Treegrpo Tree Advantage Rl

Achieve 2.4× faster RL training for diffusion models by restructuring denoising as tree search with shared computation. TreeGRPO computes fine-grained step advantages instead of trajectory-level rewards—crucial for efficient diffusion model optimization.

adu2021 Updated 7 repo stars

File contents

adu2021/skillxiv commit 7f64e023f1

Frequently asked questions

npx get-skill add adu2021/treegrpo-tree-advantage-rl