0.0.2 • Published 3 months ago
minbpe v0.0.2
Minbpe
This package is a port of https://github.com/karpathy/minbpe in pure JavaScript. Right now, it only supports server environments, but we will also grant support for browsers too.
Usage
Install minbpe with npm:
npm i minbpe
Use in your JavaScript file:
import { BasicTokenizer } from 'minbpe'
const tokenizer = new BasicTokenizer()
const text = 'The quick brown fox jumps over the lazy dog'
// optionally train a tokenizer:
// tokenizer.train(text, 256 + 3)
const ids = tokenizer.encode(text)
const decoded = tokenizer.decode(ids)
Progress status
- Base Tokenizer
- Regex Tokenizer (work in progress)
GPT4 Tokenizer (work in progress)
Node.js support
- Deno support
- Bun support
- Cloudflare Workers support (work in progress)