Git commit
347710f
Operating System & Version
Ubuntu 24.04
GGML backends
Vulkan
Command-line arguments used
./sd --diffusion-model /home/user/sd.cpp-webui/models/unet/Qwen_Image_Edit-Q8_0.gguf --vae /home/user/sd.cpp-webui/models/vae/Qwen_Image-VAE.safetensors --qwen2vl /home/user/sd.cpp-webui/models/clip/Qwen2.5-VL-7B-Instruct-Q8_0.gguf --cfg-scale 2.5 --sampling-method euler -v --offload-to-cpu --diffusion-fa --flow-shift 3 -r /home/user/qwen_image_edit.png -p "change 'edit.cpp' to 'flux.cpp'" --seed 1118877715456453
Steps to reproduce
I followed the instructions in the Qwen image guide and copied the same command as in the example, with same image and seed.
What you expected to happen
To get the same resulting image as in the example
What actually happened
I get an image with mixed up colours and letters :
Logs / error messages / stack trace
No response
Additional context / environment details
I am using a ryzen APU - GCN 5.0 with Vulkan which works well with all other models - Flux, Flux Kontext and SD
Git commit
347710f
Operating System & Version
Ubuntu 24.04
GGML backends
Vulkan
Command-line arguments used
./sd --diffusion-model /home/user/sd.cpp-webui/models/unet/Qwen_Image_Edit-Q8_0.gguf --vae /home/user/sd.cpp-webui/models/vae/Qwen_Image-VAE.safetensors --qwen2vl /home/user/sd.cpp-webui/models/clip/Qwen2.5-VL-7B-Instruct-Q8_0.gguf --cfg-scale 2.5 --sampling-method euler -v --offload-to-cpu --diffusion-fa --flow-shift 3 -r /home/user/qwen_image_edit.png -p "change 'edit.cpp' to 'flux.cpp'" --seed 1118877715456453
Steps to reproduce
I followed the instructions in the Qwen image guide and copied the same command as in the example, with same image and seed.
What you expected to happen
To get the same resulting image as in the example
What actually happened
I get an image with mixed up colours and letters :
Logs / error messages / stack trace
No response
Additional context / environment details
I am using a ryzen APU - GCN 5.0 with Vulkan which works well with all other models - Flux, Flux Kontext and SD