RuntimeError: The size of tensor a (2) must match the size of tensor b (4) at non-singleton dimension 1

在 InvokeAI 打补丁时,如果 LoHa 被作用到 Conv2d 层,而该 LoHa 的因子仍保留卷积形状(例如由 OneTrainer 生成的 [rank, in, kh, kw] 与 [out, rank, 1, 1] 组合),就会触发 RuntimeError: The size of
![[Question]: model verification attempts failed: mineru no such method](https://www.chat-gpts.plus/wp-content/uploads/2026/10/20407-3de6db30-768x403.jpg)


![[Bug]: Streamed usage.cost leaves out the Claude web search fee (server_tool_use dropped)](https://www.chat-gpts.plus/wp-content/uploads/2026/10/46365-d0bf6294-768x403.jpg)
![[Bug]: case-sensitive header lookup in _get_extra_header_tags drops configured spend tags](https://www.chat-gpts.plus/wp-content/uploads/2026/10/46847-b0f35143-768x403.jpg)



