lebchen/nb-sbert-base-onnx

ONNX version of NbAiLab/nb-sbert-base

This repository contains ONNX-converted weights for NbAiLab/nb-sbert-base, compatible with Transformers.js.

It includes both:

1. Quantized (int8): Faster, smaller (default).
2. Full Precision (float32): Higher theoretical accuracy, larger file size.

Usage (Node.js)

First, install the library:

npm install @huggingface/transformers

Option 1: Use Quantized Model (Recommended)

This is the default behavior. It loads model_quantized.onnx (approx. 4x smaller).

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

const embedder = await pipeline(
  'feature-extraction', 
  'lebchen/nb-sbert-base-onnx', 
  { device: 'auto' } // Defaults to { quantized: true }
);

const output = await embedder("Dette er en test.", { pooling: 'mean', normalize: true });

Option 2: Use Full Precision Model

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

const embedder = await pipeline(
  'feature-extraction', 
  'lebchen/nb-sbert-base-onnx', 
  { 
    device: 'auto', 
    quantized: false // Forces loading of model.onnx
  }
);

Credits & License

The original model was developed by the National Library of Norway (AI Lab). Original repository: https://huggingface.co/NbAiLab/nb-sbert-base

This distribution is licensed under Apache 2.0.