distilbert-base-spanish-uncased-finetuned-qa-mlqa-onnx

Model Description

Spanish DistilBERT model fine-tuned for question answering on the MLQA dataset, exported to ONNX and quantized for use with Transformers.js.

Files

• config.json
• tokenizer.json
• tokenizer_config.json
• vocab.txt
• onnx/model_quantized.onnx

Live Demo

This model is deployed in production at plaudere.com as part of a bilingual Question Answering system.
Visitors can seamlessly switch between the Spanish QnA model (this one) and an English QnA model via the website’s language selector.
You can see both models in action by requesting support directly on the site.

For a guided experience, explore this Plaudere article:
AI Assistance with QnA Models
It demonstrates how the author integrated QnA models into a real-world application on the Plaudere website.
Here, you can interact with the agent, ask questions in English about the article itself, and watch it respond in real time.

Usage in Transformers.js

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

// Load the Spanish QA pipeline
pipeline('question-answering', 'jestevesv/distilbert-base-spanish-uncased-finetuned-qa-mlqa-onnx', { quantized: true })
  .then(qa => qa(
    '¿Cuál es la capital de Francia?', 
    'La capital de Francia es París.'
  ))
  .then(result => console.log(result))
  .catch(err => console.error(err));