A comprehensive guide for working with the NVIDIA Cosmos ecosystem—a suite of World Foundation Models (WFMs) for real-world, domain-specific applications across robotics, simulation, autonomous systems, and physical scene understanding.

This cookbook provides step-by-step workflows, technical recipes, and concrete examples for the complete AI development lifecycle with Cosmos models:

• Inference: Quick-start examples with pre-trained models
• Data Curation: Scalable data processing pipelines with Cosmos Curator
• Post-Training: Custom fine-tuning for domain-specific adaptation
• Evaluation: Quality control and model assessment workflows

The Cosmos ecosystem includes core model families: Curator, Predict (versions 2 and 2.5), Transfer (versions 1 and 2.5), Reason 1, and RL, each targeting specific capabilities in the AI development workflow.

Before getting started, ensure you have the following requirements:

NVIDIA GPUs: Not required for local documentation rendering. For running cookbook recipes and workflows: Ampere architecture or newer (A100, H100) - minimum 1 GPU, recommended 8 GPUs

• Operating System: Ubuntu 24.04, 22.04, or 20.04
• Python: Version 3.10+
• NVIDIA Container Toolkit: 1.16.2 or later
• CUDA: 12.4 or later
• Docker Engine
• Access: Internet connection for downloading models and dependencies

# Install uv (fast Python package manager)
curl -LsSf https://astral.sh/uv/install.sh | sh
source $HOME/.local/bin/env

# Install just (command runner)
uv tool install -U rust-just

# Optional useful tools
uv tool install -U s5cmd      # High-performance S3 operations
uv tool install -U streamlit  # Web app framework
uv tool install -U yt-dlp     # Video downloading

# Clone the repository
git clone https://github.com/nvidia-cosmos/cosmos-cookbook.git
cd cosmos-cookbook

# Install dependencies and setup
just install

# Serve documentation locally
just serve-external  # For public documentation
# or
just serve-internal   # For internal documentation (if applicable)

Then open http://localhost:8000 in your browser.

The Cosmos Cookbook is organized into two main directories:

Contains the source documentation in markdown files:

• Technical guides and workflows
• End-to-end examples and case studies
• API references and tutorials
• Getting started guides

Contains executable scripts referenced throughout the cookbook:

• Data processing and curation pipelines
• Model evaluation and quality control scripts
• Configuration files for post-training tasks
• Automation tools and utilities

This structure separates documentation from implementation, making it easy to navigate between reading about workflows and executing the corresponding scripts.

# Development
just install          # Install dependencies and setup
just setup            # Setup pre-commit hooks
just serve-external   # Serve public documentation locally
just serve-internal   # Serve internal documentation locally

# Quality Control
just lint            # Run linting and formatting
just test            # Run all tests and validation

# Continuous Integration
just ci-lint         # Run CI linting checks
just ci-deploy-internal         # Deploy internal documentation
just ci-deploy-external         # Deploy external documentation

• End-to-End Examples: Complete workflows from data to deployment
• Quick Start Templates: Get up and running in minutes
• Modular Scripts: Reusable components for custom workflows
• Evaluation Tools: Built-in quality assessment and metrics
• Production Ready: Scalable pipelines for real-world deployment
• Comprehensive Docs: Detailed guides and API references

• Full Documentation - Complete guides and examples
• Getting Started - Environment setup and first steps
• Contributing Guide - How to contribute to the cookbook

• Share Success Stories: We love hearing how you use Cosmos models creatively
• Report Issues: Use GitHub issues for bugs and feature requests
• Discussions: Join our community discussions
• Documentation: Check our comprehensive guides first

This project will download and install additional third-party open source software projects. Review the license terms of these open source projects before use.

NVIDIA Cosmos source code is released under the Apache 2 License.

NVIDIA Cosmos models are released under the NVIDIA Open Model License. For a custom license, please contact cosmos-license@nvidia.com.