We might eventually see the code for the models and inference, but I do doubt we’ll see training code or be granted access to training data. Google is pretty bad at this.