- cuquantum.cutensornet.create_network_descriptor(intptr_t handle, int32_t n_inputs, n_modes_in, extents_in, strides_in, modes_in, alignments_in, int32_t n_modes_out, extents_out, strides_out, modes_out, uint32_t alignment_out, int data_type, int compute_type) intptr_t ¶
Create a tensor network descriptor.
handle (intptr_t) – The library handle.
n_inputs (int) – The number of input tensors.
A host array of the number of modes for each input tensor. It can be
A host array of extents for each input tensor. It can be
A host array of strides for each input tensor. It can be
A host array of modes for each input tensor. It can be
A host array of alignments for each input tensor. It can be
n_modes_out (int32_t) – The number of modes of the output tensor. If this is set to -1 and
modes_outis set to 0 (not provided), the output modes will be inferred. If this is set to 0, the network is force reduced.
The extents of the output tensor (on host). It can be
The strides of the output tensor (on host). It can be
The modes of the output tensor (on host). It can be
alignment_out (uint32_t) – The alignment for the output tensor.
data_type (cuquantum.cudaDataType) – The data type of the input and output tensors.
compute_type (cuquantum.ComputeType) – The compute type of the tensor contraction.
An opaque descriptor handle (as Python
- Return type
strides_out) is set to 0 (
NULL), it means the input tensors (output tensor) are in the Fortran layout (F-contiguous).