def test_fsim_loss_reduction(x, y) -> None:
loss = FSIMLoss(reduction='mean')
measure = loss(x, y)
assert measure.dim(
) == 0, f'FSIM with `mean` reduction must return 1 number, got {len(measure)}'
loss = FSIMLoss(reduction='sum')
measure = loss(x, y)
assert measure.dim(
) == 0, f'FSIM with `mean` reduction must return 1 number, got {len(measure)}'
loss = FSIMLoss(reduction='none')
measure = loss(x, y)
assert len(measure) == x.size(0), \
f'FSIM with `none` reduction must have length equal to number of images, got {len(measure)}'
loss = FSIMLoss(reduction='random string')
with pytest.raises(ValueError):
loss(x, y)
def test_fsim_loss_reduction(prediction: torch.Tensor,
target: torch.Tensor) -> None:
loss = FSIMLoss(reduction='mean')
measure = loss(prediction, target)
assert measure.dim(
) == 0, f'FSIM with `mean` reduction must return 1 number, got {len(measure)}'
loss = FSIMLoss(reduction='sum')
measure = loss(prediction, target)
assert measure.dim(
) == 0, f'FSIM with `mean` reduction must return 1 number, got {len(measure)}'
loss = FSIMLoss(reduction='none')
measure = loss(prediction, target)
assert len(measure) == prediction.size(0), \
f'FSIM with `none` reduction must have length equal to number of images, got {len(measure)}'
loss = FSIMLoss(reduction='random string')
with pytest.raises(KeyError):
loss(prediction, target)
def test_fsim_loss_computes_grad(x, y, device: str) -> None:
x.requires_grad_()
loss_value = FSIMLoss()(x.to(device), y.to(device))
loss_value.backward()
assert x.grad is not None, 'Expected non None gradient of leaf variable'
def test_fsim_loss_computes_grad(prediction: torch.Tensor,
target: torch.Tensor, device: str) -> None:
prediction.requires_grad_()
loss_value = FSIMLoss()(prediction, target)
loss_value.backward()
assert prediction.grad is not None, 'Expected non None gradient of leaf variable'