Closed sroet closed 1 year ago
Hi! This is the full output from running python -m unittest discover
from the tests/E2ETests
on my laptop with GPU. Many errors sadly.
pytomvolume_vs_numpy.py
and pytom_FullPipeline.py
are not executed because their filename does not start with test_
.I made it foldable because its long, also the formatting is weird due to # in the output.
I tried running autofocus class test after the mpi4py update, and noticed two things:
mpirun -np 2 {self.installdir}/pytom {self.installdir}/classification/auto_focus_classify.py
which should be changed to mpirun -np 2 python {self.installdir}/classification/auto_focus_classify.py
to load pytom.lib.pytom_volume correctly. classification/auto_focus_classify.py
starts with this assertion assert np.__version__ >= '1.7.0'
, which for me crashed because the string '1.22.3'
is not larger than '1.7.0'
of course... So statement should be removed and numpy version dependency added in the setup.I can also make a new pull request for fixing the autofocus test?
I will make another pull request here to incorporate the data from the unittest data repo (zipped). And also fix the alignment test using it
I will make another pull request here to incorporate the data from the unittest data repo (zipped). And also fix the alignment test using it
Turns out it doesnt compress well. When we downloaded the repo zipped, it didnt incorporate the git-lfs files correctly (see: https://github.com/git-lfs/git-lfs/issues/903). So we were just compressing pointers to the actual data (haha).
If I compress the stack of tilt-series data now, it goes from 1.9 Gb to 1.7 Gb. That seems (1) too large to ship with the main repository and (2) exceeds github filesize limit of 100 Mb (without git-lfs). I could circumvent the file size limit by splitting the tilt-series in individual files of ~55 Mb, that can be gzipped to 50 Mb. But that does not change the added data size of the whole repo.
@McHaillet do you know what the intended behavior is of test_Binscripts.pypytom_MyFunctionTest.addProjectDirToParticleList
? Should the test be updated to the current behavior, or should the current test pass (and this the behavior updated)?
@McHaillet do you know what the intended behavior is of
test_Binscripts.pypytom_MyFunctionTest.addProjectDirToParticleList
? Should the test be updated to the current behavior, or should the current test pass (and this the behavior updated)?
What is the error?
I would assume it passes as it should just append a path to a filename.
After an offline discussion with @McHaillet we decided to merge this partial fix and will continue in another PR
This PR will fix all the e2e test on a machine without a gpu