Open cohenjer opened 1 month ago
Not sure why the pytorch test for constrained_parafac is returning an error, I did not change the contents of the code there... seems to be a type mismatch between floats and integers but I have no idea why it happens.
Attention: Patch coverage is 74.66468%
with 170 lines
in your changes are missing coverage. Please review.
Project coverage is 80.10%. Comparing base (
63fc5c5
) to head (4386510
). Report is 6 commits behind head on main.:exclamation: Current head 4386510 differs from pull request most recent head c553d69. Consider uploading reports for the commit c553d69 to get more accurate results
Files | Patch % | Lines |
---|---|---|
tensorly/solvers/proximal.py | 67.50% | 91 Missing :warning: |
tensorly/solvers/nnls.py | 65.07% | 44 Missing :warning: |
tensorly/tenalg/proximal.py | 0.00% | 29 Missing :warning: |
tensorly/solvers/penalizations.py | 92.00% | 6 Missing :warning: |
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.
This is a copy of the part of #542 related to solvers.
What this PR solves:
What this PR does:
(I left the old proximal file with the modified nnls_hals and fista solvers to show the changes, we can remove it before merging.)
solvers/penalizations.py
with a utility function to process input ridge/sparsity regularization (transform 1d input to list of correct length, avoid no regularization on only some modes which makes the factorization ill-posed).Possible improvements:
Note: If this PR is merged, I will proceed with the nonnegative algorithms improvements in #542.