문서
카테고리
단어
분 읽기
관련 카테고리: 없음
Production configuration for running NeMo-RL (GRPO) and TRL (DPO) training jobs with labeled preference datasets on Karpenter Spot node pools and Volcano Gang Scheduling.