Trainer 支持 Verl GRPO（目前单轮）、持久化训练进程、去除swanlab#184

RYY177 wants to merge 1 commit into

dev/v2OpenDCAI/Dataflow-LoopAI:dev/v2from

ryy/trainer0720-verlOpenDCAI/Dataflow-LoopAI:ryy/trainer0720-verlCopy head branch name to clipboard