unifying art tools in verbose interface

Signed-off-by: GiulioZizzo <[email protected]>

Author: GiulioZizzo

art/attacks/extraction/knockoff_nets.py

@@ -155,7 +155,7 @@ def _random_extraction(self, x: np.ndarray, thieved_classifier: "CLASSIFIER_TYPE
             y=fake_labels,
             batch_size=self.batch_size_fit,
             nb_epochs=self.nb_epochs,
-            verbose=0,
+            verbose=False,
         )
 
         return thieved_classifier
@@ -243,7 +243,7 @@ def _adaptive_extraction(
                 y=fake_label,
                 batch_size=self.batch_size_fit,
                 nb_epochs=1,
-                verbose=0,
+                verbose=False,
             )
 
             # Test new labels

art/attacks/poisoning/sleeper_agent_attack.py

@@ -360,7 +360,7 @@ def _create_model(
             for layer in model_pt.model.children():
                 if hasattr(layer, "reset_parameters"):
                     layer.reset_parameters()  # type: ignore
-            model_pt.fit(x_train, y_train, batch_size=batch_size, nb_epochs=epochs, verbose=1)
+            model_pt.fit(x_train, y_train, batch_size=batch_size, nb_epochs=epochs, verbose=True)
             predictions = model_pt.predict(x_test)
             accuracy = np.sum(np.argmax(predictions, axis=1) == np.argmax(y_test, axis=1)) / len(y_test)
             logger.info("Accuracy of retrained model : %s", accuracy * 100.0)
@@ -370,7 +370,7 @@ def _create_model(
 
             self.substitute_classifier.model.trainable = True
             model_tf = self.substitute_classifier.clone_for_refitting()
-            model_tf.fit(x_train, y_train, batch_size=batch_size, nb_epochs=epochs, verbose=0)
+            model_tf.fit(x_train, y_train, batch_size=batch_size, nb_epochs=epochs, verbose=False)
             predictions = model_tf.predict(x_test)
             accuracy = np.sum(np.argmax(predictions, axis=1) == np.argmax(y_test, axis=1)) / len(y_test)
             logger.info("Accuracy of retrained model : %s", accuracy * 100.0)

art/defences/trainer/adversarial_trainer.py

@@ -188,7 +188,9 @@ def fit_generator(self, generator: "DataGenerator", nb_epochs: int = 20, **kwarg
                     x_batch[adv_ids] = x_adv
 
                 # Fit batch
-                self._classifier.fit(x_batch, y_batch, nb_epochs=1, batch_size=x_batch.shape[0], verbose=0, **kwargs)
+                self._classifier.fit(
+                    x_batch, y_batch, nb_epochs=1, batch_size=x_batch.shape[0], verbose=False, **kwargs
+                )
                 attack_id = (attack_id + 1) % len(self.attacks)
 
     def fit(  # pylint: disable=W0221
@@ -260,7 +262,9 @@ def fit(  # pylint: disable=W0221
                     x_batch[adv_ids] = x_adv
 
                 # Fit batch
-                self._classifier.fit(x_batch, y_batch, nb_epochs=1, batch_size=x_batch.shape[0], verbose=0, **kwargs)
+                self._classifier.fit(
+                    x_batch, y_batch, nb_epochs=1, batch_size=x_batch.shape[0], verbose=False, **kwargs
+                )
                 attack_id = (attack_id + 1) % len(self.attacks)
 
     def predict(self, x: np.ndarray, **kwargs) -> np.ndarray:

art/defences/trainer/dp_instahide_trainer.py

@@ -155,7 +155,7 @@ def fit(  # pylint: disable=W0221
                 x_aug = self._generate_noise(x_aug)
 
                 # fit batch
-                self._classifier.fit(x_aug, y_aug, nb_epochs=1, batch_size=x_aug.shape[0], verbose=0, **kwargs)
+                self._classifier.fit(x_aug, y_aug, nb_epochs=1, batch_size=x_aug.shape[0], verbose=False, **kwargs)
 
                 # get metrics
                 loss = self._classifier.compute_loss(x_aug, y_aug, reduction="mean")
@@ -234,7 +234,7 @@ def fit_generator(self, generator: "DataGenerator", nb_epochs: int = 20, **kwarg
                 x_aug = self._generate_noise(x_aug)
 
                 # fit batch
-                self._classifier.fit(x_aug, y_aug, nb_epochs=1, batch_size=x_aug.shape[0], verbose=0, **kwargs)
+                self._classifier.fit(x_aug, y_aug, nb_epochs=1, batch_size=x_aug.shape[0], verbose=False, **kwargs)
 
                 # get metrics
                 loss = self._classifier.compute_loss(x_aug, y_aug, reduction="mean")

art/estimators/certification/randomized_smoothing/macer/pytorch.py

@@ -75,7 +75,6 @@ def __init__(
         gamma: float = 8.0,
         lmbda: float = 12.0,
         gaussian_samples: int = 16,
-        verbose: bool = False,
     ) -> None:
         """
         Create a MACER classifier.
@@ -105,7 +104,6 @@ def __init__(
         :param gamma: The hinge factor.
         :param lmbda: The trade-off factor.
         :param gaussian_samples: The number of gaussian samples per input.
-        :param verbose: Show progress bars.
         """
         super().__init__(
             model=model,
@@ -122,7 +120,6 @@ def __init__(
             sample_size=sample_size,
             scale=scale,
             alpha=alpha,
-            verbose=verbose,
         )
         self.beta = beta
         self.gamma = gamma

art/estimators/certification/randomized_smoothing/macer/tensorflow.py

@@ -75,7 +75,6 @@ def __init__(
         gamma: float = 8.0,
         lmbda: float = 12.0,
         gaussian_samples: int = 16,
-        verbose: bool = False,
     ) -> None:
         """
         Create a MACER classifier.
@@ -108,7 +107,6 @@ def __init__(
         :param gamma: The hinge factor.
         :param lmbda: The trade-off factor.
         :param gaussian_samples: The number of gaussian samples per input.
-        :param verbose: Show progress bars.
         """
         super().__init__(
             model=model,
@@ -125,21 +123,14 @@ def __init__(
             sample_size=sample_size,
             scale=scale,
             alpha=alpha,
-            verbose=verbose,
         )
         self.beta = beta
         self.gamma = gamma
         self.lmbda = lmbda
         self.gaussian_samples = gaussian_samples
 
     def fit(
-        self,
-        x: np.ndarray,
-        y: np.ndarray,
-        batch_size: int = 128,
-        nb_epochs: int = 10,
-        verbose: bool = False,
-        **kwargs
+        self, x: np.ndarray, y: np.ndarray, batch_size: int = 128, nb_epochs: int = 10, verbose: bool = False, **kwargs
     ) -> None:
         """
         Fit the classifier on the training set `(x, y)`.

art/estimators/certification/randomized_smoothing/pytorch.py

@@ -71,7 +71,6 @@ def __init__(
         sample_size: int = 32,
         scale: float = 0.1,
         alpha: float = 0.001,
-        verbose: bool = False,
     ):
         """
         Create a randomized smoothing classifier.
@@ -97,7 +96,6 @@ def __init__(
         :param sample_size: Number of samples for smoothing.
         :param scale: Standard deviation of Gaussian noise added.
         :param alpha: The failure probability of smoothing.
-        :param verbose: Show progress bars.
         """
         if preprocessing_defences is not None:
             warnings.warn(
@@ -120,7 +118,6 @@ def __init__(
             sample_size=sample_size,
             scale=scale,
             alpha=alpha,
-            verbose=verbose,
         )
 
     def _predict_classifier(self, x: np.ndarray, batch_size: int, training_mode: bool, **kwargs) -> np.ndarray:

art/estimators/certification/randomized_smoothing/randomized_smoothing.py

@@ -49,7 +49,6 @@ def __init__(
         *args,
         scale: float = 0.1,
         alpha: float = 0.001,
-        verbose: bool = False,
         **kwargs,
     ) -> None:
         """
@@ -58,13 +57,11 @@ def __init__(
         :param sample_size: Number of samples for smoothing.
         :param scale: Standard deviation of Gaussian noise added.
         :param alpha: The failure probability of smoothing.
-        :param verbose: Show progress bars.
         """
         super().__init__(*args, **kwargs)  # type: ignore
         self.sample_size = sample_size
         self.scale = scale
         self.alpha = alpha
-        self.verbose = verbose
 
     def _predict_classifier(self, x: np.ndarray, batch_size: int, training_mode: bool, **kwargs) -> np.ndarray:
         """
@@ -77,12 +74,13 @@ def _predict_classifier(self, x: np.ndarray, batch_size: int, training_mode: boo
         """
         raise NotImplementedError
 
-    def predict(self, x: np.ndarray, batch_size: int = 128, **kwargs) -> np.ndarray:
+    def predict(self, x: np.ndarray, batch_size: int = 128, verbose: bool = False, **kwargs) -> np.ndarray:
         """
         Perform prediction of the given classifier for a batch of inputs, taking an expectation over transformations.
 
         :param x: Input samples.
         :param batch_size: Batch size.
+        :param verbose: Display training progress bar.
         :param is_abstain: True if function will abstain from prediction and return 0s. Default: True
         :type is_abstain: `boolean`
         :return: Array of predictions of shape `(nb_inputs, nb_classes)`.
@@ -98,7 +96,7 @@ def predict(self, x: np.ndarray, batch_size: int = 128, **kwargs) -> np.ndarray:
         logger.info("Applying randomized smoothing.")
         n_abstained = 0
         prediction = []
-        for x_i in tqdm(x, desc="Randomized smoothing", disable=not self.verbose):
+        for x_i in tqdm(x, desc="Randomized smoothing", disable=not verbose):
             # get class counts
             counts_pred = self._prediction_counts(x_i, batch_size=batch_size)
             top = counts_pred.argsort()[::-1]

art/estimators/certification/randomized_smoothing/smooth_adv/pytorch.py

@@ -77,7 +77,6 @@ def __init__(
         num_noise_vec: int = 1,
         num_steps: int = 10,
         warmup: int = 1,
-        verbose: bool = False,
     ) -> None:
         """
         Create a SmoothAdv classifier.
@@ -107,7 +106,6 @@ def __init__(
         :param num_noise_vec: The number of noise vectors.
         :param num_steps: The number of attack updates.
         :param warmup: The warm-up strategy that is gradually increased up to the original value.
-        :param verbose: Show progress bars.
         """
         super().__init__(
             model=model,
@@ -124,7 +122,6 @@ def __init__(
             sample_size=sample_size,
             scale=scale,
             alpha=alpha,
-            verbose=verbose,
         )
         self.epsilon = epsilon
         self.num_noise_vec = num_noise_vec

art/estimators/certification/randomized_smoothing/smooth_adv/tensorflow.py

@@ -77,7 +77,6 @@ def __init__(
         num_noise_vec: int = 1,
         num_steps: int = 10,
         warmup: int = 1,
-        verbose: bool = False,
     ) -> None:
         """
         Create a MACER classifier.
@@ -110,7 +109,6 @@ def __init__(
         :param num_noise_vec: The number of noise vectors.
         :param num_steps: The number of attack updates.
         :param warmup: The warm-up strategy that is gradually increased up to the original value.
-        :param verbose: Show progress bars.
         """
         super().__init__(
             model=model,
@@ -150,13 +148,7 @@ def __init__(
         self.attack = ProjectedGradientDescent(classifier, eps=self.epsilon, max_iter=1, verbose=False)
 
     def fit(
-        self,
-        x: np.ndarray,
-        y: np.ndarray,
-        batch_size: int = 128,
-        nb_epochs: int = 10,
-        verbose: bool = False,
-        **kwargs
+        self, x: np.ndarray, y: np.ndarray, batch_size: int = 128, nb_epochs: int = 10, verbose: bool = False, **kwargs
     ) -> None:
         """
         Fit the classifier on the training set `(x, y)`.