A Problem encountered when using data parallel

[Aziily]

Hi, I think I may need someone to help me solve the problem
I am trying to use two nodes which each has a GPU, and I have written such config in config.py

while when I use colossalai run to start the distributed training, I get such WARNING

and by watching the logs of training, I believe it doesn't train a model, instead it starts training a model per process

[YuliangLiu0306]

Did you set the config about 'zero'?