Only call get_output_embeddings when tie_word_embeddings is set #16667
Conversation
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
This avoids an unnecessary call and avoids problems during initialization of class hierarchies.
|
The documentation is not available anymore as the PR was closed or merged. |
sgugger
approved these changes
Apr 8, 2022
There was a problem hiding this comment.
Looks like a reasonable change. Thanks for your PR!
Also pinging @LysandreJik since it touches the core model file.
elusenji
pushed a commit
to elusenji/transformers
that referenced
this pull request
Jun 12, 2022
…ingface#16667) This avoids an unnecessary call and avoids problems during initialization of class hierarchies. Co-authored-by: Samuel Melm <samuel.melm@stud.uni-heidelberg.de>
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Outdated suggestions cannot be applied.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?
This PR just changes the order in which two conditions in
tie_weightsin thePreTrainedModelclass are checked. This would allow us (and probably others) to avoid problems when initializing child classes of models.In our (maybe somewhat specific use case) we encounter a problem when initializing a child class of a model. During the call of the parent constructor (
PreTrainedModel)get_ourput_embeddingsof the child class is called when it is not completely initialized. This also happens whentie_word_embeddings=Falseand the output embeddings are not even needed. Changing the order of the conditions fixes the problem without changing the functionality.Who can review?
Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.
@patrickvonplaten, @thomwolf, @sgugger