diff --git a/tutorials/source_zh_cn/advanced_use/differential_privacy.md b/tutorials/source_zh_cn/advanced_use/differential_privacy.md index cea3059d609fb0881faf3806aeb25393531d7bca..e93111295233d07c14b70fc807d486483f048201 100644 --- a/tutorials/source_zh_cn/advanced_use/differential_privacy.md +++ b/tutorials/source_zh_cn/advanced_use/differential_privacy.md @@ -56,6 +56,9 @@ from mindarmour.diff_privacy import PrivacyMonitorFactory from mindarmour.utils.logger import LogUtil from lenet5_net import LeNet5 from lenet5_config import mnist_cfg as cfg + +LOGGER.set_level('INFO') +TAG = 'Lenet5_train' ``` ### 配置环境信息 @@ -72,17 +75,17 @@ from lenet5_config import mnist_cfg as cfg - initial_noise_multiplier：差分隐私参数，高斯噪声的标准差等于initial_noise_multiplier乘以l2_norm_bound。 ```python - parser = argparse.ArgumentParser(description='MindSpore Example') + parser = argparse.ArgumentParser(description='MindSpore MNIST Example') parser.add_argument('--device_target', type=str, default="Ascend", choices=['Ascend', 'GPU', 'CPU'], help='device where the code will be implemented (default: Ascend)') - parser.add_argument('--data_path', type=str, default="./_unzip", + parser.add_argument('--data_path', type=str, default="./MNIST_unzip", help='path where the dataset is saved') parser.add_argument('--dataset_sink_mode', type=bool, default=False, help='dataset_sink_mode is False or True') - parser.add_argument('--micro_batches', type=int, default=None, + parser.add_argument('--micro_batches', type=int, default=32, help='optional, if use differential privacy, need to set micro_batches') - parser.add_argument('--l2_norm_bound', type=float, default=1, + parser.add_argument('--l2_norm_bound', type=float, default=1.0, help='optional, if use differential privacy, need to set l2_norm_bound') - parser.add_argument('--initial_noise_multiplier', type=float, default=0.1, + parser.add_argument('--initial_noise_multiplier', type=float, default=1.5, help='optional, if use differential privacy, need to set initial_noise_multiplier') args = parser.parse_args() ``` @@ -100,7 +103,7 @@ from lenet5_config import mnist_cfg as cfg 加载数据集并处理成MindSpore数据格式。 ```python -def generate__dataset(data_path, batch_size=32, repeat_size=1, +def generate_mnist_dataset(data_path, batch_size=32, repeat_size=1, num_parallel_workers=1, sparse=True): """ create dataset for training or testing @@ -167,7 +170,7 @@ def fc_with_initialize(input_channels, out_channels): def weight_variable(): - return TruncatedNormal(0.02) + return TruncatedNormal(0.05) class LeNet5(nn.Cell): @@ -201,7 +204,7 @@ class LeNet5(nn.Cell): return x ``` -加载`LeNet`网络，定义损失函数、配置checkpoint、用上述定义的数据加载函数`generate__dataset`载入数据。 +加载`LeNet`网络，定义损失函数、配置checkpoint、用上述定义的数据加载函数`generate_mnist_dataset`载入数据。 ```python network = LeNet5() @@ -212,7 +215,7 @@ ckpoint_cb = ModelCheckpoint(prefix="checkpoint_lenet", directory='./trained_ckpt_file/', config=config_ck) -ds_train = generate__dataset(os.path.join(args.data_path, "train"), +ds_train = generate_mnist_dataset(os.path.join(args.data_path, "train"), cfg.batch_size, cfg.epoch_size) ``` @@ -221,12 +224,15 @@ ds_train = generate__dataset(os.path.join(args.data_path, "train"), 1. 配置差分隐私优化器的参数。 + - 判断micro_batches和batch_size参数是否符合要求。 - 实例化差分隐私工厂类。 - 设置差分隐私的噪声机制，目前支持固定标准差的高斯噪声机制：'Gaussian'和自适应调整标准差的自适应高斯噪声机制：'AdaGaussian'。 - 设置优化器类型，目前支持'SGD'和'Momentum'。 - 设置差分隐私预算监测器RDP，用于观测每个step中的差分隐私预算$\epsilon$的变化。 ```python + if args.micro_batches and cfg.batch_size % args.micro_batches != 0: + raise ValueError("Number of micro_batches should divide evenly batch_size") gaussian_mech = DPOptimizerClassFactory(args.micro_batches) gaussian_mech.set_mechanisms('Gaussian', norm_bound=args.l2_norm_bound, @@ -236,9 +242,9 @@ ds_train = generate__dataset(os.path.join(args.data_path, "train"), momentum=cfg.momentum) rdp_monitor = PrivacyMonitorFactory.create('rdp', num_samples=60000, - batch_size=16, - initial_noise_multiplier=5, - target_delta=0.5, + batch_size=cfg.batch_size, + initial_noise_multiplier=args.initial_noise_multiplier* + args.l2_norm_bound, per_print_times=10) ``` @@ -262,13 +268,13 @@ ds_train = generate__dataset(os.path.join(args.data_path, "train"), dataset_sink_mode=args.dataset_sink_mode) LOGGER.info(TAG, "============== Starting Testing ==============") - ckpt_file_name = 'trained_ckpt_file/checkpoint_lenet-10_1875.ckpt' + ckpt_file_name = 'trained_ckpt_file/checkpoint_lenet-10_234.ckpt' param_dict = load_checkpoint(ckpt_file_name) load_param_into_net(network, param_dict) - ds_eval = generate__dataset(os.path.join(args.data_path, 'test'), batch_size=cfg.batch_size) + ds_eval = generate_mnist_dataset(os.path.join(args.data_path, 'test'), batch_size=cfg.batch_size) acc = model.eval(ds_eval, dataset_sink_mode=False) LOGGER.info(TAG, "============== Accuracy: %s ==============", acc) - + ``` 4. 运行命令。 @@ -290,11 +296,9 @@ ds_train = generate__dataset(os.path.join(args.data_path, "train"), ... ============== Starting Testing ============== ... - ============== Accuracy: 0.9635 ============== + ============== Accuracy: 0.9091 ============== ``` - ![dp_res](images/dp_res.png) - ### 引用 [1] C. Dwork and J. Lei. Differential privacy and robust statistics. In STOC, pages 371–380. ACM, 2009. diff --git a/tutorials/source_zh_cn/advanced_use/images/dp_res.png b/tutorials/source_zh_cn/advanced_use/images/dp_res.png deleted file mode 100644 index c661414fcb6dc82206c1d76f813f9cf4e7fdae01..0000000000000000000000000000000000000000 Binary files a/tutorials/source_zh_cn/advanced_use/images/dp_res.png and /dev/null differ