Construct a layernorm module (See citation for details).

Class that represents a train/validation/test dataset that's readable for PyTorch Note that this class inherits torch.utils.data.Dataset

ScaleNorm

Describes an attributed edge in an undirected graph

Core encoder is a stack of N layers

Encoder is made up of self-attn and feed forward (defined below)

Define standard linear + softmax generation step.

Describes an undirected graph class

Class that represents a train/validation/test datum - self.label: 0 neg, 1 pos -1 missing for different target.

Describes an attributed node in an undirected graph

Implements FFN equation.

The molecule whose reward is the QED.

A namedtuple defines the result of a step for the molecule class. The namedtuple contains the following fields: state: Chem.RWMol. The mole

A residual connection followed by a layer norm. Note for code simplicity the norm is first as opposed to last.

Abstract base class for different ML models.

Defines the Markov decision process of generating a molecule.

Define standard linear + softmax generation step.