attention_decoder.py

"""
This file defines the decoder.
"""

import tensorflow as tf
from tensorflow.python.ops import array_ops, math_ops, nn_ops, variable_scope


# Note: this function is based on tf.contrib.legacy_seq2seq_attention_decoder, which is now
# outdated. In the future, it would make more sense to write variants on the attention mechanism
# using the new seq2seq library for tensorflow 1.0:
# https://www.tensorflow.org/api_guides/python/contrib.seq2seq#Attention
def attention_decoder(
    decoder_inputs, initial_state, encoder_states, cell, initial_state_attention=False,
    use_coverage=False, prev_coverage=None, entity_tokens=None
):
    """
    Args:
        decoder_inputs:
            A list of 2D Tensors [batch_size x input_size].
        initial_state:
            2D Tensor [batch_size x cell.state_size]. If two layers, then tuple of such tensors.
        encoder_states:
            3D Tensor [batch_size x attn_length x attn_size].
        cell:
            rnn_cell.RNNCell defining the cell function and size.
        initial_state_attention:
            Note that this attention decoder passes each decoder input through a linear layer with
            the previous step's context vector to get a modified version of the input. If
            initial_state_attention is False, on the first decoder step the "previous context
            vector" is just a zero vector. If initial_state_attention is True, we use initial_state
            to (re)calculate the previous step's context vector. We set this to False for
            train/eval mode (because we call attention_decoder once for all decoder steps) and True
            for decode mode (because we call attention_decoder once for each decoder step).
        use_coverage:
            boolean. If True, use coverage mechanism.
        prev_coverage:
            If not None, a tensor with shape (batch_size, attn_length). The previous step's
            coverage vector. This is only not None in decode mode when using coverage.
        entity_tokens:
            optional tensor with shape [batch_size, attn_length]. 1 if token is an entity else 0.
  
    Returns:
        outputs:
            A list of the same length as decoder_inputs of 2D Tensors of shape
            [batch_size x cell.output_size]. The output vectors.
        state:
            The final state of the decoder. A tensor shape [batch_size x cell.state_size], or if
            two layers, then a tuple of such tensors.
        attn_dists:
            A list containing tensors of shape (batch_size,attn_length). The attention
            distributions for each decoder step.
        p_gens:
            List of scalars. The values of p_gen for each decoder step.
        coverage:
            Coverage vector on the last step computed. None if use_coverage=False.
    """
    with variable_scope.variable_scope("attention_decoder"):
        batch_size = encoder_states.get_shape()[0].value
        attn_size = encoder_states.get_shape()[2].value

        # Reshape encoder_states (need to insert a dim) to shape
        # (batch_size, attn_len, 1, attn_size)
        encoder_states = tf.expand_dims(encoder_states, axis=2)

        # To calculate attention, we calculate
        # v^T tanh(W_h h_i + W_s s_t + b_attn)
        # where h_i is an encoder state, and s_t a decoder state.
        # attn_vec_size is the length of the vectors v, b_attn, (W_h h_i) and (W_s s_t).
        # We set it to be equal to the size of the encoder states.
        attention_vec_size = attn_size

        # Get the weight matrix W_h and apply it to each encoder state to get (W_h h_i), the
        # encoder features
        W_h = variable_scope.get_variable("W_h", [1, 1, attn_size, attention_vec_size])
        # shape (batch_size, attn_length, 1, attention_vec_size)
        encoder_features = nn_ops.conv2d(encoder_states, W_h, [1, 1, 1, 1], "SAME")

        # Get the weight vectors v and w_c (w_c is for coverage)
        v = variable_scope.get_variable("v", [attention_vec_size])
        if use_coverage:
            with variable_scope.variable_scope("coverage"):
                w_c = variable_scope.get_variable("w_c", [1, 1, 1, attention_vec_size])

        if prev_coverage is not None:
            # for beam search mode with coverage
            # reshape from (batch_size, attn_length) to (batch_size, attn_len, 1, 1)
            prev_coverage = tf.expand_dims(tf.expand_dims(prev_coverage,2),3)

        def attention(decoder_state, coverage=None):
            """
            Calculate the context vector and attention distribution from the decoder state.
      
            Args:
                decoder_state:
                    state of the decoder
                coverage:
                    Optional. Previous timestep's coverage vector, shape
                    (batch_size, attn_len, 1, 1).
      
            Returns:
                context_vector:
                    weighted sum of encoder_states
                attn_dist:
                    attention distribution
                coverage:
                    new coverage vector. shape (batch_size, attn_len, 1, 1)
            """
            if not isinstance(decoder_state, tf.contrib.rnn.LSTMStateTuple):
                # Is two layer LSTM. Use just the top layer state to compute attention.
                decoder_state = decoder_state[1]

            with variable_scope.variable_scope("Attention"):
                # Pass the decoder state through a linear layer (this is W_s s_t + b_attn in the
                # paper)
                # shape (batch_size, attention_vec_size)
                decoder_features = linear(decoder_state, attention_vec_size, bias=True)
                # reshape to (batch_size, 1, 1, attention_vec_size)
                decoder_features = tf.expand_dims(tf.expand_dims(decoder_features, 1), 1)

                if use_coverage and coverage is not None:
                    # Multiply coverage vector by w_c to get coverage_features.
                    # c has shape (batch_size, attn_length, 1, attention_vec_size)
                    coverage_features = nn_ops.conv2d(coverage, w_c, [1, 1, 1, 1], "SAME")

                    # Calculate v^T tanh(W_h h_i + W_s s_t + w_c c_i^t + b_attn)
                    # shape (batch_size, attn_length)
                    e = math_ops.reduce_sum(
                        v * math_ops.tanh(encoder_features + decoder_features + coverage_features),
                        [2, 3]
                    )

                    # Take softmax of e to get the attention distribution
                    # shape (batch_size, attn_length)
                    attn_dist = nn_ops.softmax(e)
                    if entity_tokens is not None:
                        attn_dist *= entity_tokens
                        attn_sum = tf.reduce_sum(attn_dist, axis=1, keep_dims=True)
                        attn_dist /= attn_sum

                    # Update coverage vector
                    coverage += array_ops.reshape(attn_dist, [batch_size, -1, 1, 1])

                else:
                    # Calculate v^T tanh(W_h h_i + W_s s_t + b_attn)
                    e = math_ops.reduce_sum(
                        v * math_ops.tanh(encoder_features + decoder_features),
                        [2, 3]
                    )

                    # Take softmax of e to get the attention distribution
                    # shape (batch_size, attn_length)
                    attn_dist = nn_ops.softmax(e)
                    if entity_tokens is not None:
                        attn_dist *= entity_tokens
                        attn_sum = tf.reduce_sum(attn_dist, axis=1, keep_dims=True)
                        attn_dist /= attn_sum

                    if use_coverage:
                        # first step of training
                        coverage = tf.expand_dims(tf.expand_dims(attn_dist, 2), 2)

                # Calculate the context vector from attn_dist and encoder_states
                # shape (batch_size, attn_size)
                context_vector = math_ops.reduce_sum(
                    array_ops.reshape(attn_dist, [batch_size, -1, 1, 1]) * encoder_states,
                    [1, 2]
                )
                context_vector = array_ops.reshape(context_vector, [-1, attn_size])

            return context_vector, attn_dist, coverage

        outputs = []
        attn_dists = []
        p_gens = []
        state = initial_state
        # initialize coverage to None or whatever was passed in
        coverage = prev_coverage
        context_vector = array_ops.zeros([batch_size, attn_size])
        # Ensure the second shape of attention vectors is set.
        context_vector.set_shape([None, attn_size])

        if initial_state_attention:
            # true in decode mode
            # Re-calculate the context vector from the previous step so that we can pass it
            # through a linear layer with this step's input to get a modified version of the input
            # in decode mode, this is what updates the coverage vector
            context_vector, _, coverage = attention(initial_state, coverage)
        for i, inp in enumerate(decoder_inputs):
            tf.logging.info("Adding attention_decoder timestep %i of %i", i, len(decoder_inputs))
            if i > 0:
                variable_scope.get_variable_scope().reuse_variables()

            # Merge input and previous attentions into one vector x of the same size as inp
            input_size = inp.get_shape().with_rank(2)[1]
            if input_size.value is None:
                raise ValueError("Could not infer input size from input: %s" % inp.name)
            x = linear([inp] + [context_vector], input_size, bias=True)

            # Run the decoder RNN cell. cell_output = decoder state
            cell_output, state = cell(x, state)

            # Run the attention mechanism.
            if i == 0 and initial_state_attention:  # always true in decode mode
                # you need this because you've already run the initial attention(...) call
                with variable_scope.variable_scope(variable_scope.get_variable_scope(), reuse=True):
                    # don't allow coverage to update
                    context_vector, attn_dist, _ = attention(state, coverage)
            else:
                context_vector, attn_dist, coverage = attention(state, coverage)
            attn_dists.append(attn_dist)

            # Calculate p_gen
            with tf.variable_scope('calculate_pgen'):
                if isinstance(state, tf.contrib.rnn.LSTMStateTuple):
                    p_gen_c, p_gen_h = state.c, state.h
                else:
                    # Is two layer LSTM. Just use top layer's state.
                    p_gen_c, p_gen_h = state[1].c, state[1].h

                p_gen = linear([context_vector, p_gen_c, p_gen_h, x], 1, bias=True)
                p_gen = tf.sigmoid(p_gen)
                p_gens.append(p_gen)

            # Concatenate the cell_output (= decoder state) and the context vector, and pass them
            # through a linear layer. This is V[s_t, h*_t] + b in the paper.
            with variable_scope.variable_scope("AttnOutputProjection"):
                output = linear([cell_output] + [context_vector], cell.output_size, bias=True)
            outputs.append(output)

        # If using coverage, reshape it
        if coverage is not None:
            coverage = array_ops.reshape(coverage, [batch_size, -1])

        return outputs, state, attn_dists, p_gens, coverage


def linear(args, output_size, bias, bias_start=0.0, scope=None):
    """
    Linear map: sum_i(args[i] * W[i]), where W[i] is a variable.
  
    Args:
        args:
            a 2D Tensor or a list of 2D, batch x n, Tensors.
        output_size:
            int, second dimension of W[i].
        bias:
            boolean, whether to add a bias term or not.
        bias_start:
            starting value to initialize the bias; 0 by default.
        scope:
            VariableScope for the created subgraph; defaults to "Linear".
  
    Returns:
        A 2D Tensor with shape [batch x output_size] equal to sum_i(args[i] * W[i]), where
        W[i]s are newly created matrices.
  
    Raises:
        ValueError:
            if some of the arguments has unspecified or wrong shape.
    """
    if args is None or (isinstance(args, (list, tuple)) and not args):
        raise ValueError("`args` must be specified")
    if not isinstance(args, (list, tuple)):
        args = [args]

    # Calculate the total size of arguments on dimension 1.
    total_arg_size = 0
    shapes = [a.get_shape().as_list() for a in args]
    for shape in shapes:
        if len(shape) != 2:
            raise ValueError("Linear is expecting 2D arguments: %s" % str(shapes))
        if not shape[1]:
            raise ValueError("Linear expects shape[1] of arguments: %s" % str(shapes))
        else:
            total_arg_size += shape[1]

    # Now the computation.
    with tf.variable_scope(scope or "Linear"):
        matrix = tf.get_variable("Matrix", [total_arg_size, output_size])
        if len(args) == 1:
            res = tf.matmul(args[0], matrix)
        else:
            res = tf.matmul(tf.concat(axis=1, values=args), matrix)
        if not bias:
            return res
        bias_term = tf.get_variable(
            "Bias", [output_size], initializer=tf.constant_initializer(bias_start))

    return res + bias_term