INT8 Mix Quantization of One Obsession v24 by maxfeifei8, I got really curious about how INT8 quantization would fare with Illustrious. Turns out pretty good.70 layers in INT8 and rest in INT8 Convrot for the diffusion model.v2.0<ul><li>Text Encoders are now quantized with a mix of INT8 and INT8 Convrot, total checkpoint size is now down to 3.92 GB</li><li><a target="_blank" rel="ugc" href="https://github.com/bakapotatolord/ComfyUI-PotatoForge-Encoder-Quant">A patch is required</a> to run the model as ComfyUI does not support quantization methods for the SDXL Clip Loaders. YOU WILL GET GARBAGE OUTPUT WITHOUT IT.</li><li>Same prompt and seed from v1.0 so you can compare the images in terms of quality</li></ul>v1.0<ul><li>Text Encoders are untouched</li><li>Checkpoint runs natively on ComfyUI</li></ul>Clone the below repo into your ComfyUI's custom_nodes folder. No need to add any nodes, it initializes and patches the CLIP Loader at runtime, please let me know if you encounter any issues.<a target="_blank" rel="ugc" href="https://github.com/bakapotatolord/ComfyUI-PotatoForge-Encoder-Quant">https://github.com/bakapotatolord/ComfyUI-PotatoForge-Encoder-Quant</a>Sample images once again just the first stage (T2V), used prompts from the original model.Tested on:<ul><li>NVIDIA GTX 1660 Super 6 GB</li><li>32 GB System RAM</li></ul>Test settings:<ul><li>25 steps</li><li>CFG 4.5</li><li>Euler Ancestral / Normall</li><li>696 × 1048 resolution</li><li>Approximately 2.5 s/it on a GTX 1660 Super, about 45% faster than the usual model</li></ul>

Files with SHA256: 46ee52b0dd16f5dd078a3371c3f405188cc756e50a1b2af29bba0c1e835f0a32