使用tensorflow进行NLP任务，我通常将3D张量作为输入传递，等级为3，例如[batch_size，time_steps，embedding_dimension]
RNN的代码如下：
lstm_cell =tf.contrib.rnn.GRUCell(n_hidden)
lstm_cell = tf.contrib.rnn.DropoutWrapper(lstm_cell, output_keep_prob=dropout)
outputs, _ = tf.nn.dynamic_rnn(cell=lstm_cell, inputs=x,sequence_length=seq_len, dtype=tf.float32,time_major=False) ## outputs is a list of matrices
seq_len的形状为[batch_size]，输出是一个形状的张量（batch_size，rnn_dimension）。这很好。
是否可以输入4D张量的形状：[batch_size，time_steps，num_features，embedding_dimension]并提供形状[batch_size，num_features]的rank2 sequence_length？
作为输出，我期望形状的张量：[num_features，batch_size，rnn_dimension]
是否有人使用排名高于3的输入张量？你将如何调整代码，只使用4级张量直接失败。
Dimension must be 4 but is 3 for 'test_scope/transpose' (op: 'Transpose') with input shapes: [?,156,12,300], [3].
Dimension must be 4 but is 3 for 'test_scope_1/transpose' (op: 'Transpose') with input shapes: [?,12,156,300], [3].