Error raise RuntimeError( RuntimeError: Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check ao Instalar

[aleffdaniel]

Is there an existing issue for this?

• I have searched the existing issues and checked the recent builds/commits

What would your feature do ?

When trying to install Automatic 1111 on my PC, after opening the webui-user.bat file, CMD show the error command " raise RuntimeError(
RuntimeError: Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check" My GPU is an AMD 570 series 4GB, can anyone help me?

Proposed workflow

1. Go to ....
2. Press ....
3. ...

Additional information

No response

[lshqqytiger]

Launch webui with --use-directml or --use-zluda. AMD GPUs will not be detected automatically as there are several ways to run stable diffusion on AMD GPUs and you should select one.

[Etwas1111]

how exactly one can launch it with them? do i edit the webui-user.bat file?

[knrh8r]

how exactly one can launch it with them? do i edit the webui-user.bat file?

Either add the flags in webui-user.bat, ex. set COMMANDLINE_ARGS=--use-directml
Or create a shortcut to webui.bat, then go into shortcut properties and edit the Target field, ex. path\to\webui.bat --use-directml

[Draagoo]

how exactly one can launch it with them? do i edit the webui-user.bat file?

Either add the flags in webui-user.bat, ex. set COMMANDLINE_ARGS=--use-directml Or create a shortcut to webui.bat, then go into shortcut properties and edit the Target field, ex. path\to\webui.bat --use-directml

I tried this, i just get the error: AttributeError: module 'torch' has no attribute 'dml'

[lshqqytiger]

Remove venv folder and try again.

[juliannevela]

Remove venv folder and try again.

This was what finally worked for me. I am now successfully able to run stable diffusion on my AMD GPU on Windows. :) Thanks!

[InfernalSeraph]

It has been so difficult trying to find complete working instructions to get this running on AMD GPU's. It sure would be appreciated if this was part of the steps for installation. Or if there was maybe a batch for AMD users that guided us through the necessary steps?

It isn't quite clear whether I needed Olive or not, nor which pip package(s) for Olive are actually required. And somewhere suggests adding --skip-torch-cuda-test to the COMMANDLINE_ARGS is supposed to make it work, but that was not the case and I was just stumbling around blindly trying to figure out what unfamiliar settings might be needed to get this to finally successfully generate something.

Regardless, thank you @lshqqytiger for enlightening us.

[JosephDavid2]

my board is rx 5500 xt and it worked with the set COMMANDLINE_ARGS=----use-zluda