nltk
accelerate
optimum>=1.12.0
auto-gptq
torch==2.1.0
transformers>=4.32.0
selfcheckgpt