[None][perf] Triton FP8-activation MoE for decode-sized batches on Hopper - #21250

[None][perf] Triton FP8-activation MoE for decode-sized batches on Hopper#21250

ruocheng-nv wants to merge 4 commits into

NVIDIA:mainNVIDIA/TensorRT-LLM:mainfrom

ruocheng-nv:perf/hopper-fp8-moe-triton-decoderuocheng-nv/TensorRT-LLM:perf/hopper-fp8-moe-triton-decodeCopy head branch name to clipboard