Converts a sparse tensor from a standardized sparse layout to a device specific sparse layout used by FullyConnected
- Parameter batchSize: The number of input-output pairs to process.
- Parameter inputDenseShape: The dense shape of the sparse 2D input. You must define the sizes and layout.
- Parameter inputLayout: The layout and nonzero indices of the sparse input.
- Parameter output: The destination array descriptor that contains the output device optimized BNNS Sparse Fully Connected weights.
If the output's `data` is `nil`, BNNS allocates the memory, otherwise, you may preallocate this memory to the size of the dense input (
use `BNNSNDArrayDescriptor.getter:dataSize` to calculate this value).
- Parameter sparseParameters: An optional data structure that provides a hint to the sparsity function.
- Parameter workspace: An optional pointer to a memory region that the function uses as scratch space. Set this to `nil` to have
BNNS allocate scratch space, otherwise, you must provide a buffer at least equal to the size of the dense input.
- seealso: `BNNSNDArrayFullyConnectedSparsifySparseCOO`.
- seealso: `BNNSNDArrayFullyConnectedSparsifySparseCSR`.
Omit the output parameters from the call. The result contains $ret and $out, with output inside $out.
