Do you plan to integrate this algorithm into the vllm project?

[Alienfeel]

RT

[RyeYuan]

I had tried to use the sageAttention in prefilling phase of vllm based on latest 2.0.0 branch, however, after replacing fa2, it doesn't seem to have much effect, and the end-to-end throughput performance (tokens/s) remains almost the same

[jason-huang03]

We are discussing with sglang team, and there is possibility that sageattention will be used in sglang in the future.