view article Article Performant local mixture-of-experts CPU inference with GPU acceleration in llama.cpp Doctor-Shotgun β’ Jan 30 β’ 33