Skip to content

remove pos emb from state dict for old models #7068

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 11 commits into from
Jul 19, 2023
Merged

remove pos emb from state dict for old models #7068

Changes from all commits
Commits
Show all changes
11 commits
Select commit Hold shift + click to select a range
19a4e35
remove pos emb from state dict
ekmb Jul 19, 2023
a1fcab3
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot] Jul 19, 2023
0d29878
move to nlp_model
ekmb Jul 19, 2023
e2d8388
move to nlp_model
ekmb Jul 19, 2023
1b5884f
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot] Jul 19, 2023
e063bfc
update comment
ekmb Jul 19, 2023
8b7ce58
update comment
ekmb Jul 19, 2023
1178c82
fix nmt test
ekmb Jul 19, 2023
e0374f1
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot] Jul 19, 2023
8f26f98
fix nmt test
ekmb Jul 19, 2023
9d3beb2
Merge branch 'bert_pos_emb_fix' of https://github.com/NVIDIA/NeMo int…
ekmb Jul 19, 2023
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
12 changes: 11 additions & 1 deletion nemo/collections/nlp/models/nlp_model.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,7 +16,7 @@
import hashlib
import json
import os
from typing import Any, Optional
from typing import Any, Mapping, Optional

from omegaconf import DictConfig, OmegaConf
from pytorch_lightning import Trainer
Expand Down Expand Up @@ -385,3 +385,13 @@ def load_from_checkpoint(
finally:
cls._set_model_restore_state(is_being_restored=False)
return checkpoint

def load_state_dict(self, state_dict: Mapping[str, Any], strict: bool = True):
# starting with trasformers v4.31.0, buffer for position_ids is persistent=False
if (
self.bert_model is not None
and "position_ids" not in self.bert_model.embeddings._modules
and "bert_model.embeddings.position_ids" in state_dict
):
del state_dict["bert_model.embeddings.position_ids"]
super(NLPModel, self).load_state_dict(state_dict, strict=strict)