ValueError: LoRA adapter 1 contains fused 3D MoE weights for ‘language_model.model.layers.0.experts’, but the model uses a 2D MoE LoRA wrapper. Start the engine with enable_mixed_moe_lora_form

在 vLLM 中加载 Gemma4 MoE 模型并提供 PEFT 专家 LoRA 适配器时,若未启用混合 MoE LoRA 格式,引擎会因 LoRA 适配器含 3D 融合 MoE 权重、而模型侧使用 2D MoE LoRA 包装器而报错。优先排查启动时是否启用了 enable_mixed_moe_l




![[Bug]: Wiki REFINE does not recover after transient LLM retries are exhausted](https://www.chat-gpts.plus/wp-content/uploads/2026/11/20929-0b02e3f8-768x403.jpg)
![[Bug]: Admin user list returns only the first 10 users when no page_size is given](https://www.chat-gpts.plus/wp-content/uploads/2026/11/21722-e4a3b681-768x403.jpg)
![[Bug]: GET /api/v1/models/all-models returns null rows or panics when a page straddles the end, and page=1 skips the first page](https://www.chat-gpts.plus/wp-content/uploads/2026/11/21727-74cc2885-768x403.jpg)

![[Bug]: rust vllm-bench result is quite different from vllm bench serve](https://www.chat-gpts.plus/wp-content/uploads/2026/11/59154-ea9f2c53-768x403.jpg)