python vllmtask.py --judge /home/aiscuser/fhw/model_weights/llama --start 0 --end 100000
python vllmtask.py --judge /home/aiscuser/fhw/model_weights/llama --start 100000 --end 200000
python vllmtask.py --judge /home/aiscuser/fhw/model_weights/llama --start 200000 --end 311720
python /data/local/zhangdi/DPO/DPO_train.py