[英]How to print (trace) within tensorflow's source code
I have a small trial to print out something within tensorflow's kernel.我有一个小的尝试来打印 tensorflow 内核中的一些东西。
ex, I would like to print something within例如,我想在里面打印一些东西
https://www.tensorflow.org/api_docs/python/tf/train/GradientDescentOptimizer https://www.tensorflow.org/api_docs/python/tf/train/GradientDescentOptimizer
Then I search for the corresponding sources in my laptop where Ubuntu 16.04 has been installed.然后我在安装了 Ubuntu 16.04 的笔记本电脑中搜索相应的源。
Below are what I found.下面是我发现的。
I think gradient_descent.py within either我认为gradient_descent.py 要么
/usr/local/lib/python3.5/dist-packages/tensorflow/contrib/optimizer_v2 /usr/local/lib/python3.5/dist-packages/tensorflow/contrib/optimizer_v2
or或者
/usr/local/lib/python3.5/dist-packages/tensorflow/python/training /usr/local/lib/python3.5/dist-packages/tensorflow/python/training
are used.被使用。 Hence I modify them both.
因此我修改了它们。
However, I cannot see the modifications.但是,我看不到修改。
I wonder why it failed.我想知道为什么它失败了。
Is there any way the trace Tensorflow source code (I know C & C++ part needs to be re-compile so focus on Python now)?有没有办法跟踪 Tensorflow 源代码(我知道 C & C++ 部分需要重新编译,所以现在专注于 Python)?
Thanks in advance.提前致谢。
Check CurrentStackTrace
API provided by TensorFlow util .检查TensorFlow util提供的
CurrentStackTrace
API。 An example use can be found here .可以在此处找到示例用法。
声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.