Does DDP always split one model across multiple GPU?

Question

I\'m new in pytorch.
Because I want to increase the batch size, and the model is too heavy, so I received feedback that using pytorch DP(DataParallel) and DDP(Distribute