Skip to content

perf(aggregation): Prevent cuda sync in normalize #557

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

Merged
ValerianRey merged 2 commits into from
Feb 5, 2026
Merged

perf(aggregation): Prevent cuda sync in normalize #557

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
fcd865a
perf(aggregation): Prevent cuda sync in normalize
ValerianRey Feb 4, 2026
bdc566a
Merge branch 'main' into prevent-cuda-sync-normalize
ValerianRey Feb 5, 2026
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
9 changes: 5 additions & 4 deletions src/torchjd/_linalg/_gramian.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -51,11 +51,12 @@ def normalize(gramian: PSDMatrix, eps: float) -> PSDMatrix:
sqrt of the sum of the diagonal elements. The gramian of the (Frobenius) normalization of `A` is
therefore `G` divided by the sum of its diagonal elements.
"""

squared_frobenius_norm = gramian.diagonal().sum()
if squared_frobenius_norm < eps:
output = torch.zeros_like(gramian)
else:
output = gramian / squared_frobenius_norm
condition = squared_frobenius_norm < eps

# Use torch.where rather than a if-else to avoid cuda synchronization.
output = torch.where(condition, torch.zeros_like(gramian), gramian / squared_frobenius_norm)
return cast(PSDMatrix, output)


Expand Down