Skip to content

Pull requests: verl-project/verl

New pull request New
362 Open 2,835 Closed
362 Open 2,835 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

[fsdp] fix: use non-inplace div for activation-offloaded logits views
#5513 opened Mar 7, 2026 by cavities12 Loading…
5 of 8 tasks
3
[data] fix: use %-style format placeholders in logger.warning()
#5512 opened Mar 7, 2026 by cavities12 Loading…
4 of 8 tasks
[fsdp] fix: pass dp_group to prepare_dynamic_batch to fix potential hang
#5511 opened Mar 6, 2026 by dubin555 Loading…
5 tasks done
3
[ci] chore: add npu one step off megatron ci and fix fsdp ci
#5510 opened Mar 6, 2026 by wucong25 Loading…
8 tasks
2
[rollout] feat: support eagle3 speculative decode in rollout
#5509 opened Mar 6, 2026 by miracle0517 Loading…
8 tasks done
8
[ci] feat: change runner
#5507 opened Mar 6, 2026 by tfhddd • Draft
8 tasks
1
[data] fix: Fix tensor size mismatch in AgentLoopWorker
#5506 opened Mar 6, 2026 by gaotiexinqu Loading…
5 of 8 tasks
3
[WIP][algo] feat: add GDPO (Group reward-Decoupled Normalization Policy Optimization) algorithm
#5503 opened Mar 5, 2026 by Rhetee • Draft
4 of 8 tasks
6
[doc] fix: fix te pip install instructions
#5501 opened Mar 5, 2026 by TKONIY Loading…
8 tasks done
2
[algo,trainer,worker,fsdp,cfg,doc] feat: add SDPO (Self-Distillation Policy Optimization)
#5499 opened Mar 5, 2026 by jonhue • Draft
2 of 8 tasks
4
[Bug] fix vllm-ascend load weight issue with fully async
#5496 opened Mar 5, 2026 by wangshuyang31 Loading…
8 tasks
2
[megatron] fix: fix use_remove_padding when using mtp in sft
#5490 opened Mar 4, 2026 by arvyanh Loading…
8 tasks done
4
[trainer] fix: support nsys when using sft_trainer_ray.py
#5489 opened Mar 4, 2026 by arvyanh Loading…
8 tasks done
2
[fully_async, one_step_off] feat: support auto resume on abort when using fully_async
#5487 opened Mar 4, 2026 by ArronHZG Loading…
8 tasks
4
[ckpt] fix: Fix checkpoint engine backend unset error
#5473 opened Mar 3, 2026 by ZLiao097 Loading…
4 of 8 tasks
5
[trainer] fix: add trainable flag for masking assistant messages
#5472 opened Mar 3, 2026 by heiseon Loading…
1
4
[doc] chore: npu docker support vllm013
#5471 opened Mar 3, 2026 by yyyy2000 Loading…
8 tasks
9
Integration Agentbay Cloud Sanbox
#5470 opened Mar 3, 2026 by piaoyunHW Loading…
3 tasks
4
[single_controller]fix: correct spelling error 'procecss' -> 'process'
#5464 opened Mar 2, 2026 by Fan-Yunfan Loading…
2
[BREAKING][single_controller] refactor: remove Ray leakage from base interface
#5463 opened Mar 2, 2026 by JenniferWang • Draft
4 of 6 tasks
10
[rollout] feat: add inter-node rollout support for TRTLLMReplica
#5444 opened Mar 1, 2026 by Superjomn • Draft
8 tasks
2
[trainer (agent loop)] feat: add support for multi trajectory in one rollout in agent loop
#5443 opened Feb 28, 2026 by huaiyizhao Loading…
8 tasks
5
[misc] feat: support dynamic bsz using group size
#5438 opened Feb 28, 2026 by Kite0011 Loading…
8 tasks
5
[model] feat: add qwen3.5 fsdp grpo training support with npu optimization
#5434 opened Feb 28, 2026 by Kirrito-k423 • Draft
8 tasks
7
[tool, rollout] refactor: cache tool list initialization with MCP timeout
#5426 opened Feb 27, 2026 by denismegerle Loading…
7 of 8 tasks
3
ProTip! Find all pull requests that aren't related to any open issues with -linked:issue.