The result is a machine learning framework that is easier to work with—for example, by using the relatively simple Keras API for model training—and more performant. Distributed training is ...