Tiny inference-only implementation of LLaMA - View it on GitHub