GraphVite can be installed from either conda or source. You can also easily install the library on Google Colab for demonstration.

Install from conda

To install GraphVite from conda, you only need one line.

conda install -c milagraph -c conda-forge graphvite cudatoolkit=$(nvcc -V | grep -Po "(?<=V)\d+.\d+")

By default, this will install all dependencies, including PyTorch and matplotlib. If you only need embedding training without evaluation, there is an alternative with minimum dependencies.

conda install -c milagraph -c conda-forge graphvite-mini cudatoolkit=$(nvcc -V | grep -Po "(?<=V)\d+.\d+")

Install from source

First, clone GraphVite from GitHub.

git clone
cd graphvite

Install compilation and runtime dependencies via conda.

conda install -y --file conda/requirements.txt

Compile the code using the following directives. If you have faiss installed from source, you can pass -DFAISS_PATH=/path/to/faiss to cmake.

mkdir build
cd build && cmake .. && make && cd -

Finally, install Python bindings.

cd python && python install && cd -

Install on Colab

First, install Miniconda on Colab.

!wget -c
!chmod +x
!./ -b -p /usr/local -f

Then we install GraphVite and some tools for Jupyter Notebook.

!conda install -y -c milagraph -c conda-forge graphvite \
    python=3.6 cudatoolkit=$(nvcc -V | grep -Po "(?<=V)\d+\.\d+")
!conda install -y wurlitzer ipykernel

Load the installed packages. Now you are ready to go.

import site
%reload_ext wurlitzer