Store ert<->everest realization mapping #9767
Conversation
CodSpeed Performance ReportMerging #9767 will not alter performanceComparing Summary
|
yngve-sk
force-pushed
the
25.01.16.add-everest-realization-info-to-ert
branch
8 times, most recently
from
4e70ffd to
be7341e
Compare
yngve-sk
force-pushed
the
25.01.16.add-everest-realization-info-to-ert
branch
4 times, most recently
from
d4d8a92 to
2202375
Compare
yngve-sk
force-pushed
the
25.01.16.add-everest-realization-info-to-ert
branch
from
2202375 to
3990831
Compare
|
|
||
| realization_mapping: dict[int, EverestRealizationInfo] = { | ||
| idx: EverestRealizationInfo( | ||
| geo_realization=self._everest_config.model.realizations[real], |
There was a problem hiding this comment.
Can we internally already move from geo to model, I think we should encourage that in the long run.
| class EverestRealizationInfo(TypedDict): | ||
| geo_realization: int | ||
| perturbation: int | None # None means it stems from unperturbed controls | ||
| # Q: Maybe we also need result ID, or no? Ref if we have multiple evaluations |
There was a problem hiding this comment.
Yes, we will probably, let's not worry about it now.
| for e in self.ensembles: | ||
| ens_parameters = { | ||
| group: e.ert_ensemble.load_parameters(group) | ||
| .to_dataarray() | ||
| .data.reshape((e.ert_ensemble.ensemble_size, -1)) | ||
| for group in parameter_values | ||
| } |
There was a problem hiding this comment.
you think may need some caching here, or is there already in the storage?
There was a problem hiding this comment.
This just points to parameters saved in xr files in ERT storage
| else None | ||
| ), | ||
| ) | ||
| for idx, real in enumerate(evaluator_context.realizations) |
There was a problem hiding this comment.
This maps the indices of the control vectors to realizations/perturbations. Is that the intention? Because if the intention is to map simulation ID (= ert realization?) then this will fail if some control vectors are not evaluated due to being cached or being marked as inactive. You would need to do the same as I did in my recent PR to get the simulation ID's for the simulations that actually did run.
| )["values"] | ||
|
|
||
| cached_data = ( | ||
| objectives.to_numpy() * -1, |
There was a problem hiding this comment.
Be careful after rebasing, the -1 is probably not needed anymore.
| if constraints is not None: | ||
| assert cached_constraints is not None | ||
| constraints[control_idx, ...] = cached_constraints | ||
| for control_idx, ( |
There was a problem hiding this comment.
I think caching is now always assumed to be on? We do have a flag for in in the configuration, maybe we should kill it.
Note: Should not be merged until a consistent realization mapping is received from ROPT, replacing the logic in the second commit of this PR.
Issue
Resolves #9751
Resolves #9674
Approach
Short description of the approach
(Screenshot of new behavior in GUI if applicable)
git rebase -i main --exec 'pytest tests/ert/unit_tests -n logical -m "not integration_test"')When applicable