写Tensorflow遇到的bug和常用Python方法总结

1. 问题：ValueError: Only call `softmax_cross_entropy_with_logits` with named arguments (labels=..., logits=..., ...)

解决办法：将tf.nn.softmax_cross_entropy_with_logits(y, out)改写为：

                     tf.nn.softmax_cross_entropy_with_logits(labels=y, logits=out)

2.问题：ValueError: Cannot evaluate tensor using `eval()`

解决办法：将print(x.eval())改写为：
                     with tf.Session() as sess:
                            print(x.eval())

3. 问题：Input ‘split_dim’ of ‘Split’ Op has type float32 that does not match expected type of int32

解决办法：将hsplit = tf.split(0, nsteps, h)改写为：

                      hsplit = tf.split(h, nsteps, 0)

有关slip方法：

import tensorflow as tf

A = [[1, 2, 3], [4, 5, 6]]
x = tf.split(A, 3, 1)

with tf.Session() as sess:
    c = sess.run(x)
    for ele in c:
        print(ele)

打印效果如下：

[[1]
 [4]]
[[2]
 [5]]
[[3]
 [6]]

4.返回数组（List）长度的方法

array = [0,1,2,3,4,5]
print len(array)
6

即用：len(list)，而不是用list.length

5. numpy库数组属性查看：类型、尺寸、形状、维度

import numpy as np  
  
a1 = np.array([1,2,3,4,5,6],dtype=np.float32)  
print(a1)  
print("数据类型",type(a1))           #打印数组数据类型  
print("数组元素数据类型：",a1.dtype) #打印数组元素数据类型  
print("数组元素总数：",a1.size)      #打印数组尺寸，即数组元素总数  
print("数组形状：",a1.shape)         #打印数组形状  
print("数组的维度数目",a1.ndim)      #打印数组的维度数目 

6.问题：AttributeError:_parse_flags

解决办法：将FLAGS._parse_flags()改写为：

                      FLAGS.flag_values_dict()

7.问题：TypeError: can't concat str to bytes

解决办法：将out.write(result.encode('utf-8').strip()+"\n") 更改为：

                     out.write(str(result.encode('utf-8').strip()+b"\n"))

8. 问题：TypeError: write() argument must be str, not bytes

解决办法：将filehandle = open(WAV_FILE, 'w')更改为：
                      filehandle = open(WAV_FILE, 'wb+')

未完待续！！