Retrievers

Base Retriever.

EncoderType

Bases: TypedDict

Type definition for encoder modules supporting different modalities.

Source code in src/fed_rag/base/retriever.py

class EncoderType(TypedDict):
    """Type definition for encoder modules supporting different modalities."""

    text: torch.nn.Module | None
    image: torch.nn.Module | None
    audio: torch.nn.Module | None
    video: torch.nn.Module | None

BaseRetriever

Bases: BaseModel, ABC

Base Retriever Class.

This abstract class provides the interface for creating Retriever objects that encode strings into numerical vector representations.

Source code in src/fed_rag/base/retriever.py

class BaseRetriever(BaseModel, ABC):
    """Base Retriever Class.

    This abstract class provides the interface for creating Retriever objects that
    encode strings into numerical vector representations.
    """

    model_config = ConfigDict(arbitrary_types_allowed=True)

    @abstractmethod
    def encode_query(
        self, query: str | list[str] | Query | list[Query], **kwargs: Any
    ) -> torch.Tensor | EncodeResult:
        """Encode a string query into a torch.Tensor.

        Args:
            query (str | list[str] | Query | list[Query]): The query or list of queries to encode.

        Returns:
            torch.Tensor: The vector representation(s) of the encoded query/queries.
        """

    @abstractmethod
    def encode_context(
        self, context: str | list[str] | Context | list[Context], **kwargs: Any
    ) -> torch.Tensor | EncodeResult:
        """Encode a string context into a torch.Tensor.

        Args:
            context (str | list[str] | Context | list[Context]): The context or list of contexts to encode.

        Returns:
            torch.Tensor: The vector representation(s) of the encoded context(s).
        """

    @property
    @abstractmethod
    def encoder(self) -> torch.nn.Module | EncoderType | None:
        """PyTorch model associated with the encoder associated with retriever."""

    @property
    @abstractmethod
    def query_encoder(self) -> torch.nn.Module | EncoderType | None:
        """PyTorch model associated with the query encoder associated with retriever."""

    @property
    @abstractmethod
    def context_encoder(self) -> torch.nn.Module | EncoderType | None:
        """PyTorch model associated with the context encoder associated with retriever."""

encoder abstractmethod property

encoder

PyTorch model associated with the encoder associated with retriever.

query_encoder abstractmethod property

query_encoder

PyTorch model associated with the query encoder associated with retriever.

context_encoder abstractmethod property

context_encoder

PyTorch model associated with the context encoder associated with retriever.

encode_query abstractmethod

encode_query(query, **kwargs)

Encode a string query into a torch.Tensor.

Parameters:

Name	Type	Description	Default
query	str | list[str] | Query | list[Query]	The query or list of queries to encode.	required

Returns:

Type	Description
Tensor | EncodeResult	torch.Tensor: The vector representation(s) of the encoded query/queries.

Source code in src/fed_rag/base/retriever.py

@abstractmethod
def encode_query(
    self, query: str | list[str] | Query | list[Query], **kwargs: Any
) -> torch.Tensor | EncodeResult:
    """Encode a string query into a torch.Tensor.

    Args:
        query (str | list[str] | Query | list[Query]): The query or list of queries to encode.

    Returns:
        torch.Tensor: The vector representation(s) of the encoded query/queries.
    """

encode_context abstractmethod

encode_context(context, **kwargs)

Encode a string context into a torch.Tensor.

Parameters:

Name	Type	Description	Default
context	str | list[str] | Context | list[Context]	The context or list of contexts to encode.	required

Returns:

Type	Description
Tensor | EncodeResult	torch.Tensor: The vector representation(s) of the encoded context(s).

Source code in src/fed_rag/base/retriever.py

@abstractmethod
def encode_context(
    self, context: str | list[str] | Context | list[Context], **kwargs: Any
) -> torch.Tensor | EncodeResult:
    """Encode a string context into a torch.Tensor.

    Args:
        context (str | list[str] | Context | list[Context]): The context or list of contexts to encode.

    Returns:
        torch.Tensor: The vector representation(s) of the encoded context(s).
    """