以RNN中的lstm为例： 首先说一个普遍的错误方式，该方式会引起维数不匹配报错！
# 错误方式！！！ lstm_cell = tf.nn.rnn_cell.BasicLSTMCell(num_units) stacked_lstm = tf.nn.rnn_cell.MultiRNNCell([lstm_cell for _ in range(layers_nums)])正确的多层lstm的定义
lstm_cell = [tf.nn.rnn_cell.BasicLSTMCell(num_units) for _ in range(layers_nums)] stacked_lstm = tf.nn.rnn_cell.MultiRNNCell(lstm_cell)