Should I use tf.keras APIs or TF's low- and mid-level APIs?
Check the chapter and try finding the answer to this.
When should I use tf.keras Sequential and Functional APIs? Why do I need model subclassing?
In general, for simpler models, tf.keras Sequential should be used. Most of models can be written using Sequential APIs. However, for those models that require multiple inputs and outputs and some specific connections, such as residual, Functional APIs should be used. For really customized models, you can use model subclassing.
