Multi-GPU Demos Scripts Single GPU $ python3 main.py Data Parallel $ python3 main_dp.py Distributed Data Parallel $ torchrun --standalone --nproc_per_node=NUM_GPUS main_ddp.py or $ python3 -m torch.distributed.launch --nproc_per_node=NUM_GPUS main_ddp.py