ROCm Cluster Validation Suite (CVS) documentation

2026-10-26

2 min read time

Applies to Linux

CVS is a collection of test suites that validate AMD ROCm clusters end to end, from single-node burn-in health tests to cluster-wide distributed training and inference. CVS runs over SSH from a head node, or on managed compute with no SSH between nodes for orchestrator-based suites. Slurm and SPUR are supported; see Run on managed compute.

The component public repository is located at ROCm/cvs.

Install

• Quickstart

• Install CVS

• Upgrade CVS

• Uninstall or downgrade CVS

How to

• Configure the cluster file

• Configure test configuration files

• Run test suites

• Run on managed compute

• Run cluster commands

• Copy files to cluster nodes

• Monitor cluster health

Reference

• Scalability and performance

• CVS cluster file

• Run with containers

• Passwordless SSH

• Configuration files

• Command-line interface reference

To contribute to the documentation, refer to Contributing to ROCm.

You can find licensing information on the License page.