README.md

---
license: cc-by-nc-4.0
language:
- ar
---

## Checkpoints

### Pre-Trained Models

 Model | Pre-train Dataset | Model | Tokenizer |
| --- | --- | --- | --- |
| ArTST v2 base | Dialects | [Hugging Face](https://huggingface.co/MBZUAI/ArTSTv2/blob/main/pretrain_checkpoint.pt) | [Hugging Face](https://huggingface.co/MBZUAI/ArTSTv2/blob/main/tokenizer_artstv2.model)

### Finetuned Models
 Model | Finetune Dataset | Model | Tokenizer |
| --- | --- | --- | --- |
| ArTST v2 ASR | MGB2 | [Hugging Face](https://huggingface.co/MBZUAI/ArTSTv2/blob/main/ASR_MGB2_best.pt_hf.pt) | [Hugging Face](https://huggingface.co/MBZUAI/ArTSTv2/blob/main/tokenizer_artstv2.model) |
| ArTST v2 ASR | QASR | [Hugging Face](https://huggingface.co/MBZUAI/ArTSTv2/blob/main/ASR_QASR_best.pt_hf.pt) | [Hugging Face](https://huggingface.co/MBZUAI/ArTSTv2/blob/main/tokenizer_artstv2.model) |
| ArTST v2 ASR | MGB2 - Dialects | [Hugging Face](https://huggingface.co/MBZUAI/ArTSTv2/blob/main/ASR_Dialects_MGB2_best.pt_hf.pt) | [Hugging Face](https://huggingface.co/MBZUAI/ArTSTv2/blob/main/tokenizer_artstv2.model) |

# Acknowledgements

ArTST is built on [SpeechT5](https://arxiv.org/abs/2110.07205) Architecture. If you use any of ArTST models, please cite 

``` 
@inproceedings{toyin2023artst,
  title={ArTST: Arabic Text and Speech Transformer},
  author={Toyin, Hawau and Djanibekov, Amirbek and Kulkarni, Ajinkya and Aldarmaki, Hanan},
  booktitle={Proceedings of ArabicNLP 2023},
  pages={41--51},
  year={2023}
}

@misc{djanibekov2024dialectalcoveragegeneralizationarabic,
      title={Dialectal Coverage And Generalization in Arabic Speech Recognition}, 
      author={Amirbek Djanibekov and Hawau Olamide Toyin and Raghad Alshalan and Abdullah Alitr and Hanan Aldarmaki},
      year={2024},
      eprint={2411.05872},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2411.05872}, 
}
```