Update to qibolab 0.2

[stavros11]

I am starting this by squashing the commits of #809, because many of them was keeping up-to-date with the many API changes during the qibolab 0.2 development and these are not very relevant for the final commit history.

The update requires roughly the following steps:

• Update the _acquisition in all routines (list of routines below).
• Update the methods that update the Platform after each routine. This should mainly affect the update.py file and not the routines themselves.
• Deal with parameters that are no longer in the qibolab Platform. These are not required for execution, but usually calculated/updated by qibocal (T1, T2, fidelities, etc.). Adding calibration parameters to qibocal #1021
• Fix pylint and tests. Most like this has to come after the first two (or at least the first), because currently even just the PulseSequence manipulation is broken almost everywhere.
• Fix any direct references to qibolab, which are based on the 0.1 API, in the docs.

This update should not affect the qibocal interface. I called the branch 0.2 because I assumed that we will eventually release it as such, in order to maintain agreement between qibolab-qibocal versions, but this is not strictly required if we don't break the qibocal interface (we could keep going with qibocal 0.1.*). I could change the branch name if it is confusing.

@Edoardo-Pedicillo mentioned that a potential qibocal refactoring was being discussed, however I would suggest to keep the two independent unless there is an important reason not to. If you want the "refactored qibocal" to support qibolab 0.1, you need to do this refactor first, however based on the discussed priorities, I think we should go ahead with 0.2 and forget 0.1 as soon as possible.

To elaborate a bit more on the above points:
(1) requires the most work, however it is also the tedious kind of work, so there should be no significant blocking points. I will add below a list of the routines, to split them among the people involved and keep track what is updated. In every routine we need to update the PulseSequence creation, how the native pulses are obtained from the Platform, how Sweepers are defined and potentially how results are obtained and post-processed. Qibolab devs could potentially give a tutorial with more details. Some routines are already updated (see below) so they could serve as template.

(2) may be a bit trickier because, unlike (1), we don't have a POC yet and qibolab 0.2 is not very user-friendly when it comes to Platform parameter updates. I am planning to open another PR (on top of this) with a POC to discuss and once this is finalized we can propagate to all updates. The good news is that this can also progress independently of point 1.

(3) is not strictly required for things to work and the simplest solution is to just drop all the updates of such parameters. If users want to report them, they can refer to the corresponding reports, until we come up with a better way to store them. However, alternative solutions are welcome (@andrea-pasquale @alecandido ).

Points 4 and 5 should come naturally as 1 and 2 are completed.

List of routines to be updated (taken from `__init__` ):

• "allxy",
  - [ ] "allxy_drag_pulse_tuning", (because it could be just a qibocal script)
• "single_shot_classification",
• "spin_echo" (saving single shots not tested),
• "spin_echo_signal" (saving single shots not tested),
• "t1" (saving single shots not tested),
• "t1_sequences", Porting some protocols to qibolab 0.2 #1036 (we can drop it since all drivers in 0.2 should support sweepers over duration)
• "t1_signal" (saving single shots not tested),
• "t2", Porting some protocols to qibolab 0.2 #1036
• "t2_sequences", Porting some protocols to qibolab 0.2 #1036 (we can drop it since all drivers in 0.2 should support sweepers over duration)
• "t2_signal", Porting some protocols to qibolab 0.2 #1036
• "zeno", Porting some protocols to qibolab 0.2 #1036
• "zeno_signal", Porting some protocols to qibolab 0.2 #1036 (we can drop it since the version with probability should be enough)
• "coupler_chevron" Restore coupler protocols #1039 ,
• "coupler_qubit_spectroscopy" Restore coupler protocols #1039 ,
• "coupler_resonator_spectroscopy" Restore coupler protocols #1039 ,
• "dispersive_shift", Porting some protocols to qibolab 0.2 #1036
• "dispersive_shift_qutrit", Porting some protocols to qibolab 0.2 #1036
• "drag_tuning", Porting some protocols to qibolab 0.2 #1036
• "fast_reset", (we can drop it since no driver in 0.2 supports fast reset)
• "flipping", Porting some protocols to qibolab 0.2 #1036
• "flipping_signal", Porting some protocols to qibolab 0.2 #1036
• "avoided_crossing", (we can drop it since it will be replaced Add experiment for qubit frequency vs flux amplitude  #1037)
• "qubit_crosstalk",Porting some protocols to qibolab 0.2 #1036
• "qubit_flux" (see Update to qibolab 0.2 #990 (comment)),
• "qubit_flux_tracking", (we can drop it temporarily since it performs correction of the order of kHz)
• "resonator_crosstalk", (we can drop it since we can evaluate the crosstalk matrix just with qubit crosstalk)
• "resonator_flux" (see Update to qibolab 0.2 #990 (comment)),
• "qubit_spectroscopy",
• "qubit_spectroscopy_ef", Porting some protocols to qibolab 0.2 #1036
• "qutrit_classification", Porting some protocols to qibolab 0.2 #1036
• "rabi_amplitude",
• "rabi_amplitude_signal",
• "rabi_length",
• "rabi_amplitude_ef", Porting some protocols to qibolab 0.2 #1036
• "rabi_length_sequences", (to be dropped see other sequences experiments)
• "rabi_length_signal",
• "ramsey" (unrolling not tested),
• "ramsey_signal" (unrolling not tested),
• "filtered_rb", Porting some protocols to qibolab 0.2 #1036
• "standard_rb", Porting some protocols to qibolab 0.2 #1036
• "readout_characterization", Porting some protocols to qibolab 0.2 #1036
• "readout_mitigation_matrix", Porting some protocols to qibolab 0.2 #1036
• "resonator_amplitude", Porting some protocols to qibolab 0.2 #1036
• "resonator_frequency", Porting some protocols to qibolab 0.2 #1036 (to be dropped since to evaluate the best frequency the dispersive shift protocol should enough)
• "twpa_frequency", (to be converted into a script)
• "twpa_frequency_power", (to be converted into a script)
• "twpa_frequency_snr", (to be converted into a script)
• "twpa_power", (to be converted into a script)
• "twpa_power_snr", (to be converted into a script)
• "resonator_punchout",
• "resonator_punchout_attenuation", (to be reintroduced after qblox driver is ready for 0.2)
• "resonator_spectroscopy",
• "calibrate_state_discrimination", Porting some protocols to qibolab 0.2 #1036
• "time_of_flight_readout", Porting some protocols to qibolab 0.2 #1036
• "chevron",
• "chevron_signal",
• "chsh_circuits", (we will add it later)
• "chsh_pulses", (we will add it later)
• "correct_virtual_z_phases",
• "correct_virtual_z_phases_signal", (to be removed)
• "state_tomography", Porting some protocols to qibolab 0.2 #1036
• "allxy_resonator_depletion_tuning", (to be discussed)
• "two_qubit_state_tomography", Porting some protocols to qibolab 0.2 #1036
• "qubit_power_spectroscopy", Porting some protocols to qibolab 0.2 #1036
• "rabi_amplitude_frequency", Porting some protocols to qibolab 0.2 #1036
• "rabi_amplitude_frequency_signal", Porting some protocols to qibolab 0.2 #1036
• "rabi_length_frequency", Porting some protocols to qibolab 0.2 #1036
• "rabi_length_frequency_signal", Porting some protocols to qibolab 0.2 #1036
• "standard_rb_2q", Porting some protocols to qibolab 0.2 #1036
• "standard_rb_2q_inter", Porting some protocols to qibolab 0.2 #1036
• "optimize_two_qubit_gate", Porting some protocols to qibolab 0.2 #1036

[stavros11]

Note that testing updated routines on hardware also requires a working platform that complies with the qibolab 0.2 interface. As of now the only available is qiboteam/qibolab_platforms_qrc#153.

[stavros11]

In 4154ea9 I updated the resonator and qubit flux routines. Here are some examples on

• qw11q: http://login.qrccluster.com:9000/zCADyOH9REGlihNzbs5wVQ==, http://login.qrccluster.com:9000/thAfEvfURuK21lagxVlkMQ==
• qw5q_platinum: http://login.qrccluster.com:9000/wf3OKDQuTDiOxUVCoOrNag==, http://login.qrccluster.com:9000/-dW4I4e6SgeToDRh1IxNUw==

As mentioned elsewhere, these routines are making use of some parameters that are no longer in the qibocal platform, such as the crosstalk matrix and bare resonator frequency for the fit of resonator flux and the anharmonicity for the qubit flux of the 02 state. I have temporarily disabled/set dummy values in the related points, however we should probably find a solution on how to deal with these parameters before merging this.

[stavros11]

I have updated most routines that are needed for basic calibration of a chip from scratch (see list in first post). I have tested these on qw11q and they seem to work. If anyone tries to use this and encounters any issues or if some important routine is missing let us know. For TII, it should work with the platforms on qibolab_platforms_qrc main and the latest qibolab main (qw11q should also work with 0.2.0 release).

Note that the spin echo routine has been changed to use sweepers, instead of unrolling, since now there is no discrepancy with how delay durations are swept and they should be uniformly supported by all drivers in qibolab 0.2.

For the rest of routines I will probably need some help. I will now shift focus on #996 to implement the updates using the method discussed.

[alecandido]

Whenever someone can take a look, please consider merging/rebasing on main to solve the conflicts (even discarding everything, if they do not apply to the new version).

[stavros11]

I have done some comparisons between 0.1 and 0.2 (using this branch) on qw11q:

• Single shot: http://login.qrccluster.com:9000/6Iisf1W1T1KrStCypqIR-Q==, http://login.qrccluster.com:9000/iQCjykMTSOOHd8Jq7knHtA==
• Rabi amplitude: http://login.qrccluster.com:9000/wAeqh9o8Q0ebEfbTslCzZQ==/, http://login.qrccluster.com:9000/2P5wVGx6TV6k0Aw9kAcfUQ==/
• T1: http://login.qrccluster.com:9000/fJ6Qi3ABTruVIOgCe4d78w==, http://login.qrccluster.com:9000/3Sd_pNOwSCes-qIuwOQgHA==
• Ramsey detuned: http://login.qrccluster.com:9000/OR3tXHH_T62xSM6orw8AaQ==, http://login.qrccluster.com:9000/1-QRamA7TJSiMTpqaIbYGg==
• AllXY: http://login.qrccluster.com:9000/C9djroZ8QJWrDQOLtftbuQ==/, http://login.qrccluster.com:9000/TyMHnbHoRpSsY2Iic3n_zw==
• Chevron: http://login.qrccluster.com:9000/uxAuiz8wTzC7b0W1D9fS1Q==, http://login.qrccluster.com:9000/aMzhVDeoQ722XyU9jbthZw==
• CZ phases: http://login.qrccluster.com:9000/jgnsuJs9Sp-Bh8fwPVWbeQ==, http://login.qrccluster.com:9000/jvzUSBsXRaOk0tuxTbM7dg==
• CHSH: http://login.qrccluster.com:9000/jCZJNfrHQiCrlrOC8n342Q==, http://login.qrccluster.com:9000/tjSYbmTSS6O4d7rl3TtlNg==/

These tests were done using the qw11q platform from the 0.1 branch of qibolab and the corresponding converted platform for 0.2 (which can be found in qiboteam/qibolab_platforms_qrc#205).

Most routines I tried look similar among the two versions. There are known discrepancies on

• Rabi length with interpolated sweepers (to be fixed in driver - done in Fix duration interpolated sweeper for QM qibolab#1115),
• Ramsey detuned (see above - not investigated yet),
• CHSH (partly known issue that will be fixed in the driver). Also, this routine is not part of this PR so I would postpone the discussion for Update CHSH pulses routine to 0.2 #1049.

Note that these tests are an in a small subset of routines. However, I would still ask others to try using 0.2 so that we can identify other problems, if there are any. Maybe merging this could be a good step towards that direction.

[andrea-pasquale]

Thanks for the detailed comparison @stavros11!
Something else that was missing which I added in d7745ae is the update for the DRAG pulse.
I'm intrigued by the Ramsey, I will try to have a look.

Note that these tests are an in a small subset of routines. However, I would still ask others to try using 0.2 so that we can identify other problems, if there are any. Maybe merging this could be a good step towards that direction.

Generally I agree with this, since if we merge people (at least developers) will be more or less forced to use this version :)