Skip to content

Disabled post_process when calling MCoreBert + added missing add_lm_h… #14548

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
shaltielshmid wants to merge 2 commits into
base: main
Choose a base branch
from
Open

Disabled post_process when calling MCoreBert + added missing add_lm_h… #14548

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
41fabcb
Disabled post_process when calling MCoreBert + added missing add_lm_h…
shaltielshmid Aug 21, 2025
7cda8a1
Apply isort and black reformatting
shaltielshmid Aug 21, 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
8 changes: 6 additions & 2 deletions nemo/collections/llm/bert/model/base.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -207,7 +207,9 @@ class MCoreBertModelWrapperWithPostLNSupport(MCoreBert):
def __init__(
self, bert_type='megatron', add_pooler=True, tokenizer: Optional["TokenizerSpec"] = None, *args, **kwargs
):
super(MCoreBertModelWrapperWithPostLNSupport, self).__init__(*args, **kwargs)
orig_post_process = kwargs.pop('post_process', True)
super(MCoreBertModelWrapperWithPostLNSupport, self).__init__(*args, **kwargs, post_process=False)
self.post_process = orig_post_process
self.add_pooler = add_pooler
self.bert_type = bert_type
self.tokenizer = tokenizer
Expand Down Expand Up @@ -315,7 +317,9 @@ def forward(
if self.share_embeddings_and_output_weights:
output_weight = self.shared_embedding_or_output_weight()

hidden_states_after_lm_head = self.lm_head(hidden_states=hidden_states)
hidden_states_after_lm_head = (
self.lm_head(hidden_states=hidden_states) if self.config.add_lm_head else hidden_states
)
logits, _ = self.output_layer(hidden_states_after_lm_head, weight=output_weight)

binary_logits = None
Expand Down
Loading