TI Deep Learning Product User Guide
|
This structure define the parameters of data or kernel buffer used by TIDL layers (In,Out)
Data Fields | |
int32_t | dataId |
int32_t | elementType |
int32_t | numDim |
int32_t | dataQ |
int32_t | minValue |
int32_t | maxValue |
float32_tidl | minTensorValue |
float32_tidl | maxTensorValue |
float32_tidl | tensorScale |
int32_t | padW |
int32_t | padH |
int32_t | batchPadW |
int32_t | batchPadH |
int32_t | numBatchW |
int32_t | numBatchH |
int32_t | roundBits |
int32_t | pitch [TIDL_DIM_MAX-1] |
int32_t | dimValues [TIDL_DIM_MAX] |
int32_t sTIDL_DataParams_t::dataId |
Unique Data ID for the tensor holding feature vector
int32_t sTIDL_DataParams_t::elementType |
Element type of the feature vector as defined in eTIDL_ElementType
int32_t sTIDL_DataParams_t::numDim |
Number of dimensions supported by the tensor
int32_t sTIDL_DataParams_t::dataQ |
Q factor used for fixed point : NOT used now
int32_t sTIDL_DataParams_t::minValue |
Minimum value of Accumulator in fixed point : NOT used now
int32_t sTIDL_DataParams_t::maxValue |
Maximum value of Accumulator in fixed point : NOT used now
float32_tidl sTIDL_DataParams_t::minTensorValue |
Minimum value of Accumulator in floating point
float32_tidl sTIDL_DataParams_t::maxTensorValue |
Maximum value of Accumulator in floating point
float32_tidl sTIDL_DataParams_t::tensorScale |
Scale value applied on the tensor for final fixed point
int32_t sTIDL_DataParams_t::padW |
Horizontal Padding
int32_t sTIDL_DataParams_t::padH |
Vertical Padding
int32_t sTIDL_DataParams_t::batchPadW |
Padding between horizontal Batches in stitched image
int32_t sTIDL_DataParams_t::batchPadH |
Padding between vertical Batches in the stitched image
int32_t sTIDL_DataParams_t::numBatchW |
Number of Batches stitched in the width direction for batch processing
int32_t sTIDL_DataParams_t::numBatchH |
Number of Batches stitched in the height direction for batch processing
int32_t sTIDL_DataParams_t::roundBits |
Rounding bit used while converting to small bit depth (8 or 16 bit)
int32_t sTIDL_DataParams_t::pitch[TIDL_DIM_MAX-1] |
Pitch for each dimension as defined by eTIDL_PitchDimIndex
int32_t sTIDL_DataParams_t::dimValues[TIDL_DIM_MAX] |
Values of dimensions as defined by eTIDL_DataDimIndex