-
Notifications
You must be signed in to change notification settings - Fork 3.8k
Pull requests: verl-project/verl
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
[ray_trainer] fix: skip extra step when resume already exceeds total_training_steps
#6232
opened May 1, 2026 by
startju
Contributor
Loading…
[trainer] fix: update TorchTitanEngine for latest torchtitan API
#6231
opened May 1, 2026 by
acisseJZhong
Collaborator
Loading…
3 tasks
[worker] feat: QAT with FP8 (w8a8 & w8a16)
#6229
opened Apr 30, 2026 by
HollowMan6
Collaborator
•
Draft
8 tasks done
[reward, trainer] feat: support multi-output trajectories in async reward scoring
#6228
opened Apr 30, 2026 by
guillemgt
Contributor
Loading…
7 of 8 tasks
[trainer] fix: dump all outputs in validation in main_ppo_sync
#6227
opened Apr 30, 2026 by
guillemgt
Contributor
Loading…
7 of 8 tasks
[reward] fix: compute correct rollout world size
#6226
opened Apr 30, 2026 by
guillemgt
Contributor
Loading…
6 of 8 tasks
[ci] refactor: try use new ci system
#6220
opened Apr 30, 2026 by
ETOgaosion
Collaborator
Loading…
8 tasks
[ci, vllm] test: add CI for process_weights_after_loading correctness
#6219
opened Apr 30, 2026 by
Kamleecoder
•
Draft
4 of 8 tasks
[rollout] fix: guard sglang profiling when self.tokenizer_manager is …
#6217
opened Apr 30, 2026 by
LeiDing191
Loading…
[tool] fix: In the memory snapshot collection logic, opening history records is not compatible with NPU
#6216
opened Apr 30, 2026 by
shaanjiangcun
Loading…
2 of 8 tasks
[model] fix: Ulysses SP support for Qwen3.5/Qwen3.5-MoE
#6212
opened Apr 29, 2026 by
nev8rz
Contributor
Loading…
[fix] fix moe models _update_weights error
#6210
opened Apr 29, 2026 by
jamindy
Loading…
4 of 8 tasks
[fix]: fix seq_len pad len, and adapt to new mtp_loss api (for megatron dev brance)
#6206
opened Apr 29, 2026 by
zpltys
Contributor
Loading…
8 tasks
[ci] fix: fix cpu_unit_test and tensordict CI on vllm_omni test, preparing for CI refactor
#6205
opened Apr 29, 2026 by
ETOgaosion
Collaborator
Loading…
8 tasks
[npu] feat: add true_on_policy_npu runtime consistency patch package
#6204
opened Apr 29, 2026 by
ZhangRan-Zora
•
Draft
[trainer] feat: add mindspeedmm backend engine support on NPU.support Qwen3.5-27B、Qwen3.5-35B
Ascend
#6199
opened Apr 29, 2026 by
OneMondy
Loading…
8 tasks
[fsdp, megatron, vllm, trainer, algo, cfg] feat: Nitrobrew on-policy distillation (teacher hidden states communication + constant-memory fused KL)
#6194
opened Apr 28, 2026 by
tea-more
Loading…
5 of 8 tasks
[tool] feat: simpler function-based tool registration
#6189
opened Apr 28, 2026 by
Begunner
Collaborator
Loading…
2 tasks done
[fsdp] fix: FSDP2 CPUOffloadPolicy crashes in get_per_tensor_param during weight sync
#6188
opened Apr 28, 2026 by
xiefan46
Contributor
Loading…
2 of 8 tasks
[veomni] feat: bump veomni to v0.1.8 in 0.7.1 branch and add veomni 30b npu script
#6187
opened Apr 28, 2026 by
wangshuyang31
Contributor
Loading…
8 tasks
Previous Next
ProTip!
Filter pull requests by the default branch with base:main.