Repository navigation
Keep the cross-approximation bases when the SVD does not converge - #49
Merged
Merged
Conversation
Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01MFB6QLC6geNXiEorQzmJ6r
…converge Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_01MFB6QLC6geNXiEorQzmJ6r
Codecov Report❌ Patch coverage is
📢 Thoughts on this report? Let us know! |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Requested by Matteo · project thread
On the GitHub runners, two
_cross_couplingtests fail on CPU withlinalg.svd: The algorithm failed to converge. This happens on both dependabot PRs (#47 and #48) and repeated on a re-run. Each round the body and coil bases are extended with sampled columns that they already partly span, so the singular values cluster at one and near zero. The same torch 2.15.0+cpu wheel converges on an Intel Xeon here, so the failure depends on the CPU. On the runners it fails even on the small triangular factor after QR._left_singular_vectorsreduces the matrix with QR and takes the SVD of the triangular factor. If that SVD raises, it takes the vectors from the eigendecomposition of the factor's Gram matrix instead, strongest first. The two agree on every vector whose singular value stands above √ε, and the core fit in_fitted_coredrops everything below that anyway.Tests, on CPU and CUDA: an extended basis keeps its span in its strongest singular vectors on both paths, with the SVD made to fail for the second. Locally, all of
tests/test_implicit.pyalso passes with the eigendecomposition path forced on. Results:format_and_lint.shclean, andtests/test_implicit.pygives 13 passed (13 CUDA skipped).🤖 Generated with Claude Code
https://claude.ai/code/session_01MFB6QLC6geNXiEorQzmJ6r