Issues

• [RFC] LoRA refactor: one general adapter abstraction, no per-model patches

  verl-project/verl-omni#685 · Filter by author knlnguyen1802 opened on Sep 30, 2026
  1

  

• [Bug] Multimodal video payloads cross-contaminated between concurrent requests in multi-replica (2-node) deployments — vLLM-Omni 0.28.0rc1

  bug
  verl-project/verl-omni#660 · Filter by author zyxzyxzyx opened on Sep 23, 2026
  4

• [recipe] Qwen3-Omni thinker separate-async recipe: colocated offload flags are pure overhead on the dedicated trainer pool and silently freeze training at parameter_sync_step>1 (pinned-verl snapshot aliasing)

  Bug

  verl-project/verl-omni#645 · Filter by author zhtmike opened on Sep 21, 2026
  

• choice_reward.compute_score scores empty/malformed responses as correct when ground_truth has no <answer> tag

  verl-project/verl-omni#643 · Filter by author OnePunchMonk opened on Sep 20, 2026
  1

• [Feature] Report per-sample progress inside a diffusion training step

  verl-project/verl-omni#638 · Filter by author ruiling-smartbear opened on Sep 20, 2026

• [RFC] 0-diff train-rollout log-probs for the Qwen3-Omni thinker

  enhancement
  verl-project/verl-omni#637 · Filter by author cr-gao opened on Sep 20, 2026
  5

  

• [RFC] Add GPU and NPU profiling support to the diffusion V1 trainer

  verl-project/verl-omni#632 · Filter by author SelfDLer opened on Sep 18, 2026

• [RFC] MiniMax-H3 production-level training roadmap

  enhancementhelp wanted
  verl-project/verl-omni#631 · Filter by author SamitHuang opened on Sep 18, 2026
  5

  

• [RFC]: MiniCPM-o 4.5 thinker GSPO on the omni separate-async V1 trainer (standalone rollout)

  enhancement
  Task

  verl-project/verl-omni#628 · Filter by author zhtmike opened on Sep 18, 2026
  

• [RFC] Save inference-ready LoRA adapters during diffusion training

  verl-project/verl-omni#626 · Filter by author mengchengTang opened on Sep 17, 2026
  1

  

• [RFC] Support MiniMax-H3 T2VA FlowGRPO on Cambricon MLU

  verl-project/verl-omni#619 · Filter by author SeptPonts opened on Sep 17, 2026
  3

• [Bug] Single-GPU FSDP1 NO_SHARD LoRA weight synchronization fails with CPU offload

  verl-project/verl-omni#618 · Filter by author 0z5a opened on Sep 17, 2026

• [bug] diffusion separate_async: the step-end weight sync waits for a running batch and discards results that are already computed

  bug
  verl-project/verl-omni#610 · Filter by author cr-gao opened on Sep 17, 2026
  4

  

• [discussion] Train/infer consistency: collection + analysis skills

  verl-project/verl-omni#599 · Filter by author mengchengTang opened on Sep 15, 2026

• [RFC] Unified model merge entrypoint for pipeline-specific publishing

  enhancement
  verl-project/verl-omni#596 · Filter by author NancyFyong opened on Sep 14, 2026

• [Bug] Online async rollout aborts at first generation: GlobalRequestLoadBalancer actor worker is killed while importing transformers (before __init__)

  verl-project/verl-omni#587 · Filter by author roooobotx opened on Sep 12, 2026
  3

• [RFC] Support MiniCPM-o 4.5 streaming full-duplex on-policy distillation

  enhancement
  verl-project/verl-omni#566 · Filter by author NancyFyong opened on Sep 8, 2026

• [RFC] Support MiniCPM-o 4.5 simplex on-policy distillation

  enhancement
  verl-project/verl-omni#565 · Filter by author NancyFyong opened on Sep 8, 2026

• minimax h3 t2va diffusionnft reward issue

  verl-project/verl-omni#563 · Filter by author charon-agi opened on Sep 8, 2026
  10

  

• [Bug]: BAGEL FlowGRPO v1 (sync) trainer diverges mid-training where v0 with identical hyperparameters stays stable

  bug
  verl-project/verl-omni#561 · Filter by author zhtmike opened on Sep 8, 2026
  8

  

• [CI] Retriggering ready-for-ci after a PR update can cause GPU smoke port conflicts

  bug
  verl-project/verl-omni#554 · Filter by author Sky-Trigger opened on Sep 7, 2026

• [RFC] Support DiffusionNFT for BOOGU-Image and BOOGU-Image-Edit (T2I & TI2I)

  help wanted
  verl-project/verl-omni#551 · Filter by author SamitHuang opened on Sep 7, 2026
  5

  

• [RFC] Qwen-Image DMD2 adapter and MVP validation

  verl-project/verl-omni#543 · Filter by author NancyFyong opened on Sep 4, 2026

• [RFC] Model-neutral extension points for multi-stage AR training and pointwise audio reward

  enhancementrefactor
  verl-project/verl-omni#538 · Filter by author dongbo910220 opened on Sep 4, 2026

• [RFC] Support LTX-2.5 text-to-audio-video FlowGRPO

  enhancement
  verl-project/verl-omni#537 · Filter by author NancyFyong opened on Sep 4, 2026