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
I really like using smart_dispatch and I have a few ideas how it could be improved:
The location for creating logs should configurable, now SMART_DISPATCH_LOGS seems to be hardcoded.
The names of the log folder in SMART_DISPATCH_LOGS should be configurable. When I run my 99th experiment it becomes quite hard to enter the path to its folder with logs. Would be cool to be able to do smth like
which could create a folder SMART_DISPATCH_LOGS/my_experiment_with_batch_norm
The by far most important deliverables of a launch are .err and .out files. I think they should have simpler names (now their names contain some kind of a hash) and that they should be in the top-level folder, not in /logs.
I think that the default current directory for the launched job should be not the home folder, but rather SMART_DISPATCH_LOGS/my_experiment_.... This would be very helpful when one wants to repeat an experiment multiple times but is too lazy to invent a handful of different names for experiments of the series.
So far this it it, maybe more is coming.
The text was updated successfully, but these errors were encountered:
I really like using smart_dispatch and I have a few ideas how it could be improved:
SMART_DISPATCH_LOGS
seems to be hardcoded.The names of the log folder in SMART_DISPATCH_LOGS should be configurable. When I run my 99th experiment it becomes quite hard to enter the path to its folder with logs. Would be cool to be able to do smth likewhich could create a folderSMART_DISPATCH_LOGS/my_experiment_with_batch_norm
/logs
.I think that the default current directory for the launched job should be not the home folder, but ratherSMART_DISPATCH_LOGS/my_experiment_...
. This would be very helpful when one wants to repeat an experiment multiple times but is too lazy to invent a handful of different names for experiments of the series.So far this it it, maybe more is coming.
The text was updated successfully, but these errors were encountered: