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
As of my knowledge cutoff in September 2021, PyTorch's SVD implementation had some limitations. It uses a batched SVD routine, which can cause problems with larger matrices due to GPU memory limitations. PyTorch's SVD also may return singular values that are not sorted, which is non-standard behavior that can surprise users.
The text was updated successfully, but these errors were encountered:
On the other hand, CuPy uses a more standard, non-batched routine for SVD, which is less likely to run into memory issues for large matrices. Also, CuPy's SVD returns singular values in descending order, which is the typical behavior expected by users.
Accordiong to ChatGPT
As of my knowledge cutoff in September 2021, PyTorch's SVD implementation had some limitations. It uses a batched SVD routine, which can cause problems with larger matrices due to GPU memory limitations. PyTorch's SVD also may return singular values that are not sorted, which is non-standard behavior that can surprise users.
The text was updated successfully, but these errors were encountered: