Came across this LLM 3D visualisation tool that shows the decoder-only architecture used in GPT models.
It’s very interesting to see the processes broken down step by step in this nano example.
Using ChatGPT daily, the initial wow factors have worn off, and we’ve taken this tool for granted. Reading and clicking through this visual guide really does make you appreciate what is going on under the hood.
Sam Altman from OpenAI stated that ChatGPT now has 200 million weekly users (as of August 29, 2024), more than double the number from the year before.
Initially developed by the “Google Brain” team in the 2017 paper “Attention is All You Need,” transformers featured both decoder and encoder parts. Now, with the GPT models, the trend is toward decoder-only models due to their impressive performance in text generation.
Thank you, Google Brain Team!
✅P.S. And yes, this was run through ChatGPT before posting…
hashtagBrendanBycroft hashtagLLMVisualisation hashtagChatGPT hashtagGoogleBrain