Why the result of IDM-VTON on VITONHD is so bad?

I have no idea of the inference result by IDM-VTON on VITONHD which is so bad. Like this.  I run the inferecnce code following the official setting `CUDA_VISIBLE_DEVICES=4 accelerate launch inference.py --pretrained_model_name_or_path "./models_IDM_VTON" --width 768 --height 1024 --num_inference_steps 30 --output_dir "/data/lizhiying/IDM-VTON/output/VITON-HD_paired" --data_dir "/data/lizhiying/CatVTON/data/VITON-HD" --seed 42 --test_batch_size 2 --guidance_scale 2.0`. Could you please help me to solve this problem? Thanks a lot!!!