I am running Gemma 4 31B for a project using LlamaCPP. There is no integrated main model + MTP drafter GGUF. And from what I can tell, LlamaCPP was updated to not accept a separate MTP drafter GGUF but instead to use a combined GGUF for main+drafter. So how can I use Gemma 4 31B with MTP on LlamaCPP
