Simple way to create a large-scale corpus of synth sounds and their parameters?

[turian]

Is your feature request related to a problem? Please describe.
I would like to drive odin2 from Python, so I can easily create a large-scale corpus of synth sounds and their parameters.
I would like to use this to train an ML system that, given audio, can predict the underlying synthesis parameters. I have already been researching this actively.

Describe the solution you'd like
I want something under 25 lines of Python code that I can easily generate a large-scale corpus using "human-designed" presets (and maybe variations) with a variety of velocity, f0, note duration, etc.

Describe alternatives you've considered
Using Surge synth (which has a nice Python API) or torchsynth

Additional context
Large scale datasets are useful for studying synthesizers using AI. However, nearly all existing softsynths assume the user is a human, not an API. :(