speculative decoding in llama.cpp: anyone testing this?

[mangotinker73]

A couple questions before I guess too hard. Are you testing this on mini PC or something else?

There are a few different conversations hiding inside this thread, so being explicit about the goal helps.