Explore Continuous Integration and Continuous Deployment (CI/CD) for Machine Learning in this 58-minute webinar from Open Data Science. Learn how to automate ML model training and evaluation using CML (Continuous Machine Learning) and best practices from software engineering. Discover techniques for automatically allocating and shutting down cloud instances, generating performance reports in pull/merge requests, transferring data between cloud storage and computing instances, and customizing automation workflows with GitLab CI/CD. Gain insights into common problems, assumptions, and practical implementations of CI/CD for ML projects. Cover topics such as two-step workflows, neural style transfer, GPU usage, automation importance, handling streaming data, running notebooks in the cloud, and comparisons with other ML tools like MLflow and DVC. By the end of this webinar, acquire the knowledge to streamline your ML development process and improve collaboration within data science teams.
Overview
Syllabus
Introduction
Agenda
Common Problems
Assumptions
What is CML
Getting started with CML
Example report
CML Runner
TwoStep Workflow
Do we need reports
CICD workflow
Live demo
Training script
CI configuration
Change hyper parameters
Commit changes
Neural style transfer
CI script
Change style image
Pipelines
GPU
QA
Automation
Is automation important
Can we follow standard practices without CML
How do you configure AWS resources
What if I submit multiple changes near each other
Streaming data
Running notebooks
CloudSpot
Installing CML
CML vs ML Flow
Docker Image
Mobile ML
DVC vs ML Flow
Preventing Emergence
CML vs Terraform
Taught by
Open Data Science
