Compatibility with Parallel Processing Methods

[mxhsin]

Hello,
Regarding GASearchCV, is it amenable and compatible with MPI or any other parallel processing method? Currently the fit method of GASearchCV takes very long to compute and is computationally expensive. Will it work with any parallel processing methods (eg. MPI) to speed up the process? Appreciate if you could help!

[rodrigo-arenas]

Hi @mxhsin, at this point, the only parallelism that is enabled is in the training of each individual, cross-validation is performed to calculate the fitness of the solution, and the level of parallelism is controlled by the parameter n_jobs of the GASearchCV class. The default value is None which means no parallelism, you can set it to -1 to use all the available CPUs.
Although this will increase the speed of training, it is still a generally slow process depending on how many individuals and generations you set.
You can also try to use callbacks to stop the training with early stopping strategies: Callbacks Tutorial

[yeodynasty]

Hi, can the individual runs be parallelized also?