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

[dynamic shapes] make backed_size_oblivious behavior consistent b/w symbolic_shapes/inductor (#164796)

Browse Source 
Summary: call guard_or_ directly to enable backed_size_obl in inductor calls to guard_or

Test Plan: CI and unit test added.

Differential Revision: D84009392

Pull Request resolved: https://github.com/pytorch/pytorch/pull/164796
Approved by: https://github.com/laithsakka
This commit is contained in:
Pian Pawakapan
2025-10-08 00:19:03 +00:00
committed by PyTorch MergeBot
parent e98c4e835b
commit bd3b98a8a5
2 changed files with 23 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
test/test_dynamic_shapes.py
View File

@ -1426,6 +1426,15 @@ class f(torch.nn.Module):
f(torch.tensor([1]), torch.tensor([1])), torch.tensor([20])
)
@fresh_cache()
def test_slice_backed_size_oblivious(self):
@torch.compile(backend="inductor", fullgraph=True, dynamic=True)
def f(x):
return x[:5]
with torch.fx.experimental._config.patch(backed_size_oblivious=True):
f(torch.randn(10, 10))
def test_baddbmm_symint(self):
from torch._subclasses.fake_tensor import FakeTensorMode