rnn.rst 15.0 KB
Newer Older
W
wangjiang03 已提交
1
Recurrent Neural Network Configuration
2 3
======================================

W
wangjiang03 已提交
4
This tutorial will guide you how to configure recurrent neural network in PaddlePaddle. PaddlePaddle supports highly flexible and efficient recurrent neural network configuration. In this tutorial, you will learn how to:
5

W
wangjiang03 已提交
6 7 8 9
- prepare sequence data for learning recurrent neural networks.
- configure recurrent neural network architecture.
- generate sequence with learned recurrent neural network models.

10
We will use vanilla recurrent neural network, and sequence to sequence model to guide you through these steps. The code of sequence to sequence model can be found at :code:`demo/seqToseq`.
W
wangjiang03 已提交
11

12 13 14
=====================
Prepare Sequence Data
=====================
W
wangjiang03 已提交
15

16
PaddlePaddle does not need any preprocessing to sequence data, such as padding. The only thing that needs to be done is to set the type of the corresponding type to input. For example, the following code snippets defines three input. All of them are sequences, and the size of them are :code:`src_dict`, :code:`trg_dict`, and :code:`trg_dict`:
W
wangjiang03 已提交
17

18
.. code-block:: python
W
wangjiang03 已提交
19 20 21 22

    settings.slots = [
      integer_value_sequence(len(settings.src_dict)),
      integer_value_sequence(len(settings.trg_dict)),
23
      integer_value_sequence(len(settings.trg_dict))]
W
wangjiang03 已提交
24 25


26
Then at the :code:`process` function, each :code:`yield` function will return three integer lists. Each integer list is treated as a sequence of integers:
W
wangjiang03 已提交
27

28
.. code-block:: python
W
wangjiang03 已提交
29 30 31 32

    yield src_ids, trg_ids, trg_ids_next


33
For more details description of how to write a data provider, please refer to `PyDataProvider2 <../../ui/data_provider/index.html>`_. The full data provider file is located at :code:`demo/seqToseq/dataprovider.py`.
W
wangjiang03 已提交
34

35
===============================================
W
wangjiang03 已提交
36
Configure Recurrent Neural Network Architecture
37
===============================================
W
wangjiang03 已提交
38

39
-------------------------------------
W
wangjiang03 已提交
40
Simple Gated Recurrent Neural Network
41 42
-------------------------------------

W
wangjiang03 已提交
43 44 45
Recurrent neural network process a sequence at each time step sequentially. An example of the architecture of LSTM is listed below.

.. image:: ./bi_lstm.jpg
46
	 :align: center
W
wangjiang03 已提交
47

48
Generally speaking, a recurrent network perform the following operations from :math:`t=1` to :math:`t=T`, or reversely from :math:`t=T` to :math:`t=1`.
W
wangjiang03 已提交
49 50 51 52 53 54

.. math::

    x_{t+1} = f_x(x_t), y_t = f_y(x_t)


55
where :math:`f_x(.)` is called **step function**, and :math:`f_y(.)` is called **output function**. In vanilla recurrent neural network, both of the step function and output function are very simple. However, PaddlePaddle supports the configuration of very complex architectures by modifying these two functions. We will use the sequence to sequence model with attention as an example to demonstrate how you can configure complex recurrent neural network models. In this section, we will use a simple vanilla recurrent neural network as an example of configuring simple recurrent neural network using :code:`recurrent_group`. Notice that if you only need to use simple RNN, GRU, or LSTM, then :code:`grumemory` and :code:`lstmemory` is recommended because they are more computationally efficient than :code:`recurrent_group`.
W
wangjiang03 已提交
56

57
For vanilla RNN, at each time step, the **step function** is:
W
wangjiang03 已提交
58 59 60 61 62 63

.. math::

    x_{t+1} = W_x x_t + W_i I_t + b

where :math:`x_t` is the RNN state, and :math:`I_t` is the input, :math:`W_x` and :math:`W_i` are transformation matrices for RNN states and inputs, respectively. :math:`b` is the bias.
64
Its **output function** simply takes :math:`x_t` as the output.
W
wangjiang03 已提交
65

66
:code:`recurrent_group` is the most important tools for constructing recurrent neural networks. It defines the **step function**, **output function** and the inputs of the recurrent neural network. Notice that the :code:`step` argument of this function implements both the :code:`step function` and the :code:`output function`:
W
wangjiang03 已提交
67

68
.. code-block:: python
W
wangjiang03 已提交
69 70 71 72 73 74 75 76 77 78 79 80

    def simple_rnn(input,
                   size=None,
                   name=None,
                   reverse=False,
                   rnn_bias_attr=None,
                   act=None,
                   rnn_layer_attr=None):
        def __rnn_step__(ipt):
           out_mem = memory(name=name, size=size)
           rnn_out = mixed_layer(input = [full_matrix_projection(ipt),
                                          full_matrix_projection(out_mem)],
81 82 83 84 85
                                 name = name,
                                 bias_attr = rnn_bias_attr,
                                 act = act,
                                 layer_attr = rnn_layer_attr,
                                 size = size)
W
wangjiang03 已提交
86 87 88 89 90 91 92
           return rnn_out
        return recurrent_group(name='%s_recurrent_group' % name,
                               step=__rnn_step__,
                               reverse=reverse,
                               input=input)


93
PaddlePaddle uses memory to construct step function. **Memory** is the most important concept when constructing recurrent neural networks in PaddlePaddle. A memory is a state that is used recurrently in step functions, such as :math:`x_{t+1} = f_x(x_t)`. One memory contains an **output** and a **input**. The output of memory at the current time step is utilized as the input of the memory at the next time step. A memory can also has a **boot layer**, whose output is utilized as the initial value of the memory. In our case, the output of the gated recurrent unit is employed as the output memory. Notice that the name of the layer :code:`rnn_out` is the same as the name of :code:`out_mem`. This means the output of the layer :code:`rnn_out` (:math:`x_{t+1}`) is utilized as the **output** of :code:`out_mem` memory.
W
wangjiang03 已提交
94 95 96

A memory can also be a sequence. In this case, at each time step, we have a sequence as the state of the recurrent neural network. This can be useful when constructing very complex recurrent neural network. Other advanced functions include defining multiple memories, and defining hierarchical recurrent neural network architecture using sub-sequence.

97
We return :code:`rnn_out` at the end of the function. It means that the output of the layer :code:`rnn_out` is utilized as the **output** function of the gated recurrent neural network.
W
wangjiang03 已提交
98

99
-----------------------------------------
W
wangjiang03 已提交
100
Sequence to Sequence Model with Attention
101
-----------------------------------------
W
wangjiang03 已提交
102 103 104
We will use the sequence to sequence model with attention as an example to demonstrate how you can configure complex recurrent neural network models. An illustration of the sequence to sequence model with attention is shown in the following figure.

.. image:: ./encoder-decoder-attention-model.png
105
 	 :align: center
W
wangjiang03 已提交
106

107
In this model, the source sequence :math:`S = \{s_1, \dots, s_T\}` is encoded with a bidirectional gated recurrent neural networks. The hidden states of the bidirectional gated recurrent neural network :math:`H_S = \{H_1, \dots, H_T\}` is called *encoder vector* The decoder is a gated recurrent neural network. When decoding each token :math:`y_t`, the gated recurrent neural network generates a set of weights :math:`W_S^t = \{W_1^t, \dots, W_T^t\}`, which are used to compute a weighted sum of the encoder vector. The weighted sum of the encoder vector is utilized to condition the generation of the token :math:`y_t`.
W
wangjiang03 已提交
108

109
The encoder part of the model is listed below. It calls :code:`grumemory` to represent gated recurrent neural network. It is the recommended way of using recurrent neural network if the network architecture is simple, because it is faster than :code:`recurrent_group`. We have implemented most of the commonly used recurrent neural network architectures, you can refer to `Layers <../../ui/api/trainer_config_helpers/layers_index.html>`_  for more details.
W
wangjiang03 已提交
110

111
We also project the encoder vector to :code:`decoder_size` dimensional space, get the first instance of the backward recurrent network, and project it to :code:`decoder_size` dimensional space:
W
wangjiang03 已提交
112

113
.. code-block:: python
W
wangjiang03 已提交
114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141

    # Define the data layer of the source sentence.
    src_word_id = data_layer(name='source_language_word', size=source_dict_dim)
    # Calculate the word embedding of each word.
    src_embedding = embedding_layer(
        input=src_word_id,
        size=word_vector_dim,
        param_attr=ParamAttr(name='_source_language_embedding'))
    # Apply forward recurrent neural network.
    src_forward = grumemory(input=src_embedding, size=encoder_size)
    # Apply backward recurrent neural network. reverse=True means backward recurrent neural network.
    src_backward = grumemory(input=src_embedding,
                              size=encoder_size,
                              reverse=True)
    # Mix the forward and backward parts of the recurrent neural network together.
    encoded_vector = concat_layer(input=[src_forward, src_backward])

    # Project encoding vector to decoder_size.
    encoder_proj = mixed_layer(input = [full_matrix_projection(encoded_vector)],
                               size = decoder_size)

    # Compute the first instance of the backward RNN.
    backward_first = first_seq(input=src_backward)

    # Project the first instance of backward RNN to decoder size.
    decoder_boot = mixed_layer(input=[full_matrix_projection(backward_first)], size=decoder_size, act=TanhActivation())


142
The decoder uses :code:`recurrent_group` to define the recurrent neural network. The step and output functions are defined in :code:`gru_decoder_with_attention`:
W
wangjiang03 已提交
143

144
.. code-block:: python
145 146
    group_inputs=[StaticInput(input=encoded_vector,is_seq=True),
                  StaticInput(input=encoded_proj,is_seq=True)]
W
wangjiang03 已提交
147 148 149 150 151
    trg_embedding = embedding_layer(
        input=data_layer(name='target_language_word',
                         size=target_dict_dim),
        size=word_vector_dim,
        param_attr=ParamAttr(name='_target_language_embedding'))
152 153
    group_inputs.append(trg_embedding)

W
wangjiang03 已提交
154 155 156 157 158 159 160 161
    # For decoder equipped with attention mechanism, in training,
    # target embedding (the groudtruth) is the data input,
    # while encoded source sequence is accessed to as an unbounded memory.
    # StaticInput means the same value is utilized at different time steps.
    # Otherwise, it is a sequence input. Inputs at different time steps are different.
    # All sequence inputs should have the same length.
    decoder = recurrent_group(name=decoder_group_name,
                              step=gru_decoder_with_attention,
162
                              input=group_inputs)
W
wangjiang03 已提交
163 164


165
The implementation of the step function is listed as below. First, it defines the **memory** of the decoder network. Then it defines attention, gated recurrent unit step function, and the output function:
W
wangjiang03 已提交
166

167
.. code-block:: python
W
wangjiang03 已提交
168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199

    def gru_decoder_with_attention(enc_vec, enc_proj, current_word):
        # Defines the memory of the decoder.
        # The output of this memory is defined in gru_step.
        # Notice that the name of gru_step should be the same as the name of this memory.
        decoder_mem = memory(name='gru_decoder',
                             size=decoder_size,
                             boot_layer=decoder_boot)
        # Compute attention weighted encoder vector.
        context = simple_attention(encoded_sequence=enc_vec,
                                   encoded_proj=enc_proj,
                                   decoder_state=decoder_mem)
        # Mix the current word embedding and the attention weighted encoder vector.
        decoder_inputs = mixed_layer(inputs = [full_matrix_projection(context),
                                               full_matrix_projection(current_word)],
                                     size = decoder_size * 3)
        # Define Gated recurrent unit recurrent neural network step function.
        gru_step = gru_step_layer(name='gru_decoder',
                                  input=decoder_inputs,
                                  output_mem=decoder_mem,
                                  size=decoder_size)
        # Defines the output function.
        out = mixed_layer(input=[full_matrix_projection(input=gru_step)],
                          size=target_dict_dim,
                          bias_attr=True,
                          act=SoftmaxActivation())
        return out


=================
Generate Sequence
=================
200
After training the model, we can use it to generate sequences. A common practice is to use **beam search** to generate sequences. The following code snippets defines a beam search algorithm. Notice that :code:`beam_search` function assumes the output function of the :code:`step` returns a softmax normalized probability vector of the next token. We made the following changes to the model.
W
wangjiang03 已提交
201

202 203
* use :code:`GeneratedInput` for trg_embedding. :code:`GeneratedInput` computes the embedding of the generated token at the last time step for the input at the current time step.
* use :code:`beam_search` function. This function needs to set:
W
wangjiang03 已提交
204

205 206 207 208
  - :code:`bos_id`: the start token. Every sentence starts with the start token.
  - :code:`eos_id`: the end token. Every sentence ends with the end token.
  - :code:`beam_size`: the beam size used in beam search.
  - :code:`max_length`: the maximum length of the generated sentences.
W
wangjiang03 已提交
209

210 211 212 213 214 215
* use :code:`seqtext_printer_evaluator` to print text according to index matrix and dictionary. This function needs to set:

  - :code:`id_input`: the integer ID of the data, used to identify the corresponding output in the generated files.
  - :code:`dict_file`: the dictionary file for converting word id to word.
  - :code:`result_file`: the path of the generation result file.
    
216
The code is listed below:
W
wangjiang03 已提交
217

218
.. code-block:: python
W
wangjiang03 已提交
219

220 221
    group_inputs=[StaticInput(input=encoded_vector,is_seq=True),
                  StaticInput(input=encoded_proj,is_seq=True)]
W
wangjiang03 已提交
222 223 224 225 226 227 228 229 230 231
    # In generation, decoder predicts a next target word based on
    # the encoded source sequence and the last generated target word.
    # The encoded source sequence (encoder's output) must be specified by
    # StaticInput which is a read-only memory.
    # Here, GeneratedInputs automatically fetchs the last generated word,
    # which is initialized by a start mark, such as <s>.
    trg_embedding = GeneratedInput(
        size=target_dict_dim,
        embedding_name='_target_language_embedding',
        embedding_size=word_vector_dim)
232
    group_inputs.append(trg_embedding)
W
wangjiang03 已提交
233 234
    beam_gen = beam_search(name=decoder_group_name,
                           step=gru_decoder_with_attention,
235
                           input=group_inputs,
W
wangjiang03 已提交
236 237 238
                           bos_id=0, # Beginnning token.
                           eos_id=1, # End of sentence token.
                           beam_size=beam_size,
239 240 241 242 243 244
                           max_length=max_length)

    seqtext_printer_evaluator(input=beam_gen,
                              id_input=data_layer(name="sent_id", size=1),
                              dict_file=trg_dict_path,
                              result_file=gen_trans_file)
W
wangjiang03 已提交
245 246 247
    outputs(beam_gen)


248
Notice that this generation technique is only useful for decoder like generation process. If you are working on sequence tagging tasks, please refer to `Semantic Role Labeling Demo <../../demo/semantic_role_labeling/index.html>`_ for more details.
W
wangjiang03 已提交
249

250
The full configuration file is located at :code:`demo/seqToseq/seqToseq_net.py`.