使用tensorflow时，我收到了一条TypeError消息

Question

当我使用tensorflow RNN时，我遇到了这个问题。 这是我的包的版本：

Python version: 2.7.12 (default, Nov 19 2016, 06:48:10) 
[GCC 5.4.0 20160609]

Packages versions: 
numpy 1.12.1
tensorflow 1.1.0

代码是：

import tensorflow as tf
import numpy as np
from tensorflow.contrib import rnn

batches = 28

permutations = np.random.permutation(batches)

cell_class = rnn.BasicLSTMCell
cell = cell_class(200)
zero_state = cell.zero_state(batches, tf.float32)

with tf.Session() as sess:
    state = sess.run(zero_state)
    state[np.arange(batches)] = state[permutations]

我收到的消息：

Traceback (most recent call last):
  File "test.py", line 15, in <module>
    state[np.arange(batches)] = state[permutations]
TypeError: only integer scalar arrays can be converted to a scalar index

Answer 1

您可以尝试使用以下行替换zero_state值：

zero_state = cell.zero_state(batches, tf.float32)[0]

这似乎与Numpy的最新版本有关。最近的更改使得将单元素数组视为用于索引的标量的错误。