The expanded size of the tensor (3072) must match the existing size (3264) at non-singleton dimension 1. Target sizes: [3072, 3072]. Tensor sizes: [3072, 3264]
The GGUF quantized model cannot be used in ComfyUI. I encountered the following error.
https://github.com/city96/ComfyUI-GGUF/issues/423
In the Load CLIP GGUF node, I noticed there is a “Longcat image” option, which suggests that city96’s ComfyUI GGUF node may already support the Longcat image model. I’m not sure whether there was an issue during the conversion process on your end.
Could you please help check this? Thank you.
Which model version did you download — the primary one or the one inside the comfy folder?
For native ComfyUI support, you need to download the version from the comfy folder. That one works properly with the UNET GGUF Loader.
Make sure ComfyUI is updated to latest version.
Which model version did you download — the primary one or the one inside the comfy folder?
For native ComfyUI support, you need to download the version from the comfy folder. That one works properly with the UNET GGUF Loader.
Make sure ComfyUI is updated to latest version.
I downloaded the q8_0 quantized model you provided. I’m not sure whether the issue is that City96’s ComfyUI GGUF node is not yet compatible, or if something went wrong during your model conversion. My ComfyUI version is the latest. If convenient, you can use the workflow JSON file I provided in the GitHub issue above to check whether the problem can be reproduced.
https://huggingface.co/vantagewithai/LongCat-Image-GGUF/blob/main/LongCat-Image-Q8_0.gguf
Please use this version of the model instead. The link you shared points to the GGUF version of the original model with the original tensor names, which is meant to be used with their GitHub repo. For ComfyUI, you need the version inside the comfy folder, since it has tensor names adjusted to match ComfyUI requirements.
Use this one:
https://huggingface.co/vantagewithai/LongCat-Image-GGUF/blob/main/comfy/LongCat-Image-Q8_0.gguf
Please use this version of the model instead. The link you shared points to the GGUF version of the original model with the original tensor names, which is meant to be used with their GitHub repo. For ComfyUI, you need the version inside the comfy folder, since it has tensor names adjusted to match ComfyUI requirements.
Use this one:
https://huggingface.co/vantagewithai/LongCat-Image-GGUF/blob/main/comfy/LongCat-Image-Q8_0.gguf
Thank you so much.