NVIDIA Run:ai Self-Hosted Product Documentation

Find all the product information, step-by-step guides, and reference documentation for deploying and operating NVIDIA Run:ai in on-premises and private cloud environments.

New to NVIDIA Run:ai

Begin with the quick start guides to set up the platform and run your first workloads.

Infrastructure Administrators Install the platform, complete the onboarding wizard, and set up cluster infrastructure

Platform Administrators Configure org structure, access control, resource quotas, and governance policies

AI Practitioners Launch workspaces, run training, and deploy inference

Already Managing a Deployment

Review the latest changes and upgrade guidance before making updates to your environment.

What's New New features, enhancements, and hotfixes

Upgrade Step-by-step instructions to upgrade your control plane and cluster

Product Version Life Cycle Plan upgrades based on support timelines

Workloads (AI Lifecycle )

Step-by-step guides for running workloads at each stage of the AI lifecycle - from interactive experimentation to model training and production deployment.

Workspaces (Experiment) Create an interactive environment for model development and experimentation

Training (Standard) Submit a standard training workload to build and train your model

Training (Distributed) Run a distributed training workload across multiple nodes

Inference (deploy) Deploy and serve AI models with autoscaling and multi-node support

Supported workload types (via YAML ) Submit supported workload types (such as Dynamo) using a YAML manifest via the UI

Tutorials (end-to-end) Complete workflows that connect workspace → training → deployment