Debugging tensorflow Source code

[saras26]

Hi everyone, I am trying to understand the internals of TensorFlow in-depth, the only way I could figure out was to use a python debugger and navigate into the code creating breakpoints in my example script, however, I was wondering if there is any other way that would be more intensive to understand in-depth, I am just looking for another supporting debugging method or something like that to do this in-depth analysis.Please let me know if you guys used any other ways or did the same as above .Is there also a way to understand the core c++ functions while performing the debugging as mentioned above

Thank you in advance

[Aniket_Dubey]

Hi @saras26 ,

@saras26

I am just looking for another supporting debugging method or something like that to do this in-depth analysis.

• Leverage TensorBoard debugger to monitor training progress and inspect intermediate activations, gradients, and other data within your model during training. This can provide valuable clues about how TensorFlow processes information within your script.

@saras26

Is there also a way to understand the core c++ functions while performing the debugging as mentioned above

• The official TensorFlow C++ API reference documents the core C++ functions used internally. By studying the reference and understanding the function signatures and functionalities, you can gain insights into how Python interacts with the C++ backend.

• The TensorFlow GitHub repository holds the source code for both the Python and C++ parts. Exploring the relevant C++ files can give you a deeper understanding of the underlying implementation details

Thank You .