diff --git a/python/paddle/fluid/tests/unittests/npu/CMakeLists.txt b/python/paddle/fluid/tests/unittests/npu/CMakeLists.txt
index f71e04c09aa38b8cf7b3a167b84d4dc0e6cc3ec7..4ab9262f248a21d454f9d32520fe084553cadfe7 100644
--- a/python/paddle/fluid/tests/unittests/npu/CMakeLists.txt
+++ b/python/paddle/fluid/tests/unittests/npu/CMakeLists.txt
@@ -1,6 +1,8 @@
 file(GLOB TEST_OPS RELATIVE "${CMAKE_CURRENT_SOURCE_DIR}" "test_*.py")
 string(REPLACE ".py" "" TEST_OPS "${TEST_OPS}")
 
-foreach(TEST_OP ${TEST_OPS})
-    py_test_modules(${TEST_OP} MODULES ${TEST_OP})
-endforeach(TEST_OP)
+if (WITH_ASCEND_CL)
+    foreach(TEST_OP ${TEST_OPS})
+        py_test_modules(${TEST_OP} MODULES ${TEST_OP})
+    endforeach(TEST_OP)
+endif()
diff --git a/python/paddle/fluid/tests/unittests/npu/test_abs_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_abs_op_npu.py
index 9382cf2162ef2598192e0a8e0f1bd630cbb9a6a4..3c16a24b33191d4fc6942aeaceb4dfc8d04a5e48 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_abs_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_abs_op_npu.py
@@ -25,8 +25,6 @@ import paddle.fluid as fluid
 paddle.enable_static()
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestNPUAbs(OpTest):
     def setUp(self):
         self.op_type = "abs"
diff --git a/python/paddle/fluid/tests/unittests/npu/test_accuracy_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_accuracy_op_npu.py
index 5aeca5abd9f8315e2793e6cd5ba53a283c850b85..0f55c8b5914870e775e5e78783d7fcdf3b6fc07b 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_accuracy_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_accuracy_op_npu.py
@@ -27,8 +27,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestAccuracy(OpTest):
     def setUp(self):
         self.op_type = "accuracy"
@@ -60,7 +58,7 @@ class TestAccuracy(OpTest):
         self.dtype = np.float32
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
 class TestAccuracy2(TestAccuracy):
diff --git a/python/paddle/fluid/tests/unittests/npu/test_adam_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_adam_op_npu.py
index 17dad036185d844ecb3ebcc154d78b349bab33d0..02d4002f72c492bf7ec05be9539e8e5f8bedb5a0 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_adam_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_adam_op_npu.py
@@ -25,8 +25,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestAdam(OpTest):
     def setUp(self):
         self.set_npu()
@@ -75,11 +73,9 @@ class TestAdam(OpTest):
         self.dtype = np.float32
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, atol=1e-5, check_dygraph=False)
+        self.check_output_with_place(self.place, atol=1e-5)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestAdamWithEpsilonTensor(OpTest):
     def setUp(self):
         self.set_npu()
@@ -131,11 +127,9 @@ class TestAdamWithEpsilonTensor(OpTest):
         self.dtype = np.float32
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, atol=1e-5, check_dygraph=False)
+        self.check_output_with_place(self.place, atol=1e-5)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestAdamOpWithSkipUpdate(OpTest):
     def setUp(self):
         self.set_npu()
@@ -185,11 +179,9 @@ class TestAdamOpWithSkipUpdate(OpTest):
         self.dtype = np.float32
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, atol=1e-5, check_dygraph=False)
+        self.check_output_with_place(self.place, atol=1e-5)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestAdamOpWithGlobalBetaPow(OpTest):
     def setUp(self):
         self.set_npu()
@@ -244,11 +236,9 @@ class TestAdamOpWithGlobalBetaPow(OpTest):
         self.dtype = np.float32
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, atol=1e-5, check_dygraph=False)
+        self.check_output_with_place(self.place, atol=1e-5)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestNet(unittest.TestCase):
     def _test(self, run_npu=True):
         main_prog = paddle.static.Program()
@@ -309,8 +299,6 @@ class TestNet(unittest.TestCase):
         self.assertTrue(np.allclose(npu_loss, cpu_loss, rtol=1e-3))
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestNetWithEpsilonTensor(unittest.TestCase):
     def _test(self, run_npu=True):
         main_prog = paddle.static.Program()
diff --git a/python/paddle/fluid/tests/unittests/npu/test_amp_check_finite_and_scale_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_amp_check_finite_and_scale_op_npu.py
index e92bfbb4d77d33aece688655a19847bc372f39ca..604eb32db0a6c939f8d2b1e01a5f6df1ac28e7b0 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_amp_check_finite_and_scale_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_amp_check_finite_and_scale_op_npu.py
@@ -25,8 +25,6 @@ from paddle.fluid.contrib.mixed_precision.amp_nn import check_finite_and_unscale
 paddle.enable_static()
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestCheckFiniteAndUnscale(unittest.TestCase):
     def get_prog(self):
         paddle.enable_static()
@@ -39,11 +37,11 @@ class TestCheckFiniteAndUnscale(unittest.TestCase):
                 name="status", shape=[8], dtype='float32')
             main_program.global_block().append_op(
                 type="alloc_float_status",
-                outputs={"FloatStatus": float_status}, )
+                outputs={"FloatStatus": float_status})
             main_program.global_block().append_op(
                 type="clear_float_status",
                 inputs={"FloatStatus": float_status},
-                outputs={"FloatStatusOut": float_status}, )
+                outputs={"FloatStatusOut": float_status})
             c = paddle.fluid.layers.elementwise_div(a, b)
             out, found_inf = check_finite_and_unscale(
                 [c], scale, float_status=float_status)
@@ -95,8 +93,6 @@ class TestCheckFiniteAndUnscale(unittest.TestCase):
         self.assertFalse(found_inf[0])
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestCheckFiniteAndUnscaleClearFloatStatus(unittest.TestCase):
     def get_prog(self):
         paddle.enable_static()
@@ -109,21 +105,21 @@ class TestCheckFiniteAndUnscaleClearFloatStatus(unittest.TestCase):
                 name="status", shape=[8], dtype='float32')
             main_program.global_block().append_op(
                 type="alloc_float_status",
-                outputs={"FloatStatus": float_status}, )
+                outputs={"FloatStatus": float_status})
             main_program.global_block().append_op(
                 type="clear_float_status",
                 inputs={"FloatStatus": float_status},
-                outputs={"FloatStatusOut": float_status}, )
+                outputs={"FloatStatusOut": float_status})
             c = paddle.fluid.layers.elementwise_div(a, b)
             out, found_inf = check_finite_and_unscale(
                 [c], scale, float_status=float_status)
             main_program.global_block().append_op(
                 type="alloc_float_status",
-                outputs={"FloatStatus": float_status}, )
+                outputs={"FloatStatus": float_status})
             main_program.global_block().append_op(
                 type="clear_float_status",
                 inputs={"FloatStatus": float_status},
-                outputs={"FloatStatusOut": float_status}, )
+                outputs={"FloatStatusOut": float_status})
             d = paddle.fluid.layers.elementwise_add(a, b)
             out, found_inf = check_finite_and_unscale(
                 [d], scale, float_status=float_status)
diff --git a/python/paddle/fluid/tests/unittests/test_assign_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_assign_op_npu.py
similarity index 88%
rename from python/paddle/fluid/tests/unittests/test_assign_op_npu.py
rename to python/paddle/fluid/tests/unittests/npu/test_assign_op_npu.py
index ed21549b7e01fddcf30c8a53f1a96081cb3bb8f0..14133d5a385ff84536ab414cd0388ab4edabb722 100644
--- a/python/paddle/fluid/tests/unittests/test_assign_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_assign_op_npu.py
@@ -27,8 +27,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestAssign(OpTest):
     def setUp(self):
         self.set_npu()
@@ -49,7 +47,7 @@ class TestAssign(OpTest):
         self.dtype = np.float32
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
 if __name__ == '__main__':
diff --git a/python/paddle/fluid/tests/unittests/npu/test_bce_loss_npu.py b/python/paddle/fluid/tests/unittests/npu/test_bce_loss_npu.py
index eb00e777fe0eef3811c9749d477677ce794d095d..16db952533437c699fb7993350f280a03ccbe76a 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_bce_loss_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_bce_loss_npu.py
@@ -147,8 +147,6 @@ def calc_bceloss(input_np, label_np, reduction='mean', weight_np=None):
     return expected
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestBCELoss(unittest.TestCase):
     def test_BCELoss(self):
         input_np = np.random.uniform(0.1, 0.8, size=(20, 30)).astype(np.float32)
@@ -220,8 +218,6 @@ def bce_loss(input, label):
     return -1 * (label * np.log(input) + (1. - label) * np.log(1. - input))
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestBceLossOp(OpTest):
     def setUp(self):
         self.set_npu()
@@ -248,15 +244,11 @@ class TestBceLossOp(OpTest):
         self.shape = [10, 10]
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestBceLossOpCase1(OpTest):
     def init_test_cast(self):
         self.shape = [2, 3, 4, 5]
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestBceLossOpCase2(OpTest):
     def init_test_cast(self):
         self.shape = [2, 3, 20]
diff --git a/python/paddle/fluid/tests/unittests/npu/test_cast_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_cast_op_npu.py
index ae48866b7b969d5e7f2d7bf0dc9ed93c46aed4bb..f522eb10d92e6b7e377f072cc153f8b275a053e5 100755
--- a/python/paddle/fluid/tests/unittests/npu/test_cast_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_cast_op_npu.py
@@ -27,8 +27,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestCast1(OpTest):
     def setUp(self):
         self.set_npu()
@@ -48,7 +46,7 @@ class TestCast1(OpTest):
         self.__class__.use_npu = True
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
 class TestCast2(OpTest):
@@ -70,7 +68,7 @@ class TestCast2(OpTest):
         self.__class__.use_npu = True
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False, atol=1e-3)
+        self.check_output_with_place(self.place, atol=1e-3)
 
 
 class TestCast3(OpTest):
@@ -92,7 +90,7 @@ class TestCast3(OpTest):
         self.__class__.use_npu = True
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False, atol=1e-3)
+        self.check_output_with_place(self.place, atol=1e-3)
 
 
 if __name__ == '__main__':
diff --git a/python/paddle/fluid/tests/unittests/npu/test_coalesce_tensor_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_coalesce_tensor_op_npu.py
index 37fa5f8cad2abee06438f2d27da5e27ff5bbf963..f1bbf0becf1950360735a60838257c2084a53416 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_coalesce_tensor_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_coalesce_tensor_op_npu.py
@@ -28,8 +28,6 @@ SEED = 2021
 alignment = 512
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestAllocContinuousSpace(OpTest):
     def setUp(self):
         self.__class__.use_npu = True
@@ -82,12 +80,9 @@ class TestAllocContinuousSpace(OpTest):
         self.check_output_with_place(
             place=paddle.NPUPlace(0),
             no_check_set=["FusedOutput"],
-            atol=1e-5,
-            check_dygraph=False)
+            atol=1e-5, )
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestAllocContinuousSpace2(TestAllocContinuousSpace):
     def init_attr(self):
         return {
@@ -102,8 +97,7 @@ class TestAllocContinuousSpace2(TestAllocContinuousSpace):
         self.check_output_with_place(
             place=paddle.NPUPlace(0),
             no_check_set=["FusedOutput"],
-            atol=1e-5,
-            check_dygraph=False)
+            atol=1e-5, )
 
 
 if __name__ == '__main__':
diff --git a/python/paddle/fluid/tests/unittests/npu/test_compare_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_compare_op_npu.py
index 54a2c1e7163a9f122927bc2781eb1a13a84a124e..909bfaaa07fd5894eda6a8a8998d0fd897724413 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_compare_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_compare_op_npu.py
@@ -26,8 +26,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestEqual(OpTest):
     def setUp(self):
         self.set_npu()
@@ -53,11 +51,9 @@ class TestEqual(OpTest):
         self.dtype = np.float32
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestLessthan(OpTest):
     def setUp(self):
         self.set_npu()
@@ -83,7 +79,7 @@ class TestLessthan(OpTest):
         self.dtype = np.float32
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
 class TestEqual2(TestEqual):
diff --git a/python/paddle/fluid/tests/unittests/npu/test_concat_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_concat_op_npu.py
index a2ec1c7a9eef6ebc516cdc94062dd628c77a1f81..8f11d00ccabf67df24073755da2b744467b22825 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_concat_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_concat_op_npu.py
@@ -26,8 +26,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestConcat(OpTest):
     def setUp(self):
         self.set_npu()
@@ -56,7 +54,7 @@ class TestConcat(OpTest):
         self.dtype = np.float32
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
     def init_test_data(self):
         self.x0 = np.random.random((1, 4, 50)).astype(self.dtype)
@@ -65,12 +63,9 @@ class TestConcat(OpTest):
         self.axis = 0
 
     def test_check_grad(self):
-        self.check_grad_with_place(
-            self.place, ['x0', 'x2'], 'Out', check_dygraph=False)
-        self.check_grad_with_place(
-            self.place, ['x1'], 'Out', check_dygraph=False)
-        self.check_grad_with_place(
-            self.place, ['x2'], 'Out', check_dygraph=False)
+        self.check_grad_with_place(self.place, ['x0', 'x2'], 'Out')
+        self.check_grad_with_place(self.place, ['x1'], 'Out')
+        self.check_grad_with_place(self.place, ['x2'], 'Out')
 
 
 class TestConcatFP16(OpTest):
@@ -102,7 +97,7 @@ class TestConcatFP16(OpTest):
         self.dtype = np.float16
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
     def init_test_data(self):
         self.x0 = np.random.random((1, 4, 50)).astype(self.dtype)
diff --git a/python/paddle/fluid/tests/unittests/npu/test_dropout_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_dropout_op_npu.py
index 6b936514452f74df2a1692538ddb0d4128365fc2..4c434561e1cc11a84bef6a2f8c07ed192d028452 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_dropout_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_dropout_op_npu.py
@@ -28,8 +28,6 @@ SEED = 2021
 EPOCH = 100
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestDropoutOp(OpTest):
     def setUp(self):
         self.op_type = "dropout"
@@ -55,17 +53,14 @@ class TestDropoutOp(OpTest):
         self.place = paddle.NPUPlace(0)
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
     def test_check_grad_normal(self):
         if self.dtype == np.float16:
             return
-        self.check_grad_with_place(
-            self.place, ['X'], 'Out', check_dygraph=False)
+        self.check_grad_with_place(self.place, ['X'], 'Out')
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestDropoutOpInput1d(TestDropoutOp):
     # change input shape
     def setUp(self):
@@ -85,15 +80,13 @@ class TestDropoutOpInput1d(TestDropoutOp):
         }
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestDropoutOpInput1d(TestDropoutOp):
     # the input is 1-D
     def setUp(self):
         self.op_type = "dropout"
         self.set_npu()
         self.init_dtype()
-        self.inputs = {'X': np.random.random((2000, )).astype(self.dtype)}
+        self.inputs = {'X': np.random.random((2000)).astype(self.dtype)}
         self.attrs = {
             'dropout_prob': 0.0,
             'fix_seed': True,
@@ -106,8 +99,6 @@ class TestDropoutOpInput1d(TestDropoutOp):
         }
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestDropoutOp2(TestDropoutOp):
     # the dropout_prob is 1.0
     def setUp(self):
@@ -127,8 +118,6 @@ class TestDropoutOp2(TestDropoutOp):
         }
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestDropoutOp3(TestDropoutOp):
     # the input dim is 3
     def setUp(self):
@@ -148,8 +137,6 @@ class TestDropoutOp3(TestDropoutOp):
         }
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 @skip_check_grad_ci(reason="For inference, check_grad is not required.")
 class TestDropoutOpInference(OpTest):
     # is_test = True
@@ -174,11 +161,9 @@ class TestDropoutOpInference(OpTest):
         self.place = paddle.NPUPlace(0)
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 @skip_check_grad_ci(reason="For inference, check_grad is not required.")
 class TestDropoutOpInference2(TestDropoutOpInference):
     def setUp(self):
@@ -194,8 +179,6 @@ class TestDropoutOpInference2(TestDropoutOpInference):
         self.outputs = {'Out': self.inputs['X']}
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestDropoutOpWithSeed(TestDropoutOp):
     # the seed is a Tensor
     def setUp(self):
@@ -218,8 +201,6 @@ class TestDropoutOpWithSeed(TestDropoutOp):
         }
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestDropoutOpFp16(TestDropoutOp):
     # float16
     def init_dtype(self):
@@ -231,8 +212,6 @@ class TestDropoutOpFp16(TestDropoutOp):
         self.place = paddle.NPUPlace(0)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestDropoutAPI(unittest.TestCase):
     def setUp(self):
         np.random.seed(123)
diff --git a/python/paddle/fluid/tests/unittests/npu/test_elementwise_add_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_elementwise_add_op_npu.py
index 6a82157faaec41d9abaffa9b68e3a3e80b6b2fb3..5288db5ceb1c6fe33af912ed4408352997876943 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_elementwise_add_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_elementwise_add_op_npu.py
@@ -25,8 +25,6 @@ import paddle.fluid as fluid
 paddle.enable_static()
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestElementwiseAddOp(OpTest):
     def setUp(self):
         self.set_npu()
@@ -62,34 +60,32 @@ class TestElementwiseAddOp(OpTest):
         self.axis = -1
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
     def test_check_grad_normal(self):
         self.check_grad_with_place(
-            self.place, ['X', 'Y'],
+            self.place,
+            ['X', 'Y'],
             'Out',
-            max_relative_error=0.006,
-            check_dygraph=False)
+            max_relative_error=0.006, )
 
     def test_check_grad_ingore_x(self):
         self.check_grad_with_place(
-            self.place, ['Y'],
+            self.place,
+            ['Y'],
             'Out',
             no_grad_set=set("X"),
-            max_relative_error=0.006,
-            check_dygraph=False)
+            max_relative_error=0.006, )
 
     def test_check_grad_ingore_y(self):
         self.check_grad_with_place(
-            self.place, ['X'],
+            self.place,
+            ['X'],
             'Out',
             no_grad_set=set("Y"),
-            max_relative_error=0.006,
-            check_dygraph=False)
+            max_relative_error=0.006, )
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestAddAPI(unittest.TestCase):
     def test_name(self):
         with paddle.static.program_guard(paddle.static.Program()):
@@ -134,8 +130,6 @@ class TestAddAPI(unittest.TestCase):
                 msg="z_value = {}, but expected {}".format(z_value, z_expected))
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestAddError(unittest.TestCase):
     def test_errors(self):
         with paddle.static.program_guard(paddle.static.Program()):
diff --git a/python/paddle/fluid/tests/unittests/npu/test_elementwise_div_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_elementwise_div_op_npu.py
index 0ae2678d10b47c8998882e3ee00d177e86236a06..ee72ee74065e3b1dd417a5d3adac8711a2464cba 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_elementwise_div_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_elementwise_div_op_npu.py
@@ -26,8 +26,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestElementwiseDiv(OpTest):
     def setUp(self):
         self.set_npu()
@@ -54,30 +52,28 @@ class TestElementwiseDiv(OpTest):
         self.dtype = np.float32
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
     def test_check_grad_normal(self):
         self.check_grad_with_place(
-            self.place, ['X', 'Y'],
+            self.place,
+            ['X', 'Y'],
             'Out',
-            max_relative_error=0.007,
-            check_dygraph=False)
+            max_relative_error=0.007, )
 
     def test_check_grad_ingore_x(self):
         self.check_grad_with_place(
-            self.place, ['Y'],
+            self.place,
+            ['Y'],
             'Out',
             max_relative_error=0.007,
-            no_grad_set=set("X"),
-            check_dygraph=False)
+            no_grad_set=set("X"), )
 
     def test_check_grad_ingore_y(self):
         self.check_grad_with_place(
-            self.place, ['X'], 'Out', no_grad_set=set("Y"), check_dygraph=False)
+            self.place, ['X'], 'Out', no_grad_set=set("Y"))
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestElementwiseDivFp16(OpTest):
     def setUp(self):
         self.set_npu()
@@ -105,11 +101,9 @@ class TestElementwiseDivFp16(OpTest):
         self.dtype = np.float16
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False, atol=1e-5)
+        self.check_output_with_place(self.place, atol=1e-5)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestElementwiseDivNet(unittest.TestCase):
     def _test(self, run_npu=True):
         main_prog = paddle.static.Program()
diff --git a/python/paddle/fluid/tests/unittests/npu/test_elementwise_floordiv_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_elementwise_floordiv_op_npu.py
index 93538e938670f07ab78f33e0c9749b702854b7d6..36d282a3d06f77369024d5c08e3f5d467d0a76d4 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_elementwise_floordiv_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_elementwise_floordiv_op_npu.py
@@ -24,8 +24,6 @@ import paddle
 paddle.enable_static()
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestElementwiseFloorDiv(OpTest):
     def setUp(self):
         self.op_type = "elementwise_floordiv"
@@ -53,11 +51,9 @@ class TestElementwiseFloorDiv(OpTest):
         self.dtype = "int64"
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestElementwiseFloorDiv2(TestElementwiseFloorDiv):
     def init_dtype(self):
         self.dtype = "int32"
diff --git a/python/paddle/fluid/tests/unittests/test_elementwise_max_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_elementwise_max_op_npu.py
similarity index 91%
rename from python/paddle/fluid/tests/unittests/test_elementwise_max_op_npu.py
rename to python/paddle/fluid/tests/unittests/npu/test_elementwise_max_op_npu.py
index 6475caf970cba7be5efad60b9a4c094e112175c3..6c325b020206c4e6c96c83a318edbebf79a3768f 100644
--- a/python/paddle/fluid/tests/unittests/test_elementwise_max_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_elementwise_max_op_npu.py
@@ -26,8 +26,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestElementwiseMax(OpTest):
     def setUp(self):
         self.set_npu()
@@ -54,7 +52,7 @@ class TestElementwiseMax(OpTest):
         self.dtype = np.float32
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
     # TODO(ascendrc): Max grad test
     # def test_check_grad(self):
@@ -64,8 +62,6 @@ class TestElementwiseMax(OpTest):
     #
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestElementwiseMaxFp16(OpTest):
     def setUp(self):
         self.set_npu()
@@ -93,11 +89,9 @@ class TestElementwiseMaxFp16(OpTest):
         self.dtype = np.float16
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False, atol=1e-5)
+        self.check_output_with_place(self.place, atol=1e-5)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestElementwiseMaxNet(unittest.TestCase):
     def _test(self, run_npu=True):
         main_prog = paddle.static.Program()
diff --git a/python/paddle/fluid/tests/unittests/npu/test_elementwise_min_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_elementwise_min_op_npu.py
index b4d9c7285b2b556f76f94241cc0b9373319f7753..2034a12c5c0feb92e351889c3ca7ec25c0b29611 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_elementwise_min_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_elementwise_min_op_npu.py
@@ -26,8 +26,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestElementwiseMin(OpTest):
     def setUp(self):
         self.set_npu()
@@ -54,7 +52,7 @@ class TestElementwiseMin(OpTest):
         self.dtype = np.float32
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
     # TODO(ascendrc): Min grad test
     # def test_check_grad(self):
@@ -64,8 +62,6 @@ class TestElementwiseMin(OpTest):
     #
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestElementwiseMinFp16(OpTest):
     def setUp(self):
         self.set_npu()
@@ -93,11 +89,9 @@ class TestElementwiseMinFp16(OpTest):
         self.dtype = np.float16
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False, atol=1e-5)
+        self.check_output_with_place(self.place, atol=1e-5)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestElementwiseMinNet(unittest.TestCase):
     def _test(self, run_npu=True):
         main_prog = paddle.static.Program()
diff --git a/python/paddle/fluid/tests/unittests/npu/test_elementwise_mul_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_elementwise_mul_op_npu.py
index 9bfb7e033e7ea454223c683877bb30f02506be75..ea94661e8a51e680b23e6bb095caa813c29ab74f 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_elementwise_mul_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_elementwise_mul_op_npu.py
@@ -26,8 +26,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestElementwiseMul(OpTest):
     def setUp(self):
         self.set_npu()
@@ -54,7 +52,7 @@ class TestElementwiseMul(OpTest):
         self.dtype = np.float32
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
     # TODO(ascendrc): Mul grad test
     # def test_check_grad(self):
@@ -64,8 +62,6 @@ class TestElementwiseMul(OpTest):
     #
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestElementwiseMulFp16(OpTest):
     def setUp(self):
         self.set_npu()
@@ -93,11 +89,9 @@ class TestElementwiseMulFp16(OpTest):
         self.dtype = np.float16
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False, atol=1e-5)
+        self.check_output_with_place(self.place, atol=1e-5)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestElementwiseMulNet(unittest.TestCase):
     def _test(self, run_npu=True):
         main_prog = paddle.static.Program()
diff --git a/python/paddle/fluid/tests/unittests/npu/test_elementwise_pow_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_elementwise_pow_op_npu.py
index 862c546b8e05ebe2046e9c5aeb52178fa47f59ab..dea1828a6d75fca3e6a871207e8a746305169a6c 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_elementwise_pow_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_elementwise_pow_op_npu.py
@@ -26,8 +26,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestElementwisePow(OpTest):
     def setUp(self):
         self.set_npu()
@@ -54,7 +52,7 @@ class TestElementwisePow(OpTest):
         self.dtype = np.float32
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
     # TODO(ascendrc): Pow grad test
     # def test_check_grad(self):
@@ -64,8 +62,6 @@ class TestElementwisePow(OpTest):
     #
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestElementwisePowFp16(OpTest):
     def setUp(self):
         self.set_npu()
@@ -93,11 +89,9 @@ class TestElementwisePowFp16(OpTest):
         self.dtype = np.float16
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False, atol=1e-5)
+        self.check_output_with_place(self.place, atol=1e-5)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestElementwisePowNet(unittest.TestCase):
     def _test(self, run_npu=True):
         main_prog = paddle.static.Program()
diff --git a/python/paddle/fluid/tests/unittests/npu/test_elementwise_sub_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_elementwise_sub_op_npu.py
index 8c6c7b46f49f2725b646202998095adef3a65e63..6faa77b4602137ad0e870fa8b7bd213e50c49022 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_elementwise_sub_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_elementwise_sub_op_npu.py
@@ -27,8 +27,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestElementwiseSubOp(OpTest):
     def setUp(self):
         self.set_npu()
@@ -64,7 +62,7 @@ class TestElementwiseSubOp(OpTest):
         self.axis = 0
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
     # TODO(ascendrc): For grad tests, OpTest raises FatalError:Segmentation fault
     #  when call op.run, which may be caused by system environment exception
@@ -74,7 +72,7 @@ class TestElementwiseSubOp(OpTest):
     #         self.place, ['X', 'Y'],
     #         'Out',
     #         max_relative_error=0.006,
-    #         check_dygraph=False)
+    #         )
     #
     # def test_check_grad_ingore_x(self):
     #     self.check_grad_with_place(
@@ -82,18 +80,16 @@ class TestElementwiseSubOp(OpTest):
     #         'Out',
     #         no_grad_set=set("X"),
     #         max_relative_error=0.006,
-    #         check_dygraph=False)
+    #         )
     #
     # def test_check_grad_ingore_y(self):
     #     self.check_grad_with_place(
     #         self.place, ['X'],
     #         'Out',
     #         no_grad_set=set("Y"),
-    #         max_relative_error=0.006,check_dygraph=False)
+    #         max_relative_error=0.006,)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestSubtractAPI(unittest.TestCase):
     def test_name(self):
         with paddle.static.program_guard(paddle.static.Program()):
@@ -138,8 +134,6 @@ class TestSubtractAPI(unittest.TestCase):
                 msg="z_value = {}, but expected {}".format(z_value, z_expected))
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestSubtractError(unittest.TestCase):
     def test_errors(self):
         with paddle.static.program_guard(paddle.static.Program()):
@@ -158,8 +152,6 @@ class TestSubtractError(unittest.TestCase):
             self.assertRaises(TypeError, paddle.subtract, x2, y2)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestSubtractNet(unittest.TestCase):
     def _test(self, run_npu=True):
         main_prog = paddle.static.Program()
diff --git a/python/paddle/fluid/tests/unittests/npu/test_expand_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_expand_op_npu.py
index f6a84d3be5c100c7324f2ed62ce8c934f4318b7f..375003f79e500f99ceaf374ce898d998263700e0 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_expand_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_expand_op_npu.py
@@ -26,8 +26,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestExpand(OpTest):
     def setUp(self):
         self.set_npu()
@@ -50,7 +48,7 @@ class TestExpand(OpTest):
         self.dtype = np.float32
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
     # TODO(ascendrc): Add grad test
     # def test_check_grad(self):
@@ -60,8 +58,6 @@ class TestExpand(OpTest):
     #
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestExpandV2(TestExpand):
     def setUp(self):
         self.set_npu()
@@ -82,8 +78,6 @@ class TestExpandV2(TestExpand):
         self.outputs = {'Out': out}
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestExpandFp16(TestExpand):
     no_need_check_grad = True
 
@@ -91,8 +85,6 @@ class TestExpandFp16(TestExpand):
         self.dtype = np.float16
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestExpandNet(unittest.TestCase):
     def _test(self, run_npu=True):
         main_prog = paddle.static.Program()
diff --git a/python/paddle/fluid/tests/unittests/npu/test_fill_constant_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_fill_constant_op_npu.py
index 6e619bfd11fb901994ad3a91187a716b014dab41..c8d7f2f9dc9873739075ff5808a856cd3e8a6c0a 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_fill_constant_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_fill_constant_op_npu.py
@@ -27,8 +27,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestFillConstant(OpTest):
     def setUp(self):
         self.set_npu()
@@ -47,7 +45,7 @@ class TestFillConstant(OpTest):
         self.dtype = np.float32
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
 class TestFillConstantInt(OpTest):
@@ -71,7 +69,7 @@ class TestFillConstantInt(OpTest):
         self.dtype = np.int32
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
 class TestFillConstantFP16(OpTest):
@@ -95,7 +93,7 @@ class TestFillConstantFP16(OpTest):
         self.dtype = np.float16
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False, atol=1e-3)
+        self.check_output_with_place(self.place, atol=1e-3)
 
 
 if __name__ == '__main__':
diff --git a/python/paddle/fluid/tests/unittests/npu/test_gather_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_gather_op_npu.py
index 008422ffd21188327fa938734928a0dc62187824..daca3d884600a09da244d4ec2952b2028c7e9762 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_gather_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_gather_op_npu.py
@@ -34,8 +34,6 @@ def gather_numpy(x, index, axis):
     return gather
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestGatherOp(OpTest):
     def setUp(self):
         self.set_npu()
@@ -53,14 +51,14 @@ class TestGatherOp(OpTest):
         self.__class__.use_npu = True
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
     def test_check_grad(self):
         self.check_grad_with_place(
-            self.place, ['X'],
+            self.place,
+            ['X'],
             'Out',
-            max_relative_error=0.006,
-            check_dygraph=False)
+            max_relative_error=0.006, )
 
     def config(self):
         """
@@ -72,8 +70,6 @@ class TestGatherOp(OpTest):
         self.index_type = "int32"
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestCase1(TestGatherOp):
     def config(self):
         """
@@ -85,8 +81,6 @@ class TestCase1(TestGatherOp):
         self.index_type = "int32"
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class API_TestGather(unittest.TestCase):
     def test_out1(self):
         with fluid.program_guard(fluid.Program(), fluid.Program()):
@@ -120,8 +114,6 @@ class API_TestGather(unittest.TestCase):
         self.assertTrue(np.allclose(result, expected_output))
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestGatherGrad(unittest.TestCase):
     def _test(self, run_npu=True):
         main_prog = paddle.static.Program()
diff --git a/python/paddle/fluid/tests/unittests/npu/test_gaussian_random_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_gaussian_random_op_npu.py
index 07e214e06003f3fa0bdca0c73911124126e347d2..11f64b8fc7d269b908c2cec026b2d3abfb3c48a8 100755
--- a/python/paddle/fluid/tests/unittests/npu/test_gaussian_random_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_gaussian_random_op_npu.py
@@ -26,8 +26,6 @@ from test_gaussian_random_op import TestGaussianRandomOp
 paddle.enable_static()
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestNPUGaussianRandomOp(OpTest):
     def setUp(self):
         self.set_npu()
diff --git a/python/paddle/fluid/tests/unittests/npu/test_gelu_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_gelu_op_npu.py
index 4127c7382880e261ce4a376ccc0660e486fdc699..760ce59812ea2e9516d3f4da39938c4b59eee838 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_gelu_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_gelu_op_npu.py
@@ -32,8 +32,6 @@ def np_gelu(x):
     return y
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestGelu(OpTest):
     def setUp(self):
         self.set_npu()
@@ -56,18 +54,13 @@ class TestGelu(OpTest):
         self.dtype = np.float32
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False, atol=1e-3)
+        self.check_output_with_place(self.place, atol=1e-3)
 
     def test_check_grad(self):
         self.check_grad_with_place(
-            self.place, ['X'],
-            'Out',
-            check_dygraph=False,
-            max_relative_error=0.007)
+            self.place, ['X'], 'Out', max_relative_error=0.007)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestGeluFp16(OpTest):
     def setUp(self):
         self.set_npu()
@@ -91,11 +84,9 @@ class TestGeluFp16(OpTest):
         self.dtype = np.float16
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False, atol=1e-3)
+        self.check_output_with_place(self.place, atol=1e-3)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestGeluNet(unittest.TestCase):
     def _test(self, run_npu=True):
         main_prog = paddle.static.Program()
diff --git a/python/paddle/fluid/tests/unittests/npu/test_increment_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_increment_op_npu.py
index 3e2e8f944b84c6f2c97b7d834e6f2b6b7511583d..dfb9b26d64ea852614bba1a30bf9b9555fb110e6 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_increment_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_increment_op_npu.py
@@ -29,8 +29,6 @@ SEED = 2021
 NPUPlace = 0
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestIncrement(OpTest):
     def setUp(self):
         self.set_npu()
@@ -54,11 +52,9 @@ class TestIncrement(OpTest):
         self.dtype = np.int64
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestIncrementFP16(OpTest):
     def setUp(self):
         self.set_npu()
@@ -82,11 +78,9 @@ class TestIncrementFP16(OpTest):
         self.dtype = np.float16
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestIncrementInplace(unittest.TestCase):
     def test_npu(self):
         main_prog = paddle.static.Program()
diff --git a/python/paddle/fluid/tests/unittests/npu/test_layer_norm_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_layer_norm_op_npu.py
index d447dfb8d4d031e6f29fdfedab285066d4dea565..0345ac1f2065b1b6dfa769dd7e6fd64c7790df6f 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_layer_norm_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_layer_norm_op_npu.py
@@ -36,8 +36,6 @@ from op_test import _set_use_system_allocator
 _set_use_system_allocator(False)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestLayerNormOp(unittest.TestCase):
     def setUp(self):
         self.use_cudnn = True
@@ -191,8 +189,6 @@ class TestLayerNormOp(unittest.TestCase):
         self.check_forward_backward(shape=[2, 3, 4, 5], begin_norm_axis=3)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestLayerNormOpFP16(TestLayerNormOp):
     def init_dtype(self):
         self.dtype = np.float16
diff --git a/python/paddle/fluid/tests/unittests/npu/test_log_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_log_op_npu.py
index 3cdd2448628a0b0f1900cc8b15d884d578a445ca..9534431e99a7a2e0218fe08dfd95a770b9924915 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_log_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_log_op_npu.py
@@ -26,8 +26,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestLog(OpTest):
     def setUp(self):
         self.set_npu()
@@ -50,7 +48,7 @@ class TestLog(OpTest):
         self.dtype = np.float32
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
     # TODO(ascendrc): Add grad test
     # def test_check_grad(self):
@@ -60,8 +58,6 @@ class TestLog(OpTest):
     #
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestLogFp16(OpTest):
     def setUp(self):
         self.set_npu()
@@ -85,11 +81,9 @@ class TestLogFp16(OpTest):
         self.dtype = np.float16
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False, atol=1e-5)
+        self.check_output_with_place(self.place, atol=1e-5)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestLogNet(unittest.TestCase):
     def _test(self, run_npu=True):
         main_prog = paddle.static.Program()
diff --git a/python/paddle/fluid/tests/unittests/npu/test_logical_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_logical_op_npu.py
index 6d1327f068a528c97b8b3e882bc37f2c07635acb..f695eeb0f27743144bb67486d6435fcc21604617 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_logical_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_logical_op_npu.py
@@ -220,8 +220,6 @@ def type_map_factory():
     } for x_type in x_type_list for y_type in y_type_list]
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestCPU(unittest.TestCase):
     def test(self):
         test(self)
@@ -235,8 +233,6 @@ class TestCPU(unittest.TestCase):
             test_type_error(self, False, type_map)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestNPU(unittest.TestCase):
     def test(self):
         test(self, True)
diff --git a/python/paddle/fluid/tests/unittests/npu/test_lookup_table_v2_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_lookup_table_v2_op_npu.py
index 41fe0636bd7790433dee33dd358ec7ed6d7ae9e5..56f04a6e993f3aba98d6d3ed874e460888412e1d 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_lookup_table_v2_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_lookup_table_v2_op_npu.py
@@ -26,8 +26,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestLookupTableV2(OpTest):
     def setUp(self):
         self.set_npu()
@@ -67,17 +65,14 @@ class TestLookupTableV2(OpTest):
         self.dim = 20
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
     def test_check_grad(self):
         if self.dtype == np.float16:
             return
-        self.check_grad_with_place(
-            self.place, ['W'], 'Out', check_dygraph=False)
+        self.check_grad_with_place(self.place, ['W'], 'Out')
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestLookupTableV2FP16(TestLookupTableV2):
     no_need_check_grad = True
 
@@ -89,16 +84,12 @@ class TestLookupTableV2FP16(TestLookupTableV2):
         self.__class__.no_need_check_grad = True
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestLookupTableV2Dim32(TestLookupTableV2):
     def init_dim(self):
         # embedding_dim is multiple of 32
         self.dim = 64
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestLookupTableV2Dim32FP16(TestLookupTableV2):
     no_need_check_grad = True
 
diff --git a/python/paddle/fluid/tests/unittests/npu/test_matmulv2_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_matmulv2_op_npu.py
index b093fa4f2caa4abc1262c150d84ebff7dae14328..53766c5eb61b7a429636551b5e7b1c926f9b38e4 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_matmulv2_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_matmulv2_op_npu.py
@@ -26,15 +26,13 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 def reference_matmul(X, Y, transpose_X=False, transpose_Y=False):
     """Reference forward implementation using np.matmul."""
     # np.matmul does not support the transpose flags, so we manually
     # transpose X and Y appropriately.
     if transpose_X:
         if X.ndim == 1:
-            X = X.reshape((X.size, ))
+            X = X.reshape((X.size))
         elif X.ndim == 2:
             X = X.T
         else:
@@ -43,7 +41,7 @@ def reference_matmul(X, Y, transpose_X=False, transpose_Y=False):
             X = np.transpose(X, tuple(dim))
     if transpose_Y:
         if Y.ndim == 1:
-            Y = Y.reshape((Y.size, ))
+            Y = Y.reshape((Y.size))
         else:
             dim = [i for i in range(len(Y.shape))]
             dim[-1], dim[len(Y.shape) - 2] = dim[len(Y.shape) - 2], dim[-1]
@@ -53,7 +51,7 @@ def reference_matmul(X, Y, transpose_X=False, transpose_Y=False):
     if not Out.shape:
         # We do not support 0-dimensional Tensors (scalars). So where
         # np.matmul outputs a scalar, we must convert to a Tensor of
-        # shape (1, ) instead.
+        # shape (1) instead.
         # Everywhere else, we are compatible with np.matmul.
         Out = np.array([Out], dtype="float64")
     return Out
@@ -95,7 +93,7 @@ class TestMatMul(OpTest):
         self.dtype = np.float32
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False, atol=1e-5)
+        self.check_output_with_place(self.place, atol=1e-5)
 
 
     # TODO(ascendrc): Add grad test
@@ -137,8 +135,6 @@ class TestMatMul4(TestMatMul):
         self.trans_y = False
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestMatMulNet(unittest.TestCase):
     def _test(self, run_npu=True):
         main_prog = paddle.static.Program()
@@ -207,8 +203,8 @@ class TestMatMulNet(unittest.TestCase):
 
 
 # The precision is aligned in NPU and GPU separately, which is only used for the usage method.
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
+
+
 class TestMatMulNet3_2(unittest.TestCase):
     def _test(self, run_npu=True):
         main_prog = paddle.static.Program()
diff --git a/python/paddle/fluid/tests/unittests/npu/test_mean_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_mean_op_npu.py
index 6e8f99a9dbb19785094ad6a94d9f371fe409fc69..e69c2fd84dd9dbb8ed31ba4071ef3d22eb4916a8 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_mean_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_mean_op_npu.py
@@ -27,8 +27,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestMean(OpTest):
     def setUp(self):
         self.set_npu()
@@ -50,15 +48,12 @@ class TestMean(OpTest):
         self.dtype = np.float32
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
     def test_check_grad(self):
-        self.check_grad_with_place(
-            self.place, ['X'], 'Out', check_dygraph=False)
+        self.check_grad_with_place(self.place, ['X'], 'Out')
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestMeanFP16(OpTest):
     def setUp(self):
         self.set_npu()
@@ -81,7 +76,7 @@ class TestMeanFP16(OpTest):
         self.dtype = np.float16
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
 if __name__ == '__main__':
diff --git a/python/paddle/fluid/tests/unittests/npu/test_memcpy_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_memcpy_op_npu.py
index 63c4fb8e5885eaa33ba18227c3e89ce3b0c97b84..a421e2667347c5b046e424a59f2efd2022a3199d 100755
--- a/python/paddle/fluid/tests/unittests/npu/test_memcpy_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_memcpy_op_npu.py
@@ -28,8 +28,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestMemcpy_FillConstant(unittest.TestCase):
     def get_prog(self):
         paddle.enable_static()
diff --git a/python/paddle/fluid/tests/unittests/npu/test_mul_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_mul_op_npu.py
index 07f187a0f0de9d6f570ff610ca412a37bfae895e..cb58a2a8d440936c50b009dbdece5587ee94961f 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_mul_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_mul_op_npu.py
@@ -52,30 +52,30 @@ class TestMul(OpTest):
         self.dtype = np.float32
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False, atol=1e-5)
+        self.check_output_with_place(self.place, atol=1e-5)
 
     def test_check_grad_normal(self):
         self.check_grad_with_place(
-            self.place, ['X', 'Y'],
+            self.place,
+            ['X', 'Y'],
             'Out',
-            max_relative_error=0.0065,
-            check_dygraph=False)
+            max_relative_error=0.0065, )
 
     def test_check_grad_ingore_x(self):
         self.check_grad_with_place(
-            self.place, ['Y'],
+            self.place,
+            ['Y'],
             'Out',
             no_grad_set=set("X"),
-            max_relative_error=0.0065,
-            check_dygraph=False)
+            max_relative_error=0.0065, )
 
     def test_check_grad_ingore_y(self):
         self.check_grad_with_place(
-            self.place, ['X'],
+            self.place,
+            ['X'],
             'Out',
             no_grad_set=set("Y"),
-            max_relative_error=0.0065,
-            check_dygraph=False)
+            max_relative_error=0.0065, )
 
 
 @skip_check_grad_ci(
@@ -170,8 +170,6 @@ class TestMul3FP16(TestMul3):
         pass
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestMulNet(unittest.TestCase):
     def init_dtype(self):
         self.dtype = np.float32
@@ -243,8 +241,6 @@ class TestMulNet(unittest.TestCase):
         self.assertTrue(np.allclose(npu_loss, cpu_loss))
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestMulNet3_2(unittest.TestCase):
     def init_dtype(self):
         self.dtype = np.float32
@@ -317,8 +313,6 @@ class TestMulNet3_2(unittest.TestCase):
         self.assertTrue(np.allclose(npu_loss, cpu_loss, atol=1e-5))
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestMulNet3_2_xc2(unittest.TestCase):
     def init_dtype(self):
         self.dtype = np.float32
diff --git a/python/paddle/fluid/tests/unittests/npu/test_npu_place.py b/python/paddle/fluid/tests/unittests/npu/test_npu_place.py
index 3f71fad2b9c1084148d8b0a28e556cc0bf5f366e..91e0c29e10609b7cbe41a657e62d7a1b2f3c8d7d 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_npu_place.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_npu_place.py
@@ -22,8 +22,6 @@ from paddle.fluid import core
 paddle.enable_static()
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestNpuPlace(unittest.TestCase):
     def test(self):
         p = core.Place()
@@ -33,8 +31,6 @@ class TestNpuPlace(unittest.TestCase):
         self.assertEqual(p.npu_device_id(), 0)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestNpuPlaceError(unittest.TestCase):
     def test_static(self):
         # NPU is not supported in ParallelExecutor
diff --git a/python/paddle/fluid/tests/unittests/npu/test_pow_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_pow_op_npu.py
index 8c67766b31184a36446c4fa39f64f760fa23912c..a188953d70c93cc0853b0f58d65c9a464688c6dc 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_pow_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_pow_op_npu.py
@@ -26,8 +26,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestPow(OpTest):
     def setUp(self):
         self.set_npu()
@@ -50,15 +48,12 @@ class TestPow(OpTest):
         self.dtype = np.float32
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
     def test_check_grad(self):
-        self.check_grad_with_place(
-            self.place, ['X'], 'Out', check_dygraph=False)
+        self.check_grad_with_place(self.place, ['X'], 'Out')
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestPowFp16(OpTest):
     def setUp(self):
         self.set_npu()
@@ -82,11 +77,9 @@ class TestPowFp16(OpTest):
         self.dtype = np.float16
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False, atol=1e-5)
+        self.check_output_with_place(self.place, atol=1e-5)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestPowNet(unittest.TestCase):
     def _test(self, run_npu=True):
         main_prog = paddle.static.Program()
diff --git a/python/paddle/fluid/tests/unittests/npu/test_reduce_any_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_reduce_any_op_npu.py
index 583a648224d7309bb1f2aa29db6871020091867f..1a30d1395283ec9bc60dcceaa1b5963b9176b60a 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_reduce_any_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_reduce_any_op_npu.py
@@ -28,8 +28,6 @@ from paddle.fluid.framework import convert_np_dtype_to_dtype_
 paddle.enable_static()
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestAny8DOp(OpTest):
     def setUp(self):
         self.set_npu()
@@ -46,11 +44,9 @@ class TestAny8DOp(OpTest):
         self.__class__.use_npu = True
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestAnyOpWithDim(OpTest):
     def setUp(self):
         self.set_npu()
@@ -64,11 +60,9 @@ class TestAnyOpWithDim(OpTest):
         self.__class__.use_npu = True
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestAny8DOpWithDim(OpTest):
     def setUp(self):
         self.set_npu()
@@ -85,18 +79,16 @@ class TestAny8DOpWithDim(OpTest):
         self.__class__.use_npu = True
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestAnyOpWithKeepDim(OpTest):
     def setUp(self):
         self.set_npu()
         self.op_type = "reduce_any"
         self.place = paddle.NPUPlace(0)
         self.inputs = {'X': np.random.randint(0, 2, (5, 6, 10)).astype("bool")}
-        self.attrs = {'dim': (1, ), 'keep_dim': True}
+        self.attrs = {'dim': (1), 'keep_dim': True}
         self.outputs = {
             'Out': np.expand_dims(
                 self.inputs['X'].any(axis=self.attrs['dim']), axis=1)
@@ -106,7 +98,7 @@ class TestAnyOpWithKeepDim(OpTest):
         self.__class__.use_npu = True
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
 class TestAny8DOpWithKeepDim(OpTest):
@@ -118,7 +110,7 @@ class TestAny8DOpWithKeepDim(OpTest):
             'X': np.random.randint(0, 2,
                                    (2, 5, 3, 2, 2, 3, 4, 2)).astype("bool")
         }
-        self.attrs = {'dim': (1, ), 'keep_dim': True}
+        self.attrs = {'dim': (1), 'keep_dim': True}
         self.outputs = {
             'Out': np.expand_dims(
                 self.inputs['X'].any(axis=self.attrs['dim']), axis=1)
@@ -128,7 +120,7 @@ class TestAny8DOpWithKeepDim(OpTest):
         self.__class__.use_npu = True
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
 if __name__ == '__main__':
diff --git a/python/paddle/fluid/tests/unittests/npu/test_reduce_sum_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_reduce_sum_op_npu.py
index d3861bf0780cb58f7362ff9dbd05c99a222bc21b..bd7ce2a040c93b406ac136738a4a32a3c32e840c 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_reduce_sum_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_reduce_sum_op_npu.py
@@ -26,8 +26,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestReduceSum(OpTest):
     def setUp(self):
         np.random.seed(SEED)
@@ -66,10 +64,10 @@ class TestReduceSum(OpTest):
 
     def initTestCase(self):
         self.shape = (5, 6)
-        self.axis = (0, )
+        self.axis = (0)
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
     # TODO(ascendrc): Add grad test
     # def test_check_grad(self):
@@ -84,8 +82,6 @@ class TestReduceSum2(OpTest):
         self.dtype = np.int32
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestReduceSumNet(unittest.TestCase):
     def set_reduce_sum_function(self, x):
         # keep_dim = False
@@ -151,16 +147,12 @@ class TestReduceSumNet(unittest.TestCase):
         self.assertTrue(np.allclose(npu_loss, cpu_loss))
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestReduceSumNet2(TestReduceSumNet):
     def set_reduce_sum_function(self, x):
         # keep_dim = True
         return paddle.fluid.layers.reduce_sum(x, dim=-1, keep_dim=True)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestReduceSumNet3(TestReduceSumNet):
     def _test(self, run_npu=True):
         main_prog = paddle.static.Program()
diff --git a/python/paddle/fluid/tests/unittests/npu/test_relu_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_relu_op_npu.py
index 9273d01299d8f564ee0ae575b47bb30e939c3d76..a2547808e6f161ae1cdac5ea5944863d7c640d24 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_relu_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_relu_op_npu.py
@@ -26,8 +26,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestRelu(OpTest):
     def setUp(self):
         self.set_npu()
@@ -50,11 +48,9 @@ class TestRelu(OpTest):
         self.dtype = np.float32
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestReluFp16(OpTest):
     def setUp(self):
         self.set_npu()
@@ -78,11 +74,9 @@ class TestReluFp16(OpTest):
         self.dtype = np.float16
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False, atol=1e-5)
+        self.check_output_with_place(self.place, atol=1e-5)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestReluNeg(OpTest):
     def setUp(self):
         self.set_npu()
@@ -105,13 +99,13 @@ class TestReluNeg(OpTest):
         self.dtype = np.float32
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
 #
 #
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
+
+
 class TestReluNet(unittest.TestCase):
     def _test(self, run_npu=True):
         main_prog = paddle.static.Program()
diff --git a/python/paddle/fluid/tests/unittests/npu/test_reshape2_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_reshape2_op_npu.py
index 885c990c702bd35d2052b3cb79abf11a74b3efc2..520de15f4df62be60faf910cd8ddb515a66c4f7e 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_reshape2_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_reshape2_op_npu.py
@@ -26,8 +26,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestReshape2(OpTest):
     def setUp(self):
         self.set_npu()
@@ -51,12 +49,10 @@ class TestReshape2(OpTest):
         self.infered_shape = (20, 10)
 
     def test_check_output(self):
-        self.check_output_with_place(
-            self.place, check_dygraph=False, no_check_set=['XShape'])
+        self.check_output_with_place(self.place, no_check_set=['XShape'])
 
     def test_check_grad_normal(self):
-        self.check_grad_with_place(
-            self.place, ['X'], 'Out', check_dygraph=False)
+        self.check_grad_with_place(self.place, ['X'], 'Out')
 
 
 class TestReshape2_case2(TestReshape2):
diff --git a/python/paddle/fluid/tests/unittests/npu/test_save_load_npu.py b/python/paddle/fluid/tests/unittests/npu/test_save_load_npu.py
index e7e7fb39c913b28f1e3597a5a5ba4f57f98c108d..3bdf8146fb22846369f484f92a6208e3883e65ad 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_save_load_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_save_load_npu.py
@@ -36,56 +36,42 @@ from test_static_save_load import *
 paddle.enable_static()
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestNPUSaveLoadBase(TestSaveLoadBase):
     def set_place(self):
         return fluid.CPUPlace() if not core.is_compiled_with_npu(
         ) else paddle.NPUPlace(0)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestNPUSaveLoadPartial(TestSaveLoadPartial):
     def set_place(self):
         return fluid.CPUPlace() if not core.is_compiled_with_npu(
         ) else paddle.NPUPlace(0)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestNPUSaveLoadSetStateDict(TestSaveLoadSetStateDict):
     def set_place(self):
         return fluid.CPUPlace() if not core.is_compiled_with_npu(
         ) else paddle.NPUPlace(0)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestNPUProgramStatePartial(TestProgramStatePartial):
     def set_place(self):
         return fluid.CPUPlace() if not core.is_compiled_with_npu(
         ) else paddle.NPUPlace(0)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestNPULoadFromOldInterface(TestLoadFromOldInterface):
     def set_place(self):
         return fluid.CPUPlace() if not core.is_compiled_with_npu(
         ) else paddle.NPUPlace(0)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestNPULoadFromOldInterfaceSingleFile(TestLoadFromOldInterfaceSingleFile):
     def set_place(self):
         return fluid.CPUPlace() if not core.is_compiled_with_npu(
         ) else paddle.NPUPlace(0)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestNPUProgramStateOldSave(TestProgramStateOldSave):
     def setUp(self):
         self.test_dygraph = False
@@ -95,8 +81,6 @@ class TestNPUProgramStateOldSave(TestProgramStateOldSave):
         ) else paddle.NPUPlace(0)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestNPUProgramStateOldSaveSingleModel(TestProgramStateOldSaveSingleModel):
     def set_place(self):
         return fluid.CPUPlace() if not core.is_compiled_with_npu(
diff --git a/python/paddle/fluid/tests/unittests/npu/test_scale_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_scale_op_npu.py
index 9b4547bc24474afccf2454f992a1c92c3dd22605..65ec28fbf7d3a394ce0ac93c81a651445d584c75 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_scale_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_scale_op_npu.py
@@ -26,8 +26,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestScale(OpTest):
     def setUp(self):
         self.set_npu()
@@ -51,7 +49,7 @@ class TestScale(OpTest):
         self.dtype = np.float32
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
 class TestFP16Scale(TestScale):
@@ -82,7 +80,7 @@ class TestBiasAfterScale(OpTest):
         self.dtype = np.float32
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
 if __name__ == '__main__':
diff --git a/python/paddle/fluid/tests/unittests/npu/test_scatter_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_scatter_op_npu.py
index c3e52c9bfad533bdef724cff7e447f991fa2d6b2..c05b53d9a48621c61c8c958c360a13d2e9bf9466 100755
--- a/python/paddle/fluid/tests/unittests/npu/test_scatter_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_scatter_op_npu.py
@@ -27,8 +27,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestCast1(OpTest):
     def setUp(self):
         self.set_npu()
@@ -49,7 +47,7 @@ class TestCast1(OpTest):
         self.__class__.use_npu = True
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
 class TestCast2(OpTest):
@@ -72,7 +70,7 @@ class TestCast2(OpTest):
         self.__class__.use_npu = True
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
 class TestCast3(OpTest):
@@ -95,7 +93,7 @@ class TestCast3(OpTest):
         self.__class__.use_npu = True
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
 class TestCast4(OpTest):
@@ -119,7 +117,7 @@ class TestCast4(OpTest):
         self.__class__.use_npu = True
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
 if __name__ == '__main__':
diff --git a/python/paddle/fluid/tests/unittests/npu/test_seed_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_seed_op_npu.py
index 29914d21e2673ca92440c5346bfc2d6b5522849c..85a1e0594ba945cd267ea886f1979c25f14ef8d8 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_seed_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_seed_op_npu.py
@@ -26,8 +26,6 @@ import paddle.fluid.core as core
 paddle.enable_static()
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestSeedOpFixSeed(OpTest):
     def setUp(self):
         self.set_npu()
@@ -43,8 +41,6 @@ class TestSeedOpFixSeed(OpTest):
         self.check_output_with_place(paddle.NPUPlace(0))
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestSeedOpDiffSeed(OpTest):
     def setUp(self):
         self.set_npu()
diff --git a/python/paddle/fluid/tests/unittests/npu/test_sgd_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_sgd_op_npu.py
index af0dea4776d23fdebe26f68b5c84c7d3d07d2940..99061cba8d27000c1917b61a7efeb8215316cec2 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_sgd_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_sgd_op_npu.py
@@ -24,8 +24,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestSGD(OpTest):
     def setUp(self):
         self.set_npu()
@@ -50,11 +48,9 @@ class TestSGD(OpTest):
         self.w = 15
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestNet(unittest.TestCase):
     def _test(self, run_npu=True):
         main_prog = paddle.static.Program()
diff --git a/python/paddle/fluid/tests/unittests/npu/test_shape_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_shape_op_npu.py
index 7b9a74b2be98dee86f2b3192d746cc56895ca1d9..cb1b0c458fcaa95d16f22c8f61e6bfbf92aae89e 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_shape_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_shape_op_npu.py
@@ -26,8 +26,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestShape(OpTest):
     def setUp(self):
         self.set_npu()
@@ -50,7 +48,7 @@ class TestShape(OpTest):
         self.dtype = np.float32
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
 if __name__ == '__main__':
diff --git a/python/paddle/fluid/tests/unittests/npu/test_slice_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_slice_op_npu.py
index b56ee8c8c0748b1e8afacd5cab3ab1b721d35cfa..5a38f14868bb8a0feb93a47b00b591edaf0a5855 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_slice_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_slice_op_npu.py
@@ -27,8 +27,6 @@ SEED = 2021
 EPOCH = 100
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestSliceOp(OpTest):
     def setUp(self):
         self.op_type = "slice"
@@ -60,13 +58,12 @@ class TestSliceOp(OpTest):
         self.place = paddle.NPUPlace(0)
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
     def test_check_grad_normal(self):
         if self.dtype == np.float16:
             return
-        self.check_grad_with_place(
-            self.place, ['Input'], 'Out', check_dygraph=False)
+        self.check_grad_with_place(self.place, ['Input'], 'Out')
 
 
 class TestSliceOp2(TestSliceOp):
@@ -79,8 +76,6 @@ class TestSliceOp2(TestSliceOp):
         self.out = self.input[:, 0:1, :]
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestSliceOpFp16(TestSliceOp):
     def init_dtype(self):
         self.dtype = np.float16
@@ -147,8 +142,6 @@ class TestSliceOpTensor2(TestSliceOpTensor):
         self.out = self.input[:, 0:1, :]
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestSliceOpFp16Tensor(TestSliceOpTensor):
     def init_dtype(self):
         self.dtype = np.float16
@@ -237,8 +230,6 @@ class TestSliceOpTensorList2(TestSliceOpTensorList):
         self.out = self.input[:, 0:1, :]
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestSliceOpFp16TensorList(TestSliceOpTensorList):
     def init_dtype(self):
         self.dtype = np.float16
@@ -249,8 +240,6 @@ class TestSliceOpFp16TensorList(TestSliceOpTensorList):
         self.place = paddle.NPUPlace(0)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestSliceNet(unittest.TestCase):
     def _test(self, run_npu=True):
         main_prog = paddle.static.Program()
diff --git a/python/paddle/fluid/tests/unittests/npu/test_softmax_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_softmax_op_npu.py
index c1ba41943a359ba2103bfd34c722c697d6b01b2f..f2a9ef2bee074def9bd5cd8d3f4ae0a6ac913815 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_softmax_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_softmax_op_npu.py
@@ -27,8 +27,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestSoftmax(OpTest):
     def setUp(self):
         self.set_npu()
@@ -51,11 +49,9 @@ class TestSoftmax(OpTest):
         self.dtype = np.float32
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestSoftmaxNet(unittest.TestCase):
     def _test(self, run_npu=True):
         main_prog = paddle.static.Program()
diff --git a/python/paddle/fluid/tests/unittests/npu/test_softmax_with_cross_entropy_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_softmax_with_cross_entropy_op_npu.py
index 2ee089360e6dd2f62aabdd25179e7e9410b365e4..8d78ee6a97efdd1df99c9636e8e18a2905d858a5 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_softmax_with_cross_entropy_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_softmax_with_cross_entropy_op_npu.py
@@ -28,8 +28,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestSoftmaxWithCrossEntropyOp(OpTest):
     def set_npu(self):
         self.__class__.use_npu = True
@@ -86,7 +84,7 @@ class TestSoftmaxWithCrossEntropyOp(OpTest):
             self.attrs['axis'] = self.axis
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
     def test_check_grad(self):
         if self.dtype == np.float16:
@@ -95,13 +93,10 @@ class TestSoftmaxWithCrossEntropyOp(OpTest):
         self.check_grad_with_place(
             self.place, ['Logits'],
             'Loss',
-            check_dygraph=False,
             numeric_grad_delta=0.001,
             max_relative_error=0.5)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestPowNet(unittest.TestCase):
     def _test(self, run_npu=True):
         main_prog = paddle.static.Program()
diff --git a/python/paddle/fluid/tests/unittests/npu/test_sqrt_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_sqrt_op_npu.py
index 556fa76424b8b60f2efff371c833f57bdc341e40..acb99746d231ded16032bfdc1839b6b0f3120f62 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_sqrt_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_sqrt_op_npu.py
@@ -26,8 +26,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestSqrt(OpTest):
     def setUp(self):
         self.set_npu()
@@ -50,7 +48,7 @@ class TestSqrt(OpTest):
         self.dtype = np.float32
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
     # TODO(ascendrc): Add grad test
     # def test_check_grad(self):
@@ -60,8 +58,6 @@ class TestSqrt(OpTest):
     #
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestSqrtFp16(OpTest):
     def setUp(self):
         self.set_npu()
@@ -85,11 +81,9 @@ class TestSqrtFp16(OpTest):
         self.dtype = np.float16
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False, atol=1e-5)
+        self.check_output_with_place(self.place, atol=1e-5)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestSqrtNet(unittest.TestCase):
     def _test(self, run_npu=True):
         main_prog = paddle.static.Program()
diff --git a/python/paddle/fluid/tests/unittests/npu/test_square_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_square_op_npu.py
index 8c1a8d0070484a3b536256a6e8aafeb20fcf0ae0..8262c3b94d6a8531fe104eed6c4c447d0e9edef1 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_square_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_square_op_npu.py
@@ -26,8 +26,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestSquare(OpTest):
     def setUp(self):
         self.set_npu()
@@ -50,7 +48,7 @@ class TestSquare(OpTest):
         self.dtype = np.float32
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
     # TODO(ascendrc): Add grad test
     # def test_check_grad(self):
@@ -60,8 +58,6 @@ class TestSquare(OpTest):
     #
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestSquareFp16(OpTest):
     def setUp(self):
         self.set_npu()
@@ -85,11 +81,9 @@ class TestSquareFp16(OpTest):
         self.dtype = np.float16
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False, atol=1e-5)
+        self.check_output_with_place(self.place, atol=1e-5)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestSquareNet(unittest.TestCase):
     def _test(self, run_npu=True):
         main_prog = paddle.static.Program()
diff --git a/python/paddle/fluid/tests/unittests/npu/test_squeeze_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_squeeze_op_npu.py
index 7a725b3b9d5d31f174a2b1c818480f95d8c4274d..2e741c8d8a5ef00f23fb5171dff0f5e147203094 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_squeeze_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_squeeze_op_npu.py
@@ -26,10 +26,9 @@ from paddle.fluid import Program, program_guard
 
 paddle.enable_static()
 
-
 # Correct: General.
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
+
+
 class TestSqueezeOp(OpTest):
     def setUp(self):
         self.set_npu()
@@ -58,8 +57,8 @@ class TestSqueezeOp(OpTest):
 
 
 # Correct: There is mins axis.
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
+
+
 class TestSqueezeOp1(TestSqueezeOp):
     def init_test_case(self):
         self.ori_shape = (1, 3, 1, 40)
@@ -68,8 +67,8 @@ class TestSqueezeOp1(TestSqueezeOp):
 
 
 # Correct: No axes input.
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
+
+
 class TestSqueezeOp2(TestSqueezeOp):
     def init_test_case(self):
         self.ori_shape = (1, 20, 1, 5)
@@ -78,8 +77,8 @@ class TestSqueezeOp2(TestSqueezeOp):
 
 
 # Correct: Just part of axes be squeezed. 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
+
+
 class TestSqueezeOp3(TestSqueezeOp):
     def init_test_case(self):
         self.ori_shape = (6, 1, 5, 1, 4, 1)
@@ -88,8 +87,8 @@ class TestSqueezeOp3(TestSqueezeOp):
 
 
 # Correct: The demension of axis is not of size 1 remains unchanged.
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
+
+
 class TestSqueezeOp4(TestSqueezeOp):
     def init_test_case(self):
         self.ori_shape = (6, 1, 5, 1, 4, 1)
diff --git a/python/paddle/fluid/tests/unittests/npu/test_stack_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_stack_op_npu.py
index 721fb95dd9b72f989746bbe1a7e27596a6b18a34..bdfc7a03c6c8324ce8fe8cc4cd211a058d7fcb75 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_stack_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_stack_op_npu.py
@@ -26,8 +26,6 @@ import paddle.fluid.core as core
 paddle.enable_static()
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestStackOpBase(OpTest):
     def initDefaultParameters(self):
         self.num_inputs = 4
@@ -77,50 +75,36 @@ class TestStackOpBase(OpTest):
         self.check_grad_with_place(self.place, self.get_x_names(), 'Y')
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestStackOp1(TestStackOpBase):
     def initParameters(self):
         self.num_inputs = 16
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestStackOp2(TestStackOpBase):
     def initParameters(self):
         self.num_inputs = 20
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestStackOp3(TestStackOpBase):
     def initParameters(self):
         self.axis = -1
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestStackOp4(TestStackOpBase):
     def initParameters(self):
         self.axis = -4
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestStackOp5(TestStackOpBase):
     def initParameters(self):
         self.axis = 1
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestStackOp6(TestStackOpBase):
     def initParameters(self):
         self.axis = 3
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestStackAPIWithLoDTensorArray(unittest.TestCase):
     """
     Test stack api when the input(x) is a LoDTensorArray.
@@ -157,8 +141,6 @@ class TestStackAPIWithLoDTensorArray(unittest.TestCase):
                     [self.x] * self.iter_num, axis=self.axis)))
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestTensorStackAPIWithLoDTensorArray(unittest.TestCase):
     """
     Test stack api when the input(x) is a LoDTensorArray.
@@ -195,8 +177,6 @@ class TestTensorStackAPIWithLoDTensorArray(unittest.TestCase):
                     [self.x] * self.iter_num, axis=self.axis)))
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class API_test(unittest.TestCase):
     def test_out(self):
         with fluid.program_guard(fluid.Program(), fluid.Program()):
@@ -223,8 +203,6 @@ class API_test(unittest.TestCase):
             self.assertRaises(TypeError, paddle.stack, x)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class API_DygraphTest(unittest.TestCase):
     def test_out(self):
         data1 = np.array([[1.0, 2.0]])
diff --git a/python/paddle/fluid/tests/unittests/npu/test_sum_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_sum_op_npu.py
index 21b42814c07b062937019e36c67f236d1c5e1c5d..1ea8504ceec01f7b8349ee3d4837757c4dd5fa29 100755
--- a/python/paddle/fluid/tests/unittests/npu/test_sum_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_sum_op_npu.py
@@ -27,8 +27,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestSum1(OpTest):
     def setUp(self):
         self.set_npu()
@@ -52,7 +50,7 @@ class TestSum1(OpTest):
         self.__class__.use_npu = True
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
 class TestSum2(OpTest):
@@ -86,7 +84,7 @@ class TestSum2(OpTest):
         self.__class__.use_npu = True
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
 class TestSum3(OpTest):
@@ -111,7 +109,7 @@ class TestSum3(OpTest):
         self.__class__.use_npu = True
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
 if __name__ == '__main__':
diff --git a/python/paddle/fluid/tests/unittests/npu/test_tanh_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_tanh_op_npu.py
index 235fa2783fb3c8c507ebfa73c5631c551fce4f1a..55be94da2b7e0346d8c6783d244c9d3a2c43273e 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_tanh_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_tanh_op_npu.py
@@ -26,8 +26,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestTanh(OpTest):
     def setUp(self):
         self.set_npu()
@@ -50,7 +48,7 @@ class TestTanh(OpTest):
         self.dtype = np.float32
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
     # TODO(ascendrc): Add grad test
     # def test_check_grad(self):
@@ -60,8 +58,6 @@ class TestTanh(OpTest):
     #
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestTanhFp16(OpTest):
     def setUp(self):
         self.set_npu()
@@ -85,11 +81,9 @@ class TestTanhFp16(OpTest):
         self.dtype = np.float16
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False, atol=1e-3)
+        self.check_output_with_place(self.place, atol=1e-3)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestTanhNet(unittest.TestCase):
     def _test(self, run_npu=True):
         main_prog = paddle.static.Program()
diff --git a/python/paddle/fluid/tests/unittests/npu/test_top_k_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_top_k_op_npu.py
index 04d4565f7485808daa706a9781c2b7159ab9222a..b735adf76d6c1296197960b084aac660c688c5cd 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_top_k_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_top_k_op_npu.py
@@ -27,8 +27,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestTopk(OpTest):
     def setUp(self):
         self.set_npu()
@@ -56,11 +54,9 @@ class TestTopk(OpTest):
         self.dtype = np.float16
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestTopkV2(OpTest):
     def setUp(self):
         self.set_npu()
@@ -88,7 +84,7 @@ class TestTopkV2(OpTest):
         self.dtype = np.float16
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
 if __name__ == '__main__':
diff --git a/python/paddle/fluid/tests/unittests/npu/test_transpose_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_transpose_op_npu.py
index 17f6a0ae1ca9bffcb78b3526f78f9a26e4546fc4..e95f3cc83cfb31ee955d65c9f722eb2c0180db86 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_transpose_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_transpose_op_npu.py
@@ -25,8 +25,6 @@ import paddle.fluid as fluid
 paddle.enable_static()
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestTransposeOp(OpTest):
     def setUp(self):
         self.set_npu()
@@ -58,11 +56,9 @@ class TestTransposeOp(OpTest):
         self.axis = -1
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestTransposeOpFP16(TestTransposeOp):
     no_need_check_grad = True
 
diff --git a/python/paddle/fluid/tests/unittests/npu/test_truncated_gaussian_random_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_truncated_gaussian_random_op_npu.py
index ff89508d196235a8e50678908938ba0fc24d6981..de94e7febaca7692543cb9748f1481557f53d208 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_truncated_gaussian_random_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_truncated_gaussian_random_op_npu.py
@@ -29,8 +29,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestTruncatedNormal(unittest.TestCase):
     def _test(self, run_npu=True):
         main_prog = paddle.static.Program()
diff --git a/python/paddle/fluid/tests/unittests/npu/test_uniform_random_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_uniform_random_op_npu.py
index 7c358c244f34ddf1c9b3d9e3f055ea333f9b435f..0e21c59432badc13e13a4eaef7871f8ee888050f 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_uniform_random_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_uniform_random_op_npu.py
@@ -39,8 +39,6 @@ def output_hist(out):
     return hist, prob
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestNPUUniformRandomOp(OpTest):
     def setUp(self):
         self.set_npu()
@@ -76,8 +74,6 @@ class TestNPUUniformRandomOp(OpTest):
                 hist, prob, rtol=0, atol=0.01), "hist: " + str(hist))
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestNPUUniformRandomOpSelectedRows(unittest.TestCase):
     def get_places(self):
         places = [core.CPUPlace()]
diff --git a/python/paddle/fluid/tests/unittests/npu/test_unstack_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_unstack_op_npu.py
index 6dd3c30c272c237b9aff31274b2274cdfa08cf8e..097f31c72467c049dca6f246343a7c79f6ad4561 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_unstack_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_unstack_op_npu.py
@@ -24,8 +24,6 @@ import paddle
 paddle.enable_static()
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestUnStackOpBase(OpTest):
     def initDefaultParameters(self):
         self.input_dim = (5, 6, 7)
@@ -75,29 +73,21 @@ class TestUnStackOpBase(OpTest):
         self.check_grad_with_place(self.place, ['X'], self.get_y_names())
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestStackOp3(TestUnStackOpBase):
     def initParameters(self):
         self.axis = -1
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestStackOp4(TestUnStackOpBase):
     def initParameters(self):
         self.axis = -3
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestStackOp5(TestUnStackOpBase):
     def initParameters(self):
         self.axis = 1
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestStackOp6(TestUnStackOpBase):
     def initParameters(self):
         self.axis = 2
diff --git a/python/paddle/fluid/tests/unittests/npu/test_update_loss_scaling_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_update_loss_scaling_op_npu.py
index cae3239229f441812a42be11ee8d8f34253cff05..1388adf609ff62a48832836e461debe09fc9bdca 100644
--- a/python/paddle/fluid/tests/unittests/npu/test_update_loss_scaling_op_npu.py
+++ b/python/paddle/fluid/tests/unittests/npu/test_update_loss_scaling_op_npu.py
@@ -25,8 +25,6 @@ paddle.enable_static()
 SEED = 2021
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestUpdateLossScalingOp(OpTest):
     def setUp(self):
         self.set_npu()
@@ -71,7 +69,7 @@ class TestUpdateLossScalingOp(OpTest):
         }
 
     def test_check_output(self):
-        self.check_output_with_place(self.place, check_dygraph=False)
+        self.check_output_with_place(self.place)
 
 
 class TestUpdateLossScalingOpBad(TestUpdateLossScalingOp):
@@ -103,8 +101,6 @@ class TestUpdateLossScalingOpBad(TestUpdateLossScalingOp):
         }
 
 
-@unittest.skipIf(not paddle.is_compiled_with_npu(),
-                 "core is not compiled with NPU")
 class TestUpdateLossScalingLayer(unittest.TestCase):
     def loss_scaling_check(self, use_npu=True, scope=fluid.Scope()):
         a = fluid.data(name="a", shape=[1024, 1024], dtype='float32')