diff --git a/python/paddle/fluid/contrib/slim/tests/imperative_test_utils.py b/python/paddle/fluid/contrib/slim/tests/imperative_test_utils.py
index 1a5f52b0406d8a2392463165c880be652666296c..c0637bc7de480e60bc30b40dfaa2cde4f67a5876 100644
--- a/python/paddle/fluid/contrib/slim/tests/imperative_test_utils.py
+++ b/python/paddle/fluid/contrib/slim/tests/imperative_test_utils.py
@@ -153,7 +153,7 @@ class ImperativeLenet(fluid.dygraph.Layer):
         x = self.quant_stub(inputs)
         x = self.features(x)
 
-        x = fluid.layers.flatten(x, 1)
+        x = paddle.flatten(x, 1, -1)
         x = self.add(x, paddle.to_tensor(0.0))  # For CI
         x = self.fc(x)
         return x
@@ -238,7 +238,7 @@ class ImperativeLenetWithSkipQuant(fluid.dygraph.Layer):
         x = self.relu6_0(x)
         x = self.pool2d_1(x)
 
-        x = fluid.layers.flatten(x, 1)
+        x = paddle.flatten(x, 1, -1)
 
         x = self.linear_0(x)
         x = self.leaky_relu_0(x)
diff --git a/python/paddle/fluid/contrib/slim/tests/test_imperative_out_scale.py b/python/paddle/fluid/contrib/slim/tests/test_imperative_out_scale.py
index 02b19947ec9f60a5b207a93246b9c7a014032e9f..e978b636452078915b30a74b1a3383929d94b1ee 100644
--- a/python/paddle/fluid/contrib/slim/tests/test_imperative_out_scale.py
+++ b/python/paddle/fluid/contrib/slim/tests/test_imperative_out_scale.py
@@ -121,7 +121,7 @@ class ImperativeLenet(fluid.dygraph.Layer):
     def forward(self, inputs):
         x = self.features(inputs)
 
-        x = fluid.layers.flatten(x, 1)
+        x = paddle.flatten(x, 1, -1)
         x = self.fc(x)
         return x
 
diff --git a/python/paddle/fluid/contrib/slim/tests/test_imperative_qat_lsq.py b/python/paddle/fluid/contrib/slim/tests/test_imperative_qat_lsq.py
index fbf1f4e7f5180fc5c271d376d58ab80f176ef0a2..1b54a5b55b1083c2fcb8d11260ad2b0ec01fe313 100644
--- a/python/paddle/fluid/contrib/slim/tests/test_imperative_qat_lsq.py
+++ b/python/paddle/fluid/contrib/slim/tests/test_imperative_qat_lsq.py
@@ -117,7 +117,7 @@ class ImperativeLenet(fluid.dygraph.Layer):
 
     def forward(self, inputs):
         x = self.features(inputs)
-        x = fluid.layers.flatten(x, 1)
+        x = paddle.flatten(x, 1, -1)
         x = self.fc(x)
         return x
 
diff --git a/python/paddle/fluid/layers/detection.py b/python/paddle/fluid/layers/detection.py
index 1519b04243e6ba0c848b221bfae06bd235cb0bc6..543f63b639ea7ca2fb611967cd4c2361a5e1fb9b 100644
--- a/python/paddle/fluid/layers/detection.py
+++ b/python/paddle/fluid/layers/detection.py
@@ -1739,7 +1739,9 @@ def ssd_loss(
     conf_shape = nn.shape(confidence)
 
     def __reshape_to_2d(var):
-        return nn.flatten(x=var, axis=2)
+        out = paddle.flatten(var, 2, -1)
+        out = paddle.flatten(out, 0, 1)
+        return out
 
     # 1. Find matched bounding box by prior box.
     #   1.1 Compute IOU similarity between ground-truth boxes and prior boxes.
@@ -2335,8 +2337,15 @@ def multi_box_head(
     """
 
     def _reshape_with_axis_(input, axis=1):
-        out = nn.flatten(x=input, axis=axis)
-        return out
+        # Note : axis!=0 in current references to this func
+        # if axis == 0:
+        #     x = paddle.flatten(input, 0, -1)
+        #     x = paddle.unsqueeze(x, 0)
+        #     return x
+        # else:
+        x = paddle.flatten(input, axis, -1)
+        x = paddle.flatten(x, 0, axis - 1)
+        return x
 
     def _is_list_or_tuple_(data):
         return isinstance(data, list) or isinstance(data, tuple)
@@ -2445,7 +2454,7 @@ def multi_box_head(
         )
 
         mbox_loc = paddle.transpose(mbox_loc, perm=[0, 2, 3, 1])
-        mbox_loc_flatten = nn.flatten(mbox_loc, axis=1)
+        mbox_loc_flatten = paddle.flatten(mbox_loc, 1, -1)
         mbox_locs.append(mbox_loc_flatten)
 
         # get conf
@@ -2457,8 +2466,9 @@ def multi_box_head(
             padding=pad,
             stride=stride,
         )
+
         conf_loc = paddle.transpose(conf_loc, perm=[0, 2, 3, 1])
-        conf_loc_flatten = nn.flatten(conf_loc, axis=1)
+        conf_loc_flatten = paddle.flatten(conf_loc, 1, -1)
         mbox_confs.append(conf_loc_flatten)
 
     if len(box_results) == 1:
diff --git a/python/paddle/fluid/layers/nn.py b/python/paddle/fluid/layers/nn.py
index 821a974dd7863b697114fb5a9eee8c34b1d1286c..3e40a80216ac67e73cf6a29d82a388b2b5c013d4 100644
--- a/python/paddle/fluid/layers/nn.py
+++ b/python/paddle/fluid/layers/nn.py
@@ -109,7 +109,6 @@ __all__ = [
     'log',
     'crop_tensor',
     'prelu',
-    'flatten',
     'unique',
     'unique_with_counts',
     'elementwise_add',
@@ -6842,98 +6841,6 @@ def prelu(x, mode, param_attr=None, data_format="NCHW", name=None):
     return out
 
 
-def flatten(x, axis=1, name=None):
-    r"""
-    **Flatten op**
-
-    Flatten the input tensor into a 2D matrix.
-
-    For Example:
-
-    .. code-block:: text
-
-        Case 1:
-
-          Given
-            X.shape = (3, 100, 100, 4)
-
-          and
-            axis = 2
-
-          We get:
-            Out.shape = (3 * 100, 4 * 100)
-
-        Case 2:
-
-          Given
-            X.shape = (3, 100, 100, 4)
-
-          and
-            axis = 0
-
-          We get:
-            Out.shape = (1, 3 * 100 * 100 * 4)
-
-    Args:
-        x (Variable): A tensor of rank >= axis. A tensor with type float32,
-                      float64, int8, int32, int64, uint8.
-        axis (int): Indicate up to which input dimensions (exclusive) should
-                    be flattened to the outer dimension of the output.
-                    The value for axis must be in the range [0, R], where R
-                    is the rank of the input tensor. Default: 1.
-        name(str, Optional): For details, please refer to :ref:`api_guide_Name`.
-                        Generally, no setting is required. Default: None.
-
-    Returns:
-        Variable: A 2D tensor with the contents of the input tensor, with input \
-                  dimensions up to axis flattened to the outer dimension of \
-                  the output and remaining input dimensions flattened into the \
-                  inner dimension of the output. A Tensor with type same as input x.
-
-    Raises:
-        ValueError: If x is not a variable.
-        ValueError: If axis is not in range [0, rank(x)].
-
-    Examples:
-
-        .. code-block:: python
-
-            import paddle
-            import paddle.fluid as fluid
-            paddle.enable_static()
-            x = fluid.data(name="x", shape=[4, 4, 3], dtype="float32")
-            # x shape is [4, 4, 3]
-            out = fluid.layers.flatten(x=x, axis=2)
-            # out shape is [16, 3]
-    """
-    check_variable_and_dtype(
-        x,
-        'x',
-        ['float32', 'float64', 'int8', 'int32', 'int64', 'uint8'],
-        'flatten',
-    )
-    if _non_static_mode():
-        return _legacy_C_ops.flatten2(x, 'axis', axis)[0]
-
-    helper = LayerHelper('flatten', **locals())
-
-    if not (isinstance(x, Variable)):
-        raise ValueError("The input x should be a Variable")
-
-    if not (isinstance(axis, int)) or axis > len(x.shape) or axis < 0:
-        raise ValueError("The axis should be a int, and in range [0, rank(x)]")
-
-    out = helper.create_variable_for_type_inference(x.dtype)
-    x_shape = helper.create_variable_for_type_inference(x.dtype)
-    helper.append_op(
-        type='flatten2',
-        inputs={"X": x},
-        outputs={'Out': out, 'XShape': x_shape},
-        attrs={"axis": axis},
-    )
-    return out
-
-
 from paddle.fluid.framework import convert_np_dtype_to_dtype_
 
 
diff --git a/python/paddle/fluid/tests/unittests/ipu/test_dy2static_fp16_ipu.py b/python/paddle/fluid/tests/unittests/ipu/test_dy2static_fp16_ipu.py
index d5049979c0f1040459f4efcbf85d33b4909d4ee6..8a13e5abb552379e418ed5cf46b3fba93eed53e9 100644
--- a/python/paddle/fluid/tests/unittests/ipu/test_dy2static_fp16_ipu.py
+++ b/python/paddle/fluid/tests/unittests/ipu/test_dy2static_fp16_ipu.py
@@ -31,7 +31,7 @@ class SimpleLayer(paddle.nn.Layer):
 
     def forward(self, x, target=None):
         x = self.conv(x)
-        x = paddle.fluid.layers.flatten(x, axis=1)
+        x = paddle.flatten(x, 1, -1)
         if target is not None:
             x = paddle.fluid.layers.softmax(x)
             loss = paddle.fluid.layers.cross_entropy(x, target)
diff --git a/python/paddle/fluid/tests/unittests/ipu/test_dy2static_ipu.py b/python/paddle/fluid/tests/unittests/ipu/test_dy2static_ipu.py
index 4e16b0efdf822ea8e3e20245f39bcba985055d6e..fd1c762f2095caa9316de4b8d6999d9cf8023602 100644
--- a/python/paddle/fluid/tests/unittests/ipu/test_dy2static_ipu.py
+++ b/python/paddle/fluid/tests/unittests/ipu/test_dy2static_ipu.py
@@ -46,7 +46,7 @@ class SimpleLayer(paddle.nn.Layer):
     @to_static()
     def forward(self, x, target=None):
         x = self.conv(x)
-        x = paddle.fluid.layers.flatten(x, axis=1)
+        x = paddle.flatten(x, 1, -1)
         if target is not None:
             if self.use_softmax:
                 x = paddle.fluid.layers.softmax(x)
diff --git a/python/paddle/fluid/tests/unittests/ipu/test_flatten_op_ipu.py b/python/paddle/fluid/tests/unittests/ipu/test_flatten_op_ipu.py
index 58108b262e0ffc2e81cb7797830c7f1ca19d4454..fe89e81854a1bae385c5ff0f0c2632885788b8fa 100644
--- a/python/paddle/fluid/tests/unittests/ipu/test_flatten_op_ipu.py
+++ b/python/paddle/fluid/tests/unittests/ipu/test_flatten_op_ipu.py
@@ -47,7 +47,12 @@ class TestBase(IPUOpTest):
         x = paddle.static.data(
             name=self.feed_list[0], shape=self.feed_shape[0], dtype='float32'
         )
-        out = paddle.fluid.layers.flatten(x=x, **self.attrs)
+        if self.attrs['axis'] == 0:
+            x = paddle.flatten(x, 0, -1)
+            out = paddle.unsqueeze(x, 0)
+        else:
+            x = paddle.flatten(x, self.attrs['axis'], -1)
+            out = paddle.flatten(x, 0, self.attrs['axis'] - 1)
         self.fetch_list = [out.name]
 
     def run_model(self, exec_mode):
diff --git a/python/paddle/fluid/tests/unittests/ipu/test_modelruntime_ipu.py b/python/paddle/fluid/tests/unittests/ipu/test_modelruntime_ipu.py
index 383b2632aa0c5aa3af382cd2ecabf1f5eddbb7ee..9fda7f780e86d4bad2505025b03f82e42009edff 100644
--- a/python/paddle/fluid/tests/unittests/ipu/test_modelruntime_ipu.py
+++ b/python/paddle/fluid/tests/unittests/ipu/test_modelruntime_ipu.py
@@ -30,7 +30,7 @@ class SimpleLayer(paddle.nn.Layer):
 
     def forward(self, x, target=None):
         x = self.conv(x)
-        x = paddle.fluid.layers.flatten(x, axis=1)
+        x = paddle.flatten(x, 1, -1)
         if target is not None:
             x = paddle.fluid.layers.softmax(x)
             loss = paddle.fluid.layers.cross_entropy(x, target)
diff --git a/python/paddle/fluid/tests/unittests/ipu/test_print_op_ipu.py b/python/paddle/fluid/tests/unittests/ipu/test_print_op_ipu.py
index 6a0d384fa9c71e44a810efa36fe7a088e9d90c96..ccf0a38bbf88c74636a10f932ca1f5c09b69731d 100644
--- a/python/paddle/fluid/tests/unittests/ipu/test_print_op_ipu.py
+++ b/python/paddle/fluid/tests/unittests/ipu/test_print_op_ipu.py
@@ -117,7 +117,7 @@ class SimpleLayer(paddle.nn.Layer):
     def forward(self, x, target=None):
         x = self.conv(x)
         print(x)
-        x = paddle.fluid.layers.flatten(x, axis=1)
+        x = paddle.flatten(x, 1, -1)
         if target is not None:
             x = paddle.fluid.layers.softmax(x)
             loss = paddle.fluid.layers.cross_entropy(x, target)
diff --git a/python/paddle/fluid/tests/unittests/ir/inference/test_trt_flatten_op.py b/python/paddle/fluid/tests/unittests/ir/inference/test_trt_flatten_op.py
index 8beb2000c688f2258e96d5f02ba55bf951300e1b..4ed648ed9c8062384cdbf11ad90142d12b752cbb 100644
--- a/python/paddle/fluid/tests/unittests/ir/inference/test_trt_flatten_op.py
+++ b/python/paddle/fluid/tests/unittests/ir/inference/test_trt_flatten_op.py
@@ -17,6 +17,7 @@ import unittest
 import numpy as np
 from inference_pass_test import InferencePassTest
 
+import paddle
 import paddle.fluid as fluid
 import paddle.fluid.core as core
 from paddle.fluid.core import AnalysisConfig, PassVersionChecker
@@ -40,7 +41,7 @@ class TRTFlattenTest(InferencePassTest):
         self.fetch_list = [out]
 
     def append_flatten(self, data):
-        return fluid.layers.flatten(data, axis=1)
+        return paddle.flatten(data, 1, -1)
 
     def test_check_output(self):
         if core.is_compiled_with_cuda():
@@ -75,7 +76,7 @@ class TRTFlattenDynamicTest(InferencePassTest):
         self.fetch_list = [out]
 
     def append_flatten(self, data):
-        return fluid.layers.flatten(data, axis=1)
+        return paddle.flatten(data, 1, -1)
 
     def test_check_output(self):
         if core.is_compiled_with_cuda():
diff --git a/python/paddle/fluid/tests/unittests/ir/inference/test_trt_transpose_flatten_concat_fuse_pass.py b/python/paddle/fluid/tests/unittests/ir/inference/test_trt_transpose_flatten_concat_fuse_pass.py
index 8e0fa3a0eba2d7da4b4d81f177795630d065d8cd..192274ef34106e4958db0f1e61619df473e7df6e 100644
--- a/python/paddle/fluid/tests/unittests/ir/inference/test_trt_transpose_flatten_concat_fuse_pass.py
+++ b/python/paddle/fluid/tests/unittests/ir/inference/test_trt_transpose_flatten_concat_fuse_pass.py
@@ -32,10 +32,12 @@ class TransposeFlattenConcatFusePassTRTTest(InferencePassTest):
             data2 = fluid.data(
                 name="data2", shape=[8, 32, 128], dtype="float32"
             )
+
             trans1 = paddle.transpose(data1, perm=[0, 2, 1])
             trans2 = paddle.transpose(data2, perm=[0, 2, 1])
-            flatt1 = fluid.layers.flatten(trans1)
-            flatt2 = fluid.layers.flatten(trans2)
+            flatt1 = paddle.flatten(trans1, 1, -1)
+            flatt2 = paddle.flatten(trans2, 1, -1)
+
             concat_out = fluid.layers.concat([flatt1, flatt2], axis=1)
             # There is no parameters for above structure.
             # Hence, append a batch_norm to avoid failure caused by load_combined.
diff --git a/python/paddle/fluid/tests/unittests/mlu/test_flatten2_op_mlu.py b/python/paddle/fluid/tests/unittests/mlu/test_flatten2_op_mlu.py
index ace458ff42c0576d73ec02b588665a4864df44b9..c6ae69b2d57e0c031fad82e9316d99b3d019f023 100644
--- a/python/paddle/fluid/tests/unittests/mlu/test_flatten2_op_mlu.py
+++ b/python/paddle/fluid/tests/unittests/mlu/test_flatten2_op_mlu.py
@@ -75,41 +75,5 @@ class TestFlattenOpSixDims(TestFlattenOp):
         self.new_shape = (36, 16)
 
 
-class TestStaticFlattenInferShapePythonAPI(unittest.TestCase):
-    def execute_api(self, x, axis=1):
-        return fluid.layers.flatten(x, axis=axis)
-
-    def test_static_api(self):
-        paddle.enable_static()
-        main_prog = paddle.static.Program()
-        with paddle.static.program_guard(main_prog, paddle.static.Program()):
-            x = paddle.static.data(
-                name="x", shape=[-1, 3, -1, -1], dtype='float32'
-            )
-            out = self.execute_api(x, axis=2)
-        self.assertTrue((-1, -1) == out.shape)
-
-
-class TestFlatten2OpError(unittest.TestCase):
-    def test_errors(self):
-        with fluid.program_guard(fluid.Program(), fluid.Program()):
-            input_data = np.random.random((3, 2, 4, 5)).astype("float64")
-
-        def test_Variable():
-            # the input type must be Variable
-            fluid.layers.flatten(input_data, axis=1)
-
-        self.assertRaises(TypeError, test_Variable)
-
-        def test_type():
-            # dtype must be float32, float64, int8, int32, int64, uint8.
-            x2 = fluid.layers.data(
-                name='x2', shape=[3, 2, 4, 5], dtype='float16'
-            )
-            fluid.layers.flatten(x2, axis=1)
-
-        self.assertRaises(TypeError, test_type)
-
-
 if __name__ == "__main__":
     unittest.main()
diff --git a/python/paddle/fluid/tests/unittests/test_flatten2_op.py b/python/paddle/fluid/tests/unittests/test_flatten2_op.py
index fe5aad118af46e5792a110d7ce5c4d7b94f10d84..d67b873084d5a02b8237dc22a641639a3dd7776d 100644
--- a/python/paddle/fluid/tests/unittests/test_flatten2_op.py
+++ b/python/paddle/fluid/tests/unittests/test_flatten2_op.py
@@ -14,8 +14,6 @@
 
 import unittest
 import numpy as np
-import paddle.fluid as fluid
-import paddle
 from op_test import OpTest
 
 
@@ -68,41 +66,5 @@ class TestFlattenOpSixDims(TestFlattenOp):
         self.new_shape = (36, 16)
 
 
-class TestStaticFlattenInferShapePythonAPI(unittest.TestCase):
-    def execute_api(self, x, axis=1):
-        return fluid.layers.flatten(x, axis=axis)
-
-    def test_static_api(self):
-        paddle.enable_static()
-        main_prog = paddle.static.Program()
-        with paddle.static.program_guard(main_prog, paddle.static.Program()):
-            x = paddle.static.data(
-                name="x", shape=[-1, 3, -1, -1], dtype='float32'
-            )
-            out = self.execute_api(x, axis=2)
-        self.assertTrue((-1, -1) == out.shape)
-
-
-class TestFlatten2OpError(unittest.TestCase):
-    def test_errors(self):
-        with fluid.program_guard(fluid.Program(), fluid.Program()):
-            input_data = np.random.random((3, 2, 4, 5)).astype("float64")
-
-        def test_Variable():
-            # the input type must be Variable
-            fluid.layers.flatten(input_data, axis=1)
-
-        self.assertRaises(TypeError, test_Variable)
-
-        def test_type():
-            # dtype must be float32, float64, int8, int32, int64, uint8.
-            x2 = fluid.layers.data(
-                name='x2', shape=[3, 2, 4, 5], dtype='float16'
-            )
-            fluid.layers.flatten(x2, axis=1)
-
-        self.assertRaises(TypeError, test_type)
-
-
 if __name__ == "__main__":
     unittest.main()
diff --git a/python/paddle/fluid/tests/unittests/test_imperative_layer_apply.py b/python/paddle/fluid/tests/unittests/test_imperative_layer_apply.py
index 88b91e76b112a0ac7049693b8c8abbb3554ce2b0..156cac0b3a3243d42f13aab2e6fbaf34fbe38004 100644
--- a/python/paddle/fluid/tests/unittests/test_imperative_layer_apply.py
+++ b/python/paddle/fluid/tests/unittests/test_imperative_layer_apply.py
@@ -47,7 +47,7 @@ class LeNetDygraph(fluid.dygraph.Layer):
         x = self.features(inputs)
 
         if self.num_classes > 0:
-            x = fluid.layers.flatten(x, 1)
+            x = paddle.flatten(x, 1, -1)
             x = self.fc(x)
         return x
 
diff --git a/python/paddle/fluid/tests/unittests/test_layers.py b/python/paddle/fluid/tests/unittests/test_layers.py
index 22b4e22061b9be8133be498c0f23ced930e1411c..aff461f8786c9634b2cddd5a064a5a53f5d3bd24 100644
--- a/python/paddle/fluid/tests/unittests/test_layers.py
+++ b/python/paddle/fluid/tests/unittests/test_layers.py
@@ -4129,7 +4129,7 @@ class TestBook(LayerTest):
                 shape=[4, 4, 3],
                 dtype="float32",
             )
-            out = layers.flatten(x, axis=1, name="flatten")
+            out = paddle.flatten(x, 1, -1, name="flatten")
             return out
 
     def test_linspace(self):
diff --git a/python/paddle/tests/test_model.py b/python/paddle/tests/test_model.py
index 91ffb3377d1082dfc8dd80f15f50beb3246ca596..58c88e56482fa3188b56f70c97b757913cc8b72b 100644
--- a/python/paddle/tests/test_model.py
+++ b/python/paddle/tests/test_model.py
@@ -61,7 +61,7 @@ class LeNetDygraph(paddle.nn.Layer):
         x = self.features(inputs)
 
         if self.num_classes > 0:
-            x = fluid.layers.flatten(x, 1)
+            x = paddle.flatten(x, 1, -1)
             x = self.fc(x)
         return x