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.

Note

Our documentation is being updated and will be online soon.

Latest News

• What’s New in v0.5

Getting Started

• Installation

Basics

• Design Philosophy
• Assets
• Building Recipes
• Runtime Extension

Guides

• Add Your Own Model
• Debugging with PuDB

Concepts

• Gangs

API Reference

• fairseq2.assets
• fairseq2.checkpoint
• fairseq2.data
• fairseq2.datasets
• fairseq2.device
• fairseq2.gang
• fairseq2.model_checkpoint
• fairseq2.models
• fairseq2.recipe.optim
• fairseq2.utils.validation
• fairseq2.logging
• fairseq2.metrics
• fairseq2.models
• fairseq2.nn
• fairseq2.optim
• fairseq2.recipe

Other

• Contributing to fairseq2
• Bibliography

Indices and tables

• Index

• Module Index

• Search Page