Single controller hackathon smd #149
Merged
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
bowenyang008
reviewed
Sep 2, 2025
bowenyang008
reviewed
Sep 2, 2025
bowenyang008
reviewed
Sep 2, 2025
bowenyang008
reviewed
Sep 2, 2025
bowenyang008
reviewed
Sep 2, 2025
bowenyang008
reviewed
Sep 2, 2025
bowenyang008
reviewed
Sep 2, 2025
bowenyang008
reviewed
Sep 3, 2025
bowenyang008
reviewed
Sep 3, 2025
bowenyang008
approved these changes
Sep 3, 2025
There was a problem hiding this comment.
LGTM with some minor comments, maybe worth another algo expert (@jdchang1 or Jailu) to cross check it?
jdchang1
approved these changes
Sep 3, 2025
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Outdated suggestions cannot be applied.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.
(1) Add the sequence level RL algorithm Squared Mirror Descent (SMD) used by Kimi 2 and kimi 1.5.
Two sample runs with different hyperparameters:
single-controller-hackathon-smd-B73Na5single-controller-hackathon-smd-qBJppKBoth have 84% on math 500 and 78% on math hard.
(2) Added decoupled GRPO (importance weighting using logp problem vllm engine), and the importance weighted SMD. Mlflow runs