Tiktoken

tiktoken is a fast BPE tokeniser created by OpenAI.

1. How the text is split: by tiktoken tokens

2. How the chunk size is measured: by tiktoken tokens

#!pip install tiktoken

# This is a long document we can split up.
with open('../../../state_of_the_union.txt') as f:
    state_of_the_union = f.read()

from langchain.text_splitter import TokenTextSplitter

text_splitter = TokenTextSplitter(chunk_size=10, chunk_overlap=0)

texts = text_splitter.split_text(state_of_the_union)
print(texts[0])

Madam Speaker, Madam Vice President, our