Bootstrap from a prior run
Optional. Point at a saved *_client_state.json to inherit
search space, fitted GP, and trial history. The current YAML is
deep-merged on top.
Add a specialize map to pin context
parameters from a robust run to concrete values.
Parent: ( trials)
Parent experiment is marked immutable. foamBO will clear that flag at load time so specialization can mutate the search space.
Context groups:
| Pin | Parameter | Context? | Bounds | Value |
|---|---|---|---|---|
Experiment
Trial Generation
Baseline
Set baseline values for each parameter. Keys auto-populated from experiment parameters.
Optimization
Prefix with - to minimize. Comma-separate for multi-objective.
Case Runner
Orchestration Settings
Global Stopping Strategy
Early Stopping Strategy
Store
Dimensionality Reduction
Automatically fix low-importance parameters based on Sobol sensitivity analysis, shrinking the search space after enough trials are run.
Trial Dependencies
Define trial-to-trial dependencies (warm start, mesh inheritance, etc.)
Source Trial
Resolution results are written to .foambo_deps.json ($FOAMBO_DEPS). Your runner script reads the manifest to decide what to do.