How would you parallelize the training of a large deep learning model across a multi-node GPU cluster using PyTorch Distributed Data Parallel (DDP)?
