By default, the dtype of int tensor is `tf.dtypes.int64`, and the dtype of float tensor is `tf.dtypes.float64`, which need more resources of calculation. Making the dtypes configurable is a meaningful feature.
By default, the dtype of int tensor is
tf.dtypes.int64, and the dtype of float tensor istf.dtypes.float64, which need more resources of calculation.Making the dtypes configurable is a meaningful feature.