Anvil Workflows

Workflow implementations for Anvil models.

class openadmet.models.anvil.workflow.AnvilDeepLearningWorkflow(*, metadata: ~openadmet.models.anvil.specification.Metadata, data_spec: ~openadmet.models.anvil.specification.DataSpec, transform: ~openadmet.models.transforms.transform_base.TransformBase | None = None, split: ~openadmet.models.split.split_base.SplitterBase, feat: ~openadmet.models.features.feature_base.FeaturizerBase, model: ~openadmet.models.architecture.model_base.ModelBase, ensemble: ~openadmet.models.active_learning.ensemble_base.EnsembleBase | None = None, trainer: ~openadmet.models.trainer.trainer_base.TrainerBase, evals: list[openadmet.models.eval.eval_base.EvalBase], model_kwargs: dict = <factory>, ensemble_kwargs: dict = <factory>, debug: bool = False, resolved_output_dir: ~pathlib._local.Path | None = None)[source]

Bases: AnvilWorkflowBase

Workflow for running deep learning Anvil configuration.

check_finetuning_paths()[source]

Check that finetuning path pairs are consistent and exist on disk.

Both param_path and serial_path must be provided together (or neither). When both are provided, both paths must exist before training begins. The same requirement applies to param_paths / serial_paths for ensemble workflows, which must additionally be equal-length lists.

Raises:: ValueError – If exactly one of the path pair is provided, if provided paths do not exist on disk, or if ensemble path lists have unequal length.

check_if_val_needed()[source]

Check if validation set is needed or not.

Raises:: ValueError – If ensemble is specified but no validation set is requested.

check_no_transform()[source]

Check that no transform step is specified.

Raises:: ValueError – If a transform step is specified in the recipe.

model_config: ClassVar[ConfigDict] = {}: Configuration for the model, should be a dictionary conforming to [ConfigDict][pydantic.config.ConfigDict].

model_post_init(context: Any, /) → None

This function is meant to behave like a BaseModel method to initialize private attributes.

It takes context as an argument since that’s what pydantic-core passes when calling it.

Args:: self: The BaseModel instance. context: The context.

run(output_dir: PathLike = 'anvil_training', debug: bool = False, tag: str = None) → Any[source]

Run the workflow.

Parameters:

output_dir (PathLike, optional) – Directory to save outputs, by default “anvil_training”
debug (bool, optional) – Whether to run in debug mode, by default False
tag (str, optional) – Tag to override the one in the recipe, by default None

Returns:

Result of the workflow run

Return type:

Any

class openadmet.models.anvil.workflow.AnvilWorkflow(*, metadata: ~openadmet.models.anvil.specification.Metadata, data_spec: ~openadmet.models.anvil.specification.DataSpec, transform: ~openadmet.models.transforms.transform_base.TransformBase | None = None, split: ~openadmet.models.split.split_base.SplitterBase, feat: ~openadmet.models.features.feature_base.FeaturizerBase, model: ~openadmet.models.architecture.model_base.ModelBase, ensemble: ~openadmet.models.active_learning.ensemble_base.EnsembleBase | None = None, trainer: ~openadmet.models.trainer.trainer_base.TrainerBase, evals: list[openadmet.models.eval.eval_base.EvalBase], model_kwargs: dict = <factory>, ensemble_kwargs: dict = <factory>, debug: bool = False, resolved_output_dir: ~pathlib._local.Path | None = None)[source]