Reduce number of samples in {svd,pca}_lowrank OpInfos (#127199)

We don't need to generate so many samples for these very expensive ops.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/127199
Approved by: https://github.com/peterbell10, https://github.com/zou3519

test/functorch/test_ops.py

@@ -457,11 +457,7 @@ class TestOperators(TestCase):
                 {torch.float32: tol(atol=1e-04, rtol=1e-04)},
             ),
             tol1("masked.cumprod", {torch.float32: tol(atol=1e-05, rtol=1e-05)}),
-            tol1(
-                "svd_lowrank",
-                {torch.float32: tol(atol=3e-04, rtol=3e-04)},
-                device_type="cuda",
-            ),
+            tol1("svd_lowrank", {torch.float32: tol(atol=3e-04, rtol=3e-04)}),
             tol1(
                 "linalg.multi_dot",
                 {torch.float32: tol(atol=1e-05, rtol=8e-04)},

torch/testing/_internal/common_methods_invocations.py

@@ -2108,14 +2108,13 @@ def sample_inputs_singular_matrix_factors(op_info, device, dtype, requires_grad=
     """
 
     make_arg = partial(make_tensor, device=device, dtype=dtype, requires_grad=requires_grad)
-    batches = [(), (0, ), (2, ), (1, 1)]
-    size = [1, 5, 10]
-
+    batches = [(), (2,)]
+    size = [3, 4]
     for batch, m, n in product(batches, size, size):
-        for k in range(min(3, m, n)):
-            a = make_arg((*batch, m, k))
-            b = make_arg((*batch, n, k))
-            yield a, b
+        k = 2
+        a = make_arg((*batch, m, k))
+        b = make_arg((*batch, n, k))
+        yield a, b
 
 
 def sample_inputs_svd_lowrank(op_info, device, dtype, requires_grad=False, **kwargs):
@@ -17980,10 +17979,6 @@ op_db: List[OpInfo] = [
                                     'TestFwdGradients',
                                     'test_fn_fwgrad_bwgrad',
                                     dtypes=[torch.complex128]),
-                       DecorateInfo(unittest.skip("See comment above"),
-                                    'TestBwdGradientsCUDA',
-                                    'test_fn_gradgrad',
-                                    dtypes=[torch.complex128]),
                        ],
            skips=(
                # test does not work with passing lambda for op