README.md

Project

Diffusion-based vector image generator. Aims for a practical application. Click here to see more.

Prerequisite

• Python 3.9
• CUDA Toolkit
• NVIDIA cuDNN

Install modules

If you don't use anaconda, use pip to install the modules. (Not recommended)

Make sure the CUDA Toolkit 12.4 is installed before you proceed!

conda install pytorch torchvision torchaudio pytorch-cuda=12.4 -c pytorch -c nvidia
conda install -c conda-forge accelerate
conda install -c conda-forge diffusers
conda install conda-forge::transformers
pip install xformers --index-url https://download.pytorch.org/whl/cu124
pip install compel sentencepiece protobuf nltk

Update submodules

git submodule update --init --recursive

Refer this document to install and build DiffVG.

Troubleshoot

• If system fails to load libcudnn_cnn_infer.so.8 (WSL only)