search

Leonidas-Z / ExaConstit_CP_Framework

Use ExaConstit on the framwork
2 stars 0 forks source link

Branch that will be moved into ExaConstit #10

Closed rcarson3 closed 2 years ago

rcarson3 commented 2 years ago

@Leonidas-Z just to let you know I'm going to try and use the clean_history_exaconstit branch to bring into ExaConstit. I had to remove the DEAP and pyMOO directories from that branch as we can't have them in exaconstit proper due to licensing issues. I tried to clean up the git history a bit, but it was a bit of a lost battle so I just ended up with one massive squashed commit that you're contributed in. It has all of your commits listed in the commit message at least. I'll probably add a 2-3 more commits to clean everything up, but now that the framework works and I've tested it on real data I feel comfortable moving it out into the open into ExaConstit.

Thanks again for all of the hardwork you put into this as it's definitely going to be helpful to quite a few people!

Leonidas-Z commented 2 years ago

@rcarson3 Thank you so much! I am very happy that this is going forward! What did you do as a workaround regarding the licensing issues? I am asking this since the DEAP module didn't have the U-NSGA-III algorithm implemented.

rcarson3 commented 2 years ago

@Leonidas-Z so I removed the visualization folder but kept the ExaPlots.py file. Next, I removed the DEAP_mod folder. In order for us to still be able to use DEAP I made a fork of the project that merged in a conversion to python3 branch as well as your modifications which added the U-NSGA-III algo. You can find that fork here: https://github.com/rcarson3/deap/tree/master . On our code side of things, I did have to modify things slightly where the DEAP_mod import was replaced with deap afterwords things pretty much just work.

rcarson3 commented 2 years ago

@Leonidas-Z so I think this almost ready, but one of the last question I kinda had is did you intend for the U-NSGA-III algo to also be used in the selection process for the best solution as well as seen in these lines: https://github.com/Leonidas-Z/ExaConstit_CP_Framework/blob/main/Exaopt_DEAP/ExaConstit_NSGA3.py#L434-L459 ? Since it currently looks like that select method is using the NSGA-III algorithm instead or is the U-NSGA-III only intended for the mutation portion of things?

rcarson3 commented 2 years ago

@Leonidas-Z just wanted to let you know that this has now been pulled into the main branch of ExaConstit here: https://github.com/LLNL/ExaConstit/tree/exaconstit-dev/workflows .