Z-image GGUF detection not working correctly

[nabileisa]

When a z-image gguf deffusion model is selected the architecture switches to Automatic, flux,flux kontext only
but when an fp8 models is selected it shows z-image correctly
and when left at automatic it runs as flux workflow which throws an error which is expected since it's not the correct workflow.

dumped workflow is attached
workflow.json

[Acly]

Where is the file from? The ones linked in #2188 are working

[nabileisa]

The GGUF file I got from here
https://huggingface.co/gguf-org/z-image-gguf/tree/main
I think I might have found out the issue, Looking at the detection code in this commit It expects the general.architecture string lumina2 which is correct for the quants you linked
but the quant I have I looks like they had set the general.architecture string to pig for some reason , but it still does runs fine in comfy as a regular z-image model.

[Acly]

yea that's the reason. I dunno what pig is supposed to mean and why they chose that as architecture. This "gguf-org" also seems kinda scammy to me.

ComfyUI doesn't use arch for detection and also expects you to set stuff up manually.

[nabileisa]

Yea you are probably right , but I wanted a q4 that was a little bit smaller than the one you linked and his was the smallest I could find 😂 of course at the time I had no idea about the whole pig business.
Anyway I got it working by modifying the arch to lumina2 using this gradio app running locally.