Andrej Karpathy
5 HN recommenders · recommended in 6 months, 2024–2025
2:13:35 · 1.3M views · Feb 20, 2024
Andrej Karpathy builds the GPT tokenizer from scratch, implementing byte pair encoding along with encode and decode. He shows how many odd LLM behaviors trace back to tokenization, and covers Unicode byte encodings along the way.
Recommended as the best deep dive on tokenizers, worth the two hours for anyone working with LLMs.
8 comments from Hacker News
I finally understood the weirdness of tokenizers after watching the video Andrej Karpathy made: "Let's build the GPT Tokenizer" ( this video ). He goes through why we need them instead of raw byte sequences (too expensive) and how the Byte Pair Encoding algorithm works. Worth spending 2h for the deeper understanding if you deal with LLMs.
Andrej Karpathy goes into quite some depth about how tokenization works here: this video tl;dr many of the LLMs use byte-pair encoding to create tokens. You take a set of documents, and then form tokens by repeatedly merging the most common pair of tokens. The initial set of tokens is 256 raw bytes. And the text is typically represented in utf-8. I expect that although the LLMs can understand arbitrarily but cleanly offset unicode code points by (eventually) noticing the final byte of each sequence, they would do markedly worse on actually processing and completing on them, because they will not have been reduced to the normal set of tokens. However, if the text is actually output cleanly converted, either in internal thinking tokens or in the beginning of the response, they should do fine. Understanding tokenization is surprisingly useful, even if that video seems awfully long to devote to such a tedious subject. Even Karpathy doesn't like it!
ChatGPT is specifically bad at these kinds of tasks because of tokenization. If you plug your query into [LINK] , you can see that"egregious" is a single token, so the LLM doesn't actually see any "e" characters -- to answer your question it would have had to learn a fact about how the word was spelled from it's training data, and I imagine texts explicitly talking about how words are spelled are not very common. Good explanation here if this still doesn't make sense: [LINK] , or check out Andrej Karpathy's latest video if you have 2 hours for a deep dive: this video IMO questions about spelling or number sense are pretty tired as gotchas, because they are all basically just artifacts of this implementation detail. There are other language models available that don't have this issue. BTW this is also the reason DALL-E etc suck at generating text in images.
This is just an artefact of tokenisation though. The model simply isn’t ever shown the letters that make up words, unless they are spelled out explicitly. It sees tokens representing groups of words. This is a little like saying a human isn’t intelligent because they couldn’t answer your question that you asked in an ultrasonic wavelength. If you’d like to learn more this video is a great resource: this video
Karpathy's "Let's build the GPT Tokenizer" also contains a good introduction to Unicode byte encodings, ASCII, UTF-8, UTF-16, UTF-32 in the first 20 minutes: this video