Skip to content

fix: restore weights_only=False when reloading best checkpoint #15867

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

Open
nileshpatil6 wants to merge 1 commit into
base: main
Choose a base branch
from
Open

fix: restore weights_only=False when reloading best checkpoint #15867

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
6 changes: 5 additions & 1 deletion nemo/utils/callbacks/nemo_model_checkpoint.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -237,7 +237,11 @@ def on_save_checkpoint(self, trainer, pl_module, checkpoint):

self.previous_best_path = self.best_model_path
old_state_dict = deepcopy(pl_module.state_dict())
checkpoint = torch.load(maybe_injected_best_model_path, map_location='cpu')
# This checkpoint was just written to disk by this same training run (see
# `_save_checkpoint` below), so it is trusted. It also contains the model's
# hparams (e.g. OmegaConf DictConfig), which `weights_only=True` (the default
# since torch 2.6) cannot unpickle, so it must be loaded with weights_only=False.
checkpoint = torch.load(maybe_injected_best_model_path, map_location='cpu', weights_only=False)
if 'state_dict' in checkpoint:
checkpoint = checkpoint['state_dict']
# get a new instanace of the model
Expand Down
Loading