Whisper Large Catalan
This model is a fine-tuned version of openai/whisper-large on the mozilla-foundation/common_voice_13_0 ca dataset. It achieves the following results on the evaluation set:
- Loss: 0.1458
- Wer: 5.0700
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: 16
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 64
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- training_steps: 20000
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.1059 | 1.02 | 1000 | 0.1744 | 7.6342 |
| 0.0159 | 3.02 | 2000 | 0.1943 | 7.3850 |
| 0.0526 | 5.02 | 3000 | 0.1899 | 6.8522 |
| 0.058 | 7.02 | 4000 | 0.1782 | 6.7802 |
| 0.0161 | 9.02 | 5000 | 0.1995 | 6.6339 |
| 0.065 | 11.02 | 6000 | 0.1563 | 6.4544 |
| 0.082 | 13.02 | 7000 | 0.1789 | 6.0309 |
| 0.0339 | 15.02 | 8000 | 0.1509 | 5.7554 |
| 0.0581 | 17.01 | 9000 | 0.1573 | 6.0446 |
| 0.0181 | 19.01 | 10000 | 0.1838 | 5.5913 |
| 0.0188 | 21.01 | 11000 | 0.1610 | 5.4804 |
| 0.0134 | 23.01 | 12000 | 0.1821 | 5.3953 |
| 0.008 | 25.01 | 13000 | 0.1748 | 5.3804 |
| 0.0071 | 27.01 | 14000 | 0.1858 | 5.4701 |
| 0.0371 | 29.01 | 15000 | 0.1610 | 5.6599 |
| 0.0076 | 31.01 | 16000 | 0.1571 | 5.1655 |
| 0.0181 | 33.01 | 17000 | 0.1449 | 5.4558 |
| 0.0522 | 35.0 | 18000 | 0.1340 | 5.8388 |
| 0.0356 | 37.0 | 19000 | 0.1458 | 5.0700 |
| 0.0132 | 39.0 | 20000 | 0.1310 | 5.1941 |
Framework versions
- Transformers 4.33.0.dev0
- Pytorch 2.0.1+cu117
- Datasets 2.14.4
- Tokenizers 0.13.3
Citation
If you use these models in your research, please cite:
@misc{dezuazo2025whisperlmimprovingasrmodels,
title={Whisper-LM: Improving ASR Models with Language Models for Low-Resource Languages},
author={Xabier de Zuazo and Eva Navas and Ibon Saratxaga and Inma Hernáez Rioja},
year={2025},
eprint={2503.23542},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2503.23542},
}
Please, check the related paper preprint in arXiv:2503.23542 for more details.
Licensing
This model is available under the Apache-2.0 License. You are free to use, modify, and distribute this model as long as you credit the original creators.