More style changes

Signed-off-by: Kevin Eykholt <[email protected]>

Author: keykholt

art/attacks/poisoning/hidden_trigger_backdoor/hidden_trigger_backdoor.py

@@ -103,7 +103,8 @@ def __init__(
                        Otherwise, it is an array of indicies.
         :param source: The class/indicies which will have a trigger added to cause misclassification
                        If an int, it represents a label. Otherwise, it is an array of indicies.
-        :param feature_layer: The name of the feature representation layer.
+        :param feature_layer: The name of the feature representation layer
+        :param backdoor: A PoisoningAttackBackdoor that adds a backdoor trigger to the input.
         :param eps: Maximum perturbation that the attacker can introduce.
         :param learning_rate: The learning rate of clean-label attack optimization.
         :param decay_coeff: The decay coefficient of the learning rate.

art/attacks/poisoning/hidden_trigger_backdoor/hidden_trigger_backdoor_keras.py

@@ -85,6 +85,7 @@ def __init__(
         :param source: The class/indicies which will have a trigger added to cause misclassification
                        If an int, it represents a label. Otherwise, it is an array of indicies.
         :param feature_layer: The name of the feature representation layer.
+        :param backdoor: A PoisoningAttackBackdoor that adds a backdoor trigger to the input.
         :param eps: Maximum perturbation that the attacker can introduce.
         :param learning_rate: The learning rate of clean-label attack optimization.
         :param decay_coeff: The decay coefficient of the learning rate.
@@ -277,12 +278,14 @@ def poison(  # pylint: disable=W0221
                 poison_samples = np.clip(poison_samples, *self.estimator.clip_values)
 
                 if i % self.print_iter == 0:
-                    f"Batch: {batch_id} | i: {i:5d} | \
+                    print(
+                        f"Batch: {batch_id} | i: {i:5d} | \
                         LR: {learning_rate:2.5f} | \
                         Loss Val: {losses.val:5.3f} | Loss Avg: {losses.avg:5.3f}"
+                    )
 
                 if loss < self.stopping_threshold or i == (self.max_iter - 1):
-                    f"Max_Loss: {loss}"
+                    print(f"Max_Loss: {loss}")
                     final_poison[cur_index : cur_index + offset] = poison_samples
                     break
 
@@ -305,7 +308,8 @@ def _get_keras_tensor(self):
                 self.estimator._layer_names  # pylint: disable=W0212
             ):
                 raise ValueError(
-                    f"Layer index {self.feature_layer} is outside of range (0 to {len(self.estimator._layer_names) - 1} included)."  # pylint: disable=W0212
+                    f"Layer index {self.feature_layer} is outside of range [0 to \
+                    {len(self.estimator._layer_names) - 1}])."  # pylint: disable=W0212
                 )
             layer_name = self.estimator._layer_names[self.feature_layer]  # pylint: disable=W0212
         else:

art/attacks/poisoning/hidden_trigger_backdoor/hidden_trigger_backdoor_pytorch.py

@@ -101,6 +101,7 @@ def __init__(
         :param source: The class/indicies which will have a trigger added to cause misclassification
                        If an int, it represents a label. Otherwise, it is an array of indicies.
         :param feature_layer: The name of the feature representation layer.
+        :param backdoor: A PoisoningAttackBackdoor that adds a backdoor trigger to the input.
         :param eps: Maximum perturbation that the attacker can introduce.
         :param learning_rate: The learning rate of clean-label attack optimization.
         :param decay_coeff: The decay coefficient of the learning rate.
@@ -254,12 +255,14 @@ def poison(  # pylint: disable=W0221
                 poison_samples = poison_samples.clamp(*self.estimator.clip_values)
 
                 if i % self.print_iter == 0:
-                    f"Batch: {batch_id} | i: {i:5d} | \
+                    print(
+                        f"Batch: {batch_id} | i: {i:5d} | \
                         LR: {learning_rate:2.5f} | \
                         Loss Val: {losses.val:5.3f} | Loss Avg: {losses.avg:5.3f}"
+                    )
 
                 if loss.item() < self.stopping_threshold or i == (self.max_iter - 1):
-                    f"Max_Loss: {loss.item()}"
+                    print(f"Max_Loss: {loss.item()}")
                     final_poison[cur_index : cur_index + offset] = poison_samples.detach().cpu().numpy()
                     break
 

art/estimators/classification/pytorch.py

@@ -362,7 +362,7 @@ def fit(  # pylint: disable=W0221
         batch_size: int = 128,
         nb_epochs: int = 10,
         training_mode: bool = True,
-        **kwargs
+        **kwargs,
     ) -> None:
         """
         Fit the classifier on the training set `(x, y)`.