Repository Issue Activity (beta)

#293 recap sglang-omni version bump, from 0.5.12 to 0.5.16

#253 learn plan skill

#252 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

#113 Multi-turn rollout Update #1

#106 SGLang veRL trouble shooting

#162 search-r1

#211 slime docs for FSDP baseline

#169 RL SGLang Update Weight in Tensor Slow for MOE

#200 Review Feedback of Code-Walk-Through

#201 [RFC] Partial Rollout Improvements in slime

#197 ValueError: Pointer argument (at 0) cannot be accessed from Triton (cpu tensor?)

#194 希望verl参数加一列默认值

#193 Over sample in SGLang

#191 SLIME 使用体验

#192 SGLang rollout error while running multi-turn RL [CUDA error: an illegal memory access was encountered]

#184 Memory Fragmentation Issue of FSDP

#178 Dynamic Sampling support on Verl

#136 Scheduler 接收请求以及处理批次 (Scheduler Receive Requests and Process Batches) 的图片有误

#177 May I ask how model weights being loaded while starting a launch server?

#131 Multi-turn rollout & agentic RL Status & Roadmap

#137 Multi-turn Update #2 VLM Support Tracker

#165 [Typos] Typos and unclear expressions in `rlhf/verl/multi-turn/code-walk-through/readme-4.md`

#160 multi-turn 如何只进行推理 不训练？

#159 Search-R1 GRPO training stability

#103 TRL SGLang Support