Can it use integrated GPU?

[kun-create]

I've AMD integrated GPU 4GB, but it's still showing "No GPU being used. Careful, inference might be extremely slow!"

Does this program not use integrated GPU?

[Advitica]

It's looking for Nvidia and CUDA installed. I have multiple GPU's with 6GB but this is not enough to load the model :(

[gkucsko]

have you tried using small versions of model (specified in preload_models?) not sure about amd support, right now only cuda and cpu are supported

[Tom-Neverwinter]

T Neverwinter 18 minutes ago (edited) update to cuda 11.8, no gpu found? torch 2.0.0 no ++117/118 solved based on this: https://github.com/AUTOMATIC1111/stable-diffusion-webui/discussions/8691 command: pip3 install clean-fid numba numpy torch==2.0.0+cu118 torchvision --force-reinstall --extra-index-url https://download.pytorch.org/whl/cu118

I didnt fix the gpu as my generation times are still astronomical, however it seems to install the appropriate versions?

[gkucsko]

are the issues here related? is everyone on AMD or is this mixed with cuda issues?

[Tom-Neverwinter]

AMD CPU with Nvidia GPU. seems to be an AMD CPU issue. https://pastebin.com/wRVCpcep

I dont see more correlation than that at this time sadly

[kun-create]

So it will not work until I use the Nvidia Graphics card!

[gkucsko]

it should work on cpu, albeit slower of course..

[limapedro]

how to run on cpu? it doesn't seem to have a option to pass device on init of the model!

[gkucsko]

preload_models has arguments to disable gpu, but either way it should happen automatically if you don't have a supported gpu

[gkucsko]

Closing for now since I can’t reproduce the issue