-
Notifications
You must be signed in to change notification settings - Fork 5.7k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Fix Frontend Failing Test: torch - blas_and_lapack_ops.svd #28769
Comments
Now the failing tests shows that the ground truth is somehow thought to be all zeros, while the parameters of some = False, compute = False cause the full_matrices to be computed while zero-filled tensors should be returned for U and V (according to torch documentation https://pytorch.org/docs/stable/generated/torch.svd.html) |
But according to the same documentation, the output should be a tuple of (U, S, V), instead of the current supposed ground truth of (U, V) (as shown here https://github.com/Transpile-AI/ivy/actions/runs/9589504921/job/26443403485#step:6:127). So I believe the current implementations of the torch.svd and the test are both wrong. |
Currently, both svd function in torch frontend have certain behaviors unimplemented, so will start working on that first. |
Also found that the backend implementation of torch.linear_algebra.svd does not produce the namedtuple output of (U,S,Vh) as described by the ivy function svd when compute_uv is false. |
No description provided.
The text was updated successfully, but these errors were encountered: