Modes files (background.txt) are stored alongside all datasets.

How do we want to refer to these?

• Recommended modes?
• Default modes?

What I actually want is to store a schema listing

1. type information
2. the variables

The actual way modes should be set depends on the problem being solved.

How might they be accessed?

Assuming we're already in a workflow like:

from relational_datasets import load train, test = load("toy_cancer") # Maybe? modes = load_default_modes("toy_cancer") modes = load_modes("toy_cancer")

Some part of them should be implied by the data itself. For example, when implementing the conversion from vector-based datasets this felt like a natural way to implement things:

from relational_datasets.convert import from_numpy import numpy as np data, modes = from_numpy( # ... )