[fix] Fix math reward hanging [WIP] #109
Open
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
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Outdated suggestions cannot be applied.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.
Problem
The training hangs during the math reward computation phase when meeting timeouts.
Cause
The suspected cause is attributed to hanging worker processes within the math reward function. For example, calls to time-consuming operations like sympy.simplify() may enter an indefinite state on certain inputs if not properly killed.
The existing asyncio timeout mechanism captures the top-level task but fails to terminate the orphaned sympy subprocess. This might result in the ProcessPoolExecutor being unable to shut down cleanly, leading to a deadlock. Also, since the reward calculation is a CPU-bound operation, the asyncio event loop might not bring benefits.
Solution
This PR implements
mp_reward_manager.py, a multiprocessing reward manager that replaces the previous asyncio implementation.The implementation is currently undergoing testing to validate the fix.