MLflow Tensorflow Integration

Introduction

TensorFlow is an end-to-end open source platform for machine learning. It has a comprehensive, flexible ecosystem of tools, libraries, and community resources that lets researchers push the state-of-the-art in ML and developers easily build and deploy ML-powered applications.

MLflow has built-in support (we call it MLflow Tensorflow flavor) for Tensorflow workflow, at a high level in MLflow we provide a set of APIs for:

  • Simplified Experiment Tracking: Log parameters, metrics, and models during model training.

  • Experiments Management: Store your Tensorflow experiments in MLflow server, and you can view and share them from MLflow UI.

  • Effortless Deployment: Deploy Tensorflow models with simple API calls, catering to a variety of production environments.

Developer Guide of Tensorflow with MLflow

To learn more about the nuances of the tensorflow flavor in MLflow, please read the developer guide. It will walk you through the following topics:

View the Developer Guide
  • Autologging Tensorflow Experiments with MLflow: How to left MLflow autolog Tensorflow experiments, and what metrics are logged.

  • Control MLflow Logging with Keras Callback: For people who don’t like autologging, we offer an option to log experiments to MLflow using a custom Keras callback.

  • Log Your Tensorflow Models with MLflow: How to log your Tensorflow models with MLflow and how to load them back for inference.