MolecularDiffusion.modules.layers.esen.nn.radial¶

This source code is licensed under the MIT license found in the LICENSE file in the root directory of this source tree.

Classes¶

`EnvelopedBesselBasis`
`GaussianSmearing`
`PolynomialEnvelope`	Polynomial envelope function that ensures a smooth cutoff.
`RadialMLP`	Contruct a radial function (linear layers + layer normalization + SiLU) given a list of channels
`SphericalBesselBasis`	1D spherical Bessel basis

Functions¶

gaussian(→ torch.Tensor)

Module Contents¶

class MolecularDiffusion.modules.layers.esen.nn.radial.EnvelopedBesselBasis(num_radial: int, cutoff: float)¶

Bases: torch.nn.Module

forward(d)¶

envelope¶

inv_cutoff¶

rbf¶

class MolecularDiffusion.modules.layers.esen.nn.radial.GaussianSmearing(start: float = -5.0, stop: float = 5.0, num_gaussians: int = 50, basis_width_scalar: float = 1.0)¶

Bases: torch.nn.Module

forward(dist) → torch.Tensor¶

coeff¶

num_output = 50¶

class MolecularDiffusion.modules.layers.esen.nn.radial.PolynomialEnvelope(exponent: int = 5)¶

Bases: torch.nn.Module

Polynomial envelope function that ensures a smooth cutoff.

forward(d_scaled: torch.Tensor) → torch.Tensor¶

a: float¶

b: float¶

c: float¶

p: float¶

class MolecularDiffusion.modules.layers.esen.nn.radial.RadialMLP(channels_list)¶

Bases: torch.nn.Module

Contruct a radial function (linear layers + layer normalization + SiLU) given a list of channels

forward(inputs: torch.Tensor) → torch.Tensor¶

net¶

class MolecularDiffusion.modules.layers.esen.nn.radial.SphericalBesselBasis(num_radial: int, cutoff: float)¶

Bases: torch.nn.Module

1D spherical Bessel basis

Parameters:

num_radial (int) – Controls maximum frequency.
cutoff (float) – Cutoff distance in Angstrom.

forward(d_scaled: torch.Tensor) → torch.Tensor¶

frequencies¶

norm_const¶

MolecularDiffusion.modules.layers.esen.nn.radial.gaussian(x: torch.Tensor, mean, std) → torch.Tensor¶