HAMi
Apache-2.0 CNCF project that shares and slices GPUs and other accelerators across Kubernetes workloads
AI Agents / GPU Scheduling · Apache-2.0 · 4.9k
Reviewed 2026-10-12 · last push 2026-10-19 · repo checked 2026-10-19
What's Good
Apache-2.0, a CNCF Incubating project, release v2.10.0. Installs with Helm as a scheduler extender plus device plugins; pods ask for a slice with resource requests such as GPU memory and core percentage. Supports NVIDIA plus several other accelerator vendors, and needs no changes to your application images. ★~4.7k.
The Catch
It is a scheduling layer for a GPU cluster you already run, not a model server — you still bring vLLM, Triton or whatever serves the model. How limits are enforced depends on each vendor’s device backend, so isolation and overcommit behaviour differ between NVIDIA and other accelerators; test the limits on your hardware before you rely on them.
Verdict
The CNCF way to pack many small GPU jobs onto shared cards — validate isolation per accelerator vendor.
Embed
Maintainer? Add the badge to your README. It is free, links to this review, and there is no paid version.
Markdown
[](https://stackgems.com/gems/hami)
HTML
<a href="https://stackgems.com/gems/hami"><img src="https://stackgems.com/badge/hami.svg" alt="HAMi reviewed on Stack Gems" height="20"></a>