De-Bazel-ing TensorFlow Service

Question

While I admire, and I am somewhat confused, the documentary commitment to mediate everything related to TensorFlow Serving through Bazel, my understanding of this is at best insignificant. I would like to minimize my interaction with him.

I implement my own TF Serving server, adapting the code from the Inception + TF Serving tutorial. I found that the BUILD files are intimidating enough, and instead of doing a long debugging process, I decided to just edit the BUILD to access the file .cc, instead of creating python things that (as I understand it?) Are not strictly necessary.

However, my functional TF Serving installation cannot be imported into python. Using regular TensorFlow, you create a file .whland set it that way; is there something similar you can do with TF Serving? That way, I could save the design and export the models in the python friendly interactive shell area, and not edit it, crossing all available fingers, build in basel, and then /bazel-bin/path/running/whatever.

Just adding a directory to my PYTHONPATH has not yet succeeded.

Thank!

Answer 1

You are close, you need to update the environment, as in this script

... / portion / Basel bin / tensorflow_serving / example / mnist_export

, export PYTHONPATH =...

shadoworflow_serving