Skip to content


A deep learning library focused on graph representation learning for real-world chemical tasks.

  • โœ… State-of-the-art GNN architectures.
  • ๐Ÿ Extensible API: build your own GNN model and train it with ease.
  • โš—๏ธ Rich featurization: powerful and flexible built-in molecular featurization.
  • ๐Ÿง  Pretrained models: for fast and easy inference or transfer learning.
  • โฎ” Read-to-use training loop based on Pytorch Lightning.
  • ๐Ÿ”Œ Have a new dataset? Graphium provides a simple plug-and-play interface. Change the path, the name of the columns to predict, the atomic featurization, and youโ€™re ready to play!


For CPU or GPU

Use mamba:

# Install Graphium
mamba install -c conda-forge graphium

or pip:

pip install graphium


# Install Graphcore's SDK and Graphium dependencies in a new environment called `.graphium_ipu`
./ .graphium_ipu

The above step needs to be done once. After that, enable the SDK and the environment as follows:

source .graphium_ipu

Finally, you will need to install graphium with pip

pip install graphium