torch==2.1.2
accelerate
bitsandbytes
transformers
elasticsearch
openai
vllm