Infrastructure / Compilers & Model Optimization

Attention Gym

Attention-Gym is a Triton-based collection of sparse and quantized attention kernel implementations - Flash Attention 2, sliding-tile attention and several SageAttention variants - built for researchers to rapidly implement, test and validate new attention mechanisms.

Verified 2026-09-02 via the GitHub API and the LICENSE body.

Openness

5 high confidence

5.0

license: Apache-2.0(OSI)
source: public
core-gated: ungated

Apache-2.0 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://github.com/RiseAI-Sys/attention-gym/blob/main/LICENSE recorded 2026-09-02
LICENSE body carries the verbatim Apache License 2.0 text.
https://api.github.com/repos/RiseAI-Sys/attention-gym recorded 2026-09-02
Repo metadata for attention-gym's canonical repository - private false, archived false, fork false.
https://github.com/RiseAI-Sys/attention-gym/blob/main/README.md recorded 2026-09-02
README describes Attention-Gym as a flexible framework built on Triton for implementing, testing and validating sparse and quantized attention mechanisms, listing the specific kernels it currently supports, with no paid tier, enterprise edition or license-gated build beside the published source.

Adoption

1 low confidence

1.0

43 GitHub stars, which lands in the <1K stars band of the stars scale, level 1. The `attn-gym` package on PyPI is pytorch-labs' flex-attention examples project, not this repository, and Attention-Gym itself has no registry package, so no download count exists and the stars scale applies.

https://api.github.com/repos/RiseAI-Sys/attention-gym recorded 2026-09-02
Repo metadata - stargazers_count = 43 - for RiseAI-Sys/attention-gym.

Capability

2 medium confidence

2.0

Banded on the category feature matrix as a narrow kernel set: a small collection of already-published kernels re-packaged for experimentation, rather than a broad optimization family or a full compiler. Placed two bands below the tensorrt rung.

https://github.com/RiseAI-Sys/attention-gym/blob/main/README.md recorded 2026-09-02
README describes Attention-Gym as a flexible framework built on Triton for implementing, testing and validating sparse and quantized attention mechanisms, listing the specific kernels it currently supports, with no paid tier, enterprise edition or license-gated build beside the published source.

Verified 2026-09-02