tensorflow RuntimeError:Graph is finalized and cannot be modified

有可能的致错原因

   tf.summary.scalar('loss',loss)
    merged = tf.summary.merge_all()
 
    #save
    saver = tf.train.Saver(tf.global_variables())
    sv = tf.train.Supervisor(logdir=xx,saver=saver）

    with sv.managed_session(config=config) as sess:

这样写没有报上述错误

然而：下述顺序会报错

    tf.summary.scalar('loss',loss)

    #save
    saver = tf.train.Saver(tf.global_variables())
    sv = tf.train.Supervisor(logdir=xx,saver=saver）

    merged = tf.summary.merge_all()

    with sv.managed_session(config=config) as sess:

   可能的原因是：https://stackoverflow.com/questions/41798311/tensorflow-graph-is-finalized-and-cannot-be-modified中提到The root cause for your error seems to be that MonitoredTrainingSession has finalized (frozen) the graph and your tf.global_variable_initializer() is no longer able to modify it. 

可能是saver中涉及冻结图标的操作，但是这和merge_all有何关系扔不懂:-D....