QPU performance bot
scarrazza opened this issue · comments
We need a performance measurement tool which will extract parameters after execution on QPU for a given platform runcard. This program should store at least T1, T2 and fidelities.
Actually @Jacfomg already coded something similar for the selfhosted in qiboteam/qibolab_platforms_qrc#73
From qibocal this is achieved by using a specific action runcard, which is then executed using qq auto.
From a practical perspective should we just include this runcard here or do you want something more @scarrazza ?
Yes, this works, please add the runcard here.
This will work for T1, T2 and assigment fidelity. For the RB I still need to fix a couple things on the routine and find some nice parameters to get a fidelity with a short scan.
Yes, this works, please add the runcard here.
Concerning the runcard I need to specify already the platform and the qubits...
If we want something which is platform agnostic we may want to add the possibility to pass the platform and the qubits directly from the cli (instead of putting them in the runcard).
Something like:
qq auto <runcard.yml> --platform qw5q_gold --qubits [0,1,2,3] -o <output_folder>Let me know what you think
@stavros11 this could also be useful to simplify the selfhosted workflow I guess
Alternatively, we could have a qq qpu-monitor which automatically uses your predefined runcard and delegates to qibolab the decision of the platform name.
Alternatively, we could have a
qq qpu-monitorwhich automatically uses your predefined runcard and delegates to qibolab the decision of the platform name.
I see. In order to change the platform this program should be able to edit yaml files or dictionaries. Or we need to put a placeholder for the platform
@andrea-pasquale please do whatever is simpler and faster, from my side, I have just coded a strategy to display the information starting from the output of qq.