tensorFlow（cifar10官方例子）部分问题记录

1. cifar10的变量重复定义问题：

argparse.ArgumentError: argument --batch_size: conflicting option string(s): --batch_size

答案：

Solved the problem. One is not supposed to execute python cifar10.py directly, but one should run python cifar10_train.py or python cifar10_eval.py instead.

 

That's not surprising. The error indicated that something was trying to add the same commandline argument twice, which could result if one script didn't 'know' that a parser had already been defined.

简单的说，就是重复定义，将重复定义的内容删除即可。

2. cifar10

ValueError: All shapes must be fully defined: [TensorShape([Dimension(24), Dimension(24), Dimension(3)]), TensorShape([Dimension(None)])]

答案：

官方的bug，目前已经被修复了（2017年1月初修复）。

相关内容：

https://github.com/MartinThoma/algorithms/commit/38ce1f87d6e4396cde64fe831c2ead2507781270