Skip to content

[SW-223106] change code with robust implementation #2153

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
merged 1 commit into from
Mar 28, 2025
Merged

[SW-223106] change code with robust implementation #2153

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
29 changes: 7 additions & 22 deletions neural_compressor/torch/algorithms/weight_only/save_load.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -873,8 +873,6 @@ def _load_remaining_pretrained_weight(self, model):
resolved_archive_file = self.kwargs.pop("resolved_archive_file", None)
torch_dtype = self.kwargs.pop("torch_dtype", torch.float32)
dtype_orig = self.kwargs.pop("dtype_orig", None)
offload_folder = self.kwargs.pop("offload_folder", None)
offload_state_dict = self.kwargs.pop("offload_state_dict", False)

# restore default dtype
if dtype_orig is not None:
Expand All @@ -884,26 +882,13 @@ def _load_remaining_pretrained_weight(self, model):
resolved_archive_file = [resolved_archive_file]
for shard_file in resolved_archive_file:
state_dict = load_state_dict(shard_file)

params_dict = {
"model": model,
"state_dict": state_dict,
"start_prefix": "",
"expected_keys": self.loaded_state_dict_keys,
"device_map": {"": self.device},
"offload_folder": offload_folder,
"state_dict_folder": tempfile.mkdtemp() if offload_state_dict else None,
"state_dict_index": {} if offload_state_dict else None,
"dtype": torch_dtype,
"keep_in_fp32_modules": [],
}

import transformers

if transformers.__version__ < "4.45.0":
params_dict["loaded_state_dict_keys"] = self.loaded_state_dict_keys

_load_state_dict_into_meta_model(**params_dict)
keys = list(state_dict.keys())
for k in keys:
if k not in self.loaded_state_dict_keys:
state_dict.pop(k)
for k, v in state_dict.items():
state_dict[k] = v.to(torch_dtype)
model.load_state_dict(state_dict, strict=False, assign=True)

# make sure token embedding weights are still tied if needed
model.tie_weights()
Expand Down
Loading