-
Notifications
You must be signed in to change notification settings - Fork 3.1k
Pull requests: volcengine/verl
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
[ci] feat: add npu workflow，e2e_sft_llm&model&reward_model_vllm
#5039
opened Jan 25, 2026 by
yyyy2000
Loading…
8 tasks
[megatron, training_utils] fix: router replay R3 align router replay data with global layer indices
#5037
opened Jan 24, 2026 by
HollowMan6
Loading…
8 tasks done
[trainer] fix: resolve dataset config in agent loop
#5034
opened Jan 24, 2026 by
yyDing1
Loading…
8 tasks
[fsdp, megatron] Refactor fully-async training to support multiple checkpoint engine backends
#5029
opened Jan 23, 2026 by
Shangwei-Li
•
Draft
8 tasks
[rollout, vllm, sglang] fix: forward max_tokens/max_new_tokens from rollout config to vllm/sglang backends
#5028
opened Jan 23, 2026 by
psyloy
Loading…
1 task
[feat] Atropos integration with GRPO (#1782)
#5026
opened Jan 22, 2026 by
vyomakesh0728
Loading…
4 of 6 tasks
[rollout, perf, cfg] fix: Add global step info and support more profile control params for rollout profiling (sglang backend)
#5025
opened Jan 22, 2026 by
bithighrr
Loading…
5 of 8 tasks
[megatron] fix: megatron async save ckpt fix
#5016
opened Jan 22, 2026 by
Leem-Li
Loading…
8 tasks done
[rollout] feat: support filter for fully_async_policy
#5014
opened Jan 22, 2026 by
sl-1314
Loading…
8 tasks
Bug Report: filter_overlong_prompts Fails for Multimodal Data
#5004
opened Jan 21, 2026 by
bizhongan414
Loading…
[ray] feat: use get_device_name() for automatic device detection in RayWorkerGroup instead of by parameter passing
#5000
opened Jan 21, 2026 by
jianjunzhong
Loading…
4 of 8 tasks
[vllm, sglang] feat: opt for FP8 rollout memory
#4997
opened Jan 21, 2026 by
Agoniii
Loading…
2 of 8 tasks
[WIP][data] feat: TransferQueue - integrate TransferQueue into main codebase
#4987
opened Jan 20, 2026 by
0oshowero0
•
Draft
6 of 8 tasks
[megatron, training_utils] fix: Patch MoEAlltoAllTokenDispatcher.preprocess for router replay
#4986
opened Jan 19, 2026 by
HollowMan6
Loading…
6 of 8 tasks
[reward] fix: support RemoteRewardManager in load_reward_manager when use_reward_loop=True for fix math-verify issue #3407
#4985
opened Jan 19, 2026 by
DtYXs
Loading…
2 of 8 tasks
[model,doc] feat: add NPU GRPO training scripts for Qwen2.5-32B/Qwen3-30B (Megaton/vLLM backends)
#4984
opened Jan 19, 2026 by
psyloy
Loading…
[ci, doc] feat: Update Ascend Dockerfile and docker build workflow to 8.3.RC1 version for VeRL + Sglang
#4983
opened Jan 19, 2026 by
xiazhahe
Loading…
2 of 7 tasks
[megatron] feat: Support MTP training in SFT
#4981
opened Jan 19, 2026 by
arvyanh
Loading…
8 tasks done
[doc, trainer] fix: shoudn't use rollout routing replay data for R2
#4973
opened Jan 18, 2026 by
HollowMan6
Loading…
5 of 8 tasks
[training_utils] refactor: Extend response slicing to handle multi-dimensional model outputs
#4964
opened Jan 17, 2026 by
JacobHelwig
Loading…
Previous Next
ProTip!
Type g i on any issue or pull request to go back to the issue listing page.