INT8 weights get re-quantized as plain tensorwise on LoRA offload (convrot, per-channel params dropped)

用户在 ComfyUI 中加载 Ideogram 4 或其他使用 *_convrot_simple 的 INT8 量化模型,并同时加载一个普通 LoRA。在默认动态显存管理下生成大图(如 2048x2048)时,图像质量明显下降,而使用非 INT8 模型或去掉 LoRA 时无此问题。小分辨率(如 5
![[Model3D] 3D model preview doesn't work when the app is embedded using web components](https://www.chat-gpts.plus/wp-content/uploads/2026/06/7485-5a5ebf69-768x403.jpg)
![[Bug]: websearch_interception silently truncates streaming response on /v1/messages — follow-up call always uses stream=False](https://www.chat-gpts.plus/wp-content/uploads/2026/06/27721-388279ff-768x403.jpg)
![[1.15.0] Accessing a published Web App redirects users to the Dify login page](https://www.chat-gpts.plus/wp-content/uploads/2026/06/38043-2a7ca6b6-768x403.jpg)
![[1.15.0] Under certain interactions, objects underneath a toast become unclickable](https://www.chat-gpts.plus/wp-content/uploads/2026/06/38047-55a7553c-768x403.jpg)
![issue: file['path'] field not set when attaching old files](https://www.chat-gpts.plus/wp-content/uploads/2026/06/26260-6fdb75e4-768x403.jpg)

![[BUG]: Magic Beacon context window expanding endlessly after translation request.](https://www.chat-gpts.plus/wp-content/uploads/2026/06/5918-c710b5fc-768x403.jpg)

![[Question]: Knowledge graph extraction problem](https://www.chat-gpts.plus/wp-content/uploads/2026/06/7119-a01a8c11-768x403.jpg)