Entering edit mode
3.2 years ago
rakinitopo
▴
10
Hi! I'm having trouble using Admixture, as it is struggling to converge. I have used it before and I don't recall it taking so long. The main difference between runs is that I've now increased the number of threads up to 10 (--j10). I'm running it on a server with 70GB RAM and 24 cores and it's already been 8 hours.
The following are the last two iterations:
5 (QN/Block) Elapsed: 3077.51 Loglikelihood: -5.19583e+09 (delta): 1.79089e+06
6 (QN/Block) Elapsed: 3085.11 Loglikelihood: -5.19545e+09 (delta): 380648
This post reported similar issues, but I'm not using --haploid, so I don't think it's the same bug: ADMIXTURE haploid mode does not converge
Thank you in advance!
Sometimes you can actually slow the program down if you use too many threads since they all have to read from the same disk at the same time. If you use e.g. 1 or 2 threads, do you still have the same issue?
I'll try with just 2 threads then, thank you!!!