# ConStellaration: Quasi-Isodynamic Stellarator Plasma Boundaries with Ideal MHD Equilibria

## Facts

- Provider: Proxima Fusion
- License: MIT (https://huggingface.co/datasets/proxima-fusion/constellaration)
- Size: 7,095 files
- Format: PNG, Parquet (in sampled files)
- Stored at: Hugging Face, https://huggingface.co/datasets/proxima-fusion/constellaration
- Access: public, readable without an account
- Category: Energy
- Source: https://arxiv.org/abs/2506.19583
- Page: https://scigantic.com/archive/df0d2012-b100-5a83-b117-ac414887e314
- Keywords: Energy, benchmark, PNG, Parquet, Hugging Face

## About

A dataset of diverse quasi-isodynamic (QI) stellarator plasma boundary shapes with their performance metrics, ideal magnetohydrodynamic (MHD) equilibria and the settings used to generate them, released by Proxima Fusion to let optimization and machine learning researchers work on stellarator design for magnetic confinement fusion energy. Metrics and equilibria are evaluated in vacuum and with plasma inside at volume-averaged beta of 1 to 5 percent. The default subset holds boundaries, ideal MHD metrics, omnigenous field targets and sampling settings (DESC, VMEC, near-axis expansion); vmecpp_wout holds, per boundary, a JSON string of the VMEC++ WOut file that converts to a VMEC2000 output file; desc_equilibrium holds DESC re-solves of those equilibria. Listing the repository (measured 2026-11-13) shows data/ with 3 Parquet shards (0.60 GB), vmecpp_wout/ with Parquet parts totalling about 382 GB plus an id_to_file_map, desc_equilibrium/, finite_beta_1pct/ to finite_beta_5pct/ and matching vmecpp_wout_finite_beta folders; about 2.7 TB in all. Open with Hugging Face datasets, pandas or pyarrow, and the constellaration Python package for forward models and benchmark problems. Released under the MIT licence; cite arXiv:2506.19583.

## Load the data

```python
from datasets import load_dataset
ds = load_dataset("proxima-fusion/constellaration")
print(ds)
```

## Data structure

File types in sampled files: PNG, Parquet

Example file: `data/train-00000-of-00003.parquet`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/df0d2012-b100-5a83-b117-ac414887e314
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/df0d2012-b100-5a83-b117-ac414887e314
- Over MCP: `get_schema_card` with id `df0d2012-b100-5a83-b117-ac414887e314` at https://api.scigantic.com/mcp

## Related topics

- [Parquet datasets](https://scigantic.com/topics/parquet)
- [Benchmark datasets](https://scigantic.com/topics/benchmark)
- [PNG datasets](https://scigantic.com/topics/png)
- [Energy datasets](https://scigantic.com/topics/energy)
