Addcmul

English | 简体中文

Product Support Status

Product	Supported
Ascend 950PR&950DT products	√
Atlas A3 products	√
Atlas A2 products	√
Atlas 200I/500 A2 inference products	×
Atlas inference products	×
Atlas training products	√

Function Description

• Operator function: a tensor computation function that performs the combined multiply-divide-add operation and merges tensor multiplication (with scaling) + tensor addition into a single operation.

• Computation formula:

$out=self+value\times inpu{t}_1\times inpu{t}_2$

Parameter Description

Parameter Name	Input/Output/Attribute	Description	Data Type	Data Format
self	Input	Input parameter for the addcmul computation, which is self in the formula.	FLOAT16, BFLOAT16, FLOAT, INT32	ND
input_1	Input	Input parameter for the addcmul computation, which is input_1 in the formula.	FLOAT16, BFLOAT16, FLOAT, INT32	ND
input_2	Input	Input parameter for the addcmul computation, which is input_2 in the formula.	FLOAT16, BFLOAT16, FLOAT, INT32	ND
value	Input	Input parameter for the addcmul computation, which is value in the formula.	FLOAT16, BFLOAT16, FLOAT, INT32	ND
out	Output	Output parameter for the addcmul computation, which is out in the formula.	FLOAT16, BFLOAT16, FLOAT, INT32	ND

• Atlas training products and Atlas inference products: BFLOAT16 is not supported.

Constraint Description

None

Invocation Description

Invocation Method	Invocation Sample	Description
aclnn invocation	test_aclnn_addcmul	Invokes the Addcmul operator through the aclnnAddcmul API.
Graph-mode invocation	-	Invokes the Addcmul operator by building a graph through the operator IR.