循环使用TimeseriesGenerator多个时间序列

Question

我有多个串联的时间序列，我想用它们来训练我的LSTM模型。我想避免一个时间序列的预测是基于另一个时间序列的特征的。因此，我尝试在循环中使用TimeseriesGenerator。我的代码看起来像这样

def data_generator(X, y, batch_size, look_back):

generators = []

for train_session in np.unique(X[:, -1]):

    mask = X[:, -1] == train_session

    X = X[mask]
    y = y[mask]

    generators.append(TimeseriesGenerator(X,
                                    y,
                                    length=look_back,
                                    batch_size=batch_size))

yield next(generators)

运行model.fit_generator时出现以下错误：

ValueError: `start_index+length=10 > end_index=-1` is disallowed, as no part of the sequence would be left to be used as current step.

这是什么问题，我该如何解决？

谢谢