convert_to_audio_data.py

from datasets import Dataset, Audio
from glob import glob

audio_dataset = Dataset.from_dict(
    {"audio": glob("audio/*.mp3")}
).cast_column("audio", Audio())
audio_dataset.push_to_hub("kotoba-tech/kotoba-whisper-eval")