Closed kmdalton closed 1 year ago
Merging #139 (c469268) into main (42f8c96) will decrease coverage by
9.39%
. The diff coverage isn/a
.
@@ Coverage Diff @@
## main #139 +/- ##
==========================================
- Coverage 79.18% 69.79% -9.39%
==========================================
Files 52 52
Lines 2344 2344
==========================================
- Hits 1856 1636 -220
- Misses 488 708 +220
Flag | Coverage Δ | |
---|---|---|
unittests | 69.79% <ø> (-9.39%) |
:arrow_down: |
Flags with carried forward coverage won't be shown. Click here to find out more.
see 6 files with indirect coverage changes
:mega: We’re building smart automated test selection to slash your CI/CD build times. Learn more
CI was failing as of the tensorflow 2.14 release. The underlying cause was a memory leak in the test_cli.py
submodule. Tracing this back, it seems to be an inherent issue with tensorflow a la this issue. This issue was always present, but it didn't become a problem until 2.14 started to allocate more memory than previous versions. Running careless in a separate process for the purposes of testing bypasses the memory leak and forces tf to clean up after the process exits. So, from now on, I'll be using subprocess to call the CLI during testing. It's not the most satisfying solution and will probably make debugging harder by dint of less specific error messages. However, it is required to stop the memory leak in order to stay within the ram limits of github's default runner.
I'm closing this PR in favor of #140, because this solution breaks codecov.
Update the versions of actions in build and publish workflows.