adds available_device to test_entropy.py #3335 (#3358)

* adds available_device to test_entropy.py #3335

* fix test_case producing large values

Author: BanzaiTokyo

tests/ignite/metrics/test_entropy.py

@@ -45,8 +45,9 @@ def test_case(request):
 
 
 @pytest.mark.parametrize("n_times", range(5))
-def test_compute(n_times, test_case):
-    ent = Entropy()
+def test_compute(n_times, test_case, available_device):
+    ent = Entropy(device=available_device)
+    assert ent._device == torch.device(available_device)
 
     y_pred, y, batch_size = test_case
 
@@ -59,14 +60,15 @@ def test_compute(n_times, test_case):
     else:
         ent.update((y_pred, y))
 
-    np_res = np_entropy(y_pred.numpy())
+    np_res = np_entropy(y_pred.cpu().numpy())
 
     assert isinstance(ent.compute(), float)
     assert pytest.approx(ent.compute()) == np_res
 
 
-def test_accumulator_detached():
-    ent = Entropy()
+def test_accumulator_detached(available_device):
+    ent = Entropy(device=available_device)
+    assert ent._device == torch.device(available_device)
 
     y_pred = torch.tensor([[2.0, 3.0], [-2.0, -1.0]], requires_grad=True)
     y = torch.zeros(2)

tests/ignite/metrics/test_hsic.py

@@ -77,7 +77,7 @@ def test_case(request) -> Tuple[Tensor, Tensor, int]:
         x = torch.randn(N, 5)
         y = x @ torch.normal(0.0, math.pi, size=(5, 3))
         y = (
-            torch.stack([torch.sin(y[:, 0]), torch.cos(y[:, 1]), torch.exp(y[:, 2])], dim=1)
+            torch.stack([torch.sin(y[:, 0]), torch.cos(y[:, 1]), torch.exp(y[:, 2]) / 10], dim=1)
             + torch.randn_like(y) * 1e-4
         )