Set default stage in adapter to "inference" #418
Merged
Conversation
This file contains hidden or 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
- In addition, make the stage used in the datasets configurable with a parameter. Defaults to "training" for now.
Codecov ReportAttention: Patch coverage is
|
Contributor
|
Datasets will pretty much only ever be used directly in training. I am fine with this change, although I do wish we had a better alternative available (such as simply not requiring a stage argument for the Adapter, since it should not be trainable in the first place). |
LarsKue
reviewed
Apr 17, 2025
Contributor
LarsKue
left a comment
LarsKue
left a comment
There was a problem hiding this comment.
I don't think we should change the stage argument to an enum, because either users will then have to make another import (cumbersome), or we make it compatible with strings anyway, in which case this becomes the primary use case.
Contributor
|
I suggest we stick to the simplest solution as proposed by Valentin. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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Outdated suggestions cannot be applied.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.
Concerns #417.
Question:
As Lars already highlighted when creating the current implementation, we cannot pass the stage to the
__getitem__methods of the datasets. How do we want to handle this?For now, I have made the stage that will be used configurable in the
__init__function.What should its default be? For now, I have set it to "training". The alternative would be to set it to "inference" and dynamically change it to "training" in
Approximator.fit.What do you think, @LarsKue and @stefanradev93? Setting this to a draft until we have discussed this...
Minor point: Do we want to convert the stages to an enum. This would be a bit more typing (but autocomplete should help), and might prevent future bugs due to typos in the strings. Not sure whether it is worth it though, what do you think?