Heterogeneous GPU Sharing on Kubernetes

HAMi webiste, join the community https://discord.gg/Amhy7XmbNq

Open-source GPU resource management and observability platform for Kubernetes, built on HAMi.

resource isolator for KAI-scheduler, use hami-core to provide resource isolation inside container for NVIDIA devices

A transparent, in-container GPU resource controller that enforces memory and compute limits by intercepting CUDA calls without application or driver changes.

Kubernetes device plugin for Ascend NPUs, supporting hard and soft NPU slicing with HAMi and Volcano.

Community hub for HAMi governance, membership, meetings, meetups, webinars, talks, and contribution guides.

Fast migration from Device Plugin to DRA for HAMi users.

Project metadata for HAMi - CNCF .project automation

HAMi GPU virtualization demo on GKE: vLLM + YOLOv8 multi-workload inference with fractional A100 sharing, Terraform IaC