Optimizers for noisy quantum devices (scikit-quant) #1240
Conversation
There was a problem hiding this comment.
Is this intended to be an optional install that would not be installed normally when the user installs Aqua? Is there some reason for it to be optional?
If the intent is for it to be optional then the imports and unit tests need to handle the case where its not installed as is done in the various other optionally installed parts.
If it should be part of install then setup.py and requirements.txt should be changed and dropped from the -dev.txt one.
A reno release note should be added that lists these as a new feature addition.
Now I will note that scikit-quant already includes an interop module for qiskit - though because of the recent OptimizerSupportLevel change, that is unfortunately a breaking change if you implemented an optimizer, that interop logic would need revision to be compatible to master. Maybe if the logic is more directly included here then there is no need for an interop module for qiskit in the scikit-quant code - was there any communication with the scikit-quant devs at all?
woodsp-ibm
added
Changelog: New Feature
manoelmarques
dismissed stale reviews from pbark and adekusar-drl
via
9a3e84e
|
Having just looked at this further the installation of scikit-quant installs the supported set of optimizers from their own packages as dependencies. These dependent packages come in different licenses such as BOBYQA in GPL and SnobFit, ImFil are listed as Other/Proprietary licence. As such I think the scikit-quant should be a purely optional install which means changing imports to check for installation and ensuring that Aqua does not fail to load etc if not installed - same for unit tests they should just skip. So basically this would go into extra requires in setup.py, and be removed from the requirements-dev. The build can be altered to install these so the function gets tested as is done for other optional installs. |
There was a problem hiding this comment.
@manoelmarques Thanks for taking care of the optional dependence changes.
|
Great thanks guys ! :) |
…/qiskit-aqua#1240) * snobfit first * bobyqa optimizer * imfil optimizer * orbit optimizer * orbit added * snobfit first * bobyqa optimizer * imfil optimizer * orbit optimizer * orbit added * snobfit first * bobyqa optimizer * imfil optimizer * orbit optimizer * orbit added * final pre-pull * final pre-pull 2 * added req and docstrings * copyright corrected * utf-8 supressed * missing commas, docstrings. * make scikit-quant optional * Add to optimizers documentation * Spelling: licences (UK variant) to licenses Co-authored-by: Anton Dekusar <[email protected]> Co-authored-by: Panagiotis Barkoutsos <[email protected]> Co-authored-by: Manoel Marques <[email protected]> Co-authored-by: Steve Wood <[email protected]> Co-authored-by: woodsp <[email protected]>
…/qiskit-aqua#1240) * snobfit first * bobyqa optimizer * imfil optimizer * orbit optimizer * orbit added * snobfit first * bobyqa optimizer * imfil optimizer * orbit optimizer * orbit added * snobfit first * bobyqa optimizer * imfil optimizer * orbit optimizer * orbit added * final pre-pull * final pre-pull 2 * added req and docstrings * copyright corrected * utf-8 supressed * missing commas, docstrings. * make scikit-quant optional * Add to optimizers documentation * Spelling: licences (UK variant) to licenses Co-authored-by: Anton Dekusar <[email protected]> Co-authored-by: Panagiotis Barkoutsos <[email protected]> Co-authored-by: Manoel Marques <[email protected]> Co-authored-by: Steve Wood <[email protected]> Co-authored-by: woodsp <[email protected]>
Summary
Interface to python implementation of SNOBFIT, IMFIL, BOBYQA classical optimizers from the scikit-quant package (https://github.com/scikit-quant/scikit-quant).
Also see: https://qat4chem.lbl.gov/software
Details and comments
pip install scikit-quantto install the optimizers (sckit-quant has been added to the requirements-dev.txt).The implementation in Qiskit follows exact template of already implemented scipy optimizers.
A test is included for the 3 optimizers.
The keywords were added to pylintdic.