Yeah, unfortunately it seems this can’t be converted to a llama.cpp compatible format yet, and that’s pretty big a tradeoff right now. Not surprising with how new it is, but we’ll have to wait to combine it with other improvements. Pretty exciting for the future though.