Docs: treat KTO as a stable trainer#6345
Merged
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. |
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
KTOTrainerwas promoted to the stable API (#6175), but the docs and docstrings still treated it as experimental. This aligns KTO's visibility and references withDPOTrainer.Changes
README.md,docs/source/index.md): add a "KTO is now stable" highlight.docs/source/index.md): drop the stale 🧪 marker fromKTOTrainerand move it next toDPOTrainer.KTOTrainerto the Highlights trainer list, aKTOTrainercode example, and atrl ktoCLI example.experimental.kto.KTOTrainer/experimental.kto.KTOConfigrefs withKTOTrainer/KTOConfigacross docs (kto_trainer.md,example_overview.md,paper_index.md,dataset_formats.md) and source docstrings (kto_config.py,kto_trainer.py), including the collator paths (~trainer.kto_trainer.DataCollator...).No code behavior changes; the
trl.experimental.ktodeprecation shim is untouched.Note
Low Risk
Documentation and docstring-only updates; training behavior and the
trl.experimental.ktoshim are unchanged.Overview
Documentation and docstrings now present KTO as a stable trainer alongside DPO, matching the API promotion in #6175 (no runtime changes).
What's New in
README.mdanddocs/source/index.mdadds a KTO is now stable highlight; the docs taxonomy movesKTOTrainerout of experimental (drops 🧪) and lists it with offline methods next toDPOTrainer. The README gains KTO parity with DPO: Highlights trainer list, aKTOTrainerquick-start, and atrl ktoCLI example.Across docs (
kto_trainer.md,dataset_formats.md,example_overview.md,paper_index.md) and source (kto_config.py,kto_trainer.py), references switch fromexperimental.kto.KTOTrainer/KTOConfigto stableKTOTrainer/KTOConfig, including autodoc targets and collator paths under~trainer.kto_trainer.*.dataset_formats.mdlistsKTOTrainerin the main trainer table instead of under experimental.Reviewed by Cursor Bugbot for commit 4577043. Bugbot is set up for automated code reviews on this repo. Configure here.