Kubernetes-based GPU orchestration platform fully deployed on-premises, including the control plane and clusters, enabling dynamic AI workload scheduling and resource optimization within your infrastructure.
NVIDIA Run:ai
NVIDIA Run:ai is an AI workload and GPU orchestration platform for teams running training, inference, or development workloads on Kubernetes. It provides advanced scheduling, workload optimization, RBAC, governance, and auditing, so IT and data science teams get more out of their GPUs. It is built for shared GPU clusters used by many teams. You need working Kubernetes knowledge, and installation takes some infrastructure planning.
Open source
This collection provides the NVIDIA Run:ai open-source release. The source code is available on GitHub under the Apache 2.0 license. The collection includes the control plane and cluster components, which you run on your own Kubernetes infrastructure: on premises, in a private cloud, or in a public cloud.
You do not need an NGC account, subscription, or API key to download and use the artifacts in this collection.
Getting Started
Start with the NVIDIA Run:ai overview to understand the platform architecture, system components, and installation options.
The NVIDIA Run:ai repository on GitHub is open for exploration and contribution.
Documentation
Find all the product guides, quickstarts, tutorials, configuration references, and API documentation in the NVIDIA Run:ai Documentation Portal.
Support
NVIDIA Run:ai open source is community-supported. Report issues, ask questions, and contribute on GitHub.
NVIDIA Run:ai is also available through NVIDIA software offerings: NVIDIA AI Enterprise, NVIDIA Mission Control, NCP software and support offers, CSP marketplace listings, and NVIDIA Run:ai standalone. These provide the same platform together with NVIDIA support.
