Open
Conversation
Implements SDPO from Hübotter et al. (2026), "Reinforcement Learning via Self-Distillation" (arXiv:2601.20802). Signed-off-by: Celine Tan <tancelinetan@gmail.com>
- SDPOLossFn: next_token_logprobs from the training forward already has shape [B, S-1] (next-token convention), so do not apply [:, 1:] again. teacher_logprobs / prev_logprobs come from get_logprobs() [B, S] and still need [:, 1:]. Without this fix training crashes with a tensor size mismatch (e.g. 1022 vs 1023). - SDPOLossFn: add num_valid_samples to the returned metrics dict; DTensorPolicyWorkerV2.train() requires this key. - sdpo_train: replace the non-existent checkpointer.maybe_save_checkpoint / checkpointer.save_checkpoint / timeout.update / timeout.timed_out calls with the CheckpointManager / TimeoutChecker API that actually exists: init_tmp_checkpoint + finalize_checkpoint, mark_iteration + check_save. - sdpo_train: pass step argument to validate() calls. Signed-off-by: Celine Tan <tancelinetan@gmail.com>
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
3 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What does this PR do ?
Add a one line overview of what this PR aims to accomplish.
This PR aims to add an implementation for Self-Distillation Policy Optimization (https://arxiv.org/pdf/2601.20802).
Issues
List issues that this PR closes (syntax):
Usage
# Add a code snippet demonstrating how to use thisBefore your PR is "Ready for review"
Pre checks:
Additional Information