Skip to content

Pull requests: modelscope/ms-swift

New pull request New
86 Open 3,872 Closed
86 Open 3,872 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

fix(trainer): support selective logits with sequence parallel
#10039 opened Sep 4, 2026 by zjn20030811 Loading…
1
2
Support mixed text and raw-token rollout content in Swift template
#10029 opened Sep 2, 2026 by kerbeans Loading…
refactor(sequence_parallel): centralize SP wiring behind SPStrategy facade
#10028 opened Sep 2, 2026 by werwrewe Loading…
1 of 5 tasks
feat(gkd): add agentic monitoring metrics
#10026 opened Sep 2, 2026 by Casuallkk Loading…
1 of 4 tasks
perf(rlhf): vectorize padding-free log-probability restore
#10021 opened Sep 1, 2026 by Ruihan11 Contributor Loading…
1 of 5 tasks
@hjh0119
perf(npu): vectorize ring attention LSE extractio
#10002 opened Aug 28, 2026 by Ruihan11 Contributor Loading…
1 of 5 tasks
3
docs: document OrcaRouter as an OpenAI-compatible sampling provider
#9995 opened Aug 27, 2026 by nissrin2020ali-ux Loading…
1 of 4 tasks
feat(fsdp2): load model on meta device for non-rank0 ranks (0 CPU RAM per worker)
#9982 opened Aug 24, 2026 by cben484 Contributor Loading…
fix: align channel loss with sequence parallel labels
#9977 opened Aug 24, 2026 by taking-lying-flat Contributor Loading…
1 of 4 tasks
1
@hjh0119
feat(grpo): add M2PO for stale-rollout training
#9965 opened Aug 21, 2026 by primorLee Contributor Loading…
2 of 4 tasks
@hjh0119
Add --dataloader_multiprocessing_context to work around Python 3.14 incompatibility
#9941 opened Aug 18, 2026 by sliedes Contributor Loading…
1 of 4 tasks
6
[Megatron] Preserve RNG state across checkpoint resume
#9935 opened Aug 17, 2026 by taking-lying-flat Contributor Loading…
1
Fix reward model margin broadcasting and alignment
#9927 opened Aug 16, 2026 by taking-lying-flat Contributor Loading…
Fix DPO IPO log-prob normalization
#9925 opened Aug 16, 2026 by taking-lying-flat Contributor Loading…
1 of 4 tasks
Fix BatchSamplerShard tail sampling
#9907 opened Aug 14, 2026 by taking-lying-flat Contributor Loading…
[Train] Reduce padding-free embedding output memory
#9893 opened Aug 11, 2026 by taking-lying-flat Contributor Loading…
1
fix(deploy): handle reasoning-only stream chunks
#9881 opened Aug 11, 2026 by taking-lying-flat Contributor Loading…
1 of 4 tasks
Fix IterablePackingDataset workers to use spawn
#9876 opened Aug 9, 2026 by zupengwang Loading…
1 of 4 tasks
1
2
fix ppo callbacks
#9872 opened Aug 7, 2026 by leuitong Loading…
3 tasks
2
Fix local/unfused Megatron attention and LoRA on Ascend
#9864 opened Aug 6, 2026 by hazelduan Contributor • Draft
[Feature] Add --streaming_shard: split the streaming dataset across data-parallel ranks
#9860 opened Aug 5, 2026 by Rapisurazurite Loading…
1 of 4 tasks
Fix Qwen3.5 multimodal packing kwargs compatibility
#9838 opened Aug 3, 2026 by taking-lying-flat Contributor Loading…
1
Support model initialization without loading pretrained weights
#9821 opened Jul 30, 2026 by LiXinYuECNU Loading…
2
feat(sft):Add optional fused linear ce for qwen2/qwen3 on npu
#9784 opened Jul 22, 2026 by yangguang-zhang Loading…
1 of 4 tasks
2
feat: support epoch-based checkpoint intervals
#9782 opened Jul 22, 2026 by tongchen126 Contributor Loading…
2 of 4 tasks
1
ProTip! Type g p on any issue or pull request to go back to the pull request listing page.