TensorFlow：使用Attention与BeamSearch的Seq2Seq模型中.clone()方法问题

最近我在TensorFlow 1.6.0里捣鼓Seq2Seq模型，打算借助bidirectional_dynamic_decode、Attention机制以及BeamSearchDecoder来实现。为了简化内容，下面只贴出核心的编码器部分代码：

编码器部分 #

def make_lstm(rnn_size, keep_prob):
    lstm = tf.nn.rnn_cell.LSTMCell(rnn_size, initializer = tf.truncated_normal_initializer(mean = 0.0, stddev = 1.0))
    lstm_dropout = tf.nn.rnn_cell.DropoutWrapper(lstm, input_keep_prob = keep_prob)
    return lstm_dropout

cell_fw = tf.nn.rnn_cell.MultiRNNCell([make...

内容的提问来源于stack exchange，提问作者tmmmmmi