[None][refactor] Unify sparse attention framework with clean backend interfaces - #12733

[None][refactor] Unify sparse attention framework with clean backend interfaces#12733

lfr-0531 merged 22 commits into

NVIDIA:mainNVIDIA/TensorRT-LLM:mainfrom

lfr-0531:user/fanrongl/sparse-attention-refactorlfr-0531/TensorRT-LLM:user/fanrongl/sparse-attention-refactorCopy head branch name to clipboard