Underlying all this are projects like ml_metadata, protobufs, and Beam. If you wished, you could pick and chose to use only those libraries and forget about TFX. I agree even those restrict you to some degree. Its then your decision if the benefits of reusing their code outweigh the risks associated with depending on a "3rd party" library.
"Learning from the pitfall of others is very much important to build your own system"
This presents more ways to lock-in yourself to Tensorflow.
As opposed to what?
Underlying all this are projects like ml_metadata, protobufs, and Beam. If you wished, you could pick and chose to use only those libraries and forget about TFX. I agree even those restrict you to some degree. Its then your decision if the benefits of reusing their code outweigh the risks associated with depending on a "3rd party" library.
Is there sample code available for all of this somewhere?
best place to start is the TFX Developer Tutorial: www.tensorflow.org/tfx/tutorials/tfx/workshop
Where’s the slide view? Oh saw a quick pullback :)
pickled tree in a flask, up and running less than an hour :)