TensorFlow:记住下一批次的LSTM状态(有状态LSTM)

Question

给定训练有素的LSTM模型,我想对单个时间步进行推理,即seq_length = 1在下面的示例中.在每个时间步之后,需要记住内部LSTM(内存和隐藏)状态以用于下一个"批处理".在推理的最开始,init_c, init_h给定输入计算内部LSTM状态.然后将它们存储在LSTMStateTuple传递给LSTM 的对象中.在训练期间,每个时间步都更新此状态.然而,对于推理,我希望state在批次之间保存,即初始状态只需要在开始时计算,然后在每个"批次"(n = 1)之后保存LSTM状态.

我发现这个相关的StackOverflow问题:Tensorflow,在RNN中保存状态的最佳方法？.但是,这仅在以下情况下有效state_is_tuple=False,但TensorFlow很快就会弃用此行为(请参阅rnn_cell.py).Keras似乎有一个很好的包装器可以使有状态的 LSTM成为可能,但我不知道在TensorFlow中实现这一目标的最佳方法.TensorFlow GitHub上的这个问题也与我的问题有关:https://github.com/tensorflow/tensorflow/issues/2838

有关构建有状态LSTM模型的任何好建议吗？

inputs  = tf.placeholder(tf.float32, shape=[None, seq_length, 84, 84], name="inputs")
targets = tf.placeholder(tf.float32, shape=[None, seq_length], name="targets")

num_lstm_layers = 2

with tf.variable_scope("LSTM") as scope:

    lstm_cell  = tf.nn.rnn_cell.LSTMCell(512, initializer=initializer, state_is_tuple=True)
    self.lstm  = tf.nn.rnn_cell.MultiRNNCell([lstm_cell] * num_lstm_layers, state_is_tuple=True)

    init_c = # compute initial LSTM memory state using contents in placeholder 'inputs'
    init_h = # compute initial LSTM hidden state using contents in placeholder 'inputs'
    self.state = [tf.nn.rnn_cell.LSTMStateTuple(init_c, init_h)] * num_lstm_layers

    outputs = []

    for step in range(seq_length):

        if step != 0:
            scope.reuse_variables()

        # CNN features, as input for LSTM
        x_t = # ... 

        # LSTM step through time
        output, self.state = self.lstm(x_t, self.state)
        outputs.append(output)

Answer 1

我发现在占位符中保存所有图层的整个状态是最简单的.

init_state = np.zeros((num_layers, 2, batch_size, state_size))

...

state_placeholder = tf.placeholder(tf.float32, [num_layers, 2, batch_size, state_size])

然后解压缩并创建一个LSTMStateTuples元组,然后再使用本机tensorflow RNN Api.

l = tf.unpack(state_placeholder, axis=0)
rnn_tuple_state = tuple(
[tf.nn.rnn_cell.LSTMStateTuple(l[idx][0], l[idx][1])
 for idx in range(num_layers)]
)

RNN通过API:

cell = tf.nn.rnn_cell.LSTMCell(state_size, state_is_tuple=True)
cell = tf.nn.rnn_cell.MultiRNNCell([cell]*num_layers, state_is_tuple=True)
outputs, state = tf.nn.dynamic_rnn(cell, x_input_batch, initial_state=rnn_tuple_state)

然后state- 变量将作为占位符被送到下一批.

Answer 2

Tensorflow,在RNN中保存状态的最佳方法？实际上是我原来的问题.下面的代码是我如何使用状态元组.

with tf.variable_scope('decoder') as scope:
    rnn_cell = tf.nn.rnn_cell.MultiRNNCell \
    ([
        tf.nn.rnn_cell.LSTMCell(512, num_proj = 256, state_is_tuple = True),
        tf.nn.rnn_cell.LSTMCell(512, num_proj = WORD_VEC_SIZE, state_is_tuple = True)
    ], state_is_tuple = True)

    state = [[tf.zeros((BATCH_SIZE, sz)) for sz in sz_outer] for sz_outer in rnn_cell.state_size]

    for t in range(TIME_STEPS):
        if t:
            last = y_[t - 1] if TRAINING else y[t - 1]
        else:
            last = tf.zeros((BATCH_SIZE, WORD_VEC_SIZE))

        y[t] = tf.concat(1, (y[t], last))
        y[t], state = rnn_cell(y[t], state)

        scope.reuse_variables()

而不是使用tf.nn.rnn_cell.LSTMStateTuple我只是创建一个工作正常的列表列表.在这个例子中,我没有保存状态.但是,您可以很容易地从变量中取出状态,只使用assign来保存值.