ComfyUI wrapper for Kwai-Kolors

Rudimentary wrapper that runs Kwai-Kolors text2image pipeline using diffusers.

Models (fp16, 16.5GB) are automatically downloaded from https://huggingface.co/Kwai-Kolors/Kolors/tree/main

to ComfyUI/models/diffusers/Kolors

For now the text enconder seems to take ~13-15GB VRAM, while the inference after that only ~9GB.