transformers
torch
einops
accelerate
tiktoken
scipy
transformers_stream_generator==0.0.4 
peft 
deepspeed
bitsandbytes
optimum
vllm==0.3.2
chromadb
sentence_transformers
difflib