This is HQ Int8 Row ConvRot of Krea 2 Turbo (fast 8 steps and High Quality) model.Made from official BF16 model with SECourses Musubi Trainer Quantization appYou can download and use Musubi Trainer app for both training and quantization from here : <a target="_blank" rel="ugc" href="https://www.patreon.com/SECourses/posts/secourses-musubi-137551634">https://www.patreon.com/SECourses/posts/secourses-musubi-137551634</a>To be able to use this model with very best performance please use our Torch 2.13 CUDA 13 ComfyUI installer with ready presets : <a target="_blank" rel="ugc" href="https://www.patreon.com/SECourses/posts/download-comfyui-installers-and-presets-105023709">https://www.patreon.com/SECourses/posts/download-comfyui-installers-and-presets-105023709</a>I also recommend our SwarmUI installer with ready SwarmUI presets : <a target="_blank" rel="ugc" href="https://www.patreon.com/posts/download-swarmui-installer-and-presets-114517862">https://www.patreon.com/posts/download-swarmui-installer-and-presets-114517862</a>Int8 ConvRot is 96.2% similar to BF16 meanwhile GGUF Q8 is only 90.0% and FP8 Scaled is 82.2% and NVFP4 is 63.7%<ul><li>Moreover, Int8 ConvRot generates the output in 3.05 seconds, making it 1.82× faster than BF16, which takes 5.56 seconds.</li><li>NVFP4 takes 3.8 seconds and is 1.46× faster than BF16, whereas GGUF Q8 takes 6.06 seconds and is approximately 8.3% slower than BF16.</li><li>So Int8 ConvRot generated with our Musubi Trainer app at high quality is almost 100% faster and almost same quality as BF16</li><li>High quality generation takes few hours on RTX 5090</li></ul>With our ComfyUI backend, Int8 Row ConvRot is able to generate faster than FP8 Scaled literally 100% faster on RTX 3000, 4000 and 5000 series GPUsModel quantization is taking around 3-4 hours on RTX 5090 since we do training like quantization with prodigy optimizerCheck model screenshots to see and learn more

Files with SHA256: ed963f042b2b0d87b5c0ad9e34a20ddcd0b87fd0f587025bc7f69257f2835d2c