Fix resized LM head weights being overwritten by post_init#45079
Merged
ArthurZucker merged 1 commit intohuggingface:mainfrom Apr 2, 2026
Merged
Conversation
When `tie_word_embeddings=False`, `_get_resized_lm_head()` creates a new `nn.Linear` without `_is_hf_initialized`, causing `post_init()` to reinitialize its weights. Set the flag after weight copying is done. Fixes huggingface#35141
javierdejesusda
force-pushed
the
fix/resize-embeddings-hf-initialized
branch
from
7b10fd4 to
e59da7d
Compare
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
github-merge-queue
Bot
removed this pull request from the merge queue due to failed status checks
ArthurZucker
approved these changes
Mar 31, 2026
Collaborator
ArthurZucker
left a comment
ArthurZucker
left a comment
There was a problem hiding this comment.
Good catch! thanks 🤗
github-merge-queue
Bot
removed this pull request from the merge queue due to no response for status checks
marvinzh
pushed a commit
to marvinzh/transformers
that referenced
this pull request
Apr 3, 2026
…ce#45079) When `tie_word_embeddings=False`, `_get_resized_lm_head()` creates a new `nn.Linear` without `_is_hf_initialized`, causing `post_init()` to reinitialize its weights. Set the flag after weight copying is done. Fixes huggingface#35141
SangbumChoi
pushed a commit
to SangbumChoi/transformers
that referenced
this pull request
Apr 4, 2026
…ce#45079) When `tie_word_embeddings=False`, `_get_resized_lm_head()` creates a new `nn.Linear` without `_is_hf_initialized`, causing `post_init()` to reinitialize its weights. Set the flag after weight copying is done. Fixes huggingface#35141
sirzechs66
pushed a commit
to sirzechs66/transformers
that referenced
this pull request
Apr 18, 2026
…ce#45079) When `tie_word_embeddings=False`, `_get_resized_lm_head()` creates a new `nn.Linear` without `_is_hf_initialized`, causing `post_init()` to reinitialize its weights. Set the flag after weight copying is done. Fixes huggingface#35141
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What does this PR do?
Fixes #35141
When
tie_word_embeddings=False, callingresize_token_embeddings()thenpost_init()overwritesthe LM head weights with random values. This happens because
_get_resized_lm_head()returns a newnn.Linearwithout setting_is_hf_initialized, sopost_inittreats it as uninitialized.The fix adds
new_lm_head._is_hf_initialized = Trueat the end of_get_resized_lm_head(), afterall weight copying is done.
_get_resized_embeddings()doesn't need this because it reuses theoriginal module object.
Test added to
ModelTesterMixinintest_modeling_common.py, following the existingtest_resize_embeddings_untiedpattern. Verified on GPT2, LLaMA, and OPT.Supersedes #36221 (stale since Feb 2025, reviewer feedback never addressed).
Who can review?
@Rocketknight1 @ArthurZucker