ot-triton-lab / flash-sinkhorn
The official repository of FlashSinkhorn [ICML 2026 Oral]
Why this job: works in the areas we cover, but does not say which job it does
What it does
This package provides the streaming kernels of FlashSinkhorn and the multiscale, block-sparse method of FlashSinkhorn 2. Dense backends support balanced and relaxed-marginal OT, gradients, supported Hessian-vector products, and transport-plan application. A streaming c-transform supports semi-dual objectives.
Get started
pip install --upgrade flash-sinkhorn- Maintenance
- Active
- Last commit 9 days ago
- Stars
- 226
- On GitHub
- Licence
- MIT
- Allows business use
- Downloads for
- None listed
- No release files
- Latest version
- v0.4.1
- Released 38 days ago
- Built with
- Python
- 100% of the code
Is it looked after?
- Contributors
- Not stated
- Last 90 days
- Issues
- 100% closed
- 0 still open
- Pull requests
- 6 merged
- 1 waiting
- Releases
- 8
- 8 in the last year
- No contributing guide
- No security policy
- No code of conduct
Similar repos
| Repo | Stars | Last commit | Licence | Downloads for |
|---|---|---|---|---|
| BobMcDear / attorch | 607 | 3 months ago | MIT | No downloads |
| NVIDIA / cuml | 5,321 | 6 days ago | Apache-2.0 | No downloads |
| AMD-AGI / GEAK | 204 | today | Apache-2.0 | No downloads |
More: languages, topics and project record
Languages
- Python 100%
- Nix <1%
Topics
- cuda
- entropic-optimal-transport
- flash-attention
- gpu
- machine-learning
- optimal-transport
- pytorch
- sinkhorn
- triton
- flashsinkhorn
Project record
- JobOther AI tools
- StatusListed
- Commits56
- Forks22
- Watching12
- Stars in 28 days+2
- CreatedDec 21, 2025
- First seen by usNov 7, 2026
- CheckedNov 8, 2026