diff --git a/python/paddle/distribution/normal.py b/python/paddle/distribution/normal.py
index 3eb3fd2d59009c48cf28a64fd6129ea5e0acaf06..7eb9fb597d3a2763a32e93925a7d34ce86f9f499 100644
--- a/python/paddle/distribution/normal.py
+++ b/python/paddle/distribution/normal.py
@@ -180,7 +180,7 @@ class Normal(distribution.Distribution):
                 self.loc + self.scale, batch_shape + shape, self.dtype, 0.0
             )
             zero_tmp_reshape = paddle.reshape(zero_tmp, output_shape)
-            zero_tmp_shape = nn.shape(zero_tmp_reshape)
+            zero_tmp_shape = paddle.shape(zero_tmp_reshape)
             normal_random_tmp = nn.gaussian_random(
                 zero_tmp_shape, mean=0.0, std=1.0, seed=seed, dtype=self.dtype
             )
diff --git a/python/paddle/fluid/contrib/layers/rnn_impl.py b/python/paddle/fluid/contrib/layers/rnn_impl.py
index b28cac87950c9f75cffd4bdfa21113bc6b1004ed..4e23057fc4680d2e8147e7acaa8727a5310d701a 100644
--- a/python/paddle/fluid/contrib/layers/rnn_impl.py
+++ b/python/paddle/fluid/contrib/layers/rnn_impl.py
@@ -330,7 +330,7 @@ def basic_gru(
 
     mask = None
     if sequence_length:
-        max_seq_len = layers.shape(input)[0]
+        max_seq_len = paddle.shape(input)[0]
         mask = layers.sequence_mask(
             sequence_length, maxlen=max_seq_len, dtype='float32'
         )
@@ -614,7 +614,7 @@ def basic_lstm(
 
     mask = None
     if sequence_length:
-        max_seq_len = layers.shape(input)[0]
+        max_seq_len = paddle.shape(input)[0]
         mask = layers.sequence_mask(
             sequence_length, maxlen=max_seq_len, dtype='float32'
         )
diff --git a/python/paddle/fluid/layers/detection.py b/python/paddle/fluid/layers/detection.py
index d490b0457d98cc651db9afec80729aa315227705..dddd2fa386d2b8fb9c7681ed75f49f6a73d2d667 100644
--- a/python/paddle/fluid/layers/detection.py
+++ b/python/paddle/fluid/layers/detection.py
@@ -1588,7 +1588,7 @@ def ssd_loss(
         raise ValueError("Only support mining_type == max_negative now.")
 
     num, num_prior, num_class = confidence.shape
-    conf_shape = nn.shape(confidence)
+    conf_shape = paddle.shape(confidence)
 
     def __reshape_to_2d(var):
         out = paddle.flatten(var, 2, -1)
diff --git a/python/paddle/fluid/layers/nn.py b/python/paddle/fluid/layers/nn.py
index 4dab44ebe54feb8ba695cc21e53289126906fdbd..fda4c24e510685f1669c1f0a51a737574b85ceb9 100644
--- a/python/paddle/fluid/layers/nn.py
+++ b/python/paddle/fluid/layers/nn.py
@@ -97,7 +97,6 @@ __all__ = [
     'elementwise_mul',
     'gaussian_random',
     'sampling_id',
-    'shape',
     'clip',
     'clip_by_norm',
     'mean',
@@ -5010,95 +5009,6 @@ def sampling_id(x, min=0.0, max=1.0, seed=0, dtype='float32'):
     return out
 
 
-def shape(input):
-    """
-    :alias_main: paddle.shape
-        :alias: paddle.shape,paddle.tensor.shape,paddle.tensor.attribute.shape
-        :old_api: paddle.fluid.layers.shape
-
-    **Shape Layer**
-
-    Get the shape of the input.
-
-    .. code-block:: text
-
-        Case1:
-            Given N-D Tensor:
-                input = [ [1, 2, 3, 4], [5, 6, 7, 8] ]
-
-            Then:
-                input.shape = [2, 4]
-
-        Case2:
-            Given SelectedRows:
-                input.rows = [0, 4, 19]
-                input.height = 20
-                input.value = [ [1, 2], [3, 4], [5, 6] ]  # inner tensor
-            Then:
-                input.shape = [3, 2]
-
-    Args:
-        input (Variable): The input can be N-D Tensor or SelectedRows with data type bool, float16, float32, float64, int32, int64.
-                          If input variable is type of SelectedRows, returns the shape of it's inner tensor.
-
-    Returns:
-        Variable (Tensor): The shape of the input variable.
-
-    Examples:
-        .. code-block:: python
-
-            import paddle.fluid as fluid
-            import numpy as np
-            import paddle
-            paddle.enable_static()
-
-            inputs = fluid.data(name="x", shape=[3, 100, 100], dtype="float32")
-            output = fluid.layers.shape(inputs)
-
-            exe = fluid.Executor(fluid.CPUPlace())
-            exe.run(fluid.default_startup_program())
-
-            img = np.ones((3, 100, 100)).astype(np.float32)
-
-            res = exe.run(fluid.default_main_program(), feed={'x':img}, fetch_list=[output])
-            print(res) # [array([  3, 100, 100], dtype=int32)]
-    """
-    if in_dygraph_mode():
-        out = _C_ops.shape(input)
-        out.stop_gradient = True
-        return out
-    if _in_legacy_dygraph():
-        out = _legacy_C_ops.shape(input)
-        out.stop_gradient = True
-        return out
-
-    check_variable_and_dtype(
-        input,
-        'input',
-        [
-            'bool',
-            'float16',
-            'float32',
-            'float64',
-            'int32',
-            'int64',
-            'complex64',
-            'complex128',
-        ],
-        'shape',
-    )
-    helper = LayerHelper('shape', **locals())
-    out = helper.create_variable_for_type_inference(dtype='int32')
-    helper.append_op(
-        type='shape',
-        inputs={'Input': input},
-        outputs={'Out': out},
-        stop_gradient=True,
-    )
-
-    return out
-
-
 def _elementwise_op(helper):
     op_type = helper.layer_type
     x = helper.kwargs.get('x', None)
diff --git a/python/paddle/fluid/layers/rnn.py b/python/paddle/fluid/layers/rnn.py
index 60ac537ffc6d408589381b1bd2785ecd2cd7d0d0..8b5721438d2e5351752ebf78551e27527fea247c 100644
--- a/python/paddle/fluid/layers/rnn.py
+++ b/python/paddle/fluid/layers/rnn.py
@@ -673,7 +673,7 @@ def _rnn_static_graph(
         inputs = map_structure(_transpose_batch_time, inputs)
 
     if sequence_length:
-        max_seq_len = nn.shape(flatten(inputs)[0])[0]
+        max_seq_len = paddle.shape(flatten(inputs)[0])[0]
         mask = sequence_lod.sequence_mask(
             sequence_length,
             maxlen=max_seq_len,
@@ -1215,7 +1215,7 @@ class BeamSearchDecoder(Decoder):
         """
         self.kinf = 1e9
         state = flatten(initial_cell_states)[0]
-        self.batch_size = nn.shape(state)[0]
+        self.batch_size = paddle.shape(state)[0]
 
         self.start_token_tensor = tensor.fill_constant(
             shape=[1], dtype="int64", value=self.start_token
diff --git a/python/paddle/fluid/tests/unittests/dygraph_to_static/ifelse_simple_func.py b/python/paddle/fluid/tests/unittests/dygraph_to_static/ifelse_simple_func.py
index 2fa012559cc77fef07e754cc03f3720ffe1fe162..985d091d6b9c46d8b5307471123e99f8734566c0 100644
--- a/python/paddle/fluid/tests/unittests/dygraph_to_static/ifelse_simple_func.py
+++ b/python/paddle/fluid/tests/unittests/dygraph_to_static/ifelse_simple_func.py
@@ -151,7 +151,7 @@ def nested_if_else(x_v):
         #  `x_v.shape[0]` is not Tensor, and `batch_size` is the return value of `true_fn` after transformed.
         # col = -1
         # batch_size = x_v.shape[0]
-        batch_size = fluid.layers.shape(x_v)[0]
+        batch_size = paddle.shape(x_v)[0]
 
     # if tensor.shape is [1], now support to compare with numpy.
     if paddle.mean(x_v).numpy() < 0:
@@ -180,7 +180,7 @@ def nested_if_else_2(x):
         z = y
     x_shape_0 = x.shape[0]
     if x_shape_0 < 1:
-        if fluid.layers.shape(y).numpy()[0] < 1:
+        if paddle.shape(y).numpy()[0] < 1:
             res = fluid.layers.fill_constant(
                 value=2, shape=x.shape, dtype="int32"
             )
@@ -212,7 +212,7 @@ def nested_if_else_3(x):
         else:
             out = x - 1
     else:
-        y_shape = fluid.layers.shape(y)
+        y_shape = paddle.shape(y)
         if y_shape.numpy()[0] < 1:
             res = fluid.layers.fill_constant(
                 value=2, shape=x.shape, dtype="int32"
@@ -290,7 +290,7 @@ class NetWithControlFlowIf(fluid.dygraph.Layer):
 
 
 def if_with_and_or(x_v, label=None):
-    batch_size = fluid.layers.shape(x_v)
+    batch_size = paddle.shape(x_v)
     if (
         x_v is not None
         and (paddle.mean(x_v).numpy()[0] > 0 or label is not None)
@@ -308,7 +308,7 @@ def if_with_and_or(x_v, label=None):
 
 
 def if_with_and_or_1(x, y=None):
-    batch_size = fluid.layers.shape(x)
+    batch_size = paddle.shape(x)
     if batch_size[0] > 1 and y is not None:
         x = x + 1
     if y is not None or batch_size[0] > 1:
@@ -317,7 +317,7 @@ def if_with_and_or_1(x, y=None):
 
 
 def if_with_and_or_2(x, y=None):
-    batch_size = fluid.layers.shape(x)
+    batch_size = paddle.shape(x)
     if x is not None and batch_size[0] > 1 and y is not None:
         x = x + 1
     if batch_size[0] > 1 or y is not None or x is not None:
@@ -326,7 +326,7 @@ def if_with_and_or_2(x, y=None):
 
 
 def if_with_and_or_3(x, y=None):
-    batch_size = fluid.layers.shape(x)
+    batch_size = paddle.shape(x)
     mean_res = paddle.mean(x)
     if (
         x is not None
@@ -341,7 +341,7 @@ def if_with_and_or_3(x, y=None):
 
 
 def if_with_and_or_4(x, y=None):
-    batch_size = fluid.layers.shape(x)
+    batch_size = paddle.shape(x)
     mean_res = paddle.mean(x)
     if (x is not None and batch_size[0] > 1) or (
         y is not None and mean_res.numpy()[0] > 0
@@ -361,7 +361,7 @@ def if_with_class_var(x, y=None):
             self.b = 2
 
     foo = Foo()
-    batch_size = fluid.layers.shape(x)
+    batch_size = paddle.shape(x)
     mean_res = paddle.mean(x)
 
     if batch_size[0] > foo.a:
diff --git a/python/paddle/fluid/tests/unittests/dygraph_to_static/seq2seq_dygraph_model.py b/python/paddle/fluid/tests/unittests/dygraph_to_static/seq2seq_dygraph_model.py
index d364b8a1a5d602f3f97acfa5f52c71943a582992..5c04aecd1a87fcbf45df20fadc41fe5e62b6c408 100644
--- a/python/paddle/fluid/tests/unittests/dygraph_to_static/seq2seq_dygraph_model.py
+++ b/python/paddle/fluid/tests/unittests/dygraph_to_static/seq2seq_dygraph_model.py
@@ -297,8 +297,8 @@ class BaseModel(fluid.dygraph.Layer):
         loss = paddle.nn.functional.softmax_with_cross_entropy(
             logits=dec_output, label=label, soft_label=False
         )
-        loss = paddle.squeeze(loss, axis=[2])
-        max_tar_seq_len = fluid.layers.shape(tar)[1]
+        loss = paddle.squeeze(loss, axes=[2])
+        max_tar_seq_len = paddle.shape(tar)[1]
         tar_mask = fluid.layers.sequence_mask(
             tar_sequence_length, maxlen=max_tar_seq_len, dtype='float32'
         )
@@ -833,8 +833,8 @@ class AttentionModel(fluid.dygraph.Layer):
         loss = paddle.nn.functional.softmax_with_cross_entropy(
             logits=dec_output, label=label, soft_label=False
         )
-        loss = paddle.squeeze(loss, axis=[2])
-        max_tar_seq_len = fluid.layers.shape(tar)[1]
+        loss = paddle.squeeze(loss, axes=[2])
+        max_tar_seq_len = paddle.shape(tar)[1]
         tar_mask = fluid.layers.sequence_mask(
             tar_sequence_length, maxlen=max_tar_seq_len, dtype='float32'
         )
diff --git a/python/paddle/fluid/tests/unittests/dygraph_to_static/simnet_dygraph_model.py b/python/paddle/fluid/tests/unittests/dygraph_to_static/simnet_dygraph_model.py
index 8c3d62feacc62c31d86f3752f34b4c2240af79fe..d16f07d9a2e343c258706050ba5a9199835d5777 100644
--- a/python/paddle/fluid/tests/unittests/dygraph_to_static/simnet_dygraph_model.py
+++ b/python/paddle/fluid/tests/unittests/dygraph_to_static/simnet_dygraph_model.py
@@ -210,7 +210,7 @@ class ConstantLayer:
         operation
         """
         shape = list(shape)
-        input_shape = fluid.layers.shape(input)
+        input_shape = paddle.shape(input)
         shape[0] = input_shape[0]
         constant = fluid.layers.fill_constant(shape, dtype, value)
         return constant
diff --git a/python/paddle/fluid/tests/unittests/dygraph_to_static/test_bmn.py b/python/paddle/fluid/tests/unittests/dygraph_to_static/test_bmn.py
index 0cb3e333045f5807481af5825f42b3737a1f1f14..b7461b21aa612eeb730188f1adc5de8bf728a325 100644
--- a/python/paddle/fluid/tests/unittests/dygraph_to_static/test_bmn.py
+++ b/python/paddle/fluid/tests/unittests/dygraph_to_static/test_bmn.py
@@ -321,9 +321,7 @@ def bmn_loss_func(
             gt_label = paddle.reshape(x=gt_label, shape=[-1])
             gt_label.stop_gradient = True
             pmask = fluid.layers.cast(x=(gt_label > 0.5), dtype=DATATYPE)
-            num_entries = fluid.layers.cast(
-                fluid.layers.shape(pmask), dtype=DATATYPE
-            )
+            num_entries = fluid.layers.cast(paddle.shape(pmask), dtype=DATATYPE)
             num_positive = fluid.layers.cast(paddle.sum(pmask), dtype=DATATYPE)
             ratio = num_entries / num_positive
             coef_0 = 0.5 * ratio / (ratio - 1)
diff --git a/python/paddle/fluid/tests/unittests/dygraph_to_static/test_dict.py b/python/paddle/fluid/tests/unittests/dygraph_to_static/test_dict.py
index 742e828aa9acb4e18b3bf3f564cabaf7bc7bdc44..597580eedc7655ba30920413db63c3dc9b191ad7 100644
--- a/python/paddle/fluid/tests/unittests/dygraph_to_static/test_dict.py
+++ b/python/paddle/fluid/tests/unittests/dygraph_to_static/test_dict.py
@@ -97,8 +97,8 @@ class MainNetWithDict(fluid.dygraph.Layer):
             ),
         }
         # TODO(Aurelius84): The following code will be converted into:
-        # max_len = layers.cond(layers.shape(input)[0] != max_len,
-        #                       lambda: layers.shape(input)[0], lambda: max_len)
+        # max_len = layers.cond(paddle.shape(input)[0] != max_len,
+        #                       lambda: paddle.shape(input)[0], lambda: max_len)
         # But max_len should be wrapped into tensor, which is not supported.
 
         # Comment out this line of code for now.
diff --git a/python/paddle/fluid/tests/unittests/dygraph_to_static/test_lac.py b/python/paddle/fluid/tests/unittests/dygraph_to_static/test_lac.py
index 54b97c9280a404f650bdc93773690daf4fd92b2c..5aff8c710ae7b5da9f1da0708cbc1578e8cd0024 100644
--- a/python/paddle/fluid/tests/unittests/dygraph_to_static/test_lac.py
+++ b/python/paddle/fluid/tests/unittests/dygraph_to_static/test_lac.py
@@ -79,7 +79,7 @@ class DynamicGRU(fluid.dygraph.Layer):
         res = []
         for i in range(inputs.shape[1]):
             if self.is_reverse:
-                j = fluid.layers.shape(inputs)[1] - 1 - i
+                j = paddle.shape(inputs)[1] - 1 - i
             else:
                 j = i
 
diff --git a/python/paddle/fluid/tests/unittests/dygraph_to_static/test_tensor_shape.py b/python/paddle/fluid/tests/unittests/dygraph_to_static/test_tensor_shape.py
index 4e29f2bf6b44a662bb0668c03cae53cc6e0a9d8d..1ad55d3fbaa412e2ae1831433fee4ad0acbe245e 100644
--- a/python/paddle/fluid/tests/unittests/dygraph_to_static/test_tensor_shape.py
+++ b/python/paddle/fluid/tests/unittests/dygraph_to_static/test_tensor_shape.py
@@ -38,7 +38,7 @@ def dyfunc_tensor_shape_2(x):
 def dyfunc_tensor_shape_3(x):
     # Transform y.shape but run y.shape actually because y is not Tensor
     x = fluid.dygraph.to_variable(x)
-    y = np.ones(5)
+    y = paddle.ones([1, 5])
     res = paddle.reshape(x, shape=y.shape)
     return res
 
@@ -97,7 +97,7 @@ def dyfunc_paddle_shape_api(x):
     a = paddle.shape(x)[0]
     # alias api will also not be converted.
     alias_old_api = paddle.fluid.layers
-    b = alias_old_api.shape(x)[1]
+    b = paddle.shape(x)[1]
     res = paddle.reshape(x, shape=(b, a))
     return res
 
@@ -199,7 +199,7 @@ def dyfunc_with_while_3(x):
 
 def dyfunc_with_while_4(x):
     x = paddle.to_tensor(x)
-    y = np.ones(5)
+    y = paddle.ones([1, 5])
     y_shape_0 = y.shape[0]
     i = 1
 
@@ -309,6 +309,11 @@ class TestTensorShapeBasic3(TestTensorShapeBasic):
     def init_test_func(self):
         self.dygraph_func = dyfunc_tensor_shape_3
 
+    def _set_expected_op_num(self):
+        self.expected_op_num = 3
+        self.expected_shape_op_num = 0
+        self.expected_slice_op_num = 0
+
 
 class TestTensorShapeBasic4(TestTensorShapeBasic):
     def init_test_func(self):
@@ -475,7 +480,7 @@ class TestTensorShapeInWhile4(TestTensorShapeBasic):
         self.dygraph_func = dyfunc_with_while_4
 
     def _set_expected_op_num(self):
-        self.expected_op_num = 4
+        self.expected_op_num = 1
         self.expected_shape_op_num = 0
         self.expected_slice_op_num = 0
 
diff --git a/python/paddle/fluid/tests/unittests/dygraph_to_static/yolov3.py b/python/paddle/fluid/tests/unittests/dygraph_to_static/yolov3.py
index 5f894744700f0bcb85adf83654625d7e7ad8f67b..5cf1f0f0f533a0c6efefc447b4bd9d87b00b5a0c 100644
--- a/python/paddle/fluid/tests/unittests/dygraph_to_static/yolov3.py
+++ b/python/paddle/fluid/tests/unittests/dygraph_to_static/yolov3.py
@@ -203,7 +203,7 @@ class Upsample(fluid.dygraph.Layer):
 
     def forward(self, inputs):
         # get dynamic upsample output shape
-        shape_nchw = fluid.layers.shape(inputs)
+        shape_nchw = paddle.shape(inputs)
         shape_hw = paddle.slice(shape_nchw, axes=[0], starts=[2], ends=[4])
         shape_hw.stop_gradient = True
         in_shape = fluid.layers.cast(shape_hw, dtype='int32')
diff --git a/python/paddle/fluid/tests/unittests/test_dynamic_rnn_stop_gradient.py b/python/paddle/fluid/tests/unittests/test_dynamic_rnn_stop_gradient.py
index 9774ea32e44f68e03278a5f0043351ea6f3436d4..3e3eefd5d278d3c2a80f8aaddd0b736685534300 100644
--- a/python/paddle/fluid/tests/unittests/test_dynamic_rnn_stop_gradient.py
+++ b/python/paddle/fluid/tests/unittests/test_dynamic_rnn_stop_gradient.py
@@ -40,7 +40,7 @@ def build_and_run_program(place, batch_size, beam_size, stop_gradient=False):
     while_op = layers.While(cond)
     scores = layers.array_write(x, step_idx)
     with while_op.block():
-        bs = layers.cast(layers.shape(x)[0], "int64")
+        bs = layers.cast(paddle.shape(x)[0], "int64")
         for _ in range(20):
             bs = layers.cast(bs, 'int64')
         bs.stop_gradient = stop_gradient
diff --git a/python/paddle/fluid/tests/unittests/test_layers.py b/python/paddle/fluid/tests/unittests/test_layers.py
index 25b6d0513d1c87527dd5f8617f6d3a79cdffb333..64e671c8ee9bd7706cddc8b949a3f32d873120f3 100644
--- a/python/paddle/fluid/tests/unittests/test_layers.py
+++ b/python/paddle/fluid/tests/unittests/test_layers.py
@@ -3307,7 +3307,7 @@ class TestBook(LayerTest):
             input = self._get_data(
                 name="input", shape=[3, 100, 100], dtype="float32"
             )
-            out = layers.shape(input)
+            out = paddle.shape(input)
             return out
 
     def make_pad2d(self):
diff --git a/python/paddle/fluid/tests/unittests/test_rnn_cell_api.py b/python/paddle/fluid/tests/unittests/test_rnn_cell_api.py
index 6b2383ed56933ebce02a0606604d532b99ac8eff..2302e6f0d241fddb3faf2d187a7122ba56a1e9f4 100644
--- a/python/paddle/fluid/tests/unittests/test_rnn_cell_api.py
+++ b/python/paddle/fluid/tests/unittests/test_rnn_cell_api.py
@@ -635,7 +635,7 @@ def def_seq2seq_model(
         logits=logits, label=label, soft_label=False
     )
     loss = layers.unsqueeze(loss, axes=[2])
-    max_tar_seq_len = layers.shape(target)[1]
+    max_tar_seq_len = paddle.shape(target)[1]
     tar_mask = layers.sequence_mask(
         target_length, maxlen=max_tar_seq_len, dtype="float32"
     )
diff --git a/python/paddle/fluid/tests/unittests/test_rnn_decode_api.py b/python/paddle/fluid/tests/unittests/test_rnn_decode_api.py
index 3b3539c4861f125de9dc14638dd9d0fa16f342ff..67657071db83f36f1bc747f54032e769e148bb72 100644
--- a/python/paddle/fluid/tests/unittests/test_rnn_decode_api.py
+++ b/python/paddle/fluid/tests/unittests/test_rnn_decode_api.py
@@ -248,7 +248,7 @@ class Seq2SeqModel:
             ),
         ]
         src_mask = layers.sequence_mask(
-            src_length, maxlen=layers.shape(src)[1], dtype="float32"
+            src_length, maxlen=paddle.shape(src)[1], dtype="float32"
         )
         encoder_padding_mask = (src_mask - 1.0) * 1e9
         encoder_padding_mask = layers.unsqueeze(encoder_padding_mask, [1])
@@ -400,7 +400,7 @@ class MLE:
 
     def learn(self, probs, label, weight=None, length=None):
         loss = layers.cross_entropy(input=probs, label=label, soft_label=False)
-        max_seq_len = layers.shape(probs)[1]
+        max_seq_len = paddle.shape(probs)[1]
         mask = layers.sequence_mask(length, maxlen=max_seq_len, dtype="float32")
         loss = loss * mask
         loss = paddle.mean(loss, axis=[0])
diff --git a/python/paddle/fluid/tests/unittests/test_static_shape_inferrence_for_shape_tensor.py b/python/paddle/fluid/tests/unittests/test_static_shape_inferrence_for_shape_tensor.py
index 17e7f69a3b49e4dbff77ab39f4a5d882b64ed3a6..6f70e553cc2bc54064c7aa74f3cc265928f70911 100644
--- a/python/paddle/fluid/tests/unittests/test_static_shape_inferrence_for_shape_tensor.py
+++ b/python/paddle/fluid/tests/unittests/test_static_shape_inferrence_for_shape_tensor.py
@@ -23,7 +23,7 @@ class StaticShapeInferrenceTest(unittest.TestCase):
         data = paddle.fluid.layers.data(
             name="x", shape=[-1, 2], dtype='float32'
         )
-        shape = paddle.fluid.layers.shape(data)  # shape should be [-1, 2]
+        shape = paddle.shape(data)  # shape should be [-1, 2]
         x = paddle.fluid.layers.uniform_random(shape)
         self.assertEqual(x.shape, data.shape)
         paddle.disable_static()
diff --git a/python/paddle/fluid/tests/unittests/test_while_loop_op.py b/python/paddle/fluid/tests/unittests/test_while_loop_op.py
index 8e733ef9208b3161481c0748c5b4e52fa7146e52..8c94834c9a28b122a37d0724485de059a1928dfa 100644
--- a/python/paddle/fluid/tests/unittests/test_while_loop_op.py
+++ b/python/paddle/fluid/tests/unittests/test_while_loop_op.py
@@ -595,7 +595,7 @@ class TestApiWhileLoopSliceInBody(unittest.TestCase):
         with program_guard(main_program, startup_program):
             x = fluid.layers.data(name='x', shape=[5], dtype='int32')
             z = fluid.layers.fill_constant([1], 'int32', 0)
-            x_shape = fluid.layers.shape(x)
+            x_shape = paddle.shape(x)
             i = fluid.layers.fill_constant([1], 'int32', 0)
             z, _ = fluid.layers.while_loop(cond, body, [z, i])
 
diff --git a/python/paddle/jit/dy2static/ast_transformer.py b/python/paddle/jit/dy2static/ast_transformer.py
index 2e244d6f341833c53111d83890972dec6bff9d11..826232e723f6078f4a7e86037a1268dff1b93fa1 100644
--- a/python/paddle/jit/dy2static/ast_transformer.py
+++ b/python/paddle/jit/dy2static/ast_transformer.py
@@ -127,7 +127,7 @@ class DygraphToStaticAst(BaseTransformer):
         transformers = [
             EarlyReturnTransformer,
             BasicApiTransformer,  # Basic Api
-            TensorShapeTransformer,  # Tensor.shape -> layers.shape(Tensor)
+            TensorShapeTransformer,  # Tensor.shape -> paddle.shape(Tensor)
             BreakContinueTransformer,  # break/continue in loops
             ReturnTransformer,  # return in functions
             LogicalTransformer,  # logical and/or/not
diff --git a/python/paddle/jit/dy2static/convert_operators.py b/python/paddle/jit/dy2static/convert_operators.py
index fa622b14094de0e841ddb81546c2ea768f0c2039..df20a5c4e0c98c1c284abd2e9cb83f2ea6fbc658 100644
--- a/python/paddle/jit/dy2static/convert_operators.py
+++ b/python/paddle/jit/dy2static/convert_operators.py
@@ -31,7 +31,6 @@ from paddle.fluid.layers import (
 from paddle.fluid.layers import (
     cast,
     control_flow,
-    nn,
 )
 from paddle.fluid.layers.control_flow import (
     cond,
@@ -524,7 +523,7 @@ def convert_len(var):
             # so we return a variable dynamically inferred from var.shape.
             if var.shape[0] > 0 and var.type == core.VarDesc.VarType.LOD_TENSOR:
                 return var.shape[0]
-            return nn.shape(var)[0]
+            return paddle.shape(var)[0]
         elif var.type == core.VarDesc.VarType.LOD_TENSOR_ARRAY:
             return paddle.tensor.array_length(var)
         else:
@@ -607,7 +606,7 @@ def convert_shape(x):
     if isinstance(x, Variable):
         values = list(x.shape)
         if has_negative(values):
-            shape_tensor = nn.shape(x)
+            shape_tensor = paddle.shape(x)
             for i, v in enumerate(values):
                 if v is None or v < 0:
                     values[i] = shape_tensor[i]
diff --git a/python/paddle/tensor/attribute.py b/python/paddle/tensor/attribute.py
index d6f08f676b8ae2b23dbb249293af34dffac37ec2..ac5f99a0b1c3c17dabe14d920e6d9e446c24c871 100644
--- a/python/paddle/tensor/attribute.py
+++ b/python/paddle/tensor/attribute.py
@@ -93,7 +93,7 @@ def shape(input):
             paddle.enable_static()
 
             inputs = fluid.data(name="x", shape=[3, 100, 100], dtype="float32")
-            output = fluid.layers.shape(inputs)
+            output = paddle.shape(inputs)
 
             exe = fluid.Executor(fluid.CPUPlace())
             exe.run(fluid.default_startup_program())