OSSForks/pytorch
1
0
Fork 0
mirror of https://github.com/zebrajr/pytorch.git synced 2025-12-06 00:20:18 +01:00
Code Issues Actions 38 Packages Projects Releases Wiki Activity

[CUDA] xFail max-autotune grouped gemm tests on devices with insufficient SM count (#165921)

Browse Source 
Pull Request resolved: https://github.com/pytorch/pytorch/pull/165921
Approved by: https://github.com/ngimel
This commit is contained in:
Eddie Yan 2025-10-30 20:05:03 +00:00 committed by PyTorch MergeBot
parent 52db60170d
commit f911d64750
1 changed files with 8 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
test/test_matmul_cuda.py
View File

@ -49,6 +49,8 @@ from torch.testing._internal.common_utils import (
decorateIf,
)
from torch.testing._internal.inductor_utils import IS_BIG_GPU
from torch._inductor.test_case import TestCase as InductorTestCase
_IS_SM8X = False
@ -619,8 +621,12 @@ class TestMatmulCuda(InductorTestCase):
raise AssertionError(f"Invalid op: {op}")
C_ref = f_ref(A, B.transpose(-2, -1), offs=offs)
C = f(A, B.transpose(-2, -1), offs=offs)
torch.testing.assert_close(C, C_ref)
if not IS_BIG_GPU and max_autotune:
with self.assertRaisesRegex(torch._inductor.exc.InductorError, "NoValidChoicesError"):
C = f(A, B.transpose(-2, -1), offs=offs)
else:
C = f(A, B.transpose(-2, -1), offs=offs)
self.assertEqual(C, C_ref)
@onlyCUDA