HAMi Workshop

HAMi Workshop is a community learning resource for HAMi.

Welcome to the HAMi Workshop! This hands-on workshop teaches you how to install, configure, and operate HAMi on Kubernetes.

What You'll Learn

• Concepts: Understand GPU software stack, drivers, and HAMi architecture
• Installation: Deploy HAMi on real clusters or local environments
• GPU Sharing: Share GPUs across multiple workloads
• Scheduling: Control GPU placement with binpack/spread and priority policies
• Monitoring: Observe GPU utilization across your cluster

Prerequisites

• Kubernetes cluster (v1.23+)
• kubectl configured and connected to your cluster
• Nodes with NVIDIA GPUs (or use our Fake GPU for local testing)

Workshop Sections

Concepts

Background knowledge for understanding HAMi.

• GPU Software Stack Overview
• Understanding GPU Drivers
• HAMi Cluster Architecture

Labs

Hands-on exercises.

• Lab 1: Online Installation: Install HAMi on a live cluster.
• Lab 2: Local Fake GPU Setup: Set up HAMi locally on macOS.

About HAMi

HAMi is a Kubernetes GPU virtualization and scheduling middleware. It supports GPU sharing, memory isolation, and fine-grained scheduling.

• GitHub: Source code and contributions.
• Official Website: Documentation and community.