Replies: 1 comment
-
|
@timothyas Thanks for sharing this recipe! I'm sure it will be useful to others, even after we get to the bottom of whatever was causing the concurrency issues you observed. I'll be testing it out myself to make sure I understand the technique. |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
-
As noted in #49, I ran into some trouble while scaling up the stats computation via wxvx. I'm not really sure if some of the trouble stemmed from using multiple tasks/processes within wxvx, but I decided to try something else to parallelize the operations and I thought I'd share them here in case its useful to others.
I ended up using GNU Parallel basically because that's what Perlmutter recommends since it enables one big slurm job rather than overloading the scheduler. My process was as follows:
Create a slurm script that pulls in the observations and then creates an input file to be processed with GNU parallel.
The cycles directory has 1 separate yaml for each cycle processed by wxvx, created with a separate python script. Then, the parallel_wxvx.sh script is what does the main tasks:
This ended up working pretty well for me. Probably not the cleanest approach, but I just thought I'd share!
Beta Was this translation helpful? Give feedback.
All reactions