Closed dotsdl closed 1 year ago
I see the CI tests are failing because the environment is not using the latest openff-toolkit and openff-units.
Ah actually is the lack of openff-models
, rather. I'll fix that.
:exclamation: No coverage uploaded for pull request base (
protocol-neqcyc@7351d9c
). Click here to learn what that means. The diff coverage isn/a
.
I guess retrying didn't really help https://github.com/choderalab/perses/actions/runs/4736629137/jobs/8408413871?pr=1177#step:9:2086
@ijpulidos how's this looking for unit handling on get_estimate
and get_uncertainty
?
I updated the default settings to match recommendations for production settings from @jchodera. This will avoid users submitting very short runs by default; I'm doing live of testing these choices now. Aiming to merge this PR in the morning.
@ijpulidos if you get a chance to look over my choices in the default settings, please do! Just trying to get things reasonable and you have more intuition for this than I do!
@dotsdl They seem good! Thanks for changing these.
@ijpulidos I think we're good to go here. Ran a burn test on my local box to check for memory leaks (host and GPU), and didn't observe behavior that strongly suggested any.
Can you give this a thumbs up then I'll merge?
Got one more fix to put in.
Nvm, all good I think.
@dotsdl This looks good to me, great work! Feel free to merge
Evaluating coverage of unit tests for NonEquilibriumCyclingProtocol, adding tests where we still have gaps.