Skip to content

Conversation

@vertumnal
Copy link

@vertumnal vertumnal commented Dec 9, 2025

Negative weights lead to NaN values during normalization and may cause an indefinite hang when executed on CUDA. This adds a small check to validate the edge weights supplied as input are positive. Related to #10548 Fixes #795, #61

Add error handling for negative edge weights in GCN.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Weighted GCN layer produces some nans

1 participant