Skip to content

Accept compact MXFP8 scales in GEMM preparation - #3519

Draft
wujingyue wants to merge 3 commits into
NVIDIA:mainfrom
wujingyue:fix/mxfp8-compact-scales-gemm
Draft

wujingyue wants to merge 3 commits into
NVIDIA:mainfrom
wujingyue:fix/mxfp8-compact-scales-gemm

Swizzle compact MXFP8 scales directly into padded output

9b6e9b0
Select commit
Loading
Failed to load commit list.