Skip to content

Pull requests: THUDM/slime

New pull request New
88 Open 1,218 Closed
88 Open 1,218 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

support offloading non-updatable server run-ci-changed
#1668 opened Mar 5, 2026 by zhuzilin Loading…
fix: use semantic version comparison for PyTorch >= 2.6 detection
#1667 opened Mar 4, 2026 by abatilo Loading…
1
internv3.5 support
#1660 opened Mar 3, 2026 by samaritan1998 Loading…
fix: normalize rewards per-group when sample counts are unequal
#1655 opened Mar 2, 2026 by dubin555 Loading…
2 of 3 tasks
1
feat: Add knowledge distillation example with offline support
#1654 opened Mar 2, 2026 by tourzhao Loading…
3 tasks
Fix the Rotary Position Embedding (RoPE) parameter passing in the GLM5 mode
#1650 opened Mar 2, 2026 by hanxdmech-ship-it Loading…
Fix missing packed_seq_params in bshd qkv_format
#1649 opened Mar 1, 2026 by coding-famer Loading…
Fix: pass return_tensors in text_kwargs for transformers>=5.0.0 compatibility
#1648 opened Mar 1, 2026 by coding-famer Loading…
[WIP] fix transforrmers api change at 5.2.0 run-ci-megatron
#1647 opened Feb 28, 2026 by UbeCc Loading…
Refactor code safety checks by removing patterns
#1643 opened Feb 28, 2026 by Rohan5commit Loading…
add npu patch for qwen3-vl-8b grpo multi-turn based on tag v0.2.2
#1637 opened Feb 27, 2026 by ascend-slime Loading…
Autofix/issue 1578 hf2megatron arg suffix
#1636 opened Feb 27, 2026 by yitianlian Loading…
feat: add --lazy-multimodal-load to defer image process to rollout time
#1623 opened Feb 25, 2026 by yzlnew Loading…
2
fix(r3,vlm): remove orphaned RoutingReplay from decoder rebuild.
#1620 opened Feb 24, 2026 by yxyOo Loading…
Fix pipeline parallel and multimodal input handling for Kimi-VL
#1598 opened Feb 19, 2026 by Chokoyo • Draft
Add Kimi-VL/K25 Megatron-HF conversion support
#1597 opened Feb 19, 2026 by Chokoyo • Draft
[Feature] Add configurable arguments for rollout manager actor
#1596 opened Feb 18, 2026 by TSunny007 Loading…
1
[Feature] Add curriculum learning example with dynamic multi-task training and online prompt filtering
#1594 opened Feb 18, 2026 by zhangzx-uiuc Loading…
1
[Fix] Minor fix for properly finishing / flushing wandb logging metrics at exit
#1592 opened Feb 17, 2026 by silunw Loading…
1
[Feature] Add modular tracking interface with MLflow backend
#1591 opened Feb 17, 2026 by mouad-hpc Loading…
4 tasks done
2
Add retries to the Remote Reward Model, do not fail on connection drops or endpoint instability
#1582 opened Feb 12, 2026 by joyliu-q Loading…
add rollout_data to rollout_data_postprocess
#1581 opened Feb 12, 2026 by zui-jiang Loading…
[Fix] Add missing space between --save and extra_args in convert_checkpoint
#1580 opened Feb 12, 2026 by kaysonyu Loading…
1 task done
fix: Handle IPv6 addresses in weight sync init_method
#1576 opened Feb 11, 2026 by zx3xyy Loading…
Support qwen3-next MTP Training
#1575 opened Feb 11, 2026 by zx3xyy Loading…
14
ProTip! Updated in the last three days: updated:>2026-03-02.