Merge pull request #1219 from Trusted-AI/development_issue_1217

Remove duplicate application of preprocessing in fallback BaseEstimator.fit_generator

Author: beat-buesser

art/estimators/estimator.py

@@ -443,11 +443,8 @@ def fit_generator(self, generator: "DataGenerator", nb_epochs: int = 20, **kwarg
             ):  # type: ignore
                 x, y = generator.get_batch()
 
-                # Apply preprocessing and defences
-                x_preprocessed, y_preprocessed = self._apply_preprocessing(x, y, fit=True)  # type: ignore
-
                 # Fit for current batch
-                self.fit(x_preprocessed, y_preprocessed, nb_epochs=1, batch_size=generator.batch_size, **kwargs)
+                self.fit(x, y, nb_epochs=1, batch_size=generator.batch_size, **kwargs)
 
     @abstractmethod
     def get_activations(