Personal projects at https://git.kokoham.com/sleepy . Trying to optimize LLM inference frameworks for local models in speed, and explore interesting quants for size.