[dofm]

I never did quite get round to a llama.cpp build but support is there now anyway and I have the MTP drafter working with the QAT 26B build:

https://news.ycombinator.com/item?id=48441450