parallel GP with example <test_GP_LJ_fresh.yaml>

[TSMtang]

Dear Flare developer,

I've been using flare SGP and it's excellent, but when I try to use the full GP, I find that it's not doing parallel calculation.

Here is what I did:

1. use example file <test_GP_LJ_fresh.yaml> as template
2. modify "n_cpus: 1" to "n_cpus: 16"
3. export OMP_NUM_THREADS=2
4. use command $ flare-otf input.yaml

In the flare output file (.out) it knows how many cores I'm giving, but when I check my system monitor, only two CPUs are working. And no matter how I modify that "n_cpus" and "OMP_NUM_THREADS", it's not doing parallel calculation.

Could you please give me a hint of what might be the problem? I'm confused because SGP has no problem, whereas GP has this issue.

Best regards
Shengming Tang

[YuuuXie]

Hi @TSMtang, sorry for the very delayed reply. I never receive notifications from our GitHub Discussions so I did not notice there was a question... Are you still facing the problem? Based on the documentation, it seems that there is also the arg parallel that needs to be True:

flare/flare/bffs/gp/gp.py

Line 66 in 960757e

parallel (bool, optional): If True, the covariance matrix K of the GP is

[YuuuXie]

P.S. Just modified my notification settings so I won't miss anything... Did not realize when did I "unwatch"ed our repo mistakenly