This is the primary Java API for building and training neural networks with TensorFlow. This API deliberately mirrors the overall structure of Python Keras. However, it is intended as a comfortable, idiomatic Java API for developers who may or may not be familiar with Keras.
This API is intended to provide convenient, sensible defaults, while still allowing you to exercise fine control over the details of your model, training, and inference when necessary.
More specifically, the following goals drive API evolution:
-
If either you know how to implement a model in the Python Keras API, or you are reimplementing an existing Python Keras model in Java, you should be able to cleanly and naturally follow the same high-level structure in the framework API.
-
Also, given some familiarity with patterns followed throughout the framework API, you should be able to easily translate every detail of a Python Keras implementation into the framework API.
-
However, the framework API is not intended to literally mimic the Python Keras API. Rather, it should expose the same capabilities in an API that feels natural and idiomatic to a Java programmer who does not know Keras. If we ever find ourselves unable to reconcile this goal with easy translation from Python Keras, we may split out a Keras layer.
-
Also, the framework API should support fine control over all aspects of modeling, training, and inference. Unlike with Python Keras, we want this to feel like staying in the same API rather than diving into a separate layer. But here again, if we are ever unable to reconcile this goal with easy translation from Python Keras, we may split the framework API into two layers.