cupyx.distributed.array.distributed_array#

cupyx.distributed.array.distributed_array(array, index_map, mode=None)[source]#

Creates a distributed array from the given data.

This function does not check if all elements of the given array are stored in some of the chunks.

Parameters:

  • array (array_like) – DistributedArray object, cupy.ndarray object or any other object that can be passed to numpy.array().

  • index_map (dict from int to array indices) – Indices for the chunks that devices with designated IDs own. One device can have multiple chunks, which can be specified as a list of array indices.

  • mode (mode object, optional) – Mode that determines how overlaps of the chunks are interpreted. Defaults to cupyx.distributed.array.REPLICA.

Return type:

DistributedArray

See also

DistributedArray.mode for details about modes.

Example

>>> array = cupy.arange(9).reshape(3, 3)
>>> A = distributed_array(
...     array,
...     {0: [(slice(2), slice(2)),  # array[:2, :2]
...          slice(None, None, 2)], # array[::2]
...      1:  (slice(1, None), 2)})  # array[1:, 2]