version of openai/whisper-small on the OpenSLR Chilean Spanish dataset. It achieves the following results on the evaluation set:
- Loss: 0.1563
- Wer: 5.7685
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: 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: 500
- training_steps: 4000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.0012 | 8.6207 | 1000 | 0.1316 | 5.5476 |
| 0.0002 | 17.2414 | 2000 | 0.1478 | 5.7295 |
| 0.0001 | 25.8621 | 3000 | 0.1539 | 5.7945 |
| 0.0001 | 34.4828 | 4000 | 0.1563 | 5.7685 |
Framework versions
- Transformers 4.48.3
- Pytorch 2.8.0+cu126
- Datasets 3.6.0
- Tokenizers 0.21.4