Gotcha, that context is more informative thanks.
Perhaps instead of running on every iteration of your optimization cycle you just do a tunable fraction at random
thr=0.02
rand(Uniform(0,1)) < thr && GC.gc();
Gotcha, that context is more informative thanks.
Perhaps instead of running on every iteration of your optimization cycle you just do a tunable fraction at random
thr=0.02
rand(Uniform(0,1)) < thr && GC.gc();