automated tuning based on dependency graph
AutoDepGraph is a framework for using dependency graphs to calibrate a system. It is heavily inspired by "Physical qubit calibration on a directed acyclic graph".
AutoDepGraph consists of two main classes, the CalibrationNode and the Graph. Calibration is done by calling a node that one wants to execute, the node contains the logic required to satisfy the nodes it depends on (parents).
A CalibrationNode contains:
- Good (green): check passes
- needs calibration (yellow): calibration is not up to date anymore and needs to be updated
- Bad (red): calibration or check has failed
- unknown (grayed): checks of the node should be run
- active (blue): calibration or check in progress
- parents: the nodes it depends on
- children: nodes that depend on this node
- check_function : name of function to be executed when check is called. This can be a method of another instrument.
- calibrate_function : name of function to be executed when calibrate is called. This can be a method of another instrument.
- calibration_timeout: time in (s) after which a calibration times out.
- execute or call
- Performs the logic of a node (check state, satisfy requirements) with the goal of moving to a "good" state
- Performs checks to determine and the state of a node
- Executes the calibration routines of the node
- execute or call
A Graph is a container of nodes, it is used for:
- new graphs can be created by instantiating a graph and then using the add_node method to define new nodes.
- loading and saving the graph
- real-time visualization using pyqtgraph
- state of the node determines color of a node
- if a node has no calibrate function defined it is a manual node and has a hexagonal instead of a circle as symbol
- mouseover information lists more properties (planned)
For an introductory example see the example notebook. If you want to see how to use a specific function, see the tests located in the autodepgraph/tests folder.
- Clone the repository
- install the requirements
- navigate to the repository and run
pip install -e .
- verify success of installation by running
N.B. windows can be "problematic"
Installation on windows is a bit more difficult, this relates mostly to the installation of pygraphviz. To install graphviz and pygraphviz on windows follow these steps:
- get the 64 bit version of , copy it to e.g., program files and add the bin folder to the system path.
- the 64 bit version lacks the libxml2.dll, you most likely have this from some other program. You can find this by searching for
libxml2.dllin the program files folder. After that just copy paste it to the bin folder of graphviz.
- get pygraphviz by downloading the master from github.
- Now you will need to edit pygraphviz/graphviz.i and pygraphviz/graphviz_wrap.c according to the changes at https://github.com/Kagami/pygraphviz/tree/py3-windows-iobase. A reference can be found in the _install folder
- Next install using
python setup.py install --include-path="C:\Program Files\graphviz-2.38_x64\include" --library-path="C:\Program Files\graphviz-2.38_x64\lib"
- then install autodepgraph and test the installation using
I would like to thank Julian Kelly for the idea of using a dependency graph for calibrations and for early discussions. I would like to thank Joe Weston for discussions and help in working out the initial design. I would like to acknowledge Livio Ciorciaro for disucssions and as a coauthor of this project.
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
|Filename, size||File type||Python version||Upload date||Hashes|
|Filename, size autodepgraph-0.3.4.tar.gz (9.1 kB)||File type Source||Python version None||Upload date||Hashes View|