transformers
accelerate
bitsandbytes
optimum
huggingface_hub
scikit-build-core
llama-cpp-python
llama-cpp-agent>=0.2.25