Why is XLA compilation used automatically with uint32 in Tensorflow

Question

I am having problems with TF 2.3, in particular with tensors of typetf.uint32. It seems that several operations don\'t support tf.uint32, for examp