python - 调用 sess.run() 时 Tensorflow 挂起

Question

下面的代码会挂起(只有 CTRLz 让我退出)。

import tensorflow as tf
import cifar10 # from https://github.com/tensorflow/models/tree/master/tutorials/image/cifar10 (both cifar10.py & cifar10_input.py)

def main():
    print 'TensorFlow version: ',tf.__version__

    with tf.Session() as sess:

    with tf.device('/cpu:0'):
        images, labels = cifar10.distorted_inputs()

    input = tf.constant([[[1, 2, 3], [5, 5, 5]], [[4, 5, 6], [7, 7, 7]], [[7, 8, 9], [9, 9, 9]]])

    one=input[0]
    print "X1 ",type(input), one
    oneval = sess.run(one)
    print "X2 ",type(one), one, type(oneval), oneval

    two=images[0]
    print "Y1 ",type(images), two
    twoval = sess.run(two)
    print "Y2 ",type(two), two, type(twoval), twoval

main()

我得到以下输出(使用 Python 2.7.5):

[gpu@centos-7-4 demo]$ python demo.py
TensorFlow version:  1.2.1
2017-07-31 16:06:45.503157: W tensorflow/core/platform/cpu_feature_guard.cc:45] >The TensorFlow library wasn't compiled to use SSE4.1 instructions, but these are available on your machine and could speed up CPU computations.
2017-07-31 16:06:45.503182: W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE4.2 instructions, but these are available on your machine and could speed up CPU computations.
2017-07-31 16:06:45.503187: W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use AVX instructions, but these are available on your machine and could speed up CPU computations.
Filling queue with 20000 CIFAR images before starting to train. This will take a few minutes.
X1  class 'tensorflow.python.framework.ops.Tensor'> Tensor("strided_slice:0", shape=(2, 3), dtype=int32)
X2  class 'tensorflow.python.framework.ops.Tensor'> Tensor("strided_slice:0", shape=(2, 3), dtype=int32) <type 'numpy.ndarray'> [[1 2 3] [5 5 5]]
Y1  class 'tensorflow.python.framework.ops.Tensor'> Tensor("strided_slice_1:0", shape=(24, 24, 3), dtype=float32)
^Z

有人有任何建议(或解决方案)吗？

如果有人对背景感兴趣，我的最终目标是将 distorted_inputs() 返回的张量转换为一组 JSON 对象。因此，天真的计划是遍历图像的每个元素并提取值。

Answer 1

我在这里找到了答案Printing tensorflow tensor in Python hangs forever

关键是两行:

coord = tf.train.Coordinator()
threads = tf.train.start_queue_runners(sess=sess, coord=coord)