diff --git a/paddle/phi/api/yaml/legacy_backward.yaml b/paddle/phi/api/yaml/legacy_backward.yaml
index 3028e0c6c73c63855de19799d51ea6265e1311df..cef9e44eeaf7f977871e2e012c5d4663cce8ef28 100755
--- a/paddle/phi/api/yaml/legacy_backward.yaml
+++ b/paddle/phi/api/yaml/legacy_backward.yaml
@@ -797,10 +797,7 @@
   forward : expand_grad (Tensor x, Tensor grad_out, IntArray shape) -> Tensor(grad_x)
   args : (Tensor grad_x_grad, IntArray shape)
   output : Tensor(grad_out_grad)
-  infer_meta :
-    func : ExpandInferMeta
-  kernel :
-    func : expand
+  invoke : expand(grad_x_grad, shape)
 
 - backward_api : expand_grad
   forward : expand (Tensor x, IntArray shape) -> Tensor(out)
diff --git a/python/paddle/fluid/tests/unittests/test_expand_v2_op.py b/python/paddle/fluid/tests/unittests/test_expand_v2_op.py
index 6fc6fc8f7eb6b49148b19acf0ca7b28c038ee760..82fb8284fe78592b20a827febd74b6d4dd4edaed 100644
--- a/python/paddle/fluid/tests/unittests/test_expand_v2_op.py
+++ b/python/paddle/fluid/tests/unittests/test_expand_v2_op.py
@@ -18,9 +18,12 @@ import unittest
 import numpy as np
 from op_test import OpTest
 import paddle.fluid as fluid
-from paddle.fluid import compiler, Program, program_guard
+from paddle.fluid import compiler, Program, program_guard, core
 import paddle
 from paddle.fluid.framework import _test_eager_guard
+import gradient_checker
+from decorator_helper import prog_scope
+import paddle.fluid.layers as layers
 
 
 # Situation 1: shape is a list(without tensor)
@@ -284,6 +287,80 @@ class TestExpandV2DygraphAPI(unittest.TestCase):
                                           egr_expand_1.numpy())
 
 
+class TestExpandDoubleGradCheck(unittest.TestCase):
+
+    def expand_wrapper(self, x):
+        return paddle.expand(x[0], [2, 3])
+
+    @prog_scope()
+    def func(self, place):
+        # the shape of input variable should be clearly specified, not inlcude -1.
+        eps = 0.005
+        dtype = np.float32
+
+        data = layers.data('data', [2, 3], False, dtype)
+        data.persistable = True
+        out = paddle.expand(data, [2, 3])
+        data_arr = np.random.uniform(-1, 1, data.shape).astype(dtype)
+
+        gradient_checker.double_grad_check([data],
+                                           out,
+                                           x_init=[data_arr],
+                                           place=place,
+                                           eps=eps)
+        fluid.set_flags({"FLAGS_retain_grad_for_all_tensor": True})
+        gradient_checker.double_grad_check_for_dygraph(self.expand_wrapper,
+                                                       [data],
+                                                       out,
+                                                       x_init=[data_arr],
+                                                       place=place)
+
+    def test_grad(self):
+        paddle.enable_static()
+        places = [fluid.CPUPlace()]
+        if core.is_compiled_with_cuda():
+            places.append(fluid.CUDAPlace(0))
+        for p in places:
+            self.func(p)
+
+
+class TestExpandTripleGradCheck(unittest.TestCase):
+
+    def expand_wrapper(self, x):
+        return paddle.expand(x[0], [2, 3])
+
+    @prog_scope()
+    def func(self, place):
+        # the shape of input variable should be clearly specified, not inlcude -1.
+        eps = 0.005
+        dtype = np.float32
+
+        data = layers.data('data', [2, 3], False, dtype)
+        data.persistable = True
+        out = paddle.expand(data, [2, 3])
+        data_arr = np.random.uniform(-1, 1, data.shape).astype(dtype)
+
+        gradient_checker.triple_grad_check([data],
+                                           out,
+                                           x_init=[data_arr],
+                                           place=place,
+                                           eps=eps)
+        fluid.set_flags({"FLAGS_retain_grad_for_all_tensor": True})
+        gradient_checker.triple_grad_check_for_dygraph(self.expand_wrapper,
+                                                       [data],
+                                                       out,
+                                                       x_init=[data_arr],
+                                                       place=place)
+
+    def test_grad(self):
+        paddle.enable_static()
+        places = [fluid.CPUPlace()]
+        if core.is_compiled_with_cuda():
+            places.append(fluid.CUDAPlace(0))
+        for p in places:
+            self.func(p)
+
+
 if __name__ == "__main__":
     paddle.enable_static()
     unittest.main()