Implementation of average- and worst-case robust flatness measures for adversarial training.
Primary LanguagePython