StreamingPipelineWrapper 类

流式处理场景中最终管道的包装器。

此包装器旨在实现以下目标:

  1. 将 FeaturizationInfoProvider 公开为运行的最终管道的 pipeline.steps[0]。 在非流式处理场景中公开记录了此 API 预期,我们添加它以进行奇偶校验。

  2. 为最终管道对象启用 pipeline.fit() 和 pipeline.transform()。 FeaturizationInfoProvider 本身不能直接作为 Nimbus 管道的组件(我们会遇到各种与 Nimbus 相关的特定错误)。 我们改为在内部构建一个 Nimbus 管道 (self._pipeline),并使用它进行操作。

继承
azureml.automl.runtime.shared.model_wrappers._AbstractModelWrapper
StreamingPipelineWrapper

构造函数

StreamingPipelineWrapper(streaming_featurization_transformer: Optional[azureml.automl.runtime.featurization.streaming.streaming_featurization_transformer.StreamingFeaturizationTransformer], fitted_pipeline: azureml.automl.runtime.shared.nimbus_wrappers.NimbusMlPipelineWrapper)

参数

streaming_featurization_transformer
fitted_pipeline

方法

fit
fit_transform
get_model

用于获取内部原始模型对象的抽象方法。

predict
predict_proba

使用最终估计器将转换应用于数据并预测类概率。

transform

fit

fit(training_data, **kwargs)

参数

training_data
必需

fit_transform

fit_transform(training_data)

参数

training_data
必需

get_model

用于获取内部原始模型对象的抽象方法。

get_model()

返回

内部模型对象。

predict

predict(X, *args, **kwargs)

参数

X
必需

predict_proba

使用最终估计器将转换应用于数据并预测类概率。

predict_proba(X, verbose=0, **kwargs)

参数

X
必需
verbose
默认值: 0

transform

transform(X, **kwargs)

参数

X
必需

属性

classes_

named_steps

steps