Class_weight: TypeError: unhashable type: ‘numpy.ndarray‘

笔者在运行下面这段代码时：

class_weights = compute_class_weight(class_weight = 'balanced',
classes = np.unique(train_label_), y = train_label_)

遇到报错：

TypeError: unhashable type: 'numpy.ndarray'

 此时：train_label.shape : (sample ,2)：此时用了

train_label =np_utils.to_categorical(train_label,num_classes=2)

后将shape改为（sample ,1）问题即解决。

此时又出现新的报错：

ValueError: classes should include all valid labels that can be in y

将代码做出如下更改即可:

ss_weights = compute_class_weight(class_weight = 'balanced',
classes = np.unique(train_label), y = np.ravel(train_label))

这里要注意的是，class_weight参数的接收形式：clases=[0, 1];y=[0,0,0,0,1,1,1,1]。shape:(1,sample)。