freq_dit

## Environment Preparation ```bash conda env create -f environment.yml conda activate freq_dit ``` ## Download Models ```bash hf download Tongyi-MAI/Z-Image --local-dir ./z_image ``` ## Inference on demo_dataset We provide the `inference.py` script to offer a simplest and fastest way for you to run our model. ```bash tar -xvf demo_dataset.tar --strip-components=7 bash scripts/inference.sh ``` ## Train - Before training, you **must** organize the structure of your dataset as consistent with that in `demo_dataset.tar`. - For the purpose of saving gpu memory, we vectorize the text in advance. You can run `scripts/vectorize.sh` to do so. ```bash bash scripts/train.sh ```