ModelInferenceTime

class ModelInferenceTime

Measure model average inference time (in seconds) per sample.

Parameters

n_samplesint , default: 1000

number of samples to use for inference, but if actual dataset is smaller then all samples will be used

__init__(n_samples: int = 1000, **kwargs)

__new__(*args, **kwargs)

Methods

ModelInferenceTime.add_condition(name, ...)	Add new condition function to the check.
ModelInferenceTime.add_condition_inference_time_is_not_greater_than([value])	Add condition - checking that the average model inference time (in seconds) per sample is not greater than X.
ModelInferenceTime.clean_conditions()	Remove all conditions from this check instance.
ModelInferenceTime.conditions_decision(result)	Run conditions on given result.
ModelInferenceTime.finalize_check_result(...)	Finalize the check result by adding the check instance and processing the conditions.
ModelInferenceTime.metadata([with_doc_link])	Return check metadata.
ModelInferenceTime.name()	Name of class in split camel case.
ModelInferenceTime.params([show_defaults])	Return parameters to show when printing the check.
ModelInferenceTime.remove_condition(index)	Remove given condition by index.
ModelInferenceTime.run(dataset[, model])	Run check.
ModelInferenceTime.run_logic(context[, ...])	Run check.

Examples

Model Inference Time

previous

IdentifierLeakage.run_logic

next

ModelInferenceTime.add_condition