{"success":true,"data":{"author":"Proxima Fusion","bucketAccessType":"public","bucketName":"proxima-fusion/coilstellaration","bucketPrefix":"","category":"Energy","cloudProvider":"huggingface","createdAt":"2026-11-14T01:01:18.700Z","description":"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.","displayTitle":"CoilStellaration: Optimized Stellarator Coilsets and Coil Design Benchmark","id":"abe36dea-ff32-5e26-a989-482107361213","license":"MIT","licenseEvidenceUrl":"https://huggingface.co/datasets/proxima-fusion/coilstellaration","objectCountEstimate":220,"probeStatus":"ok","referenceUrl":"https://arxiv.org/abs/2611.12689","region":"us","schemaCard":{"accessHints":[{"language":"python-datasets","snippet":"from datasets import load_dataset\nds = load_dataset(\"proxima-fusion/coilstellaration\")"},{"language":"python-hub","snippet":"from huggingface_hub import snapshot_download\nsnapshot_download(repo_id=\"proxima-fusion/coilstellaration\", repo_type=\"dataset\")"}],"caveats":["HuggingFace metadata: author=proxima-fusion, license:mit, size_categories:1M<n<10M, format:parquet, downloads=2751"],"format":"mixed","generatedAt":"2026-11-14T01:01:19.363Z","layout":{"example":"coilsets/id_to_file_map.parquet","pattern":"{repoId}/{filename}"},"mountPath":"/mnt/archive","sampleHeaders":[{"bytes":"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","encoding":"hex","key":"coilsets/id_to_file_map.parquet","truncatedAt":512}],"sampleObjects":["assets/coilstellaration.gif","coilsets/id_to_file_map.parquet","coilsets/part-0.parquet","coilsets/part-1.parquet","coilsets/part-2.parquet"],"sidecars":[{"bytes":"---\nconfigs:\n- config_name: results\n  default: true\n  data_files:\n  - split: train\n    path: results/train-part-*.parquet\n  - split: eval\n    path: results/eval-part-*.parquet\n- config_name: metrics\n  data_files: metrics/part-*.parquet\n- config_name: requirements\n  data_files: requirements/part-*.parquet\n- config_name: coilsets\n  data_files: coilsets/part-*.parquet\nlicense: mit\nlanguage:\n- en\ntags:\n- coilsets\n- stellarators\n- fusion\n- optimization\n- engineering\npretty_name: CoilStellaration\nsize_categories:\n- 1M<n<10M\n---\n\n# CoilStellaration\n\n![DESC-optimized coilsets with 4 to 8 coils per half-period around ConStellaration plasma boundaries with 2 to 5 field periods, colored by coil current](assets/coilstellaration.gif)\n\nCoilStellaration is a dataset of optimized stellarator coilsets for 14,496 plasma boundaries from [ConStellaration](https://huggingface.co/datasets/proxima-fusion/constellaration), together with a benchmark for models that design coils.\n\nEach sample pairs a plasma boundary with a randomly sampled set of engineering requirements, such as how far the coils must stay from each other and from the plasma. For every sample we designed a coilset with REGCOIL and, for two thirds of them, refined it with DESC. We then measured how accurately each coilset reproduces the plasma's magnetic field and how buildable it is.\n\n## Dataset Details\n\nThe plasma of a stellarator is shaped by the magnetic field of external coils. Once a plasma boundary is chosen, the coils have to reproduce its field accurately while staying buildable: spaced apart, clear of the plasma, and without sharp bends. Finding such coils takes an expensive optimization for every new design. This dataset lets optimization and machine learning researchers learn that step, for example to propose coils for a new plasma boundary in one shot or to warm-start a coil optimizer.\n\n| | |\n| --- | --- |\n| Plasma boundaries | 14,496, with 2 to 5 field periods |\n| Samples | 266,549: 253,250 `train`, 13,299 `eval`, with no plasma boundary in both |\n| DESC-optimized coilsets | 177,341 |\n| REGCOIL coilsets | 266,549 |\n| Coils per half-period | 4 to 8 |\n\n- **Curated by:** Proxima Fusion\n- **License:** MIT\n- **Code:** https://github.com/proximafusion/coilstellaration\n- **Paper:** https://arxiv.org/abs/2611.12689\n- **Plasma boundaries and equilibria:** https://huggingface.co/datasets/proxima-fusion/constellaration\n- **Benchmark leaderboard:**: https://huggingface.co/spaces/proxima-fusion/coilstellaration-bench\n\n### Coilset metrics\n\nFour metrics describe a coilset throughout the dataset and the benchmark:\n\n| Metric | Meaning | Better |\n| --- | --- | --- |\n| Field error | \\|B·n\\| / \\|B\\| on the plasma boundary: how much the coils' field crosses the boundary instead of running along it | Lower |\n| Coil-coil distance | Smallest gap between any two coils | Higher |\n| Coil-plasma distance | Smallest gap between any coil and the plasma | Higher |\n| Curvature | Sharpest bend of any coil | Lower |\n\nDistances and curvatures are normalized by the plasma's minor radius (`normalized_*`), so devices of different sizes compare directly.\n\n## Dataset Structure\n\n| Subset | Rows | Size | Contents |\n| --- | ---: | ---: | --- |\n| **results** (default) | 266,549 | 0.3 GB | One row per sample: its requirements, summaries of its metrics, benchmark labels, and the IDs of its full objects in the other subsets. **Start here.** |\n| **coilsets** | 443,890 | 2.3 GB | Coil shapes and currents, REGCOIL and DESC |\n| **requirements** | 266,549 | 0.08 GB | The sampled requirements |\n| **metrics** | 343,272 | 213 GB | Full metrics of each coilset: values along every coil and over the plasma boundary |\n\n**coilsets**, **requirements** and **metrics** each hold an `id` and a `json` column. The plasma boundaries are not copied here: `constellaration_boundary_id` is the `plasma_config_id` of a boundary in ConStellaration.\n\nThe main columns of **results**:\n\n| Column | Contents |\n| --- | --- |\n| `constellaration_boundary_id` | The plasma boundary, in ConStellaration |\n| `regcoil_coilset_id`, `desc_coilset_id` | T","encoding":"utf8","key":"README.md","name":"README.md","size":16435,"truncated":true}],"sourceObjectKey":"coilsets/id_to_file_map.parquet","starterCell":"from datasets import load_dataset\nds = load_dataset(\"proxima-fusion/coilstellaration\")\nprint(ds)","starterSamplePath":"coilsets/id_to_file_map.parquet","totalObjects":220},"summary":"Optimized stellarator coilsets for 14,496 plasma boundaries from ConStellaration (2 to 5 field periods), with a benchmark for models that design coils.","title":"CoilStellaration: Optimized Stellarator Coilsets and Coil Design Benchmark","updatedAt":"2026-11-14T02:37:13.987Z","visibility":"public","websiteUrl":"https://github.com/proximafusion/coilstellaration"}}