Pad

Description

Padding the input tensor.

Input

Two inputs:

x: tensor of type float16, float, or int32.

pads: int32, int64.

constant_value: optional. Defaults to 0, an empty string, or False. If the selected mode is constant, the scalar value is used.

Output

One output:

y: tensor of the identical data type as input x.

Property

mode: string type. The supported modes are constant, reflect, and edge.

Constraints

When the value of mode is constant, only constant_value=0 is supported.

This operator supports only the Atlas inference products .

ONNX Opset Support

Opset v11/v13/v14/v15/v16/v17/v18