Experiments

metasurface_py.experiments

Reproducible experiment management.

ExperimentConfig

Bases: BaseModel

Serializable experiment configuration.

Describes a complete experiment: geometry, cell model, optimization objective, and solver settings. Can be saved/loaded as TOML.

ExperimentResult dataclass

Result of an experiment run.

Parameters:

Name	Type	Description	Default
config	ExperimentConfig	The experiment configuration used.	required
optimization	OptimizationResult	The optimization result.	required
environment	dict[str, Any]	Environment metadata for reproducibility.	dict()

build_from_config(config)

Instantiate objects from an experiment config.

Parameters:

Name	Type	Description	Default
config	ExperimentConfig	Experiment configuration.	required

Returns:

Type	Description
tuple[Metasurface, Any, AngleGrid]	Tuple of (Metasurface, Objective, AngleGrid).

capture_environment()

Capture current environment for reproducibility.

Returns:

Type	Description
dict[str, Any]	Dict with package version, Python version, platform,
dict[str, Any]	git commit hash, and timestamp.

run_experiment(config)

Execute an experiment from a configuration.

Builds the metasurface, objective, and angles from config, then runs relax_then_quantize optimization.

Parameters:

Name	Type	Description	Default
config	ExperimentConfig	Experiment configuration.	required

Returns:

Type	Description
ExperimentResult	ExperimentResult with optimization result and provenance.

set_global_seed(seed)

Set global random seed for reproducibility.

Parameters:

Name	Type	Description	Default
seed	int	Random seed value.	required