Skip to content

kijai/ComfyUI-KwaiKolorsWrapper

Repository files navigation

ComfyUI wrapper for Kwai-Kolors

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

Update - safetensors

Added alternative way to load the ChatGLM3 model from single safetensors file (the configs are included in this repo already). Including already quantized models:

image

https://huggingface.co/Kijai/ChatGLM3-safetensors/upload/main

goes into:

ComfyUI\models\LLM\checkpoints image

image

Installation:

Clone this repository to 'ComfyUI/custom_nodes` folder.

Install the dependencies in requirements.txt, transformers version 4.38.0 minimum is required:

pip install -r requirements.txt

or if you use portable (run this in ComfyUI_windows_portable -folder):

python_embeded\python.exe -m pip install -r ComfyUI\custom_nodes\ComfyUI-KwaiKolorsWrapper\requirements.txt

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

to ComfyUI/models/diffusers/Kolors

Model folder structure needs to be the following:

PS C:\ComfyUI_windows_portable\ComfyUI\models\diffusers\Kolors> tree /F
│   model_index.json
│
├───scheduler
│       scheduler_config.json
│
├───text_encoder
│       config.json
│       pytorch_model-00001-of-00007.bin
│       pytorch_model-00002-of-00007.bin
│       pytorch_model-00003-of-00007.bin
│       pytorch_model-00004-of-00007.bin
│       pytorch_model-00005-of-00007.bin
│       pytorch_model-00006-of-00007.bin
│       pytorch_model-00007-of-00007.bin
│       pytorch_model.bin.index.json
│       tokenizer.model
│       tokenizer_config.json
│       vocab.txt
│
└───unet
        config.json
        diffusion_pytorch_model.fp16.safetensors

To run this, the text enconder is what takes most of the VRAM, but can be quantized to fit approximately these amounts:

Model Size
fp16 ~13 GB
quant8 ~8 GB
quant4 ~4 GB

After that, the sampling single image at 1024 can be expected to take similar amounts than SDXL. For VAE the base SDXL VAE is used.

image

image

About

Diffusers wrapper to run Kwai-Kolors model

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages