dagzoo

Reproducible synthetic tabular data from latent causal structure.

Start Here GitHub

Quickstart

Install dagzoo, inspect the recipe catalog, generate a first run, publish a handoff root, and use the same recipe in Python.

Read more

Reference Packs

Start from named recipe:<name> configs and published recipes you can inspect, reuse, and cite.

Read more

Publish to Hub

Generate a portable handoff root, keep dagzoo-only sidecars local, and publish the public corpus to Hugging Face Hub.

Read more

Output Format

Stable shard layout, metadata contract, and in-memory sample shape.

Read more

Advanced Controls

Move from named recipes to repo-local YAML authoring, optional dagzoo filter runs, and deeper workflow controls.

Read more

How It Works

See the runtime model, terminology, and the latent-DAG-to-tabular generation flow.

Read more

Features

Explore focused guides for diagnostics, interventions, missingness, shift, noise, and benchmark guardrails.

Read more