schnetpack.nn
Basic layers
Fully connected linear layer with activation function. |
Equivariant layers
Cartesian:
Gated equivariant block as used for the prediction of tensorial properties by PaiNN. |
Irreps:
Generates the real spherical harmonics for a batch of vectors. |
|
SO3-equivariant Clebsch-Gordon tensor product. |
|
SO3-equivariant convolution using Clebsch-Gordon tensor product. |
|
SO3-equivariant gated nonlinearity. |
|
SO3-equivariant parametric gated nonlinearity. |
Radial basis
Gaussian radial basis functions. |
|
Gaussian radial basis functions centered at the origin. |
|
Sine for radial basis functions with coulomb decay (0th order bessel). |
Cutoff
Behler-style cosine cutoff module. |
|
Mollifier cutoff module scaled to have a value of 1 at \(r=0\). |
Activations
Compute shifted soft-plus activation function. |
Ops
Sum over values with the same indices. |
Factory functions
Build multiple layer fully connected perceptron neural network. |
|
Build neural network analog to MLP with `GatedEquivariantBlock`s instead of dense layers. |
|