torch==2.1.2
transformers
accelerate
optimum
scipy
numpy
huggingface-hub