You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
MetaOpt's global timeout currently stops all worker processes after a certain clock/realword time has passed. To compare the solution quality of optimizers on machines with load, it would be neccessary to stop their worker processes after a certain processor time has been used by them. That could be done using the plugin mechanism by polling the used user and system CPU time of each worker process routinely and terminating them when a certain limit is reached. Since Python 3.3, there is tme.process_time() which returns exactly that value. For the sake of comparability between interpreters MetaOpt should most likely use one implementation with all of them. There is a gist with a backport to Python 2.7 that could be used as inspiration for a backwards compatible implementation. See also the PEP418 on this process time function
The text was updated successfully, but these errors were encountered:
MetaOpt's global timeout currently stops all worker processes after a certain clock/realword time has passed. To compare the solution quality of optimizers on machines with load, it would be neccessary to stop their worker processes after a certain processor time has been used by them. That could be done using the plugin mechanism by polling the used user and system CPU time of each worker process routinely and terminating them when a certain limit is reached. Since Python 3.3, there is tme.process_time() which returns exactly that value. For the sake of comparability between interpreters MetaOpt should most likely use one implementation with all of them. There is a gist with a backport to Python 2.7 that could be used as inspiration for a backwards compatible implementation. See also the PEP418 on this process time function
The text was updated successfully, but these errors were encountered: