Fix flakiness on StochasticDepth test (#4758)

* Fix flakiness on the TestStochasticDepth test.

* Fix minor bug when p=1.0

* Remove device and dtype setting.

Author: datumbox

test/test_ops.py

@@ -1149,13 +1149,15 @@ def _create_masks(image, masks):
 
 
 class TestStochasticDepth:
+    @pytest.mark.parametrize("seed", range(10))
     @pytest.mark.parametrize("p", [0.2, 0.5, 0.8])
     @pytest.mark.parametrize("mode", ["batch", "row"])
-    def test_stochastic_depth(self, mode, p):
+    def test_stochastic_depth_random(self, seed, mode, p):
+        torch.manual_seed(seed)
         stats = pytest.importorskip("scipy.stats")
         batch_size = 5
         x = torch.ones(size=(batch_size, 3, 4, 4))
-        layer = ops.StochasticDepth(p=p, mode=mode).to(device=x.device, dtype=x.dtype)
+        layer = ops.StochasticDepth(p=p, mode=mode)
         layer.__repr__()
 
         trials = 250
@@ -1173,7 +1175,22 @@ def test_stochastic_depth(self, mode, p):
                 num_samples += batch_size
 
         p_value = stats.binom_test(counts, num_samples, p=p)
-        assert p_value > 0.0001
+        assert p_value > 0.01
+
+    @pytest.mark.parametrize("seed", range(10))
+    @pytest.mark.parametrize("p", (0, 1))
+    @pytest.mark.parametrize("mode", ["batch", "row"])
+    def test_stochastic_depth(self, seed, mode, p):
+        torch.manual_seed(seed)
+        batch_size = 5
+        x = torch.ones(size=(batch_size, 3, 4, 4))
+        layer = ops.StochasticDepth(p=p, mode=mode)
+
+        out = layer(x)
+        if p == 0:
+            assert out.equal(x)
+        elif p == 1:
+            assert out.equal(torch.zeros_like(x))
 
 
 class TestUtils:

torchvision/ops/stochastic_depth.py

@@ -34,7 +34,9 @@ def stochastic_depth(input: Tensor, p: float, mode: str, training: bool = True)
     else:
         size = [1] * input.ndim
     noise = torch.empty(size, dtype=input.dtype, device=input.device)
-    noise = noise.bernoulli_(survival_rate).div_(survival_rate)
+    noise = noise.bernoulli_(survival_rate)
+    if survival_rate > 0.0:
+        noise.div_(survival_rate)
     return input * noise