Lighter is a YAML-driven deep learning framework built on PyTorch Lightning. Define your model, data, and training in config files instead of writing boilerplate code.

Quick Start

pip install lighter

Create config.yaml:

trainer:
  _target_: pytorch_lightning.Trainer
  max_epochs: 10

system:
  _target_: lighter.System
  model:
    _target_: torchvision.models.resnet18
    num_classes: 10
  criterion:
    _target_: torch.nn.CrossEntropyLoss
  optimizer:
    _target_: torch.optim.Adam
    params: "$@system::model.parameters()"
    lr: 0.001
  dataloaders:
    train:
      _target_: torch.utils.data.DataLoader
      batch_size: 32
      dataset:
        _target_: torchvision.datasets.CIFAR10
        root: ./data
        train: true
        download: true
        transform:
          _target_: torchvision.transforms.ToTensor

Run:

lighter fit config.yaml

Override from CLI:

lighter fit config.yaml system::optimizer::lr=0.01

→ Full tutorial

Documentation

• 📚 Get Started
• ⚙️ Configuration Guide
• 🔌 Adapters
• 🏗️ Architecture

Projects Using Lighter

• 🏥 Foundation Models for Cancer Imaging
• 🧠 Vision Foundation Models for CT

Community

• 💬 Discord
• 🐛 GitHub Issues
• 📺 YouTube
• 🤝 Contributing

Citation

@article{lighter,
    doi = {10.21105/joss.08101},
    year = {2025}, publisher = {The Open Journal}, volume = {10}, number = {111}, pages = {8101},
    author = {Hadzic, Ibrahim and Pai, Suraj and Bressem, Keno and Foldyna, Borek and Aerts, Hugo JWL},
    title = {Lighter: Configuration-Driven Deep Learning},
    journal = {Journal of Open Source Software}
}