How to design sampler of dataloader in pytorch if I want labels of every samples in one batch be the same?

Question

My task can be simply regarded as a binary classification. My model demand the labels in one batch have to be the same. And labels between batches should be random chosen. How s