Skip to content

fix(grpo): clamp log-ratio and k3 KL for numerical stability#4

Open
WyldeCat wants to merge 2 commits into
feat/flce-num-chunks-override-v2from
fix/grpo-clamp-numerical-stability
Open

fix(grpo): clamp log-ratio and k3 KL for numerical stability#4
WyldeCat wants to merge 2 commits into
feat/flce-num-chunks-override-v2from
fix/grpo-clamp-numerical-stability

Commits

Commits on May 14, 2026

Commits on May 20, 2026