if (existing != accumulated_gradients_.end()) {
        // This is a somewhat odd case to be in, since it means we have two
        // operations which supposedly both created the same Tensor. It comes up
        // in recompute_grad, where the gradients have the same value. However,
        // only the original gradient is connected to everything else, so we
        // should still use that.

  virtual Status NumElements(int64_t* num_elements) const = 0;
  // Returns size of specified dimension
  //
  // -1 indicates an unknown axis length; this is unreachable for most standard
  // ImmediateExecutionTensorHandles, but comes up for example when computing
  // the shape of a parallel tensor with component shapes differing across
  // devices.
  virtual Status Dim(int dim_index, int64_t* dim) const = 0;