out.patch

diff --git a/test/common_quantization.py b/test/common_quantization.py
index 188fbf200..13dbc6358 100644
--- a/test/common_quantization.py
+++ b/test/common_quantization.py
@@ -55,6 +55,10 @@ class QuantizationTestCase(TestCase):
         self.train_data = [(torch.rand(20, 5, dtype=torch.float), torch.randint(0, 1, (20,), dtype=torch.long)) for _ in range(20)]
         self.img_data = [(torch.rand(20, 3, 10, 10, dtype=torch.float), torch.randint(0, 1, (20,), dtype=torch.long))
                          for _ in range(20)]
+        fullname = self.id().lower()
+        is_cuda_test = 'gpu' in fullname or 'cuda' in fullname or 'cuda' in torch.tensor([]).device.type
+        if not is_cuda_test:
+            raise unittest.SkipTest("Test is probably not a CUDA test. Skipping to save time.")
 
     def checkNoPrepModules(self, module):
         r"""Checks the module does not contain child
diff --git a/test/common_utils.py b/test/common_utils.py
index 6384b8683..6606d9c60 100644
--- a/test/common_utils.py
+++ b/test/common_utils.py
@@ -465,6 +465,10 @@ class TestCase(expecttest.TestCase):
                 raise unittest.SkipTest("test is fast; we disabled it with PYTORCH_TEST_SKIP_FAST")
 
         set_rng_seed(SEED)
+        fullname = self.id().lower()
+        is_cuda_test = 'gpu' in fullname or 'cuda' in fullname or 'cuda' in torch.tensor([]).device.type
+        if not is_cuda_test:
+            raise unittest.SkipTest("Test is probably not a CUDA test. Skipping to save time.")
 
     def assertTensorsSlowEqual(self, x, y, prec=None, message=''):
         max_err = 0
diff --git a/test/jit_utils.py b/test/jit_utils.py
index e98f64fcb..384fc9ed3 100644
--- a/test/jit_utils.py
+++ b/test/jit_utils.py
@@ -72,6 +72,10 @@ class JitTestCase(TestCase):
             torch.jit.TracerWarning.ignore_lib_warnings()
             JitTestCase._restored_warnings = True
         self.setHooks()
+        fullname = self.id().lower()
+        is_cuda_test = 'gpu' in fullname or 'cuda' in fullname or 'cuda' in torch.tensor([]).device.type
+        if not is_cuda_test:
+            raise unittest.SkipTest("Test is probably not a CUDA test. Skipping to save time.")
 
     def tearDown(self):
         super(JitTestCase, self).tearDown()
diff --git a/test/test_c10d.py b/test/test_c10d.py
index 3196cd7bc..e3ec199c5 100644
--- a/test/test_c10d.py
+++ b/test/test_c10d.py
@@ -241,6 +241,10 @@ class FileStoreTest(TestCase, StoreTestBase):
     def setUp(self):
         super(FileStoreTest, self).setUp()
         self.file = tempfile.NamedTemporaryFile(delete=False)
+        fullname = self.id().lower()
+        is_cuda_test = 'gpu' in fullname or 'cuda' in fullname or 'cuda' in torch.tensor([]).device.type
+        if not is_cuda_test:
+            raise unittest.SkipTest("Test is probably not a CUDA test. Skipping to save time.")
 
     def _create_store(self):
         store = c10d.FileStore(self.file.name, 1)
@@ -255,6 +259,10 @@ class PrefixFileStoreTest(TestCase, StoreTestBase):
         self.filestore = c10d.FileStore(self.file.name, 1)
         self.prefix = "test_prefix"
         self.filestore.set_timeout(timedelta(seconds=300))
+        fullname = self.id().lower()
+        is_cuda_test = 'gpu' in fullname or 'cuda' in fullname or 'cuda' in torch.tensor([]).device.type
+        if not is_cuda_test:
+            raise unittest.SkipTest("Test is probably not a CUDA test. Skipping to save time.")
 
     def _create_store(self):
         return c10d.PrefixStore(self.prefix, self.filestore)
@@ -301,6 +309,10 @@ class PrefixTCPStoreTest(TestCase, StoreTestBase):
         self.tcpstore = create_tcp_store('localhost')
         self.prefix = "test_prefix"
         self.tcpstore.set_timeout(timedelta(seconds=300))
+        fullname = self.id().lower()
+        is_cuda_test = 'gpu' in fullname or 'cuda' in fullname or 'cuda' in torch.tensor([]).device.type
+        if not is_cuda_test:
+            raise unittest.SkipTest("Test is probably not a CUDA test. Skipping to save time.")
 
     def _create_store(self):
         return c10d.PrefixStore(self.prefix, self.tcpstore)
@@ -519,6 +531,10 @@ class MultiProcessTestCase(TestCase):
         self.rank = self.MAIN_PROCESS_RANK
         self.file = tempfile.NamedTemporaryFile(delete=False)
         self.processes = [self._spawn_process(rank) for rank in range(int(self.world_size))]
+        fullname = self.id().lower()
+        is_cuda_test = 'gpu' in fullname or 'cuda' in fullname or 'cuda' in torch.tensor([]).device.type
+        if not is_cuda_test:
+            raise unittest.SkipTest("Test is probably not a CUDA test. Skipping to save time.")
 
     def tearDown(self):
         super(MultiProcessTestCase, self).tearDown()
@@ -2781,6 +2797,10 @@ class ReducerTest(TestCase):
     def setUp(self):
         self.store = c10d.FileStore("/dev/null", 1)
         self.process_group = c10d.ProcessGroupGloo(self.store, 0, 1)
+        fullname = self.id().lower()
+        is_cuda_test = 'gpu' in fullname or 'cuda' in fullname or 'cuda' in torch.tensor([]).device.type
+        if not is_cuda_test:
+            raise unittest.SkipTest("Test is probably not a CUDA test. Skipping to save time.")
 
     def test_single_dtype_single_bucket(self):
         model = ReducerModule()
diff --git a/test/test_cpp_extensions.py b/test/test_cpp_extensions.py
index 72fc899e4..a05d0a56a 100644
--- a/test/test_cpp_extensions.py
+++ b/test/test_cpp_extensions.py
@@ -54,6 +54,10 @@ class TestCppExtension(common.TestCase):
         default_build_root = torch.utils.cpp_extension.get_default_build_root()
         if os.path.exists(default_build_root):
             shutil.rmtree(default_build_root)
+        fullname = self.id().lower()
+        is_cuda_test = 'gpu' in fullname or 'cuda' in fullname or 'cuda' in torch.tensor([]).device.type
+        if not is_cuda_test:
+            raise unittest.SkipTest("Test is probably not a CUDA test. Skipping to save time.")
 
     def test_extension_function(self):
         x = torch.randn(4, 4)
diff --git a/test/test_dataloader.py b/test/test_dataloader.py
index 38e3d13d3..1db5c7948 100644
--- a/test/test_dataloader.py
+++ b/test/test_dataloader.py
@@ -647,6 +647,10 @@ class TestDataLoader(TestCase):
         self.data = torch.randn(100, 2, 3, 5)
         self.labels = torch.randperm(50).repeat(2)
         self.dataset = TensorDataset(self.data, self.labels)
+        fullname = self.id().lower()
+        is_cuda_test = 'gpu' in fullname or 'cuda' in fullname or 'cuda' in torch.tensor([]).device.type
+        if not is_cuda_test:
+            raise unittest.SkipTest("Test is probably not a CUDA test. Skipping to save time.")
 
     def _test_sequential(self, loader):
         batch_size = loader.batch_size
@@ -1458,6 +1462,10 @@ class TestStringDataLoader(TestCase):
     def setUp(self):
         super(TestStringDataLoader, self).setUp()
         self.dataset = StringDataset()
+        fullname = self.id().lower()
+        is_cuda_test = 'gpu' in fullname or 'cuda' in fullname or 'cuda' in torch.tensor([]).device.type
+        if not is_cuda_test:
+            raise unittest.SkipTest("Test is probably not a CUDA test. Skipping to save time.")
 
     @unittest.skipIf(not TEST_CUDA, "CUDA unavailable")
     def test_shuffle_pin_memory(self):
@@ -1484,6 +1492,10 @@ class TestDictDataLoader(TestCase):
     def setUp(self):
         super(TestDictDataLoader, self).setUp()
         self.dataset = DictDataset()
+        fullname = self.id().lower()
+        is_cuda_test = 'gpu' in fullname or 'cuda' in fullname or 'cuda' in torch.tensor([]).device.type
+        if not is_cuda_test:
+            raise unittest.SkipTest("Test is probably not a CUDA test. Skipping to save time.")
 
     def test_sequential_batch(self):
         loader = DataLoader(self.dataset, batch_size=2, shuffle=False)
@@ -1528,6 +1540,10 @@ class TestNamedTupleDataLoader(TestCase):
     def setUp(self):
         super(TestNamedTupleDataLoader, self).setUp()
         self.dataset = NamedTupleDataset()
+        fullname = self.id().lower()
+        is_cuda_test = 'gpu' in fullname or 'cuda' in fullname or 'cuda' in torch.tensor([]).device.type
+        if not is_cuda_test:
+            raise unittest.SkipTest("Test is probably not a CUDA test. Skipping to save time.")
 
     @unittest.skipIf(not TEST_CUDA, "CUDA unavailable")
     def test_collate_and_pin_memory_with_namedtuple(self):
@@ -1576,6 +1592,10 @@ class TestCustomPinFn(TestCase):
         inps = torch.arange(10 * 5, dtype=torch.float32).view(10, 5)
         tgts = torch.arange(10 * 5, dtype=torch.float32).view(10, 5)
         self.dataset = TensorDataset(inps, tgts)
+        fullname = self.id().lower()
+        is_cuda_test = 'gpu' in fullname or 'cuda' in fullname or 'cuda' in torch.tensor([]).device.type
+        if not is_cuda_test:
+            raise unittest.SkipTest("Test is probably not a CUDA test. Skipping to save time.")
 
     @unittest.skipIf(not TEST_CUDA, "CUDA unavailable")
     @skipIfRocm
@@ -1627,6 +1647,10 @@ class TestIndividualWorkerQueue(TestCase):
     def setUp(self):
         super(TestIndividualWorkerQueue, self).setUp()
         self.dataset = TestWorkerQueueDataset([i for i in range(128)])
+        fullname = self.id().lower()
+        is_cuda_test = 'gpu' in fullname or 'cuda' in fullname or 'cuda' in torch.tensor([]).device.type
+        if not is_cuda_test:
+            raise unittest.SkipTest("Test is probably not a CUDA test. Skipping to save time.")
 
     def _run_ind_worker_queue_test(self, batch_size, num_workers):
         loader = DataLoader(
diff --git a/test/test_distributed.py b/test/test_distributed.py
index 5c4fedad2..5a818734a 100644
--- a/test/test_distributed.py
+++ b/test/test_distributed.py
@@ -1617,6 +1617,10 @@ if BACKEND == "gloo" or BACKEND == "nccl":
             Barrier.init()
             for rank in range(int(WORLD_SIZE)):
                 self.processes.append(self._spawn_process(rank))
+            fullname = self.id().lower()
+            is_cuda_test = 'gpu' in fullname or 'cuda' in fullname or 'cuda' in torch.tensor([]).device.type
+            if not is_cuda_test:
+                raise unittest.SkipTest("Test is probably not a CUDA test. Skipping to save time.")
 
         def tearDown(self):
             super(TestDistBackend, self).tearDown()
diff --git a/test/test_distributions.py b/test/test_distributions.py
index 62a19cd1b..2ca1e1b62 100644
--- a/test/test_distributions.py
+++ b/test/test_distributions.py
@@ -2727,6 +2727,10 @@ class TestDistributionShapes(TestCase):
         self.tensor_sample_1 = torch.ones(3, 2)
         self.tensor_sample_2 = torch.ones(3, 2, 3)
         Distribution.set_default_validate_args(True)
+        fullname = self.id().lower()
+        is_cuda_test = 'gpu' in fullname or 'cuda' in fullname or 'cuda' in torch.tensor([]).device.type
+        if not is_cuda_test:
+            raise unittest.SkipTest("Test is probably not a CUDA test. Skipping to save time.")
 
     def tearDown(self):
         super(TestDistributionShapes, self).tearDown()
@@ -3111,6 +3115,10 @@ class TestKL(TestCase):
 
     def setUp(self):
         super(TestKL, self).setUp()
+        fullname = self.id().lower()
+        is_cuda_test = 'gpu' in fullname or 'cuda' in fullname or 'cuda' in torch.tensor([]).device.type
+        if not is_cuda_test:
+            raise unittest.SkipTest("Test is probably not a CUDA test. Skipping to save time.")
 
         class Binomial30(Binomial):
             def __init__(self, probs):
@@ -3628,6 +3636,10 @@ class TestLazyLogitsInitialization(TestCase):
         super(TestLazyLogitsInitialization, self).setUp()
         self.examples = [e for e in EXAMPLES if e.Dist in
                          (Categorical, OneHotCategorical, Bernoulli, Binomial, Multinomial)]
+        fullname = self.id().lower()
+        is_cuda_test = 'gpu' in fullname or 'cuda' in fullname or 'cuda' in torch.tensor([]).device.type
+        if not is_cuda_test:
+            raise unittest.SkipTest("Test is probably not a CUDA test. Skipping to save time.")
 
     def test_lazy_logits_initialization(self):
         for Dist, params in self.examples:
@@ -3772,6 +3784,10 @@ class TestAgainstScipy(TestCase):
                 scipy.stats.weibull_min(c=positive_var2[0], scale=positive_var[0])
             )
         ]
+        fullname = self.id().lower()
+        is_cuda_test = 'gpu' in fullname or 'cuda' in fullname or 'cuda' in torch.tensor([]).device.type
+        if not is_cuda_test:
+            raise unittest.SkipTest("Test is probably not a CUDA test. Skipping to save time.")
 
     def test_mean(self):
         for pytorch_dist, scipy_dist in self.distribution_pairs:
@@ -3870,6 +3886,10 @@ class TestTransforms(TestCase):
             self.transforms += transforms
             if cache_size == 0:
                 self.unique_transforms = transforms[:]
+        fullname = self.id().lower()
+        is_cuda_test = 'gpu' in fullname or 'cuda' in fullname or 'cuda' in torch.tensor([]).device.type
+        if not is_cuda_test:
+            raise unittest.SkipTest("Test is probably not a CUDA test. Skipping to save time.")
 
     def _generate_data(self, transform):
         domain = transform.domain
@@ -4312,6 +4332,10 @@ class TestValidation(TestCase):
     def setUp(self):
         super(TestCase, self).setUp()
         Distribution.set_default_validate_args(True)
+        fullname = self.id().lower()
+        is_cuda_test = 'gpu' in fullname or 'cuda' in fullname or 'cuda' in torch.tensor([]).device.type
+        if not is_cuda_test:
+            raise unittest.SkipTest("Test is probably not a CUDA test. Skipping to save time.")
 
     def test_valid(self):
         for Dist, params in EXAMPLES:
diff --git a/test/test_nn.py b/test/test_nn.py
index 0d663c1c4..b5155d819 100644
--- a/test/test_nn.py
+++ b/test/test_nn.py
@@ -8905,6 +8905,10 @@ class TestNNInit(TestCase):
     def setUp(self):
         super(TestNNInit, self).setUp()
         random.seed(123)
+        fullname = self.id().lower()
+        is_cuda_test = 'gpu' in fullname or 'cuda' in fullname or 'cuda' in torch.tensor([]).device.type
+        if not is_cuda_test:
+            raise unittest.SkipTest("Test is probably not a CUDA test. Skipping to save time.")
 
     def _is_normal(self, tensor, mean, std):
         samples = tensor.view(-1).tolist()
diff --git a/test/test_optim.py b/test/test_optim.py
index e334962c9..aa0b1c11e 100644
--- a/test/test_optim.py
+++ b/test/test_optim.py
@@ -541,6 +541,10 @@ class TestLRScheduler(TestCase):
         self.opt = SGD(
             [{'params': self.net.conv1.parameters()}, {'params': self.net.conv2.parameters(), 'lr': 0.5}],
             lr=0.05)
+        fullname = self.id().lower()
+        is_cuda_test = 'gpu' in fullname or 'cuda' in fullname or 'cuda' in torch.tensor([]).device.type
+        if not is_cuda_test:
+            raise unittest.SkipTest("Test is probably not a CUDA test. Skipping to save time.")
 
     def test_old_pattern_warning(self):
         epochs = 35
diff --git a/test/test_sparse.py b/test/test_sparse.py
index e105f9113..201f3d61a 100644
--- a/test/test_sparse.py
+++ b/test/test_sparse.py
@@ -60,6 +60,10 @@ class TestSparse(TestCase):
         self.sparse_tensor = sparse_tensor_factory
         self.legacy_sparse_tensor = torch.sparse.DoubleTensor
         super(TestSparse, self).setUp()
+        fullname = self.id().lower()
+        is_cuda_test = 'gpu' in fullname or 'cuda' in fullname or 'cuda' in torch.tensor([]).device.type
+        if not is_cuda_test:
+            raise unittest.SkipTest("Test is probably not a CUDA test. Skipping to save time.")
 
     def _gen_sparse(self, sparse_dim, nnz, with_size):
         if isinstance(with_size, Number):
diff --git a/test/test_utils.py b/test/test_utils.py
index 0970f6656..c2e3b2fa8 100644
--- a/test/test_utils.py
+++ b/test/test_utils.py
@@ -287,6 +287,10 @@ class TestDataLoader(TestCase):
     def setUp(self):
         self.dataset = torch.randn(5, 3, 3, 2)
         self.batch_size = 3
+        fullname = self.id().lower()
+        is_cuda_test = 'gpu' in fullname or 'cuda' in fullname or 'cuda' in torch.tensor([]).device.type
+        if not is_cuda_test:
+            raise unittest.SkipTest("Test is probably not a CUDA test. Skipping to save time.")
 
     def test_random_seed(self):
         def run():