`F.normalize` returns ~1e12 gradient at zero-vector input instead of NaN, while the forward pass already returns NaN
May 20, 2026 · #184575
Python
Difficulty: Medium
Labels
triage review module: autograd module: correctness (silent) module: norms and normalization module: edge cases
Parent Repository
pytorch/pytorch
Python repository
100,083 27,851
Ready to work on this? Walk through the full fork-to-PR workflow so your first contribution goes smoothly.