ValueError: (InvalidArgument) If Attr(soft_label) == true, the axis dimension of Input(X) and Input(Label) should be equal.
[Hint: Expected logits_dims[axis] == labels_dims[axis], but received logits_dims[axis]:102 != labels_dims[axis]:1.] (at /paddle/paddle/fluid/operators/softmax_with_cross_entropy_op.cc:171)

使用硬标签的交叉熵可以跑起来，但是使用软标签就出现问题，分类数为102，标签维度为1，维度不统一,我是用paddle.vision.DatasetFolder(train_image_dir)读取数据集的。

应该怎么修改能跑起来？谢谢大家！