llama.cpp : add --n-cpu-moe option

[anzax]

Please give us ability to set new --n-cpu-moe value for llama.cpp
ggml-org/llama.cpp#15077

It is mandatory for new MOE OpenAI open models and GLM-4.5
It would be really nice to be able to pass args to llama.cpp so you don't need to add new controls to UI all the time.

[EzequielDM]

I agree with being able to pass args directly to llama.cpp, that should be an option.

[x0wllaar]

This would actually make large MoEs usable for me

[anzax]

I’ll add a link to the Reddit post, just to highlight the real demand: https://www.reddit.com/r/LocalLLaMA/comments/1mr7m2r/lm_studio_now_supports_llamacpp_cpu_offload_for/