Computationally Efficient Visual Transformers (Visage Technologies internship)

We implemented quantization methods for visual transformers and analyzed tradeoffs between model performance and accuracy (Visage Technologies internship)

To reproduce results, consult the scripts/ directory. The general workflow is:

1. Finetune the models (start_finetune.sh small finetune-small etc.)
2. Run the experiments (start_experiments.sh)