facebookresearch/recipes

Recipes are a standard, well supported set of blueprints for machine learning engineers to rapidly train models using the latest research techniques without significant engineering overhead.Specifically, recipes aims to provide- Consistent access to pre-trained SOTA models ready for production- Reference implementations for SOTA research reproducibility, and infrastructure to guarantee correctness, efficiency, and interoperability.

What it solves

It removes the need to write repetitive "glue code" when training deep learning models. Instead of building training loops, configuration parsers, and data loaders from scratch, users can start with production-ready, reproducible baselines for common machine learning tasks.

How it works

The library provides "recipes"—ready-to-run Python applications that bundle together a model architecture, a trainer, and a configuration. These recipes are modular, allowing users to swap components (like changing a model from ResNet18 to ResNet52) via configuration files or by directly modifying the code for deeper customization.

Who it’s for

Machine learning researchers and developers who want to quickly experiment with high-quality baselines or need a reliable starting point to extend for their own specific projects.

Highlights

  • Ready-to-run recipes for vision (image classification) and text (document classification).
  • Supports scaling from local laptops to large clusters.
  • Uses Hydra for flexible configuration and hyper-parameter management.
  • Focuses on accurately reproducing research paper baselines.

Related

  • Project
    google/uncertainty-baselines

    A collection of high-quality, standardized baselines and templates for researchers to benchmark uncertainty and robustness in deep learning models.

  • Project
    cvg/glue-factory

    A library for training and evaluating deep neural networks that extract and match local visual features, supporting state-of-the-art models like LightGlue and GlueStick.

  • Project
    dotnet/machinelearning

    A cross-platform open-source machine learning framework for .NET developers to build, train, and deploy custom models without needing Python or R.

  • Project
    thinking-machines-lab/tinker-cookbook

    A library of abstractions and recipes for fine-tuning language models via the Tinker API, supporting SFT, RL, and multimodal training.

  • Project
    facebookresearch/stopes

    A library for preparing machine translation data, providing tools for cleaning monolingual data, mining bitext, and running knowledge distillation pipelines.