Skip to content

fix _pointwise_loss for target gradients #10018

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
li-roy wants to merge 2 commits into from
Closed

fix _pointwise_loss for target gradients #10018

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
34d5f80
fix _pointwise_loss for target gradients
Jul 30, 2018
c224552
add test
Jul 31, 2018
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 6 additions & 0 deletions test/test_nn.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4812,6 +4812,12 @@ def test_triplet_margin_loss_swap_no_reduce(self):
self.assertEqual(F.triplet_margin_loss(input1, input2, input3, swap=True, reduction='none'),
loss_reference_fns['TripletMarginLoss'](input1, input2, input3, swap=True, reduction='none'))

def test_pointwise_loss_target_grad_none_reduction(self):
i = torch.randn(5, 10)
t = torch.randn(5, 10, requires_grad=True)
self.assertEqual(F.mse_loss(i, t, reduction='none').size(), t.size())
self.assertEqual(F.l1_loss(i, t, reduction='none').size(), t.size())

def test_cosine_similarity(self):
input1 = torch.randn(4, 4, requires_grad=True)
input2 = torch.randn(4, 4, requires_grad=True)
Expand Down
10 changes: 3 additions & 7 deletions torch/nn/functional.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1671,7 +1671,7 @@ def _pointwise_loss(lambd, lambd_optimized, input, target, reduction='elementwis
return d
return torch.mean(d) if reduction == 'elementwise_mean' else torch.sum(d)
else:
return lambd_optimized(input, target, reduction)
return lambd_optimized(input, target, _Reduction.get_enum(reduction))


def smooth_l1_loss(input, target, size_average=None, reduce=None, reduction='elementwise_mean'):
Expand All @@ -1695,9 +1695,7 @@ def l1_loss(input, target, size_average=None, reduce=None, reduction='elementwis
See :class:`~torch.nn.L1Loss` for details.
"""
if size_average is not None or reduce is not None:
reduction = _Reduction.legacy_get_enum(size_average, reduce)
else:
reduction = _Reduction.get_enum(reduction)
reduction = _Reduction.legacy_get_string(size_average, reduce)
return _pointwise_loss(lambda a, b: torch.abs(a - b), torch._C._nn.l1_loss,
input, target, reduction)

Expand All @@ -1710,9 +1708,7 @@ def mse_loss(input, target, size_average=None, reduce=None, reduction='elementwi
See :class:`~torch.nn.MSELoss` for details.
"""
if size_average is not None or reduce is not None:
reduction = _Reduction.legacy_get_enum(size_average, reduce)
else:
reduction = _Reduction.get_enum(reduction)
reduction = _Reduction.legacy_get_string(size_average, reduce)
return _pointwise_loss(lambda a, b: (a - b) ** 2, torch._C._nn.mse_loss, input, target, reduction)


Expand Down