[feature request] [pytorch] Truncated SVD #8049
Description
Facebook published reference code of fast NumPy randomized SVD (based on power iteration): https://github.com/facebook/fbpca/blob/master/fbpca.py#L1503
Only top eigenvectors are useful e.g. for spectral clustering, it would be nice to have it supported for CUDA.
Metadata
Metadata
Assignees
Labels
A request for a proper, new feature.Related to torch.cuda, and CUDA support in generalRelated to numpy support, and also numpy compatibility of our operatorsNot as important as medium or high priority tasks, but we will work on these.This issue has been looked at a team member, and triaged and prioritized into an appropriate module