SDV

SDV copied to clipboard

There have been a few questions, such as #515 and #299 about how to set a random state for the models. External workarounds exist, like the one shared in those questions, but we may want to add random_state arguments that take care of doing this internally.

The arguments should be added in all the places where it makes sense doing so, such as the __init__, fit and sample methods of the models, or any other steps that involve any kind of randomness, such as the Constraints or the RDT transformers.

Note that as of SDV 0.15.0, all single-table sampling functions now have a randomize_samples parameter to control the randomness.

Remaining places to control randomness are:

• Multi-table sampling functions
• Sequential model sampling functions
• All fitting functions