Question

我在尝试使用joblib取消对象时遇到属性错误。我应该单独挑选我的课程然后拆开它们吗？

Class MainWindow
Private Sub Button1_Click(sender As System.Object, e As System.Windows.RoutedEventArgs) Handles Button1.Click
    For Each window As Window In Application.Current.Windows
        window.Background = New SolidColorBrush(Colors.Yellow)
    Next
End Sub
Private Sub Button2_Click(sender As System.Object, e As System.Windows.RoutedEventArgs) Handles Button2.Click
    Dim myWindow As New Window1()
    myWindow.Owner = Application.Current.MainWindow
    myWindow.ShowDialog()
End Sub
End Class

在此处加载腌制文件

class LengthTransformer(TransformerMixin):

    def transform(self, X, **transform_params):
        length = pd.DataFrame(X.apply(lambda x: len(x)))
        return length

    def fit(self, X, y=None, **fit_params):
        return self

pipeline = Pipeline([('features', FeatureUnion([
                                 ('ngram_tf_idf', Pipeline([
                                     ('vect', CountVectorizer()),
                                     ('tfidf', TfidfTransformer())
                                 ])),
                                 ('length',LengthTransformer())
                            ])),
                             ('clf', clf),])

vectorizer_name = 'fpath'+name

joblib.dump(pipeline.steps[0][1],vectorizer_name+'_vect_'+str(int(accuracy*100)) +'.pkl')

我收到以下错误。我相信这是因为它没有腌制我的课程。我读过pickle可以用来序列化类，但我使用joblib来获得更好的性能。

vect_file = "MultinomialNB_vect_80.pkl"
vect = joblib.load(vect_file)

Answer 1

在一个单独的模块中定义您的类，并将该模块导入正在执行pickle转储的程序和执行pickle加载的程序。

Joblib无法正确地解开课程

1 个答案: