Flash Attention provides 2-4x speedup and 10-20x memory reduction for transformer attention through IO-aware tiling and recomputation.
Works with
AI-first code editor with Composer
Before installing skills in Cursor, ensure your development environment meets these requirements:
node --versionoptimizing-attention-flashExecute the skills CLI command in your project's root directory to begin installation:
Fetches optimizing-attention-flash from davila7/claude-code-templates and configures it for Cursor.
The CLI shows a list of agents. Use arrow keys and space to select Cursor:
Confirm successful installation by checking the skill directory location:
Restart Cursor to activate optimizing-attention-flash. Access via /optimizing-attention-flash in your agent's command palette.
We perform automated surface-level scans (Gen AI Scanner, Socket, Snyk) during installation. These checks detect common vulnerabilities but do not guarantee complete security. Always review skill source code and verify the publisher's reputation before production use.
Skills execute code in your environment. Always review source, verify the publisher, and test in isolation before production.
Submit your Claude Code skill and start earning
Create detailed user stories, acceptance criteria, and feature specs
Example
Generate user stories for 'password reset feature' with acceptance criteria, edge cases, and test scenarios
Reduce spec writing time by 50%, ensure comprehensive coverage
Research competitors, compare features, identify gaps
Example
Analyze 5 competitor products, create feature comparison matrix, suggest differentiation opportunities
Complete competitive research in 2 hours instead of 2 days
Evaluate features using frameworks (RICE, ICE, Kano) and create prioritized backlogs
Example
Score 20 feature ideas using RICE framework, generate prioritized roadmap with rationale
0
total installs
0
this week
24.2K
GitHub stars
0
upvotes
Run in your terminal
0
installs
0
this week
24.2K
stars
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+):
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):
pip install flash-attn --no-build-isolation
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)
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
Step 1: Check PyTorch version
python -c "import torch; print(torch.__version__)"
# Should be ≥2.2.0
If <2.2, upgrade:
pip install --upgrade torch
Step 2: Enable Flash Attention backend
Replace standard attention:
# Before (standard attention)
attn_weights = torch.softmax(q @ k.transpose(-2, -1) / math.sqrt(d_k), dim=-1)
out = attn_weights @ v
# After (Flash Attention)
import torch.nn.functional as F
out = F.scaled_dot_product_attention(q, k, v, attn_mask=mask)
Force Flash Attention backend:
with torch.backends.cuda.sdp_kernel(
enable_flash=True,
enable_math=False,
enable_mem_efficient=False
):
out = F.scaled_dot_product_attention(q, k, v)
Step 3: Verify speedup with profiling
import torch.utils.benchmark as benchmark
def test_attention(use_flash):
q, k, v = [torch.randn(2, 8, 2048, 64, device='cuda', dtype=torch.float16) for _ in range(3)]
if use_flash:
with torch.backends.cuda.sdp_kernel(enable_flash=True):
return F.scaled_dot_product_attention(q, k, v)
else:
attn = (q @ k.transpose(-2, -1) / 8.0).softmax(dim=-1)
return attn @ v
# Benchmark
t_flash = benchmark.Timer(stmt='test_attention(True)', globals=globals())
t_standard = benchmark.Timer(stmt='test_attention(False)', globals=globals())
print(f"Flash: {t_flash.timeit(100).mean:.3f}s")
print(f"Standard: {t_standard.timeit(100).mean:.3f}s")
Expected: 2-4x speedup for sequences >512 tokens.
Step 4: Test accuracy matches baseline
# Compare outputs
q, k, v = [torch.randn(1, 8, 512, 64, device='cuda', dtype=torch.float16) for _ in range(3)]
# Flash Attention
out_flash = F.scaled_dot_product_attention(q, k, v)
# Standard attention
attn_weights = torch.softmax(q @ k.transpose(-2, -1) / 8.0, dim=-1)
out_standard = attn_weights @ v
# Check difference
diff = (out_flash - out_standard).abs().max()
print(f"Max difference: {diff:.6f}")
# Should be <1e-3 for float16
For multi-query attention, sliding window, or H100 FP8.
Copy this checklist:
flash-attn Library Setup:
- [ ] Step 1: Install flash-attn library
- [ ] Step 2: Modify attention code
- [ ] Step 3: Enable advanced features
- [ ] Step 4: Benchmark performance
Step 1: Install flash-attn library
# NVIDIA GPUs (CUDA 12.0+)
pip install flash-attn --no-build-isolation
# Verify installation
python -c "from flash_attn import flash_attn_func; print('Success')"
Step 2: Modify attention code
from flash_attn import flash_attn_func
# Input: [batch_size, seq_len, num_heads, head_dim]
# Transpose from [batch, heads, seq, dim] if needed
q = q.transpose(1, 2) # [batch, seq, heads, dim]
k = k.transpose(1, 2)
v = v.transpose(1, 2)
out = flash_attn_func(
q, k, v,
dropout_p=0.1,
causal=True, # For autoregressive models
window_size=(-1, -1), # No sliding window
softmax_scale=None # Auto-scale
)
out = out.transpose(1, 2) # Back to [batch, heads, seq, dim]
Step 3: Enable advanced features
Make data-driven prioritization decisions faster
Draft PRDs, status updates, and stakeholder presentations
Example
Create executive summary of Q3 roadmap, monthly progress report, feature launch announcement
Save 3-5 hours/week on communication overhead
Prerequisites
Time Estimate
30-60 minutes to see productivity improvements
Steps
Common Pitfalls
✓ Do
✗ Don't
💡 Pro Tips
✓ Use when
Use for user story writing, competitive research, roadmap prioritization, stakeholder communication, and PRD drafting. Best for reducing repetitive documentation and research work.
✗ Avoid when
Avoid for strategic product vision (requires deep customer empathy), pricing decisions (needs market and financial expertise), or when face-to-face customer discovery is more valuable than speed.
davila7/claude-code-templates
mattpocock/skills
parcadei/continuous-claude-v3
cursor/plugins
ailabs-393/ai-labs-claude-skills
ailabs-393/ai-labs-claude-skills
optimizing-attention-flash fits our agent workflows well — practical, well scoped, and easy to wire into existing repos.
optimizing-attention-flash reduced setup friction for our internal harness; good balance of opinion and flexibility.
I recommend optimizing-attention-flash for anyone iterating fast on agent tooling; clear intent and a small, reviewable surface area.
optimizing-attention-flash is among the better-maintained entries we tried; worth keeping pinned for repeat workflows.
Keeps context tight: optimizing-attention-flash is the kind of skill you can hand to a new teammate without a long onboarding doc.
Useful defaults in optimizing-attention-flash — fewer surprises than typical one-off scripts, and it plays nicely with `npx skills` flows.
optimizing-attention-flash reduced setup friction for our internal harness; good balance of opinion and flexibility.
optimizing-attention-flash has been reliable in day-to-day use. Documentation quality is above average for community skills.
We added optimizing-attention-flash from the explainx registry; install was straightforward and the SKILL.md answered most questions upfront.
Solid pick for teams standardizing on skills: optimizing-attention-flash is focused, and the summary matches what you get after install.
showing 1-10 of 29