[None][feat] Add per-encoder attention backend selection for multimodal models - #20802

[None][feat] Add per-encoder attention backend selection for multimodal models#20802

yechank-nvidia wants to merge 6 commits into

NVIDIA:mainNVIDIA/TensorRT-LLM:mainfrom

yechank-nvidia:feat/mm-encoder-attention-backend-20261021yechank-nvidia/TensorRT-LLM:feat/mm-encoder-attention-backend-20261021Copy head branch name to clipboard