Hi,
I'm learning to work with argo+optuna, and yous repository is a gem.

Right now I have multiple models being trained (in parallel, using argo workflows) that are later combined and evaluated on a testset.
I'm trying to use optuna for hyperparameter tuning.
Although I could optuna to optimize each training by itself, I would like that a single trial consists the training of all models.
For this case, I would have to create a trial before calling the training scripts (and use trial.suggest... inside of them), and later evaluate the loss on the evaluate script.

Could you help me on how can I 'breakdown' the study.optimize( for creating the trial on one script (before training scripts), and evaluating it in another script (on the one that performs the emsemble)?