def build_permutation(self):

        with tf.variable_scope("encoder"):
            
            with tf.variable_scope("embedding"):
                # Embed input sequence
                W_embed =tf.get_variable("weights", [1,self.input_dimension+2, self.input_embed], initializer=self.initializer) # +2 for TW feat. here too
                embedded_input = tf.nn.conv1d(self.input_, W_embed, 1, "VALID", name="embedded_input")
                # Batch Normalization
                embedded_input = tf.layers.batch_normalization(embedded_input, axis=2, training=self.is_training, name='layer_norm', reuse=None)

            with tf.variable_scope("dynamic_rnn"):
                # Encode input sequence
                cell1 = LSTMCell(self.num_neurons, initializer=self.initializer)  # BNLSTMCell(self.num_neurons, self.training) or cell1 = DropoutWrapper(cell1, output_keep_prob=0.9)
                # Return the output activations [Batch size, Sequence Length, Num_neurons] and last hidden state as tensors.
                encoder_output, encoder_state = tf.nn.dynamic_rnn(cell1, embedded_input, dtype=tf.float32)

        with tf.variable_scope('decoder'):
            # Ptr-net returns permutations (self.positions), with their log-probability for backprop
            self.ptr = Pointer_decoder(encoder_output, self.config)
            self.positions, self.log_softmax, self.attending, self.pointing = self.ptr.loop_decode(encoder_state)
            variable_summaries('log_softmax',self.log_softmax, with_max_min = True) 

def get_rnn_cell_trainable_variables(cell):
    """Returns the list of trainable variables of an RNN cell.

    Args:
        cell: an instance of :tf_main:`RNNCell <nn/rnn_cell/RNNCell>`.

    Returns:
        list: trainable variables of the cell.
    """
    cell_ = cell
    while True:
        try:
            return cell_.trainable_variables
        except AttributeError:
        # Cell wrappers (e.g., `DropoutWrapper`) cannot directly access to
        # `trainable_variables` as they don't initialize superclass
        # (tf==v1.3). So try to access through the cell in the wrapper.
            cell_ = cell._cell  # pylint: disable=protected-access 

def apply_dropout(
    cell, input_keep_probability, output_keep_probability, random_seed=None):
  """Apply dropout to the outputs and inputs of `cell`.

  Args:
    cell: An `RNNCell`.
    input_keep_probability: Probability to keep inputs to `cell`. If `None`,
      no dropout is applied.
    output_keep_probability: Probability to keep outputs of `cell`. If `None`,
      no dropout is applied.
    random_seed: Seed for random dropout.

  Returns:
    An `RNNCell`, the result of applying the supplied dropouts to `cell`.
  """
  input_prob_none = input_keep_probability is None
  output_prob_none = output_keep_probability is None
  if input_prob_none and output_prob_none:
    return cell
  if input_prob_none:
    input_keep_probability = 1.0
  if output_prob_none:
    output_keep_probability = 1.0
  return contrib_rnn.DropoutWrapper(
      cell, input_keep_probability, output_keep_probability, random_seed) 

def build_lstm(self):
        def build_cell():
            cell = rnn.BasicLSTMCell(self._hidden_size, forget_bias=1.0, state_is_tuple=True)
            cell = rnn.DropoutWrapper(cell, output_keep_prob=self._keep_prob)
            return cell
        mul_cell = rnn.MultiRNNCell([build_cell() for _ in range(self._num_layer)], 
                                    state_is_tuple=True)
        self._init_state = mul_cell.zero_state(self._num_seq, dtype=tf.float32)
        outputs, self._final_state = tf.nn.dynamic_rnn(mul_cell, self._inputs, 
                                                       initial_state=self._init_state)
        outputs = tf.reshape(outputs, [-1, self._hidden_size])
        W = tf.Variable(tf.truncated_normal([self._hidden_size, self._corpus.word_num],
                                            stddev=0.1, dtype=tf.float32))
        bais = tf.Variable(tf.zeros([1, self._corpus.word_num], 
                                    dtype=tf.float32), dtype=tf.float32)
        self._prediction = tf.nn.softmax(tf.matmul(outputs, W) + bais) 

def _create_rnn_cell(self):
        """
        Creates a single RNN cell according to the architecture of this RNN.
        
        Returns
        -------
        rnn cell
            A single RNN cell according to the architecture of this RNN
        """
        keep_prob = 1.0 if self.keep_prob is None else self.keep_prob

        if self.cell_type == CellType.GRU:
            return DropoutWrapper(GRUCell(self.num_units), keep_prob, keep_prob)
        elif self.cell_type == CellType.LSTM:
            return DropoutWrapper(LSTMCell(self.num_units), keep_prob, keep_prob)
        else:
            raise ValueError("unknown cell type: {}".format(self.cell_type)) 

def inference(self):
        """main computation graph here: 1. embeddding layer, 2.Bi-LSTM layer, 3.concat, 4.FC layer 5.softmax """
        #1.get emebedding of words in the sentence
        self.embedded_words = tf.nn.embedding_lookup(self.Embedding,self.input_x) #shape:[None,sentence_length,embed_size]
        #2. Bi-lstm layer
        # define lstm cess:get lstm cell output
        lstm_fw_cell=rnn.BasicLSTMCell(self.hidden_size) #forward direction cell
        lstm_bw_cell=rnn.BasicLSTMCell(self.hidden_size) #backward direction cell
        if self.dropout_keep_prob is not None:
            lstm_fw_cell=rnn.DropoutWrapper(lstm_fw_cell,output_keep_prob=self.dropout_keep_prob)
            lstm_bw_cell=rnn.DropoutWrapper(lstm_bw_cell,output_keep_prob=self.dropout_keep_prob)
        # bidirectional_dynamic_rnn: input: [batch_size, max_time, input_size]
        #                            output: A tuple (outputs, output_states)
        #                                    where:outputs: A tuple (output_fw, output_bw) containing the forward and the backward rnn output `Tensor`.
        outputs,_=tf.nn.bidirectional_dynamic_rnn(lstm_fw_cell,lstm_bw_cell,self.embedded_words,dtype=tf.float32) #[batch_size,sequence_length,hidden_size] #creates a dynamic bidirectional recurrent neural network
        print("outputs:===>",outputs) #outputs:(<tf.Tensor 'bidirectional_rnn/fw/fw/transpose:0' shape=(?, 5, 100) dtype=float32>, <tf.Tensor 'ReverseV2:0' shape=(?, 5, 100) dtype=float32>))
        #3. concat output
        output_rnn=tf.concat(outputs,axis=2) #[batch_size,sequence_length,hidden_size*2]
        #self.output_rnn_last=tf.reduce_mean(output_rnn,axis=1) #[batch_size,hidden_size*2] 
        self.output_rnn_last=output_rnn[:,-1,:] ##[batch_size,hidden_size*2] #TODO
        print("output_rnn_last:", self.output_rnn_last) # <tf.Tensor 'strided_slice:0' shape=(?, 200) dtype=float32>
        #4. logits(use linear layer)
        with tf.name_scope("output"): #inputs: A `Tensor` of shape `[batch_size, dim]`.  The forward activations of the input network.
            logits = tf.matmul(self.output_rnn_last, self.W_projection) + self.b_projection  # [batch_size,num_classes]
        return logits 

def inference(self):
        """main computation graph here: 1. embeddding layer, 2.Bi-LSTM layer, 3.mean pooling, 4.FC layer, 5.softmax """
        #1.get emebedding of words in the sentence
        self.embedded_words = tf.nn.embedding_lookup(self.Embedding,self.input_x) #shape:[None,sentence_length,embed_size]
        #2. Bi-lstm layer
        # define lstm cess:get lstm cell output
        lstm_fw_cell=rnn.BasicLSTMCell(self.hidden_size) #forward direction cell
        lstm_bw_cell=rnn.BasicLSTMCell(self.hidden_size) #backward direction cell
        if self.dropout_keep_prob is not None:
            lstm_fw_cell=rnn.DropoutWrapper(lstm_fw_cell,output_keep_prob=self.dropout_keep_prob)
            lstm_bw_cell==rnn.DropoutWrapper(lstm_bw_cell,output_keep_prob=self.dropout_keep_prob)
        # bidirectional_dynamic_rnn: input: [batch_size, max_time, input_size]
        #                            output: A tuple (outputs, output_states)
        #                                    where:outputs: A tuple (output_fw, output_bw) containing the forward and the backward rnn output `Tensor`.
        outputs,_=tf.nn.bidirectional_dynamic_rnn(lstm_fw_cell,lstm_bw_cell,self.embedded_words,dtype=tf.float32) #[batch_size,sequence_length,hidden_size] #creates a dynamic bidirectional recurrent neural network
        print("outputs:===>",outputs) #outputs:(<tf.Tensor 'bidirectional_rnn/fw/fw/transpose:0' shape=(?, 5, 100) dtype=float32>, <tf.Tensor 'ReverseV2:0' shape=(?, 5, 100) dtype=float32>))
        #3. concat output
        output_rnn=tf.concat(outputs,axis=2) #[batch_size,sequence_length,hidden_size*2]
        output_rnn_pooled=tf.reduce_mean(output_rnn,axis=1) #[batch_size,hidden_size*2] #output_rnn_last=output_rnn[:,-1,:] ##[batch_size,hidden_size*2] #TODO
        print("output_rnn_pooled:", output_rnn_pooled) # <tf.Tensor 'strided_slice:0' shape=(?, 200) dtype=float32>
        #4. logits(use linear layer)
        with tf.name_scope("output"): #inputs: A `Tensor` of shape `[batch_size, dim]`.  The forward activations of the input network.
            logits = tf.matmul(output_rnn_pooled, self.W_projection) + self.b_projection  # [batch_size,num_classes]
        return logits 

def get_rnn_cell_trainable_variables(cell):
    """Returns the list of trainable variables of an RNN cell.

    Args:
        cell: an instance of :tf_main:`RNNCell <nn/rnn_cell/RNNCell>`.

    Returns:
        list: trainable variables of the cell.
    """
    cell_ = cell
    while True:
        try:
            return cell_.trainable_variables
        except AttributeError:
            # Cell wrappers (e.g., `DropoutWrapper`) cannot directly access to
            # `trainable_variables` as they don't initialize superclass
            # (tf==v1.3). So try to access through the cell in the wrapper.
            cell_ = cell._cell  # pylint: disable=protected-access 

def inference(self):
        """main computation graph here: 1. embeddding layer, 2.Bi-LSTM layer, 3.concat, 4.FC layer 5.softmax """
        #1.get emebedding of words in the sentence
        self.embedded_words = tf.nn.embedding_lookup(self.Embedding,self.input_x) #shape:[None,sentence_length,embed_size]
        #2. Bi-lstm layer
        # define lstm cess:get lstm cell output
        lstm_fw_cell=rnn.BasicLSTMCell(self.hidden_size) #forward direction cell
        lstm_bw_cell=rnn.BasicLSTMCell(self.hidden_size) #backward direction cell
        if self.dropout_keep_prob is not None:
            lstm_fw_cell=rnn.DropoutWrapper(lstm_fw_cell,output_keep_prob=self.dropout_keep_prob)
            lstm_bw_cell=rnn.DropoutWrapper(lstm_bw_cell,output_keep_prob=self.dropout_keep_prob)
        # bidirectional_dynamic_rnn: input: [batch_size, max_time, input_size]
        #                            output: A tuple (outputs, output_states)
        #                                    where:outputs: A tuple (output_fw, output_bw) containing the forward and the backward rnn output `Tensor`.
        outputs,_=tf.nn.bidirectional_dynamic_rnn(lstm_fw_cell,lstm_bw_cell,self.embedded_words,dtype=tf.float32) #[batch_size,sequence_length,hidden_size] #creates a dynamic bidirectional recurrent neural network
        print("outputs:===>",outputs) #outputs:(<tf.Tensor 'bidirectional_rnn/fw/fw/transpose:0' shape=(?, 5, 100) dtype=float32>, <tf.Tensor 'ReverseV2:0' shape=(?, 5, 100) dtype=float32>))
        #3. concat output
        output_rnn=tf.concat(outputs,axis=2) #[batch_size,sequence_length,hidden_size*2]
        #self.output_rnn_last=tf.reduce_mean(output_rnn,axis=1) #[batch_size,hidden_size*2] 
        self.output_rnn_last=output_rnn[:,-1,:] ##[batch_size,hidden_size*2] #TODO
        print("output_rnn_last:", self.output_rnn_last) # <tf.Tensor 'strided_slice:0' shape=(?, 200) dtype=float32>
        #4. logits(use linear layer)
        with tf.name_scope("output"): #inputs: A `Tensor` of shape `[batch_size, dim]`.  The forward activations of the input network.
            logits = tf.matmul(self.output_rnn_last, self.W_projection) + self.b_projection  # [batch_size,num_classes]
        return logits 

def apply_dropout(
    cell, input_keep_probability, output_keep_probability, random_seed=None):
  """Apply dropout to the outputs and inputs of `cell`.

  Args:
    cell: An `RNNCell`.
    input_keep_probability: Probability to keep inputs to `cell`. If `None`,
      no dropout is applied.
    output_keep_probability: Probability to keep outputs of `cell`. If `None`,
      no dropout is applied.
    random_seed: Seed for random dropout.

  Returns:
    An `RNNCell`, the result of applying the supplied dropouts to `cell`.
  """
  input_prob_none = input_keep_probability is None
  output_prob_none = output_keep_probability is None
  if input_prob_none and output_prob_none:
    return cell
  if input_prob_none:
    input_keep_probability = 1.0
  if output_prob_none:
    output_keep_probability = 1.0
  return contrib_rnn.DropoutWrapper(
      cell, input_keep_probability, output_keep_probability, random_seed) 

def horizontal_cell(images, num_filters_out, cell_fw, cell_bw, keep_prob=1.0, scope=None):
  """Run an LSTM bidirectionally over all the rows of each image.

  Args:
    images: (num_images, height, width, depth) tensor
    num_filters_out: output depth
    scope: optional scope name

  Returns:
    (num_images, height, width, num_filters_out) tensor, where
  """
  with tf.variable_scope(scope, "HorizontalGru", [images]):
    sequence = images_to_sequence(images)

    shapeT = tf.shape(sequence)
    sequence_length = shapeT[0]
    batch_sizeRNN = shapeT[1]
    sequence_lengths = tf.to_int64(
      tf.fill([batch_sizeRNN], sequence_length))
    forward_drop1 = DropoutWrapper(cell_fw, output_keep_prob=keep_prob)
    backward_drop1 = DropoutWrapper(cell_bw, output_keep_prob=keep_prob)
    rnn_out1, _ = tf.nn.bidirectional_dynamic_rnn(forward_drop1, backward_drop1, sequence, dtype=tf.float32,
                                                  sequence_length=sequence_lengths, time_major=True,
                                                  swap_memory=True, scope=scope)
    rnn_out1 = tf.concat(rnn_out1, 2)
    rnn_out1 = tf.reshape(rnn_out1, shape=[-1, batch_sizeRNN, 2, num_filters_out])
    output_sequence = tf.reduce_sum(rnn_out1, axis=2)
    batch_size=tf.shape(images)[0]
    output = sequence_to_images(output_sequence, batch_size)
    return output 

def apply_dropout(cells, dropout_keep_probabilities, random_seed=None):
  """Applies dropout to the outputs and inputs of `cell`.

  Args:
    cells: A list of `RNNCell`s.
    dropout_keep_probabilities: a list whose elements are either floats in
    `[0.0, 1.0]` or `None`. It must have length one greater than `cells`.
    random_seed: Seed for random dropout.

  Returns:
    A list of `RNNCell`s, the result of applying the supplied dropouts.

  Raises:
    ValueError: If `len(dropout_keep_probabilities) != len(cells) + 1`.
  """
  if len(dropout_keep_probabilities) != len(cells) + 1:
    raise ValueError(
        'The number of dropout probabilites must be one greater than the '
        'number of cells. Got {} cells and {} dropout probabilities.'.format(
            len(cells), len(dropout_keep_probabilities)))
  wrapped_cells = [
      contrib_rnn.DropoutWrapper(cell, prob, 1.0, seed=random_seed)
      for cell, prob in zip(cells[:-1], dropout_keep_probabilities[:-2])
  ]
  wrapped_cells.append(
      contrib_rnn.DropoutWrapper(cells[-1], dropout_keep_probabilities[-2],
                                 dropout_keep_probabilities[-1]))
  return wrapped_cells 

def cell_create(self,scope_name):
         with tf.variable_scope(scope_name):
             if self.cell_type == 'tanh':
                 cells = rnn.MultiRNNCell([rnn.BasicRNNCell(self.n_hidden[i]) for i in range(self.n_layers)], state_is_tuple=True)
             elif self.cell_type == 'LSTM': 
                 cells = rnn.MultiRNNCell([rnn.BasicLSTMCell(self.n_hidden[i]) for i in range(self.n_layers)], state_is_tuple=True)
             elif self.cell_type == 'GRU':
                 cells = rnn.MultiRNNCell([rnn.GRUCell(self.n_hidden[i]) for i in range(self.n_layers)], state_is_tuple=True)
             elif self.cell_type == 'LSTMP':
                 cells = rnn.MultiRNNCell([rnn.LSTMCell(self.n_hidden[i]) for i in range(self.n_layers)], state_is_tuple=True)
             cells = rnn.DropoutWrapper(cells, input_keep_prob=self.dropout_ph,output_keep_prob=self.dropout_ph) 
         return cells 

def define_graph(self):
        essays = tf.placeholder(tf.float32, [None, self.hp.d_e_len,
                                             self.hp.w_dim])
        scores = tf.placeholder(tf.float32, [None])

        # Long Short-Term Memory layer
        lstm_cell = tfrnn.BasicLSTMCell(num_units=self.hp.lstm_hidden_size)
        lstm_cell = tfrnn.DropoutWrapper(
            cell=lstm_cell,
            output_keep_prob=self.hp.dropout_keep_prob)
        init_state = lstm_cell.zero_state(self.hp.batch_size, dtype=tf.float32)
        lstm, _ = tf.nn.dynamic_rnn(lstm_cell, essays, dtype=tf.float32)

        # Mean over Time pooling
        mot = tf.reduce_mean(lstm, axis=1)

        # Dense layer
        dense = tf.layers.dense(inputs=mot, units=1, activation=tf.nn.sigmoid)

        # Prediction and Loss
        preds = tf.reshape(dense, [-1])
        tf.summary.histogram('preds', preds)
        loss = tf.losses.mean_squared_error(scores, preds)
        tf.summary.scalar('loss', loss)
        merged_summary = tf.summary.merge_all()

        return (essays,
                scores,
                merged_summary,
                loss,
                preds) 

def getLayeredCell(layer_size, num_units, input_keep_prob,
        output_keep_prob=1.0):
    return rnn.MultiRNNCell([rnn.DropoutWrapper(rnn.BasicLSTMCell(num_units),
        input_keep_prob, output_keep_prob) for i in range(layer_size)]) 

def func():
    lstm_cell = rnn.BasicLSTMCell(hidden_size, forget_bias=1.0, state_is_tuple=True)
    lstm_cell = rnn.DropoutWrapper(lstm_cell, output_keep_prob=keep_prob)
    return lstm_cell 

def func():
    lstm_cell = rnn.BasicLSTMCell(num_units=hidden_size, forget_bias=1.0, state_is_tuple=True)
    lstm_cell = rnn.DropoutWrapper(lstm_cell, output_keep_prob=keep_prob)
    return lstm_cell 

def define_rnn_cell(cell_class, num_units, num_layers=1, keep_prob=1.0,
                    input_keep_prob=None, output_keep_prob=None):
    if input_keep_prob is None:
        input_keep_prob = keep_prob
    if output_keep_prob is None:
        output_keep_prob = keep_prob

    cells = []
    for _ in range(num_layers):
        if cell_class == 'GRU':
            cell = GRUCell(num_units=num_units)
        elif cell_class == 'LSTM':
            cell = LSTMCell(num_units=num_units)
        else:
            cell = RNNCell(num_units=num_units)

        if keep_prob < 1.0:
            cell = DropoutWrapper(cell=cell, input_keep_prob=input_keep_prob, output_keep_prob=output_keep_prob)
        cells.append(cell)

    if len(cells) > 1:
        final_cell = MultiRNNCell(cells)
    else:
        final_cell = cells[0]

    return final_cell 

def _make_cell(self, hidden_size=None):
        """Create a single RNN cell"""
        cell = self.cell_type(hidden_size or self.hidden_size)
        if self.dropout:
            cell = rnn.DropoutWrapper(cell, self.keep_prob)
        if self.residual:
            cell = rnn.ResidualWrapper(cell)
        return cell 

def build_single_cell(self, n_hidden, use_residual):
        """构建一个单独的rnn cell
        Args:
            n_hidden: 隐藏层神经元数量
            use_residual: 是否使用residual wrapper
        """

        if self.cell_type == 'gru':
            cell_type = GRUCell
        else:
            cell_type = LSTMCell

        cell = cell_type(n_hidden)

        if self.use_dropout:
            cell = DropoutWrapper(
                cell,
                dtype=tf.float32,
                output_keep_prob=self.keep_prob_placeholder,
                seed=self.seed
            )

        if use_residual:
            cell = ResidualWrapper(cell)

        return cell 

def create_cell(unit_type, hidden_units, num_layers, use_residual=False, input_keep_prob=1.0, output_keep_prob=1.0, devices=None):
    if unit_type == 'lstm':
        def _new_cell():
            return tf.nn.rnn_cell.BasicLSTMCell(hidden_units)
    elif unit_type == 'gru':
        def _new_cell():
            return tf.contrib.rnn.GRUCell(hidden_units)
    else:
        raise ValueError('cell_type must be either lstm or gru')

    def _new_cell_wrapper(residual_connection=False, device_id=None):
        c = _new_cell()

        if input_keep_prob < 1.0 or output_keep_prob < 1.0:
            c = rnn.DropoutWrapper(c, input_keep_prob=input_keep_prob, output_keep_prob=output_keep_prob)

        if residual_connection:
            c = rnn.ResidualWrapper(c)

        if device_id:
            c = rnn.DeviceWrapper(c, device_id)

        return c

    if num_layers > 1:
        cells = []

        for i in range(num_layers):
            is_residual = True if use_residual and i > 0 else False
            cells.append(_new_cell_wrapper(is_residual, devices[i] if devices else None))

        return tf.contrib.rnn.MultiRNNCell(cells)
    else:
        return _new_cell_wrapper(device_id=devices[0] if devices else None) 

def __call__(self, inputs, mask):
        '''
        inputs: the embeddings of a batch of sequences. (batch_size, seq_length, emb_size)
        mask: mask for imcomplete sequences. (batch_size, seq_length, 1)
        '''
        cells = []
        for _ in range(self.layers):
            cell = rnn.BasicLSTMCell(self.hidden_units, activation=self.hidden_activation)
            cell = rnn.DropoutWrapper(cell, output_keep_prob=1.-self.dropout)
            cells.append(cell)
        self.cell = cell = rnn.MultiRNNCell(cells)
        zero_state = cell.zero_state(tf.shape(inputs)[0], dtype=tf.float32)
        sequence_length = tf.count_nonzero(tf.squeeze(mask, [-1]), -1)
        outputs, state = tf.nn.dynamic_rnn(cell, inputs, sequence_length=sequence_length, initial_state=zero_state)
        return outputs 

def lstm_cell(self):
        cell = rnn.LSTMCell(self.cfg.getint('net_work', 'hidden_size'), reuse=tf.get_variable_scope().reuse)
        return rnn.DropoutWrapper(cell, output_keep_prob=self.keep_prob) 

def _witch_cell(self):
        """
        RNN 类型
        :return: 
        """
        cell_tmp = None
        if self.cell_type == 'lstm':
            cell_tmp = rnn.BasicLSTMCell(self.hidden_unit)
        elif self.cell_type == 'gru':
            cell_tmp = rnn.GRUCell(self.hidden_unit)
        # 是否需要进行dropout
        if self.dropout_rate is not None:
            cell_tmp = rnn.DropoutWrapper(cell_tmp, output_keep_prob=self.dropout_rate)
        return cell_tmp 

def input_encoder_bi_lstm(self):
        """use bi-directional lstm to encode query_embedding:[batch_size,sequence_length,embed_size]
                                         and story_embedding:[batch_size,story_length,sequence_length,embed_size]
        output:query_embedding:[batch_size,hidden_size*2]  story_embedding:[batch_size,self.story_length,self.hidden_size*2]
        """
        #1. encode query: bi-lstm layer
        lstm_fw_cell = rnn.BasicLSTMCell(self.hidden_size)  # forward direction cell
        lstm_bw_cell = rnn.BasicLSTMCell(self.hidden_size)  # backward direction cell
        if self.dropout_keep_prob is not None:
            lstm_fw_cell = rnn.DropoutWrapper(lstm_fw_cell, output_keep_prob=self.dropout_keep_prob)
            lstm_bw_cell == rnn.DropoutWrapper(lstm_bw_cell, output_keep_prob=self.dropout_keep_prob)
        query_hidden_output, _ = tf.nn.bidirectional_dynamic_rnn(lstm_fw_cell, lstm_bw_cell, self.query_embedding,dtype=tf.float32,scope="query_rnn")  # [batch_size,sequence_length,hidden_size] #creates a dynamic bidirectional recurrent neural network
        query_hidden_output = tf.concat(query_hidden_output, axis=2) #[batch_size,sequence_length,hidden_size*2]
        self.query_embedding=tf.reduce_sum(query_hidden_output,axis=1) #[batch_size,hidden_size*2]
        print("input_encoder_bi_lstm.self.query_embedding:",self.query_embedding)

        #2. encode story
        # self.story_embedding:[batch_size,story_length,sequence_length,embed_size]
        self.story_embedding=tf.reshape(self.story_embedding,shape=(-1,self.story_length*self.sequence_length,self.embed_size)) #[self.story_length*self.sequence_length,self.embed_size]
        lstm_fw_cell_story = rnn.BasicLSTMCell(self.hidden_size)  # forward direction cell
        lstm_bw_cell_story = rnn.BasicLSTMCell(self.hidden_size)  # backward direction cell
        if self.dropout_keep_prob is not None:
            lstm_fw_cell_story = rnn.DropoutWrapper(lstm_fw_cell_story, output_keep_prob=self.dropout_keep_prob)
            lstm_bw_cell_story == rnn.DropoutWrapper(lstm_bw_cell_story, output_keep_prob=self.dropout_keep_prob)
        story_hidden_output, _ = tf.nn.bidirectional_dynamic_rnn(lstm_fw_cell_story, lstm_bw_cell_story, self.story_embedding,dtype=tf.float32,scope="story_rnn")
        story_hidden_output=tf.concat(story_hidden_output,axis=2) #[batch_size,story_length*sequence_length,hidden_size*2]
        story_hidden_output=tf.reshape(story_hidden_output,shape=(-1,self.story_length,self.sequence_length,self.hidden_size*2))
        self.story_embedding = tf.reduce_sum(story_hidden_output, axis=2)  # [batch_size,self.story_length,self.hidden_size*2] 

def inference(self):
        """main computation graph here: 1. embeddding layer, 2.Bi-LSTM layer-->dropout,3.LSTM layer-->dropout 4.FC layer 6.softmax layer """
        #1.get emebedding of words in the sentence
        self.embedded_words = tf.nn.embedding_lookup(self.Embedding,self.input_x) #shape:[None,sentence_length,embed_size]
        
        #2. Bi-lstm layer
        # define lstm cess:get lstm cell output
        lstm_fw_cell=rnn.BasicLSTMCell(self.hidden_size) #forward direction cell
        lstm_bw_cell=rnn.BasicLSTMCell(self.hidden_size) #backward direction cell
        if self.dropout_keep_prob is not None:
            lstm_fw_cell=rnn.DropoutWrapper(lstm_fw_cell,output_keep_prob=self.dropout_keep_prob)
            lstm_bw_cell=rnn.DropoutWrapper(lstm_bw_cell,output_keep_prob=self.dropout_keep_prob)
        # bidirectional_dynamic_rnn: input: [batch_size, max_time, input_size]
        #                            output: A tuple (outputs, output_states)
        #                                    where:outputs: A tuple (output_fw, output_bw) containing the forward and the backward rnn output `Tensor`.
        outputs,_=tf.nn.bidirectional_dynamic_rnn(lstm_fw_cell,lstm_bw_cell,self.embedded_words,dtype=tf.float32) #[batch_size,sequence_length,hidden_size] #creates a dynamic bidirectional recurrent neural network
        print("outputs:===>",outputs) #outputs:(<tf.Tensor 'bidirectional_rnn/fw/fw/transpose:0' shape=(?, 5, 100) dtype=float32>, <tf.Tensor 'ReverseV2:0' shape=(?, 5, 100) dtype=float32>))
        output_rnn=tf.concat(outputs,axis=2) #[batch_size,sequence_length,hidden_size*2]

        #3. Second LSTM layer
        rnn_cell=rnn.BasicLSTMCell(self.hidden_size*2)
        if self.dropout_keep_prob is not None:
            rnn_cell=rnn.DropoutWrapper(rnn_cell,output_keep_prob=self.dropout_keep_prob)
        _,final_state_c_h=tf.nn.dynamic_rnn(rnn_cell,output_rnn,dtype=tf.float32)
        final_state=final_state_c_h[1]

        #4 .FC layer
        output=tf.layers.dense(final_state,self.hidden_size*2,activation=tf.nn.tanh)
        
        #5. logits(use linear layer)
        with tf.name_scope("output"): #inputs: A `Tensor` of shape `[batch_size, dim]`.  The forward activations of the input network.
            logits = tf.matmul(output, self.W_projection) + self.b_projection  # [batch_size,num_classes]
        return logits 

def _single_cell(unit_type,
                 num_units,
                 forget_bias,
                 dropout,
                 mode,
                 residual_connection=False,
                 residual_fn=None,
                 trainable=True):
  """Create an instance of a single RNN cell."""
  # dropout (= 1 - keep_prob) is set to 0 during eval and infer
  dropout = dropout if mode == tf.estimator.ModeKeys.TRAIN else 0.0

  # Cell Type
  if unit_type == "lstm":
    single_cell = contrib_rnn.LSTMCell(
        num_units, forget_bias=forget_bias, trainable=trainable)
  elif unit_type == "gru":
    single_cell = contrib_rnn.GRUCell(num_units, trainable=trainable)
  elif unit_type == "layer_norm_lstm":
    single_cell = contrib_rnn.LayerNormBasicLSTMCell(
        num_units,
        forget_bias=forget_bias,
        layer_norm=True,
        trainable=trainable)
  elif unit_type == "nas":
    single_cell = contrib_rnn.NASCell(num_units, trainable=trainable)
  else:
    raise ValueError("Unknown unit type %s!" % unit_type)

  # Dropout (= 1 - keep_prob).
  if dropout > 0.0:
    single_cell = contrib_rnn.DropoutWrapper(
        cell=single_cell, input_keep_prob=(1.0 - dropout))

  # Residual.
  if residual_connection:
    single_cell = contrib_rnn.ResidualWrapper(
        single_cell, residual_fn=residual_fn)

  return single_cell 

def rnn_cell(rnn_cell_size, dropout_keep_prob, residual, is_training=True):
  """Builds an LSTMBlockCell based on the given parameters."""
  dropout_keep_prob = dropout_keep_prob if is_training else 1.0
  cells = []
  for i in range(len(rnn_cell_size)):
    cell = rnn.LSTMBlockCell(rnn_cell_size[i])
    if residual:
      cell = rnn.ResidualWrapper(cell)
      if i == 0 or rnn_cell_size[i] != rnn_cell_size[i - 1]:
        cell = rnn.InputProjectionWrapper(cell, rnn_cell_size[i])
    cell = rnn.DropoutWrapper(
        cell,
        input_keep_prob=dropout_keep_prob)
    cells.append(cell)
  return rnn.MultiRNNCell(cells) 

def _witch_cell(self):
        """
        RNN 类型
        :return: 
        """
        cell_tmp = None
        if self.cell_type == 'lstm':
            cell_tmp = rnn.BasicLSTMCell(self.hidden_unit)
        elif self.cell_type == 'gru':
            cell_tmp = rnn.GRUCell(self.hidden_unit)
        # 是否需要进行dropout
        if self.dropout_rate is not None:
            cell_tmp = rnn.DropoutWrapper(cell_tmp, output_keep_prob=self.dropout_rate)
        return cell_tmp 

def _biLSTMBlock(self, inputs, num_units, scope, seq_len=None, isReuse=False):
        with tf.variable_scope(scope, reuse=isReuse):
            lstmCell = LSTMCell(num_units=num_units)
            dropLSTMCell = lambda: DropoutWrapper(lstmCell, output_keep_prob=self.dropout_keep_prob)
            fwLSTMCell, bwLSTMCell = dropLSTMCell(), dropLSTMCell()
            output = tf.nn.bidirectional_dynamic_rnn(cell_fw=fwLSTMCell,
                                                     cell_bw=bwLSTMCell,
                                                     inputs=inputs,
                                                     # sequence_length=seq_len,
                                                     dtype=tf.float32)
            return output

    # data encoding block ("3.1 Input Encoding" in paper)