Pass attn_implementation when using AutoXXX.from_config#30507
Merged
amyeroberts merged 2 commits intohuggingface:mainfrom Apr 29, 2024
Merged
Conversation
|
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. |
younesbelkada
approved these changes
Apr 27, 2024
Contributor
younesbelkada
left a comment
younesbelkada
left a comment
There was a problem hiding this comment.
Makes sense ! Thanks a lot for the investigation and fix ! 🚀
Contributor
|
thank you! |
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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?
Passes
attn_implementation=config._attn_implementationwhen creating a model usingAutoXxx.from_config. This ensures the attention implementation selected is complementary with the parent model.Not added for some models, as the default attn implementation selected for the model isn't compatible with the component model c.f. a57b9b5
If this is not passed, then breaking changes can occur when the attention implementation in the component model being loaded changes -- see #30506 for an example of breaking changes.
For example, when SDPA was added to BERT, this attention mechanism was then selected by default. This however wasn't supported by other models using it as an encoder.