fasterkernels
High-performance FlashAttention-2 and Flash Decoding (GQA + varlen) kernels implemented in Triton, TileLang, and CUDA
Activity
- Latest release
- 1mo ago
- Total releases
- 1
- Cadence
- —
- Last 12 months
- 1
Details
- License
- custom
- First release
- Jul 19, 2026
Releases
| Version | Released | |
|---|---|---|
0.2.0
initial
|
0.2.0
initial
Dependencies (7)
|