Build your own deep learning library

Hi there, I made a small and easy to read deep learning library from scratch using only numpy and python. The goal is to demystify how auto-diff engine such as Pytorch works under the hood. It teaches how backpropagation operate in practice using vector-Jacobian products. It features examples written with the library of implementation for popular algorithms such as GANs, digit classification and PPO for Reinforcement Learning. There is an API doc and a blog post explaining the implementation. It is still in active development as I will keep adding more features (GPU support and more operations). The end goal is to be able to build any of the recent models (LLMs, diffusion models) using this lib, allowing a user to dive into the low level implementation details as will. Any interest in it ? [link to the repo]( https://github.com/sabeaussan/EaZyGrad?tab=readme-ov-file) [link to the doc](https://eazygrad.readthedocs.io/en/latest/index.html)