Hello! :)
I’m trying to use within-chain parallelisation to try and speed up my model but unfortunately don’t think I’m doing something right here. So,
AFAIK I have 16 CPUs to use and have 4 chains. I used threads = threading(2)
, in the brm function, to use 8 CPUs and that seemed to work (when I say ‘seemed to work’, I mean I saw 8 CPUs at ~ 100% usage). But when I tried using threads = threading(3)
or threads = threading(4)
, it still only seems to use 8 CPUs (CPU 1 to 6 and CPU 11 and 12),
I’d like to use at least 12 CPUs (i.e., 3 threads per chain) and would appreciate any help. Details of the virtual computer I am using on the server is below,
- Operating System: CentOS Linux (Release 7 (Core) 64-bit; Kernel Linux 3.10.0-1160.36.2.el7.x86_64 x86_64) and Hardware (Memory: 31.3 GiB; Processor: AMD EPYC 7501 32-Core Processor × 16)
- brms Version: 2.16.2
Thank you for reading, cheers,
Rohan