Apply Black formatting

Signed-off-by: Beat Buesser <[email protected]>

Author: Beat Buesser

art/attacks/evasion/deepfool.py

@@ -120,9 +120,8 @@ def generate(self, x: np.ndarray, y: Optional[np.ndarray] = None, **kwargs) -> n
 
         # Compute perturbation with implicit batching
         for batch_id in trange(
-                int(np.ceil(x_adv.shape[0] / float(self.batch_size))),
-                desc="DeepFool",
-                disable=not self.verbose):
+            int(np.ceil(x_adv.shape[0] / float(self.batch_size))), desc="DeepFool", disable=not self.verbose
+        ):
             batch_index_1, batch_index_2 = batch_id * self.batch_size, (batch_id + 1) * self.batch_size
             batch = x_adv[batch_index_1:batch_index_2].copy()
 

art/attacks/evasion/projected_gradient_descent/projected_gradient_descent_pytorch.py

@@ -277,7 +277,7 @@ def _apply_perturbation(self, x: "torch.Tensor", perturbation: "torch.Tensor", e
             clip_min, clip_max = self.estimator.clip_values
             x = torch.max(
                 torch.min(x, torch.tensor(clip_max).to(self.estimator.device)),
-                torch.tensor(clip_min).to(self.estimator.device)
+                torch.tensor(clip_min).to(self.estimator.device),
             )
 
         return x
@@ -328,7 +328,7 @@ def _compute_torch(
                 clip_min, clip_max = self.estimator.clip_values
                 x_adv = torch.max(
                     torch.min(x_adv, torch.tensor(clip_max).to(self.estimator.device)),
-                    torch.tensor(clip_min).to(self.estimator.device)
+                    torch.tensor(clip_min).to(self.estimator.device),
                 )
 
         else:

art/defences/preprocessor/mp3_compression.py

@@ -120,7 +120,7 @@ def wav_to_mp3(x, sample_rate):
             elif x.dtype != np.int16 and normalized:
                 # x is not of type np.int16 and seems to be normalized. Therefore undoing normalization and
                 # casting to np.int16.
-                x = (x * 2**15).astype(np.int16)
+                x = (x * 2 ** 15).astype(np.int16)
 
             tmp_wav, tmp_mp3 = BytesIO(), BytesIO()
             write(tmp_wav, sample_rate, x)
@@ -134,7 +134,7 @@ def wav_to_mp3(x, sample_rate):
 
             if normalized:
                 # x was normalized. Therefore normalizing x_mp3.
-                x_mp3 = x_mp3 * 2**-15
+                x_mp3 = x_mp3 * 2 ** -15
             return x_mp3
 
         if x.ndim != 3:

tests/attacks/test_wasserstein.py

@@ -472,7 +472,7 @@ def set_learning_phase(self):
             kernel_size=3,
             max_iter=1,
             conjugate_sinkhorn_max_iter=10,
-            projected_sinkhorn_max_iter=10
+            projected_sinkhorn_max_iter=10,
         )
 
         x = np.random.normal(size=(1, 3, 33, 32))