---
library_name: transformers
language:
- ar
license: mit
base_model: openai/whisper-large-v3-turbo
tags:
- generated_from_trainer
datasets:
- fixie-ai/common_voice_17_0
- google/fleurs
- UBC-NLP/Casablanca
- deepdml/Tunisian_MSA
- ymoslem/MediaSpeech
metrics:
- wer
model-index:
- name: Whisper Turbo ar
  results:
  - task:
      name: Automatic Speech Recognition
      type: automatic-speech-recognition
    dataset:
      name: Common Voice 17.0
      type: fixie-ai/common_voice_17_0
    metrics:
    - name: Wer
      type: wer
      value: 18.89976313325132
---
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->

# Whisper Turbo ar

This model is a fine-tuned version of [openai/whisper-large-v3-turbo](https://huggingface.co/openai/whisper-large-v3-turbo) on the Common Voice 17.0 dataset.
It achieves the following results on the evaluation set:
- Loss: 0.1973
- Wer: 18.8998
- Cer: 5.0561

## 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: 16
- seed: 42
- optimizer: Use 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_ratio: 0.04
- training_steps: 18000

### Training results

| Training Loss | Epoch  | Step  | Validation Loss | Wer     | Cer    |
|:-------------:|:------:|:-----:|:---------------:|:-------:|:------:|
| 0.5358        | 0.0556 | 1000  | 0.3047          | 26.8192 | 8.1187 |
| 0.3875        | 0.1111 | 2000  | 0.2829          | 27.2654 | 7.5340 |
| 0.2546        | 0.1667 | 3000  | 0.2629          | 24.2008 | 6.7543 |
| 0.1702        | 0.2222 | 4000  | 0.2628          | 23.4884 | 6.5769 |
| 0.1075        | 0.2778 | 5000  | 0.2584          | 23.9566 | 6.6370 |
| 0.0859        | 0.3333 | 6000  | 0.2569          | 24.5221 | 6.6761 |
| 0.06          | 0.3889 | 7000  | 0.2479          | 22.1828 | 6.1018 |
| 0.0539        | 0.4444 | 8000  | 0.2461          | 22.6143 | 6.2866 |
| 0.0427        | 0.5    | 9000  | 0.2402          | 23.1083 | 6.3401 |
| 0.0341        | 0.5556 | 10000 | 0.2356          | 22.2012 | 6.0513 |
| 0.0275        | 0.6111 | 11000 | 0.2338          | 20.7378 | 5.6669 |
| 0.0204        | 0.6667 | 12000 | 0.2296          | 21.1381 | 5.7997 |
| 0.0156        | 0.7222 | 13000 | 0.2324          | 21.9037 | 5.8359 |
| 0.0162        | 0.7778 | 14000 | 0.2214          | 20.4825 | 5.5345 |
| 0.0163        | 0.8333 | 15000 | 0.2131          | 21.0426 | 5.6430 |
| 0.0127        | 0.8889 | 16000 | 0.2093          | 19.5791 | 5.2782 |
| 0.006         | 0.9444 | 17000 | 0.2083          | 19.8197 | 5.2719 |
| 0.0072        | 1.0    | 18000 | 0.1973          | 18.8998 | 5.0561 |


### Framework versions

- Transformers 4.48.0.dev0
- Pytorch 2.5.1+cu121
- Datasets 3.6.0
- Tokenizers 0.21.0

## Citation

Please cite the model using the following BibTeX entry:

```bibtex
@misc{deepdml/whisper-large-v3-turbo-ar-mix-norm,
      title={Fine-tuned Whisper turbo ASR model for speech recognition in Arabic},
      author={Jimenez, David},
      howpublished={\url{https://huggingface.co/deepdml/whisper-large-v3-turbo-ar-mix-norm}},
      year={2026}
    }
```