I've recently bought Macbook Pro with M1 Pro, and finally managed to install arm64 libs for data science with miniforge, partially following this article (P.S. remember to use only one package manager either conda (preferrably) or pip for package installation) For now running stacking regressor with lightgbm and xgboost using n_jobs = 8 (number of performance cores) gives the best performance output. This way it is using 8 performance cores. If you use n_jobs = -1 setting, it makes Mac use 2 efficiency cores and 4 performance cores for some reason.
So the question is, is it possible to run it on all 10 cores?