Repository · GitHub

thu-ml/SageAttention

[ICLR2025, ICML2025, NeurIPS2025 Spotlight] Quantized Attention achieves speedup of 2-5x compared to FlashAttention, without losing end-to-end metrics across language, image, and video models.

View on GitHubReport an issue

Owner	thu-ml
License (as declared)	Apache-2.0
Stars	3,969
Forks	528
Main language	Cuda
Created	3 October 2024
Last updated at source	17 January 2026

cudaattentionefficient-attentioninference-accelerationllmllm-inframlsysquantizationtritonvideo-generatevideo-generationvit

Embed this entry

<iframe src="https://silytics.si/embed/item/822c5995-0232-4a0a-87bd-c6ae13914751" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Free, no tracking. More about feeds and embeds.

More from thu-ml

tianshouRepository
TurboDiffusionRepository
zhusuanRepository
SpargeAttnRepository

Silytics shows public metadata from GitHub and links to the original. Information may be incomplete or out of date. Last refreshed 15 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.