[Feat] add chunked-attention kernels on Hopper (for llama4) - #4291

[Feat] add chunked-attention kernels on Hopper (for llama4)#4291

schetlur-nv merged 3 commits into

NVIDIA:mainNVIDIA/TensorRT-LLM:mainfrom

PerkzZheng:user/perkzz/chunked-attentionPerkzZheng/TensorRT-LLM:user/perkzz/chunked-attentionCopy head branch name to clipboard