Issues

• Verl version

  PeterGriffinJin/Search-R1#191 · Filter by author ChaojunNie opened on Sep 28, 2026

• Training collapsion

  PeterGriffinJin/Search-R1#188 · Filter by author lyx0727 opened on Jul 30, 2026

• infer.py: get_query() on full sequence may match <search> block from previous turns

  PeterGriffinJin/Search-R1#187 · Filter by author azure-z77 opened on Jul 29, 2026

• LLDS: A solution to stablize the search R1 training

  PeterGriffinJin/Search-R1#186 · Filter by author vengdeng opened on Jun 4, 2026

• 请问 sm_120框架的 blackwell gpu可以使用吗

  PeterGriffinJin/Search-R1#182 · Filter by author Triumfoo opened on Apr 30, 2026

• 训练稳定性与信用分配：针对长检索序列实现“跳过检索内容”的 GAE 机制

  PeterGriffinJin/Search-R1#175 · Filter by author hyntdlemon1900 opened on Mar 7, 2026
  2

• Support for "Qwen3" architecture and vLLM versions >= 0.11.0

  PeterGriffinJin/Search-R1#173 · Filter by author HardenGale opened on Jan 6, 2026
  1

• Quick question: Has anyone solved “NCCL communicator was aborted on rank 0” error? Need some help here, thanks!

  PeterGriffinJin/Search-R1#170 · Filter by author WenB777 opened on Dec 19, 2025

• Urgent, training encountered an error: batch size must be positive

  PeterGriffinJin/Search-R1#166 · Filter by author WanFan11 opened on Nov 8, 2025
  2

• Unable to Resume Training from Checkpoints – Possible Issue in Checkpoint Loading Logic

  PeterGriffinJin/Search-R1#160 · Filter by author xytsakura opened on Oct 27, 2025
  1

• Question about UID assignment logic in rollout generation (do_search vs. not do_search)

  PeterGriffinJin/Search-R1#159 · Filter by author passionate11 opened on Oct 24, 2025

• model support issues

  PeterGriffinJin/Search-R1#156 · Filter by author AlbertQiSun opened on Oct 17, 2025
  1

• Top-k Parameter Set to 3, but 5 Documents Returned for Some Queries During Training

  PeterGriffinJin/Search-R1#155 · Filter by author seTalent opened on Sep 24, 2025
  4

• Cannot train search-r1 model based on skyrl

  PeterGriffinJin/Search-R1#154 · Filter by author HelloWorldLTY opened on Sep 14, 2025

• Upgrade Search-R1 to support latest vLLM and Verl

  PeterGriffinJin/Search-R1#153 · Filter by author qinhaneee opened on Sep 7, 2025
  5

• The training time

  PeterGriffinJin/Search-R1#152 · Filter by author ghh1125 opened on Aug 21, 2025
  4

• Do you have plans to support the following composite scenarios?

  PeterGriffinJin/Search-R1#148 · Filter by author excelsimon opened on Aug 15, 2025

• Model stuck

  PeterGriffinJin/Search-R1#145 · Filter by author Data-Designer opened on Aug 4, 2025

• Is that possible to use the latest version of verl directly?

  PeterGriffinJin/Search-R1#143 · Filter by author RustDesert opened on Jul 29, 2025
  2

• PPO training fails after 50-100 steps usually before it reaches the trainer.save_freq/trainer.test_freq

  PeterGriffinJin/Search-R1#142 · Filter by author thakur-nandan opened on Jul 25, 2025
  5

• not logging in multi-node training (works in single-node)

  PeterGriffinJin/Search-R1#141 · Filter by author yazdanbakhsh opened on Jul 24, 2025
  2

• Ray-related problem

  PeterGriffinJin/Search-R1#139 · Filter by author Data-Designer opened on Jul 21, 2025
  2

• Request SearchR1-nq_hotpotqa_train-qwen2.5-7b-it-em-ppo-v0.3

  PeterGriffinJin/Search-R1#130 · Filter by author Tree-Shu-Zhao opened on Jul 4, 2025
  1

• verl/third_party/vllm modifications and support for newer vLLM versions

  PeterGriffinJin/Search-R1#129 · Filter by author TigeR0se opened on Jul 3, 2025
  2

• Self-built data output is abnormal

  PeterGriffinJin/Search-R1#122 · Filter by author xjtupy opened on Jun 25, 2025
  9