Skip to content

Export unquantized FP8Linear layers in the export dtype, not fp32 - #2651

Merged
shengliangxu merged 3 commits into
NVIDIA:mainfrom
tritsystem:fix/fp8-unpack-export-dtype
Oct 6, 2026
Merged

Export unquantized FP8Linear layers in the export dtype, not fp32#2651
shengliangxu merged 3 commits into
NVIDIA:mainfrom
tritsystem:fix/fp8-unpack-export-dtype

Commits

  1. Commits on Oct 3, 2026

  2. Commits on Oct 6, 2026