SeizureTransformer (competition checkpoint) โ€” research-only release

This repository contains a converted model.safetensors checkpoint for the SeizureTransformer time-step seizure detection model.

โš ๏ธ Critical safety and regulatory notice

NOT FDA APPROVED. NOT FOR MEDICAL USE.

  • This model is NOT approved by the U.S. FDA (or any medical regulator) for clinical use.
  • Do NOT use this model for diagnosis, treatment decisions, triage, emergency response, or any real-world patient care.
  • Outputs can be wrong, incomplete, biased, unstable, or unsafe.
  • Human expert oversight is mandatory in all contexts.

Research-only and non-commercial terms

By using these weights, you agree:

  1. Research/Education only โ€” no clinical deployment.
  2. No commercial use without explicit written permission from original rights holders/authors.
  3. No harmful use โ€” do not use this model to harm, surveil, discriminate against, or otherwise negatively impact people.
  4. No safety-critical use โ€” no autonomous decision-making affecting human health/safety.

If these terms conflict with upstream licensing/rights, upstream rights prevail.

Provenance

  • Source project: keruiwu/SeizureTransformer
  • Public container source used to obtain checkpoint: yujjio/seizure_transformer
  • Extracted checkpoint path in container: wu_2025/model.pth
  • Converted to model.safetensors for Rust/Burn inference.

Files

  • model.safetensors โ€” converted checkpoint
  • config.json โ€” model architecture/inference config
  • LICENSE โ€” research-only usage notice for this distribution

Intended use

  • Offline research benchmarking
  • Reproducibility and parity testing
  • Non-clinical experimentation

Out of scope / prohibited

  • Clinical workflows
  • Medical device integration
  • Commercial product integration without permission
  • Any use intended to harm humans
Downloads last month
10
Safetensors
Model size
44.1M params
Tensor type
F32
ยท
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support