NodePiece

class NodePiece(*, triples_factory: ~pykeen.triples.triples_factory.CoreTriplesFactory, num_tokens: int | ~collections.abc.Sequence[int] = 2, tokenizers: str | ~pykeen.nn.node_piece.tokenization.Tokenizer | type[~pykeen.nn.node_piece.tokenization.Tokenizer] | None | ~collections.abc.Sequence[str | ~pykeen.nn.node_piece.tokenization.Tokenizer | type[~pykeen.nn.node_piece.tokenization.Tokenizer] | None] = None, tokenizers_kwargs: ~collections.abc.Mapping[str, ~typing.Any] | None | ~collections.abc.Sequence[~collections.abc.Mapping[str, ~typing.Any] | None] = None, embedding_dim: int = 64, interaction: str | ~pykeen.nn.modules.Interaction | type[~pykeen.nn.modules.Interaction] | None = <class 'pykeen.nn.modules.DistMultInteraction'>, aggregation: str | ~collections.abc.Callable[[~torch.Tensor, int], ~torch.Tensor] | None = None, entity_initializer: str | ~collections.abc.Callable[[~torch.Tensor], ~torch.Tensor] | None = None, entity_normalizer: str | ~collections.abc.Callable[[~torch.Tensor], ~torch.Tensor] | None = None, entity_constrainer: str | ~collections.abc.Callable[[~torch.Tensor], ~torch.Tensor] | None = None, entity_regularizer: str | ~pykeen.regularizers.Regularizer | None = None, relation_initializer: str | ~collections.abc.Callable[[~torch.Tensor], ~torch.Tensor] | None = None, relation_normalizer: str | ~collections.abc.Callable[[~torch.Tensor], ~torch.Tensor] | None = None, relation_constrainer: str | ~collections.abc.Callable[[~torch.Tensor], ~torch.Tensor] | None = None, relation_regularizer: str | ~pykeen.regularizers.Regularizer | None = None, **kwargs)[source]

Bases: ERModel[Tensor, Tensor, Tensor]

A wrapper which combines an interaction function with NodePiece entity representations from [galkin2021].

This model uses the pykeen.nn.NodePieceRepresentation instead of a typical pykeen.nn.representation.Embedding to more efficiently store representations.

Initialize the model.

Parameters:

  • triples_factory (CoreTriplesFactory) – the triples factory. Must have create_inverse_triples set to True.

  • num_tokens (int | Sequence[int]) – the number of relations to use to represent each entity, cf. pykeen.nn.NodePieceRepresentation.

  • tokenizers (str | Tokenizer | type[Tokenizer] | None | Sequence[str | Tokenizer | type[Tokenizer] | None]) – the tokenizer to use, cf. pykeen.nn.node_piece.tokenizer_resolver.

  • tokenizers_kwargs (Mapping[str, Any] | None | Sequence[Mapping[str, Any] | None]) – additional keyword-based parameters passed to the tokenizer upon construction.

  • embedding_dim (int) – the embedding dimension. Only used if embedding_specification is not given.

  • interaction (str | Interaction | type[Interaction] | None) – the interaction module, or a hint for it.

  • aggregation (str | Callable[[Tensor, int], Tensor] | None) –

    aggregation of multiple token representations to a single entity representation. By default, this uses torch.mean(). If a string is provided, the module assumes that this refers to a top-level torch function, e.g. “mean” for torch.mean(), or “sum” for func:torch.sum. An aggregation can also have trainable parameters, .e.g., MLP(mean(MLP(tokens))) (cf. DeepSets from [zaheer2017]). In this case, the module has to be created outside of this component.

    Moreover, we support providing “mlp” as a shortcut to use the MLP aggregation version from [galkin2021].

    We could also have aggregations which result in differently shapes output, e.g. a concatenation of all token embeddings resulting in shape (num_tokens * d,). In this case, shape must be provided.

    The aggregation takes two arguments: the (batched) tensor of token representations, in shape (*, num_tokens, *dt), and the index along which to aggregate.

  • entity_initializer (str | Callable[[Tensor], Tensor] | None) – a hint for initializing anchor embeddings

  • entity_normalizer (str | Callable[[Tensor], Tensor] | None) – a hint for normalizing anchor embeddings

  • entity_constrainer (str | Callable[[Tensor], Tensor] | None) – a hint for constraining anchor embeddings

  • entity_regularizer (str | Regularizer | None) – a hint for regularizing anchor embeddings

  • relation_initializer (str | Callable[[Tensor], Tensor] | None) – a hint for initializing relation embeddings

  • relation_normalizer (str | Callable[[Tensor], Tensor] | None) – a hint for normalizing relation embeddings

  • relation_constrainer (str | Callable[[Tensor], Tensor] | None) – a hint for constraining relation embeddings

  • relation_regularizer (str | Regularizer | None) – a hint for regularizing relation embeddings

  • kwargs – additional keyword-based arguments passed to ERModel.__init__()

Raises:

ValueError – if the triples factory does not create inverse triples

Attributes Summary

hpo_default

The default strategy for optimizing the model's hyper-parameters

Attributes Documentation

hpo_default: ClassVar[Mapping[str, Any]] = {'embedding_dim': {'high': 256, 'low': 16, 'q': 16, 'type': <class 'int'>}}

The default strategy for optimizing the model’s hyper-parameters