-
Notifications
You must be signed in to change notification settings - Fork 3.9k
Pull requests: verl-project/verl
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
[fully_async] fix: initialize _dump_executor in FullyAsyncTrainer and FullyAsyncRollouter
#6438
opened May 21, 2026 by
nanastassacos
Contributor
Loading…
5 of 6 tasks
[docker] chore: upgrade sglang to 0.5.12
#6435
opened May 21, 2026 by
ETOgaosion
Collaborator
Loading…
8 tasks
[rollout, agent-loop] feat: add reasoning parser to strip think blocks before tool extraction (fixes #6424)
#6434
opened May 21, 2026 by
abinggo
Contributor
Loading…
[megatron,rollout] fix: align MTP loss and rollout metrics
#6432
opened May 21, 2026 by
xhx1022
Collaborator
Loading…
Generic remote backend abstraction for RL Training
#6422
opened May 20, 2026 by
sfc-gh-truwase
Loading…
8 tasks
[BREAKING][worker, ckpt] refactor: rename ActorRolloutRefWorker.update_weights to actor_weights_to_rollout
#6417
opened May 20, 2026 by
wengeezhang
Loading…
3 tasks done
[worker]fix: restore actor engine GPU memory logging
#6416
opened May 20, 2026 by
wlhgtc
Contributor
Loading…
6 of 8 tasks
[ckpt, model] fix: save LoRA train metadata for PPO actor checkpoint
#6409
opened May 20, 2026 by
Yatogaii
Contributor
Loading…
6 of 8 tasks
[megatron] feat: NCCL communicator suspend/resume for Megatron in colocated mode
#6408
opened May 19, 2026 by
xiefan46
Contributor
Loading…
4 of 8 tasks
[rollout, tool, cfg] feat: add per-assistant-turn cap for multi-turn agent loop
#6404
opened May 19, 2026 by
nev8rz
Contributor
Loading…
5 of 7 tasks
[ci] fix: the qwen3 model replaces the qwen25 model
Ascend
#6398
opened May 19, 2026 by
daikang6
Contributor
Loading…
8 tasks
[fully_async] fix rollout failure propagation
#6395
opened May 19, 2026 by
anzhsoft
Contributor
Loading…
[docker] chore: update vllm 0.20.2 image
#6393
opened May 19, 2026 by
ETOgaosion
Collaborator
Loading…
8 tasks
[perf] feat: Add Qwen3.5 MFU flops estimation
#6389
opened May 18, 2026 by
liangxuZhang
Contributor
Loading…
7 of 8 tasks
[megatron] feat: ascend bump into megatron 016
Ascend
#6374
opened May 16, 2026 by
wangshuyang31
Contributor
Loading…
8 tasks
[rollout] feat: enable MooncakeStoreConnector with hard-reset on weight update
#6373
opened May 16, 2026 by
aoshen02
Loading…
5 of 7 tasks
[algo] fix: mask invalid sequence loss values
#6349
opened May 14, 2026 by
haoyang9804
Contributor
Loading…
5 of 7 tasks
[rollout] feat: support suffix speculative decoding in vLLM rollout
#6342
opened May 14, 2026 by
walterchenchn
Contributor
Loading…
8 tasks
[trainer, fsdp, megatron, veomni, torchtitan] feat: add freeze_module_pattern for partial parameter freezing
#6320
opened May 12, 2026 by
wzhgba
Loading…
5 of 6 tasks
Previous Next
ProTip!
Type g p on any issue or pull request to go back to the pull request listing page.