fast-stable-diffusion
fast-stable-diffusion copied to clipboard
TheLastBen
Xformers Error
Discussed in https://github.com/TheLastBen/fast-stable-diffusion/discussions/1936
Originally posted by HaikenEdge April 3, 2023 I'm getting this error when I try to run "Start Stable-Diffusion" cell of the notebook:
/usr/local/lib/python3.9/dist-packages/torchvision/transforms/functional_tensor.py:5: UserWarning: The torchvision.transforms.functional_tensor module is deprecated in 0.15 and will be **removed in 0.17**. Please don't rely on it. You probably just need to use APIs in torchvision.transforms.functional or in torchvision.transforms.v2.functional.
warnings.warn(
WARNING[XFORMERS]: xFormers can't load C++/CUDA extensions. xFormers was built for:
PyTorch 1.13.1+cu117 with CUDA 1107 (you have 2.0.0+cu118)
Python 3.9.16 (you have 3.9.16)
Please reinstall xformers (see https://github.com/facebookresearch/xformers#installing-xformers)
Memory-efficient attention, SwiGLU, sparse and more won't be available.
Set XFORMERS_MORE_DETAILS=1 for more details
Not sure what's causing it or how to fix it.
installing newer version of Xformer from terminal removes the error but i still get this error
/usr/local/lib/python3.9/dist-packages/torchvision/transforms/functional_tensor.py:5: UserWarning: The torchvision.transforms.functional_tensor module is deprecated in 0.15 and will be removed in 0.17. Please don't rely on it. You probably just need to use APIs in torchvision.transforms.functional or in torchvision.transforms.v2.functional. warnings.warn(
https://github.com/TheLastBen/fast-stable-diffusion/issues/1935#issuecomment-1494698624
1 open the code in the Dependencies tab 2 insert- !pip install -U xformers --no-deps -qq 3 reinstall Dependencies 4 profit
1 open the code in the Dependencies tab 2 insert- !pip install -U xformers --no-deps -qq 3 reinstall Dependencies 4 profit
The version I'm using doesn't seem to have a Dependencies tab.
#1935 (comment)
The version I'm using doesn't seem to have a Dependencies tab.
the meaning is the same, run it pip install -U xformers
Tornado, i dont see that dependices tab are you on the training module, because on the main there is no tab for that
fixed
NansException: A tensor with all NaNs was produced in Unet. This could be either because there's not enough precision to represent the picture, or because your video card does not support half type. Try setting the "Upcast cross attention layer to float32" option in Settings > Stable Diffusion or using the --no-half commandline argument to fix this. Use --disable-nan-check commandline argument to disable this check.
possibly related to the xformers, before update it works. now i get this
Steps: generate image 768x768 with Hi Res Fix x2 Highest i can go without receiving that error is 720x720, which is 1440x1440 in Hires. fix x2
try a different model
So far i tried default 1.5 model, realistic vision, illuminati diffusion, all result the same error.
I'm also getting the NansException exception when trying to generate large images, with or without hires fix. Seems a lot of people are too. I haven't seen a reliable fix yet. I'm prompting using the exact same prompt, model, loras, and settings as I did a few days ago.
NansException: A tensor with all NaNs was produced in Unet. This could be either because there's not enough precision to represent the picture, or because your video card does not support half type. Try setting the "Upcast cross attention layer to float32" option in Settings > Stable Diffusion or using the --no-half commandline argument to fix this. Use --disable-nan-check commandline argument to disable this check.
Same, are local intallations (not on colab) having these problems? Seems like something breaks every week, but I never see people complaining anywhere else
Same, are local intallations (not on colab) having these problems? Seems like something breaks every week, but I never see people complaining anywhere else
tested, no issue
