-
-
Notifications
You must be signed in to change notification settings - Fork 149
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
Test out newer versions of PyTorch (torch>=2.0.0
) for compatibility with downstream projects (SCT, ADS)
#1304
Test out newer versions of PyTorch (torch>=2.0.0
) for compatibility with downstream projects (SCT, ADS)
#1304
Conversation
torch==2.0.0
(base branch: kk/chore_imagio
)torch==2.0.0
(base branch: kk/chore_imageio
)
ab00a2a
to
7d839e4
Compare
This version will be necessary for compatibility with MONAI/nnUNet, so I'm testing it here.
`torch._six` was removed, and we should be accessing an internal API anyway.
7d839e4
to
4ce4c05
Compare
torch==2.0.0
(base branch: kk/chore_imageio
)torch>=2.0.0
) for compatibility with downstream projects (SCT, ADS)
Now that #1297 has been merged, I've updated this PR to remove the now-redundant commits from the I'm starting out by testing only |
Pull Request Test Coverage Report for Build 7104163036
💛 - Coveralls |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Although I did a sanity check on my end using a GPU, would recommend running the test suite once on the lab's GPU. Just going to pre-approve 🚀
Also, add explanatory comment about why these specific versions are referenced.
Testing on ADS now |
It's going to take me a bit longer than expected to test, because since imageio was updated from version 2->3 in #1297, I'm going to have to update ADS to make it compatible with this new imageio as well (basically converting all our as_gray alls) before I can attempt to run the tests for this branch |
@joshuacwnewton everything looks like it's working on the ADS side - I tested it out in this PR by pointing the installation to the latest commit: https://github.com/axondeepseg/axondeepseg/pull/739/files#diff-9efd195f4e9bfb79ccd456a1d8370fafcc4bcb0b00ea3799222667d2ae818533R29 Everything looked good using our napari GUI as well. So it seems it's a green light from the ADS side! Thanks! |
Thank you so much for your testing (with imageio as well!) |
Because ivadomed's inference functionality is incorporated downstream in many of NeuroPoly's projects (e.g. SCT, ADS), it must "play nicely" with other packages installed into those downstream projects.
One issue with this is that ivadomed currently pins a (relatively old) version of PyTorch, while many other deep learning packages use a newer version of PyTorch. Because of ivadomed's restriction, it often can't be installed into the same environments as other deep learning packages.
So, this PR tests out newer versions of torch (
>=2.0.0
), for the purposes of compatibility with libraries such as nnUnet, MONAI, etc.