triton kernel that implements the Flash-Decoding algorithm

triton-lang / triton

Development repository for the Triton language and compiler

https://triton-lang.org/

MIT License

13.27k stars 1.63k forks source link

Open zhangxiao-stack opened 8 months ago

zhangxiao-stack commented 8 months ago

Is there a Flash-Decoding algorithm implemented based on Triton?

xinji1 commented 7 months ago

IIUC, lightllm has implemented a flash-decoding triton kernel. Maybe you can refer it.