How to use fp16 version of the model file?

[cyio]

Question

example files: https://huggingface.co/Xenova/modnet/tree/main/onnx

[xenova]

This requires the experimental v3 branch: #545
where you can specify dtype:

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

// Create feature extraction pipeline
const extractor = await pipeline('feature-extraction', 'Xenova/all-MiniLM-L6-v2', {
    // device: 'webgpu', (Optional)
    dtype: 'fp32', // or 'fp16'
});

[cyio]

step:

1. npm install xenova/transformers.js#v3
2. for vite project, resolve Top await error with vite build error -- Transform failed with 3 errors - Top-level await is not available in the configured target environment remix-run/remix#7969 (comment)