diff --git a/python/paddle/v2/framework/tests/gradient_checker.py b/python/paddle/v2/framework/tests/gradient_checker.py
index 5f9e54837edae9fea8f564eb8974572f3d243403..d251f14b9d15804b0e64fd72ed9f780c321bc816 100644
--- a/python/paddle/v2/framework/tests/gradient_checker.py
+++ b/python/paddle/v2/framework/tests/gradient_checker.py
@@ -156,6 +156,7 @@ class GradientChecker(unittest.TestCase):
 
     def compare_grad(self, forward_op, inputs):
         backward_op = core.Operator.backward(forward_op, set())
+        # return if not compile with GPU or not implementing GPU kernel
         if not (core.is_compile_gpu() and backward_op.support_gpu()):
             return
 
@@ -239,7 +240,7 @@ class GradientChecker(unittest.TestCase):
         for place in places:
             # get analytical gradients according to different device
             analytic_grads = self.get_grad(forward_op, backward_op, input_vars,
-                                           check_grad_names, place)
+                                           check_names, place)
             self.assert_is_close(numeric_grads, analytic_grads, check_names,
                                  max_relative_error,
                                  "Gradient Check On %s" % str(place))
diff --git a/python/paddle/v2/framework/tests/test_sigmoid_op.py b/python/paddle/v2/framework/tests/test_sigmoid_op.py
index 1a6d395be662bc7c20bc405e6d269462e8de0eca..c3bd79f5dc0ba4b087b681b3f8c241def3ba33c8 100644
--- a/python/paddle/v2/framework/tests/test_sigmoid_op.py
+++ b/python/paddle/v2/framework/tests/test_sigmoid_op.py
@@ -17,15 +17,10 @@ class TestSigmoidGradOp(GradientChecker):
     def test_compare_grad(self):
         op = create_op("sigmoid")
         inputs = {"X": np.random.random((11, 17)).astype("float32")}
-
-        # compare gpu and cpu results for backward op
+        # compare gpu and cpu results for backward op.
+        # skip this test if only compiling CPU version.
         self.compare_grad(op, inputs)
-
-    def test_check_grad(self):
-        op = create_op("sigmoid")
-        inputs = {"X": np.random.random((11, 17)).astype("float32")}
-
-        # check gradients
+        # check gradients 
         self.check_grad(op, inputs, set("X"), "Y")