Trainers: support binary, multiclass, and multilabel tasks #2219
+82
−309
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
github-actions
bot
added
testing
adamjstewart
commented
Aug 12, 2024
| num_classes: int = 1000, | ||
| task: str = 'multiclass', | ||
| num_classes: int | None = None, | ||
| num_labels: int | None = None, |
There was a problem hiding this comment.
The defaults here match torchmetrics. If task='multiclass', only num_classes is used. If task='multilabel', only num_labels is used. If task='binary', both are ignored. Honestly, we could have a single num_classes if we want and simply use it for both.
adamjstewart
commented
Aug 12, 2024
| @@ -266,147 +262,3 @@ def predict_step( | |||
| x = batch['image'] | |||
| y_hat: Tensor = self(x).softmax(dim=-1) | |||
| return y_hat | |||
|
|
|||
|
|
|||
| class MultiLabelClassificationTask(ClassificationTask): | |||
There was a problem hiding this comment.
The right thing to do would be to deprecate this first and remove it in 0.7.0. Not sure how widely used it is. Deprecation is kind of annoying because you need to change all tests to acknowledge the warning message.
|
I think we first need to add a multilabel semantic segmentation dataset to properly test this. |
|
Alternatively, skip multilabel semantic segmentation and only support multilabel classification. |
adamjstewart
added
the
backwards-incompatible
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Outdated suggestions cannot be applied.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.
Instead of having separate trainers for binary, multiclass, and multilabel, let's create a single trainer that can handle all 3.
This applies to both Classification and Semantic Segmentation but not to our other trainers.
Closes #2205 @robmarkcole
Closes #245 @calebrob6