Option to use a smaller model for faster computation

[marl]

Option to use a smaller model for faster computation

[justinsalamon]

This sounds like a good plan to me. A few things to note:

1. Since we need a solution for all models, I'm not sure it makes sense to just add tiny to pypi (this is not an urgent matter) - it feels messy. We might as well work directly on the final solution which downloads all the models from the GH repo (maybe still include the full model on pypi, and download all the other models from gH) and only then push to pypi.
2. We should ensure that the option to choose a smaller model is also available via the python API, not just the CLI
3. Once all the above is implemented we should bump the version, create a new release on GH and push to pypi (using twine to ensure correct markdown display on pypi).

[jongwook]

1. Once I add the implementation for handling multiple models, the updated Python code under /crepe/ can stay the same when we change the storage to GH - so I'd say it's more gradual path toward the solution.

I'm inclined to downloading all models from GH if possible, because including the same model for every release is a waste/abuse of PyPI storage.

2. Sure; the CLI will call the appropriate python API.

3. Of course.

[justinsalamon]

OK, sounds good. Let me know if you need my help with any of this, cheers.