Is k-diffusion included with this optimized version of SD?

[JoshuaKimsey]

Apologizes if this is the wrong place to ask. Was just curious if K-Diffusion/k_lms was added for the 0.5 release. I've been using a different setup, a fork or companion of this version with Gradio, that uses it and quite like the results.

If K-Diffusion isn't available yet, any chance it could be added as a sampling option? :)

[Orthestral]

I'd be very interested in using k_lms and k_euler_a as well without being limited to 512x512

[TheEnhas]

Yeah I'd like more samplers too, PLMS is okay but sometimes k_euler_ancestral and others are better depending on the image.

[JoshuaKimsey]

It would be awesome to see more samplers implemented here directly. One variant of this by hlky, which is used by the guide the Stable Diffusion Subreddit suggests, adds a large number of samplers, including k-lms, k_euler_a, and more. It might be worth looking into to add a number of different options for samplers in the base Optimized SD. It also adds the Gradio modifications to support that functionality too. Would definitely love to see some/all of these features make their way back to the base form of Optimized SD, for a number of reasons.

I'm also not sure how much it is needed compared to the other one I linked to, I need to test out this version of SD fully on Linux, but I'd be happy to help add any additional support for Linux-based systems as well. :)

[throwaway-mezzo-mix]

+1 on this, I would love more samplers.