diff --git a/paddle/fluid/API.spec b/paddle/fluid/API.spec
index 82dd3d3e65de78ce0b416e0604a500477529496c..5c92a29bb33436c6e8d2be5c2365f04e0a39926b 100644
--- a/paddle/fluid/API.spec
+++ b/paddle/fluid/API.spec
@@ -236,6 +236,7 @@ paddle.fluid.layers.pixel_shuffle (ArgSpec(args=['x', 'upscale_factor'], varargs
 paddle.fluid.layers.fsp_matrix (ArgSpec(args=['x', 'y'], varargs=None, keywords=None, defaults=None), ('document', 'b76ccca3735bea4a58a0dbf0d77c5393'))
 paddle.fluid.layers.continuous_value_model (ArgSpec(args=['input', 'cvm', 'use_cvm'], varargs=None, keywords=None, defaults=(True,)), ('document', 'a07a44c2bacdcd09c1f5f35a96a0514e'))
 paddle.fluid.layers.where (ArgSpec(args=['condition'], varargs=None, keywords=None, defaults=None), ('document', '3126e3039e752ce26077f1efaca355c6'))
+paddle.fluid.layers.sign (ArgSpec(args=['x'], varargs=None, keywords=None, defaults=None), ('document', 'ccf6bb7912afd2818d24bc45461e807a'))
 paddle.fluid.layers.data (ArgSpec(args=['name', 'shape', 'append_batch_size', 'dtype', 'lod_level', 'type', 'stop_gradient'], varargs=None, keywords=None, defaults=(True, 'float32', 0, VarType.LOD_TENSOR, True)), ('document', 'adf285346e23316097f7789b572491e9'))
 paddle.fluid.layers.open_files (ArgSpec(args=['filenames', 'shapes', 'lod_levels', 'dtypes', 'thread_num', 'buffer_size', 'pass_num', 'is_test'], varargs=None, keywords=None, defaults=(None, None, 1, None)), ('document', 'cf12066a3139026119f97f9d4381a1bd'))
 paddle.fluid.layers.read_file (ArgSpec(args=['reader'], varargs=None, keywords=None, defaults=None), ('document', 'b0a1c2fc51c27a106da28f3308c41f5e'))
diff --git a/python/paddle/fluid/layers/nn.py b/python/paddle/fluid/layers/nn.py
index ea8ddc91eca938edd607aee0556c2212bba2f584..bab76fdfcb1f2fa80cca6678d3f00fd7efcbede7 100644
--- a/python/paddle/fluid/layers/nn.py
+++ b/python/paddle/fluid/layers/nn.py
@@ -201,6 +201,7 @@ __all__ = [
     'fsp_matrix',
     'continuous_value_model',
     'where',
+    'sign',
 ]
 
 kIgnoreIndex = -100
@@ -11396,3 +11397,34 @@ def where(condition):
     helper.append_op(
         type='where', inputs={'Condition': condition}, outputs={'Out': [out]})
     return out
+
+
+def sign(x):
+    """
+    **sign**
+
+    This function returns sign of every element in `x`: 1 for positive, -1 for negative and 0 for zero.
+
+    Args:
+        x(Variable|numpy.ndarray): The input tensor.
+
+    Returns:
+        Variable: The output sign tensor with identical shape and dtype to `x`.
+
+    Examples:
+        .. code-block:: python
+
+          # [1, 0, -1]
+          data = fluid.layers.sign(np.array([3, 0, -2])) 
+    """
+
+    helper = LayerHelper("sign", **locals())
+
+    if not isinstance(x, Variable):
+        x = assign(x)
+
+    out = helper.create_variable_for_type_inference(dtype=x.dtype)
+
+    helper.append_op(type='sign', inputs={'X': [x]}, outputs={'Out': [out]})
+
+    return out