# tilegym-monkey-patch-kernels-to-transformers

This skill wires TileGym kernels into Hugging Face Transformers models by swapping components before model creation. Use it to integrate existing kernels or build new cuTile ones.

- skill · `type=skill` · Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, OpenCode · 1600 installs · ★ 3735
- Author: TileGym Team
- License: not stated · policy=link-only · description only; the file stays with the author
- Source: https://github.com/NVIDIA/skills/tree/HEAD/skills/tilegym-monkey-patch-kernels-to-transformers/SKILL.md
- Updated: 2026-10-23
- HTML: https://affpapa.org/en/nevibecode/skills/tilegym-monkey-patch-kernels-to-transformers

## What it does

The skill integrates TileGym kernels with a selected model from `transformers`. It replaces specific modules and classes, and patches initialization, forward, and weight-loading methods before the model is created. The library's source stays untouched; at runtime, the model calls TileGym implementations.

The workflow covers environment setup, integration of existing kernels, creation of cuTile kernels for uncovered PyTorch code, and a final report. Reusable kernels are tracked with FlashInfer-Bench-style metadata. The patches must run before model creation; otherwise, the model components will not be replaced.

To start, give the agent the skill name and the target model ID. It may ask follow-up questions. The agent then follows the environment and integration instructions in the `references` files.

## Install

### Claude Code (native platform)

1. **Skills CLI (Vercel)**
```bash
npx skills add NVIDIA/skills --skill tilegym-monkey-patch-kernels-to-transformers -a claude-code
```

2. **Manually: put the files in place**
- File: `~/.claude/skills/tilegym-monkey-patch-kernels-to-transformers/SKILL.md` — Copy the whole skill folder; for one project use .claude/skills/tilegym-monkey-patch-kernels-to-transformers/

### Codex

1. **Skills CLI (Vercel)**
```bash
npx skills add NVIDIA/skills --skill tilegym-monkey-patch-kernels-to-transformers -a codex
```

2. **Manually: put the files in place**
- File: `.agents/skills/tilegym-monkey-patch-kernels-to-transformers/SKILL.md` — For all projects use ~/.agents/skills/

### Cursor

1. **Skills CLI (Vercel)**
```bash
npx skills add NVIDIA/skills --skill tilegym-monkey-patch-kernels-to-transformers -a cursor
```

2. **Manually: put the files in place**
- File: `.cursor/skills/tilegym-monkey-patch-kernels-to-transformers/SKILL.md` — Cursor also reads .agents/skills/ and .claude/skills/

### GitHub Copilot

1. **GitHub CLI**
```bash
gh skill install NVIDIA/skills tilegym-monkey-patch-kernels-to-transformers
```

2. **Manually: put the files in place**
- File: `.github/skills/tilegym-monkey-patch-kernels-to-transformers/SKILL.md`

### Gemini CLI

1. **Skills CLI (Vercel)**
```bash
npx skills add NVIDIA/skills --skill tilegym-monkey-patch-kernels-to-transformers -a gemini-cli
```

2. **Platform CLI**
```bash
gemini skills install https://github.com/NVIDIA/skills --path skills/tilegym-monkey-patch-kernels-to-transformers
```

### Windsurf (Devin Desktop)

**Manually: put the files in place**
- File: `.devin/skills/tilegym-monkey-patch-kernels-to-transformers/SKILL.md` — The legacy .windsurf/skills/ path still works

### OpenCode

**Skills CLI (Vercel)**
```bash
npx skills add NVIDIA/skills --skill tilegym-monkey-patch-kernels-to-transformers -a opencode
```

### 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/NVIDIA/skills/tree/HEAD/skills/tilegym-monkey-patch-kernels-to-transformers/SKILL.md

## When to use it
- Connect TileGym kernels to a `transformers` model
- Replace model components without editing the library source
- Create cuTile kernels for uncovered PyTorch code
- Check end-to-end model functionality and performance

## Example prompts
- Use this skill to connect TileGym kernels to `Qwen/Qwen3.5-0.8B`.
- Set up the environment and integrate existing TileGym kernels with my `transformers` model.
- Find uncovered PyTorch code and create cuTile kernels by following this skill's workflow.

## Similar
- [skill-creator](https://affpapa.org/en/nevibecode/skills/skill-creator.md) — skill · 414180 installs
- [writing-for-agents](https://affpapa.org/en/nevibecode/skills/writing-for-agents.md) — skill · 466719 installs
- [media-use](https://affpapa.org/en/nevibecode/skills/media-use.md) — skill · 1100000 installs
- [mcp-builder](https://affpapa.org/en/nevibecode/skills/mcp-builder.md) — skill · 130949 installs
- [subagent-driven-development](https://affpapa.org/en/nevibecode/skills/subagent-driven-development.md) — skill · 237746 installs
- [orchestration](https://affpapa.org/en/nevibecode/skills/orchestration.md) — skill · 349800 installs

From the same set: [nvidia/skills](https://affpapa.org/en/nevibecode/by/nvidia/skills.md) — 195 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