Use W&B Artifacts for dataset versioning, model versioning, and tracking dependencies and results across machine learning pipelines. Think of an artifact as a versioned folder of data.