diff --git a/paddle/fluid/eager/auto_code_generator/final_state_generator/codegen_utils.py b/paddle/fluid/eager/auto_code_generator/final_state_generator/codegen_utils.py
index 61ed1deb27f95659370079b3d4be4d0c84b92343..8c98d9fa275dc12188f38210ce043e9788cebb6e 100644
--- a/paddle/fluid/eager/auto_code_generator/final_state_generator/codegen_utils.py
+++ b/paddle/fluid/eager/auto_code_generator/final_state_generator/codegen_utils.py
@@ -27,7 +27,7 @@ ops_to_fill_zero_for_empty_grads = set([
     "add_triple_grad", "multiply_double_grad", "multiply_triple_grad",
     "conv2d_grad_grad", "batch_norm_double_grad", "tanh_double_grad",
     "tanh_triple_grad", "subtract_double_grad", "divide_double_grad",
-    "log_double_grad", "elu_double_grad"
+    "log_double_grad", "elu_double_grad", "leaky_relu_double_grad"
 ])
 
 # For API dispatch used at python-level
diff --git a/python/paddle/fluid/tests/unittests/test_activation_nn_grad.py b/python/paddle/fluid/tests/unittests/test_activation_nn_grad.py
index 9fcb38641850edf475c1f92d8aa2aa6a96b559de..570551e82646f8501b2d50d0e4c612e114f7c8b6 100644
--- a/python/paddle/fluid/tests/unittests/test_activation_nn_grad.py
+++ b/python/paddle/fluid/tests/unittests/test_activation_nn_grad.py
@@ -161,6 +161,9 @@ class TestReluDoubleGradCheck(unittest.TestCase):
 
 
 class TestLeakyReluDoubleGradCheck(unittest.TestCase):
+    def leaky_relu_wrapper(self, x):
+        return paddle.nn.functional.leaky_relu(x[0], negative_slope=0.2)
+
     @prog_scope()
     def func(self, place):
         shape = [2, 3, 7, 9]
@@ -177,6 +180,8 @@ class TestLeakyReluDoubleGradCheck(unittest.TestCase):
 
         gradient_checker.double_grad_check(
             [x], y, x_init=x_arr, place=place, eps=eps)
+        gradient_checker.double_grad_check_for_dygraph(
+            self.leaky_relu_wrapper, [x], y, x_init=x_arr, place=place)
 
     def test_grad(self):
         paddle.enable_static()
diff --git a/python/paddle/fluid/tests/unittests/test_norm_nn_grad.py b/python/paddle/fluid/tests/unittests/test_norm_nn_grad.py
index 49fe397644dc6e79d1b0d436b31c706d77906b6a..1452b869d4f8b8569d211c083d1c3afc736d0b29 100644
--- a/python/paddle/fluid/tests/unittests/test_norm_nn_grad.py
+++ b/python/paddle/fluid/tests/unittests/test_norm_nn_grad.py
@@ -43,6 +43,7 @@ class TestInstanceNormDoubleGradCheck(unittest.TestCase):
                 [x], z, x_init=x_arr, atol=atol, place=place, eps=eps)
 
     def test_grad(self):
+        paddle.enable_static()
         places = [fluid.CPUPlace()]
         if core.is_compiled_with_cuda():
             places.append(fluid.CUDAPlace(0))
@@ -77,6 +78,14 @@ class TestBatchNormDoubleGradCheck(unittest.TestCase):
         self.data_layout = 'NCHW'
         self.use_global_stats = False
         self.shape = [2, 3, 4, 5]
+        self.channel_index = 1
+
+    def batch_norm_wrapper(self, x):
+        batch_norm = paddle.nn.BatchNorm2D(
+            self.shape[self.channel_index],
+            data_format=self.data_layout,
+            use_global_stats=self.use_global_stats)
+        return batch_norm(x[0])
 
     @prog_scope()
     def func(self, place):
@@ -94,8 +103,15 @@ class TestBatchNormDoubleGradCheck(unittest.TestCase):
             x_arr = np.random.uniform(-1, 1, self.shape).astype(dtype)
             gradient_checker.double_grad_check(
                 [x], z, x_init=x_arr, atol=atol, place=place, eps=eps)
+            gradient_checker.double_grad_check_for_dygraph(
+                self.batch_norm_wrapper, [x],
+                z,
+                x_init=x_arr,
+                atol=atol,
+                place=place)
 
     def test_grad(self):
+        paddle.enable_static()
         places = [fluid.CPUPlace()]
         if core.is_compiled_with_cuda():
             places.append(fluid.CUDAPlace(0))
@@ -108,6 +124,7 @@ class TestBatchNormDoubleGradCheckCase1(TestBatchNormDoubleGradCheck):
         self.data_layout = 'NHWC'
         self.use_global_stats = False
         self.shape = [2, 3, 4, 5]
+        self.channel_index = 3
 
 
 class TestBatchNormDoubleGradCheckCase2(TestBatchNormDoubleGradCheck):
@@ -115,6 +132,7 @@ class TestBatchNormDoubleGradCheckCase2(TestBatchNormDoubleGradCheck):
         self.data_layout = 'NCHW'
         self.use_global_stats = True
         self.shape = [2, 3, 4, 5]
+        self.channel_index = 1
 
 
 class TestBatchNormDoubleGradCheckCase3(TestBatchNormDoubleGradCheck):
@@ -122,6 +140,7 @@ class TestBatchNormDoubleGradCheckCase3(TestBatchNormDoubleGradCheck):
         self.data_layout = 'NHWC'
         self.use_global_stats = True
         self.shape = [2, 3, 4, 5]
+        self.channel_index = 3
 
 
 class TestBatchNormDoubleGradCheckCase4(TestBatchNormDoubleGradCheck):
@@ -129,6 +148,14 @@ class TestBatchNormDoubleGradCheckCase4(TestBatchNormDoubleGradCheck):
         self.data_layout = 'NCHW'
         self.use_global_stats = False
         self.shape = [2, 2, 3, 4, 5]
+        self.channel_index = 1
+
+    def batch_norm_wrapper(self, x):
+        batch_norm = paddle.nn.BatchNorm3D(
+            self.shape[self.channel_index],
+            data_format=self.data_layout,
+            use_global_stats=self.use_global_stats)
+        return batch_norm(x[0])
 
 
 class TestBatchNormDoubleGradCheckCase5(TestBatchNormDoubleGradCheck):
@@ -165,8 +192,8 @@ class TestBatchNormDoubleGradCheckCase6(TestBatchNormDoubleGradCheckCase5):
         self.data_layout = 'NCHW'
         self.use_global_stats = True
         self.shape = [2, 3, 4, 5]
+        self.channel_index = 1
 
 
 if __name__ == "__main__":
-    paddle.enable_static()
     unittest.main()