Coilstellaration Bench

Coilstellaration Bench

Scores ONNX models that predict a stellarator's coilset from a plasma boundary and requirement metrics, so researchers can submit models and compare their results.

Made by proxima-fusion on Hugging Face. These notes are Mindket's, written from its Space page and README; the app's own files and license are on Hugging Face.

About

What it is: This is a leaderboard for ONNX models that design the coils of a stellarator. Given the shape of a plasma boundary and a list of engineering requirements, each model outputs a predicted coilset. The Space accompanies a research paper on a dataset and benchmark for this task.

Submissions go into a request queue dataset, an evaluator Space scores them, and the results are published to a results dataset.

Who it suits: It suits researchers and engineers who train models for stellarator design and want a shared way to compare them.

Industry | Who uses it | How they use it | Example
Fusion research | Stellarator design researchers | Submit an ONNX model and read how it scores | A group submits a model and reads its published results
Machine learning | ML engineers | Train on the coilstellaration dataset and benchmark the model here | An engineer checks a new model against the scoring rules
Engineering | Coil design engineers | Read the scoring and the ONNX contract in the About tab | An engineer studies how a design is judged
Education | Students | Study a dataset and benchmark project in a niche field | A student follows how a submission moves from queue to results

You can skip it if you do not work with stellarators or ONNX models.

Where to use it: You use it in these places.

• A web browser, on the Space's page on Hugging Face
• A local copy on Windows, macOS or Linux, which needs Python 3.12
• The Space's About tab, for the ONNX contract, the scoring and the citation

Use case | Who | What they get back
Compare models | Researchers | Scores for ONNX models on the benchmark
Check the rules | Model builders | The ONNX contract and scoring in the About tab
Run the board locally | Developers | A copy of the Gradio leaderboard app on their own computer

How to set it up: The online page needs only a web browser. A local copy needs Python 3.12 and pip.

• A web browser, for the online page
• Python 3.12 with pip, for a local copy
• Python packages from requirements.txt
• System packages from packages.txt, which Hugging Face installs with apt-get

1. Open the Space's page on Hugging Face to use it online, with nothing to install.
2. For a local copy, install the Hugging Face Hub package.
3. Download the Space's files into a folder and move into it.
4. Install Gradio 6.17.3 and the packages in requirements.txt.
5. Start the app.

pip install -U huggingface_hub
hf download proxima-fusion/coilstellaration-bench --repo-type space --local-dir coilstellaration-bench
cd coilstellaration-bench
pip install gradio==6.17.3
pip install -r requirements.txt
python leaderboard/app.py

How to use it: Start the app, then open the Gradio page it serves in your browser. Read the About tab for the ONNX contract, the scoring and the citation.

python leaderboard/app.py

Tool | What it does
hf download | Copies the Space's files to your computer
pip install -r requirements.txt | Installs the Python packages the app needs
python leaderboard/app.py | Starts the leaderboard app on your computer

Example: You run python leaderboard/app.py after installing the packages. The Gradio app starts on your computer, and you open it in your browser to see the leaderboard and its About tab.

For newcomers and experts: The page ranks models that take a plasma boundary plus some requirements and answer with a predicted coilset. Start by opening the Space's page on Hugging Face and reading the About tab.

Under the hood, models must be ONNX files that follow the contract in the About tab. A request is queued in one dataset, an evaluator Space scores it, and the result is published to another dataset. The Space is synced from a GitHub project and runs on cpu-basic hardware at Hugging Face.

Submit an ONNX model
Request is queued in the requests dataset
Evaluator Space scores it
Result is published to the results dataset
Leaderboard shows the scores

Good to know: The page is a leaderboard first, and its rules live in the About tab.

• License: mit.
• Cost to run: none named. The Space itself uses cpu-basic hardware at Hugging Face.
• Data it reaches: the requests and results datasets on Hugging Face, plus the coilstellaration dataset it names.
• Not for: editing the README here, since it is generated from leaderboard/space.leaderboard.md.

On Hugging Face, 13 November 2026

• Likes: 4
• Built with: gradio
• License: mit

Links

• Space: https://huggingface.co/spaces/proxima-fusion/coilstellaration-bench
• Files: https://huggingface.co/spaces/proxima-fusion/coilstellaration-bench/tree/main

Try it

Open the Space in your browser, or copy it to your own Hugging Face account from its page. Nothing in this kit runs by itself.