https://github.com/Dao-AILab/flash-attention
Score: 29.668073443613654
Last synced: 1 day ago
JSON representation
Repository metadata:
Fast and memory-efficient exact attention
- Host: GitHub
- URL: https://github.com/Dao-AILab/flash-attention
- Owner: Dao-AILab
- License: bsd-3-clause
- Created: 2022-05-19T21:22:06.000Z (over 4 years ago)
- Default Branch: main
- Last Pushed: 2026-10-29T03:51:33.000Z (5 days ago)
- Last Synced: 2026-10-29T18:17:20.525Z (4 days ago)
- Language: Python
- Homepage:
- Size: 22.2 MB
- Stars: 25,221
- Watchers: 145
- Forks: 3,194
- Open Issues: 1,394
-
Metadata Files:
- Readme: README.md
- License: LICENSE
- Authors: AUTHORS
- Agents: AGENTS.md
- Claude: CLAUDE.md
Owner metadata:
- Name: Dao AI Lab
- Login: Dao-AILab
- Email:
- Kind: organization
- Description: We are an AI research group led by Prof. Tri Dao
- Website: https://tridao.me/
- Location: United States of America
- Twitter:
- Company:
- Icon url: https://avatars.githubusercontent.com/u/139507659?v=4
- Repositories: 1
- Last Synced at: 2026-10-31T17:34:39.243Z
- Profile URL: https://github.com/Dao-AILab
GitHub Events
Total
- Commit comment event: 10
- Create event: 82
- Delete event: 51
- Discussion event: 2
- Fork event: 683
- Issue comment event: 2170
- Issues event: 584
- Member event: 3
- Pull request event: 379
- Pull request review comment event: 206
- Pull request review event: 271
- Push event: 876
- Release event: 17
- Watch event: 6030
- Total: 11364
Last Year
- Create event: 58
- Delete event: 44
- Discussion event: 2
- Fork event: 141
- Issue comment event: 392
- Issues event: 85
- Member event: 2
- Pull request event: 173
- Pull request review comment event: 124
- Pull request review event: 185
- Push event: 473
- Release event: 1
- Watch event: 1104
- Total: 2784
Committers metadata
Last synced: 2 days ago
Total Commits: 1,577
Total Committers: 219
Avg Commits per committer: 7.201
Development Distribution Score (DDS): 0.413
Commits in past year: 488
Committers in past year: 98
Avg Commits per committer in past year: 4.98
Development Distribution Score (DDS) in past year: 0.742
| Name | Commits | |
|---|---|---|
| Tri Dao | t****q@g****m | 926 |
| Driss Guessous | 3****g | 92 |
| Johnsonms | l****u@g****m | 35 |
| jayhshah | j****h@g****m | 32 |
| Pierce Freeman | p****n@g****m | 25 |
| Reuben Stern | 1****s | 23 |
| Henry Tsang | h****g@m****m | 21 |
| Guilherme Leobas | g****s@g****m | 18 |
| rocking | C****i@a****m | 15 |
| Kirthi Shankar Sivamani | k****i@n****m | 15 |
| Ying Zhang | y****z@m****m | 12 |
| Dan Fu | d****9@g****m | 11 |
| Michael Melesse | m****e@g****m | 9 |
| Lequn Chen | l****n@p****i | 9 |
| Johnny | j****4@g****m | 9 |
| dan_the_3rd | 4****d | 7 |
| Yunwei Li | y****3@g****m | 7 |
| Alkaid | g****o@m****m | 6 |
| oliver könig | o****g@n****m | 6 |
| Jean-Luc Duprat | j****d@a****g | 5 |
| Jane (Yuan) Xu | 3****9 | 5 |
| liangel-02 | l****l@m****m | 5 |
| Markus Hoehnerbach | m****h@m****m | 5 |
| Ted Zadouri | t****i@g****m | 5 |
| timmy-feng | 7****g | 5 |
| Aman Gupta Karmani | a****n@t****t | 5 |
| Xuechen Li | 1****n | 4 |
| Grigory Sizov | g****v@f****m | 4 |
| NanoCode012 | n****o@a****i | 4 |
| Phil Wang | l****s@g****m | 4 |
| and 189 more... | ||
Issue and Pull Request metadata
Last synced: 2 days ago
Total issues: 1,063
Total pull requests: 771
Average time to close issues: 30 days
Average time to close pull requests: 19 days
Total issue authors: 846
Total pull request authors: 311
Average comments per issue: 2.19
Average comments per pull request: 1.21
Merged pull request: 307
Bot issues: 0
Bot pull requests: 0
Past year issues: 94
Past year pull requests: 394
Past year average time to close issues: about 1 month
Past year average time to close pull requests: 20 days
Past year issue authors: 80
Past year pull request authors: 148
Past year average comments per issue: 2.38
Past year average comments per pull request: 1.15
Past year merged pull request: 108
Past year bot issues: 0
Past year bot pull requests: 0
Top Issue Authors
- ziyuhuang123 (33)
- phantaurus (10)
- complexfilter (8)
- endurehero (6)
- FurkanGozukara (6)
- TyGu1 (5)
- gaodaheng (5)
- DefTruth (5)
- goldhuang (5)
- janelu9 (5)
- DianCh (4)
- eric-tc-wong (4)
- ArtificialZeng (4)
- Johnsonms (4)
- foreverpiano (4)
Top Pull Request Authors
- drisspg (122)
- Johnsonms (24)
- rocking5566 (20)
- imitater-dou (19)
- johnnynunez (14)
- guoriyue (13)
- ipiszy (11)
- yunweili3 (11)
- ksivaman (10)
- LucasWilkinson (9)
- jayhshah (8)
- reubenconducts (7)
- micmelesse (7)
- abcdabcd987 (7)
- NanoCode012 (6)
Top Issue Labels
Top Pull Request Labels
- size/M (1)
- bug (1)
- cla:yes (1)
Package metadata
- Total packages: 21
-
Total downloads:
- pypi: 1,194,691 last-month
- conda: 177 total
- Total docker downloads: 120,271
- Total dependent packages: 44 (may contain duplicates)
- Total dependent repositories: 317 (may contain duplicates)
- Total versions: 305
- Total maintainers: 12
pypi.org: flash-attn
Flash Attention: Fast and Memory-Efficient Exact Attention
- Homepage: https://github.com/Dao-AILab/flash-attention
- Documentation: https://flash-attn.readthedocs.io/
- Licenses: BSD License
- Latest release: 2.8.3.post2 (published 20 days ago)
- Last Synced: 2026-10-31T17:39:35.383Z (2 days ago)
- Versions: 79
- Dependent Packages: 44
- Dependent Repositories: 317
- Downloads: 292,988 Last month
- Docker Downloads: 120,271
-
Rankings:
- Stargazers count: 0.335%
- Dependent packages count: 0.475%
- Dependent repos count: 0.836%
- Downloads: 0.894%
- Average: 1.371%
- Forks count: 2.111%
- Docker downloads count: 3.575%
- Maintainers (1)
proxy.golang.org: github.com/Dao-AILab/flash-attention
- Homepage:
- Documentation: https://pkg.go.dev/github.com/Dao-AILab/flash-attention#section-documentation
- Licenses: bsd-3-clause
- Latest release: v2.8.3+incompatible (published about 1 year ago)
- Last Synced: 2026-10-31T17:38:27.475Z (2 days ago)
- Versions: 71
- Dependent Packages: 0
- Dependent Repositories: 0
-
Rankings:
- Dependent packages count: 9.558%
- Average: 10.17%
- Dependent repos count: 10.783%
proxy.golang.org: github.com/dao-ailab/flash-attention
- Homepage:
- Documentation: https://pkg.go.dev/github.com/dao-ailab/flash-attention#section-documentation
- Licenses: bsd-3-clause
- Latest release: v2.8.3+incompatible (published about 1 year ago)
- Last Synced: 2026-10-31T17:38:01.033Z (2 days ago)
- Versions: 71
- Dependent Packages: 0
- Dependent Repositories: 0
-
Rankings:
- Dependent packages count: 9.558%
- Average: 10.17%
- Dependent repos count: 10.783%
pypi.org: pi-flash-attn
Flash Attention: Fast and Memory-Efficient Exact Attention
- Homepage: https://github.com/Dao-AILab/flash-attention
- Documentation: https://pi-flash-attn.readthedocs.io/
- Licenses: BSD License
- Latest release: 2.8.2 (published about 1 year ago)
- Last Synced: 2026-10-31T17:39:10.783Z (2 days ago)
- Versions: 1
- Dependent Packages: 0
- Dependent Repositories: 0
- Downloads: 4 Last month
-
Rankings:
- Stargazers count: 0.482%
- Forks count: 0.771%
- Dependent packages count: 8.696%
- Average: 14.739%
- Dependent repos count: 49.007%
- Maintainers (1)
pypi.org: flash-attn-wheels-test
Flash Attention: Fast and Memory-Efficient Exact Attention
- Homepage: https://github.com/Dao-AILab/flash-attention
- Documentation: https://flash-attn-wheels-test.readthedocs.io/
- Licenses: BSD License
- Latest release: 2.0.8.post17 (published about 3 years ago)
- Last Synced: 2026-10-31T17:38:26.209Z (2 days ago)
- Versions: 14
- Dependent Packages: 0
- Dependent Repositories: 0
- Downloads: 41 Last month
-
Rankings:
- Stargazers count: 0.36%
- Forks count: 2.344%
- Dependent packages count: 7.565%
- Average: 19.918%
- Dependent repos count: 69.402%
- Maintainers (1)
pypi.org: fa4
Flash Attention CUTE (CUDA Template Engine) implementation
- Homepage: https://github.com/Dao-AILab/flash-attention
- Documentation: https://fa4.readthedocs.io/
- Licenses: BSD 3-Clause License
- Latest release: 4.0.0b3 (published 8 months ago)
- Last Synced: 2026-10-31T17:38:42.896Z (2 days ago)
- Versions: 1
- Dependent Packages: 0
- Dependent Repositories: 0
- Downloads: 28 Last month
-
Rankings:
- Dependent packages count: 7.716%
- Average: 25.672%
- Dependent repos count: 43.628%
- Maintainers (1)
pypi.org: sgl-fa4
Flash Attention CUTE (CUDA Template Engine) implementation
- Homepage: https://github.com/Dao-AILab/flash-attention
- Documentation: https://sgl-fa4.readthedocs.io/
- Licenses: BSD 3-Clause License
- Latest release: 4.0.5 (published 8 months ago)
- Last Synced: 2026-10-31T17:38:00.911Z (2 days ago)
- Versions: 5
- Dependent Packages: 0
- Dependent Repositories: 0
- Downloads: 40,421 Last month
-
Rankings:
- Dependent packages count: 7.729%
- Average: 25.717%
- Dependent repos count: 43.704%
- Maintainers (2)
pypi.org: tokenspeed-fa4
Flash Attention CUTE (CUDA Template Engine) implementation
- Homepage: https://github.com/Dao-AILab/flash-attention
- Documentation: https://tokenspeed-fa4.readthedocs.io/
- Licenses: BSD 3-Clause License
- Latest release: 4.0.0.post20260923 (published about 1 month ago)
- Last Synced: 2026-10-31T17:38:43.435Z (2 days ago)
- Versions: 9
- Dependent Packages: 0
- Dependent Repositories: 0
- Downloads: 1,989 Last month
-
Rankings:
- Dependent packages count: 7.555%
- Average: 25.857%
- Downloads: 27.293%
- Dependent repos count: 42.723%
- Maintainers (1)
pypi.org: flash-attn-4
Flash Attention CUTE (CUDA Template Engine) implementation
- Homepage: https://github.com/Dao-AILab/flash-attention
- Documentation: https://flash-attn-4.readthedocs.io/
- Licenses: BSD 3-Clause License
- Latest release: 4.0.0b37 (published 6 days ago)
- Last Synced: 2026-10-31T17:39:44.902Z (2 days ago)
- Versions: 35
- Dependent Packages: 0
- Dependent Repositories: 0
- Downloads: 859,207 Last month
-
Rankings:
- Dependent packages count: 7.872%
- Average: 26.189%
- Dependent repos count: 44.505%
- Maintainers (2)
pypi.org: qifa
flash-attn quick installer
- Homepage: https://github.com/Dao-AILab/flash-attention
- Documentation: https://qifa.readthedocs.io/
- Licenses: BSD-3-Clause
- Latest release: 0.1.1 (published 11 months ago)
- Last Synced: 2026-10-31T17:39:35.190Z (2 days ago)
- Versions: 2
- Dependent Packages: 0
- Dependent Repositories: 0
- Downloads: 13 Last month
-
Rankings:
- Dependent packages count: 8.34%
- Average: 27.74%
- Dependent repos count: 47.14%
- Maintainers (1)
pypi.org: crossing-flash-attn
Flash Attention: Fast and Memory-Efficient Exact Attention
- Homepage: https://github.com/Dao-AILab/flash-attention
- Documentation: https://crossing-flash-attn.readthedocs.io/
- Licenses: BSD License
- Latest release: 2.6.3 (published about 2 years ago)
- Last Synced: 2026-10-31T17:40:03.040Z (2 days ago)
- Versions: 1
- Dependent Packages: 0
- Dependent Repositories: 0
-
Rankings:
- Dependent packages count: 10.467%
- Average: 34.696%
- Dependent repos count: 58.925%
- Maintainers (1)
pypi.org: hstu-attn
HSTU Attention: The HSTU structure mainly serves generative recommendation
- Homepage: https://github.com/Dao-AILab/flash-attention
- Documentation: https://hstu-attn.readthedocs.io/
- Licenses: BSD License
- Latest release: 2.5.9.post1 (published over 2 years ago)
- Last Synced: 2026-10-31T17:39:51.016Z (2 days ago)
- Versions: 1
- Dependent Packages: 0
- Dependent Repositories: 0
-
Rankings:
- Dependent packages count: 10.542%
- Average: 34.954%
- Dependent repos count: 59.365%
- Maintainers (1)
anaconda.org: flash-attn
FlashAttention is a fast and memory-efficient implementation of the exact attention mechanism for Transformer models. It supports Ampere, Ada Lovelace, Hopper, and Blackwell NVIDIA GPUs with fp16/bf16 precision.
- Homepage: https://github.com/Dao-AILab/flash-attention
- Licenses: BSD-3-Clause
- Latest release: 2.8.3 (published 8 months ago)
- Last Synced: 2026-06-23T15:01:31.411Z (4 months ago)
- Versions: 1
- Dependent Packages: 0
- Dependent Repositories: 0
- Downloads: 121 Total
-
Rankings:
- Dependent packages count: 38.177%
- Average: 41.173%
- Dependent repos count: 41.627%
- Downloads: 43.716%
anaconda.org: flash-attn-layer-norm
FlashAttention is a fast and memory-efficient implementation of the exact attention mechanism for Transformer models. It supports Ampere, Ada Lovelace, Hopper, and Blackwell NVIDIA GPUs with fp16/bf16 precision.
- Homepage: https://github.com/Dao-AILab/flash-attention
- Licenses: BSD-3-Clause
- Latest release: 2.8.3 (published 8 months ago)
- Last Synced: 2026-07-27T23:01:27.198Z (3 months ago)
- Versions: 1
- Dependent Packages: 0
- Dependent Repositories: 0
- Downloads: 28 Total
-
Rankings:
- Dependent packages count: 38.177%
- Average: 41.173%
- Dependent repos count: 41.627%
- Downloads: 43.716%
anaconda.org: flash-attn-fused-dense
FlashAttention is a fast and memory-efficient implementation of the exact attention mechanism for Transformer models. It supports Ampere, Ada Lovelace, Hopper, and Blackwell NVIDIA GPUs with fp16/bf16 precision.
- Homepage: https://github.com/Dao-AILab/flash-attention
- Licenses: BSD-3-Clause
- Latest release: 2.8.3 (published 8 months ago)
- Last Synced: 2026-07-27T23:01:27.501Z (3 months ago)
- Versions: 1
- Dependent Packages: 0
- Dependent Repositories: 0
- Downloads: 28 Total
-
Rankings:
- Dependent packages count: 38.177%
- Average: 41.173%
- Dependent repos count: 41.627%
- Downloads: 43.716%
nixpkgs-unstable: python313Packages.flash-attn-4
CuTeDSL-based implementation of FlashAttention for Hopper and Blackwell GPUs
- Homepage: https://github.com/Dao-AILab/flash-attention/tree/main/flash_attn/cute
- Documentation: https://github.com/NixOS/nixpkgs/blob/nixos-unstable/pkgs/development/python-modules/flash-attn-4/default.nix#L71
- Licenses: BSD-3-Clause
- Latest release: 4.0.0.beta37 (published 2 days ago)
- Last Synced: 2026-11-01T10:53:23.372Z (2 days ago)
- Versions: 2
- Dependent Packages: 0
- Dependent Repositories: 0
-
Rankings:
- Dependent repos count: 0.0%
- Dependent packages count: 0.0%
- Average: 100%
- Maintainers (2)
nixpkgs-unstable: python313Packages.flash-attn
Official implementation of FlashAttention and FlashAttention-2
- Homepage: https://github.com/Dao-AILab/flash-attention/
- Documentation: https://github.com/NixOS/nixpkgs/blob/nixos-unstable/pkgs/development/python-modules/flash-attn/default.nix#L159
- Licenses: BSD-3-Clause
- Latest release: 2.8.2 (published 9 months ago)
- Last Synced: 2026-11-01T10:53:22.609Z (2 days ago)
- Versions: 2
- Dependent Packages: 0
- Dependent Repositories: 0
-
Rankings:
- Dependent repos count: 0.0%
- Dependent packages count: 0.0%
- Average: 100%
- Maintainers (1)
nixpkgs-unstable: python313Packages.flash-attn-3
Official implementation of FlashAttention-3
- Homepage: https://github.com/Dao-AILab/flash-attention/blob/main/hopper
- Documentation: https://github.com/NixOS/nixpkgs/blob/nixos-unstable/pkgs/development/python-modules/flash-attn-3/default.nix#L107
- Licenses: BSD-3-Clause
- Latest release: 3.0.0-unstable-2026-06-10 (published 2 days ago)
- Last Synced: 2026-11-01T10:53:22.973Z (2 days ago)
- Versions: 2
- Dependent Packages: 0
- Dependent Repositories: 0
-
Rankings:
- Dependent repos count: 0.0%
- Dependent packages count: 0.0%
- Average: 100%
- Maintainers (1)
nixpkgs-unstable: python314Packages.flash-attn-4
CuTeDSL-based implementation of FlashAttention for Hopper and Blackwell GPUs
- Homepage: https://github.com/Dao-AILab/flash-attention/tree/main/flash_attn/cute
- Documentation: https://github.com/NixOS/nixpkgs/blob/nixos-unstable/pkgs/development/python-modules/flash-attn-4/default.nix#L71
- Licenses: BSD-3-Clause
- Latest release: 4.0.0.beta37 (published 1 day ago)
- Last Synced: 2026-11-01T11:28:39.007Z (1 day ago)
- Versions: 2
- Dependent Packages: 0
- Dependent Repositories: 0
-
Rankings:
- Dependent repos count: 0.0%
- Dependent packages count: 0.0%
- Average: 100%
- Maintainers (2)
nixpkgs-unstable: python314Packages.flash-attn-3
Official implementation of FlashAttention-3
- Homepage: https://github.com/Dao-AILab/flash-attention/blob/main/hopper
- Documentation: https://github.com/NixOS/nixpkgs/blob/nixos-unstable/pkgs/development/python-modules/flash-attn-3/default.nix#L107
- Licenses: BSD-3-Clause
- Latest release: 3.0.0-unstable-2026-06-10 (published 1 day ago)
- Last Synced: 2026-11-01T11:28:38.673Z (1 day ago)
- Versions: 2
- Dependent Packages: 0
- Dependent Repositories: 0
-
Rankings:
- Dependent repos count: 0.0%
- Dependent packages count: 0.0%
- Average: 100%
- Maintainers (1)
nixpkgs-unstable: python314Packages.flash-attn
Official implementation of FlashAttention and FlashAttention-2
- Homepage: https://github.com/Dao-AILab/flash-attention/
- Documentation: https://github.com/NixOS/nixpkgs/blob/nixos-unstable/pkgs/development/python-modules/flash-attn/default.nix#L159
- Licenses: BSD-3-Clause
- Latest release: 2.8.2 (published 9 months ago)
- Last Synced: 2026-11-01T11:28:38.339Z (1 day ago)
- Versions: 2
- Dependent Packages: 0
- Dependent Repositories: 0
-
Rankings:
- Dependent repos count: 0.0%
- Dependent packages count: 0.0%
- Average: 100%
- Maintainers (1)
Dependencies
- Jimver/cuda-toolkit v0.2.11 composite
- actions/checkout v3 composite
- actions/create-release v1 composite
- actions/setup-python v4 composite
- actions/upload-release-asset v1 composite
- joutvhu/get-release v1 composite
- base latest build
- env-${PERSONAL} latest build
- env-0 latest build
- nvcr.io/nvidia/pytorch 22.12-py3 build
- einops *
- ninja *
- packaging *
- torch *
- apache-tvm-ffi >=0.1.5,<0.2
- einops *
- nvidia-cutlass-dsl ==4.3.4
- torch *
- torch-c-dlpack-ext *
- typing-extensions *
- Jimver/cuda-toolkit v0.2.29 composite
- actions/cache/restore v4 composite
- actions/cache/save v4 composite
- actions/checkout v5 composite
- actions/setup-python v5 composite
- actions/upload-release-asset v1 composite
- joutvhu/get-release v1 composite
- pierotofy/set-swap-space v1.0 composite
- actions/checkout v5 composite
- actions/setup-python v6 composite
- pre-commit/action v3.0.1 composite
- einops *
- ninja *
- packaging *
- torch *
- ./.github/actions/gpu-test * composite
- actions/checkout v4 composite
- docker/login-action v3 composite
- actions/checkout v4 composite
- actions/download-artifact v4 composite
- actions/setup-python v5 composite
- actions/upload-artifact v4 composite
- pypa/gh-action-pypi-publish release/v1 composite
- softprops/action-gh-release v2 composite
- ubuntu 24.04 build