Skip to content
Other

CoilStellaration ResMLP: Physics-Informed Model for Magnetic Coil Calculation

2 min read

This model is a fast ResMLP (residual multilayer perceptron) architecture trained to solve plasma magnetic confinement problems for stellarators within the CoilStellaration benchmark. In a single forward pass (without iterative loops, taking roughly 1.8 ms per coil set), it predicts configuration and currents while strictly accounting for physical constraints like curvature and clearances. Built on 6 layers with 3.22 million parameters, the network is designed for fusion energy engineers and researchers who need instant calculation of magnetic system geometry.

Key features

  • ResMLP architecture: a 6-layer pre-normalized network weighing 3.22M parameters with a size of 12.32 MiB in ONNX.
  • Physics-informed training: two-phase training with JAX losses based on the Biot-Savart law and geometric constraints.
  • Zero validation leakage: strict training on the train split with a safety margin for physical clearances.
  • High speed: instant calculation in ~1.8 ms thanks to eliminating iterative loops (Loop = 0).

What you can do

  • Integrate the model into CoilStellaration benchmark pipelines for rapid evaluation of coil shapes in the fixed_shape track.
  • Perform preliminary filtering and optimization of stellarator coil geometry prior to heavy numerical simulations.
  • Explore physical metrics for magnetic field accuracy and plasma clearances in real time.

Example

Using the predictor for instant evaluation of coil configurations and currents in the fixed_shape track based on an ONNX model for plasma physics and stellarator tasks.

Read moreDetails
Average local eval score: 0.3704
Strict feasibility: 21.67% (loose 58.0%, medium 7.0%)
Feasibility: coil–coil 92.3%, coil–plasma 93.0%, curvature 97.0%
Training: Apple Silicon CPU, ~7.3 minutes in two phases
Supported track: fixed_shape (N_fp = 3)
Discuss Telegram 0

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 0.0 GB · ready files0.5 GB
  • FP8 / INT8recommendedweights 0.0 GB0.5 GB
  • 4-bitweights 0.0 GB0.5 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · BF16 — from RTX 2060

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # petro-popov/coilstellaration-direct-phys-resmlp
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download petro-popov/coilstellaration-direct-phys-resmlp --local-dir ~/models/coilstellaration-direct-phys-resmlp
    # The model is in ~/models. How to run it is in the model card (link above).
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of coilstellaration-direct-phys-resmlp (quants, finetunes, LoRAs) will show up under the bell.

No known relatives of this model yet. All derivatives on Hugging Face

About the model

petro-popov/coilstellaration-direct-phys-resmlp
Author
petro-popov
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

00

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments