cupynumeric.linalg.eig#

cupynumeric.linalg.eig( a: ndarray, ) → tuple[ndarray, ...]#

Compute the eigenvalues and right eigenvectors of a square array.

Parameters:

a ((..., M, M) array_like) – Matrices for which the eigenvalues and right eigenvectors will be computed, at least dimension 2.

Returns:

eigenvalues ((…, M) array_like) – The eigenvalues, each repeated according to its multiplicity.
eigenvectors ((…, M, M) array) – The normalized (unit “length”) eigenvectors, such that the column eigenvectors[:,i] is the eigenvector corresponding to the eigenvalue eigenvalues[i].

Raises:

LinAlgError – If the eigenvalue computation does not converge.

Notes

Unlike NumPy, cuPyNumeric always returns complex-dtype results, even if the imaginary part is zero.

Multi-GPU/CPU usage is limited to data parallel matrix-wise batching.