Question

所以我一直在阅读有关排队和排队如何在tensorflow中工作的问题，而且我一直在努力创建一个文件名队列并从中拉出来。

不幸的是，如果我在启动线程后立即尝试出列，我会收到错误。是否有一个原因？如果我进入1秒钟的睡眠计时器，它将会出错。如果没有，它有时会起作用，但通常会抛出异常（如下面的代码所示）

import tensorflow as tf
import time
with tf.Graph().as_default():
    filename_list = ['data_batch_{}.mat'.format(i+1) for i in range(5)]
    filename_queue = tf.train.string_input_producer(filename_list)

    with tf.Session() as sess:
        coord = tf.train.Coordinator()
        threads = tf.train.start_queue_runners(sess=sess, coord=coord)

        #time.sleep(1) # If I uncomment this it works
        for i in range(5):
            print(sess.run(filename_queue.dequeue()))

        coord.request_stop()
        coord.join(threads)

抛出异常：

---------------------------------------------------------------------------
NotFoundError                             Traceback (most recent call last)
<ipython-input-28-cf6ab7b71f22> in <module>()
     10         #time.sleep(1)
     11         for i in range(5):
---> 12             print(sess.run(filename_queue.dequeue()))
     13 
     14         coord.request_stop()

/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.pyc in run(self, fetches, feed_dict, options, run_metadata)
    331     try:
    332       result = self._run(None, fetches, feed_dict, options_ptr,
--> 333                          run_metadata_ptr)
    334       if run_metadata:
    335         proto_data = tf_session.TF_GetBuffer(run_metadata_ptr)

/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.pyc in _run(self, handle, fetches, feed_dict, options, run_metadata)
    571     try:
    572       results = self._do_run(handle, target_list, unique_fetches,
--> 573                              feed_dict_string, options, run_metadata)
    574     finally:
    575       # The movers are no longer used. Delete them.

/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.pyc in _do_run(self, handle, target_list, fetch_list, feed_dict, options, run_metadata)
    646     if handle is None:
    647       return self._do_call(_run_fn, self._session, feed_dict, fetch_list,
--> 648                            target_list, options, run_metadata)
    649     else:
    650       return self._do_call(_prun_fn, self._session, handle, feed_dict,

/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.pyc in _do_call(self, fn, *args)
    666         except KeyError:
    667           pass
--> 668       raise type(e)(node_def, op, message)
    669 
    670   def _extend_graph(self):

NotFoundError: FetchOutputs node input_producer_Dequeue:0: not found

Answer 1

感谢您告诉我们这个问题。我已经提交了相应的 GitHub issue，并准备了一个修复程序，它应该很快出现在存储库中。

与此同时，通过在开始会话之前创建单个dequeue()操作，以下代码应该有效：

import tensorflow as tf
with tf.Graph().as_default():
    filename_list = ['data_batch_{}.mat'.format(i+1) for i in range(5)]
    filename_queue = tf.train.string_input_producer(filename_list)
    dequeued_t = filename_queue.dequeue()

    with tf.Session() as sess:
        coord = tf.train.Coordinator()
        threads = tf.train.start_queue_runners(sess=sess, coord=coord)

        for i in range(5):
            print(sess.run(dequeued_t))

        coord.request_stop()
        coord.join(threads)

启动线程失败后立即出列

1 个答案: