pytorch

a30b3595 - fix double backward for `binary_cross_entropy` loss function when `reduction=sum`. (#59479)

fix double backward for `binary_cross_entropy` loss function when `reduction=sum`. (#59479) Summary: Fixes https://github.com/pytorch/pytorch/issues/59477. ```python In [1]: import torch In [2]: x = torch.rand(3, 3, dtype=torch.double, requires_grad=True) In [3]: y = torch.rand(3, 3, dtype=torch.double) In [4]: torch.autograd.gradgradcheck(lambda x, y: torch.nn.functional.binary_cross_entropy(x, y, reduction='sum'), [x, y]) Out[4]: True In [5]: torch.autograd.gradgradcheck(lambda x, y: torch.nn.functional.binary_cross_entropy(x, y, reduction='mean'), [x, y]) Out[5]: True In [6]: torch.autograd.gradcheck(lambda x, y: torch.nn.functional.binary_cross_entropy(x, y, reduction='sum'), [x, y]) Out[6]: True ``` More comprehensive testing could be added in https://github.com/pytorch/pytorch/pull/59447 where explicit `gradcheck` and `gradgradcheck` tests are added. Pull Request resolved: https://github.com/pytorch/pytorch/pull/59479 Reviewed By: ejguan Differential Revision: D28934354 Pulled By: albanD fbshipit-source-id: 12ce68e3c5c499b2531f7cdba3c22548d67e07e9