gradio
torch
transformers
datasets
accelerate
bitsandbytes
ctransformers
autoawq>=0.1.8
optimum
auto-gptq