This is a write up of one of my first machine learning side projects that I am particularly proud of, a protein structure prediction model based on AlphaFold 3 designed to be trained on a single mid tier GPU (RTX 3060 in my case).
I talk about the successes and failures of this project, in particular its merits and shortcomings as a first project for learning ML.
For those more curious about technical details, I believe this is one of the first implementations of AlphaFold 3 in PyTorch (adapted to only consider proteins and ignoring other molecules) - see https://github.com/ogchen/nanofold for the project code if you are curious!
I talk about the successes and failures of this project, in particular its merits and shortcomings as a first project for learning ML.
For those more curious about technical details, I believe this is one of the first implementations of AlphaFold 3 in PyTorch (adapted to only consider proteins and ignoring other molecules) - see https://github.com/ogchen/nanofold for the project code if you are curious!