CalibrationReport#
- class hydromodpy.calibration.CalibrationReport(session_id, method, n_iterations, best_objective, best_sim_id, duration_s, save_runs, promoted, best_parameters=None, parameter_uncertainty=(), workspace=None, extra=<factory>, store_factory=None)[source]#
Bases:
objectStructured summary of one calibration session.
Returned by
Project.calibrateand by therun_calibration_clihelper (both keep ato_dictshim so existing callers keep working).Attributes#
- session_id
UUID (hex) of the calibration session in the catalog.
- method
Optimizer method name (e.g.
"optuna").- n_iterations
Number of iterations actually run (could be <
max_iterif the optimizer converged).- best_objective
Best (minimum) objective value achieved.
- best_sim_id
UUID of the promoted best run when
save_runs != "none", otherwiseNone.- best_parameters
Physical values of the best candidate, keyed by calibrated parameter name.
Nonewhen no candidate was evaluated. A staged calibration reads it to freeze what a phase calibrated.- duration_s
Wall-clock duration of the calibration loop in seconds.
- save_runs
The
save_runsmode used ("none","best_n"or"all").- promoted
Count of iterations promoted to full simulations after the loop.
- workspace
Workspace root the session was written to.
- extra
Free-form metadata (callers may attach anything extra here).
- param session_id:
- type session_id:
- param method:
- type method:
- param n_iterations:
- type n_iterations:
- param best_objective:
- type best_objective:
- param best_sim_id:
- type best_sim_id:
- param duration_s:
- type duration_s:
- param save_runs:
- type save_runs:
- param promoted:
- type promoted:
- param best_parameters:
- type best_parameters:
- param parameter_uncertainty:
- type parameter_uncertainty:
- param workspace:
- type workspace:
Path|None- param extra:
- type extra:
- param store_factory:
- type store_factory:
-
parameter_uncertainty:
tuple[Any,...] = ()# Width beside each calibrated value, when the file asked for one.
Empty unless
[calibration.uncertainty]declared a method that produces it. The values themselves are inbest_parametersand are never touched by it.
- property iterations#
Return the iteration history as a
pandas.DataFrame.Loads lazily through the configured calibration store.
- property best#
Return the best promoted
RunorNone.