The original PyTorch implementation of the EXACT loss function for accuracy optimization via gradient descent.
See EXACT: How to Train Your Accuracy (2022) for more details.
The short version of the paper was presented at the TAG in Machine Learning ICML Workshop (2022).
If you use EXACT in your research, please cite the paper above.