Provide documentation on how to train a Transformers4Rec model using multiple GPU with DataParallel (DP) and DistributedDataParallel (DDP).

• Short explanation of DP and DDP with links to PyTorch documentation
• Describe code snippets, command line examples and environment variables necessary to use the Trainer with DP and DDP
• Table comparing runtimes x number of GPUs for one of the integration tests for single GPU, DP and DDP
• Describe that LR needs to be increased when using DP and DDP