Skip to content

config

Declarative (YAML / dict) configuration of a calibration problem.

The schema keys on the same dotted "<container>.<field>" target paths used throughout the calibration package, so a calibration setup can live next to a crop preset as data:

1
2
3
4
5
6
7
8
parameters:
  crop.tsum1:       {bounds: [1000, 2000]}
  crop.tbasem:      {bounds: [-2, 5], kind: continuous}
  crop.fltb@0.5:    {bounds: [0.30, 0.70]}
  crop.fltb@0.646:  {bounds: [0.10, 0.50]}
constraints:
  - members: [crop.fltb@0.5, crop.fltb@0.646]
    order: descending

The loader returns (specs, groups) ready to pass to ~torchcrop.calibration.manager.CalibrationManager.

load_calibration_config(source)

Load a calibration problem from a YAML file, path, or dict.

Parameters:

Name Type Description Default
source str | Path | dict[str, Any]

Path to a YAML file, or an already-parsed config dict.

 required

Returns:

Type Description
tuple[list[ParameterSpec], list[ConstraintGroup]]

(specs, groups) ready for ~torchcrop.calibration.manager.CalibrationManager.
Source code in torchcrop/calibration/config.py 
def load_calibration_config(
    source: str | Path | dict[str, Any],
) -> tuple[list[ParameterSpec], list[ConstraintGroup]]:
    """Load a calibration problem from a YAML file, path, or dict.

    Args:
        source: Path to a YAML file, or an already-parsed config dict.

    Returns:
        ``(specs, groups)`` ready for
        `~torchcrop.calibration.manager.CalibrationManager`.
    """
    if isinstance(source, dict):
        return specs_from_config(source)
    yaml = _import_yaml()
    with open(source) as fh:
        return specs_from_config(yaml.safe_load(fh) or {})

specs_from_config(config)

Build specs and groups from an already-parsed config mapping.

Parameters:

Name Type Description Default
config dict[str, Any]

Mapping with optional parameters and constraints keys (see module docstring for the schema).

 required

Returns:

Type Description
tuple[list[ParameterSpec], list[ConstraintGroup]]

(specs, groups).

Exceptions:

Type Description
ValueError

If the mapping is malformed.
Source code in torchcrop/calibration/config.py 
def specs_from_config(
    config: dict[str, Any],
) -> tuple[list[ParameterSpec], list[ConstraintGroup]]:
    """Build specs and groups from an already-parsed config mapping.

    Args:
        config: Mapping with optional ``parameters`` and ``constraints`` keys
            (see module docstring for the schema).

    Returns:
        ``(specs, groups)``.

    Raises:
        ValueError: If the mapping is malformed.
    """
    specs: list[ParameterSpec] = []
    for name, opts in (config.get("parameters") or {}).items():
        opts = opts or {}
        cats = opts.get("categories")
        specs.append(
            ParameterSpec(
                name=name,
                bounds=_as_bounds(opts.get("bounds")),
                kind=opts.get("kind", "continuous"),
                transform=opts.get("transform"),
                init=opts.get("init"),
                categories=tuple(cats) if cats is not None else None,
            )
        )

    groups: list[ConstraintGroup] = []
    for entry in config.get("constraints") or []:
        groups.append(
            ConstraintGroup(
                members=tuple(entry["members"]),
                order=entry.get("order", "ascending"),
            )
        )
    return specs, groups