Lightning-cli

This guide explores building a scalable training pipeline using PyTorch Lightning and LightningCLI. It covers handling dataset preparation, where different transformations for training and validation are applied. The post details how Lightning organizes code into data and model modules, streamlining training but complicating parameter sharing. By using LightningCLI and link_arguments, key parameters are passed between modules. Based on the discussed aspects a final blueprint is proposed.