def test_entropy_for_uniform_distribution(self, device: str):
metric = Entropy()
logits = torch.tensor([[1, 1, 1, 1], [1, 1, 1, 1]], dtype=torch.float, device=device)
metric(logits)
assert_allclose(metric.get_metric()["entropy"], 1.38629436)
# actual values shouldn't effect uniform distribution:
logits = torch.tensor([[2, 2, 2, 2], [2, 2, 2, 2]], dtype=torch.float, device=device)
metric(logits)
assert_allclose(metric.get_metric()["entropy"], 1.38629436)
metric.reset()
assert metric._entropy == 0.0
assert metric._count == 0.0
def test_entropy_for_uniform_distribution(self):
metric = Entropy()
logits = torch.Tensor([[1, 1, 1, 1], [1, 1, 1, 1]])
metric(logits)
numpy.testing.assert_almost_equal(metric.get_metric(), 1.38629436)
# actual values shouldn't effect uniform distribution:
logits = torch.Tensor([[2, 2, 2, 2], [2, 2, 2, 2]])
metric(logits)
numpy.testing.assert_almost_equal(metric.get_metric(), 1.38629436)
metric.reset()
assert metric._entropy == 0.0
assert metric._count == 0.0
def test_entropy_for_uniform_distribution(self):
metric = Entropy()
logits = torch.Tensor([[1, 1, 1, 1],
[1, 1, 1, 1]])
metric(logits)
numpy.testing.assert_almost_equal(metric.get_metric(), 1.38629436)
# actual values shouldn't effect uniform distribution:
logits = torch.Tensor([[2, 2, 2, 2],
[2, 2, 2, 2]])
metric(logits)
numpy.testing.assert_almost_equal(metric.get_metric(), 1.38629436)
metric.reset()
assert metric._entropy == 0.0
assert metric._count == 0.0
