Learning Timeline
Key Insights
Model Selection Guide by VRAM
Choose a GGUF variant based on your GPU capacity: Use Q2 (lightest) for 8GB VRAM, medium variants for 12GB VRAM, and any variant if you have 16GB VRAM or more.
Quality vs Performance
The Q2 version is the smallest and most VRAM-efficient, but it may produce noise or image quality degradation compared to the larger original models.
Step by Step
Setting Up GGUF Custom Nodes
- Open your ComfyUI interface.
- Click the 'Manager' button on the main menu panel.
- Select the 'Custom Nodes Manager' menu from the list of options.
- In the search bar, type 'gguf'.
- Look for the node named 'ComfyUI-GGUF' created by 'city96'.
- Click the 'Install' button (or 'Update' if you already have it to ensure you have the latest version).
- Click the 'Restart' button on the prompt that appears and press 'OK' to restart ComfyUI.
Configuring the Qwen GGUF Workflow
- Ensure the GGUF model files downloaded from Quonstack have been placed in the 'ComfyUI/models/unet/' folder.
- In your Qwen Image Edit workflow, locate the existing 'Load Diffusion Model' node.
- Delete or disconnect that 'Load Diffusion Model' node.
- Double-click on any empty space in the ComfyUI interface.
- Type and select the 'Unet Loader (GGUF)' node to add it to the canvas.
- On the 'Unet Loader (GGUF)' node, click the 'unet_name' dropdown menu and select the Qwen GGUF model you downloaded (e.g., Q2_K).
- Connect the 'MODEL' output from the 'Unet Loader (GGUF)' node to the model input of the next node in the workflow.
- Click the 'Queue Prompt' button to start generating images using the quantized model.
