Add multiprocessing to medium-range lagged ensemble - #18
Conversation
|
|
||
| # Set list of medium range lagged ensemble runs and hours of forcing lag | ||
| ens_members = { | ||
| 'no_da': 0, |
There was a problem hiding this comment.
What is the purpose of making this configurable as a global variable, rather than just using 'no_da'?
There was a problem hiding this comment.
In the pending usage of nwm-rte, the var is imported by nwm-rte to build the first realization before passing things off to nwm-fcst-mgr for the rest of the lagged ensemble sequence. However since it would be better to avoid building the first realization twice (from nwm-rte and then from nwm-fcst-mgr), this could be reverted when that change is made before merging.
|
Closing but not deleting the branch yet. Will reopen a new PR from this branch after making changes there: https://github.com/NGWPC/nwm-fcst-mgr/tree/maxkipp-laggedens-n2-2 Which was based on branch |
2 participants
This enables multiprocessing for medium-range lagged ensemble forecasts, leveraging outputs of
partitionGeneratorand using the number of processes specified in that intermediary output.This should be reviewed in conjunction with related PRs on other repositories:
https://github.com/NGWPC/nwm-rte/pull/52
NGWPC/nwm-msw-mgr#54
NGWPC/ngen-forcing#138