graph2mat.bindings.e3nn

Interface to e3nn, as well as functions that use irreps.

e3nn uses torch.

Therefore, the interface with e3nn takes as a starting point the interface with torch, defined in graph2mat.bindings.torch.

This interface has two goals:

• Wrap the core functionality. The main addition to the core functionality is two thin wrappers around TorchGraph2Mat and TorchMatrixBlock to handle irreps. There is no need to wrap the data containers for now, as we don’t define data in terms of irreps. In the future, one could envision for example training directly on irreps, and that would require a wrapper around TorchBasisMatrixData.

• Implement equivariant functions that use e3nn irreps and therefore can not be implemented in graph2mat.core without importing e3nn. These functions can be used as the blocks within Graph2Mat to make the model equivariant.

Modules

irreps_tools	Utility tools to deal with e3nn irreps.
modules	E3nn based functions to use in graph2mat.