Otter Documentation

Learn how to compose transformer architectures visually and export production-ready PyTorch code.

Getting Started

5 min read

Learn how to create your first transformer architecture in Otter.

Component Reference

Reference

Explore all available components: embeddings, attention, FFN, normalization, and more.

Training Guide

10 min read

Run your exported model on local GPU, Google Colab, or Lambda Labs.

Quick Start

Ready to dive in? Here's the fastest path to your first architecture:

  1. 1. Open the Canvas
  2. 2. Drag an Embedding component onto the canvas
  3. 3. Add Attention, FFN, and Output components
  4. 4. Connect them in order
  5. 5. Click Export when the architecture is valid