fix: math reward fn #26
Merged
+2
−1
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
|
Except training, the bug also effects test score, causing ~0.1 test score higher than the acutal one at the intial stage of training. |
hiyouga
pushed a commit
that referenced
this pull request
Oct 4, 2025
malhajar17
pushed a commit
to malhajar17/EasyR1_ex
that referenced
this pull request
Oct 21, 2025
malhajar17
pushed a commit
to malhajar17/EasyR1_ex
that referenced
this pull request
Oct 21, 2025
- Update the CSS - Change the UI: - use tabs - use file upload Refs: ART-95
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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Outdated suggestions cannot be applied.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.
Currently, the reward fn intends to give +0.1 reward to encourage output answer in the format of
\boxed{}.However, in
CustomRewardManagerclass, it usessequences = torch.cat((valid_prompt_ids, valid_response_ids))as the input ofself.compute_score. In this way, varseqencesincludes the system prompt "Please reason step by step, and put your final answer within \boxed{}.".Note that the system prompt already includes
\boxed{}. Thus the reward fn gives +0.1 even the model acutally output without \boxed{}. An example is below:Fix this bug by passing the response without system prompt into
self.compute_score