-
Notifications
You must be signed in to change notification settings - Fork 4.5k
Pull requests: verl-project/verl
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
[misc] feat: uv support deep-ep
#7655
opened Sep 1, 2026 by
ETOgaosion
Collaborator
Loading…
8 tasks
[doc] fix: update Megatron-LM version in install docs to v0.18.0
#7649
opened Sep 1, 2026 by
imitater-dou
Loading…
6 tasks done
[distillation] fix: Disable detokenization for teacher prompt_logprobs
#7648
opened Sep 1, 2026 by
HSYZhang
Loading…
[rollout] feat: continuous_token support Assistant Role in Merge and Encode Context
#7643
opened Sep 1, 2026 by
gxlvera
Contributor
Loading…
[rollout] fix: DeepSeek tool outputs after the first tool turn follow the template
#7636
opened Aug 31, 2026 by
ruiling-smartbear
Contributor
Loading…
6 of 7 tasks
[rollout] feat: tool parser for the DeepSeek-V3 family
#7635
opened Aug 31, 2026 by
ruiling-smartbear
Contributor
Loading…
5 of 7 tasks
[megatron,worker] fix: offload actor parameters after disaggregated weight sync
#7631
opened Aug 31, 2026 by
msz12345
Loading…
8 tasks done
[rollout] FlexKV KV cache offload for dynamic resource
#7621
opened Aug 31, 2026 by
DylanChen-NV
Loading…
8 tasks
[model] fix: use the transformers 5 rotary API in the GLM-4V attention patch
#7620
opened Aug 31, 2026 by
alanhuangyoo
Loading…
6 tasks done
[vllm, rollout] feat: support Ascend prefill-decode disaggregation
Ascend
#7616
opened Aug 30, 2026 by
chifeichi
Loading…
4 of 7 tasks
[reward] fix: DisRM returns reward_extra_info so async scoring does not KeyError
#7608
opened Aug 29, 2026 by
amanyagami
Loading…
6 tasks done
[reward] fix: remove eval() RCE in prime_math grader (CVE-2026-6878)
#7607
opened Aug 29, 2026 by
amanyagami
Loading…
5 of 6 tasks
[megatron, ci] chore: upgrade Ascend stack to Megatron 0.18.0
Ascend
#7604
opened Aug 29, 2026 by
lxb007981
Contributor
Loading…
8 tasks
[ray, docker] fix: pin Ray 2.58.0 (#7599)
#7600
opened Aug 29, 2026 by
jameswong-hai
Loading…
6 of 8 tasks
[cfg, v1] fix: validate authoritative actor FSDP strategy
#7597
opened Aug 28, 2026 by
yuchenwang3
Loading…
6 tasks done
[BREAKING][perf] feat: overlap bucketed weight transfer with receiver-side processing
#7591
opened Aug 28, 2026 by
zjchenn
Contributor
Loading…
6 of 8 tasks
[trainer] feat: add reverse KL top-k distillation loss
#7590
opened Aug 28, 2026 by
YichenZW
Loading…
6 of 7 tasks
[trainer] fix: dump reward_extra_info in v1 train rollout JSONL
#7588
opened Aug 27, 2026 by
jmaicher
Loading…
5 of 8 tasks
[vllm] fix: Use dedicated Ray concurrency group for vLLMHttpServer control methods
#7587
opened Aug 27, 2026 by
rao-ashish
Contributor
Loading…
7 of 8 tasks
[trainer] fix: pass min/max seqlen when rebuilding jagged nested tensors
#7586
opened Aug 27, 2026 by
jmaicher
Loading…
5 of 8 tasks
[engine] fix: honor entropy_from_logits_with_chunking when remove_padding is off
#7582
opened Aug 27, 2026 by
alanhuangyoo
Loading…
[algo] fix: keep single-sample groups out of the rloo_vectorized zero path
#7581
opened Aug 27, 2026 by
alanhuangyoo
Loading…
[rollout, trainer, training_utils, doc] feat: support weighted multi-output agent-loop trajectories
#7580
opened Aug 27, 2026 by
dafu-wu
Contributor
Loading…
2 tasks done
[rollout, vllm] feat: add Laminar session routing, replica port, and KV snapshot hooks
#7570
opened Aug 26, 2026 by
tongyx361
Collaborator
Loading…
5 of 8 tasks
Previous Next
ProTip!
Type g p on any issue or pull request to go back to the pull request listing page.