Documentation

mmad

Function

Performs matrix multiplication and addition.

Prototype

1
class mmad(x, y, b, is_inited=False)

Parameters

Parameter

Data Type

Description

x

Tensor variable

Left matrix, in the L0A space. FP16 is supported.

y

Tensor variable

Right matrix, in the L0B space. FP16 is supported.

b

Tensor variable

Bias term, which can be in the L0C space or Bias Table space. FP32 is supported

is_inited

bool

When the input is in the L0C space, is_inited=True needs to be added because there is no direct channel to transfer data from the GM to the L0C.

Constraints

When the bias term is in the Bias Table space, the tensor data format must be ND and shape must be [n,].

Example

1
2
3
4
5
from mskpp import mmad, Tensor
in_x = Tensor("GM", "FP16", [32, 48], format="ND")
in_y = Tensor("GM", "FP16", [48, 16], format="ND")
in_z = Tensor("GM", "FP32", [32, 16], format="NC1HWC0")
out_z = mmad(in_x, in_y, in_z)()