Temporary fix of error mitigation routine if QIBO_PLATFORM is not set

[stavros11]

The error mitigation routine relies on the GlobalBackend in a few places. This causes issues when the QIBO_PLATFORM environment variable is not properly set. Usually we do not see these issues because this environment variable is set in the slurm prologue, however this is not the case for the qw5q_platinum node.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 97.35%. Comparing base (293e2a9) to head (53a94ca).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1063      +/-   ##
==========================================
+ Coverage   97.32%   97.35%   +0.03%     
==========================================
  Files         124      124              
  Lines        9893     9894       +1     
==========================================
+ Hits         9628     9632       +4     
+ Misses        265      262       -3     

Flag	Coverage Δ
unittests	97.35% <100.00%> (+0.03%)	⬆️

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

Files with missing lines	Coverage Δ
src/qibocal/protocols/readout_mitigation_matrix.py	100.00% <100.00%> (+2.50%)	⬆️

... and 2 files with indirect coverage changes

[alecandido]

Wait: if you directly use QibolabBackend [*], then $QIBO_PLAFORM should be irrelevant. Isn't it?

If that's the case, this should be sufficient

Suggested change

	os.environ["QIBO_PLATFORM"] = platform.name
	backend = QibolabBackend(platform)
	backend = QibolabBackend(platform)

[*]: which is a great idea - until we'll reintroduce support in Qibocal for backends other than Qibolab #913 (but that's a separate issue, to tackle later on)

[stavros11]

Wait: if you directly use QibolabBackend [*], then $QIBO_PLAFORM should be irrelevant. Isn't it?

That is also what I initially expected, however it seems that another instance of GlobalBackend is instantiated within the

qibocal/src/qibocal/protocols/readout_mitigation_matrix.py

Line 81 in 53a94ca

_, results = execute_transpiled_circuit(

call which causes a failure if the environment variable is not properly set (which is the case if the slurm prologue is wrong). From the error trace, it seemed that this second GlobalBackend is related to the transpilation process, but I did not investigate in detail.

In any case, this PR was just a quick hotfix for something that @Luca-Ben-Herrmann was running, caused by the wrong slurm prologue of qw5q_platinum partition (which has also been fixed). Ideally, I would like to make all routines independent of the GlobalBackend to avoid such issues, but I am not sure how much of rework that requires.

[*]: which is a great idea - until we'll reintroduce support in Qibocal for backends other than Qibolab #913 (but that's a separate issue, to tackle later on)

I guess hard-coding QibolabBackend here is also not optimal as it would not work for simulation, which I am not sure if is relevant for that particular routine. Anyway, this PR is mostly to raise these issues, maybe we don't even need to merge as it is, unless you are all fine with restricting with qibolab for now and introducing other backends later, as you are proposing.

[alecandido]

I guess hard-coding QibolabBackend here is also not optimal as it would not work for simulation,

As said, I would not worry too much about this. Not yet.

call which causes a failure if the environment variable is not properly set (which is the case if the slurm prologue is wrong). From the error trace, it seemed that this second GlobalBackend is related to the transpilation process, but I did not investigate in detail.

Actually, I failed to find it, since in that call the backend you're instantiating is passed in (and transitively passed down into the called transpile_circuits() function as well).