Skip to content

[PyTorch debug] FakeQuant: support Float8BlockScaling and fix MoE / w…#3040

Draft
shangxiaokang wants to merge 2 commits into
NVIDIA:mainfrom
shangxiaokang:fake_quant_bwfp8
Draft

[PyTorch debug] FakeQuant: support Float8BlockScaling and fix MoE / w…#3040
shangxiaokang wants to merge 2 commits into
NVIDIA:mainfrom
shangxiaokang:fake_quant_bwfp8

Commits