NF4 is significantly faster and more memory-efficient than FP8 due to its use of native bnb.matmul_4bit, which avoids casting and leverages low-bit CUDA tricks. It achieves better numerical precision and dynamic range by storing weights in multiple tensors of varying precisions, unlike FP8's single-tensor approachthe list of all (nf4) models (lora not supported yet)<a target="_blank" rel="ugc" href="https://huggingface.co/silveroxides/flux1-nf4-weights/tree/main">https://huggingface.co/silveroxides/flux1-nf4-weights/tree/main</a>what you have to do justgo to ComfyUI/custom_nodes/then (for full checkpoint)<a target="_blank" rel="ugc" href="https://github.com/comfyanonymous/ComfyUI_bitsandbytes_NF4">https://github.com/comfyanonymous/ComfyUI_bitsandbytes_NF4</a><ul><li><code>git clone https://github.com/comfyanonymous/ComfyUI_bitsandbytes_NF4</code></li></ul>or for unet only <ul><li><code>git clone https://github.com/DenkingOfficial/ComfyUI_UNet_bitsandbytes_NF4.git</code></li></ul>and then run this workflow in comfyui"<a target="_blank" rel="ugc" href="https://civarchive.com/models/644552/flux-ai-stylish-prompternf4-supported?modelVersionId=738925">more advance"</a> <a target="_blank" rel="ugc" href="https://civarchive.com/models/644552/flux-ai-stylish-prompternf4-supported?modelVersionId=738925">workflow</a> "just disable checkpoint and activate NF4 checkpoint loader"for flux resolutionin custom node , or cancel it<a target="_blank" rel="ugc" href="https://github.com/al-swaiti/ComfyUI-OllamaGemini">https://github.com/al-swaiti/ComfyUI-OllamaGemini</a>

Files with SHA256: aa6fc75626c1be19057c64e63702f96db2149d6113ac2b3d9d7bc86a226864cb