如何使用tensorflow在两个lstm之间分配重量？

Question

我想在两个lstm（或双向-lstm）之间共享权重。说，我定义了2个lstms：

with tf.variable_scope("lstm_1") as _scope:
    output_fw, output_state_fw = tf.nn.dynamic_rnn(cell = lstm_cell1,...)

with tf.variable_scope("lstm_2") as _scope:
        output_fw, output_state_fw = tf.nn.dynamic_rnn(cell = lstm_cell2,...)

如何在lstm_1和lstm_2之间分享权重？如果让lstm_cell1=lstm_cell2，那该怎么办？如果没有，我该怎么办？ THX。