class AbstractLayer(*args: Any, **kwargs: Any)

Bases: Module, ABC

Abstract layer with KG and Basic GCN layers

distance(triplets: torch.Tensor) → torch.Tensor

Get distance between triples

Parameters:

triplets – (Tensor) triplets to calculate

Returns:

(Tensor) distances

forward_gcn(x: torch.Tensor) → torch.Tensor

Forward GCN

Parameters:

x – Tensor

Returns:

Tensor

get_n_hop(entity_id: int) → Dict[int, List[int]]

Get n_hops from entity

Parameters:

entity_id – id of entity

Returns:

Dict with N-hop neighbours

class GCNGRU(*args: Any, **kwargs: Any)

Bases: AbstractLayer

GCN + GRU layer

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

Forward GCN + GRU layers

Parameters:

x – Tensor

Returns:

Tensor

class GRU(*args: Any, **kwargs: Any)

Bases: AbstractLayer

GRU layer

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

Forward

Parameters:

x – Tensor

Returns:

Tensor

transe_forward(pos_triplet: torch.Tensor, neg_triplet: torch.Tensor) → torch.Tensor

Forward with KGE model

Parameters:

• pos_triplet – (Tensor) positive triplets

• neg_triplet – (Tensor) negative triplets

Returns:

TransE scores