The environment is rather unusual to set up, where all attempts to create an environment.yaml file have failed to correctly setup the environment. For this reason, please follow the following instructions to setup a working environment manually:

-First load the conda environment from clean.yml, containing the basic parameters. Then run the following commands from within the environment:

conda env create -f clean.yml
conda activate D-SD
git clone https://github.com/CompVis/taming-transformers.git
pip install -e ./taming-transformers
pip install torch-fidelity
pip install einops
pip install tqdm
pip install transformers==4.19.2 kornia==0.6.4 torchmetrics==0.7.0
conda install tensorboard
conda install -c conda-forge charset-normalizer
pip install omegaconf -U
pip install hydra-core --upgrade
pip install diffusers
pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu117
pip3 install Cython
pip install pytorch-lightning==1.7.7
conda install sympy
conda install matplotlib -y
pip install pytorch-fid
pip install wandb
pip install chardet
conda install -c "nvidia/label/cuda-11.7.0" cuda-toolkit
conda env config vars set WANDB_API_KEY= <- WANDB Key

This should provide a working base environment, with some packages possibly required when dealing with FID/IS calculations, etc.
For distill_diffusion, aiming to distill during training of a diffusion model, please also install the packages provided in the requirements.txt file in that directory.
