terminator.models.layers.condense.ResidueFeatures

class terminator.models.layers.condense.ResidueFeatures(hparams)[source]

Bases: Module

Module which featurizes TERM match residue information

Variables:

  • embedding (nn.Embedding) – Embedding layer for residue identities (represented as int)

  • relu (nn.ReLU) – ReLU activation layer

  • tanh (nn.Tanh) – tanh activation layer

  • lin2 (lin1,) – Embedding layers

  • bn (nn.BatchNorm2d) – Batch Normalization for features

__init__(hparams)[source]
Parameters:

hparams (dict) – Dictionary of model hparams (see ~/scripts/models/train/default_hparams.json for more info)

Methods

__init__(hparams)

Parameters:

hparams (dict) -- Dictionary of model hparams (see ~/scripts/models/train/default_hparams.json for more info)

forward(X, features)

Featurize TERM matches and their associated features

Attributes

T_destination

alias of TypeVar('T_destination', bound=Mapping[str, Tensor])

dump_patches

This allows better BC support for load_state_dict().
forward(X, features)[source]

Featurize TERM matches and their associated features

Parameters:

  • X (torch.LongTensor) – Match residue identities Shape: n_batches x n_matches x sum_term_len

  • features (torch.Tensor) – Features associated with match residues (e.g. torsion angles, RMSD, environment value Shape: n_batches x n_matches x sum_term_len x NUM_TERM_FEATURES

Returns:

out – Featurized TERM match residues Shape: n_batches x n_hidden x sum_term_len x n_alignments

Return type:

torch.Tensor