NodeJS implementation of today's most used tokenizers, with a focus on performance and versatility. Bindings over the Rust implementation. If you are interested in the High-level design, you can go check it there.
- Train new vocabularies and tokenize using 4 pre-made tokenizers (Bert WordPiece and the 3 most common BPE versions).
- Extremely fast (both training and tokenization), thanks to the Rust implementation. Takes less than 20 seconds to tokenize a GB of text on a server's CPU.
- Easy to use, but also extremely versatile.
- Designed for research and production.
- Normalization comes with alignments tracking. It's always possible to get the part of the original sentence that corresponds to a given token.
- Does all the pre-processing: Truncate, Pad, add the special tokens your model needs.
npm install tokenizersimport { BertWordPieceTokenizer } from "tokenizers";
const wordPieceTokenizer = await BertWordPieceTokenizer.fromOptions({ vocabFile: "./vocab.txt" });
const wpEncoded = await wordPieceTokenizer.encode("Who is John?", "John is a teacher");
console.log(wpEncoded.getTokens());
console.log(wpEncoded.getIds());
console.log(wpEncoded.getAttentionMask());
console.log(wpEncoded.getOffsets());
console.log(wpEncoded.getOverflowing());
console.log(wpEncoded.getOriginalString());
console.log(wpEncoded.getSpecialTokensMask());
console.log(wpEncoded.getTypeIds());CharBPETokenizer: The original BPEByteLevelBPETokenizer: The byte level version of the BPESentencePieceBPETokenizer: A BPE implementation compatible with the one used by SentencePieceBertWordPieceTokenizer: The famous Bert tokenizer, using WordPiece