diff --git a/paddle/phi/api/yaml/legacy_backward.yaml b/paddle/phi/api/yaml/legacy_backward.yaml
index 2050653904d16f5293c65c378659fd6072ab0c96..9810242037bcec02dee243f74a0af434b334440e 100755
--- a/paddle/phi/api/yaml/legacy_backward.yaml
+++ b/paddle/phi/api/yaml/legacy_backward.yaml
@@ -411,6 +411,16 @@
   kernel :
     func : einsum_grad
 
+- backward_op : elementwise_heaviside_grad
+  forward : heaviside (Tensor x, Tensor y) -> Tensor(out)
+  args : (Tensor x, Tensor y, Tensor out_grad, int axis = -1)
+  output : Tensor(x_grad), Tensor(y_grad)
+  infer_meta :
+    func : GeneralBinaryGradInferMeta
+    param : [x, y]
+  kernel :
+    func : elementwise_heaviside_grad
+
 - backward_op : elementwise_pow_grad
   forward : elementwise_pow(Tensor x, Tensor y) -> Tensor(out)
   args : (Tensor x, Tensor y, Tensor out_grad, int axis=-1)
diff --git a/paddle/phi/api/yaml/legacy_ops.yaml b/paddle/phi/api/yaml/legacy_ops.yaml
index b448dadbd58ca2b141b6f752ab8f292c5220365f..3be9cdf371d317a1d671b2bb606d1f1b76f52cba 100755
--- a/paddle/phi/api/yaml/legacy_ops.yaml
+++ b/paddle/phi/api/yaml/legacy_ops.yaml
@@ -564,6 +564,15 @@
     func : einsum
   backward : einsum_grad
 
+- op : elementwise_heaviside
+  args : (Tensor x, Tensor y)
+  output : Tensor
+  infer_meta :
+    func : ElementwiseInferMeta
+  kernel :
+    func : elementwise_heaviside
+  backward : elementwise_heaviside_grad
+
 - op : elementwise_pow
   args : (Tensor x, Tensor y)
   output : Tensor(out)
diff --git a/python/paddle/fluid/tests/unittests/test_elementwise_heaviside_op.py b/python/paddle/fluid/tests/unittests/test_elementwise_heaviside_op.py
index d917a6bc33a661c0e448b1a25c22f3a7d3f52935..4e80ef918704d6973cebe843466c999884418b63 100644
--- a/python/paddle/fluid/tests/unittests/test_elementwise_heaviside_op.py
+++ b/python/paddle/fluid/tests/unittests/test_elementwise_heaviside_op.py
@@ -32,20 +32,21 @@ class TestElementwiseOp(OpTest):
         self.op_type = "elementwise_heaviside"
         x = np.random.random((13, 17)).astype("float64")
         y = np.random.random((13, 17)).astype("float64")
+        self.python_api = paddle.heaviside
         self.inputs = {'X': x, 'Y': y}
         self.outputs = {'Out': np.heaviside(self.inputs['X'], self.inputs['Y'])}
 
     def test_check_output(self):
-        self.check_output()
+        self.check_output(check_eager=True)
 
     def test_check_grad_normal(self):
-        self.check_grad(['X', 'Y'], 'Out')
+        self.check_grad(['X', 'Y'], 'Out', check_eager=True)
 
     def test_check_grad_ingore_x(self):
-        self.check_grad(['Y'], 'Out', no_grad_set=set("X"))
+        self.check_grad(['Y'], 'Out', no_grad_set=set("X"), check_eager=True)
 
     def test_check_grad_ingore_y(self):
-        self.check_grad(['X'], 'Out', no_grad_set=set('Y'))
+        self.check_grad(['X'], 'Out', no_grad_set=set('Y'), check_eager=True)
 
 
 class TestHeavisideBroadcast(unittest.TestCase):
diff --git a/python/paddle/tensor/math.py b/python/paddle/tensor/math.py
index 32978bc30dd994aaa2d08c87b988bf1d2a9054ed..81b092f4c38b4e177a47abc496c85ccae79d409c 100644
--- a/python/paddle/tensor/math.py
+++ b/python/paddle/tensor/math.py
@@ -4857,14 +4857,10 @@ def heaviside(x, y, name=None):
             #    [[0.        , 0.20000000, 1.        ],
             #     [0.        , 1.        , 0.30000001]]
     """
-    op_type = 'elementwise_heaviside'
-    axis = -1
-    act = None
     if in_dygraph_mode():
-        return _elementwise_op_in_dygraph(
-            x, y, axis=axis, act=act, op_name=op_type
-        )
+        return _C_ops.elementwise_heaviside(x, y)
     else:
+        op_type = 'elementwise_heaviside'
         return _elementwise_op(LayerHelper(op_type, **locals()))