-
Notifications
You must be signed in to change notification settings - Fork 130
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Fix conservation tracking tests #2716
Conversation
(see these six tests xfailing e.g. in this run: https://github.com/ilastik/ilastik/actions/runs/4698963212/jobs/8331947378 or most recently in https://app.circleci.com/pipelines/github/ilastik/ilastik/1340/workflows/130b9cc3-3761-43b7-bd3a-ea7609e32c4e/jobs/5647) |
Codecov ReportPatch coverage:
Additional details and impacted files@@ Coverage Diff @@
## main #2716 +/- ##
==========================================
+ Coverage 54.35% 54.59% +0.23%
==========================================
Files 519 530 +11
Lines 60863 61425 +562
Branches 8398 8453 +55
==========================================
+ Hits 33081 33532 +451
- Misses 26105 26140 +35
- Partials 1677 1753 +76
☔ View full report in Codecov by Sentry. |
A small tip: Github now supports draft pull requests, so it's not necessary to add |
Unclear why the previous rev (8710dfa) changed the import to `ilastik.__main__` in a bunch of other test files, but in this file went for a now broken `import main`. `ilastik.__main__` seems to work fine here too.
6b5a218
to
5849287
Compare
I've found and fixed the remaining tests that had the same file path issues, but those tests suffer from their own additional problem: They contained a switch that would pass all tests if they are unable to import either I changed the pass-if-missing to xfail-if-missing, and now all of the |
Instead of xfailing, we probably should conditionally skip those tests if a module cannot be imported. |
5849287
to
d1efdf9
Compare
859c97f
to
999d3ec
Compare
I managed to create a new input .ilp that passes (with Gurobi 9.5.2 on my system) Turns out the GUI gives you no way to actually see all existing track labels that have been put in Training, and I only found the ones that weren't visible when I inspected the h5 manually. On CI the tests will be skipped because no Gurobi there :) |
Hi @btbest, sorry this took longer as I had to build the tracking with learning dependencies on apple silicon first (was pushing this off, so thanks for the nudge). Two tests fail for me:
it appears that the outputs cannot be found, can you double check that in a clean repo, you can run these tests on windows? e.g. I get an error wrt to not being able to find the following file:
|
With the old .ilp file, the test errors out in vigra with "RandomForestn::predictProbabilities(): Too few columns in feature matrix". Probably because the file contains old feature names for the 2D convex hull location.
999d3ec
to
025785c
Compare
🤦 I had "updated" the expected output file names to what seemed to be new default file names with I have no idea how I obtained output files with that name pattern at any point, but anyway I've reverted the name change. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
test success confirmed on my machine, too :) sooo good to have this running again!
The six tests in testConservationTrackingHeadless.py have been xfailing for quite a while because file paths were broken. Having fixed that, it turned out also the main function import didn't work as implemented.