llamacpp_python container may not work on all CPUs

[tiran]

llama-cpp detects CPU features like AVX, AVX2, FMA3, and F16C at build time. If the container is built on a machine that supports these instruction sets, then the binary won't work on CPUs without these instructions.

ai-lab-recipes/model_servers/llamacpp_python/cuda/Containerfile

Lines 4 to 7 in 96555a1

	RUN pip install --upgrade pip
	ENV CMAKE_ARGS="-DLLAMA_CUBLAS=on"
	ENV FORCE_CMAKE=1
	RUN pip install --no-cache-dir --upgrade -r /locallm/requirements.txt

References:

• https://github.com/ggerganov/llama.cpp/blob/d26e8b669dbf1f5f5a0afe4d2d885e86cf566302/CMakeLists.txt#L73-L78
• Core dumped on trying to import from llama_cpp module when built with CUBLAS=on abetlen/llama-cpp-python#412

Credits to @bbrowning for figuring this out. He suggested to build llama-cpp-python with CMAKE_ARGS="-DLLAMA_CUBLAS=on -DLLAMA_AVX2=OFF -DLLAMA_FMA=OFF -DLLAMA_F16C=OFF"

[Gregory-Pereira]

I would consider this as a feature to include in release 1.0 and so linking to #262

[rhatdan]

Looks like this is in the main branch.