Using a data flow graph

TensorFlow has TensorBoard to provide a graphical image of the computation graph. This makes it convenient to understand, debug, and optimize complex neural network programs. TensorBoard can also provide quantitative metrics about the execution of the network. It reads TensorFlow event files, which contain the summary data that you generate while running the TensorFlow Session.