Add NVFP4 per-token quantization recipe#3045

Draft

cael-ling wants to merge 14 commits into

NVIDIA:mainfrom

cael-ling:feature/nvfp4-per-token-recipe

Commits on May 27, 2026

Add NVFP4 per-token GEMM, fused grouped amax, cast, tests and benches

cael-ling
and
zhongbozhu
committed
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot]
committed

Commits on May 28, 2026

Add optional col-wise RHT to NVFP4 per-token amax+quant (single + grouped)

cael-ling
and
zhongbozhu
committed
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot]
committed

Commits on May 29, 2026

Commits on May 30, 2026

Commits on May 31, 2026

Commits on Jun 2, 2026