how to apply gradient clipping in TensorFlow when distributed training?

Question

I would like to know how to apply gradient clipping in TensorFlow when distributed training. Here\'s my code:

    @lazy_property
    def optimize(self):
        #