atlas_q.cuquantum_backend.CuQuantumBackend

class atlas_q.cuquantum_backend.CuQuantumBackend(config=None)

Bases: object

Optional cuQuantum backend for accelerated tensor operations.

Automatically falls back to PyTorch if cuQuantum is not available.

Methods

contract(tensors, indices[, optimize])	Tensor contraction with optional cuQuantum acceleration.
svd(tensor[, chi_max, cutoff])	Compute SVD with optional cuQuantum acceleration.

__init__(config=None)

Initialize cuQuantum backend.

Args:

config: Configuration options (uses defaults if None)

Methods

__init__([config])	Initialize cuQuantum backend.
contract(tensors, indices[, optimize])	Tensor contraction with optional cuQuantum acceleration.
svd(tensor[, chi_max, cutoff])	Compute SVD with optional cuQuantum acceleration.

__init__(config=None)

Initialize cuQuantum backend.

Args:

config: Configuration options (uses defaults if None)

svd(tensor, chi_max=None, cutoff=1e-14)

Compute SVD with optional cuQuantum acceleration.

Args:

tensor: Input tensor (2D after reshaping) chi_max: Maximum bond dimension (truncation) cutoff: Singular value cutoff threshold

Returns:

U, S, Vdagger tensors

contract(tensors, indices, optimize='auto')

Tensor contraction with optional cuQuantum acceleration.

Args:

tensors: List of tensors to contract indices: Einsum-style index notation optimize: Contraction path optimization strategy

Returns:

Contracted tensor

__del__()

Cleanup cuQuantum resources