[None][feat] add generic PrimTS block-sparse FMHA and unify sparse attention runtime inputs - #18815

[None][feat] add generic PrimTS block-sparse FMHA and unify sparse attention runtime inputs#18815

heyuhhh wants to merge 4 commits into

NVIDIA:mainNVIDIA/TensorRT-LLM:mainfrom

heyuhhh:user/yuhangh/prims-ts-block-sparse-fmhaheyuhhh/TensorRT-LLM:user/yuhangh/prims-ts-block-sparse-fmhaCopy head branch name to clipboard