Merge pull request #759 from TKiteRunner/fix/grpo-oom-reorder-rewards

fix: resolve CUDA OOM in train_grpo.py on GPUs with <=8GB VRAM

jingyaogong committed 3mo ago

e73a407f7abcfe7d6fd4c660592b6ed7a0fbc059

Committed by GitHub <noreply@github.com> on 5/6/2026, 7:00:17 AM