| FazBrowse GitHub Viewer | Trending | | Home |
| Tools: [Download Repo ZIP] [Original HTTPS Page] |
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
- isunitspace: require dim(V) == 1 for GenericUnit sectors (#537) - GradedSpace ⊕/supremum: check unit homogeneity of the result (#538) - isconj(::ComplexSpace): return isdual(V) instead of always true (#539) - multi_associator: return a vector, not a scalar, on early-exit for GenericFusion (#540) - split(f, 0): use leftunit(f.coupled) instead of indexing an empty uncoupled tuple (#541) - repartition: return a Pair in the identity branch, matching every other branch (#542) - Mooncake scalar_pullback: accumulate into the tangent instead of overwriting it (#543) - rand/randn/randexp/randisometry(rng, T, space): fix one(domain) typo (#544) - pinv(::DiagonalTensorMap): fix inverted atol/rtol defaulting and empty-tensor throw (#545) - t1 / t2: promote to a float scalartype, matching t1 \ t2 (#546) Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
- split(f, 0): also guard innerlines_extended construction, which still indexed the empty uncoupled tuple for a 0-leg tree - pinv(::DiagonalTensorMap): use eps (not sqrt(eps)) for the default rtol, matching _default_rtol's convention and dense LinearAlgebra.pinv Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
- pinv(::DiagonalTensorMap): reuse _default_rtol instead of duplicating its formula - Add regression tests for split(f, 0) on a genuine 0-leg tree and for multi_associator's early-exit branch on a GenericFusion sector Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
|
This is a PR with the release, along with a number of small fixes for issues that got uncovered by doing a generic 🤖-audit. I'm still running the MPSKit testsuite locally to make sure things are now settled correctly, but after that confirmation I think this is ready to release. |
Sorry, something went wrong.
Codecov Report❌ Patch coverage is 74.28571% with 9 lines in your changes missing coverage. Please review.
|
Sorry, something went wrong.
| Back | FazBrowse Home | New Git URL |
TensorKit.jl v0.17.2
A performance-and-correctness patch release fixing an up-to-80x adjoint permute! slowdown, a @planar contraction bug, and a batch of small bugs found in a pre-release audit.
Highlights
Closes #537, #538, #539, #540, #541, #542, #543, #544, #545, #546
Full Changelog
See CHANGELOG for the complete list of changes.