Compose Hydra Training Config

Compose Hydra Training Configs Without Launching Training.

class scripts.training.compose_hydra_training_config.MaterializedConfigBundle(training_config, dataset_config)

Bases: object

Store the resolved training and dataset configuration payloads.

Parameters:

• training_config (dict[str, Any])

• dataset_config (dict[str, Any])

training_config: dict[str, Any]

dataset_config: dict[str, Any]

scripts.training.compose_hydra_training_config.build_argument_parser()

Build the command-line argument parser.

Return type:

ArgumentParser

scripts.training.compose_hydra_training_config.parse_command_line_arguments()

Parse command-line arguments.

Return type:

Namespace

scripts.training.compose_hydra_training_config.import_hydra_dependencies()

Import Hydra dependencies with a repository-specific error message.

Returns:

Tuple containing compose, initialize_config_dir, and OmegaConf.

Raises:

RuntimeError – If hydra-core or omegaconf is not installed.

scripts.training.compose_hydra_training_config.resolve_path_argument(path_value)

Resolve one repository-relative or absolute path argument.

Parameters:

path_value (str) – Path supplied by the command line or default settings.

Returns:

Absolute path resolved against the repository root when needed.

Return type:

Path

scripts.training.compose_hydra_training_config.remove_hydra_internal_keys(configuration_dictionary)

Remove Hydra-only composition keys from the training payload.

Parameters:

configuration_dictionary (dict[str, Any]) – Resolved Hydra configuration dictionary.

Returns:

Training configuration dictionary without composition helper keys.

Return type:

dict[str, Any]

scripts.training.compose_hydra_training_config.validate_materialized_training_config(training_config)

Validate the minimum repository training-config contract.

Parameters:

training_config (dict[str, Any]) – Materialized training configuration payload.

Raises:

AssertionError – If the materialized config is missing required fields.

Return type:

None

scripts.training.compose_hydra_training_config.validate_materialized_dataset_config(dataset_config, training_config)

Validate the minimum dataset-config contract.

Parameters:

• dataset_config (dict[str, Any]) – Materialized dataset configuration payload.

• training_config (dict[str, Any]) – Materialized training configuration payload.

Raises:

AssertionError – If required dataset sections are missing.

Return type:

None

scripts.training.compose_hydra_training_config.build_default_training_output_path(training_config)

Build the default materialized training-config output path.

Parameters:

training_config (dict[str, Any]) – Materialized training configuration payload.

Returns:

Absolute default output path for the materialized training config.

Return type:

Path

scripts.training.compose_hydra_training_config.build_default_dataset_output_path(training_config)

Build the default materialized dataset-config output path.

Parameters:

training_config (dict[str, Any]) – Materialized training configuration payload.

Returns:

Absolute default output path for the materialized dataset config.

Return type:

Path

scripts.training.compose_hydra_training_config.compose_hydra_training_config(config_dir, config_name, override_list)

Compose the Hydra config and extract repository payloads.

Parameters:

• config_dir (Path) – Hydra config directory.

• config_name (str) – Hydra config name.

• override_list (Sequence[str]) – Hydra command-line override list.

Returns:

Materialized config bundle containing training and dataset payloads.

Return type:

MaterializedConfigBundle

scripts.training.compose_hydra_training_config.write_yaml_file(payload, output_path)

Write one YAML payload with repository-standard formatting.

Parameters:

• payload (dict[str, Any]) – YAML-serializable dictionary payload.

• output_path (Path) – Absolute output path.

Return type:

None

scripts.training.compose_hydra_training_config.main()

Run the Hydra materialization workflow.

Return type:

None