SangBin Cho

there's an odd ci failure

@LiuXiaoxuanPKU what's the latest supported version from flash infer? Also, is this something we can just simply build it with torch 2.3?

Let's merge it without adding tests to CI? I think we don't need to be blocked by them (to merge). I will also create an issue to their repo for...

https://github.com/flashinfer-ai/flashinfer/issues/227

awseom! super excited to see performance with cuda graph!

@MichoChan I think it is because it doesn't have cuda graph support yet (at large qps, the cpu overhead cuda graph removes is negligible usually)

cc @comaniac

oops. sorry .can you resolve the merge conflict?

amd failure seems unrelated. cc @simon-mo I am merging this

Hmm is there a way to just exclude it in actual wheel?