# SentencePiece

A library that trains subword tokenizers directly from raw text, without pre-splitting into words.

## What it is used for

Tokenizers for multilingual and domain-specific models.