Merge pull request #6787 from luotao1/seq_pool_doc

add python wrapper for sequence_pool

doc/api/v2/fluid/layers.rst

@@ -170,6 +170,18 @@ sequence_pool
     :noindex:
 
 
+sequence_first_step
+-------------------
+..  autofunction:: paddle.v2.fluid.layers.sequence_first_step
+    :noindex:
+
+
+sequence_last_step
+------------------
+..  autofunction:: paddle.v2.fluid.layers.sequence_last_step
+    :noindex:
+
+
 pool2d
 ------
 ..  autofunction:: paddle.v2.fluid.layers.pool2d

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

@@ -13,7 +13,8 @@ __all__ = [
     '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'
+    'lstm_unit', 'reduce_sum', 'reduce_mean', 'sequence_first_step',
+    'sequence_last_step'
 ]
 
 
@@ -575,8 +576,52 @@ def conv2d(input,
 def sequence_pool(input, pool_type, **kwargs):
     """
     This function add the operator for sequence pooling. 
-    This is applied on top of the input using pool_type mentioned
-    in the parameters.
+    It pools features of all time-steps of each instance, and is applied 
+    on top of the input using pool_type mentioned in the parameters. 
+
+    It supports four pool_type:
+
+    - average: :math:`Out[i] = \\frac{\sum_i X_i}{N}`
+    - sum:     :math:`Out[i] = \sum_jX_{ij}`
+    - sqrt:    :math:`Out[i] = \\frac{\sum_jX_{ij}}{\sqrt{len(X_i)}}`
+    - max:     :math:`Out[i] = max(X_i)`
+
+    .. code-block:: text
+
+       x is a 1-level LoDTensor:
+         x.lod = [[0, 2, 5, 7]]
+         x.data = [1, 3, 2, 4, 6, 5, 1]
+         x.dims = [7, 1]
+
+       then output is a Tensor:
+         out.dim = [3, 1]
+         with condition len(x.lod[-1]) - 1 == out.dims[0]
+
+       for different pool_type:
+         average: out.data = [2, 4, 3], where 2=(1+3)/2, 4=(2+4+6)/3, 3=(5+1)/2
+         sum    : out.data = [4, 12, 6], where 4=1+3, 12=2+4+6, 6=5+1
+         sqrt   : out.data = [2.82, 6.93, 4.24], where 2.82=(1+3)/sqrt(2),
+                    6.93=(2+4+6)/sqrt(3), 4.24=(5+1)/sqrt(2)
+         max    : out.data = [3, 6, 5], where 3=max(1,3), 6=max(2,4,6), 5=max(5,1)
+         
+    Args:
+        input(variable): The input variable which is a LoDTensor.
+        pool_type (string): The pooling type of sequence_pool. 
+            It supports average, sum, sqrt and max.
+
+    Returns:
+        The sequence pooling variable which is a Tensor.
+
+    Examples:
+
+        .. code-block:: python
+             
+             x = fluid.layers.data(name='x', shape=[7, 1], 
+                              dtype='float32', lod_level=1)
+             avg_x = fluid.layers.sequence_pool(input=x, pool_type='average')
+             sum_x = fluid.layers.sequence_pool(input=x, pool_type='sum')
+             sqrt_x = fluid.layers.sequence_pool(input=x, pool_type='sqrt')
+             max_x = fluid.layers.sequence_pool(input=x, pool_type='max')
     """
     helper = LayerHelper('sequence_pool', input=input, **kwargs)
     dtype = helper.input_dtype()
@@ -593,6 +638,72 @@ def sequence_pool(input, pool_type, **kwargs):
     return pool_out
 
 
+def sequence_first_step(input, **kwargs):
+    """
+    This funciton get the first step of sequence.
+
+    .. code-block:: text
+
+       x is a 1-level LoDTensor:
+         x.lod = [[0, 2, 5, 7]]
+         x.data = [1, 3, 2, 4, 6, 5, 1]
+         x.dims = [7, 1]
+
+       then output is a Tensor:
+         out.dim = [3, 1]
+         with condition len(x.lod[-1]) - 1 == out.dims[0]
+         out.data = [1, 2, 5], where 1=first(1,3), 2=first(2,4,6), 5=first(5,1)
+         
+    Args:
+        input(variable): The input variable which is a LoDTensor.
+
+    Returns:
+        The sequence's first step variable which is a Tensor.
+
+    Examples:
+
+        .. code-block:: python
+             
+             x = fluid.layers.data(name='x', shape=[7, 1], 
+                              dtype='float32', lod_level=1)
+             x_first_step = fluid.layers.sequence_first_step(input=x)
+    """
+    return sequence_pool(input=input, pool_type="first")
+
+
+def sequence_last_step(input, **kwargs):
+    """
+    This funciton get the last step of sequence.
+
+    .. code-block:: text
+
+       x is a 1-level LoDTensor:
+         x.lod = [[0, 2, 5, 7]]
+         x.data = [1, 3, 2, 4, 6, 5, 1]
+         x.dims = [7, 1]
+
+       then output is a Tensor:
+         out.dim = [3, 1]
+         with condition len(x.lod[-1]) - 1 == out.dims[0]
+         out.data = [3, 6, 1], where 3=last(1,3), 6=last(2,4,6), 1=last(5,1)
+         
+    Args:
+        input(variable): The input variable which is a LoDTensor.
+
+    Returns:
+        The sequence's last step variable which is a Tensor.
+
+    Examples:
+
+        .. code-block:: python
+             
+             x = fluid.layers.data(name='x', shape=[7, 1], 
+                              dtype='float32', lod_level=1)
+             x_last_step = fluid.layers.sequence_last_step(input=x)
+    """
+    return sequence_pool(input=input, pool_type="last")
+
+
 def pool2d(input,
            pool_size,
            pool_type,

python/paddle/v2/fluid/tests/book/test_machine_translation.py

@@ -33,7 +33,7 @@ def encoder_decoder():
 
     fc1 = fluid.layers.fc(input=src_embedding, size=hidden_dim * 4, act='tanh')
     lstm_hidden0, lstm_0 = layers.dynamic_lstm(input=fc1, size=hidden_dim * 4)
-    encoder_out = layers.sequence_pool(input=lstm_hidden0, pool_type="last")
+    encoder_out = layers.sequence_last_step(input=lstm_hidden0)
 
     # decoder
     trg_language_word = layers.data(

python/paddle/v2/fluid/tests/test_dyn_rnn.py

@@ -63,8 +63,7 @@ class TestDynRNN(unittest.TestCase):
 
             all_timesteps = fluid.layers.array_to_lod_tensor(
                 x=out, table=rank_table)
-            last = fluid.layers.sequence_pool(
-                input=all_timesteps, pool_type='last')
+            last = fluid.layers.sequence_last_step(input=all_timesteps)
             logits = fluid.layers.fc(input=last, size=1, act=None)
             loss = fluid.layers.sigmoid_cross_entropy_with_logits(
                 x=logits, label=label)
@@ -101,7 +100,7 @@ class TestDynRNN(unittest.TestCase):
                 rnn.update_memory(mem, out_)
                 rnn.output(out_)
 
-            last = fluid.layers.sequence_pool(input=rnn(), pool_type='last')
+            last = fluid.layers.sequence_last_step(input=rnn())
             logits = fluid.layers.fc(input=last, size=1, act=None)
             label = fluid.layers.data(name='label', shape=[1], dtype='float32')
             loss = fluid.layers.sigmoid_cross_entropy_with_logits(