before the first exercise:
For the first part of this assignment, you’ll create a model using TF Keras’ Sequential API, which allows you to build layer by layer, and is ideal for building models where each layer has exactly one input tensor and one output tensor.
how to understand the"one tensor"？
In this exercise, you will learn “Sequential API” and “Functional API”.
“Sequential API” is to write tensorflow/keras APIs as a list, and, expects the output of the previous list entry goes to the next list entry. So, just passing an output to the next layer as an input sequentially. We can not pass an output to two different layers, skip one layers, and so on. In this sense, "
is ideal for building models where each layer has exactly one input tensor and one output tensor
But, sometimes, we want to have two inputs, share outputs for multiple layers, skip outputs as “short cut” and bring to other layer, etc. In that case, we use “Functional API”. In Functional API, each layer is “instantiated” at first. Then, input/output can be specified later to connect each layers.
That’s flexible.
The above is actually what you learn in this exercise. I think you already catch an important point. 