Whisper Small Welsh - Robust
This model is a fine-tuned version of openai/whisper-small on the mozilla-foundation/common_voice_11_0 cy dataset. It achieves the following results on the evaluation set:
- Loss: 0.4569
- Wer: 23.0736
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: 5e-05
- train_batch_size: 64
- eval_batch_size: 32
- 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 |
|---|---|---|---|---|
| 0.0597 | 6.1 | 1000 | 0.4690 | 29.4666 |
| 0.0107 | 12.2 | 2000 | 0.4707 | 26.2671 |
| 0.0026 | 18.29 | 3000 | 0.4643 | 24.6763 |
| 0.0007 | 24.39 | 4000 | 0.4629 | 23.8024 |
| 0.0004 | 30.49 | 5000 | 0.4610 | 23.0616 |
Framework versions
- Transformers 4.26.0.dev0
- Pytorch 1.13.1+cu117
- Datasets 2.8.0
- Tokenizers 0.13.2
- Downloads last month
- 4
Inference Providers
NEW
This model is not currently available via any of the supported third-party Inference Providers, and
the model is not deployed on the HF Inference API.
Dataset used to train ALM/whisper-cy-small-augmented
Evaluation results
- Wer on mozilla-foundation/common_voice_11_0 cytest set self-reported23.074
- WER on mozilla-foundation/common_voice_11_0 cytest set self-reported20.330