OpenMed-PII-French-SuperClinical-Small-44M-v1-ONNX

OpenMed-PII-French-SuperClinical-Small-44M-v1 (ONNX)

ONNX export of OpenMed/OpenMed-PII-French-SuperClinical-Small-44M-v1, ready to run in the browser (or Node.js) with Transformers.js.

This is a DeBERTa-v2 token-classification model for French clinical PII / NER detection (76 labels).

Files

Usage (Transformers.js)

import { pipeline } from '@huggingface/transformers';

const pipe = await pipeline(
  'token-classification',
  'OpenMed-PII-French-SuperClinical-Small-44M-v1-ONNX',
  { dtype: 'q8' }   // use 'fp32' for full accuracy
);

const result = await pipe(
  "Le patient Jean Dupont, né le 12/03/1980, habite à Lyon."
);
console.log(result);
// [
//   { entity: 'B-FIRSTNAME', word: 'Jean', ... },
//   { entity: 'B-LASTNAME',  word: 'Dupont', ... },
//   { entity: 'B-DATEOFBIRTH', word: '12', ... },
//   ...
// ]

Conversion

Exported with 🤗 Optimum:

optimum-cli export onnx \
  --model OpenMed/OpenMed-PII-French-SuperClinical-Small-44M-v1 \
  --task token-classification \
  output/

The int8 variant was produced with ONNX Runtime dynamic quantization (QUInt8 weights).