Together with my colleague Alexander Neubeck we wrote a blog post about a fast byte-pair encoding algorithm he developed. Read it here: So many tokens, so little time: Introducing a faster, more flexible byte-pair tokenizer.