tensorflow 问题汇总，持续更新...

• （1） TypeError: Expected int32, got list containing Tensors of type '_Message' instead.
• （2） TypeError: Expected int32, got of type 'Variable' instead.
   

解决方法：

查了很多都说要重新装TensorFlow，然后找到了下面的解决方法：

https://stackoverflow.com/questions/41813665/tensorflow-slim-typeerror-expected-int32-got-list-containing-tensors-of-type

You just have to update all the line with tf.concat

for example

net = tf.concat(3, [branch_0, branch_1, branch_2, branch_3])

should be changed to

net = tf.concat([branch_0, branch_1, branch_2, branch_3], 3)

•  （3） AttributeError: module 'tensorflow' has no attribute 'mul'.

https://stackoverflow.com/questions/42217059/tensorflowattributeerror-module-object-has-no-attribute-mul

According to the tensorflow 1.0.0 release notes,

tf.mul, tf.sub and tf.neg are deprecated in favor of tf.multiply, tf.subtract and tf.negative.

You'll need to replace tf.mul with tf.multiply.

•  （4） TypeError: Input 'split_dim' of 'Split' Op has type float32 that does not match expected type of int32.

https://github.com/tensorflow/models/issues/933

The documentation clearly indicates the calling order for tf.split should be:

tf.split(split_dim, num_split, value, name='split')

But the actual code indicates:

def split(value, num_or_size_splits, axis=0, num=None, name="split"):

• （5）AttributeError: module 'tensorflow.tools.api.generator.api.nn' has no attribute 'rnn'

在新版本中，tf.nn.rnn 被 tf.nn.static_rnn 替换。

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

tf.nn.softmax_cross_entropy_with_logits(logits_expanded,labels)

更换为：

tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=logits_expanded,labels=labels))

• （7）python版本问题：

Is generator.next() visible in python 3.0?

But instead of calling g.__next__(), as Paolo says, use next(g).

• （8）