tf.estimator is a high-level API for TensorFlow. It is used to simplify machine learning programming by providing the means for the straightforward training, evaluation, predicting, and exporting of models for serving.
Estimators confer many advantages on the TensorFlow developer. It is easier and more intuitive to develop models with Estimators than with low-level APIs. In particular, the same model can be run on a local machine or on a distributed multi-server system. The model is also agnostic to the processor it finds itself on, that is, either CPUs, GPUs, or TPUs. Estimators also simplify the development process by making it easier for model developers to share implementations and, being built on Keras layers, make customization simpler.
Estimators take care of all of the background plumbing that goes into working with a TensorFlow model. They support...
