Run Cluster Validation Suite (CVS) inference and benchmarking test suites

2026-10-21

3 min read time

Applies to Linux

LLM serving, disaggregated prefill/decode, and diffusion workloads for cluster-scale inference validation.

Suite	How to run	Config reference
vLLM	Run vLLM LLM inference tests with Cluster Validation Suite (CVS)	vLLM inference benchmark configuration file for Cluster Validation Suite (CVS)
ATOM	Run ATOM LLM inference benchmarks with Cluster Validation Suite (CVS)	ATOM inference benchmark configuration file for Cluster Validation Suite (CVS)
SGLang	Run SGLang LLM inference benchmarks with Cluster Validation Suite (CVS)	SGLang inference benchmark configuration for Cluster Validation Suite (CVS)
xDiT	Run xDiT diffusion inference tests with Cluster Validation Suite (CVS)	xDiT inference benchmark configuration for Cluster Validation Suite (CVS)

See also Run Cluster Validation Suite (CVS) test suites on AMD Instinct GPU clusters for common cvs run flags and workflow.

Common prerequisites

These prerequisites apply to every inference suite. Each suite page lists only its additional requirements.

On every GPU node:

• Docker, with the SSH user able to run it through passwordless sudo docker or membership in the docker group.

• The host ROCm driver loaded, so /dev/kfd, /dev/dri/*, and /dev/infiniband/* when the workload uses RDMA are available for device passthrough.

• The suite container image loaded on the node or pullable from a reachable registry.

• Model weights on storage mounted at the same path on every participating node.

• A Hugging Face token file when the model requires gated access.

On the machine where you run cvs run:

• CVS installed (see Install Cluster Validation Suite (CVS) on ROCm).

• SSH key access to every cluster node.

Test results

Sample output snapshots from inference and diffusion runs.

Important

The numbers in these sample reports and screenshots (throughput, latency, and every other metric) are illustrative sample values, not final tuned results. They are shown only to illustrate the report layout.

Inference

HTML test report

Run Deck report

The Run Deck dashboard: run card, lifecycle timeline, sweep cells, cross-shape comparison, and full results.

Run Deck interactive viewer

The interactive viewer: token throughput per GPU versus interactivity, and gate margin versus concurrency.

Diffusion

HTML test report

Run Deck report

The Run Deck dashboard: run card and full results.