Whisper Tiny Basque
This model is a fine-tuned version of openai/whisper-tiny on the mozilla-foundation/common_voice_17_0 eu dataset. It achieves the following results on the evaluation set:
- Loss: 0.4350
- Wer: 20.1244
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: 3.75e-05
- train_batch_size: 256
- eval_batch_size: 128
- 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: 40000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.0348 | 9.3458 | 1000 | 0.3382 | 22.7152 |
| 0.0021 | 18.6916 | 2000 | 0.4092 | 21.7844 |
| 0.0009 | 28.0374 | 3000 | 0.4509 | 21.9026 |
| 0.0023 | 37.3832 | 4000 | 0.4062 | 20.7181 |
| 0.0003 | 46.7290 | 5000 | 0.4350 | 20.1244 |
| 0.0002 | 56.0748 | 6000 | 0.4546 | 20.1702 |
| 0.0001 | 65.4206 | 7000 | 0.4745 | 20.2179 |
| 0.0001 | 74.7664 | 8000 | 0.4941 | 20.1995 |
| 0.0 | 84.1121 | 9000 | 0.5142 | 20.3342 |
| 0.0 | 93.4579 | 10000 | 0.5353 | 20.4386 |
| 0.0 | 102.8037 | 11000 | 0.5567 | 20.5495 |
| 0.0 | 112.1495 | 12000 | 0.5788 | 20.6100 |
| 0.0 | 121.4953 | 13000 | 0.6023 | 20.6988 |
| 0.0 | 130.8411 | 14000 | 0.6253 | 20.7767 |
| 0.0 | 140.1869 | 15000 | 0.6486 | 20.8683 |
Framework versions
- Transformers 4.52.3
- Pytorch 2.6.0+cu124
- Datasets 3.6.0
- Tokenizers 0.21.1