AnalysisVisual AI
Jun 28, 10:59 AM
ComfyUI plugin QuantFunc accelerates diffusion models 11.7x on 4090
Generates 1024px Qwen-Image in 0.9s on a 4090 using custom low-bit CUDA kernels (NVFP4/INT4), claiming 11.7× speedup over FP8 while maintaining close quality. Supports Z-Image and Flux.2 as well.
·
Jun 28, 10:59 AM