chronocratic-datasets

Ready-to-use time series datasets for PyTorch Lightning.

chronocratic-datasets provides pre-built, Lightning-compatible data modules for popular time series benchmarks. It covers forecasting (ETT, Weather, Electricity) and classification (UCR, UEA) with caching, scaling, and distributed-data-parallel (DDP) compliance out of the box.

Features

• Forecasting datasets: ETT, Weather, Electricity with sliding-window and raw-series modes

• Classification datasets: UCR (univariate) and UEA (multivariate) benchmarks

• Lightning integration: Ready-to-use LightningDataModule subclasses

• Caching: Automatic dataset caching for fast reloads

• Scaling: Configurable data normalization via scikit-learn scalers

• DDP compliant: Works seamlessly with DataParallel and Strategy("ddp")

Quick Install

pip install chronocratic-datasets

Documentation Contents

Guides

• Quick Start
  • Installation
  • Basic Workflow
    • Forecasting Example
    • Classification Example
  • Next Steps
• Forecasting Modules
  • Weather Module
  • ETT Data Module
  • Electricity Load Module
  • Dataset Classes
  • Loader Mode
  • Forecasting Mode
  • Scaling
  • Next Steps
• Classification Modules
  • UCR Classification Data Module
  • UEA Classification Data Module
  • Dataset Classes
  • Loader Mode
  • Splitting Strategy
  • Scaling
  • Next Steps

API Reference

• API Reference
  • Enum API Reference
  • Dataset API Reference
  • DataModule API Reference
  • Utility API Reference

Project

• Changelog
• Contributing