!18 Modify the inputs batch_size of defense tests to adapt to current version MindSpore.

Merge pull request !18 from jxlang910/master

example/mnist_demo/lenet5_mnist_fuzzing.py

@@ -41,7 +41,7 @@ def test_lenet_mnist_fuzzing():
     model = Model(net)
 
     # get training data
-    data_list = "./MNIST_datasets/train"
+    data_list = "./MNIST_unzip/train"
     batch_size = 32
     ds = generate_mnist_dataset(data_list, batch_size, sparse=True)
     train_images = []
@@ -55,7 +55,7 @@ def test_lenet_mnist_fuzzing():
 
     # fuzz test with original test data
     # get test data
-    data_list = "./MNIST_datasets/test"
+    data_list = "./MNIST_unzip/test"
     batch_size = 32
     ds = generate_mnist_dataset(data_list, batch_size, sparse=True)
     test_images = []

tests/ut/python/defenses/test_ad.py

@@ -39,7 +39,7 @@ TAG = 'Ad_Test'
 def test_ad():
     """UT for adversarial defense."""
     num_classes = 10
-    batch_size = 16
+    batch_size = 32
 
     sparse = False
     context.set_context(mode=context.GRAPH_MODE)

tests/ut/python/defenses/test_ead.py

@@ -41,7 +41,7 @@ TAG = 'Ead_Test'
 def test_ead():
     """UT for ensemble adversarial defense."""
     num_classes = 10
-    batch_size = 16
+    batch_size = 64
 
     sparse = False
     context.set_context(mode=context.GRAPH_MODE)
@@ -53,7 +53,7 @@ def test_ead():
     if not sparse:
         labels = np.eye(num_classes)[labels].astype(np.float32)
 
-    net = SimpleNet()
+    net = Net()
     loss_fn = nn.SoftmaxCrossEntropyWithLogits(is_grad=False, sparse=sparse)
     optimizer = Momentum(net.trainable_params(), 0.001, 0.9)
 

tests/ut/python/defenses/test_nad.py

@@ -39,7 +39,7 @@ TAG = 'Nad_Test'
 def test_nad():
     """UT for natural adversarial defense."""
     num_classes = 10
-    batch_size = 16
+    batch_size = 32
 
     sparse = False
     context.set_context(mode=context.GRAPH_MODE)

tests/ut/python/defenses/test_pad.py

@@ -39,7 +39,7 @@ TAG = 'Pad_Test'
 def test_pad():
     """UT for projected adversarial defense."""
     num_classes = 10
-    batch_size = 16
+    batch_size = 32
 
     sparse = False
     context.set_context(mode=context.GRAPH_MODE)