transformers==4.38.2 accelerate bitsandbytes optimum huggingface_hub==0.22.2 scikit-build-core https://github.com/abetlen/llama-cpp-python/releases/download/v0.2.81-cu124/llama_cpp_python-0.2.81-cp310-cp310-linux_x86_64.whl llama-cpp-agent>=0.2.25