diff --git a/paddle/phi/backends/xpu/xpu2_op_list.cc b/paddle/phi/backends/xpu/xpu2_op_list.cc
index 77b113c61b005180a240303fd4b73320781ea7f0..a18683f6b52e62f03d14c5f15a615a1962a2ba82 100644
--- a/paddle/phi/backends/xpu/xpu2_op_list.cc
+++ b/paddle/phi/backends/xpu/xpu2_op_list.cc
@@ -269,12 +269,6 @@ XPUOpMap& get_kl2_ops() {
                      phi::DataType::INT8,
                      phi::DataType::FLOAT16,
                      phi::DataType::FLOAT32})},
-      {"flatten",
-       XPUKernelSet({phi::DataType::INT64,
-                     phi::DataType::INT32,
-                     phi::DataType::INT8,
-                     phi::DataType::FLOAT16,
-                     phi::DataType::FLOAT32})},
       {"flatten_grad",
        XPUKernelSet({phi::DataType::INT64,
                      phi::DataType::INT32,
diff --git a/python/paddle/fluid/tests/unittests/xpu/test_zero_dim_tensor_xpu.py b/python/paddle/fluid/tests/unittests/xpu/test_zero_dim_tensor_xpu.py
index 64085e2fcdc353565b8de906a5f6ce04e2956bf2..97925b72beaa76f7c2e69a5f6d467a6ec1b7c976 100644
--- a/python/paddle/fluid/tests/unittests/xpu/test_zero_dim_tensor_xpu.py
+++ b/python/paddle/fluid/tests/unittests/xpu/test_zero_dim_tensor_xpu.py
@@ -747,6 +747,7 @@ class TestSundryAPI(unittest.TestCase):
             out1.numpy(),
         )
 
+        out0.retain_grads()
         out0.backward()
         self.assertEqual(out0.grad.shape, [1])
         self.assertEqual(logit.grad.shape, [2, 3])