Quantization toolkit for ComfyUI with native W4A4, W4A8, and W8A8 conversion, mixed-precision loading, LoRA handling, and low-VRAM runtime controls.
https://github.com/SparknightLLC/ComfyUI-QuantizationToolkit
git clone https://github.com/SparknightLLC/ComfyUI-QuantizationToolkit
comfy node install comfyui-int8-fast-fork