modosaic.core.modality¶

modosaic.core.modality ¶

Modality ¶

Modality(model, postprocessor, validators=None)

Bases: ABC

Base class for an executable modality.

A modality bundles the three pieces needed by the pipeline: a generator, a postprocessor, and optional validators. Subclasses provide the concrete Modalities value.

Attributes:

Name	Type	Description
`model`	`ModalityGenerator[T]`	Generator that produces the modality output.
`validators`	`list[ValidatorStep[T, Any]]`	Validation steps executed after generation.
`postprocessor`	`ModalityPostprocessor[T]`	Converts accepted generated output into artifacts.

Initialize a modality from generator, postprocessor, and validators.

Parameters:

Name	Type	Description	Default
`model`	`ModalityGenerator[T]`	Generator used to produce the modality output.	required
`postprocessor`	`ModalityPostprocessor[T]`	Artifact builder for accepted outputs.	required
`validators`	`Iterable[ModalityValidator[T, Any] \| ValidatorStep[T, Any]] \| None`	Optional validator objects or fully configured `ValidatorStep` instances.	`None`

Source code in modosaic/core/modality.py

def __init__(
        self,
        model: ModalityGenerator[T],
        postprocessor: ModalityPostprocessor[T],
        validators: Iterable[
                        ModalityValidator[T, Any] | ValidatorStep[T, Any]
                        ] | None = None,
) -> None:
    """Initialize a modality from generator, postprocessor, and validators.

    Args:
        model: Generator used to produce the modality output.
        postprocessor: Artifact builder for accepted outputs.
        validators: Optional validator objects or fully configured
            `ValidatorStep` instances.
    """
    self.model = model
    self.postprocessor = postprocessor
    self.validators = [
        self._build_validator_step(validator)
        for validator in validators or []
    ]

modality `abstractmethod` `property` ¶

modality

Return the modality type represented by this object.

generate ¶

generate(record)

Generate this modality for a record.

Parameters:

Name	Type	Description	Default
`record`	`ImageRecord`	Input image record.	required

Returns:

Type	Description
`T`	Generated modality output.

Source code in modosaic/core/modality.py

def generate(self, record: ImageRecord) -> T:
    """Generate this modality for a record.

    Args:
        record: Input image record.

    Returns:
        Generated modality output.
    """
    return self.model.generate(record)

postprocess ¶

postprocess(record, generated)

Create experiment artifacts for an accepted modality output.

Parameters:

Name	Type	Description	Default
`record`	`ImageRecord`	Input image record.	required
`generated`	`T`	Accepted generated output.	required

Returns:

Type	Description
`list[ExperimentArtifact[Any]]`	Artifacts ready to be saved by `ExperimentService`.

Source code in modosaic/core/modality.py

def postprocess(self, record: ImageRecord, generated: T) -> list[ExperimentArtifact[Any]]:
    """Create experiment artifacts for an accepted modality output.

    Args:
        record: Input image record.
        generated: Accepted generated output.

    Returns:
        Artifacts ready to be saved by `ExperimentService`.
    """
    return self.postprocessor.process(record, generated)

validate ¶

validate(record, generated, generated_modalities=None)

Run validators for a generated modality output.

Parameters:

Name	Type	Description	Default
`record`	`ImageRecord`	Input image record.	required
`generated`	`T`	Output produced by `generate`.	required
`generated_modalities`	`Mapping[Modalities, Any] \| None`	Accepted outputs from earlier modalities, used to satisfy validator dependencies.	`None`

Returns:

Type	Description
`list[ValidationResult[Any]]`	Validation results in configured validator order.

Source code in modosaic/core/modality.py

def validate(
        self,
        record: ImageRecord,
        generated: T,
        generated_modalities: Mapping[Modalities, Any] | None = None,
) -> list[ValidationResult[Any]]:
    """Run validators for a generated modality output.

    Args:
        record: Input image record.
        generated: Output produced by `generate`.
        generated_modalities: Accepted outputs from earlier modalities,
            used to satisfy validator dependencies.

    Returns:
        Validation results in configured validator order.
    """
    context = generated_modalities or {}
    return [
        validator.validate(record, generated, context)
        for validator in self.validators
    ]

modosaic.core.modality¶