Disentangled Latent Diffusion Model (DLDM)

This repo contains Pytorch implemented model definitions, pre-trained weights and training/sampling code for DLDMs.

Installation guide

1. git clone this repo
2. Install the pkgs and activate envorionment

$ git clone git@github.com:Yoonho-Na/DLDM.git
$ cd DLDM
$ conda env create -f environment.yaml
$ conda activate dldm

Download pre-trained model

We provide pretrained weights.

$ python scripts/pretrained_dldm.py

Sampling

$ python sample.py

Custom dataset

1. put your files (.jpg, .npy, .png, ...) in a folder custom_folder
2. create 2 text files a xx_train.txt and xx_valid.txt that point to the files in your training and test set respectively
   find $(pwd)/custom_folder/train -name "*.npy" > xx_train.txt
find $(pwd)/custom_folder/valid -name "*.npy" > xx_valid.txt

${pwd}/custom_folder/train/
├── class1
│   ├── filename1.npy
│   ├── filename2.npy
│   ├── ...
├── class2
│   ├── filename1.npy
│   ├── filename2.npy
│   ├── ...
├── ...

${pwd}/custom_folder/valid/
├── class1
│   ├── filename1.npy
│   ├── filename2.npy
│   ├── ...
├── class2
│   ├── filename1.npy
│   ├── filename2.npy
│   ├── ...
├── ...

3. adapt configs/custom_DAE.yaml to point to these 2 files
4. run python main.py --base configs/custom_DAE.yaml -t True --gpus 0,1 to train on two GPUs. Use --gpus 0, (with a trailing comma) to train on a single GPU.

Generated dataset

• https://drive.google.com/drive/folders/1o2G57UrSs3v4whlVAVDVAGO-N3iRraEk?usp=sharing