Reticulate not creating multiple threads inside docker?

Is there any reason why the R script below

library(reticulate)

sk <- import('sklearn')
np <- import('numpy')

a <- np$random$random(c(as.integer(5000), as.integer(5000)))
pca <- sk$decomposition$PCA(as.integer(100))

pca$fit(a)

runs on 4 cores when I try it locally, but it uses only 1 core inside a docker container? I am using --cpuset-cpus="0-3" and have verified by running the same task in python that the container has access to all 4 cores. I also tried a bunch of other things, but I can't seem to figure out a reason for this behavior. Anybody has any idea?