Fix tests for cupy and cuquantum backends

[renatomello]

Checklist:

• Reviewers confirm new code works as expected.
• Tests are passing.
• Coverage does not decrease.
• Documentation is updated.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 99.81%. Comparing base (81c173b) to head (7e04b68).
Report is 6 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #1238   +/-   ##
=======================================
  Coverage   99.81%   99.81%           
=======================================
  Files          72       72           
  Lines       10520    10528    +8     
=======================================
+ Hits        10501    10509    +8     
  Misses         19       19           

Flag	Coverage Δ
unittests	99.81% <100.00%> (+<0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[alecandido]

Thanks @renatomello, I see that it is not completed yet, but it already seems a simplification.

As soon as you complete, I'll be glad to approve!

[renatomello]

Some tests in tests/test_models_error_mitigation.py fail on GPU only if the tests are ran in sequence. If the tests are ran individually, the tests pass. This is probably an issue with GlobalBackend being set up by a previous test

[renatomello]

@MatteoRobbiati the majority of DBI tests are failing on GPU. It seems that it is because self.backend is not setup properly in the DoubleBracketIteration class. That would also explain the need for all those self.backend.to_numpy calls inside the class.

[alecandido]

Some tests in tests/test_models_error_mitigation.py fail on GPU only if the tests are ran in sequence. If the tests are ran individually, the tests pass. This is probably an issue with GlobalBackend being set up by a previous test

If you are receiving the global backend from a fixture, as in here:

qibo/tests/conftest.py

Lines 57 to 59 in 31bbbb4

	@pytest.fixture
	def backend(backend_name):
	yield get_backend(backend_name)

you could add some cleanup code after the yield.

[andrea-pasquale]

Some tests in tests/test_models_error_mitigation.py fail on GPU only if the tests are ran in sequence. If the tests are ran individually, the tests pass. This is probably an issue with GlobalBackend being set up by a previous test

I also saw the same behavior some time ago, very weird.

@MatteoRobbiati the majority of DBI tests are failing on GPU. It seems that it is because self.backend is not setup properly in the DoubleBracketIteration class. That would also explain the need for all those self.backend.to_numpy calls inside the class.

You are right, given that we don't tests on GPU with the CI sometimes we forget. I can also have a look if you want @MatteoRobbiati.

[renatomello]

It's better to move forward with this PR only after #1202 is merged

[renatomello]

All tests should be passing on GPUs now. Someone else should test locally too. Thanks @andrea-pasquale for making DBI tests faster and GPU compatible.