frozenleaves/pytorch
1
0
Fork 0
mirror of https://github.com/pytorch/pytorch.git synced 2025-10-20 12:54:11 +08:00
Code Issues Packages Projects Releases Wiki Activity

[CUDA] try to abate some flakiness in test_stream_event_nogil (#148796)

Browse Source 
threshold twiddling as one in a few dozen runs tend to fail the current threshold

Pull Request resolved: https://github.com/pytorch/pytorch/pull/148796
Approved by: https://github.com/Skylion007
This commit is contained in:
eqy
2025-03-12 19:12:47 +00:00
committed by PyTorch MergeBot
parent 215f856142
commit b90698f5ba
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
test/test_cuda_multigpu.py
View File

@ -815,8 +815,8 @@ class TestCudaMultiGPU(TestCase):
# it may vary on different hardware in different environments.
# Therefore, this test uses relative comparisons, checking if the
# sum of parent and child threads execution time is greater than the
# real execution time by least 40%.
self.assertGreater(parent_time + child_time, total_time * 1.4)
# real execution time by least 30%.
self.assertGreater(parent_time + child_time, total_time * 1.3)
# This test is flaky for ROCm, see issue #62602
@skipIfRocm