438 check workers available causes execution to hang #440
Conversation
- if acme is around, basically just set parallel=True - if not, wait for only a fraction of the workers before starting the computations Changes to be committed: modified: ../syncopy/shared/dask_helpers.py modified: ../syncopy/shared/kwarg_decorators.py
Codecov ReportBase: 68.02% // Head: 68.17% // Increases project coverage by
Additional details and impacted files@@ Coverage Diff @@
## dev #440 +/- ##
==========================================
+ Coverage 68.02% 68.17% +0.15%
==========================================
Files 81 81
Lines 9833 9845 +12
Branches 2032 2034 +2
==========================================
+ Hits 6689 6712 +23
+ Misses 2607 2597 -10
+ Partials 537 536 -1
Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here. ☔ View full report at Codecov. |
Changes to be committed: modified: syncopy/shared/dask_helpers.py modified: syncopy/shared/kwarg_decorators.py
There was a problem hiding this comment.
Thanks for this work. As discussed, it seems that the way Dask works and with our independent trials, there really is no point in waiting for more workers to show up instead of simply starting to work right away.
Of course, we cannot add tests that will check this behavior on the Github CI.
dfsp-spirit
deleted the
438-check_workers_available-causes-execution-to-hang
branch
Changes Summary
acmeis around earlier, and let it handle all parallel issuesReviewer Checklist