HF does this for ggufs, and it’ll show you what quantizations will work on the GPU(s) you’ve selected. Hopefully that feature gets expanded to support more mode

HF does this for ggufs, and it’ll show you what quantizations will work on the GPU(s) you’ve selected. Hopefully that feature gets expanded to support more model types.