when we run the experiments (e.g., the branch learners) for n_jobs (run in a core). each run in a core will consume a lot of memory. After a lot of testing and meet a lot of weird problems, I finally found the rough consume memory for each run in a core.
It is about 4 Gbits each n_jobs (run in a core). Therefore, generally, do not run more than 16 cores for 64 Gbits memory on ripper1/ ripper2/ripper3.
If you run more cores, you will probably meet the problem like:
If you run 16 cores on ripper1/ripper2/ripper3, you will see the perfect ending like:
note that 1500 is the terminal iteration.
for more computing platform, we could calculate it with 4 Gbits per cores at least.
when we run the experiments (e.g., the branch learners) for
n_jobs(run in a core). each run in a core will consume a lot of memory. After a lot of testing and meet a lot of weird problems, I finally found the rough consume memory for each run in a core.It is about
4 Gbitseachn_jobs(run in a core). Therefore, generally, do not run more than 16 cores for 64 Gbits memory on ripper1/ ripper2/ripper3.If you run more cores, you will probably meet the problem like:
If you run 16 cores on ripper1/ripper2/ripper3, you will see the perfect ending like:
note that 1500 is the terminal iteration.
for more computing platform, we could calculate it with 4 Gbits per cores at least.