Requires:
- openapi
- cryptography
- fernet
To get access to options such as collapsable cells, install nbextensions:
[Needs updating]
There are two demo Jupyter Notebook files.
The first was written by Clay Morrison, called
ASKEM_TA1_demo_2022_10.ipynb
. It contains a candidate workflow and a
number of markers for groups to contribute content and code.
The second was written by Mike Cafarella, called
ASKEM-Simple-Walkthrough-v1.ipynb
. It contains enough code to do a
simple walkthrough of TA-1 components. It will:
- Load model source text from XDD
- Execute the original model and show results
- Analyze it using Gromet
- Load a Pandemic Ontology
- Attempt to find locations in the model where parameters from the Pandemic Ontology are found in the model code.
- Expose these recovered knobs to the user, for adjusting model parameters. The user can then execute the altered model and see results.
- After tweaking user can choose to store the new altered model in XDD.
As of now, this is a cheap-and-cheerful end-to-end demo. It works, but there are a number of placeholders where we need to do more work. For example, we have a hard-coded pandemic model instead of loading or discovering it. Also, right now the gromet representation is generated but not actually used to alter and execute the model; we use the source code instead. Also, the model/knob discoveries are hard-coded instead of dynamically discovered.
Finally, there are a number of places where you can and should add your own team's work. I have not identified all of these; Clay's file above is probably a better guide to that.
This simple walkthrough relies on a number of pretty standard Python data science packages (shutil, requests, matplotlib, pandas, etc). The only thing you probably have to install is the gromet package. You can find instructions for this in the Arizona repo here: https://github.com/ml4ai/automates.
Also, note that you'll need the latest version of plotly (5.10).