python在赋值前引用的局部变量

我在想要使用保存的模型提取模型的SHAP 值来解释经过训练的 ML 模型时遇到了赋值前引用的局部变量这个问题
import tensorflow as tf
from tensorflow.keras.models import save_model, load_model

tf.keras.backend.set_floatx('float64')
import numpy as np
import pandas as pd
from sklearn.preprocessing import StandardScaler
import pickle
import argparse
import os
import shap
from DNN_functions import nanmean_squared_error

'''
python step35.py --modelname 'DNN_BLR_fing'
'''


def SHAP(args):
    modeltype = args.modelname.split('_')[0]
    imputation = args.modelname.split('_')[1]
    features = args.modelname.split('_')[2]

    maindirectory = os.getcwd() + '/models/' + args.modelname

    X_df = pd.read_csv(os.getcwd() + '/datasets/datasetAX_' + features + '.csv')
    os.chdir(maindirectory)

    # obtain the proper Xscaler
    X = np.array(X_df)
    if features == 'desc':
        Xscaler = StandardScaler()
        X = Xscaler.fit_transform(X)

    if modeltype == 'RF':
        filename = args.modelname + '.sav'
        model = pickle.load(open(filename, 'rb'))

        background = X
        explainer = shap.TreeExplainer(model)
        shap_values = explainer.shap_values(background)

    if modeltype == 'DNN':
        max_index = 15

        background = X
        avg_shap_values = np.zeros((6, X.shape[0], X.shape[1]))
        for i in range(max_index + 1):
            print('Calculating SHAP values ' + str(i) + '/' + str(max_index + 1))
            directory = maindirectory + '/DNN_' + str(i)
            model = tf.keras.models.load_model(directory,
                                               custom_objects={'nanmean_squared_error': nanmean_squared_error})
            e = shap.explainers._deep.Deep(model, background)
            shap_values = e.shap_values(background)
            avg_shap_values = avg_shap_values + shap_values

        avg_shap_values = avg_shap_values / (max_index + 1)
        shap_values = avg_shap_values

    Columns = ['H2', 'He', 'O2', 'N2', 'CO2', 'CH4']
    for i in range(len(shap_values)):
        gas = Columns[i]
        np.savetxt('shap_' + str(i) + '.csv', shap_values[i], delimiter=",")
    print('SHAP values saved to ' + maindirectory)


if __name__ == '__main__':
    parser = argparse.ArgumentParser()
    parser.add_argument('--modelname', type=str, required=True,
                        help='provide the name of the folder in the "model" folder')
    parsed_args = parser.parse_args()

    SHAP(parsed_args)
运行结果及报错内容
File "step35.py", line 81, in 
    SHAP(parsed_args)
  File "step35.py", line 69, in SHAP
    for i in range(len(shap_values)):
UnboundLocalError: local variable 'shap_values' referenced before assignment
请问大家要怎样修改我的这个脚本中的变量呢

img


如果下面要用到这个变量，那么就要在所有的条件判断里都要定义，进入判读之前给一个默认值，否则程序是会有问题的

在这里加个定义就好了，然后后续如果不走if的话，可能还要判断一下空指针
或者如果是列表类型的话，直接定义shape_values = []也可以

img

shap_values是定义在if里面的
由于python没有块的概念，只要你if正常走，这样是没有问题的
但是你的if没有走，导致shap_values没有被赋值

代码没问题哟，你传参有问题，--modelname 需要“modeltype”你没传或者有误，应该是RF_BLR_fing或者DNN_BLR_fing
你的代码的调用规则写的很明确了

img