NVIDIA Docs Hub Homepage NVIDIA Networking Networking Software CloudAI Benchmark Framework 1.7.1 Workloads

Workloads

This chapter contains automatically generated documentation for all CloudAI workloads. Each workload provides specific functionality for running different types of tests and benchmarks.

Available Workloads

Test	Slurm	Kubernetes	RunAI	Standalone
AIConfigurator	❌	❌	❌	✅
AI Dynamo	✅	✅	❌	❌
Bash Command	✅	❌	❌	❌
Chakra Replay	✅	❌	❌	❌
DDLB	✅	❌	❌	❌
DeepEP Benchmark	✅	❌	❌	❌
DynamoMocker	❌	❌	❌	✅
fio	✅	❌	❌	✅
JaxToolbox workloads (DEPRECATED)	✅	❌	❌	❌
MegatronRun	✅	❌	❌	❌
MegatronBridge	✅	❌	❌	❌
NCCL	✅	✅	✅	❌
NeMo v1.0 aka NemoLauncher (DEPRECATED)	✅	❌	❌	❌
Nemo Run	✅	❌	❌	❌
NIXL Bench	✅	❌	❌	❌
NIXL EP	✅	❌	❌	❌
NIXL KVBench	✅	❌	❌	❌
NIXL CTPerf	✅	❌	❌	❌
Sleep	✅	✅	❌	✅
SGLang	✅	❌	❌	❌
Slurm Container	✅	❌	❌	❌
Triton Inference	✅	❌	❌	❌
UCC	✅	❌	❌	❌
vLLM	✅	❌	❌	❌

Adding New Workloads

To add documentation for a new workload:

Add docstrings to your Python classes and methods.
Create a reStructuredText file in doc/workloads/ (e.g., my_workload.rst).
Add it to the table above.

The documentation will be automatically generated during the build process.

Previous Systems

Next AI Dynamo