Skip to content

NVIDIA/srt-slurm-recipes

srt-slurm-recipes

This repository is the source of truth for official srt-slurm sweep configurations across node counts, models, GPU platforms, and inference frameworks.

The upstream tool is NVIDIA/srt-slurm. Keep benchmark execution logic in that project; keep reusable sweep configuration in this repository.

Layout

Recipes are organized by benchmark shape:

recipes/
  single-node/
    <model>/
      <gpu>/
        <framework>/
  multi-node/
    <model>/
      <gpu>/
        <framework>/

Each leaf directory is reserved for one or more *.yaml sweep configs for that exact combination.

Current recipes:

Run python scripts/update_recipe_table.py after adding or removing recipe directories. The same check is available as a pre-commit hook at .githooks/pre-commit. Enable it for a checkout with scripts/install_hooks.sh.

Support Matrix

Legend:

  • ✅ Recipe is present (links to containing directory)
  • ⚠️ Recipe is in progress
  • ➖ Not supported (does not fit a single node — use the multi-node recipe, not on InferenceX)

Multi-Node

Model GB200 / GB300 B200 / B300 H100 / H200
DeepSeek-R1
DeepSeek-V4-Pro
GLM-5
gpt-oss-120b
Kimi-K2.5
MiniMax-M2.5
MiniMax-M3

Single-Node

Model B200 / B300 H100 / H200
DeepSeek-R1
DeepSeek-V4-Pro
gpt-oss-120b
Kimi-K2.5
MiniMax-M2.5
MiniMax-M3 ⚠️
NVIDIA-Nemotron-3-Super-120B-A12B NVFP4 FP8
NVIDIA-Nemotron-3-Ultra
Qwen3.5

Running a Sweep

Install srt-slurm from upstream, then point srtctl at a recipe YAML in this repo:

git clone https://github.com/NVIDIA/srt-slurm.git
cd srt-slurm
uv pip install -e .

RECIPES_PATH=/path/to/srt-slurm-recipes
FRAMEWORK=<framework>
uv run srtctl dry-run -f "${RECIPES_PATH}/recipes/single-node/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4/B200/${FRAMEWORK}/1k1k.yaml"
uv run srtctl apply -f "${RECIPES_PATH}/recipes/single-node/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4/B200/${FRAMEWORK}/1k1k.yaml" \
  --tags official,NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4,B200,"${FRAMEWORK}"

The dry-run command should be used before submitting jobs.

Contributing

See CONTRIBUTING.md for how to add or update official sweep recipes.

Support, Security & Governance

License

This repository follows the upstream srt-slurm license. See LICENSE.

About

Official NVIDIA/srt-slurm sweep configs for benchmarking LLMs across NVIDIA GPUs and frameworks, spanning aggregated and disaggregated serving on single- and multi-node setups.

Resources

License

Code of conduct

Contributing

Security policy

Stars

5 stars

Watchers

1 watching

Forks

Releases

No releases published

Packages

 
 
 

Contributors