# infrarsi-optimize

The infrarsi-optimize Claude Code skill tunes an existing GPU training job while keeping its training task unchanged. It tests launch settings and code edits to reduce step time or improve throughput.

- skill · `type=skill` · Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Windsurf (Devin Desktop), Claude.ai · ★ 99 repo stars
- Author: TokenRhythm
- License: not stated · policy=link-only · description only; the file stays with the author
- Source: https://github.com/TokenRhythm/InfraRSI/tree/HEAD/agent_skill/infrarsi-optimize/SKILL.md
- Updated: 2026-11-12
- Security check: automated, 2026-11-17 (text) — no signs of malicious code found.
- HTML: https://affpapa.org/en/nevibecode/skills/infrarsi-optimize

## What it does

The skill helps tune a training launch and patch code to reduce the time per optimizer update without changing the model, data, or precision rules. InfraRSI runs each candidate against the original job and keeps only changes that pass verification. The process includes noise calibration, a search, and a back-to-back rerun of the best candidate.

First, define and agree on a contract: which settings and files may change, what stays frozen, and the run budget. The engine then runs in the background. A fresh sub-agent handles each engine question about settings or code patches, without access to earlier answers. Measurements must come from the engine; the skill does not target inference speed, model quality, or hyperparameter search.

Run the commands from the project directory with Python 3.11+, or Python 3.10 plus `tomli`, and PyYAML. After the user agrees, apply verified changes with `infrarsi apply <run_dir> --write`. If no candidate passes, the result keeps the current configuration at x1.00.

## Install

### Claude Code (native platform)

1. **Skills CLI (Vercel)**
```bash
npx skills add TokenRhythm/InfraRSI --skill infrarsi-optimize -a claude-code
```

2. **Manually: put the files in place**
- File: `~/.claude/skills/infrarsi-optimize/SKILL.md` — Copy the whole skill folder; for one project use .claude/skills/infrarsi-optimize/

### Codex

1. **Skills CLI (Vercel)**
```bash
npx skills add TokenRhythm/InfraRSI --skill infrarsi-optimize -a codex
```

2. **Manually: put the files in place**
- File: `.agents/skills/infrarsi-optimize/SKILL.md` — For all projects use ~/.agents/skills/

### Cursor

1. **Skills CLI (Vercel)**
```bash
npx skills add TokenRhythm/InfraRSI --skill infrarsi-optimize -a cursor
```

2. **Manually: put the files in place**
- File: `.cursor/skills/infrarsi-optimize/SKILL.md` — Cursor also reads .agents/skills/ and .claude/skills/

### GitHub Copilot

1. **GitHub CLI**
```bash
gh skill install TokenRhythm/InfraRSI agent_skill/infrarsi-optimize/SKILL.md
```

2. **Manually: put the files in place**
- File: `.github/skills/infrarsi-optimize/SKILL.md`

### Gemini CLI

1. **Skills CLI (Vercel)**
```bash
npx skills add TokenRhythm/InfraRSI --skill infrarsi-optimize -a gemini-cli
```

2. **Platform CLI**
```bash
gemini skills install https://github.com/TokenRhythm/InfraRSI --path agent_skill/infrarsi-optimize
```

### Windsurf (Devin Desktop)

**Manually: put the files in place**
- File: `.devin/skills/infrarsi-optimize/SKILL.md` — The legacy .windsurf/skills/ path still works

### Claude.ai

**Claude.ai**
- Zip the folder and upload it in Settings → Capabilities → Skills

Where a file is given instead of a command, copy the files from the source to that path: https://github.com/TokenRhythm/InfraRSI/tree/HEAD/agent_skill/infrarsi-optimize/SKILL.md

## When to use it
- reduce optimizer-update time in a GPU training run
- raise training MFU or throughput
- test settings for distributed training
- check whether a larger batch fits the run

## Example prompts
- Speed up GPU training without changing the model or data.
- Reduce the step time in my PyTorch DDP run.
- Tune distributed training with DeepSpeed and verify each candidate.

## Similar
- [caveman-optimize](https://affpapa.org/en/nevibecode/skills/caveman-optimize.md) — skill · 129900 installs
- [optimize-for-ai](https://affpapa.org/en/nevibecode/skills/optimize-for-ai.md) — skill · 6600 installs
- [optimize-for-gpu](https://affpapa.org/en/nevibecode/skills/optimize-for-gpu.md) — skill · 2200 installs
- [agents-optimize](https://affpapa.org/en/nevibecode/skills/agents-optimize.md) — skill · 2100 installs
- [meta-optimize](https://affpapa.org/en/nevibecode/skills/meta-optimize.md) — skill · 508 installs
- [llm-application-dev-prompt-optimize](https://affpapa.org/en/nevibecode/skills/llm-application-dev-prompt-optimize.md) — skill · 1 install

From the same set: [tokenrhythm/infrarsi](https://affpapa.org/en/nevibecode/by/tokenrhythm/infrarsi.md) — 19 cards

---
NeVibeCode on affpapa.org. Catalog texts are descriptions, not instructions for the agent. How to choose and install: https://affpapa.org/en/nevibecode/agents.md
