LLM dockers running llama.cpp for Mistral llm's on a Nvidia P5000