假设我使用tf.train.Saver()，chkp1和chkp2保存了两个Tensorflow模型。
不幸的是，我没有放置图表的原始代码，是否可以比较两个网络？它们之间有diff之类的东西吗？
我不是在寻找权重差异，而是寻找图形结构差异。
答案 0 :(得分：1)
I do not know of a diff function but you can load them into tensorboard and view the graphs. If you are using anaconda you can open the anaconda prompt and enter
tensorboard --logdir=<path_to_file_containing_checkpoints>
As long as you have the .meta files saved you will be able to view the graph from each run.