Collaboration/Sponsorship: Improving SDXL Inference Performance in stable-diffusion.cpp

[JustMaier]

We’re exploring stable-diffusion.cpp at Civitai to better serve SDXL requests, but we’ve found inference times aren’t as fast as expected. We’d love to help improve this and are open to sponsoring development or collaborating with contributors here. Would you be interested in discussing ways to optimize performance together?

[wbruna]

Sorry, @JustMaier , I didn't really notice this separate issue until now; probably because of the similar title on the discussion on the main repo. But I believe it got some traction there, at least at the beginning.

I don't actually have a lot of experience with ML, so from the items you mentioned we could focus on, I'm afraid I could only really help with the model loading part - and rmatif is already advancing a lot on that front. But if those priorities change, I'd be happy to discuss other approaches.