MolecularDiffusion.modules.models.ligandiff.gvp

Geometric Vector Perceptron.

Verbatim port of LigandDiff’s src/gvp.py (itself from DiffHopp, https://github.com/jostorge/diffusion-hopping).

Attributes

s_V

Classes

GVP	Geometric vector perceptron (Jing et al.).

Module Contents

class MolecularDiffusion.modules.models.ligandiff.gvp.GVP(in_dims: Tuple[int, int], out_dims: Tuple[int, int], activations=(F.relu, torch.sigmoid), vector_gate: bool = False, eps: float = 0.0001)

Bases: torch.nn.Module

Geometric vector perceptron (Jing et al.).

forward(x: torch.Tensor | s_V) → torch.Tensor | s_V

reset_parameters() → None

W_h

W_m

W_mu

eps = 0.0001

h

m

mu

n

v

vector_gate = False

MolecularDiffusion.modules.models.ligandiff.gvp.s_V