Align KTO with DPO: Remove enforcement of causal language models#5701
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. |
qgallouedec
approved these changes
May 5, 2026
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
3 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Align KTO with DPO: Remove enforcement of causal language models.
This PR makes a small change to the
KTOTrainerby removing the check that raised an error when an encoder-decoder model was used. As a result, the restriction that KTO only supports causal language models is no longer enforced in the code.Part of:
@qgallouedec, shouldn't I enforce this in DPO instead?
Note
Medium Risk
Removes a guardrail in
KTOTrainerthat previously blocked encoder-decoder models, which may expose unsupported/untested model architectures to KTO training and lead to runtime errors or incorrect loss computation.Overview
KTOTrainerno longer raises an error when the provided model is configured as encoder-decoder, aligning its initialization behavior with DPO by removing the causal-LM-only enforcement.Reviewed by Cursor Bugbot for commit 577037c. Bugbot is set up for automated code reviews on this repo. Configure here.