Question

使用Tensorflow 0.12时，MultiRNNCell的工作方式发生了变化，对于初学者，state_is_tuple现在默认设置为True，此外，还有对此进行讨论：

state_is_tuple：如果True，则接受和返回的状态为n - 元组，其中n = len(cells)。如果False，则状态都沿列轴连接。后一种行为很快就会被弃用。

我想知道我是如何使用GRU单元的多层RNN的，到目前为止我的代码是：

def _run_rnn(self, inputs):
        # embedded inputs are passed in here
        self.initial_state = tf.zeros([self._batch_size, self._hidden_size], tf.float32)
        cell = tf.nn.rnn_cell.GRUCell(self._hidden_size)
        cell = tf.nn.rnn_cell.DropoutWrapper(cell, output_keep_prob=self._dropout_placeholder)
        cell = tf.nn.rnn_cell.MultiRNNCell([cell] * self._num_layers, state_is_tuple=False)

        outputs, last_state = tf.nn.dynamic_rnn(
            cell = cell,
            inputs = inputs,
            sequence_length = self.sequence_length,
            initial_state = self.initial_state
        ) 

        return outputs, last_state

我的输入查找单词ID并返回相应的嵌入向量。现在，运行上面的代码我受到以下错误的欢迎：

ValueError: Dimension 1 in both shapes must be equal, but are 100 and 200 for 'rnn/while/Select_1' (op: 'Select') with input shapes: [?], [64,100], [64,200]

我有?的地方在我的占位符中：

def _add_placeholders(self):
        self.input_placeholder = tf.placeholder(tf.int32, shape=[None, self._max_steps])
        self.label_placeholder = tf.placeholder(tf.int32, shape=[None, self._max_steps])
        self.sequence_length = tf.placeholder(tf.int32, shape=[None])
        self._dropout_placeholder = tf.placeholder(tf.float32)

Answer 1

您的主要问题在于initial_state的设置。由于您的州现在是一个元组（更具体地说是LSTMStateTuple），因此您无法直接将其分配给tf.zeros。而是使用，

self.initial_state = cell.zero_state(self._batch_size, tf.float32)

请查看documentation了解更多信息。

要在代码中使用此功能，您需要在feed_dict中传递此 tensor 。做这样的事情，

state = sess.run(model.initial_state)
for batch in batches:
    # Logic to add input placeholder in `feed_dict`
    feed_dict[model.initial_state] = state
    # Note I'm re-using `state` below
    (loss, state) = sess.run([model.loss, model.final_state], feed_dict=feed_dict)

在张量流0.12中使用MultiRNNCell

1 个答案: