Includes tqdm bar

[mreismendes]

It would be interesting to see an estimation of the number of models to be trained and the concluded percentage.

Using TQDM would be great.

PS: I'm not a software engineer. I'm just a data scientist that sometimes goes deep into under the hood issues.

Thank you!

[winedarksea]

I could do a bar to show number of generations but not number of models. The reason for that is models are generated dynamically for each generation. Actual number of models per generation is approximately the same - say N models (it's about 50 max if I remember) are created, these can include duplicates of previous models, these duplicates are then removed which leaves you usually with N minus some random amount in each generation.
I feel I kind-of do this already with the print messages (if verbose > 0). I am not sure how much tqdm would actually add to the experience - sure it would be prettier, but it's also another dependency.

[mreismendes]

I know there is an option to built more than one level of "TQDM" as far as you know the n models previously. But as you said I think this enhancement surely is not on the top of the bucket list, it's more related to aesthetical issues.