如果我不想使用类，如何从另一个函数访问动态变量？

Question

我有 2 个函数是 train 函数和 logreg 函数，主要函数是在其中运行 logreg 函数的 train。

当我执行 train 函数时，它给了我错误，

NameError: name 'clf_hyper' is not defined

我认为我没有得到导致 logreg 函数的clf_hyper变量，

对数函数，

from sklearn import model_selection

def logreg(clf,xtrain, ytrain):

    # define a grid of parameter
    # this can be a dictionary or a list of
    # dictionaries
    param_grid = {
        #"solver": ['newton-cg', 'lbfgs', 'liblinear', 'sag', 'saga'],
        #"penalty": ['none', 'l1', 'l2', 'elasticnet'],
        "C": [100, 10]
    }

    # initialize grid search
    model = model_selection.GridSearchCV(
        estimator = clf,
        param_grid = param_grid,
        scoring = "accuracy",
        verbose = 10,
        n_jobs = 1
    )

    # fit the model and extract best score
    model.fit(xtrain, ytrain)

    best_parameters = model.best_estimator_.get_params()
    for param_name in sorted(param_grid.keys()):
        print(f"\t{param_name}: {best_parameters[param_name]}")

    # initialize model with best_estimator_
    clf_hyper = model.best_estimator_

训练功能，

import argparse
import os

import config
#import model_dispatcher
#import vectorizer_dispatcher
import dispatcher
import use_function
import hyperparameter

import pandas as pd
import joblib

from nltk.tokenize import word_tokenize

from sklearn import linear_model
from sklearn import metrics
from sklearn import model_selection
from sklearn.feature_extraction.text import CountVectorizer

def run(fold, model, vectorizer):
    #read the training data
    df = pd.read_csv(config.TRAINING_FILE)

    # applying clean_text to Revies column
    df.loc[:, 'Review'] = df.Review.apply(use_function.clean_text)

    # training data is where kfold is not equal to provided fold
    # also, note that we reset the index
    df_train = df[df.kfold != fold].reset_index(drop=True)

    # validation data is where kfold is equal to provided fold
    df_test = df[df.kfold == fold].reset_index(drop=True)

    # initialize CountVectorizer with NLTK,s word_tokenize
    # function as tokenizer
    vectorizer = dispatcher.vectorizers[vectorizer]

    #fit count_vec on training data reviews
    vectorizer.fit(df_train.Review)

    #transform training and validation data reviews
    xtrain = vectorizer.transform(df_train.Review)
    xtest = vectorizer.transform(df_test.Review)

    ytrain = df_train.Rating

    # initialize model
    clf = dispatcher.models[model]

    #initialize hyperparameter if you want use
    # if not just give # sign in
    hyperparameter.logreg(clf,xtrain,ytrain)

    #return clf value from hyperparameter function
    #return clf_hyper


    #fit the model on training data reviews and Rating
    clf_hyper.fit(xtrain, df_train.Rating)

    # make prediction on test data
    # threshold for predictions is 0.5
    preds = clf_hyper.predict(xtest)

    #calculate accuracy
    accuracy = metrics.accuracy_score(df_test.Rating, preds)

    print(f"Fold={fold}")
    print(f"Accuracy = {accuracy}")
    print("")

    # save the model
    joblib.dump(clf,os.path.join(config.MODEL_OUTPUT, "dt_{fold}.bin")
    )

那么如何在 logreg 函数中获取变量 clf_hyper，如果我不想使用类，我将在训练函数中使用该函数？ 谢谢

Answer 1

如果你真的想这样做（我不推荐）你可以将 clf_hyper 定义为全局变量

def logreg(clf,xtrain, ytrain):
    global clf_hyper 

Answer 2

我会尝试（如果可能？）直接从函数 Logreg() 返回变量 clf_hyper。 然后您可以通过调用该函数来获取 clf_hyper 的值。

def logreg(clf, xtrain, ytrain):
    ...
    return clf_hyper
def run(fold, model, vectorizer):
    ...
    #return clf_hyper and exec the func
    return hyperparameter.logreg(clf,xtrain,ytrain)