Module aivis_engine_v2_sp_sdk_python.inference

Classes

class SignalPredictionInference (handle, key)

Class for creating and handling signal prediction inference. This is your entry point for all inference operations.

Private constructor. To create new instances use any of:

• SignalPredictionInference.create_by_training()
• SignalPredictionInference.create_by_model()

Static methods

def create_by_incremental_learning(cls, learning: SignalPredictionIncrementalLearning, config_json: str) ‑> SignalPredictionInference

@FlavourRequirement([Flavour.FULL])

Create signal prediction inference for the given learning handle and config.

To create predictions with this inference instance use any of:

• SignalPredictionInference.infer_boolean()
• SignalPredictionInference.infer_float()
• SignalPredictionInference.infer_string()

Parameters

learning : aivis_engine_v2_sp_sdk_python.learning.SignalPredictionIncrementalLearning

Instance of signal prediction incremental learning

config_json : str

DtoInferenceConfig as JSON string

Returns

SignalPredictionInference

Instance of signal prediction inference

def create_by_model(cls, model_json: str, config_json: str) ‑> SignalPredictionInference

@FlavourRequirement([Flavour.FULL, Flavour.INFERENCE])

Create signal prediction inference for given model and config.

To create predictions with this inference instance use any of:

• SignalPredictionInference.infer_boolean()
• SignalPredictionInference.infer_float()
• SignalPredictionInference.infer_string()

Parameters

model_json : str

DtoModel as JSON string

config_json : str

DtoInferenceConfig as JSON string

Returns

SignalPredictionInference

Instance of signal prediction inference

def create_by_training(cls, training: SignalPredictionTraining, config_json: str) ‑> SignalPredictionInference

@FlavourRequirement([Flavour.FULL])

Create signal prediction inference for given training handle and config.

To create predictions with this inference instance use any of:

• SignalPredictionInference.infer_boolean()
• SignalPredictionInference.infer_float()
• SignalPredictionInference.infer_string()

Parameters

training : SignalPredictionTraining

Instance of signal prediction training

config_json : str

DtoInferenceConfig as JSON string

Returns

SignalPredictionInference

Instance of signal prediction inference

Methods

def destroy(self)

@FlavourRequirement([Flavour.FULL, Flavour.INFERENCE])

Destroy this signal prediction inference. It's always safe to destroy an inference. Internally the destruction only takes place after all references to this object have been released.

def get_data_specification(self) ‑> str

@FlavourRequirement([Flavour.FULL, Flavour.INFERENCE])

Get this inference's DtoInferenceDataSpecification

Returns

str

DtoInferenceDataSpecification as JSON string

def get_output_specification(self) ‑> str

@FlavourRequirement([Flavour.FULL, Flavour.INFERENCE])

Get this inference's DtoInferenceOutputSpecification

Returns

str

DtoInferenceOutputSpecification as JSON string

def infer_boolean(self, data: SignalPredictionData, timestamps: List[int]) ‑> List[DtoBooleanDataPoint]

@FlavourRequirement([Flavour.FULL, Flavour.INFERENCE])

Calculate boolean-valued inference for given data context and timestamps. The result will be a list of DtoBooleanDataPoint instances corresponding to given timestamps. Make sure to fill the data context with all data needed for calculation. Check the model's DtoInferenceDataSpecification for needed signals and time ranges.

Parameters

data :  SignalPredictionData

Instance of signal prediction data

timestamps : List[int]

List of timestamps

Returns

List[DtoBooleanDataPoint]

List of boolean data points

def infer_boolean_with_category_probabilities(self, data: SignalPredictionData, timestamps: List[int]) ‑> List[DtoBooleanDataPointWithCategoryProbabilities]

@FlavourRequirement([Flavour.FULL, Flavour.INFERENCE])

Calculate boolean-valued category-probability inference list for given signal prediction inference, data context and timestamps. Method can only be called if underlying model was trained with categorical interpreter. The result will be a list of DtoBooleanDataPointWithCategoryProbabilities instances corresponding to given timestamps. Make sure to fill the data context with all data needed for calculation. Check the model's DtoInferenceDataSpecification for needed signals and time ranges.

Parameters

data :  SignalPredictionData

Instance of signal prediction data

timestamps : List[int]

List of timestamps

Returns

List[DtoBooleanDataPointWithCategoryProbabilities]

List of data points consisting of list of category-probability pairs for boolean categories

def infer_float(self, data: SignalPredictionData, timestamps: List[int]) ‑> List[DtoFloatDataPoint]

@FlavourRequirement([Flavour.FULL, Flavour.INFERENCE])

Calculate float-valued inference for given data context and timestamps. The result will be a list of DtoFloatDataPoint instances corresponding to given timestamps. Make sure to fill the data context with all data needed for calculation. Check the model's DtoInferenceDataSpecification for needed signals and time ranges.

Parameters

data :  SignalPredictionData

Instance of signal prediction data

timestamps : List[int]

List of timestamps

Returns

List[DtoFloatDataPoint]

List of float data points

def infer_float_with_category_probabilities(self, data: SignalPredictionData, timestamps: List[int]) ‑> List[DtoFloatDataPointWithCategoryProbabilities]

@FlavourRequirement([Flavour.FULL, Flavour.INFERENCE])

Calculate float-valued category-probability inference list for given signal prediction inference, data context and timestamps. Method can only be called if underlying model was trained with categorical interpreter. The result will be a list of DtoFloatDataPointWithCategoryProbabilities instances corresponding to given timestamps. Make sure to fill the data context with all data needed for calculation. Check the model's DtoInferenceDataSpecification for needed signals and time ranges.

Parameters

data :  SignalPredictionData

Instance of signal prediction data

timestamps : List[int]

List of timestamps

Returns

List[DtoFloatDataPointWithCategoryProbabilities]

List of data points consisting of list of category-probability pairs for float categories

def infer_float_with_next_normal(self, data: SignalPredictionData, timestamps: List[int], config_json: str) ‑> List[DtoFloatDataPointWithNextNormal]

@FlavourRequirement([Flavour.FULL, Flavour.INFERENCE])

For given data, timestamps and next normal config, calculate the inference and the closest point that is normal as defined by the normal config. The result will be a list of DtoFloatDataPointWithNextNormal instances corresponding to given timestamps. Make sure to fill the aivis_engine_v2_ad_sdk_python.data.AnomalyDetectionData with all data needed for calculation. Check the model's inference data specification for needed signals and time ranges.

@experimental: Might change in future releases. See the user guide for known performance issues.

Parameters

data :  SignalPredictionData

Instance of signal prediction data

timestamps : List[int]

List of timestamps

config_json : str

DtoFloatNextNormalConfig as JSON string

Returns

List[DtoFloatDataPointWithNextNormal]

List of enhanced float data points

def infer_string(self, data: SignalPredictionData, timestamps: List[int]) ‑> List[DtoStringDataPoint]

@FlavourRequirement([Flavour.FULL, Flavour.INFERENCE])

Calculate string-valued inference for given data context and timestamps. The result will be a list of DtoStringDataPoint instances corresponding to given timestamps. Make sure to fill the data context with all data needed for calculation. Check the model's DtoInferenceDataSpecification for needed signals and time ranges.

Parameters

data :  SignalPredictionData

Instance of signal prediction data

timestamps : List[int]

List of timestamps

Returns

List[DtoStringDataPoint]

List of string data points

def infer_string_with_category_probabilities(self, data: SignalPredictionData, timestamps: List[int]) ‑> List[DtoStringDataPointWithCategoryProbabilities]

@FlavourRequirement([Flavour.FULL, Flavour.INFERENCE])

Calculate string-valued category-probability inference list for given signal prediction inference, data context and timestamps. Method can only be called if underlying model was trained with categorical interpreter. The result will be a list of DtoStringDataPointWithCategoryProbabilities instances corresponding to given timestamps. Make sure to fill the data context with all data needed for calculation. Check the model's DtoInferenceDataSpecification for needed signals and time ranges.

Parameters

data :  SignalPredictionData

Instance of signal prediction data

timestamps : List[int]

List of timestamps

Returns

List[DtoStringDataPointWithCategoryProbabilities]

List of data points consisting of list of category-probability pairs for string categories