frozenleaves/pytorch
1
0
Fork 0
mirror of https://github.com/pytorch/pytorch.git synced 2025-10-20 21:14:14 +08:00
Code Issues Packages Projects Releases Wiki Activity

Work around buggy use_const_ref_for_mutable_tensors (#145530)

Browse Source 
See https://github.com/pytorch/pytorch/issues/145522 for context

This doesn't fix the problem with use_const_ref_for_mutable_tensors and the boxed wrapper, instead it just gets all of our out kernels off of this flag so that the mutable matching pattern works correctly. I also add a check in torchgen to prevent people from making this mistake in the future.

Signed-off-by: Edward Z. Yang <ezyang@meta.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/145530
Approved by: https://github.com/albanD, https://github.com/bdhirsh
This commit is contained in:
Edward Z. Yang
2025-01-23 20:37:27 -08:00
committed by PyTorch MergeBot
parent 9d6927715f
commit bc62930765
5 changed files with 33 additions and 15 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
torchgen/model.py
View File

@ -626,6 +626,11 @@ class NativeFunction:
)
assert isinstance(use_const_ref_for_mutable_tensors, bool)
if use_const_ref_for_mutable_tensors:
assert (
not func.arguments.out
), "see https://github.com/pytorch/pytorch/issues/145522"
variants_s = e.pop("variants", "function")
assert isinstance(variants_s, str)
variants: set[Variant] = set()