Skip to content

fix: use find_tied_parameters api from HF for tied weight keys #250

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
parthchadha merged 1 commit into from
Apr 23, 2025
Merged

fix: use find_tied_parameters api from HF for tied weight keys #250

Show file tree
Hide file tree
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
4 changes: 2 additions & 2 deletions nemo_reinforcer/models/policy/dtensor_policy_worker.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,7 +25,7 @@
FSDPModule,
)
from transformers import AutoModelForCausalLM, AutoTokenizer
from transformers.modeling_utils import _get_tied_weight_keys
from transformers.integrations.accelerate import find_tied_parameters
from nemo_reinforcer.models.dtensor.parallelize import _parallelize_model

from nemo_reinforcer.algorithms.interfaces import LossFunction
Expand Down Expand Up @@ -256,7 +256,7 @@ def train(
mbs: Optional[int] = None,
) -> Dict[str, Any]:
"""Train the policy on a batch of data with a given loss function."""
num_tied_weights = len(_get_tied_weight_keys(self.model))
num_tied_weights = len(find_tied_parameters(self.model))
skip_tie_check = os.environ.get("NRL_SKIP_TIED_WEIGHT_CHECK")
if (
num_tied_weights != 0
Expand Down
4 changes: 2 additions & 2 deletions nemo_reinforcer/models/policy/fsdp1_policy_worker.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -39,7 +39,7 @@
)

from transformers import AutoModelForCausalLM, AutoTokenizer
from transformers.modeling_utils import _get_tied_weight_keys
from transformers.integrations.accelerate import find_tied_parameters
from nemo_reinforcer.models.policy import PolicyConfig
from nemo_reinforcer.models.policy.utils import import_class_from_path
from nemo_reinforcer.distributed.virtual_cluster import (
Expand Down Expand Up @@ -229,7 +229,7 @@ def train(
) -> Dict[str, Any]:
"""Train the policy on a batch of data with a given loss function."""
# Check if the model has tied weights
num_tied_weights = len(_get_tied_weight_keys(self.model))
num_tied_weights = len(find_tied_parameters(self.model))
skip_tie_check = os.environ.get("NRL_SKIP_TIED_WEIGHT_CHECK")
if num_tied_weights != 0 and not skip_tie_check:
raise ValueError(
Expand Down