Any chance of being able to build models in Go (instead of just running them) in Tensorflow 2.0?