use more cores than I set for the job when I use mlxtend for feature selection on a cluster

[王晓康]

I was doing feature selection using mlxtend. When I run it on my local computer, it is fine. However, when I run on a cluster, the administrator of the cluster informed me that my job was using more cores than the number of cores I requested for the job.  So I was thinking whether there is a way to set how many cores used by mlxtend.

[Sebastian Raschka]

Hi there,

this could be either a bug in joblib or you didn't specify the number of jobs. I assume you were using SequentialFeatureSelector?

You could try setting n_jobs to a certain integer rather than n_jobs=-1.

Best,
Sebastian

> On Feb 3, 2020, at 12:01 PM, 王晓康 <wmm...@gmail.com> wrote:
>
> I was doing feature selection using mlxtend. When I run it on my local computer, it is fine. However, when I run on a cluster, the administrator of the cluster informed me that my job was using more cores than the number of cores I requested for the job. So I was thinking whether there is a way to set how many cores used by mlxtend.
>
>
>

> --
> You received this message because you are subscribed to the Google Groups "mlxtend" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to mlxtend+u...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/mlxtend/bb658f3e-d594-4829-a633-d351f0b00940%40googlegroups.com.