I.       Overview:

In this course, you will learn about TensorFlow Extended (or TFX), which is Google’s production machine learning platform based on TensorFlow for management of ML pipelines and metadata. The first few modules discuss pipeline components, pipeline orchestration with TFX, how you can automate your pipeline through CI/CD, and how to manage ML metadata. Then we will discuss how to automate and reuse ML pipelines across multiple ML frameworks such as tensorflow, pytorch, scikit learn, and xgboost. You will also learn how to use Cloud Composer to orchestrate your continuous training pipelines, and MLflow for managing the complete machine learning life cycle.

II.     Duration: 01 day

III. Objectives

• Orchestrate model training and deployment with TFX and Cloud AI Platform.
• Operate deployed machine learning models effectively and efficiently.
• Perform continuous training using various frameworks (Scikit Learn, XGBoost, PyTorch) and orchestrate pipelines using Cloud Composer and MLFlow.
• Integrate ML workflows with upstream and downstream data management workflows to maintain end-to-end lineage and metadata management.

IV. Intended Audience

• This course is primarily intended for the following participants:
• Data Scientists looking to deliver business impact by quickly converting from Machine Learning prototype to production.
• Software Engineers looking to develop Machine Learning Engineering skills.
• ML Engineers who want to adopt Google Cloud.

V.    Prerequisites

To get the most out of this course, participants should have:

• Completed Machine Learning with Google Cloud or have equivalent experience.
• Completed MLOps Fundamentals course

VI.  Outline

The course includes presentations, demonstrations, and hands-on labs.

Module 1: Introduction to TFX

• Develop a high-level understanding of TFX standard pipeline components.
• Learn how to use a TFX Interactive Context for prototype development of TFX pipelines.
• Work with the Tensorflow Data Validation (TFDV) library to check and analyze input data.
• Utilize the Tensorflow Transform (TFT) library for scalable data preprocessing and feature transformations.
• Use the KerasTuner library for model hyperparameter tuning.
• Employ the Tensorflow Model Analysis (TFMA) library for model evaluation.

Module 2: Pipeline orchestration with TFX

• Use the TFX CLI and Kubeflow UI to build and deploy TFX pipelines to a hosted AI Platform Pipelines instance on Google Cloud.
• Deploy a TensorFlow model trained using AI Platform Training to AI Platform Prediction.
• Perform advanced distributed hyperparameter tuning using CloudTuner and Cloud AI Platform Vizier.

Module 3: Custom components and CI/CD for TFX pipelines

• Develop a CI/CD workflow with Cloud Build to build and deploy a TFX Pipeline.
• Integrate Github trigger to trigger Cloud Build CI/CD workflow for a TFX pipeline.

Module 4: ML Metadata with TFX

• Access and analyze pipeline artifacts in ML Metadata store.

Module 5: Continuous Training with multiple SDKs, KubeFlow & AI Platform Pipelines

• Perform continuous training with Scikit-learn and AI Platform Pipelines.
• Perform continuous training with PyTorch and AI Platform Pipelines.
• Perform continuous training with XGBoost and AI Platform Pipelines.
• Perform continuous training with TensorFlow and AI Platform Pipelines.

Module 6: Continuous Training with Cloud Composer

• Perform continuous training with Cloud Composer.

Module 7: ML Pipelines with MLflow

• Manage Machine Learning lifecycle with MLflow.

Module 8: Summary

Summarize the course.