CoilStellaration: Optimized Stellarator Coilsets and Coil Design Benchmark

Provider
Proxima Fusion
License
MIT (source)
Size
220 files
Format
Parquet (in sampled files)
Stored at
Hugging Face (https://huggingface.co/datasets/proxima-fusion/coilstellaration)
Access
Free and public. No account with the data provider is needed.
Category
Energy
Source
https://arxiv.org/abs/2611.12689

About CoilStellaration: Optimized Stellarator Coilsets and Coil Design Benchmark

Optimized stellarator coilsets for 14,496 plasma boundaries from ConStellaration (2 to 5 field periods), with a benchmark for models that design coils. Each of 266,549 samples (253,250 train, 13,299 eval, no boundary in both) pairs a boundary with randomly sampled engineering requirements such as coil-coil and coil-plasma clearance. Every sample has a REGCOIL coilset and 177,341 were refined with DESC, with 4 to 8 coils per half-period. Each coilset is scored by normalized field error |B.n|/|B| on the plasma boundary, minimum coil-coil distance, minimum coil-plasma distance and maximum curvature, normalized by minor radius. Coils are closed curves given by Fourier coefficients of x, y and z. Listing the repository (measured 2026-11-13) shows results/ (4 Parquet files, 0.31 GB, the starting table), coilsets/ (6 Parquet files, 2.36 GB), metrics/ (203 Parquet files, 213 GB of per-coil and per-surface values) and requirements/. Boundaries are referenced by plasma_config_id in ConStellaration, not copied. Open with Hugging Face datasets, pandas or pyarrow. Released by Proxima Fusion under the MIT licence; cite arXiv:2611.12689.

How to load this dataset in Python

from datasets import load_dataset
ds = load_dataset("proxima-fusion/coilstellaration")
print(ds)

Data structure

File types in sampled files: Parquet

Files

Example file: coilsets/id_to_file_map.parquet

Analyze this dataset without downloading it

On Scigantic, CoilStellaration: Optimized Stellarator Coilsets and Coil Design Benchmark opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.

This page is also available as Markdown and as JSON.

Related Energy datasets