Building End-to-end Machine Learning Workflows with Kubeflow 1
In this course, you will learn how to effectively build end-to-end scalable, production grade machine learning workflows using Kubeflow.
What you'll learn
Building production grade, scalable machine learning workflows is a complex and time-consuming task. In this course, Building End-to-end Machine Learning Workflows with Kubeflow 1, you will learn to use Kubeflow and discover how it can enable data scientists and machine learning engineers to build end-to-end machine learning workflows and perform rapid experimentation. First, you will delve into performing large scale distributed training. Next, you will explore hyperparameter tuning, model versioning, serverless model serving, and canary rollouts. Finally, you will learn how to build reproducible pipelines using various Kubeflow components, such as notebook server, fairing, metadata, katib, and Kubeflow pipelines. When you are finished with the course, you will be able to build end-to-end workflows for your machine learning and deep learning projects.
Table of contents
- Introduction 1m
- Overview 2m
- Model Development Process and Challenges 2m
- Kubeflow Components for Training 2m
- Fashion-MNIST Training Workflow 2m
- Kubeflow Notebook 1m
- Demo: Setting up Notebook Server with a Pre-built Image 5m
- Demo: Setting up Notebook Server with a Custom Image 6m
- Deep Learning Model Overview for Fashion-MNIST 4m
- Demo: Training in Kubeflow Notebook 8m
- Metadata Overview 1m
- Demo: Metadata Tracking 3m
- Kubeflow Fairing Overview 1m
- Demo: Kubeflow Fairing 4m
- Distributed Training 2m
- Demo: Distributed Training with GPU 4m
- Demo: Distributed Training with TFJob 8m
- Hyperparameter Tuning with Katib 2m
- Demo: Performing Hyperparameter Tuning with Katib 4m
- Summary 2m
- Introduction 1m
- Overview 1m
- Model Serving Process and Challenges 3m
- Kubeflow Components for Serving 1m
- KFServing Overview 1m
- Demo: Serving Model Using KFServing 5m
- Demo: Pre and Post-processing Using KFServing 4m
- Canary Rollout Overview 1m
- Demo: Canary Rollout Using KFServing 3m
- Demo: Performance Monitoring Using KFServing, Prometheus, and Grafana 3m
- Demo: Auto Scaling and Load Testing 2m
- Summary 2m
- Introduction 1m
- Overview 1m
- Machine Learning Workflow Pipeline and Challenges 3m
- Kubeflow Components for Building Pipeline 1m
- Kubeflow Pipeline Overview 2m
- Fashion-MNIST Use Case Pipeline 1m
- Demo: Building Kubeflow Pipeline with Hyperparameter Tuning Step 6m
- Demo: Adding Training Step to Kubeflow Pipeline 3m
- Demo: Adding Serving Step to Kubeflow Pipeline 3m
- Demo: Building Kubeflow Pipeline from Notebook 3m
- Summary 2m
About the author
Abhishek Kumar is a data science consultant, author, and Google Developers Expert (GDE) in machine learning. He holds a master’s degree from the University of California, Berkeley, and has been featured in the "Top 40 under 40 Data Scientist" list. He is also a public speaker and has delivered talks in top data conferences across the globe including Strata Data, AI conference, ODSC, and Fifth Elephant. His focus area is machine learning and deep learning at scale and is also a recipient of the... more