flash-moe-inference

Run 397B parameter Mixture-of-Experts LLMs on a MacBook using pure C/Metal with SSD streaming

Independent GitHub snapshot · Automated classification · Original documentation governs