Skip to content

Remove custom get_train/eval_dataloader from OnlineDPO#5291

Merged
albertvillanova merged 2 commits into
huggingface:mainfrom
albertvillanova:fu-5276
Mar 16, 2026
Merged

Remove custom get_train/eval_dataloader from OnlineDPO#5291
albertvillanova merged 2 commits into
huggingface:mainfrom
albertvillanova:fu-5276

Conversation

@albertvillanova

@albertvillanova albertvillanova commented Mar 16, 2026 •
edited by cursor Bot
Loading

Copy link
Copy Markdown
Member

Remove custom get_train/eval_dataloader from OnlineDPO,

This PR updates the configuration and trainer logic for Online DPO implementation by adding a new configuration option default (remove_unused_columns=False), and removing custom dataloader methods to simplify the codebase.

Note that transformers.Trainer.get_train_dataloader (and get_eval_dataloader) support skipping the removal of unused columns by setting args.remove_unused_columns=False since v4.54.1 and we support >=4.56.2:

Related to:

Changes

Configuration improvements:

  • Added the remove_unused_columns option to OnlineDPOConfig, defaulting to False instead of the previous default of True, and updated the documentation to reflect this change.

Codebase simplification:

  • Removed custom implementations of get_train_dataloader and get_eval_dataloader from OnlineDPOTrainer, which previously skipped the remove_unused_columns logic, reverting to the standard Trainer behavior.
  • Cleaned up imports in online_dpo_trainer.py by removing unused or redundant imports, including wraps, Trainer, seed_worker, and DataLoader.

Note

Medium Risk
Changes Online DPO’s dataloader behavior by relying on transformers.Trainer with remove_unused_columns=False, which may affect training/eval if downstream datasets previously depended on column stripping or the prior custom loader behavior.

Overview
Online DPO now defaults to keeping all dataset columns by adding remove_unused_columns=False to OnlineDPOConfig (and documenting the overridden default).

OnlineDPOTrainer drops its custom get_train_dataloader/get_eval_dataloader overrides that existed solely to bypass unused-column removal, and cleans up the related now-unused imports—delegating dataloader construction back to transformers.Trainer.

Written by Cursor Bugbot for commit d9bce4c. This will update automatically on new commits. Configure here.

@HuggingFaceDocBuilderDev

HuggingFaceDocBuilderDev commented Mar 16, 2026

Copy link
Copy Markdown

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
qgallouedec approved these changes Mar 16, 2026
View reviewed changes

@qgallouedec qgallouedec left a comment

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks, I think it makes sense. I'm not even sure that online dpo works with remove_unused_columns=True. If we decide to refactor it in the future we should aim to something closer to GRPO where this feature is properly supported

@albertvillanova albertvillanova merged commit d46131f into huggingface:main Mar 16, 2026
3 of 4 checks passed
qgallouedec added a commit that referenced this pull request Mar 18, 2026
@qgallouedec
Squashed commit of the following:
56cff26 
commit 52cd0cc
Author: Albert Villanova del Moral <8515462+albertvillanova@users.noreply.github.com>
Date:   Tue Mar 17 15:31:26 2026 +0100

    Fix UNEXPECTED lm_head.weight warning when loading a CausalLM as a reward model (#5295)

commit 7b42fc4
Author: Albert Villanova del Moral <8515462+albertvillanova@users.noreply.github.com>
Date:   Tue Mar 17 15:29:11 2026 +0100

    Prevent corruption of DPO VLM training if "keep_end" truncation_mode (#5286)

commit 3acb8e8
Author: Albert Villanova del Moral <8515462+albertvillanova@users.noreply.github.com>
Date:   Tue Mar 17 15:27:10 2026 +0100

    Support max_length in DPO VLM training (#5284)

commit ee339a0
Author: Carlos Miguel Patiño <carlos.patino@huggingface.co>
Date:   Tue Mar 17 14:01:44 2026 +0100

    [GKD] Buffer Implementation for Distillation Trainer (#5137)

    Co-authored-by: lewtun <lewis.c.tunstall@gmail.com>

commit d46131f
Author: Albert Villanova del Moral <8515462+albertvillanova@users.noreply.github.com>
Date:   Mon Mar 16 15:27:19 2026 +0100

    Remove custom get_train/eval_dataloader from OnlineDPO (#5291)

commit 85cf8f4
Author: Albert Villanova del Moral <8515462+albertvillanova@users.noreply.github.com>
Date:   Mon Mar 16 15:24:24 2026 +0100

    Remove TrainingArguments import from experimental trainers (#5290)

commit 91e3da0
Author: Quentin Gallouédec <45557362+qgallouedec@users.noreply.github.com>
Date:   Mon Mar 16 07:19:51 2026 -0600

    Fix `accuracy_reward` crash when called from non-main thread (#5281)

commit 4996631
Author: Albert Villanova del Moral <8515462+albertvillanova@users.noreply.github.com>
Date:   Mon Mar 16 07:44:28 2026 +0100

    Fix support for model_init_kwargs in MiniLLM when passed as CLI JSON string (#5274)

commit 5fceaa7
Author: Albert Villanova del Moral <8515462+albertvillanova@users.noreply.github.com>
Date:   Mon Mar 16 07:43:34 2026 +0100

    Simplify structured outputs logic across vLLM versions in scripts/vllm_serve (#5273)

commit 406d406
Author: casinca <47400729+casinca@users.noreply.github.com>
Date:   Sat Mar 14 04:12:49 2026 +0100

    feat(`grpo_trainer.py`): Variational Sequence-Level Soft Policy Optimization (VESPO) (#5199)

commit d0ac7ef
Author: LeonEricsson <70749762+LeonEricsson@users.noreply.github.com>
Date:   Sat Mar 14 02:53:33 2026 +0100

    Allow nullable logprobs in vLLM serve responses  (#5203)

    Co-authored-by: Quentin Gallouédec <45557362+qgallouedec@users.noreply.github.com>
    Co-authored-by: Quentin Gallouédec <gallouedec.quentin@gmail.com>

commit c0eabc4
Author: Quentin Gallouédec <45557362+qgallouedec@users.noreply.github.com>
Date:   Fri Mar 13 18:19:15 2026 -0600

    Change default `vllm_mode` to `"colocate"` and add v0→v1 migration guide (#5255)

    Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

commit 6c0fccd
Author: Mario Šaško <mariosasko777@gmail.com>
Date:   Sat Mar 14 00:19:38 2026 +0100

    35% faster packing + rename `bfd-requeue` to `bfd_split` (#5189)

    Co-authored-by: Quentin Gallouédec <45557362+qgallouedec@users.noreply.github.com>
    Co-authored-by: Quentin Gallouédec <gallouedec.quentin@gmail.com>
qgallouedec added a commit that referenced this pull request Mar 18, 2026
@qgallouedec
Squashed commit of the following:
14dbaf1 
commit 3972d66
Author: Quentin Lhoest <42851186+lhoestq@users.noreply.github.com>
Date:   Wed Mar 18 22:26:44 2026 +0100

    Suggest the `Json()` type for tool calling dataset format (#5307)

    Co-authored-by: Quentin Gallouédec <gallouedec.quentin@gmail.com>
    Co-authored-by: Quentin Gallouédec <45557362+qgallouedec@users.noreply.github.com>

commit 5c6e915
Author: Quentin Gallouédec <45557362+qgallouedec@users.noreply.github.com>
Date:   Wed Mar 18 14:55:19 2026 -0600

    Update `RewardFunc` type annotation to allow `None`values in reward list (#5297)

commit ee96845
Author: Albert Villanova del Moral <8515462+albertvillanova@users.noreply.github.com>
Date:   Wed Mar 18 17:03:54 2026 +0100

    Fix DPOTrainer collators to truncate sequences before padding (#5305)

commit 435c2ae
Author: Quentin Gallouédec <45557362+qgallouedec@users.noreply.github.com>
Date:   Wed Mar 18 08:09:42 2026 -0600

    Add guidance to avoid `hasattr` and `getattr` with defaults in `AGENTS.md` (#5294)

commit 26ce6a3
Author: Quentin Gallouédec <45557362+qgallouedec@users.noreply.github.com>
Date:   Wed Mar 18 00:44:12 2026 -0600

    Apply docstyle (#5296)

commit 52cd0cc
Author: Albert Villanova del Moral <8515462+albertvillanova@users.noreply.github.com>
Date:   Tue Mar 17 15:31:26 2026 +0100

    Fix UNEXPECTED lm_head.weight warning when loading a CausalLM as a reward model (#5295)

commit 7b42fc4
Author: Albert Villanova del Moral <8515462+albertvillanova@users.noreply.github.com>
Date:   Tue Mar 17 15:29:11 2026 +0100

    Prevent corruption of DPO VLM training if "keep_end" truncation_mode (#5286)

commit 3acb8e8
Author: Albert Villanova del Moral <8515462+albertvillanova@users.noreply.github.com>
Date:   Tue Mar 17 15:27:10 2026 +0100

    Support max_length in DPO VLM training (#5284)

commit ee339a0
Author: Carlos Miguel Patiño <carlos.patino@huggingface.co>
Date:   Tue Mar 17 14:01:44 2026 +0100

    [GKD] Buffer Implementation for Distillation Trainer (#5137)

    Co-authored-by: lewtun <lewis.c.tunstall@gmail.com>

commit d46131f
Author: Albert Villanova del Moral <8515462+albertvillanova@users.noreply.github.com>
Date:   Mon Mar 16 15:27:19 2026 +0100

    Remove custom get_train/eval_dataloader from OnlineDPO (#5291)

commit 85cf8f4
Author: Albert Villanova del Moral <8515462+albertvillanova@users.noreply.github.com>
Date:   Mon Mar 16 15:24:24 2026 +0100

    Remove TrainingArguments import from experimental trainers (#5290)

commit 91e3da0
Author: Quentin Gallouédec <45557362+qgallouedec@users.noreply.github.com>
Date:   Mon Mar 16 07:19:51 2026 -0600

    Fix `accuracy_reward` crash when called from non-main thread (#5281)

commit 4996631
Author: Albert Villanova del Moral <8515462+albertvillanova@users.noreply.github.com>
Date:   Mon Mar 16 07:44:28 2026 +0100

    Fix support for model_init_kwargs in MiniLLM when passed as CLI JSON string (#5274)

commit 5fceaa7
Author: Albert Villanova del Moral <8515462+albertvillanova@users.noreply.github.com>
Date:   Mon Mar 16 07:43:34 2026 +0100

    Simplify structured outputs logic across vLLM versions in scripts/vllm_serve (#5273)

commit 406d406
Author: casinca <47400729+casinca@users.noreply.github.com>
Date:   Sat Mar 14 04:12:49 2026 +0100

    feat(`grpo_trainer.py`): Variational Sequence-Level Soft Policy Optimization (VESPO) (#5199)

commit d0ac7ef
Author: LeonEricsson <70749762+LeonEricsson@users.noreply.github.com>
Date:   Sat Mar 14 02:53:33 2026 +0100

    Allow nullable logprobs in vLLM serve responses  (#5203)

    Co-authored-by: Quentin Gallouédec <45557362+qgallouedec@users.noreply.github.com>
    Co-authored-by: Quentin Gallouédec <gallouedec.quentin@gmail.com>

commit c0eabc4
Author: Quentin Gallouédec <45557362+qgallouedec@users.noreply.github.com>
Date:   Fri Mar 13 18:19:15 2026 -0600

    Change default `vllm_mode` to `"colocate"` and add v0→v1 migration guide (#5255)

    Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

commit 6c0fccd
Author: Mario Šaško <mariosasko777@gmail.com>
Date:   Sat Mar 14 00:19:38 2026 +0100

    35% faster packing + rename `bfd-requeue` to `bfd_split` (#5189)

    Co-authored-by: Quentin Gallouédec <45557362+qgallouedec@users.noreply.github.com>
    Co-authored-by: Quentin Gallouédec <gallouedec.quentin@gmail.com>
songhappy pushed a commit to songhappy/trl that referenced this pull request Apr 20, 2026
@albertvillanova
Remove custom get_train/eval_dataloader from OnlineDPO (huggingface#5291
ae7444b 
)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@qgallouedec qgallouedec qgallouedec approved these changes

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@albertvillanova @HuggingFaceDocBuilderDev @qgallouedec