C++ API Reference for Intel® Data Analytics Acceleration Library 2020 Update 1
2D transposed convolution layer parameters More...
Parameter | ( | ) |
Default constructor
size_t groupDimension |
Dimension for which the grouping is applied. groupDimension=1 is supported now
Indices indices |
Data structure representing the dimension for convolution kernels. (2,3) is supported now
KernelSizes kernelSizes |
Data structure representing the sizes of the two-dimensional kernel subtensor
size_t nGroups |
Number of groups which the input data is split in groupDimension dimension
size_t nKernels |
Number of kernels applied to the input layer data
Paddings paddings |
Data structure representing the number of data to be implicitly added to the subtensor
Strides strides |
Data structure representing the intervals on which the kernel should be applied to the input
ValueSizes valueSizes |
Data structure representing the dimension sizes of value tensor Y. If this value contains (0, 0) the size of value dimensions will be computed automatically.
For more complete information about compiler optimizations, see our Optimization Notice.