Merge branch 'dev_1.9.0' into development_maintenance_190

Author: beat-buesser

.github/workflows/ci-lingvo.yml

@@ -47,7 +47,7 @@ jobs:
           sudo apt-get update
           sudo apt-get -y -q install ffmpeg libavcodec-extra
           python -m pip install --upgrade pip setuptools wheel
-          pip install -q -r <(sed '/^scipy/d;/^matplotlib/d;/^pandas/d;/^statsmodels/d;/^numba/d' requirements_test.txt)
+          pip install -q -r <(sed '/^scipy/d;/^matplotlib/d;/^pandas/d;/^statsmodels/d;/^numba/d;/^jax/d' requirements_test.txt)
           pip install scipy==1.5.4
           pip install matplotlib==3.3.4
           pip install pandas==1.1.5
@@ -58,6 +58,7 @@ jobs:
           pip install lingvo==${{ matrix.lingvo }}
           pip install tensorflow-addons==0.9.1
           pip install model-pruning-google-research==0.0.3
+          pip install jax[cpu]==0.2.17
           pip list
       - name: Run ${{ matrix.name }} Tests
         run: ./run_tests.sh ${{ matrix.framework }}

art/attacks/evasion/pixel_threshold.py

@@ -156,9 +156,11 @@ def generate(self, x: np.ndarray, y: Optional[np.ndarray] = None, **kwargs) -> n
                 raise ValueError(
                     "This attack has not yet been tested for binary classification with a single output classifier."
                 )
-            if len(y.shape) > 1:
+            if y.ndim > 1 and y.shape[1] > 1:
                 y = np.argmax(y, axis=1)
 
+        y = np.squeeze(y)
+
         if self.th is None:
             logger.info(
                 "Performing minimal perturbation Attack. \

art/defences/preprocessor/mp3_compression_pytorch.py

@@ -62,22 +62,19 @@ def __init__(
         :param device_type: Type of device on which the classifier is run, either `gpu` or `cpu`.
         :param verbose: Show progress bars.
         """
-        import torch  # lgtm [py/repeated-import]
         from torch.autograd import Function
 
-        super().__init__(is_fitted=True, apply_fit=apply_fit, apply_predict=apply_predict)
+        super().__init__(
+            device_type=device_type,
+            is_fitted=True,
+            apply_fit=apply_fit,
+            apply_predict=apply_predict,
+        )
         self.channels_first = channels_first
         self.sample_rate = sample_rate
         self.verbose = verbose
         self._check_params()
 
-        # Set device
-        if device_type == "cpu" or not torch.cuda.is_available():
-            self._device = torch.device("cpu")
-        else:  # pragma: no cover
-            cuda_idx = torch.cuda.current_device()
-            self._device = torch.device("cuda:{}".format(cuda_idx))
-
         self.compression_numpy = Mp3Compression(
             sample_rate=sample_rate,
             channels_first=channels_first,

art/defences/preprocessor/preprocessor.py

@@ -140,6 +140,18 @@ class PreprocessorPyTorch(Preprocessor):
     Abstract base class for preprocessing defences implemented in PyTorch that support efficient preprocessor-chaining.
     """
 
+    def __init__(self, device_type: str = "gpu", **kwargs):
+        import torch  # lgtm [py/repeated-import]
+
+        super().__init__(**kwargs)
+
+        # Set device
+        if device_type == "cpu" or not torch.cuda.is_available():
+            self._device = torch.device("cpu")
+        else:  # pragma: no cover
+            cuda_idx = torch.cuda.current_device()
+            self._device = torch.device("cuda:{}".format(cuda_idx))
+
     @abc.abstractmethod
     def forward(
         self, x: "torch.Tensor", y: Optional["torch.Tensor"] = None

art/defences/preprocessor/spatial_smoothing_pytorch.py

@@ -72,22 +72,18 @@ def __init__(
         :param apply_predict: True if applied during predicting.
         :param device_type: Type of device on which the classifier is run, either `gpu` or `cpu`.
         """
-        import torch  # lgtm [py/repeated-import]
 
-        super().__init__(apply_fit=apply_fit, apply_predict=apply_predict)
+        super().__init__(
+            device_type=device_type,
+            apply_fit=apply_fit,
+            apply_predict=apply_predict,
+        )
 
         self.channels_first = channels_first
         self.window_size = window_size
         self.clip_values = clip_values
         self._check_params()
 
-        # Set device
-        if device_type == "cpu" or not torch.cuda.is_available():
-            self._device = torch.device("cpu")
-        else:  # pragma: no cover
-            cuda_idx = torch.cuda.current_device()
-            self._device = torch.device("cuda:{}".format(cuda_idx))
-
         from kornia.filters import MedianBlur
 
         class MedianBlurCustom(MedianBlur):

art/defences/preprocessor/video_compression_pytorch.py

@@ -68,23 +68,20 @@ def __init__(
         :param device_type: Type of device on which the classifier is run, either `gpu` or `cpu`.
         :param verbose: Show progress bars.
         """
-        import torch  # lgtm [py/repeated-import]
         from torch.autograd import Function
 
-        super().__init__(is_fitted=True, apply_fit=apply_fit, apply_predict=apply_predict)
+        super().__init__(
+            device_type=device_type,
+            is_fitted=True,
+            apply_fit=apply_fit,
+            apply_predict=apply_predict,
+        )
         self.video_format = video_format
         self.constant_rate_factor = constant_rate_factor
         self.channels_first = channels_first
         self.verbose = verbose
         self._check_params()
 
-        # Set device
-        if device_type == "cpu" or not torch.cuda.is_available():
-            self._device = torch.device("cpu")
-        else:  # pragma: no cover
-            cuda_idx = torch.cuda.current_device()
-            self._device = torch.device("cuda:{}".format(cuda_idx))
-
         self.compression_numpy = VideoCompression(
             video_format=video_format,
             constant_rate_factor=constant_rate_factor,

art/estimators/classification/keras.py

@@ -562,13 +562,14 @@ def fit(self, x: np.ndarray, y: np.ndarray, batch_size: int = 128, nb_epochs: in
                `fit_generator` function in Keras and will be passed to this function as such. Including the number of
                epochs or the number of steps per epoch as part of this argument will result in as error.
         """
+        y_ndim = y.ndim
         y = check_and_transform_label_format(y, self.nb_classes)
 
         # Apply preprocessing
         x_preprocessed, y_preprocessed = self._apply_preprocessing(x, y, fit=True)
 
         # Adjust the shape of y for loss functions that do not take labels in one-hot encoding
-        if self._reduce_labels:
+        if self._reduce_labels or y_ndim == 1:
             y_preprocessed = np.argmax(y_preprocessed, axis=1)
 
         self._model.fit(x=x_preprocessed, y=y_preprocessed, batch_size=batch_size, epochs=nb_epochs, **kwargs)

art/experimental/__init__.py

@@ -0,0 +1,4 @@
+"""
+This module contains the experimental Estimator API.
+"""
+from art.experimental.estimators.jax import JaxEstimator

art/experimental/estimators/__init__.py

@@ -0,0 +1,4 @@
+"""
+Experimental classifiers.
+"""
+from art.experimental.estimators.classification.jax import JaxClassifier