def test_basic():
levels = torch.tensor([[1., 1., 0., 0.], [1., 0., 0., 0.],
[1., 1., 1., 1.]])
logits = torch.tensor([[2.1, 1.8, -2.1, -1.8], [1.9, -1., -1.5, -1.3],
[1.9, 1.8, 1.7, 1.6]])
got_val = coral_loss(logits, levels, reduction=None)
expect_val = torch.tensor([0.5370, 0.8951, 0.6441])
assert torch.allclose(got_val, expect_val, rtol=1e-03, atol=1e-05)
got_val = coral_loss(logits, levels, reduction='sum')
expect_val = torch.tensor(2.0761)
assert torch.allclose(got_val, expect_val, rtol=1e-03, atol=1e-05)
got_val = coral_loss(logits, levels)
expect_val = torch.tensor(0.6920)
assert torch.allclose(got_val, expect_val, rtol=1e-03, atol=1e-05)
def shared_step(self, batch):
x, y = batch
levels = levels_from_labelbatch(y, num_classes=self.num_classes)
logits, probas = self(x)
loss = coral_loss(logits, levels)
predicted_labels = proba_to_label(probas).float()
n_samples = x.shape[0]
mae = torch.sum(torch.abs(predicted_labels - y))
mse = torch.sum((predicted_labels - y)**2)
return loss, mae, mse, n_samples
def f():
coral_loss(logits, levels, reduction='something')
def f():
coral_loss(logits, levels)