在Tensorflow中计算批量中的成对距离而不复制张量？

I want to compute the pairwise square distance of a batch of feature in Tensorflow. I have a simple implementation using + and * operations by tiling the original tensor :I want to compute the pairwise square distance