Parameters

Member	Type	Default Value	Description
elewiseType	ElewiseType	ELEWISE_UNDEFINED	Computation type ELEWISE_UNDEFINED: default value, which is undefined. This default type is unavailable. You need to set this parameter. ELEWISE_CAST: data type conversion ELEWISE_MULS: element-wise vector multiplication ELEWISE_COS: element-wise cosine computation ELEWISE_SIN: element-wise sine computation ELEWISE_NEG: opposite number computation element-wise ELEWISE_QUANT: quantization. ELEWISE_LOGICAL_NOT: element-wise logical NOT ELEWISE_ADD: element-wise addition. ELEWISE_MUL: element-wise vector-to-vector multiplication. ELEWISE_REALDIV: element-wise vector-to-vector division. ELEWISE_LOGICAL_AND: element-wise logical AND ELEWISE_LOGICAL_OR: element-wise logical OR ELEWISE_LESS: element-wise less than comparison ELEWISE_GREATER: element-wise greater than comparison ELEWISE_SUB: element-wise subtraction. ELEWISE_EQUAL: element-wise equal comparison ELEWISE_QUANT_PER_CHANNEL: quantization per channel. ELEWISE_DEQUANT_PER_CHANNEL: dequantization per channel. ELEWISE_DYNAMIC_QUANT: element-wise dynamic quantization. ELEWISE_TANH: element-wise hyperbolic tangent computation. ELEWISE_TYPE_MAX: boundary value, which is used only to determine whether the value exceeds the boundary. This value cannot be used.
quantParam	struct	-	Quantization parameter. For details, see Table 1.
mulsParam	struct	-	Multiplication parameter. For details, see Table 2.
outTensorType	aclDataType	ACL_DT_UNDEFINED	Output data type after conversion.
rsv[8]	uint8_t	{0}	Reserved

**Table 1** QuantParam members
Member	Type	Default Value	Description
inputScale	float	1.0f	Quantization step
asymmetric	bool	false	Quantization mode. false indicates symmetric quantization, and true indicates asymmetric quantization.
inputOffset	int	0	Quantization offset
rsv[20]	uint8_t	{0}	Reserved

**Table 2** MulsParam members
Member	Type	Default Value	Description
varAttr	float	0.0f	Value of vector multiplication
rsv[12]	uint8_t	{0}	Reserved

Parent topic: ElewiseOperation