Using custom dummy platform

[andrea-pasquale]

In this PR I'm attempting to define a custom "dummy" platform in Qibocal (which is currently called fake but if there are better proposal I'm happy to change the name). The motivation started from #1099 where I've noticed that the dummy was missing a local oscillator making it impossible to run the integration weights protocol with the dummy available in Qibolab.

In general I believe that this could be beneficial to further remove some nested dependencies between Qibolab and Qibocal.
For example currently the calibration parameters for dummy where stored in Qibocal while the parameters.json is currently in Qibolab.
The new fake platform only has 2 qubits with a coupler, since I don't believe that for testing purposes we need more than that.

Currently the new platform is added at import time by adding the corresponding directly to QIBOLAB_PLAFORMS.

• update documentation

EDIT: after some internal discussions we have decided to rename fake to mock

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 97.77%. Comparing base (fc89963) to head (e7b2996).
Report is 2184 commits behind head on kernels.

Additional details and impacted files

@@             Coverage Diff             @@
##           kernels    #1101      +/-   ##
===========================================
+ Coverage    96.05%   97.77%   +1.71%     
===========================================
  Files          107       99       -8     
  Lines         7432     8088     +656     
===========================================
+ Hits          7139     7908     +769     
+ Misses         293      180     -113     

Flag	Coverage Δ
unittests	97.77% <100.00%> (+1.71%)	⬆️

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

Files with missing lines	Coverage Δ
src/qibocal/__init__.py	100.00% <100.00%> (ø)
src/qibocal/auto/execute.py	95.38% <100.00%> (-3.37%)	⬇️
src/qibocal/auto/runcard.py	97.50% <100.00%> (-0.50%)	⬇️
src/qibocal/calibration/platform.py	100.00% <100.00%> (ø)

... and 34 files with indirect coverage changes

[andrea-pasquale]

With the latest changes I've defined the mockplatform which generate random numbers only in tests.
For this reason I had to drop the DEFAULT_EXECUTOR which was using dummy as discussed in #926.
Similarly to what is already happening in #1103 I've added a platforms folder in qibocal where I've added a symlink to mock.

[alecandido]

Thanks @andrea-pasquale, everything looks good.

I'm only providing minor suggestions.

[alecandido]

This is only needed for dummy, to avoid breaking the API (since we introduced it in Qibolab after v0.2.0). However, you can just drop this function, and rename the former create(), unless you import this file directly. create_platform() will detect it and provide the Platform.load() call on its own.

[andrea-pasquale]

I tried to do exactly as you said but when running test I'm getting an error RuntimeError: Cannot establish connection to instrument dummy. Error captured: ''dict' object has no attribute 'connect''.
Given that, if I am not wrong, this feature is only available in qibolab main and is not available in a release I suggest to open an issue and skip this comment for now.

[alecandido]

Given that, if I am not wrong, this feature is only available in qibolab main and is not available in a release I suggest to open an issue and skip this comment for now.

Actually, it should be available in v0.2.3, since qiboteam/qibolab#1123 was introduced in that release.
However, I fixed something after that release in qiboteam/qibolab#1151, and I was trying to remember whether it was needed sometimes or always...