For an ongoing research we need to handle human feedback between generations.
What would be the best way to implement this?
At the moment I implemented with a low number of evaluations, and let the user decide which candidates are good and start a new evolution with the selected candidates as seeds.
But I feel like this solution is losing the internal model parameters.
What would be the best approach to implement some kind of pausing of the evolution with a custom selector?