Change TensorAdd to Add

examples/model_security/model_attacks/cv/faster_rcnn/src/FasterRcnn/resnet50.py

@@ -224,7 +224,7 @@ class ResidualBlockUsing(nn.Cell):
                 self.bn_down_sample = self.bn_down_sample.set_train()
             if not weights_update:
                 self.conv_down_sample.weight.requires_grad = False
-        self.add = P.TensorAdd()
+        self.add = P.Add()
 
     def construct(self, x):
         identity = x

examples/model_security/model_attacks/cv/yolov3_darknet53/src/darknet.py

@@ -64,7 +64,7 @@ class ResidualBlock(nn.Cell):
         out_chls = out_channels//2
         self.conv1 = conv_block(in_channels, out_chls, kernel_size=1, stride=1)
         self.conv2 = conv_block(out_chls, out_channels, kernel_size=3, stride=1)
-        self.add = P.TensorAdd()
+        self.add = P.Add()
 
     def construct(self, x):
         identity = x

examples/model_security/model_defenses/mnist_evaluation.py

@@ -22,7 +22,7 @@ from mindspore import context
 from mindspore import nn
 from mindspore.nn import Cell
 from mindspore.nn import SoftmaxCrossEntropyWithLogits
-from mindspore.ops.operations import TensorAdd
+from mindspore.ops.operations import Add
 from mindspore.train.serialization import load_checkpoint, load_param_into_net
 from scipy.special import softmax
 
@@ -58,7 +58,7 @@ class EncoderNet(Cell):
     def __init__(self, encode_dim):
         super(EncoderNet, self).__init__()
         self._encode_dim = encode_dim
-        self.add = TensorAdd()
+        self.add = Add()
 
     def construct(self, inputs):
         """

examples/model_security/model_defenses/mnist_similarity_detector.py

@@ -18,7 +18,7 @@ from mindspore import Model
 from mindspore import Tensor
 from mindspore import context
 from mindspore.nn import Cell
-from mindspore.ops.operations import TensorAdd
+from mindspore.ops.operations import Add
 from mindspore.train.serialization import load_checkpoint, load_param_into_net
 
 from mindarmour import BlackModel
@@ -72,7 +72,7 @@ class EncoderNet(Cell):
     def __init__(self, encode_dim):
         super(EncoderNet, self).__init__()
         self._encode_dim = encode_dim
-        self.add = TensorAdd()
+        self.add = Add()
 
     def construct(self, inputs):
         """

mindarmour/privacy/diff_privacy/mechanisms/mechanisms.py

@@ -142,7 +142,7 @@ class NoiseMechanismsFactory:
             >>>                 loss_fn=loss,
             >>>                 optimizer=net_opt,
             >>>                 metrics=None)
-            >>> ms_ds = ds.GeneratorDataset(dataset_generator(batch_size, batches),
+            >>> ms_ds = ds.GeneratorDataset(dataset_generator,
             >>>                            ['data', 'label'])
             >>> model.train(epochs, ms_ds, dataset_sink_mode=False)
         """
@@ -325,7 +325,7 @@ class _MechanismsParamsUpdater(Cell):
         self._init_noise_multiplier = init_noise_multiplier
 
         self._div = P.Div()
-        self._add = P.TensorAdd()
+        self._add = P.Add()
         self._assign = P.Assign()
         self._sub = P.Sub()
         self._one = Tensor(1, mstype.float32)
@@ -414,7 +414,7 @@ class AdaClippingWithGaussianRandom(Cell):
                                                  mstype.float32)
 
         self._zero = Tensor(0, mstype.float32)
-        self._add = P.TensorAdd()
+        self._add = P.Add()
         self._sub = P.Sub()
         self._mul = P.Mul()
         self._exp = P.Exp()

mindarmour/privacy/diff_privacy/optimizer/optimizer.py

@@ -42,7 +42,7 @@ def tensor_grad_scale(scale, grad):
 class _TupleAdd(nn.Cell):
     def __init__(self):
         super(_TupleAdd, self).__init__()
-        self.add = P.TensorAdd()
+        self.add = P.Add()
         self.hyper_map = C.HyperMap()
 
     def construct(self, input1, input2):

mindarmour/privacy/diff_privacy/train/model.py

@@ -112,7 +112,7 @@ class DPModel(Model):
         >>>                 loss_fn=loss,
         >>>                 optimizer=net_opt,
         >>>                 metrics=None)
-        >>> ms_ds = ds.GeneratorDataset(dataset_generator(batch_size, batches),
+        >>> ms_ds = ds.GeneratorDataset(dataset_generator,
         >>>                             ['data', 'label'])
         >>> model.train(epochs, ms_ds, dataset_sink_mode=False)
     """
@@ -133,7 +133,7 @@ class DPModel(Model):
         opt_name = opt.__class__.__name__
         # Check whether noise_mech and DPOptimizer are both None or not None, if so, raise ValueError.
         # And check whether noise_mech or DPOtimizer's mech method is adaptive while clip_mech is not None,
-        # if so, rasie ValuerError too.
+        # if so, raise ValuerError too.
         if noise_mech is not None and "DPOptimizer" in opt_name:
             msg = 'DPOptimizer is not supported while noise_mech is not None'
             LOGGER.error(TAG, msg)
@@ -323,7 +323,7 @@ class _ClipGradients(nn.Cell):
 class _TupleAdd(nn.Cell):
     def __init__(self):
         super(_TupleAdd, self).__init__()
-        self.add = P.TensorAdd()
+        self.add = P.Add()
         self.hyper_map = C.HyperMap()
 
     def construct(self, input1, input2):
@@ -422,7 +422,7 @@ class _TrainOneStepWithLossScaleCell(Cell):
         self._clip_by_global_norm = _ClipGradients()
         self._noise_mech = noise_mech
         self._clip_mech = clip_mech
-        self._add = P.TensorAdd()
+        self._add = P.Add()
         self._norm = nn.Norm()
         self._tuple_add = _TupleAdd()
         self._hyper_map = C.HyperMap()
@@ -508,7 +508,7 @@ class _TrainOneStepWithLossScaleCell(Cell):
                                                     GRADIENT_CLIP_TYPE,
                                                     self._norm_bound)
             grads = self._tuple_add(grads, record_grad)
-            total_loss = P.TensorAdd()(total_loss, loss)
+            total_loss = P.Add()(total_loss, loss)
         loss = P.Div()(total_loss, self._micro_float)
         beta = self._div(beta, self._micro_batches)
 
@@ -626,7 +626,7 @@ class _TrainOneStepCell(Cell):
         self._noise_mech = noise_mech
         self._clip_mech = clip_mech
         self._tuple_add = _TupleAdd()
-        self._add = P.TensorAdd()
+        self._add = P.Add()
         self._norm = nn.Norm()
         self._hyper_map = C.HyperMap()
         self._zero = Tensor(0, mstype.float32)
@@ -698,7 +698,7 @@ class _TrainOneStepCell(Cell):
                                                     GRADIENT_CLIP_TYPE,
                                                     self._norm_bound)
             grads = self._tuple_add(grads, record_grad)
-            total_loss = P.TensorAdd()(total_loss, loss)
+            total_loss = P.Add()(total_loss, loss)
         loss = self._div(total_loss, self._micro_float)
 
         if self._noise_mech is not None:

mindarmour/privacy/sup_privacy/train/model.py

@@ -206,7 +206,7 @@ class _TupleAdd(nn.Cell):
     """
     def __init__(self):
         super(_TupleAdd, self).__init__()
-        self.add = P.TensorAdd()
+        self.add = P.Add()
         self.hyper_map = C.HyperMap()
 
     def construct(self, input1, input2):

tests/st/resnet50/resnet_cifar10.py

@@ -121,7 +121,7 @@ class ResidualBlock(nn.Cell):
         self.bn3 = bn_with_initialize_last(out_channels)
 
         self.relu = P.ReLU()
-        self.add = P.TensorAdd()
+        self.add = P.Add()
 
     def construct(self, x):
         identity = x
@@ -168,7 +168,7 @@ class ResidualBlockWithDown(nn.Cell):
 
         self.conv_down_sample = conv1x1(in_channels, out_channels, stride=stride, padding=0)
         self.bn_down_sample = bn_with_initialize(out_channels)
-        self.add = P.TensorAdd()
+        self.add = P.Add()
 
     def construct(self, x):
         identity = x

tests/ut/python/adv_robustness/attacks/black/test_hsja.py

@@ -23,7 +23,7 @@ from mindarmour import BlackModel
 from mindarmour.adv_robustness.attacks import HopSkipJumpAttack
 from mindarmour.utils.logger import LogUtil
 
-from ut.python.utils.mock_net import Net
+from tests.ut.python.utils.mock_net import Net
 
 context.set_context(mode=context.GRAPH_MODE)
 context.set_context(device_target="Ascend")

tests/ut/python/adv_robustness/attacks/black/test_nes.py

@@ -23,7 +23,7 @@ from mindarmour import BlackModel
 from mindarmour.adv_robustness.attacks import NES
 from mindarmour.utils.logger import LogUtil
 
-from ut.python.utils.mock_net import Net
+from tests.ut.python.utils.mock_net import Net
 
 context.set_context(mode=context.GRAPH_MODE)
 context.set_context(device_target="Ascend")

tests/ut/python/adv_robustness/attacks/black/test_pointwise_attack.py

@@ -26,7 +26,7 @@ from mindarmour import BlackModel
 from mindarmour.adv_robustness.attacks import PointWiseAttack
 from mindarmour.utils.logger import LogUtil
 
-from ut.python.utils.mock_net import Net
+from tests.ut.python.utils.mock_net import Net
 
 context.set_context(mode=context.GRAPH_MODE, device_target="Ascend")
 

tests/ut/python/adv_robustness/attacks/black/test_pso_attack.py

@@ -144,7 +144,7 @@ def test_pso_attack_targeted():
 
 
 @pytest.mark.level0
-@pytest.mark.platform_x86_gpu_inference
+@pytest.mark.platform_x86_gpu_training
 @pytest.mark.env_card
 @pytest.mark.component_mindarmour
 def test_pso_attack_gpu():

tests/ut/python/adv_robustness/attacks/test_gradient_method.py

@@ -133,7 +133,7 @@ def test_fast_gradient_method():
 
 
 @pytest.mark.level0
-@pytest.mark.platform_x86_gpu_inference
+@pytest.mark.platform_x86_gpu_training
 @pytest.mark.env_card
 @pytest.mark.component_mindarmour
 def test_fast_gradient_method_gpu():

tests/ut/python/adv_robustness/attacks/test_jsma.py

@@ -108,7 +108,7 @@ def test_jsma_attack_2():
 
 
 @pytest.mark.level0
-@pytest.mark.platform_x86_gpu_inference
+@pytest.mark.platform_x86_gpu_training
 @pytest.mark.env_card
 @pytest.mark.component_mindarmour
 def test_jsma_attack_gpu():

tests/ut/python/adv_robustness/attacks/test_lbfgs.py

@@ -24,7 +24,7 @@ from mindspore.train.serialization import load_checkpoint, load_param_into_net
 from mindarmour.adv_robustness.attacks import LBFGS
 from mindarmour.utils.logger import LogUtil
 
-from ut.python.utils.mock_net import Net
+from tests.ut.python.utils.mock_net import Net
 
 context.set_context(mode=context.GRAPH_MODE, device_target="Ascend")
 

tests/ut/python/adv_robustness/defenses/test_ad.py

@@ -26,7 +26,7 @@ from mindspore.nn.optim.momentum import Momentum
 from mindarmour.adv_robustness.defenses import AdversarialDefense
 from mindarmour.utils.logger import LogUtil
 
-from ut.python.utils.mock_net import Net
+from tests.ut.python.utils.mock_net import Net
 
 LOGGER = LogUtil.get_instance()
 TAG = 'Ad_Test'

tests/ut/python/adv_robustness/defenses/test_ead.py

@@ -28,7 +28,7 @@ from mindarmour.adv_robustness.attacks import \
 from mindarmour.adv_robustness.defenses import EnsembleAdversarialDefense
 from mindarmour.utils.logger import LogUtil
 
-from ut.python.utils.mock_net import Net
+from tests.ut.python.utils.mock_net import Net
 
 LOGGER = LogUtil.get_instance()
 TAG = 'Ead_Test'

tests/ut/python/adv_robustness/defenses/test_nad.py

@@ -25,7 +25,7 @@ from mindspore.nn.optim.momentum import Momentum
 from mindarmour.adv_robustness.defenses import NaturalAdversarialDefense
 from mindarmour.utils.logger import LogUtil
 
-from ut.python.utils.mock_net import Net
+from tests.ut.python.utils.mock_net import Net
 
 LOGGER = LogUtil.get_instance()
 TAG = 'Nad_Test'

tests/ut/python/adv_robustness/defenses/test_pad.py

@@ -25,7 +25,7 @@ from mindspore.nn.optim.momentum import Momentum
 from mindarmour.adv_robustness.defenses import ProjectedAdversarialDefense
 from mindarmour.utils.logger import LogUtil
 
-from ut.python.utils.mock_net import Net
+from tests.ut.python.utils.mock_net import Net
 
 LOGGER = LogUtil.get_instance()
 TAG = 'Pad_Test'

tests/ut/python/adv_robustness/detectors/black/test_similarity_detector.py

@@ -20,7 +20,7 @@ import pytest
 from mindspore.nn import Cell
 from mindspore import Model
 from mindspore import context
-from mindspore.ops.operations import TensorAdd
+from mindspore.ops.operations import Add
 
 from mindarmour.adv_robustness.detectors import SimilarityDetector
 
@@ -35,7 +35,7 @@ class EncoderNet(Cell):
     def __init__(self, encode_dim):
         super(EncoderNet, self).__init__()
         self._encode_dim = encode_dim
-        self.add = TensorAdd()
+        self.add = Add()
 
     def construct(self, inputs):
         """

tests/ut/python/adv_robustness/detectors/test_ensemble_detector.py

@@ -18,7 +18,7 @@ import numpy as np
 import pytest
 
 from mindspore.nn import Cell
-from mindspore.ops.operations import TensorAdd
+from mindspore.ops.operations import Add
 from mindspore.train.model import Model
 from mindspore import context
 
@@ -35,7 +35,7 @@ class Net(Cell):
     """
     def __init__(self):
         super(Net, self).__init__()
-        self.add = TensorAdd()
+        self.add = Add()
 
     def construct(self, inputs):
         """
@@ -53,7 +53,7 @@ class AutoNet(Cell):
     """
     def __init__(self):
         super(AutoNet, self).__init__()
-        self.add = TensorAdd()
+        self.add = Add()
 
     def construct(self, inputs):
         """

tests/ut/python/adv_robustness/detectors/test_mag_net.py

@@ -19,7 +19,7 @@ import pytest
 
 import mindspore.ops.operations as P
 from mindspore.nn import Cell
-from mindspore.ops.operations import TensorAdd
+from mindspore.ops.operations import Add
 from mindspore import Model
 from mindspore import context
 
@@ -36,7 +36,7 @@ class Net(Cell):
 
     def __init__(self):
         super(Net, self).__init__()
-        self.add = TensorAdd()
+        self.add = Add()
 
     def construct(self, inputs):
         """

tests/ut/python/adv_robustness/detectors/test_region_based_detector.py

@@ -20,7 +20,7 @@ import pytest
 from mindspore.nn import Cell
 from mindspore import Model
 from mindspore import context
-from mindspore.ops.operations import TensorAdd
+from mindspore.ops.operations import Add
 
 from mindarmour.adv_robustness.detectors import RegionBasedDetector
 
@@ -34,7 +34,7 @@ class Net(Cell):
     """
     def __init__(self):
         super(Net, self).__init__()
-        self.add = TensorAdd()
+        self.add = Add()
 
     def construct(self, inputs):
         """

tests/ut/python/privacy/diff_privacy/test_model_train.py

@@ -26,11 +26,13 @@ from mindarmour.privacy.diff_privacy import NoiseMechanismsFactory
 from mindarmour.privacy.diff_privacy import ClipMechanismsFactory
 from mindarmour.privacy.diff_privacy import DPOptimizerClassFactory
 
-from ut.python.utils.mock_net import Net
+from tests.ut.python.utils.mock_net import Net
 
 
-def dataset_generator(batch_size, batches):
+def dataset_generator():
     """mock training data."""
+    batch_size = 32
+    batches = 128
     data = np.random.random((batches*batch_size, 1, 32, 32)).astype(
         np.float32)
     label = np.random.randint(0, 10, batches*batch_size).astype(np.int32)
@@ -49,8 +51,6 @@ def test_dp_model_with_pynative_mode():
     norm_bound = 1.0
     initial_noise_multiplier = 0.01
     network = Net()
-    batch_size = 32
-    batches = 128
     epochs = 1
     micro_batches = 2
     loss = nn.SoftmaxCrossEntropyWithLogits(sparse=True)
@@ -73,7 +73,7 @@ def test_dp_model_with_pynative_mode():
                     loss_fn=loss,
                     optimizer=net_opt,
                     metrics=None)
-    ms_ds = ds.GeneratorDataset(dataset_generator(batch_size, batches),
+    ms_ds = ds.GeneratorDataset(dataset_generator,
                                 ['data', 'label'])
     model.train(epochs, ms_ds, dataset_sink_mode=False)
 
@@ -88,8 +88,6 @@ def test_dp_model_with_graph_mode():
     norm_bound = 1.0
     initial_noise_multiplier = 0.01
     network = Net()
-    batch_size = 32
-    batches = 128
     epochs = 1
     loss = nn.SoftmaxCrossEntropyWithLogits(sparse=True)
     noise_mech = NoiseMechanismsFactory().create('Gaussian',
@@ -110,7 +108,7 @@ def test_dp_model_with_graph_mode():
                     loss_fn=loss,
                     optimizer=net_opt,
                     metrics=None)
-    ms_ds = ds.GeneratorDataset(dataset_generator(batch_size, batches),
+    ms_ds = ds.GeneratorDataset(dataset_generator,
                                 ['data', 'label'])
     model.train(epochs, ms_ds, dataset_sink_mode=False)
 
@@ -125,8 +123,6 @@ def test_dp_model_with_graph_mode_ada_gaussian():
     norm_bound = 1.0
     initial_noise_multiplier = 0.01
     network = Net()
-    batch_size = 32
-    batches = 128
     epochs = 1
     alpha = 0.8
     loss = nn.SoftmaxCrossEntropyWithLogits(sparse=True)
@@ -146,6 +142,6 @@ def test_dp_model_with_graph_mode_ada_gaussian():
                     loss_fn=loss,
                     optimizer=net_opt,
                     metrics=None)
-    ms_ds = ds.GeneratorDataset(dataset_generator(batch_size, batches),
+    ms_ds = ds.GeneratorDataset(dataset_generator,
                                 ['data', 'label'])
     model.train(epochs, ms_ds, dataset_sink_mode=False)

tests/ut/python/privacy/diff_privacy/test_monitor.py

@@ -25,13 +25,16 @@ import mindspore.context as context
 from mindarmour.privacy.diff_privacy import PrivacyMonitorFactory
 from mindarmour.utils.logger import LogUtil
 
-from ut.python.utils.mock_net import Net
+from tests.ut.python.utils.mock_net import Net
 
 LOGGER = LogUtil.get_instance()
 TAG = 'DP-Monitor Test'
 
 
-def dataset_generator(batch_size, batches):
+def dataset_generator():
+    batch_size = 16
+    batches = 128
+
     data = np.random.random((batches * batch_size, 1, 32, 32)).astype(
         np.float32)
     label = np.random.randint(0, 10, batches * batch_size).astype(np.int32)
@@ -48,7 +51,6 @@ def dataset_generator(batch_size, batches):
 def test_dp_monitor():
     context.set_context(mode=context.GRAPH_MODE, device_target="Ascend")
     batch_size = 16
-    batches = 128
     epochs = 1
     rdp = PrivacyMonitorFactory.create(policy='rdp', num_samples=60000,
                                        batch_size=batch_size,
@@ -64,19 +66,18 @@ def test_dp_monitor():
     model = Model(network, net_loss, net_opt)
 
     LOGGER.info(TAG, "============== Starting Training ==============")
-    ds1 = ds.GeneratorDataset(dataset_generator(batch_size, batches),
+    ds1 = ds.GeneratorDataset(dataset_generator,
                               ["data", "label"])
     model.train(epochs, ds1, callbacks=[rdp], dataset_sink_mode=False)
 
 
 @pytest.mark.level0
-@pytest.mark.platform_x86_gpu_inference
+@pytest.mark.platform_x86_gpu_training
 @pytest.mark.env_card
 @pytest.mark.component_mindarmour
 def test_dp_monitor_gpu():
     context.set_context(mode=context.GRAPH_MODE, device_target="GPU")
     batch_size = 16
-    batches = 128
     epochs = 1
     rdp = PrivacyMonitorFactory.create(policy='rdp', num_samples=60000,
                                        batch_size=batch_size,
@@ -92,7 +93,7 @@ def test_dp_monitor_gpu():
     model = Model(network, net_loss, net_opt)
 
     LOGGER.info(TAG, "============== Starting Training ==============")
-    ds1 = ds.GeneratorDataset(dataset_generator(batch_size, batches),
+    ds1 = ds.GeneratorDataset(dataset_generator,
                               ["data", "label"])
     model.train(epochs, ds1, callbacks=[rdp], dataset_sink_mode=False)
 
@@ -104,7 +105,6 @@ def test_dp_monitor_gpu():
 def test_dp_monitor_cpu():
     context.set_context(mode=context.GRAPH_MODE, device_target="CPU")
     batch_size = 16
-    batches = 128
     epochs = 1
     rdp = PrivacyMonitorFactory.create(policy='rdp', num_samples=60000,
                                        batch_size=batch_size,
@@ -120,7 +120,7 @@ def test_dp_monitor_cpu():
     model = Model(network, net_loss, net_opt)
 
     LOGGER.info(TAG, "============== Starting Training ==============")
-    ds1 = ds.GeneratorDataset(dataset_generator(batch_size, batches),
+    ds1 = ds.GeneratorDataset(dataset_generator,
                               ["data", "label"])
     model.train(epochs, ds1, callbacks=[rdp], dataset_sink_mode=False)
 
@@ -133,7 +133,6 @@ def test_dp_monitor_cpu():
 def test_dp_monitor_zcdp():
     context.set_context(mode=context.GRAPH_MODE, device_target="Ascend")
     batch_size = 16
-    batches = 128
     epochs = 1
     zcdp = PrivacyMonitorFactory.create(policy='zcdp', num_samples=60000,
                                         batch_size=batch_size,
@@ -149,19 +148,18 @@ def test_dp_monitor_zcdp():
     model = Model(network, net_loss, net_opt)
 
     LOGGER.info(TAG, "============== Starting Training ==============")
-    ds1 = ds.GeneratorDataset(dataset_generator(batch_size, batches),
+    ds1 = ds.GeneratorDataset(dataset_generator,
                               ["data", "label"])
     model.train(epochs, ds1, callbacks=[zcdp], dataset_sink_mode=False)
 
 
 @pytest.mark.level0
-@pytest.mark.platform_x86_gpu_inference
+@pytest.mark.platform_x86_gpu_training
 @pytest.mark.env_card
 @pytest.mark.component_mindarmour
 def test_dp_monitor_zcdp_gpu():
     context.set_context(mode=context.GRAPH_MODE, device_target="GPU")
     batch_size = 16
-    batches = 128
     epochs = 1
     zcdp = PrivacyMonitorFactory.create(policy='zcdp', num_samples=60000,
                                         batch_size=batch_size,
@@ -177,7 +175,7 @@ def test_dp_monitor_zcdp_gpu():
     model = Model(network, net_loss, net_opt)
 
     LOGGER.info(TAG, "============== Starting Training ==============")
-    ds1 = ds.GeneratorDataset(dataset_generator(batch_size, batches),
+    ds1 = ds.GeneratorDataset(dataset_generator,
                               ["data", "label"])
     model.train(epochs, ds1, callbacks=[zcdp], dataset_sink_mode=False)
 
@@ -189,7 +187,6 @@ def test_dp_monitor_zcdp_gpu():
 def test_dp_monitor_zcdp_cpu():
     context.set_context(mode=context.GRAPH_MODE, device_target="CPU")
     batch_size = 16
-    batches = 128
     epochs = 1
     zcdp = PrivacyMonitorFactory.create(policy='zcdp', num_samples=60000,
                                         batch_size=batch_size,
@@ -205,6 +202,6 @@ def test_dp_monitor_zcdp_cpu():
     model = Model(network, net_loss, net_opt)
 
     LOGGER.info(TAG, "============== Starting Training ==============")
-    ds1 = ds.GeneratorDataset(dataset_generator(batch_size, batches),
+    ds1 = ds.GeneratorDataset(dataset_generator,
                               ["data", "label"])
     model.train(epochs, ds1, callbacks=[zcdp], dataset_sink_mode=False)

tests/ut/python/privacy/diff_privacy/test_optimizer.py

@@ -19,7 +19,7 @@ from mindspore.train.model import Model
 
 from mindarmour.privacy.diff_privacy import DPOptimizerClassFactory
 
-from ut.python.utils.mock_net import Net
+from tests.ut.python.utils.mock_net import Net
 
 
 @pytest.mark.level0
@@ -42,7 +42,7 @@ def test_optimizer():
 
 
 @pytest.mark.level0
-@pytest.mark.platform_x86_gpu_inference
+@pytest.mark.platform_x86_gpu_training
 @pytest.mark.env_card
 @pytest.mark.component_mindarmour
 def test_optimizer_gpu():

tests/ut/python/privacy/evaluation/test_inversion_attack.py

@@ -22,7 +22,7 @@ import mindspore.context as context
 
 from mindarmour.privacy.evaluation.inversion_attack import ImageInversionAttack
 
-from ut.python.utils.mock_net import Net
+from tests.ut.python.utils.mock_net import Net
 
 
 context.set_context(mode=context.GRAPH_MODE)

tests/ut/python/privacy/evaluation/test_membership_inference.py

@@ -25,14 +25,16 @@ import mindspore.context as context
 
 from mindarmour.privacy.evaluation import MembershipInference
 
-from ut.python.utils.mock_net import Net
+from tests.ut.python.utils.mock_net import Net
 
 
 context.set_context(mode=context.GRAPH_MODE)
 
 
-def dataset_generator(batch_size, batches):
+def dataset_generator():
     """mock training data."""
+    batch_size = 16
+    batches = 1
     data = np.random.randn(batches*batch_size, 1, 32, 32).astype(
         np.float32)
     label = np.random.randint(0, 10, batches*batch_size).astype(np.int32)
@@ -74,11 +76,9 @@ def test_membership_inference_object_train():
             "n_neighbors": [3, 5, 7],
         }
     }]
-    batch_size = 16
-    batches = 1
-    ds_train = ds.GeneratorDataset(dataset_generator(batch_size, batches),
+    ds_train = ds.GeneratorDataset(dataset_generator,
                                    ["image", "label"])
-    ds_test = ds.GeneratorDataset(dataset_generator(batch_size, batches),
+    ds_test = ds.GeneratorDataset(dataset_generator,
                                   ["image", "label"])
     inference_model.train(ds_train, ds_test, config)
 
@@ -96,11 +96,9 @@ def test_membership_inference_eval():
     inference_model = MembershipInference(model, -1)
     assert isinstance(inference_model, MembershipInference)
 
-    batch_size = 16
-    batches = 1
-    eval_train = ds.GeneratorDataset(dataset_generator(batch_size, batches),
+    eval_train = ds.GeneratorDataset(dataset_generator,
                                      ["image", "label"])
-    eval_test = ds.GeneratorDataset(dataset_generator(batch_size, batches),
+    eval_test = ds.GeneratorDataset(dataset_generator,
                                     ["image", "label"])
 
     metrics = ["precision", "accuracy", "recall"]

tests/ut/python/privacy/sup_privacy/test_model_train.py

@@ -25,15 +25,18 @@ from mindspore.train.callback import LossMonitor
 from mindspore.nn.metrics import Accuracy
 import mindspore.dataset as ds
 
-from ut.python.utils.mock_net import Net as LeNet5
-
 from mindarmour.privacy.sup_privacy import SuppressModel
 from mindarmour.privacy.sup_privacy import SuppressMasker
 from mindarmour.privacy.sup_privacy import SuppressPrivacyFactory
 from mindarmour.privacy.sup_privacy import MaskLayerDes
 
-def dataset_generator(batch_size, batches):
+from tests.ut.python.utils.mock_net import Net as LeNet5
+
+
+def dataset_generator():
     """mock training data."""
+    batches = 10
+    batch_size = 32
     data = np.random.random((batches*batch_size, 1, 32, 32)).astype(
         np.float32)
     label = np.random.randint(0, 10, batches*batch_size).astype(np.int32)
@@ -51,7 +54,6 @@ def test_suppress_model_with_pynative_mode():
     networks_l5 = LeNet5()
     epochs = 5
     batch_num = 10
-    batch_size = 32
     mask_times = 10
     lr = 0.01
     masklayers_lenet5 = []
@@ -79,7 +81,7 @@ def test_suppress_model_with_pynative_mode():
     ckpoint_cb = ModelCheckpoint(prefix="checkpoint_lenet",
                                  directory="./trained_ckpt_file/",
                                  config=config_ck)
-    ds_train = ds.GeneratorDataset(dataset_generator(batch_size, batch_num), ['data', 'label'])
+    ds_train = ds.GeneratorDataset(dataset_generator, ['data', 'label'])
 
     model_instance.train(epochs, ds_train, callbacks=[ckpoint_cb, LossMonitor(), suppress_masker],
                          dataset_sink_mode=False)