garyakimoto

> @notkmikiya Yes I tried to come back to the [a9fed7c](https://github.com/AUTOMATIC1111/stable-diffusion-webui/commit/a9fed7c364061ae6efb37f797b6b522cb3cf7aa2) did it work? thanks

> No sorry, it is not working, the only thing working so far is to add `--opt-sub-quad-attention` but it is reducing the GPU performance. thank you for your report ,...

> @Kobeb33f Ok, managed to get ahold of a 2060 12GB to test some stuff out. Here are some results of the testing. > > When testing on this: `python:...

> > > @Kobeb33f Ok, managed to get ahold of a 2060 12GB to test some stuff out. Here are some results of the testing. > > > When testing...

I think this is really Automatic1111 version problem, because I was try to use kohya_ss they really speed up my training and photo create and I can set batch size...

> I've tried everything and I'm getting this error too. There is something wonky in the program that starts collecting stray memory. I've downloaded torch 2 and 2.1 with cu117...

I have same problem, even I try "set COMMANDLINE_ARGS=--skip-install --xformers " and it still try to download / update

REMOVE ALL OF THEM (pytorch, torchvision, torchaudio, cudatoolkit) and run: conda install pytorch=1.11.0 torchvision=0.12 torchaudio=0.11 cudatoolkit=11.5 -c pytorch -c conda-forge I tested myself and confirm this will work for me...

do you install CUDA Toolkit before? https://developer.nvidia.com/cuda-11.3.0-download-archive also if you're runing dual display card (example : some notebook run intel display card + nvidia ), you better to disable non-...