Merge pull request #7485 from lcy-seso/wrapper_for_elementwise_math_op

add wrapper for element-wise math operator.

doc/api/v2/fluid/layers.rst

@@ -358,3 +358,132 @@ reduce_min
 ..  autofunction:: paddle.v2.fluid.layers.reduce_min
     :noindex:
 
+logsigmoid
+----------
+..  autofunction:: paddle.v2.fluid.layers.logsigmoid
+    :noindex:
+
+exp
+---
+..  autofunction:: paddle.v2.fluid.layers.exp
+    :noindex:
+
+relu
+----
+..  autofunction:: paddle.v2.fluid.layers.relu
+    :noindex:
+
+tanh
+----
+..  autofunction:: paddle.v2.fluid.layers.tanh
+    :noindex:
+
+tanh_shrink
+-----------
+..  autofunction:: paddle.v2.fluid.layers.tanh_shrink
+    :noindex:
+
+softshrink
+----------
+..  autofunction:: paddle.v2.fluid.layers.softshrink
+    :noindex:
+
+sqrt
+----
+..  autofunction:: paddle.v2.fluid.layers.sqrt
+    :noindex:
+
+abs
+----
+..  autofunction:: paddle.v2.fluid.layers.abs
+    :noindex:
+
+ceil
+----
+..  autofunction:: paddle.v2.fluid.layers.ceil
+    :noindex:
+
+floor
+-----
+..  autofunction:: paddle.v2.fluid.layers.floor
+    :noindex:
+
+round
+-----
+..  autofunction:: paddle.v2.fluid.layers.round
+    :noindex:
+
+reciprocal
+----------
+..  autofunction:: paddle.v2.fluid.layers.reciprocal
+    :noindex:
+
+log
+---
+..  autofunction:: paddle.v2.fluid.layers.log
+    :noindex:
+
+square
+------
+..  autofunction:: paddle.v2.fluid.layers.square
+    :noindex:
+
+softplus
+--------
+..  autofunction:: paddle.v2.fluid.layers.softplus
+    :noindex:
+
+softsign
+---------
+..  autofunction:: paddle.v2.fluid.layers.softsign
+    :noindex:
+
+brelu
+-----
+..  autofunction:: paddle.v2.fluid.layers.brelu
+    :noindex:
+
+leaky_relu
+----------
+..  autofunction:: paddle.v2.fluid.layers.leaky_relu
+    :noindex:
+
+soft_relu
+---------
+..  autofunction:: paddle.v2.fluid.layers.soft_relu
+    :noindex:
+
+elu
+----
+..  autofunction:: paddle.v2.fluid.layers.elu
+    :noindex:
+
+relu6
+-----
+..  autofunction:: paddle.v2.fluid.layers.relu6
+    :noindex:
+
+pow
+----
+..  autofunction:: paddle.v2.fluid.layers.pow
+    :noindex:
+
+hard_shrink
+-----------
+..  autofunction:: paddle.v2.fluid.layers.hard_shrink
+    :noindex:
+
+thresholded_relu
+----------------
+..  autofunction:: paddle.v2.fluid.layers.thresholded_relu
+    :noindex:
+
+hard_sigmoid
+-------------
+..  autofunction:: paddle.v2.fluid.layers.hard_sigmoid
+    :noindex:
+
+swish
+------
+..  autofunction:: paddle.v2.fluid.layers.swish
+    :noindex:

python/paddle/v2/fluid/__init__.py

@@ -23,9 +23,22 @@ from memory_optimization_transpiler import memory_optimize
 
 Tensor = LoDTensor
 __all__ = framework.__all__ + executor.__all__ + [
-    'io', 'initializer', 'layers', 'nets', 'optimizer', 'backward',
-    'regularizer', 'LoDTensor', 'CPUPlace', 'CUDAPlace', 'Tensor', 'ParamAttr'
-    'DataFeeder', 'clip', 'DistributeTranspiler', 'memory_optimize'
+    'io',
+    'initializer',
+    'layers',
+    'nets',
+    'optimizer',
+    'backward',
+    'regularizer',
+    'LoDTensor',
+    'CPUPlace',
+    'CUDAPlace',
+    'Tensor',
+    'ParamAttr'
+    'DataFeeder',
+    'clip',
+    'DistributeTranspiler',
+    'memory_optimize',
 ]
 
 

python/paddle/v2/fluid/backward.py

@@ -3,7 +3,10 @@ from . import core
 import collections
 import copy
 
-__all__ = ['append_backward', 'calc_gradient']
+__all__ = [
+    'append_backward',
+    'calc_gradient',
+]
 
 
 def _rename_arg_(op_descs, old_name, new_name, begin_idx=None, end_idx=None):

python/paddle/v2/fluid/clip.py

@@ -3,7 +3,9 @@ import layers
 from . import core
 
 __all__ = [
-    'GradientClipByValue', 'append_gradient_clip_ops', 'error_clip_callback'
+    'GradientClipByValue',
+    'append_gradient_clip_ops',
+    'error_clip_callback',
 ]
 
 

python/paddle/v2/fluid/default_scope_funcs.py

 """
 Default scope function.
 
-`Paddle` manages Scope as programming language's scope.  It just a 
-thread-local stack of Scope. Top of that stack is current scope, the bottom 
-of that stack is all scopes' parent. 
+`Paddle` manages Scope as programming language's scope.  It just a
+thread-local stack of Scope. Top of that stack is current scope, the bottom
+of that stack is all scopes' parent.
 
-Invoking `var/find_var`  can `new/find` variable in current scope. 
-Invoking `enter_local_scope/leave_local_scope` can create or destroy local 
-scope. 
+Invoking `var/find_var`  can `new/find` variable in current scope.
+Invoking `enter_local_scope/leave_local_scope` can create or destroy local
+scope.
 
-A `scoped_function` will take a `function` as input. That function will be 
-invoked in a new local scope. 
+A `scoped_function` will take a `function` as input. That function will be
+invoked in a new local scope.
 """
 
 import paddle.v2.fluid.core
@@ -19,8 +19,12 @@ import threading
 __tl_scope__ = threading.local()
 
 __all__ = [
-    'get_cur_scope', 'enter_local_scope', 'leave_local_scope', 'var',
-    'find_var', 'scoped_function'
+    'get_cur_scope',
+    'enter_local_scope',
+    'leave_local_scope',
+    'var',
+    'find_var',
+    'scoped_function',
 ]
 
 
@@ -71,7 +75,7 @@ def find_var(name):
 def scoped_function(func):
     """
     invoke `func` in new scope.
-    
+
     :param func: a callable function that will be run in new scope.
     :type func: callable
     """

python/paddle/v2/fluid/evaluator.py

@@ -4,7 +4,10 @@ import layers
 from framework import Program, unique_name, Variable, program_guard
 from layer_helper import LayerHelper
 
-__all__ = ['Accuracy', 'ChunkEvaluator']
+__all__ = [
+    'Accuracy',
+    'ChunkEvaluator',
+]
 
 
 def _clone_var_(block, var):
@@ -21,19 +24,19 @@ def _clone_var_(block, var):
 class Evaluator(object):
     """
     Base Class for all evaluators
-    
+
     Args:
-        name(str): The name of evaluator. such as, "accuracy". Used for generate 
+        name(str): The name of evaluator. such as, "accuracy". Used for generate
             temporary variable name.
-        main_program(Program, optional): The evaluator should be added to this 
+        main_program(Program, optional): The evaluator should be added to this
             main_program. Default default_main_program()
-        startup_program(Program, optional):The parameter should be added to this 
+        startup_program(Program, optional):The parameter should be added to this
             startup_program. Default default_startup_program()
-            
+
     Attributes:
-        states(list): The list of state variables. states will be reset to zero 
+        states(list): The list of state variables. states will be reset to zero
             when `reset` is invoked.
-        metrics(list): The list of metrics variables. They will be calculate 
+        metrics(list): The list of metrics variables. They will be calculate
             every mini-batch
     """
 
@@ -66,14 +69,14 @@ class Evaluator(object):
 
     def create_state(self, suffix, dtype, shape):
         """
-        Create state variable. 
-        
+        Create state variable.
+
         NOTE: It is not a public API.
-        
+
         Args:
-            suffix(str): the state suffix. 
-            dtype(str|core.DataType): the state data type 
-            shape(tuple|list): the shape of state 
+            suffix(str): the state suffix.
+            dtype(str|core.DataType): the state data type
+            shape(tuple|list): the shape of state
 
         Returns: State variable
 
@@ -127,8 +130,8 @@ class Accuracy(Evaluator):
 
 class ChunkEvaluator(Evaluator):
     """
-    Accumulate counter numbers output by chunk_eval from mini-batches and 
-    compute the precision recall and F1-score using the accumulated counter 
+    Accumulate counter numbers output by chunk_eval from mini-batches and
+    compute the precision recall and F1-score using the accumulated counter
     numbers.
     """
 

python/paddle/v2/fluid/framework.py

@@ -7,9 +7,15 @@ import proto.framework_pb2 as framework_pb2
 from . import core
 
 __all__ = [
-    'Block', 'Variable', 'Program', 'Operator', 'default_startup_program',
-    'default_main_program', 'program_guard', 'switch_startup_program',
-    'switch_main_program'
+    'Block',
+    'Variable',
+    'Program',
+    'Operator',
+    'default_startup_program',
+    'default_main_program',
+    'program_guard',
+    'switch_startup_program',
+    'switch_main_program',
 ]
 
 EMPTY_VAR_NAME = core.kEmptyVarName()

python/paddle/v2/fluid/initializer.py

 import framework
 import numpy as np
 
-__all__ = ['Constant', 'Uniform', 'Normal', 'Xavier']
+__all__ = [
+    'Constant',
+    'Uniform',
+    'Normal',
+    'Xavier',
+]
 
 
 class Initializer(object):

python/paddle/v2/fluid/io.py

@@ -4,9 +4,15 @@ import cPickle as pickle
 from paddle.v2.fluid.framework import Program, Parameter, default_main_program, Variable
 
 __all__ = [
-    'save_vars', 'save_params', 'save_persistables', 'load_vars', 'load_params',
-    'load_persistables', "save_inference_model", "load_inference_model",
-    "get_inference_program"
+    'save_vars',
+    'save_params',
+    'save_persistables',
+    'load_vars',
+    'load_params',
+    'load_persistables',
+    'save_inference_model',
+    'load_inference_model',
+    'get_inference_program',
 ]
 
 

python/paddle/v2/fluid/layers/nn.py

@@ -9,12 +9,33 @@ from ..param_attr import ParamAttr
 from tensor import concat
 
 __all__ = [
-    'fc', 'embedding', 'dynamic_lstm', 'gru_unit', 'linear_chain_crf',
-    'crf_decoding', 'cos_sim', 'cross_entropy', 'square_error_cost', 'accuracy',
-    'chunk_eval', 'sequence_conv', 'conv2d', 'sequence_pool', 'pool2d',
-    'batch_norm', 'beam_search_decode', 'conv2d_transpose', 'sequence_expand',
-    'lstm_unit', 'reduce_sum', 'reduce_mean', 'reduce_max', 'reduce_min',
-    'sequence_first_step', 'sequence_last_step', 'dropout'
+    'fc',
+    'embedding',
+    'dynamic_lstm',
+    'gru_unit',
+    'linear_chain_crf',
+    'crf_decoding',
+    'cos_sim',
+    'cross_entropy',
+    'square_error_cost',
+    'accuracy',
+    'chunk_eval',
+    'sequence_conv',
+    'conv2d',
+    'sequence_pool',
+    'pool2d',
+    'batch_norm',
+    'beam_search_decode',
+    'conv2d_transpose',
+    'sequence_expand',
+    'lstm_unit',
+    'reduce_sum',
+    'reduce_mean',
+    'reduce_max',
+    'reduce_min',
+    'sequence_first_step',
+    'sequence_last_step',
+    'dropout',
 ]
 
 
@@ -248,13 +269,13 @@ def gru_unit(input,
             h_t & = dot((1-u_t), m_t) + dot(u_t, h_{t-1})
 
     The inputs of gru unit includes :math:`z_t`, :math:`h_{t-1}`. In terms
-    of the equation above, the :math:`z_t` is split into 3 parts - 
-    :math:`xu_t`, :math:`xr_t` and :math:`xm_t`. This means that in order to 
-    implement a full GRU unit operator for an input, a fully 
+    of the equation above, the :math:`z_t` is split into 3 parts -
+    :math:`xu_t`, :math:`xr_t` and :math:`xm_t`. This means that in order to
+    implement a full GRU unit operator for an input, a fully
     connected layer has to be applied, such that :math:`z_t = W_{fc}x_t`.
 
-    The terms :math:`u_t` and :math:`r_t` represent the update and reset gates 
-    of the GRU cell. Unlike LSTM, GRU has one lesser gate. However, there is 
+    The terms :math:`u_t` and :math:`r_t` represent the update and reset gates
+    of the GRU cell. Unlike LSTM, GRU has one lesser gate. However, there is
     an intermediate candidate hidden output, which is denoted by :math:`m_t`.
     This layer has three outputs :math:`h_t`, :math:`dot(r_t, h_{t-1})`
     and concatenation of :math:`u_t`, :math:`r_t` and :math:`m_t`.
@@ -276,7 +297,7 @@ def gru_unit(input,
         .. code-block:: python
 
              # assuming we have x_t_data and prev_hidden of size=10
-             x_t = fluid.layers.fc(input=x_t_data, size=30) 
+             x_t = fluid.layers.fc(input=x_t_data, size=30)
              hidden_val, r_h_val, gate_val = fluid.layers.gru_unit(input=x_t,
                                                     hidden = prev_hidden)
 

python/paddle/v2/fluid/layers/ops.py

 from ..registry import register_layer
 
 __activations__ = [
-    'abs', 'tanh', 'sigmoid', 'relu', 'sqrt', 'ceil', 'floor', 'log', 'round'
+    'sigmoid',
+    'logsigmoid',
+    'exp',
+    'relu',
+    'tanh',
+    'tanh_shrink',
+    'softshrink',
+    'sqrt',
+    'abs',
+    'ceil',
+    'floor',
+    'round',
+    'reciprocal',
+    'log',
+    'square',
+    'softplus',
+    'softsign',
+    'brelu',
+    'leaky_relu',
+    'soft_relu',
+    'elu',
+    'relu6',
+    'pow',
+    'stanh',
+    'hard_shrink',
+    'thresholded_relu',
+    'hard_sigmoid',
+    'swish',
 ]
 
 __all__ = [

python/paddle/v2/fluid/layers/tensor.py

@@ -6,8 +6,16 @@ from ..core import DataType
 import numpy
 
 __all__ = [
-    'create_tensor', 'create_parameter', 'cast', 'concat', 'sums', 'assign',
-    'fill_constant_batch_size_like', 'fill_constant', 'ones', 'zeros'
+    'create_tensor',
+    'create_parameter',
+    'cast',
+    'concat',
+    'sums',
+    'assign',
+    'fill_constant_batch_size_like',
+    'fill_constant',
+    'ones',
+    'zeros',
 ]
 
 

python/paddle/v2/fluid/memory_optimization_transpiler.py

@@ -121,8 +121,10 @@ class ControlFlowGraph(object):
                                 # and dtype_to_size[cache_dtype]
                                 if x_dtype == cache_dtype:
                                     print(
-                                        "Hit Cache !!!! cache pool index is %d, var name is %s, cached var name is %s, var shape is %s "
-                                        %
+                                        ("Hit Cache !!!! cache pool index "
+                                         "is %d, var name is %s, "
+                                         "cached var name is %s, "
+                                         "var shape is %s ") %
                                         (index, x, cache_var, str(cache_shape)))
                                     self.pool.pop(index)
                                     _rename_arg_(

python/paddle/v2/fluid/nets.py

 import layers
 
-__all__ = ["simple_img_conv_pool", "sequence_conv_pool"]
+__all__ = [
+    "simple_img_conv_pool",
+    "sequence_conv_pool",
+]
 
 
 def simple_img_conv_pool(input,

python/paddle/v2/fluid/registry.py

@@ -8,7 +8,11 @@ import proto.framework_pb2 as framework_pb2
 from framework import OpProtoHolder, Variable, Program, Operator
 from paddle.v2.fluid.layer_helper import LayerHelper, unique_name
 
-__all__ = ['deprecated', 'register_layer', 'autodoc']
+__all__ = [
+    'deprecated',
+    'register_layer',
+    'autodoc',
+]
 
 
 def _convert_(name):
@@ -80,11 +84,10 @@ def _generate_doc_string_(op_proto):
 
 
 def register_layer(op_type):
-    """
-    Register an Python layer for an Operator
+    """Register the Python layer for an Operator.
 
     Args:
-       op_type: The name of the operator to be created
+       op_type: The name of the operator to be created.
 
     This function takes in the operator type (sigmoid, mean , average etc) and
     creates the operator functionality.
@@ -98,16 +101,16 @@ def register_layer(op_type):
 
     if len(not_intermediate_outputs) != 1:
         raise ValueError("Only one non intermediate output operator can be",
-                         "automatically generated")
+                         "automatically generated.")
 
     if not_intermediate_outputs[0].duplicable:
         raise ValueError(
-            "Only non duplicable op can be automatically generated")
+            "Only non duplicable op can be automatically generated.")
 
     for output in intermediate_outputs:
         if output.duplicable:
             raise ValueError("The op can be automatically generated only when ",
-                             "all intermediate ops are not duplicable")
+                             "all intermediate ops are not duplicable.")
 
     o_name = not_intermediate_outputs[0].name
     intermediate_output_names = [output.name for output in intermediate_outputs]

python/paddle/v2/fluid/regularizer.py

 import framework
 
-__all__ = ['append_regularization_ops', 'L1Decay', 'L2Decay']
+__all__ = [
+    'append_regularization_ops',
+    'L1Decay',
+    'L2Decay',
+]
 
 
 def append_regularization_ops(parameters_and_grads, regularization=None):