flogo Support multiple versions of Tensorflow

Current behavior: The inference activity supports the concept of multiple frameworks and the current version of the tensorflow implementation is coded against 1.4.x which is has currently been tested against 1.9.x. Tensorflow 1.10 does not currently work with the Inference activity.

Expected behavior: Evaluate the changes in tf 1.10, either implement them against the current tensorflow framework in the inference activity assuming backward compatibility can be maintained. Otherwise, implement a tf1.10 framework and allow the developer to choose their desired implementation at flogo build time.

What is the motivation / use case for changing the behavior? Maintain support for the latest TensorFlow versions.

Additional information you deem important (e.g. I need this tomorrow):

Oct 04 '18 23:10 mellistibco

I am currently running TF1.12 (and I ran TF1.10 without problems as well when it was the most recent) on my machine and it works just fine with Flogo inferencing.

Feb 06 '19 14:02 abramvandergeest

TF2.0 was recently released for python. When THE TF GoLib is released to 2.0 we will need to support it as well.

Dec 11 '19 19:12 abramvandergeest