-
Notifications
You must be signed in to change notification settings - Fork 4.2k
Pull requests: verl-project/verl
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
[doc, algo] feat: add Ascend NPU GSPO script for Qwen3-32B in releases/v0.8.0
#6891
opened Jun 29, 2026 by
chengminhua
Contributor
Loading…
8 tasks
[doc, algo] feat: add Ascend NPU GSPO script for Qwen3-32B
#6890
opened Jun 29, 2026 by
chengminhua
Contributor
Loading…
8 tasks
[data] fix: make overlong prompt filter picklable
#6888
opened Jun 29, 2026 by
zhangdw156
Loading…
7 of 8 tasks
[trainer, fully_async] Adapt vLLM 0.19+ for Ascend NPU
Ascend
#6886
opened Jun 29, 2026 by
fh188
Contributor
Loading…
8 tasks
[trainer, fully_async] Adapt vLLM 0.19+ for Ascend NPU
Ascend
#6881
opened Jun 29, 2026 by
fh188
Contributor
Loading…
8 tasks
[fsdp] fix: Fix Qwen3 MoE FSDP weight sync for vLLM rollout in Transformers 5
#6879
opened Jun 29, 2026 by
lxb007981
Loading…
4 of 8 tasks
[ci] fix: Add more test cases in e2e_ppo_trainer_megatron_sglang_ascend.yml
Ascend
#6877
opened Jun 29, 2026 by
xiazhahe
Contributor
Loading…
8 tasks
[trainer, rollout] feat: opt-in rollout-level dispatch for the V1 agent-loop trainer
#6874
opened Jun 28, 2026 by
huaiyizhao
Contributor
Loading…
8 tasks
[fully_async, reward, docs] fix: skip reward function call when use_task_rewards=False
#6870
opened Jun 28, 2026 by
kenkenpa2126
Contributor
Loading…
7 of 8 tasks
[megatron] fix: balanced sequence split in dynamic_cp_split_batch (#6786)
#6869
opened Jun 27, 2026 by
ajinkyajawale14499
Loading…
[trainer, fully_async] feat: add streaming rollouter mode to the V1 PPO trainer
#6868
opened Jun 27, 2026 by
huaiyizhao
Contributor
Loading…
8 tasks
[rollout, trainer, cfg] feat: per-request abort hooks and AbortableLLMServerClient
#6865
opened Jun 27, 2026 by
cr-gao
Loading…
6 of 7 tasks
[vllm] fix: crash in start_profile/stop_profile on non-master nodes when nnodes > 1
#6861
opened Jun 26, 2026 by
kyle-zhangchi
Loading…
6 of 8 tasks
[sglang] fix seed collisions in deterministic GRPO rollouts
#6857
opened Jun 26, 2026 by
tntnnlrw
Loading…
[misc] feat: support fsspec (gs:// , s3://) sources in copy_to_local
#6850
opened Jun 25, 2026 by
dkondoetsy
Loading…
7 tasks done
[data, rollout, worker] feat: add Open-R1 multimodal and TinyLLaVA-Video-R1 preprocessing and training scripts
#6849
opened Jun 25, 2026 by
lihanwen7
Loading…
3 of 4 tasks
[opd, fsdp, megatron] fix: enhance mem footprint for forward_kl_topk OPD
#6848
opened Jun 25, 2026 by
dimjava
Loading…
5 of 8 tasks
[trainer] Fix process_validation_metrics crash on None-filled sparse reward keys
#6845
opened Jun 25, 2026 by
abinggo
Contributor
Loading…
[ci] chore: fix vllm_ascend ci
Ascend
#6839
opened Jun 24, 2026 by
wucong25
Collaborator
Loading…
8 tasks
[megatron] fix: return 3-tuple under calculate_per_token_loss to fix MoE aux/z-loss grad blowup at CP>1
#6836
opened Jun 24, 2026 by
EricMarcus-ai
Contributor
Loading…
6 of 8 tasks
[rollout, trainer, cfg] feat: privileged-context teacher scoring for OPSD
#6833
opened Jun 24, 2026 by
HaozheZhang6
Contributor
•
Draft
[ci] chore: add some NPU's UT/ST
Ascend
#6831
opened Jun 24, 2026 by
daikang6
Contributor
Loading…
8 tasks
Previous Next
ProTip!
What’s not been updated in a month: updated:<2026-05-29.