将密层输出到lstm单元时出现问题

Question

我想将形状为（1,70,70）的隐藏层的输出张量馈送到lstm层。 lstm层具有70个单位和1个时间步。如何调整张量以使其与张量流中的lstm输入兼容。     input_layer = layer.dense（self.xs，70，activation ='relu'，kernel_initializer ='random_uniform'）

    hidden1 = layer.dense(input_layer, 70, activation='relu', 
kernel_initializer='random_uniform')
    hidden2 = layer.dense(hidden1, 70, activation='relu', 
kernel_initializer='random_uniform')
    hidden3 = layer.dense(hidden2, 70, activation='relu', 
kernel_initializer='random_uniform')
    hidden4 = layer.dense(hidden3, 70, activation='relu', 
kernel_initializer='random_uniform')

    lstm_cell_1 = tf.contrib.rnn.BasicLSTMCell(self.num_units, forget_bias=1.0, state_is_tuple=True)
    lstm_cell_2 = tf.contrib.rnn.BasicLSTMCell(self.num_units, forget_bias=1.0, state_is_tuple=True)
    lstm_cell = tf.contrib.rnn.MultiRNNCell([lstm_cell_1, lstm_cell_2], state_is_tuple=True)
    initial_state = lstm_cell.zero_state(self.batch_size, dtype=tf.float32)
    with tf.variable_scope('lstm1'):
        outputs, final_state = tf.nn.dynamic_rnn(lstm_cell, hidden4, initial_state=initial_state)