For a distributed computing/BitTorrent-style method of running these LLMs, see: https://github.com/bigscience-workshop/petals.