[Question] How to add a custom auxiliary SFT loss to the GRPO training loop （Chord）?
Maintainers usually reply within 1 day
Nobody has claimed this yet.
Assessment
- Difficulty
- 5/5
- Estimated time
- Over a week
- Newbie friendliness
- 25/100
- Issue type
- Feature
- Clarity
- Needs clarification
- Activity status
- Quiet
- Tech stack
- python
- Domain
- machine-learning
Research direction
No source file, test, or entry point is named in the issue. Start by locating the GRPO loss computation and routing-replay implementation, then determine how an SFT forward pass would interact with both; done would be a maintainer-confirmed design for combining the losses and handling SFT routing.
Written by the indexing model from the issue text.
Description
Your Question
How to add a custom auxiliary SFT loss to the GRPO training loop?
I'm training DeepSeek-V3.2 on 128×H200 with GRPO and would like to mix in an SFT loss term during training, similar to CHORD (https://arxiv.org/abs/2508.11408):
$$\mathcal{L} = (1 - \mu) \cdot \mathcal{L}{\text{GRPO}} + \mu \cdot \mathcal{L}{\text{SFT}}$$
Could you point me to where in the codebase the GRPO loss is computed so I can add the SFT forward pass and combine the two losses?
Also, one question about MoE: routing replay forces the same expert routing from rollout during training, but SFT data has no corresponding rollout. Should I just skip routing replay for the SFT forward pass and use live routing?
Thanks!
What I've Tried
Verified CHORD on smaller models using ms-swift, but ms-swift only supports vLLM for GRPO rollout and lacks routing replay / off-policy sequence masking, so it doesn't work for DeepSeek-V3.2 at this scale.
Environment (if relevant)
- slime version:
- Python version:
- PyTorch version:
- CUDA/ROCm version:
- GPU type and count:
- OS:
Additional Context
No response
Pre-submission Checklist
- I have read the CONTRIBUTING.md and understand the collaboration scope.
- I have read the documentation and FAQ and my question is not answered there.
- I have searched for existing issues and my question has not been asked before.
- Dominant language
- Python
- Stars
- 8.8k
- Forks
- 1.3k
- Avg merge
- 20h 16m
- Merged PRs (30d)
- 24
Getting set up
- No Dockerfile or Docker Compose file
- No pull request template
- Read the contributing guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
More from THUDM/slime
-
[Bug] Math reward accepts distinct large integers after float normalizationPossibly taken @lll-hhh claimed this 2 days ago. Open
Difficulty 2/5 1-3 hours Newbie friendliness 78/100
Maintainers usually reply within 1 day
-
gpqa: when no option letter can be extracted, a single-letter gold label is substring-matched against the whole reply, so a reply that never answers can score 1.0Possibly taken @shaurya416 claimed this 39 days ago. Open
Difficulty 2/5 1-3 hours Newbie friendliness 88/100
Maintainers usually reply within 1 day
-
[Bug] shipped dynamic-sampling filters crash on fan-out groups (list[list[Sample]]) from multi-turn agent rolloutsPossibly taken @QIU-Guanzong claimed this 50 days ago. Open
Difficulty 2/5 1-3 hours Newbie friendliness 84/100
Maintainers usually reply within 1 day
-
[Bug] W&B initialization fails with wandb 0.28.2 because wandb.util.generate_id was removedPossibly taken @cyyecao-lappland claimed this 68 days ago. Openbug
Difficulty 1/5 1-3 hours Newbie friendliness 90/100
THUDM/slime#2332 · 1 comment ·
Maintainers usually reply within 1 day
-
[Bug] solution_str = sample.prompt + sample.responsePossibly taken @mvanhorn claimed this 132 days ago. Openbug
Difficulty 2/5 1-3 hours Newbie friendliness 68/100
THUDM/slime#1829 · 1 comment ·
Maintainers usually reply within 1 day
Similar issues
-
area:scripts area:tests bug priority:later
Difficulty 2/5 1-3 hours Newbie friendliness 72/100
Ikalus1988/MisakaNet#4358 ·
Maintainers usually reply within 1 day
-
Discovered ports from service discovery are never propagated to modules in Module.load()Possibly taken @vishalmore90 claimed this today. Open
Difficulty 2/5 1-3 hours Newbie friendliness 70/100
OWASP/Nettacker#1828 · 1 comment ·
Maintainers usually reply within 1 day
-
Difficulty 1/5 Under an hour Newbie friendliness 88/100
aws-samples/sample-per-model-bedrock#708 ·
Maintainers usually reply within 1 day
-
Difficulty 2/5 1-3 hours Newbie friendliness 78/100
-
Difficulty 2/5 1-3 hours Newbie friendliness 75/100
sjkim1127/Reversecore_MCP#365 ·
Maintainers usually reply within 1 day