search

DEAP / deap

Distributed Evolutionary Algorithms in Python
http://deap.readthedocs.org/
GNU Lesser General Public License v3.0
5.75k stars 1.12k forks source link

any plans allowing deap to run on gpu #209

Open kirk86 opened 7 years ago

kirk86 commented 7 years ago

Hi, may I ask if there are any plans on allowing deap to harness gpu power? With all those gpu libraries out there it might be worth considering whether we could harvest some of those in order to speed up deap. I've noticed that when I run my example using deap during the computations it transitions from using a single process to utilizing all the process. I'm trying to figure out how can I mitigate this transition so that it executes fully on all the processes. I've also made a profile of the bottlenecks profile_ga And it seems that eaSimple is the one causing the bottleneck. Do you guys have any suggestions on how to speed things up or possibly mitigate this issue? I don't understand why the generation process is causing this bottleneck since usually it should be fast. Given that the computations included in the generation process are not too expensive.

  1. Evaluate fitness
  2. Select best
  3. Crossover
  4. Mutate
  5. Replace old with new
gfviegas commented 4 years ago

that'd be a great deal, and I wonder if anyone ever done that using DEAP. I'm working in a heavy recursive operator in-between the generation proccess and any extra resource with performance would help a ton

yxliang01 commented 4 years ago

Wondering, can you simply apply numba? :D

fmder commented 4 years ago

Yeah you should look to apply GPU in you operators, like initialization, evaluation and variation.

Le sam. 12 oct. 2019 04 h 25, Xiao Liang notifications@github.com a écrit :

Wondering, can you simply apply numba? :D

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/DEAP/deap/issues/209?email_source=notifications&email_token=AAHKXQXN4VG5PDCVTXYB3C3QOGCWJA5CNFSM4DNT5GLKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEBBZQLI#issuecomment-541300781, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAHKXQU2L7OOOQBZJDJKRODQOGCWJANCNFSM4DNT5GLA .