HardSigmoidBackward

General

HardSigmoidBackward operation computes gradient for HardSigmoid. The formula is defined as follows:

\[\begin{split}diff\_src = \begin{cases} diff\_dst \cdot \alpha & \text{if}\ 0 < \alpha src + \beta < 1 \\ 0 & \text{otherwise}\ \end{cases}\end{split}\]

Operation attributes

Attribute Name	Descr
alpha	\(\alpha\) in the formula.	f32	Arbitrary f32 value.	Required
beta	\(\beta\) in the formula.	f32	Arbitrary f32 value.	Required

Execution arguments

The inputs and outputs must be provided according to the index order shown below when constructing an operation.

Inputs

Index	Argument Name	Re
0	src	Required
1	diff_dst	Required

Outputs

Index	Argument Name	Re
0	diff_src	Required

Supported data types

HardSigmoidBackward operation supports the following data type combinations.

Src	Diff_dst	Di
f32	f32	f32
f16	f16	f16
bf16	bf16	bf16