Documentation

vector_dup

Function

vector_dup instruction abstraction.

y = vector_dup(x), where x and y are filled with elements by element.

Prototype

1
class vector_dup(x, y, fill_shape)

Parameters

Parameter

Input/Output

Data Type

Description

x

Input

Tensor variable

Input x-vector tensor. FP16, FP32, INT16, INT32, UINT16, and UINT32 are supported.

y

Output

Tensor variable

Output y-vector tensor. FP16, FP32, INT16, INT32, UINT16, and UINT32 are supported.

fill_shape

Input

list

Shape value to be expanded of the target tensor.

Constraints

Since the input to this instruction is only a scalar, you need to create a tensor whose shape is [1] and is_inited is True as the simulated scalar input, without increasing performance overhead.

Example

1
2
3
4
from mskpp import vector_dup, Tensor
ub_x = Tensor("UB", "FP16", [1], format="ND", is_inited=True)
ub_y = Tensor("UB")
out = vector_dup(ub_x, ub_y, [8, 2048])()