You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The "divide_and_conquer" SVD algorithm (LAPACK's gesdd) is currently the default. It is a good balance of speed, accuracy, and reliability. However, it can occasionally fail to converge. In those cases, I think we should catch it and automatically try gesvd (which would be a simple modification to the logic here: https://github.com/ITensor/NDTensors.jl/blob/v0.1.28/src/linearalgebra.jl#L137-L145).
The "divide_and_conquer" SVD algorithm (LAPACK's
gesdd
) is currently the default. It is a good balance of speed, accuracy, and reliability. However, it can occasionally fail to converge. In those cases, I think we should catch it and automatically trygesvd
(which would be a simple modification to the logic here: https://github.com/ITensor/NDTensors.jl/blob/v0.1.28/src/linearalgebra.jl#L137-L145).Similar design in the C++ ITensor code: ITensor/ITensor#386
The text was updated successfully, but these errors were encountered: