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

Failed to build wheel in Nvidia CUDA 10.1 Ubuntu 16.04 Docker image #284

Closed
MiroPsota opened this issue Jul 23, 2020 · 2 comments
Closed

Comments

@MiroPsota
Copy link

MiroPsota commented Jul 23, 2020

Instructions To Reproduce the Issue:

Build Dockerfile and run commands in build_wheel_pytorch3d.sh.

Relevant error:

Parts with errors in error.log and full log in full.log.

Notes:

I can build pytorch3d wheel with cuda 10.1 Ubuntu 18.04 and cuda 11.0 Ubuntu 20.04 as base images so it's likely some incompatibility with GCC 5.4?
Also buildable with cuda 10.1 Ubuntu 16.04 at commit dc08c30.

@bottler
Copy link
Contributor

bottler commented Jul 23, 2020

This is the problem with cloneable not working in nvcc. It can be fixed by making a change like 85c396f to the code recently added in bce396d.

With pytorch master I think this won't be a problem because of a fix like pytorch/pytorch#37221. It also might be working with latest pytorch (1.5.1) rather than the 1.5.0 which you might be picking up - I'm not sure about this, but I suspect that's why our CI is happy at the moment.

@MiroPsota
Copy link
Author

MiroPsota commented Jul 23, 2020

I confirm it's building with pytorch 1.5.1. I suggest to add note to INSTALL.md.

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

2 participants