Implied timescales calculation hangs with multiprocessing
yuxuanzhuang opened this issue · 0 comments
yuxuanzhuang commented
It seems to be a recurring question (related #1576 #1553 #1582 and maybe more). And I know you guys will replace the implementation of it in a future version of pyemma, but I just want to point out that the issue seems to be related to the combination of threadpool_limits and get_context("spawn").Pool---for some weird reason, they don't play nice with the serialization of MaximumLikelihoodMSM.
Either removing threadpool_limits or using get_context().Pool can resolve the hanging.
ubuntu 20.04
python: 3.10
pyemma: 2.5.12
threadpoolctl: 3.1.0