Use raise .. from .. to explicitly chain exceptions

pytorch_lightning/accelerators/ddp2_backend.py

@@ -47,9 +47,9 @@ def _resolve_task_idx(self):
             # torchelastic or general non_slurm ddp2
             try:
                 self.task_idx = int(os.environ['LOCAL_RANK'])
-            except Exception as e:
+            except Exception as exp:
                 m = 'ddp2 only works in SLURM or via torchelastic with the WORLD_SIZE, LOCAL_RANK, GROUP_RANK flags'
-                raise MisconfigurationException(m)
+                raise MisconfigurationException(m) from exp
 
     def train(self):
         model = self.trainer.model

pytorch_lightning/metrics/metric.py

@@ -165,15 +165,15 @@ def aggregate(self, *tensors: torch.Tensor) -> torch.Tensor:
         """
         try:
             return torch.cat(tensors).mean(0)
-        except (ValueError, TypeError):
+        except (ValueError, TypeError) as e:
             if isinstance(tensors[0], Mapping):
                 return {k: torch.stack([tensor[k] for tensor in tensors]).mean(0) for k in tensors[0].keys()}
             elif isinstance(tensors[0], Sequence) and not isinstance(tensors[0], torch.Tensor):
                 return tuple([torch.stack(tmp).mean(0) for tmp in zip(*tensors)])
             elif isinstance(tensors[0], torch.Tensor):
                 return torch.stack(tensors).mean(0)
             else:
-                raise TypeError("unknown metric value format to aggregate")
+                raise TypeError("unknown metric value format to aggregate") from e
 
     @staticmethod
     def compute(self, data: Any, output: Any):

pytorch_lightning/trainer/logging.py

@@ -141,13 +141,13 @@ def process_dict_result(self, output, train=False):
         if train:
             try:
                 loss = output['loss']
-            except Exception:
+            except Exception as exp:
                 if isinstance(output, torch.Tensor):
                     loss = output
                 else:
                     raise RuntimeError(
                         'No `loss` value in the dictionary returned from `model.training_step()`.'
-                    )
+                    ) from exp
 
             # when using dp need to reduce the loss
             if self.use_dp or self.use_ddp2:

pytorch_lightning/utilities/parsing.py

@@ -158,8 +158,8 @@ class AttributeDict(Dict):
     def __getattr__(self, key):
         try:
             return self[key]
-        except KeyError:
-            raise AttributeError(f'Missing attribute "{key}"')
+        except KeyError as exp:
+            raise AttributeError(f'Missing attribute "{key}"') from exp
 
     def __setattr__(self, key, val):
         self[key] = val

tests/base/models.py

@@ -10,9 +10,9 @@
 
 try:
     from test_tube import HyperOptArgumentParser
-except ImportError:
+except ImportError as exp:
     # TODO: this should be discussed and moved out of this package
-    raise ImportError('Missing test-tube package.')
+    raise ImportError('Missing test-tube package.') from exp
 
 from pytorch_lightning.core.lightning import LightningModule