Skip to content

fix macro when ignore_index is set #2163

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
wants to merge 9 commits into from
Closed

fix macro when ignore_index is set #2163

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
69970f3
fix macro when ignore_index is set
edumotya Oct 10, 2023
9945e08
Merge branch 'master' into bugfix/macro_ignore_index
Borda Oct 17, 2023
28ad59c
Merge branch 'master' into bugfix/macro_ignore_index
Borda Oct 20, 2023
08cdbe9
Merge branch 'master' into bugfix/macro_ignore_index
Borda Nov 8, 2023
e2fbb7a
Merge branch 'master' into bugfix/macro_ignore_index
Borda Nov 26, 2023
47a1d68
Merge branch 'master' into bugfix/macro_ignore_index
Borda Dec 18, 2023
83021e4
Merge branch 'master' into bugfix/macro_ignore_index
Borda Jan 4, 2024
4e9388c
Merge branch 'master' into bugfix/macro_ignore_index
Borda Apr 10, 2024
f542991
Merge branch 'master' into bugfix/macro_ignore_index
Borda Oct 31, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions src/torchmetrics/functional/classification/stat_scores.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -416,6 +416,8 @@ def _multiclass_stat_scores_update(
fp = confmat.sum(0) - tp
fn = confmat.sum(1) - tp
tn = confmat.sum() - (fp + fn + tp)
if ignore_index is not None:
fp[ignore_index] = 0
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

this is probably correct, but since many metrics derive from the stat_scores class that means that basically all would need to have their unittests fixed

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

what do you mean by it needs to be fixed?

return tp, fp, tn, fn


Expand Down
7 changes: 2 additions & 5 deletions tests/unittests/classification/test_accuracy.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -190,12 +190,9 @@ def _reference_sklearn_accuracy_multiclass(preds, target, ignore_index, multidim
return _reference_sklearn_accuracy(target, preds)
confmat = sk_confusion_matrix(target, preds, labels=list(range(NUM_CLASSES)))
acc_per_class = confmat.diagonal() / confmat.sum(axis=1)
acc_per_class[np.isnan(acc_per_class)] = 0.0
if average == "macro":
acc_per_class = acc_per_class[
(np.bincount(preds, minlength=NUM_CLASSES) + np.bincount(target, minlength=NUM_CLASSES)) != 0.0
]
return acc_per_class.mean()
return np.nanmean(acc_per_class)
acc_per_class[np.isnan(acc_per_class)] = 0.0
if average == "weighted":
weights = confmat.sum(1)
return ((weights * acc_per_class) / weights.sum()).sum()
Expand Down
Loading