Files
pytorch/test/distributed
Tugsbayasgalan Manlaibaatar afeec56a5a Fix replacement reconstruct (#164937)
If we return Dtensor, the object is created via fx graph call so we never needed to reconstruct them. But if there is side effect, we do need to reconstruct it.

Differential Revision: [D84159000](https://our.internmc.facebook.com/intern/diff/D84159000)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/164937
Approved by: https://github.com/StrongerXi
2025-10-09 15:31:23 +00:00
..
2025-01-22 04:48:28 +00:00