Python objects & constants

Objects

tensor_qualifiers_dtype	NumPy dtype object that represents the cutensornetTensorQualifiers_t struct.
contraction_path_dtype	NumPy dtype object that represents the cutensornetContractionPath_t struct.
slice_info_pair_dtype	NumPy dtype object that represents the cutensornetSliceInfoPair_t struct.
slicing_config_dtype	NumPy dtype object that represents the cutensornetSlicingConfig_t struct.

Enums

Status(value)	See cutensornetStatus_t.
NetworkAttribute(value)	See cutensornetNetworkAttributes_t.
ContractionAutotunePreferenceAttribute(value)	See cutensornetContractionAutotunePreferenceAttributes_t.
ContractionOptimizerConfigAttribute(value)	See cutensornetContractionOptimizerConfigAttributes_t.
ContractionOptimizerInfoAttribute(value)	See cutensornetContractionOptimizerInfoAttributes_t.
GraphAlgo(value)	See cutensornetGraphAlgo_t.
MemoryModel(value)	See cutensornetMemoryModel_t.
OptimizerCost(value)	See cutensornetOptimizerCost_t.
SmartOption(value)	See cutensornetSmartOption_t.
WorksizePref(value)	See cutensornetWorksizePref_t.
Memspace(value)	See cutensornetMemspace_t.
WorkspaceKind(value)	See cutensornetWorkspaceKind_t.
TensorSVDConfigAttribute(value)	See cutensornetTensorSVDConfigAttributes_t.
TensorSVDInfoAttribute(value)	See cutensornetTensorSVDInfoAttributes_t.
TensorSVDNormalization(value)	See cutensornetTensorSVDNormalization_t.
TensorSVDPartition(value)	See cutensornetTensorSVDPartition_t.
TensorSVDAlgo(value)	See cutensornetTensorSVDAlgo_t.
GateSplitAlgo(value)	See cutensornetGateSplitAlgo_t.
StatePurity(value)	See cutensornetStatePurity_t.
BoundaryCondition(value)	See cutensornetBoundaryCondition_t.
StateAttribute(value)	See cutensornetStateAttributes_t.
StateMPOApplication(value)	See cutensornetStateMPOApplication_t.
MarginalAttribute(value)	See cutensornetMarginalAttributes_t.
SamplerAttribute(value)	See cutensornetSamplerAttributes_t.
AccessorAttribute(value)	See cutensornetAccessorAttributes_t.
ExpectationAttribute(value)	See cutensornetExpectationAttributes_t.

Python functions

Handle Management API

create()	Initializes the cuTensorNet library.
destroy(intptr_t handle)	Destroys the cuTensorNet library handle.

Network Descriptor API

create_network_descriptor(intptr_t handle, ...)	Initializes a cutensornetNetworkDescriptor_t, describing the connectivity (i.e., network topology) between the tensors.
destroy_network_descriptor(intptr_t desc)	Frees all the memory associated with the network descriptor.
get_output_tensor_descriptor(...)	Creates a cutensornetTensorDescriptor_t representing the output tensor of the network.
network_get_attribute(intptr_t handle, ...)	Gets attributes of network_descriptor.
network_set_attribute(intptr_t handle, ...)	Sets attributes of network_descriptor.

Tensor Descriptor API

create_tensor_descriptor(intptr_t handle, ...)	Initializes a cutensornetTensorDescriptor_t, describing the information of a tensor.
destroy_tensor_descriptor(intptr_t desc_tensor)	Frees all the memory associated with the tensor descriptor.
get_tensor_details(intptr_t handle, ...)	Get the tensor's metadata.

Contraction Optimizer API

create_contraction_optimizer_info(...)	Allocates resources for optimizerInfo.
create_contraction_optimizer_info_from_packed_data(...)	Create an optimizerInfo object from the provided buffer.
update_contraction_optimizer_info_from_packed_data(...)	Update the provided optimizer_info object from the provided buffer.
destroy_contraction_optimizer_info(...)	Frees all the memory associated with optimizer_info.
get_contraction_optimizer_info_attribute_dtype(...)	Get the Python data type of the corresponding ContractionOptimizerInfoAttribute attribute.
contraction_optimizer_info_get_attribute(...)	Gets attributes of optimizer_info.
contraction_optimizer_info_set_attribute(...)	Sets attributes of optimizer_info.
contraction_optimizer_info_get_packed_size(...)	Gets the packed size of the optimizer_info object.
contraction_optimizer_info_pack_data(...)	Packs the optimizer_info object into the provided buffer.
create_contraction_optimizer_config(...)	Sets up the required hyper-optimization parameters for the contraction order solver (see contraction_optimize()).
destroy_contraction_optimizer_config(...)	Frees all the memory associated with optimizer_config.
get_contraction_optimizer_config_attribute_dtype(...)	Get the Python data type of the corresponding ContractionOptimizerConfigAttribute attribute.
contraction_optimizer_config_get_attribute(...)	Gets attributes of optimizer_config.
contraction_optimizer_config_set_attribute(...)	Sets attributes of optimizer_config.
contraction_optimize(intptr_t handle, ...)	Computes an "optimized" contraction order as well as slicing info (for more information see Overview section) for a given tensor network such that the total time to solution is minimized while adhering to the user-provided memory constraint.

Contraction Plan API

create_contraction_plan(intptr_t handle, ...)	Initializes a cutensornetContractionPlan_t.
destroy_contraction_plan(intptr_t plan)	Frees all resources owned by plan.
create_contraction_autotune_preference(...)	Sets up the required auto-tune parameters for the contraction plan.
destroy_contraction_autotune_preference(...)	Frees all the memory associated with autotune_preference.
get_contraction_autotune_preference_attribute_dtype(...)	Get the Python data type of the corresponding ContractionAutotunePreferenceAttribute attribute.
contraction_autotune_preference_get_attribute(...)	Gets attributes of autotune_preference.
contraction_autotune_preference_set_attribute(...)	Sets attributes of autotune_preference.
contraction_autotune(intptr_t handle, ...)	Auto-tunes the contraction plan to find the best cutensorContractionPlan_t for each pair-wise contraction.

Workspace Management API

create_workspace_descriptor(intptr_t handle)	Creates a workspace descriptor that holds information about the user provided memory buffer.
destroy_workspace_descriptor(intptr_t desc)	Frees the workspace descriptor.
workspace_compute_contraction_sizes(...)	Computes the workspace size needed to contract the input tensor network using the provided contraction path.
workspace_compute_qr_sizes(intptr_t handle, ...)	Computes the workspace size needed to perform the tensor QR operation.
workspace_compute_svd_sizes(intptr_t handle, ...)	Computes the workspace size needed to perform the tensor SVD operation.
workspace_compute_gate_split_sizes(...)	Computes the workspace size needed to perform the gating operation.
workspace_get_memory_size(intptr_t handle, ...)	Retrieves the needed workspace size for the given workspace preference, memory space, workspace kind.
workspace_set_memory(intptr_t handle, ...)	Sets the memory address and workspace size of the workspace provided by user.
workspace_get_memory(intptr_t handle, ...)	Retrieves the memory address and workspace size of workspace hosted in the workspace descriptor.
workspace_purge_cache(intptr_t handle, ...)	Purges the cached data in the specified memory space.

Memory Management API

set_device_mem_handler(intptr_t handle, handler)	Set the device memory handler for cuTensorNet.
get_device_mem_handler(intptr_t handle)	Get the device memory handler for cuTensorNet.

Network Contraction API

contract_slices(intptr_t handle, ...)

Performs the actual contraction of the tensor network.

Gradient Computation API

compute_gradients_backward(intptr_t handle, ...)

Computes the gradients of the network w.r.t.

Slice Group API

create_slice_group_from_id_range(...)	Create a cutensornetSliceGroup_t object from a range, which produces a sequence of slice IDs from the specified start (inclusive) to the specified stop (exclusive) values with the specified step.
create_slice_group_from_ids(intptr_t handle, ...)	Create a slice group from a sequence of slice IDs.
destroy_slice_group(intptr_t slice_group)	Releases the resources associated with a cutensornetSliceGroup_t object and sets its value to null.

Approximate Tensor Network Execution API

tensor_qr(intptr_t handle, ...)	Performs QR decomposition of a tensor.
tensor_svd(intptr_t handle, ...)	Performs SVD decomposition of a tensor.
gate_split(intptr_t handle, ...)	Performs gate split operation.

Tensor SVD Config API

create_tensor_svd_config(intptr_t handle)	Sets up the options for singular value decomposition and truncation.
destroy_tensor_svd_config(intptr_t svd_config)	Frees all the memory associated with the tensor svd configuration.
get_tensor_svd_config_attribute_dtype(int attr)	Get the Python data type of the corresponding TensorSVDConfigAttribute attribute.
tensor_svd_config_get_attribute(...)	Gets attributes of svd_config.
tensor_svd_config_set_attribute(...)	Sets attributes of svd_config.

Tensor SVD Info API

create_tensor_svd_info(intptr_t handle)	Sets up the information for singular value decomposition.
destroy_tensor_svd_info(intptr_t svd_info)	Frees all the memory associated with the TensorSVDInfo object.
get_tensor_svd_info_attribute_dtype(int attr)	Get the Python data type of the corresponding TensorSVDInfoAttribute attribute.
tensor_svd_info_get_attribute(...)	Gets attributes of svd_info.

Distributed Parallelization API

distributed_reset_configuration(...)	Resets the distributed MPI parallelization configuration.
distributed_get_num_ranks(intptr_t handle)	Queries the number of MPI ranks in the current distributed MPI configuration.
distributed_get_proc_rank(intptr_t handle)	Queries the rank of the current MPI process in the current distributed MPI configuration.
distributed_synchronize(intptr_t handle)	Globally synchronizes all MPI processes in the current distributed MPI configuration, ensuring that all preceding cutensornet API calls have completed across all MPI processes.
get_mpi_comm_pointer(comm)	Simple helper to get the address to and size of a MPI_Comm handle.

High-Level Tensor Network API

create_state(intptr_t handle, int purity, ...)	Creates an empty tensor network state of a given shape defined by the number of primary tensor modes and their extents.
destroy_state(intptr_t tensor_network_state)	Frees all resources owned by the tensor network state.
state_apply_tensor(intptr_t handle, ...)	DEPRECATED: Applies a tensor operator to the tensor network state.
state_apply_tensor_operator(intptr_t handle, ...)	Applies a tensor operator to the tensor network state.
state_apply_controlled_tensor_operator(...)	Applies a controlled tensor operator to the tensor network state.
state_apply_network_operator(...)	Applies a tensor network operator to a tensor network state.
state_update_tensor(intptr_t handle, ...)	Registers an external update of the elements of the specified tensor operator that was previously applied to the tensor network state.
state_update_tensor_operator(...)	Registers an external update of the elements of the specified tensor operator that was previously applied to the tensor network state.
get_state_attribute_dtype(int attr)	Get the Python data type of the corresponding StateAttribute attribute.
state_configure(intptr_t handle, ...)	Configures computation of the full tensor network state, either in the exact or a factorized form.
state_prepare(intptr_t handle, ...)	Prepares computation of the full tensor network state, either in the exact or a factorized form.
state_get_info(intptr_t handle, ...)	Retrieves an attribute related to computation of the full tensor network state, either in the exact or a factorized form.
state_compute(intptr_t handle, ...)	Computes the tensor network state representation.
state_initialize_mps(intptr_t handle, ...)	Imposes a user-defined MPS (Matrix Product State) factorization on the initial tensor network state with the given shape and data.
state_finalize_mps(intptr_t handle, ...)	Imposes a user-defined MPS (Matrix Product State) factorization on the final tensor network state with the given shape.
create_marginal(intptr_t handle, ...)	Creates a representation of the specified marginal tensor for a given tensor network state.
destroy_marginal(...)	Destroys the tensor network state marginal.
get_marginal_attribute_dtype(int attr)	Get the Python data type of the corresponding MarginalAttribute attribute.
marginal_configure(intptr_t handle, ...)	Configures computation of the requested tensor network state marginal tensor.
marginal_prepare(intptr_t handle, ...)	Prepares computation of the requested tensor network state marginal tensor.
marginal_get_info(intptr_t handle, ...)	Retrieves an attribute related to computation of the requested tensor network state marginal tensor.
marginal_compute(intptr_t handle, ...)	Computes the requested tensor network state marginal tensor.
create_sampler(intptr_t handle, ...)	Creates a tensor network state sampler.
destroy_sampler(intptr_t tensor_network_sampler)	Destroys the tensor network state sampler.
get_sampler_attribute_dtype(int attr)	Get the Python data type of the corresponding SamplerAttribute attribute.
sampler_configure(intptr_t handle, ...)	Configures the tensor network state sampler.
sampler_prepare(intptr_t handle, ...)	Prepares the tensor network state sampler.
sampler_get_info(intptr_t handle, ...)	Retrieves an attribute related to tensor network state sampling.
sampler_sample(intptr_t handle, ...)	Performs sampling of the tensor network state, that is, generates the requested number of samples.
create_accessor(intptr_t handle, ...)	Creates a tensor network state amplitudes accessor.
destroy_accessor(...)	Destroyes the tensor network state amplitudes accessor.
get_accessor_attribute_dtype(int attr)	Get the Python data type of the corresponding AccessorAttribute attribute.
accessor_configure(intptr_t handle, ...)	Configures computation of the requested tensor network state amplitudes tensor.
accessor_prepare(intptr_t handle, ...)	Prepares computation of the requested tensor network state amplitudes tensor.
accessor_get_info(intptr_t handle, ...)	Retrieves an attribute related to computation of the requested tensor network state amplitudes tensor.
accessor_compute(intptr_t handle, ...)	Computes the amplitudes of the tensor network state.
create_expectation(intptr_t handle, ...)	Creates a representation of the tensor network state expectation value.
destroy_expectation(...)	Destroyes the tensor network state expectation value representation.
get_expectation_attribute_dtype(int attr)	Get the Python data type of the corresponding ExpectationAttribute attribute.
expectation_configure(intptr_t handle, ...)	Configures computation of the requested tensor network state expectation value.
expectation_prepare(intptr_t handle, ...)	Prepares computation of the requested tensor network state expectation value.
expectation_get_info(intptr_t handle, ...)	Retrieves an attribute related to computation of the requested tensor network state expectation value.
expectation_compute(intptr_t handle, ...)	Computes an (unnormalized) expectation value of a given tensor network operator over a given tensor network state.

Tensor Network Operator API

create_network_operator(intptr_t handle, ...)	Creates an uninitialized tensor network operator of a given shape defined by the number of state modes and their extents.
network_operator_append_product(...)	Appends a tensor product operator component to the tensor network operator.
network_operator_append_mpo(intptr_t handle, ...)	Appends a Matrix Product Operator (MPO) component to the tensor network operator.
destroy_network_operator(...)	Frees all resources owned by the tensor network operator.

Versioning API

get_version()	Returns Version number of the cuTensorNet library.
get_cudart_version()	Returns version number of the CUDA runtime that cuTensorNet was compiled against.