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

fix segmentation demo #874

Closed
williamFalcon opened this issue Feb 17, 2020 · 7 comments · Fixed by #876
Closed

fix segmentation demo #874

williamFalcon opened this issue Feb 17, 2020 · 7 comments · Fixed by #876
Assignees
Labels
bug Something isn't working help wanted Open to be worked on

Comments

@williamFalcon
Copy link
Contributor

williamFalcon commented Feb 17, 2020

@akshaykvnit @Borda

 ERROR collecting pl_examples/full_examples/semantic_segmentation/semseg.py

 No module named 'torchvision.models.segmentation'
oops, looks like this test fails. Need to fix asap
@williamFalcon williamFalcon added bug Something isn't working help wanted Open to be worked on labels Feb 17, 2020
@williamFalcon williamFalcon assigned Borda and unassigned Borda Feb 17, 2020
@akshaykulkarni07
Copy link
Contributor

@williamFalcon
According to this page, it does exist. Not sure why it fails.

@williamFalcon
Copy link
Contributor Author

probably has to do with the version of torchvision which i noticed isn't being checked or anything

@akshaykulkarni07
Copy link
Contributor

@williamFalcon
According to this page, segmentation models were introduced in version 0.3 of torchvision. So, I think the tests might be using older versions (for backwards compatibility maybe?)

What I can do is add a custom implementation of U-Net (popular architecture) and replace the torchvision model. But even that'll take a few hours because I have some work right now.

@williamFalcon
Copy link
Contributor Author

@akshaykvnit yes please. That would be ideal. This is breaking the core tests, so we'll have to roll this back otherwise.

Thanks!

@Borda
Copy link
Member

Borda commented Feb 17, 2020

@williamFalcon major question is what torchvision version do you want to support, in past we had there torchvision>=0.4

@akshaykulkarni07
Copy link
Contributor

@williamFalcon major question is what torchvision version do you want to support, in past we had there torchvision>=0.4

I would also suggest keeping torchvision>=0.3 at least because a lot of object detection and segmentation models were introduced in 0.3. These are very good architectures which will be good for examples, instead of using custom models.

@Borda
Copy link
Member

Borda commented Feb 17, 2020

Note that this package is not dependent on torchvision any more, just some tests and examples...

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
bug Something isn't working help wanted Open to be worked on
Projects
None yet
Development

Successfully merging a pull request may close this issue.

3 participants