qlinks.backends package

Submodules

qlinks.backends.sparse module

class qlinks.backends.sparse.SparseBackend(*args, **kwargs)

Bases: Protocol

Minimal backend protocol for sparse Hamiltonian construction.

name: str

as_data_array(data, dtype)

as_index_array(indices)

coo_matrix(data, rows, cols, shape, dtype)

empty_csr(shape, dtype)

to_cpu_array(array)

max_abs_data(sparse_matrix)

__init__(*args, **kwargs)

class qlinks.backends.sparse.ScipySparseBackend(name='scipy')

Bases: object

SciPy sparse backend used by default Hamiltonian builders.

The backend normalizes sparse-matrix construction behind a small protocol so model/build code can also target CuPy when available.

name

Backend name, always "scipy".

Type:

str

name: str

as_data_array(data, dtype)

as_index_array(indices)

coo_matrix(data, rows, cols, shape, dtype)

empty_csr(shape, dtype)

to_cpu_array(array)

max_abs_data(sparse_matrix)

__init__(name='scipy')

class qlinks.backends.sparse.CupySparseBackend(name='cupy')

Bases: object

CuPy/CuPyX sparse backend.

Imported lazily so qlinks does not require CuPy unless backend=”cupy” is used.

Note

cupyx.scipy.sparse currently follows the scipy sparse matrix-style API more closely than scipy’s newer sparse array API. Therefore this backend creates cupyx.scipy.sparse.coo_matrix/csr_matrix, not csr_array.

name: str

property cp

property cupyx_sparse

as_data_array(data, dtype)

as_index_array(indices)

coo_matrix(data, rows, cols, shape, dtype)

empty_csr(shape, dtype)

to_cpu_array(array)

max_abs_data(sparse_matrix)

__init__(name='cupy')

qlinks.backends.sparse.get_sparse_backend(backend='scipy')

Resolve a sparse backend.

backend=”auto” currently chooses scipy. This keeps behavior deterministic. Later, you can change “auto” to use cupy if an input basis lives on GPU.

Module contents

class qlinks.backends.CupySparseBackend(name='cupy')

Bases: object

CuPy/CuPyX sparse backend.

Imported lazily so qlinks does not require CuPy unless backend=”cupy” is used.

Note

cupyx.scipy.sparse currently follows the scipy sparse matrix-style API more closely than scipy’s newer sparse array API. Therefore this backend creates cupyx.scipy.sparse.coo_matrix/csr_matrix, not csr_array.

name: str

property cp

property cupyx_sparse

as_data_array(data, dtype)

as_index_array(indices)

coo_matrix(data, rows, cols, shape, dtype)

empty_csr(shape, dtype)

to_cpu_array(array)

max_abs_data(sparse_matrix)

__init__(name='cupy')

class qlinks.backends.ScipySparseBackend(name='scipy')

Bases: object

SciPy sparse backend used by default Hamiltonian builders.

The backend normalizes sparse-matrix construction behind a small protocol so model/build code can also target CuPy when available.

name

Backend name, always "scipy".

Type:

str

name: str

as_data_array(data, dtype)

as_index_array(indices)

coo_matrix(data, rows, cols, shape, dtype)

empty_csr(shape, dtype)

to_cpu_array(array)

max_abs_data(sparse_matrix)

__init__(name='scipy')

class qlinks.backends.SparseBackend(*args, **kwargs)

Bases: Protocol

Minimal backend protocol for sparse Hamiltonian construction.

name: str

as_data_array(data, dtype)

as_index_array(indices)

coo_matrix(data, rows, cols, shape, dtype)

empty_csr(shape, dtype)

to_cpu_array(array)

max_abs_data(sparse_matrix)

__init__(*args, **kwargs)

qlinks.backends.get_sparse_backend(backend='scipy')

Resolve a sparse backend.

backend=”auto” currently chooses scipy. This keeps behavior deterministic. Later, you can change “auto” to use cupy if an input basis lives on GPU.