Use cuDNN for row-scaled NVFP4 grouped GEMM#3042

zianglih wants to merge 2 commits into

NVIDIA:mainNVIDIA/TransformerEngine:mainfrom

zianglih:codex/cudnn-row-scale-nvfp4-grouped-gemmzianglih/TransformerEngine:codex/cudnn-row-scale-nvfp4-grouped-gemmCopy head branch name to clipboard