Search results

Qwen3-next-80B GSPO训练 rollout结果乱码

bugSomething isn't workingSomething isn't working

Status: Open.

#4917 In volcengine/verl;

· yangyuxiang1996 opened on Jan 14, 2026

求助，累计生成卡住了怎么办

Status: Open.

#4913 In volcengine/verl;

· iuaku opened on Jan 13, 2026

bypass_mode 开启失败

bugSomething isn't workingSomething isn't working

Status: Open.

#4912 In volcengine/verl;

· zhangsheng377 opened on Jan 13, 2026

Might be time to update the ascend_sglang_quick_start.rst

Status: Open.

#4911 In volcengine/verl;

· zheliuyu opened on Jan 13, 2026

How to constrain LLM's output to strictly follow Pydantic format during RL training?

Status: Open.

#4908 In volcengine/verl;

· zhangxixi551 opened on Jan 13, 2026

Qwen3-next 80B-A3B GRPO training error, invalid tensor size

bugSomething isn't workingSomething isn't working

Status: Open.

#4907 In volcengine/verl;

· FangxuLiu opened on Jan 13, 2026

Qwen3-VL-8B-Instruct + fsdp: RuntimeError: Expected all tensors to be on the same device, but got index is on cpu, different from other tensors on cuda

Status: Open.

#4906 In volcengine/verl;

· ImmortalSdm opened on Jan 13, 2026

[RFC] Cross-node placement group with IP-Based worker sorting

Status: Open.

#4903 In volcengine/verl;

· jianjunzhong opened on Jan 13, 2026

reward loop example？

Status: Open.

#4899 In volcengine/verl;

· Kedaya66 opened on Jan 13, 2026

FullyAsyncPipeline, Question about why sometime timing_s/param_sync cost enormous time.

Status: Open.

#4896 In volcengine/verl;

· hjko0530 opened on Jan 12, 2026

Fatal Python error: Aborted

bugSomething isn't workingSomething isn't working

Status: Open.

#4895 In volcengine/verl;

· zhangzef opened on Jan 12, 2026

TypeError: cannot pickle 'SSLContext' object

bugSomething isn't workingSomething isn't working

Status: Open.

#4894 In volcengine/verl;

· zhangzef opened on Jan 12, 2026