whisper-small-nl-vl
This model is a fine-tuned version of openai/whisper-small on the flemish-mozilla-common-voice dataset. It achieves the following results on the evaluation set:
- Loss: 0.1278
- Wer: 6.1734
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: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- training_steps: 5000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.0579 | 1.33 | 1000 | 0.1254 | 7.7204 |
| 0.0235 | 2.67 | 2000 | 0.1159 | 6.5565 |
| 0.0124 | 4.0 | 3000 | 0.1209 | 6.4843 |
| 0.0019 | 5.33 | 4000 | 0.1249 | 6.2891 |
| 0.0012 | 6.67 | 5000 | 0.1278 | 6.1734 |
Framework versions
- Transformers 4.39.3
- Pytorch 2.2.2+cu121
- Datasets 2.18.0
- Tokenizers 0.15.2