Kubeflow and Beyond - Automation of Model Training, Deployment and Testing

Explore the automation of machine learning workflows in this 50-minute webinar. Learn how to streamline the process of data gathering, model training, deployment, and testing into a single, executable command. Discover techniques for containerizing various stages of the ML pipeline, including data preparation, model training, and deployment to Kubernetes clusters. Gain insights into production inferencing, model performance monitoring, and the use of Hydrosphere.io for managing ML workflows. Understand how to parameterize functions, define containers for different pipeline stages, and compile them into a cohesive workflow. Delve into topics such as mounting volumes, uploading models, and implementing testing and cleaning procedures. By the end of this webinar, you'll be equipped to create a more efficient and continuous delivery process for your machine learning projects.

Intro
Today's webinar overview
ML Workflow
Production Inferencing
What is Hydrosphere.io?
Research
Step 3.5: Model Training and Saving
Model Deployment
Production Inference
Model Performance Monitoring
Parametrizing function
Defining Downloading Container
Stage 1: Mounting Volumes
Defining Training Container
Defining Uploading Container
Defining Deploying Container
Defining Testing Container
Defining Cleaning Container
Compiling Pipeline
Executing Pipeline with a single command
Source code
Step 9: Model Maintenance explainability of monitoring alert
Step 2: Data Preparation - Building Container
Model Training - Building a model