Enable NVFP4 fused grouped MLP #3048

Open

sraman-rgb wants to merge 17 commits into

NVIDIA:mainfrom

sraman-rgb:nvfp4-grouped-mlp-main

Commits on May 27, 2026

Enable NVFP4 fused grouped MLP follow-up

sraman-rgb
authored and
a
committed
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot]
committed

Commits on May 29, 2026

Merge branch 'main' into nvfp4-grouped-mlp-main
timmoon10
committed
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot]
committed
[PyTorch] Drop `_MXFP8` suffix from fused grouped MLP op classes

timmoon10
and
claude
committed
[PyTorch] Add `ceil_div` utility and use it in fused grouped MLP ops

timmoon10
and
claude
committed
[PyTorch] Drop defensive `getattr` defaults in fused grouped MLP ops

timmoon10
and
claude
committed
Fix NVFP4 RHT grouped MLP reference
sraman-rgb
committed
Simplify NVFP4 RHT grouped MLP reference
sraman-rgb
committed

Commits on May 30, 2026