1 article in this category
This article details training a WordPiece tokenizer for BERT models, achieving a vocabulary size of 30,522 tokens.