Assigning n_jobs

[fabiogeraci]

Hi,

below it is the definition of the selector class, n_jobs is by default set to 2, and two screen shot before and after running the code. It appears that n_jobs is running on all available cpus

class Selector:
    def __init__(self, estimator=None, n_jobs: int = 2):
        self.estimator = estimator
        self.n_jobs = n_jobs
        self.selector_model = None
        self.select_estimator()

    def select_estimator(self):

        self.selector_model = GeneticSelectionCV(
            self.estimator, cv=5, verbose=0,
            scoring="f1_weighted", max_features=25,
            n_population=50, crossover_proba=0.5,
            mutation_proba=0.2, n_generations=50,
            crossover_independent_proba=0.5,
            mutation_independent_proba=0.04,
            tournament_size=3, n_gen_no_change=10,
            caching=True, n_jobs=self.n_jobs)

[manuel-calzolari]

Hi Fabio,

using the example in the documentation https://sklearn-genetic.readthedocs.io/en/latest/examples.html (changing the value of the n_jobs parameter) I am not able to reproduce this issue.

Which estimator do you use? If you configure the n_jobs parameter of the estimator (e.g. of a Random Forest) to use multiple jobs, then additional CPUs can be used by the estimator itself. Basically, the total parallelism would be given by the n_jobs of the genetic selector multiplied by the n_jobs of the estimator.