TorchWM Documentation

TorchWM is a modular PyTorch library for world models, latent-dynamics planning, and representation learning workflows (Dreamer, JEPA, IRIS, DiT, and more).

Get Started

• Getting Started
• Installation

User Guides

• Public API Quick Reference
• Tutorial notebooks
• Using Operators for Inference
• Training Guide
• Inference Guide
• Environments Guide
• Environment Backends
• TorchWM CLI
• Package Overview
• World Models Study Guide

Algorithms

• Dreamer: Model-Based RL with Latent Dynamics
• JEPA: Joint Embedding Predictive Architecture
• IRIS: Transformers for Sample-Efficient World Models
• DiT: Diffusion Transformer
• Genie: Generative Interactive Environment

Reference

• API Reference
• Configs Reference

Development

• Contributing
• Benchmarking World Models