lightning-hydra-template and pytorch-lightning-template
These two tools are competitors, as both provide templates for PyTorch Lightning projects, with ashleve/lightning-hydra-template offering additional integration with Hydra for configuration management, while miracleyoo/pytorch-lightning-template focuses on a simpler adaptation of existing PyTorch code.
About lightning-hydra-template
ashleve/lightning-hydra-template
PyTorch Lightning + Hydra. A very user-friendly template for ML experimentation. ⚡🔥⚡
Combines Hydra's hierarchical config composition with PyTorch Lightning's training abstractions to enable rapid experimentation through command-line overrides and config-driven instantiation. Includes built-in support for multiple experiment tracking backends (W&B, MLFlow, Neptune, Comet), hyperparameter search via Hydra plugins like Optuna, and automated logging/checkpointing with dynamically-generated folder structures. Provides a structured project layout with pre-commit hooks, CI/CD workflows, and generic test utilities to accelerate ML prototyping on prepared datasets.
About pytorch-lightning-template
miracleyoo/pytorch-lightning-template
An easy/swift-to-adapt PyTorch-Lighting template. 套壳模板，简单易用，稍改原来Pytorch代码，即可适配Lightning。You can translate your previous Pytorch code much easier using this template, and keep your freedom to edit all the functions as well. Big-project-friendly as well. No need to rewrite your config in hydra.
Decouples models and datasets through interface abstractions (MInterface/DInterface), allowing multiple implementations to coexist without code duplication while maintaining full control over training logic like `training_step` and `configure_optimizers`. Provides specialized templates for classification and super-resolution tasks with pre-configured project structures, reducing boilerplate while supporting extensibility through command-line argument passing to dynamically instantiated components.
Scores updated daily from GitHub, PyPI, and npm data. How scores work