fix: replace 35 bare except clauses with except Exception (#7873)

## What
Replace 35 bare `except:` clauses with `except Exception:`.

## Why
Bare `except:` catches `BaseException`, including `KeyboardInterrupt`
and `SystemExit`, which can prevent clean process shutdown and mask
critical errors. Using `except Exception:` catches all application-level
errors while allowing system-level exceptions to propagate correctly.

Co-authored-by: haosenwang1018 <haosenwang1018@users.noreply.github.com>

Author: haosenwang1018

accelerator/cpu_accelerator.py

@@ -232,7 +232,7 @@ def is_fp16_supported(self):
         try:
             if torch.ops.mkldnn._is_mkldnn_fp16_supported():
                 return True
-        except:
+        except Exception:
             return False
 
     def supported_dtypes(self):

csrc/aio/py_test/parse_aio_stats.py

@@ -50,7 +50,7 @@ def extract_value(key, file):
                 return int(v[0]) * 1024 * 1024
             else:
                 return int(key[2:])
-    except:
+    except Exception:
         print(f"{file}: extract_value fails on {key}")
         return None
 

deepspeed/autotuning/autotuner.py

@@ -69,7 +69,7 @@ def __init__(self, args, active_resources):
             try:
                 os.makedirs(self.exps_dir, exist_ok=True)
                 logger.info(f"Created autotuning experiments directory: {self.exps_dir}")
-            except:
+            except Exception:
                 logger.error(
                     f"Failed to create {self.exps_dir}, please check exps_dir in the autotuning config file is accessible by all the nodes in the job."
                 )
@@ -82,7 +82,7 @@ def __init__(self, args, active_resources):
             try:
                 os.makedirs(self.results_dir, exist_ok=True)
                 logger.info(f"Created autotuning results directory: {self.results_dir}")
-            except:
+            except Exception:
                 logger.error(
                     f"Failed to create {self.results_dir}, please check results_dir in the autotuning config file is accessible by all the nodes in the job."
                 )

deepspeed/autotuning/tuner/base_tuner.py

@@ -67,6 +67,6 @@ def tune(self, sample_size=1, n_trials=1000, early_stopping=None):
                     )
                     break
             return i
-        except:
+        except Exception:
             logger.info("Tuner Error:", sys.exc_info()[0])
             return i

deepspeed/compile/input_storage.py

@@ -44,7 +44,7 @@ def _extract_tensor_metadata(self, tensor: torch.Tensor) -> TensorMetadata:
         # Get memory format safely
         try:
             memory_format = tensor.memory_format() if hasattr(tensor, 'memory_format') else torch.contiguous_format
-        except:
+        except Exception:
             memory_format = torch.contiguous_format
 
         # Store real data for tensors if configured to do so

deepspeed/compile/profilers/graph_profile.py

@@ -67,7 +67,7 @@ def _get_mem_usage_out_of_torch():
 
         torch_alloc = get_accelerator().memory_allocated()
         adjust = info.used - torch_alloc
-    except:
+    except Exception:
         # pynvml not available
         pass
 

deepspeed/inference/v2/inference_parameter.py

@@ -47,7 +47,7 @@ def to(self, *args, **kwargs):
                 new_attr = attr.to(*args, **kwargs)
                 setattr(new_tensor, name, new_attr)
                 new_tensor.aux_attrs[name] = new_attr
-        except:
+        except Exception:
             pass
 
         return new_tensor

deepspeed/module_inject/auto_tp.py

@@ -68,7 +68,7 @@ def strided_copy(self,
             if src_shape[outer_dim] == dst_shape[self.out_dim]:
                 try:
                     dst = dst.reshape(-1).data.copy_(src.data.reshape(-1)).reshape(src.shape)
-                except:
+                except Exception:
                     print(dst.shape, src.shape)
                     exit()
                 dst = torch.nn.parameter.Parameter(dst, requires_grad=False)

deepspeed/module_inject/containers/bert.py

@@ -41,7 +41,7 @@ def __init__(self, client_module, inference=False):
                     transformers.models.bert.modeling_bert.BertLayer,
                     transformers.models.roberta.modeling_roberta.RobertaLayer
                 ]
-            except:
+            except Exception:
                 HFBertLayerPolicy._orig_layer_class = None
 
     def get_hidden_heads(self):

deepspeed/module_inject/containers/clip.py

@@ -35,7 +35,7 @@ def __init__(self, client_module, inference=False):
             try:
                 import transformers
                 HFCLIPLayerPolicy._orig_layer_class = transformers.models.clip.modeling_clip.CLIPEncoderLayer
-            except:
+            except Exception:
                 HFCLIPLayerPolicy._orig_layer_class = None
 
     def get_hidden_heads(self):