Welcome to fairseq2 Documentation

fairseq2 is a sequence modeling toolkit that allows researchers and developers to train custom models for translation, summarization, language modeling, and other content generation tasks.

Quick Start

Run a quick start tutorial.

Basics

Get familiar with fairseq2.

API Reference

Jump to the code.

Documentation

Getting Started

• Installation
• Quick Start

Basics

• Overview
• Design Philosophy
• CLI
• Assets
• Data Pipeline
• Checkpoint Management
• Recipes
• Runtime Extensions
• Gang
• Trainer

Tutorials

• End-to-End Fine-Tuning
• Monitor Your Experiments
• Working with Presets
• Efficient Benchmarking
• Debugging with PuDB
• Add Your Own Model

FAQ

• Contributing to fairseq2

Reference

• API Reference
• Bibliography

Indices and tables

• Index

• Module Index