https://huggingface.co/WhereIsAI/UAE-Large-V1 with ONNX weights to be compatible with Transformers.js.
Usage (Transformers.js)
If you haven't already, you can install the Transformers.js JavaScript library from NPM using:
npm i @huggingface/transformers
You can then use the model to compute embeddings like this:
import { pipeline } from '@huggingface/transformers';
// Create a feature-extraction pipeline
const extractor = await pipeline('feature-extraction', 'Xenova/UAE-Large-V1', {
dtype: "fp32" // Options: "fp32", "fp16", "q8", "q4"
});
// Compute sentence embeddings
const sentences = ['That is a happy person', 'That is a very happy person'];
const output = await extractor(sentences, { pooling: 'cls' });
console.log(output);
// Tensor {
// dims: [ 2, 1024 ],
// type: 'float32',
// data: Float32Array(2048) [ -0.1308155655860901, 0.44334232807159424, ... ],
// size: 2048
// }
Compute cosine similarity between the two sentences:
import { cos_sim } from '@huggingface/transformers';
console.log(cos_sim(output[0].data, output[1].data))
// 0.9586893906734091
You can convert the output Tensor to a nested JavaScript array using .tolist():
console.log(output.tolist());
// [
// [ -0.1308155655860901, 0.44334232807159424, -0.12212765961885452, ... ],
// [ 0.03931744396686554, 0.30553528666496277, -0.19462820887565613, ... ]
// ]
Note: Having a separate repo for ONNX weights is intended to be a temporary solution until WebML gains more traction. If you would like to make your models web-ready, we recommend converting to ONNX using 🤗 Optimum and structuring your repo like this one (with ONNX weights located in a subfolder named onnx).