Cores=x doesn't impact the number of cores actually used

Questions v5
1

Hi,

bambi==0.9.0.
pymc==4.0.0

I have an issue similar to #187 - no matter what the number of cores is in the fit() (bambi) / sample() (pymc) method the process utilized maximum number of cores on my computer.

For example, here I specified cores=4.

Screenshot from 2022-07-18 09-24-33
Screenshot from 2022-07-18 09-24-33703×78 9.53 KB

Now, I suspect that this relates to this debate but I am still not sure how can one restrict the actual amount of comp resources used. Any pointers? Thanks!

P.S.: This jams up the pc, so I have to run even smallish models on HPC which isn’t super practical.
P.S.II: See also this issue for ref.

2

Can you replicate this on a pure PyMC model (no bambi)?