A 🤗-compatible version of the
GPT-4 tokenizer (adapted from
openai/tiktoken). This means it can be used with Hugging Face libraries including
Transformers,
Tokenizers, and
Transformers.js.
1from transformers import GPT2TokenizerFast
2
3tokenizer = GPT2TokenizerFast.from_pretrained('Xenova/gpt-4')
4assert tokenizer.encode('hello world') == [15339, 1917]
1import { AutoTokenizer } from '@xenova/transformers';
2
3const tokenizer = await AutoTokenizer.from_pretrained('Xenova/gpt-4');
4const tokens = tokenizer.encode('hello world'); // [15339, 1917]
