以y_true所依赖的方式自定义Keras的损失函数

I'm working on a multi-label classifier. I have many output labels [1, 0, 0, 1...] where 1 indicates that the input belongs to that label and 0 means otherwise. I'm working on a multi-label classifier. I have