DynSeqOuter

Description

Sums up by offset.

Input

Four inputs:

x1: tensor of type float16 or float, with shape [bs, feature_dim]. bs: number of data records in a single batch; feature_dim: feature dimension.

x2: tensor of type float16 or float, with shape [bs, feature_dim]. bs: number of data records in a single batch; feature_dim: feature dimension.

seq_len1: tensor of type int32, with shape [batch_size] and bs=sum (seq_len1).

seq_len2: tensor of type int32, with shape [batch_size] and bt=sum (seq_len2).

Output

One output:

y: tensor of type float16 or float, with shape [bst, feature_dim] and bst=sum(seq_len1 * seq_len2).

Constraints

This operator is customized for specific user scenarios and is not recommended in irrelevant scenarios.

This operator supports only the Atlas inference products .

ONNX Opset Support

Opset v8/v9/v10/v11/v12/v13/v14/v15/v16