mat1 and mat2 shapes error when attempting to use use this model.
I've successfully created a Qwen Image Edit workflow and produced several successful edited images using the base 'Qwen2.5-VL-7B-Instruct-Q4_K_M.gguf' model, but when I try this 'Qwen2.5-VL-7B-Abliterated-Caption-it' version I get the following error:
Requested to load QwenImageTEModel_
loaded completely 9438.675122070312 7360.4169921875 True
!!! Exception during processing !!! mat1 and mat2 shapes cannot be multiplied (5376x1280 and 3840x1280)
Any advice would be appreciated. Go easy please, I've only been using ComfyUI a week or so, maybe I missed something obvious.
Thank you for your time reading.
☹️
You tried this quant format with the image edit functionality and got an exception, right? Is your workflow public so I can look into what went wrong? @Frowningface
You tried this quant format with the image edit functionality and got an exception, right? Is your workflow public so I can look into what went wrong? @Frowningface
Thank you for contacting me back! I searched for clues across the web for solutions to no avail. Yes that is the one I'm using, I should have specified better! Anyways, here is my workflow.
The non-abliterated model has no issues:
The Q4 KM quant of your model (quantized by Mrader). The ONLY part I changed was the CLIP Loader:
Looking forward to knowing what I did wrong!
EDIT, FIXED: I messed up the file names and having too many mmprojs in the same folder I guess. Once I fix that file name and ended the model name with period and moved everything out of that folder, it loaded!
Sources in case any other noobs like me get the wrong idea:
I was brought here from CivitAI: https://civitai.com/models/1841581/qwen-image-wf (they said your model was their favorite).
https://github.com/city96/ComfyUI-GGUF/issues/317
https://github.com/comfyanonymous/ComfyUI/issues/9246
https://github.com/comfyanonymous/ComfyUI/issues/9423
You fixed that, that’s great.
Thank you!