通过java运行时使用pickled python类

我使用scikit learn从一些字符串构建了一个二进制SVM分类器:

count_vect = TfidfVectorizer(sublinear_tf=True, min_df=10, norm='l2', encoding='latin-1', ngram_range=(1, 4))
X_counts = count_vect.fit_transform(X)
tfidf_transformer = TfidfTransformer()
X_tfidf = tfidf_transformer.fit_transform(X_counts)
clf = SGDClassifier(loss='hinge', penalty='l2', tol=1e-3)
clf.fit(X_tfidf, Y)

然后,我围绕经过训练的分类器创建了一个包装类并对其进行了pickle:

class Classifier:
    def __init__(self, clf, vect):
        self.classifier = clf
        self.vectorizer = vect

    def classify(self, s):
        return self.classifier.predict(self.vectorizer.transform([s]))[0]

with open('my_classifier.pkl', 'wb') as fout:
    pickle.dump(Classifier(clf, count_vect), fout)

with open('my_classifier.pkl', 'rb') as fin:
    clf = pickle.load(fin)

result = clf.classify(sys.argv[1])
print(result)

但是,当我试图通过java运行时执行脚本时,它显示了不正确的输出。

public boolean classify(String s) throws IOException {
        String cmd = "python3 pkl_classifier.py \"" + s + "\"";
        Process p = Runtime.getRuntime().exec(cmd);
        BufferedReader stdIn = new BufferedReader(new InputStreamReader(p.getInputStream()));

        String out = stdIn.readLine();
        if (out != null) {
            switch (Integer.parseInt(out)) {
                case 0: return false;
                case 1: return true;
                default: throw new RuntimeException("Error with classifier script:" + out);
            }
        }

        return false;
    }

分类器的输出可以是0或1。但是这个java代码总是产生一个0。我已经把整个命令打印出来了( cmd )直接在终端执行,产生正确的输出。但是java运行时生成的输出总是0。