[TRTLLM-10022][feat] Add hopper xqa decode support for skip softmax attention#10264

pengbowang-nv merged 11 commits into

NVIDIA:mainNVIDIA/TensorRT-LLM:mainfrom

pengbowang-nv:dev-add-hopper-xqa-skip-softmaxpengbowang-nv/TensorRT-LLM:dev-add-hopper-xqa-skip-softmaxCopy head branch name to clipboard