[Draft] add class-wise metrics #2800
Open
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
adamjstewart
reviewed
Jun 4, 2025
| @@ -180,33 +196,159 @@ def configure_losses(self) -> None: | |||
| def configure_metrics(self) -> None: | |||
| """Initialize the performance metrics. | |||
|
|
|||
| * :class:`~torchmetrics.Accuracy`: Overall accuracy | |||
| (OA) using 'micro' averaging. The number of true positives divided by the | |||
There was a problem hiding this comment.
These need to be indented for Sphinx to parse the bulleted list correctly
| dataset size. Higher values are better. | ||
| * :class:`~torchmetrics.JaccardIndex`: Intersection | ||
| over union (IoU). Uses 'micro' averaging. Higher valuers are better. | ||
| * :class:`~torchmetrics.classification.MulticlassAccuracy`: Overall accuracy |
There was a problem hiding this comment.
This change seems like going in the wrong direction, shouldn't we keep the generic class links?
Comment on lines
+221
to
+227
| expected_classes = 2 if task == 'binary' else num_classes | ||
| if len(labels) != expected_classes: | ||
| raise ValueError( | ||
| f"Length of labels ({len(labels)}) must equal number of classes " | ||
| f"({expected_classes}). Either provide {expected_classes} labels or " | ||
| f"set labels=None to use default integer names." | ||
| ) |
There was a problem hiding this comment.
This is currently missing test coverage
|
|
||
| # Validate labels length if provided | ||
| if labels is not None and task in ['binary', 'multiclass']: | ||
| expected_classes = 2 if task == 'binary' else num_classes |
There was a problem hiding this comment.
In the above docs, we say that "labels" is only used for task=multiclass, this disagrees with that
| self.test_metrics = metrics.clone(prefix='test_') | ||
|
|
||
| # For binary and multilabel tasks, we only need micro averaging | ||
| if task in ['binary', 'multilabel']: |
There was a problem hiding this comment.
This whole if branch can be replaced with:
for average in ['micro', 'macro']:
# For binary and multilabel tasks, we only need micro averaging
if task in ['binary', 'multilabel'] and average == 'macro':
continue
...Then we don't have to duplicate.
| ) | ||
|
|
||
| # Add classwise metrics for binary and multiclass (not multilabel) | ||
| if task in ['binary', 'multiclass'] and labels is not None: |
There was a problem hiding this comment.
Can we combine this too, something like:
for average in ['micro', 'macro', 'global']:
metric = ...
if average == 'global':
metric = ClasswiseWrapper(metric)
...
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.
Replaces #2130
Adds support for
labelsto be used inClasswiseWrapper