_, predicted = torch.max(outputs.data, 1) total += labels.size(0) loss = criterion(outputs, labels) loss_sum += loss.data.cpu().item() * images.size(0) correct += (predicted == labels).sum().item() cnt += int(images.size()[0]) print('Accuracy of the network on the 10000 test images: %f %%' % (100 * correct / total)) print("loss=", loss_sum / float(cnt)) elapsed_time = time.time() - inference_start print("Elapsed time for Prediction", elapsed_time) if __name__ == "__main__": _, _, _, test_to_run = argparser_distributed() print("====== New Tests ======") print("Test To run:", test_to_run) net_state_name, config_name = get_net_config_name(test_to_run) print(f"net_state going to load: {net_state_name}") print(f"store_configs going to load: {config_name}") store_configs = np.load(config_name, allow_pickle="TRUE").item() def input_shift(data): first_layer_name = "conv1" return data_shift(data, store_configs[first_layer_name + "ForwardX"]) transform_test = transforms.Compose([ transforms.ToTensor(), transforms.Normalize((0.4914, 0.4822, 0.4465),
p = Process(target=run_secure_nn_client_with_random_data, args=[secure_nn, correctness_func, master_address, master_port]) p.start() processes.append(p) for p in processes: p.join() if party == Config.server_rank: run_secure_nn_server_with_random_data(secure_nn, correctness_func, master_address, master_port) if party == Config.client_rank: run_secure_nn_client_with_random_data(secure_nn, correctness_func, master_address, master_port) print(f"\nTest for {test_name}: End") if __name__ == "__main__": input_sid, master_addr, master_port, test_to_run = argparser_distributed() sys.stdout = Logger() print("====== New Tests ======") print("Test To run:", test_to_run) num_repeat = 5 for _ in range(num_repeat): if test_to_run in ["small", "all"]: marshal_secure_nn_parties(input_sid, master_addr, master_port, generate_small_nn(), correctness_small_nn) if test_to_run in ["relu", "all"]: marshal_secure_nn_parties(input_sid, master_addr, master_port, generate_relu_only_nn(), correctness_relu_only_nn) if test_to_run in ["maxpool", "all"]: marshal_secure_nn_parties(input_sid, master_addr, master_port, generate_maxpool2x2(), correctness_maxpool2x2) if test_to_run in ["conv2d", "all"]: