In order to serve the models, they need to be saved first. In this section, we demonstrate a slightly modified version of the MNIST example from the official TensorFlow documentation, available at the following link: https://www.tensorflow.org/serving/serving_basic.
The TensorFlow team recommends using SavedModel for saving and restoring models built and trained in TensorFlow. According to the TensorFlow documentation:
SavedModel is a language-neutral, recoverable, hermetic serialization format. SavedModel enables higher-level systems and tools to produce, consume, and transform TensorFlow models.
You can follow along with the code in the Jupyter notebook ch-11b_Saving_TF_Models_with_SavedModel_for_TF_Serving ...
