Conversion of MoE version of 1.6

#2
by Jack-771a - opened

https://huggingface.co/LanguageBind/MoE-LLaVA-Phi2-2.7B-4e
Can you please convert this model as well into .gguf file format?
Or tell the way how to do this. All (2) scripts I found doesn't work and can't convert models into .GGUF

Owner

have you followed the convert and quantize steps similar to what's in this PR? https://github.com/ggerganov/llama.cpp/pull/4406

this should work for MoE

@Jack-771a llama cpp doesnt have support for that model yet
the steps that cjpais gave is for normal moe but that one is llava moe.

so you either ask for support in llama cpp or try to create it by yourself

Sign up or log in to comment