[Stable Diffusion] Prompt Sharing and Learning Thread

[felldude]

Using the merge I made and messing about with the sampler and scheduler and some prompts I managed to get slightly better images.

I don't know if the attention would affect the image quality, but did you compare your Iterations per second vs XFormers

 

[Synalon]

I don't know if the attention would affect the image quality, but did you compare your Iterations per second vs XFormers

It seems slower overall now, it was taking roughly 320 seconds for 4 images before now its taking 1200 seconds

 

[felldude]

It seems slower overall now, it was taking roughly 320 seconds for 4 images before now its taking 1200 seconds

Interesting thanks for testing, in my test I had no speed difference on any model, But I do have the CuDnn files in the Comfy Torch build.

Good to know Xformers works for an FP8 model at least when its upcast with BF16