如何在 Azure 机器学习设计器中为 Python Model 设置学习器类型？

Question

I am testing Azure Machine Learning Designer by having a custom Python Model (a simple kNN classification). I would like to tune the value of 'k' and get the best performing model but "Tune Model Hyperparameters" module gives following error when giving output from my "Create Python Model" as input.

ModuleExceptionMessage:LearnerTypesNotCompatible: Got incompatible learner type: "None". Expected learner types are: "(<TaskType.BinaryClassification: 1>, <TaskType.MultiClassification: 2>, <TaskType.Regression: 3>)".

How I can set the learner type of my own Python model? Is it even possible? Should I just code the parameter tuning myself with "Execute Python Script"-module?

My "Create Python model"-module script:

import pandas as pd
from sklearn.neighbors import KNeighborsClassifier

class AzureMLModel:
    def __init__(self, k = 3):
        self.model = KNeighborsClassifier(n_neighbors = k)
        self.feature_column_names = list()

    def train(self, df_train, df_label):
        self.feature_column_names = df_train.columns.tolist()
        self.model.fit(df_train, df_label)

    def predict(self, df):
        return pd.DataFrame({'Scored Labels': self.model.predict(df[self.feature_column_names])})

Answer 1

Tune Model Hyperparameters module does not support using the Create Python Model module as an input model.

My recommendation, would be to explore using the azureml-sdk 's HyperDrive to tune hyper-parameters. Here's a link that gives an introduction, there's a Jupyter notebook at the end as well.

Answer 2

Thanks for your question. Tune Model Hyperparameters does not support Create Python Model. Please use Execute Python Script to tune the parameters.