GRPO/DPO Training Job
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.
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.