style corrections

Signed-off-by: Muhammad Zaid Hameed <Zaid.Hameed@ibm.com>

Author: Muhammad Zaid Hameed

art/defences/trainer/adversarial_trainer_awp_pytorch.py

@@ -375,7 +375,7 @@ def _batch_process(self, x_batch: np.ndarray, y_batch: np.ndarray) -> Tuple[floa
         return train_loss, train_acc, train_n
 
     def _weight_perturbation(
-        self, x_batch: np.ndarray, x_batch_pert: np.ndarray, y_batch: np.ndarray
+        self, x_batch: "torch.Tensor", x_batch_pert: "torch.Tensor", y_batch: "torch.Tensor"
     ) -> Dict[str, "torch.Tensor"]:
         """
         Calculate wight perturbation for a batch of data.
@@ -416,15 +416,15 @@ def _weight_perturbation(
                 "Incorrect mode provided for base adversarial training. 'mode' must be among 'PGD' and 'TRADES'."
             )
 
-        self._proxy_classifier._optimizer.zero_grad()  # pylint: disable=W0212
+        self._proxy_classifier._optimizer.zero_grad()  # type: ignore  # pylint: disable=W0212
         loss.backward()
-        self._proxy_classifier._optimizer.step()  # pylint: disable=W0212
+        self._proxy_classifier._optimizer.step()  # type: ignore  # pylint: disable=W0212
 
         params_dict_proxy, _ = self._calculate_model_params(self._proxy_classifier)
 
         for name in list_keys:
             perturbation = params_dict_proxy[name]["param"] - params_dict[name]["param"]
-            perturbation = perturbation.reshape(params_dict[name]["size"])
+            perturbation = torch.reshape(perturbation, params_dict[name]["size"])  # type: ignore
             scale = params_dict[name]["norm"] / (perturbation.norm() + EPS)
             w_perturb[name] = scale * perturbation
 
@@ -444,7 +444,7 @@ def _calculate_model_params(
 
         import torch
 
-        params_dict = OrderedDict()
+        params_dict = OrderedDict()  # type: ignore
         list_params = []
         for name, param in p_classifier._model.state_dict().items():  # pylint: disable=W0212
             if len(param.size()) <= 1:

examples/adversarial_training_awp.py

@@ -18,15 +18,16 @@
 from art.attacks.evasion import ProjectedGradientDescent
 
 """
-For this example we choose the PreActResNet18 model as used in the paper 
+For this example we choose the PreActResNet18 model as used in the paper
 (https://proceedings.neurips.cc/paper/2020/file/1ef91c212e30e14bf125e9374262401f-Paper.pdf)
 The code for the model architecture has been adopted from
 https://github.com/csdongxian/AWP/blob/main/AT_AWP/preactresnet.py
 """
 
 
 class PreActBlock(nn.Module):
-    '''Pre-activation version of the BasicBlock.'''
+    """Pre-activation version of the BasicBlock."""
+
     expansion = 1
 
     def __init__(self, in_planes, planes, stride=1):
@@ -36,22 +37,23 @@ def __init__(self, in_planes, planes, stride=1):
         self.bn2 = nn.BatchNorm2d(planes)
         self.conv2 = nn.Conv2d(planes, planes, kernel_size=3, stride=1, padding=1, bias=False)
 
-        if stride != 1 or in_planes != self.expansion*planes:
+        if stride != 1 or in_planes != self.expansion * planes:
             self.shortcut = nn.Sequential(
-                nn.Conv2d(in_planes, self.expansion*planes, kernel_size=1, stride=stride, bias=False)
+                nn.Conv2d(in_planes, self.expansion * planes, kernel_size=1, stride=stride, bias=False)
             )
 
     def forward(self, x):
         out = F.relu(self.bn1(x))
-        shortcut = self.shortcut(out) if hasattr(self, 'shortcut') else x
+        shortcut = self.shortcut(out) if hasattr(self, "shortcut") else x
         out = self.conv1(out)
         out = self.conv2(F.relu(self.bn2(out)))
         out += shortcut
         return out
 
 
 class PreActBottleneck(nn.Module):
-    '''Pre-activation version of the original Bottleneck module.'''
+    """Pre-activation version of the original Bottleneck module."""
+
     expansion = 4
 
     def __init__(self, in_planes, planes, stride=1):
@@ -61,16 +63,16 @@ def __init__(self, in_planes, planes, stride=1):
         self.bn2 = nn.BatchNorm2d(planes)
         self.conv2 = nn.Conv2d(planes, planes, kernel_size=3, stride=stride, padding=1, bias=False)
         self.bn3 = nn.BatchNorm2d(planes)
-        self.conv3 = nn.Conv2d(planes, self.expansion*planes, kernel_size=1, bias=False)
+        self.conv3 = nn.Conv2d(planes, self.expansion * planes, kernel_size=1, bias=False)
 
-        if stride != 1 or in_planes != self.expansion*planes:
+        if stride != 1 or in_planes != self.expansion * planes:
             self.shortcut = nn.Sequential(
-                nn.Conv2d(in_planes, self.expansion*planes, kernel_size=1, stride=stride, bias=False)
+                nn.Conv2d(in_planes, self.expansion * planes, kernel_size=1, stride=stride, bias=False)
             )
 
     def forward(self, x):
         out = F.relu(self.bn1(x))
-        shortcut = self.shortcut(out) if hasattr(self, 'shortcut') else x
+        shortcut = self.shortcut(out) if hasattr(self, "shortcut") else x
         out = self.conv1(out)
         out = self.conv2(F.relu(self.bn2(out)))
         out = self.conv3(F.relu(self.bn3(out)))
@@ -89,10 +91,10 @@ def __init__(self, block, num_blocks, num_classes=10):
         self.layer3 = self._make_layer(block, 256, num_blocks[2], stride=2)
         self.layer4 = self._make_layer(block, 512, num_blocks[3], stride=2)
         self.bn = nn.BatchNorm2d(512 * block.expansion)
-        self.linear = nn.Linear(512*block.expansion, num_classes)
+        self.linear = nn.Linear(512 * block.expansion, num_classes)
 
     def _make_layer(self, block, planes, num_blocks, stride):
-        strides = [stride] + [1]*(num_blocks-1)
+        strides = [stride] + [1] * (num_blocks - 1)
         layers = []
         for stride in strides:
             layers.append(block(self.in_planes, planes, stride))
@@ -113,7 +115,7 @@ def forward(self, x):
 
 
 def PreActResNet18(num_classes=10):
-    return PreActResNet(PreActBlock, [2,2,2,2], num_classes=num_classes)
+    return PreActResNet(PreActBlock, [2, 2, 2, 2], num_classes=num_classes)
 
 
 class CIFAR10_dataset(Dataset):
@@ -202,7 +204,9 @@ def __len__(self):
 )
 
 # Step 4: Create the trainer object - AdversarialTrainerAWPPyTorch
-trainer = AdversarialTrainerAWPPyTorch(classifier, proxy_classifier, attack, mode="PGD", gamma=gamma, beta=6.0, warmup=warmup)
+trainer = AdversarialTrainerAWPPyTorch(
+    classifier, proxy_classifier, attack, mode="PGD", gamma=gamma, beta=6.0, warmup=warmup
+)
 
 
 # Build a Keras image augmentation object and wrap it in ART