enable unit tests (#25963)

Summary:
These unit tests pass after landing all the warp size awareness patches.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/25963

Differential Revision: D17319124

Pulled By: bddppq

fbshipit-source-id: 22f5d5f1ca9c67e66a7ccf983b2d2f889a74e729

test/test_cuda.py

@@ -2660,7 +2660,6 @@ class TestCuda(TestCase):
         tensor = tensor.unsqueeze(1)
         self.assertEqual(tensor.var(0), 0.03125)
 
-    @skipIfRocm
     def test_digamma(self):
         def test(use_double=False):
             cpu_tensor = torch.randn(10, 10, 10)
@@ -2689,7 +2688,6 @@ class TestCuda(TestCase):
         norm_errors = (gpu_out - cpu_out.cuda()) / gpu_out
         self.assertEqual(norm_errors, expected_errors)
 
-    @skipIfRocm
     def test_polygamma(self):
         def test(use_double=False):
             cpu_tensor = torch.randn(10, 10, 10)

test/test_jit.py

@@ -1875,7 +1875,6 @@ graph(%Ra, %Rb):
 
     @unittest.skipIf(not RUN_CUDA, "test_dropout_cuda require CUDA")
     @unittest.skipIf(IS_WINDOWS, "NYI: fuser support for Windows")
-    @skipIfRocm
     def test_dropout_cuda(self):
         # Dropout AD is dispatched to _fused_dropout in CUDA case,
         # which is not included in TestJitGeneratedFunctional

test/test_jit_fuser.py

@@ -47,7 +47,6 @@ class TestFuser(JitTestCase):
         self._test_fused_abs(device="cuda")
 
     @unittest.skipIf(not RUN_CUDA, "requires CUDA")
-    @skipIfRocm
     def test_zero_element_tensors(self):
         def decode(sin_t, cos_t):
             theta = torch.atan2(sin_t.float(), cos_t.float())
@@ -358,7 +357,6 @@ class TestFuser(JitTestCase):
     # If this is a real problem, we'll need to revisit Torchscript Function
     # lifetimes in Python.
     @unittest.skipIf(not RUN_CUDA, "fuser requires CUDA")
-    @skipIfRocm
     def test_lerp(self):
         start = torch.randn(4, 1, dtype=torch.float, device='cuda')
         end = torch.randn(1, 4, dtype=torch.float, device='cuda')

test/test_nn.py

@@ -3336,7 +3336,6 @@ class TestNN(NNTestCase):
         self._test_InstanceNorm_general(nn.InstanceNorm3d, input, dtype=torch.float)
 
     @unittest.skipIf(not TEST_CUDA, "CUDA unavailable")
-    @skipIfRocm
     def test_InstanceNorm3d_general_cuda(self):
         b = random.randint(3, 5)
         c = random.randint(2, 5)
@@ -4007,7 +4006,6 @@ class TestNN(NNTestCase):
         self._test_batchnorm_grad()
 
     @unittest.skipIf(not TEST_CUDA, "CUDA unavailable")
-    @skipIfRocm
     def test_batchnorm_grad_cuda(self):
         self._test_batchnorm_grad("cuda")
         if TEST_CUDNN:

test/test_optim.py

@@ -14,8 +14,7 @@ from torch import sparse
 from torch.optim.lr_scheduler import LambdaLR, StepLR, MultiStepLR, \
     ExponentialLR, CosineAnnealingLR, ReduceLROnPlateau, _LRScheduler, \
     CyclicLR, CosineAnnealingWarmRestarts, OneCycleLR
-from common_utils import TestCase, run_tests, TEST_WITH_UBSAN, load_tests, \
-    skipIfRocm
+from common_utils import TestCase, run_tests, TEST_WITH_UBSAN, load_tests
 
 # load_tests from common_utils is used to automatically filter tests for
 # sharding on sandcastle. This line silences flake warnings
@@ -285,7 +284,6 @@ class TestOptim(TestCase):
             [lambda opt: StepLR(opt, gamma=0.99999, step_size=300)]
         )
 
-    @skipIfRocm
     def test_adam(self):
         self._test_basic_cases(
             lambda weight, bias: optim.Adam([weight, bias], lr=1e-3)
@@ -401,7 +399,6 @@ class TestOptim(TestCase):
              lambda opt: ReduceLROnPlateau(opt, threshold=1e-4)]
         )
 
-    @skipIfRocm
     def test_adamax(self):
         self._test_basic_cases(
             lambda weight, bias: optim.Adamax([weight, bias], lr=1e-1)
@@ -426,7 +423,6 @@ class TestOptim(TestCase):
         with self.assertRaisesRegex(ValueError, "Invalid momentum value: -1.0"):
             optim.RMSprop(None, lr=1e-2, momentum=-1.0)
 
-    @skipIfRocm
     def test_asgd(self):
         self._test_basic_cases(
             lambda weight, bias: optim.ASGD([weight, bias], lr=1e-3, t0=100)
@@ -451,7 +447,6 @@ class TestOptim(TestCase):
         with self.assertRaisesRegex(ValueError, "Invalid eta values: 1.0, 0.5"):
             optim.Rprop(None, lr=1e-2, etas=(1.0, 0.5))
 
-    @skipIfRocm
     def test_lbfgs(self):
         self._test_basic_cases(
             lambda weight, bias: optim.LBFGS([weight, bias]),