G github.com Show HN: Llama.cpp fork with 2-4x multiGPU speed for MoE models bigger than VRAM 2 points neuralll 1 month ago 3 comments
