Description
Adversarial examples can be generated against PyTorch models for robustness research and evaluation. Researchers and ML engineers can test classifiers against perturbation methods and compare defensive techniques. It is dual-use model-evaluation tooling and should be used on models and data you are authorized to test.