/mockGPT

an attempt to create a Transformer based LLM, like GPT from scratch

MIT LicenseMIT

mockGPT

The whole world is taken by storm with the launch of ChatGPT by open.ai.

While big Tech giants Google, Microsoft, Apple, Meta, ... are jumping into the space to make an AGI or LLM based products, this project is a small attempt to create a Transformer based LLM, like GPT from scratch.

This project is heavily inspired by Andrej Karpathy's series on Neural Networks, and the video Let's build GPT

Note: The chances of succeeding with this project is low in my first attempt, and even if I do, I believe that the model will turn out to be small, with limited training dataset. Hence there should't be any safety concerns as Sam Altman describes in Lex Fridman Podcast #367, in the section: Fear (1:09:05)