Intel® Math Kernel Library 2018 Developer Reference - C
Creates split layers.
dnnError_t dnnSplitCreate_F32 (dnnPrimitive_t *pSplit, dnnPrimitiveAtributes_t attributes, const size_t nDstTensors, dnnLayout_t dataLayout, size_t dstChannelSize[]);
dnnError_t dnnSplitCreate_F64 (dnnPrimitive_t *pSplit, dnnPrimitiveAtributes_t attributes, const size_t nDstTensors, dnnLayout_t dataLayout, size_t dstChannelSize[]);
attributes |
The set of attributes for the primitive. |
nDstTensors |
The number of output tensors. |
dataLayout |
The layout of the input tensor. |
dstChannelSize |
dstChannelSize[i] is the number of channels for the i-th output tensor. |
pSplit |
Pointer to the primitive to create. |
Each dnnSplitCreate function creates a split primitive. The split of a tensor into M tensors with C0, C1, ..., CM-1 channels, respectively, and same other sizes is defined as:
where