GH-18: feat(pipeline): Add serialization for pipeline scores

src/getml_io/getml/scores.py

@@ -0,0 +1,33 @@
+from collections.abc import Sequence
+from datetime import datetime
+from typing import Annotated, Literal
+
+from pydantic import BaseModel, Field
+
+
+class _Score(BaseModel):
+    date_time: datetime
+    set_used: str
+    target: str
+
+
+class ClassificationScore(_Score):
+    accuracy: float
+    auc: float
+    cross_entropy: float
+    type: Literal["classification"] = "classification"
+
+
+class RegressionScore(_Score):
+    mae: float
+    rmse: float
+    rsquared: float
+    type: Literal["regression"] = "regression"
+
+
+Score = Annotated[
+    ClassificationScore | RegressionScore,
+    Field(discriminator="type"),
+]
+
+Scores = Sequence[ClassificationScore] | Sequence[RegressionScore]

src/getml_io/metadata/pipeline_information.py

@@ -14,6 +14,7 @@
 from getml_io.getml.features import Features
 from getml_io.getml.predictors import FeatureSelector, Predictor
 from getml_io.getml.preprocessors import Preprocessor
+from getml_io.getml.scores import Scores
 from getml_io.metadata.data_model_information import DataModelInformation
 from getml_io.metadata.dataframe_information import DataFrameInformationByName
 from getml_io.metadata.placeholder_information import PlaceholderInformation
@@ -44,7 +45,7 @@ class PipelineInformation(BaseModel):
     targets: Sequence[str]
     data_model: DataModelInformation
     features: Features
-    # scores # TODO @urfoex: #18
+    scores: Scores
     # columns # TODO @urfoex: #50
     # metadata # TODO @urfoex: #51
     # tables # TODO @urfoex: #52

src/getml_io/serialize/exception.py

@@ -1,5 +1,7 @@
 from pathlib import Path
 
+from getml.pipeline.score import Score as GetMLScore
+
 from getml_io.getml.roles import Role
 from getml_io.metadata.dataframe_information import (
     ROLE_TO_COLUMN_STATISTICS_TYPE_MAPPING,
@@ -91,3 +93,19 @@ def __init__(
             f"Supported are: {list(ROLE_TO_COLUMN_STATISTICS_TYPE_MAPPING.keys())}."
         )
         super().__init__(message)
+
+
+class WrongPipelineScoreTypeError(GetMLIOError):
+    """Exception raised when the type of a score does not match the expected type."""
+
+    def __init__(
+        self,
+        expected_type: type[GetMLScore],
+        received_type: type[GetMLScore],
+    ) -> None:
+        """Initialize the exception with a custom message."""
+        message = (
+            f"Expected score type {expected_type.__name__}, "
+            f"but received {received_type.__name__}."
+        )
+        super().__init__(message)

src/getml_io/serialize/pipeline.py

@@ -13,6 +13,10 @@
 )
 from getml.pipeline import Features as GetMLFeatures
 from getml.pipeline import Pipeline
+from getml.pipeline import Scores as GetMLScores
+from getml.pipeline.score import ClassificationScore as GetMLClassificationScore
+from getml.pipeline.score import RegressionScore as GetMLRegressionScore
+from getml.pipeline.score import Score as GetMLScore
 from numpy.typing import NDArray
 
 from getml_io.getml.feature_learning import (
@@ -43,6 +47,7 @@
     Substring,
     TextFieldSplitter,
 )
+from getml_io.getml.scores import ClassificationScore, RegressionScore, Scores
 from getml_io.metadata.dataframe_information import DataFrameInformationByName
 from getml_io.metadata.pipeline_information import (
     LossFunction,
@@ -51,6 +56,7 @@
 from getml_io.serialize.data_model import serialize_data_model
 from getml_io.serialize.dataframe_information import derive_instances_with_relative_path
 from getml_io.serialize.dataframe_or_view import serialize_dataframe_or_view
+from getml_io.serialize.exception import WrongPipelineScoreTypeError
 from getml_io.serialize.pipeline_information import serialize_pipeline_information
 from getml_io.serialize.placeholder import serialize_placeholder
 from getml_io.utils.convert import (
@@ -123,7 +129,7 @@ def serialize_pipeline(
         targets=pipeline.targets,
         data_model=serialize_data_model(pipeline.data_model),
         features=serialize_features(pipeline.features),
-        # scores # TODO @urfoex: #18
+        scores=serialize_scores(pipeline.scores),
         # columns # TODO @urfoex: #50
         # metadata # TODO @urfoex: #51
         # tables # TODO @urfoex: #52
@@ -331,3 +337,64 @@ def serialize_features(features: GetMLFeatures) -> Features:
         )
         for feature in features
     }
+
+
+def serialize_scores(scores: GetMLScores) -> Scores:
+    """Serialize getML Scores into a Scores object.
+
+    Args:
+        scores: The getML Scores to serialize.
+
+    Returns:
+        Scores: The serialized Scores information.
+
+    """
+    return (
+        _serialize_classification_scores(list(scores))
+        if scores.is_classification
+        else _serialize_regression_scores(list(scores))
+    )
+
+
+def _serialize_classification_scores(
+    scores: list[GetMLScore],
+) -> list[ClassificationScore]:
+    classification_scores: list[ClassificationScore] = []
+    for score in scores:
+        if not isinstance(score, GetMLClassificationScore):
+            raise WrongPipelineScoreTypeError(
+                GetMLClassificationScore,
+                type(score),
+            )
+        classification_scores.append(
+            ClassificationScore(
+                date_time=score.date_time,
+                set_used=score.set_used,
+                target=score.target,
+                accuracy=score.accuracy,
+                auc=score.auc,
+                cross_entropy=score.cross_entropy,
+            ),
+        )
+    return classification_scores
+
+
+def _serialize_regression_scores(scores: list[GetMLScore]) -> list[RegressionScore]:
+    regression_scores: list[RegressionScore] = []
+    for score in scores:
+        if not isinstance(score, GetMLRegressionScore):
+            raise WrongPipelineScoreTypeError(
+                GetMLRegressionScore,
+                type(score),
+            )
+        regression_scores.append(
+            RegressionScore(
+                date_time=score.date_time,
+                set_used=score.set_used,
+                target=score.target,
+                mae=score.mae,
+                rmse=score.rmse,
+                rsquared=score.rsquared,
+            ),
+        )
+    return regression_scores

tests/integration/assertions.py

@@ -2,6 +2,7 @@
 from pathlib import Path
 
 from getml_io.getml.features import Features
+from getml_io.getml.scores import Scores
 from getml_io.metadata.container_information import ContainerInformation
 from getml_io.metadata.dataframe_information import (
     ColumnProfile,
@@ -213,6 +214,10 @@ def assert_pipeline_information(
         pipeline_information.features,
         expected_pipeline_information.features,
     )
+    assert_scores(
+        pipeline_information.scores,
+        expected_pipeline_information.scores,
+    )
 
 
 def assert_features(
@@ -229,3 +234,23 @@ def assert_features(
         assert feature.sql is not None
         assert feature.importance is not None
         assert feature.correlation is not None
+
+
+def assert_scores(
+    scores: Scores,
+    expected_scores: Scores,
+) -> None:
+    assert len(scores) == len(expected_scores)
+    for score, expected_score in zip(scores, expected_scores, strict=True):
+        assert score.type == expected_score.type
+        assert score.target == expected_score.target
+        assert score.date_time is not None
+        assert score.set_used == expected_score.set_used
+        if score.type == "classification":
+            assert score.accuracy is not None
+            assert score.auc is not None
+            assert score.cross_entropy is not None
+        elif score.type == "regression":
+            assert score.mae is not None
+            assert score.rmse is not None
+            assert score.rsquared is not None

tests/integration/data/loans/expected.pipeline.json

@@ -631,11 +631,11 @@
   "feature_learners": [
     {
       "aggregation": [
-        "SUM",
+        "COUNT",
         "AVG",
+        "SUM",
         "MAX",
-        "MIN",
-        "COUNT"
+        "MIN"
       ],
       "allow_sets": true,
       "delta_t": 0.0,
@@ -649,19 +649,19 @@
       "num_threads": 0,
       "propositionalization": {
         "aggregation": [
-          "SUM",
-          "AVG",
           "MEDIAN",
-          "TREND",
-          "COUNT MINUS COUNT DISTINCT",
-          "MODE",
           "STDDEV",
-          "MAX",
-          "MIN",
+          "COUNT",
+          "AVG",
+          "TREND",
           "COUNT DISTINCT",
+          "SUM",
           "FIRST",
+          "MODE",
           "LAST",
-          "COUNT"
+          "COUNT MINUS COUNT DISTINCT",
+          "MAX",
+          "MIN"
         ],
         "delta_t": 0.0,
         "loss_function": "CrossEntropyLoss",
@@ -1062,5 +1062,25 @@
       "correlation": 0.06660193285904141,
       "sql": ""
     }
-  }
+  },
+  "scores": [
+    {
+      "date_time": "2025-08-19T22:33:06",
+      "set_used": "train",
+      "target": "default",
+      "accuracy": 0.9825708061002179,
+      "auc": 0.9952295229522934,
+      "cross_entropy": 0.08200917844672241,
+      "type": "classification"
+    },
+    {
+      "date_time": "2025-08-19T22:33:07",
+      "set_used": "test",
+      "target": "default",
+      "accuracy": 0.9551569506726457,
+      "auc": 0.8903818953323912,
+      "cross_entropy": 0.1751293856880503,
+      "type": "classification"
+    }
+  ]
 }

tests/integration/data/numerical/expected.pipeline.json

@@ -570,19 +570,19 @@
       "num_threads": 0,
       "propositionalization": {
         "aggregation": [
-          "SUM",
-          "AVG",
           "MEDIAN",
-          "TREND",
-          "COUNT MINUS COUNT DISTINCT",
-          "MODE",
           "STDDEV",
-          "MAX",
-          "MIN",
+          "COUNT",
+          "AVG",
+          "TREND",
           "COUNT DISTINCT",
+          "SUM",
           "FIRST",
+          "MODE",
           "LAST",
-          "COUNT"
+          "COUNT MINUS COUNT DISTINCT",
+          "MAX",
+          "MIN"
         ],
         "delta_t": 0.0,
         "loss_function": "SquareLoss",
@@ -817,5 +817,25 @@
       "correlation": -0.09593210355580734,
       "sql": ""
     }
-  }
+  },
+  "scores": [
+    {
+      "date_time": "2025-08-19T22:33:16",
+      "set_used": "train",
+      "target": "targets",
+      "mae": 0.16209001296605818,
+      "rmse": 0.25502448762753166,
+      "rsquared": 0.999961056217837,
+      "type": "regression"
+    },
+    {
+      "date_time": "2025-08-19T22:33:18",
+      "set_used": "test",
+      "target": "targets",
+      "mae": 0.47556934465061534,
+      "rmse": 0.6929843656362203,
+      "rsquared": 0.9996960826789243,
+      "type": "regression"
+    }
+  ]
 }