lightning-hydra-template

Why we included this project

Most deep learning projects start as a pile of ad-hoc scripts, and that gets painful the moment you want to reproduce a run or try ten variations. This template fixes that by wiring PyTorch Lightning together with Hydra so models, datasets, callbacks, and loggers are all defined in config files. Changing a dataset, adding a model, or tweaking an optimizer becomes a YAML edit instead of a code change. The code is heavily commented and ships with practical MLOps pieces like logging, checkpointing, and support for multi-GPU, TPU, or SLURM, which makes it a decent learning resource for teams new to structured ML engineering. It is a solid scaffold for turning a research idea into something you can actually track and rerun.