Home / orchestra-research / ai-research-skills · GitHub

optimizing-attention-flash SKILL.md skillA

This file is byte-identical to the first copy the registry indexed, one of 3 indexed copies. Same content hash, so the same checks give both the same grade.

optimizing-attention-flash from orchestra-research/ai-research-skills is an agent skill: a SKILL.md folder an AI agent loads when a task calls for it.

Install it, pinned

mdr add orchestra-research/ai-research-skills/optimizing-attention-flash@v1.0.0

Pins v1.0.0 in mdr.lock. Pin the fingerprint instead.

Audit grade: A, 17 of 17 checks passed
Kind: SKILL.md skill
Repository: orchestra-research/ai-research-skills
Path: 10-optimization/flash-attention/SKILL.md
Versions recorded: 5
Latest version: v1.0.0, committed Nov 20, 2025
First version recorded: Nov 19, 2025
Size: 10,201 bytes
Repository stars: 13,718
License: MIT

Optimizes transformer attention with Flash Attention for 2-4x speedup and 10-20x memory reduction. Use when training/running transformers with long sequences (>512 tokens), encountering GPU memory issues with attention, or need faster inference. Supports PyTorch native SDPA, flash-attn library, H100 FP8, and sliding window attention.

Collected from public GitHub. Every version is saved exactly as it was and never changes. Install it at an exact version with the mdr tool, and every file is checked before it reaches your agent. The main file is checked against the SHA-256 fingerprint recorded here, and the others against the fingerprints GitHub keeps for its source commit. The automatic checks below grade the latest version, and the same checks always give the same file the same grade.

What the file says

# Flash Attention - Fast Memory-Efficient Attention

## Quick start

Flash Attention provides 2-4x speedup and 10-20x memory reduction for transformer attention through IO-aware tiling and recomputation.

**PyTorch native (easiest, PyTorch 2.2+)**:
```python
import torch
import torch.nn.functional as F

q = torch.randn(2, 8, 512, 64, device='cuda', dtype=torch.float16)  # [batch, heads, seq, dim]
k = torch.randn(2, 8, 512, 64, device='cuda', dtype=torch.float16)
v = torch.randn(2, 8, 512, 64, device='cuda', dtype=torch.float16)

# Automatically uses Flash Attention if available
out = F.scaled_dot_product_attention(q, k, v)
```

**flash-attn library (more features)**:
```bash
pip install flash-attn --no-build-isolation
```

```python
from flash_attn import flash_attn_func

# q, k, v: [batch, seqlen, nheads, headdim]
out = flash_attn_func(q, k, v, dropout_p=0.0, causal=True)
```

## Common workflows

### Workflow 1: Enable in existing PyTorch model

Copy this checklist:

```
Flash Attention Integration:
- [ ] Step 1: Check PyTorch version (≥2.2)
- [ ] Step 2: Enable Flash Attention backend
- [ ] Step 3: Verify speedup with profiling
- [ ] Step 4: Test accuracy matches baseline
```
…

Read the whole file at its exact version.

How to install

The exact contents, by fingerprint

mdr add orchestra-research/ai-research-skills/optimizing-attention-flash@sha256:e0c1da83766332e0

The line at the top of this page pins the label v1.0.0, which is what most people want. This one pins the sha256 fingerprint of the contents instead, so it survives a label being moved. Either way the exact version is written to mdr.lock, and mdr install downloads and checks those contents again, so it installs them exactly or stops with an error.

Badge

[![mdr](https://markdownregistry.com/badge/art_clfb7j7cuou6fsfx.svg)](https://markdownregistry.com/a/art_clfb7j7cuou6fsfx)

1 badge image fetches in 30 days. What the badge shows.

Versions

The registry has recorded 5 versions of this file, between Nov 19, 2025 and Nov 20, 2025, more than one a day.

version	committed	commit	size	audit
v1.0.0 latest	2025-11-20	0a9ded4	10,201 B	A	view · diff
v1.0.0	2025-11-20	72c8b2c	10,187 B	A	view · diff
v1.0.0	2025-11-20	51fc3a3	10,186 B	A	view · diff
v1.0.0	2025-11-20	81c38fe	10,185 B	A	view · diff
v1.0.0	2025-11-19	bd4461f	10,184 B	A	view

Audit of the latest version

A 17 of 17 checks passed. Fixed automatic checks with no AI involved, so the same checks give the same answer every run.

pass: Frontmatter block present
pass: Frontmatter declares a name
pass: Frontmatter declares a description
pass: Size between 200 bytes and 200 KB (10201 bytes)
pass: No zero-width or bidi control characters
pass: No instruction hidden inside an HTML comment
pass: No link to an exfiltration or paste host
pass: No credential-shaped string
pass: No instruction to send local credentials anywhere
pass: No text hidden with inline styles
pass: No prompt-injection phrasing
pass: No curl or wget piped into a shell
pass: No recursive delete of root, home or parent
pass: No instruction to read or print local credentials
pass: No base64 blob over 200 characters
pass: No link to a raw IP address
pass: No script tag

Source

GitHub

orchestra-research/ai-research-skills · 13,718 stars · license MIT · pushed 2026-10-11 · branch main

API

GET https://markdownregistry.com/api/v1/artifacts/art_clfb7j7cuou6fsfx
GET https://markdownregistry.com/api/v1/resolve?ref=orchestra-research/ai-research-skills/optimizing-attention-flash
GET https://markdownregistry.com/api/v1/blob/e0c1da83766332e0f3db4bcca954c3a70be399632758dcec2680aa82e236438d

Your AI assistant does the legwork. You hear about the deals worth your word. Hand yours the standing instructions at modelranch.com and it joins the network that reads files like this one.

Guides: the SKILL.md format and its required fields · what to check before you install a skill · how to pin a skill to an exact version · where to find agent skills · keeping a team on the same skills · which agent file does what.

More from orchestra-research/ai-research-skills

autoresearch SKILL.md skill

orchestra-research/ai-research-skills · 0-autoresearch-skill/SKILL.md · Orchestrates end-to-end autonomous AI research projects using a two-loop architecture. The inner loop runs rapid…