pdmlabs.optimization.smac_adapter

pdmlabs.optimization.smac_adapter#

SMAC3 optimizer adapter for PdMLabs.

All smac and ConfigSpace imports are lazy (inside method bodies) so this module is importable even when smac is not installed.

Install SMAC3 support with:

pip install pdmlabs[smac]

The adapter uses HyperparameterOptimizationFacade as the single entry point. Hyperparameter space conversion rules:

  • list with string values → CategoricalHyperparameter

  • list with numeric values → OrdinalHyperparameter (preserves the exact candidate set defined by the user)

  • scipy.stats frozen dist → UniformFloatHyperparameter

A unique smac_output_<hex>/ directory is created per maximize() call to prevent file-system collisions across parallel or repeated runs.

Classes

SMAC3Adapter()

Adapter for SMAC3 HyperparameterOptimizationFacade.
class pdmlabs.optimization.smac_adapter.SMAC3Adapter#

Bases: BaseOptimizerAdapter

Adapter for SMAC3 HyperparameterOptimizationFacade.

Requires smac >= 2.0.0 (pip install pdmlabs[smac]).

maximize(param_space: dict, objective_fn, n_iterations: int, n_jobs: int, initial_random: int, constraint_fn=None) → dict#

Maximize objective_fn using SMAC3.

SMAC3 minimises internally; the objective is negated so SMAC effectively maximises. Results are negated back before returning.

minimize(param_space: dict, objective_fn, n_iterations: int, n_jobs: int, initial_random: int, constraint_fn=None) → dict#

Minimise objective_fn using SMAC3 directly (no negation).

supports_categorical: bool = True#
Contents