laguna-xs-2.1:q4_K_M crashes on second run – GGML_ASSERT(tensor->data != NULL) + STATUS_STACK_BUFFER_OVERRUN

在 Ollama 0.32.4 上,使用 laguna-xs-2.1:q4_K_M 模型连续第二次发起推理请求时, llama-server 崩溃,报错 GGML_ASSERT(tensor->data != NULL) 和 STATUS_STACK_BUFFER_OVERRUN 。优先排查 pro



![[Bug or Missing Feature] Memory feature cannot be deleted – clicking delete and refreshing still shows the memory](https://www.chat-gpts.plus/wp-content/uploads/2026/07/15576-c668d8a0-768x403.jpg)
![[Question]: Build docker Image Fail](https://www.chat-gpts.plus/wp-content/uploads/2026/07/8101-4e47a719-768x403.jpg)
![[Bug]: LLM classifier failed - NoneType has no attribute "update"](https://www.chat-gpts.plus/wp-content/uploads/2026/07/34487-12b40810-768x403.jpg)


