Skip to content

Pull requests: modelscope/ms-swift

New pull request New
80 Open 3,683 Closed
80 Open 3,683 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(npu): handle frozen weights in MindSpeed gradient accumulation fusion
#9790 opened Jul 23, 2026 by addsubmuldiv Collaborator Loading…
1 of 4 tasks
4
fix(ci): refresh wandb in the CUDA test container setup
#9789 opened Jul 22, 2026 by he-yufeng Contributor Loading…
1
1
Fix Gemma 3 mixed-batch token type IDs
#9786 opened Jul 22, 2026 by GoGiants1 Loading…
1 of 4 tasks
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
1
feat: support epoch-based checkpoint intervals
#9782 opened Jul 22, 2026 by tongchen126 Contributor Loading…
2 of 4 tasks
1
Fix TRL optional dependency checks with Transformers 5
#9770 opened Jul 20, 2026 by GodRayyy Loading…
1 of 4 tasks
8
fix(dataset): raise a clear error on an empty dataset entry
#9756 opened Jul 16, 2026 by he-yufeng Contributor Loading…
5
Add optional fast lora path for qwen2/qwen3 on npu
#9754 opened Jul 16, 2026 by burm95 Loading…
1 of 4 tasks
11
fix(dataset): raise a clear error when custom dataset_info is not a list
#9751 opened Jul 15, 2026 by he-yufeng Contributor Loading…
5
fix(security): block SSRF via multimodal media URLs in load_file
#9741 opened Jul 13, 2026 by he-yufeng Contributor Loading…
1
5
[model] support TeleChat3
#9727 opened Jul 10, 2026 by allplepine Loading…
1 of 4 tasks
9
fix: Call destroy_process_group() at the end of the trainer pipelines
#9726 opened Jul 10, 2026 by gusario Contributor Loading…
1 of 4 tasks
5
fix: flush asynchronous JSONL logs on training end
#9710 opened Jul 9, 2026 by Alwin4Zhang Loading…
4
fix(tuners/lora): re-apply lora_dtype after peft _cast_adapter_dtype (#9694)
#9699 opened Jul 7, 2026 by Anai-Guo Contributor Loading…
1
2
Add language-model-only export option
#9670 opened Jun 30, 2026 by prefersung Loading…
2
Fix: MiniCPM-V 4.6 training hangs on text-only samples with DeepSpeed
#9639 opened Jun 24, 2026 by randydl Contributor Loading…
6
[bugfix] guard _repair_ms_bench against an empty messages list
#9608 opened Jun 20, 2026 by he-yufeng Contributor Loading…
1
[bugfix] guard load_audio against bytes input
#9606 opened Jun 20, 2026 by he-yufeng Contributor Loading…
1 of 4 tasks
5
Update load mcore checkpoint
#9595 opened Jun 18, 2026 by Jintao-Huang Collaborator Loading…
2
Update gkd_loss.py teacher_prompt supports two formats:
#9562 opened Jun 15, 2026 by yhy19 Loading…
1 of 4 tasks
4
[bugfix] add optional MEDIA_DECODE_TIMEOUT to prevent silent media-decode hang
#9541 opened Jun 11, 2026 by HaozheZhang6 Contributor Loading…
1 of 4 tasks
1
2
[WIP] support tps
#9513 opened Jun 8, 2026 by Jintao-Huang Collaborator Loading…
6
[WIP] refactor trainer
#9510 opened Jun 8, 2026 by Jintao-Huang Collaborator Loading…
14
fix(megatron): pre-initialize NCCL communicator for MoE expert DP group to prevent lazy-init deadlock
#9486 opened Jun 3, 2026 by zb2313 Loading…
3
Add lightweight TGS and MFU logging for training
#9465 opened Jun 1, 2026 by WendaDeng Loading…
3 tasks
9
ProTip! Updated in the last three days: updated:>2026-07-20.