TI Deep Learning Product User Guide
|
This structure define the parameters of Inner Product (Fully connected) layer in TIDL.
Data Fields | |
int32_t | weights |
int32_t | bias |
int32_t | activationType |
int32_t | numInNodes |
int32_t | numOutNodes |
int32_t | weightsQ |
float32_tidl | weightScale |
float32_tidl | biasScale |
int32_t | zeroWeightValue |
int32_t | biasQ |
int32_t | inDataQ |
int32_t | outDataQ |
int32_t | interDataQ |
int32_t | biasB |
int32_t sTIDL_InnerProductParams_t::weights |
Offset for kernel parameters
int32_t sTIDL_InnerProductParams_t::bias |
Offset for bias parameters
int32_t sTIDL_InnerProductParams_t::activationType |
activation type to be used
int32_t sTIDL_InnerProductParams_t::numInNodes |
Number of elements in the flattened input
int32_t sTIDL_InnerProductParams_t::numOutNodes |
Number of elements in the output
int32_t sTIDL_InnerProductParams_t::weightsQ |
Q value of Kernel weights
float32_tidl sTIDL_InnerProductParams_t::weightScale |
floating point scale for weight
float32_tidl sTIDL_InnerProductParams_t::biasScale |
floating point scale for bias
int32_t sTIDL_InnerProductParams_t::zeroWeightValue |
value of weights added for dynamic quantSytle
int32_t sTIDL_InnerProductParams_t::biasQ |
Q value kernel Bias
int32_t sTIDL_InnerProductParams_t::inDataQ |
Q value of the in data
int32_t sTIDL_InnerProductParams_t::outDataQ |
Q value expected for out data
int32_t sTIDL_InnerProductParams_t::interDataQ |
Q value intermediate output data
int32_t sTIDL_InnerProductParams_t::biasB |
Newly added