Embedding

Function

Map discrete indexes to continuous vectors (for table lookup).

Prototype

torch.nn.Embedding(
    num_embeddings: int,
    embedding_dim: int,
    padding_idx: Optional[int] = None,
    max_norm: Optional[float] = None,
    norm_type: float = 2.0,
    scale_grad_by_freq: bool = False,
    sparse: bool = False,
    _weight: Optional[Tensor] = None,
    device=None,
    dtype=None
)

Parameters

**Table 1** Parameter description
Parameter	Type	Mandatory (Yes/No)	Description
num_embeddings	int	Yes	Size of the embedding dictionary (index range: [0, num_embeddings-1])
embedding_dim	int	Yes	Embedding vector dimension
padding_idx	int	No	Padding index (the vector gradient corresponding to this index is fixed at 0)
max_norm	float	No	Maximum norm constraint (scaling is performed when the constraint is exceeded)
norm_type	float	No	Norm calculation type (2.0 by default, indicating the L2 norm)
scale_grad_by_freq	bool	No	Whether to scale gradients by frequency (default value: False)
sparse	bool	No	Whether to use sparse gradients (default value: False)

Parent topic: Operator Interfaces