[NVFP4][MOE] Add unfused quantization fallback when input shape is not aligned #2747

ptrendx merged 2 commits intoNVIDIA:mainNVIDIA/TransformerEngine:mainfrom

zhongbozhu:add_fallback_to_nvfp4_split_quantizezhongbozhu/TransformerEngine:add_fallback_to_nvfp4_split_quantizeCopy head branch name to clipboard