(solved) I can generate images with these parameters with A1111 but not in ruined fooocus (Model: ponyRealism_v21VAE)

[Giribot]

Hello !

I can generate images with these settings with A1111:
Steps: 60, Sampler: DPM++ 2M Karras, CFG scale: 6, Seed: 687976846748, Face restoration: CodeFormer, Size: 990x540, Model hash: 059934ff58, Model: ponyRealism_v21VAE, Version: f0.0.17v1.8.0rc-latest-277-g0af28699

But in Ruined Foocus, even in "Custom", All diffusers or samplers (i try all) does'nt work .......

So i go to:
https://civitai.com/models/372465

I learned that you have to put “clip skip” at two.

So, it's work.....

Is it possible to detect if a model need to clip skip or not like A1111 ?

thanks

[runew0lf]

Its always clipskip 1 with sdxl, with pony its always 2, there is no need to put in automatic detection of clipskip. There is also a pony preset performance!

[yownas]

When you select the model you can see the model type. It doesn't give you all the information but hopefully it helps without running into information-overload.