Skip to content

Pull requests: verl-project/verl

New pull request New
417 Open 3,071 Closed
417 Open 3,071 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

[trainer,perf] fix: enable profiler for SFT trainer
#5909 opened Apr 8, 2026 by wuxibin89 Loading…
3
[doc] chore: Bug fixes for the qwen3-235b model in 256k scenarios
#5908 opened Apr 8, 2026 by autbuster Loading…
8 tasks
3
[ckpt] feat: add WPI (Weight Propagation Interface) checkpoint engine for zero-copy cross-node weight transfer
#5907 opened Apr 7, 2026 by yangspirit Loading…
8 tasks done
7
[megatron] fix: Adjust the attention mask shape for VLM with Megatron on NPU
#5904 opened Apr 7, 2026 by ZLiao097 • Draft
3 of 8 tasks
5
[chat_template]: handle system message in apply_chat_template fall…
#5903 opened Apr 7, 2026 by khazic Loading…
6 of 8 tasks
4
[bugfix] Fix modeling_qwen2_5_vl missing attribute 'Qwen2RMSNorm'
#5901 opened Apr 7, 2026 by ZhuYajun-AI Loading…
8 tasks
3
[veomni] fix: improve VeOmniEngine and add flash attention kwargs support
#5900 opened Apr 7, 2026 by deerlu Loading…
8 tasks
3
[trainer] fix: return NaN for empty tensors in compute_data_metrics
#5899 opened Apr 7, 2026 by Jackie2049 • Draft
5 tasks done
1
5
[model] feat: support qwen35 mtp sft/rl
#5898 opened Apr 7, 2026 by zpltys Loading…
1 of 8 tasks
7
feat(opd): support SGLang as teacher server for online policy distillation
#5897 opened Apr 7, 2026 by nathon-lee Loading…
19
[tool, perf] feat: add reward timing metrics in agent loop
#5896 opened Apr 7, 2026 by guillemgt Loading…
6 of 7 tasks
2
[megatron] fix: MTP loss deadlock when using context parallelism
#5895 opened Apr 7, 2026 by xhx1022 Loading…
5
[Q2 Roadmap] Remove legacy workers
#5890 opened Apr 7, 2026 by nathon-lee Loading…
4
fix: flatten multi-component position_ids to 1D for nested tensor compatibility
#5886 opened Apr 6, 2026 by yifannnwu Loading…
1 of 2 tasks
4
[doc] feat: Add Qwen3vl-8B NPU Optimization Practice
#5873 opened Apr 3, 2026 by Rhetee • Draft
4 of 8 tasks
4
[megatron] fix: dynamic context parallel batch splitting and loss normalization
#5869 opened Apr 2, 2026 by Kite0011 Loading…
8 tasks
9
[sglang] Adapting the use of _launch_subprocesses to the latest SGLang branch
#5868 opened Apr 2, 2026 by xiazhahe Loading…
8 tasks
2
feature: Enhance Ray subprocess error handling system
#5855 opened Apr 2, 2026 by abeiabeiqq Loading…
7
[megatron] feat: enable Megatron FSDP for SFT training
#5854 opened Apr 1, 2026 by yxs Loading…
4 tasks done
7
[megatron] fix: always patch actor postprocess on unfused path for MTP models
#5845 opened Apr 1, 2026 by AkiRusProd Loading…
5
[doc] feat: add Claude Code skills for add-dataset, add-reward, add-trainer
#5844 opened Apr 1, 2026 by khazic Loading…
3 tasks done
6
[doc, misc] chore: add Claude Code skills and CLAUDE.md for AI-assisted development
#5843 opened Apr 1, 2026 by khazic Loading…
4 tasks done
2
[trainer] feat: add group reward std and gradient SNR metrics to compute_data_metrics
#5842 opened Apr 1, 2026 by KLGR123 Loading…
3 tasks done
3
[rollout] chore: bump up trtllm image version to 1.3.0rc10
#5841 opened Apr 1, 2026 by Superjomn Loading…
6 of 8 tasks
@Superjomn @hchings
10
[perf, fsdp, trainer] feat: Skip training for zero-advantage responses to speed up RL.
#5838 opened Apr 1, 2026 by sheilaliuxl Loading…
7 tasks done
5
ProTip! Adding no:label will show everything without a label.