Added recommendation target

aligned/compiler/feature_factory.py

@@ -30,6 +30,7 @@
 from aligned.schemas.target import ClassificationTarget as ClassificationTargetSchemas
 from aligned.schemas.target import ClassTargetProbability
 from aligned.schemas.target import RegressionTarget as RegressionTargetSchemas
+from aligned.schemas.target import RecommendationTarget as RecommendationTargetSchemas
 from aligned.schemas.transformation import EmbeddingModel, Transformation
 from aligned.schemas.vector_storage import VectorStorage
 
@@ -171,6 +172,39 @@ def sort_key(x: tuple[int, FeatureFactory]) -> int:
     return features, derived_features
 
 
+@dataclass
+class RecommendationTarget(FeatureReferencable):
+
+    feature: FeatureFactory
+    rank_feature: FeatureFactory | None = field(default=None)
+
+    _name: str | None = field(default=None)
+    _location: FeatureLocation | None = field(default=None)
+
+    def __set_name__(self, owner, name):
+        self._name = name
+
+    def feature_referance(self) -> FeatureReferance:
+        if not self._name:
+            raise ValueError('Missing name, can not create reference')
+        if not self._location:
+            raise ValueError('Missing location, can not create reference')
+        return FeatureReferance(self._name, self._location, self.feature.dtype)
+
+    def estemating_rank(self, feature: FeatureFactory) -> RecommendationTarget:
+        self.rank_feature = feature
+        return self
+
+    def compile(self) -> RecommendationTargetSchemas:
+        self_ref = self.feature_referance()
+
+        return RecommendationTargetSchemas(
+            self.feature.feature_referance(),
+            feature=self_ref.as_feature(),
+            estimating_rank=self.rank_feature.feature_referance() if self.rank_feature else None,
+        )
+
+
 @dataclass
 class RegressionLabel(FeatureReferencable):
     feature: FeatureFactory
@@ -375,6 +409,9 @@ def as_classification_label(self) -> ClassificationLabel:
     def as_regression_label(self) -> RegressionLabel:
         return RegressionLabel(self)
 
+    def as_recommendation_target(self) -> RecommendationTarget:
+        return RecommendationTarget(self)
+
     def compile(self) -> DerivedFeature:
 
         if not self.transformation:

aligned/compiler/model.py

@@ -14,6 +14,7 @@
     EventTimestamp,
     FeatureFactory,
     FeatureReferencable,
+    RecommendationTarget,
     RegressionLabel,
     TargetProbability,
     ModelVersion,
@@ -203,6 +204,7 @@ class MyModel(ModelContract):
         stream_source=metadata.prediction_stream,
         classification_targets=set(),
         regression_targets=set(),
+        recommendation_targets=set(),
     )
     probability_features: dict[str, set[TargetProbability]] = {}
 
@@ -261,6 +263,8 @@ class MyModel(ModelContract):
             probability_features[feature_name] = probability_features.get(feature_name, set()).union(
                 {feature}
             )
+        elif isinstance(feature, RecommendationTarget):
+            inference_view.recommendation_targets.add(feature.compile())
         elif isinstance(feature, Entity):
             inference_view.entities.add(feature.feature())
         elif isinstance(feature, FeatureFactory):
@@ -299,4 +303,5 @@ class MyModel(ModelContract):
         tags=metadata.tags,
         description=metadata.description,
         dataset_store=metadata.dataset_store,
+        exposed_at_url=metadata.exposed_at_url,
     )

aligned/feature_store.py

@@ -10,7 +10,6 @@
 from prometheus_client import Histogram
 
 from aligned.compiler.model import ModelContractWrapper
-from aligned.schemas.model import FeatureInputVersions
 from aligned.data_file import DataFileReference, upsert_on_column
 from aligned.data_source.batch_data_source import BatchDataSource
 from aligned.enricher import Enricher
@@ -736,11 +735,8 @@ def location(self) -> FeatureLocation:
 
     def raw_string_features(self, except_features: set[str]) -> set[str]:
 
-        if isinstance(self.model.features, FeatureInputVersions):
-            version = self.selected_version or self.model.features.default_version
-            features = self.model.features.features_for(version)
-        else:
-            features = self.model.features
+        version = self.selected_version or self.model.features.default_version
+        features = self.model.features.features_for(version)
 
         return {
             f'{feature.location.identifier}:{feature.name}'

aligned/schemas/feature.py

@@ -286,6 +286,13 @@ def renamed(self, new_name: str) -> Feature:
             constraints=self.constraints,
         )
 
+    def as_reference(self, location: FeatureLocation) -> FeatureReferance:
+        return FeatureReferance(
+            name=self.name,
+            location=location,
+            dtype=self.dtype,
+        )
+
     def __hash__(self) -> int:
         return hash(self.name)
 
@@ -354,6 +361,15 @@ class FeatureReferance(Codable):
     dtype: FeatureType
     # is_derived: bool
 
+    def as_feature(self) -> Feature:
+        return Feature(
+            name=self.name,
+            dtype=self.dtype,
+            description=None,
+            tags=None,
+            constraints=None,
+        )
+
     def __hash__(self) -> int:
         return hash(self.name)
 

aligned/schemas/model.py

@@ -7,7 +7,7 @@
 from aligned.schemas.feature import EventTimestamp, Feature, FeatureReferance
 from aligned.data_source.stream_data_source import StreamDataSource
 from aligned.schemas.event_trigger import EventTrigger
-from aligned.schemas.target import ClassificationTarget, RegressionTarget
+from aligned.schemas.target import ClassificationTarget, RecommendationTarget, RegressionTarget
 from aligned.schemas.derivied_feature import DerivedFeature
 from aligned.data_source.batch_data_source import BatchDataSource
 from aligned.schemas.folder import DatasetStore
@@ -57,6 +57,7 @@ class PredictionsView(Codable):
 
     regression_targets: set[RegressionTarget] | None = field(default=None)
     classification_targets: set[ClassificationTarget] | None = field(default=None)
+    recommendation_targets: set[RecommendationTarget] | None = field(default=None)
 
     @property
     def full_schema(self) -> set[Feature]:
@@ -110,6 +111,8 @@ def labels_estimates_refs(self) -> set[FeatureReferance]:
             return {feature.estimating for feature in self.classification_targets}
         elif self.regression_targets:
             return {feature.estimating for feature in self.regression_targets}
+        elif self.recommendation_targets:
+            return {feature.estimating for feature in self.recommendation_targets}
         else:
             raise ValueError('Found no targets in the model')
 
@@ -118,6 +121,8 @@ def labels(self) -> set[Feature]:
             return {feature.feature for feature in self.classification_targets}
         elif self.regression_targets:
             return {feature.feature for feature in self.regression_targets}
+        elif self.recommendation_targets:
+            return {feature.feature for feature in self.recommendation_targets}
         else:
             raise ValueError('Found no targets in the model')
 
@@ -137,10 +142,7 @@ def __hash__(self) -> int:
         return self.name.__hash__()
 
     def feature_references(self, version: str | None = None) -> set[FeatureReferance]:
-        if isinstance(self.features, FeatureInputVersions):
-            return set(self.features.features_for(version or self.features.default_version))
-        else:
-            return self.features
+        return set(self.features.features_for(version or self.features.default_version))
 
     @property
     def request_all_predictions(self) -> FeatureRequest:

aligned/schemas/target.py

@@ -47,3 +47,15 @@ class ClassificationTarget(Codable):
 
     def __hash__(self) -> int:
         return self.feature.name.__hash__()
+
+
+@dataclass
+class RecommendationTarget(Codable):
+
+    estimating: FeatureReferance
+    feature: Feature
+
+    estimating_rank: FeatureReferance | None = field(default=None)
+
+    def __hash__(self) -> int:
+        return self.feature.name.__hash__()

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "aligned"
-version = "0.0.61"
+version = "0.0.62"
 description = "A data managment and lineage tool for ML applications."
 authors = ["Mats E. Mollestad <mats@mollestad.no>"]
 license = "Apache-2.0"

test_data/titanic-sets.json

@@ -1 +1 @@
-{"raw_data": [], "train_test": [], "train_test_validation": [{"id": "titanic_test", "name": null, "request_result": {"entities": [{"name": "passenger_id", "dtype": {"name": "int32"}, "description": null, "tags": null, "constraints": null}], "features": [{"name": "survived", "dtype": {"name": "bool"}, "description": "If the passenger survived", "tags": null, "constraints": null}, {"name": "is_mr", "dtype": {"name": "bool"}, "description": null, "tags": null, "constraints": null, "depending_on": [{"name": "name", "location": {"name": "titanic", "location": "feature_view"}, "dtype": {"name": "string"}}], "transformation": {"name": "contains", "dtype": {"name": "bool"}, "key": "name", "value": "Mr."}, "depth": 1}, {"name": "age", "dtype": {"name": "float"}, "description": "A float as some have decimals", "tags": null, "constraints": [{"name": "lower_bound_inc", "value": 0.0}, {"name": "upper_bound_inc", "value": 100.0}]}, {"name": "cabin", "dtype": {"name": "string"}, "description": null, "tags": null, "constraints": [{"name": "optional"}]}, {"name": "name", "dtype": {"name": "string"}, "description": null, "tags": null, "constraints": [{"name": "optional"}]}, {"name": "sibsp", "dtype": {"name": "int32"}, "description": "Number of siblings on titanic", "tags": null, "constraints": [{"name": "lower_bound_inc", "value": 0.0}, {"name": "optional"}, {"name": "upper_bound_inc", "value": 20.0}]}, {"name": "is_male", "dtype": {"name": "bool"}, "description": null, "tags": null, "constraints": null, "depending_on": [{"name": "sex", "location": {"name": "titanic", "location": "feature_view"}, "dtype": {"name": "string"}}], "transformation": {"name": "equals", "dtype": {"name": "bool"}, "key": "sex", "value": {"name": "string", "value": "male"}}, "depth": 1}, {"name": "is_female", "dtype": {"name": "bool"}, "description": null, "tags": null, "constraints": null, "depending_on": [{"name": "sex", "location": {"name": "titanic", "location": "feature_view"}, "dtype": {"name": "string"}}], "transformation": {"name": "equals", "dtype": {"name": "bool"}, "key": "sex", "value": {"name": "string", "value": "female"}}, "depth": 1}, {"name": "has_siblings", "dtype": {"name": "bool"}, "description": null, "tags": null, "constraints": null, "depending_on": [{"name": "sibsp", "location": {"name": "titanic", "location": "feature_view"}, "dtype": {"name": "int32"}}], "transformation": {"name": "not-equals", "dtype": {"name": "bool"}, "key": "sibsp", "value": {"name": "int", "value": 0}}, "depth": 1}, {"name": "sex", "dtype": {"name": "string"}, "description": null, "tags": null, "constraints": [{"name": "in_domain", "values": ["male", "female"]}, {"name": "optional"}]}], "event_timestamp": null}, "train_dataset": {"mapping_keys": {}, "type_name": "csv", "path": "test_data/titanic-train.csv", "csv_config": {"seperator": ",", "compression": "infer", "should_write_index": false}}, "test_dataset": {"mapping_keys": {}, "type_name": "csv", "path": "test_data/titanic-test.csv", "csv_config": {"seperator": ",", "compression": "infer", "should_write_index": false}}, "validation_dataset": {"mapping_keys": {}, "type_name": "csv", "path": "test_data/titanic-validate.csv", "csv_config": {"seperator": ",", "compression": "infer", "should_write_index": false}}, "train_size_fraction": 0.6, "test_size_fraction": 0.20000000000000007, "validate_size_fraction": 0.19999999999999996, "target": ["survived"], "description": null, "tags": null}], "active_learning": []}
+{"raw_data": [], "train_test": [], "train_test_validation": [{"id": "titanic_test", "name": null, "request_result": {"entities": [{"name": "passenger_id", "dtype": {"name": "int32"}, "description": null, "tags": null, "constraints": null}], "features": [{"name": "survived", "dtype": {"name": "bool"}, "description": "If the passenger survived", "tags": null, "constraints": null}, {"name": "has_siblings", "dtype": {"name": "bool"}, "description": null, "tags": null, "constraints": null, "depending_on": [{"name": "sibsp", "location": {"name": "titanic", "location": "feature_view"}, "dtype": {"name": "int32"}}], "transformation": {"name": "not-equals", "dtype": {"name": "bool"}, "key": "sibsp", "value": {"name": "int", "value": 0}}, "depth": 1}, {"name": "sex", "dtype": {"name": "string"}, "description": null, "tags": null, "constraints": [{"name": "optional"}, {"name": "in_domain", "values": ["male", "female"]}]}, {"name": "sibsp", "dtype": {"name": "int32"}, "description": "Number of siblings on titanic", "tags": null, "constraints": [{"name": "upper_bound_inc", "value": 20.0}, {"name": "optional"}, {"name": "lower_bound_inc", "value": 0.0}]}, {"name": "is_male", "dtype": {"name": "bool"}, "description": null, "tags": null, "constraints": null, "depending_on": [{"name": "sex", "location": {"name": "titanic", "location": "feature_view"}, "dtype": {"name": "string"}}], "transformation": {"name": "equals", "dtype": {"name": "bool"}, "key": "sex", "value": {"name": "string", "value": "male"}}, "depth": 1}, {"name": "name", "dtype": {"name": "string"}, "description": null, "tags": null, "constraints": [{"name": "optional"}]}, {"name": "is_mr", "dtype": {"name": "bool"}, "description": null, "tags": null, "constraints": null, "depending_on": [{"name": "name", "location": {"name": "titanic", "location": "feature_view"}, "dtype": {"name": "string"}}], "transformation": {"name": "contains", "dtype": {"name": "bool"}, "key": "name", "value": "Mr."}, "depth": 1}, {"name": "age", "dtype": {"name": "float"}, "description": "A float as some have decimals", "tags": null, "constraints": [{"name": "upper_bound_inc", "value": 100.0}, {"name": "lower_bound_inc", "value": 0.0}]}, {"name": "is_female", "dtype": {"name": "bool"}, "description": null, "tags": null, "constraints": null, "depending_on": [{"name": "sex", "location": {"name": "titanic", "location": "feature_view"}, "dtype": {"name": "string"}}], "transformation": {"name": "equals", "dtype": {"name": "bool"}, "key": "sex", "value": {"name": "string", "value": "female"}}, "depth": 1}, {"name": "cabin", "dtype": {"name": "string"}, "description": null, "tags": null, "constraints": [{"name": "optional"}]}], "event_timestamp": null}, "train_dataset": {"mapping_keys": {}, "type_name": "csv", "path": "test_data/titanic-train.csv", "csv_config": {"seperator": ",", "compression": "infer", "should_write_index": false}}, "test_dataset": {"mapping_keys": {}, "type_name": "csv", "path": "test_data/titanic-test.csv", "csv_config": {"seperator": ",", "compression": "infer", "should_write_index": false}}, "validation_dataset": {"mapping_keys": {}, "type_name": "csv", "path": "test_data/titanic-validate.csv", "csv_config": {"seperator": ",", "compression": "infer", "should_write_index": false}}, "train_size_fraction": 0.6, "test_size_fraction": 0.20000000000000007, "validate_size_fraction": 0.19999999999999996, "target": ["survived"], "description": null, "tags": null}], "active_learning": []}

test_data/titanic-test.csv

@@ -1,21 +1,21 @@
-passenger_id,has_siblings,is_mr,age,cabin,name,sibsp,is_male,is_female,survived,sex
-61,False,True,22.0,,"Sirayanian, Mr. Orsen",0,True,False,False,male
-62,False,False,38.0,B28,"Icard, Miss. Amelie",0,False,True,True,female
-63,True,True,45.0,C83,"Harris, Mr. Henry Birkhardt",1,True,False,False,male
-64,True,False,4.0,,"Skoog, Master. Harald",3,True,False,False,male
-65,False,True,,,"Stewart, Mr. Albert A",0,True,False,False,male
-66,True,False,,,"Moubarek, Master. Gerios",1,True,False,True,male
-67,False,True,29.0,F33,"Nye, Mrs. (Elizabeth Ramell)",0,False,True,True,female
-68,False,True,19.0,,"Crease, Mr. Ernest James",0,True,False,False,male
-69,True,False,17.0,,"Andersson, Miss. Erna Alexandra",4,False,True,True,female
-70,True,True,26.0,,"Kink, Mr. Vincenz",2,True,False,False,male
-71,False,True,32.0,,"Jenkin, Mr. Stephen Curnow",0,True,False,False,male
-72,True,False,16.0,,"Goodwin, Miss. Lillian Amy",5,False,True,False,female
-73,False,True,21.0,,"Hood, Mr. Ambrose Jr",0,True,False,False,male
-74,True,True,26.0,,"Chronopoulos, Mr. Apostolos",1,True,False,False,male
-75,False,True,32.0,,"Bing, Mr. Lee",0,True,False,True,male
-76,False,True,25.0,F G73,"Moen, Mr. Sigurd Hansen",0,True,False,False,male
-77,False,True,,,"Staneff, Mr. Ivan",0,True,False,False,male
-78,False,True,,,"Moutal, Mr. Rahamin Haim",0,True,False,False,male
-79,False,False,0.83,,"Caldwell, Master. Alden Gates",0,True,False,True,male
-80,False,False,30.0,,"Dowdell, Miss. Elizabeth",0,False,True,True,female
+survived,has_siblings,sex,sibsp,is_male,name,passenger_id,is_mr,age,is_female,cabin
+False,False,male,0,True,"Sirayanian, Mr. Orsen",61,True,22.0,False,
+True,False,female,0,False,"Icard, Miss. Amelie",62,False,38.0,True,B28
+False,True,male,1,True,"Harris, Mr. Henry Birkhardt",63,True,45.0,False,C83
+False,True,male,3,True,"Skoog, Master. Harald",64,False,4.0,False,
+False,False,male,0,True,"Stewart, Mr. Albert A",65,True,,False,
+True,True,male,1,True,"Moubarek, Master. Gerios",66,False,,False,
+True,False,female,0,False,"Nye, Mrs. (Elizabeth Ramell)",67,True,29.0,True,F33
+False,False,male,0,True,"Crease, Mr. Ernest James",68,True,19.0,False,
+True,True,female,4,False,"Andersson, Miss. Erna Alexandra",69,False,17.0,True,
+False,True,male,2,True,"Kink, Mr. Vincenz",70,True,26.0,False,
+False,False,male,0,True,"Jenkin, Mr. Stephen Curnow",71,True,32.0,False,
+False,True,female,5,False,"Goodwin, Miss. Lillian Amy",72,False,16.0,True,
+False,False,male,0,True,"Hood, Mr. Ambrose Jr",73,True,21.0,False,
+False,True,male,1,True,"Chronopoulos, Mr. Apostolos",74,True,26.0,False,
+True,False,male,0,True,"Bing, Mr. Lee",75,True,32.0,False,
+False,False,male,0,True,"Moen, Mr. Sigurd Hansen",76,True,25.0,False,F G73
+False,False,male,0,True,"Staneff, Mr. Ivan",77,True,,False,
+False,False,male,0,True,"Moutal, Mr. Rahamin Haim",78,True,,False,
+True,False,male,0,True,"Caldwell, Master. Alden Gates",79,False,0.83,False,
+True,False,female,0,False,"Dowdell, Miss. Elizabeth",80,False,30.0,True,