Rearrange DPOTrainer #3501
Open
Rearrange DPOTrainer #3501
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
|
Thanks! I like it! Please let us know when it's ready for review! |
|
Thanks for your attention! It's ready for review. cc @qgallouedec |
qgallouedec
reviewed
May 29, 2025
trl/trainer/dpo_trainer.py
Outdated
Comment on lines
220
to
221
| if model is None: | ||
| raise ValueError("No model provided. Please provide a model to train.") |
There was a problem hiding this comment.
You can remove this, and replace
- model: Optional[Union[PreTrainedModel, nn.Module, str]] = None,
+ model:Union[PreTrainedModel, nn.Module, str],
trl/trainer/dpo_trainer.py
Outdated
Comment on lines
285
to
286
| if processing_class is None: | ||
| raise ValueError("processing_class must be specified to tokenize a DPO dataset.") |
There was a problem hiding this comment.
This could be nice to have instead
if processing_class is None:
processing_class = AutoTokenizer.from_pretrained(model_id)
|
Thank you for your suggestions! I've updated the mentioned parts and more to sync the |
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Outdated suggestions cannot be applied.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.
What does this PR do?
Rearranged the code structure in
DPOTrainer. Details include:_create_model_from_path,_prepare_peft_model, and_prepare_gradient_checkpointing, while keeping the original logic unchanged. This follows the implementation inSFTTrainer, which also splits model preparation into various functional parts for clarity.disable_dropout_in_model(model)andmodel.warnings_issued["estimate_tokens"]ahead so that all operations on the model are grouped together. This is safe and would not influence the execution.modelandprocessing_class. Now we can passNoneas the value formodelandprocessing_classinDPOTrainer. This feature is adapted fromSFTTrainer.SFTTrainer.SFTTrainer.Before submitting
Pull Request section?
to it if that's the case.
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.