AttributeError:无法从'app.py'>

Question

我已经运行代码来加载pickle保存的变量。这是我的密码

import pickle

app = Flask(__name__)
nb_model = pickle.load(open('model_pickle.pkl', 'rb'))

这是一个物体

class NaiveBayesSentiment():

def __init__(self, datasentimen, stopword, norm, datatest):
    self.datasentimen = datasentimen
    self.stop_word = stopword.T.values
    self.norm_words =  pd.Series(norm['kata kbbi'].values,index=norm['kata singkatan']).to_dict()
    self.datatest = pd.DataFrame([datatest], columns=['comment'])
def cleansing_data(self, datasentimen):
            datasentimen = datasentimen.replace('\n', ' ',regex = True)
            datasentimen = datasentimen.str.replace("[^a-zA-Z]+", " ", regex = True)
            datasentimen = datasentimen.str.replace('\s{2,}', ' ', regex=True)
            datasentimen = datasentimen.str.lower()
            return datasentimen
            cleanposdata = cleansing_data(datasentimen[datasentimen['class sentiment'] == 'positif']['comment'])
            cleannegdata = cleansing_data(datasentimen[datasentimen['class sentiment'] == 'negatif']['comment'])  

if __name__ == "__main__":
  app.run(debug=True)

我得到的错误如下：

nb_model = pickle.load(open('model_pickle.pkl', 'rb'))
AttributeError: Can't get attribute 'NaiveBayesSentiment' on <module '__main__' from 'app.py'>

这是我在Pyton中执行pickle时的代码

nb_model = NaiveBayesSentiment(datasentimen, stopword, norm, datatest)

with open('model_pickle.pkl', 'wb') as pickle_out:
pickle.dump(nb_model, pickle_out)

with open('model_pickle.pkl', 'rb') as pickle_in:
unpickled_nb_model = pickle.load(pickle_in)

print(unpickled_nb_model.finalclassification())

我该怎么办？请帮帮我

Answer 1

Python的pickle序列化数据，但不序列化类或函数。它们只保存了名称，所以反序列化时不存在具有相同名称的类或函数，出现上述错误。因此，同样要加载pickle对象，必须存在相同的定义

添加：

这意味着调用pickle.load时必须定义NaiveBayesSentiment。换句话说，导入一个具有NaiveBayesSentiment的模块