Hands-On Machine Learning with TensorFlow.js

Simple save

You should use this API if you do not have any requirement to do complicated things such as saving multiple models in one file. It is a simple and quick way to export your model in SavedModel format. The parameters to be given to the simple save API are as follows:

• Session
• Input tensors
• Output tensors

The code looks as follows. Your operation graph should be constructed in the session (sess):

import tensorflow as tfwith tf.Session() as sess:  tf.saved_model.simple_save(sess,            './my_tensorflow_model',            inputs={"x": x, "y": y},            outputs={"z": z})

You will see the my_tensorflow_model directory in your current directory where the program runs. It should contain a graph definition in the protocol buffer format and variables in the binary ...