Repository Issues

agronholm/SageAttention

Quantized Attention that achieves speedups of 2.1-3.1x and 2.7-5.1x compared to FlashAttention2 and xformers, respectively, without lossing end-to-end metrics across various models.

View on GitHub
Stars
 (0 stars)
Forks
 (1 fork)
Indexed issues
 (0 indexed issues)
open beginner issues
 (0 open beginner issues)
Latest indexed
Aug 14, 2026
Last GitHub push
Mar 23, 2025
License
Apache License 2.0
Contributing guide
No contributing guide
Code of conduct
No code of conduct
Dominant language
Cuda
PR merge metrics
 (No merged PRs in 30d)
Beginner labels
No beginner labels indexed

Issues

0 open indexed issues

No open indexed issues found for this repository.