Pedro Cuenca

The problem is that timesteps reversal had been moved to the new function `calculateTimesteps`, but only for the PNDM scheduler. See [this PR](https://github.com/littleowl/ml-stable-diffusion/pull/1) for a proposed fix. With this change,...

> would be able to get a compiled split_ensum with the VAE encoder in it for version 2 or earlier pls? @pcuenca I'm having difficulty generating them locally ("Error computing...

Strange, the current PyTorch versions include cuda and should work out of the box. You may need to add the following to the `requirements.txt` file after you duplicate the space:...

@kesar can you post the URL for your Space, if it's public? I can try to take a quick look.

@i3oc9i The repro snippet would allow us to easily test on 0.11 and 0.12.1. I tried to reproduce but couldn't; what I did was run a stable diffusion pipeline and...

I'm starting to work on it, perhaps you can help if you want :) My idea is to make it simple and visual: retrieve a list of compatible models from...

Hello, @algal! I have the feeling that 1.5 GB of GPU RAM won't be enough to run the model, but I haven't been able to test on such a configuration...

Yes, please! I haven't been able to test it on mixed GPU systems myself. It'd also be helpful if you could select a random seed and post which one it...

Thanks a lot. That certainly doesn't look right. I suppose the app is using the discrete GPU, but it'd be interesting to confirm it. Could you please try to follow...

Thanks @nafrinafri! I'll try to force the use of the discrete GPU to see if it solves this problem; unfortunately, I don't have a Mac right now that uses two...