I am still figuring out a lot of stuff and wanted to adopt the best practices. Is it wiser to always choose Model.parameters? or Global Variables are going to stick around?
If you have an input to your model, you should use a Model Parameter. Especially if you want to experiment with that value using the Experimenter, Optimizer, or Reinforcement Learning.
If you want to store the state of your model's logic globally, then global variables are still a good solution.
14 People are following this question.
