frozenleaves/pytorch
1
0
Fork 0
mirror of https://github.com/pytorch/pytorch.git synced 2025-10-21 21:49:24 +08:00
Code Issues Packages Projects Releases Wiki Activity

[PyTorch] Make TensorImpl::sizes() customizable and disable it for NestedTensorImpl (#73817)

Browse Source 
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/73817

NestedTensorImpl doesn't have sizes(). Silently getting wrong results back from it is not conducive to efficient software development. Make it throw while allowing sizes() to be inlined in the common case anyway, just like is_contiguous(). Thanks ezyang for the reminder that we could do this.
ghstack-source-id: 151302903

Test Plan: Updated test_nestedtensor.py

Reviewed By: ezyang

Differential Revision: D34660829

fbshipit-source-id: 1289f21127d6a8359893f9174f3c430a290f2c7f
(cherry picked from commit 7098b9fcfbd25a03bac19e1148426ff073810edd)
This commit is contained in:
Scott Wolchok
2022-03-15 11:40:44 -07:00
committed by PyTorch MergeBot
parent 4646caede9
commit 90be8fa279
5 changed files with 71 additions and 15 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
test/test_nestedtensor.py
View File

@ -133,14 +133,15 @@ class TestNestedTensor(TestCase):
RuntimeError, "numel is disabled", lambda: a1.numel(),
)
@unittest.skipIf(IS_FBCODE, "size is not virtual in fbcode.")
@torch.inference_mode()
def test_size(self):
for constructor in _iter_constructors():
a1 = constructor([])
self.assertRaisesRegex(
RuntimeError,
"NestedTensorImpl doesn't support sizes",
"Tensors of type NestedTensorImpl do not have sizes"
if IS_FBCODE
else "NestedTensorImpl doesn't support sizes",
lambda: a1.size(),
)