(1) Define training data: input tensor and target tensor.
(2) Define the network (or model) composed of layers, and map the input to the target.
(3) Configuration learning process: select loss function, optimizer and indicators to be monitored.
(4) Call the fit method of the model to iterate on the training data.
Definition Model There are two methods: one is to use the Sequential class (linear only for the stack of layers, which is the most
common network architecture), the other is a functional API (functional API, there is a layer comprising To acyclic graph, so that you can
build any form of architecture).
