---
library_name: transformers
language:
- ar
license: apache-2.0
base_model: openai/whisper-base
tags:
- generated_from_trainer
datasets:
- google/fleurs
- fixie-ai/common_voice_17_0
- UBC-NLP/Casablanca
- ymoslem/MediaSpeech
- deepdml/Tunisian_MSA
metrics:
- wer
model-index:
- name: Whisper Base ar
  results:
  - task:
      name: Automatic Speech Recognition
      type: automatic-speech-recognition
    dataset:
      name: Common Voice 17.0
      type: google/fleurs
    metrics:
    - name: Wer
      type: wer
      value: 40.550118433374344
---
<!-- 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 Base ar

This model is a fine-tuned version of [openai/whisper-base](https://huggingface.co/openai/whisper-base) on the Common Voice 17.0 dataset.
It achieves the following results on the evaluation set:
- Loss: 0.5179
- Wer: 40.5501
- Cer: 13.2382

## 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: 64
- eval_batch_size: 64
- 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     |
|:-------------:|:------:|:-----:|:---------------:|:-------:|:-------:|
| 0.7397        | 0.0556 | 1000  | 0.6305          | 54.8668 | 18.9365 |
| 0.3962        | 0.1111 | 2000  | 0.5805          | 50.5793 | 16.9481 |
| 0.1913        | 0.1667 | 3000  | 0.5593          | 48.8019 | 16.2853 |
| 0.1031        | 0.2222 | 4000  | 0.5390          | 46.7766 | 15.6262 |
| 0.0743        | 0.2778 | 5000  | 0.5193          | 46.1321 | 15.5048 |
| 0.0463        | 0.3333 | 6000  | 0.5074          | 44.1857 | 14.5137 |
| 0.0296        | 1.0197 | 7000  | 0.5135          | 43.6074 | 14.0715 |
| 0.0288        | 1.0752 | 8000  | 0.5119          | 43.6514 | 14.6808 |
| 0.0232        | 1.1308 | 9000  | 0.4999          | 41.8538 | 13.6624 |
| 0.022         | 1.1863 | 10000 | 0.4930          | 41.8813 | 13.6632 |
| 0.0226        | 1.2419 | 11000 | 0.4779          | 41.8208 | 13.8859 |
| 0.0213        | 1.2974 | 12000 | 0.4795          | 41.0569 | 13.3648 |
| 0.0194        | 1.353  | 13000 | 0.4831          | 41.0881 | 13.3223 |
| 0.0148        | 2.0393 | 14000 | 0.5064          | 41.2644 | 13.4050 |
| 0.0131        | 2.0949 | 15000 | 0.5116          | 41.2570 | 13.5709 |
| 0.0116        | 2.1504 | 16000 | 0.5102          | 40.6860 | 13.2589 |
| 0.0088        | 2.206  | 17000 | 0.5196          | 40.4859 | 13.2482 |
| 0.0129        | 2.2616 | 18000 | 0.5179          | 40.5501 | 13.2382 |


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