jeffrey-newman / charles-metronamica-optimiser Goto Github PK

Last run appears to have stopped due to:

slurmstepd: error: Exceeded job memory limit

Taking doubly as long.
Test of single OF = 11 minutes.
Yet 39 minutes for two generations (equal numbe rof population as workers)

Test runs on Phoenix are not sending emails of hypervolume status

To allow for better issue identification, should put host name in log files, esp. for mpi jobs on clusters

Boost has function to get host name

https://www.quora.com/Using-Boost-C++-libraries-how-do-I-obtain-local-machine-hostnames-and-IPs-for-each-network-adapter

MCK seems to be opening a window when running on Phoenix. Does not do this when running MCK on l02.phoenix.ade......

Registry value in users are correct.

Mutation rate is not being set in atleast one of the parallel calibrators

Always receive the same values back from MetronamicaOF for each objective for each decisision variable string

Seems to be (on one hand) maximising clump and minimising FKS.

Potential improvements for speeding up metrics (e.g. use vector instead of map; have template of neighbourhood for each cell for entire map, rather than recalc neighbours for every cell for each iteration.

Massage current class for Charles' smaller case study