mml.core.scripts.pipeline_configuration

class PipelineCfg

Bases: object

__init__(pipeline_cfg: DictConfig, restrict_keys: List[str] | None = None) → None

PipelineCfg holds relevant configuration elements of a training pipeline to store, reproduce and leverage knowledge at a later point. The intended usage is by invoking from_cfg() on the full mml config, which will produce a masked copy only focussing on a subset of config keys.

Parameters:

• pipeline_cfg (DictConfig) – a config

• restrict_keys (Optional[List[str]]) – which config keys to focus upon

activate(current_cfg: DictConfig) → None

To be used as a config manager, activates this pipeline upon the currently active mml config. When the context exits, the original configuration is restored.

Parameters:

current_cfg (DictConfig) – the currently active mml config

Returns:

no return value, the mml config is modified in place

clone() → PipelineCfg

Convenience method to deepcopy the configuration.

Returns:

a deepcopy of the configuration

classmethod from_cfg(current_cfg: DictConfig, restrict_keys: List[str] | None = None) → PipelineCfg

Extracts relevant pipeline keys from current config determined by restrict_keys.

Parameters:

• current_cfg (DictConfig) – the FULL config to derive the pipeline configuration from

• restrict_keys (Optional[List[str]]) – which config keys to focus upon

Returns:

classmethod load(path: Path, pipeline_keys: List[str] | None = None) → PipelineCfg

Load a stores pipeline configuration (or blueprint) from path.

Parameters:

• path (Path) – the path to the stored file

• pipeline_keys (Optional[List[str]]) – which config keys to focus upon

Returns:

the loaded pipeline configuration (restricted to provided pipeline keys)

store(task_struct: TaskStruct, as_blueprint: bool = False) → Path

Store this pipeline. Requires a task struct to determine task name. If blueprint is set, this will be stored in the BLUEPRINTS folder instead of PIPELINES. This allows for easier re-usage.

Parameters:

• task_struct (TaskStruct) – struct of the task this pipeline has or should be applied upon

• as_blueprint (bool) – if true store as blueprint otherwise as pipeline

Returns:

the path to the stored file