search

MICS-Lab / lincs

Learn and Infer Non Compensatory Sortings
https://mics-lab.github.io/lincs/
GNU Lesser General Public License v3.0
2 stars 1 forks source link

Make learning deterministic #18

Closed jacquev6 closed 1 year ago

jacquev6 commented 1 year ago

Currently the result of a call to learndepends on OMP_NUM_THREADS. It should be possible to make it independent of the number of threads and produce identical results on any machine, regardless of CPU or GPU and number of threads.

Independence from number of threads should be achievable using a RNG per model instead of per thread.

jacquev6 commented 1 year ago

This was an issue in the previous implementation, before lincs was even initiated. Learning is deterministic in this implementation.