diff --git a/src/torchattack/admix.py b/src/torchattack/admix.py
index 42f899c..86432b1 100644
--- a/src/torchattack/admix.py
+++ b/src/torchattack/admix.py
@@ -11,6 +11,21 @@ class Admix(Attack):
 
     From the paper 'Admix: Enhancing the Transferability of Adversarial Attacks',
     https://arxiv.org/abs/2102.00436
+
+    Args:
+        model: The model to attack.
+        normalize: A transform to normalize images.
+        device: Device to use for tensors. Defaults to cuda if available.
+        eps: The maximum perturbation. Defaults to 8/255.
+        steps: Number of steps. Defaults to 10.
+        alpha: Step size, `eps / steps` if None. Defaults to None.
+        decay: Decay factor for the momentum term. Defaults to 1.0.
+        portion: Portion for the mixed image. Defaults to 0.2.
+        size: Number of randomly sampled images. Defaults to 3.
+        num_classes: Number of classes of the dataset used. Defaults to 1001.
+        clip_min: Minimum value for clipping. Defaults to 0.0.
+        clip_max: Maximum value for clipping. Defaults to 1.0.
+        targeted: Targeted attack if True. Defaults to False.
     """
 
     def __init__(
@@ -29,24 +44,6 @@ def __init__(
         clip_max: float = 1.0,
         targeted: bool = False,
     ) -> None:
-        """Initialize the Admix attack.
-
-        Args:
-            model: The model to attack.
-            normalize: A transform to normalize images.
-            device: Device to use for tensors. Defaults to cuda if available.
-            eps: The maximum perturbation. Defaults to 8/255.
-            steps: Number of steps. Defaults to 10.
-            alpha: Step size, `eps / steps` if None. Defaults to None.
-            decay: Decay factor for the momentum term. Defaults to 1.0.
-            portion: Portion for the mixed image. Defaults to 0.2.
-            size: Number of randomly sampled images. Defaults to 3.
-            num_classes: Number of classes of the dataset used. Defaults to 1001.
-            clip_min: Minimum value for clipping. Defaults to 0.0.
-            clip_max: Maximum value for clipping. Defaults to 1.0.
-            targeted: Targeted attack if True. Defaults to False.
-        """
-
         super().__init__(normalize, device)
 
         self.model = model
diff --git a/src/torchattack/deepfool.py b/src/torchattack/deepfool.py
index 99c6861..0c1fab7 100644
--- a/src/torchattack/deepfool.py
+++ b/src/torchattack/deepfool.py
@@ -11,6 +11,16 @@ class DeepFool(Attack):
 
     From the paper 'DeepFool: A Simple and Accurate Method to Fool Deep Neural Networks'
     https://arxiv.org/abs/1511.04599
+
+    Args:
+        model: The model to attack.
+        normalize: A transform to normalize images.
+        device: Device to use for tensors. Defaults to cuda if available.
+        steps: Number of steps. Defaults to 100.
+        overshoot: Overshoot parameter for noise control. Defaults to 0.02.
+        num_classes: Number of classes to consider. Defaults to 10.
+        clip_min: Minimum value for clipping. Defaults to 0.0.
+        clip_max: Maximum value for clipping. Defaults to 1.0.
     """
 
     def __init__(
@@ -24,19 +34,6 @@ def __init__(
         clip_min: float = 0.0,
         clip_max: float = 1.0,
     ) -> None:
-        """Initialize the DeepFool attack.
-
-        Args:
-            model: The model to attack.
-            normalize: A transform to normalize images.
-            device: Device to use for tensors. Defaults to cuda if available.
-            steps: Number of steps. Defaults to 100.
-            overshoot: Overshoot parameter for noise control. Defaults to 0.02.
-            num_classes: Number of classes to consider. Defaults to 10.
-            clip_min: Minimum value for clipping. Defaults to 0.0.
-            clip_max: Maximum value for clipping. Defaults to 1.0.
-        """
-
         super().__init__(normalize, device)
 
         self.model = model
diff --git a/src/torchattack/difgsm.py b/src/torchattack/difgsm.py
index f5955b0..1463a04 100644
--- a/src/torchattack/difgsm.py
+++ b/src/torchattack/difgsm.py
@@ -12,6 +12,26 @@ class DIFGSM(Attack):
 
     From the paper 'Improving Transferability of Adversarial Examples with Input
     Diversity' https://arxiv.org/abs/1803.06978
+
+    Note:
+        Key parameters include `resize_rate` and `diversity_prob`, which defines the
+        scale size of the resized image and the probability of applying input
+        diversity. The default values are set to 0.9 and 1.0 respectively (implying
+        that input diversity is always applied).
+
+    Args:
+        model: The model to attack.
+        normalize: A transform to normalize images.
+        device: Device to use for tensors. Defaults to cuda if available.
+        eps: The maximum perturbation. Defaults to 8/255.
+        steps: Number of steps. Defaults to 10.
+        alpha: Step size, `eps / steps` if None. Defaults to None.
+        decay: Decay factor for the momentum term. Defaults to 1.0.
+        resize_rate: The resize rate. Defaults to 0.9.
+        diversity_prob: Applying input diversity with probability. Defaults to 1.0.
+        clip_min: Minimum value for clipping. Defaults to 0.0.
+        clip_max: Maximum value for clipping. Defaults to 1.0.
+        targeted: Targeted attack if True. Defaults to False.
     """
 
     def __init__(
@@ -29,29 +49,6 @@ def __init__(
         clip_max: float = 1.0,
         targeted: bool = False,
     ) -> None:
-        """Initialize the DI-FGSM attack.
-
-        Note:
-            Key parameters include `resize_rate` and `diversity_prob`, which defines the
-            scale size of the resized image and the probability of applying input
-            diversity. The default values are set to 0.9 and 1.0 respectively (implying
-            that input diversity is always applied).
-
-        Args:
-            model: The model to attack.
-            normalize: A transform to normalize images.
-            device: Device to use for tensors. Defaults to cuda if available.
-            eps: The maximum perturbation. Defaults to 8/255.
-            steps: Number of steps. Defaults to 10.
-            alpha: Step size, `eps / steps` if None. Defaults to None.
-            decay: Decay factor for the momentum term. Defaults to 1.0.
-            resize_rate: The resize rate. Defaults to 0.9.
-            diversity_prob: Applying input diversity with probability. Defaults to 1.0.
-            clip_min: Minimum value for clipping. Defaults to 0.0.
-            clip_max: Maximum value for clipping. Defaults to 1.0.
-            targeted: Targeted attack if True. Defaults to False.
-        """
-
         super().__init__(normalize, device)
 
         self.model = model
diff --git a/src/torchattack/fgsm.py b/src/torchattack/fgsm.py
index 539b4d1..5ab5984 100644
--- a/src/torchattack/fgsm.py
+++ b/src/torchattack/fgsm.py
@@ -11,6 +11,15 @@ class FGSM(Attack):
 
     From the paper 'Explaining and Harnessing Adversarial Examples',
     https://arxiv.org/abs/1412.6572
+
+    Args:
+        model: A torch.nn.Module network model.
+        normalize: A transform to normalize images.
+        device: Device to use for tensors. Defaults to cuda if available.
+        eps: Maximum perturbation measured by Linf. Defaults to 8/255.
+        clip_min: Minimum value for clipping. Defaults to 0.0.
+        clip_max: Maximum value for clipping. Defaults to 1.0.
+        targeted: Targeted attack if True. Defaults to False.
     """
 
     def __init__(
@@ -23,18 +32,6 @@ def __init__(
         clip_max: float = 1.0,
         targeted: bool = False,
     ) -> None:
-        """Initialize the FGSM attack.
-
-        Args:
-            model: A torch.nn.Module network model.
-            normalize: A transform to normalize images.
-            device: Device to use for tensors. Defaults to cuda if available.
-            eps: Maximum perturbation measured by Linf. Defaults to 8/255.
-            clip_min: Minimum value for clipping. Defaults to 0.0.
-            clip_max: Maximum value for clipping. Defaults to 1.0.
-            targeted: Targeted attack if True. Defaults to False.
-        """
-
         super().__init__(normalize, device)
 
         self.model = model
diff --git a/src/torchattack/mifgsm.py b/src/torchattack/mifgsm.py
index 5df6d38..ec9bf8c 100644
--- a/src/torchattack/mifgsm.py
+++ b/src/torchattack/mifgsm.py
@@ -11,6 +11,18 @@ class MIFGSM(Attack):
 
     From the paper 'Boosting Adversarial Attacks with Momentum',
     https://arxiv.org/abs/1710.06081
+
+    Args:
+        model: The model to attack.
+        normalize: A transform to normalize images.
+        device: Device to use for tensors. Defaults to cuda if available.
+        eps: The maximum perturbation. Defaults to 8/255.
+        steps: Number of steps. Defaults to 10.
+        alpha: Step size, `eps / steps` if None. Defaults to None.
+        decay: Decay factor for the momentum term. Defaults to 1.0.
+        clip_min: Minimum value for clipping. Defaults to 0.0.
+        clip_max: Maximum value for clipping. Defaults to 1.0.
+        targeted: Targeted attack if True. Defaults to False.
     """
 
     def __init__(
@@ -26,21 +38,6 @@ def __init__(
         clip_max: float = 1.0,
         targeted: bool = False,
     ) -> None:
-        """Initialize the MI-FGSM attack.
-
-        Args:
-            model: The model to attack.
-            normalize: A transform to normalize images.
-            device: Device to use for tensors. Defaults to cuda if available.
-            eps: The maximum perturbation. Defaults to 8/255.
-            steps: Number of steps. Defaults to 10.
-            alpha: Step size, `eps / steps` if None. Defaults to None.
-            decay: Decay factor for the momentum term. Defaults to 1.0.
-            clip_min: Minimum value for clipping. Defaults to 0.0.
-            clip_max: Maximum value for clipping. Defaults to 1.0.
-            targeted: Targeted attack if True. Defaults to False.
-        """
-
         super().__init__(normalize, device)
 
         self.model = model
diff --git a/src/torchattack/nifgsm.py b/src/torchattack/nifgsm.py
index 507e516..7f0f40c 100644
--- a/src/torchattack/nifgsm.py
+++ b/src/torchattack/nifgsm.py
@@ -15,6 +15,18 @@ class NIFGSM(Attack):
 
     From the paper 'Nesterov Accelerated Gradient and Scale Invariance for Adversarial
     Attacks' https://arxiv.org/abs/1908.06281
+
+    Args:
+        model: The model to attack.
+        normalize: A transform to normalize images.
+        device: Device to use for tensors. Defaults to cuda if available.
+        eps: The maximum perturbation. Defaults to 8/255.
+        steps: Number of steps. Defaults to 10.
+        alpha: Step size, `eps / steps` if None. Defaults to None.
+        decay: Decay factor for the momentum term. Defaults to 1.0.
+        clip_min: Minimum value for clipping. Defaults to 0.0.
+        clip_max: Maximum value for clipping. Defaults to 1.0.
+        targeted: Targeted attack if True. Defaults to False.
     """
 
     def __init__(
@@ -30,21 +42,6 @@ def __init__(
         clip_max: float = 1.0,
         targeted: bool = False,
     ) -> None:
-        """Initialize the NI-FGSM attack.
-
-        Args:
-            model: The model to attack.
-            normalize: A transform to normalize images.
-            device: Device to use for tensors. Defaults to cuda if available.
-            eps: The maximum perturbation. Defaults to 8/255.
-            steps: Number of steps. Defaults to 10.
-            alpha: Step size, `eps / steps` if None. Defaults to None.
-            decay: Decay factor for the momentum term. Defaults to 1.0.
-            clip_min: Minimum value for clipping. Defaults to 0.0.
-            clip_max: Maximum value for clipping. Defaults to 1.0.
-            targeted: Targeted attack if True. Defaults to False.
-        """
-
         super().__init__(normalize, device)
 
         self.model = model
diff --git a/src/torchattack/pgd.py b/src/torchattack/pgd.py
index 88713bf..48b0e9f 100644
--- a/src/torchattack/pgd.py
+++ b/src/torchattack/pgd.py
@@ -11,6 +11,18 @@ class PGD(Attack):
 
     From the paper 'Towards Deep Learning Models Resistant to Adversarial Attacks'
     https://arxiv.org/abs/1706.06083
+
+    Args:
+        model: The model to attack.
+        normalize: A transform to normalize images.
+        device: Device to use for tensors. Defaults to cuda if available.
+        eps: The maximum perturbation. Defaults to 8/255.
+        steps: Number of steps. Defaults to 10.
+        alpha: Step size, `eps / steps` if None. Defaults to None.
+        random_start: Start from random uniform perturbation. Defaults to True.
+        clip_min: Minimum value for clipping. Defaults to 0.0.
+        clip_max: Maximum value for clipping. Defaults to 1.0.
+        targeted: Targeted attack if True. Defaults to False.
     """
 
     def __init__(
@@ -26,21 +38,6 @@ def __init__(
         clip_max: float = 1.0,
         targeted: bool = False,
     ) -> None:
-        """Initialize the PGD attack.
-
-        Args:
-            model: The model to attack.
-            normalize: A transform to normalize images.
-            device: Device to use for tensors. Defaults to cuda if available.
-            eps: The maximum perturbation. Defaults to 8/255.
-            steps: Number of steps. Defaults to 10.
-            alpha: Step size, `eps / steps` if None. Defaults to None.
-            random_start: Start from random uniform perturbation. Defaults to True.
-            clip_min: Minimum value for clipping. Defaults to 0.0.
-            clip_max: Maximum value for clipping. Defaults to 1.0.
-            targeted: Targeted attack if True. Defaults to False.
-        """
-
         super().__init__(normalize, device)
 
         self.model = model
diff --git a/src/torchattack/pgdl2.py b/src/torchattack/pgdl2.py
index 34a651d..042a226 100644
--- a/src/torchattack/pgdl2.py
+++ b/src/torchattack/pgdl2.py
@@ -13,6 +13,18 @@ class PGDL2(Attack):
 
     From the paper 'Towards Deep Learning Models Resistant to Adversarial Attacks',
     https://arxiv.org/abs/1706.06083
+
+    Args:
+        model: The model to attack.
+        normalize: A transform to normalize images.
+        device: Device to use for tensors. Defaults to cuda if available.
+        eps: The maximum perturbation, measured in L2. Defaults to 1.0.
+        steps: Number of steps. Defaults to 10.
+        alpha: Step size, `eps / steps` if None. Defaults to None.
+        random_start: Start from random uniform perturbation. Defaults to True.
+        clip_min: Minimum value for clipping. Defaults to 0.0.
+        clip_max: Maximum value for clipping. Defaults to 1.0.
+        targeted: Targeted attack if True. Defaults to False.
     """
 
     def __init__(
@@ -28,21 +40,6 @@ def __init__(
         clip_max: float = 1.0,
         targeted: bool = False,
     ) -> None:
-        """Initialize the PGD attack.
-
-        Args:
-            model: The model to attack.
-            normalize: A transform to normalize images.
-            device: Device to use for tensors. Defaults to cuda if available.
-            eps: The maximum perturbation, measured in L2. Defaults to 1.0.
-            steps: Number of steps. Defaults to 10.
-            alpha: Step size, `eps / steps` if None. Defaults to None.
-            random_start: Start from random uniform perturbation. Defaults to True.
-            clip_min: Minimum value for clipping. Defaults to 0.0.
-            clip_max: Maximum value for clipping. Defaults to 1.0.
-            targeted: Targeted attack if True. Defaults to False.
-        """
-
         super().__init__(normalize, device)
 
         self.model = model
diff --git a/src/torchattack/sinifgsm.py b/src/torchattack/sinifgsm.py
index 85567b0..717b7d5 100644
--- a/src/torchattack/sinifgsm.py
+++ b/src/torchattack/sinifgsm.py
@@ -9,8 +9,21 @@
 class SINIFGSM(Attack):
     """The SI-NI-FGSM (Scale-invariant Nesterov-accelerated Iterative FGSM) attack.
 
-    From the paper 'Nesterov Accelerated Gradient and Scale Invariance for Adversarial
-    Attacks' https://arxiv.org/abs/1908.06281
+        From the paper 'Nesterov Accelerated Gradient and Scale Invariance for Adversarial
+        Attacks' https://arxiv.org/abs/1908.06281
+
+    Args:
+        model: The model to attack.
+        normalize: A transform to normalize images.
+        device: Device to use for tensors. Defaults to cuda if available.
+        eps: The maximum perturbation. Defaults to 8/255.
+        steps: Number of steps. Defaults to 10.
+        alpha: Step size, `eps / steps` if None. Defaults to None.
+        decay: Decay factor for the momentum term. Defaults to 1.0.
+        m: Number of scaled copies of the image. Defaults to 3.
+        clip_min: Minimum value for clipping. Defaults to 0.0.
+        clip_max: Maximum value for clipping. Defaults to 1.0.
+        targeted: Targeted attack if True. Defaults to False.
     """
 
     def __init__(
@@ -27,22 +40,6 @@ def __init__(
         clip_max: float = 1.0,
         targeted: bool = False,
     ) -> None:
-        """Initialize the SI-NI-FGSM attack.
-
-        Args:
-            model: The model to attack.
-            normalize: A transform to normalize images.
-            device: Device to use for tensors. Defaults to cuda if available.
-            eps: The maximum perturbation. Defaults to 8/255.
-            steps: Number of steps. Defaults to 10.
-            alpha: Step size, `eps / steps` if None. Defaults to None.
-            decay: Decay factor for the momentum term. Defaults to 1.0.
-            m: Number of scaled copies of the image. Defaults to 3.
-            clip_min: Minimum value for clipping. Defaults to 0.0.
-            clip_max: Maximum value for clipping. Defaults to 1.0.
-            targeted: Targeted attack if True. Defaults to False.
-        """
-
         super().__init__(normalize, device)
 
         self.model = model
diff --git a/src/torchattack/tifgsm.py b/src/torchattack/tifgsm.py
index 28d2215..a8fd8bb 100644
--- a/src/torchattack/tifgsm.py
+++ b/src/torchattack/tifgsm.py
@@ -13,6 +13,25 @@ class TIFGSM(Attack):
 
     From the paper 'Evading Defenses to Transferable Adversarial Examples by
     Translation-Invariant Attacks' https://arxiv.org/abs/1904.02884
+
+    Note:
+        Key parameters include `kernel_len` and `n_sig`, which defines the size and
+        the radius of the gaussian kernel. The default values are set to 15 and 3
+        respectively, which are best according to the paper.
+
+    Args:
+        model: The model to attack.
+        normalize: A transform to normalize images.
+        device: Device to use for tensors. Defaults to cuda if available.
+        eps: The maximum perturbation. Defaults to 8/255.
+        steps: Number of steps. Defaults to 10.
+        alpha: Step size, `eps / steps` if None. Defaults to None.
+        decay: Decay factor for the momentum term. Defaults to 1.0.
+        kern_len: Length of the kernel (should be an odd number). Defaults to 15.
+        n_sig: Radius of the gaussian kernel. Defaults to 3.
+        clip_min: Minimum value for clipping. Defaults to 0.0.
+        clip_max: Maximum value for clipping. Defaults to 1.0.
+        targeted: Targeted attack if True. Defaults to False.
     """
 
     def __init__(
@@ -30,28 +49,6 @@ def __init__(
         clip_max: float = 1.0,
         targeted: bool = False,
     ) -> None:
-        """Initialize the TI-FGSM attack.
-
-        Note:
-            Key parameters include `kernel_len` and `n_sig`, which defines the size and
-            the radius of the gaussian kernel. The default values are set to 15 and 3
-            respectively, which are best according to the paper.
-
-        Args:
-            model: The model to attack.
-            normalize: A transform to normalize images.
-            device: Device to use for tensors. Defaults to cuda if available.
-            eps: The maximum perturbation. Defaults to 8/255.
-            steps: Number of steps. Defaults to 10.
-            alpha: Step size, `eps / steps` if None. Defaults to None.
-            decay: Decay factor for the momentum term. Defaults to 1.0.
-            kern_len: Length of the kernel (should be an odd number). Defaults to 15.
-            n_sig: Radius of the gaussian kernel. Defaults to 3.
-            clip_min: Minimum value for clipping. Defaults to 0.0.
-            clip_max: Maximum value for clipping. Defaults to 1.0.
-            targeted: Targeted attack if True. Defaults to False.
-        """
-
         super().__init__(normalize, device)
 
         self.model = model
diff --git a/src/torchattack/vmifgsm.py b/src/torchattack/vmifgsm.py
index 65190b8..9b2e82b 100644
--- a/src/torchattack/vmifgsm.py
+++ b/src/torchattack/vmifgsm.py
@@ -12,6 +12,25 @@ class VMIFGSM(Attack):
     From the paper
     'Enhancing the Transferability of Adversarial Attacks through Variance Tuning'
     https://arxiv.org/abs/2103.15571
+
+    Note:
+        Key parameters are `n` and `beta`, where `n` is the number of sampled
+        examples for variance tuning and `beta` is the upper bound of the
+        neighborhood for varying the perturbation.
+
+    Args:
+        model: The model to attack.
+        normalize: A transform to normalize images.
+        device: Device to use for tensors. Defaults to cuda if available.
+        eps: The maximum perturbation. Defaults to 8/255.
+        steps: Number of steps. Defaults to 10.
+        alpha: Step size, `eps / steps` if None. Defaults to None.
+        decay: Decay factor for the momentum term. Defaults to 1.0.
+        n: Number of sampled examples for variance tuning. Defaults to 5.
+        beta: The upper bound of the neighborhood. Defaults to 1.5.
+        clip_min: Minimum value for clipping. Defaults to 0.0.
+        clip_max: Maximum value for clipping. Defaults to 1.0.
+        targeted: Targeted attack if True. Defaults to False.
     """
 
     def __init__(
@@ -29,28 +48,6 @@ def __init__(
         clip_max: float = 1.0,
         targeted: bool = False,
     ) -> None:
-        """Initialize the VMI-FGSM attack.
-
-        Note:
-            Key parameters are `n` and `beta`, where `n` is the number of sampled
-            examples for variance tuning and `beta` is the upper bound of the
-            neighborhood for varying the perturbation.
-
-        Args:
-            model: The model to attack.
-            normalize: A transform to normalize images.
-            device: Device to use for tensors. Defaults to cuda if available.
-            eps: The maximum perturbation. Defaults to 8/255.
-            steps: Number of steps. Defaults to 10.
-            alpha: Step size, `eps / steps` if None. Defaults to None.
-            decay: Decay factor for the momentum term. Defaults to 1.0.
-            n: Number of sampled examples for variance tuning. Defaults to 5.
-            beta: The upper bound of the neighborhood. Defaults to 1.5.
-            clip_min: Minimum value for clipping. Defaults to 0.0.
-            clip_max: Maximum value for clipping. Defaults to 1.0.
-            targeted: Targeted attack if True. Defaults to False.
-        """
-
         super().__init__(normalize, device)
 
         self.model = model
diff --git a/src/torchattack/vnifgsm.py b/src/torchattack/vnifgsm.py
index d8103a7..9a6e894 100644
--- a/src/torchattack/vnifgsm.py
+++ b/src/torchattack/vnifgsm.py
@@ -11,6 +11,25 @@ class VNIFGSM(Attack):
 
     From the paper 'Enhancing the Transferability of Adversarial Attacks through
     Variance Tuning' https://arxiv.org/abs/2103.15571
+
+    Note:
+        Key parameters are `n` and `beta`, where `n` is the number of sampled
+        examples for variance tuning and `beta` is the upper bound of the
+        neighborhood for varying the perturbation.
+
+    Args:
+        model: The model to attack.
+        normalize: A transform to normalize images.
+        device: Device to use for tensors. Defaults to cuda if available.
+        eps: The maximum perturbation. Defaults to 8/255.
+        steps: Number of steps. Defaults to 10.
+        alpha: Step size, `eps / steps` if None. Defaults to None.
+        decay: Decay factor for the momentum term. Defaults to 1.0.
+        n: Number of sampled examples for variance tuning. Defaults to 5.
+        beta: The upper bound of the neighborhood. Defaults to 1.5.
+        clip_min: Minimum value for clipping. Defaults to 0.0.
+        clip_max: Maximum value for clipping. Defaults to 1.0.
+        targeted: Targeted attack if True. Defaults to False.
     """
 
     def __init__(
@@ -28,28 +47,6 @@ def __init__(
         clip_max: float = 1.0,
         targeted: bool = False,
     ) -> None:
-        """Initialize the VNI-FGSM attack.
-
-        Note:
-            Key parameters are `n` and `beta`, where `n` is the number of sampled
-            examples for variance tuning and `beta` is the upper bound of the
-            neighborhood for varying the perturbation.
-
-        Args:
-            model: The model to attack.
-            normalize: A transform to normalize images.
-            device: Device to use for tensors. Defaults to cuda if available.
-            eps: The maximum perturbation. Defaults to 8/255.
-            steps: Number of steps. Defaults to 10.
-            alpha: Step size, `eps / steps` if None. Defaults to None.
-            decay: Decay factor for the momentum term. Defaults to 1.0.
-            n: Number of sampled examples for variance tuning. Defaults to 5.
-            beta: The upper bound of the neighborhood. Defaults to 1.5.
-            clip_min: Minimum value for clipping. Defaults to 0.0.
-            clip_max: Maximum value for clipping. Defaults to 1.0.
-            targeted: Targeted attack if True. Defaults to False.
-        """
-
         super().__init__(normalize, device)
 
         self.model = model