Use wrapped model for reference completions in WinRateCallback and set default freq to eval_steps in LogCompletionsCallback`
#2074
Conversation
This file contains 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
qgallouedec
reviewed
Sep 16, 2024
trl/trainer/callbacks.py
Outdated
| @@ -253,7 +253,8 @@ def on_train_begin(self, args: TrainingArguments, state: TrainerState, control: | |||
| tokenizer = kwargs["tokenizer"] | |||
| tokenizer.padding_side = "left" | |||
| accelerator = self.trainer.accelerator | |||
| model = getattr(self.trainer, "ref_model", kwargs["model"]) # get the ref model if any, else use the model | |||
| # Use the reference model if available, otherwise use the initial model | |||
| model = getattr(self.trainer, "ref_model", self.trainer.model_wrapped) | |||
There was a problem hiding this comment.
Does it work in the peft case? ie when ref_model is None?
There was a problem hiding this comment.
Oh good catch - let me see if I can write a unit test for this case
There was a problem hiding this comment.
Test added here with some minor cleanup to reuse the same expected winrates across tests: 592a088
There was a problem hiding this comment.
OK to merge @qgallouedec or would you like some final changes?
|
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. |
kashif
approved these changes
Sep 18, 2024
qgallouedec
reviewed
Sep 18, 2024
qgallouedec
approved these changes
Sep 18, 2024
Co-authored-by: Quentin Gallouédec <45557362+qgallouedec@users.noreply.github.com>
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Outdated suggestions cannot be applied.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?
While training some models with
GKDTrainerI ran into an issue with ZeRO-3 andWinRateCallbackwhere the initial model provided by the callback is unwrapped and this incompatible withunwrap_model_for_generation(). I also realised that usingeval_stepsis far more intuitive for bothWinRateCallbackandLogCompletionsCallbackbecause we usually reserve logging for fast computations like the loss, andLogCompletionsCallbackproduces a major bottleneck in the logging if we generate everylogging_steps.This PR fixes both issues
Before submitting
Pull Request section?
to it if that's the case.
documentation guidelines.
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.