import waitGPU waitGPU.wait(utilization=20, available_memory=10000, interval=10) import problems as pblm from trainer import * import setproctitle if __name__ == "__main__": args = pblm.argparser(prefix='mnist', method='task_spec_robust', opt='adam', starting_epsilon=0.05, epsilon=0.2, thres=0.035) kwargs = pblm.args2kwargs(args) setproctitle.setproctitle('python') print("threshold for classification error: {:.1%}".format(args.thres)) print('Matrix type: {0}\t\t' 'Category: {1}\t\t' 'Epoch number: {2}\t\t' 'Targeted epsilon: {3}\t\t' 'Starting epsilon: {4}\t\t' 'Sechduled length: {5}'.format(args.type, args.category, args.epochs, args.epsilon, args.starting_epsilon, args.schedule_length), end='\n') if args.l1_proj is not None: print('Projection vectors: {0}\t\t' 'Train estimation: {1}\t\t' 'Test estimation: {2}'.format(args.l1_proj, args.l1_train,
model = pblm.mnist_model_wide(args.model_factor).cuda() elif m == 'deep': #print("Using deep model with model_factor={}".format(args.model_factor)) #_, test_loader = pblm.mnist_loaders(64//(2**args.model_factor)) #model = pblm.mnist_model_deep(args.model_factor).cuda() print('using customised deep model') model = pblm.mnist_model_deep_custom().cuda() elif m == '500': model = pblm.mnist_500().cuda() else: model = pblm.mnist_model().cuda() return model if __name__ == "__main__": args = pblm.argparser(opt='adam', verbose=200, starting_epsilon=0.01) print("saving file to {}".format(args.prefix)) setproctitle.setproctitle(args.prefix) train_log = open(args.prefix + "_train.log", "w") test_log = open(args.prefix + "_test.log", "w") train_loader, _ = pblm.mnist_loaders(args.batch_size) _, test_loader = pblm.mnist_loaders(args.test_batch_size) torch.manual_seed(args.seed) torch.cuda.manual_seed(args.seed) for X, y in train_loader: break kwargs = pblm.args2kwargs(args, X=Variable(X.cuda())) best_err = 1
from cifar import select_model import problems as pblm from trainer import * import setproctitle import random if __name__ == "__main__": args = pblm.argparser(prefix='cifar', method='task_spec_robust', epsilon=0.03486, l1_proj=50, l1_train='median', starting_epsilon=0.001, opt='sgd', lr=0.05, thres=0.35) kwargs = pblm.args2kwargs(args) setproctitle.setproctitle('python') print("threshold for classification error: {:.1%}".format(args.thres)) print('Matrix type: {0}\t\t' 'Category: {1}\t\t' 'Epoch number: {2}\t\t' 'Targeted epsilon: {3}\t\t' 'Starting epsilon: {4}\t\t' 'Sechduled length: {5}'.format(args.type, args.category, args.epochs, args.epsilon, args.starting_epsilon, args.schedule_length), end='\n') if args.l1_proj is not None:
def select_model(m): if m == 'large': # raise ValueError model = pblm.cifar_model_large().cuda() elif m == 'resnet': model = pblm.cifar_model_resnet(N=args.resnet_N, factor=args.resnet_factor).cuda() else: model = pblm.cifar_model().cuda() return model if __name__ == "__main__": args = pblm.argparser(epsilon=0.0347, starting_epsilon=0.001, batch_size=50, opt='sgd', lr=0.05) print("saving file to {}".format(args.prefix)) setproctitle.setproctitle(args.prefix) train_log = open(args.prefix + "_train.log", "w") test_log = open(args.prefix + "_test.log", "w") train_loader, test_loader = pblm.cifar_loaders(args.batch_size) torch.manual_seed(args.seed) torch.cuda.manual_seed_all(args.seed) random.seed(0) numpy.random.seed(0)
import waitGPU waitGPU.wait(utilization=20, available_memory=10000, interval=10) import problems as pblm from trainer import * import setproctitle if __name__ == "__main__": args = pblm.argparser(prefix='mnist', opt='adam', starting_epsilon=0.05, epsilon=0.2, thres=0.04) kwargs = pblm.args2kwargs(args) setproctitle.setproctitle('python') print("saving file to {}".format(args.proctitle)) if args.method == 'overall_robust': print("threshold for classification error: {:.1%}".format(args.thres)) elif args.method != 'baseline': raise ValueError("Unknown training method.") saved_filepath = ('../saved_log/' + args.proctitle) model_filepath = os.path.dirname('../models/' + args.proctitle) if not os.path.exists(saved_filepath): os.makedirs(saved_filepath) if not os.path.exists(model_filepath): os.makedirs(model_filepath) model_path = ('../models/' + args.proctitle + '.pth') train_log = open(saved_filepath + '/train_log.txt', "w")
elif m == 'large': model = pblm.cifar_model_large().cuda() # elif m == 'resNet': # model = pblm.cifar_model_resnet().cuda() else: raise ValueError('model argument not recognized for imagenet') return model if __name__ == "__main__": args = pblm.argparser(prefix='imagenet', gan_type='biggan', starting_epsilon=0.01, opt='sgd', lr=0.05, batch_size_test=8, proj=50, norm_train='l2_normal', norm_test='l2', epsilon=0.1412, seed=0) setproctitle.setproctitle('python') kwargs = pblm.args2kwargs(args) print("saving file to {}".format(args.proctitle)) saved_filepath = ('./saved_log/' + args.proctitle) model_filepath = os.path.dirname('./models/' + args.proctitle) if not os.path.exists(saved_filepath): os.makedirs(saved_filepath) if not os.path.exists(model_filepath):
def select_model(m): if m == 'large': # raise ValueError model = pblm.cifar_model_large().cuda() else: model = pblm.cifar_model().cuda() return model if __name__ == "__main__": args = pblm.argparser(prefix='cifar', epsilon=0.03486, starting_epsilon=0.001, l1_proj=50, l1_train='median', opt='sgd', lr=0.05, ratio=0) setproctitle.setproctitle('python') kwargs = pblm.args2kwargs(args) print("saving file to {}".format(args.proctitle)) if args.method == 'overall_robust': print("threshold for classification error: {:.1%}".format(args.thres)) elif args.method != 'baseline': raise ValueError("Unknown training method.") saved_filepath = ('../saved_log/' + args.proctitle) model_filepath = os.path.dirname('../models/' + args.proctitle) if not os.path.exists(saved_filepath):
print("Using deep model with model_factor={}".format( args.model_factor)) _, test_loader = pblm.mnist_loaders(64 // (2**args.model_factor)) model = pblm.mnist_model_deep(args.model_factor).cuda() elif m == '500': model = pblm.mnist_500().cuda() else: print('``` Use default MNIST model.') model = pblm.mnist_model().cuda() return model if __name__ == "__main__": args = pblm.argparser(opt='adam', verbose=200, starting_epsilon=0.01, batch_size=30, epsilon=0.3) print("saving file to {}".format(args.prefix)) setproctitle.setproctitle(args.prefix) train_log = open(args.prefix + "_train.log", "w") test_log = open(args.prefix + "_test.log", "w") train_loader, _ = pblm.mnist_loaders(args.batch_size) _, test_loader = pblm.mnist_loaders(args.test_batch_size) torch.manual_seed(args.seed) torch.cuda.manual_seed(args.seed) for X, y in train_loader: break
import waitGPU waitGPU.wait(utilization=40, available_memory=8000, interval=20) import problems as pblm from trainer import * import setproctitle if __name__ == "__main__": args = pblm.argparser(prefix='mnist', gan_type='ACGAN', opt='adam', batch_size_test=10, proj=50, norm_train='l2_normal', norm_test='l2', epsilon=1.58, seed=0) kwargs = pblm.args2kwargs(args) setproctitle.setproctitle('python') print("saving file to {}".format(args.proctitle)) saved_filepath = ('./saved_log/' + args.proctitle) model_filepath = os.path.dirname('./models/' + args.proctitle) if not os.path.exists(saved_filepath): os.makedirs(saved_filepath) if not os.path.exists(model_filepath): os.makedirs(model_filepath) model_path = ('./models/' + args.proctitle) train_res = open(saved_filepath + '/train_res.txt', "w")