Bridging Data Gaps in Diffusion Models with Adversarial Noise-Based Transfer Learning (ICML 2024)
_{Official Pytorch Implementation}

Updates

• [11/2024] Release paper.

Setup

conda create -n ant python=3.9
conda activate ant
# Install torch, torchvision (https://pytorch.org/get-started/locally/)
pip install -e .

Download the dataset and put it in the dataset folder and Download the ckpt and put it in initial_checkpoint folder. Then run the following command.

Run

(1) Train classifier model

python train_discriminator.py --cfg configs/base_cls.yaml --source_data_path [path_to_source_data] --target_data_path [path_to_target_data] --opts model.classifier.initial_checkpoint [path__to_pretrain_classifier_checkpoint]

(2) Train ANT model

python train.py --cfg configs/base.yaml --source_data_path [path_to_source_data] --target_data_path [path_to_target_data] --opts model.classifier.initial_checkpoint [path_to_path_to_finetuned_classifier_checkpoint]  model.ddpm.initial_checkpoint [path_to_path_to_pretrain_ddpm_checkpoint]

BibTeX

If this repo is useful to you, please cite our corresponding technical paper.

@inproceedings{wang2024bridging,
  title={Bridging Data Gaps in Diffusion Models with Adversarial Noise-Based Transfer Learning},
  author={Wang, Xiyu and Lin, Baijiong and Liu, Daochang and Chen, Ying-Cong and Xu, Chang},
  booktitle={Forty-first International Conference on Machine Learning},
  year={2024}
}

Acknowledgement

We would like to express our gratitude for the contributions of several previous works to the development of VGen. This includes, but is not limited to Stable Diffusion, OpenCLIP, guided-diffusion, and DDPM. We are committed to building upon these foundations in a way that respects their original contributions.