matrix_attention

[ allennlp.modules.matrix_attention.matrix_attention ]

---

MatrixAttention Objects

class MatrixAttention(torch.nn.Module,  Registrable)

MatrixAttention takes two matrices as input and returns a matrix of attentions.

We compute the similarity between each row in each matrix and return unnormalized similarity scores. Because these scores are unnormalized, we don't take a mask as input; it's up to the caller to deal with masking properly when this output is used.

Input: - matrix_1 : (batch_size, num_rows_1, embedding_dim_1) - matrix_2 : (batch_size, num_rows_2, embedding_dim_2)

Output: - (batch_size, num_rows_1, num_rows_2)

forward

 | def forward(
 |     self,
 |     matrix_1: torch.Tensor,
 |     matrix_2: torch.Tensor
 | ) -> torch.Tensor