The package in kgcnn contains several layer classes to build up graph convolution models in Keras with Tensorflow, PyTorch or Jax as backend. Some models are given as an example in literature. Focus of kgcnn is (batched) graph learning for molecules kgcnn.molecule and materials kgcnn.crystal. Below you can find explanations and information on how to use kgcnn . See Reference under Package Content for code documentation.
intro installation data.ipynb models.ipynb layers.ipynb literature.ipynb molecules.ipynb forces.ipynb
kgcnn
genindexmodindexsearch