def test_compute_loss_no_scaling(self):
exam = torch.Tensor([4.0]).repeat(2, 20, 5)
exam2 = torch.Tensor([1.0]).repeat(2, 20, 5)
exam11 = torch.Tensor([4.0]).repeat(2, 20)
exam1 = torch.Tensor([1.0]).repeat(2, 20)
d = DilateLoss()
compute_loss(exam11, exam1, torch.rand(1, 20), d, None)
# compute_loss(exam, exam2, torch.rand(2, 20), DilateLoss(), None)
result = compute_loss(exam, exam2, torch.rand(2, 20), torch.nn.MSELoss(), None)
self.assertEqual(float(result), 9.0)
def test_compute_loss(self):
crit = self.model.crit[0]
loss = compute_loss(torch.ones(2, 20), torch.zeros(2, 20), torch.rand(3, 20, 1), crit, None, None)
self.assertEqual(loss.item(), 1.0)