Building an LLM on ChatGPT-2 level.
Build so it can be trained and ran locally on ordinary hardware.
Creating an LLM from scratch