This project was made for studying LLMs and how they're implemented without any abstraction libraries such as those made by Hugging Face. The goal is to implement a basic GPT-2-esque LLM with libraries consisting of the standard library, PyTorch and Pandas. The idea of using 4chan as the basis for the training dataset came from this Kitboga-lookalike Yannic Kilcher's YouTube video.
ymcx/llmfs
Folders and files
| Name | Name | Last commit date | ||
|---|---|---|---|---|