Use explicit exception chaining

pl_bolts/callbacks/vision/confused_logit.py

@@ -64,12 +64,12 @@ def on_train_batch_end(self, trainer, pl_module, batch, batch_idx, dataloader_id
         x, y = batch
         try:
             logits = pl_module.last_logits
-        except AttributeError as e:
+        except AttributeError as err:
             m = """please track the last_logits in the training_step like so:
                 def training_step(...):
                     self.last_logits = your_logits
             """
-            raise AttributeError(m)
+            raise AttributeError(m) from err
 
         # only check when it has opinions (ie: the logit > 5)
         if logits.max() > self.min_logit_value:

pl_bolts/datamodules/base_dataset.py

@@ -54,5 +54,5 @@ def _download_from_url(self, base_url: str, data_folder: str, file_name: str):
         fpath = os.path.join(data_folder, file_name)
         try:
             urllib.request.urlretrieve(url, fpath)
-        except HTTPError:
-            raise RuntimeError(f'Failed download from {url}')
+        except HTTPError as err:
+            raise RuntimeError(f'Failed download from {url}') from err

pl_bolts/datamodules/imagenet_dataset.py

@@ -20,9 +20,9 @@
 try:
     from torchvision.datasets import ImageNet
     from torchvision.datasets.imagenet import load_meta_file
-except ImportError:
+except ImportError as err:
     raise ImportError('You want to use `torchvision` which is not installed yet,'  # pragma: no-cover
-                      ' install it with `pip install torchvision`.')
+                      ' install it with `pip install torchvision`.') from err
 
 
 class UnlabeledImagenet(ImageNet):

pl_bolts/datamodules/mnist_dataset.py

@@ -3,9 +3,9 @@
 try:
     from torchvision import transforms as transform_lib
     from torchvision.datasets import MNIST
-except ImportError:
+except ImportError as err:
     raise ImportError('You want to use `torchvision` which is not installed yet,'  # pragma: no-cover
-                      ' install it with `pip install torchvision`.')
+                      ' install it with `pip install torchvision`.') from err
 
 try:
     from PIL import Image

pl_bolts/models/regression/linear_regression.py

@@ -126,9 +126,9 @@ def cli_main():
     # create dataset
     try:
         from sklearn.datasets import load_boston
-    except ImportError:
+    except ImportError as err:
         raise ImportError('You want to use `sklearn` which is not installed yet,'  # pragma: no-cover
-                          ' install it with `pip install sklearn`.')
+                          ' install it with `pip install sklearn`.') from err
 
     X, y = load_boston(return_X_y=True)  # these are numpy arrays
     loaders = SklearnDataModule(X, y)

pl_bolts/models/regression/logistic_regression.py

@@ -132,9 +132,9 @@ def cli_main():
     # Example: Iris dataset in Sklearn (4 features, 3 class labels)
     try:
         from sklearn.datasets import load_iris
-    except ImportError:
+    except ImportError as err:
         raise ImportError('You want to use `sklearn` which is not installed yet,'  # pragma: no-cover
-                          ' install it with `pip install sklearn`.')
+                          ' install it with `pip install sklearn`.') from err
 
     X, y = load_iris(return_X_y=True)
     loaders = SklearnDataModule(X, y)

pl_bolts/models/self_supervised/cpc/cpc_module.py

@@ -176,9 +176,9 @@ def validation_step(self, batch, batch_nb):
     def shared_step(self, batch):
         try:
             from pl_bolts.datamodules.stl10_datamodule import STL10DataModule
-        except ImportError:
+        except ImportError as err:
             raise ImportError('You want to use `torchvision` which is not installed yet,'  # pragma: no-cover
-                              ' install it with `pip install torchvision`.')
+                              ' install it with `pip install torchvision`.') from err
 
         if isinstance(self.datamodule, STL10DataModule):
             unlabeled_batch = batch[0]