Accelerate the CI pipeline

[michaelosthege]

Right now the CI is taking ~1.5 h.

However, we can split up jobs in the job matrix to parallelize more, and also investigate whether we're unnecessarily running things multiple times (e.g. different Python versions).

Maybe the scripts/check_all_tests_are_covered.py from PyMC can be adapted to quickly diagnose redundacies?

[danhphan]

Hi @michaelosthege, let me work on this one :) Thanks!

[ricardoV94]

I think GitHub CI is running the tests marked as slow, one failed recently. That could explain why they take so long

[danhphan]

Yes, good points, I will look into it in details this weekend. Sorry for the delay :)

[ricardoV94]

The runslow seems to be on purpose:

pytensor/.github/workflows/test.yml

Line 135 in 19e1a98

python -m pytest -x -r A --verbose --runslow --cov=pytensor/ --cov-report=xml:coverage/coverage-${MATRIX_ID}.xml --no-cov-on-fail $PART

[danhphan]

Yes, the --runslow is used to run pytest on tests with @pytest.mark.slow decorator. All these slow tests are currently run on both python version 3.7 and 3.9. Running these slow tests on only one python version (3.9) should reduce CI running time.

I'm also looking into detail to see how to break these tests into smaller parts similar to PyMC workflows tests.yml

[danhphan]

It seems that most slow tests are from tests/tensor/test_math.py and tests/tensor/test_basic.py, so I'm going to separate these two test files.

[ricardoV94]

Closed via #176