Skip to content

[moe] remove DPGradScaler that over-scales expert grads under moe_dp - #6467

Merged
yuxuan-lou merged 2 commits into
mainfrom
fix/moe-dp-grad-scale
Oct 19, 2026
Merged

yuxuan-lou merged 2 commits into
mainfrom
fix/moe-dp-grad-scale

Commits

  1. Commits on Oct 14, 2026

  2. Commits on Oct 15, 2026