Reference

class Workspace(create_key, workspace: Workspace)

classmethod create_from_config(config: Config) → Workspace

classmethod create_from_json_model(model_json: Dict[str, Any]) → Workspace

classmethod create_from_model(data: bytearray) → Workspace

create_parser(format_type: Literal[FormatType.VwText]) → TextParser

create_parser(format_type: Union[Literal[FormatType.DsJson], Literal[FormatType.Json]]) → JsonParser

property entry_reduction_types: ReductionTypesDescription

learn(features: Union[SparseFeatures, CbAdfFeatures], label: Union[SimpleLabel, CbLabel]) → None

predict(features: Union[SparseFeatures, CbAdfFeatures]) → Union[ScalarPred, ActionScoresPred, ActionProbsPred]

predict_then_learn(features: Union[SparseFeatures, CbAdfFeatures], label: Union[SimpleLabel, CbLabel]) → Union[ScalarPred, ActionScoresPred, ActionProbsPred]

serialize() → bytearray

serialize_to_json() → Dict[str, Any]

class JsonParser

parse(input: Union[Dict[str, Any], str]) → Tuple[Union[SparseFeatures, CbAdfFeatures], Optional[Union[SimpleLabel, CbLabel]]]

class TextParser

parse(input: str) → Tuple[Union[SparseFeatures, CbAdfFeatures], Optional[Union[SimpleLabel, CbLabel]]]

class ReductionTypesDescription(self, input_label_type: LabelType, output_prediction_type: PredictionType, input_features_type: FeaturesType, output_label_type: Optional[LabelType] = None, input_prediction_type: Optional[PredictionType] = None, output_features_type: Optional[FeaturesType] = None)

input_features_type

The features type expected as input for this reduction

Return type:

FeaturesType

input_label_type

The label type expected as input for this reduction

Return type:

LabelType

input_prediction_type

The prediction type expected as input for this reduction. If this is a base reduction, this will be None

Return type:

Optional[PredictionType]

output_features_type

The features type expected as output for this reduction. If this is a base reduction, this will be None

Return type:

Optional[FeaturesType]

output_label_type

The label type expected as output for this reduction. If this is a base reduction, this will be None

Return type:

Optional[LabelType]

output_prediction_type

The prediction type expected as output for this reduction

Return type:

PredictionType

class FormatType

DsJson = FormatType.DsJson

Json = FormatType.Json

VwText = FormatType.VwText

Configuration

class Config

While this shows up as a class, it is actually a TypedDict. So it can be instantiated as a dict or as a class.

entryReduction: EntryReductionConfig

globalConfig: GlobalConfig

class GlobalConfig

While this shows up as a class, it is actually a TypedDict. So it can be instantiated as a dict or as a class.

constantFeatureEnabled: NotRequired[bool]

hashSeed: NotRequired[int]

interactions: NotRequired[List[Union[Literal['Default'], NameInteraction]]]

numBits: NotRequired[int]

class EntryReductionConfig

While this shows up as a class, it is actually a TypedDict. So it can be instantiated as a dict or as a class.

config: NotRequired[Dict[str, Any]]

typename: str

Interaction

alias of Union[Literal[‘Default’], NameInteraction]

class NameInteraction

name: str

Labels

class LabelType

Type of label

• Simple - Corresponds with reductionml.SimpleLabel

• CB - Corresponds with reductionml.CbLabel

• Binary - not implemented

Binary = LabelType.Binary

CB = LabelType.CB

Simple = LabelType.Simple

class SimpleLabel(value: float, weight: float = 1.0)

Parameters:

• value (float) – Label value

• weight (float) – Label weight

value

The label’s value

Return type:

float

weight

Weight of example to be used in update

Return type:

float

class CbLabel(action: int, cost: float, probability: float)

Parameters:

• action (int) – Chosen action (zero based)

• cost (float) – Cost of chosen action

• probability (float) – Probability of chosen action

action

The label’s action

Return type:

int

cost

The label’s cost

Return type:

float

probability

The label’s probability

Return type:

float

Predictions

class PredictionType

Type of prediction

• Scalar - Corresponds with reductionml.ScalarPred

• ActionScores - Corresponds with reductionml.ActionScoresPred

• ActionProbs - Corresponds with reductionml.ActionProbsPred

• Binary - not implemented

ActionProbs = PredictionType.ActionProbs

ActionScores = PredictionType.ActionScores

Binary = PredictionType.Binary

Scalar = PredictionType.Scalar

class ScalarPred(prediction: float, raw_prediction: float)

Parameters:

• prediction (float) – Prediction value (including clamping by the seen range and link function)

• raw_prediction (float) – Raw prediction value

prediction

raw_prediction

class ActionScoresPred(value: List[Tuple[int, float]])

Parameters:

value – A list of tuples of the form (action, score)

value

class ActionProbsPred(value: List[Tuple[int, float]])

Parameters:

value – A list of tuples of the form (action, probability)

value

Features

class FeaturesType

Type of features

• SparseSimple - Corresponds with reductionml.SparseFeatures

• SparseCbAdf - Corresponds with reductionml.CbAdfFeatures

SparseCbAdf = FeaturesType.SparseCbAdf

SparseSimple = FeaturesType.SparseSimple

class CbAdfFeatures

class SparseFeatures

class typing_extensions.NotRequired

See typing_extensions.NotRequired