auto-attack
auto-attack copied to clipboard
Evaluation of Defended Models
Is there any way I can provide my robust WRN-28-10 model and can get the AA adversarial accuracy on that model?