Distributed-Llama: Connect home devices into a cluster for LLM inference