Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

torch_geometric OS_Error #3310

Closed
fffahim opened this issue Oct 7, 2021 · 8 comments
Closed

torch_geometric OS_Error #3310

fffahim opened this issue Oct 7, 2021 · 8 comments

Comments

@fffahim
Copy link

fffahim commented Oct 7, 2021

image
How to solve this? Can anyone help?

@rusty1s
Copy link
Member

rusty1s commented Oct 8, 2021

Colab uses CUDA 11.1. As such, you need to download the wheels from https://data.pyg.org/whl/torch-1.9.0+cu111.html.

@manzar96
Copy link

manzar96 commented Nov 1, 2021

I have a similar problem that I can not figure out how to solve.
image

I am using pytorch==1.9.0+cu111
and I installed pytorch-geom according to the following command as referred at the docs:
pip install torch-scatter torch-sparse torch-cluster torch-spline-conv torch-geometric -f https://data.pyg.org/whl/torch-1.9.0+cu111.html
The Cuda version of the GPU is 1.4, however, I could not find the corresponding version for pytorch-geom.
What should I do?
@rusty1s
Thank you in advance.

@rusty1s
Copy link
Member

rusty1s commented Nov 2, 2021

Your system looks for CUDA 10.2, so I expect that installing from https://data.pyg.org/whl/torch-1.9.0+cu102.html should fix this error.

@manzar96
Copy link

manzar96 commented Nov 2, 2021

image
I am using an RTX 3090 CUDA version 11.4. How is that possible?

@rusty1s
Copy link
Member

rusty1s commented Nov 2, 2021

What does torch.version.cuda return for you?

@manzar96
Copy link

manzar96 commented Nov 2, 2021

What does torch.version.cuda return for you?

11.1

@rusty1s
Copy link
Member

rusty1s commented Nov 2, 2021

I see, thanks for confirming. I think you might have duplicated versions of torch-cluster installed (one for CUDA 10.2 and one for CUDA 11.1), other-wise it should never look for CUDA 10.2 in the first-place. Can you confirm that there do not exist multiple versions of torch-cluster by repeatedly running:

pip uninstall torch-cluster

and re-install torch-cluster+cu111 once again?

@manzar96
Copy link

manzar96 commented Nov 2, 2021

Yes, that was the problem. Thanks a lot!

@fffahim fffahim closed this as completed Nov 12, 2021
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants