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

[dynamo][numpy] Add unsigned integer dtypes (#125717)

Browse Source 
We should support these to whatever extent we can. They corresponding
`torch.uint<w>` types are defined, so I don't see an issue with
generating the various casting rules and allowing them to trace.

Pull Request resolved: https://github.com/pytorch/pytorch/pull/125717
Approved by: https://github.com/lezcano
This commit is contained in:
Andrew M. James
2024-06-04 17:56:39 +00:00
committed by PyTorch MergeBot
parent 4ce5322a1f
commit 879d01afcb
8 changed files with 536 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
test/test_numpy_interop.py
View File

@ -476,13 +476,18 @@ class TestNumPyInterop(TestCase):
self.assertTrue(r2.requires_grad)
@onlyCPU
def test_parse_numpy_int(self, device):
@skipIfTorchDynamo()
def test_parse_numpy_int_overflow(self, device):
# assertRaises uses a try-except which dynamo has issues with
# Only concrete class can be given where "Type[number[_64Bit]]" is expected
self.assertRaisesRegex(
RuntimeError,
"(Overflow|an integer is required)",
lambda: torch.mean(torch.randn(1, 1), np.uint64(-1)),
) # type: ignore[call-overload]
@onlyCPU
def test_parse_numpy_int(self, device):
# https://github.com/pytorch/pytorch/issues/29252
for nptype in [np.int16, np.int8, np.uint8, np.int32, np.int64]:
scalar = 3