gradio
librosa
loguru
numpy<2
spaces
torch
transformers