HKUST(GZ) Data Intelligence and Analytics Lab (DIAL)

license

BSD-3-Clause(OSI)

source

public

core features withheld

no

BSD-3-Clause license body confirmed. The repository is public and unarchived and builds the whole product, and the README describes no paid tier, enterprise edition or license-gated build beside it, so source is public and the core ungated.

• https://api.github.com/repos/HKUSTDial/flash-sparse-attention recorded 2026-09-02

  Repo metadata for flash-sparse-attention's canonical repository - private false, archived false, fork false.

• https://github.com/HKUSTDial/flash-sparse-attention/blob/main/LICENSE recorded 2026-10-07

  LICENSE body is the verbatim three-clause BSD text under an HKUST(GZ) DIAL copyright line.

• https://github.com/HKUSTDial/flash-sparse-attention/blob/main/README.md recorded 2026-10-07

  README describes Flash-Sparse-Attention as "a high-performance trainable sparse attention implementation", listing "Forward and backward passes for dense attention, sparse attention, and gated attention", "Grouped Query Attention and Multi Query Attention", "Paged Attention" and "Fused Quant for low-precision computation on hardware without native FP8 support"; it names no paid tier, enterprise edition or license-gated build beside the published source.

PyPI downloads of `flash-sparse-attn` are the adoption signal here; the package is published from the project's own HKUSTDial/flash-sparse-attention repository, and `pip install flash-sparse-attn` is the install path its README documents.

• https://pypi.org/pypi/flash-sparse-attn/json recorded 2026-09-02

  PyPI metadata for flash-sparse-attn 2.0.5 - project_urls Source = https://github.com/HKUSTDial/flash-sparse-attention, summary 'Flash Sparse Attention: Fast and Memory-Efficient Trainable Sparse Attention'.

• https://pypistats.org/api/packages/flash-sparse-attn/recent recorded 2026-09-02

  last_month downloads = 64 for flash-sparse-attn

Flash-Sparse-Attention applies one optimization family, attention, broadly across configurations rather than narrowing to one model, matching Flash Attention's feature breadth. But no outside project is recorded depending on it the way Flash Attention's usage page links seven integrating projects, which keeps it a step behind that library and short of Apache TVM.

• https://github.com/HKUSTDial/flash-sparse-attention/blob/main/README.md recorded 2026-10-07

  README describes Flash-Sparse-Attention as "a high-performance trainable sparse attention implementation", listing "Forward and backward passes for dense attention, sparse attention, and gated attention", "Grouped Query Attention and Multi Query Attention", "Paged Attention" and "Fused Quant for low-precision computation on hardware without native FP8 support"; it names no paid tier, enterprise edition or license-gated build beside the published source.