iGSE baseline
— Predicted volumetric lossUses waveform, frequency, and material; ignores temperature and DC bias.
Princeton Power Electronics Research Lab
MagNet data and model hub
Open resources for magnetic-material measurements, model development, and core-loss prediction.
01
02
03
04
Data
Raw steady-state and transient measurements, plus training-ready releases.
Models
Available baselines, migrated research models, and validation status.
Predict
Configure a waveform and estimate steady-state magnetic core loss.
Documentation
Data contracts, training examples, model packaging, and usage guides.
Data library
Datasets
Find raw measurements and training-ready releases. Published releases link to immutable files on Hugging Face.
Hugging Face releases
Only verified public releases show download links. Planned releases remain visible with their publishing status.
Loading datasets…
Model library
Models
Browse baselines and trained model packages. Published releases link to pinned artifacts and model cards on Hugging Face.
Hugging Face releases
Downloads are enabled only after the repository, immutable revision, license, and file checksums are verified.
Loading models…
Predict
Compare core-loss predictions
Enter one steady-state operating point to see the iGSE baseline and MagNet Transformer estimates together, plus the Transformer's predicted B–H loop. Model limitations and operating-range warnings appear with each result.
Sinusoidal waveform, 100 mT peak.
MagNet Transformer
— Predicted volumetric lossExperimental legacy model. First run downloads the browser runtime and material model.
Predicted B–H loop
128 paired samples, in cycle order. The trace is not artificially closed; Transformer loss follows the original 127-interval H dB integration.
Documentation
Use MagNet
Follow the shared data and model contracts so experiments remain reproducible and published artifacts remain traceable.
- Review the measurementsUnderstand equipment, geometry, units, calibration, and acquisition metadata.
- Load a datasetUse stable sample identifiers and published train, validation, and test splits.
- Train a baselineRun the included steady-state or transient example with a recorded configuration.
- Package a modelUse the common inference interface and document the model's operating limits.