TFX

TensorFlow Extended (TFX) was released in 2019 as an open source end-to-end (E2E) ML platform for deploying production ML pipelines. TFX is the most widely used, general purpose E2E ML platform at Alphabet, including Google.

Deploying advanced machine learning technology to serve customers and/or business needs requires a rigorous approach and production-ready systems. An ML application in production requires modern software development methodology, as well as issues unique to ML and data science. Hear about the importance of MLOps, the use of ML pipeline architectures for implementing production ML applications, rigorous analysis of model performance and sensitivity, and review Google’s experience with TensorFlow Extended (TFX):

Manage MLOps and deploy Machine Learning to production with the new and improved TFX:

ML engineering for production ML deployments with TFX (TensorFlow Fall 2020 Updates):

Questions

  • What is the main goal of TFX?


Overview

What exactly is this TFX thing? (TensorFlow Extended):

Questions

  • What are the main TFX modules?


Pipelines

How do TFX pipelines work? (TensorFlow Extended):

Questions

  • What makes a component?

  • What is a pipeline?

  • What is an orchestrator and DAG?


Metadata

Why do I need metadata? (TensorFlow Extended)

Questions

  • What is metadata?

  • What is in metadata store?


Components

Distributed Processing and Components (TensorFlow Extended)

Questions

  • What is Apache Beam?

Describe the TFX modules:

  • What is StatisticsGen?

  • What is SchemaGen?

  • What is ExampleValidator?

  • What is Transform?

  • What is Trainer and the Tensorboard?

  • What is Evaluator?

  • What is ModelValidator?

  • What is Pusher?


Model understanding

Model Understanding and Business Reality (TensorFlow Extended)

Questions

  • Why understand the model?

  • What is the ML insights triangle?

  • How to find model problems?