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.

Sign up for GitHub

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

Jump to bottom

Change BYOL task argument name and switch to timm models #918

Merged
merged 7 commits into from
Dec 15, 2022
Merged

Change BYOL task argument name and switch to timm models #918

merged 7 commits into from
Dec 15, 2022

Conversation

nilsleh
Copy link
Collaborator

@nilsleh nilsleh commented Nov 27, 2022

Change the argument name for the tasks according to comment in #854. And switch from torchvision models to timm models.

@github-actions github-actions bot added testing Continuous integration testing trainers PyTorch Lightning trainers labels Nov 27, 2022
@adamjstewart adamjstewart added the backwards-incompatible Changes that are not backwards compatible label Nov 27, 2022
@adamjstewart adamjstewart added this to the 0.4.0 milestone Nov 27, 2022
adamjstewart
adamjstewart previously approved these changes Nov 28, 2022
View reviewed changes
@nilsleh
Copy link
Collaborator Author

nilsleh commented Dec 6, 2022

Sorry, I think I messed this up by also changing something in utils.py to get the tests to pass, but probably that should be its own PR change.

@nilsleh
Copy link
Collaborator Author

nilsleh commented Dec 6, 2022

Mhm now I am actually not sure what the best way to proceed is, because the utils.py function extract_encoder (extract_backbone) needs to look for backbone because that is the argument name for the trainer now, so I think I also need to change utils.py?

@adamjstewart
Copy link
Collaborator

Yes, you can change those jointly if needed. In hindsight, maybe we should have had a single PR for all torchvision -> timm changes and a single PR for all argument renames. Changing different tasks separately can lead to inconsistencies.

Basically, do whatever you have to do and I'll review. I think at this point we've 100% decided how things should look, it's just a matter of getting PRs merged in a way that doesn't break the tests.

@adamjstewart
Copy link
Collaborator

I think this is the last torchvision -> timm/argument rename PR? Should be easier to convert everything in unison now.

adamjstewart
adamjstewart reviewed Dec 14, 2022
View reviewed changes
conf/byol.yaml Outdated Show resolved Hide resolved
tests/conf/byol.yaml Outdated Show resolved Hide resolved
tests/trainers/test_utils.py Outdated Show resolved Hide resolved
torchgeo/trainers/utils.py Outdated Show resolved Hide resolved
@adamjstewart adamjstewart merged commit 98c50af into microsoft:main Dec 15, 2022
yichiac pushed a commit to yichiac/torchgeo that referenced this pull request Apr 29, 2023
@nilsleh
Change BYOL task argument name and switch to timm models (microsoft#918)
0646ce4 
* change argument name and supporrt timm

* list_models()

* encoder -> backbone

* fix failing test

* suggested changes
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@adamjstewart adamjstewart adamjstewart approved these changes

Assignees
No one assigned
Labels
backwards-incompatible Changes that are not backwards compatible testing Continuous integration testing trainers PyTorch Lightning trainers
Projects
None yet
Milestone
0.4.0
Development

Successfully merging this pull request may close these issues.
2 participants
@nilsleh @adamjstewart