nn.GatedEquivariantBlock

class nn.GatedEquivariantBlock(*args: Any, **kwargs: Any)

Gated equivariant block as used for the prediction of tensorial properties by PaiNN. Transforms scalar and vector representation using gated nonlinearities.

References:

[1]

Schütt, Unke, Gastegger: Equivariant message passing for the prediction of tensorial properties and molecular spectra. ICML 2021 (to appear)

Parameters:

• n_sin – number of input scalar features

• n_vin – number of input vector features

• n_sout – number of output scalar features

• n_vout – number of output vector features

• n_hidden – number of hidden units

• activation – interal activation function

• sactivation – activation function for scalar outputs