Update to address the issue of slower inference speed in ComfyUI

Previously, I was using ComfyUI (0.3.68) and converted the workflow to Python using https://github.com/pydn/ComfyUI-to-Python-Extension, and inference worked perfectly.

However, after updating to the latest version [v0.3.75](https://github.com/comfyanonymous/ComfyUI/releases/tag/v0.3.75), the inference speed slowed down. For example, using Z-Image, the ComfyUI workflow only took 10 seconds, but the converted Python workflow took a whopping 5 minutes.