Add fit to Chevron

[andrea-pasquale]

This PR implements a post-processing operation for the chevron protocol in order to find the best point for the CZ.

The post-processing works as follows:

1. Perform FFT for each amplitude value
2. Find amplitude at which the freq has the minimum value
3. Compute duration by fitting cosinusoidal function at the point of minimum freq

This is the result on hardware

TODO:

• Add table

Checklist:

• Reviewers confirm new code works as expected.
• Tests are passing.
• Coverage does not decrease.
• Documentation is updated.
• Compatibility with Qibo modules (Please edit this section if the current pull request is not compatible with the following branches).
  • Qibo: master
  • Qibolab: main
  • Qibolab_platforms_qrc: main

[codecov]

Codecov Report

Merging #520 (4077637) into main (6e5b3cf) will increase coverage by 0.04%.
Report is 9 commits behind head on main.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main     #520      +/-   ##
==========================================
+ Coverage   96.42%   96.47%   +0.04%     
==========================================
  Files          66       66              
  Lines        4789     4846      +57     
==========================================
+ Hits         4618     4675      +57     
  Misses        171      171              

Flag	Coverage Δ
unittests	96.47% <100.00%> (+0.04%)	⬆️

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

Files Changed	Coverage
.../characterization/two_qubit_interaction/chevron.py	100.00%
...racterization/two_qubit_interaction/cz_virtualz.py	100.00%
...ls/characterization/two_qubit_interaction/utils.py	100.00%

[igres26]

That is a good choice for fitting. More usage might expose where it fails, but for that we might want it merged.

[alecandido]

Of course, the critical part is the fit (as for the name of the PR), and we might need to iterate a bit.

All the other modifications are fine :)

[alecandido]

I know it's annoying, but the algorithm is definitely not trivial: it would be nice to have a complete description in English (rather than Python) of the whole algorithm.

Maybe we could discuss it before going through the process. Or you can use me as a guinea pig to test the description.

[andrea-pasquale]

Feel free to have a look now.

[stavros11]

Thanks for implementing this. As a test I tried it on the chevron we got with QM on qw5q_gold last February (one of the first in TII) and it gave: amplitude=0.998, duration=28.7 which is acceptable. Here is the final curve fit in the amplitude slice (done quickly in a notebook):

We should address @alecandido's comments related to code, such as the redundant qubit variables, before merging, but I agree with @igres26 that in order to evaluate how good the fitting is we need to try it in practice. So I would not block this PR for this, particularly because now we don't have any QPUs to check. We can always improve the fitting later, as we did with other routines.

[andrea-pasquale]

@alecandido I addressed all your comments but the improvements on the median for which I opened #535.
If everything else looks good for you I will merge soon given that I already have 2 approvs

[alecandido]

Ok, since we postponed the fit improvements, the rest is pretty much innocuous, and the implementation is fine.

Let's merge :)

[andrea-pasquale]

Ok, since we postponed the fit improvements, the rest is pretty much innocuous, and the implementation is fine.

Let's merge :)

I saw that you approved the comment above so I merged...
Thanks for the feedback.