📢 Improve GRPO trainer error message for invalid num_generations #3199
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
qgallouedec
changed the title
|
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. |
Change-Id: Ia34c4f5b0c55ea62149849612a3b3c18cc291d68
AliBakly
force-pushed
the
fix-grpo-error-message
branch
from
b38b233 to
53672f3
Compare
yxliu-TAMU
pushed a commit
to mincheolseong/ECEN743-GRPO-Project-Proposal
that referenced
this pull request
Apr 20, 2025
…gingface#3199) Co-authored-by: Quentin Gallouédec <[email protected]>
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Outdated suggestions cannot be applied.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.
This PR fixes a confusing error message when users try to use
num_generations=1with the GRPO trainer.Current Message:
ValueError: The global train batch size (1 x 4) must be evenly divisible by the number of generations per prompt (1). Given the current train batch size, the valid values for the number of generations are: [2, 4].This is mathematically incorrect since 1 is a valid divisor of any number, and confused me in the beginning. The real issue is that GRPO requires at least 2 generations to function properly.
Changes
Just raise a ValueError if
num_generations<2with a more informative message.