mars.tensor.split#

mars.tensor.split(ary, indices_or_sections, axis=0)[source]#

Split a tensor into multiple sub-tensors.

Parameters

ary (Tensor) – Tensor to be divided into sub-tensors.
indices_or_sections (int or 1-D tensor) –
If indices_or_sections is an integer, N, the array will be divided into N equal tensors along axis. If such a split is not possible, an error is raised.

If indices_or_sections is a 1-D tensor of sorted integers, the entries indicate where along axis the array is split. For example, [2, 3] would, for axis=0, result in
- ary[:2]
- ary[2:3]
- ary[3:]
If an index exceeds the dimension of the tensor along axis, an empty sub-tensor is returned correspondingly.
axis (int, optional) – The axis along which to split, default is 0.

Returns

sub-tensors – A list of sub-tensors.

Return type

list of Tensors

Raises

ValueError – If indices_or_sections is given as an integer, but a split does not result in equal division.

See also

array_split: Split a tensor into multiple sub-tensors of equal or near-equal size. Does not raise an exception if an equal division cannot be made.
hsplit: Split into multiple sub-arrays horizontally (column-wise).
vsplit: Split tensor into multiple sub-tensors vertically (row wise).
dsplit: Split tensor into multiple sub-tensors along the 3rd axis (depth).
concatenate: Join a sequence of tensors along an existing axis.
stack: Join a sequence of tensors along a new axis.
hstack: Stack tensors in sequence horizontally (column wise).
vstack: Stack tensors in sequence vertically (row wise).
dstack: Stack tensors in sequence depth wise (along third dimension).

Examples

>>> import mars.tensor as mt

>>> x = mt.arange(9.0)
>>> mt.split(x, 3).execute()
[array([ 0.,  1.,  2.]), array([ 3.,  4.,  5.]), array([ 6.,  7.,  8.])]

>>> x = mt.arange(8.0)
>>> mt.split(x, [3, 5, 6, 10]).execute()
[array([ 0.,  1.,  2.]),
 array([ 3.,  4.]),
 array([ 5.]),
 array([ 6.,  7.]),
 array([], dtype=float64)]