我想用我在 azure ml 上预训练和保存的模型创建一个实验。使用此模型时,我收到此错误:
Got exception when invoking script: ''Sequential' object has no attribute '_distribution_strategy''.
它可能与已安装的库有关,但是,我不知道如何修复它。这是代码:
import pandas as pd
import numpy as np
import sys
import pickle
import io
import os
os.system(f"pip install keras")
os.system(f"pip install tensorflow=2.2")
import keras
from sklearn.feature_extraction.text import TfidfVectorizer
def azureml_main(df, dataframe2 = None):
sys.path.insert(0,"./Script Bundle")
tfidf, clf = pickle.load(open("./Script Bundle/model.pkl", 'rb'))
f = tfidf.transform(df.description)
y_pred = clf.predict_classes(f.toarray()) ### this line causes the error
y_probability = np.around(clf.predict_proba(f.toarray()),4)
df_pred = pd.DataFrame(y_pred).transpose()
df_probability = pd.DataFrame(y_probability).transpose()
df_all = pd.concat([df_pred.transpose(),df_probability.transpose()],axis=1)
return df_all,
希望有人可以提供帮助。