This is a follow-up to the previous article in which I talked about using KoboldCpp to run AI Models locally. For AMD APUs, this framework mainly uses CPU and GPU via Vulkan. I noticed that Koboldcpp doesn't