From 2f95285ba490a6cea374c7412465f6ceb747c133 Mon Sep 17 00:00:00 2001 From: Troy Raen Date: Wed, 17 Jul 2024 09:57:07 -0700 Subject: [PATCH] Add comment about potential parallelization problems (#287) --- light_curves/light_curve_classifier.md | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/light_curves/light_curve_classifier.md b/light_curves/light_curve_classifier.md index 4e792b7a..12641aa0 100644 --- a/light_curves/light_curve_classifier.md +++ b/light_curves/light_curve_classifier.md @@ -441,7 +441,10 @@ check_is_mtype(X_train, mtype="pd-multiindex", scitype="Panel", return_metadata= #this cell takes 35s to run on a sample of 267 light curves #setup the classifier -clf = Arsenal(time_limit_in_minutes=1, n_jobs = -1) +#n_jobs is the number of jobs to run in parallel. some environments have trouble with this. +#if you encounter an error such as 'BrokenProcessPool' while training or predicting, you may +#want to either set n_jobs = 1 or use a different compute environment. +clf = Arsenal(time_limit_in_minutes=1, n_jobs = -1) # '-1' n_jobs means use all processors #fit the classifier on the training dataset clf.fit(X_train, y_train)