Skip to content

Navigation Menu

Appearance settings

GitHub Copilot
Write better code with AI

GitHub Spark New
Build and deploy intelligent apps

GitHub Models New
Manage and compare prompts

GitHub Advanced Security
Find and fix vulnerabilities

Actions
Automate any workflow
Codespaces
Instant dev environments

Issues
Plan and track work

Code Review
Manage code changes

Discussions
Collaborate outside of code

Code Search
Find more, search less
Explore

Why GitHub

Documentation

GitHub Skills

Blog
Integrations

GitHub Marketplace

MCP Registry
View all features
By company size

Enterprises

Small and medium teams

Startups

Nonprofits
By use case

App Modernization

DevSecOps

DevOps

CI/CD

View all use cases
By industry

Healthcare

Financial services

Manufacturing

Government

View all industries
View all solutions
Topics

AI

DevOps

Security

Software Development

View all
Explore

Learning Pathways

Events & Webinars

Ebooks & Whitepapers

Customer Stories

Partners

Executive Insights
GitHub Sponsors
Fund open source developers
The ReadME Project
GitHub community articles
Repositories

Topics

Trending

Collections
Enterprise platform
AI-powered developer platform
Available add-ons

GitHub Advanced Security
Enterprise-grade security features

Copilot for business
Enterprise-grade AI features

Premium Support
Enterprise-grade 24/7 support
Pricing

Search code, repositories, users, issues, pull requests...

Search

Clear

Search syntax tips

Provide feedback

We read every piece of feedback, and take your input very seriously.

Include my email address so I can be contacted

Saved searches

Use saved searches to filter your results more quickly

Name

Query

To see all available qualifiers, see our documentation.

Appearance settings

You signed in with another tab or window. Reload to refresh your session. You signed out in another tab or window. Reload to refresh your session. You switched accounts on another tab or window. Reload to refresh your session.

Dismiss alert

flash-algo / flash-sparse-attention Public

Notifications You must be signed in to change notification settings
Fork 39
Star 456

Code
Issues 7
Pull requests 14
Discussions
Actions
Projects
Wiki
Security
Insights

Additional navigation options

Code
Issues
Pull requests
Discussions
Actions
Projects
Wiki
Security
Insights

Add specialized CUDA kernels for multi-head attention with various head dimensions#30

Merged

LoserCheems merged 51 commits intomainflash-algo/flash-sparse-attention:mainfrom

Support-sparse_gemmflash-algo/flash-sparse-attention:Support-sparse_gemmCopy head branch name to clipboard

Jun 26, 2025

Conversation Commits51 (51)Checks Files changed

Merged

Add specialized CUDA kernels for multi-head attention with various head dimensions#30
LoserCheems merged 51 commits intomainfrom
Support-sparse_gemm

Commits

Commits on Jun 26, 2025

Adds specialized causal attention kernel for bf16 hdim32
LoserCheems
committed
Adds specialized kernel for 32-dim bfloat16 forward pass
LoserCheems
committed
Adds causal flash attention kernel for 32-dim heads
LoserCheems
committed
Adds specialized kernel for 32-dim heads with FP16
LoserCheems
committed
Adds specialized kernel for 64-dim causal attention
LoserCheems
committed
Add bfloat16 forward kernel for 64-dim heads on SM80
LoserCheems
committed
Adds specialized kernel for causal attention with 64-dim heads
LoserCheems
committed
Adds FP16 forward kernel for 64-dim heads on SM80
LoserCheems
committed
Adds specialized CUDA kernel for bfloat16 causal attention
LoserCheems
committed
Adds specialized flash attention kernel for hdim96 bf16
LoserCheems
committed
Adds FP16 causal forward kernel for 96 head dimension
LoserCheems
committed
Adds Flash Attention forward kernel for 96-dim heads
LoserCheems
committed
Adds specialized kernel for bfloat16 causal attention
LoserCheems
committed
Adds specialized kernel for bf16 hdim128 forward pass
LoserCheems
committed
Adds specialized kernel for 128-dim causal attention
LoserCheems
committed
Adds specialized FP16 kernel for head dimension 128
LoserCheems
committed
Adds bfloat16 causal flash attention kernel for 192 head dim
LoserCheems
committed
Adds specialized CUDA kernel for bfloat16 head dimension 192
LoserCheems
committed
Adds FP16 causal flash attention kernel for 192 head dim
LoserCheems
committed
Adds specialized Flash Attention kernel for 192 head dimension
LoserCheems
committed
Adds specialized CUDA kernel for bfloat16 causal attention
LoserCheems
committed
Adds flash attention forward kernel for head dimension 256
LoserCheems
committed
Adds FP16 causal flash attention kernel for 256 head dim
LoserCheems
committed
Adds specialized kernel for head dimension 256 with FP16
LoserCheems
committed
Adds specialized kernel for 32-dim BF16 causal attention
LoserCheems
committed
Adds split kernel for bfloat16 hdim32 forward pass
LoserCheems
committed
Adds split kernel for FP16 causal attention
LoserCheems
committed
Adds auto-generated kernel for FP16 SM80 hdim32
LoserCheems
committed
Adds split kernel for bfloat16 causal attention
LoserCheems
committed
Adds split kernel for bfloat16 head dimension 64
LoserCheems
committed
Adds CUDA kernel for 64-dim causal attention
LoserCheems
committed
Adds split kernel file for 64-dim heads with FP16
LoserCheems
committed
Adds specialized kernel for 96-dim causal attention
LoserCheems
committed
Adds split kernel for bfloat16 head dimension 96
LoserCheems
committed
Adds specialized kernel for hdim96 fp16 causal attention
LoserCheems
committed
Adds split kernel for head dimension 96 with FP16
LoserCheems
committed
Adds split kernel for bf16 causal attention
LoserCheems
committed
Adds split kernel for bf16 hdim128 on SM80
LoserCheems
committed
Adds specialized CUDA kernel for FP16 causal attention
LoserCheems
committed
Adds split kernel for fp16 hdim128 on SM80
LoserCheems
committed
Adds specialized kernel for head dimension 192 with bfloat16
LoserCheems
committed
Adds specialized kernel for head dimension 192 with bfloat16
LoserCheems
committed
Adds specialized kernel for head dimension 192 with FP16
LoserCheems
committed
Adds split kernel for head dimension 192 with FP16
LoserCheems
committed
Adds specialized kernel for hdim256 bf16 causal attention
LoserCheems
committed
Adds split kernel for head dimension 256 with bfloat16
LoserCheems
committed
Adds specialized kernel for 256-dim causal attention
LoserCheems
committed
Adds split kernel for head dimension 256 with FP16
LoserCheems
committed
Enables fwd_split kernel generation and fixes data types
LoserCheems
committed
Changes is_causal field type from bool to str
LoserCheems
committed
Changes return type from List to Generator for memory efficiency
LoserCheems
committed

Footer

© 2025 GitHub, Inc.

Footer navigation

Terms
Privacy
Security
Status
Community
Docs
Contact

You can’t perform that action at this time.