Skip to content

PTQ-Level Efficiency, QAT-Level Accuracy | 训练后量化的效率,量化感知训练的精度

fix: replace bare except clauses with except Exception (#1981)

Ramakrishnan Sathyavageeswaran81b4db357d2d
diff --git a/auto_round/calib_dataset.py b/auto_round/calib_dataset.py
index 092505898..b95e38524 100644
--- a/auto_round/calib_dataset.py
+++ b/auto_round/calib_dataset.py
@@ -103,7 +103,7 @@ def apply_chat_template_to_samples(samples, tokenizer, seqlen, system_prompt=Non
tokenize=False,
add_generation_prompt=True,
)
- except:
+ except Exception:
logger.warning("Failed to apply chat template. removing the system role in chat history.")
message_modified = [msg for msg in message if msg["role"] != "system"]
chat_templated = tokenizer.apply_chat_template(
@@ -837,7 +837,7 @@ def get_dataset_len(dataset):
try:
dataset_len = len(dataset)
return dataset_len
- except:
+ except Exception:
cnt = 0
for _ in dataset:
cnt += 1
@@ -879,7 +879,7 @@ def select_dataset(dataset, indices):
"""
try:
return dataset.select(indices)
- except:
+ except Exception:
list_data = list(select(dataset, indices))
import pandas as pd
diff --git a/auto_round/inference/backend.py b/auto_round/inference/backend.py
index 1ef8ba9b1..dd6edd6fd 100644
--- a/auto_round/inference/backend.py
+++ b/auto_round/inference/backend.py
@@ -1263,7 +1263,7 @@ def process_requirement(requirements: list, target_device="cuda", logger_level="
for req in requirements:
try:
require_version(req)
- except:
+ except Exception:
missing_requirements.append(req)
gptq_req = next((f'"{req}"' for req in missing_requirements if "gptqmodel" in req), None)
diff --git a/auto_round/inference/convert_model.py b/auto_round/inference/convert_model.py
index b3fc98de4..491137040 100644
--- a/auto_round/inference/convert_model.py
+++ b/auto_round/inference/convert_model.py
@@ -66,7 +66,7 @@ def skip_not_convert_modules(model, quantization_config, layer_names, layer_conf
modules_to_not_convert = getattr(quantization_config, "modules_to_not_convert", [])
try: # transformers new api
modules_to_not_convert = get_modules_to_not_convert(model, modules_to_not_convert, add_default_skips=True)
- except:
+ except Exception:
modules_to_not_convert = _get_modules_to_not_convert(model, modules_to_not_convert)
if modules_to_not_convert:
for layer_name in layer_names:
@@ -156,7 +156,7 @@ try:
from transformers.quantizers.base import HfQuantizer
get_modules_to_not_convert = HfQuantizer.get_modules_to_not_convert
-except:
+except Exception:
get_modules_to_not_convert = _get_modules_to_not_convert
@@ -500,7 +500,7 @@ def _import_exllamav2_kernels():
"""Attempts to import ExLlamaV2 kernels for performance optimization."""
try:
from exllamav2_kernels import gemm_half_q_half, make_q_matrix # pylint: disable=E0611, E0401
- except:
+ except Exception:
logger.warning_once(
"AutoGPTQ ExLlamaV2 has not been installed, Please install it using the following command: "
"`pip install git+https://github.com/AutoGPTQ/AutoGPTQ.git@b8b4127`"
diff --git a/auto_round/modeling/unfused_moe/__init__.py b/auto_round/modeling/unfused_moe/__init__.py
index a112a2a15..15012aea9 100644
--- a/auto_round/modeling/unfused_moe/__init__.py
+++ b/auto_round/modeling/unfused_moe/__init__.py
@@ -178,7 +178,7 @@ def pre_check_config(model_name: str | torch.nn.Module, trust_remote_code: bool
for key in model_keys:
if "gate_up_proj" in key:
return False
- except:
+ except Exception:
return True
return True
diff --git a/auto_round/utils/common.py b/auto_round/utils/common.py
index 8b68351c8..556f3f505 100644
--- a/auto_round/utils/common.py
+++ b/auto_round/utils/common.py
@@ -98,7 +98,7 @@ class LazyImport(object):
try:
self.module = importlib.import_module(self.module_name)
mod = getattr(self.module, name)
- except:
+ except Exception:
spec = importlib.util.find_spec(str(self.module_name + "." + name))
mod = importlib.util.module_from_spec(spec)
spec.loader.exec_module(mod)
diff --git a/auto_round/utils/device.py b/auto_round/utils/device.py
index dcf2bf6d2..884683d08 100644
--- a/auto_round/utils/device.py
+++ b/auto_round/utils/device.py
@@ -312,7 +312,7 @@ def set_cuda_visible_devices(device: str):
indices = [int(device) for device in devices]
try:
pick_device = [current_visible_devices[i] for i in indices]
- except:
+ except Exception:
raise ValueError(
"Invalid '--device' value: It must be smaller than the number of available devices."
" For example, with CUDA_VISIBLE_DEVICES=4,5, "
diff --git a/auto_round/utils/device_manager.py b/auto_round/utils/device_manager.py
index 45b1e4dd7..f2972d733 100644
--- a/auto_round/utils/device_manager.py
+++ b/auto_round/utils/device_manager.py
@@ -396,7 +396,7 @@ class ARDevice:
if callable(fn):
try:
fn() # pylint: disable=E1102 # mps has issues
- except:
+ except Exception:
pass
def device_index(self, index: int):
diff --git a/auto_round/utils/model.py b/auto_round/utils/model.py
index 1239fd81f..df4995cd6 100644
--- a/auto_round/utils/model.py
+++ b/auto_round/utils/model.py
@@ -302,7 +302,7 @@ def _is_mxfp4_model(model_path, trust_remote_code=True):
try: # in case of config loading failure for new models
config = AutoConfig.from_pretrained(model_path, trust_remote_code=trust_remote_code)
- except:
+ except Exception:
return False
model_type = getattr(config, "model_type", "")
@@ -898,7 +898,7 @@ def diffusion_load_model(
from transformers import AutoConfig
config = AutoConfig.from_pretrained(pretrained_model_name_or_path, trust_remote_code=trust_remote_code)
- except:
+ except Exception:
config = None
model_type = getattr(config, "model_type", "")
@@ -1313,8 +1313,10 @@ def is_diffusion_model(model_or_path: Union[str, object], trust_remote_code: boo
# A special case for NextStep
if model_type == "nextstep":
return True
- except:
- pass
+ except Exception:
+ logger.warning(
+ f"Failed to load config for {model_or_path}, trying to check model_index.json for diffusion pipeline."
+ )
index_file = None
if not os.path.isdir(model_or_path):
try:
@@ -1704,7 +1706,7 @@ def _to_model_dtype(model, model_dtype):
model = model.to(torch.bfloat16)
elif model_dtype == "float32" or model_dtype == "fp32" and model.dtype != torch.bfloat32:
model = model.to(torch.float32)
- except:
+ except Exception:
logger.error("please use more device to fit the device or just use one device")
exit()
return model