• Home
  • Features
  • Pricing
  • Docs
  • Announcements
  • Sign In

f-dangel / curvlinops / 22975243719
91%

Build:
DEFAULT BRANCH: main
Ran 11 Mar 2026 09:36PM UTC
Jobs 1
Files 49
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst

11 Mar 2026 09:25PM UTC coverage: 91.197% (-0.01%) from 91.208%
22975243719

push

github

web-flow
[REF] Reduce KFAC/EKFAC test parametrization (#260)

* [REF] Reduce KFAC/EKFAC test parametrization and input sizes

Strip redundant exclude/shuffle/separate_weight_and_bias/check_deterministic
axes from property tests (trace, det, logdet, frobenius_norm), inverse tests,
and forward-only Fisher tests. These axes test parameter handling logic that
is already covered by the core correctness tests (test_kfac_type2 etc.).

Also reduce spatial dimensions for Conv2d test cases (20x20 -> 8x8) and
UnetModel tests (32x32 -> 8x8) to cut backpropagation cost in type-2 Fisher.

Reduces test count from ~8300 to ~2100 and wall-clock time from ~7min to
~4min20s.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* [REF] Strip exclude/shuffle/sep_bias parametrization from MC tests

MC tests verify convergence of Monte Carlo samples to type-2 Fisher,
not canonical space converter correctness. The exclude/shuffle/sep_bias
axes are already covered by the type-2 exactness tests.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* [REF] Strip exclude/shuffle/sep_bias from remaining non-type2 tests

Canonical space converter coverage (exclude, shuffle,
separate_weight_and_bias) is now consolidated into the four type-2
exactness tests: kfac/ekfac × standard/weight_sharing.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* [DOC] Add changelog entry for test parametrization reduction

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* [DOC] Update changelog PR link to #260

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* [REF] Remove redundant kwargs from property and closer-to-exact tests

Add defaults (separate_weight_and_bias=True, check_deterministic=False)
to _test_property and move separate_weight_and_bias to a defaulted kwarg
in _test_ekfac_closer_to_exact_than_kfac. Fix docstring typo (THe→The).

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* [CI] Parallelize tests with pytest-xdist on GitHub runner

Install pyte... (continued)

2362 of 2590 relevant lines covered (91.2%)

0.91 hits per line

Uncovered Existing Lines

Lines Coverage ∆ File
1
94.39
0.05% curvlinops/computers/io_collector/collector.py
1
98.15
-1.85% curvlinops/kfac.py
2
95.12
-2.1% curvlinops/computers/io_collector/conv.py
2
92.47
-1.57% curvlinops/utils.py
7
94.16
0.53% curvlinops/computers/kfac.py
9
91.34
-0.4% curvlinops/kfac_utils.py
16
86.44
-6.24% curvlinops/computers/io_collector/linear.py
Jobs
ID Job ID Ran Files Coverage
1 22975243719.1 11 Mar 2026 09:36PM UTC 49
91.2
GitHub Action Run
Source Files on build 22975243719
  • Tree
  • List 49
  • Changed 8
  • Source Changed 0
  • Coverage Changed 8
Coverage ∆ File Lines Relevant Covered Missed Hits/Line
  • Back to Repo
  • 9a3033c2 on github
  • Prev Build on main (#22869106971)
  • Delete
STATUS · Troubleshooting · Open an Issue · Sales · Support · CAREERS · ENTERPRISE · START FREE · SCHEDULE DEMO
ANNOUNCEMENTS · TWITTER · TOS & SLA · Supported CI Services · What's a CI service? · Automated Testing

© 2026 Coveralls, Inc