I still had this branch from some time ago. It has (almost?) everything. I was still trying to improve the test_run script to only start the container once and invoke the model multiple times; right now it restarts the container for each interface.

I had not yet figured out how to deal with the inputs and outputs when invoking multiple times. It might be easiest to add the test inputs and outputs folder to the container and swap the input for each interface to a single input folder and move the output to the outputs folder. Otherwise it's a headache with file permissions.

I don't think there is a clean way of including a test script for exec mode. It would just be confusing. If a user wants to revert to exec mode, they'll have to adjust the test script themselves.

It might be easiest to add the test inputs and outputs folder to the container and swap the input for each interface to a single input folder and move the output to the outputs folder.

In prod we do this with links from /input to the actual input location for each invocation, same with output.

I don't think there is a clean way of including a test script for exec mode. It would just be confusing. If a user wants to revert to exec mode, they'll have to adjust the test script themselves.

We should not offer templates or an option for exec mode, there is no good reason to use it over invoke.

The LABEL being used as switch finally had the 2 cents rolling down (or up!) on the way things are set up. =P

After this is merged, we might want to update the onboarding powerpoint as well as the documentation to reflect the new default.