Tensorflow Estimator API

What is the Tensorflow Estimator API?

The Tensorflow Estimator API is a powerful tool for creating machine learning models that are both complete and intuitive. With the Estimator API, one can easily train, evaluate, and generate predictions for one’s model. There are two types of Estimators to choose from: pre-made Estimators or custom Estimators that you can write yourself. Best of all, estimators-based models can be run on local or distributed servers, and on CPUs, GPUs, or TPUs without any changes to your model.

Estimators encapsulate four main features: training, evaluation, prediction, and exporting your model for serving. They also come with numerous benefits, including simplified sharing of implementations between model developers, high-level intuitive code for easier model creation, and built-in customization with tf.keras.layers. Plus, Estimators make life easier by building graphs for you and providing a safely distributed training loop that controls all aspects of the training process.