fix: filter predictions with ignore_index in MulticlassClassificationMetrics#997
Open
jonaserb-k wants to merge 1 commit intomainfrom
Open
fix: filter predictions with ignore_index in MulticlassClassificationMetrics#997jonaserb-k wants to merge 1 commit intomainfrom
jonaserb-k wants to merge 1 commit intomainfrom
Conversation
…Metrics torchmetrics' ignore_index only filters targets, not predictions. When a model produces unmappable answers (e.g. "n/a"), the extraction maps them to -1, which causes torch.bincount to crash on negative values. This overrides update() to filter out predictions matching ignore_index before they reach the underlying torchmetrics computation.
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
ignore_indexonly filters targets, not predictions. When a model produces unmappable answers (e.g."n/a"), the answer extraction maps them to-1, which causestorch.bincountto crash on negative values.update()inMulticlassClassificationMetricsto filter out predictions matchingignore_indexbefore they reach the underlying torchmetrics computation.Context
Discovered during CFMPB breast tubule formation benchmark evaluation —
alibaba/qwen2-5-vl-7b-instruct-vllmreturned"n/a"for some samples, which was mapped to-1byExtractDiscreteAnswer, crashingMulticlassAccuracy/MulticlassF1Score.Test plan
test_multiclass_classification_metrics)