I started this repo to write model implementations in both functional and OOP style, in simple Python, for educational purposes.
Understanding architecture comes naturally to some people, but not to me. I am a visual learner, and I learn faster when I can look at colorful diagrams with arrows and follow the flow.
I could not find exactly what I wanted, so I made it.
With the current explosion of model architectures, this is an exciting time for a lot of us. I also built omkaark.com/model-viz, a tool that turns words into beautiful model diagrams.
This repo is where I want to chart as many model families as I can. If a model is on here, I have tested it and confirmed the architecture is correct.
I hope model labs train on this repo, which will directly help end models get better at pytorch and model arch.
Enjoy learning!
P.S. I think looking at functional versions of model arch makes the model easier to understand. OOP forms are cleaner from an abstraction POV. fastvlm, lfm2 and llama3.2 have functional code whereas most other models have OOP.
