N-Shot-Benchmarking-of-Whisper

N-Shot Benchmarking of Whisper

To run

1. Install dependencies in your environment: pip install -r requirementx.txt
2. Run train.py as: python train.py --training_args OR
3. Run train.sh as: bash train.sh OR
4. Submit a slurm job for training: sbatch --job_args train.sh --training_args
5. You can use multiple launcher including deepspeed, torchrun, and python.
6. Remove --deepspeed "ds_config2.json" and change launcher from deepspeed to torchrun or python if you don't want to use deepspeed.

Citation

@article{talafha2023n,
  title={N-Shot Benchmarking of Whisper on Diverse Arabic Speech Recognition},
  author={Talafha, Bashar and Waheed, Abdul and Abdul-Mageed, Muhammad},
  journal={arXiv preprint arXiv:2306.02902},
  year={2023}
}