This is a repository for the term project in IEOR 4540 at Columbia University.

The Data Mining (Transformers) course covers vision transformer theory and applications.

data

• Contains model checkpoints, history log files, and sample images for part of the evaluation process. Also has params.json with optimal model hyperparameters.

src

• Contains util.py with helper functions as well as vit.py with Vision Transformer and Performer-f implementations. Also contains a Trainer class to perform model training.

notebooks

• For model training and data analysis. The goal of this project is to benchmark various transformers.