[Bug]: cumem_tag sleep/wake corrupts cross-rank PP broadcast (TP+PP, varied prompts) → cudaErrorIllegalAddress
![[Bug]: cumem_tag sleep/wake corrupts cross-rank PP broadcast (TP+PP, varied prompts) → cudaErrorIllegalAddress](https://www.chat-gpts.plus/wp-content/uploads/2026/06/45519-fa9d4f9d-768x403.jpg)
在 vLLM 中启用 --sleep-mode-backend=cumem_tag ,配置 TP=2, PP=2(4 个 worker)部署,使用 AWQ-INT4 量化模型(如 cyankiwi/Qwen3.6-27B-AWQ-BF16-INT4 ),通过 /sleep 和 /wake_up 接口



![[Bug]: Error in previewing DOCX files in knowledge base search results.](https://www.chat-gpts.plus/wp-content/uploads/2026/06/3264-4ecea5ed-768x403.jpg)
![[Question]: a question about an Error](https://www.chat-gpts.plus/wp-content/uploads/2026/06/13461-7bcdb5b6-768x403.jpg)
![[Bug]: `ContextVar` not propagated on non-`async` tool execution.](https://www.chat-gpts.plus/wp-content/uploads/2026/06/21555-b799ab6c-768x403.jpg)
![[Question]: How to make multi_agents have separate Memoreis while share the same Context?](https://www.chat-gpts.plus/wp-content/uploads/2026/06/21888-297ef423-768x403.jpg)

![[Gemma 4] Gemma4UnifiedForConditionalGeneration text-only inference produces degenerate output (token repetition collapse)](https://www.chat-gpts.plus/wp-content/uploads/2026/06/46531-3f809be1-768x403.jpg)