Project-HAMi/HAMi — Heterogeneous GPU Sharing on Kubernetes

Description and use

Project‑HAMi (Heterogeneous GPU Sharing on Kubernetes) is an open‑source framework that enables efficient, fine‑grained sharing of GPU resources across diverse workloads in a Kubernetes cluster. By abstracting heterogeneous GPU hardware and providing a scheduler‑aware device plugin, it lets multiple pods concurrently utilize portions of a single GPU, boosting utilization and simplifying deployment for AI, ML, and graphics workloads. The project also includes tooling and documentation for DevOps teams to integrate, monitor, and manage shared GPU resources at scale.

Problem

Project-HAMi/HAMi helps make deployment and operations more repeatable.

Outcome

Medium: useful for prototypes or internal workflows, with dependency and maintenance checks before production.

Workflow

• standardize deployment
• automate operations
• improve platform reliability

Evidence

Heterogeneous GPU Sharing on Kubernetes

Source