-
Notifications
You must be signed in to change notification settings - Fork 27.7k
Cholesky: return info instead of RuntimeError #47608
Copy link
Copy link
Closed
Labels
module: linear algebraIssues related to specialized linear algebra operations in PyTorch; includes matrix multiply matmulIssues related to specialized linear algebra operations in PyTorch; includes matrix multiply matmultriagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module
Metadata
Metadata
Assignees
Labels
module: linear algebraIssues related to specialized linear algebra operations in PyTorch; includes matrix multiply matmulIssues related to specialized linear algebra operations in PyTorch; includes matrix multiply matmultriagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module
🚀 Feature
choleskyraises aRuntimeErrorif the matrix is not (numerically) positive-definite. This makes it impossible to use it with TorchScript (for loop over increasing regularization of try-except with cholesky).Similar to
lu, returninfoto indicate which of the batches matrices were successfully decomposed and which weren't.Motivation
Use
choleskywith TorchScript.Pitch
Return
lu-likeinfoinstead of raising aRuntimeError.cc @vishwakftw @jianyuh @nikitaved @pearu @mruberry @heitorschueroff