-
Notifications
You must be signed in to change notification settings - Fork 52
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
fix: Make sure that checkpoints in PBT are removed once they are no longer needed #600
Conversation
1bb58c5
to
c489c8f
Compare
I should also check what happens for synchronous multi-fidelity methods. |
Converting this to draft. I'd still like to resoved the situation for synchronous MF schedulers |
Codecov ReportPatch coverage:
Additional details and impacted files@@ Coverage Diff @@
## main #600 +/- ##
==========================================
+ Coverage 64.98% 65.09% +0.11%
==========================================
Files 353 354 +1
Lines 26238 26279 +41
==========================================
+ Hits 17051 17107 +56
+ Misses 9187 9172 -15
... and 6 files with indirect coverage changes 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 in Codecov by Sentry. |
OK, this should go in, since it fixed the checkpoint situation for PBT and the synchronous schedulers. Pre-emptive removal for ASHA promotion is more complicated, I'll do it later. |
40deb1e
to
c9d404a
Compare
…lso fixes this issue for PBT
c9d404a
to
b514dda
Compare
… needed
This should solve the issue about checkpoints not being removed. I just return STOP instead of PAUSE. See the code comment why I think this is fine.
I also implemented early checkpoint removal for the synchronous schedulers, which requires the scheduler to flag paused trials for which the checkpoint is no longer needed.
By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.