#openai
gradio
langchain
transformers
torch
#`low_cpu_mem_usage=True` or a `device_map` requires Accelerate: `pip install accelerate`
accelerate