Data and model paralelisation for optimal training

[crangelsmith]

As models and datasets get bigger we should test strategies for optimising the training.

pytorch-lighting is a good option as it allows you to test different strategies in data and model parallelisation under the same framework, also changes to the code are minimal if Fabric is used. See example here: https://github.com/Lightning-AI/pytorch-lightning?tab=readme-ov-file#lightning-fabric-expert-control