will be nondeterministic if `indices` contains duplicates.

`indices` is an integer tensor containing indices into a new tensor of shape
`shape`.

* `indices` must have at least 2 axes: `(num_updates, index_depth)`.
* The last axis of `indices` is how deep to index into `tensor` so  this index
  depth must be less than the rank of `tensor`: `indices.shape[-1] <= tensor.ndim`