Tensorflow:检查点简单加载 [英] Tensorflow: checkpoints simple load

问题描述

我有一个检查点文件:

checkpoint-20001 checkpoint-20001.meta

如何从这个空间中提取变量，而不必加载以前的模型和启动会话等

how do I extract variables from this space, without having to load the previous model and starting session etc.

我想做类似的事情

cp = load(checkpoint-20001)
cp.var_a

推荐答案

它没有记录，但您可以使用类 tf.train.NewCheckpointReader 检查 Python 中检查点的内容.

It's not documented, but you can inspect the contents of a checkpoint from Python using the class tf.train.NewCheckpointReader.

这是一个使用它的测试用例，因此您可以看到该类是如何工作的.https://github.com/tensorflow//blob/861644c0bcae5d56f7b3f439696eefa6df8580ec/tensorflow/python/training/saver_test.py#L1203

Here's a test case that uses it, so you can see how the class works. https://github.com/tensorflow/tensorflow/blob/861644c0bcae5d56f7b3f439696eefa6df8580ec/tensorflow/python/training/saver_test.py#L1203

由于它不是一个文档化的类，它的 API 在未来可能会发生变化.

Since it isn't a documented class, its API may change in the future.

这篇关于Tensorflow:检查点简单加载的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！