gradio
transformers
torch
sentence-transformers
PyPDF2
numpy
accelerate>=0.26.0
optimum
bitsandbytes