On Fri, 2025-03-07 at 14:40 +0100, Christian Kastner wrote: > > FYI, Mathieu Baudier contributed the Vulkan backend. Thanks, Mathieu! > > In the ideal case, this should make ggml usable with Nvidia cards, > without requiring non-free packages to build. I haven't tested this yet, > but I am now converting llama.cpp to use our package ggml. > Thanks, Mathieu! BTW, I can test the vulkan backend on nvidia GPU when it is ready.