xla

e123f924 - Promote int to float for tanh operation (consistent with Pytorch) (#6166)