Tensorflow:通过协调器停止线程似乎不可行?

Following a standard prefetching queue, I extended the aformentioned example by some additional validation code, see attached code. That is, every ith training step, the learned model is evaluated on a validation set (several, in my case). The validation set can't be fed through the queue, so one possible idea is to build an additional inference graph using the shared variables., I exten