tokenizers==0.22.*
regex
requests
markdownify
beautifulsoup4
lxml
tiktoken   # only for analysis.py (comparison with OpenAI tokenizers)
