The previously released SD fine-tuning model DPO has released a Checkpoint file, so you can try it out.
This model uses LLM-like RlHF-like technology to fine-tune the model to improve its aesthetic performance, and from their tests, the human evaluation is 70% better than the original SDXL and SD70 models.
Tell me how to use it: Download the model from Huggingface and place it in the /models/unet directory of the ComfyUI root directory.
Then open ComfyUI and use the UNETLoader node to load the model, VAE can be SDXL or VAE 1.5.
SDXL-DPO：https://huggingface.co/mhdang/dpo-sdxl-text2image-v1
SD1.5-DPO：https://huggingface.co/mhdang/dpo-sd1.5-text2image-v1
