I have various subclasses of statistical models in my repo that typically take similar arguments but some may require special arguments. I’m writing a function to create, train, and score each model and then compare their accuracy. That leads to two linked questions:

1. What is the best data structure to use to loop through models; and
2. How would I pass the specific special arguments if/when they occur? I think this is a case for kwargs but am still fairly new to implementing kwargs.

My current planned architecture is:

def compare_models(dataframe, models):
results = {}
for model in models:
model.train(dataframe, )
results.update({model: []})

 return results

I can make it work but looking for most elegant solution.