Views
No views yet
Dec 2025: Experiment checkpoints are released here! 🎉Nov 2025: The paper is now available on arXiv. ☕️1conda env create -f environment.yml
2conda activate CLDgit clone https://github.com/monkek123King/CLD.gitfrom huggingface_hub import snapshot_download
repo_id = "black-forest-labs/FLUX.1-dev"
snapshot_download(repo_id, local_dir=Path_to_pretrained_FLUX_model)from huggingface_hub import snapshot_download
repo_id = "alimama-creative/FLUX.1-dev-Controlnet-Inpainting-Alpha"
snapshot_download(repo_id, local_dir=Path_to_pretrained_FLUX_adapter)ckpt
├── decouple_LoRA
│ ├── adapter
│ │ └── pytorch_lora_weights.safetensors
│ ├── layer_pe.pth
│ └── transformer
│ └── pytorch_lora_weights.safetensors
├── pre_trained_LoRA
│ └── pytorch_lora_weights.safetensors
├── prism_ft_LoRA
│ └── pytorch_lora_weights.safetensors
└── trans_vae
└── 0008000.ptpretrained_model_name_or_path: Path_to_pretrained_FLUX_model
pretrained_adapter_path: Path_to_pretrained_FLUX_adapter
transp_vae_path: "ckpt/trans_vae/0008000.pt"
pretrained_lora_dir: "ckpt/pre_trained_LoRA"
artplus_lora_dir: "ckpt/prism_ft_LoRA"
lora_ckpt: "ckpt/decouple_LoRA/transformer"
layer_ckpt: "ckpt/decouple_LoRA"
adapter_lora_dir: "ckpt/decouple_LoRA/adapter"python -m train.train -c train/train.yamlpython -m infer.infer -c infer/infer.yamlpython -m eval.prepare_gtpython evaluate.py --pred-dir "Path_to_predict_results" --gt-dir "Path_to_gt_samples" --output-dir "Path_to_save_eval_results"1@article{liu2025controllable,
2 title={Controllable Layer Decomposition for Reversible Multi-Layer Image Generation},
3 author={Liu, Zihao and Xu, Zunnan and Shu, Shi and Zhou, Jun and Zhang, Ruicheng and Tang, Zhenchao and Li, Xiu},
4 journal={arXiv preprint arXiv:2511.16249},
5 year={2025}
6}