---
library_name: transformers
language:
- ar
license: apache-2.0
base_model: openai/whisper-tiny
tags:
- generated_from_trainer
datasets:
- pain/MASC
- google/fleurs
- deepdml/Tunisian_MSA
- deepdml/mtedx
- ymoslem/MediaSpeech
- UBC-NLP/Casablanca
- fixie-ai/common_voice_17_0
metrics:
- wer
model-index:
- name: Whisper Tiny ar
  results:
  - task:
      name: Automatic Speech Recognition
      type: automatic-speech-recognition
    dataset:
      name: Common Voice 17.0
      type: pain/MASC
    metrics:
    - name: Wer
      type: wer
      value: 52.36224086961312
---
<!-- 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 Tiny ar

This model is a fine-tuned version of [openai/whisper-tiny](https://huggingface.co/openai/whisper-tiny) on the Common Voice 17.0 dataset.
It achieves the following results on the evaluation set:
- Loss: 0.5962
- Wer: 52.3622
- Cer: 18.7245

## 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     |
|:-------------:|:------:|:-----:|:---------------:|:-------:|:-------:|
| 1.1234        | 0.0556 | 1000  | 0.8106          | 67.0994 | 25.5658 |
| 0.8957        | 0.1111 | 2000  | 0.7234          | 62.5310 | 23.3911 |
| 0.7114        | 0.1667 | 3000  | 0.6871          | 59.7675 | 21.8722 |
| 0.6346        | 0.2222 | 4000  | 0.6637          | 58.1976 | 20.7336 |
| 0.4961        | 0.2778 | 5000  | 0.6545          | 57.7404 | 20.7048 |
| 0.4354        | 0.3333 | 6000  | 0.6473          | 56.7948 | 20.2061 |
| 0.3924        | 0.3889 | 7000  | 0.6325          | 55.8400 | 20.0139 |
| 0.3466        | 0.4444 | 8000  | 0.6274          | 55.4176 | 20.1441 |
| 0.2979        | 0.5    | 9000  | 0.6206          | 54.6997 | 19.6005 |
| 0.3099        | 0.5556 | 10000 | 0.6150          | 54.0166 | 19.3231 |
| 0.2681        | 0.6111 | 11000 | 0.6120          | 53.5980 | 19.1106 |
| 0.2383        | 0.6667 | 12000 | 0.6113          | 53.5576 | 19.4238 |
| 0.2582        | 0.7222 | 13000 | 0.6060          | 52.7515 | 18.7573 |
| 0.1543        | 0.7778 | 14000 | 0.6018          | 52.6175 | 18.4895 |
| 0.2356        | 0.8333 | 15000 | 0.6023          | 52.9902 | 18.9782 |
| 0.2031        | 0.8889 | 16000 | 0.5984          | 52.5165 | 18.8550 |
| 0.2437        | 0.9444 | 17000 | 0.5951          | 52.4926 | 18.7514 |
| 0.2269        | 1.0    | 18000 | 0.5962          | 52.3622 | 18.7245 |


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