replaced with by writing regular Python functions. Using the `tf.function`
    decorator, such functions can be turned into graphs which can be executed
    remotely, serialized, and optimized for performance.
*   Unification of `tf.train.Optimizers` and `tf.keras.Optimizers`. Use
    `tf.keras.Optimizers` for TF2.0. `compute_gradients` is removed as public
    API, use `GradientTape` to compute gradients.

 * remain in their original namespace. This reduces the final Gradle distribution
 * size and makes us more conscious of which parts of a library we really need.
 *
 * WARNING: if you decide to do the minification by hand, make sure that you cover all paths of loading classes:
 * reflection, dynamic loading, etc. and understand how the library works internally.
 * These changes might break things in subtle ways otherwise.
 */