Get the latest tech news
A nanoGPT pipeline packed in a spreadsheet
A nanoGPT pipeline packed in a spreadsheet. Contribute to dabochen/spreadsheet-is-all-you-need development by creating an account on GitHub.
While reading about LLMs, I realised that the internal mechanisms of a transformer is basically a range of matrices calculations being connected in a certain order. Firstly, all the blocks are the values or parameters that is processed through the GPT architecture, they are being color coded as purple, green and orange. Lastly, this demo is built with great help from the LLM visualization project ( https://bbycroft.net/llm) by Brendan Bycroft which uses 3D animations to explain transformers.
Or read this on Hacker News
