# open3d.ml.tf.ops.reduce_subarrays_sum¶

open3d.ml.tf.ops.reduce_subarrays_sum(values, row_splits, name=None)

Computes the sum for each subarray in a flat vector of arrays.

The start and end of the subarrays are defined by an exclusive prefix sum. Zero length subarrays are allowed as shown in the following example:

import open3d.ml.tf as ml3d

ml3d.ops.reduce_subarrays_sum(
values = [1,2,3,4],
row_splits=[0,2,2,4] # defines 3 subarrays with starts and ends 0-2,2-2,2-4
)
# returns [3,0,7]

# or with pytorch
import torch
import open3d.ml.torch as ml3d

ml3d.ops.reduce_subarrays_sum(
values = torch.Tensor([1,2,3,4]),
row_splits=torch.LongTensor([0,2,2,4]) # defines 3 subarrays with starts and ends 0-2,2-2,2-4
)
# returns [3,0,7]

Parameters
• values – A Tensor. Must be one of the following types: int32, int64, float32, float64. Linear memory which stores the values for all arrays.

• row_splits – A Tensor of type int64. Defines the start and end of each subarray. This is an exclusive prefix sum with 0 as the first element and the length of values as additional last element. If there are N subarrays the length of this vector is N+1.

• name – A name for the operation (optional).

Returns

A Tensor. Has the same type as values. The sum of each subarray. The sum of an empty subarray is 0. sums is a zero length vector if values is a zero length vector.