Document how to load an experiment without the cloud

[wshanks]

It would be nice to document how users who don't have access to the cloud database can load old experiments.

I think this can be done roughly as:

from qiskit_experiments.framework import ExperimentData


# Somehow the user must retain and run the same code that created the experiment originally
experiment = Experiment(**opts)

# Somehow the user must retain and provide the list of job IDs
data = ExperimentData(job_ids=job_ids)
experiment.analysis.run(data)
data.block_for_results()

I think that should retrieve the raw data from the IBM Quantum service and run the same analysis on it again. I am not sure if there are some caveats that need to be added on cases where some information could be missing here.

Perhaps documentation for this could fill in the "somehow"s above -- maybe BaseExperiment.config() and BaseExperiment.from_config() could be used to save and restore the experiment from a file instead of re-running the original experiment code. Also, the qiskit-ibm-experiment package provides a local option for IBMExperimentService (see here). This local option is provided primarily for testing (it is not very scalable), but perhaps it would (or could) provide a better option for a user to save and reload a handful of experiments rather than storing job IDs by hand.

This topic came up in this Slack discussion.

[eliarbel]

I think we should 'promote' the option to save experiments locally to be easy to use, and handle all the required nitty-gritty, such as job-ids handling as Will mentioned. Currently the option is hidden and requires code tweaking to use. Enabling some like this:

exp_date = some_exp.run(...)

... 

exp_data.save(local_file=<local db>) 

could be very convenient

[coruscating]

Guide added in #983. We should continue the save interface discussion in a new thread.