Skip to content

Model

Bases: StaticCollection

A class describing a model that was trained on a particular dataset.

Attributes:

Name Type Description
name String

The name of the model.

metadata Dictionary

A dictionary of metadata that describes the model.

Examples:

>>> Model.create(name="model1")
>>> Model.create(name="model1", metadata={})
>>> Model.create(name="model1", metadata={"foo": "bar", "pi": 3.14})
Source code in valor/coretypes.py
 639
 640
 641
 642
 643
 644
 645
 646
 647
 648
 649
 650
 651
 652
 653
 654
 655
 656
 657
 658
 659
 660
 661
 662
 663
 664
 665
 666
 667
 668
 669
 670
 671
 672
 673
 674
 675
 676
 677
 678
 679
 680
 681
 682
 683
 684
 685
 686
 687
 688
 689
 690
 691
 692
 693
 694
 695
 696
 697
 698
 699
 700
 701
 702
 703
 704
 705
 706
 707
 708
 709
 710
 711
 712
 713
 714
 715
 716
 717
 718
 719
 720
 721
 722
 723
 724
 725
 726
 727
 728
 729
 730
 731
 732
 733
 734
 735
 736
 737
 738
 739
 740
 741
 742
 743
 744
 745
 746
 747
 748
 749
 750
 751
 752
 753
 754
 755
 756
 757
 758
 759
 760
 761
 762
 763
 764
 765
 766
 767
 768
 769
 770
 771
 772
 773
 774
 775
 776
 777
 778
 779
 780
 781
 782
 783
 784
 785
 786
 787
 788
 789
 790
 791
 792
 793
 794
 795
 796
 797
 798
 799
 800
 801
 802
 803
 804
 805
 806
 807
 808
 809
 810
 811
 812
 813
 814
 815
 816
 817
 818
 819
 820
 821
 822
 823
 824
 825
 826
 827
 828
 829
 830
 831
 832
 833
 834
 835
 836
 837
 838
 839
 840
 841
 842
 843
 844
 845
 846
 847
 848
 849
 850
 851
 852
 853
 854
 855
 856
 857
 858
 859
 860
 861
 862
 863
 864
 865
 866
 867
 868
 869
 870
 871
 872
 873
 874
 875
 876
 877
 878
 879
 880
 881
 882
 883
 884
 885
 886
 887
 888
 889
 890
 891
 892
 893
 894
 895
 896
 897
 898
 899
 900
 901
 902
 903
 904
 905
 906
 907
 908
 909
 910
 911
 912
 913
 914
 915
 916
 917
 918
 919
 920
 921
 922
 923
 924
 925
 926
 927
 928
 929
 930
 931
 932
 933
 934
 935
 936
 937
 938
 939
 940
 941
 942
 943
 944
 945
 946
 947
 948
 949
 950
 951
 952
 953
 954
 955
 956
 957
 958
 959
 960
 961
 962
 963
 964
 965
 966
 967
 968
 969
 970
 971
 972
 973
 974
 975
 976
 977
 978
 979
 980
 981
 982
 983
 984
 985
 986
 987
 988
 989
 990
 991
 992
 993
 994
 995
 996
 997
 998
 999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
class Model(StaticCollection):
    """
    A class describing a model that was trained on a particular dataset.

    Attributes
    ----------
    name : String
        The name of the model.
    metadata : Dictionary
        A dictionary of metadata that describes the model.

    Examples
    --------
    >>> Model.create(name="model1")
    >>> Model.create(name="model1", metadata={})
    >>> Model.create(name="model1", metadata={"foo": "bar", "pi": 3.14})
    """

    name: String = String.symbolic(owner="model", name="name")
    metadata: Dictionary = Dictionary.symbolic(owner="model", name="metadata")

    def __init__(
        self,
        *,
        name: str,
        metadata: Optional[dict] = None,
        connection: Optional[ClientConnection] = None,
    ):
        """
        Creates a local instance of a model.

        Use 'Model.create' classmethod to create a model with persistence.

        Parameters
        ----------
        name : String
            The name of the model.
        metadata : Dictionary
            A dictionary of metadata that describes the model.
        connection : ClientConnection, optional
            An initialized client connection.
        """
        self.conn = connection
        super().__init__(name=name, metadata=metadata if metadata else dict())

    @classmethod
    def create(
        cls,
        name: str,
        metadata: Optional[Dict[str, Any]] = None,
        connection: Optional[ClientConnection] = None,
        **_,
    ) -> Model:
        """
        Creates a model that persists in the back end.

        Parameters
        ----------
        name : str
            The name of the model.
        metadata : dict, optional
            A dictionary of metadata that describes the model.
        connection : ClientConnection, optional
            An initialized client connection.
        """
        model = cls(name=name, metadata=metadata, connection=connection)
        Client(connection).create_model(model)
        return model

    @classmethod
    def get(
        cls,
        name: str,
        connection: Optional[ClientConnection] = None,
    ) -> Union[Model, None]:
        """
        Retrieves a model from the back end database.

        Parameters
        ----------
        name : str
            The name of the model.
        connection : ClientConnnetion, optional
            An optional Valor client object for interacting with the API.

        Returns
        -------
        Union[valor.Model, None]
            The model or 'None' if it doesn't exist.
        """
        return Client(connection).get_model(name)

    def add_prediction(
        self,
        dataset: Dataset,
        prediction: Prediction,
    ) -> None:
        """
        Add a prediction to the model.

        Parameters
        ----------
        dataset : valor.Dataset
            The dataset that is being operated over.
        prediction : valor.Prediction
            The prediction to create.
        """
        Client(self.conn).create_predictions(
            dataset=dataset,
            model=self,
            predictions=[prediction],
        )

    def add_predictions(
        self,
        dataset: Dataset,
        predictions: List[Prediction],
        timeout: Optional[float] = 10.0,
    ) -> None:
        """
        Add multiple predictions to the model.

        Parameters
        ----------
        dataset : valor.Dataset
            The dataset that is being operated over.
        predictions : List[valor.Prediction]
            The predictions to create.
        timeout : float, optional
            The number of seconds the client should wait until raising a timeout.
        """
        Client(self.conn).create_predictions(
            dataset=dataset,
            model=self,
            predictions=predictions,
            timeout=timeout,
        )

    def get_prediction(
        self, dataset: Union[Dataset, str], datum: Union[Datum, str]
    ) -> Union[Prediction, None]:
        """
        Get a particular prediction.

        Parameters
        ----------
        dataset : Union[Dataset, str]
            The dataset the datum belongs to.
        datum : Union[Datum, str]
            The desired datum.

        Returns
        ----------
        Union[Prediction, None]
            The matching prediction or 'None' if it doesn't exist.
        """
        return Client(self.conn).get_prediction(
            dataset=dataset, model=self, datum=datum
        )

    def finalize_inferences(self, dataset: Union[Dataset, str]) -> None:
        """
        Finalizes the model over a dataset such that new predictions cannot be added to it.
        """
        return Client(self.conn).finalize_inferences(
            dataset=dataset, model=self
        )

    def _create_label_map(
        self,
        label_map: Optional[Dict[Label, Label]],
    ) -> Union[List[List[List[str]]], None]:
        """Convert a dictionary of label maps to a serializable list format."""
        if not label_map:
            return None

        if not isinstance(label_map, dict) or not all(
            [
                isinstance(key, Label) and isinstance(value, Label)
                for key, value in label_map.items()
            ]
        ):
            raise TypeError(
                "label_map should be a dictionary with valid Labels for both the key and value."
            )

        return_value = []
        for key, value in label_map.items():
            if not all(
                [
                    (isinstance(v.key, str) and isinstance(v.value, str))
                    for v in [key, value]
                ]
            ):
                raise TypeError
            return_value.append(
                [
                    [key.key, key.value],
                    [value.key, value.value],
                ]
            )
        return return_value

    def evaluate_classification(
        self,
        datasets: Union[Dataset, List[Dataset]],
        filters: Optional[Filter] = None,
        label_map: Optional[Dict[Label, Label]] = None,
        pr_curve_max_examples: int = 1,
        metrics_to_return: Optional[List[MetricType]] = None,
        *_,
        allow_retries: bool = False,
        timeout: Optional[float] = None,
    ) -> Evaluation:
        """
        Start a classification evaluation job.

        Parameters
        ----------
        datasets : Union[Dataset, List[Dataset]]
            The dataset or list of datasets to evaluate against.
        filters : Filter, optional
            Optional set of constraints to filter evaluation by.
        label_map : Dict[Label, Label], optional
            Optional mapping of individual labels to a grouper label. Useful when you need to evaluate performance using labels that differ across datasets and models.
        metrics_to_return : List[MetricType], optional
            The list of metrics to compute, store, and return to the user.
        allow_retries : bool, default = False
            Option to retry previously failed evaluations.
        timeout : float, optional
            The number of seconds the client should wait until raising a timeout.

        Returns
        -------
        Evaluation
            A job object that can be used to track the status of the job and get the metrics of it upon completion.
        """
        if metrics_to_return and not set(metrics_to_return).issubset(
            MetricType.classification()
        ):
            raise ValueError(
                f"The following metrics are not supported for classification: '{set(metrics_to_return) - MetricType.classification()}'"
            )

        # format request
        datasets = datasets if isinstance(datasets, list) else [datasets]
        filters = filters if filters else Filter()
        request = EvaluationRequest(
            dataset_names=[dataset.name for dataset in datasets],  # type: ignore - issue #604
            model_names=[self.name],  # type: ignore - issue #604
            filters=filters,
            parameters=EvaluationParameters(
                task_type=TaskType.CLASSIFICATION,
                label_map=self._create_label_map(label_map=label_map),
                pr_curve_max_examples=pr_curve_max_examples,
                metrics_to_return=metrics_to_return,
            ),
        )

        # create evaluation
        evaluation = Client(self.conn).evaluate(
            request, allow_retries=allow_retries, timeout=timeout
        )
        if len(evaluation) != 1:
            raise RuntimeError
        return evaluation[0]

    def evaluate_detection(
        self,
        datasets: Union[Dataset, List[Dataset]],
        filters: Optional[Filter] = None,
        convert_annotations_to_type: Optional[AnnotationType] = None,
        iou_thresholds_to_compute: Optional[List[float]] = None,
        iou_thresholds_to_return: Optional[List[float]] = None,
        label_map: Optional[Dict[Label, Label]] = None,
        recall_score_threshold: float = 0,
        metrics_to_return: Optional[List[MetricType]] = None,
        pr_curve_iou_threshold: float = 0.5,
        pr_curve_max_examples: int = 1,
        *_,
        allow_retries: bool = False,
        timeout: Optional[float] = None,
    ) -> Evaluation:
        """
        Start an object-detection evaluation job.

        Parameters
        ----------
        datasets : Union[Dataset, List[Dataset]]
            The dataset or list of datasets to evaluate against.
        filters : Filter, optional
            Optional set of constraints to filter evaluation by.
        convert_annotations_to_type : enums.AnnotationType, optional
            Forces the object detection evaluation to compute over this type.
        iou_thresholds_to_compute : List[float], optional
            Thresholds to compute mAP against.
        iou_thresholds_to_return : List[float], optional
            Thresholds to return AP for. Must be subset of `iou_thresholds_to_compute`.
        label_map : Dict[Label, Label], optional
            Optional mapping of individual labels to a grouper label. Useful when you need to evaluate performance using labels that differ across datasets and models.
        recall_score_threshold : float, default=0
            The confidence score threshold for use when determining whether to count a prediction as a true positive or not while calculating Average Recall.
        metrics_to_return : List[MetricType], optional
            The list of metrics to compute, store, and return to the user.
        pr_curve_iou_threshold : float, optional
            The IOU threshold to use when calculating precision-recall curves. Defaults to 0.5.
        pr_curve_max_examples : int, optional
            The maximum number of datum examples to store when calculating PR curves.
        allow_retries : bool, default = False
            Option to retry previously failed evaluations.
        timeout : float, optional
            The number of seconds the client should wait until raising a timeout.

        Returns
        -------
        Evaluation
            A job object that can be used to track the status of the job and get the metrics of it upon completion.
        """
        if metrics_to_return and not set(metrics_to_return).issubset(
            MetricType.object_detection()
        ):
            raise ValueError(
                f"The following metrics are not supported for object detection: '{set(metrics_to_return) - MetricType.object_detection()}'"
            )

        if iou_thresholds_to_compute is None:
            iou_thresholds_to_compute = [
                round(0.5 + 0.05 * i, 2) for i in range(10)
            ]
        if iou_thresholds_to_return is None:
            iou_thresholds_to_return = [0.5, 0.75]

        # format request
        parameters = EvaluationParameters(
            task_type=TaskType.OBJECT_DETECTION,
            convert_annotations_to_type=convert_annotations_to_type,
            iou_thresholds_to_compute=iou_thresholds_to_compute,
            iou_thresholds_to_return=iou_thresholds_to_return,
            label_map=self._create_label_map(label_map=label_map),
            recall_score_threshold=recall_score_threshold,
            metrics_to_return=metrics_to_return,
            pr_curve_iou_threshold=pr_curve_iou_threshold,
            pr_curve_max_examples=pr_curve_max_examples,
        )
        datasets = datasets if isinstance(datasets, list) else [datasets]
        filters = filters if filters else Filter()
        request = EvaluationRequest(
            dataset_names=[dataset.name for dataset in datasets],  # type: ignore - issue #604
            model_names=[self.name],  # type: ignore - issue #604
            filters=filters,
            parameters=parameters,
        )

        # create evaluation
        evaluation = Client(self.conn).evaluate(
            request, allow_retries=allow_retries, timeout=timeout
        )
        if len(evaluation) != 1:
            raise RuntimeError
        return evaluation[0]

    def evaluate_segmentation(
        self,
        datasets: Union[Dataset, List[Dataset]],
        filters: Optional[Filter] = None,
        label_map: Optional[Dict[Label, Label]] = None,
        metrics_to_return: Optional[List[MetricType]] = None,
        *_,
        allow_retries: bool = False,
        timeout: Optional[float] = None,
    ) -> Evaluation:
        """
        Start a semantic-segmentation evaluation job.

        Parameters
        ----------
        datasets : Union[Dataset, List[Dataset]]
            The dataset or list of datasets to evaluate against.
        filters : Filter, optional
            Optional set of constraints to filter evaluation by.
        label_map : Dict[Label, Label], optional
            Optional mapping of individual labels to a grouper label. Useful when you need to evaluate performance using labels that differ across datasets and models.
        metrics_to_return : List[MetricType], optional
            The list of metrics to compute, store, and return to the user.
        allow_retries : bool, default = False
            Option to retry previously failed evaluations.
        timeout : float, optional
            The number of seconds the client should wait until raising a timeout.

        Returns
        -------
        Evaluation
            A job object that can be used to track the status of the job and get the metrics of it upon completion
        """
        if metrics_to_return and not set(metrics_to_return).issubset(
            MetricType.semantic_segmentation()
        ):
            raise ValueError(
                f"The following metrics are not supported for semantic segmentation: '{set(metrics_to_return) - MetricType.semantic_segmentation()}'"
            )

        # format request
        datasets = datasets if isinstance(datasets, list) else [datasets]
        filters = filters if filters else Filter()
        request = EvaluationRequest(
            dataset_names=[dataset.name for dataset in datasets],  # type: ignore - issue #604
            model_names=[self.name],  # type: ignore - issue #604
            filters=filters,
            parameters=EvaluationParameters(
                task_type=TaskType.SEMANTIC_SEGMENTATION,
                label_map=self._create_label_map(label_map=label_map),
                metrics_to_return=metrics_to_return,
            ),
        )

        # create evaluation
        evaluation = Client(self.conn).evaluate(
            request, allow_retries=allow_retries, timeout=timeout
        )
        if len(evaluation) != 1:
            raise RuntimeError
        return evaluation[0]

    def evaluate_text_generation(
        self,
        datasets: Union[Dataset, List[Dataset]],
        metrics_to_return: List[MetricType],
        filters: Optional[Filter] = None,
        llm_api_params: Optional[Dict[str, Union[str, dict]]] = None,
        metric_params: Optional[Dict[MetricType, Dict[str, Any]]] = None,
    ) -> Evaluation:
        """
        Start a classification evaluation job.

        Parameters
        ----------
        datasets : Union[Dataset, List[Dataset]]
            The dataset or list of datasets to evaluate against.
        metrics_to_return : List[MetricType]
            The list of metrics to compute, store, and return to the user. This is not optional for text generation evaluations.
        filters : Filter, optional
            Optional set of constraints to filter evaluation by.
        llm_api_params : Dict[str, Union[str,dict]], optional
            A dictionary of parameters for the LLM API.
        metric_params : Dict[MetricType, Dict[str,Any]], optional
            A dictionary of parameters for the metrics used in the evaluation. The keys should be the metrics and the values should be dictionaries of parameters for those metrics.

        Returns
        -------
        Evaluation
            A job object that can be used to track the status of the job and get the metrics of it upon completion.
        """
        if not set(metrics_to_return).issubset(MetricType.text_generation()):
            raise ValueError(
                f"The following metrics are not supported for text generation: '{set(metrics_to_return) - MetricType.text_generation()}'"
            )

        # If no api_key is provided, check the environment variables for an api key.
        if llm_api_params is not None:
            if "api_key" not in llm_api_params:
                if "client" not in llm_api_params:
                    raise ValueError(
                        "The client must be specified in the llm_api_params."
                    )

                if llm_api_params["client"] == "openai":
                    api_key = os.getenv("OPENAI_API_KEY", None)
                elif llm_api_params["client"] == "mistral":
                    api_key = os.getenv("MISTRAL_API_KEY", None)
                elif llm_api_params["client"] == "mock":
                    api_key = ""
                else:
                    raise ValueError(
                        "The client specified in llm_api_params is not supported."
                    )

                if api_key is not None:
                    llm_api_params["api_key"] = api_key

        bleu_weights = None
        rouge_types = None
        rouge_use_stemmer = None
        if metric_params is not None:
            if not all(
                metric in metrics_to_return for metric in metric_params.keys()
            ):
                raise ValueError(
                    "All metrics in metric_params must be in metrics_to_return."
                )

            if MetricType.BLEU in metric_params:
                bleu_weights = metric_params[MetricType.BLEU].get("weights")

            if MetricType.ROUGE in metric_params:
                rouge_types = metric_params[MetricType.ROUGE].get(
                    "rouge_types"
                )
                rouge_use_stemmer = metric_params[MetricType.ROUGE].get(
                    "use_stemmer"
                )

        # format request
        datasets = datasets if isinstance(datasets, list) else [datasets]
        filters = filters if filters else Filter()
        request = EvaluationRequest(
            dataset_names=[dataset.name for dataset in datasets],  # type: ignore - issue #604
            model_names=[self.name],  # type: ignore - issue #604
            filters=filters,
            parameters=EvaluationParameters(
                task_type=TaskType.TEXT_GENERATION,
                metrics_to_return=metrics_to_return,
                llm_api_params=llm_api_params,
                bleu_weights=bleu_weights,
                rouge_types=rouge_types,
                rouge_use_stemmer=rouge_use_stemmer,
            ),
        )

        # create evaluation
        evaluation = Client(self.conn).evaluate(request)
        if len(evaluation) != 1:
            raise RuntimeError
        return evaluation[0]

    def delete(self, timeout: int = 0):
        """
        Delete the `Model` object from the back end.

        Parameters
        ----------
        timeout : int, default=0
            Sets a timeout in seconds.
        """
        Client(self.conn).delete_model(self.name, timeout)  # type: ignore

    def get_labels(
        self,
    ) -> List[Label]:
        """
        Get all labels associated with a given model.

        Returns
        ----------
        List[Label]
            A list of `Labels` associated with the model.
        """
        return Client(self.conn).get_labels_from_model(self)

    def get_evaluations(
        self,
        metrics_to_sort_by: Optional[
            Dict[str, Union[Dict[str, str], str]]
        ] = None,
        *_,
        timeout: Optional[float] = None,
    ) -> List[Evaluation]:
        """
        Get all evaluations associated with a given model.

        Parameters
        ----------
        metrics_to_sort_by : dict[str, str | dict[str, str]], optional
            An optional dict of metric types to sort the evaluations by.
        timeout : float, optional
            The number of seconds the client should wait until raising a timeout.

        Returns
        ----------
        List[Evaluation]
            A list of `Evaluations` associated with the model.
        """
        return Client(self.conn).get_evaluations(
            models=[self],
            metrics_to_sort_by=metrics_to_sort_by,
            timeout=timeout,
        )

Functions

valor.Model.__init__(*, name, metadata=None, connection=None)

Creates a local instance of a model.

Use 'Model.create' classmethod to create a model with persistence.

Parameters:

Name Type Description Default
name String

The name of the model.

required
metadata Dictionary

A dictionary of metadata that describes the model.

None
connection ClientConnection

An initialized client connection.

None
Source code in valor/coretypes.py
def __init__(
    self,
    *,
    name: str,
    metadata: Optional[dict] = None,
    connection: Optional[ClientConnection] = None,
):
    """
    Creates a local instance of a model.

    Use 'Model.create' classmethod to create a model with persistence.

    Parameters
    ----------
    name : String
        The name of the model.
    metadata : Dictionary
        A dictionary of metadata that describes the model.
    connection : ClientConnection, optional
        An initialized client connection.
    """
    self.conn = connection
    super().__init__(name=name, metadata=metadata if metadata else dict())

valor.Model.add_prediction(dataset, prediction)

Add a prediction to the model.

Parameters:

Name Type Description Default
dataset Dataset

The dataset that is being operated over.

required
prediction Prediction

The prediction to create.

required
Source code in valor/coretypes.py
def add_prediction(
    self,
    dataset: Dataset,
    prediction: Prediction,
) -> None:
    """
    Add a prediction to the model.

    Parameters
    ----------
    dataset : valor.Dataset
        The dataset that is being operated over.
    prediction : valor.Prediction
        The prediction to create.
    """
    Client(self.conn).create_predictions(
        dataset=dataset,
        model=self,
        predictions=[prediction],
    )

valor.Model.add_predictions(dataset, predictions, timeout=10.0)

Add multiple predictions to the model.

Parameters:

Name Type Description Default
dataset Dataset

The dataset that is being operated over.

required
predictions List[Prediction]

The predictions to create.

required
timeout float

The number of seconds the client should wait until raising a timeout.

10.0
Source code in valor/coretypes.py
def add_predictions(
    self,
    dataset: Dataset,
    predictions: List[Prediction],
    timeout: Optional[float] = 10.0,
) -> None:
    """
    Add multiple predictions to the model.

    Parameters
    ----------
    dataset : valor.Dataset
        The dataset that is being operated over.
    predictions : List[valor.Prediction]
        The predictions to create.
    timeout : float, optional
        The number of seconds the client should wait until raising a timeout.
    """
    Client(self.conn).create_predictions(
        dataset=dataset,
        model=self,
        predictions=predictions,
        timeout=timeout,
    )

valor.Model.create(name, metadata=None, connection=None, **_) classmethod

Creates a model that persists in the back end.

Parameters:

Name Type Description Default
name str

The name of the model.

required
metadata dict

A dictionary of metadata that describes the model.

None
connection ClientConnection

An initialized client connection.

None
Source code in valor/coretypes.py
@classmethod
def create(
    cls,
    name: str,
    metadata: Optional[Dict[str, Any]] = None,
    connection: Optional[ClientConnection] = None,
    **_,
) -> Model:
    """
    Creates a model that persists in the back end.

    Parameters
    ----------
    name : str
        The name of the model.
    metadata : dict, optional
        A dictionary of metadata that describes the model.
    connection : ClientConnection, optional
        An initialized client connection.
    """
    model = cls(name=name, metadata=metadata, connection=connection)
    Client(connection).create_model(model)
    return model

valor.Model.delete(timeout=0)

Delete the Model object from the back end.

Parameters:

Name Type Description Default
timeout int

Sets a timeout in seconds.

0
Source code in valor/coretypes.py
def delete(self, timeout: int = 0):
    """
    Delete the `Model` object from the back end.

    Parameters
    ----------
    timeout : int, default=0
        Sets a timeout in seconds.
    """
    Client(self.conn).delete_model(self.name, timeout)  # type: ignore

valor.Model.evaluate_classification(datasets, filters=None, label_map=None, pr_curve_max_examples=1, metrics_to_return=None, *_, allow_retries=False, timeout=None)

Start a classification evaluation job.

Parameters:

Name Type Description Default
datasets Union[Dataset, List[Dataset]]

The dataset or list of datasets to evaluate against.

required
filters Filter

Optional set of constraints to filter evaluation by.

None
label_map Dict[Label, Label]

Optional mapping of individual labels to a grouper label. Useful when you need to evaluate performance using labels that differ across datasets and models.

None
metrics_to_return List[MetricType]

The list of metrics to compute, store, and return to the user.

None
allow_retries bool

Option to retry previously failed evaluations.

= False
timeout float

The number of seconds the client should wait until raising a timeout.

None

Returns:

Type Description
Evaluation

A job object that can be used to track the status of the job and get the metrics of it upon completion.

Source code in valor/coretypes.py
def evaluate_classification(
    self,
    datasets: Union[Dataset, List[Dataset]],
    filters: Optional[Filter] = None,
    label_map: Optional[Dict[Label, Label]] = None,
    pr_curve_max_examples: int = 1,
    metrics_to_return: Optional[List[MetricType]] = None,
    *_,
    allow_retries: bool = False,
    timeout: Optional[float] = None,
) -> Evaluation:
    """
    Start a classification evaluation job.

    Parameters
    ----------
    datasets : Union[Dataset, List[Dataset]]
        The dataset or list of datasets to evaluate against.
    filters : Filter, optional
        Optional set of constraints to filter evaluation by.
    label_map : Dict[Label, Label], optional
        Optional mapping of individual labels to a grouper label. Useful when you need to evaluate performance using labels that differ across datasets and models.
    metrics_to_return : List[MetricType], optional
        The list of metrics to compute, store, and return to the user.
    allow_retries : bool, default = False
        Option to retry previously failed evaluations.
    timeout : float, optional
        The number of seconds the client should wait until raising a timeout.

    Returns
    -------
    Evaluation
        A job object that can be used to track the status of the job and get the metrics of it upon completion.
    """
    if metrics_to_return and not set(metrics_to_return).issubset(
        MetricType.classification()
    ):
        raise ValueError(
            f"The following metrics are not supported for classification: '{set(metrics_to_return) - MetricType.classification()}'"
        )

    # format request
    datasets = datasets if isinstance(datasets, list) else [datasets]
    filters = filters if filters else Filter()
    request = EvaluationRequest(
        dataset_names=[dataset.name for dataset in datasets],  # type: ignore - issue #604
        model_names=[self.name],  # type: ignore - issue #604
        filters=filters,
        parameters=EvaluationParameters(
            task_type=TaskType.CLASSIFICATION,
            label_map=self._create_label_map(label_map=label_map),
            pr_curve_max_examples=pr_curve_max_examples,
            metrics_to_return=metrics_to_return,
        ),
    )

    # create evaluation
    evaluation = Client(self.conn).evaluate(
        request, allow_retries=allow_retries, timeout=timeout
    )
    if len(evaluation) != 1:
        raise RuntimeError
    return evaluation[0]

valor.Model.evaluate_detection(datasets, filters=None, convert_annotations_to_type=None, iou_thresholds_to_compute=None, iou_thresholds_to_return=None, label_map=None, recall_score_threshold=0, metrics_to_return=None, pr_curve_iou_threshold=0.5, pr_curve_max_examples=1, *_, allow_retries=False, timeout=None)

Start an object-detection evaluation job.

Parameters:

Name Type Description Default
datasets Union[Dataset, List[Dataset]]

The dataset or list of datasets to evaluate against.

required
filters Filter

Optional set of constraints to filter evaluation by.

None
convert_annotations_to_type AnnotationType

Forces the object detection evaluation to compute over this type.

None
iou_thresholds_to_compute List[float]

Thresholds to compute mAP against.

None
iou_thresholds_to_return List[float]

Thresholds to return AP for. Must be subset of iou_thresholds_to_compute.

None
label_map Dict[Label, Label]

Optional mapping of individual labels to a grouper label. Useful when you need to evaluate performance using labels that differ across datasets and models.

None
recall_score_threshold float

The confidence score threshold for use when determining whether to count a prediction as a true positive or not while calculating Average Recall.

0
metrics_to_return List[MetricType]

The list of metrics to compute, store, and return to the user.

None
pr_curve_iou_threshold float

The IOU threshold to use when calculating precision-recall curves. Defaults to 0.5.

0.5
pr_curve_max_examples int

The maximum number of datum examples to store when calculating PR curves.

1
allow_retries bool

Option to retry previously failed evaluations.

= False
timeout float

The number of seconds the client should wait until raising a timeout.

None

Returns:

Type Description
Evaluation

A job object that can be used to track the status of the job and get the metrics of it upon completion.

Source code in valor/coretypes.py
def evaluate_detection(
    self,
    datasets: Union[Dataset, List[Dataset]],
    filters: Optional[Filter] = None,
    convert_annotations_to_type: Optional[AnnotationType] = None,
    iou_thresholds_to_compute: Optional[List[float]] = None,
    iou_thresholds_to_return: Optional[List[float]] = None,
    label_map: Optional[Dict[Label, Label]] = None,
    recall_score_threshold: float = 0,
    metrics_to_return: Optional[List[MetricType]] = None,
    pr_curve_iou_threshold: float = 0.5,
    pr_curve_max_examples: int = 1,
    *_,
    allow_retries: bool = False,
    timeout: Optional[float] = None,
) -> Evaluation:
    """
    Start an object-detection evaluation job.

    Parameters
    ----------
    datasets : Union[Dataset, List[Dataset]]
        The dataset or list of datasets to evaluate against.
    filters : Filter, optional
        Optional set of constraints to filter evaluation by.
    convert_annotations_to_type : enums.AnnotationType, optional
        Forces the object detection evaluation to compute over this type.
    iou_thresholds_to_compute : List[float], optional
        Thresholds to compute mAP against.
    iou_thresholds_to_return : List[float], optional
        Thresholds to return AP for. Must be subset of `iou_thresholds_to_compute`.
    label_map : Dict[Label, Label], optional
        Optional mapping of individual labels to a grouper label. Useful when you need to evaluate performance using labels that differ across datasets and models.
    recall_score_threshold : float, default=0
        The confidence score threshold for use when determining whether to count a prediction as a true positive or not while calculating Average Recall.
    metrics_to_return : List[MetricType], optional
        The list of metrics to compute, store, and return to the user.
    pr_curve_iou_threshold : float, optional
        The IOU threshold to use when calculating precision-recall curves. Defaults to 0.5.
    pr_curve_max_examples : int, optional
        The maximum number of datum examples to store when calculating PR curves.
    allow_retries : bool, default = False
        Option to retry previously failed evaluations.
    timeout : float, optional
        The number of seconds the client should wait until raising a timeout.

    Returns
    -------
    Evaluation
        A job object that can be used to track the status of the job and get the metrics of it upon completion.
    """
    if metrics_to_return and not set(metrics_to_return).issubset(
        MetricType.object_detection()
    ):
        raise ValueError(
            f"The following metrics are not supported for object detection: '{set(metrics_to_return) - MetricType.object_detection()}'"
        )

    if iou_thresholds_to_compute is None:
        iou_thresholds_to_compute = [
            round(0.5 + 0.05 * i, 2) for i in range(10)
        ]
    if iou_thresholds_to_return is None:
        iou_thresholds_to_return = [0.5, 0.75]

    # format request
    parameters = EvaluationParameters(
        task_type=TaskType.OBJECT_DETECTION,
        convert_annotations_to_type=convert_annotations_to_type,
        iou_thresholds_to_compute=iou_thresholds_to_compute,
        iou_thresholds_to_return=iou_thresholds_to_return,
        label_map=self._create_label_map(label_map=label_map),
        recall_score_threshold=recall_score_threshold,
        metrics_to_return=metrics_to_return,
        pr_curve_iou_threshold=pr_curve_iou_threshold,
        pr_curve_max_examples=pr_curve_max_examples,
    )
    datasets = datasets if isinstance(datasets, list) else [datasets]
    filters = filters if filters else Filter()
    request = EvaluationRequest(
        dataset_names=[dataset.name for dataset in datasets],  # type: ignore - issue #604
        model_names=[self.name],  # type: ignore - issue #604
        filters=filters,
        parameters=parameters,
    )

    # create evaluation
    evaluation = Client(self.conn).evaluate(
        request, allow_retries=allow_retries, timeout=timeout
    )
    if len(evaluation) != 1:
        raise RuntimeError
    return evaluation[0]

valor.Model.evaluate_segmentation(datasets, filters=None, label_map=None, metrics_to_return=None, *_, allow_retries=False, timeout=None)

Start a semantic-segmentation evaluation job.

Parameters:

Name Type Description Default
datasets Union[Dataset, List[Dataset]]

The dataset or list of datasets to evaluate against.

required
filters Filter

Optional set of constraints to filter evaluation by.

None
label_map Dict[Label, Label]

Optional mapping of individual labels to a grouper label. Useful when you need to evaluate performance using labels that differ across datasets and models.

None
metrics_to_return List[MetricType]

The list of metrics to compute, store, and return to the user.

None
allow_retries bool

Option to retry previously failed evaluations.

= False
timeout float

The number of seconds the client should wait until raising a timeout.

None

Returns:

Type Description
Evaluation

A job object that can be used to track the status of the job and get the metrics of it upon completion

Source code in valor/coretypes.py
def evaluate_segmentation(
    self,
    datasets: Union[Dataset, List[Dataset]],
    filters: Optional[Filter] = None,
    label_map: Optional[Dict[Label, Label]] = None,
    metrics_to_return: Optional[List[MetricType]] = None,
    *_,
    allow_retries: bool = False,
    timeout: Optional[float] = None,
) -> Evaluation:
    """
    Start a semantic-segmentation evaluation job.

    Parameters
    ----------
    datasets : Union[Dataset, List[Dataset]]
        The dataset or list of datasets to evaluate against.
    filters : Filter, optional
        Optional set of constraints to filter evaluation by.
    label_map : Dict[Label, Label], optional
        Optional mapping of individual labels to a grouper label. Useful when you need to evaluate performance using labels that differ across datasets and models.
    metrics_to_return : List[MetricType], optional
        The list of metrics to compute, store, and return to the user.
    allow_retries : bool, default = False
        Option to retry previously failed evaluations.
    timeout : float, optional
        The number of seconds the client should wait until raising a timeout.

    Returns
    -------
    Evaluation
        A job object that can be used to track the status of the job and get the metrics of it upon completion
    """
    if metrics_to_return and not set(metrics_to_return).issubset(
        MetricType.semantic_segmentation()
    ):
        raise ValueError(
            f"The following metrics are not supported for semantic segmentation: '{set(metrics_to_return) - MetricType.semantic_segmentation()}'"
        )

    # format request
    datasets = datasets if isinstance(datasets, list) else [datasets]
    filters = filters if filters else Filter()
    request = EvaluationRequest(
        dataset_names=[dataset.name for dataset in datasets],  # type: ignore - issue #604
        model_names=[self.name],  # type: ignore - issue #604
        filters=filters,
        parameters=EvaluationParameters(
            task_type=TaskType.SEMANTIC_SEGMENTATION,
            label_map=self._create_label_map(label_map=label_map),
            metrics_to_return=metrics_to_return,
        ),
    )

    # create evaluation
    evaluation = Client(self.conn).evaluate(
        request, allow_retries=allow_retries, timeout=timeout
    )
    if len(evaluation) != 1:
        raise RuntimeError
    return evaluation[0]

valor.Model.evaluate_text_generation(datasets, metrics_to_return, filters=None, llm_api_params=None, metric_params=None)

Start a classification evaluation job.

Parameters:

Name Type Description Default
datasets Union[Dataset, List[Dataset]]

The dataset or list of datasets to evaluate against.

required
metrics_to_return List[MetricType]

The list of metrics to compute, store, and return to the user. This is not optional for text generation evaluations.

required
filters Filter

Optional set of constraints to filter evaluation by.

None
llm_api_params Dict[str, Union[str, dict]]

A dictionary of parameters for the LLM API.

None
metric_params Dict[MetricType, Dict[str, Any]]

A dictionary of parameters for the metrics used in the evaluation. The keys should be the metrics and the values should be dictionaries of parameters for those metrics.

None

Returns:

Type Description
Evaluation

A job object that can be used to track the status of the job and get the metrics of it upon completion.

Source code in valor/coretypes.py
def evaluate_text_generation(
    self,
    datasets: Union[Dataset, List[Dataset]],
    metrics_to_return: List[MetricType],
    filters: Optional[Filter] = None,
    llm_api_params: Optional[Dict[str, Union[str, dict]]] = None,
    metric_params: Optional[Dict[MetricType, Dict[str, Any]]] = None,
) -> Evaluation:
    """
    Start a classification evaluation job.

    Parameters
    ----------
    datasets : Union[Dataset, List[Dataset]]
        The dataset or list of datasets to evaluate against.
    metrics_to_return : List[MetricType]
        The list of metrics to compute, store, and return to the user. This is not optional for text generation evaluations.
    filters : Filter, optional
        Optional set of constraints to filter evaluation by.
    llm_api_params : Dict[str, Union[str,dict]], optional
        A dictionary of parameters for the LLM API.
    metric_params : Dict[MetricType, Dict[str,Any]], optional
        A dictionary of parameters for the metrics used in the evaluation. The keys should be the metrics and the values should be dictionaries of parameters for those metrics.

    Returns
    -------
    Evaluation
        A job object that can be used to track the status of the job and get the metrics of it upon completion.
    """
    if not set(metrics_to_return).issubset(MetricType.text_generation()):
        raise ValueError(
            f"The following metrics are not supported for text generation: '{set(metrics_to_return) - MetricType.text_generation()}'"
        )

    # If no api_key is provided, check the environment variables for an api key.
    if llm_api_params is not None:
        if "api_key" not in llm_api_params:
            if "client" not in llm_api_params:
                raise ValueError(
                    "The client must be specified in the llm_api_params."
                )

            if llm_api_params["client"] == "openai":
                api_key = os.getenv("OPENAI_API_KEY", None)
            elif llm_api_params["client"] == "mistral":
                api_key = os.getenv("MISTRAL_API_KEY", None)
            elif llm_api_params["client"] == "mock":
                api_key = ""
            else:
                raise ValueError(
                    "The client specified in llm_api_params is not supported."
                )

            if api_key is not None:
                llm_api_params["api_key"] = api_key

    bleu_weights = None
    rouge_types = None
    rouge_use_stemmer = None
    if metric_params is not None:
        if not all(
            metric in metrics_to_return for metric in metric_params.keys()
        ):
            raise ValueError(
                "All metrics in metric_params must be in metrics_to_return."
            )

        if MetricType.BLEU in metric_params:
            bleu_weights = metric_params[MetricType.BLEU].get("weights")

        if MetricType.ROUGE in metric_params:
            rouge_types = metric_params[MetricType.ROUGE].get(
                "rouge_types"
            )
            rouge_use_stemmer = metric_params[MetricType.ROUGE].get(
                "use_stemmer"
            )

    # format request
    datasets = datasets if isinstance(datasets, list) else [datasets]
    filters = filters if filters else Filter()
    request = EvaluationRequest(
        dataset_names=[dataset.name for dataset in datasets],  # type: ignore - issue #604
        model_names=[self.name],  # type: ignore - issue #604
        filters=filters,
        parameters=EvaluationParameters(
            task_type=TaskType.TEXT_GENERATION,
            metrics_to_return=metrics_to_return,
            llm_api_params=llm_api_params,
            bleu_weights=bleu_weights,
            rouge_types=rouge_types,
            rouge_use_stemmer=rouge_use_stemmer,
        ),
    )

    # create evaluation
    evaluation = Client(self.conn).evaluate(request)
    if len(evaluation) != 1:
        raise RuntimeError
    return evaluation[0]

valor.Model.finalize_inferences(dataset)

Finalizes the model over a dataset such that new predictions cannot be added to it.

Source code in valor/coretypes.py
def finalize_inferences(self, dataset: Union[Dataset, str]) -> None:
    """
    Finalizes the model over a dataset such that new predictions cannot be added to it.
    """
    return Client(self.conn).finalize_inferences(
        dataset=dataset, model=self
    )

valor.Model.get(name, connection=None) classmethod

Retrieves a model from the back end database.

Parameters:

Name Type Description Default
name str

The name of the model.

required
connection ClientConnnetion

An optional Valor client object for interacting with the API.

None

Returns:

Type Description
Union[Model, None]

The model or 'None' if it doesn't exist.

Source code in valor/coretypes.py
@classmethod
def get(
    cls,
    name: str,
    connection: Optional[ClientConnection] = None,
) -> Union[Model, None]:
    """
    Retrieves a model from the back end database.

    Parameters
    ----------
    name : str
        The name of the model.
    connection : ClientConnnetion, optional
        An optional Valor client object for interacting with the API.

    Returns
    -------
    Union[valor.Model, None]
        The model or 'None' if it doesn't exist.
    """
    return Client(connection).get_model(name)

valor.Model.get_evaluations(metrics_to_sort_by=None, *_, timeout=None)

Get all evaluations associated with a given model.

Parameters:

Name Type Description Default
metrics_to_sort_by dict[str, str | dict[str, str]]

An optional dict of metric types to sort the evaluations by.

None
timeout float

The number of seconds the client should wait until raising a timeout.

None

Returns:

Type Description
List[Evaluation]

A list of Evaluations associated with the model.

Source code in valor/coretypes.py
def get_evaluations(
    self,
    metrics_to_sort_by: Optional[
        Dict[str, Union[Dict[str, str], str]]
    ] = None,
    *_,
    timeout: Optional[float] = None,
) -> List[Evaluation]:
    """
    Get all evaluations associated with a given model.

    Parameters
    ----------
    metrics_to_sort_by : dict[str, str | dict[str, str]], optional
        An optional dict of metric types to sort the evaluations by.
    timeout : float, optional
        The number of seconds the client should wait until raising a timeout.

    Returns
    ----------
    List[Evaluation]
        A list of `Evaluations` associated with the model.
    """
    return Client(self.conn).get_evaluations(
        models=[self],
        metrics_to_sort_by=metrics_to_sort_by,
        timeout=timeout,
    )

valor.Model.get_labels()

Get all labels associated with a given model.

Returns:

Type Description
List[Label]

A list of Labels associated with the model.

Source code in valor/coretypes.py
def get_labels(
    self,
) -> List[Label]:
    """
    Get all labels associated with a given model.

    Returns
    ----------
    List[Label]
        A list of `Labels` associated with the model.
    """
    return Client(self.conn).get_labels_from_model(self)

valor.Model.get_prediction(dataset, datum)

Get a particular prediction.

Parameters:

Name Type Description Default
dataset Union[Dataset, str]

The dataset the datum belongs to.

required
datum Union[Datum, str]

The desired datum.

required

Returns:

Type Description
Union[Prediction, None]

The matching prediction or 'None' if it doesn't exist.

Source code in valor/coretypes.py
def get_prediction(
    self, dataset: Union[Dataset, str], datum: Union[Datum, str]
) -> Union[Prediction, None]:
    """
    Get a particular prediction.

    Parameters
    ----------
    dataset : Union[Dataset, str]
        The dataset the datum belongs to.
    datum : Union[Datum, str]
        The desired datum.

    Returns
    ----------
    Union[Prediction, None]
        The matching prediction or 'None' if it doesn't exist.
    """
    return Client(self.conn).get_prediction(
        dataset=dataset, model=self, datum=datum
    )