torchscalers

A simple collection of scalers for PyTorch pipelines.

All scalers are torch.nn.Module subclasses. Their fitted statistics are stored as module buffers, which means they:

• are included in state_dict() and saved/restored with every checkpoint automatically,

• move to the correct device with to(),

• work inside torch.nn.Sequential pipelines (calling scaler(x) is equivalent to scaler.transform(x)).

Contents

• Installation
  • Example scripts
• Examples
  • Pure PyTorch
  • PyTorch Lightning
• API Reference
  • Abstract base class
  • Scalers
  • Composite scalers