Steward Garcia

@leejet It seems to have almost the same architecture as SD 2.1 but includes some temporal consistency blocks called "time_stack." We'll need to see how they work and whether new...

> I'm currently reviewing the SVD implementation code in comfyui. Perhaps I can learn how to conveniently implement SVD within sd.cpp from this. Amazing! Good luck!!, Unfortunately, my time is...

@leejet this can be closed

> I'm still seeing it with the latest code. It could be a metal only issue. Will try on Cuda and report back. You should have mentioned what was happening...

download this vae [sdxl-vae-fix](https://huggingface.co/madebyollin/sdxl-vae-fp16-fix/blob/main/sdxl_vae.safetensors), and add the argument `--vae sdxl_vae.safetensors`

@SunGreen777 wrong repo! We don't support these facts

It seems that I have already found the problem. In the last pull request, which is almost ready to merge, the converter bug is fixed.

@Cyberhan123 At the moment, the current version of the project is highly limited in exposing functions externally. Significant refactoring and testing would be required. Honestly, I don't currently have the...

> If by "last pull request", you mean #88, it compiles but it's still broken as the original report describes Could you give me more details, model name, where you...

Currently, the metal backend is not optimized for the matrix multiplications performed by stable diffusion, and it seems that no one wants to take responsibility for fixing that in the...