Skip to content

fix vulnerability #2149

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
XuehaoSun merged 4 commits into from
Mar 25, 2025
+3 −76
Merged

fix vulnerability #2149

Changes from 1 commit
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
944ebb1
fix vulnerability
xin3he Mar 24, 2025
608af90
Update load_huggingface.py
xin3he Mar 24, 2025
aa8da34
change a better way to resolve it
xin3he Mar 24, 2025
707b767
Merge branch 'master' into xinhe/1
XuehaoSun Mar 25, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
32 changes: 2 additions & 30 deletions neural_compressor/utils/load_huggingface.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -96,33 +96,8 @@ def from_pretrained(cls, model_name_or_path: str, **kwargs) -> torch.nn.Module:
)
return model
else:
logger.info("the quantization optimized model is loading.")
keys_to_ignore_on_load_unexpected = copy.deepcopy(
getattr(model_class, "_keys_to_ignore_on_load_unexpected", None)
)
keys_to_ignore_on_load_missing = copy.deepcopy(
getattr(model_class, "_keys_to_ignore_on_load_missing", None)
)

# Avoid unnecessary warnings resulting from quantized model initialization
quantized_keys_to_ignore_on_load = [
r"zero_point",
r"scale",
r"packed_params",
r"constant",
r"module",
r"best_configure",
]
if keys_to_ignore_on_load_unexpected is None:
model_class._keys_to_ignore_on_load_unexpected = quantized_keys_to_ignore_on_load
else:
model_class._keys_to_ignore_on_load_unexpected.extend(quantized_keys_to_ignore_on_load)
missing_keys_to_ignore_on_load = [r"weight", r"bias"]
if keys_to_ignore_on_load_missing is None:
model_class._keys_to_ignore_on_load_missing = missing_keys_to_ignore_on_load
else: # pragma: no cover
model_class._keys_to_ignore_on_load_missing.extend(missing_keys_to_ignore_on_load)

# only show logs of error level, since keys_to_ignore_on_load_unexpected is not working without specific model_class
Comment thread
xin3he marked this conversation as resolved.
Outdated
transformers.logging.set_verbosity_error()
if not os.path.isdir(model_name_or_path) and not os.path.isfile(model_name_or_path): # pragma: no cover
from transformers.utils import cached_file

Expand Down Expand Up @@ -169,9 +144,6 @@ def from_pretrained(cls, model_name_or_path: str, **kwargs) -> torch.nn.Module:
**kwargs,
)

model_class._keys_to_ignore_on_load_unexpected = keys_to_ignore_on_load_unexpected
model_class._keys_to_ignore_on_load_missing = keys_to_ignore_on_load_missing

if not os.path.isdir(model_name_or_path) and not os.path.isfile(model_name_or_path): # pragma: no cover
# pylint: disable=E0611
from packaging.version import Version
Expand Down