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

[BE] Unskip some tensor creation tests on Mac (#146952)

Browse Source 
Followup after https://github.com/pytorch/pytorch/pull/145367

One should never use skip, but rather xfail otherwise one never knows when test is finally fixed.

`test_float_to_int_conversion_finite` were fixed on MacOS a while back (guess since the time Intel builds were disbaled), while `test_float_to_int_conversion_nonfinite` is fixed by https://github.com/pytorch/pytorch/pull/145367 that selects architecture-appropriate reference values for Arm ISA

Note, that results of floating to integral types cast are undefined if floating point value is outside of integral dynamic range

"Fixes" https://github.com/pytorch/pytorch/issues/38752

Pull Request resolved: https://github.com/pytorch/pytorch/pull/146952
Approved by: https://github.com/atalman, https://github.com/seemethere
This commit is contained in:
Nikita Shulga
2025-02-12 01:59:15 +00:00
committed by PyTorch MergeBot
parent 78ebd3c502
commit 0acbf8039a
1 changed files with 1 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
test/test_tensor_creation_ops.py
View File

@ -26,9 +26,7 @@ from torch.testing._internal.common_utils import (
set_default_dtype,
set_default_tensor_type,
TEST_SCIPY,
IS_MACOS,
IS_PPC,
IS_JETSON,
IS_WINDOWS,
IS_FBCODE,
IS_SANDCASTLE,
@ -1051,8 +1049,6 @@ class TestTensorCreation(TestCase):
# errors with UBSAN. These casts are deliberate in PyTorch, however, and
# NumPy may have the same behavior.
@onlyNativeDeviceTypes
@unittest.skipIf(IS_MACOS or IS_JETSON, "Test is broken on MacOS and Jetson, \
see https://github.com/pytorch/pytorch/issues/38752")
@unittest.skipIf(IS_PPC, "Test is broken on PowerPC, see https://github.com/pytorch/pytorch/issues/39671")
@dtypes(torch.bool, torch.uint8, torch.int8, torch.int16, torch.int32, torch.int64)
def test_float_to_int_conversion_finite(self, device, dtype):
@ -1081,10 +1077,10 @@ class TestTensorCreation(TestCase):
self._float_to_int_conversion_helper(vals, device, dtype, refs)
# Note: CUDA will fail this test on most dtypes, often dramatically.
# Note: This test validates undefined behavior consistency in float-to-ints casts
# NB: torch.uint16, torch.uint32, torch.uint64 excluded as this
# nondeterministically fails, warning "invalid value encountered in cast"
@onlyCPU
@unittest.skipIf(IS_MACOS, "Nonfinite conversion results on MacOS are different from others.")
@unittest.skipIf(IS_S390X, "Test fails for int16 on s390x. Needs investigation.")
@dtypes(torch.bool, torch.uint8, torch.int8, torch.int16, torch.int32, torch.int64)
def test_float_to_int_conversion_nonfinite(self, device, dtype):