kutacc_af2_linear
KuTACC implementation for linear change, corresponding to the following formula:
Interface Definition
void kutacc_af2_linear(kutacc_tensor_h act, kutacc_tensor_h weight, float* bias_data, kutacc_tensor_h result, int64_t beta);
Parameters
Parameter |
Type |
Description |
Input/Output |
|---|---|---|---|
act |
kutacc_tensor_h |
Basic data A |
Input |
weight |
kutacc_tensor_h |
Weight parameter B |
Input |
bias_data |
float * |
Bias data |
Input |
result |
kutacc_tensor_h |
Calculation result |
Output |
beta |
int64_t |
Coefficient of C |
Input |
Parameter constraints:
The dimension size of bias must equal the dimension size of weight minus 1.
The dimension sizes of act and weight are both greater than or equal to 2.
Parent topic: GEMM Operators