Adding transformation adaptor #193
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
vale-salvatelli
changed the title
maxilse
previously approved these changes
Mar 3, 2022
harshita-s
reviewed
Mar 3, 2022
dccastro
reviewed
Mar 3, 2022
| if k_output is None: | ||
| ditems = ret | ||
| else: | ||
| if isinstance(ret, type(ditems[k_output])): |
There was a problem hiding this comment.
For my information, I'm not sure I understand this. Is the dictionary expected to already contain k_output? And why do the types need to match? This seems overly restrictive—consider e.g. loading an image tensor from a path.
Another comment is that the MONAI docs suggest that transforms shouldn't modify input dictionaries in-place, but rather call ditems.copy() before assigning the outputs.
There was a problem hiding this comment.
I think they're both good comments:
- correct we could extend this adaptor to be compatible with a broader set of transformations and also give the option to add a key rather than just overwriting. I wanted to keep this as simple as possible to solve our current case.
- do you know why MONAI suggest to copy? here I am a bit worried about memory consumption if every transformation is creating a copy but happy to follow up on this
There was a problem hiding this comment.
On the second point, it's just a shallow copy of the dictionary, in case the original needs to be used elsewhere (e.g. if passed to multiple parallel transformations).
dccastro
approved these changes
Mar 3, 2022
harshita-s
approved these changes
Mar 3, 2022
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
PR to add a transformation adaptor to allow transformations working only on Tensors/Images to work with a dict in input