Closed Andrew-S-Rosen closed 1 year ago
@wjcunningham7, @santoshkumarradha, @cjao: This should be good to go I think! I just tested it out, and it worked as expected.
Since we will want to mimic this approach in other executors, please make sure the naming and such are to your liking. When it's merged, can we also get a version update? Thanks!
Patch coverage: 100.00
% and project coverage change: +0.36
:tada:
Comparison is base (
ba35b6c
) 84.06% compared to head (cf1da17
) 84.43%.:exclamation: Current head cf1da17 differs from pull request most recent head 81b687a. Consider uploading reports for the commit 81b687a to get more accurate results
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Do you have feedback about the report comment? Let us know in this issue.
Wow this was super fast and perfect ! 🙏
CC : @wjcunningham7
EDIT: Fixed and ready to go!! I confirmed that everything works correctly in a "real" scenario on Perlmutter.
Alright, I'm a bit stumped. Entirely unrelated to this PR, I'm having that issue where I cannot unpickle things on the remote machine (Perlmutter). I tried nuking my entire conda env again on both local and remote but with no luck. That's making it a bit hard to test this and give me confidence that everything is working perfectly.
I also think I must have introduced an error somewhere when switching from self._current_remote_workdir
to passing the currente_remote_workdir
around because the unique folders aren't being made anymore when I tried it out "for real" (despite the tests passing here).
Going to have to tackle this tomorrow because I have no brain cells left right now...
Closes #56 and combines a separate bugfix I was going to introduce in #58.