SVDInfo#

class cuquantum.tensornet.tensor.SVDInfo(
reduced_extent: int,
full_extent: int,
discarded_weight: float,
algorithm: str,
gesvdj_residual: float | None = None,
gesvdj_sweeps: int | None = None,
gesvdp_err_sigma: float | None = None,
)[source]#

A data class for holding information regarding SVD truncation at runtime.

full_extent#

The total number of singular values after matricization (before truncation).

Type:

int

reduced_extent#

The number of remaining singular values after truncation.

Type:

int

discarded_weight#

The discarded weight for the truncation.

Type:

float

algorithm#

The algorithm used in the SVD execution.

Type:

str

gesvdj_residual#

The residual for full gesvdj execution.

Type:

float | None

gesvdj_sweeps#

The number of iterations used in the gesvdj execution.

Type:

int | None

gesvdp_err_sigma#

The error sigma in the gesvdp execution.

Type:

float | None

Note

When the SVD algorithm is set to "gesvdr" with fixed extent truncation enabled in cuquantum.cutensornet.tensor.SVDMethod, the discarded weight will not be computed.