Code to run the computer vision benchmark in the paper "Rethinking Early Stopping: Refine, Then Calibrate" https://arxiv.org/abs/2501.19195.
main.py: Launch runs and log results.utils.py: Contains our pytorch-lightning module that allows benchmarking TS-Refinement against other early stopping metrics.figures.ipynb: Generate figures for the paper.resnet.pyandwide_resnet.py: Deep learning model used, from https://github.com/uoguelph-mlrg/Cutout