Serialization

ModelPipeline Serialization

Save and load pipeline configurations for reproducibility and version control.

Save

from epftoolbox2.pipelines import ModelPipeline
from epftoolbox2.models import OLSModel, LassoCVModel
from epftoolbox2.evaluators import MAEEvaluator
from epftoolbox2.exporters import TerminalExporter, ExcelExporter

predictors = [
    "load_actual",
    "is_monday_d+{horizon}",
    "price_d-1",
    "warsaw_temperature_2m_d+{horizon}",
]

pipeline = (
    ModelPipeline()
    .add_model(OLSModel(predictors=predictors, training_window=365, name="OLS"))
    .add_model(LassoCVModel(predictors=predictors, training_window=365, cv=5, name="LassoCV"))
    .add_evaluator(MAEEvaluator())
    .add_exporter(TerminalExporter())
    .add_exporter(ExcelExporter("results.xlsx"))
)

pipeline.save("model_pipeline.yaml")

Load

pipeline = ModelPipeline.load("model_pipeline.yaml")
report = pipeline.run(data=df, test_start="2024-04-01", test_end="2024-06-01")

YAML Format

models:
- class: OLSModel
  params:
    predictors: &id001
    - load_actual
    - is_monday_d+{horizon}
    - price_d-1
    - warsaw_temperature_2m_d+{horizon}
    training_window: 365
    name: OLS
- class: LassoCVModel
  params:
    predictors: *id001
    training_window: 365
    cv: 5
    name: LassoCV
evaluators:
- class: MAEEvaluator
  params: {}
exporters:
- class: TerminalExporter
  params:
    show: [summary, horizon]
- class: ExcelExporter
  params:
    path: results.xlsx

Serialization Rules

Type	Behavior
str, int, float, bool, list, dict, None	Serialized as-is
Path	Converted to str
Callable predictors	Raises ValueError — use "{horizon}" string templates instead
Private attrs (_name)	Excluded

Callable Predictors

Callable predictors cannot be serialized to YAML. Replace them with "{horizon}" string templates before saving:

# Instead of:
lambda h: f"warsaw_temperature_2m_d+{h}"

# Use:
"warsaw_temperature_2m_d+{horizon}"

YAML Aliases

If the same predictor list object is passed to multiple models, PyYAML emits a YAML anchor (&id001) and alias (*id001). This is valid YAML — the aliased list is reconstructed independently for each model on load.

Registering Custom Models/Evaluators/Exporters

Add to COMPONENT_REGISTRY in model_pipeline.py:

COMPONENT_REGISTRY["models"]["MyModel"] = "mypackage.models.my_model"
COMPONENT_REGISTRY["evaluators"]["MyEvaluator"] = "mypackage.evaluators.my_evaluator"
COMPONENT_REGISTRY["exporters"]["MyExporter"] = "mypackage.exporters.my_exporter"