Open TalRub104 opened 3 days ago
Hello, you may leave your email address, and we can use it to get in touch. I believe this will facilitate receiving notifications and solving problems more conveniently.
Hi, My email is:
talr5@mail.tau.ac.il
Thanks!
Hi, The scripts train_trades_mnist.py and train_trades_cifar10.py don't include the following: Code to save the best checkpoints The folllwing evaluation:
train_loss, train_acc = eval_train(args, model, device, train_loader) test_loss, test_acc = eval_test(args, model, device, test_loader) adv_test_loss, adv_test_acc = adv_test(args, model, device, test_loader, args.attack_type)
*Call for AA_eval(model, args)
Additionally, I don't understand why the AA_adv_test function doesn't perform adversarial accuracy evaluation as the adv_test function does, nor why it doesn't return any value or print any result.