NVIDIA/srt-slurm

Latest recorded source activity: October 6, 2026 at 21:04

SkillsMP catalog refreshed: November 5, 2026 at 19:57

skills collected: 7

GitHub stars: 95
GitHub forks: 109

Install with an AI assistant

Copy this prompt into the AI assistant you're using.

Follow https://skillsmp.com/skill-install/prompt.md to install Agent Skills from https://github.com/NVIDIA/srt-slurm.

Repository README

Repository README source: README.md

Read the source README

Expand full README

# srtctl Command-line tool for distributed LLM inference benchmarks on SLURM clusters using SGLang, vLLM, TensorRT LLM, TileRT, TokenSpeed and AMD's ATOM. Replace complex shell scripts and 50+ CLI flags with a declarative `schema: 2` YAML recipe: `engine:` names the engine, `roles:` describes each worker role, and `services:` covers everything launched next to the workers. **Documentation:** https://nvidia.github.io/srt-slurm/ (source in [docs/](docs/README.md); agents: [llms.txt](https://nvidia.github.io/srt-slurm/llms.txt)) ## Quick Start ```bash git clone https://github.com/NVIDIA/srt-slurm.git cd srt-slurm uv sync --no-dev # One-time setup: downloads etcd, NATS, uv and the tachometer binaries for the # compute nodes' architecture and writes srtslurm.yaml (account, partition, GPUs per node) make setup ARCH=aarch64 # or ARCH=x86_64 uv run srtctl dry-run -f examples/sglang/dynamo-agg.yaml # render without submitting uv run srtctl apply -f examples/sglang/dynamo-agg.yaml # submit ``` [Installation](docs/installation.md) covers cluster-specific setup; [examples/](examples/README.md) has runnable recipes, one per frontend and topology. ## Let an agent drive it srtctl ships a skill that teaches Claude Code, Codex or Cursor how to set up a checkout, write `srtslurm.yaml`, author and validate recipes, submit them, and read the results. Install it into the checkout and hand the agent the checkout: ```bash uv run srtctl skill --target claude # .claude/skills/srtctl/SKILL.md uv run srtctl skill --target codex # .codex/skills/srtctl/SKILL.md uv run srtctl skill --target cursor # .cursor/rules/srtctl.mdc ``` The `srtctl-mcp` server, the recipe JSON Schema for editors, and `llms.txt` are described in the [documentation home](docs/README.md#for-agents-and-editors). ## Where things are documented - Write a recipe: [Recipe Guide](docs/config-reference.md), then the topic pages it links (engines, topology, frontends, benchmarks, services, ...) - Every field, type, default and allowed value: [Schema Reference](docs/schema-reference.md) (generated from the code) - Every command and flag: [CLI Guide](docs/cli.md) and [CLI Reference](docs/cli-reference.md) (generated) - Run and debug: [Monitoring](docs/monitoring.md), [SLURM FAQ](docs/slurm-faq.md) - Analyze: [Profiling](docs/profiling.md), [DSight](docs/dsight.md) - Old recipes: [Legacy (v1) layout](docs/legacy-v1.md); `srtctl migrate` rewrites them

Loading the repository README…

Skills in this repository

classification pending

skill

occupation

description

updated

srtctl

unclassified

Run srt-slurm (srtctl) inference benchmarks on a Slurm cluster. Use when setting up a checkout, writing or migrating a recipe, submitting a job, or reading its results.

2026-10-06

perf-goal-writer

unclassified

Write or refine a ready-to-use /goal prompt for a performance benchmark, regression investigation or optimization campaign. Define measurable targets, a matched baseline, change boundaries, diagnostic evidence and completion criteria. Use for drafting the…

2026-10-03

add-config-field

unclassified

Add, rename, or re-type a recipe or srtslurm.yaml field in srt-slurm so that it ships complete (schema, resolver, dry-run, docs, example, tests).

2026-10-03

dsight-query

unclassified

Analyze existing DSight reports through the read-only SQLite-backed CLI, Python or MCP query interface. Use when inspecting request latency, lifecycle, metrics, Nsight ranges or CPU samples, comparing runs, or gathering evidence for a dashboard view.

2026-10-03

slurm-job-sizing

unclassified

Size Slurm allocations, warmup and traffic for short debugging jobs or full srt-slurm benchmarks.

2026-10-02

design-rule-sweep

unclassified

Periodic cleanup of srt-slurm drift - shrink the Design Rule baselines, fix stale agent docs, and open small focused PRs.

2026-09-29

validate-without-cluster

unclassified

Verify an srt-slurm change end to end without SLURM or GPUs, using dry-run, the mock orchestrator, and the launch snapshots.

2026-09-29

Showing 7 of 7 collected skills.