Experimental INT4 GGUF of Krea 2 Turbo offering decent visual output quality, 42% reduction in model size (7GB) and 35-59% performance improvement over INT8 safetensors.IMPORTANT!Please update the ComfyUI GGUF loader nodes (this fork: <a target="_blank" rel="ugc" href="https://registry.comfy.org/publishers/molbal/nodes/comfyui-gguf-reboot">https://registry.comfy.org/publishers/molbal/nodes/comfyui-gguf-reboot</a>) to load this properly. You will need version v28.09.04 or newer and use the Unet loader (GGUF, Dynamic VRAM) node to load it.INT4 GGUF speed on my Windows and Linux machines, compared to INT8 ConvRot safetensors:<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/9ce050bf-211f-4c5a-acac-0c2ca43dbcc5/original=true/9ce050bf-211f-4c5a-acac-0c2ca43dbcc5.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/eb26c916-6b35-4bfc-8842-2490d2ea7a6f/original=true/eb26c916-6b35-4bfc-8842-2490d2ea7a6f.jpeg" />(Numbers, hardware, and testing setup in the <a target="_blank" rel="ugc" href="https://civarchive.com/articles/34840/introducing-int4-ggufs-for-krea2-turbo-and-flux2klein-faster-and-smaller-than-int8">article</a>)More information in the information page here: <a target="_blank" rel="ugc" href="https://molbal.github.io/gguf/ecosystem/quant-formats.html#q4-cr-int4-convrot">https://molbal.github.io/gguf/ecosystem/quant-formats.html#q4-cr-int4-convrot</a>

Files with SHA256: 7945698bcc494c228edd54ef0a5ffc18cc8aec31455eac727d2f120054eab3f9