1390 Enhance DynUNet with get_feature_maps API

[Nic-Ma]

Fixes #1390 .

Description

As more than 3 users reported the problem with DynUNet, we need to change the return data of forward function.
This PR added get_feature_maps API according to @rijobro 's suggestion in #1263 (comment).

Status

Ready

Types of changes

• Non-breaking change (fix or new feature that would not break existing functionality).
• Breaking change (fix or new feature that would cause existing functionality to change).
• New tests added to cover the changes.
• Integration tests passed locally by running ./runtests.sh --codeformat --coverage.
• Quick tests passed locally by running ./runtests.sh --quick.
• In-line docstrings updated.
• Documentation updated, tested make html command in the docs/ folder.

[Nic-Ma]

/black

[Nic-Ma]

/black

[wyli]

thanks, this is a breaking change, may need to update the tutorial as well

[DrAlaafeef]

I got an error when trying to run the DynUNet tutorial using the docker image "projectmonai/monai:latest" on 21/2/2021.

"torch.nn.modules.module.ModuleAttributeError: 'DynUNet' object has no attribute 'get_feature_maps'"

[rijobro]

Thanks @DrAlaafeef, you're seeing this because we modified the DynUNet public interface again, see #1596. We deleted the get_feature_maps method and returned a concatenated tensor containing both the output of the forward pass and the feature maps (if deep_supervision==True).

Would you be able to create an issue on the Tutorials repository? We're currently in the process of setting up our CI/CD to automatically run all of our tutorials, so we should avoid this sort of problem in the future.

@yiheng-wang-nv do you think you could take care of the Tutorials PR?

[rijobro]

@DrAlaafeef Alternatively, you can checkout the 0.4.0 tag for both MONAI and the tutorials to ensure they are in sync with each other.